IJC

International Journal of Cancer

Telomere dysfunction and chromothripsis

!lie Ernst1, David T.W. Jones2, Kendra K. Maass3, Agata Rode1, Katharina I. Deeg4, Billy Michael Chelliah Jebaraj5,
Aure
Andrey Korshunov6, Volker Hovestadt1, Michael A. Tainsky7, Kristian W. Pajtler2, Sebastian Bender2, Sebastian Brabetz2,
Susanne Gro €bner2, Marcel Kool2, Frauke Devens1, Jennifer Edelmann5, Cindy Zhang8, Pedro Castelo-Branco8, Uri Tabori8,
David Malkin9, Karsten Rippe4, Stephan Stilgenbauer5, Stefan M. Pfister2, Marc Zapatka1 and Peter Lichter1
1
Division of Molecular Genetics, German Cancer Research Center (DKFZ), Heidelberg, Germany
2
Division of Pediatric Neurooncology, German Cancer Research Center (DKFZ), Heidelberg, Germany
3
Division Functional Architecture of the Cell, German Cancer Research Center (DKFZ), Heidelberg, Germany
4
Genome Organization and Function, German Cancer Research Center (DKFZ), Heidelberg, Germany
5
Department of Internal Medicine III, University of Ulm, Germany
6
Department of Neuropathology University Hospital, Clinical Cooperation Unit Neuropathology, German Cancer Research Center (DKFZ), Heidelberg,
Germany
7
Department of Oncology, Karmanos Cancer Institute, Wayne State University School of Medicine, Detroit, MI
8
Division of Pediatric Hematology-Oncology and the Labatt Brain Tumour Research Centre, The Hospital for Sick Children, Toronto, ON, Canada

Cancer Genetics and Epigenetics

9
Division of Hematology/Oncology and Department of Pediatrics, The Hospital for Sick Children and Department of Pediatrics, University of Toronto,
Toronto, ON, Canada

Chromothripsis is a recently discovered form of genomic instability, characterized by tens to hundreds of clustered DNA rear-
rangements resulting from a single dramatic event. Telomere dysfunction has been suggested to play a role in the initiation of
this phenomenon, which occurs in a large number of tumor entities. Here, we show that telomere attrition can indeed lead to
catastrophic genomic events, and that telomere patterns differ between cells analyzed before and after such genomic catastro-
phes. Telomere length and telomere stabilization mechanisms diverge between samples with and without chromothripsis in a
given tumor subtype. Longitudinal analyses of the evolution of chromothriptic patterns identify either stable patterns between
matched primary and relapsed tumors, or loss of the chromothriptic clone in the relapsed specimen. The absence of additional
chromothriptic events occurring between the initial tumor and the relapsed tumor sample points to telomere stabilization after
the initial chromothriptic event which prevents further shattering of the genome.

Chromothripsis is a recently discovered form of genome in 2–3% of all cancers, across various tumor types, and has
instability, whereby one or a few chromosomes are affected been associated with poor prognosis in certain entities.2,5
by tens to hundreds of clustered DNA rearrangements.1,2 Chromothripsis is thought to promote, and in some cases
Localized chromosome shattering occurs as a one-time cata- perhaps even cause, cancer development, since it can lead to
strophic genomic event, followed by inaccurate repair of the the loss of tumor suppressor genes, to the formation of onco-
derivative fragments.3,4 This phenomenon has been observed genic fusions and to oncogene amplification.1,2 Several fea-
tures in tumor cells have been linked to chromothripsis,
Key words: telomere, chromothripsis, genome instability, genomic which is presumed to be a context-dependent event. Notably,
catastrophe we reported an association of chromothripsis with TP53
Additional Supporting Information may be found in the online mutation in subsets of medulloblastoma and acute myeloid
version of this article. leukemia.2 However, the precise genomic context, the trig-
Grant sponsor: Project CancerTelSys; Grant number: 01ZX1302 ger(s) for the DNA double-strand breaks as well as the mech-
(K. Rippe and S.M. Pfister, in the E:med program of the German anistic basis of the shattering and repair process are currently
Federal Ministry of Education and Research (BMBF); Grant unknown. Several nonexclusive mechanisms have been pro-
sponsor: Sys Glio; Grant number: 031A425A (P. Lichter, within posed, among which are DNA damage in micronuclei,6 pre-
BMBF); Grant sponsor: DFG; Grant number: SFB1074, subproject mature chromosome condensation, breakage-fusion-bridge
B1 (S. Stilgenbauer) (BFB) cycles and telomere dysfunction.2,7,8
DOI: 10.1002/ijc.30033 Telomeres are essential structures for genomic stability.9,10
History: Received 12 Jan 2016; Accepted 25 Jan 2016; Online 9 Feb These specialized nucleoprotein complexes stabilize chromo-
2016 some ends by protecting them from end-to-end fusions and
Correspondence to: Peter Lichter, German Cancer Research Center DNA degradation. In humans, telomeres are comprized
(DKFZ), Division of Molecular Genetics, Im Neuenheimer Feld 580, largely of (TTAGGG)n tandem repeats associated with cap-
69120 Heidelberg, Germany, Fax: 149 6221 42 4639, ping proteins, which constitute the shelterin complex. Telo-
E-mail: [email protected] meric repeat sequences are added by the telomerase enzyme,

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
 2906 Telomere dysfunction and chromothripsis

