www.nature.

com/scientificreports

OPEN Gut microbiome alterations in

Alzheimer’s disease
Nicholas M. Vogt   1, Robert L. Kerby2, Kimberly A. Dill-McFarland   2, Sandra J. Harding1,
Andrew P. Merluzzi1, Sterling C. Johnson3,1,4, Cynthia M. Carlsson3,1,4, Sanjay Asthana3,1,
Received: 21 June 2017 Henrik Zetterberg5,6,7,8, Kaj Blennow5,6, Barbara B. Bendlin1,4 & Federico E. Rey2
Accepted: 27 September 2017 Alzheimer’s disease (AD) is the most common form of dementia. However, the etiopathogenesis of this
Published: xx xx xxxx devastating disease is not fully understood. Recent studies in rodents suggest that alterations in the
gut microbiome may contribute to amyloid deposition, yet the microbial communities associated with
AD have not been characterized in humans. Towards this end, we characterized the bacterial taxonomic
composition of fecal samples from participants with and without a diagnosis of dementia due to AD. Our
analyses revealed that the gut microbiome of AD participants has decreased microbial diversity and is
compositionally distinct from control age- and sex-matched individuals. We identified phylum- through
genus-wide differences in bacterial abundance including decreased Firmicutes, increased Bacteroidetes,
and decreased Bifidobacterium in the microbiome of AD participants. Furthermore, we observed
correlations between levels of differentially abundant genera and cerebrospinal fluid (CSF) biomarkers
of AD. These findings add AD to the growing list of diseases associated with gut microbial alterations, as
well as suggest that gut bacterial communities may be a target for therapeutic intervention.

Despite decades of research, the etiology underlying the development of dementia due to Alzheimer’s disease
(AD) remains unknown, and there are currently no preventative or disease-modifying treatments available. In
the brain, AD pathology is characterized by extracellular plaques composed of amyloid-β (Aβ) peptide and intra-
cellular neurofibrillary tangles composed of hyperphosphorylated tau protein1. However, what causes these hall-
mark features is unexplained. In recent years, researchers have proposed a potential role for pathogenic microbes,
including those derived from the gut, in the development or exacerbation of AD pathology2–5.
Humans harbor complex communities of microbes, with the vast majority of the microbial population resid-
ing in the distal gut. Gut microbes perform key functions for human health including energy extraction, biosyn-
thesis of vitamins, protection against pathogen overgrowth, and education of the immune system6. Microbial
colonization of the gut occurs during birth, is highly dynamic through infancy, and resembles adult structure
by about 3 years of age7. Thereafter, the composition of the microbiome within an individual remains generally
stable8, albeit with substantial interpersonal variation, particularly in elderly individuals9.
Alterations in the composition of this complex ecosystem have been associated with the development of a
variety of gastrointestinal and metabolic diseases including inflammatory bowel disease (IBD), obesity, diabetes,
and insulin resistance10. More recently, the influence of gut microbiota on central nervous system function –
often referred to as the gut-brain axis – has received significant attention, and alterations in the gut microbiome
have been associated with neurological conditions including autism spectrum disorder, multiple sclerosis, and
Parkinson’s disease11–13.
With respect to dementia, a recent study in cognitively impaired elderly participants investigated a limited
number of pro- and anti-inflammatory gut bacterial taxa and found altered abundance in individuals with

1
Wisconsin Alzheimer’s Disease Research Center, University of Wisconsin School of Medicine and Public Health, 600
Highland Avenue J5/1 Mezzanine, Madison, WI 53792, USA. 2Department of Bacteriology, University of Wisconsin-
Madison, 1550 Linden Drive, Madison, WI 53706, USA. 3Geriatric Research Education and Clinical Center, William S.
Middleton Memorial Veterans Hospital, 2500 Overlook Terrace, Madison, WI 53705, USA. 4Wisconsin Alzheimer’s
Institute, University of Wisconsin School of Medicine and Public Health, WARF Building, 610 Walnut Street, 9th Floor,
Suite 957, Madison, WI 53726, USA. 5Department of Psychiatry and Neurochemistry, Institute of Neuroscience and
Physiology, The Sahlgrenska Academy at the University of Gothenburg, Mölndal, Sweden. 6Clinical Neurochemistry
Laboratory, Sahlgrenska University Hospital, Mölndal, Sweden. 7Department of Molecular Neuroscience, University
College London Institute of Neurology, Queen Square, London, United Kingdom. 8UK Dementia Research Institute
at University College London, London, United Kingdom. Correspondence and requests for materials should be
addressed to B.B.B. (email: [email protected]) or F.E.R. (email: [email protected])

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 1

www.nature.com/scientificreports/

Control AD p value
n 25 25
Age (yrs, mean ± SD) 69.3 ± 7.5 71.3 ± 7.3 0.346
Sex (% Female) 72% (18/25) 68% (17/25) 0.785
Clinical dementia rating (CDR) score NA
0–normal 100% (25/25) 0
0.5–very mild dementia 40% (10/25)
1–mild dementia 36% (9/25)
2–moderate dementia 24% (6/25)
APOE ε4 genotype 20% (5/25) 72% (18/25) <0.001*
Ethnicity (% Caucasian) 96% (24/25) 92% (23/25) 0.552
BMI (kg/m2, median [IQR]) 26.1 [24.3–33.2] 26.0 [22.9–29.1] 0.467
Diabetes diagnosis 2/25 2/25 1.000
MIND Diet total score (mean ± SD) 7.6 ± 2.2 6.6 ± 2.7 0.160
Bristol stool scale score (mean ± SD) 3.5 ± 1.7 3.8 ± 1.2 0.561

Table 1.  Participant characteristics.

positive amyloid positron emission tomography (PET) imaging14. In addition, recent studies in transgenic mouse
models of AD have demonstrated that manipulating gut microbiota can influence cerebral amyloid deposi-
tion15,16. However, to date there have been no comprehensive surveys of whole gut microbiota in humans with
AD. In this study, we performed bacterial 16S ribosomal RNA (rRNA) gene sequencing of DNA isolated from
fecal samples in order to characterize the gut microbial communities in individuals with and without a clinical
diagnosis of dementia due to AD. In addition, we examined the relationship between gut microbiota and AD
pathology as measured by cerebrospinal fluid (CSF) biomarkers of AD.

Results
Study Design and Participant Characteristics.  Participants were recruited from the Wisconsin
Alzheimer’s Disease Research Center (ADRC) and the Wisconsin Registry for Alzheimer’s Prevention (WRAP)
study (see Methods). Gut microbiome compositional analysis was performed on fecal samples collected from
home-dwelling participants with dementia due to AD (n = 25), and age- and sex-matched Control participants
(n = 25). Table 1 reports participant characteristics. AD and Control groups did not differ with respect to age,
sex, ethnicity, BMI, or diabetes status. There was no difference between groups in total score on a 15-item food
questionnaire based on the Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND) Diet17,
which provided a semi-quantitative measure of dietary intake and allowed us to asses dietary differences. As
expected, the APOE ε4 genotype was more prevalent in the AD group. The majority of AD participants had very
mild or mild dementia, with clinical dementia rating (CDR) scores ranging from 0.5–2. Medication information
is reported in Supplementary Table S1. The AD group had a greater number of participants taking selective ser-
otonin reuptake inhibitors (SSRIs), and all but one AD participant was taking an acetylcholinesterase inhibitor
(donepezil or rivastigmine) and/or memantine.

