EFSA Report 2017

EXTERNAL SCIENTIFIC REPORT
APPROVED: 13 December 2016

doi:10.2903/sp.efsa.2017.EN-1151
Closing gaps for performing a risk assessment on Listeria

monocytogenes in ready-to-eat (RTE) foods: activity 3, the
comparison of isolates from different compartments along
the food chain, and from humans using whole genome
sequencing (WGS) analysis
Eva Mller Nielsen1, Jonas T. Bjrkman1, Kristoffer Kiil1, Kathie Grant2, Tim Dallman2,
Anas Painset2, Corinne Amar2, Sophie Roussel3, Laurent Guillier3, Benjamin Flix3,
Ovidiu Rotariu4, Francisco Perez-Reche4, Ken Forbes4, Norval Strachan4
1
Statens Serum Institut, Copenhagen, Denmark; 2Public Health England, Colindale, UK; 3Anses,
Maison-Alfort, France; 4University of Aberdeen, UK
Abstract
This report presents the results of the project Closing gaps for performing a risk assessment on
Listeria monocytogenes in ready-to-eat (RTE) foods: activity 3, the comparison of isolates from
different compartments along the food chain, and from humans using whole genome sequencing
(WGS) analysis. The main objective was to compare L. monocytogenes isolates collected in the EU
from ready-to-eat (RTE) foods, compartments along the food chain and from human cases by the use
of WGS. A total of 1,143 L. monocytogenes isolates were selected for the study, including 333 human
clinical isolates and 810 isolates from the food chain. The isolates were whole genome sequenced.
The phylogeny showed a clear delineation between L. monocytogenes lineages and between clonal
complexes within lineages. A range of typing methods were applied to the sequence data, providing
the framework to answer questions on genetic diversity and epidemiological relationships.
Retrospective analysis of nine outbreaks showed that WGS is a powerful tool in national and
international outbreak investigations as WGS can accurately rule isolates in or out of outbreaks.
Source attribution models showed bovine reservoir to be the main source of human disease although
other sources also contributed and generally confidence intervals were high. Numerous consistent
genetic linkages between a priori unlinked strains were identified, some of which involved isolates
from multiple countries. The presence of putative markers conferring the potential to survive/multiply
in the food chain and/or cause disease in humans was explored by detecting the presence of putative
virulence genes, AMR genes and factors conferring the ability to persist in the food processing chain.
This study has demonstrated one of the major benefits of WGS, which is the ability to address a wide
range of questions including those on virulence, antimicrobial resistance, source attribution,
surveillance and outbreak detection and investigation, in a single experiment.
European Food Safety Authority, 2017
Key words: Listeria monocytogenes, Whole Genome Sequencing, Genetic Diversity, Phylogeny, Food,
Human
Question number: EFSA-Q-2014-00026

Correspondence: [email protected]
www.efsa.europa.eu/publications EFSA Supporting publication 2017:EN-1151

LISEQ WGS analysis of Listeria from food and human sources
Disclaimer: The present document has been produced and adopted by the bodies identified above
as author(s). This task has been carried out exclusively by the author(s) in the context of a contract
between the European Food Safety Authority and the author(s), awarded following a tender
procedure. The present document is published complying with the transparency principle to which the
Authority is subject. It may not be considered as an output adopted by the Authority. The European
Food Safety Authority reserves its rights, view and position as regards the issues addressed and the
conclusions reached in the present document, without prejudice to the rights of the authors.
Acknowledgements:
We would like to thank all the persons and institutes that have provided the project with isolates and
accompanying information. Without them, this project would not have been possible.
Lin Cathrine T. Brandal, Norwegian Institute of Public Health, Norway
Julio Vzquez Moreno and Raquel Abad Torreblanca, Instituto de Salud Carlos III, Spain
Marc Lecuit, Institut Pasteur, France
Alexandre Leclercq, Institut Pasteur, France
Iva Hristova, National Center of Infectious and Parasitic Diseases, Bulgaria
Marija Trkov, National Laboratory of Health, Environment and Food, Slovenia
Cecilia Jernberg, Public Health Agency of Sweden, Sweden
Ariane Pietzka, Austrian Agency for Health and Food Safety, Austria
Eelco Franz and Ingrid Friesema, RIVM, The Netherlands
Carlo Spanu, University of Sassari Sardinia
Ifip, French Institute for Pig and Pork Industry, Maisons-Alfort, France
All the NRLs for providing the isolates from the EU baseline study
Special thanks to Sylvain Brisse and Alexandra Moura, Institut Pasteur, France, for providing cgMLST
data.
The authors would also like to thank the EFSA staff members: Maria Teresa da Silva Felicio, Beatriz
Guerra, Ernesto Lebana and Valentina Rizzi as well as the members of the Working Group on Listeria
monocytogenes contamination of ready-to-eat foods: Kostas Koutsoumanis, Roland Lindqvist, Moez
Sanaa, Panagiotis Skandamis, Niko Speybroek, Johanna Takkinen and Martin Wagner for the support,
revisions and suggestions during the development of the present procurement activity and report.
Suggested citation: Eva Mller Nielsen, Jonas T. Bjrkman, Kristoffer Kiil, Kathie Grant,
Tim Dallman, Anas Painset, Corinne Amar, Sophie Roussel, Laurent Guillier, Benjamin Flix,
Ovidiu Rotariu, Francisco Perez-Reche, Ken Forbes and Norval Strachan, 2017. Closing gaps for
performing a risk assessment on Listeria monocytogenes in ready-to-eat (RTE) foods: activity 3, the
comparison of isolates from different compartments along the food chain, and from humans using
whole genome sequencing (WGS) analysis. EFSA supporting publication 2017:EN-1151. 170 pp.
doi:10.2903/sp.efsa.2017.EN-1151
ISSN: 2397-8325
European Food Safety Authority, 2017
Reproduction is authorised provided the source is acknowledged.
www.efsa.europa.eu/publications 2 EFSA Supporting publication 2017:EN-1151

The present document has been produced and adopted by the bodies identified above as author(s). This task has been carried out exclusively by the author(s)
in the context of a contract between the European Food Safety Authority and the author(s), awarded following a tender procedure. The present document is
published complying with the transparency principle to which the Authority is subject. It may not be considered as an output adopted by the Authority. The
European Food Safety Authority reserves its rights, view and position as regards the issues addressed and the conclusions reached in the present document,
without prejudice to the rights of the author(s).
Summary
This report presents the results of the project Closing gaps for performing a risk assessment on
Listeria monocytogenes in ready-to-eat (RTE) foods: activity 3, the comparison of isolates from
different compartments along the food chain, and from humans using whole genome sequencing
(WGS) analysis. The project acronym, LISEQ, will be used in this report.
The main objective of the study was to compare L. monocytogenes isolates collected in the EU from
ready-to-eat (RTE) foods, compartments along the food chain and from human cases by the use of
whole genome sequencing (WGS).
A total of 1,143 L. monocytogenes isolates were selected for the study and these included 333 human
clinical isolates and 810 isolates from the food chain. The food chain isolates were acquired as part of
the EU-wide Baseline survey (BLS) on ready-to-eat food conducted in 2010-2011 (353 isolates),
obtained as part of national surveys, control programmes or research projects (423 isolates) or in
connection to outbreak investigations (34 isolates). The human clinical isolates were supplied by
national public health laboratories and represented sporadic cases (262 isolates) and outbreak-related
isolates (71 isolates) from eleven European countries, mainly in the years 2010-2011.
A database was constructed with the available metadata for the isolates with links to the genome
sequences. The isolates were whole genome sequenced (WGS) at Public Health Englands sequencing
facility. To achieve the goals of the specific objectives we applied a range of microbial typing methods
to the sequence data providing the framework to answer questions on genetic diversity and
epidemiological relationships. Three allelic based typing methodologies were used, multi-locus
sequence typing (MLST), core genome MLST (cgMLST) and ribosomal MLST (rMLST) as well as Single
Nucleotide Polymorphisms (SNPs). SNPs inference and analysis was performed using software
developed at PHE with each L. monocytogenes clonal complex (CC) having a separate database
instance. Short read sequences from strains selected in the strains of the study were mapped against
an appropriate reference genome. The resultant sequence alignment maps were processed and high
quality variants extracted. From 1,143 isolates sequenced, 42 different CC and 13 singleton sequence
types (STs) were identified. One isolate had a novel ST and could not be assigned to any CC. Ten
clonal complexes accounted for 70% of the isolates.
It was a specific objective to perform a retrospective analysis of outbreak strains to investigate the
suitability of WGS as a tool in outbreak investigations. Isolates from nine food-borne outbreaks,
representing a range of different characteristics, were selected. The relationship between the human
clinical isolates, isolates available from the suspected sources as well as similar background isolates
were characterised by SNP and cgMLST analyses. Most of the outbreaks were tightly clustered; five
out of nine had a maximum pairwise distance of <5 SNP, four outbreaks had a maximum SNP
difference between 8 and 21. The cgMLST showed for the most part concordant results with the SNP
analysis. In 8 outbreaks the median and maximum sizes of branches within a whole CC were shorter
compared to SNP branches.
To explore the genetic diversity of L. monocytogenes within and between the different sources and of
human origin, different indices and measures were applied to characterise and describe the variation
of isolates in the collection. Simpsons index for humans and the five sources exhibited high diversity
(>0.8) for both MLST and rMLST. Simpsons index of diversity between each of the sources was
indistinguishable. Rarefaction curves showed for both MLST and rMLST that all of the genotypes had
not been sampled. Bovine and human isolates had the highest number of new STs per isolate. Neis
genetic distance showed that there were significant differences between human and all sources at all
levels of molecular analysis explored, but the distance between humans and bovine was the smallest.
Two approaches were used to assess the epidemiological relationship of L. monocytogenes from the
different sources and of human origin considering the genomic information and the metadata
available for each isolate. The first approach was using the method of source attribution, i.e.,
partitioning of the human disease burden of listeriosis to specific sources. Because of the relatively
small number of isolates, all of the isolates along the food chain that originate from a particular

reservoir were combined. Human clinical cases were attributed to these sources by utilising five
different mathematical models and the genomic typing data that was generated. Source attribution
was applied utilising 5 models (Hald, Dutch, STRUCTURE, Asymmetric Island and Aberdeen) for
5 sources [fish, swine, ovine, bovine and poultry] and four sources [removing poultry]). All of the
models showed bovine as the main source of human disease (32-64% for 5 sources and 33-61% for
4 sources) but for a number of the models there were broad confidence intervals.
The second approach to assess the possible epidemiological relationship between strains was to
identify clusters of clinical and food isolates based on SNP differences. The WGS data was analysed
along with the epidemiological information of the food and clinical isolate to assess, retrospectively,
relationships between circulating strains of L. monocytogenes in EU within 2010-2012 period. The
retrospective analysis showed that numerous consistent genetic linkages, between a priori unlinked
strains, can be established with WGS. By the use of SNP pairwise distances, 124 clusters of isolates
were identified and 27 of these included both human and food isolates, potentially relating sporadic
human cases to contemporary food isolates that circulate in EU. All three categories of RTE food
products were involved, but most of the clusters were related to smoked fish.
Another specific objective of the study was to identify the presence of putative markers conferring the
potential to survive/multiply in the food chain and/or cause disease in humans (e.g. virulence and
antimicrobial resistance). We analysed the WGS data for the presence of 115 putative markers of
virulence. More than 80% of markers were present in greater than 95% of the isolates suggesting
that most putative markers described in the literature are ubiquitous across L. monocytogenes
lineages I and II. The majority of markers not present in all isolates were over-represented in food
and/or lineage II isolates with markers associated with stress survival or cell wall modification being
particularly enriched. Conversely, the recently discovered Listeria pathogenicity island 3 and the
surface protein VIP were more likely to be found in clinical and/or lineage I isolates. Similar to the
virulence genes, the presence of genes related to resistance to antimicrobials and detergents were
searched in the LISEQ collection. There was found remarkable low resistance to tetracycline (<0.1%)
and penicillin (1%). Resistance to detergents and antiseptics via efflux activity was significant with
mechanisms detected at a prevalence approaching 20%. Some studies have suggested that
L. monocytogenes strains that are able to persist in the food production environment are genetically
distinct from transient strains that do not have this capability. The presence or absence of genes
thought to promote persistence was not found to be pertinent for predicting persistent phenotype.
However, it was shown that WGS SNP-based analysis is well suited and valuable for investigating
persistence and contamination routes within food processing facilities and within the food chain.
In conclusion, this study carried out WGS of a large unique collection of L. monocytogenes isolates
from foods, food processing environments and clinical cases from a large number of European
countries. This study has demonstrated one of the major benefits of WGS, which is the ability to
address a wide range of questions including those on virulence, antimicrobial resistance, source
attribution, surveillance and outbreak detection and investigation, in a single experiment. This study
illustrates one of the major strengths of WGS in comparison to conventional molecular typing
methods, which is its ability to provide high quality, unambiguous data. WGS analysis such as cgMLST
and cgSNP-based typing approaches have been shown to have unparalleled strain typing resolution
and it has been demonstrated here how WGS is able to link previously undetected cases to outbreaks
and detect clusters of cases that were previously undetected. It has also been shown, however, that
knowledge of the accessory genome can contribute to the interpretation of strain relatedness. The
limitations of WGS have less to do with the actual sequencing and the analyses themselves but more
dependent on representative sampling of isolates and requirement for good epidemiological data to
further investigate genetically linked by WGS. This study supports the use of WGS for
L. monocytogenes outbreak investigations although experience from more complex outbreaks would
be valuable. However, is difficult to recreate outbreak investigations accurately retrospectively and in
order to maximise the advantages of using WGS for outbreak detection it would be highly valuable to
use WGS prospectively for the surveillance of listeriosis across Europe.

Table of Contents
Abstract.........................................................................................................................................1
Acknowledgements: .......................................................................................................................2
Summary .......................................................................................................................................3
1. Introduction........................................................................................................................8
1.1. Background and Terms of Reference as provided by the requestor ......................................10
2. Isolate collection ...............................................................................................................12
2.1. Selection of strains to fulfil the main and specific objectives of the study..............................12
2.2. Strains from the baseline survey (BLS) ...............................................................................14
2.2.1. Selection criteria BLS 1. Availability of the strains................................................................14
2.2.2. Selection criterion BLS 2: Isolates from same origin (food and Member State) ......................14
2.2.3. Selection criterion BLS 3. Multiple isolates from a sample (all categories) .............................15
2.3. Strains from other foods (OF) ............................................................................................15
2.3.1. Selection criterion OF 1. Food origin...................................................................................15
2.3.2. Selection criterion OF 2. Temporal criterion ........................................................................15
2.3.3. Selection criterion OF 3. Further selection...........................................................................15
2.3.4. Fruit and vegetables..........................................................................................................15
2.4. Strains from food chain production stages ..........................................................................16
2.5. Strains from sporadic human clinical cases .........................................................................16
2.5.1. Selection criterion C1. Availability of the strains. .................................................................16
2.5.2. Selection criterion C2. Temporal and geographical criterion .................................................16
2.5.3. Selection criterion C3. Incidence ........................................................................................16
2.5.4. Selection criterion C4. Subtype information.........................................................................17
2.6. Strains from outbreaks ......................................................................................................17
2.7. Strain selection summary...................................................................................................17
2.8. Database ..........................................................................................................................21
2.8.1. Database structure............................................................................................................21
2.8.2. Core information on strains................................................................................................22
2.8.3. Data export ......................................................................................................................22
3. Methodologies ..................................................................................................................23
4. Sequencing and Phylogentic Analysis .................................................................................25
4.1. Methods ...........................................................................................................................25
4.1.1. DNA extraction .................................................................................................................25
4.1.2. DNA sequencing and validation ..........................................................................................25
4.1.3. Assembly and annotation...................................................................................................26
4.1.4. Gene by gene based typing ...............................................................................................26
4.1.5. Phylogenetic analysis ........................................................................................................26
4.1.6. Genetic distance: SNP address ...........................................................................................28
4.1.7. Genetic distance: SNP address ...........................................................................................28
4.2. Results .............................................................................................................................29
4.2.1. Processing of isolates ........................................................................................................29
4.2.2. Gene by gene based typing: MLST and clonal complex assignment ......................................30
4.2.3. Developing the Framework for Phylogenetic Analysis...........................................................32
4.2.4. Phylogenetic analysis of major clonal complexes .................................................................34
4.3. Conclusion........................................................................................................................38
5. Retrospective analysis of outbreaks....................................................................................39
5.1. Methods ...........................................................................................................................39
5.2. Results .............................................................................................................................40
5.2.1. Outbreak 1 CC155..........................................................................................................40
5.2.2. Outbreak 2 CC1 .............................................................................................................42
5.2.3. Outbreak 3 CC7 .............................................................................................................44
5.2.4. Outbreak 4 CC59............................................................................................................45

5.2.5. Outbreak 5 CC415..........................................................................................................47

5.2.6. Outbreak 6 CC398..........................................................................................................48
5.2.7. Outbreak 7 CC87............................................................................................................49
5.2.8. Outbreak 8 ST14............................................................................................................52
5.2.9. Outbreak 9 CC4 .............................................................................................................54
5.3. Conclusions ......................................................................................................................56
6. Genetic diversity ...............................................................................................................57
6.1. Methods ...........................................................................................................................57
6.1.1. Simpson's Diversity Index ..................................................................................................57
6.1.2. Rarefaction .......................................................................................................................58
6.1.3. Nei's genetic distance ........................................................................................................58
6.1.4. Graphical visualisation and cluster analysis .........................................................................58
6.1.5. Analyses...........................................................................................................................58
6.1.6. Selection of Genomes for analysis suitable for genetic diversity and source attribution analysis59
6.2. Results and Discussion ......................................................................................................60
6.2.1. 7 locus MLST ....................................................................................................................60
6.2.2. 30 locus rMLST .................................................................................................................62
6.2.3. 1,748 locus cgMLST ..........................................................................................................65
6.2.4. 39,529 cgSNP ...................................................................................................................66
6.2.5. Graphical Visualisation.......................................................................................................66
6.3. Conclusions ......................................................................................................................69
7. Epidemiological relationship: Source attribution ..................................................................69
7.1. Methods ...........................................................................................................................70
7.1.1. Source Attribution Methods................................................................................................70
7.1.2. Self-Attribution .................................................................................................................74
7.1.3. Analyses...........................................................................................................................74
7.2. Results and Discussion ......................................................................................................74
7.2.1. Source Attribution of 5 sources ..........................................................................................74
7.2.2. Source Attribution of 4 Sources (Excluding Poultry) .............................................................81
7.2.3. Discussion ........................................................................................................................86
7.3. Conclusions ......................................................................................................................86
8. Epidemiological relationship linking of genetically related isolates......................................87
8.1. Methods ...........................................................................................................................87
8.1.1. Definition of genetically clustered strains ............................................................................87
8.1.2. R packages and software...................................................................................................88
8.2. Results .............................................................................................................................89
8.2.1. Epidemiological analysis of genetically clustered strains: link between human sporadic strains
and potential relation with food strains...............................................................................91
8.2.2. Geographical and temporal widespread of genetically clustered strains.................................94
8.2.3. Consistency of clusters established.....................................................................................95
8.3. Conclusion........................................................................................................................96
9. Putative markers...............................................................................................................96
9.1. Methods ...........................................................................................................................97
9.1.1. Antibiotic resistance genes.................................................................................................97
9.1.2. Published virulence factors ................................................................................................97
9.1.3. Genes implicated in persistence .........................................................................................97
9.1.4. Markers of host association................................................................................................98
9.2. Results .............................................................................................................................98
9.2.1. Antimicrobial resistance .....................................................................................................98
9.2.2. Published virulence factors ................................................................................................99
9.2.3. Genes implicated in persistence ....................................................................................... 101
9.2.4. Markers of host association.............................................................................................. 109
9.3. Conclusion...................................................................................................................... 109
10. Conclusions .................................................................................................................... 110

11. Additional supporting information..................................................................................... 115

References................................................................................................................................. 116
Glossary .................................................................................................................................... 124
LISEQ database glossary............................................................................................................. 124
List of abbreviations used in the report ........................................................................................ 125
Appendix 1: Isolates from the EU-wide baseline survey on prevalence of L. monocytogenes in certain
RTE foods conducted in 2010-2012 .................................................................................. 126
Appendix 2: Isolates other food, ready-to-eat meat and cheese .................................................... 139
Appendix 3: Isolates from other food, fruits and vegetables.......................................................... 147
Appendix 4: Isolates from the food production chain .................................................................... 148
Appendix 5: Isolates from sporadic clinical cases.......................................................................... 156
Appendix 6: Isolates from outbreaks ........................................................................................... 163
Appendix 7: Rarefaction and Simpsons diversity index of 7 locus MLST clinical data stratified by age166
Appendix 8: Links to Attribution model software........................................................................... 168

1. Introduction
Listeria monocytogenes causes a range of clinical illnesses from mild diarrhoea to severe invasive
infection (listeriosis) including bacteraemia, meningitis, encephalitis, abortion and stillbirth. In EU in
2014, a total number of 2,161 confirmed human cases were reported by 27 MS, corresponding to an
EU notification rate of 0.52 cases per 100,000 population. The highest notification rates were
observed in Denmark, Sweden, Finland and Spain (1.64, 1.30, 1.19 and 1.15 cases per
100,000 population respectively) (EFSA and ECDC, 2015). Listeriosis generally affects individuals who
have a weakened immune system including the elderly, those who are immunosuppressed due to
existing medical conditions or their treatment, pregnant women and neonates. Whilst listeriosis is a
relatively rare disease, it has a high fatality rate of 20-30% and the burden of disease is high. The
majority of cases appear to be sporadic, although outbreaks are not uncommon. Listeriosis is almost
exclusively transmitted by contaminated food with ready-to-eat meat and fish products and soft and
semi-soft cheeses often identified as sources of infection. Due to its ability to survive under conditions
of stress, L. monocytogenes has the capacity to persist in food processing environments, sometimes
for years and often this is the route by which ready-to-eat food becomes contaminated. Identifying
the food vehicle and tracing the origin of contamination are paramount in developing and
implementing effective control and preventative measures and the typing of L. monocytogenes
isolates continues to play a crucial role in such investigations.
A number of phenotypic and genotypic methods are used for typing L. monocytogenes. Traditionally,
serotyping, based on the agglutination of somatic (O) and flagellar (H) antigens, classifying
L. monocytogenes into at least 13 serotypes, has been the first level of subtype discrimination. More
recently a multiplex PCR scheme based on the amplification of four specific marker genes (lmo0737;
ORF2110; lmo1118 and ORF2819), has been used to distinguish L. monocytogenes into four molecular
groups that correlate well with known L. monocytogenes serotypes (1/2a-3a; 1/2b-3b-7; 1/2c-3c and
4b-3b-7) (Doumith et al., 2004). Since at least 95% of isolates from food and clinical cases are of
serotypes 1/2a, 1/2b, 1/2c and 4b this robust, reproducible PCR-based method has been adopted by
many reference laboratories as a rapid alternative that overcomes the many drawbacks of serotyping.
However, because both traditional serotyping and serogrouping by multiplex PCR separate
L. monocytogenes into only four groups, they lack discriminatory power which has led to the
development of a variety of molecular typing techniques for higher resolution subtyping for outbreak
detection, for linking human cases to food sources and for tracking strains along the food chain. The
most widely used method in reference laboratories is pulsed-field gel electrophoresis (PFGE), but
other methods including multi-locus variable number of tandem-repeat analysis (MLVA) and fAFLP are
also used (recently reviewed (Camargo et al., 2016)). PFGE using two restriction enzymes is the
standard method in PulseNet laboratories in North America, PulseNet International, as well as in the
European surveillance system for human infections (TESSy). PFGE is also the first method to be
included in the on-going project - coordinated by ECDC and EFSA - that aims to include molecular
typing data from humans and food/animal/environment into a joint molecular typing database (EFSA,
2014).
The molecular typing methods commonly in use such as PFGE suffer a number of practical limitations
including labour intensiveness, reproducibility and inter laboratory comparability. Whilst such methods
have been extremely valuable in assisting epidemiological investigations of listeriosis, they also lack
the ability to inform evolutionary relationships which would provide valuable insight for source
tracking and source attribution (Orsi et al., 2011). Such information is available through single
nucleotide polymorphism (SNP)based typing approaches including multilocus sequencing typing
(MLST) and whole genome sequencing (WGS) analysis. MLST is a well-established sequencing-based
method whereby the unique variation in specific fragments, of a set of housekeeping genes (usually
7) are assigned an allele number and the alleles at each loci provide an allelic profile or sequence

type. This technique has been used to study and describe the population structure and phylogeny of
many bacterial pathogens and has shown that L. monocytogenes forms a structured population
consisting of four divergent lineages (I-IV) (Ragon et al., 2008). Each lineage comprises specific
serotypes, with Lineage I containing serotypes: 1/2b, 3b, 4b, 4e and 7; lineage II: serotypes 1/2a,
1/2c, 3a and 3c; lineage III: serotypes 4b, 1/2a, 4a and 4c and lineage IV: 4a and 4c. The genetic
lineages have different, although at times overlapping, genetic, phenotypic and epidemiological
characteristics with the majority human illness caused by strains in lineages I and II. Thus 7-locus
MLST, in common with other current molecular typing techniques, in focusing on only a small portion
of the bacterial genome, provides insufficient strain resolution for detailed epidemiological
investigations and the use of other molecular typing techniques are also required.
With the advent of next generation sequencing technologies, entire bacterial genomes are now readily
available for analysis affording the highest level of strain discrimination, the ability to infer
phylogenetic relationships and access to a wealth of additional information such as virulence and
resistance markers. It is anticipated that in the near future whole genome sequencing (WGS) will
replace the currently used typing methods for foodborne pathogens, as it is now possible to obtain a
multitude of different characteristics based on WGS data in real time at a reasonable cost. The value
of WGS as a bacterial typing tool has already been assessed in a number of specific settings and WGS
is increasingly used for outbreak investigations and source tracing of L. monocytogenes. Thus, WGS
was used in several recent national studies for outbreak detection and investigations, e.g. in Austria
(Rychli et al., 2014), Australia (Kwong et al., 2016), USA (Jackson et al., 2016), Denmark (Kvistholm
Jensen et al., 2016). The improved resolution obtained by WGS enabled linking of isolates and more
robust case definitions enabling isolates to be ruled in or out of outbreaks. Furthermore, WGS makes
it possible to recognise extended time-period outbreaks and link clinical cases to food products and
food production facilities (Gillesberg Lassen et al., 2016). Such studies demonstrate the advantages of
using WGS analysis for national surveillance and outbreak detection and investigation. There have
also been studies where WGS analysis has been used to investigate the strains that persist in the food
processing environment and shown to distinguish those that are persistent from ones that are
repeatedly reintroduced (Stasiewicz et al., 2015).
However, the experience of using WGS for international comparison of isolates and linking human
cases to food products or food production facilities across borders is still very limited. This study
aimed at supporting the future European-wide use of WGS for improved food safety. The main
objective of the study was to compare L. monocytogenes isolates collected in the EU from ready-to-
eat (RTE) foods, along the food chain and from human cases. The isolates were whole genome
sequenced and the data were used for a number of analyses aiming at describing the phylogeny and
diversity of isolates from foods and humans in Europe, to evaluate the use of WGS for outbreak
investigations, to evaluate the possible epidemiological relationship between isolates, and to identify
possible markers of virulence/survival. More than 1,100 isolates were selected for the study (Section
2). The sequences were analysed with a variety of methods in order to fulfil the defined objectives of
the study. This has facilitated the comparison of a variety of WGS analysis methods using a large data
set from different European countries. This type of study has not been performed previously but is
essential to ensure the adoption at the European level of robust, harmonised WGS analysis methods
for national and international surveillance, outbreak detection and investigation.

1.1. Background and Terms of Reference as provided by the requestor
In the European Union (EU), listeriosis continues to be a serious food-borne illness, with high
morbidity, hospitalisation and mortality in vulnerable populations. For example in 2012,
1,642 confirmed human cases of listeriosis were reported including 198 deaths.1 The trend in reported
human listeriosis cases has been gradually increasing over the past four years.
The main route of transmission to humans is through consumption of contaminated food. The
bacterium can be found in raw foods and in processed foods that are contaminated during and/or
after processing. Because L. monocytogenes is able to multiply at low temperatures (2 to 4C), ready-
to-eat (RTE) foods with a relatively long shelf-life (such as fishery products, meat products and
cheese) are of particular public health concern.
An EU-wide baseline survey (BLS) was conducted in 2010 and 2011 to estimate the prevalence and
contamination levels in three RTE foods at retail in accordance with Decision 2010/678/EU: packaged
(not frozen) smoked or gravad fish (3,053 samples), packaged heat-treated meat products
(3,530 samples) and soft or semi-soft cheeses (3,452 samples). The Part A report (prevalence
estimates) was published in 2013.2 The EU prevalence of fish samples at the time of sampling was
10.4 % and at the end of shelf-life 10.3 %, while for meat and cheese samples at the end of shelf-life
these prevalences were 2.07 % and 0.47 %, respectively. The Terms of Reference of the subsequent
Part B report are (a) the analysis of (risk) factors related to the prevalence of contaminated foods,
(b) the development of predictive models for the microbial growth of L. monocytogenes under various
storage conditions, and (c) the development of predictive models for compliance with
L. monocytogenes food safety criteria in RTE foods. Publication of this report is expected in June
2014.
In a self-task mandate by the BIOHAZ Panel, information on current and prospective molecular sub-
typing methods for food-borne pathogens (among which L. monocytogenes) has been reviewed in
terms of discriminatory capability, reproducibility, and capability for international harmonisation. The
opinion was published at the end of 2013.3 Molecular approaches to characterise isolates, specifically
using sequence-based approaches as those based on whole genome sequence (WGS) analyses,
provide the means of describing and characterising the variation of bacterial populations with the
highest resolution possible, enhancing substantially our ability to understand and trace the sources
and spread of the diseases that they may cause.
Main objective:
The main objective of the contract resulting from the present procurement procedure is to compare
L. monocytogenes isolates collected in the EU from RTE foods, compartments along the food chain
and humans using whole genome sequencing (WGS) analysis.
The specific objectives are as follows:
1
EFSA and ECDC, 2014. The European Union Summary Report on Trends and Sources of Zoonoses, Zoonotic Agents and
Food-borne Outbreaks in 2012. EFSA Journal 2014;12(2):3547, 312 pp. doi:10.2903/j.efsa.2014.3547
2
European Food Safety Authority, 2013. Analysis of the baseline survey on the prevalence of Listeria monocytogenes in certain
ready-to-eat (RTE) foods in the EU, 2010-2011 Part A: Listeria monocytogenes prevalence estimates. EFSA Journal
2013;11(6):3241, 75 pp. doi:10.2903/j.efsa.2013.3241
3
EFSA BIOHAZ Panel (EFSA Panel on Biological Hazards), 2013. Scientific Opinion on the evaluation of molecular typing
methods for major food-borne microbiological hazards and their use for attribution modelling, outbreak investigation and
scanning surveillance: Part 1 (evaluation of methods and applications). EFSA Journal 2013;11(12):3502, 84 pp.
doi:10.2903/j.efsa.2013.3502

Specific objective 1: to carry out the molecular characterisation of a selection of L. monocytogenes

isolates from different sources, i.e. RTE foods, compartments along the food chain (e.g. food
producing animals, food processing environment), and humans employing WGS analysis.
The tenderer is expected, as part of the project, to present first a list of the L. monocytogenes isolates
from different sources, i.e. RTE foods, compartments along the food chain (e.g. food producing
animals, food processing environment), and humans that are accessible for WGS analysis. Isolates
from RTE foods shall at least consist of L. monocytogenes isolates collected from the BLS in RTE
foods. In the BLS 134 packaged smoked or gravad fish samples were found contaminated with
L. monocytogenes at the time of sampling, 133 at the end-of-shelf-life, and 176 at both stages. Also,
72 and 16 positive samples were detected in packaged heat-treated meat products and soft and semi-
soft cheeses, respectively. These isolates are stored at the EU Reference Laboratory for
L. monocytogenes and/or individual National Reference Laboratories for L. monocytogenes.4 EFSA will
ensure access for the successful contractor to these isolates, however, all the costs related to the
preparation and transport of these isolates to their own premises shall be covered by the contractor.
The successful contractor will not be authorised to use these isolates for any purpose outside the
remit of this project and will have to destroy them after completion of the project. An example of
Agreement for the transfer of Materials to be signed between the Contractor and the EURL and/or
the NRL before the shipment of the isolates can be found in Annex 7 of these tender specifications.
Further, L. monocytogenes isolates from other EU sources should also be included (e.g. RTE foods,
food producing animals, food processing environment) as well as human isolates from the same
period of the baseline survey (i.e. 2010-2011). If necessary, the isolate collection can be
complemented with other isolates from recent years (2012-2014). For each isolate relevant metadata
should be available in order to fulfil the objective. The tenderer is responsible for the identification of
the origin (e.g. geography, laboratory) of these isolates. Consideration of the sources, including the
number of isolates and related metadata, should be made with the overall aim of maximising the
outcome of the objectives of the study.
Then the criteria employed by the tenderer for the selection of the isolates for WGS analysis should be
clearly described in the offer. The minimum and/or maximum number of isolates to be included is left
to the discretion of the tenderer, with an estimate of this number to be provided in the offer. The total
number of isolates proposed should soundly represent isolates from different sources as stated above
and humans as well as from various geographical regions in the EU. Both the selection criteria and the
number of isolates proposed should be indicated in the offer and applied with the overall aim of
maximising the outcome of the objectives of the study.
Although the primary responsibility to list and select L. monocytogenes isolates is with the tenderer,
EFSA may provide, during the implementation of the contract, information on potential available
sources of isolates in particular from humans. The successful contractor should consult and agree with
EFSA on the final selection of the set of isolates to be typed. The WGS typing should be carried out
with state-of-the-art equipment and methodologies which conform with current laboratory standards
and that can be referred to or reported in a clear and concise manner. Robust annotation pipelines for
the WGS data generated should be designed and implemented with the aim of getting a harmonised
framework for subsequent data analysis.
Specific objective 2: to analyse the WGS typing data of the selected L. monocytogenes isolates with
three goals:
4
No objections were raised at the Meeting of the SCoFCAH (Section Biological Safety of the Food Chain), held in Brussels on
16 October 2013 for accessing the isolates from the EU BLS and its epidemiological data for this activity. Storage of isolates
by the NRLs or EU RL Lm beyond the minimum duration of 2 years has been requested.

i. to explore the genetic diversity of L. monocytogenes within and between the different sources and
human origin;
ii. to assess the epidemiological relationship of L. monocytogenes from the different sources and of
human origin considering the genomic information and the metadata available for each isolate;
iii. to identify the presence of putative markers conferring the potential to survive/multiply in the food
chain and/or cause disease in humans (e.g. virulence and antimicrobial resistance).
Specific objective 3: to perform a retrospective analysis of outbreak strains (i.e. using a subset of
epidemiologically linked human and food isolates) to investigate the suitability of WGS as a tool in
outbreak investigations:
Strains from known food-borne outbreaks of human listeriosis should be characterised employing WGS
methods and analysed following the methodological frame employed under objective 1 above. Next,
the available WGS data should be analysed for establishing and/or supporting links between the
different strains. The outcome of this analysis should provide an evaluation on the advantages and
limitations of employing WGS data for investigating outbreaks of food-borne listeriosis.
This contract was awarded by EFSA to:

Contractor: Statens Serum Institut (SSI), French Agency for Food, Environmental and Occupational
Health & Safety (ANSES), Public Health England (PHE), University of Aberdeen (UA)
Contract: Closing gaps for performing a risk assessment on Listeria monocytogenes in ready-to-eat
(RTE) foods: activity 3, the comparison of isolates from different compartments along the food chain,
and from humans using whole genome sequencing (WGS) analysis.
Contract number: OC/EFSA/BIOCONTAM/2014/01-CT 1
2. Isolate collection
2.1. Selection of strains to fulfil the main and specific objectives of the
study
A geographically wide selection of isolates from human cases and RTE foods that are likely to be the
direct source of human Listeria monocytogenes infections forms the main basis for the analyses to
fulfil objectives 2i and 2ii (genetic diversity and epidemiological relationship). High priority was given
to this selection by including 262 and 610 isolates from sporadic human cases and retail food samples,
respectively, to enable meaningful epidemiological linking as well as robust source attribution
modelling.
As stated in the tender specifications, isolates from the EU-wide baseline survey (BLS) conducted in
2010 and 2011 (EFSA, 2013) should be included in this project. Therefore, the selection of RTE food
isolates was based on the availability of isolates from the BLS, which included three types of RTE
food: smoked and gravad fish, packaged heat-treated meat products, soft and semi-soft cheeses.
These types of food are considered important sources of L. monocytogenes causing human illness
(Greig and Ravel, 2009; Batz et al., 2012). A total of 353 isolates from the BLS were selected for this
project according to their availability and the criteria described in the following sections. However,
most of the available BLS isolates were from fish products whereas a much smaller portion was from
meat products and cheeses (EFSA, 2013). Therefore, additional isolates from RTE meat products and
cheeses were obtained from as many different EU Member States as possible to restore the collection
of RTE food isolates to a more balanced set of the three RTE food categories. The choice of focusing

on the food categories represented in the BLS will strengthen the conclusions in relation to these
specific and highly relevant sources, but the potential of making strong conclusions on the significance
of other potential food sources is limited. A few isolates from fruits and vegetables were also selected.
It was anticipated that a substantial number of isolates from live animals would be needed to fully
represent live animals as a source in the epidemiological analyses, especially considering that the
genetic variation in the bacterial population is likely to decline with transfer to the next stage in the
food chain (environment animal production facility food human). Since the study was limited
to approximately 1,000 isolates to be sequenced within the project it was decided not to include
isolates from live animals as in source attribution strains isolated at production facility stage or in
foods are generally used (Pires et al., 2009). An additional reason to avoid the animal samples is that
it is very hard to find animal samples with proven epidemiological links to illness or contaminated
products (e.g. foods). An exception was made for samples obtained from specific fisheries.
Source attribution modelling was based on the host animal (e.g. ovine, bovine, fish etc.). Since
information is available on the animal origin of the food matrix (e.g. the origin of the milk is known,
bovine/sheep/goat, for milk and milk products) the different foods can be linked to their host animal.
Isolates from raw food sampled at production sites of meat, milk and fish production as well as
environmental isolates from such production sites (potentially persistent strains) are included in this
project (Figure 2.1 #a). It was not possible in all instances to obtain raw products and environmental
isolates taken at the same production plant. This was mainly due to raw products and end products
not being produced in a single factory as it occurs for example with cheese.
Data for the specific objective 2iii was obtained by comparing isolates from human infections and RTE
foods (Figure 2.1 #c) by looking for genes/alleles overrepresented in those from human cases
compared to RTE food as well as specific animal reservoirs as the bovine, ovine, etc. (putative
markers conferring the potential to cause disease in humans). Likewise, isolates from raw food were
compared to the isolates from potentially persistent strains and RTE food (Figure 2.1 #b) to look for
putative markers conferring a better potential to survive/multiply in the food chain. The chance of
identifying types of L. monocytogenes overrepresented in human cases compared to the presence in
the main groups of RTE food is considered reasonably high (at least giving the possibility of identifying
potential markers for human disease). Identifying markers for potential to survive/multiply in the food
chain is less straight forward due to the expected high diversity of strains entering the food chain
(e.g. (Rckerl et al., 2014)) and most likely different strains of Listeria are continuously entering a
specific facility. In this study, data is analysed on a general population level based on a sample of
isolates representing the different compartments. It is not possible, in this study, to investigate the
dynamics and circulation of strains within a specific factory or food processing chain (Figure 2.1,
dashed lines). This would require very large data sets and a large complete study involving sampling,
see e.g. (Muhterem-Uyar et al., 2015), and this is not part of the scope of the present study.
To fulfil specific objective 3, isolates from nine retrospective outbreaks, including human cases and
related food isolates, were selected to represent different sources of outbreaks in different
geographical regions. The results of the analysis of the nine epidemiologically described outbreaks
show the degree of genetic variation between strains associated with a single outbreak and give
indications on the usefulness of WGS in outbreak investigations.
The specific criteria applied for the selection of isolates are described in the following sections.