What’s new?
Chromothripsis is characterized by extensive locally clustered genomic rearrangements, whereby chromosome shattering is
followed by rejoining of the DNA fragments by error-prone repair mechanisms. The present study elaborates on a previously
proposed role in the initiation of chromothripsis for telomere erosion and breakage-fusion-bridge (BFB) cycles, in which chro-
mosomes repeatedly break and are rejoined. In cells lacking normal mechanisms for genome preservation, telomere attrition
and BFB cycles induced chromothripsis. Subsequent activation of tumor-specific telomere maintenance mechanisms prevented
further chromosomal shattering. The findings suggest that telomere maintenance pathways may represent therapeutic targets
in chromothripsis-positive tumors.

which compensates for the loss of chromosome ends during Material and Methods
cell division. In contrast to germline cells, most somatic cells Whole-genome sequencing of fibroblasts from LFS patients
lack telomerase expression.11 As a result, telomeres in Purified DNA was quantified using the Qubit Broad Range
somatic cells become progressively shorter with each cell divi- double-stranded DNA assay (Life Technologies). Genomic
Cancer Genetics and Epigenetics

sion. When they reach a critical length, telomeres are no lon- DNA was sheared using an S2 Ultrasonicator (Covaris).
ger protected by the shelterin complex, and are then Whole genome sequencing library preparations were per-
recognized as DNA double-strand breaks that trigger the formed according to the manufacturer’s instructions (NEB-
DNA damage response. As a consequence, the cells undergo Next, NEB). The quality of the libraries was assessed using a
replicative senescence and/or apoptosis as a biological barrier Bioanalyzer (Agilent). Sequencing was performed using the
to prevent neoplastic transformation.11 However, if protective Illumina MiSeq platform. Read alignment was done with
mechanisms such as p53 function are compromized, cells BWA17 to the human hg19 reference genome. Copy number
may continue to proliferate; the further erosion of telomeres analysis was performed using library cn.mops in paired mode
hinders their function of protecting the chromosome ends and applying DNAcopy for segmentation18 with R (R Core
and eventually leads to chromosomal instability.9 Unpro- Team (2015). R: A language and environment for statistical
tected or broken chromosome ends can fuse to other chro- computing. R Foundation for Statistical Computing, Vienna,
mosomal extremities via nonhomologous end joining Austria. https://www.R-project.org/). Raw reads have been
(NHEJ).12 These fusions create di-centric chromosomes that submitted to the EGA genome phenome archive with
eventually break during anaphase, when the two centromeres restricted access. The use of these cultures for research was
are pulled in opposite directions.13 The BFB cycle then con- an IRB approved project from MD Anderson Cancer Center.
tinues in the following cell cycle, when sister chromatids fuse
after DNA replication. To overcome telomere shortening,
Tumor material and patient characteristics
cancer cells must regain the ability to maintain telomeres,
Clinical samples and data were collected after receiving writ-
either through the activation of telomerase activity or via the
ten informed consent according to protocols approved by the
alternative lengthening of telomeres (ALT) pathway, which
respective institutional review boards. Tumor samples of CLL
involves DNA recombination.14 Telomere length, expression
patients were collected at the University of Ulm and were
of the telomerase reverse transcriptase (TERT) and the pres-
described in a previous study.19 Ependymoma samples were
ence of ALT are prognostic factors in several tumor
collected from various institutions and were described previ-
entities.15,16
ously.20 Glioblastoma samples were collected and analyzed
A link between telomere attrition, BFB cycles and chro-
within the ICGC PedBrain consortium (Bender, Gronych,
mothripsis was shown in the context of childhood acute
Hutter, Warnatz et al. in revision). Medulloblastoma samples
lymphoblastic leukaemia with recurrent amplification of meg-
were also collected and analyzed within the ICGC PedBrain
abase regions of chromosome 21.8 In individuals born with
consortium and were described previously.21 At least 80% of
the rare constitutional Robertsonian translocation between
tumor cell content was estimated in all tumor samples by
chromosomes 15 and 21, amplification is initiated by a chro-
H&E staining of cryosections of the piece from which nucleic
mothripsis event involving the Robertsonian chromosome.8
acid extraction was performed. Diagnoses were confirmed by
In sporadic iAMP21, BFB cycles typically represent the ini-
histopathological assessment by at least two independent
tiating event, often followed by chromothripsis.8 We
pathologists. Clinical and patient characteristics are shown in
hypothesized that telomere dysfunction, followed by BFB
Supporting Information Tables 1 and 2.
cycles, can also be a precursor to chromothripsis in other
tumor entities. We show that telomere erosion can lead to
chromothripsis and that tumors of a given entity with and qPCR analysis of telomere length
without chromothripsis differ in terms of their telomere Measurement of telomere length was carried out using a Q-
length and telomere stabilization mechanisms. PCR-based technique as previously described.22,23

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
Ernst et al. 2907

Propensity score matching mark) as previously described.26 The ALT phenotype is char-
In order to identify samples matching the chromothriptic acterized by a variable number of large, ultra-bright telomere
CLL cases according to the clinical characteristics we used DNA aggregates. Tumors containing !1% tumor cells dis-
the library MatchIt within R (http://www.jstatsoft.org/v42/ playing such aggregates were considered as ALT-positive.
i08/).24 We applied the method “nearest” to select four con-
trol samples not showing chromothripsis per affected case Measurement of telomere length by terminal restriction
being best matched according to age, mutation status of fragment analysis
TP53, SF3B1 and Notch1 as well as del 11q22.q23, trisomy Telomere lengths were determined by a terminal restriction
12q13, del 13q14 and del 17p13. fragment (TRF) kit (Roche Diagnostics, Mannheim, Ger-
many) as previously described.27 DNA samples were digested
Analysis of public TCGA datasets with the restriction enzymes RsaI and HinfI and run on aga-
Public TCGA datasets were downloaded from the TCGA rose gels. A biotinylated g-DNA molecular weight marker
website (https://tcga-data.nci.nih.gov/tcga/dataAccessMatrix. was used as DNA length standard. The DNA samples were
htm) for bladder carcinoma (BLCA), cervical squamous cell depurinated, denatured and transferred to a positively
carcinoma (CESC), colon adenocarcinoma (COAD), lung charged nylon membrane Hybond-N (Amersham Pharmacia