Composition of the Gut Microbiome of Control and AD Groups.  Sequencing of the V4 region of
the 16S rRNA gene generated a total of 4.8 million sequence reads (mean ± SD: ~96,000 ± 32,000 reads/par-
ticipant), which were clustered into operational taxonomic units (OTUs) at 97% similarity and assigned taxo-
nomic classifications down to the lowest phylogenetic level possible (see Methods). The final OTU dataset for
AD and Control groups consisted of 972 OTUs classified to 95 genera, 46 families, 24 orders, 17 classes, and 9
phyla. Between groups, there were no differences in percentages of sequences classified to the phylum (Control:
99.6 ± 0.5%, AD: 99.6 ± 0.9%) or genus level (Control: 79.6 ± 7.4%, AD: 83.1 ± 8.1%). Across all 50 participants,
the dominant phyla were Firmicutes and Bacteroidetes, which respectively made up 78% (78.1 ± 8.7%) and 15%
(14.9 ± 8.4%) of total abundance, with lower contributions from Actinobacteria (2.6%), Verrucomicrobia (2.6%),
and Proteobacteria (1.1%) (Supplementary Fig. S1). The predominant bacterial families for all participants were
Lachnospiraceae (39.1%), Ruminococcaceae (29.6%), and Bacteroidaceae (9.8%), followed by Verrucomicrobiaceae
(2.6%), Clostridiales (1.9%), and Bifidobacteriaceae (1.5%) (Supplementary Fig. S1).

AD is Associated with Changes in the Gut Microbiome.  The composition of the gut microbiome was
characterized using traditional ecological measures including richness (the number of unique OTUs present in
a participant), alpha diversity (the richness and abundance of OTUs within each participant), and beta diversity
(the similarity or difference in composition between participants). For microbiome richness estimates, we used
the Abundance-based coverage estimator (ACE) and Chao1; these metrics use non-parametric modeling to cal-
culate a conservative estimate of total OTU richness for each participant. The microbiome of AD participants had
reduced richness, with both ACE and Chao1 significantly decreased in the AD group compared to the Control
group (t-test; ACE: DF = 48, t = 3.05, p = 0.004; Chao1: DF = 48, t = 2.98, p = 0.004) (Supplementary Fig. S2). For
alpha diversity metrics, we used the Inverse Simpson and Shannon Indexes, and Faith’s Phylogenetic Diversity
(PD), an alpha diversity metric that also incorporates phylogenetic relationships. While there was a trend towards
a decrease in the Inverse Simpson Index between Control and AD groups (Mann-Whitney; U = 227.0, p = 0.097),

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 2

www.nature.com/scientificreports/

A B Weighted UniFrac C Dialister

Phylum
Turicibacter
0.2
Diagnosis Firmicutes

Phylogenetic Diversity
Control SMB53
* Bacteroidetes
AD c115
10 Actinobacteria

(mean ± SEM)
0.1 Firmicutes (p)

NMDS2
Verrucomicrobia
Haemophilus
Proteobacteria
0.0 Christensenellaceae (f)
Bacteria (k)

Taxonomic classification
5 Lactococcus
-0.1 Oscillospira
Lachnospira
[Mogibacteriaceae] (f)
0 -0.1 0.0 0.1 0.2
Control AD Clostridiales (o)
NMDS1 Ruminococcus
D Bacteria (k)
Adlercreutzia
Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

40 Ruminococcaceae (f)
90
Streptococcus
10
30 [Ruminococcus]
80
Clostridium
20 Lachnospiraceae (f)
70 5 Coprococcus
Blautia
10
60 Bacteroides
Akkermansia
0 0
Dorea
Control AD Control AD Control AD
-8 -4 0 4
Firmicutes Bacteroidetes Actinobacteria
Log2 Fold Change

Figure 1.  Alzheimer’s disease is associated with alterations in gut microbiome composition. (A) Faith’s
Phylogenetic Diversity is decreased in the microbiome of AD participants. *p < 0.05. (B) Non-metric
multidimensional scaling (NMDS) plot of weighted UniFrac analysis of relative sample OTU composition.
NMDS analysis was limited to two dimensions, with a stress measurement of 0.17. Each dot represents a
scaled measure of the composition of a given participant, color- and shape-coded by cohort. (C) Differential
abundance analysis identified 14 OTUs that were increased and 68 OTUs that were decreased in AD relative
to Control participants (p < 0.05, FDR-corrected). Each point represents an OTU. Data plotted as log2 fold
change; OTUs to the right of the zero line are more abundant and OTUs to the left of the zero line are less
abundant in AD compared to Control groups. OTUs are organized on the y-axis according to the lowest
taxonomic classification possible. (D) OTUs grouped at the phylum level and analyzed using Metastats show
that AD participants have decreased abundance of Firmicutes and Actinobacteria, and increased abundance of
Bacteroidetes compared to Control participants (p < 0.05, FDR-corrected). Tukey plots show median, IQR, and
participant data points for phylum relative abundance.

both the Shannon Index and Faith’s PD were significantly decreased in AD participants compared to Control par-
ticipants (t-test; Shannon: DF = 48, t = 2.44, p = 0.019, PD: DF = 48, t = 2.59, p = 0.013) (Fig. 1A, Supplementary
Fig. S2). With respect to beta diversity, Bray-Curtis dissimilarity and UniFrac analysis (weighted and unweighted)
demonstrated compositional differences in the microbiome between AD and Control groups (PERMANOVA,
Bray-Curtis: F = 2.87, p < 0.001, weighted UniFrac: F = 3.84, p < 0.001; unweighted UniFrac: F = 2.60, p < 0.005)
(Fig. 1B, Supplementary Fig. S3).
Differential abundance analysis of taxa at the OTU level revealed that the microbiome of AD participants
showed significantly altered abundance of 82 OTUs relative to the Control group, with 14 OTUs more abundant
and 68 OTUs less abundant in AD (Fig. 1C, Supplementary Table S2). Plotting these 82 OTUs as a phylogenetic
tree-ordered and diagnosis-grouped heat map shows clear differences in OTU abundance distributions between
Control and AD groups (Supplementary Fig. S4).
OTUs were taxonomically grouped and differential abundance was analyzed at the phylum, family, and genus
levels using Metastats (Supplementary Table S3). At the phylum level, AD participants had decreased abun-
dance of Firmicutes and Actinobacteria, and increased abundance of Bacteroidetes compared to Control par-
ticipants (Fig. 1D). Within Firmicutes, the families Ruminococcaceae, Turicibacteraceae, Peptostreptococcaceae,
Clostridiaceae, and Mogibacteriaceae, and the genera SMB53 (family Clostridiaceae), Dialister, Clostridium,
Turicibacter, and cc115 (family Erysipelotrichaceae) were all less abundant in AD participants, while the family
Gemellaceae and the genera Blautia, Phascolarctobacterium, and Gemella were more abundant in AD participants
(Fig. 2). Within Bacteroidetes, Bacteroidaceae and Rikenellaceae at the family level, and Bacteroides and Alistipes
at the genus level were more abundant in AD participants. The decrease in Actinobacteria was reflected by
decreased Bifidobacteriaceae at the family level and by decreased Bifidobacterium and Adlercreutzia at the genus
level. Additionally, the genus Bilophila in the phylum Proteobacteria was more abundant in AD participants.
Predictive metagenomics analysis (PICRUSt18) identified potential functional changes in the gut microbiome
of AD participants. These include increases in predicted gene content in KEGG pathways related to metabolism
and biosynthesis, including oxidative phosphorylation, carbohydrate metabolism, and amino acid metabolism
(Supplementary Fig. S5), and decreases in predicted gene content in KEGG pathways related to signal transduc-
tion and cell motility, including bacterial chemotaxis pathways, secretion systems, bacterial motility proteins, and
two-component signal transduction systems.