The links between groups of strains #a, #b and #c will help to fulfil specific objectives (see text).
Figure 2.1. Representation of the different categories of isolates from different stages of food chain
and isolates implicated in clinical stages selected in the study
2.2. Strains from the baseline survey (BLS)

The isolates that have been collected in the baseline survey were the target priority for sequencing.
Not all the isolates were sequenced. The selection criteria of the isolates are described in Sections
2.2.1, 2.2.2 and 2.2.3.
2.2.1. Selection criteria BLS 1. Availability of the strains
Some strains in the BLS were not sent to the EURL and therefore not available for this study. The
available strains came from 22 MSs and 1 non-MS out of the 24 participants in the BLS.
2.2.2. Selection criterion BLS 2: Isolates from same origin (food and
Member State)
2.2.2.1. Selection criterion BLS 2a: Isolates from paired samples

Analyses of L. monocytogenes were made at the end of shelf-life for all three types of the surveyed
RTE foods and, also, at the time of sampling for the fish samples. Isolates from paired fishery product
samples with the same molecular profile (PFGE) at initial sampling and at end of shelf life were usually
considered to be the same strain and therefore sequencing both isolates would not provide additional

useful data. The strains isolated from the product at the end of shelf life were the ones selected for
sequencing.
2.2.2.2. Selection criterion BLS 2b: Isolates from same origin (food and Member State)
It is also likely that the same strains are represented amongst isolates provided by the same MS from
the same type of food, e.g. in the BLS, out of a total of 12 isolates from country B fishery product
there were only 3 different PFGE profiles. This information on diversity of PFGE types amongst BSL
isolates indicated that the same product (that is coming from a unique factory) was sampled in a
unique retailer. Therefore, not all of these isolates were selected.
2.2.3. Selection criterion BLS 3. Multiple isolates from a sample (all

categories)
Although one isolate per positive food sample should have been selected, several laboratories
collected more than one isolate (2 to 6). It was decided to include different isolates that originated
from the same sample as long as typing information confirmed that these isolates were different (e.g.
based on their PFGE profile). This was the case for nine samples. Accordingly, when typing data was
unable to distinguish them, one of the isolates was selected randomly.
The selected BLS isolates consist of 353 isolates coming from 22 MSs + 1 non-MS. The 297, 49 and
7 strains were respectively isolated from RTE fishery products, meat products, and soft and semi-soft
cheeses. The complete list of isolates selected after applying selection criteria is given in Appendix 1.
2.3. Strains from other foods (OF)

The BLS provides numerous isolates for fishery products but far less for cheese and meat products. In
order to increase the number of strains from cheese and meat, eight Member States provided isolates
from these food products. The following criteria were applied to the available strains.
2.3.1. Selection criterion OF 1. Food origin
Strains from meat and dairy products were selected. Only isolates from RTE foods were considered.
2.3.2. Selection criterion OF 2. Temporal criterion
In order to match with BLS and clinical isolates, these were selected from the 2010-2012 period.
2.3.3. Selection criterion OF 3. Further selection
As more than 500 strains were available for RTE meat products and more than 300 for cheese
products after application of selection criteria OF 1 and OF 2, an additional selection was needed. For
MSs and non-MS where subtyping information (PFGE, AFLP, agglutination/molecular serotype) was
available, the most prevalent groups were selected. For the other MSs a random sampling (each
isolate has the same probability to be sampled) has been carried out. From the country Q, 12
sequenced isolates from soft and semi-soft cheese were already available.
2.3.4. Fruit and vegetables
Few strains were available from fruits and vegetables. Fruits and vegetables are usually not routinely
tested for L. monocytogenes. Five isolates from the country B that were isolated within the time frame
of the BLS were included see Appendix 3.
The category of isolates constituting "other foods" is composed of 218 isolates (including 12 already
sequenced) coming from 8 MSs (A, B, C, G, Q, V, X and Z) with a 126 and 80 isolates from RTE meat
products and cheese, respectively. In addition, 12 strains were assigned to the combined food

category. The complete list of isolates selected after applying selection criteria is given in Appendix 2.
The six isolates from fruits and vegetables are listed in Appendix 3.
2.4. Strains from food chain production stages

In order to compare isolates from both the baseline survey and clinical cases to isolates persisting or
circulating in food processing chains, we chose four sets of strains corresponding to the three types of
RTE food in the BLS. In order to match with the BLS and clinical isolates, these were preferentially
selected from the 2010-2012 period. The period was sequentially extended until the desired number
of isolates was reached. The stains are presented in Appendix 4.
The first set constitutes 62 strains isolated from pork meat cuts and meat products and from food
processing environments (between 2003 and 2014) in various regions of country C and from food
processing environments in country B (2010-2011). The list of strains is given in Appendix 4. The
selection of pork production isolates from country C includes isolates that will be sequenced in the
context of another study and made available for this study: 6 isolates from pork processing
environment (potentially persistent strains) and 2 from raw products.
The second set constitutes 21 strains isolated from raw milk and cheese and semi-soft cheese
production environments in the country B.
The third set consists of 100 strains isolated from cheese and semi-soft cheese factories in the country
Q. These isolates were already sequenced and the genome sequences were made available for this
project.
The fourth set constitutes 29 strains isolated from raw fish and smoked salmon production
environments (e.g. environmental swab samples). These strains come from the country B.
The category of food chain production isolates consists of a total of 200 isolates coming from 3 MSs
(countries B, C and Q). Of these, 142 were isolated from environmental samples. The complete list of
isolates is given in Appendix 4. Ninety-six of these isolates were already sequenced before this
project.
2.5. Strains from sporadic human clinical cases

Clinical isolates from assumed sporadic human cases were included in the study. Isolates were from
the baseline survey period, 2010-2011, and priority was given according to the criteria stated below.
It has not been possible to obtain clinical isolates from all relevant MSs, as some laboratories/MSs
were not willing to contribute to the project.
2.5.1. Selection criterion C1. Availability of the strains.
Isolates from the country A and B were directly available for the project since the national clinical
reference laboratories from these MSs were partners in the project (NPHLs of country A and B,
respectively). Furthermore, we got positive response from nine MSs that were willing to contribute
clinical isolates: country C, D, F, Q, T, W, X, Y, Z.
2.5.2. Selection criterion C2. Temporal and geographical criterion
In order to match with BLS and other food isolates, the 2010-2011 period was preferred. The
selection strategy was to encompass a wide geographical distribution. From some MSs (F, Q, W), only
a small number of isolates from these years were available and all of these were therefore included.
For MSs and one non-MS with a smaller population size (A, D, T, Y, Z), 20 isolates were selected for
sequencing, while from MSs with large populations (B, C, X) - and a correspondingly higher number of
listeriosis cases - 35 isolates were selected from each. In addition, a number of isolates that have
already been sequenced were included.
2.5.3. Selection criterion C3. Incidence

An effort was made to get not only MSs with large areas/populations represented in the project, but
also those with high incidence. According to the ECDC Annual epidemiological report 2014 - food-
and waterborne diseases and zoonoses the Nordic MSs and non-MS have among the highest
incidences in Europe (five of the six highest), and in this study three are included. These high
incidences could be an artefact of different surveillance systems in the Nordic vs other MSs and non-
MSs, but at the European level, this publication was the most reliable estimate available.
2.5.4. Selection criterion C4. Subtype information.
For MSs where subtyping information (PFGE, AFLP, agglutination/molecular serotype) was available,
the most prevalent groups were selected. For the other MSs a random sampling (each isolate has the
same probability to be sampled) has been carried out.
The clinical isolates constitute 262 isolates from 10 MSs + 1 non-MS. A total of 250 isolates were
sequenced in this project whilst whole genome sequences were already available for 16 isolates. The
complete list of isolates selected after applying selection criteria is given in Appendix 5.
2.6. Strains from outbreaks

To investigate the suitability of WGS as a tool in outbreak investigations, isolates from
epidemiologically confirmed retrospective outbreaks were selected. Isolates from nine well-described
outbreaks in three MSs + one non-MS are available for the project. Clinical isolates as well as isolates
from suspected or confirmed sources are included. From some outbreaks, all available isolates were
included. From large outbreaks, a random selection of up to 25 isolates was made.
Table 2.1: Number of human and food isolates in each of the nine outbreaks
Country Human Food Vehicle of infection
Outbreak 1 B 5 10 Beef
Outbreak 2 B 5 3 Crab meat
Outbreak 3 B 5 4 Sandwiches
Outbreak 4 B 2 2 Ox tongue
Outbreak 5 B 9 1 Unknown
Outbreak 6 T 4 1 Rakfisk
Outbreak 7 X 13 6 Foie gras
Outbreak 8 X 4 9 Cheese
Outbreak 9 C 25 0 Brie cheese
The set of outbreak isolates (Table 2.1) consists of 105 isolates from nine outbreaks; from eight of
these, both clinical and food isolates were available. The sample size per outbreak ranged from 5 to
25 isolates. Whole genome sequences were already available for 13 isolates. The remaining isolates
have been sequenced in the project. The complete list of isolates is given in Appendix 6.
2.7. Strain selection summary

A summary of the final set of strains included in the project is given in Table 2.2. In total, 676 food-
related strains were selected for sequencing within the project: BLS, other foods" and food chain
production stages. PHE contributed with 100 additional already sequenced strains from cheese
procession plants in country Q (including isolates from 12 food samples and 88 environmental
samples; listed in Appendix 4). In total, 262 assumed sporadic human clinical isolates were selected

for the study - 16 of these are already sequenced. Nine epidemiologically confirmed outbreaks were
represented by 105 isolates from either human infections or food samples related to the outbreak (13
of these were already sequenced). The strain collection was sent to PHE for sequencing in batches.
The final complete strain collection was held at PHE during the project.
Table 2.2.: Repartition of the 1,143 isolates according to country and context of isolation
Country Baseline Other foods Food production Clinical, Outbreaks Total

Study Appendix 2+3 chain sporadic Appendix 6
Appendix 1 Appendix 4 Appendix 5
A 7 29 35 71
B 4 28 68 31 43 174
C 35 83 32 35 25 210
D 4 20 24
E 6 6
F 15 8 23
G 4 4 8
H 5 5
J 10 10
K 14 14
L 54 54
M 2 2
N 9 9
P 3 4 7
Q 33 100 23 156
R 4 4
S 4 4
T 4 20 5 29
U 62 62
V 6 28 34
W 7 15 22
X 38 34 35 32 139
Y 8 20 28
Z 15 13 20 48
Total 353 223 200 262 105 1,143
Table 2.3 presents in detail the repartition of the 776 strains corresponding to Appendices 1, 2, and 3.
Table 2.3.: Repartition of the 776 food isolates according to country, food matrix and source
Country Food matrix Source Number of Total per

code isolates country
A Elaborated food products combining Mixed sources 1
several food categories
Fish and fishery products Fish 6
Meat and meat products Bovine 2


Meat and meat products Gallus gallus (fowl) 3
Meat and meat products Sheep 1
Meat and meat products Swine 8
Meat and meat products Unspecified 14
Milk and milk products Bovine 1 36
B Elaborated food products combining Mixed sources 5
Fruit, vegetables, cereals and herbs Vegetal 5
Meat and meat products Mixed animal source 3
Meat and meat products Poultry not specified 1
Milk and milk products Bovine 5
Milk and milk products Goat 1
Milk and milk products Unspecified 24 100
C Fish and fishery products Fish 31
Meat and meat products Ducks 3
Meat and meat products Mixed sources 3
Meat and meat products Poultry not specified 1
Milk and milk products Bovine 35
D Fish and fishery products Fish 3
E Fish and fishery products Fish 6 6
F Fish and fishery products Fish 15 15
G Fish and fishery products Fish 3
Meat and meat products Swine 5 8
H Fish and fishery products Fish 3
J Fish and fishery products Fish 9
Meat and meat products Gallus gallus (fowl) 1 10
K Fish and fishery products Fish 14 14
L Fish and fishery products Fish 42


M Meat and meat products Fish 1
Meat and meat products Swine 1 2
N Fish and fishery products Fish 9 9
P Fish and fishery products Fish 2
Q Fish and fishery products Fish 30
Milk and milk products Sheep 100
R Fish and fishery products Fish 1
Meat and meat products Unspecified 3 4
S Fish and fishery products Fish 4 4
T Fish and fishery products Fish 4 4
U Fish and fishery products Fish 55
Meat and meat products Turkeys 2
V Fish and fishery products Fish 6
Milk and milk products Sheep 16
W Fish and fishery products Fish 7 7
X Elaborated food products combining Mixed sources 9
Meat and meat products Ducks 1
Meat and meat products Geese 1
Meat and meat products Turkeys 1
Y Fish and fishery products Fish 8 8


Z Elaborated food products combining Mixed sources 2
Meat and meat products Unspecified 1 28
2.8. Database
All available information on isolate characteristics and associated descriptive epidemiological
information have been collected from isolate providers and organised in a database. This database
links the WGS typing data and metadata associated for each strain. The database is specific to this
project and will be used only for this study. The database was developed and managed by
Bionumerics software (version 7, Applied Maths, Sint-Martens-Latem, Belgium). Supplementary Excel
file contains all the information included in the database (Annex A).
2.8.1. Database structure
The database has a structure similar to the EURL Lm DB.

Most of the fields contain predefined pick lists to avoid errors in reporting. The food matrices and the
description of the food nature listed within the LISEQ database respected the EFSA standard sample
description code (SSD2) (Flix et al., 2014). Food product description followed three level hierarchal
information. Food products were first classified according to the food matrix type (see Figure 2.2).
Then for each food matrix, food products are listed. Finally, information on process is provided. Figure
2 describes the information fields of the project database.
The database (both in Excel and in Bionumerics format) also includes strain-typing-results extracted
from the genome analysis, such as MLST, Clonal complex, SNP address and cgMLST (only descriptive
fields are shown in Figure 2.2).

Figure 2.2.: Description of the information fields in the LISEQ-DB
2.8.2. Core information on strains
All the strains (Table 2.2) do not have the same degree of associated information. For some strains
full information is available with three levels information on food Food matrix/Food products/Process
(e.g. Meat and Meat Products/Deli products-pate/sliced) for other strains the information is less
detailed. Yet, the following core information is at least present for the food strains:
Sample type
Geographical information (at least Country)
Sampling date (at least year)
Food matrix/Food products
Food origin
For strains from sporadic human listeriosis cases, the information is limited to sampling date (at least
year) and geographical information. Food outbreak strains shared the same degree of information as
other food strains. For all the clinical strains, an additional information field contains the clinical
symptoms data with five options (bacteremia, meningitis, pregnancy related, other, unknown).
2.8.3. Data export
For each specific objective (epidemiological relationship and source attribution modelling), outputs
with the required information field were exported in appropriate format (csv, etc.).

3. Methodologies
In order to fulfil the main objective of the contract, i.e. to compare L. monocytogenes isolates
collected in the EU from RTE foods, compartments along the food chain and humans using whole
genome sequencing (WGS) analysis, a range of methodologies were employed. This Section provides
an overview of the methods and the rationale for choosing these.
The first specific objective was to carry out the molecular characterisation of a selection of
L. monocytogenes isolates from different sources, i.e. RTE foods, compartments along the food chain
(e.g. food producing animals, food processing environment), and humans employing WGS analysis. As
a first step towards this goal, DNA was extracted from the selected isolates and subjected to WGS.
The platform employed for sequencing was the Illumina HiSeq, which is the most commonly employed
cost effective, rapid method for sequencing high numbers of bacterial genomes and thus as close to a
standardized sequencing method as possible. Additionally, sequence data obtained by Illumina
sequencing contains less sequencing errors as compared to several other sequencing platforms. The
current sequencing methods rely on massive parallel sequencing, meaning that instead of sequencing
the genomes from one end to the other, each genome is fragmented and these small fragments are
sequenced simultaneously, in parallel. Thus, the sequence data produced from this type of setup
consists of millions of sequence reads per isolate genome, with each read typically of a size around
100-200 nucleotides. Subsequently, the millions of reads are pieced together thereby assembling an
almost complete genome. The procedures for sample preparation, sequencing, quality control and
assembly are described in Section 4.
The WGS data generated was analysed by different bioinformatics procedures (described in Section 4)
to explore the phylogeny and to produce data sets (typing data) that could form the basis for the
further analysis and interpretation of data for the specific objectives 2 and 3.
In order to perform molecular characterisation of isolates several gene-by-gene approaches were
employed in this project. Firstly, 7-locus MLST as defined in Ragon et al. (2008) was extracted from
the WGS data, and although now based on WGS this way of performing MLST is completely
comparable with the conventional form of MLST based on Sanger sequencing of the seven loci.
Sequence types (STs) were defined on basis of the allelic sequences of the seven loci and employed
to assign isolates to clonal complexes (CCs).
Listeria monocytogenes is a very diverse species and contains four divergent lineages with a high
number of variants, and therefore to be able to perform fine resolution phylogenies. Single Nucleotide
Polymorphism (SNP) analysis was performed separately for distinct clonal complexes, employing the
different reference genomes for each clonal complex, i.e. the best suitable for each clonal complex.
This was done in order to obtain the maximal phylogenetic resolution. In brief, SNP analysis is a
method in which phylogeny is inferred on the basis of sequence variations between isolates, across
the parts of the genome that are shared by all isolates included in the analysis, i.e. the core genome.
SNPs are assigned by comparing (mapping) sequence reads for each isolate against a common
reference genome, here a reference genome specific for each clonal complex, and subsequently
documenting the difference between any two isolates.
SNP analysis is at present the most widely used method for WGS-based discrimination for public
health surveillance and outbreak detection, but at present the method is not standardised, which
results in SNP analysis being variable between laboratories and results difficult to directly compare.
The number of SNPs is dependent on sequencing technology, sequence data quality and reference
genome employed, and even varies depending on the number of isolates included in the analysis.
Due to the SNP analysis not yet being standardised, gene-by-gene methods such as ribosomal MLST
(rMLST) and core-genome MLST (cgMLST) may be more suited to public health surveillance and
outbreak investigation of gastrointestinal bacterial pathogens, since they can be carried out in a more
standardised way and especially because the results are easier to communicate. cgMLST is a gene-by-
gene approach similar to the conventional MLST but instead of being based on seven loci it is based
on the core genome of the species, with the present cgMLST scheme including 1748 loci (Moura et al.,

2016). In cgMLST allelic variations within each of these loci make up the final type, and is used to
differentiate between clonal isolates. Typically, cgMLST (and rMLST) are performed on assembled
genome data, in contrast to SNP analysis.
Further analysis and interpretation of data for specific objectives 2 and 3 were based on the WGS
typing data sets produced under objective 1 described above. For Specific Objective 3, to perform a
retrospective analysis of outbreak strains to investigate the suitability of WGS as a tool in outbreak
investigations, isolates from nine food-borne outbreaks of human listeriosis were selected. These
outbreaks represented a range of different characteristics with respect to a number of factors such as
food source, time span, geography, number of cases, etc. For each outbreak, the relation between the
human clinical isolates, isolates available from the suspected sources as well as similar background
isolates were characterised (Section 5). For outbreak investigations, high-discriminatory methods are
desirable and thus SNP and cgMLST analyses were employed for characterising the outbreaks. The
SNP and allele differences (cgMLST) seen for epidemiologically related and un-related isolates in this
retrospective analysis of outbreaks provided valuable input to the analysis parameters used for
assessing the epidemiological relationship between L. monocytogenes isolates of human origin and
those from different sources (Objective 2). Therefore, the analysis of retrospective outbreaks
(Objective 3) is presented before the analyses related to Objective 2.
For the first goal of Specific Objective 2, to explore the genetic diversity of L. monocytogenes within
and between the different sources and human origin, different indices and measures were applied to
characterise and describe the variation of isolates in the collection (presented in Section 6). These
included diversity index, rarefaction curves and Neis genetic distance. These measures can give an
overall understanding of the diversity within and between sources/reservoirs and differences can be
statistically tested. To give useful description of diversity, diversity index and rarefaction curves must
be based on typing methods producing a limited number of distinct types in the data set. We chose
MLST (7 loci) as well as rMLST (30 loci) as these are both well-established typing methods with an
internationally recognised nomenclature. In addition, cgMLST and SNP data was also used for
assessing the genetic distance between populations.
We employed two different approached for the second goal of Specific Objective 2, to assess the
epidemiological relationship of L. monocytogenes from the different sources and of human origin
considering the genomic information and the metadata available for each isolate. The first approach
was using the method of source attribution (Section 7, i.e., partitioning of the human disease burden
of listeriosis to specific sources, where the term source includes animal reservoirs and vehicles (e.g.
foods). Because of the relatively small number of isolates, all of the isolates along the food chain that
originate from a particular reservoir were combined. This enabled the following sources of isolates and
their respective genomes to be determined: bovine, ovine, swine, fish and poultry. Human clinical
cases were attributed to these sources by utilising five different mathematical models and the
genomic typing data that was generated as described in Section 4.
The second approach to assess the possible epidemiological relationship between strains was to
identify clusters of clinical and food isolates based on SNP differences (Section 8). The WGS data was
analysed along with the epidemiological information of the food and clinical isolate to assess,
retrospectively, relationships between circulating strains of L. monocytogenes in EU within 2010-2012
period. Clusters of interest were further investigated by focusing on metadata associated to each
strain.
The third goal of Specific Objective 2 was to identify the presence of putative markers conferring the
potential to survive/multiply in the food chain and/or cause disease in humans (e.g. virulence and
antimicrobial resistance). In recent years, numerous L. monocytogenes virulence factors have been
suggested. We analysed the genomes for the presence/absence of a large set of putative virulence
genes and compared the representation of these in clinical and food chain isolates as well as the
representation according to phylogeny, i.e. lineage, clonal complex and sequence type (Section 9).
Although the majority of L. monocytogenes isolates are generally susceptible to antimicrobials, a small
proportion are found to demonstrate resistance to certain clinically relevant antimicrobials and the
antimicrobial resistance determinants have been described genetically. Similar to the virulence genes,

the presence of antimicrobial resistance genes was searched in the LISEQ collection. Some studies
have suggested that L. monocytogenes strains that are able to persist in the food production
environment are genetically distinct from transient strains that do not have this capability. Recently,
a number of genes potentially involved in persistence have been suggested. To explore this
hypothesis, we first tested the ability of WGS to differentiate potential persistent strains from other
strains collected in a cheese plant. Secondly, we compared the presence/absence of potential
persistance genes in strains isolated from food processing environment (potentially persistent
strains) to those in strains isolated in raw product (potentially non persistent, or transient strains)
(Section 9). Along with the study of putative markers involved for virulence and persistence,
investigation of potential host specific markers was carried out (Section 9).
The methods, results and conclusions of each subject and specific objective are described in detail in
the Sections 5-9. An overall conclusion is given in Section 10.
4. Sequencing and Phylogentic Analysis

This section details the methodology undertaken to produce high quality whole genome sequencing
data from the Listeria monocytogenes isolates selected in this study. To achieve the goals of specific
objective 2 we then applied a range of microbial typing methods to the sequence data providing the
framework to answer questions on genetic diversity and epidemiological relationships.
4.1. Methods
4.1.1. DNA extraction
DNA extraction of all isolates was performed at PHE using a pre-lysis procedure optimised for
L. monocytogenes followed by automated DNA extraction. In brief, bacterial growth is harvested into
a 96 deep well processing plate and treated with lysozyme at 37C for 1 h followed by Proteinase K
overnight at 56C with gentle shaking. Lysates are then heated to 95-100C for 10 minutes to ensure
any unlysed organisms are killed and enzymes are destroyed. Samples are then treated with
Ribonuclease A for 15 minutes at 37C, centrifuged and the supernatants transferred to an automated
nucleic acid extraction platform, presently the QiaSymphony. The yield and purity of extracted DNA is
assessed using the Life Technologies Quant-iT high sensitivity 96-well assay and the GloMax
Multi+ Detection and LabChip DX Systems. DNA is diluted to 10-30ng/l and submitted for whole
genome sequencing to the PHE Genomic Development and Services Unit.
The Pasteur Institute provided DNA from clinical and outbreak isolates for sequencing at PHE. The
extracted DNA complied with PHE's quality criteria (10-30ng/l and OD260/280 = 1.8-2.0).
4.1.2. DNA sequencing and validation
Paired-end libraries were generated using the Illumina Nextera XT sample preparation kit. Assessment
of fragment sizes was performed on the Perkin Elmer Labchip GX after fragmentation and clean-up.
After normalisation, samples were pooled by hand and library quantification was performed using the
KAPA library quantification kit for Illumina sequencing, on an ABI Viia7. Paired-end sequencing was
performed on the Illumina HiSeq 2500 instrument using the TruSeq Rapid SBS kit (200 cycle) and
TruSeq Paired-end rapid cluster kit. The following cycle parameters were used for sequencing: Read
1: 101, Index read 1: 8, Index read 2: 8 and Read 2: 101. RTA version 1.17.21.3 was used for
generation of base call files.
FASTQ creation and de-multiplexing via CASAVA was performed on a dedicated high performance
cluster (HPC). FASTQ reads were quality trimmed using Trimomatic (Bolger et al., 2014) with bases
removed from the trailing end that fall below a PHRED score of 30. If the read length post trimming
was less than 50 the read and its pair were discarded. If the post trimmed yield was less than
150 megabases the sample was discarded. A kmer (a short string of DNA of length k) based approach
was used (https://github.com/phe-bioinformatics/kmerid) to confirm the identity of the sample and to

ensure the sequence was free from contamination. If any non-Listeria kmers were identified in the
FASTQ reads the sample was discarded.
4.1.3. Assembly and annotation
Short reads were assembled using appropriate de novo assembly tools (e.g. SPADES). Spades
assembly (version 3.5.0) run with Kmer 21,33,55,77,83, and the only-assembler option (Bankevich et
al., 2012).
Assembled genomes were annotated in terms of protein coding features and RNA features. Prokka
software (version 1.11) (Seemann, 2014) was used to annotate all the isolates. Preselected kingdom
Bacteria and genus Listeria was performed to insure the accuracy of the annotation.
4.1.4. Gene by gene based typing
Gene by gene, or Allelic based methods have long been used to characterise microbial organisms and
provide the opportunity for a common nomenclature based on the presence of specific allele types.
The correlation of three different gene based typing methodologies with the underlying phylogeny
elucidated above was carried out.
Multi-locus Sequence Typing (MLST)
The international MLST database for L. monocytogenes is maintained by Institut Pasteur
(http://bigsdb.pasteur.fr/listeria/listeria.html). The database holds 2,944 isolates representing 739 STs
(as of 24 Apr 2015) and represents a globally representative collection from some 70 countries. More
than 80 different sample sources are included of which more than 900 isolates are human-associated
with the remainder from food, animal and environmental sources. The population genetic structure is
characterised by major clonal groups. The top five clonal groups comprise one third of the described
STs.
The MLST sequence type as defined by the Pasteur Scheme (Ragon et al., 2008) was extracted from
each sequences using MOST (https://github.com/phe-bioinformatics/MOST) (Tewolde et al., 2016)
and assigned a clonal complex in accordance with the Institut Pasteur international MLST database for
L. monocytogenes (http://bigsdb.pasteur.fr/listeria/listeria.html) designation.
Core genome MLST (cgMLST)
A cgMLST scheme extends the concept of MLST to the core loci present in majority of genomes from
a representative collection of isolates from a species/family. We employed a recently developed Lm
cgMLST method by Moura and colleagues (Moura et al., 2016) to assign an allelic designation to the
1,748 loci in the scheme for each of the genomes sequenced. The scheme was implemented using
BIGSdb (Jolley and Maiden, 2010).
Ribosomal MLST (rMLST)
Ribosomal Multilocus Sequence Typing (rMLST) is an approach that indexes variation of the 53 genes
encoding the bacterial ribosome protein subunits (rps genes) as a means of integrating microbial
taxonomy and typing. The rMLST allelic variants were extracted as a subset of the cgMLST set based
on the loci names, resulting in a total set of 30 ribosomal genes (Annex A).
4.1.5. Phylogenetic analysis
Phylogenetic methods exploiting nucleotide resolution variation (Single Nucleotide Polymorphisms

(SNPs)) between bacterial isolates can be used to elucidate the relatedness and ancestry of strains
under robust evolutionary models and provide a framework to explore the genetic diversity. SNPs
inference and analysis was performed using software developed at PHE: SnapperDB
(https://github.com/phe-bioinformatics/snapperdb), with each L.monocytogenes clonal complex
having a separate database instance. In summary short read sequences from strains selected in this

study were mapped against an appropriate reference genome of L. monocytogenes (see Table 4.1)
using BWA-MEM (Li and Durbin, 2010). The resultant sequence alignment maps (SAMs) are processed
and high quality variants (MQ>30, AD>0.9, DP>10) extracted using GATK2 (McKenna et al., 2010).
Variants and uncertain positions are stored for further analysis. Maximum likelihood phylogenies were
produced using RAxML-8.17 (Stamatakis, 2014) using the GTRGAMMA model
(https://github.com/lguillier/LISEQ-codes#chapter-4).
Recombination allows for rapid introduction of new genetic material between strains and such
evolutionary events can impact on the ancestral inference provided by phylogenetic methods. Gubbins
(Croucher et al., 2014) was be used to identify recombinant regions of the genome within clonal
groups. These positions can then be filtered as appropriate when inferring the ancestry of strains.
These robust phylogenetic representations of the population structure of strains provided the
framework to assess the diversity of L. monocytogenes within and between the different sources and
human origin at the lineage, clonal complex and strain level resolution.
Table 4.1.: Complete genome sequences that were used as reference genomes for each
corresponding clonal complex
Clonal L. monocytogenes Genbank

Complex strain Accession
CC1 F2365 AE017262
CC2 J1-220 CP006047
CC3 R2-502 CP006594
CC4 Clip80459 FM242711
CC6 J1816 CP006046
CC7 10403S CP002002
CC8 88-0478 CP006862
CC9 SLCC2479 FR733649
CC5 J2-064 CP006592
CC11 J0161 CP002001
CC14 SLCC7179 FR733650
CC69 HCC23 CP001175
CC121 6179 HG813249
CC131 SLCC2376 FR733651
CC155 1998 CP002004
CC361 J2-031 CP006593

Samples are sequenced and checked if they are contaminant free and have a FASTQ yield greater than 150 MB. The FASTQ
reads are assembled and submitted to the Pasteur BIGSDB instance where cgMLST and rMLST allele calls assigned. The 7 locus
MLST is extracted using MOST and if unambiguous the reads are mapped to the appropriate clonal complex specific reference
genome and a SNP address assigned. A summary report containing the metrics and results of this process is produced.
Samples are sequenced and checked they are contaminant free and have a FASTQ yield greater than 150 MB.
Figure 4.1: Schematic diagram of sequence analysis pipeline
4.1.6. Genetic distance: SNP address
As new strains are added to SnapperDB they are compared to the database and a distance matrix
maintained of all pairwise SNP distances. The distance represents those positions that differ between
a pair of isolates with respect to the reference genome. Single linkage clustering of genetic distance is
an effective method of describing phylogenetic groups as it is inclusive of clonal expansion events.
Using hierarchical single linkage clustering of the pairwise SNP distances we are able to define a
nomenclature that corresponds to the tree structure from deep branches through to clades through to
identical strains. This enables an isolate level nomenclature to be derived for each genome sequence
allowing efficient searching of the population studied as well as automated cluster detection. Single
linkage clustering was performed at 7 descending SNP distance thresholds (250, 100, 50, 25,
10, 5, 0) which generates a 7 digit address where each number represents a unique cluster at
that threshold.
4.1.7. Genetic distance: SNP address
As new strains are added to SnapperDB they are compared to the database and a distance matrix
maintained of all pairwise SNP distances. The distance represents those positions that differ between
a pair of isolates with respect to the reference genome. Single linkage clustering of genetic distance is
an effective method of describing phylogenetic groups as it is inclusive of clonal expansion events.