Cancer Genetics and Epigenetics

adenocarcinoma (LUAD), stomach adenocarcinoma (STAD), Biotech, Little Chalfont, UK) by capillary blotting. The mem-
uterine corpus endometrioid carcinoma (UCEC), n > 400 brane was washed in saline–sodium citrate buffer. The blot
cases for each tumor entity. TERT gene expression levels of was hybridized with a (TTAGGG)3 telomere probe and
tumor samples were extracted from normalized RNASeqV2 washed. Chemiluminescent detection was performed accord-
Level 3 data files. Analysis of copy-number alterations was ing to the Detection Kit (Roche Diagnostics, Basel, Switzer-
performed using the Affymetrix Human SNP Array 6.0 land). Detection was carried out on an X-ray Hyperfilm ECL.
Calculation of the mean telomere length followed the manu-
’nocnv_hg19’ Level 3 data files (containing only somatic
facturer’s instruction and formula (TeloTAGGG Telomere
alterations). TERT copy-number states were derived by
Length Assay, Roche). Alpha Ease FC version 6.0.2 was used
extracting the mean intensity of the segment containing the
to measure the band intensity or chemiluminescent signal.
TERT gene. To identify the number of breakpoints per chro-
mosome for each sample, we first filtered segments that were
TERT expression in ependymoma
supported by <100 probes, as these likely represent technical
Expression values for TERT for ependymoma cases of the
artifacts. We then counted the number of breakpoints for
RELA subgroup (which includes cases with chromothripsis)
which consecutive segments had a mean intensity difference
were based on Affymetrix GeneChip Human Genome U133
larger than 0.2. This number was then normalized for the
Plus 2.0 arrays from a publically available dataset from a pre-
respective chromosome length. vious study of our group.20 Gene expression data have been
deposited at the NCBI Gene Expression Omnibus (GEO;
FISH
http://www.ncbi.nlm.nih.gov/geo) and are accessible under
We selected a medulloblastoma patient-derived xenograft accession number GSE64415.
case with high-level amplification of the TERT locus and
chromothripsis on chromosome 17. We designed a FISH TERT expression and breakpoint analysis in glioblastoma
probe to a region of chromosome 17 for which we expected Expression values for the TERT gene were from RNA
>2 signals in this tumor because of the rearrangements asso- sequencing analyses from a glioblastoma cohort analyzed
ciated with chromothripsis, based on the paired-end sequenc- within ICGC (Bender, Gronych, Hutter, Warnatz et al. in
ing data. Two-color interphase FISH25 was performed using a revision). Chromothripsis scoring and number of breakpoints
rhodamine-labeled probe (RP11-98205; chromosome 17) and per chromosome were based on whole-genome sequencing
a Cy5-labeled probe (TERT; 117B23). Pretreatment of slides, data from the same cohort of patients. Analysis of the stabil-
hybridization, posthybridization processing and signal detec- ity of the chromothriptic patterns over time was performed
tion were performed as described previously.25 Samples based on copy-number plots derived from 450k array data
showing sufficient FISH efficiency (>90% nuclei with signals) and whole-genome sequencing data.
were evaluated by two independent investigators. Metaphase
FISH for verifying clone-mapping position was performed C-circle assay
using peripheral blood cell cultures of healthy donors as out- The C-circle assay was performed as described previously.28
lined previously. Amplifications presented as innumerable Genomic DNA was quantified using a Qubit Fluorometer
tight clusters of signals for the probe for TERT. (Life Technologies). A 30 ng DNA was combined with 10 ml
2 3 U29 Buffer, 7.5 U U29 DNA polymerase (both NEB),
Telomere-specific fluorescent in situ hybridization 0.2 mg/ml BSA, 0.1% (v/v) Tween 20, 1 mM each dATP,
Telomere-specific FISH was performed using a FITC-labeled dGTP and dTTP and incubated for 8 hr at 308C and then at
peptide nucleic acid telomere probe (Dako, Glostrup, Den- 658C for 20 min. Reactions without addition of polymerase

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
 2908 Telomere dysfunction and chromothripsis
Cancer Genetics and Epigenetics

Figure 1. (a) Copy number plots from whole-genome sequencing analysis of fibroblast cultures from four LFS patients at early and late pas-
sage for each case. One chromosome harboring copy-number oscillations characteristic of catastrophic genomic events is shown in each
case (the copy-number profiles for all chromosomes are shown in Supporting Information Figure 1).(b) C-circle levels in early and late pas-
sage fibroblasts from LFS patients. As references, ALT-negative HeLa cells and ALT-positive U2OS cells were used. For each sample, reac-
tions without addition of polymerase (- pol) were included as controls. MDAH041 cells, which are ALT-negative even at the post-crisis
stage, have been shown previously to have telomerase activity at late passages.30(c) Analysis of telomere length by qPCR in fibroblasts
from LFS patients. T/S ratios show the ratios between telomere repeat copy number and single copy gene copy number.