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 3

www.nature.com/scientificreports/

Families less abundant in AD

Ruminococcaceae Bifidobacteriaceae Clostridiaceae Mogibacteriaceae Turicibacteraceae Peptostreptococcaceae
12.5 2.5
50

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)
2.0
10.0 7.5 2.0 0.9
40
1.5
7.5 1.5
30 5.0 0.6
5.0 1.0
1.0
20
2.5 0.3
2.5 0.5 0.5
10
0.0 0.0 0.0 0.0 0.0
Control AD Control AD Control AD Control AD Control AD Control AD

Families more abundant in AD

Bacteroidaceae Rikenellaceae Gemellaceae

0.03
Relative Abundance (%)
Relative Abundance (%)

Relative Abundance (%)

4
20

3
15 0.02

2
10
0.01

5 1

0.00
0 0
Control AD Control AD Control AD

Genera less abundant in AD

Bifidobacterium SMB53 Dialister Clostridium Turicibacter Adlercreutzia cc115
12.5 6 3

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

Relative Abundance (%)

4
10.0 0.20
0.9 0.3
4 2 3
7.5 0.15
0.6 0.2
2 0.10
5.0
2 1
0.3 0.1
2.5 1 0.05

0.0 0 0 0 0.0 0.00 0.0

Control AD Control AD Control AD Control AD Control AD Control AD Control AD

Genera more abundant in AD

Blautia Bacteroides Alistipes Phascolarctobacterium Bilophila Gemella

30 3

Relative Abundance (%)

Relative Abundance (%)

4 0.03
Relative Abundance (%)
Relative Abundance (%)

Relative Abundance (%)

0.8
Relative Abundance (%)

20

3
Phylum
2 0.6
20 15 0.02
Firmicutes
2 0.4 Bacteroidetes
10
1 0.01 Actinobacteria
10
1 0.2
5 Proteobacteria

0 0.0 0.00
0 0
Control AD Control AD Control AD Control AD Control AD Control AD

Figure 2.  Bacterial families and genera differentially represented in feces from AD participants compared to
Control participants (p < 0.05, FDR-corrected). Tukey plots are colored by phylum and show median, IQR, and
participant data points for genus or family relative abundance.

Differentially Abundant Genera Are Correlated with CSF Biomarkers of AD Pathology.  Our
primary microbiome compositional analysis identified 13 genera as differentially abundant between AD and
Control groups. We next examined the relationship between the relative abundance of these 13 taxa and levels
of CSF biomarkers in a subset of participants who had also undergone lumbar puncture. This analysis included
microbiome and CSF data from 9 AD group participants, and 31 non-demented (ND) participants (10 from the
Control group, and an additional 21 largely younger participants not selected in the original random matching
with AD participants). Correlations were calculated using data from the total group of 40 participants, as well
as separately for ND participants and AD participants. CSF biomarkers included Aβ42/Aβ40, phosphorylated tau
(p-tau), the ratio of p-tau/Aβ42, and chitinase-3-like protein 1 (YKL-40). CSF Aβ42/Aβ40 is an indicator of amy-
loid burden, with lower levels in the CSF reflecting greater amyloid deposition in the brain; p-tau is a marker of
neurofibrillary tangles, with higher levels reflecting greater tangle pathology in the brain; the ratio of p-tau/Aβ42
incorporates both aspects of pathology, with higher values implying greater AD pathology19. YKL-40 is a marker
of astroglial and/or microglial activation, and has been shown to be elevated in CSF of individuals with dementia
due to AD20–22. Supplementary Table S4 reports participant characteristics, CSF biomarker levels, and relative
abundances of genera.
Across all 40 participants included in this analysis, we observed generally consistent trends between bacterial
relative abundance and CSF biomarkers of AD pathology (Fig. 3). The direction of these trends was largely similar

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 4

www.nature.com/scientificreports/

Aβ 42/Aβ 40 p-tau p-tau/Aβ 42

All ND AD All ND AD All ND AD
Increased in AD
Blautia -0.55 -0.30 -0.62 0.35 0.09 0.41 0.46 0.18 0.57

Bacteroides -0.34 -0.28 0.18 0.36 0.29 -0.01 0.33 0.20 0.23

Phascolarctobacterium -0.21 0.06 0.33 0.31 0.28 -0.43 0.45 0.32 -0.10

Alistipes -0.01 0.15 0.33 0.27 0.34 -0.19 0.27 0.25 -0.15

Bilophila -0.12 -0.07 0.05 -0.10 -0.18 -0.09 0.05 -0.05 0.03

Gemella -0.22 -0.14 0.15 -0.09 -0.19 -0.37 0.04 -0.09 -0.62

Decreased in AD
Clostridium 0.16 0.04 -0.42 -0.13 -0.09 0.38 -0.24 -0.15 0.25

Bifidobacterium 0.10 -0.02 0.68 -0.09 0.03 -0.60 -0.17 -0.09 -0.53

Dialister 0.34 0.18 0.12 -0.35 -0.29 -0.10 -0.51 -0.43 -0.42

Turicibacter 0.30 0.07 0.37 -0.28 -0.03 -0.50 -0.38 -0.18 -0.64

SMB53 0.36 0.17 0.37 -0.45 -0.26 -0.50 -0.48 -0.36 -0.64

Adlercrutzia 0.13 -0.06 -0.01 -0.06 0.04 0.05 -0.17 -0.01 -0.16

cc115 0.62 0.49 NA -0.41 -0.20 NA -0.44 -0.20 NA

-1 -0.75 -0.5 0 0.5 0.75 1 p <0.05

Spearman’s rho

Figure 3.  Bacterial taxa correlate with CSF biomarkers of AD pathology. 13 genera identified as differentially
abundant in AD were correlated with CSF biomarkers of AD including the Aβ42/Aβ40 ratio (with lower CSF
levels reflecting greater amyloid deposition in the brain), phosphorylated tau (p-tau), and the p-tau/Aβ42 ratio
(which incorporates both facets of AD pathology). Correlations were calculated separately for all 40 participants
(All), 31 non-demented participants (ND), and 9 AD participants (AD). In general, genera identified as more
abundant in AD were associated with greater AD pathology, while genera identified as less abundant in AD
were associated with less AD pathology. Genera are ordered from most to least abundant. Shape and color of
ellipses represent strength of Spearman’s rank correlation coefficients (rho). Bolded ellipse borders represent
significant correlations (two-sided, p < 0.05 uncorrected).

for both AD participants and healthy non-demented participants. For genera that are more abundant in AD, we
observed a relationship between increased bacterial abundance and greater AD pathology, which was indicated
by predominantly negative correlations between bacterial abundance and CSF Aβ42/Aβ40 (suggesting greater
abundance is associated with greater amyloid burden in the brain), and predominantly positive correlations
between bacterial abundance and CSF p-tau and p-tau/Aβ42. These relationships were especially strong in those
genera with overall greater relative abundance, particularly in Bacteroides and Blautia. Similarly, for those genera
that are less abundant in AD, we observed a relationship between decreased bacterial abundance and greater
AD pathology, with levels of SMB53 and Dialister showing the strongest correlations with CSF AD biomarkers.
Notably, in both more and less abundant genera, the strongest correlations between bacterial abundance and CSF
AD biomarkers were in the same direction whether including only AD participants, only healthy non-demented
participants, or all participants. Additionally, in AD participants, we observed a relationship between increased
abundance of Bacteroides and increased CSF YKL-40 levels, and a relationship between decreased abundance of
both Turicibacter and SMB53 and increased CSF YKL-40 levels (Supplementary Fig. S6).

