architecture and evolution of neochromosomes

Architecture and evolution of cancer neochromosomes

Tony PapenfussBioinformatics DivisionThe Walter and Eliza Hall Institute of Medical Research

Bioinformatics and Cancer Genomics LabPeter MacCallum Cancer Centre

What are chromosome-scale mutations?

• Large-scale changes to chromosomes found in cancers (& congenitally)

• Includes structures and processes

• Often complex; require ways of thinking and analysing data to make sense of them

Kataegis

• Greek for thunderstorm

• Localized hypermutation

Single nucleotide variant index

Geno

mic

dist

ance

Alexandrov et al, Nature 2013

Chromothripsis: chromosomal shattering

Stephens et al. Cell 2011

“Criteria” for chromothripsis

Korbel & Campbell, Cell 2013

Linear Breakage-Fusion-Bridge cycle

Gisselsson et al, Hum Genet 1999McClintock, Genetics 1941

Telomere loss Replication Di-centric Di-centric

Torn apart at cell division

Daughter cells inherit different chromosomes

lacking telomeres

BFB generates inverted duplications

Giant cancer-associated neochromosomes

• Giant super-numerary chromosomes found in some cancers

• Little studied• Linear or circular (rings)• Gigantic• NCs have centromeres &

telomeres• Harbour known oncogenes

Sandberg, Cancer Genet Cytogenet (2004)

Prevalence of neochromosomes in cancers

Class Tumour type NC prevalence

Mesenchymal Parosteal osteosarcoma 90%Well-differentiated liposarcoma 85%De-differentiated liposarcoma 82%Dermatofibrosarcoma protuberans 67%Overall 14%

Haematological Dendritic cell neoplasm 24%Large B cell lymphoma 18%

Overall 3%Garsed, Holloway & Thomas, Bioessays (2009)

Hereditas, Volume 42, Issue 3-4, 1956

The remarkable case of four interlocked rings forming a chain (Fig.2v) was found in a cell with quadruple chromosome number, the rest of the chromosomes being arranged in quartets. This structure is unparalleled in chromosome experience, as far as we know…

Low resolution studies of neochromosomes in well-differentiated liposarcoma (WD-LPS)

• Neochromosomes are composed of material from multiple chromosomes

• Chr12 always present

• High level of amplification of known oncogenes: MDM2, CDK4, HMGA2mFISH 778 WD-LPS cell line

Pedeutour et al. Genes Chromosomes Cancer, 1999

Amplification of neochromosomal material may involve circular breakage-fusion-bridge cycles

Gisselsson et al, Hum Genet 1999

Sequencing of WD-LPS neochromosomes

• Flow enriched neochromosome isoforms from 5 cell lines: 449 (primary), 778 (recurrence), GOT3, T1000, LPS141

• Sequenced enriched neochromosomes to 5X-30X coverage per cell line

• Performed RNA-seq in all cell lines

• Sequenced 2 patient primary WD-LPS tumours to ~30X coverage

• FISH and ChIP-seq (CENP-A) studies of centromeres in cell lines

Chromosome size

778 chrA

778 chrB

GC c

onte

nt

778 cell line

Normal chromosomes

Initial analysis

• Align the reads back to the human reference genome (hg19)

• No recurrent single nucleotide variants or indels on the neochromsome

• Several fusion genes containing novel exons

• No recurrent fusions genes

Estimation of copy number

GC (%)

GC bias

Num

ber o

f rea

ds

Position (Mb)

778 Chr12 - Read depth

Num

ber o

f rea

ds(5

kb w

indo

ws)

Estimation of copy number: Background correction and calibration

Copy

num

ber

778 Chr12 - Copy number

Position (Mb)

778 Chr22 - Copy number

Position (Mb)

Neochromosomes are composed of 100s of highly amplified genomic intervals derived focally from nearly every chromosome

778 Chr12 - Copy number ST059 Chr12 - Copy number

Position (Mb) Position (Mb)

Copy

num

ber

Amplified genomic intervals show high level of internal rearrangement

778 Chr12

How are these connected?