were included as controls. After adding 2 3 SSC, the sample might survive because of the compromized p53 damage
was dot-blotted onto a 2 3-SSC-soaked Roti-Nylon plus response.
membrane (pore size 0.45 mm, Carl Roth). The membrane To detect potential chromothriptic events, we performed
was baked for 20 min at 1208C, hybridized and developed low-coverage whole-genome sequencing of matched early
using the TeloTAGGG Telomere Length Assay Kit (Roche). passage (p13–19) and late passage (p84–308) fibroblasts from
Chemiluminescent signals were detected using a ChemiDoc four LFS patients. For each of the four postcrisis cultures,
MP imaging system (Bio-Rad). several chromosomes harboured copy-number gains and/or
losses (Supporting Information Fig. 1). Interestingly, massive
DNA rearrangements clustered on one chromosome and typ-
Results ical oscillations between two copy-number states characteris-
Telomere shortening can lead to chromothripsis tic of chromothripsis were also observed in all four post-
Healthy human fibroblasts eventually senesce in culture. Due crisis cultures (Fig. 1a). Fibroblasts from the same patients
to the absence of p53, however, fibroblasts from some analyzed at early passages showed no or very minor copy-
patients affected by Li-Fraumeni syndrome (LFS)—who har- number aberrations (Supporting Information Figure 1).
bour heterozygous germline TP53 mutations—can bypass this Analysis of DNA C-circles, partially single-stranded circu-
senescence, and continue to proliferate with further telomere lar DNA elements containing telomeric repeats that are
shortening.29 The cells acquire an altered morphology, show reported to be specific markers of ALT activity,28 demon-
chromosomal anomalies and escape from the growth crisis.29 strated that two of three analyzed postcrisis cultures for
We therefore hypothesized that telomere shortening in LFS which material was available showed activation of the ALT
fibroblasts might lead to chromothripsis, and that the cells pathway, whereas the two analyzed early passage cultures

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
Ernst et al. 2909

were ALT-negative (Fig. 1b). At late passages, most cultures fusion positive ependymomas.32 Alterations affecting other
remain negative for TERT and have long and heterogeneous genes related to TP53, such as CDKN2A (which is frequently
telomeres, while a minority demonstrates strong TERT expres- deleted in ependymomas, see Supporting Information Table
sion and short and stable telomere lengths.30 In one of the four 1) might therefore be associated with chromothripsis in this
analyzed cultures (MDAH041), which was shown to be ALT- tumor entity.
negative with the C-circle assay, telomerase activity has previ- In a cohort of patients with chronic lymphocytic leukemia
ously been reported at late passages.30 The other three cultures (CLL) analyzed on Affymetrix 6.0 SNP arrays 19 (Edelmann
were negative for TERT even at late passages, and this held true et al. in preparation), we identified seven cases with chromo-
in independent postcrisis clones from the same culture.30 thripsis and 26 control cases without chromothripsis, but
Therefore, the ability to re-activate telomerase or to use the with comparable clinical parameters (e.g. same age distribu-
ALT pathway in order to bypass senescence may be strain- tion, same status for nonchromothriptic chromosomal aber-
specific in fibroblasts derived from individuals with LFS,30 or rations and mutations known to affect the prognosis of CLL
may be related to the type of underlying TP53 mutation, which patients, including the same proportions of mutations in
may confer differential dysfunction. Quantitative real-time TP53, see Supporting Information Table 1). Unlike in medul-
PCR analysis showed that telomeres were longer in all four cul- loblastoma and ependymoma, cancer cells had shorter telo-

Cancer Genetics and Epigenetics

tures at the late passages as compared to early passages (Fig. meres in the cases with chromothripsis as compared to the
1c). These findings suggest that in such cells, telomere attrition cases without chromothripsis (Fig. 2c).
leads to catastrophic genomic events, followed by telomere re- The underlying mechanisms linking telomere maintenance
stabilization. Based on these results, we hypothesized that telo- with chromothripsis therefore seems to be entity-specific,
mere attrition might lead to chromothripsis and that telomere which prompted us to further investigate the putative role of
length might differ between tumors affected by chromothripsis telomere stabilization mechanisms in chromothripsis.
and control cases without chromothripsis.
Telomere stabilization mechanisms in tumors with
Telomere patterns differ between tumors affected by chromothripsis
chromothripsis and control cases without chromothripsis In order to proliferate indefinitely, cancer cells must elongate
We measured telomere length by terminal restriction frag- their telomeres through telomerase up-regulation or by acti-
ment (TRF) analysis in four medulloblastoma cases with vation of the ALT pathway. In the particular case of chromo-
chromothripsis and six cases without chromothripsis. All thriptic cancer cells, it is presumed that the derivative
samples were obtained from patients prior to treatment. chromosomes resulting from chromothripsis are likely stabi-
Three of the four medulloblastomas with chromothripsis lized in order to prevent further progressive chromosomal
were from patients with LFS and a Sonic Hedgehog (SHH) catastrophes that would be incompatible with cell survival.
subgroup tumor. The fourth case was also from the SHH In a cohort of ependymoma patients for whom we previ-
subgroup and harbored a somatic TP53 mutation. The con- ously published the molecular classification,20 TERT expres-
trol group included two SHH medulloblastomas, 3 WNT sion was higher in cases affected by chromothripsis as
medulloblastomas and one Group 4 medulloblastoma (all compared to cases without chromothripsis from the same
with wild-type TP53 both in the tumor and germline, see molecular subgroup (Fig. 3a). Mutations in ATRX or DAXX
Supporting Information Table 1). The medulloblastomas have been associated with ALT activation, but no such muta-
without chromothripsis showed shorter telomeres and less tion was detected in any of the tumors of the RELA sub-
frequent activation of ALT as compared to the cases for group using whole-exome sequencing. None of the samples
which chromothripsis was detected (Fig. 2a, Supporting was ALT positive as assessed by the C-circle assay (Support-
Information Fig. 2). ing Information Fig. 3), suggesting that ALT might not repre-
In order to compare telomere length between cases with sent a dominant mechanism of telomere maintenance in this
and without chromothripsis in additional tumor entities, we subgroup of ependymomas. In line with these results, no
measured telomere length by quantitative PCR in two further ependymoma showed evidence of ALT activation in a previ-
cohorts. ous study on 57 pediatric ependymomas.33 Therefore, stabili-
From a cohort of patients with ependymomas, which have zation of the chromothripsis-derived chromosomes is likely
previously been genomically characterized,20 we identified achieved through the increased TERT expression seen in
eight cases with chromothripsis and nine control cases with- ependymoma. Interestingly, focal gains of 5p including the
out chromothripsis from the same molecular subgroup TERT locus have previously been reported in ependymoma.34
(RELA fusion positive). None of the patients received chemo- In glioblastoma, high-level amplification of the TERT gene
or radiotherapy prior to the surgical removal of the tumor. locus and high TERT expression were associated with chro-
Telomere length was significantly higher in cases with chro- mothriptic cases with particularly high numbers of break-
mothripsis as compared to the cases without chromothripsis points per chromosome (Figs. 3b and 3c). In addition, we
(Fig. 2b). TP53 mutations are extremely rare in ependy- also identified three glioblastoma cases affected by chromo-
moma,31 but p53 accumulation is detected in >80% of RELA thripsis in which ALT was active (Fig. 3d), showing that