Discussion
Despite the proposed role of gut microbiota in the development or progression of AD4,5,23, there have been no
comprehensive surveys of the gut microbiome in individuals with AD. In this study, we performed bacterial
16S rRNA gene sequencing on DNA isolated from fecal samples in order to compare the composition of the gut
microbiome in participants with and without a diagnosis of dementia due to AD. We discovered that the gut
microbiome of AD participants has decreased microbial richness and diversity and a distinct composition com-
pared to asymptomatic age- and sex-matched Control participants. We also identified several broad taxonomic
differences between AD and Control groups, and determined that levels of differentially abundant genera corre-
late with CSF biomarkers of AD pathology.
The decreased richness and diversity in our study broadly parallels results observed in other conditions linked
to gut microbiome alterations, including obesity, diabetes, IBD, and Parkinson’s disease13,24–26. Furthermore,
PICRUst analysis revealed broad functional changes in predicted metabolism, bacterial cell motility, and signal
transduction pathways in the gut microbiome of AD participants. While the specific bacteria responsible for

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 5

www.nature.com/scientificreports/

these compositional and functional alterations may differ between conditions, it has been proposed that these
broad-scale changes in gut microbiota (often referred to as “dysbiosis”) may play important roles in disease
progression and maintenance, potentially through immune activation and systemic inflammation27. While it is
unclear how the gut influences the development of neuropathology, substantial evidence supports the existence
of a gut-brain axis that allows bi-directional communication between the gut and brain through several pathways
including neural, endocrine, and immune mechanisms11,28. Within this framework, alterations in gut micro-
bial communities in patients with AD may result in pathophysiological changes in the brain. In support of this
hypothesis, a recent study showed that transgenic AD mice raised under germ-free conditions have less cerebral
amyloid deposition than conventionally-raised AD mice, indicating that gut microbiota influence the develop-
ment of amyloid pathology16.
In our study, the phylum Firmicutes as a whole, as well as several families, genera, and 61 OTUs classified
within Firmicutes were decreased in the AD group. A reduction in Firmicutes has been reported in the microbi-
ome of individuals with type 2 diabetes25 as well as obesity29 (although others have reported increased Firmicutes
in obesity24,30). Notably, diabetes and insulin resistance increase the risk of developing AD31–33. We have recently
reported that insulin resistance is associated with decreased cerebral glucose metabolism and increased amyloid
deposition in asymptomatic middle-aged adults enriched for risk of AD34,35. Thus, a potential mechanism by
which microbial alterations in the gut may influence AD pathology is through promoting the development of
insulin resistance and diabetes. While AD and Control groups did not differ with respect to diabetes prevalence
(Table 1), sub-clinical differences in insulin or glucose metabolism cannot be ruled out. Further investigation will
be needed to explore the relationship between microbiota and insulin resistance in AD.
In participants with AD, we observed an increase in the phylum Bacteroidetes, which was reflected by
increased Bacteroidaceae at the family level, and increased Bacteroides at the OTU and genus level. The phylum
Bacteroidetes encompasses a diverse and abundant group of gram-negative commensal bacteria in the gut36,
including the genus Bacteroides, which has been detected at higher levels in the gut of individuals with type 2
diabetes25 and in patients with Parkinson’s disease13, a neurodegenerative disorder. The major outer membrane
component of gram-negative bacteria is lipopolysaccharide (LPS), which is capable of triggering systemic inflam-
mation and the release of pro-inflammatory cytokines after translocation from the gut to systemic circulation37.
Additionally, in vitro and in vivo studies have demonstrated an association between bacterial endotoxins (e.g.
LPS) and AD pathology. Co-incubation of Aβ peptide with LPS potentiates amyloid fibrillogenesis38, and sys-
temic injection of LPS in wild-type and transgenic AD mice results in greater amyloid deposition and tau pathol-
ogy39–42. In humans, intestinal permeability increases with age43, and elderly individuals show an association
between increased LPS-binding protein (a marker of microbial translocation) and inflammation44. Moreover, a
recent study involving postmortem brain tissue from patients with AD showed that LPS and gram-negative E.
coli fragments co-localize with amyloid plaques45. Thus, increased abundance of gram-negative intestinal bac-
teria such as Bacteroides in participants with AD may result in increased translocation of LPS from the gut to
systemic circulation, which in turn may contribute to or exacerbate AD pathology through inflammation or other
mechanisms.
Additionally, compared to control participants, AD participants in our study exhibited decreased
Actinobacteria. These differences were mostly driven by changes in Bifidobacterium. Actinobacteria, particularly
members of the Bifidobacterium genus, are an important bacterial inhabitant of the human gut across the lifespan,
and their beneficial health effects have been well-documented46,47. In particular, certain species of Bifidobacterium
are associated with anti-inflammatory properties and decreased intestinal permeability48. Additionally, supple-
mentation with Bifidobacterium has been shown to decrease LPS levels in the intestine and improve gut mucosal
barrier properties in mice49,50. Interestingly, in germ-free mice colonized with human gut microbiota, increased
levels of Bifidobacterium are associated with decreased bacterial translocation to systemic circulation, while
increased levels of Bacteroides have been shown to increase bacterial translocation51. Considering our present
findings, increased Bacteroides and decreased Bifidobacterium in AD participants may represent a gut microbial
phenotype with particular propensity for translocation of pro-inflammatory bacterial components. Furthermore,
several Bifidobacterium species are widely used as probiotics. A small study of probiotics that included
Bifidobacterium demonstrated a change in Mini-Mental State Examination scores after a 12-week intervention
among participants with severe dementia52. Taken together with the decreased abundance of Bifidobacterium in
AD participants observed in our study, larger trials may be warranted, particularly in earlier disease stages.
Finally, we observed correlations between levels of differentially abundant gut microbiota and CSF biomarkers
of AD pathology in a subset of participants that had also undergone lumbar puncture. In general, genera identi-
fied as more abundant in AD were associated with greater AD pathology while genera identified as less abundant
in AD were associated with less AD pathology. These effects were most prominent when examining CSF p-tau/
Aβ42, a composite measure of AD pathology. Interestingly, even among non-demented participants who had
undergone lumbar puncture, we found a relationship between genera that were either more or less abundant
in AD and markers of amyloid and tau protein, even in the absence of dementia. In particular, Dialister and
SMB53 showed the strongest correlations in non-demented participants, with greater abundance of these bacte-
ria associated with less AD pathology, suggesting these bacterial taxa may be protective against development or
progression of AD pathology. We also observed significant associations in AD participants between CSF YKL-40
and abundance of Bacteroides, Turicibacter, and SMB53 (family Clostridiaceae). While these findings support a
link between altered gut bacterial abundance and glial activation in AD, this relationship is less clear in healthy
non-demented individuals and requires further investigation.
A limited number of studies have attempted to address the role of gut microbiota in AD. A recent inves-
tigation in cognitively-impaired older adults (without an AD diagnosis) reported increased abundance of the
pro-inflammatory bacteria Escherichia/Shigella and decreased abundance of the anti-inflammatory bacteria
Eubacterium rectale in individuals with evidence of amyloid deposition on PET imaging compared to individuals