Detecting genomic fusions

(ii) Discordant reads

(iii) Split reads

Copy number profileDiscordant read counts

Discordantly aligned read pairs

Breakdancer

778 Chr12

DRClusterSegmentation

778 neochromosome

• 260 genomic intervals

• >500 genomic fusions

• Nearly every chromosome involved

778

ST059 ST079

LPS141GOT3

T1000

Identifying material on the neochromosomeInitial segmentation by thresholding

Scaffolding amplified genomic intervals

Large-scale structure of neochromosomes•778 NC contains a 221Mb core region

•Derived from 31Mb of the genome & includes 289 genes

•Low copy regions derived from Chr7 & 22 provide telomeric caps

•Similar structures in the other lines

•Only 1.4Mb of sequence, containing 24 genes, is recurrently amplified

How did the neochromosomes form?

Circular versus linear breakage fusion bridge

• Linear breakage fusion bridge:

• Expected to generate an excess of inversions—not observed

• Generates inverted duplications—not observed

• Additionally, in 778/449 ring chromosomes were observed in the patient primary

Fusions at the edges of amplified genomic regions on Chr12 frequently connect to other other edges on Chr12

Edge to edge fusions

A B C

AB CAssembly

We can construct a walks across most of Chr12 via intra-chromosomal edge to edge fusions

Edge to edge fusions

Amplified genomic intervalsChr12

ST059

778

Edge to edge walks involving Chr12 suggest chromothripsis prior to amplification

and initially affecting only Chr12

T1000: initiated from a fusion chromosome?

Inferring chromothripsisA. Non-uniform clustering of breakpoints (✔ ; qq-

plot)

B. Regularity of oscillating copy number (½; not Chr12)

C. Interspersed LOH (½; mono-allelic amplification)

D. Rearrangements affect 1 haplotype (✗)

E. Random ordering of fragments and fusion types (✔ )

F. Walks (✔ ; edge to edge walk & formal statistical tests)

Captured telomeres of the 778 NC bear the classic signatures of chromothripsis

Position (Mb) Position (Mb)

778 Chr7 778 Chr22

Other chromothripsis-like signatures on unenriched chromosomes

T1000 Chr4 ST059 Chr3

GC c

orre

cted

read

dep

th

Copy

num

ber

Position (Mb) Position (Mb)

Can we test whether chromothripsis occurred prior to breakage fusion bridge?

Can the patterns be explained without resorting to chromothripsis?

Computational model of circular BFB

• Uses a uniform probability distribution to model breakpoints

• Run without selection, it erodes the neochromosome away

Simple model of fitness and selection

• Assume fitness is proportional to number of copies of key genes

• Require at least one copy of each selected gene to be preserved

• Scale the fitness so that it saturates at some CN

• Choose daughter cell randomly, weighted according to fitness

fitness=3

fitness=5fitness=1

Combining multiple model runs• At each cycle, we select 1 daughter cell (or the parent) and

discard others

• Each simulation traces a single lineage in the tree of possible cell fates

• Run this 500-1000 times and look at the distributions of summary statistics. e.g. Number of genomic intervals, interval lengths, …

• We explored different models of fitness/selection & different spatial distributions for the breakpoint

• Run model with and without chromothripsis

• Used uniform, biased and empirical distributions of chromothripsis breakage

Breakage-fusion-bridge only modelNu

mbe

r of g

enom

ic in

terv

als

Breakage-fusion-bridge with chromothripsis model

Num

ber o

f gen

omic

inte

rval

s

BFB-only model do not generate edge to edge walks; BFB+chromothripsis does

Model

Model of the life history of well-differentiated liposarcoma neochromosomes

Conclusions• Sequencing flow isolated neochromosomes revealed the

dynamic processes involved in their formation

• The approaches used likely have broader relevance

• The combination of mechanisms is also likely to be used in other scenarios

• For example, the emergence of resistance to targeted therapies is sometimes driven by extreme amplification of the target

Acknowledgements

Arthur HsuVincent Corbin

Papenfuss lab, WEHI

David Thomas

Peter MacCallum Cancer Centre

Owen Marshall, MCRIAndrew Holloway Zhi-Ping Feng Mark Lansdell Ben Lansdell

Dale Garsed Jan SchroederLeon di Stefano

The Lorenzo & Pamela Galli Melanoma Research Fellowship