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
 2910 Telomere dysfunction and chromothripsis
Cancer Genetics and Epigenetics

Figure 2. (a) TRF blot analysis showing telomere length in medulloblastoma specimens with chromothripsis (cases 1–4) or without chromothrip-
sis (cases 6–10). Data on molecular subgroup classification of the cases, TP53 status, telomere length and detection of ALT are presented. ALT,
alternative lengthening of telomeres; SHH, Sonic hedgehog. The original TRF blot is shown as Supporting Information Figure 2.(b) Analysis of telo-
mere length by qPCR in ependymoma. T/S ratios show the ratio between telomere repeat copy number and single copy gene copy number.
n 5 15 (7 cases with chromothripsis and 8 cases without chromothripsis). Average values for two independent experiments are shown.(c) Analy-
sis of telomere length by qPCR in chronic lymphocytic leukemia (CLL). T/S ratios show the ratio between telomere repeat copy number and single
copy gene copy number. n 5 33 (7 cases with chromothripsis and 26 cases without chromothripsis). Unpaired t-tests were used to test for statis-
tical significance. Two-tailed p-values are shown.

various stabilization mechanisms can play a role in glioblas- appears that telomere stabilization after chromothriptic
tomas affected by chromothripsis. events can be achieved either by telomerase activation or by
We then investigated potential differences in stabilization ALT. We further assessed ALT activation in four SHH
mechanisms between cases with and without chromothripsis in medulloblastomas from LFS patients and three additional
additional tumor entities outside the central nervous system. SHH medulloblastomas from cases with somatic TP53 muta-
For this, we reanalyzed publicly available datasets of bladder, tions, all with chromothripsis. All seven cases were ALT-
cervical, colon, lung, stomach and uterus cancer (n > 400 for positive (one representative example is shown in Supporting
each tumor entity). Tumors with higher copy-numbers of the Information Fig. 4). The prevalence of ALT in medulloblasto-
TERT locus and with higher TERT expression showed more mas with chromothripsis therefore seems to be much higher
DNA breakpoints per chromosome (Figs. 3e and 3f), which than the general prevalence, which was reported to be "18%
may further point to a requirement for telomere stabilization in anaplastic medulloblastomas (which is also the histological
subsequent to dramatic genomic rearrangement. subtype typically seen in TP53-mutant SHH tumors).35 TERT
Based on our analysis of the fibroblast cultures from the promoter mutations have been reported in "83% of SHH
LFS patients as well as glioblastoma tumor samples, it medulloblastomas derived from adult patients and represent

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
Ernst et al. 2911

Cancer Genetics and Epigenetics

Figure 3. (a) TERT expression levels (log2 ratios) in ependymoma cases with or without chromothripsis. Unpaired t test was used to test for
statistical significance.(b) TERT expression in glioblastoma (GBM) in relation to the highest number of breakpoints per chromosome for
each case. Red dots show cases with high-level amplifications of the TERT locus. Purple dot shows a case with TERT promoter mutation.
Green dots show cases for which ALT activation was shown. R, Pearson correlation coefficient. Transcript abundance is shown in reads per
kilobase per million (RPKM). Information on alterations affecting the p53 pathway for all cases is shown in Supporting Information Table
1.(c) Copy-number plots of two glioblastoma cases with high-level amplifications of the TERT locus, showing chromothriptic patterns with
high numbers of breakpoints on chromosome 17 and chromosome 6, respectively (cases depicted in red in Figure 3b).(d) Telomere FISH
showing ALT activation in a glioblastoma specimen for which chromothripsis was detected (telomere probe in green).(e) Analysis of the rela-
tionship between the highest number of breakpoints per chromosome (normalized for chromosome size) for each case and copy number of
TERT in six tumor entities of TCGA (n > 400 per entity). *, p < 0.05; ***, p < 0.0005. Unpaired t-tests were used to test for statistical signifi-
cance within each entity.(f) Analysis of the relationship between the percentage of cases with high numbers of breakpoints per chromo-
some (normalized to chromosome size) and TERT expression in six tumor entities of TCGA.

an alternative mechanism.36 Based on these results, we con- relapse tumors from the same patients (Supporting Informa-
clude that medulloblastomas with and without chromothrip- tion Table 2). In three glioblastoma cases for which paired
sis differ in terms of their telomere stabilization mechanisms. samples were available, the rearrangements present in the ini-
To address the question of the stability of the chromo- tial sample were also detected in the relapse specimen (one
thriptic patterns over time, we then compared primary and representative example is shown in Figs. 4ad). In three