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 6

www.nature.com/scientificreports/

who were amyloid negative14. While those results support a link between gut microbiota and brain amyloidosis,
the study only investigated the abundance of six pre-selected bacterial taxa using quantitative PCR rather than
the broader approach used here. Additionally, in a recent AD mouse microbiome study using 16S rRNA sequenc-
ing, APP/PS1 transgenic mice showed increased Helicobacteraceae and Desulfovibrionaceae at the family level,
increased Odoribacter and Helicobacter, and decreased Prevotella compared to wild-type mice53. However, while
the anatomy and physiology of the gastrointestinal tract of humans and mice share many characteristics54, there
are also substantial differences with respect to resident bacterial communities30, which makes comparing taxa and
changes in abundance between these studies difficult.
While AD participants were well-matched to our Control participants (suggesting that the gut microbiome
differences we observed were not likely the result of age, sex, BMI, or dietary differences between groups), they
did differ with respect to the use of selective serotonin reuptake inhibitors (SSRIs) and AD medications. We did
not find differences in microbial richness, diversity, or relative abundance of the 13 genera identified as altered
in AD between AD participants taking SSRIs and AD participants not taking SSRIs (Supplementary Table S5),
suggesting that these medications are not influencing our results. Nearly all AD participants in our study were
taking the AD medications donepezil or rivastigmine (acetylcholinesterase inhibitors), and/or memantine (an
NMDA receptor antagonist). It is unknown how these medications affect the gut microbiome. The most common
side effects reported for acetylcholine esterase inhibitors are gastrointestinal upset55, both nausea and diarrhea,
which could influence microbiota composition. It is worth noting that our participants did not report chronic
constipation or diarrhea, and there was no difference between groups on the Bristol stool scale (Table 1), which
can be used as a surrogate for stool transit time. Still, we recognize that we cannot completely rule out the effect
of AD medication use on our results. Further work, including animal experiments and longitudinal human stud-
ies, will be needed to determine the cause-effect relationship between gut microbiota and pathogenesis of AD.
Determining the role of gut bacteria in the progression or maintenance of AD may lead to novel interventional
approaches that alter or restore healthy gut bacterial composition, or identification of microbial metabolites that
are protective against AD.

Methods
Participants.  Participants with dementia due to AD (n = 25) were recruited from the Wisconsin Alzheimer’s
Disease Research Center (ADRC). Non-demented participants (n = 94) were recruited from both the ADRC and
the Wisconsin Registry for Alzheimer’s Prevention (WRAP) study56. The University of Wisconsin Health Science
Institutional Review Board approved all study procedures, and all experiments were performed in accordance
with relevant guidelines and regulations. All participants provided written informed consent to be involved in
this study.
Exclusion criteria for this study included any significant neurologic disease, history of alcohol/substance
dependence, major psychiatric disorders (including major depression), or any other significant medical illness.
Microbiome-specific exclusion criteria included: the use of systemic antibiotics in the previous 6 months prior
to providing the fecal sample; corticosteroid use (oral, IV, nasal, or inhaled); immune stimulating medications;
immunosuppressive agents; large doses of commercial probiotics consumed (greater than or equal to 108 cfu
or organisms per day); major dietary change during previous month (defined as eliminating or significantly
increasing a major food group); major GI tract surgery in past 5 years (with the exception of cholecystectomy
and appendectomy); major bowel resection at any time; active uncontrolled GI disorders or diseases including
inflammatory bowel disease (IBD), indeterminate colitis, irritable bowel syndrome (IBS), persistent, infectious
gastroenteritis, colitis or gastritis, persistent or chronic diarrhea of unknown etiology, Clostridium difficile infec-
tion (recurrent) or Helicobacter pylori infection (untreated), or chronic constipation.
At the time of fecal sample collection, participants completed a short questionnaire regarding recent antibi-
otic or probiotic use, as well as current and past gastrointestinal/metabolic conditions. Additionally, participants
also completed a 15-item self-report diet questionnaire developed by Martha Clare Morris (Rush University)
and based on the MIND Diet17, which allowed us to assess dietary differences between groups. We also used
information/data collected from participants at annual and bi-annual ADRC and WRAP study visits including
medication use, medical conditions/diagnoses, clinical dementia rating (CDR) scores, and CSF biomarker data
(see below). Body mass index (BMI) was calculated using the height and weight of each participant at their most
recent ADRC or WRAP study visit. APOE ε4 genotyping procedures have been described previously57, and par-
ticipants were categorized as non-carriers (zero ε4 alleles) or APOE ε4 carriers (one or two ε4 alleles). Participants
with AD were diagnosed using the NINDS/ADRDA criteria58, and confirmed by a multidisciplinary consensus
diagnostic panel.
Of the 119 participants recruited, six non-demented participants were excluded from analyses due to antibiotic
or probiotic use at the time of fecal sample collection that was not reported during initial screening. Participants
recruited from the WRAP study were largely younger than those recruited from the ADRC, thus for our primary
compositional analysis we age- and sex-matched the 25 AD participants 1-to-1 from the remaining 88 asymp-
tomatic control participants using case-control matching in SPSS with an age tolerance of 4.5 years to create an
equal-sized Control group. For secondary CSF correlational analysis, we used microbiome and CSF data from 9
AD cohort participants, and 31 non-demented (ND) participants, including 10 from the age- and sex-matched
Control cohort, and an additional 21 largely younger participants not included in the primary analysis.

