Part II · Normal biology, carcinogenesis, and inherited susceptibility · Chapter 6

The somatic genomic landscape

A mutational signature is not a label on a tumour. It is a process that is still running.

1 · Driver mutations and copy number architecture by subtype

Breast cancer has few recurrently mutated genes and a great deal of copy number change. That asymmetry is the first thing to hold, because it determines which assays are informative.

The comprehensive molecular portrait of the disease found only three genes mutated at high frequency across all cases, namely PIK3CA, TP53 and GATA3, with a long tail of low-frequency events¹. The same work showed that the four main expression subtypes differ more in copy number architecture than in point mutation burden. Luminal A disease is comparatively quiet and is dominated by PIK3CA mutation. Basal-like disease carries TP53 loss in the large majority of cases together with widespread segmental aneuploidy. HER2-enriched disease is defined by an amplification.

Copy number is also where the alternative classification lives. Clustering 2,000 tumours on joint copy number and expression produced ten integrative groups with distinct outcomes, several of which are defined by specific amplifications rather than by receptor status². Those groups differ in when they relapse as well as in whether they relapse³.

Whole-genome sequencing of 560 breast cancers extended the driver catalogue and, more usefully, placed the drivers alongside the rearrangement and signature landscape in the same tumours⁴.

The practical consequence is about panels. A targeted panel built around recurrently mutated genes captures the minority of this disease that is driven by point mutation. It reports poorly on the amplifications and deletions that define most of the rest. A negative panel result is a statement about the panel.

2 · Mutational signatures and the processes that generate them

A mutational signature is a characteristic distribution of base changes in their sequence context. It is extracted computationally from a catalogue of mutations, and it is a record of the processes that produced them.

The scale of the analyses is worth stating because it explains why the framework is stable. The first large extraction analysed 4,938,362 mutations from 7,042 cancers and recovered more than 20 distinct signatures⁵. The pan-cancer whole-genome analysis that followed used 84,729,690 somatic mutations from 4,645 whole genomes and 19,184 exomes, and defined 49 single-base substitution signatures, 11 doublet-base signatures, 4 clustered-base signatures and 17 small insertion and deletion signatures⁶.

Here is the part that is usually lost. A signature is not a descriptive category applied to a finished genome. It is the fingerprint of a mechanism, and mechanisms have tenses. Some signatures record an exposure that has ended. Some record a repair defect that is still present. Some record an enzyme that is still active in the tumour being treated.

That distinction is the whole argument of this chapter. A signature attributable to a defect in homologous recombination tells you about a current vulnerability, which is why it can be used to select therapy⁷. A signature attributable to an enzymatic mutator tells you that the tumour is still generating new variants, which is a different and more uncomfortable statement.

Two measurement caveats apply to every signature result. Extraction needs enough mutations, so signature calling is unreliable on small panels and on low tumour content. And signatures are correlated with one another, so attributing a given tumour's catalogue to a specific set of processes is an inference with uncertainty attached, not a readout.

3 · APOBEC and repair-associated signatures as worked examples

Two signature families dominate breast cancer, and they illustrate the two tenses.

APOBEC3 is the mutator. The APOBEC mutational pattern was found across 954,247 mutations in 2,680 exomes from 14 cancer types, and it reached 68% of all mutations in some individual samples⁸. Within breast cancer the HER2-enriched subtype was clearly enriched for it. The pattern also produces kataegis, meaning localised hypermutation clustered around rearrangement breakpoints⁵.

The causal chain has been established directly. Deleting APOBEC3A from human cancer cell lines that naturally accumulate these mutations diminished the APOBEC-associated signatures, and deleting APOBEC3B as well reduced the burden further without eliminating it⁹. APOBEC3A is the main driver of these mutations. The uracil glycosylase UNG was required for the transversion component, and loss of the translesion polymerase REV1 lowered the overall mutation burden. The signature therefore has a defined enzymology, and it is ongoing rather than historical.

That last word is why this section exists. APOBEC3 mutagenesis generates the specific hotspot alterations that terminate endocrine therapy with CDK4/6 inhibition, including PIK3CA E545K and E542K, ESR1 E380Q and RB1 loss, and APOBEC3-dominant tumours show shorter progression-free survival on that combination¹⁰. A mutational process is here acting as a therapeutic problem in its own right. Temporal heterogeneity and clonal evolution returns to it as a generator of acquired resistance.