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
 2912 Telomere dysfunction and chromothripsis
Cancer Genetics and Epigenetics

Figure 4. (a–d) Whole-genome copy-number plots of a glioblastoma case based on 450k data (left panels, a and b) and whole-genome
sequencing data for chromosome 7 (right panels, c and d) showing the stability of the chromothriptic patterns between the primary tumor
and the relapse specimens. (e–g) Copy-number plots of a medulloblastoma case from a LFS patient showing chromothripsis in the primary
tumor but no chromothriptic pattern in the relapse sample. The two upper panels show the entire genome (e, f) and the lower panels (g)
depict chromosomes for which chromothripsis was detected in the primary tumor.

ependymoma cases, the characteristic chromothriptic patterns relapse specimen (Supporting Information Table 2). From
also persisted (data not shown). In a SHH medulloblastoma these longitudinal studies, we can conclude that the chromo-
of a LFS patient, the chromothriptic patterns detected in the thriptic patterns are either stabilized, in which case the
primary tumor were not observed in the relapse sample relapse specimens show very similar aberrations to the pri-
despite a genome-wide sequencing depth of more than 303 mary tumors, or they are lost by clonal selection in the
(Figs. 4eh and Supporting Information Table 2). This finding relapse samples. Strikingly, we have not observed any case
is likely explained by the outgrowth of a recurrent clone lack- where there were additional chromothriptic events in the
ing the derivative pieces and/or an increased sensitivity of the relapse as compared to the primary tumor. Therefore, telo-
chromothriptic clone to therapy. Similarly, a CLL case for mere stabilization mechanisms are likely activated after the
which matched primary and relapse samples were available occurrence of chromothripsis, to prevent continued (and pre-
showed chromothripsis in the initial sample but not in the sumably therefore lethal) genomic catastrophe.

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
Ernst et al. 2913

Reconstruction of the sequence of events by selection of stem cells with defective DNA damage
In order to test whether telomere stabilization mechanisms responses that are prone to genome instability.41
are activated after the occurrence of chromothripsis, we per- Interestingly, analyses of primary and relapsed specimens
formed 2-color interphase FISH on a SHH medulloblastoma from the same patients identified only two types of evolution
specimen from a patient with LFS showing a high-level for the chromothriptic patterns. We first observed cases
amplification of the TERT locus and chromothripsis on chro- where the clone showing chromothripsis in the primary
mosome 17. The observed patterns were identical in each tumor remained as a dominant clone in the relapse tumor,
cell, with no evidence for distinct subclones. The absence of a with very similar aberrations in both specimens. These cases
subpopulation of cells showing high-level amplification of the highlight the role of the stabilization mechanisms and show
TERT locus but not additional copies of the locus on chro- minimal independent clonal evolution after the time of diag-
mosome 17 confirms that telomere stabilization occurs either nosis. Such cases have been described in other tumor entities,
concomitant with, or more likely shortly after, the chromo- such as in CLL in the first report on chromothripsis.1 In
thriptic event (Supporting Information Fig. 5). other cases, in contrast, the clone with the derivative chro-
mosome(s) was outgrown by other clones lacking the chro-
Discussion mosome(s) affected by chromothripsis. Disappearance of a

Cancer Genetics and Epigenetics

Telomere dysfunction increases with the number of divisions chromothriptic clone was also previously been reported in a
a cell goes through. Compared to the cells of healthy subjects, case study in CLL.42 Thus, in some instances, the rearrange-
these essential protective chromosome caps shorten faster in ments resulting from chromothripsis may be important for
the cells of LFS patients, leading to an increased frequency of tumor initiation but not required for tumor maintenance.1 In
telomeric fusions in such cells.37 The LFS fibroblast cultures those cases where chromothripsis does not seem to drive
analyzed in our study were previously described as showing later tumor evolution, it may be that altered genes on the
high frequencies of dicentric chromosomes and telomere derivative chromosome may somehow sensitize the cells to
treatment. The fact that no case was detected with further
associations.29 BFB cycles and erroneous centrosome segrega-
chromothriptic events occurring between the diagnosis and
tion during mitosis may trigger chromosome fragmentation,
the relapse highlights the importance of telomere stabilization
with derivative DNA fragments being stitched back together
mechanisms in preventing further rearrangements.
through error-prone repair mechanisms. The impaired p53
Several studies provided first experimental evidence for a
damage response may allow some cells to survive such a
potential link between chromothripsis and telomere dysfunc-
genomic catastrophe. Therefore, our results provide a possible
tion. Broken telomeres were shown to lead to the accumula-
explanation for the high prevalence of chromothripsis in
tion of polyploid cells.43 The formation of polyploid cells due
medulloblastomas of LFS patients. Activation of ALT seems
to dysfunctional telomeres and cytokinesis failure may
to be exceptionally frequent in LFS-associated medulloblas-
explain the link between hyperploidy and chromothripsis.44
toma, and may mark telomeres stabilized after
In pancreatic carcinomas and osteosarcomas, telomeric dys-
chromothripsis. function may trigger chromosomal fragmentation through
We show an association between telomere dysfunction persistent breakage-fusion bridge events.45 Garsed and col-
and chromothripsis not only in the context of LFS, but also leagues described the architecture and evolution of cancer-
independently of the T53 mutation status, since only a associated neochromosomes in liposarcoma.46 In this very
minority of the analyzed CLL, glioblastoma and ependymoma specific context, telomere attrition and BFB were not shown
cases had TP53 mutations. In the cases where no germline or to act as a trigger for the initial chromothriptic events, but
somatic TP53 mutation was detected, other aberrations affect- BFB cycles were implicated in the evolution of the
ing the DNA damage response and/or potentially indirectly chromothriptic-rearranged neochromosomes. In esophageal
leading to p53 dysfunction (e.g. CDKN2A/B deletion) likely adenocarcinoma, telomere shortening was shown to be more
allowed the cell to avoid apoptosis despite telomere prominent in cases bearing localized complex rearrange-
dysfunction. ments,38 and several cases therein showed evidence of both
The telomere patterns and stabilization mechanisms iden- BFB and chromothripsis.38 In hematopoietic malignancies,
tified herein are specific to each tumor entity. For instance, Campbell and colleagues suggested dicentric chromosome
chromothripsis is associated with increased telomere length formation by telomeric erosion as a potential mechanism
in medulloblastoma and ependymoma (Figs. 2a and 2b), but leading to chromothripsis in iAMP21.8 Recently, Maciejowski
with shorter telomeres in CLL (Fig. 2c) and in esophageal and colleagues suggested a telomere-based mechanism for
adenocarcinoma.38 The persistence of very short telomeres chromothripsis in cancer.47
may be a feature of cells in which ALT is active,39 but these Our study shows that telomere dysfunction and BFB
distinct patterns may also be explained by differences in telo- cycles are a general mechanism leading to chromothripsis in
merase activation capacity depending on the lineage of the several tumor entities. In the absence of functional barriers
tumor cell of origin.40 Short and dysfunctional telomeres may preserving genome integrity, telomere attrition followed by
limit normal stem cell proliferation and predispose to cancer end-to-end fusion and subsequent breakage induces