Fecal sample collection and bacterial 16S rRNA sequencing and processing.  All participants
involved in the study resided at home, where fecal sample collection occurred. Participants returned by overnight
delivery sample collection kits, packaged within insulated containers and chilled with frozen gel packs; all sam-
ples included in this study arrived chilled and 92% were processed and frozen (see below) the day following home

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 7

www.nature.com/scientificreports/

collection by an individual who was blind to the participant’s cohort/diagnosis. Upon receipt, chilled samples
were weighed, scored on the Bristol stool scale59, subsampled (~100 mg) into prepared sterile bead beating tubes,
and stored at −80 °C until processing.
Fecal samples, suspended in lysis buffer/phenol:chloroform, were processed by bead beating24 and the
genomic DNA in the recovered aqueous phase then precipitated with the addition of 0.1-volume 3 M sodium ace-
tate and 1-volume isopropanol, incubated on ice, and centrifuged (4 °C, 20 min at 18,000 x g). After rinsing with
100% ethanol and drying, the DNA pellet was dissolved in TE buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA) then
column-purified using the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel Inc., Bethlehem, PA). DNA
concentration was measured using the Qubit BR dsDNA assay (Invitrogen, Eugene, OR). The variable region V4
amplicon of the bacterial 16S rRNA gene was amplified in duplicate reactions/sample (plus a no-template control
for each primer set) using 8-bp barcoded forward and reverse primers (0.4 μM)60, 12.5 ng template, and KAPA
HiFi HotStart DNA polymerase (KAPA Biosystems, Wilmington, MA). Reactions were agarose gel-checked and
duplicates combined, purified (NucleoSpin columns), and the DNA quantified (Qubit). The final equimolar pool
was sequenced on the Illumina MiSeq platform (paired end, 2 × 250-bp).
Sequence processing and cleanup was performed using mothur v1.39.161 and a previously described protot-
col60. Briefly, quality-filtered, paired-end duplex sequence reads were combined into contigs, and sequences with
ambiguous base pairings, sequences longer that 275-bp, and homopolymers greater than 8-bp were removed.
Sequences were then aligned to the SILVA 16S rRNA gene reference alignment database, and chimeric sequences
were identified and removed. Finally, remaining sequences with 97% similarity were clustered into operational
taxonomic units (OTUs) using the OptiClust algorithm62 and assigned the lowest possible taxonomic classifica-
tions from the GreenGenes reference database (v13.8) using a naive Bayesian classifier requiring an 80% confi-
dence score. As it has been demonstrated that quality-filtering 16S amplicon sequence reads can greatly improve
accuracy of microbial community analysis63, OTUs with <0.001% of total sequence reads were filtered out from
the dataset to account for sequencing errors.

Microbial community composition and differential abundance statistical analysis.  Richness

(ACE, Chao1) and alpha diversity (Inverse Simpson, Shannon Index) metrics were calculated at the OTU-level
in mothur by performing 1000 iterations of random subsampling to 31,396 reads (the lowest single partici-
pant number of sequences) from each participant. Faith’s Phylogenetic Diversity (PD) was calculated from a
neighbor-joined phylogenetic tree created in R v3.3.2 using normalized OTU-level data and the vegan, phyloseq,
and ape packages. Beta diversity metrics were computed using normalized OTU-level data in R and included
Bray-Curtis dissimilarity, and weighted and unweighted UniFrac. To detect differences in richness and alpha
diversity between groups, we used independent two-sample t-tests for normally distributed measures or
Mann-Whitney U tests for non-normally distributed measures in SPSS. To detect statistical differences in beta
diversity metrics between groups, we used permutational multivariate analysis of variance (PERMANOVA) in
the vegan package in R.
Differential abundance of taxa between AD and Control groups was determined at the OTU level using the
DESeq2 package in R. DESeq2 is a statistical method developed to detect differential expression in RNA-seq count
data while accounting for library size differences and biological variability64. It has recently been demonstrated
that applying these methods to microbiome OTU count data leads to improvements in detecting differential
abundance compared to simple proportions or rarefying65. Results were expressed as log2 fold change in AD par-
ticipants relative to Control participants. Relative abundance comparisons at the genus, family, and phylum levels
were performed on normalized data in mothur using 10,000 iterations of Metastats, a statistical method employ-
ing non-parametric t-tests, Fisher’s exact tests, and false discovery rate (FDR) correction to detect differentially
abundant features66. OTUs present in less than 20 participants and those OTUs that could not be classified down
to the desired level were excluded from the Metastats analysis.

PICRUSt predictive functional metagenomics analysis.  We used PICRUSt18 to detect predicted

functional differences in microbial communities between AD and Control participants. Briefly, OTUs were
re-assigned PICRUSt-compatible taxonomic classifications using the GreenGenes version 13.5 reference data-
base, and then normalized by 16S rRNA copy number. The resulting normalized OTU table was then used for
prediction of KEGG orthologs (KOs) based on bacterial composition. The weighted nearest sequenced taxon
index (NSTI), a quality metric of the phylogenetic distance between the input OTUs of our samples and the
reference OTUs used for metagenomic prediction, was 0.07 ± 0.01 (mean ± SD) for our samples. KOs were col-
lapsed into hierarchical KEGG pathways using the categorize by function command in PICRUSt, and the linear
discriminate analysis (LDA) effect size (LEfSe)67 implementation in mothur was used to detect differences in level
2 and level 3 KEGG pathways. LDA scores (log 10) of significantly different pathways were plotted as bars using
ggplot2 in R.

CSF collection and microbiome correlation statistical analysis.  CSF was collected via lumbar punc-
ture in the morning after a 12hr fast with a Sprotte 25-or 24-gauge spinal needle at the L3/4 or L4/5 interspace
using gentle extraction into propylene syringes. CSF (~22 mL) was then combined, gently mixed and centrifuged
at 2,000 x g for 10 minutes. Supernatants were frozen in 0.5 mL aliquots in polypropylene tubes and stored at
−80 °C.
CSF measures included the Aβ42/Aβ40 ratio, phosphorylated tau (p-tau), and the p-tau/Aβ42 ratio. Using the
Aβ42/Aβ40 ratio normalizes CSF Aβ42 for the total amount of Aβ peptides that are present in CSF and shows better
correspondence with brain amyloid deposition as well as superior diagnostic performance than CSF Aβ42 alone68.
For the Aβ42/Aβ40 ratio, CSF Aβ42 and CSF Aβ40 were quantified separately by electrochemiluminescence (ECL)

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 8

www.nature.com/scientificreports/

using an Aβ triplex assay (MSD Human Aβ peptide Ultra-Sensitive Kit, Meso Scale Discovery, Gaithersburg,
MD). For p-tau and the p-tau/Aβ42 ratio, CSF P-tau and Aβ42 were quantified using commercially available sand-
wich ELISAs (INNOTEST β-amyloid1–42, and Phospho-Tau[181 P], respectively; Fujirebio Europe, Ghent,
Belgium). YKL-40 was quantified using sandwich ELISAs (R&D Systems, Minneapolis, Minn., USA). CSF assays
were performed in two batches and corrected for batch differences as previously described69,70. For correlational
statistical analysis, we used the “cor.test” function in R to calculate Spearman’s rank correlation coefficients
between CSF biomarker levels and normalized relative abundances for the 13 differentially abundant genera. The
correlation matrix was plotted as ellipses using the ellipse package in R.

Data Availability.  The datasets used and/or analyzed during the current study are available from the corre-
sponding author on reasonable request.