Using hierarchical single linkage clustering of the pairwise SNP distances we are able to define a
nomenclature that corresponds to the tree structure from deep branches through to clades through to
identical strains. This enables an isolate level nomenclature to be derived for each genome sequence
allowing efficient searching of the population studied as well as automated cluster detection. Single
linkage clustering was performed at 7 descending SNP distance thresholds (250, 100, 50, 25,
10, 5, 0) which generates a 7 digit address where each number represents a unique cluster at
that threshold.
The red rods represent bacterial genomes and the distance between them represents genetic distance. In this example we have
a three levels of clustering for similarity, the outer blue rings corresponds to 10 SNP differences, the inner yellow rings
corresponds to 5 SNP differences and the inner black rings represent 0 SNP differences. If two samples are identical they have
the same SNP address, i.e. two isolates that have the address 1.1.1 have 0 SNP differences between them. The isolates with
the addresses 1.2.2 and 1.2.3 have a different final digit indicating that they are not identical but both have matching 1st and
2nd digits so are within the same 10 SNP cluster and the same 5 SNP cluster. When we compare the isolate 1.2.2 against 2.3.5,
as the first digit is different, they are different by greater than 10 SNPs.
Figure 4.2.: Schematic diagram representing hierarchical clustering
4.2. Results
4.2.1. Processing of isolates
PHE received 1,048 L. monocytogenes isolates from the ANSES, the SSI, University of Aberdeen and
PHE. In addition, 108 isolates were previously sequenced (Italian cheese and cheese factories, n= 100
and UK outbreak, n=8). All isolates received by PHE were sequenced as described. Sequence data for
13 isolates were not accepted for further analysis due to poor quality or unresolvable strain
contamination.
The final dataset is composed of 1,143 sequences with a high quality of sequencing.

4.2.2. Gene by gene based typing: MLST and clonal complex assignment
From 1,143 isolates sequenced, 42 different CC and 13 singleton STs (unassigned CC) were identified.
One isolate could not be assigned to any ST or CC. 10 clonal complexes accounted for 70% of the
samples (Table 4.2, Figures 4.3 and 4.4). The population structure of the isolates in the study are
further described as a minimum spanning tree (Figure 4.5).
Table 4.2.: The clonal complexes identified and the number of isolates by isolation context as defined
in Section 2 and listed in the strain selection appendices
Clonal Lineage RTE food Food chain Clinical, Outbreak- Total

Complex Appendix 1-3 processing sporadic related
Appendix 4 Appendix 5 Appendix 6
CC121 II 144 37 6 0 187

CC9 II 81 15 14 0 110
CC8 II 69 5 24 0 98
CC1 I 10 4 50 8 72
CC2 I 19 29 20 0 68
CC101 II 10 41 16 0 67
CC6 I 30 3 28 0 61
CC155 II 32 1 8 13 54
CC7 II 16 4 16 8 44
CC14 II 13 2 9 13 37
CC4 I 1 1 10 24 36
CC87 I 10 0 4 19 33
CC31 II 24 7 1 0 32
CC3 I 18 7 6 0 31
CC37 II 9 15 5 0 29
CC204 II 17 3 1 0 21
CC59 I 10 0 4 4 18
CC5 I 7 6 4 0 17
CC21 II 13 0 2 0 15
CC20 II 8 2 2 0 12
CC415 II 0 2 0 9 11
CC18 II 0 6 4 0 10
Minor CCs LI=31 35 10 28 7 80
LII=48
L.innocua=1
Total 576 200 262 105 1,143
Note: Minor CCs included CC398, CC11, CC193, CC224, CC403, CC54, CC177, CC19, CC220, CC29, CC77, CC217, CC26, CC379,
CC207, CC218, CC388, CC475, CC88, CC89, ST184, ST200, ST32, ST382, ST392, ST560, ST570, ST602, ST736, ST773, ST839
(ordered according to occurrence).

Figure 4.3.: Distribution of clonal complexes in ready-to-eat food (isolates in Appendices 1, 2 and 3)
and from sporadic human clinical infections (isolates in Appendix 5)
Figure 4.4.: Distribution of clonal complexes from the three major food product categories (isolates
in Appendices 1-3)

Each circle represents a single sequence type (ST) that is numbered on the tree. Clonal complexes (CC) defined by single locus
variants are shaded in grey. The number of loci that differ between STs is labelled on the branches.
Figure 4.5.: Minimum spanning tree of the isolates included in this study as described by 7 locus
MLST
4.2.3. Developing the Framework for Phylogenetic Analysis
L. monocytogenes contains 4 divergent lineages with greater than 100,000 variants across the
population. This study contained isolates from lineage I & II and the population structure based on
whole genome SNPs is displayed in the phylogenetic tree in Figure 4.6. From the phylogeny it can be
seen that there is a clear delineation between lineages and clonal complexs within lineages.

Figure 4.6.: Whole genome SNP maximum likelihood phylogeny of L. monocytogenes genome
sequences with the clades annotated by 7 loci MLST Clonal Complex (CC) generated using
Parsnp (https://github.com/marbl/parsnp) with EGE-e (NC_003210.1) as the reference
genome

In order to elucidate the fine phylogenetic relations between isolates and to develop strain level
nomenclature for further analysis, strains were further analysed within each clonal complex using the
clonal complex specific reference mapping approach described in 4.1.5. Isolates not belonging to a
clonal complex were not assigned the SNP address nomenclature.
4.2.4. Phylogenetic analysis of major clonal complexes
Clonal Complex 121

This clonal complex has the most number of isolates from this study with a 187 in total, with 165 food
product isolates, 16 from the food-processing environment and 6 human clinical.
Figure 4.7.: Whole genome maximum likelihood phylogeny constructed using RAxML with the
GTRGAMMA model of Clonal Complex 121. Clinical cases are coloured red and non-clinical isolates
coloured blue. Taxa are labelled by strain identity and SNP address. High-resolution figure is available
in full size at https://github.com/lguillier/LISEQ-codes/blob/master/Chapter4/Chap4-5Trees.7z

Clonal Complex 9
Clonal complex 9 contains 110 isolates, 13 from food processing environment, 83 from food products,
14 from human clinical.
GTRGAMMA model of Clonal Complex 9. Clinical cases are coloured red and non-clinical
isolates coloured blue. Taxa are labelled by strain identity and SNP address. High-
resolution figure is available in full size at https://github.com/lguillier/LISEQ-
codes/blob/master/Chapter4/Chap4-5Trees.7z

Clonal Complex 8
The second most populated cluster is clonal complex 8. This group contains 98 isolates, 4 from food
processing environment, 70 from food products and 24 from human clinical.

Clonal Complex 101

Clonal complex 101 contains 67 strains, 35 are from food processing environment, 16 from food
products and 16 from human clinical. There is currently no complete genome from the CC101 complex
and therefore the isolate with the best assembly was used as a reference genome.
resolution figure is available available in full size at https://github.com/lguillier/LISEQ-

Clonal Complex 4
This clonal complex of 36 isolates is mainly composed of human clinical isolates (n=34) and 2 food
products. It has been show as a hypervirulent cluster by a recent publication (Maury et al., 2016).
4.3. Conclusion
The phylogenetic analysis on four large CCs (CC8, CC9, CC101 and C121) reveals that within a CC,
clinical strains are not associated to a specific clade of the tree. The CC4 phylogeny confirms recent
results that this CC is associated to clinical strains and is probably one of the most virulent (Maury et
al., 2016).
Whole genome sequencing has allowed us define the population of Listeria monocytogenes from this
study to an unprecedented resolution. It has provided the framework to answer questions on genetic

diversity in the different sources assayed in this strain collection as well as to explore possible
epidemiological links between isolates.
5. Retrospective analysis of outbreaks

Many of the traditional typing methods have inherent problems; the discriminatory power of
phenotypic methods is too low (serotyping) or suffer from biologic variability (phagetyping) (Graves et
al., 2007). Molecular methods such as PFGE have a much better discriminatory power, but are
laborious and require a lot of work for creating comparable results between different laboratories. The
most used L. monocytogenes PFGE protocol is created in the PulseNet organisation (Graves and
Swaminathan, 2001), and the protocol specifies up to three different enzymes in order to get
sufficient discriminatory power for secure outbreak detection meaning that the method is both
expensive, time consuming and laborious (Tourdjman et al., 2014).
The outbreaks in this LISEQ project were defined by each submitter based on epidemiology and their
local choice of molecular typing (e.g., PFGE and fAFLP).
WGS gives us the possibility to analyse the genome at close to the endpoint resolution of DNA typing,
and the new challenge is no longer to chase the maximum resolution, but to find out what the actual
variation within a listeriosis outbreak is. Besides the true biological variation, there will be variations in
the results dependent on both lab work and analysis approach.
As described in Section 4, there are two main approaches to analyse the relationship between
genomic sequences (SNP and gene by gene). Both approaches are in active use in reference
laboratories in Europe for outbreak detection and/or investigations. Therefore, we chose to use and
compare the two analysis approaches in order to properly address the specific objective 3 - to
perform a retrospective analysis of outbreak strains (i.e. using a subset of epidemiologically linked
human and food isolates) to investigate the suitability of WGS as a tool in outbreak investigations.
All of the outbreaks were national outbreaks without a previously known international component.
The results from this section were fed back to other parts of the project where epidemiological
relationships are reported.
5.1. Methods
In eight of the nine analysed outbreaks, isolates from both human cases and linked food isolates were
provided. The sequences from each outbreak were analysed together with all other isolates of the
same clonal complex in the study regardless of epidemiological relationship to the outbreak. The
diversity of isolates within outbreaks were explored using SNP and gene-by-gene based methods, thus
validating the thresholds for cluster definitions.
The SNP analysis was made as described previously in Section 4, using ParSNP with a complete or
high-coverage reference genome from the same clonal complex. The distance matrices shown are
created with pairwise comparisons with pairwise deletion of ambiguous positions. The SNP analyses
were visualized with maximum parsimony trees since this is a good, simple algorithm for relatively
closely relates isolates, resulting in a tree with branches of proportional lengths to the number of SNP.
The maximum parsimony trees were created based on the non-ambiguous positions for all isolates in
the clonal complex.
For the cgMLST the 1748 loci Pasteur scheme (Moura et al., 2016) was used and the analysis results
were kindly provided by Alexandra Moura and Sylvain Brisse using an initial de novo assembly using
CLC assembly cell from Qiagen Bioinformatics. cgMLST results were visualized using minimum
spanning trees since this is a fast algorithm allowing short-term divergence and micro-evolution in
populations to be reconstructed based upon sampled data.

Project isolates with the same CC, but with unknown epidemiological link to the outbreaks were
included in the analysis to disclose possible additional cases and/or sources in same country as well as
other EU countries.
5.2. Results
The outbreaks are presented individually in Sections 5.2.1 to 5.2.9. Each section contains an
epidemiological description of the outbreak, with SNP- and cgMLST-based analysis consisting of a tree
of the outbreak isolates + other isolates with the same CC. The accompanying matrices describe the
pairwise number of differences between the cluster isolates only. The SNP trees are maximum
parsimony trees and the cgMLST trees are minimum spanning trees.
The SNP and cgMLST analyses of the outbreak isolates (Table 5.1) show good concordance between
the methods and generally the same level of SNP and allele differences were found. The exception is
cluster #1 where the maximum number of allele differences is much higher than the number of
different SNPs as described further in Section 5.2.1.
Table 5.1.: Descriptive data with CC, time period and pairwise genetic distances within each outbreak
Outbreak number #1 #2 #3 #4 #5 #6 #7 #8 #9
Clonal
CC155 CC1 CC7 CC59 CC415 CC398 CC87 ST14 CC4
complex
Years 2012-13 2007-13 2013-14 2009-11 2013-14 2013 2013-14 2012 2012
All n
isolates human/ 13/41 55/17 24/20 6/12 9/3 8/1 17/16 6/16 34/2
within CC food
SNP Median 111 174 258.5 90 3 38 38 214 126
cgMLST Median 56 71 59 46 4 19,5 20 23 54
SNP Max 174 259 1368 243 93 65 74 337 183
cgMLST Max 118 119 131 119 50 36 41 47 88

n
Outbreak
human/ 5/8 5/3 4/4 2/2 8/1 4/1 13/6 4/9 24/0
isolates
food
SNP Median 0 10,5 2 9 3 0 5 2 0
cgMLST Median 1 7,5 3 7 2 0 3 2 2
SNP Mode 0 16 1 9 3 0 5 0 0
cgMLST Mode 0 5 3 10 2 0 3 1 2
SNP Max 2 21 4 12 4 1 8 8 2
cgMLST Max 51 16 6 10 4 1 7 8 4
Note: In the rows describing SNP analysis the numbers refer to the number of SNP and in the rows describing cgMLST, the
numbers indicate the number of allele differences. The table is grouped into two main parts, each headed by the tan
coloured rows. The top part includes the outbreak isolates plus all other isolates in the LISEQ project belonging to the
same CC. The bottom part concerns the outbreak isolates only. Each column corresponds to one outbreak and for each
outbreak (or outbreak plus other isolates in the same CC), median and maximal pairwise distances are shown for both
categories. For the outbreak only category the mode distance is also included. In outbreak 8 belonging to CC14 the
numbers shown are for the ST instead of the CC as the defining unit since this CC is polyphyletic. In this table, the food
category also includes some environmental isolates.
5.2.1. Outbreak 1 CC155
An outbreak of 4 cases in the North West of country B within a 3-week period with the same rare
molecular type (1/2a, XI.23). Three of the cases had a history of consuming pressed beef (a sliced
meat product consisting of meat and other ingredients moulded and set in gelatine) purchased from
different retailers but made by a single producer and a third had purchased raw meat from a butcher's
shop also supplied with cooked meat products from the same producer. Isolates of the same

molecular profile were detected in food and environmental samples from 3 retail premises where 3 of
the cases purchased pressed beef. In 2013, over one year later, a further case was identified of this
particular rare type who had eaten tongue from a retailer who also sold pressed beef from the same
producer as in the 2012 outbreak. L. monocytogenes of the same type as the recent case and the
2012 cases was isolated from pressed beef from the retailer and from pressed beef from the
producer.
The outbreak isolates form a tight cluster with a maximum pairwise distance of 2 SNP. This supports
the epidemiological investigation with a single source hypothesis. None of the 42 other isolates
belonging to CC155 in our data set were clustering together with the outbreak.
This is the only outbreak where there is a large difference between the SNP and cgMLST approaches.
Analysing the three isolates that differed showed that these had the lowest quality in the de novo
assemblies, indicated by a high number of contigs and low N50. Inspection of the specific positions
casing the variant allele calls, showed that the assembler had made erroneous decisions and that this
was the cause of the discrepancy.
0
0 0
1 0 1
0 0 0 1
0 0 0 1 0
0 0 0 1 0 0
0 0 0 1 0 0 0
0 0 0 1 0 0 0 0
0 0 0 1 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 00
1 1 1 2 1 1 1 1 1 111
1 1 1 2 1 1 1 1 1 1110
The tree is constructed using all isolates having the same CC as the outbreak isolates. To the right is shown a pairwise SNP
distance matrix of the outbreak isolates. Number on branches indicates the number of SNP differences the branch corresponds
to.
Figure 5.1.: SNP maximum parsimony tree of outbreak 1

0
2 1
0 0 0
1 1 1 0
1 1 1 0 0
2 1 2 0 0 0
1 0 0 0 0 0 0
1 1 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0
7 6 6 6 6 6 6 6 6 6
14 14 14 14 14 13 14 14 14 14 14
45 45 45 45 45 45 45 44 46 44 43 51
The tree is constructed using all isolates having the same CC as the outbreak isolates. To the right is shown a pairwise distance
matrix of the outbreak isolates indicating the number of allele differences. Number on branches indicates the number of allele
differences the branch corresponds to. Note the two food isolates diverging significantly from the other outbreak isolates.
Figure 5.2.: cgMLST minimum spanning tree
In 2013, a cluster of cases with the same type (serogroup 4, fAFLP profile V.3) was identified over a
6 month period associated with the consumption of crabmeat from a producer in the North of country
B. The food production premises had a history of contamination with L. monocytogenes of the same
type since 2011. All 3 cases reported a history of eating shellfish including Crustacea. A further
12 clinical cases were identified as having this profile since 2011, with 6 reporting consumption of
shellfish and 3 specifically reporting crabmeat. Many cases reported purchasing shellfish from mobile
vendors and the implicated crabmeat producer supplied mobile vendors. Sampling of crabmeat and
the environment at the producer's premises in response to the investigation led to the isolation of
L. monocytogenes of the same fAFLP profile as the cases. WGS was performed to determine the
genetic relatedness of the isolates sharing this particular profile including historical isolates from cases
from 2007 and 2010.
The SNP analysis showed that the maximum pairwise distance between outbreak isolates were 21 SNP
with a median of 10.5. This fits well with the long time span of the outbreak and the fact that the food
production premises had repeatedly problems with L. monocytogenes contamination. It would be
likely that the facility had a house infection with a population that diversified over the years.
There is one isolate in the LISEQ collection that was submitted as a sporadic clinical case that clusters
together with the outbreak. This was a human isolate from 2011 and from the same country as
outbreak, making this a possible missing case.

19
12 13
13 14 3
16 17 6 7
16 17 6 7 0
20 21 10 11 4 4
16 17 6 7 0 0 4
to.

7
5 5
5 6 0
7 5 0 0
9 8 2 2 2
9 8 7 8 8 10
12 14 14 14 14 16 12
differences the branch corresponds to.
Four cases in a region of country B contracted listeriosis due to rare fAFLP type (1/2a, XIV.52b); one
in May 2013 and 3 in Jan 2014. All cases were males between ages 54-67 years and all hospital
inpatients. Review of PHE food and human fAFLP database detected isolates of same rare type in
sandwiches at 3 hospitals in same geographical region as cases and from a local retailer. Only one
case reported eating hospital sandwiches. Cases were at different hospitals and only one case was at
hospital where the same type was found in sandwiches; this case did not report eating hospital
sandwiches. Subsequently during the investigation by country B, WGS showed that isolates from
4 cases, and from sandwiches at 2 hospitals and a local retailer all grouped within same 5 SNP cluster
and also identified a 5 case.
The WGS analysis shows that these isolates are very closely related, which matches the profile of a
point source outbreak within a short time frame. With the exception of one isolate, the cluster is well
separated from the other 37 CC7-isolates in the study. One other isolates in the collection clustered
relatively close (8 SNP / 5 alleles) to the outbreak isolates. This isolate is a human sporadic case from
the same country as the outbreak, but being isolated in 2010 it is not a temporal match.

4
2 2
3 3 1
1 3 12
1 3 120
1 3 1200
to.
4
3 2
4 3 0
4 4 1 1
3 3 1 2 2
3 3 1 2 2 2
6 4 1 3 4 3 2
Ox tongue incident in country B that only involved a single confirmed case. The incident was
published as a case of foodborne listeriosis linked to a contaminated food-production process (Lamden
et al., 2013). The additional case used in this outbreak analysis is a tentatively linked case reported
over 2 years prior to the published case.

An elderly immunosuppressed male with listeriosis reported regular consumption of various cooked
meats including ox tongue from a specific local food outlet. L. monocytogenes from the case was
shown to be the same type as was isolated at high levels in opened and unopened packs of ox tongue
at the same outlet. Thus, contamination was likely to have originated at the original meat producer
although sampling at the producer's facilities yielded negative results. However, surveillance data
showed that the same type of L. monocytogenes was detected 6 months earlier from an ox tongue
sample from another food outlet who used the same meat producer as the food outlet associated with
case.
The WGS analysis shows that the human isolates are almost identical, while the food isolates are
9-12 SNP distant.
9
3 12
9 2 12
to.

5
10 7
10 7 2
The outbreak MLST type ST394 belongs to CC415, for which there yet is no complete genome for
reference mapping. The SNP analysis therefore used a de novo assembled reference from ST394. All
isolates are originating from country B. Eight isolates from humans associated with the outbreak were
sequenced. One human isolate that originally was thought to be part of outbreak 5 (as listed in
appendix 6) was shown to be of a different CC and the isolate was therefore not included in the
outbreak analysis. One isolate from raw milk identified as having the same molecular typing profile by
fAFLP but for which there was no epidemiological evidence of being linked to the outbreak were also
sequenced. The initial SNP analysis showed that the raw milk isolate was quite distinct from the
human isolates with over 1,000 SNPs difference, while the distance between the isolates from the
eight cases were at most 4 SNPs. Closer inspection of the SNPs distribution showed that in the raw
milk isolate, almost all SNPs were located within a phage in the L. monocytogenes genome. All human
isolates had the same phage in this region, which was different but related to the phage in the raw
milk isolate. On initial inspection, the difference in the number of SNPs between the raw milk and
human isolates indicated that the milk isolate was not associated with the outbreak. Re-analysing the
sequences for SNPs after removing these phage regions showed that the human isolates still clustered
together, but the raw milk was now 30 SNP from the cluster. This illustrates one of the potential
pitfalls with a whole genome analysis approach where several SNP/alleles can be acquired all in one
event, and thus may not necessarily mean that strains are genetically unrelated. This is an important
point that should be taken into consideration when using SNP analysis for investigating the genetic
relatedness of strains i.e. where SNPs have occurred and not just the total number. In this instance,
however, on re-analysis, the raw milk isolate was still 30 SNPs different to the human isolates and
unlikely to be the cause of the outbreak.

4
2 4
3 1 3
3 1 3 0
3 1 3 00
3 1 3 000
0 4 2 3333
to. The food outbreak isolate is the one from raw milk, which was eventually discarded in the countrys investigation phase as
not being part of the outbreak (see text above).
1
3 2
3 2 3
2 2 3 1
2 1 2 1 0
2 4 3 1 2 0
2 1 3 1 0 0 0
This outbreak was identified as a point source cluster with the rare MLST ST-802/CC-398 in country T.
The source was epidemiologically identified as fermented fish; something that were confirmed by
isolation of the same rare ST from food. We do not yet have a common reference strain for mapping
analysis within CC398. We therefore used a de novo assembled genome from the outbreak itself as a
reference. All five isolates in the outbreak were already sequenced by the national Institute of Public
Health in country T, and were only analysed in this project. The SNP analysis of the four human and

one food isolates showed that there were only a single SNP found between all the isolates and hence
a max distance of 1 SNP and the mode 0 SNP. This corroborates the single point outbreak
epidemiology.
1
1 0
1 0 0
1 0 0 0
to.
0
0 0
0 0 0
1 1 1 1
This is analysis of two ST87 reported outbreaks between January 2013 February 2014 from a
specific region of country X (Perez-Trallero et al., 2014). ST87 belongs to the relatively rare human
serotype 1/2b. The outbreaks were unusual in that there were a high proportion of pregnancy relates
cases (7 out of 12 human cases in the one analysed here). The two described ST87 outbreaks (that
were temporally separated) were separated by a single band shift in PFGE AscI analysis (SmaI was
identical).

There are two duplicates in this study; one with isolates from both mother and daughter, and one
with both blood and CSF isolates from the same patient. These were 1 and 0 SNP apart respectively in
the SNP analysis. In the cgMLST analysis the pairs were both identical.
WGS analysis shows that the two described outbreaks cannot be separated neither by SNP, nor
cgMLST analysis. The pairwise SNP distances showed an unusual pattern with a bell shaped
distribution centred around 4-7 SNP (Figure 5.13). With the information that ST87 was common in
that geographical region throughout this year, with two described clusters and some sporadic cases, it
could be hypothesized that the original source was somewhere with a L. monocytogenes population
that have had the time to diverge for a relatively long time. The epidemiological investigation could
not pinpoint the sources more closely than that, the first outbreak seemed to be related to ham, while
in the second one the outbreak type were found in foie gras from the refrigerator of one case.
Our analysis found that two more human isolates could be assigned to the outbreak, both from
country X in 2011 - 2 years before the outbreak.

5
5 6
5 6 6
3 4 4 4
3 8 8 8 6
2 7 7 7 5 1
4 5 5 5 3 7 6
4 5 5 5 3 7 6 4
3 4 4 4 2 6 5 3 3
5 6 6 6 4 8 7 5 5 4
5 6 6 6 4 8 7 5 5 4 0
3 4 4 4 2 6 5 3 3 2 4 4
4 5 5 5 3 7 6 4 4 3 5 5 3
2 7 7 7 5 5 4 6 6 5 7 7 5 6
5 6 6 4 4 8 7 5 5 4 6 6 4 5 7
2 7 7 7 5 5 4 6 6 5 7 7 5 6 07
2 7 7 7 5 5 4 6 6 5 7 7 5 6 070
5 6 6 4 4 8 7 5 5 4 6 6 4 5 7077
The tree is constructed using all isolates having the same CC as the outbreak isolates. To the bottom right is shown a pairwise
SNP distance matrix of the outbreak isolates. Number on branches indicates the number of SNP differences the branch
corresponds to. Bottom left is a diagram of the paired SNP distances shown in the matrix bottom right.

2
1 1
1 1
2 3 1 1
2 2 1 1 2
3 3 2 2 3 3
3 3 2 2 3 3 4
3 3 2 2 3 3 4 0
2 2 1 1 2 2 3 1 1
3 3 2 2 3 3 4 4 4 2
3 3 2 2 4 3 4 4 4 2 4
3 3 2 2 3 3 4 4 4 2 4 0
2 2 1 1 3 2 3 3 3 1 3 2 1
4 4 3 3 4 4 5 5 5 3 5 3 3 2
4 4 3 3 4 5 5 5 5 4 5 3 3 2 0
4 5 3 3 4 4 5 5 5 5 5 3 3 2 0 0
4 4 3 3 5 4 5 5 5 4 5 6 5 5 6 6 6
5 4 4 3 4 4 5 4 4 2 4 4 5 3 5 7 6 2
5.2.8. Outbreak 8 ST14
This outbreak from 2012 caused by fresh cheese (de Castro et al., 2012) consisted of a total of
11 cases, and it was geographically spread over several regions of country X. The epidemiology
showed that 8 out of the 11 cases had eaten different varieties of a brand of cheese. The remaining
three cases were newborn or stillborn. Isolates were retrieved from four of the cases. The outbreak
had a clear source identified in the fresh cheese by epidemiology. The point source epidemiology was
confirmed by the WGS analysis that showed a tight cluster and that the food isolates were
intermingled with the human isolates. From the rest of the ST14 isolates in this study, one more
human isolate (submitted as a sporadic case) clustered with the outbreak isolates. This isolate was
from the same country as the outbreak, but predates the cluster by over 2 years.

Within CC14 we found three different ST. These STs are named on the separating branches. The diversity in CC14 is huge so
the analysis will just focus on the ST14 group, to which outbreak 8 belongs.
Figure 5.15.: SNP maximum parsimony tree of all CC14 isolates in the LISEQ study
7
5 2
7 4 2
6 3 1 3
5 2 0 2 1
7 4 2 4 3 2
5 2 0 2 1 0 2
5 2 0 2 1 0 2 0
5 2 0 2 1 0 2 0 0
5 2 0 2 1 0 2 0 00
8 7 5 7 6 5 7 5 555
5 2 0 2 1 0 2 0 0005
The tree is constructed using all isolates having the same ST as the outbreak isolates. To the right is shown a pairwise SNP
to.

1
2 2
1 1 1
1 0 1 1
1 1 2 1 0
1 0 1 0 0 0
1 2 2 2 0 0 0
2 1 1 1 1 0 10 1
2 2 2 1 1 1 11 11 1
0 0 0
3 2 4 3 2 2 21 12 1 31 3
5 5 6 6 4 4 52 14 1 61 26 5
1 1 1 1 2 1
6 5 6 6 5 5 51 05 1 60 16 2 70 8
2 0 2 0 1 2 1 0
2 1 1 1 2 2 1 1 1
2 2 3 2 3 4 3 2 2 3
5 4 6 5 5 6 6 4 4 6 5
5 5 6 5 6 6 6 5 5 6 7 8
The outbreak isolates cluster closely together and have the food isolates mixed in. This solidifies that
the correct source was identified.
An outbreak in country C with an epidemiologically identified source in brie cheese (Tourdjman et al.,
2014). Unfortunately, the bacterium was not isolated from the cheese samples, so the sequenced
isolates are all human. The 25 isolates submitted were a mix of actual outbreak isolates and
background isolates from the same country and period. In the original PFGE typing all the submitted
isolates were identical using both restriction enzymes AscI and ApaI. The third PFGE enzyme SmaI
could properly separate the 11 outbreak isolates from the background.
The WGS analysis confirmed the SmaI separation, but in a much more convincing way. Long branches
separate the outbreak isolates from all the isolates where only AscI and ApaI were identical.

0
0 0
0 0 0
0 0 0 0
0 0 0 0 0
0 0 0 0 0 0
1 1 1 1 1 1 1
0 0 0 0 0 0 01
0 0 0 0 0 0 010
1 1 1 1 1 1 1211
to. The isolates defined as the outbreak cluster by triple enzyme PFGE are marked with the red ellipse.
1
3 2
4 3 3
3 2 2 1
3 3 3 1 1
4 2 4 2 2 2
2 1 2 2 2 2 2
2 2 2 2 1 1 2 0
2 2 2 2 2 3 2 1 1
3 2 3 3 2 2 3 1 1 1

5.3. Conclusions
The cgMLST and SNP analyses showed concordant results with similar numbers (SNP and allele
differences) when the outbreak isolates were analysed (Table 5.1). In the following discussion the
numbers in the following conclusion will refer to SNP differences, but aside from the 2-3 deviant
cgMLST results in outbreak 1 (Section 5.2.1) the same conclusions apply for cgMLST. Most of the
outbreaks are tightly clustered; six out of nine show a typical point-source-like pattern with a median
pairwise distance of 5 SNP and a maximum pairwise distance 10 SNP.
The other three outbreaks each have their own specific profiles. Cluster 7 shows an unusual pattern
where the isolates have distances between 4 and 7 SNP. In an ideal world, it would be desirable to
use a fixed cut-off threshold for cluster definition, but in the real world, this is not possible. Using a
cut-off of 5 SNP in cluster 7 would fail to define the outbreak properly, but a cut-off of 10 would
include all of the confirmed isolates. It can be noted that country X saw an increase of this type for a
prolonged period. The type was found in diverse geographical locations and there were also small
variations found in the PFGE patterns. The presence of the outbreak isolate in an extended spatial and
a temporal space is congruent with a larger variation in the population. This is also confirmed by the
variations in WGS.
Outbreak 2 and 4 both show somewhat higher pairwise SNP distances: median of 10.5 and 9, with
maximums of 21 and 12, respectively. It should be noted that both of these clusters are occurring
during relatively long time spans. If using a single linkage model, the longest branch needed to link all
the cluster isolates would be 12 and 9 SNP, respectively.
Outbreak 5 showed the impact that the auxiliary genome can have, unless the proper reference
genomes are used. It should be noted that in the analysis of all these outbreaks, the focus was on the
core SNP and core cgMLST. The auxiliary genome can also be very valuable (Wang et al., 2015), but
currently the core analysis is the most stable and also the one amenable to tracking the isolates in
time.
The story told by the WGS analyses is reflected in the epidemiological descriptions (see individual
outbreak descriptions). Higher diversity is often linked to temporally extended outbreaks, even though
the cause of variation seen in outbreak 7 only can be speculated about.
No food isolates in the study collection, apart from those already described as part of the outbreaks,
were similar to any of the nine outbreak types. In four of the outbreaks, one or two human isolates
submitted as sporadic cases clustered together with the outbreak isolates. In all four instances, these
isolates originated from the same country as the outbreak. The isolates we have sequenced in this
project obviously is a subset of all the existing L. monocytogenes types found in food and humans in
Europe and in the world. Although the LISEQ data set was designed to give a good coverage of the
European situation in the years 2010-2011, there were not identified any international components in
these epidemiologically confirmed outbreaks. This does not mean that international outbreaks do not
occur, just that that the nine specific outbreak strains were not found among the study isolates from
other countries.
The cgMLST analysis showed for the most part very concordant results with the SNP analysis. In
8 outbreaks the median and maximum sizes of branches within a whole CC was shorter (Table 5.1)
compared to SNP branches. This is to be expected since several SNP within the same loci only results
in a single allele difference. Between the outbreak isolates where the genetic distance is a lot shorter
the differences are a lot smaller and there is no clear trend that either cgMLST or SNP results in a
higher resolution. The exception here is outbreak 1, which is described in Section 5.2.1. We have used
current state of the art bioinformatic methods, but since the two approaches use different
computational methods for defining the differences in number of SNP or number of alleles, the results
cannot be expected to be identical. In theory the number of pairwise SNP differences should always
be higher (you can fit several SNP into one allele but not vice versa) but since the methods have
computational limitations in this young research field there are still misassembles (causing false allele
calls), erroneous base calls (resulting in false SNP calls) etc. There is still a need for the scientific

community to perform targeted studies on the correlation between the methods, but both methods
work well for defining clusters in outbreaks.
The WGS analysis managed to clearly separate the outbreak isolates from the background, so WGS is
very well suited for clearly defining outbreaks. The results also indicate that every outbreak should be
considered in its own context and that one should not use a single universal cut off value for
separating an outbreak from background isolates. Outbreaks extended in time such as number 2 and
4 had up to 21 SNP or 16 cgMLST loci between individual isolates. More clonal outbreaks, such as
number 6 and 9 only had a maximum of 2 SNP/4 cgMLST alleles between any isolate, even though
outbreak 9 was the largest outbreak analysed (n=24). In the nine outbreaks analysed here, there
were no issues separating the outbreak isolates from the background, even with a fixed cut of value if
that value is set high enough. Nevertheless, with increasing number of isolates analysed, the problem
of a fixed cut-off will be exacerbated.
6. Genetic diversity
Section 4 illustrates that L. monocytogenes contains a large number of variants. The extent of this
variation, or diversity, may differ by source reservoir or from humans as has been exemplified for
Campylobacter (Strachan et al., 2013). This genetic diversity can be characterised in a number of
ways including by Simpsons diversity index (Simpson, 1949) and rarefaction (Heck et al., 1975). All of
the different microbial typing methods used earlier in this report can be utilised. However, for
pragmatic reasons only 7 locus MLST or 30 locus rMLST are practical because cgMLST and SNP based
methods would produce too many types which would not really provide useful information about
variation (practically every genome would yield a different type).
The primary aim of this section is to address Specific objective 2 part (i) - to explore the genetic
diversity of L. monocytogenes within and between different sources and human origin. This section
also investigates the genetic distance between each source by Neis genetic distance (Nei, 1975). This
methodology provides information on the genetic relatedness of isolates between sources and in
particular whether sources have distinctive populations of L. monocytogenes.
6.1. Methods
6.1.1. Simpson's Diversity Index
Simpson's Diversity Index was used to obtain an estimate of the diversity of strains by source
(Simpson, 1949),
= 1 ( )2
where fi is the relative frequency of ST i in a specific source. A value of 0 of the diversity index
indicates that all strains are the same and a value of 1 indicates that they are all different (maximum
diversity). Confidence intervals were calculated by the bootstrap method using the Pop Tools add-in
for Microsoft Excel (www.poptools.org) and significant differences between pairs of sources were
calculated using a randomisation test (Manly, 2007).
To generate bootstrapped confidence intervals the STs were resampled with replacement. This was
done 10,000 times using Pop Tools. From this the mean values of Simpsons diversity index were
calculated as well as 95% confidence intervals.
To calculate significant differences between each pair of populations (e.g. human and bovine) these
were randomized in Excel using PopTools and Simpsons diversity index calculated for each. Correction
for differences in sample size between the sources was carried out using the following method. For
each pair of sources to be compared, the one with the lowest number of isolates (Ilow) was kept

constant whilst the other was resampled without replacement for Ilow isolates. This process was then
repeated 10,000 times using the Monte Carlo Excel add-in @RISK (Palisade Ivybridge, United
Kingdom). The posterior distribution of Simpsons diversity indices was then compared with the non-
randomised diversity index to obtain the level of significance (P value).
6.1.2. Rarefaction
The extent to which the isolates from sources had sampled the maximum number of genotypes was
characterised using rarefaction. Rarefaction is a data re-sampling technique that indicates whether all
of the genotypes have been sampled which results in the curve reaching a plateau or if the curve is
still increasing there are still more genotypes in the population to be sampled (Heck et al., 1975).
Statistical significance was determined by randomisation test as described above.
6.1.3. Nei's genetic distance
Standardized genetic distances (d1) between pairs of sources were determined using the method of
Nei (Nei, 1975) and applied to genetic locus and SNP data by the method of Manly (Manly, 2007).
Briefly, for Nloci the distance is calculated as
1
1 = 0.5
,
Where pi and qi are the frequencies of the alleles (or SNPs) at each locus in source i and j,
respectively.
A Neis value of 0 indicates that the populations are identical whilst a value of 1 indicates that the two
populations have no genotypes in common. Statistical significance was determined by randomisation
test as described in 6.1.1.
6.1.4. Graphical visualisation and cluster analysis
A phylogenetic tree utilising the 50,297 SNPs generated by Parsnp (Treangen et al., 2014) was
generated by MEGA (Tamura et al., 2013) utilising the nearest-neighbour joining technique.
L. innocua was used to root the tree. This enabled visualisation of isolates around the phylogenetic
tree.
To determine whether there was clustering of isolates from each source on the phylogenetic tree the
following analysis was performed. Within each source pairwise SNP distances were calculated. The
percentage of pairwise SNP distances less than a cut-off vale (100 SNPs) was calculated. The
percentage of pairwise SNP distances between isolates outside the source was then calculated using
the same cut-off criterion as previously. If the percentage between source isolates was greater than
the percentage between outside isolates, this was taken as evidence of clustering.
6.1.5. Analyses
Table 6.1 describes the list of analyses that were performed for each of the different methods and
also the level of molecular analysis. The cgMLST for all the study isolates except for Listeria innocua
were defined with the help of the Institute Pasteur as described in Section 4. The scheme applied was
composed of 1,748 genes representing 125,029 alleles. From this 7 locus MLST and 30 locus rMLST
(30 rMLST loci are utilised in the Institute Pasteur cgMLST scheme) data were obtained. SNPs
detection was performed using the Parsnp software for all of the genomes in the database resulting in
39,529 core genome SNPs (cgSNP).