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC
 2914 Telomere dysfunction and chromothripsis

chromothripsis. This is followed by the establishment of Acknowledgments

tumor entity-specific telomere maintenance mechanisms that We thank Achim Stephan, Andrea Wittmann, Norman Mack and David
“lock in” these changes and prevent further, lethal catastro- Westermann for generous support with the whole-genome sequencing, the
FISH and the preparation of the tumor sections. Ivo Buchhalter is gratefully
phe. Our data therefore highlight that telomere maintenance
acknowledged for support with the analysis of the whole-genome sequenc-
mechanisms may represent a target for therapeutic interven- ing data. We thank Till Milde for the patient-derived xenograft model.
tion in chromothripsis-positive tumors.

References
1. Stephens PJ, Greenman CD, Fu B, et al. Massive 18. Klambauer GSK, Mayr A, Mitterecker A, et al. tion of the Doctoral Degree at the Medical
genomic rearrangement acquired in a single cata- cn.MOPS: Mixture of poissons for discovering Faculty of Heidelberg, 2014.
strophic event during cancer development. Cell copy number variations in next genereation 33. Barszczyk M, Buczkowicz P, Castelo-Branco P,
2011;144:27–40. sequencing data with a low false discovery rate. et al. Telomerase inhibition abolishes the tumorige-
2. Rausch T, Jones DT, Zapatka M, et al. Genome Nucleic Acids Res 2012;40:e69. doi:10.1093/nar/ nicity of pediatric ependymoma tumor-initiating
sequencing of pediatric medulloblastoma links gks003. Epub 2012 Feb 1. cells. Acta Neuropathol 2014;128:863–77.
catastrophic DNA rearrangements with TP53 19. Edelmann J, Holzmann K, Miller F, et al. High- 34. Modena P, Buttarelli FR, Miceli R, et al. Predic-
mutations. Cell 2012;148:59–71. resolution genomic profiling of chronic lympho- tors of outcome in an AIEOP series of childhood
3. Korbel JO, Campbell PJ. Criteria for inference of cytic leukemia reveals new recurrent genomic ependymomas: A multifactorial analysis. Neuro-
Cancer Genetics and Epigenetics