References
1. Selkoe, D. J. & Hardy, J. The amyloid hypothesis of Alzheimer’s disease at 25 years. EMBO Mol. Med. 8, 595–608 (2016).
2. Hill, J. M. et al. Pathogenic microbes, the microbiome, and Alzheimer’s disease (AD). Front. Aging. Neurosci. 6, 1–5 (2014).
3. Zhao, Y., Dua, P. & Lukiw, W. J. Microbial sources of amyloid and relevance to amyloidogenesis and Alzheimer’s disease (AD). J.
Alzheimers Dis. Parkinsonism 5, 177 (2015).
4. Bhattacharjee, S. & Lukiw, W. J. Alzheimer’s disease and the microbiome. Front. Cell. Neurosci. 7, 1–4 (2013).
5. Ghaisas, S., Maher, J. & Kanthasamy, A. Gut microbiome in health and disease: Linking the microbiome-gut-brain axis and
environmental factors in the pathogenesis of systemic and neurodegenerative diseases. Pharmacol. Ther. 158, 52–62 (2016).
6. Chow, J., Lee, S. M., Shen, Y., Khosravi, A. & Mazmanian, S. K. Host-bacterial symbiosis in health and disease. Adv. Immunol. 107,
243–274 (2010).
7. Koenig, J. E. et al. Succession of microbial consortia in the developing infant gut microbiome. Proc. Natl. Acad. Sci. USA 108(Suppl
1), 4578–4585 (2011).
8. Human Microbiome Project Consortium. Structure, function and diversity of the healthy human microbiome. Nature 486, 207–214
(2012).
9. Claesson, M. J. et al. Composition, variability, and temporal stability of the intestinal microbiota of the elderly. Proc. Natl. Acad. Sci.
USA 108(Suppl 1), 4586–4591 (2011).
10. Clemente, J. C., Ursell, L. K., Parfrey, L. W. & Knight, R. The impact of the gut microbiota on human health: an integrative view. Cell
148, 1258–1270 (2012).
11. Fung, T. C., Olson, C. A. & Hsiao, E. Y. Interactions between the microbiota, immune and nervous systems in health and disease.
Nat. Neurosci. 20, 145–155 (2017).
12. Scheperjans, F. et al. Gut microbiota are related to Parkinson’s disease and clinical phenotype. Mov. Disord. 30, 350–358 (2015).
13. Keshavarzian, A. et al. Colonic bacterial composition in Parkinson’s disease. Mov. Disord. 30, 1351–1360 (2015).
14. Cattaneo, A. et al. Association of brain amyloidosis with pro-inflammatory gut bacterial taxa and peripheral inflammation markers
in cognitively impaired elderly. Neurobiol. Aging 49, 60–68 (2017).
15. Minter, M. R. et al. Antibiotic-induced perturbations in gut microbial diversity influences neuro-inflammation and amyloidosis in
a murine model of Alzheimer’s disease. Sci. Rep. 6, 1–12 (2016).
16. Harach, T. et al. Reduction of Abeta amyloid pathology in APPPS1 transgenic mice in the absence of gut microbiota. Sci. Rep. 7,
41802 (2017).
17. Morris, M. C. et al. MIND diet associated with reduced incidence of Alzheimer’s disease. Alzheimers Dement. 11, 1007–1014 (2015).
18. Langille, M. G. I. et al. Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences. Nat.
Biotechnol. 31, 814–821 (2013).
19. Blennow, K. & Zetterberg, H. The past and the future of Alzheimer’s disease CSFbiomarkers - a journey toward validated biochemical
tests covering the whole spectrum of molecular events. Front. Neurosci. 9, 1–8 (2015).
20. Hellwig, K. et al. Neurogranin and YKL-40: independent markers of synaptic degeneration and neuroinflammation in Alzheimer’s
disease. Alz. Res. Therapy 7, 74 (2015).
21. Craig-Schapiro, R. et al. YKL-40: a novel prognostic fluid biomarker for preclinical Alzheimer’s disease. Biol. Psychiatry 68, 903–912
(2010).
22. Rosén, C. et al. Increased Levels of Chitotriosidase and YKL-40 in Cerebrospinal Fluid from Patients with Alzheimer’s disease.
Dement. Geriatr. Cogn. Dis. Extra 4, 297–304 (2014).
23. Pistollato, F. et al. Role of gut microbiota and nutrients in amyloid formation and pathogenesis of Alzheimer disease. Nutr. Rev. 74,
624–634 (2016).
24. Turnbaugh, P. J. et al. A core gut microbiome in obese and lean twins. Nature 457, 480–484 (2009).
25. Larsen, N. et al. Gut microbiota in human adults with type 2 diabetes differs from non-diabetic adults. PLoS ONE 5, 1–10 (2010).
26. Dicksved, J. et al. Molecular analysis of the gut microbiota of identical twins with Crohn’s disease. ISME J. 2, 716–727 (2008).
27. DuPont, A. W. & DuPont, H. L. The intestinal microbiota and chronic disorders of the gut. Nat. Rev. Gastroenterol. Hepatol. 8,
523–531 (2011).
28. Cryan, J. F. & Dinan, T. G. Mind-altering microorganisms: the impact of the gut microbiota on brain and behaviour. Nat. Rev.
Neurosci. 13, 701–712 (2012).
29. Schwiertz, A. et al. Microbiota and SCFA in lean and overweight healthy subjects. Obesity 18, 190–195 (2009).
30. Ley, R. E. et al. Obesity alters gut microbial ecology. Proc. Natl. Acad. Sci. USA 102, 11070–11075 (2005).
31. Rawlings, A. M. et al. Diabetes in midlife and cognitive change over 20 years: a cohort study. Ann. Intern. Med. 161, 785–793 (2014).
32. Ott, A. et al. Diabetes mellitus and the risk of dementia: The Rotterdam Study. Neurology 53, 1937–1942 (1999).
33. la Monte de, S. M. & Wands, J. R. Review of insulin and insulin-like growth factor expression, signaling, and malfunction in the
central nervous system: relevance to Alzheimer’s disease. J. Alzheimers Dis. 7, 45–61 (2005).
34. Willette, A. A. et al. Association of insulin resistance with cerebral glucose uptake in late middle-aged adults at risk for Alzheimer
disease. JAMA Neurol. 72, 1013–1020 (2015).
35. Willette, A. A. et al. Insulin resistance predicts brain amyloid deposition in late middle-aged adults. Alzheimers Dement. 11, 504–510
(2015).
36. Rajilić-Stojanović, M. & de Vos, W. M. The first 1000 cultured species of the human gastrointestinal microbiota. FEMS Microbiol.
Rev. 38, 996–1047 (2014).
37. Cani, P. D. et al. Metabolic endotoxemia initiates obesity and insulin resistance. Diabetes 56, 1761–1772 (2007).
38. Asti, A. & Gioglio, L. Can a bacterial endotoxin be a key factor in the kinetics of amyloid fibril formation? J. Alzheimers Dis. 39,
169–179 (2014).
39. Sheng, J. G. et al. Lipopolysaccharide-induced-neuroinflammation increases intracellular accumulation of amyloid precursor
protein and amyloid beta peptide in APPswe transgenic mice. Neurobiol. Dis. 14, 133–145 (2003).