Table 6.1.: Diversity, rarefaction and genetic distance analyses carried out by level of molecular
analysis
Level of molecular Simpson's Diversity Rarefaction Nei's genetic

analysis distance
7 locus MLST
30 locus rMLST
1,748 cgMLST na na
39,529 cgSNP na na
na not applicable because the large number of loci results in practically every isolate being unique.
6.1.6. Selection of Genomes for analysis suitable for genetic diversity and
source attribution analysis
Table 6.2 provides details of the number of human genomes and also the number that have cgMLST
profiles but are not part of an outbreak. In a previous source attribution analysis (Little et al., 2010)
human cases had been separated into two groups (one younger than 60 years and the other greater
or equal to 60 years). Of the human data that were not part of an outbreak and had age data there
were 50 in the <60 years age group with 121 being in the older age group. To determine whether
there is a difference in age stratification (See Appendix 7) Nei's genetic distance was calculated from
the 7 locus MLST data. No significant difference was found using a randomisation test (P=0.141). In
addition Simpson's diversity index and rarefaction were carried out but no significant differences were
found between the two groups. Hence, all of the human data were analysed as one dataset as the
results show that there is no difference in diversity and genetic relatedness by age group.
Table 6.2 also shows the number of genomes that were allocated to particular sources. The
designation of source did not depend on the part of the food chain from which the isolates originated.
For example, genomes allocated to fish would include those from a fish sampled at a fish farm, all the
way along the food chain, to those sampled at retail. The mixed category primarily comprises complex
foods made of a number of ingredients such as sandwiches etc. A number of the sources were
represented by a small number of genomes, which were insufficient for the analysis of diversity. It
was decided to only consider those distinct sources with 25 genomes available for analysis. This cut-
off was based by work done on Campylobacter (Smid et al., 2013), where they advised >25 isolates
should be used per source.
Table 6.2.: Numbers of genomes categorised to source and the subset which were not part of an
outbreak and for which cg MLST data were available
Human and source Number of Genomes Number of genomes with 7 locus

MLST and not part of an outbreak
Human* 333 261
Mixed 30 27
Poultry* 32 25
Bovine* 80 61
Shellfish 3 0
Swine* 114 112
Fish* 325 324
Unspecified 101 101
Vegetable 5 5
Ovine* 117 89
Caprine 3 3
Total 1,143 1,011
*denotes used in source attribution comprising 872 genomes. The other sources were not included as they comprise
<25 genomes or were mixed/unspecified source.

6.2. Results and Discussion

6.2.1. 7 locus MLST
Simpson's diversity index was determined for isolates from humans and each of the sources (Figure
6.1. (a)). All exhibited high diversity ranging from 0.897 (Humans) to 0.811 (ovine). However,
pairwise comparisons between each pair of sources yielded no significant differences (P>0.05).
Rarefaction was also carried out (Figure 6.1. (b)) and it can be seen that bovine and human isolates
have a higher number of new STs per isolate sampled compared with fish and ovine. Swine is
intermediate and poultry has too few isolates to determine a trend. It is also noticeable that the
rarefaction curves have not plateaued and that if a larger sample had been obtained then additional
novel genotypes would have been found.

(a)
(b)
Figure 6.1.: (a) Simpsons diversity index with 95% bootstrapped confidence intervals, and
(b) rarefaction of the 7 locus MLST human and source genome data (dashed lines show
95% bootstrapped confidence intervals)
Neis genetic distance was determined between isolates from humans and the 5 sources (Figure 6.2).
The genetic distance was significantly different between human and all of the other 5 sources
(P<0.05). Bovine had the closest genetic distance to human. All but two of the other pairwise
comparisons (i.e. those that did not involve humans) also showed significant differences. The two
which were statistically similar (P>0.05) involved poultry, which was the source present in lowest
numbers and it may be that this could be an artefact due to small sample size.

(a) (b)
(c) (d)
(e) (f)
Confidence intervals (95%) were generated by the bootstrap method. Asterisks denote significant pairwise difference between
each pair of sources (P<0.05) using randomisation test described in 6.1.1.
Figure 6.2.: Pairwise Neis genetic distance by 7 locus MLST between (a) humans, (b) fish, (c) swine,
(d) ovine, (e) bovine, (f) and poultry and the remaining sources
6.2.2. 30 locus rMLST
Simpsons diversity index was again high for the 30 locus rMLST ranging from 0.91 in humans to 0.83
in ovine (Figure 6.3). No significant differences between any of the sources was observed (P>0.05) as
was the case for 7 locus MLST.
Rarefaction by 30 locus rMLST comprises more genotypes than for 7 locus MLST (e.g. for humans
MLST provides 52 genotypes whilst rMLST has 73). Hence, the rMLST curves are steeper than for

MLST and again the curves do not plateau indicating that increasing sample size will generate
additional novel genotypes. For rMLST isolates from humans have the highest number of new STs per
isolate sampled as was the case for 7 locus MLST. However, bovine which had a similar curve to
human in 7 locus MLST is now more similar to the other sources.
Neis genetic distance shows that the L. monocytogenes population in humans is distinct compared to
that in the other source reservoirs (P<0.05) (Figure 6.4.(a)). However, again as in 7 locus MLST, the
distance between humans and bovine is the smallest. All of the comparisons between sources (Figure
6.4. (b)-(e)) all show significant pairwise differences (P<0.05).

(a)
1
Simpson's diversity index
0.8
0.6
0.4
0.2
Source
(b)
80
Human_all
70 Fish
60 Swine
Ovine
Number of new rSTs
50
Bovine
40 Poultry
30
20
10
0
0 50 100 150 200 250 300 350
Number of Genomes
Figure 6.3.: (a) Simpsons diversity index with 95% bootstrapped confidence intervals and (b)
rarefaction of the 30 locus rMLST human and source genome data using randomisation
test described in 6.1.1.

(a) (b)
Nei's genetic distance

0.6 0.6
0.4 * 0.4
* *
*
* *
* *
0.2 0.2 *
*
0
0
(c) (d)

0.6 0.6
0.4 0.4
* * * *
0.2 * * * * 0.2 * *
0 0
(e) (f)
0.6 0.6
0.4 0.4 *
* * *
* * *
0.2 0.2 * *
*
0 0
Confidence intervals (95%) were generated by the bootstrap method. Asterisks denote significant pairwise difference between
each pair of sources (P<0.05) using randomisation test described in 6.1.1.
Figure 6.4.: Pairwise Neis genetic distance by 7 locus MLST between (a) humans, (b) fish, (c) swine,
(d) ovine, (e) bovine, (f) and poultry and the remaining sources
6.2.3. 1,748 locus cgMLST
Figure 6.5 shows Neis genetic distance, comparing humans with the source reservoirs, for 1,748 locus
core genome MLST. There is a significant difference (P<0.05) in the distance between humans and all
of the sources. However, bovine has the smallest distance to human. However, bovine has the

smallest distance to human. It was not possible to perform comparisons between the source
reservoirs within the timeframe of the project.
*
0.6
* *
*
*
0.4
0.2
Asterisks denote significant pairwise difference between each pair of sources (P<0.05) using randomisation test described in
6.1.1.
Figure 6.5.: Pairwise Neis genetic distance by 1748 locus cgMLST between (a) humans and the
remaining sources
6.2.4. 39,529 cgSNP
Neis distance again shows significant differences between Humans and all of the animal reservoirs
and that the smallest distance is between humans and bovine (Figure 6.6). Comparisons between
sources were not carried out because of the long computational times required.
0.6
0.4
*
0.2 * *
* *
Confidence intervals (95%) were generated by the bootstrap method. Asterisk denote significant pairwise difference between
each pair of sources (P<0.05).
Figure 6.6.: Pairwise Neis genetic distance by 39,529 locus cgSNP between (a) humans, (b) fish,
(c) swine, (d) ovine, (e) bovine, (f) and poultry and the remaining sources
6.2.5. Graphical Visualisation
Figure 6.7 shows the phylogeny of the L. monocytogenes by source and appears to show that there is
a non-even distribution of isolates by source. For example there appear to be more human isolates
(red) in lineage I whilst there are more fish isolates (blue) in lineage II. Also, within the lineages by
visual inspection there appears that there may be clustering of sources.

Scale: L. innocua to Lineage I is 31200 SNPs. The tree is drawn to scale using the Neighbor-Joining method (See 6.1.4).
Figure 6.7.: SNP based neighbour joining tree of L. monocytogenes rooted with L. innocua (fish
blue, swine pink, ovine green, bovine brown, poultry yellow, human red and
other - white).
Figure 6.8 shows the SNP differences between all of the 872 isolates. The peaks to the right (large
SNP differences) show the differences between isolates from different lineages. There is a peak of
7.9% for pairwise comparisons <100 SNPs. This was used as the cut-off in the further analysis to
establish clustering of isolates within the phylogeny.

Here the number of SNP differences between each pair of isolates was calculated and plotted on the graph.
Figure 6.8.: Pairwise SNP difference comparisons between all 872 isolates
Table 6.3 shows the clustering analysis that was performed. The mean and median values of pairwise
SNP distance within each source are not in themselves that helpful because representatives of each
source occur across different lineages. The SNP distance between lineages is large and the relative
distribution of a source across lineages is the main driver for the size of the mean and median values.
The 95 percentiles are very broad because isolates from humans and the different sources are spread
across and within the different lineages. The final column in Table 6.3 shows the percentage of
pairwise comparisons with a cut-off of 100 SNPs. Pairwise comparisons between all isolates
(irrespective of source) indicates that 8.2% are within 100 SNPs of each other. Pairwise comparisons
within human (8.6%) and bovine (7.3%) isolates are in the same range as all and hence do not
appear to show evidence of clustering at this level. Whereas poultry (21.0%), fish (16.1%) and ovine
(16.0%) appear to all show evidence of clustering.
Table 6.3.: Pairwise SNP differences within each source
Source Mean SNP difference Median SNP % less than 100 SNPs
(95% CI) difference
Human 8,655 (0 16,190) 15,290 8.6
Fish 4,781 (0 16,072) 2,155 16.1
Swine 7,407 (0 16,192) 2,333 11.7
Ovine 7,217 (0 16,104) 2,661 16.0
Bovine 8,344 (0 16,121) 2,733 7.3
Poultry 6,849 (0 15,997) 2,153 21.0
All 7,638 (0 16,140) 2,538 8.2
CI: confidence interval; SNP: single nucleotid polymorphism.
Note: The mean, median and 95% confidence intervals are presented. Also, the percentages of pairwise comparisons within
each source that have <100 SNP difference are also given.

6.3. Conclusions
Simpsons index for humans and the 5 sources exhibited high diversity (>0.8) for both 7 locus MLST
and 30 locus rMLST. Simpsons index of diversity between each of the sources was indistinguishable.
Rarefaction showed that for both 7 locus MLST and 30 locus rMLST that all of the genotypes had not
been sampled. However, for the sources with the largest number of samples, it does show that the
number of types from humans is considerably greater than that for fish. For both 7 and 30 locus MLST
humans have the steepest slope, so have the largest number of genotypes per isolate sampled.
Bovine is second, virtually indistinguishable from humans at 7 locus MLST but lower at 30 locus MLST.
Neis genetic distance showed that there were significant differences between human and all sources
at all levels of molecular analysis explored. Also, for all levels the distance between humans and
bovine was the smallest. Computing times became long when dealing with analysis of cgMLST and
cgSNSs. As a result calculation of Neis genetic distance between sources and associated confidence
intervals and randomisation tests were not carried out.
Visualising the SNP based phylogeny tree appears to show some areas of clustering by source, though
there are many parts of the tree which are quite heterogeneous. When investigating this analytically
and comparing pairwise SNP differences with a cut-off of 100 SNPs there was no evidence that there
were no independent host clusters of bovine and human compared with chance whereas there was
evidence to show that this occurred for a number of the other sources (poultry, fish and ovine).
Whether this is a robust finding or an artefact of the sampling for this study can only be resolved
when additional isolates become available to see if the pattern continues or otherwise. If it is
generally found that there are parts of the phylogeny where there are clusters (<100 SNPs)
comprising a particular source, then this indicates that they are closely related.
It should be noted that isolates for each source came from different points in the food chain. Those
isolates obtained from sources closest to retail are likely to have had greater chance of being a result
of cross contamination from another source. It was not possible to investigate this in the current
study because there were insufficient data. However, this should be borne in mind when considering
the robustness of the results. Future work should investigate the effect that isolates from different
points (animal, factory, retail, human) along the food chain may (or may not) make to the analysis. It
was not possible to do this here because there were insufficient isolates to perform this type of
analysis.
7. Epidemiological relationship: Source attribution

The current section addresses specific objective 2(ii): Assess the epidemiological relationship of
L. monocytogenes from the different sources and of human origin considering the genomic
information and the metadata available for each isolate. It achieves this using the method of source
attribution.
The term source attribution has been defined (Pires et al., 2009) as: the partitioning of the
human disease burden of one or more foodborne infections to specific source, where the term source
includes animal reservoirs and vehicles (e.g. foods).
Attribution can be carried out at different points along the food chain (Pires et al., 2009). This can
include at production, distribution and consumption. In the current project because of the relatively
small number of isolates, all of the isolates along the food chain that originate from a particular
reservoir are combined (see Section 6). This enables the following sources of isolates and their
respective genomes to be determined: bovine, ovine, swine, fish and poultry (see Section 6). Human
clinical cases are then attributed to these sources by comparing the genotypic subtypes from the
human and source isolates. The microbial subtyping approach involves characterization of isolates of a

specific pathogen by genotypic subtyping methods (e.g., MLST, cgMLST, cgSNPs etc). These data can
then be used to perform source attribution utilising mathematical models (Mughini-Gras and van Pelt,
2014).
7.1. Methods
7.1.1. Source Attribution Methods
Availability of models
Appendix 8 provides the links to all of the attribution model programs used in this study.
Dutch Model
The Dutch model (Mughini-Gras and van Pelt, 2014) is a straight forward way to estimate the
attribution of a particular genotype (e.g. ST) to a reservoir, when the frequency distribution of each
type is known for each reservoir. If pij represents the frequency of type i (e.g. ST 19) in source j (e.g.
poultry) then the attribution score of type i in source j is given by
where the summation by j considers all the reservoirs where data exist (e.g. cattle, sheep, wild birds,
chicken, turkey etc.).
When applied at ST level this model does not guarantee that all STs will be attributed to sources. This
is because human types that are not found in the animal reservoir cannot be attributed. However, if
genetic information exists at multiple loci, as in this study, then the Dutch Model can make use of the
frequency of each individual allele at each individual locus, and estimate attribution even for STs that
are not present in the animal reservoirs. In particular, at allele level the frequencies aijk can be
p
calculated for each allele a ijk of all isolates from the animal reservoirs, where i is subtype, j source
and k the loci number.
The attribution score of bacterial subtype i in source j is
where
p aijk = BetaInv (0.5,0 + 1, N isolates + 1) if its frequency is zero (BetaInv fn in Excel). This
p
assumes that we have no prior knowledge of aijk and so is maximally noncommittal or conservative.
Sample size correction and confidence intervals: Since the sample sizes of the sources are different a
correction is incorporated. If the sample size of the smallest source is Nmin then the Dutch model is
run by sampling without replacement of Nmin isolates from each source (e.g. Nmin=25, which is the
sample size of poultry reservoir or 61 (sample size of cattle) when poultry are discarded from
analysis). This process is repeated for 10,000 iterations.
After each iteration the attribution scores of each human isolate to each source are re-calculated using
the above equation. These scores are then averaged across the number of isolates (e.g. n = 254 for
humans) and stored. The mean, standard error and 95% confidence intervals of the attribution scores
over the 10,000 iterations are then calculated.

Applicability to level of molecular analysis: This model can be readily applied to ST, 7-locus MLST,
rMLST and cgMLST. The method can also be applied to cgSNPs but there can become implementation
problems at large numbers of SNPs. In the current project an implementation of 15,000 of the
39,529 SNPs was achieved.
Hald Model
This model was developed in Denmark for the attribution of human salmonellosis (Hald et al., 2004).
This Danish Salmonella source attribution model uses a Bayesian framework with Markov Chain
Monte Carlo simulation to attribute sporadic laboratory-confirmed human Salmonella infections caused
by different Salmonella subtypes as a function of the prevalence of these subtypes in animal and food
sources and the amount of each food source consumed. The model takes into account the uncertainty
for all these factors and also includes travel as a possible risk factor.
This model was improved by (Mullner et al., 2009) to include the introduction of uncertainty in the
estimates of source prevalence and an improved strategy for identifiability and is called the Modified
Hald Model. This is the model that is used here and does not include information on amount of food
consumed as is the case for the Dutch model.
In summary, the modified Hald model achieves source attribution by comparing the frequencies of
human infections caused by different pathogenic subtypes (e.g. serotypes for Salmonella (Mullner et
al., 2009)), with the subtype frequencies found in the different sources accounting for potential
subtype- and source-dependent characteristics, that may influence their chance to cause human
illness (Hald et al., 2004)).
The model utilises a Bayesian approach to estimate and quantify the uncertainty of the parameters.
Briefly,
where oi is the observed number of human infections caused by subtype i that is assumed to be
generated by a Poisson probability distribution, whose mean parameter is given by the summation
over sources of individual ij, which are the Poisson parameters for each subtype i in source j and
are given by
ij ~ pij qi aj,
where pij is the prevalence of subtype i in source j, qi is the subtype-dependent factor, which
putatively accounts for differences in survivability, virulence and pathogenicity for subtype i, and aj is
the source-dependent factor, which putatively accounts for the ability of source j to act as a vehicle of
listeriosis.
The attribution score to each source j is calculated as follows
=
=1
=
=
=1
=
=1
=
=1
,
where I is the number of subtypes and N the number of sources.

According to Mullner et al. (Mullner et al., 2009) the following default priors were used for the above
mentioned factors.

(a) Source dependent factor

aj~dexp(0.002)
(b) Genotype dependent factor

log(qi)~Normal(0,),
where is given by a fairly diffuse Gamma(0,0.01,0.01) distribution.

(c) Prevalence
The priors for the prevalence (pij) were chosen to be independent beta distributions,
pij ~dbeta(ij, ij),
where the parameters ij and ij were determined form the posterior distributions of a separate
Bayesian analysis of the prevalence data, for each source j and subtype i (Mullner et al., 2009;
Mughini-Gras and van Pelt, 2014) (see prevalence sub-model below).
Posterior distributions of the attribution proportions Propj in each source j were obtained by a Markov
Chain Monte Carlo simulation implemented in WinBUGS1.4 (http://www.mrc-
bsu.cam.ac.uk/software/bugs/). Five independent Markov chains were run, each using 30,000
iterations (10,000 burn-ins). This was sufficient to provide convergence using the method developed
by Gelman and Rubin (Gelman and Rubin, 1992).
Prevalence sub-model:
Briefly, the prevalence was modelled as

p1ij ~ j rij,
where j ~ dbeta(1,1) is the overall prevalence of subtypes in source j and rij is the relative frequency
of genotype i in source j, which is given by
1
1 , 2 , ,1 ~ 1 , 2 ,,
=1 .
Here Xij represents the number of isolates of genotype i in source j (Mughini-Gras & van Pelt, 2014).
The mean values and the standard deviations of the posterior distributions of p1ij were used to
calculate ij and ij (the parameters of the beta distribution used in the main model) as follows
1
=
1
,
1
= .
1
.
Sample size correction and confidence intervals: Sample size correction was not implemented in this
model. The summary statistics (mean, standard deviation, median and confidence intervals denoted
as 2.5% & 97.5% percentiles) of the attribution proportions were obtained from the posterior
distributions of .
Applicability to level of molecular analysis: This model is only implemented at ST level.

STRUCTURE
This is a Bayesian clustering model designed to infer population STRUCTURE and to attribute
individuals to population groups (Pritchard et al., 2000). The program has been used successfully for
7 locus Campylobacter MLST genotyping data (Strachan et al., 2013). Each isolate is attributed on the
basis of a training dataset consisting of isolates from known populations (i.e. set USEPOPINFO to 1).
The algorithm calculates the frequency of each particular sequence type in each population. Based on
these frequencies, the probability of an isolate (e.g. a human isolate) belonging to a population group
(e.g. source includes fish, bovine, ovine, poultry, swine etc.) is calculated. This is repeated
10,000 times using the Markov Chain Monte Carlo process with 1,000 burn-in steps.
Sample size correction and confidence intervals: The model appears to have no sample size correction
within it and it was not possible to implement this either within or outwith the model. The mean and
confidence intervals of the scores were calculated as in the Dutch model.
Applicability to level of molecular analysis: This model was implemented at 7 locus MLST, rMLST and
1,748 -cgMLST. However, it was not possible to carry this out for cgSNPs because of the computation
time required.
The Asymmetric Island (AI) Model
This source attribution model incorporates a Bayesian approach and uses the allelic profile of the
sequence subtypes to reconstruct the genealogical history of the isolates (Wilson et al., 2008). The
host populations are considered to exist on separate islands (e.g. the sheep island). Mutations and
recombination occur on each island. Migrations from between each reservoir (island) into the human
population are used to estimate the degree of attribution to each source. This model has previously
been applied to Campylobacter 7 locus MLST data from England (Wilson et al., 2008), Scotland
(Sheppard et al., 2009) and New Zealand (Mullner et al., 2009).
The Asymmetric Island model assigns each human case to the potential source populations on the
basis of DNA sequence similarity. It does this by encoding the DNA sequence data for each locus as
an allele. By comparing human isolates to a panel of reference sequences of known source (e.g.
cattle, sheep, chickens, pigs, wild birds and turkey), each human case can be assigned a probability of
originating in each source population (i.e. an attribution score). The source attribution probabilities
are calculated using a statistical model of the way the DNA sequences evolve in the populations of
bacteria. In the statistical model, there are parameters representing the processes of mutation, DNA
exchange between bacteria (recombination or horizontal gene transfer) and zoonotic transmission
between populations. These processes lead to differences in gene frequencies between the source
populations, facilitating source attribution. This model also uses a MCMC process which was conducted
for 100,000 iterations, with the output file written once every 50 iterations. A symmetric Dirichlet (1)
prior is used on the proportion of human isolates attributed to sources, in which all sources are
considered equally likely a priori (Wilson et al., 2008).
Sample size correction and confidence intervals: The model appears to have no sample size correction
within it and it was not possible to implement this either within or outwith the model. The mean and
confidence intervals of the scores were calculated as in the Dutch and STRUCTURE models.
Applicability to level of molecular analysis: This model was implemented at 7 locus MLST and rMLST.
The program fails to work beyond 250 loci and so higher level analysis was not possible.
The Aberdeen Model
In this method, attribution is based on the similarity between human isolates to isolates from different
sources (e.g. fish, bovine, ovine, etc.). An isolate is attributed to the reservoir which has the
maximum number of similar loci or SNPs. This is simply calculated by summing the number of loci that
are identical. Hence, each of the 254 human isolates used in the study are allocated to a source. For
example if 30 are allocated to ovine then the attribution score to the ovine source is 30/254 = 0.12.
Sample size correction and confidence intervals: sample size was carried out as in the Dutch model as
was generation of mean, standard error and 95% confidence intervals for the attribution scores.

Applicability to level of molecular analysis: This model can be readily applied to ST, 7-locus MLST,
rMLST and cgMLST and 39,529 SNPs.
7.1.2. Self-Attribution
Self-attribution is a key performance measure for the source attribution models (Sheppard et al.,
2009). This is the average percentage accuracy that any given isolate from a source can be correctly
attributed back to its own source reservoir (e.g. the likelihood that the attribution model will assign an
ovine isolate back to ovine). This can be performed in a number of ways. Here, the attribution is
carried out and then all of the source isolates are re-introduced blind to the models and their scores to
each source are determined. Average, standard error and confidence intervals are calculated as done
for standard source attribution described above. This was carried out for all of the attribution models.
7.1.3. Analyses
Table 7.1 specifies the analyses that were performed based on what was possible to implement with
the models. Source attribution analysis was performed for the 5 main sources (fish, swine, ovine,
bovine and poultry) as described in Section 6, Table 6.2. However, since poultry was only represented
by 25 isolates and the rarefaction results in Section 6 indicate that this only represents a limited
proportion of poultry genotypes (at both 7 locus MLST and 30 locus rMLST) additional attribution
analysis was carried out with the remaining 4 sources.
Table 7.1.: Source attribution models performed according to level of molecular analysis
Number of Level of molecular analysis (number STRUCTURE Dutch Asymmetric Hald Aberdeen
sources of loci) Island
5 sources ST(1) nd nd nd nd
MLST(7) np
rMLST(30) np
cgMLST(1748) np np
cgSNP(15,000 Dutch, 39,529 np np np
Aberdeen)
4 sources ST(1) nd nd nd nd nd
(excluding MLST(7) np
poultry) rMLST(30) np
cgMLST(1748) np np
cgSNP(15,000 Dutch, 39,529 np np np
Aberdeen)
nd not done; np not possible due to software being inoperable above a certain number of loci.
7.2. Results and Discussion

7.2.1. Source Attribution of 5 sources
Single locus ST level

The results of the source attribution model at the level of single locus ST are presented in Figure 7.1.
(a). Bovine appears to be the main source that is attributed to human disease (38%). However, the
confidence intervals in the model are very large indicating that it is difficult to determine which source
is likely to be most important in terms of human infection. It may be that the relatively small sample
size of the sources could play a role in this large uncertainty and/or the fact that single locus ST may
not be a sufficient discriminating factor. Self-attribution (Sheppard et al., 2009) was used to
determine the accuracy of the Hald model Figure 7.1. (b). On average the model was correct 50% of
the time, but the problem of large confidence intervals persists, as for the attribution of human
isolates.

Only the Hald model was conducted at single locus ST because the software code that has been
developed only operates at a single locus. One disadvantage of this model is that if an ST occurs in
humans but not in the source reservoirs then it cannot be included in the analysis. There is the
potential to develop the Hald model for multiple loci. For example carrying out the analysis
independently one locus at a time and then aggregating the results. This would however be
computationally intensive and would require all the MCMC chains to converge for each individual
locus. All of the other models could be performed at single locus MLST also but it was decided not to
do this as it is already generally known that 7 locus MLST has improved performance.
(a) (b)
Hald - self attribution

80
100
Hald
Self attribution (%)

60 80
Attribution(%)
60 FISH
40
SWINE
40
OVINE
20
20 BOVINE
POULTRY
0 0
Fish Swine Ovine Bovine Poultry FISH SWINE OVINE BOVINE POULTRY
Source Source
Figure 7.1.: (a) Source attribution of human cases and (b) Self-attribution, using single locus ST data
and the Hald model (CIs = 95 percentiles)
7-Locus MLST
Self-attribution (Sheppard et al., 2009) was used to determine the accuracy of the three source
attribution methods using the 7 locus MLST data (Figure 7.2.). On average Asymmetric Island model
performed best, being correct 80% of the time, with STRUCTURE model next (45.9%) and Aberdeen
and Dutch at 41%. This should be compared with what would be expected by chance which is 20%
(i.e. 1 isolate partitioned to one of 5 sources). It is worth noting that the confidence intervals indicate
that the Asymmetric Island has the smallest whilst the Aberdeen model has the largest.

(a) (b)
STRUCTURE - self attribution Dutch - self attribution

100 100

80 80
60 FISH 60 FISH
SWINE SWINE
40 40
OVINE OVINE
20 BOVINE 20 BOVINE
POULTRY POULTRY
0 0
FISH SWINE OVINE BOVINE POULTRY FISH SWINE OVINE BOVINE POULTRY
Source Source
(c) (d)
AI - self attribution Aberdeen - self attribution

100 100

80 80
60 FISH 60 FISH
SWINE SWINE
40 40
OVINE OVINE
20 BOVINE 20 BOVINE
POULTRY POULTRY
0 0
Source Source
Figure 7.2.: Self attribution of 7 locus MLST data utilising (a) STRUCTURE, (b) Dutch, (c) Asymmetric
Island and (d) Aberdeen models (error bars denote 95% confidence intervals)
Source attribution was then carried out using human data (Figure 7.3.). All four models indicated that
the most likely source was bovine (38-64%) whilst the remaining of human isolates were shared
across the other sources. The AI model confidence intervals are again the smallest and has the
highest attribution to bovine (64%) compared to all of the other models.

80 Structure
Dutch
60 Asymetric Island
Attribution(%)
Aberdeen
40
20
0
Fish Swine Ovine Bovine Poultry
Source
Figure 7.3.: Source attribution of human cases with 7 locus MLST data utilising STRUCTURE, Dutch,
Asymmetric Island and Aberdeen models (error bars denote 95% confidence intervals)
30-locus rMLST
The self-attribution for 30-locus rMLST (Figure 7.4.) performs best for AI (80% whilst the other
models give similar levels of performance (44% STRUCTURE, 43% Aberdeen and 36% Dutch). Again,
the confidence intervals for the AI model are the smallest. The source attribution to human data
(Figure 7.5) suggests that for all 4 models the most likely source was bovine (41-59%) whilst the
remainder of human cases was shared across the other sources with poultry possibly being the
lowest. The error bars are again very small for the AI model and it attributes the highest number of
cases, of the four models, to bovine (59%). Since the self-attribution for AI is much higher than the
other models it is likely that the higher attribution to bovine is credible.

(a) (b)
Structure - self attribution Dutch - self attribution

100 100

80 80
60 FISH 60 FISH
SWINE SWINE
40 40
OVINE OVINE
20 BOVINE 20 BOVINE
POULTRY POULTRY
0 0
Source Source
(c) (d)

100 100

80 80
60 FISH 60 FISH
SWINE SWINE
40 40
OVINE OVINE
20 BOVINE 20 BOVINE
POULTRY POULTRY
0 0
Source Source
Figure 7.4.: Self attribution of 30 locus rMLST data utilising (a) STRUCTURE, (b) Dutch,
(c) Asymmetric Island and (d) Aberdeen models (error bars denote 95% confidence
intervals)
80 Structure
Dutch
60 Asymetric Island
Attribution(%)
Aberdeen
40
20
0
Source
Figure 7.5.: Source attribution of human cases using 30 locus rMLST data and STRUCTURE, Dutch,
Asymmetric Island and Aberdeen models (error bars denote 95% confidence intervals)

1748-locus cgMLST
The self-attribution results (Figure 7.6) indicate that the Aberdeen (61%) and STRUCTURE (60%)
models perform best whilst the Dutch model performs at 44%. The error bars in the Dutch model
appear to be generally larger than for the other models. Source attribution was carried out using
human data (Figure 7.7). Although, all three models indicated that the most likely source was bovine
(35-42%), this is less obvious than the findings from 7 locus MLST and 30 locus rMLST analyses.
(a) (b)
Structure - self attribution Dutch - self attribution
100 100

80 80
60 FISH 60 FISH
SWINE SWINE
40 40
OVINE OVINE
20 BOVINE 20 BOVINE
POULTRY POULTRY
0 0
Source Source
(c)
Aberdeen - self attribution
100
80
60 FISH
SWINE
40
OVINE
20 BOVINE
POULTRY
0
FISH SWINE OVINE BOVINE POULTRY
Source
Figure 7.6.: Self attribution of 1748 locus cgMLST data utilising (a) STRUCTURE, (b) Dutch, and
(c) Aberdeen models (error bars denote 95% confidence intervals)

80 Structure
Dutch
60 Aberdeen
Attribution(%)
40
20
0
Source
Figure 7.7.: Source attribution of human cases by 1748 locus cgMLST data utilising STRUCTURE,
Dutch and Aberdeen models (error bars denote 95% confidence intervals)
39,529/15,000-locus cgSNPs
The Dutch model is less accurate than the Aberdeen model in predicting the origin of the isolates by
self-attribution (30% compared with 60%) for the cgSNPs dataset. Note however that by chance the
probability to identify the right origin of an isolate from 5 sources is 20%, then the Dutch model
does considerably better than that. It is also worthy to note that the Dutch model tends to have a bias
towards fish (Figure 7.8. (a)). The confidence intervals are also very large in the Dutch model showing
that there is a high degree of variation between each of the iterations of the computer model. It
should be noted that for computational reasons the Dutch model operated on only 15,000 SNPs and
the Aberdeen model on the full 39,529 SNPs. However, the Aberdeen model self-attribution was
repeated for 15,000 SNPs and the self-attribution was found to be 62%.
Source attribution was carried out using human data (Figure 7.9). Although in both models the most
likely source was bovine (32 - 43%), the high uncertainty of the attribution results (see the size of the
confidence intervals in Figure 7.9) suggest that this result is not significant. This could be due to the
fact that many of the SNPs are not host associated and this is adding noise in the source attribution
calculations.