The repair-associated signatures are the other tense. Loss of homologous recombination leaves a compound mark across substitutions, indels and rearrangements, and that mark can be read from a genome and used to predict BRCA1 and BRCA2 deficiency without inspecting the gene⁷. Crucially it detects the phenotype produced by promoter hypermethylation as well as by mutation¹¹. A germline test answers a question about inheritance. A signature answers a question about current repair capacity, and Inherited susceptibility and familial breast cancer and DNA repair, replication stress, and genomic instability depend on the difference.

4 · Structural variation, chromothripsis, and extrachromosomal DNA

Structural variation in breast cancer is not a uniform background of rearrangement. Several distinct mechanisms operate, and they leave separable traces.

Chromothripsis is a single catastrophic event. Tens to hundreds of rearrangements occur in one cellular crisis, producing a characteristic crisscrossing pattern and oscillation between two copy number states¹². Analysis of 2,658 tumours across 38 cancer types found the phenomenon pervasive, with frequencies above 50% in several cancer types, and showed that it contributes directly to oncogene amplification¹³. A large amplification is therefore not necessarily the endpoint of gradual selection. It may be the residue of one bad mitosis.

Extrachromosomal DNA is the mechanism with the sharpest consequences. Amplified oncogenes can be carried on circular elements that lack centromeres, and such elements drive tumour evolution and genetic heterogeneity¹⁴. Across many cancer types, extrachromosomal amplification is associated with oncogene amplification and worse outcome¹⁵.

The mechanism deserves one sentence of its own, because it explains rather than describes. Centromeres are what make mitotic segregation equal. Elements without them are distributed to daughter cells without any mechanism enforcing equality, so copy number drifts at every division. Under selection the population shifts up, and when selection is removed it shifts back down.

That argument is developed in full for the ERBB2 amplicon in Amplicon topology, extrachromosomal DNA, and the mechanistic origin of instability, and it is the reason heterogeneity in HER2-positive disease has a generator rather than merely a description. The point to carry forward from here is that structural mechanism predicts behaviour under treatment. Intrachromosomal amplification is comparatively stable. Extrachromosomal amplification is reversible in the population sense, in both directions.

5 · Chromatin variants as a distinct source of heterogeneity

Sequence is one tier of variation. Chromatin state is another, and it is not reducible to the first.

Two cells with identical genomes can differ in which regions are accessible, which enhancers are active and which transcription factors are loaded. That difference is heritable through division and is reversible, which is precisely what a genetic difference is not. Chromatin and transcriptional state are increasingly treated as a distinct axis of breast cancer heterogeneity alongside genomic alteration¹⁶.

The clearest functional demonstration is the drug-tolerant state. A subpopulation survives drug exposure through a chromatin-mediated programme rather than through mutation, and sensitivity returns after passage off the drug¹⁷. Non-stem cells in breast populations can also convert spontaneously to a stem-like state, so position in the hierarchy is itself a reversible property¹⁸.

Spatial genomics has begun to place these tiers in the same tissue. Mapping clone identity and expression state together in situ shows that genetically defined clones and transcriptional states are not in one-to-one correspondence¹⁹. The same clone can occupy different states in different regions.

The measurement consequence is large and is routinely ignored. Sequencing a tumour reports the genetic tier only. A tumour with no detectable subclonal structure at the DNA level may still be composed of functionally distinct populations, and no amount of deeper sequencing will find them.

6 · Proteogenomics and the limits of inferring protein from transcript

Almost every molecular classification in routine use is built on RNA or on DNA. Drugs act on proteins. The gap between those statements is what proteogenomics measures.

Integrated analysis of breast tumours across genome, transcriptome and proteome showed that correlation between messenger RNA and protein abundance is modest and varies widely by gene²⁰. Phosphoproteomic data in the same tumours identified signalling activity that was not predictable from the underlying genomic alterations.

The follow-up analysis of 122 treatment-naive primary breast cancers made the clinical point sharper²¹. Protein-level and phosphoprotein-level data refined the assessment of retinoblastoma protein status relevant to CDK4/6 inhibitor sensitivity, and identified tumour subsets on the basis of immune features not apparent from expression alone. Detailed analysis of the ERBB2 amplicon at protein level challenged assignments made on standard diagnostics.

Three tiers therefore exist, and each answers a different question. DNA reports what alterations are present. RNA reports what is being transcribed. Protein and phosphoprotein report what is actually signalling. A claim about pathway activation made from expression data is an inference across two tiers.

This is the same discipline the assay chapters enforce elsewhere in this book. A statement about amplification is a statement about DNA, and a statement about staining is a statement about protein Receptor assessment and the measurement problem. The proteogenomic data quantify how often those two disagree.