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 9

www.nature.com/scientificreports/

40. Kahn, M. S. et al. Prolonged elevation in hippocampal Aβ and cognitive deficits following repeated endotoxin exposure in the
mouse. Behav. Brain. Res. 229, 176–184 (2012).
41. Kitazawa, M., Oddo, S., Yamasaki, T. R., Green, K. N. & LaFerla, F. M. Lipopolysaccharide-induced inflammation exacerbates tau
pathology by a cyclin-dependent kinase 5-mediated pathway in a transgenic model of Alzheimer’s disease. J. Neurosci. 25, 8843–8853
(2005).
42. Sy, M. et al. Inflammation induced by infection potentiates tau pathological features in transgenic mice. Am. J. Pathol. 178,
2811–2822 (2011).
43. Man, A. L. et al. Age-associated modifications of intestinal permeability and innate immunity in human small intestine. Clin. Sci.
129, 515–527 (2015).
44. Stehle, J. R. et al. Lipopolysaccharide-binding protein, a surrogate marker of microbial translocation, is associated with physical
function in healthy older adults. J. Gerontol. Ser. A-Biol. Sci. Med. Sci. 67, 1212–1218 (2012).
45. Zhan, X. et al. Gram-negative bacterial molecules associate with Alzheimer disease pathology. Neurology 87, 2324–2332 (2016).
46. O’Callaghan, A. & van Sinderen, D. Bifidobacteria and their role as members of the human gut microbiota. Front. Microbiol. 7, 1–23
(2016).
47. Arboleya, S., Watkins, C., Stanton, C. & Ross, R. P. Gut bifidobacteria populations in human health and aging. Front. Microbiol. 7,
1–9 (2016).
48. Underwood, M. A., German, J. B., Lebrilla, C. B. & Mills, D. A. Bifidobacterium longum subspecies infantis: champion colonizer of
the infant gut. Pediatr. Res. 77, 229–235 (2015).
49. Wang, Z. et al. The role of bifidobacteria in gut barrier function after thermal injury in rats. J. Trauma 61, 650–657 (2006).
50. Griffiths, E. A. et al. In vivo effects of bifidobacteria and lactoferrin on gut endotoxin concentration and mucosal immunity in Balb/c
mice. Dig. Dis. Sci. 49, 579–589 (2004).
51. Romond, M.-B. et al. Does the intestinal bifidobacterial colonisation affect bacterial translocation? Anaerobe 14, 43–48 (2008).
52. Akbari, E. et al. Effect of probiotic supplementation on cognitive function and metabolic status in Alzheimer’s disease: a randomized,
double-blind and controlled trial. Front. Aging. Neurosci. 8, 1–8 (2016).
53. Shen, L., Liu, L. & Ji, H.-F. Alzheimer’s disease histological and behavioral manifestations in transgenic mice correlate with specific
gut microbiome state. J. Alzheimers Dis. 56, 385–390 (2017).
54. Nguyen, T. L. A., Vieira-Silva, S., Liston, A. & Raes, J. How informative is the mouse for human gut microbiota research? Dis. Model.
Mech. 8, 1–16 (2015).
55. Hansen, R. A. et al. Efficacy and safety of donepezil, galantamine, and rivastigmine for the treatment of Alzheimer’s disease: a
systematic review and meta-analysis. Clin. Interv. Aging 3, 211–225 (2008).
56. Sager, M. A., Hermann, B. & La Rue, A. Middle-aged children of persons with Alzheimer’s disease: APOE genotypes and cognitive
function in the Wisconsin Registry for Alzheimer’s Prevention. J. Geriatr. Psychiatry Neurol. 18, 245–249 (2005).
57. Johnson, S. C. et al. The effect of TOMM40 poly-T length on gray matter volume and cognition in middle-aged persons with APOE
ε3/ε3 genotype. Alzheimers Dement. 7, 456–465 (2011).
58. McKhann, G. M. et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on
Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 7, 263–269
(2011).
59. Lewis, S. J. & Heaton, K. W. Stool form scale as a useful guide to intestinal transit time. Scand. J. Gastroenterol. 32, 920–924 (1997).
60. Kozich, J. J., Westcott, S. L., Baxter, N. T., Highlander, S. K. & Schloss, P. D. Development of a dual-index sequencing strategy and
curation pipeline for analyzing amplicon sequence data on the MiSeq Illumina sequencing platform. Appl. Environ. Microbiol. 79,
5112–5120 (2013).
61. Schloss, P. D. et al. Introducing mothur: open-source, platform-independent, community-supported software for describing and
comparing microbial communities. Appl. Environ. Microbiol. 75, 7537–7541 (2009).
62. Westcott, S. L. & Schloss, P. D. OptiClust, an improved method for assigning amplicon-based sequence data to operational
taxonomic units. mSphere 2, 1–11 (2017).
63. Bokulich, N. A. et al. Quality-filtering vastly improves diversity estimates from Illumina amplicon sequencing. Nat. Methods 10,
57–59 (2013).
64. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq. 2. Genome
Biol. 15, 31–21 (2014).
65. McMurdie, P. J. & Holmes, S. Waste not, want not: why rarefying microbiome data is inadmissible. PLoS Comput. Biol. 10, 1–12
(2014).
66. White, J. R., Nagarajan, N. & Pop, M. Statistical methods for detecting differentially abundant features in clinical metagenomic
samples. PLoS Comput. Biol. 5, 1–11 (2009).
67. Segata, N. et al. Metagenomic biomarker discovery and explanation. Genome Biol. 12(R60), 1–18 (2011).
68. Lewczuk, P., Lelental, N., Spitzer, P., Maler, J. M. & Kornhuber, J. Amyloid-β 42/40 cerebrospinal fluid concentration ratio in the
diagnostics of Alzheimer’s disease: validation of two novel assays. J. Alzheimers Dis. 43, 183–191 (2015).
69. Racine, A. M. et al. Associations between performance on an abbreviated CogState battery, other measures of cognitive function,
and biomarkers in people at risk for Alzheimer’s sisease. J. Alzheimers Dis. 54, 1395–1408 (2016).
70. Racine, A. M. et al. Association of longitudinal white matter degeneration and cerebrospinal fluid biomarkers of neurodegeneration,
inflammation and Alzheimer’s disease in late-middle-aged adults. Brain Imaging Behav. 59, 306 (2017).

Acknowledgements
This research was supported by a University of Wisconsin Institute for Clinical and Translational Research pilot
grant awarded to F.E.R. and B.B.B. as part of the Clinical and Translational Science Award (CTSA) program,
through the NIH National Center for Advancing Translational Sciences (NCATS) [UL1TR000427], NIH P50
AG033514 (S.A.), and R01 AG027161 (S.C.J.), R01 AG021155 (S.C.J.), R01 AG037639 (B.B.B.), the Swedish
Research Council, the Swedish Alzheimer’s Association, the Swedish Brain Foundation, and Torsten Söderberg
Foundation at the Royal Swedish Academy of Sciences. We would like to thank Martha Clare Morris at Rush
University for sharing her 15-item diet questionnaire, Matthew Beilfuss and Lauren Koenig for assistance with
study logistics, participant recruitment and screening, and Andrew J. Steinberger for assistance with sample
preparation. Finally, we extend thanks to our committed research participants who made this research possible.

Author Contributions
F.E.R., B.B.B., C.M.C., S.A. and S.C.J. contributed to study concept and experimental design. N.M.V., R.L.K.,
K.A.D., S.J.H., A.P.M., C.M.C., H.Z., K.B., B.B.B., and F.E.R. participated in data acquisition, analysis, or
interpretation. N.M.V., B.B.B., and F.E.R. drafted the manuscript. R.L.K., K.A.D., A.P.M., S.J.H., S.C.J., C.M.C.,
H.Z., K.B., and S.A. provided critical revision of the manuscript for important intellectual content.

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 10

www.nature.com/scientificreports/

Additional Information
Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-017-13601-y.
Competing Interests: The authors declare that they have no competing interests.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
ative Commons license, and indicate if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons license and your intended use is not per-
mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

© The Author(s) 2017

Scientific Reports | 7: 13537 | DOI:10.1038/s41598-017-13601-y 11