(a) (b)
Dutch - self attribution Aberdeen - self attribution

100 100

80 80
60 FISH 60 FISH
SWINE SWINE
40 40
OVINE OVINE
20 BOVINE 20 BOVINE
POULTRY POULTRY
0 0
Source Source
Figure 7.8.: Self attribution of cgSNPs data utilising (a) Dutch (15,000 SNPs), and (b) Aberdeen
(39,529 SNPs) models (Error bars denote 95% confidence intervals)
80 Dutch - 15kSNPs
Aberdeen - 15kSNPs
60 Aberdeen - 39kSNPs
Attribution(%)
40
20
0
Source
Figure 7.9.: Source attribution of 39,529 cgSNPs data utilising Dutch and Aberdeen models (Note
Dutch used 15,000 SNPs after removing isolates with 1 SNP difference between isolate in
reference) (error bars denote 95% confidence intervals)
7.2.2. Source Attribution of 4 Sources (Excluding Poultry)
Figure 7.10 shows the source attribution results after poultry have been removed from the analysis.
As mentioned above this was done because poultry is represented by only 25 isolates. Further, it is
worth noting that in the previous sections human attribution to poultry was generally amongst the
lowest of the 5 source reservoirs. The results at the 4 different levels of sub-typing all show that
bovine tends to have the highest rates of source attribution (7 locus MLST(35% to 61%), 30 locus
rMLST(37% to 57%), cgMLST(33% to 51%), cgSNPs(34% to 55%)). The other 3 reservoir sources
exhibit a range of attribution levels depending on the level of sub-typing and the source attribution
model.

For 7-locus MLST Self attribution improved for all of the models after poultry was dropped as a source
(Figure 7.11). For example by 8% for STRUCTURE, 11% for Dutch, 14% for AI and 13% for
Aberdeen. AI has the highest self-attribution score of 94%.
(a) (b)
(c) (d)
80 Dutch - 15kSNPs
Aberdeen - 15kSNPs
60
Attribution(%)
Aberdeen - 39kSNPs
40
20
0
Fish Swine Ovine Bovine
Source
Figure 7.10.: Source attribution excluding poultry at (a) 7 locus MLST, (b) 30 locus rMLST,
(c) 1,748 locus cgMLST and (d) 39,529 cgSNPs (error bars denote 95% confidence
intervals)

(a) (b)

100 100

80 80
60 60
FISH FISH
40 SWINE 40 SWINE
OVINE OVINE
20 20
BOVINE BOVINE
0 0
FISH SWINE OVINE BOVINE FISH SWINE OVINE BOVINE
Source Source
(c) (d)

100 100
80 80
60 60
FISH FISH
40 SWINE 40 SWINE
OVINE OVINE
20 20
BOVINE BOVINE
0 0
Source Source
Figure 7.11.: Self-attribution of 7 locus MLST data, excluding poultry, utilising (a) STRUCTURE,
(b) Dutch, (c) Asymmetric Island and (d) Aberdeen models (error bars denote 95%
confidence intervals)
For 30-locus rMLST Self attribution improved for all of the models after poultry was dropped as a
source (Figure 7.12.). For example by 8% for STRUCTURE, 6% for Dutch, 15% for AI and 13% for
Aberdeen. AI has the highest self-attribution score of 95%.

(a) (b)

100 100

80 80
60 60
FISH FISH
40 SWINE 40 SWINE
OVINE OVINE
20 20
BOVINE BOVINE
0 0
Source Source
(c) (d)
AI - self attribution
100
80 100
80
60 FISH
60
SWINE FISH
40
OVINE 40 SWINE
OVINE
20 BOVINE 20
BOVINE
0
0
FISH SWINE OVINE BOVINE
Source Source
Figure 7.12.: Self-attribution of 30 locus rMLST data, excluding poultry, utilising (a) STRUCTURE,
(b) Dutch, (c) Asymmetric Island and (d) Aberdeen models (error bars denote 95%
confidence intervals)
For 1,748-locus cgMLST self-attribution improved for all of the models after poultry was dropped as a
source (Figure 7.13), for example by 5% for STRUCTURE, 12% for Dutch and 10% for Aberdeen. The
Aberdeen model has the highest self-attribution score of 71% followed by STRUCTURE (65%) and
Dutch (56%).

(a) (b)

100 100

80 80
60 60
FISH FISH
40 SWINE 40 SWINE
OVINE OVINE
20 20
BOVINE BOVINE
0 0
Source Source
(c)

100
80
60
FISH
40 SWINE
OVINE
20
BOVINE
0
FISH SWINE OVINE BOVINE
Source
Figure 7.13.: Self attribution of 1748 locus cgMLST data, excluding poultry, utilising (a) STRUCTURE,
(b) Dutch, and (c) Aberdeen models
For cgSNPs self-attribution improved for the two models with the Dutch at 38% and the Aberdeen
model at 66% (Figure 7.14). The Dutch model appears to have a bias towards fish.
(a) (b)
Dutch - self attribution Aberdeen - self attribution

100 100
80 80
60 60
FISH FISH
40 SWINE 40 SWINE
OVINE OVINE
20 20
BOVINE BOVINE
0 0
Source Source
Figure 7.14.: Self attribution of cgSNPs data, excluding poultry, utilising (a) Dutch (15,000 cgSNPs),
and (b) Aberdeen (39,529 cgSNPs) models (error bars denote 95% confidence intervals)

7.2.3. Discussion
All of the models at all of the different levels of molecular analysis tended to place bovine as the main
source of human listeriosis (32% to 64%). The Dutch model was able to perform at all levels of
molecular analysis but was limited to 15,000 SNPs because of its current implementation in
VisualBasic. This limitation has the potential of being resolved by the software being further improved
or written using another programming platform. The Dutch model tended to have low self-attribution
compared with most of the other models and increasing the level of loci did not really improve its
performance. The Hald model was of limited utility since it could only be applied at the level of ST and
was unable to incorporate sequence types that were found in humans but not in the animal sources.
STRUCTURE was operational up to the level of cgMLST but cannot be currently implemented for
>15,000 loci as required in the cgSNP analysis. Although the AI model was operational up to
30 rMLST only, it had the highest self-attribution and tightest confidence intervals and also gave the
highest source attribution to bovine. The Asymmetric Island model incorporates recombination and
mutation. The model appears to be fairly complicated and the current explanations of its operation are
difficult to comprehend. The newly developed Aberdeen model, which operated at all levels of
molecular analysis, was relatively easy to implement and was not computationally intensive. Its self-
attribution performance was similar to a number of the other models.
Those models that were able to operate at the whole genome level (cgMLST and cgSNP) did not
appear to show improved performance from fewer loci. It is likely that a large number of the loci are
not host related and this may add noise to the analysis. It may be best to pre-select loci for host
specificity prior to source attribution. Methods need to be developed to achieve this in an unbiased
way.
The number of genomes available for some of the sources was relatively small. In the source
attribution analysis performed for 5 sources, poultry had only 25 genomes. The results indicate that
this number is probably too small and this is also seen by the relatively wide confidence intervals.
Most other published studies for source attribution tend to have at least 50, if not 100 representatives
for each source. In an analysis for campylobacter it was reported (Smid et al., 2013) that it was
preferable to have at least 100 isolates per source and the data presented showed that using
25 isolates gave a large uncertainty in the self-attribution scores as is being found in the current
study. The source attribution results for 4 sources appear to be better with higher self-attribution
scores and models producing tighter confidence intervals. Ideally selection of isolates for source
attribution should include contemporaneous sampling of isolates from sources and humans from a
fixed geographic area. In the current study, the geographic area was very broad (much of Europe)
and a fairly broad sampling timeframe with source isolates not being uniformly distributed across
Member States. Hence, the analysis should be treated with caution as there is the potential of bias.
None of the models applied utilised human consumption data. Only the Hald model has been made
operational to do this should these data become available. However the Hald model only works at the
level of ST (i.e. for a single locus).The other models provide a probability that an isolate comes from a
given source. So potentially all human clinical isolates from a country (or the EU) can be assigned
probabilistically to a source and this can be summed up to determine the likely number of cases
associated with each source. Then knowing, the exposure (i.e. amount of meals consumed associated
with each source) this can then be used to determine the average risk per meal. However, this is
simplistic as there will be a lot of different types of meals (some posing greater potential risk than
others) and also there will be variation in susceptibility of the host (immune-compromised compared
with healthy). This is an area for future research and can potentially be linked to work utilising
quantitative risk assessment.
7.3. Conclusions
Source attribution was applied utilising 5 models (Hald (ST only), Dutch (up to 15,000 cgSNPs,
STRUCTURE (up to 1748 cgMLST), Asymmetric Island (up to 30 rMLST) and Aberdeen (up to
39,529 cgSNPs) for 5 sources (fish, swine, ovine, bovine and poultry) and 4 sources (removing

poultry). All of the models showed bovine as the main source of human disease (32% - 64% for
5 sources and 33%-61% for 4 sources) but for a number of the models there were broad confidence
intervals. It was not possible to rank the relative order of importance for the other sources due to the
variation in model outputs (i.e. overlapping confidence intervals). The confidence intervals were
reduced when the poultry source was removed because of its small sample size. For all sources,
isolates from different parts of the food chain had to be combined to produce a sufficient dataset on
which source attribution could be performed. It is possible that the genetic distribution of isolates
associated with a particular source may change along the food chain and that this could affect the
source attribution results. This is an area worthy of future investigation.
The Asymmetric Island model, which was operational at only 7 locus MLST and 30 locus rMLST, had
the highest self-attribution (>80%), had the smallest confidence intervals, and had the largest
attribution to bovine (57% - 64%). The AI model therefore appears to provide the most robust results
for this dataset. However, it should be noted that the dataset used here may be biased (e.g. due to
non-uniform sampling across Europe) and would likely be more robust using larger, well-structured
datasets. The AI model is not yet operational at the cgMLST and cgSNP level. Implementation and
computational requirements of the models became more difficult the greater the number of loci that
were being processed. Currently there appeared to be no great advantage in carrying out attribution
at the highest levels of molecular resolution (i.e. cgMLST and cgSNP). New approaches are required to
select the data from across the genome to be used in the source attribution model. This is because a
number of the loci/SNPs are not informative about the source and appear to add noise to the
attribution results. There is also the potential for future research to link the source attribution results
to the risk from consuming a meal and quantitative risk assessment.
8. Epidemiological relationship linking of genetically related isolates

Several recent studies demonstrated the added value of WGS for outbreak investigations by
confirming and/or discriminating food and human isolates (Gillesberg Lassen et al., 2016; Jackson et
al., 2016; Kvistholm Jensen et al., 2016). In a context where international surveillance is increasing
(Paquet et al., 2005; Swaminathan et al., 2006), we intended to check the interest of WGS along with
epidemiological information of the food and clinical isolate to assess, retrospectively, relationships
between circulating strains of L. monocytogenes in EU within 2010-2012 period.
Within each clonal complex of L. monocytogenes the comparison of the isolates full genome to an
appropriate reference genome helped to identify individual nucleotide differences (single nucleotide
polymorphisms or SNPs). SNP differences were used to identify clusters of clinical strains and food
isolates. Clusters of interest were further investigated by focusing on metadata associated to each
strain (geographical information, timeline and isolation context).
8.1. Methods
8.1.1. Definition of genetically clustered strains
The methodology used to determine clustering of strains was based on SNP. cgMLST has proved to be
also efficient in cluster definition (Section 5). But SNPs bring currently the highest available
discriminatory information for determining genetic links between strains.
Within each CC, SNP pairwise distances were used to assess the genetic link between
L. monocytogenes strains isolated in food and strains linked to sporadic cases. Figure 8.1. -a shows,
the pairwise distance matrix for strains belonging to CC7. Some strains have less than 5 SNPs of
difference (Figure 8.1. -b). With larger SNPs difference (e.g. 10 SNPs), more links can be established.
The limit for defining genetically related strains was set to 25 SNPs according to detailed SNPs
pairwise distance observed during the retrospective analysis of known outbreaks (Section 5). Figure
8.2 shows congruent clustering of strains according to maximum likelihood phylogeny established on

SNPs or networks established on SNP distance. For such defined genetically related strains,
information on time of isolation and geography were used to retrospectively explore the links between
isolates.
(a)
Ref_CC7 RL15000089 RL15000099 RL15000072 RL15000066 RL15001370 RL15000250 RL15000351 RL15000260 RL15000275 RL15000279
Ref_CC7 268 274 162 208 429 190 585 198 194
RL15000089 268 0 1 2 266 242 251 243 241
RL15000099 274 0 1 2 264 249 261 250 248
RL15000072 162 1 1 3 172 174 178 175 174
RL15000066 208 2 2 3 218 203 212 204 203
RL15001370 429 266 264 172 218 248 410 256 253
RL15000250 190 242 249 174 203 248 187 1 0
RL15000351 585 251 261 178 212 410 187 188 186
RL15000260 198 243 250 175 204 256 1 188 1
RL15000275 194 241 248 174 203 253 0 186 1
RL15000279
(b)
5 SNPs 10 SNPs 25 SNPs
Ref_CC7 strain is public genome used for SNP calling. Upper left: sporadic isolates, lower bottom: isolates of known CC
outbreak, upper right: food isolates. Links are established between isolates if the SNP distance is lower or equal to the threshold
value indicated in the centre of circle.
Figure 8.1.: (a) Pairwise SNP distance matrix for CC7 stains (only first row column is shown).
(b) Relation between the 44 food, sporadic and outbreak strains of CC7
8.1.2. R packages and software
Distribution of SNP distance graph was generated with R software 3.2.4 and ggplot2 package. Links of
isolates within a CC, were investigated and summarized with a circular network figure produced with
functions of edgebundleR (Bostock et al., 2016) and igraph (Csardi and Nepusz, 2006) packages.
Maximum likelihood phylogenies were produced with RAxML, using the model GAMMALG, with
5,000 bootstrap replicates (Stamatakis, 2014).

(a) (b)
c7
c2 c6 c7
c5 c4
c1
c3
c2
c4
c1
c3
c6
c5
Upper left: sporadic isolates, lower left: isolates of outbreak, right: food isolates. Links established for isolates according to
distances 25 SNPs. The cluster numbers match with those listed in Table 8.1.
Figure 8.2.: (a) ML phylogeny of 45 strains of CC7. 7 clusters established on pairwise SNP distance
are presented. (b) Relation between the 45 food, sporadic and outbreak strains of CC7
8.2. Results
The links between isolates were established for 21 CCs (CC1, CC2, CC3, CC4, CC5, CC6, CC7, CC8,
CC9, CC11, CC14, CC31, CC37, CC59, CC87, CC101, CC121, CC155, CC204, CC220, CC415). For these
21 CCs, 151 "clusters" were identified according to SNP pairwise distances. Amongst them, 27 clusters
were expected as they exclusively include strains isolated in the same context (like strains belonging
to the same outbreak or strain isolated in the same factory). These clusters are not presented. Table
8.1 shows the 124 unexpected clusters that were identified based according to SNP pairwise
distance.
Table 8.1.: List of a priori non expected clusters, established according to the pairwise SNP distance
between all strains of the project
CC Cluster Outbreak Sporadic Food baseline Food other Food category*

survey
CC1 cluster 1 B(8) B(1)
CC1 cluster 10 C(2)
CC1 cluster 3 V(3) cheese
CC1 cluster 4 X(7) X(2) cheese
CC1 cluster 5 A(1) U(1) smoked and gravad fish
CC1 cluster 6 T(2)
CC1 cluster 7 Q(2)
CC1 cluster 8 B(2)
CC1 cluster 9 W(1),C(1)
CC101 cluster 10 Q(2)
CC101 cluster 11 Q(5)
CC101 cluster 7 C(2) C(1),E(1) C(2) smoked and gravad fish, dairy
CC101 cluster 8 F(1) F(1) smoked and gravad fish
CC101 cluster 9 A(1) A(1) smoked and gravad fish
CC11 cluster 1 A(2) RTE meat
CC11 cluster 2 W(1) Z(1) RTE meat
CC121 cluster 1 C(1) C(3), Q(1) smoked and gravad fish
CC121 cluster 10 P(1) B(3) smoked and gravad fish
CC121 cluster 12 D(2) smoked and gravad fish


survey
CC121 cluster 13 L(1),T(2),H(1) smoked and gravad fish
CC121 cluster 14 L(2) RTE meat
CC121 cluster 15 L(3) smoked and gravad fish
CC121 cluster 16 L(2),X(1) smoked and gravad fish
CC121 cluster 17 W(2) smoked and gravad fish
CC121 cluster 18 Q(1) B(2) smoked and gravad fish
CC121 cluster 19 K(2) smoked and gravad fish
CC121 cluster 2 T(2) U(2),L(1),T(2),H(1) smoked and gravad fish
CC121 cluster 20 J(5) RTE meat(3), smoked and gravad
fish(2)
CC121 cluster 23 C(5) RTE meat
CC121 cluster 25 A(1),J(1) RTE meat
CC121 cluster 26 X(2) RTE meat
CC121 cluster 27 B(2) FPE, RTE meat
CC121 cluster 3 C(2) smoked and gravad fish
CC121 cluster 4 C(4),S(1), E(1), smoked and gravad fish
A(1),P(1),J(1),L(2),W(1),Q(1)
CC121 cluster 5 F(3) smoked and gravad fish
CC121 cluster 6 N(2) smoked and gravad fish
CC121 cluster 7 B(2) smoked and gravad fish
CC121 cluster 8 U(1),L(7) smoked and gravad fish
CC121 cluster 9 A(1),W(1) smoked and gravad fish
CC14 cluster 1 X(13) X(1)
CC14 cluster 2 T(3)
CC14 cluster 3 U(2) smoked and gravad fish
CC155 cluster 2 X(2) LA(1),N(1) smoked and gravad fish
CC155 cluster 3 W(1),A(1) F(1),U(2),W(1),K(2),J(1) smoked and gravad fish
CC155 cluster 4 A(1) D(1) F(2) smoked and gravad fish, cheese
CC155 cluster 5 N(2),U(6),H(2) smoked and gravad fish
CC2 cluster 3 F(2)
CC2 cluster 4 A(1) LT(1)
CC2 cluster 5 B(2) FPE (1), Vegetable(1)
CC2 cluster 6 X(3) RTE meat(3)
CC2 cluster 7 C(2) RTE meat(2)
CC2 cluster 9 U(1),W(1) smoked and gravad fish(1), RTE
meat(1)
CC2 cluster 1 A(4)
CC2 cluster 2 W(2)
CC204 cluster 2 G(1) C(1),G(1),B(1) RTE meat(3), FPE(1)
CC204 cluster 3 H(2) cheese
CC204 cluster 4 B(2) RTE meat(2)
CC3 cluster 3 F(2) smoked and gravad fish
CC3 cluster 4 U(6) smoked and gravad fish (5), RTE
meat(1)
CC3 cluster 5 F(2) F(1) smoked and gravad fish
CC31 cluster 2 B(1) G(1),B(1) smoked and gravad fish, RTE
meat, FPE
CC31 cluster 3 A(1) A(1) RTE meat
CC31 cluster 4 D(1),Q(2) cheese
CC31 cluster 5 C(5) RTE meat
CC37 cluster 2 B(4),B(1),C(1) FPE, cheese, RTE meat
CC4 cluster 1 C(16) C(1)
CC5 cluster 2 C(1) C(2) RTE meat, smoked and gravad fish
CC5 cluster 4 V(2) RTE meat
CC59 cluster 1 A(1) Q(5) C(1) smoked and gravad fish, dairy


survey
CC59 cluster 2 B(1) B(1) cheese
CC6 cluster 1 C(6) cheese
CC6 cluster 10 Z(1) Z(1) smoked and gravad fish
CC6 cluster 2 Z(1) C(1) RTE meat, smoked and gravad fish
CC6 cluster 3 C(2) cheese
CC6 cluster 4 X(2) smoked and gravad fish
CC6 cluster 5 B(2) FPE, Vegetables
CC6 cluster 6 N(2),U(1),A(1),W(1) smoked and gravad fish
CC6 cluster 7 A(2) A(1) RTE meat
CC6 cluster 8 A(1),C(1), Q(1) smoked and gravad fish
D(1)
CC6 cluster 9 Z(1) B(1) FPE
CC7 cluster 3 Q(4) smoked and gravad fish
CC7 cluster 5 W(2),T(3)
CC7 cluster 6 X(3)
CC7 cluster 7 U(4) smoked and gravad fish, RTE meat
CC8 cluster 10 U(1),W(2),W(1),A(1) smoked and gravad fish
CC8 cluster 11 K(2) smoked and gravad fish
CC8 cluster 13 U(1),Q(1) smoked and gravad fish
CC8 cluster 3 Z(1) Z(1) RTE meat
CC8 cluster 4 Z(1),X(1) C(7) cheese, RTE meat
CC8 cluster 5 T(1) K(6) smoked and gravad fish
CC8 cluster 6 W(1) U(7),W(2),L(1),Q(2),W(1) smoked and gravad fish
CC8 cluster 7 C(1),X(1) B(1) smoked and gravad fish (2),
FPE(1)
CC8 cluster 8 B(3) cheese, FPE
CC8 cluster 9 S(1),W(1),J(1) smoked and gravad fish
CC8 cluster 1 T(1),A(1)
CC8 cluster 2 W(2)
CC87 cluster 1 X(13+6) X(2)
CC87 cluster 2 Q(6) smoked and gravad fish
CC87 cluster 3 X(1) V(1) RTE meat
CC9 cluster 11 Z(3) RTE meat
CC9 cluster 3 Z(1) Z(3) RTE meat
CC9 cluster 4 A(1) U(3),D(5) smoked and gravad fish
CC9 cluster 7 C(2),Z(2),Q(1) smoked and gravad fish
CC9 cluster 8 N(1),U(7),A(1),V(1) smoked and gravad fish
CC9 cluster 9 Z(2) smoked and gravad fish
CC9 cluster 1 A(3)
CC9 cluster 2 X(1) C(1) RTE meat
CC98 cluster 1 W(1),Q(1)
* FPE: food processing environment
Note: Letter identified code of countries, number in parenthesis indicates number of strains.
8.2.1. Epidemiological analysis of genetically clustered strains: link

between human sporadic strains and potential relation with food
strains
Forty-eight clusters out of the 124 included one or several sporadic human strains (representing a
total 91 sporadic human cases). For 17 out of these clusters, only human sporadic strains were
related (see e.g. cluster 5 of CC7 Figure 8.3. -a). Additionally, it was revealed that sporadic human

cases were related to four of the outbreaks studied in Section 5 (see e.g. Figure 8.3. -b). It is worth to
notice that these sporadic strains were observed in the same country where the outbreak occurred.
(a)
(b)
Red colour is associated to strains isolated in sporadic cases or in an outbreak context.
Figure 8.3.: (a) Timeline and countries of a cluster (cluster 6 of CC7) associating sporadic strains.
(b) Timeline of a sporadic strain genetically closely related to CC14 outbreak (sporadic
and outbreak strains are from the same country, that is X)
For the 27 other clusters, at least one strain isolated from food (either from baseline survey, or other
strains from national active or passive surveillance) was involved, potentially relating sporadic human
cases to contemporary food isolates that circulate in EU. Although the three categories of RTE food
products, that is smoked and gravad fish (Figure 8.4), cheese and RTE meat (Figure 8.5), were
involved, most (16) of the clusters were related to smoked fish.

(a)
(b)
Red colour is associated to strains isolated in sporadic cases or in an outbreak context. Blue colour is associated to food
isolates.
Figure 8.4.: Two clusters observed including sporadic human cases and strains isolated in smoked
salmon category from baseline survey (a) for CC8 cluster 6, (b) for CC155 cluster 3
It confirms that this smoked and gravad fish food category is of concern for the risk of listeriosis
(Pouillot et al., 2009; Tocmo et al., 2014). Yet, it cannot be concluded that sporadic cases are most
likely linked to this type of product as a majority of strains we matched against belonged to the
smoked and gravad fish category. Moreover, the majority of strains for cheese and RTE meat
categories came from passive national surveillance of a more limited number of countries. The strains
of these categories just matched less to real exposition of consumers than do the baseline survey

strains for smoked salmon. Difference in exposure could be better approached with quantitative
microbial risk assessment as this approach takes into account food exposure.
(a)
(b)
(c)
Strain isolated from RTE meat strain of CC8 cluster 2 was isolated during the baseline survey in 2011. Red colour is associated
to strains isolated in sporadic cases or in an outbreak context. Blue colour is associated to food isolates.
Figure 8.5.: Timeline and countries implicating sporadic and RTE meat strains for (a) CC6 cluster 7
and (b) CC8 cluster 3 and for (c) cheese CC59 cluster 2
8.2.2. Geographical and temporal widespread of genetically clustered

strains
Seventy-six clusters were established for food strains, i.e. not including human strains. The analysis of
these clusters revealed that strains circulated in several countries as 21 clusters involved from two up
to nine countries for cluster 4 of CC121 (Figure 8.6). This European circulation of strains is particularly
obvious for smoked and gravad fish category. As for the attribution of sporadic infections, it cannot be
inferred that trans-national circulation of strains is less present in RTE meat and cheese due to lower
number of strains available. Food exchange between EU Member States as well as consumption habits
through EU would help to determine if exposition for these two categories is country specific or not.

Blue colour is associated to food isolates.
Figure 8.6.: Timeline and countries for CC121 cluster 7 implicating the largest number of countries
(9) of all clusters of strains
The established links for sporadic strains as well as food clusters revealed that some clonal isolates
circulate for years in RTE products and confirmed the results of retrospective outbreak investigation
(see Section 5).
8.2.3. Consistency of clusters established
Consistency of genetically established clusters can be assessed with epidemiological information

associated to each strain. The countries where the strains were isolated and the type of food are the
two main elements for consistency assessment. Time of isolation is another criterion, but the relatively
short period (2 years) is not very informative.
Dealing with food categories, among the 76 clusters established between food strains and the 27
clusters where human and food strains were linked, only 13 links between more than one food
category were established (see Table 8.1, e.g. CC8 cluster_4). Different hypotheses can be advanced
to explain the contemporary presence of the same clone in different categories of RTE food. The first
is linked to cross-contamination at retail level. It has been recently shown that cross-contamination at
retail is of major importance for L. monocytogenes (Pouillot et al., 2015; Gallagher et al., 2016). Yet
cross-contamination probably concerns products of the same food category, e.g. cheeses (Heiman et
al., 2015). Another reason could be the use of common ingredients or equipment in the food chain of
the different food categories. Yet the data available for the strains of the present study is not precise
enough and there is no scientific literature that may help supporting or checking this hypothesis.
Finally, the presence of strains from different categories could be explained by false positively
associated strains. Indeed the threshold used to distinguish strain was set to 25 SNPs based on
retrospective analysis of outbreaks (Section 5). SNP pairwise distributions for some CCs show that this
threshold probably induces a loss of specificity. The first two of the multimodal distributions of
pairwise SNPs distribution (obtained of a set of diverse strains) help to distinguish closely related

strains. In our analysis, for some CCs, it appeared that the threshold of 25 is the second mode (data
not shown).
8.3. Conclusion
For any outbreak investigation, making a linkage between clinical isolates and possible food sources
requires distinguishing the suspected pathogen from the circulating background population whatever
methodologies are employed. The retrospective analysis conducted here shows that numerous
consistent genetic linkages, between a priori unlinked strains, can be established with WGS. Data to
support the establishment of the actual epidemiological linkages between the genetically related
strains was not available in this retrospective study.
With less discriminatory method (PFGE, MLST), outbreak detection is mainly based on cluster of time-
linked strains that shared the same profile (Yde et al., 2012). Systematic comparison (e.g. without
considering time-linked) with these microbiological methods is not possible as it would result in too
numerous potential epidemiological links and investigations in the field. The discriminatory power of
WGS completely changes the paradigm of outbreak investigation. Direct comparison (based on SNP or
any discriminatory method like it, e.g. cgMLST) of genomes, even in low number and/or timely
separated by several months, would result in specific and sensitive potential links.
We used a maximum SNP distance (that was confirmed by phylogenies) for establishing the link
between strains. Although we used a single rule whatever the CCs for retrospective investigation,
setting a single diversity threshold might not be the most specific and sensitive approach. According to
the diversity of subtypes in each CCs and the timeframe, this level could probably be adapted.
Furthermore, the threshold used is only valid for the workflow used to generate SNP pairwise
distances. Other variant calling workflow would result in different distributions (Sahl et al., 2016), and
thus different thresholds.
9. Putative markers
L. monocytogenes is widely found in the environment. Its ability to persist in a diverse range of niches
is supported by its ability to respond to the different stresses it encounters (Gandhi and Chikindas,
2007). These stress responses confer on it the ability to persist in environment (Gandhi and Chikindas,
2007), as well as ensuring successful transition from food into the gastrointestinal tract of hosts
(Toledo-Arana et al., 2009). Genomics data from a large collection of isolates provides the means to
identify marker genes associated with pathogen stress survival and/or virulence (Franz et al., 2014).
While the majority of L. monocytogenes isolates are generally susceptible to a large number of anti-
microbials, a small portion (Wieczorek et al., 2012; Barbosa et al., 2013; Khen et al., 2015)
demonstrate resistance to certain clinically used anti-microbials recommended for treatment of
listeriosis infection in pregnancy (Donovan, 2015). Even such low level of resistance is of concern as it
may represent an emerging pattern of developing resistance (Khen et al., 2015). Along with virulence
factors, antibiotic resistance genes that have been previously described (Charpentier and Courvalin,
1999; Lungu et al., 2011) were sought in the LISEQ collection.
Over the last 15 years, numerous virulence factors have been identified (Vzquez-Boland et al., 2001;
Toledo-Arana et al., 2009; Maury et al., 2016). We aimed at comparing the presence/absence of these
virulence factors in the genomes of the LISEQ clinical and food isolates. We sought to find any
population level differences at the lineage level which may suggest adaption or association of
particular factors to survival in the environment or to the clinical manifestation of listeriosis.
L. monocytogenes can remain on equipment or surfaces (Mettler and Carpentier, 1999) for several
months or years (Jessen and Lammert, 2003; Carpentier and Cerf, 2011). WGS has recently been
shown to be an invaluable tool to detect persistent strains in processing plants (Fagerlund et al.,
2016; Morganti et al., 2016). What is of concern is that the presence of persistent cells on food-

contact surfaces can be a source of recontamination (Lundn et al., 2002; Reij et al., 2004). One
hypothesis to explain persistence is the ability of bacteria to adapt to and survive environmental
stresses such as nutrient deprivation, hot or cold temperatures, sanitisers and preservatives,
desiccation, low pH, and high salt concentrations (Thvenot et al., 2006; Carpentier and Cerf, 2011;
Melo et al., 2015). Some studies suggested that persistent bacteria are genetically distinct from
transient strains (Autio et al., 2003; Wulff et al., 2006; Holch et al., 2013). For persistence, the first
objective was to test the ability of WGS to detect potential persistent strains among all strains
collected in a cheese plant from country Q. The second objective was to compare the
presence/absence of specific genes involved in persistence in strains isolated in food processing
environment (potentially persistent strains) to those in strains isolated in raw product (potentially non
persistent, or transient strains).
Finally, molecular genotyping techniques may also assist in the identification of potential host-
associated genetic markers. This association has already been tested for several foodborne pathogens
(Sheppard et al., 2013; Hayward et al., 2016), we aim here to carry out this search for markers for
L. monocytogenes.
9.1. Methods
9.1.1. Antibiotic resistance genes
Resistance to tetracycline, penicillin, benzalkonium chloride, quaternary ammonium sanitizers and

antiseptic were assayed in the genomes of the isolates in this study. Tetracycline resistance was
inferred from the presence of tetM and tetS, penicillin resistance inferred by the presence of penA,
benzalkonium chloride by the detection of the bcrABC locus and the Tn6188 insertion. Resistance to
quaternary ammonium sanitizers and antiseptic was inferred by the presence of the efflux pump emrE
(Charpentier and Courvalin, 1999) and qacA (Lungu et al., 2011). For detection of genes presence,
paired-end reads of each strain were mapped against the reference gene sequences using Bowtie2
v.2.2.5. (Langmead and Salzberg, 2012). The resulting alignment .sam file were then converted into
.bam files and sorted by using SAMtools (Li et al., 2009). Genes were defined as detected if they
covered greater than 80% of the query sequence with greater than 80% nucleotide identity. Genes
with coverage less than 100% were also classified as truncated.
9.1.2. Published virulence factors
A comprehensive set of 115 genes identified as putative or confirmed virulence factors were used
from the two studies (Camejo et al., 2011; Maury et al., 2016). The gene sequences were extracted
from L. monocytogenes EGD-e (accession NC_003210.1) apart from the LIPI3 clusters of gene is
extracted from L. monocytogenes F2365. Genes were detected as described in 9.1.1.
9.1.3. Genes implicated in persistence
Genetic loci involved in persistence were selected through a bibliographic research of significant genes
related to three main bacterial functions: cold growth, biofilm and resistance (Felix et al., 2015). The
list is given in Table 9.1. Genes were detected as described in 9.1.1.
Table 9.1.: Loci targeted through bibliographic research for persistent marker study
Main function Genes Gene Id Gene functions References

Biofilm actA Lmo0204 Aggregation factor (Travier et al., 2013)
Biofilm - lmo0673 Flagellar operon (Renier et al., 2011)
Biofilm bapL lmo0435 Peptidoglycane associated protein (Renier et al., 2011)
Biofilm recO lmo1460 DNA gap repare protein (Tremoulet et al.,
2002)
Biofilm - lmo2504 cell wall-binding protein (Loureno et al.,
2013)