7 · Clonal architecture at diagnosis and what one biopsy represents

A tumour at diagnosis is a population. What a biopsy returns is a sample of that population, taken from one place at one time.

Single-cell copy number sequencing showed that breast tumours are composed of discrete subclonal populations rather than a continuum²². In triple-negative disease the copy number evolution was punctuated, with an early burst of rearrangement followed by stable clonal expansion²³. Multiregion sequencing of primaries found substantial diversification, with driver alterations often confined to a subset of regions within one tumour²⁴. The subclonal structure is already established in the in situ compartment, as Early subclonal diversification already detectable in situ sets out.

Quantifying how much of a tumour any biopsy represents is itself a research problem. Pan-cancer estimates of intratumour heterogeneity depend heavily on sequencing depth, purity and the deconvolution method used, and comparisons across studies using different methods are not straightforward²⁵.

Two consequences are worth separating.

The first is about detection. A subclonal driver present in 15% of cells may fall below the confident calling threshold of a routine assay. Absence of a report is not absence of the alteration.

The second is about action. A clonal alteration, present in every cell, is a target whose removal affects the whole population. A subclonal alteration is a target whose removal affects part of it. Reports rarely distinguish the two, and the distinction determines what a response should be expected to look like. Metastatic and relapse genomes are enriched for alterations that were subclonal in the primary²⁶, which is what selection acting on this structure looks like.

8 · Selection, drift, and the difficulty of calling a driver

The word driver does a great deal of work in this field, and it is worth asking what evidence supports it in any given case.

The formal definition is evolutionary. A driver is an alteration under positive selection. Applying that test across 7,664 tumours from 29 cancer types showed that positive selection dominates in cancer and that negative selection is almost absent outside homozygous loss of essential genes²⁷. The same analysis found an average of about four coding substitutions under positive selection per tumour, and estimated that half of driver substitutions lie in genes not yet recognised as cancer genes.

The consensus catalogue built from 9,423 tumour exomes using 26 algorithms identified 299 driver genes and more than 3,400 putative missense driver mutations, of which experimental validation confirmed 60% to 85% as likely drivers²⁸. The residual uncertainty in a curated, multi-method, pan-cancer catalogue is therefore substantial.

Against that stands neutral evolution. A measurable fraction of tumours show subclonal mutation frequency distributions consistent with neutral growth, in which subclonal variation accumulates without selection²⁹. Where that model holds, a recurrent subclonal alteration may be a passenger carried up by an expanding clone rather than the reason the clone expanded.

Three failure modes follow, and they are the ones to watch for in a report.

Recurrence is not selection. A gene may be recurrently mutated because it is long, late-replicating or in an APOBEC-preferred context.

Clonality is not causation. An alteration present in every cell may have been carried along by the founding clone.

Functional annotation is not evidence in this patient. A variant classified as a driver from pan-cancer data may be doing nothing in a tumour with a different dependency structure.

One further source of false positives is not tumour biology at all. Clonal haematopoiesis becomes common with age³⁰, is shaped by prior cancer therapy³¹, and contributes variants to plasma cell-free DNA³². A variant called from blood is not automatically a tumour variant, and Liquid biopsy and longitudinal monitoring sets out how to tell them apart.