chromothripsis in cancer genomes. Cell 2013;152: alterations. Blood 2012;120:4783–94. Oncol 2012;14:1346–56.
1226–36. 20. Pajtler KW, Witt H, Sill M, et al. Molecular clas- 35. Heaphy CM, Subhawong AP, Hong SM, et al.
4. Rode A, Maass KK, Willmund KV, Lichter P, sification of ependymal tumors across all CNS Prevalence of the alternative lengthening of telo-
Ernst A. Chromothripsis in cancer cells, an compartments, histopathological grades, and age meres telomere maintenance mechanism in
update. Int J Cancer 2015. doi:10.1002/ijc.29888. groups. Cancer Cell 2015;27:728–43. human cancer subtypes. Am J Pathol 2011;179:
[Epub ahead of print] Review. 21. Jones DT, Jager N, Kool M, et al. Dissecting the 1608–15.
5. Molenaar JJ, Koster J, Zwijnenburg DA, et al. genomic complexity underlying medulloblastoma. 36. Remke M, Ramaswamy V, Peacock J, et al. TERT
Sequencing of neuroblastoma identifies chromo- Nature 2012;488:100–5. promoter mutations are highly recurrent in SHH
thripsis and defects in neuritogenesis genes. 22. O’Callaghan N, Dhillon V, Thomas P, et al. A subgroup medulloblastoma. Acta Neuropathol
Nature 2012;483:589–93. quantitative real-time PCR method for absolute 2013;126:917–29.
6. Zhang CZ, Spektor A, Cornils H, et al. Chromo- telomere length. BioTechniques 2008;44:807–9. 37. Tabori U, Nanda S, Druker H, et al. Younger age
thripsis from DNA damage in micronuclei. 23. Jebaraj BM, Kienle D, Lechel A, et al. Telomere of cancer initiation is associated with shorter telo-
Nature 2015;522:179–84. length in mantle cell lymphoma. Blood 2013;121: mere length in Li-Fraumeni syndrome. Cancer
7. Maher CA, Wilson RK. Chromothripsis and Res 2007;67:1415–8.
1184–7.
human disease: Piecing together the shattering 38. Nones K, Waddell N, Wayte N, et al. Genomic
24. Daniel E, Ho KI, Gary K, et al. Matchlt: Non-
process. Cell 2012;148:29–32. catastrophes frequently arise in esophageal adeno-
parametric preprocessing for parametric causal
8. Li Y, Schwab C, Ryan SL, et al. Constitutional carcinoma and drive tumorigenesis. Nat Commun
inference. J Stat Software 2011;42:1–28.
and somatic rearrangement of chromosome 21 in 2014;5:5224
25. Lichter P, Tang CJ, Call K, et al. High-resolution
acute lymphoblastic leukaemia. Nature 2014;508: 39. Cesare AJ, Reddel RR. Alternative lengthening of
mapping of human chromosome 11 by in situ
98–102. telomeres: Models, mechanisms and implications.
hybridization with cosmid clones. Science 1990;
9. Falandry C, Bonnefoy M, Freyer G, et al. Biology Nat Rev Genet 2010;11:319–30.
247:64–9.
of cancer and aging: A complex association with 40. Ohyashiki JH, Sashida G, Tauchi T, et al. Telo-
26. Schwartzentruber J, Korshunov A, Liu XY, et al.
cellular senescence. J Clin Oncol 2014;32:2604–10. meres and telomerase in hematologic neoplasia.
Driver mutations in histone H3.3 and chromatin
10. Wong JM, Collins K. Telomere maintenance and Oncogene 2002;21:680–7.
remodelling genes in paediatric glioblastoma.
disease. Lancet 2003;362:983–8. 41. Calado RT, Regal JA, Hills M, et al. Constitu-
Nature 2012;482:226–31.
11. Sharpless NE, Sherr CJ. Forging a signature of in tional hypomorphic telomerase mutations in
27. Tabori U, Wong V, Ma J, et al. Telomere mainte-
vivo senescence. Nat Rev Cancer 2015;15:397– patients with acute myeloid leukemia. Proc Natl
nance and dysfunction predict recurrence in
408. Acad Sci USA 2009;106:1187–92.
12. Marzec P, Armenise C, Perot G, et al. Nuclear- paediatric ependymoma. Br J Cancer 2008;99: 42. Bassaganyas L, Bea S, Escaramis G, et al. Sporadic
receptor-mediated telomere insertion leads to 1129–35. and reversible chromothripsis in chronic lympho-
genome instability in ALT cancers. Cell 2015;160: 28. Henson JD, Cao Y, Huschtscha LI, et al. DNA C- cytic leukemia revealed by longitudinal genomic
913–27. circles are specific and quantifiable markers of analysis. Leukemia 2013;27:2376–9.
13. McClintock B. The stability of broken ends of alternative-lengthening-of-telomeres activity. Nat 43. Pampalona J, Frias C, Genesca A, et al. Progres-
chromosomes in zea mays. Genetics 1941;26:234– Bio 2009;27:1181–5. sive telomere dysfunction causes cytokinesis fail-
82. 29. Bischoff FZ, Yim SO, Pathak S, et al. Spontane- ure and leads to the accumulation of polyploid
14. Henson JD, Neumann AA, Yeager TR, et al. ous abnormalities in normal fibroblasts from cells. PLoS Genet 2012;8:e1002679
Alternative lengthening of telomeres in mamma- patients with Li-Fraumeni cancer syndrome: 44. Mardin BR, Drainas AP, Waszak SM, et al. A
lian cells. Oncogene 2002;21:598–610. Aneuploidy and immortalization. Cancer Res cell-based model system links chromothripsis
15. Zhang C, Chen X, Li L, Zhou Y, Wang C, Hou S. 1990;50:7979–84. with hyperploidy. Mol Syst Biol 2015;11:828
The association between telomere length and can- 30. Gollahon LS, Kraus E, Wu TA, et al. Telomerase 45. Gisselsson D, Jonson T, Petersen A, et al. Telo-
cer prognosis: Evidence from a meta-analysis. activity during spontaneous immortalization of mere dysfunction triggers extensive DNA frag-
PLOS One 2015. doi:10.1371/journal.- Li-Fraumeni syndrome skin fibroblasts. Oncogene mentation and evolution of complex chromosome
pone.0133174. eCollection 2015. 1998;17:709–17. abnormalities in human malignant tumors. Proc
16. Hakin-Smith V, Jellinek DA, Levy D, et al. Alter- 31. Ohgaki H, Eibl RH, Schwab M, et al. Mutations Natl Acad Sci USA 2001;98:12683–8.
native lengthening of telomeres and survival in of the p53 tumor suppressor gene in neoplasms 46. Garsed DW, Marshall OJ, Corbin VD, et al. The
patients with glioblastoma multiforme. Lancet of the human nervous system. Mol Carcinog architecture and evolution of cancer neochromo-
2003;361:836–8. 1993;8:74–80. somes. Cancer Cell 2014;26:653–67.
17. LHaD R. Fast and accurate short read alignment 32. Tzaridis TD. Novel P53-targeted therapeutic 47. Maciejowski J, Li Y, Bosco N, et al. Chromothrip-
with Burrows-Wheeler Transform. Bioinformatics approaches to the treatment of high-risk ependy- sis and kataegis induced by telomere crisis. Cell
2009;25:1754–60. momas. Inaugural Dissertation for the Acquisi- 2015;163:1641–54.

Int. J. Cancer: 138, 2905–2914 (2016) V

C 2016 UICC