Main function Genes Gene Id Gene functions References

Biofilm luxS lmo1288 Quorum sensing A12 biosynthesis (Bonsaglia et al.,
protein 2014)
Cold adaptation cspB lmo2016 RNA chaperon protein (Schmid et al., 2009)
Cold adaptation cspD lmo1879 RNA chaperon protein (Schmid et al., 2009)
Dessication fliP Lmo0676 flagellum biosynthesis (Hingston et al.,
resistant 2015)
Dessication flhB lmo0679 flagellum biosynthesis (Hingston et al.,
resistant 2015)
Dessication flgD Lmo0696 flagellum biosynthesis (Hingston et al.,
resistant 2015)
Dessication flgL lmo0706 flagellum biosynthesis (Hingston et al.,
resistant 2015)
Dessication motB Lmo0686 Motor control (Hingston et al.,
resistant 2015)
Dessication fliM lmo0699 Motor control (Hingston et al.,
resistant 2015)
Dessication fliY NC_019556.1 Motor control (Hingston et al.,
resistant 2015)
No gene symbol: -
9.1.4. Markers of host association
Analyses of genetic markers between isolates from each source to isolates from humans were carried
out to identify genetic markers, which differentiated significantly between these hosts. Four
genotyping methods were used to identify genetic characters: 7-locus MLST, 30 locus rMLST,
1,748 locus cgMLST and 39,529 locus cgSNP. To reduce the size of the cgSNP dataset loci with only
1 SNP difference were removed from analysis, which left only 19,902 loci.
For each of the genotyping datasets, for each allele at each locus, odds ratios were determined for the
difference in allele abundance in one host compared to human isolates. Statistical significance
(P<0.05) of these odds ratios was determined by Fishers exact test with Bonferroni correction
incorporated for multiple comparisons.
9.2. Results
9.2.1. Antimicrobial resistance
Table 9.2 shows the percentage of strains in the study harbouring the assayed resistance genes. The
resistance profile for each strain is included in the supplementary file (Annex A). Less than 1% of
isolates showed likely resistance to tetracycline via tetM with no detection of tetS. Benzalkonium
chloride resistance was conferred in 18.5% of isolates by Tn6188 insertion and approximately 5% of
isolates by the bcrABC loci. Less than 1% of isolates harboured the efflux proteins emrE and qacA
whilst the efflux protein qacC was found in 18.3% of isolates and generally found in conjunction with
Tn6188. No isolates showed likely resistance to penicillin through the presence of penA.
Table 9.2.: Percent of isolates in the study harbouring the assayed resistance genes
Gene % Detection
tetM 0.6
tetS 0
bcrA 4.9
bcrB 4.9
bcrC 4.7
emrE 0.3
qacA 0.5

Gene % Detection
qacC 18.3
Tn6188qac 18.5
penA 0
9.2.2. Published virulence factors
The supplementary file (Annex A) shows the presence and absence of 115 putative virulence markers
across the strain collection. Of the 115 markers 2 were absent across all isolates, conversely
92 markers were present in greater than 95% of isolates. Figure 9.1 shows for each virulence marker
the proportion that was present in linage I and lineage II isolates.
Figure 9.1.: Scatter plot showing the proportion each of the 115 putative virulence markers found in
lineage I or lineage II
In total, 21 putative virulence markers had significant variability in their detection across the strain
collection. As described by Maury et al (Maury et al., 2016) the Listeria pathogenicity island 3 (LIPI-3)
was found in 60% of isolates from lineage I (ubiquitous in CC1, CC3, CC4 and CC6) but completely
absent in lineage II isolates. LIPI-3 loci 1119 showed a different presence and absence profile to the
other LIPI-3 alleles with it being found in a minority of lineage II isolates (12/187 CC121, 11/54
CC155, 14/98 CC8 and 11/110 CC9) in absence of the other LIPI-3 loci. Conversely, in lineage I
isolates some isolates do not possess loci 1119 and have an otherwise intact LIPI-3.
The known virulence surface protein Vip (Cabanes et al., 2005) was found across all isolates in lineage
I but only 70% of lineage II isolates (absent in CC204, CC21, CC31, CC37 and only present in 1/43
isolates in CC7 and 3/98 isolates in CC8). Several putative virulence factors were found in a greater
proportion in lineage II isolates compared to lineage I isolates. These included the internalins lmo2026

(absent in lineage I and ubiquitous in CC155, CC18, CC20, CC204, CC21, CC37, CC415, CC7 and CC9
in lineage II) and inlF (absent in lineage I and only absent in CC121 and CC14 of lineage II) previously
shown to be detected variably in different serotypes (Chen et al., 2009).
The five gene locus termed the stress survival islet (SSI-1) (Ryan et al., 2010) which has previously
been associated with growth of L. monocytogenes under sub-optimal conditions, contributing to
survival of certain strains in food environments, was over-represented in lineage II isolates. However,
when we consider the number of clonal complexes this association is less clear. SSI-1 is present in
CC3 and CC5 of lineage I and conversely absent in lineage II CCs 101, 121, 14, 20, 21, 415 and 7.
Ubiquitous amongst lineage II isolates was the recently described rmlADBC L-rhamanose biosynthesis
loci (lmo1081 and lmo182) (Carvalho et al., 2015) involved in producing wall teichoic acids providing
protection against the activity of antimicrobial peptides as was gtcA (Promadej et al., 1999) also
involved in ecoration of cell wall teichoic acid of L. monocytogenes. The autolysin aut was also found
across all lineage II isolates but only in CC3, CC5, CC59 and CC87 of lineage I, which is perhaps
surprising given its proposed role in entry of L. monocytogenes to non-phagocytic mammalian cells
(Cabanes et al., 2004). Finally, the surface adhesion lapB required for entry into mammalian cells is
present across all lineages but absent in all isolates of CC31.
Figure 9.2.: Scatter plot showing the proportion each of the 115 putative virulence markers found in
clinical or non-clinical isolates
Figure 9.2 shows for each virulence marker the proportion that was present in clinical and non-clinical
isolates. When compared to the assortment by lineage there seems less effect than by whether the
isolate was from a clinical sample or not.
Loss of function through partial gene deletion or miss-sense mutations is also known to be important
in virulence attenuation. To explore this, genes with less than 100% coverage of the query sequence
were designated as truncated (see supplementary file Annex A). Several genes had loss of function
truncation in lineage II but were found intact in lineage I, these include the already described inlA

deletion (Maury et al., 2016) as well as lmo0257, the terminal SSI loci lmo0478, the autolysin ami and
the actin-assembly inducing protein precursor actA. Conversely several genes were truncated in
lineage I but intact in lineage II isolates. These included the internalins inlH, inlJ, lmo1290, the stress
protein clpB and the flagellar motor switch protein lmo0698.
9.2.3. Genes implicated in persistence
For persistence, the food processing facilities of three food sectors were investigated. For two cheese
production environments, WGS was used to decipher the strain diversity and the origin of
contamination. For the pork strains, the dataset was constituted of previously identified persistent and
non-persistent strains. The presence of markers within these two subsets was investigated. For the
salmon producers of country B, WGS was used to identify the origin of contamination and to identify
persistent strains within other strains. The research of presence of putative markers was carried out
on these strains.
Persistence in dairy plants
Two different dairy plants were investigated. The first one corresponds to a cheese plant in country B,
the second one to a cheese plant in country Q.
In total there were 10 isolates that were genome sequenced from contamination of full fat semi soft
unpasteurised cheese made from bovine milk (check this) originating from country B. Six of the
isolates were from two cheese products (First product: RL15000630, RL15000631, RL15000, second
product: RL15000637, RL15000638 & RL15000639) and 4 from the factory environment (RL15000635
(rack in the chill that cheese were stored on), RL15000634 (swab of brine trolley handle),
RL15000633 (swab of top of brine trolley), RL15000632 (swab from brush used in factory)). All of the
isolates were ST37 except RL15000639 which was ST121. This isolate was the first isolated (January
2013). The remaining isolates were obtained in the following 2 months. As all ST37 strains present
less than 25 SNPs, it can be concluded that a single clone persisted in the factory environment and
was at origin of the cheese contamination.
Within the 100 strains isolated in the cheese plant from country Q in 2012, 13 different CCs were
found. The most prevalent CCs were CC101 and CC2. Figure 9.3 shows the repartition of the different
CCs.
Line thickness is proportional to number of isolates (largest CC represents 32 isolates, smallest CCs are represented by 1 strain.
Figure 9.3.: Repartition of the thirteen CCs isolated in a cheese plant from country Q

Phylogeny established with SNP analysis data of 16 CC2 strains presented on Figure 9.4 helps to
decipher which strains are related with other. At least 3 different clusters of strains of CC2 circulate in
the processing plant. Yet none of these clusters of strains in the environment match with the strain
isolated from cheese. Such an analysis would not have been possible with less discriminatory
methods.
Figure 9.4.: ML phylogeny tree of 16 strains of CC2 isolated in the same cheese factory
Persistence in the pork processing environment

The pork processing strains from cutting plants were specifically selected to look for potential
differential presence/absence of putative markers for persistence. The presence/absence analysis for
the 15 gene loci identified to be of importance for persistence in food processing environment is given
in Table 9.3.

Table 9.3.: Presence/absence of putative markers for persistence in two groups of strains: persistent
strains isolated in cutting plants, non-persistent strains isolated in raw material of the
same cutting plants (pork processing, Country C)
Presence of potential markers for persistence

RL NC_019
Gro lmo0 lmo0 lmo0 lmo1 lmo2 lmo1 lmo2 lmo1 lmo0 lmo0 lmo0 lmo0 lmo0 lmo0
num 556.1
up 204 673 435 460 504 288 016 879 676 679 696 706 686 699
ber FliY
RL1500
0543 x x x x x x x x x x x x x x
RL1500
0542 x x x x x x x x x x x x x x x
RL1500
RL1500
RL1500
0539
x x x x x x x x x x x x x x
RL1500
RL1500
RL1500
RL1500
0391
x x x x x x x x x x x x x x x
RL1500
Persistent
RL1500
RL1500
RL1500
0387
RL1500
RL1500
RL1500
RL1500
0364
RL1500
RL1500
RL1500
RL1500
RL1500
0370
RL1500
RL1500
RL1500
RL1500
Non persistent
0374
RL1500
RL1500
RL1500
RL1500
0378
Note: Empty cells indicate absence of markers.
Whatever the group of strain, i.e. the group of persistent strains isolated from food processing
environments or the group of strains isolated from recently imported raw materials into the plant, at
least 14 out of 15 gene loci were present. The gene locus lmo0435, was not present in all isolates in
either of the two groups. Yet the proportion of strains without this locus was the same both groups of
strains.

Within the 14 gene loci present, no large deletion or insertion were found in the strains of the two
groups (see the example Figure 9.5).
The upper part shows the coverage of reads along the gene. The lower part presents the mapping of each reads. For a
deletion, a section of DNA is absent in the subject genome compared to the reference genome. In the case of an insertion, a
section of DNA is present in the subject genome that is not represented in the reference genome. Position of SNP (regarding
reference gene) in aligned reads can be seen with vertical lines. Insertions are indicated by a purple I ( ) and deletions are
indicated with a black dash (). Alignments that are displayed with light gray borders and transparent or white fill, have a
mapping quality equal to zero. Green reads present a poor mapping quality. Red reads include one sequencing mutation that
corresponds to a sequencing error.
Figure 9.5.: Integrative Genomics Viewer (IGV) screen capture of pair-end reads mapping to fliM
gene. Link to figure in high quality: https://github.com/lguillier/LISEQ-codes/tree/master/
Chapter9
No allelic profile was found to explain the persistent phenotype (see two examples for actA and FliY in
Figure 9.6).
Figure 9.6.: ML phylogeny tree of persistent (light blue) and non-persistent strains (blue) based on
SNPs for gene actA (left) and fliY (right)

Persistence in salmon processing plants

The commonality between 29 isolates collected from four salmon processors in country B, over
4 years, from along the processing chain were determined using WGS.
There were nine different ST from nine different Clonal Complexes, with three multi-isolate ST
(CC121, CC101 and CC31) represented by 12, 8 and 3 isolates respectively with the remaining six ST
being singletons (Figure 9.7).

Distances on the branches in the figure are measured in SNP differences. Clonal complex 121 is spread throughout all the
processors, while processors 1 and 3 also have site-specific clones of Listeria present.
Figure 9.7.: Minimum spanning tree of strains isolated from country B Salmon Processors indicating
distribution of isolates by CC and by Processor

Phylogeny (as well as SNP address or SNP pairwise distance) for CC101 strains isolated in Processor 1
shows that the same strains circulated in the plant during the period considered, that is 2011-2013.
The same clonal group was present either in food processing environment (e.g. RL15000641 was a
strain isolated from a swab) or in raw fish. But none of the strains were isolated in final products.
Figure 9.8.: Clade of the ML phylogeny tree of CC101 (extracted from the complete phylogeny
presented in Section 4) regrouping all the strains isolated in smoked salmon Producer 1
For CC121, out of the 12 isolated strains in Producer 2, three different clonal groups were identified
(Figure 9.9.). It is worth noting that two of them match with two strains from the baseline survey
(RL15000160 and RL15000276, cf. Table. 8.1 CC121 clusters 10 and 18).
Upper right: two strains isolated in smoked salmon product sampled for baseline survey.
Figure 9.9.: Links between strains of CC121 established by pairwise SNP distance (below or equal
25)
The presence or absence of putative markers for persistence was tested in the strains of this CC. The
strains isolated more than once from Producer 2 were considered as potentially persistent and the
four strains that were isolated once were considered as non-persistent (Table 9.4). No large
insertion/deletion or SNP help to distinguish both groups of strains (Figure 9.10).

Table 9.4.: Presence/absence of putative markers for persistence in two groups of strains: persistent
strains isolated in salmon processing environment and/or finished product more than
once, non-persistent strains isolated in food processing environment only once
Presence of potential markers for persistence
NC_019556.
RL
lmo0204
lmo0673
lmo0435
lmo1460
lmo2504
lmo1288
lmo2016
lmo1879
lmo0676
lmo0679
lmo0696
lmo0706
lmo0686
lmo0699
nimber
Group
1 FliY
RL15000620 X X X X X X X X X X X X X X X
Persistent
persis
-tent
Non
The upper mapping corresponds to an a priori non-persistent strain (RL15000629), the lower to a persistent strains
(RL15000621). For a deletion, a section of DNA is absent in the subject genome compared to the reference genome. In the
case of an insertion, a section of DNA is present in the subject genome that is not represented in the reference genome.
Position of SNP in aligned reads can be seen with vertical lines (regarding the reference gene sequence shown at the bottom of
the graph). Insertions are indicated by a purple I ( ) and deletions are indicated with a black dash (). Alignments that are
displayed with light gray borders and transparent or white fill, have a mapping quality equal to zero. Green reads present a
poor mapping quality. Red reads include one sequencing mutation that corresponds to a sequencing error.
Figure 9.10.: Integrative Genomics Viewer (IGV) screen capture of pair-end reads mapping to fliM
gene. Link to figure in high quality: https://github.com/lguillier/LISEQ-codes/tree/master/
Chapter9

9.2.4. Markers of host association
The number of isolates in the different host sets and the number of loci screened for each genotyping
dataset is detailed in Table 9.5. For each method and host pair Table 9.5 indicates the total number of
alleles identified which were significantly different for that host pair and secondly the number of loci
which harboured these alleles.
Overall the number of loci which could differentiate between human isolates and a source isolate were
rare for human-bovine and uncommon for human-poultry, suggesting that the strains found in bovine
and poultry sources are genetically similar to those in human cases. Other sources had greater
numbers of distinguishing loci between the host and human isolates.
7-locus MLST and rMLST both use loci that are considered selectively neutral whilst cgMLST and
cgSNP comprise markers which span the spectrum from neutral through to those loci under selection
(both negative and positive selection). Genetic variation in 7-locus MLST, rMLST and cgMLST loci is
classified using alleles which differ from each other by sequence polymorphisms which can be
anything from a single nucleotide through to several nucleotide differences; thus these alleles are not
truly independent of each other as they may harbour common polymorphisms. cgSNP, on the other
hand, are defined as a unique site in the genome (i.e. a particular nucleotide position) and so will be
independent of each other. Secondly cgSNP, as with other genotyping schemes, have allelic variants
at each of the cgSNP loci in the form of four alternative bases.
An ideal genetic marker for molecular host attribution would be one which was found exclusively in
one host source and not in other sources. Since attribution works with several hosts all such ideal host
specific markers would be pooled together and used collectively. The benefit of this strategy is that
those markers which do not significantly contribute to host specificity - they are neutral and can only
reduce the strength of the host specific signal are excluded, and so more robust attribution scores
should result.
Table 9.5.: Abundance of putative markers to differentiate a source host isolate from a human isolate
Genotyping Number of Human isolates (N=254) vs

method Loci Bovine Fish Ovine Poultry Swine Loci found
(N=61) (N=323) (N=89) (N=25) (N=112) in any host
MLST 7 0/0 15 / 7 10 / 7 0/0 8/7 7
rMLST 30 0 45 / 21 10 / 9 1/1 32 / 19 22
cgMLST 1,748 4/4 3,900 / 1,684 / 100 / 100 2,506 / 1,748
1745 1,399 1,567
cgSNP 19,902 0 16,801 / 2,203 / 112 / 58 14,379 / 9,164
8,393 1,129 7,227
For each method and host pair, the cell indicates the total number of alleles identified which were
significantly different for that host pair and secondly the number of loci which harboured these alleles.
9.3. Conclusion
Antimicrobial resistance in Listeria sp. has been studied in various food, environmental and clinical
settings (Bertrand et al., 2005; Morvan et al., 2010; Granier et al., 2011; Jamali et al., 2015). Listeria
monocytogenes has generally been shown to be more susceptible to antimicrobial agents then other
species in the genus. In this study we found remarkable low resistance to tetracycline (<0.1%) and
penicillin (1%). Resistance to detergents and antiseptics via efflux activity was significant with
mechanisms detected at a prevalence approaching 20%. Whilst it is encouraging that the isolates in
this study show low levels of antimicrobial resistance it is important to remain vigil for emerging
resistance. Whole genome sequencing allows antimicrobial resistance monitoring to be done as a cost-

neutral activity if WGS is part of routine microbial surveillance and therefore allows this potential
threat to be reviewed going forward.
Whole genome sequencing provides the opportunity for rapid interrogation for markers of virulence.
In this study 115 putative markers of virulence were assayed for their presence or absence in this
data set. Less than 20% of markers were present in less than 95% of the isolates suggesting that
most putative markers described in the literature are fairly ubiquitous across at least lineage I and
lineage II Listeria monocytogenes. Of those that vary the majority were over-represented in food
and/or lineage II isolates with markers associated with stress survival or cell wall modification
particular enriched. Conversely the recently discovered Listeria pathogenicity island 3 and the surface
protein vip were enriched in clinical and/or lineage I isolates. Although most virulence markers were
present in all strains we do not know if the genes are in-fact expressed. Several truncations were
identified in virulence genes across the dataset with some having an increased propensity for
truncation dependent on lineage.
The present study confirms recent studies that showed that WGS and SNP-based analysis is well-
suited to investigated persistence and contamination routes of L. monocytogenes in food processing
facilities and in the food chain (Fagerlund et al., 2016).
The presence/absence of genes thought to promote persistence was not found to be pertinent for
predicting persistent phenotype. SNPs as well as insertion and deletion in these genes were not
helpful either. The study of expression of gene marker for persistence (Mazza et al., 2015) or
proteome analysis (Rychli et al., 2016) have recently appeared to be more promising for predicting
persistence phenotypes. The analysis of the accessory genome is also an important element in
persistence study as it has been recently shown that conservation of the accessory might be
associated with persistence (Fagerlund et al., 2016).
This study did not consider the accessory genome, which by definition comprises genes, which are not
present ubiquitously across the population. Such genes will make a significant contribution to the
variation in biology seen between strains and therefore should be a rich source for the discovery of
polymorphisms associated with host association, and indeed many other features. This pilot study
suggests that cgSNP (see 9.2.4), and by extension SNP in the accessory genome, are likely to be the
most fruitful source of host associated polymorphisms, which may be of use in refining molecular
attribution models.
10. Conclusions
The overall objective of this study was to compare L. monocytogenes isolates from the EU-wide BLS
on ready-to-eat foods conducted in 2010-11, with isolates from compartments along the food chain
and from human cases using WGS analysis. This was achieved by meeting the three described specific
objectives.
The first specific objective was met by assembling a fully representative isolate collection that
consisted of a total of 1,143 L. monocytogenes isolates from across the EU, including 810 isolates
from along food chain and 333 human clinical isolates. The food chain isolates comprised 353 from
the EU-wide baseline survey (BLS) on the prevalence of L. monocytogenes in certain RTE foods,
423 from national surveys, control programmes or research projects and 34 food isolates from
outbreak investigations. The clinical isolates were provided voluntarily by European national public
health laboratories and comprised 262 isolates from sporadic cases and 71 from outbreaks. Isolates
were selected within a time frame of 2010-2012 as far as possible although this was extended as
necessary to ensure the strain collection was as representative as possible within the scope of the
study. The majority of isolates were whole genome sequenced at Public Health Englands sequencing
facilities using state of the art equipment and methodologies under an accredited quality management
system. For a minority of isolates WGS data was already available with sequencing having also been
undertaken at PHE and were included in the analysis subject to WGS data meeting the same quality

metrics used for WGS data generated as part of this study. A database was constructed with the
available metadata for the isolates with links to their respective genome sequences.
In order to fulfil the second specific objective it was necessary to investigate the phylogeny of the L.
monocytogenes isolates and produce data sets, in order to provide a framework for further analyses
on the genetic diversity and potential epidemiological associations. This was carried out using a range
of bioinformatic procedures including several gene-by-gene based approaches such as 7-gene MLST,
rMLST, cgMLST as well as SNP-based methods including cg SNP analysis. This study has facilitated the
WGS analysis of a unique and large data set of L. monocytogenes isolates and has enabled the
population to be defined to an unprecedented level of resolution from linage to nucleotide. The
phylogeny showed a clear delineation between L. monocytogenes lineages and between clonal
complexes within lineages. All isolates in the study were in either lineage I or II. There was a huge
amount of diversity among the genomes sequenced and they cover the diversity in lineages I and II
as described previously by Ragon et al. (2008). There was an uneven distribution of isolates both
between the two lineages and also amongst the clonal complexes within each lineage but this is to be
expected due to the number of isolates selected from each source and due to the particular sources
themselves which were restricted to RTE foods, compartments of the food chain and human clinical
cases. A key finding from the phylogenetic analysis of four large CCs (CC8. CC9, CC101, CC121), is
that within a CC, clinical isolates are not associated to a specific clade of the tree. The phylogenetic
analysis also confirmed recent work by Maury et al. (2016) that CC4 is associated with highly virulent
clinical strains.
The third specific objective of this project was to investigate the suitability of WGS as a tool in the
investigation of listeriosis outbreaks. This was performed by a retrospective analysis of human and
food isolates that had been previously epidemiologically and microbiologically linked. The sequences
from each previously defined outbreak were analysed together with all other isolates from this study
in the same clonal complex. The CCs to which the outbreak isolates belonged were analysed by two
bioinformatic methods, SNP-based analysis and cgMLST and an important finding was that overall the
two methods gave concordant results. Nine outbreaks were studied in total of which 6 were typical
point source outbreaks. In each of these, previously epidemiologically linked isolates clustered closely
together within a maximum 8 SNP pairwise cluster, and separated from other isolates of the same CC
that were included in the outbreak analysis. The remaining 3 outbreaks showed more variation
although linked isolates could be defined within a maximal 12 SNP pairwise cluster. Two of these
outbreaks occurred over an extended time period and the variation seen may reflect diversity within
the source over a long period. The third outbreak consisted of two separate outbreaks and restricting
the SNP cluster threshold to 5 would not have included all the epidemiologically linked cases.
Increased diversity within an outbreak may be due to differences in the ecology of outbreaks e.g. the
involvement of more complex food distribution networks.
One of the outbreaks (5) demonstrated the potential impact of SNPs/alleles being acquired in a single
event (e.g. phage) on outbreak analyses. Whilst in this instance removing the particular region did not
influence the overall interpretation of the phylogenetic analysis, it demonstrates that knowledge of
where in the genome SNPs are occurring can sometimes be very important and should be taken in to
consideration when using SNP and gene by gene approaches (Wang et al, 2015). There were no
additional food isolates included as part of outbreak analyses that fell within the cluster of human
isolates for any of the outbreaks. However, in 4 outbreaks 1 or 2 human isolates submitted as
sporadic isolates did cluster together with the outbreak isolates. In all four cases these isolates
originated from the same country as the outbreak. This study therefore demonstrates the potential of
WGS analysis to detect more cases as being part of an outbreak than previous typing methods. Whilst
there was not an international aspect to the outbreaks that were analysed in this study, it does
demonstrate the ease with which WGS can accurately rule isolates in or out of outbreaks and how
valuable the method would be for international surveillance. This study included a limited number of
previously identified outbreaks and that were predominantly restricted in time and diversity. In order
to fully assess the usefulness for WGS analyses for outbreak investigation more diverse outbreaks
including those involving multiple strains and those across more than one Member State need to be
examined.

This study shows that WGS analysis clearly separates outbreak isolates from background isolates
within the same CC and thus WGS is very well suited for detecting and defining outbreaks. The results
also illustrate that when applying WGS analysis every outbreak should be considered in its own
context and that there should not be a single universal cut off value for separating outbreak and
background isolates. Analysis and interpretation of WGS clusters requires expert knowledge and
collaborative input from bioinformaticians, epidemiologists and microbiologists.
Specific objective 2 was to analyse the WGS data of the selected L. monocytogenes isolates and
thereby explore genetic diversity, epidemiological relationships and investigate putative markers of
survival and pathogenicity. Exploring the genetic diversity of L. monocytogenes within and between
different sources including those of human origin was accomplished using Simpsons Diversity index
and Rarefaction. The genetic distance between each source was investigated using Neis genetic
distance (Nei, 1975). Simpsons index indicated high genetic diversity (>0.8) within all the sources
investigated for both 7 locus MLST and 30 locus rMLST. Rarefaction demonstrated that only a small
proportion of the diversity had been sampled. Whilst we have sampled representatively across the
population structure we have not sampled deep into the diversity of the species. Isolates from clinical
cases were found to be more diverse than isolates from other sources and this may not be
unexpected as humans are most likely to be exposed to greater variety of sources. Other contributing
factors are likely to be involved, including that the food isolates were restricted to mainly fish and
meat and that the number of isolates were limited. Whilst isolates from all other sources were
different to those from humans at all levels explored, those from the bovine source were found to be
the closest genetically by Neis genetic distance. However, whether this is a robust finding or an
artefact of the sampling for this study needs to be verified using additional isolates to see whether this
pattern continues. It is important, when considering the findings here, to note that isolates for each
source came from different points in the food chain with those closest to retail being at a greater
chance of cross contamination from another source. Thus, further work using isolates more widely
distributed across the food chain is required to provide more robust data. However, because there
was not a random genetic distribution between the different sources this study demonstrates that
source attribution based on WGS has the potential to produce useful results.
The second part of specific objective 2, to assess the epidemiological relationship of
L. monocytogenes from the different sources and of human origin considering the genomic information
and the metadata available for each isolate, was investigated in two ways. Firstly by exploring source
attribution and secondly by analysing WGS data in conjunction with isolate metadata, to investigate
any potential relationships between circulating strains in the EU from 2010-2012. All of the source
attribution models showed bovine reservoir to be the main source of human disease, however, other
sources also contributed and, for most models, confidence intervals were high. For all sources,
isolates from different parts of the food chain were combined to produce a sufficient dataset to
perform source attribution. It is possible that the genetic distribution of isolates associated with a
particular source may change along the food chain and that this could affect the source attribution
results. This area merits further investigation and increasing sample size would improve the
robustness of the results and reduce biases. It was found that increasing the number of loci did not
improve source attribution for all of the models. However, the Aberdeen method tended to perform
better (i.e. be more reliable) with larger numbers of loci. New approaches need to be developed for
source attribution using the information that is available across the genome. This is because a number
of the loci/SNPs are not informative about the source and appear to add noise to the attribution
results. There is also the potential for future research to link the source attribution results to the risk
from consuming a meal and quantitative risk assessment.
Establishing links between clinical isolates and food isolates is essential for controlling listeriosis and
preventing outbreaks and the second way epidemiological relationships were investigated was by
analysing WGS data in conjunction with isolate metadata. This was conducted using a SNP-based
approach as currently this provides the highest level of strain discrimination. In this study numerous
consistent genetic linkages between a priori, unlinked strains were identified, some of which involved
isolates from multiple countries. A total of 151 clusters were detected including 124 novel clusters that
had not been detected previously. Of these, 48 included one or more sporadic human isolates of

which 17 contained only human isolates, and were thus not linked to any of the food isolates included
as part of this study. The analysis also revealed sporadic cases that were genetically related to some
of the known outbreaks investigated in Section 5, demonstrating the potential of WGS analysis to
identify previously undetected cases. The additional cases identified in this study were from the same
country in which the outbreak originally occurred and may possibly also represent earlier cases caused
by a strain that went on to cause an outbreak. For 27 novel clusters there was at least one food
isolate; potentially relating human cases to contemporary food isolates circulating in the EU.
Approximately half of all novel clusters detected contained food isolates only and the analysis revealed
that strains were circulating in several different EU countries. Whilst this was particularly evident for
smoked and gravad fish isolates there were far fewer isolates from meat and soft cheese.
This study illustrates clearly the discriminatory power of WGS, demonstrating its ability to completely
change the paradigm of outbreak investigation. WGS comparisons based on SNPs or cgMLST result in
the detection of specific and sensitive potential links between human cases and/or foods that merit
further epidemiological investigation. Epidemiological information is essential to support the
genetically defined links in outbreak investigations but data to support the epidemiological links
between the genetically related strains was not available in this retrospective study.
The analysis showed that sporadic cases can be related (i.e., putative outbreaks) and/or associated to
food isolates, even links between sporadic human cases in one country and food in another country
were identified. Possible links identified in this way would require full epidemiological investigation in
order to support the genetic data. Although this project analysed >1,100 genomes, we did not cover
all European countries and all relevant food and clinical isolates. If European wide, real time
surveillance is set up in the future, it is likely that more outbreaks will be recognised and investigated.
Earlier identification of outbreaks and possible sources will allow for more rapid interventions and the
possible prevention of more cases.
One of the many advantages of WGS is that as well as affording high resolution typing and
phylogenetic context it provides immediate access to a wealth of additional data. The third part of
specific objective 2 was to identify the presence of putative markers conferring the potential to
survive/multiply in the food chain and/or cause disease in humans. The LISEQ L. monocytogenes
genomes were mined to identify genes, or other genetic markers, known to be implicated in
antimicrobial resistance and in virulence, and also to identify genes which may play a role in
persistence and survival and in the host-specificity of different strains. In terms of antimicrobial
resistance there was remarkable low presence of tetracycline (<0.1%) and penicillin (1%) resistance
genes. Resistance to detergents and antiseptics via efflux activity was significant with mechanisms
detected at a prevalence approaching 20%. Whilst it is encouraging that the isolates in this study
show low levels of antimicrobial resistance it is important to remain vigil for emerging resistance.
Whole genome sequencing allows antimicrobial resistance monitoring to be done as a rapid cost-
neutral activity if WGS is part of routine microbial surveillance and therefore allows this potential
threat to be reviewed going forward.
WGS data were also assessed for the presence of 115 putative markers of virulence. More than 80%
of markers were present in more than 95% of the isolates suggesting that most putative markers
described in the literature are ubiquitous across L. monocytogenes lineages I and II. The majority of
markers not present in all isolates were over-represented in food and/or lineage II isolates with
markers associated with stress survival or cell wall modification being particularly enriched.
Conversely, the recently discovered Listeria pathogenicity island 3 and the surface protein VIP were
more likely to be found in clinical and/or lineage I isolates. Although most virulence markers were
present in all strains it is not known if the genes are expressed. Further work is needed including the
determination of truncation and non-sense mutations which have been shown to be associated with
changes in virulence particularly in the internalin genes (Maury et al., 2016). Several truncations were
identified in virulence genes across the dataset with some having an increased propensity for
truncation dependent on lineage.
The WGS LISEQ data, for isolates collected over long periods of time from food factories or processing
environment, was screened to determine the presence of putative markers conferring the potential to

survive and multiply in the food chain. The presence or absence of genes thought to promote
persistence was not found to be useful for predicting persistent phenotype neither was the presence
of mutations in these genes. It may be that persistent phenotype is determined by gene expression
rather than presence or absence of specific genes (Rychli et al., 2016). It is also possible that genes
or markers other than those selected in this study are important in persistence or survival including
ones in the accessory genome (Fagerlund et al., 2016). Whilst unable to demonstrate differences in
persistent gene markers in the isolates in this study, it was shown that WGS SNP-based analysis is
well suited and valuable for investigating persistence and contamination routes within food processing
facilities and within the food chain.
WGS data from human and different animal sources was used to identify host specific markers that
might be valuable for source attribution by comparing four different genotyping techniques (7-locus
MLST, rMLST, cgMLST, cgSNP). Aggregating across all hosts identified how many different loci
contributed to host specificity. For 7-locus MLST and cgMLST all loci, and for rMLST most loci
contributed towards host differentiation. In contrast, cgSNP, is the only genotyping scheme where the
loci comprise individual polymorphisms, and was the only genotyping scheme which identified a
subset of the markers across all hosts which differentiated between human and other host sources.
Whilst the work here constitutes a small study it suggests that cgSNP, and by extension SNP in the
accessory genome, are likely to be the most fruitful source of host-associated polymorphisms, which
may be of use in refining molecular attribution models. It is important to note that this study did not
investigate the accessory genome, which by definition comprises genes that are not present
ubiquitously across the population. Such genes should be a rich source for the discovery of
polymorphisms associated with host association, and indeed many other features and deserve to be
fully explored.
In conclusion, this study carried out WGS of a large unique collection of L. monocytogenes isolates
from foods, food processing environments and clinical cases from a large number of European
countries. The collection included isolates from foods that were part of the EU-wide baseline survey
(BLS) on the prevalence of L. monocytogenes in certain RTE foods and highlight the value of revisiting
well-structured surveys. This study has demonstrated one of the major benefits of WGS, which is the
ability to address a wide range of questions including those on virulence, antimicrobial resistance,
source attribution, surveillance and outbreak detection and investigation, in a single experiment. The
WGS data generated is now available for additional analysis to address a wide range of questions and
thus represents a valuable resource for further studies. The LISEQ isolates have all been typed using
current molecular methods and thus can be used to demonstrate the back compatibility of WGS with
historical data and also to assess bioinformatic programmes that are able to predict such typing
results from WGS data.
This study illustrates one of the major strengths of WGS in comparison to conventional molecular
typing methods, which is its ability to provide high quality, unambiguous data. WGS analysis such as
cgMLST and cgSNP based typing approaches have been shown to have unparalleled strain typing
resolution and it has been demonstrated here how WGS is able to link previously undetected cases to
outbreaks and detect clusters of cases that were previously undetected. It has also been shown,
however, that as well as cgMLST and cgSNP approaches, that knowledge of the accessory genome
can contribute to the interpretation of strain relatedness. The limitations of WGS are less to do with
the actual sequencing and the analyses themselves but more dependent on representative sampling
of isolates and requirement for good epidemiological data to further investigate genetically linked by
WGS. This study supports the use of WGS for L. monocytogenes outbreak investigations although
analysis of more complex outbreaks would be valuable. However, is difficult to recreate outbreak
investigations accurately retrospectively and in order to maximise the advantages of using WGS for
outbreak detection it would be highly valuable to use WGS prospectively for the surveillance of
listeriosis across Europe.