References

All 32 references

1. Cancer Genome Atlas Network. Comprehensive molecular portraits of human breast tumours. Nature 2012 490:61-70. PMID 23000897
2. Curtis C, Shah SP, Chin SF, et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012 486:346-352. PMID 22522925
3. Rueda OM, Sammut SJ, Seoane JA, et al. Dynamics of breast-cancer relapse reveal late-recurring ER-positive genomic subgroups. Nature 2019 567:399-404. PMID 30867590
4. Nik-Zainal S, Davies H, Staaf J, et al. Landscape of somatic mutations in 560 breast cancer whole-genome sequences. Nature 2016 534:47-54. PMID 27135926
5. Alexandrov LB, Nik-Zainal S, Wedge DC, et al. Signatures of mutational processes in human cancer. Nature 2013 500:415-421. PMID 23945592
6. Alexandrov LB, Kim J, Haradhvala NJ, et al. The repertoire of mutational signatures in human cancer. Nature 2020 578:94-101. PMID 32025018
7. Davies H, Glodzik D, Morganella S, et al. HRDetect is a predictor of BRCA1 and BRCA2 deficiency based on mutational signatures. Nat Med 2017 23:517-525. PMID 28288110
8. Roberts SA, Lawrence MS, Klimczak LJ, et al. An APOBEC cytidine deaminase mutagenesis pattern is widespread in human cancers. Nat Genet 2013 45:970-976. PMID 23852170
9. Petljak M, Dananberg A, Chu K, et al. Mechanisms of APOBEC3 mutagenesis in human cancer cells. Nature 2022 607:799-807. PMID 35859169
10. Gupta A, Gazzo A, Selenica P, et al. APOBEC3 mutagenesis drives therapy resistance in breast cancer. Nat Genet 2025 57:1452-1462. PMID 40379787
11. Glodzik D, Bosch A, Hartman J, et al. Comprehensive molecular comparison of BRCA1 hypermethylated and BRCA1 mutated triple negative breast cancers. Nat Commun 2020 11:3747. PMID 32719340
12. Stephens PJ, Greenman CD, Fu B, et al. Massive genomic rearrangement acquired in a single catastrophic event during cancer development. Cell 2011 144:27-40. PMID 21215367
13. Cortes-Ciriano I, Lee JJ, Xi R, et al. Comprehensive analysis of chromothripsis in 2,658 human cancers using whole-genome sequencing. Nat Genet 2020 52:331-341. PMID 32025003
14. Turner KM, Deshpande V, Beyter D, et al. Extrachromosomal oncogene amplification drives tumour evolution and genetic heterogeneity. Nature 2017 543:122-125. PMID 28178237
15. Kim H, Nguyen NP, Turner K, et al. Extrachromosomal DNA is associated with oncogene amplification and poor outcome across multiple cancers. Nat Genet 2020 52:891-897. PMID 32807987
16. Visvader JE, Rosen JM, Aparicio S. Breast cancer. Cold Spring Harb Perspect Med 2024 14:a041729. PMID 38692741
17. Sharma SV, Lee DY, Li B, et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 2010 141:69-80. PMID 20371346
18. Chaffer CL, Brueckmann I, Scheel C, et al. Normal and neoplastic nonstem cells can spontaneously convert to a stem-like state. Proc Natl Acad Sci U S A 2011 108:7950-7955. PMID 21498687
19. Lomakin A, Svedlund J, Strell C, et al. Spatial genomics maps the structure, nature and evolution of cancer clones. Nature 2022 611:594-602. PMID 36352222
20. Mertins P, Mani DR, Ruggles KV, et al. Proteogenomics connects somatic mutations to signalling in breast cancer. Nature 2016 534:55-62. PMID 27251275
21. Krug K, Jaehnig EJ, Satpathy S, et al. Proteogenomic landscape of breast cancer tumorigenesis and targeted therapy. Cell 2020 183:1436-1456. PMID 33212010
22. Navin N, Kendall J, Troge J, et al. Tumour evolution inferred by single-cell sequencing. Nature 2011 472:90-94. PMID 21399628
23. Gao R, Davis A, McDonald TO, et al. Punctuated copy number evolution and clonal stasis in triple-negative breast cancer. Nat Genet 2016 48:1119-1130. PMID 27526321
24. Yates LR, Gerstung M, Knappskog S, et al. Subclonal diversification of primary breast cancer revealed by multiregion sequencing. Nat Med 2015 21:751-759. PMID 26099045
25. Andor N, Graham TA, Jansen M, et al. Pan-cancer analysis of the extent and consequences of intratumor heterogeneity. Nat Med 2016 22:105-113. PMID 26618723
26. Yates LR, Knappskog S, Wedge D, et al. Genomic evolution of breast cancer metastasis and relapse. Cancer Cell 2017 32:169-184. PMID 28810143
27. Martincorena I, Raine KM, Gerstung M, et al. Universal patterns of selection in cancer and somatic tissues. Cell 2017 171:1029-1041. PMID 29056346
28. Bailey MH, Tokheim C, Porta-Pardo E, et al. Comprehensive characterization of cancer driver genes and mutations. Cell 2018 173:371-385. PMID 29625053
29. Williams MJ, Werner B, Barnes CP, Graham TA, Sottoriva A. Identification of neutral tumor evolution across cancer types. Nat Genet 2016 48:238-244. PMID 26780609
30. Jaiswal S, Fontanillas P, Flannick J, et al. Age-related clonal hematopoiesis associated with adverse outcomes. N Engl J Med 2014 371:2488-2498. PMID 25426837
31. Bolton KL, Ptashkin RN, Gao T, et al. Cancer therapy shapes the fitness landscape of clonal hematopoiesis. Nat Genet 2020 52:1219-1226. PMID 33106634
32. Razavi P, Li BT, Brown DN, et al. High-intensity sequencing reveals the sources of plasma circulating cell-free DNA variants. Nat Med 2019 25:1928-1937. PMID 31768066