11. Additional supporting information

Annex A - Excel file: LISEQ_DB.xlsx - Supplementary isolate list with metadata: Characteristics and
descriptive epidemiological information for all L. monocytogenes isolates included in the database of
the LISEQ tender (WGS tender analysis of L. monocytogenes from food and human sources)
Annex A can be found in the online version of this output (Supporting information section:
http://onlinelibrary.wiley.com/doi/10.2903/sp.efsa.2017.EN-1151/abstract

References
Autio T, Keto-Timonen R, Lunden J, Bjorkroth J and Korkeala H, 2003. Characterisation of persistent
and sporadic Listeria monocytogenes strains by pulsed-field gel electrophoresis (PFGE) and
amplified fragment length polymorphism (AFLP). System of Applied Microbiology, 26, 539-545.
Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, Lesin VM, Nikolenko SI, Pham S
and Prjibelski AD, 2012. SPAdes: a new genome assembly algorithm and its applications to single-
cell sequencing. Journal of Computational Biology, 19, 455-477.
Barbosa J, Magalhes R, Santos I, Ferreira V, Brando TR, Silva J, Almeida G and Teixeira P, 2013.
Evaluation of antibiotic resistance patterns of food and clinical Listeria monocytogenes isolates in
Portugal. Foodborne pathogens and disease, 10, 861-866.
Batz MB, Hoffmann S and Morris Jr JG, 2012. Ranking the disease burden of 14 pathogens in food
sources in the united states using attribution data from outbreak investigations and expert
elicitation. Journal of Food Protection, 75, 1278-1291.
Bertrand S, Huys G, Yde M, D'Haene K, Tardy F, Vrints M, Swings J and Collard J-M, 2005. Detection
and characterization of tet (M) in tetracycline-resistant Listeria strains from human and food-
processing origins in Belgium and France. Journal of medical microbiology, 54, 1151-1156.
Bolger AM, Lohse M and Usadel B, 2014. Trimmomatic: a flexible trimmer for Illumina sequence data.
Bioinformatics, btu170.
Bonsaglia E, Silva N, Jnior AF, Jnior JA, Tsunemi M and Rall V, 2014. Production of biofilm by
Listeria monocytogenes in different materials and temperatures. Food Control, 35, 386-391.
Bostock M, Patrick E, Russell K and Tarr G, 2016. Circle Plot with Bundled Edges. Package
'edgebundleR'. Available at https://cran.r-project.org/web/packages/edgebundleR/
edgebundleR.pdf.
Cabanes D, Dussurget O, Dehoux P and Cossart P, 2004. Auto, a surface associated autolysin of
Listeria monocytogenes required for entry into eukaryotic cells and virulence. Molecular
microbiology, 51, 1601-1614.
Cabanes D, Sousa S, Cebri A, Lecuit M, Garcadel Portillo F and Cossart P, 2005. Gp96 is a receptor
for a novel Listeria monocytogenes virulence factor, Vip, a surface protein. The EMBO journal, 24,
2827-2838.
Camargo AC, Woodward JJ and Nero LA, 2016. The Continuous Challenge of Characterizing the
Foodborne Pathogen Listeria monocytogenes. Foodborne pathogens and disease.
Camejo A, Carvalho F, Reis O, Leito E, Sousa S and Cabanes D, 2011. The arsenal of virulence
factors deployed by Listeria monocytogenes to promote its cell infection cycle. Virulence, 2, 379-
394.
Carpentier B and Cerf O, 2011. Review - Persistence of Listeria monocytogenes in food industry
equipment and premises. International Journal of Food Microbiology, 145, 1-8.
Carvalho F, Atilano ML, Pombinho R, Covas G, Gallo RL, Filipe SR, Sousa S and Cabanes D, 2015. L-
Rhamnosylation of Listeria monocytogenes wall teichoic acids promotes resistance to antimicrobial
peptides by delaying interaction with the membrane. PLoS Pathog, 11, e1004919.
Charpentier E and Courvalin P, 1999. Antibiotic Resistance in Listeriaspp. Antimicrobial Agents and
Chemotherapy, 43, 2103-2108.
Chen J, Luo X, Jiang L, Jin P, Wei W, Liu D and Fang W, 2009. Molecular characteristics and virulence
potential of Listeria monocytogenes isolates from Chinese food systems. Food microbiology, 26,
103-111.

Croucher NJ, Page AJ, Connor TR, Delaney AJ, Keane JA, Bentley SD, Parkhill J and Harris SR, 2014.
Rapid phylogenetic analysis of large samples of recombinant bacterial whole genome sequences
using Gubbins. Nucleic acids research, gku1196.
Csardi G and Nepusz T, 2006. The igraph software package for complex network research.
InterJournal, Complex Systems, 1695, 1-9.
de Castro V, Escudero J, Rodriguez J, Muniozguren N, Uribarri J, Saez D and Vazquez J, 2012.
Listeriosis outbreak caused by Latin-style fresh cheese, Bizkaia, Spain, August 2012. Euro Surveill,
17.
Donovan S, 2015. Listeriosis: a Rare but Deadly Disease. Clinical Microbiology Newsletter, 37, 135-
140.
Doumith M, Buchrieser C, Glaser P, Jacquet C and Martin P, 2004. Differentiation of the major Listeria
monocytogenes serovars by multiplex PCR. Journal of clinical microbiology, 42, 3819-3822.
EFSA, 2013. Analysis of the baseline survey on the prevalence of Listeria monocytogenes in certain
ready-to-eat (RTE) foods in the EU, 2010-2011 Part A: Listeria monocytogenes prevalence
estimates. EFSA Journal, 11, 75 pp.
EFSA, 2014. Technical specifications for the pilot on the collection of data on molecular testing of
food-borne pathogens from food, feed and animal samples. EFSA Supporting Publications, 11,
712E-.
EFSA and ECDC, 2015. The European Union summary report on trends and sources of zoonoses,
zoonotic agents and food-borne outbreaks in 2014. EFSA Journal, 13, 4329,
doi:4310.2903/j.efsa.2015.4329.
Engberg J, 2006. Contributions to the epidemiology of Campylobacter infections. Dan Med Bull, 53,
361-389.
European Commision, 2003. Directive 2003/99/EC of The European Parliament and of the Council of
17 November 2003 on the monitoring of zoonoses and zoonotic agents, amending Council Decision
90/424/EEC and repealing Council Directive 92/117/EEC. Official Journal of the European Union L,
325, 31-40.
European Food Safety Authority, 2010. Panel on Biological Hazards (BIOHAZ); Scientific opinion on
quantification of the risk posed by broiler meat to human campylobacteriosis in the EU. EFSA
Journal, 8, 1437.
Fagerlund A, Langsrud S, Schirmer BC, Mretr T and Heir E, 2016. Genome Analysis of Listeria
monocytogenes Sequence Type 8 Strains Persisting in Salmon and Poultry Processing
Environments and Comparison with Related Strains. PloS one, 11, e0151117.
Flix B, Danan C, Van Walle I, Lailler R, Texier T, Lombard B, Brisabois A and Roussel S, 2014.
Building a molecular Listeria monocytogenes database to centralize and share PFGE typing data
from food, environmental and animal strains throughout Europe. Journal of microbiological
methods, 104, 1-8.
Felix B, Mariet J, Maillet A, Firmesse O, Radomski N, Felten A, Touzain F, Mistou M and Roussel S,
2015. Genomic insight to understand the persistence of Listeria monocytogenes strains in
processing environments of pork products. Proceedings of the Safe Pork.
Franz E, Delaquis P, Morabito S, Beutin L, Gobius K, Rasko DA, Bono J, French N, Osek J and
Lindstedt B-A, 2014. Exploiting the explosion of information associated with whole genome
sequencing to tackle Shiga toxin-producing Escherichia coli (STEC) in global food production
systems. International Journal of Food Microbiology, 187, 57-72.
Gallagher D, Pouillot R, Hoelzer K, Tang J, Dennis SB and Kause JR, 2016. Listeria monocytogenes in
Retail Delicatessens: An Interagency Risk AssessmentRisk Mitigations. Journal of Food
Protection, 79, 1076-1088.

Gandhi M and Chikindas ML, 2007. Listeria: a foodborne pathogen that knows how to survive.
International Journal of Food Microbiology, 113, 1-15.
Gelman A and Rubin DB, 1992. Inference from iterative simulation using multiple sequences.
Statistical science, 457-472.
Gillesberg Lassen S, Ethelberg S, Bjrkman JT, Jensen T, Srensen G, Kvistholm Jensen A, Mller L,
Nielsen EM and Mlbak K, 2016. Two listeria outbreaks caused by smoked fish consumptionusing
whole-genome sequencing for outbreak investigations. Clinical Microbiology and Infection.
Granier SA, Moubareck C, Colaneri C, Lemire A, Roussel S, Dao T-T, Courvalin P and Brisabois A,
2011. Antimicrobial resistance of Listeria monocytogenes isolates from food and the environment
in France over a 10-year period. Applied and environmental microbiology, 77, 2788-2790.
Graves LM and Swaminathan B, 2001. PulseNet standardized protocol for subtyping Listeria
monocytogenes by macrorestriction and pulsed-field gel electrophoresis. International Journal of
Food Microbiology, 65, 55-62.
Graves LM, Swaminathan B and Hunter SB, 2007. Listeria, listeriosis, and food safety. In: Food
science and technology. Eds Ryser ET and Marth EH. 3rd, CRC Press, Boca Raton, 873 p.
Greig JD and Ravel A, 2009. Analysis of foodborne outbreak data reported internationally for source
attribution. International Journal of Food Microbiology, 130, 77-87.
Hald T, Vose D, Wegener HC and Koupeev T, 2004. A Bayesian Approach to Quantify the Contribution
of AnimalFood Sources to Human Salmonellosis. Risk Analysis, 24, 255-269.
Hayward MR, Petrovska L, Jansen VA and Woodward MJ, 2016. Population structure and associated
phenotypes of Salmonella enterica serovars Derby and Mbandaka overlap with host range. BMC
microbiology, 16, 1.
Heck KL, van Belle G and Simberloff D, 1975. Explicit calculation of the rarefaction diversity
measurement and the determination of sufficient sample size. Ecology, 56, 1459-1461.
Heiman KE, Garalde VB, Gronostaj M, Jackson KA, Beam S, Joseph L, Saupe A, Ricotta E, Waechter H
and Wellman A, 2015. Multistate outbreak of listeriosis caused by imported cheese and evidence of
cross-contamination of other cheeses, USA, 2012. Epidemiology and Infection, 1-11.
Hingston PA, Piercey MJ and Truelstrup Hansen L, 2015. Genes associated with desiccation and
osmotic stress in Listeria monocytogenes as revealed by insertional mutagenesis. Applied and
environmental microbiology, 81, 5350-5362.
Holch A, Webb K, Lukjancenko O, Ussery D, Rosenthal BM and Gram L, 2013. Genome sequencing
identifies two nearly unchanged strains of persistent Listeria monocytogenes isolated at two
different fish processing plants sampled 6 years apart. Applied and environmental microbiology,
79, 2944-2951.
Jackson BR, Tarr C, Strain E, Jackson KA, Conrad A, Carleton H, Katz LS, Stroika S, Gould LH and
Mody RK, 2016. Implementation of Nationwide Real-Time Whole-Genome Sequencing to Enhance
Listeriosis Outbreak Detection and Investigation. Clinical Infectious Diseases, ciw242.
Jamali H, Paydar M, Ismail S, Looi CY, Wong WF, Radmehr B and Abedini A, 2015. Prevalence,
antimicrobial susceptibility and virulotyping of Listeria species and Listeria monocytogenes isolated
from open-air fish markets. BMC microbiology, 15, 1.
Jessen B and Lammert L, 2003. Biofilm and disinfection in meat processing plants. International
Biodeterioration & Biodegradation - Hygiene and Disinfection, 51, 265-269.
Jolley KA and Maiden MC, 2010. BIGSdb: Scalable analysis of bacterial genome variation at the
population level. BMC bioinformatics, 11, 1.

Khen B, Lynch O, Carroll J, McDowell D and Duffy G, 2015. Occurrence, antibiotic resistance and
molecular characterization of Listeria monocytogenes in the beef chain in the Republic of Ireland.
Zoonoses and public health, 62, 11-17.
Kvistholm Jensen A, Nielsen EM, Bjrkman JT, Jensen T, Mller L, Persson S, Bjerager G, Perge A,
Krause TG, Kiil K, Srensen G, Andersen JK, Mlbak K and Ethelberg S, 2016. Whole-genome
Sequencing Used to Investigate a Nationwide Outbreak of Listeriosis Caused by Ready-to-eat
Delicatessen Meat, Denmark, 2014. Clinical Infectious Diseases, 63, 64-70.
Kwong JC, Stafford R, Strain E, Stinear TP, Seemann T and Howden BP, 2016. Sharing is caring:
international sharing of data enhances genomic surveillance of Listeria monocytogenes. Clinical
infectious diseases: an official publication of the Infectious Diseases Society of America.
Lamden KH, Fox AJ, Amar CF and Little CL, 2013. A case of foodborne listeriosis linked to a
contaminated food-production process. J Med Microbiol, 62, 1614-1616.
Langmead B and Salzberg SL, 2012. Fast gapped-read alignment with Bowtie 2. Nature methods, 9,
357-359.
Li H and Durbin R, 2010. Fast and accurate long-read alignment with BurrowsWheeler transform.
Bioinformatics, 26, 589-595.
Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G and Durbin R, 2009.
The sequence alignment/map format and SAMtools. Bioinformatics, 25, 2078-2079.
Little CL, Pires SM, Gillespie IA, Grant K and Nichols GL, 2010. Attribution of human Listeria
monocytogenes infections in England and Wales to ready-to-eat food sources placed on the
market: adaptation of the Hald Salmonella source attribution model. Foodborne pathogens and
disease, 7, 749-756.
Loureno A, de Las Heras A, Scortti M, Vazquez-Boland J, Frank JF and Brito L, 2013. Comparison of
Listeria monocytogenes exoproteomes from biofilm and planktonic state: Lmo2504, a protein
associated with biofilms. Applied and environmental microbiology, 79, 6075-6082.
Lundn JM, Autio TJ and Korkeala HJ, 2002. Transfer of persistent Listeria monocytogenes
contamination between food-processing plants associated with a dicing machine. Journal of Food
Protection, 65, 1129-1133.
Lungu B, O'Bryan CA, Muthaiyan A, Milillo SR, Johnson MG, Crandall PG and Ricke SC, 2011. Listeria
monocytogenes: antibiotic resistance in food production. Foodborne pathogens and disease, 8,
569-578.
Manly BF, 2007. Randomization, bootstrap and Monte Carlo methods in biology. CRC Press, pp.
Maury MM, Tsai Y-H, Charlier C, Touchon M, Chenal-Francisque V, Leclercq A, Criscuolo A, Gaultier C,
Roussel S and Brisabois A, 2016. Uncovering Listeria monocytogenes hypervirulence by harnessing
its biodiversity. Nature genetics, 48, 308313.
Mazza R, Mazzette R, McAuliffe O, Jordan K and Fox EM, 2015. Differential gene expression of three
gene targets among persistent and nonpersistent Listeria monocytogenes strains in the presence
or absence of benzethonium chloride. Journal of Food Protection, 78, 1569-1573.
McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, Altshuler D,
Gabriel S and Daly M, 2010. The Genome Analysis Toolkit: a MapReduce framework for analyzing
next-generation DNA sequencing data. Genome research, 20, 1297-1303.
Melo J, Andrew PW and Faleiro ML, 2015. Listeria monocytogenes in cheese and the dairy
environment remains a food safety challenge: The role of stress responses. Food Researche
International, 67, 75-90.
Mettler E and Carpentier B, 1999. Hygienic quality of floors in relation to surface texture. Transaction
of the Institute of Chemical Engineers, 77, 90-96.

Morganti M, Scaltriti E, Cozzolino P, Bolzoni L, Casadei G, Pierantoni M, Foni E and Pongolini S, 2016.
Processing-dependent and clonal contamination patterns of Listeria monocytogenes in the cured
ham food chain revealed by genetic analysis. Applied and environmental microbiology, 82, 822-
831.
Morvan A, Moubareck C, Leclercq A, Herve-Bazin M, Bremont S, Lecuit M, Courvalin P and Le Monnier
A, 2010. Antimicrobial resistance of Listeria monocytogenes strains isolated from humans in
France. Antimicrobial Agents and Chemotherapy, 54, 2728-2731.
Moura A, Criscuolo A, Pouseele H, Maury MM, Leclercq A, Tarr C, Bjrkman JT, Dallman T, Reimer A,
Enouf V, Larsonneur E, Carleton H, Bracq-Dieye H, Katz LS, Jones L, Touchon M, Tourdjman M,
Walker M, Stroika S, Cantinelli T, Chenal-Francisque V, Kucerova Z, Rocha EPC, Nadon C, Grant K,
Nielsen EM, Pot B, Gerner-Smidt P, Lecuit M and Brisse S, 2016. Whole genome-based population
biology and epidemiological surveillance of Listeria monocytogenes. Nature Microbiology, 2, 16185.
Mughini-Gras L and van Pelt W, 2014. Salmonella source attribution based on microbial subtyping:
Does including data on food consumption matter? International Journal of Food Microbiology, 191,
109-115.
Muhterem-Uyar M, Dalmasso M, Bolocan AS, Hernandez M, Kapetanakou AE, Kuchta T, Manios SG,
Melero B, Minaroviov J and Nicolau AI, 2015. Environmental sampling for Listeria monocytogenes
control in food processing facilities reveals three contamination scenarios. Food Control, 51, 94-
107.
Mullner P, Spencer SE, Wilson DJ, Jones G, Noble AD, Midwinter AC, Collins-Emerson JM, Carter P,
Hathaway S and French NP, 2009. Assigning the source of human campylobacteriosis in New
Zealand: a comparative genetic and epidemiological approach. Infection, Genetics and Evolution,
9, 1311-1319.
Nei M, 1975. Molecular population genetics and evolution. North-Holland Publishing Company., pp.
Orsi RH, den Bakker HC and Wiedmann M, 2011. Listeria monocytogenes lineages: Genomics,
evolution, ecology, and phenotypic characteristics. International Journal of Medical Microbiology,
301, 79-96.
Paquet C, Coulombier D, Kaiser R and Ciotti M, 2005. Epidemic intelligence: a new framework for
strengthening disease surveillance in Europe. Euro surveillance: bulletin europeen sur les maladies
transmissibles= European communicable disease bulletin, 11, 212-214.
Perez-Trallero E, Zigorraga C, Artieda J, Alkorta M and Marimon JM, 2014. Two outbreaks of Listeria
monocytogenes infection, Northern Spain. Emerg Infect Dis, 20, 2155-2157.
Pires SM, Evers EG, van Pelt W, Ayers T, Scallan E, Angulo FJ, Havelaar A and Hald T, 2009.
Attributing the human disease burden of foodborne infections to specific sources. Foodborne
pathogens and disease, 6, 417-424.
Pouillot R, Gallagher D, Tang J, Hoelzer K, Kause J and Dennis SB, 2015. Listeria monocytogenes in
Retail Delicatessens: An Interagency Risk AssessmentModel and Baseline Results. Journal of
Food Protection, 78, 134-145.
Pouillot R, Goulet V, Delignette-Muller ML, Mah A and Cornu M, 2009. Quantitative risk assessment of
Listeria monocytogenes in french cold-smoked Salmon: II. Risk characterization. Risk Analysis, 29,
806-819.
Pritchard JK, Stephens M and Donnelly P, 2000. Inference of population structure using multilocus
genotype data. Genetics, 155, 945-959.
Promadej N, Fiedler F, Cossart P, Dramsi S and Kathariou S, 1999. Cell Wall Teichoic Acid
Glycosylation inListeria monocytogenes Serotype 4b Requires gtcA, a Novel, Serogroup-Specific
Gene. Journal of bacteriology, 181, 418-425.

Ragon M, Wirth T, Hollandt F, Lavenir R, Lecuit M, Le Monnier A and Brisse S, 2008. A new
perspective on Listeria monocytogenes evolution. PLoS Pathog, 4, e1000146.
Reij M, Den Aantrekker E and ILSI Europe Risk Analysis in Microbiology Task Force, 2004.
Recontamination as a source of pathogens in processed foods. International Journal of Food
Microbiology, 91, 1-11.
Renier S, Hbraud M and Desvaux M, 2011. Molecular biology of surface colonization by Listeria
monocytogenes: an additional facet of an opportunistic Grampositive foodborne pathogen.
Environmental microbiology, 13, 835-850.
Ryan S, Begley M, Hill C and Gahan C, 2010. A fivegene stress survival islet (SSI1) that contributes
to the growth of Listeria monocytogenes in suboptimal conditions. Journal of Applied Microbiology,
109, 984-995.
Rychli K, Grunert T, Ciolacu L, Zaiser A, Razzazi-Fazeli E, Schmitz-Esser S, Ehling-Schulz M and
Wagner M, 2016. Exoproteome analysis reveals higher abundance of proteins linked to alkaline
stress in persistent Listeria monocytogenes strains. International Journal of Food Microbiology,
218, 17-26.
Rychli K, Mller A, Zaiser A, Schoder D, Allerberger F, Wagner M and Schmitz-Esser S, 2014. Genome
sequencing of Listeria monocytogenes Quargel listeriosis outbreak strains reveals two different
strains with distinct in vitro virulence potential. PloS one, 9, e89964.
Rckerl I, Muhterem-Uyar M, Muri-Klinger S, Wagner K-H, Wagner M and Stessl B, 2014. L.
monocytogenes in a cheese processing facility: Learning from contamination scenarios over three
years of sampling. International Journal of Food Microbiology, 189, 98-105.
Sahl JW, Lemmer D, Travis J, Schupp J, Gillece J, Aziz M, Driebe E, Drees K, Hicks N and Williamson
C, 2016. The Northern Arizona SNP Pipeline (NASP): accurate, flexible, and rapid identification of
SNPs in WGS datasets. bioRxiv, 037267.
Schmid B, Klumpp J, Raimann E, Loessner MJ, Stephan R and Tasara T, 2009. Role of cold shock
proteins in growth of Listeria monocytogenes under cold and osmotic stress conditions. Applied
and environmental microbiology, 75, 1621-1627.
Seemann T, 2014. Prokka: rapid prokaryotic genome annotation. Bioinformatics, btu153.
Sheppard SK, Dallas JF, Strachan NJ, MacRae M, McCarthy ND, Wilson DJ, Gormley FJ, Falush D,
Ogden ID and Maiden MC, 2009. Campylobacter genotyping to determine the source of human
infection. Clinical Infectious Diseases, 48, 1072-1078.
Sheppard SK, Didelot X, Meric G, Torralbo A, Jolley KA, Kelly DJ, Bentley SD, Maiden MC, Parkhill J
and Falush D, 2013. Genome-wide association study identifies vitamin B5 biosynthesis as a host
specificity factor in Campylobacter. Proceedings of the National Academy of Sciences, 110, 11923-
11927.
Simpson EH, 1949. Measurement of diversity. Nature, 163, 688-688.
Smid JH, Gras LM, de Boer AG, French NP, Havelaar AH, Wagenaar JA and van Pelt W, 2013.
Practicalities of using non-local or non-recent multilocus sequence typing data for source
attribution in space and time of human campylobacteriosis. PloS one, 8, e55029.
Stamatakis A, 2014. RAxML version 8: a tool for phylogenetic analysis and post-analysis of large
phylogenies. Bioinformatics, 30, 1312-1313.
Stasiewicz MJ, Oliver HF, Wiedmann M and den Bakker HC, 2015. Whole-genome sequencing allows
for improved identification of persistent Listeria monocytogenes in food-associated environments.
Applied and environmental microbiology, 81, 6024-6037.

Strachan NJ, Rotariu O, MacRae M, Sheppard SK, Smith-Palmer A, Cowden J, Maiden MC and Forbes
KJ, 2013. Operationalising factors that explain the emergence of infectious diseases: a case study
of the human campylobacteriosis epidemic. PloS one, 8, e79331.
Swaminathan B, Gerner-Smidt P, Ng L-K, Lukinmaa S, Kam K-M, Rolando S, Gutirrez EP and
Binsztein N, 2006. Building PulseNet International: an interconnected system of laboratory
networks to facilitate timely public health recognition and response to foodborne disease outbreaks
and emerging foodborne diseases. Foodbourne Pathogens & Disease, 3, 36-50.
Tamura K, Stecher G, Peterson D, Filipski A and Kumar S, 2013. MEGA6: molecular evolutionary
genetics analysis version 6.0. Molecular biology and evolution, 30, 2725-2729.
Tewolde R, Dallman T, Schaefer U, Sheppard CL, Ashton P, Pichon B, Ellington M, Swift C, Green J
and Underwood A (PeerJ Preprints), 2016. MOST: A modified MLST typing tool based on short read
sequencing. 2167-9843.
Thvenot D, Dernburg A and Vernozy-Rozand C, 2006. An updated review of Listeria monocytogenes
in the pork meat industry and its products. Journal of Applied Microbiology, 101, 7-17.
Tocmo R, Krizman K, Khoo WJ, Phua LK, Kim M and Yuk HG, 2014. Listeria monocytogenes in
VacuumPacked Smoked Fish Products: Occurrence, Routes of Contamination, and Potential
Intervention Measures. Comprehensive Reviews in Food Science and Food Safety, 13, 172-189.
Toledo-Arana A, Dussurget O, Nikitas G, Sesto N, Guet-Revillet H, Balestrino D, Loh E, Gripenland J,
Tiensuu T and Vaitkevicius K, 2009. The Listeria transcriptional landscape from saprophytism to
virulence. Nature, 459, 950-956.
Toma B, Vaillancourt J-P, Dufour B, Eloit M, Moutou F, Marsh W, Bnet J, Sanaa M and Michel P,
1999. Dictionary of veterinary epidemiology. Iowa State University Press, pp.
Tourdjman M, Leroux B, Leclercq A, Laurent E, Chenal-Francisque V, King L, Loyer S, Vaillant V,
Donguy M-P, Lecuit M and de Valk H, 2014. pidmie dinfections Listeria monocytogenes lie
la consommation de brie au lait cru France, 2012. Institut de veille sanitaire, 15 p.
Travier L, Guadagnini S, Gouin E, Dufour A, Chenal-Francisque V, Cossart P, Olivo-Marin J-C, Ghigo J-
M, Disson O and Lecuit M, 2013. ActA promotes Listeria monocytogenes aggregation, intestinal
colonization and carriage. PLoS Pathog, 9, e1003131.
Treangen TJ, Ondov BD, Koren S and Phillippy AM, 2014. The Harvest suite for rapid core-genome
alignment and visualization of thousands of intraspecific microbial genomes. Genome biology, 15,
1.
Tremoulet F, Duche O, Namane A, Martinie B, Labadie J and Consortium ELG, 2002. Comparison of
protein patterns of Listeria monocytogenes grown in biofilm or in planktonic mode by proteomic
analysis. FEMS microbiology letters, 210, 25-31.
Wang Q, Holmes N, Martinez E, Howard P, Hill-Cawthorne G and Sintchenko V, 2015. It Is Not All
about Single Nucleotide Polymorphisms: Comparison of Mobile Genetic Elements and Deletions in
Listeria monocytogenes Genomes Links Cases of Hospital-Acquired Listeriosis to the Environmental
Source. J Clin Microbiol, 53, 3492-3500.
Vzquez-Boland JA, Kuhn M, Berche P, Chakraborty T, Dom nguez-Bernal G, Goebel W, Gonzlez-Zorn
B, Wehland J and Kreft J, 2001. Listeria pathogenesis and molecular virulence determinants.
Clinical microbiology reviews, 14, 584-640.
Wieczorek K, Dmowska K and Osek J, 2012. Prevalence, characterization, and antimicrobial resistance
of Listeria monocytogenes isolates from bovine hides and carcasses. Applied and environmental
microbiology, 78, 2043-2045.
Wilson DJ, Gabriel E, Leatherbarrow AJ, Cheesbrough J, Gee S, Bolton E, Fox A, Fearnhead P, Hart CA
and Diggle PJ, 2008. Tracing the source of campylobacteriosis. PLoS Genet, 4, e1000203.

Wulff G, Gram L, Ahrens P and Fonnesbech Vogel B, 2006. One group of genetically similar Listeria
monocytogenes strains frequently dominates and persists in several fish slaughter- and
smokehouses. Applied and environmental microbiology, 72, 4313-4322.
Yde M, Naranjo M, Mattheus W, Stragier P, Pochet B, Beulens K, De Schrijver K, Van den Branden D,
Laisnez V and Flipse W, 2012. Usefulness of the European Epidemic Intelligence Information
System in the management of an outbreak of listeriosis, Belgium, 2011. Eurosurveillance, 17.

Glossary
Food vehicle: Food involved in transmitting a pathogen to a receptive host (RTE foods for L.
monocytogenes).
Food-borne outbreak: Incidence, observed under given circumstances, of two or more human
cases of the same disease and/or infection, or a situation in which the observed number of cases
exceeds the expected number and where the cases are linked, or are probably linked, to the same
food source (European Commission, 2003).
Reservoir: An animate (humans, animals, insects etc.) or inanimate object (plant, raw milk, soil,
surface in contact with food, etc.) or any combination of these serving as a habitat of a pathogen that
produces itself in such a way as to be transmitted to a susceptible host (Toma et al., 1999; European
Food Safety Authority, 2010).
Source: Origin of the pathogen causing infection, including reservoirs and food vehicles.
Source attribution: Partitioning of the human disease burden of one or more foodborne infections
to specific sources, where the term source includes animal reservoirs and vehicles (e.g. foods) (Pires
et al., 2009)
Sporadic case: Case that has not been associated with known outbreaks (Engberg, 2006)
LISEQ database glossary

Best date: gives the closest date from sampling available.
Clinical symptoms: gives the symptoms presented by the patient at the time of isolation.
Context: describes in which framework the sampling was carried out, which may be an outbreak
investigation or a specific research project.
Context level 1: gives a linear numbering of the nine outbreak strains used in this project.
Conventional serotyping: gives the serotyped obtained from serologic agglutination according to
the method of Seeliger and Hhne (1979).
Date of sampling: gives the date when the sample was taken from which listeria was isolated.

EFSA_Code.MTX.mapping: gives the mapping with the EFSA standard sample description 2 with
specific distinction between Food products isolates and Food processing environment isolates (EFSA,
2013a).
EFSA_Complete.MTX.mapping: gives the mapping with the EFSA Foodex 2 language (EFSA,
2013a).
Food matrix: the major food categories were defined according to the classification of EFSA risk-food
matrices (EFSA, 2013).
Food origin: describes the type of animal or vegetable that is the main component of the food
product. Composite food products including more than one animal or vegetable, Mixed sources is
specified. It refers to source as described in the glossary under Section 12.
Food origin level 1: specifies the type of fish species when documented.
Food product: specific category describing the type of products further obtained from a given food
matrix. These definitions follow the EFSA guidance on listeria risk (EFSA, 2012)
Geographic information: provide information on the geographic area of sampling.
Molecular serotyping: gives the serotype according to the method of the EURL for Lm. This method
respects the international reference method established by Doumith et al. (2004).
Reception date: gives the date when the strain or the sample was received in the laboratory. It may
be different from the sampling date.
Sample type: describes the isolation of the strains from food product (EFSA code: S019A Food
sample) or from food processing environment (EFSA code: S027A Environmental sample). Food
processing environments include all types of samples (food contact surfaces or non-food contact
surfaces) obtained from the place where the food product is processed.
Sampling stage: gives the level in the food chain where the sample was taken.
Sector: distinguishes clinical samples isolated from human pathology and non-human isolates.
List of abbreviations used in the report

7-MLST: 7 locus multi-locus sequence typing
BLS: base line survey
CC: clonal complex
cgMLST: core genome multi-locus sequence typing
EFSA: European Food Safety Agency
ECDC: European Centre of Disease Prevention and Control
fAFLP: fluorescent amplified-fragment length polymorphism
GWAS: genome-wide association study
IP scheme: Institute Pasteur scheme
MLST: multi-locus sequence typing
MOST: Metric oriented sequence typer
rMLST: ribosomal multi locus sequence typing
PFGE: pulsed-field gel electrophoresis
SNP: single-nucleotide polymorphism
ST: sequence type
TESSy: the European surveillance system

Appendix 1: Isolates from the EU-wide baseline survey on prevalence of L.

monocytogenes in certain RTE foods conducted in 2010-2012
RL_number Country Year Sample type Description MLST MLST

Clonal sequence
complexes types
RL15000006 C 2010 Food products Fish and fishery products-Fish-Salmo spp. CC121 121
(Salmon)-Fish origin not specified-Smoked
processing not specified
RL15000008 C 2011 Food products Fish and fishery products-Fish-Clupea CC14 14
harengus (Herring, kipper)-Wild fish-Smoked
RL15000010 C 2011 Food products Meat and meat products-Mixed sources--Deli CC121 121
products - Pate-Cooked
(Salmon)-Fish origin not specified-
RL15000013 C 2011 Food products Milk and milk products-Bovine--Soft cheese- CC20 20
Not specified
RL15000016 C 2011 Food products Fish and fishery products-Fish-Oncorhynchus CC101 101
mykiss, Salmo trutta (Trout)-Fish origin not
specified-Smoked processing not specified
harengus (Herring, kipper)-Wild fish-Filet
RL15000020 C 2011 Food products Fish and fishery products-Fish--Fish origin not CC204 204
RL15000022 C 2011 Food products Meat and meat products-Mixed sources--Deli CC204 204
products - Pate-
RL15000025 R 2011 Food products Meat and meat products-Unspecified--Deli CC5 5
product - Sliced-Cooked
products - Pate-
RL15000027 R 2011 Food products Fish and fishery products-Fish--Fish origin not CC155 155
specified-Cold smoked
product - Other product-Other stabilization


Clonal sequence
complexes types
RL15000029 F 2011 Food products Fish and fishery products-Fish--Fish origin not CC8 8
specified-Warm smoked
specified-Gravad/slightly salted
RL15000044 G 2011 Food products Meat and meat products-Swine--Deli product - CC204 204
Sliced-Cooked
RL15000045 G 2011 Food products Fish and fishery products-Fish--Fish origin not CC9 9
RL15000046 G 2011 Food products Fish and fishery products-Fish--Fish origin not ST124 124
RL15000047 G 2011 Food products Fish and fishery products-Fish--Fish origin not ST124 124
RL15000048 N 2011 Food products Fish and fishery products-Fish--Wild fish-Warm CC6 6
smoked
RL15000049 N 2011 Food products Fish and fishery products-Fish--Wild fish-Cold CC155 155
smoked
smoked
smoked
smoked
smoked
smoked
smoked
smoked
RL15000057 S 2011 Food products Fish and fishery products-Fish--Farmed fish- CC8 8
Cold smoked
RL15000058 S 2011 Food products Fish and fishery products-Fish--Wild fish-Cold CC2 2
smoked
RL15000059 S 2011 Food products Fish and fishery products-Fish--Farmed fish- CC8 8
Cold smoked


Clonal sequence
complexes types
RL15000060 S 2011 Food products Fish and fishery products-Fish--Wild fish- CC121 121
Gravad/slightly salted
RL15000061 B 2010 Food products Fish and fishery products-Fish--Fish origin not CC121 121
RL15000064 B 2011 Food products Fish and fishery products-Fish--Wild fish- CC31 31
Smoked processing not specified
RL15000065 U 2010 Food products Fish and fishery products-Fish--Fish origin not CC3 3
RL15000066 U 2010 Food products Meat and meat products-Turkeys--Deli product CC7 12
- Sliced-Cooked
RL15000081 U 2011 Food products Meat and meat products-Turkeys--Deli product CC3 3
- Sliced-Cooked
RL15000089 U 2011 Food products Meat and meat products-Swine--Deli product - CC7 12
Sliced-Cooked
RL15000090 U 2011 Food products Meat and meat products-Mixed sources--Deli CC87 87


Clonal sequence
complexes types
RL15000096 U 2011 Food products Milk and milk products-Unspecified--Semi soft CC14 91
cheese-Made from pasteurized milk
RL15000099 U 2011 Food products Meat and meat products-Swine--Deli product - CC7 12
Sliced-Cooked
RL15000103 U 2011 Food products Meat and meat products-Gallus gallus (fowl)-- CC2 145
Deli products - Pate-Other stabilization


Clonal sequence
complexes types
RL15000128 E 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121
specified-
specified-
specified-
specified-
specified-
specified-
RL15000134 Z 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121
RL15000137 Z 2011 Food products Meat and meat products-Swine--Deli product - CC9 9
Sliced-Cooked
Sliced-Cooked
Sliced-Cooked
Sliced-Cooked
RL15000142 Z 2011 Food products Fish and Fishery products-Fish--Fish origin not CC121 121
RL15000148 Z 2011 Food products Fish and fishery products-Fish--Wild fish- CC121 121
harengus (Kipper)-Fish origin not specified-
Filet
(Salmon)-Farmed fish-
(Salmon)-Fish origin not specified-Raw
RL15000152 A 2010 Food products Fish and fishery products-Fish--Wild fish- CC101 101
RL15000153 A 2010 Food products Meat and meat products-Unspecified--Deli CC31 31


Clonal sequence
complexes types
RL15000157 A 2011 Food products Fish and fishery products-Fish--Wild fish-Warm CC6 6
smoked
RL15000158 P 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121
RL15000159 P 2011 Food products Meat and meat products-Swine--Deli product - CC1 1
Other product-Cooked
RL15000160 P 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121
RL15000161 C 2011 Food products Meat and meat products-Gallus gallus (fowl)-- CC121 121
Fresh meat - Cut-Raw
harengus (Herring, kipper)-Wild fish-Raw
RL15000167 C 2011 Food products Fish and fishery products-Fish-Oncorhynchus CC121 121
specified-Raw
RL15000172 J 2010 Food products Fish and fishery products-Fish--Wild fish-Cold CC193 193
smoked
RL15000173 J 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121
RL15000174 J 2011 Food products Fish and fishery products-Fish--Wild fish-Warm CC121 121
smoked
RL15000175 J 2011 Food products Meat and meat products-Gallus gallus (fowl)-- CC155 155
Deli product - Other product-Other
stabilization
RL15000178 M 2011 Food products Meat and meat products-Fish-Salmo spp. CC121 121
(Salmon)-Deli product - Sliced-Cooked
RL15000179 M 2011 Food products Meat and meat products-Swine--Deli product - CC8 8
Sliced-Cooked
RL15000180 V 2011 Food products Fish and fishery products-Fish--Farmed fish- CC6 6
Cold smoked
RL15000181 V 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121


Clonal sequence
complexes types
RL15000185 V 2010 Food products Fish and fishery products-Fish--Farmed fish- CC193 193
RL15000186 A 2011 Food products Fish and fishery products-Fish--Farmed fish- CC121 121
RL15000189 A 2011 Food products Milk and milk products-Bovine--Semi soft CC31 325
cheese-Made from pasteurized milk
RL15000190 W 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121
RL15000197 L 2010 Food products Fish and fishery products-Fish--Fish origin not CC121 121
RL15000198 L 2010 Food products Meat and meat products-Mixed sources--Deli CC6 6
product - Sausage-Cooked
RL15000200 L 2010 Food products Milk and milk products-Bovine--Semi soft CC1 1
cheese-Made from raw milk
RL15000202 L 2010 Food products Meat and meat products-Swine--Deli products CC121 236
- Pate-Cooked
RL15000203 L 2010 Food products Meat and meat products-Swine--Deli products CC121 236
- Pate-Cooked
RL15000209 L 2010 Food products Meat and meat products-Swine--Deli product - CC3 3
Sausage-Cooked


Clonal sequence
complexes types
Sausage-Cooked
RL15000221 L 2010 Food products Meat and meat products-Mixed sources--Deli CC2 2
product - Sausage-Cooked
Sausage-Cooked
Sausage-Cooked
Sausage-Cooked
RL15000245 Y 2010 Food products Fish and fishery products-Fish--Farmed fish- CC8 120


Clonal sequence
complexes types
RL15000247 Q 2011 Food products Fish and fishery products-Fish--Farmed fish- CC59 59
Smoked (process not specify)
RL15000248 Q 2011 Food products Fish and fishery products-Fish--Fish origin not CC59 59
specified-Smoked (process not specify)
specified-
specified-
Warm smoked
RL15000271 Q 2011 Food products Milk and milk products-Unspecified--Semi soft CC31 325
cheese-
RL15000272 Q 2011 Food products Milk and milk products-Unspecified--Semi soft CC31 325
cheese-
specified-


Clonal sequence
complexes types
RL15000278 Q 2011 Food products Fish and fishery products-Fish--Wild fish-Cold CC9 9
smoked
Cold smoked
Warm smoked
specified-
RL15000286 T 2010 Food products Fish and fishery products-Fish--Fish and CC3 44
fishery products-Cold smoked
RL15000291 K 2010 Food products Fish and fishery products-Fish--Fish origin not CC8 8
RL15000305 Q 2011 Food products Meat and meat products-Swine--Deli product - CC8 8
Other meat products-
RL15000329 L 2010 Food products Fish and fishery products-Fish--Fish origin not - *


Clonal sequence
complexes types
specified-
specified-
specified-
smoked
smoked
RL15000658 X 2010 Food products Milk and milk products-Goat--Soft cheese- CC7 7
Made from pasteurized milk
RL15000659 X 2010 Food products Meat and meat products-Turkeys--Deli product CC9 9
- Sliced-Cooked
RL15000660 X 2010 Food products Meat and meat products-Swine--Deli product - CC121 121
Ham-Smoked processing not specified
Ham-Smoked processing not specified
RL15000662 X 2010 Food products Fish and fishery products-Fish-Salmo spp. CC9 9
RL15000663 X 2010 Food products Fish and fishery products-Fish-Oncorhynchus CC8 16
RL15000664 X 2011 Food products Meat and meat products-Gallus gallus (fowl)-- CC121 121
Deli product - Other product-
Sliced-Cooked
Deli product - Sliced-Cooked
RL15000673 X 2011 Food products Meat and meat products-Mixed sources--Deli CC2 2


Clonal sequence
complexes types
Sliced-Cooked
Sliced-Cooked
RL15000682 X 2011 Food products Meat and meat products-Geese--Deli product - CC121 121
Sliced-Cooked
Sliced-Cooked
Sliced-Cooked
RL15000688 X 2011 Food products Meat and meat products-Swine--Deli products CC6 6
- Pate-Cooked
RL15000692 X 2011 Food products Fish and fishery products-Fish--Fish origin not CC121 121
Sliced-Cooked
RL15000730 H 2010 Food products Milk and milk products-Unspecified--Cheese CC204 204
category not specified-
RL15000731 H 2010 Food products Fish and fishery products-Fish-Salmo spp. CC121 121
RL15000732 H 2010 Food products Milk and milk products-Unspecified--Cheese CC204 204


Clonal sequence
complexes types
*NOVEL allele. Cannot determine closest ST (SLV).

Appendix 2: Isolates other food, ready-to-eat meat and cheese

RL_number Country Year Sample type Description MLST Clonal MLST
complexes sequence
types
RL15000342 V 2010 Food products Meat and meat products-Unspecified--- CC5 5
RL15000344 Z 2014 Food products Meat and meat products-Bovine--Meat - CC77 77
Sausage-
Sausage-
RL15000346 G 2011 Food products Meat and meat products-Swine--Deli CC31 31
product - Ham-
product - Ham-
product - Ham-
Minced-
product - Ham-
RL15000351 Z 2014 Food products Elaborated food products combining CC7 7
several food categories-Mixed sources--
Sandwich-
Minced-
Minced-
Sausage-
RL15000355 Z 2013 Food products Meat and meat products-Swine--Deli CC101 101
product - Ham-Smoked processing not
specified
RL15000356 Z 2013 Food products Elaborated food products combining CC9 9
Sandwich-
RL15000357 Z 2013 Food products Meat and meat products-Gallus gallus CC9 9
(fowl)--Meat - Cut-Smoked processing
not specified
RL15000358 Z 2013 Food products Meat and meat products-Unspecified-- CC11 451
Deli product - Sausage-
Minced-
RL15000360 Z 2012 Food products Meat and meat products-Gallus gallus CC121 121
(fowl)--Meat - Cut-
RL15000394 C 2010 Food products Milk and milk products-Bovine--Semi soft CC155 155
cheese-
RL15000395 C 2010 Food products Milk and milk products-Bovine--Hard CC77 77
cheese-
RL15000396 C 2010 Food products Milk and milk products-Unspecified-- CC207 207
Cheese category not specified-
cheese-
RL15000398 C 2010 Food products Milk and milk products-Bovine--Cream- CC101 101
RL15000399 C 2010 Food products Milk and milk products-Bovine--Soft CC54 54
cheese-
RL15000402 C 2010 Food products Milk and milk products-Goat--Fresh CC2 2
cheese-


complexes sequence
types
RL15000403 C 2010 Food products Milk and milk products-Bovine--Cheese CC217 217
cheese-
cheese-
RL15000408 P 2011 Food products Milk and milk products-Bovine--Cheese CC9 9
RL15000409 P 2011 Food products Milk and milk products-Bovine--Cream- CC54 54
RL15000410 P 2011 Food products Milk and milk products-Bovine--Butter- CC3 174
RL15000411 P 2011 Food products Milk and milk products-Bovine--Cheese CC121 121
Melted cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-
cheese-


complexes sequence
types
cheese-
RL15000436 A 2012 Food products Milk and milk products-Bovine--Soft CC8 8
cheese-
RL15000437 V 2010 Food products Milk and milk products-Unspecified-- CC14 14
RL15000438 V 2010 Food products Meat and meat products-Unspecified-- CC37 37
Meat - Cut-Smoked processing not
specified
RL15000439 V 2010 Food products Milk and milk products-Sheep--Cheese CC21 21
RL15000452 V 2011 Food products Milk and milk products-Unspecified--Hard CC21 21
cheese-
RL15000459 V 2012 Food products Milk and milk products-Sheep--Cheese ST200 200
RL15000461 C 2010 Food products Meat and meat products-Swine--Deli CC9 9
product - Sausage-
RL15000462 C 2011 Food products Meat and meat products-Unspecified-- CC121 121
product - Other product-
product - Sausage-


complexes sequence
types
RL15000466 C 2011 Food products Meat and meat products-Poultry not CC121 121
specified--Deli product - Other product-
RL15000467 C 2011 Food products Meat and meat products-Ducks--Deli CC204 204
RL15000482 C 2011 Food products Meat and meat products-Mixed sources-- CC6 6
product - Sausage-
RL15000485 C 2011 Food products Meat and meat products-Gallus gallus CC121 121
(fowl)--Deli products - Pate-
product - Other product-Sliced
Deli products - Pate-
Meat - Minced-Raw
products - Pate-
product - Sausage-


complexes sequence
types
products - Pate-
product - Sausage-
RL15000506 A 2010 Food products Meat and meat products-Gallus gallus CC8 16
(fowl)--Meat - Cut-
RL15000507 A 2010 Food products Meat and meat products-Swine--Deli CC2 2
product - Ham-Sliced
RL15000508 A 2010 Food products Meat and meat products-Unspecified-- CC9 9
Meat - Minced-Cooked
product - Ham-Smoked processing not
specified
RL15000510 A 2010 Food products Meat and meat products-Swine--Meat - CC9 9
Minced-Smoked processing not specified
Other product-
Other product-
RL15000513 A 2010 Food products Meat and meat products-Sheep--Meat - CC9 9
Cut-Smoked processing not specified
RL15000514 A 2011 Food products Elaborated food products combining CC9 9
Soups-
Meat - Cut-Cooked
Meat - Cut-Cooked
Meat - Cut-Cooked
Meat - Cut-Cooked
Meat - Cut-
Meat - Cut-Sliced
Meat - Cut-
Meat - Cut-Sliced
(fowl)--Meat - Cut-
(fowl)--Meat - Cut-
Meat - Cut-Sliced


complexes sequence
types
Cut-Cooked
Deli product - Other product-Smoked
Deli product - Other product-Smoked
product - Ham-
RL15000531 A 2011 Food products Meat and meat products-Bovine--Meat - CC9 9
Cut-Cooked
Meat - Cut-Sliced
RL15000533 A 2011 Food products Meat and meat products-Bovine--Meat - CC9 9
Cut-
RL15000535 V 2010 Food products Meat and meat products-Unspecified--- CC7 *7
RL15000696 X 2010 Food products Milk and milk products-Unspecified--Hard CC1 328
cheese-
RL15000697 X 2010 Food products Milk and milk products-Unspecified-- CC1 1
RL15000698 X 2010 Food products Milk and milk products-Unspecified-- CC1 1
RL15000699 X 2011 Food products Milk and milk products-Unspecified--Hard ST382 183
cheese-
RL15000700 X 2011 Food products Milk and milk products-Unspecified--Hard CC3 3
cheese-
RL15000701 X 2010 Food products Meat and meat products-Swine--Meat - CC9 9
Cut-
RL15000702 X 2010 Food products Meat and meat products-Swine--Deli CC31 31
product - Sausage-
product - Sausage-
Sausage-
RL15000705 X 2010 Food products Meat and meat products-Gallus gallus CC121 121
(fowl)--Meat - Cut-Cooked
product - Sausage-
RL15000707 X 2010 Food products Elaborated food products combining CC121 121
Sandwich-
product - Sausage-
Ready made meal-
RL15000710 X 2010 Food products Meat and meat products-Ducks--Deli CC5 5
product - Ham-
Sausage-
product - Sausage-
product - Sausage-
product - Sausage-
Sausage-
product - Sausage-


complexes sequence
types
product - Sausage-
Ready made meal-
RL15000719 X 2010 Food products Meat and meat products-Bovine--Meat - CC3 3
Cut-
RL15000720 X 2011 Food products Meat and meat products-Unspecified-- CC121 121
Meat - Minced-
RL15000721 X 2011 Food products Meat and meat products-Unspecified-- CC2 2
Meat - Cut-Cooked
product - Sausage-
product - Sausage-
Ready to eat salad-
Ready to eat salad-
Ready to eat salad-
Ready to eat salad-
Ready to eat salad-
Ready to eat salad-
RL15001282 B 2010 Food products Milk and milk products-Bovine--Cream- CC8 8
RL15001292 B 2010 Food products Milk and milk products-Unspecified-- CC37 37
RL15001293 B 2010 Food products Meat and meat products-Swine--Meat - CC59 59
Cut-
RL15001294 B 2010 Food products Meat and meat products-Swine--Meat - CC21 21
Cut-Cooked
RL15001295 B 2010 Food products Meat and meat products-Unspecified-- CC3 3
Meat - Cut-
RL15001299 B 2010 Food products Meat and meat products-Bovine--Meat - CC121 121
Cut-
RL15001300 B 2010 Food products Meat and meat products-Poultry not CC121 121
specified--Deli product - Other product-
RL15001301 B 2011 Food products Meat and meat products-Swine--Deli CC121 121
product - Ham-
RL15001311 B 2011 Food products Milk and milk products-Unspecified--Ice CC218 218
cream-
RL15001312 B 2011 Food products Milk and milk products-Bovine--Semi soft CC224 224
cheese-
RL15001332 B 2011 Food products Milk and milk products-Bovine--Cream- CC8 8
RL15001333 B 2011 Food products Milk and milk products-Bovine--Hard CC29 427
cheese-
RL15001334 B 2011 Food products Meat and meat products-Gallus gallus CC9 9
(fowl)--Meat - Cut-
product - Sausage-
RL15001346 B 2011 Food products Milk and milk products-Unspecified--Milk- CC37 37


complexes sequence
types
Cut-Sliced
product - Ham-
RL15001349 B 2011 Food products Meat and meat products-Unspecified-- CC9 9
Meat - Cut-
RL15001352 B 2011 Food products Meat and meat products-Bovine--Deli CC6 6
Cut-Sliced
RL15001383 B 2010 Food products Milk and milk products-Goat--Soft cheese- CC59 59
RL15001384 B 2010 Food products Milk and milk products-Bovine--Semi soft CC37 37
cheese-
product - Ham-

Appendix 3: Isolates from other food, fruits and vegetables

Clonal sequence
complexes types
RL15001981 B 2010 Food products Fruit, vegetables, cereals and herbs-Vegetal- CC121 121
-Vegetable-
-Fruit-
-Vegetable-
RL15001985 B 2011 Food products Fruit, vegetables, cereals and herbs-Vegetal- ST839 839
-Vegetable-
-Fruit-

Appendix 4: Isolates from the food production chain

Clonal sequence
complexes types
RL10000011 Q Food processing Milk and milk products-Sheep-- CC9 9
environment Fresh cheese-
RL10000012 Q Food products Milk and milk products-Sheep-- CC18 18
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-
Fresh cheese-


Clonal sequence
complexes types
RL10000052 Q Food processing Milk and milk products-Sheep--Milk- CC29 29
environment
environment
environment


Clonal sequence
complexes types


Clonal sequence
complexes types
RL15000361 C 2009 Food products Meat and meat products-Swine-- CC2 2
Meat - Cut-
RL15000362 C 2010 Food products Meat and meat products- CC6 6
Unspecified--Meat - Cut-
Meat - Cut-
Meat - Cut-
Meat - Cut-
Meat - Cut-
Deli product - Ham-
RL15000371 C 2010 Food products Meat and meat products-Swine-- ST191 191
Meat - Cut-
Unspecified--Meat - Sausage-
RL15000374 C 2011 Food products Meat and meat products-Swine-- ST191 191
Meat - Cut-
Deli product - Ham-
Meat - Cut-
Meat - Cut-
RL15000387 C 2008 Food processing Meat and meat products-Swine--- CC8 8
environment
RL15000388 C 2008 Food processing Meat and meat products-Swine--- ST602 602
environment
environment
environment
environment


Clonal sequence
complexes types
environment
environment
Meat - Cut-
Deli product - Ham-
environment
environment
environment
environment
environment
environment
RL15000619 B 2002 Food products Fish and fishery products-Fish- CC121 121
Salmo spp. (Salmon)-Fish origin not
specified-
specified-
RL15000621 B 2011 Food processing Fish and fishery products-Fish- CC121 121
environment Salmo spp. (Salmon)-Fish origin not
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-


Clonal sequence
complexes types
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-
specified-


Clonal sequence
complexes types
specified-
RL15001296 B 2010 Food processing Meat and meat products- CC9 9
environment Unspecified---
RL15001297 B 2010 Food processing Meat and meat products-Mixed CC121 121
environment animal source---
RL15001298 B 2010 Food processing Milk and milk products-Unspecified-- CC31 31
environment Cheese category not specified-
RL15001305 B 2010 Food processing Elaborated food products combining CC2 2
environment several food categories-Mixed
sources--Sandwich-
sources--Sandwich-
sources--Sandwich-
sources--Sandwich-
environment Ice cream-
environment Ice cream-


Clonal sequence
complexes types
sources--Sandwich-

Appendix 5: Isolates from sporadic clinical cases
RL_number Country Year Clinical symptoms MLST Clonal MLST sequence

complexes types
RL15000306 Q 2011 Unknown CC101 101
RL15000307 Q 2010 Unknown CC101 38
RL15000308 Q 2010 Unknown CC101 38
RL15000309 Q 2010 Unknown CC6 6
RL15000310 Q 2010 Unknown CC101 38
RL15000311 Q 2010 Unknown CC398 398
RL15000312 Q 2010 Unknown CC101 38
RL15000313 Q 2010 Unknown CC1 1
RL15000314 Q 2010 Unknown CC9 9
RL15000315 Q 2010 Unknown CC8 8
RL15000316 Q 2010 Unknown CC101 38
RL15000317 Q 2010 Unknown CC101 38
RL15000318 Q 2010 Unknown ST560 560
RL15000319 Q 2010 Unknown CC101 38
RL15000320 Q 2010 Unknown CC7 7
RL15000321 Q 2010 Unknown CC1 1
RL15000322 Q 2010 Unknown CC6 6
RL15000323 Q 2010 Unknown CC8 8
RL15000324 Q 2010 Unknown CC5 5
RL15000325 Q 2010 Unknown CC3 287
RL15000326 Q 2010 Unknown CC1 1
RL15000327 Q 2010 Unknown CC101 38
RL15000328 Q 2010 Unknown CC37 37
RL15001278 B 2010 Unknown CC6 178
RL15001279 B 2010 Bacteremia CC9 9
RL15001280 B 2010 Pregnancy related CC6 6
RL15001284 B 2010 Meningitis CC1 1
RL15001286 B 2010 Unknown CC1 1
RL15001288 B 2010 Other ST736 736
RL15001290 B 2010 Other CC8 16
RL15001328 B 2011 Unknown CC220 220


complexes types
RL15001331 B 2011 Unknown CC220 220
RL15001339 B 2011 Unknown CC6 6
RL15001340 B 2011 Other CC1 1
RL15001381 B 2010 Meningitis ST392 392
RL15001414 A 2010 Bacteremia CC1 1
RL15001425 A 2011 Other CC9 9
RL15001426 A 2011 Other CC1 1
RL15001427 A 2011 Meningitis CC101 101
RL15001527 F 2010 Other CC5 5
RL15001528 F 2010 Pregnancy related CC101 101
RL15001529 F 2010 Meningitis CC2 2
RL15001530 F 2010 Unknown CC3 3


complexes types
RL15001532 F 2011 Unknown CC2 2
RL15001534 F 2011 Unknown CC3 3
RL15001535 W 2011 Bacteremia CC3 3
RL15001536 W 2010 Meningitis ST570 570
RL15001539 W 2010 Meningitis CC7 7
RL15001540 W 2010 Other CC6 6
RL15001541 W 2011 Other ST32 32
RL15001546 W 2010 Meningitis ST184 184
RL15001548 W 2010 Other CC2 2
RL15001550 Y 2011 Unknown CC1 1
RL15001551 D 2011 Meningitis CC21 21
RL15001552 Y 2011 Unknown CC1 1
RL15001553 Y 2011 Unknown CC8 8
RL15001554 Y 2011 Unknown CC8 8
RL15001555 Y 2011 Unknown CC18 18
RL15001556 Y 2011 Unknown CC398 398
RL15001557 Y 2011 Unknown CC59 59
RL15001558 Y 2011 Unknown CC14 399
RL15001559 Y 2011 Unknown CC14 14
RL15001560 Y 2011 Unknown CC155 155
RL15001561 Y 2011 Unknown CC8 8
RL15001562 Y 2011 Unknown CC11 451
RL15001563 T 2010 Bacteremia CC31 31


complexes types
RL15001574 T 2011 Unknown CC177 180
RL15001578 T 2011 Meningitis CC7 7
RL15001583 X 2010 Other CC155 155
RL15001584 X 2010 Unknown CC1 1
RL15001585 X 2010 Bacteremia CC1 1
RL15001586 X 2010 Meningitis CC1 181
RL15001589 X 2010 Pregnancy related CC14 14
RL15001590 X 2010 Other CC6 6
RL15001591 X 2010 Other CC1 1
RL15001597 X 2011 Other CC37 37
RL15001607 X 2011 Other CC204 204
RL15001611 X 2011 Other CC7 7


complexes types
RL15001617 X 2011 Other CC1 1
RL15001618 D 2010 Pregnancy related CC7 7
RL15001619 D 2010 Bacteremia CC9 9
RL15001630 D 2011 Other CC379 182
RL15001634 D 2011 Other CC1 1
RL15001636 Z 2010 Other CC1 1
RL15001637 Z 2010 Pregnancy related CC2 2
RL15001638 Z 2010 Unknown CC8 8
RL15001639 Z 2010 Unknown CC1 515
RL15001640 Z 2010 Meningitis CC8 8
RL15001642 Z 2010 Unknown CC8 8
RL15001643 Z 2010 Unknown CC6 6
RL15001644 Z 2010 Unknown CC6 6
RL15001645 Z 2010 Unknown CC2 2
RL15001646 Z 2011 Other CC379 808
RL15001650 Z 2011 Unknown CC9 9
RL15001651 Z 2011 Bacteremia CC224 224
RL15001652 Z 2011 Bacteremia CC6 6
RL15001654 Z 2011 Unknown ST773 773


complexes types
RL15001737 D 2011 Other CC6 6
RL15002422 C 2010 Bacteremia CC101 101
RL15002423 C 2010 Meningitis CC1 1
RL15002427 C 2010 Pregnancy related CC1 1
RL15003001 Y 2010 Bacteremia CC8 8
RL15003002 Y 2010 Unknown CC7 7
RL15003003 Y 2010 Unknown CC1 1
RL15003004 Y 2010 Unknown CC4 4


complexes types
RL15003005 Y 2010 Unknown CC101 101
RL15003006 Y 2010 Unknown CC398 398
RL15003007 Y 2010 Unknown CC7 7
RL15003008 Y 2010 Unknown CC8 8
RL15003021 A 2010 Pregnancy related CC2 2
RL15003022 A 2010 Pregnancy related CC2 2
RL15003025 A 2011 Other CC8 8
RL15003029 A 2011 Meningitis CC6 *6

Appendix 6: Isolates from outbreaks
RL_number Country Year Food/Human Clinical Outbreak MLST Clonal MLST

/Environment symptomsa number complexes sequence
types
RL15001313 B 2011 Food Not applicable Outbreak 4 CC59 59
RL15001336 B 2011 Human Bacteremia Outbreak 4 CC59 59
RL15001355 B 2012 Human Other Outbreak 1 CC155 155
RL15001362 B 2012 Environment Not applicable Outbreak 1 CC155 155
RL15001368 B 2014 Human Unknown Outbreak 5 CC475 504
RL15001395 B 2013 Food Not applicable Outbreak 3 CC7 *7


types
RL15001656 X 2012 Human Pregnancy Outbreak 8 CC14 14
related
related
RL15001658 X 2012 Human Meningitis Outbreak 8 CC14 14
related
RL15001660 X 2012 Food Not applicable Outbreak 8 CC14 14
related
related
related
RL15001672 X 2013 Human Bacteremia Outbreak 7 CC87 87
RL15001673 X 2013 Human Meningitis Outbreak 7 CC87 87
related
related
related
related
related


types
RL15002457 C 2012 Human Other Outbreak 9 CC4 4
RL15002458 C 2012 Human Meningitis Outbreak 9 CC4 4
RL15002460 C 2012 Human Other Outbreak 9 CC4 4
RL15002461 C 2012 Human Bacteremia Outbreak 9 CC4 4
RL15002463 C 2012 Human Pregnancy Outbreak 9 CC4 4
related
related
related
related
related
related
related
RL15003010 T 2013 Human Meningitis Outbreak 6 CC398 802
RL15003011 T 2013 Food Not applicable Outbreak 6 CC398 802
RL15003012 T 2013 Human Other Outbreak 6 CC398 802
RL15003013 T 2013 Human Bacteremia Outbreak 6 CC398 802
RL15003014 T 2013 Human Meningitis Outbreak 6 CC398 802
Clinical symptoms are not relevant for food or environmental isolates.

Appendix 7: Rarefaction and Simpsons diversity index of 7 locus MLST

clinical data stratified by age
Simpson's diversity index was determined for isolates from humans <60 years or >60 years old,
respectively (Figure 21.1). Both age groups were equally diverse (Simpsons index = 0.8810.058 for
<60 year old and 0.9230.032 for >60 years old, respectively) (P>0.05).
1
Simpson's diversity index
0.8
0.6
0.4
0.2
0
Human_<60yrs Human_>=60yrs
Source
Figure 21.1.: Simpsons diversity index of human isolates based on 7 locus MLST data stratified by
age
Rarefaction curves of clinical isolates stratified by age were plotted for 7 locus MLST data (Figure
21.2). The number of new STs per genome is similar (i.e. the two rarefaction curves appear to be
virtually identical) for both age groups (either 60 or <60 years old).

45
40
35
30
Number of new STs
25
20 Human_>60 yrs
15 Human_<60 yrs
10
0
0 20 40 60 80 100 120 140
Number of Genomes
Figure 21.2.: Rarefaction of human isolates based on 7 locus MLST data stratified by age
Neis genetic distance was determined between humans < 60 and 60 years of age. It was found
to be 0.28 (95% CIs 0.22-0.35). The distance between the two age groups was not significant by
randomisation test (P=0.141).

Appendix 8: Links to Attribution model software

There were five attribution models that were used in the study. The github link to the software for
these programs is:
https://github.com/lguillier/LISEQ-codes/tree/master/Chapter7
Listed below are brief instructions on how to access and utilse these.
The Dutch model
An example of the Dutch model for 30 locus rMLST, 864 isolates and 10,000 runs, is given in
DutchModelrMLST_v5_SampleSizeCorrection_Attribution_Human.7z
This needs to be extracted using ZIP software. This will produce a .xlsm Excel file.
The program runs under VBA Excel. The input data have to be placed in spreadsheet Program,
starting with column X in the format given in the example.
The attribution scores will be displayed in the columns J,K,L, depending on the number of sources.
These scores have to copied to the spreadsheet Results where the attribution graphic is displayed.
The modified Hald model
The modified Hald model runs under WinBugs which can be downloaded from
http://www.mrc-bsu.cam.ac.uk/software/bugs/the-bugs-project-winbugs/
The prevalence sub-model,
Attribution_Listeria_Prevalence_Hald_EFSA.odc
has to run first and the results have to be fed into the main model,
Attribution_Listeria_RealModel_Hald_EFSA.odc
STRUCTURE model
The program can be downloaded from:
http://pritchardlab.stanford.edu/structure.html
The Asymmetric Island (AI) model (iSource)
The program can be downloaded from:
http://www.danielwilson.me.uk/iSource.html

The Aberdeen model
The model was implemented in Mathematica and it can be run in any Linux system with Mathematica
installed. An example of the model and associated files (AbdnAttribution_Mathematica.tgz).
In the following instructions on how to install the files and to run the model are given.
*** Installation:
- Uncompress the file in the directory you wish. On a Unix shell,
$ tar -xzf AbdnAttribution_Mathematica.tgz
- Enter in the program directory:
$ cd AbdnAttribution_Mathematica
- The package contains the following:
- AbdnAttribution.m : Mathematica script to run the Aberdeen attribution method. It does not
need to be edited. Make sure that AbdnAttribution.m is executable. This can be achieved by executing
the following command once in the unix shell:
$ chmod a+x AbdnAttribution.m
- Input_AbdnAttribution.ini : Editable file containing the setting to run source attribution.
- Directory MLST with the data used for Listeria source attribution based on MLST information.
This will be used as a benchmark to illustrate the functioning of the program below.
*** Initialisation:
Open Input_AbdnAttribution.ini. This can be done with any text editor.
The format of the file is:
1 "**** Parameters for source attribution with AbdnAttribution.m ****"

2 "****"
3 "**** Please note: only lines 6, 8 and 10 should be edited! ****"
4 "****"
5 "---- Data directory/file ----"
6 MLST/MLST_AbdnAttribution
7 "---- Reservoir to be attributed ----"
8 Poultry
9 "---- Number of iterations for sample size correction ----"
10 10000
Line 6: indicates [data directory]/[data file], i.e. the directory where the input and output data is
stored (MLST in our example) and the name of the file containing the data (MLST_AbdnAttribution).
Line 8: Reservoir whose samples will be attributed. In the Listeria dataset, these can be: Bovine, Fish,
Human, Ovine, Poultry, Swine

Line 10: Number of iterations for sample size correction.
*** Input data format (see example in the directory MLST, file MLST_AbdnAttribution.csv)
The input data is stored in a comma-separated-values (csv) file.
Each line of the file contains:
[Reservoir name], loci 1, loci 2, ....., loci n
The isolates for each reservoir type should be consecutive in the file. However, the particular order of
reservoirs does not matter. For instance, in MLST_AbdnAttribution.csv they were grouped in the
following order: Bovine, Fish, Human, Ovine, Poultry, Swine
*** Running the code:
$ ./AbdnAttribution.m
*** Output results:
Results are output to an *.xml file which contains the name of the attributed reservoir (can be directly
open with Excel or Libreoffice calc).
The output file is stored in [data directory]/[data file]. In the MLST example, it is
MLST_AbdnAttribution_Poultry.xls


EFSA Report 2017

Uploaded by

Copyright:

Available Formats

EFSA Report 2017

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

EFSA Report 2017

Uploaded by

Copyright:

Available Formats

EXTERNAL SCIENTIFIC REPORT

APPROVED: 13 December 2016

Closing gaps for performing a risk assessment on Listeria

Question number: EFSA-Q-2014-00026

www.efsa.europa.eu/publications EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 2 EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 3 EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 4 EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 5 EFSA Supporting publication 2017:EN-1151

5.2.5. Outbreak 5 CC415..........................................................................................................47

www.efsa.europa.eu/publications 6 EFSA Supporting publication 2017:EN-1151

11. Additional supporting information..................................................................................... 115

www.efsa.europa.eu/publications 7 EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 8 EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 9 EFSA Supporting publication 2017:EN-1151

1.1. Background and Terms of Reference as provided by the requestor

www.efsa.europa.eu/publications 10 EFSA Supporting publication 2017:EN-1151

Specific objective 1: to carry out the molecular characterisation of a selection of L. monocytogenes

www.efsa.europa.eu/publications 11 EFSA Supporting publication 2017:EN-1151

This contract was awarded by EFSA to:

www.efsa.europa.eu/publications 12 EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 13 EFSA Supporting publication 2017:EN-1151

2.2. Strains from the baseline survey (BLS)

2.2.1. Selection criteria BLS 1. Availability of the strains

2.2.2.1. Selection criterion BLS 2a: Isolates from paired samples

www.efsa.europa.eu/publications 14 EFSA Supporting publication 2017:EN-1151

2.2.3. Selection criterion BLS 3. Multiple isolates from a sample (all

2.3. Strains from other foods (OF)

2.3.1. Selection criterion OF 1. Food origin

2.3.2. Selection criterion OF 2. Temporal criterion

2.3.3. Selection criterion OF 3. Further selection

2.3.4. Fruit and vegetables

www.efsa.europa.eu/publications 15 EFSA Supporting publication 2017:EN-1151

2.4. Strains from food chain production stages

2.5. Strains from sporadic human clinical cases

2.5.1. Selection criterion C1. Availability of the strains.

2.5.2. Selection criterion C2. Temporal and geographical criterion

2.5.3. Selection criterion C3. Incidence

www.efsa.europa.eu/publications 16 EFSA Supporting publication 2017:EN-1151

2.5.4. Selection criterion C4. Subtype information.

2.6. Strains from outbreaks

Country Human Food Vehicle of infection

2.7. Strain selection summary

www.efsa.europa.eu/publications 17 EFSA Supporting publication 2017:EN-1151

Country Baseline Other foods Food production Clinical, Outbreaks Total

Country Food matrix Source Number of Total per

www.efsa.europa.eu/publications 18 EFSA Supporting publication 2017:EN-1151

Country Food matrix Source Number of Total per

www.efsa.europa.eu/publications 19 EFSA Supporting publication 2017:EN-1151

Country Food matrix Source Number of Total per

www.efsa.europa.eu/publications 20 EFSA Supporting publication 2017:EN-1151

Country Food matrix Source Number of Total per

2.8.1. Database structure

The database has a structure similar to the EURL Lm DB.

www.efsa.europa.eu/publications 21 EFSA Supporting publication 2017:EN-1151

Figure 2.2.: Description of the information fields in the LISEQ-DB

2.8.2. Core information on strains

2.8.3. Data export

www.efsa.europa.eu/publications 22 EFSA Supporting publication 2017:EN-1151

www.efsa.europa.eu/publications 23 EFSA Supporting publication 2017:EN-1151