Recurrent patterns in tumour evolution

A long-standing and contentious debate in evolutionary biology concerns whether macro-evolutionary rules and trends can be deciphered, and the veracity of Gould’s famous assertion that if the tape of life were rewound and played again a different evolutionary outcome would result (Gould, 2000). Examination of rules and

patterns in evolution is an emerging theme in cancer research and has important clinical implications.

A seminal study in colorectal cancer used the frequency of somatic events across independent colorectal tumours at different stages of tumour development to infer their likely temporal order (Fearon and Vogelstein, 1990). According to this model, there are two routes resulting in colorectal cancer, one through inactivation of APC, and the other through mis-match repair deficiency. Methods that rely on (and assume) common patterns across tumours have since been extended and developed into more sophisticated mathematical models (for a review, see (Beerenwinkel et al., 2014)).

The timing of mutations can also be dissected based on their copy number. For example, if a mutation precedes a regional duplication, its copy number will be doubled. Conversely, if a mutation occurs after the duplication it will be present at only one copy (Figure 1:4). Application of such a procedure revealed mutations in TP53 mutations are early events in cutaneous squamous cell carcinomas (Durinck et al., 2011). In breast cancer, mutation data was used to time copy number events suggesting that large-scale chromosomal gains did not arise across the genome until at least 15-20% of somatic point mutations had accumulated, but subsequently were an on-going process (Nik-Zainal et al., 2012b).

Figure 1:4 Using copy number to time mutations

Mutations occurring in regions of copy number gain can be timed based on their mutational multiplicity. Mutations occurring prior to an amplification or doubling event will be present at multiple copes (red circles), whereas those occurring after amplification will only be present at one copy (blue stars).

Finally, heterogeneity itself can also illuminate the temporal sequence of somatic events in cancer. Clonal mutations, occurring on the trunk of a tumour’s

phylogenetic tree, are by definition early events, whereas subclonal events, occurring on the branches, reflect later events (summarised in Table 1-1).

Multi-region sequencing of 10 KIRCs revealed that mutations in VHL, together with loss of chromosome 3p, appear always to occur as early events in this cancer type (Gerlinger et al., 2014a). Conversely, mutations in TP53, SETD2, BAP1, PTEN and KDM5C were only ever found to be subclonal, suggesting these are often later events in KIRC evolution. In other cancer types, however, TP53 mutations have been found generally to be early events (Shah et al., 2012, Nik-Zainal et al., 2012b, Bashashati et al., 2013, Weaver et al., 2014, Yachida et al., 2010). Indeed, mutations in TP53 were found to be the only somatic events that could predict progression from Barrett’s oesophagus to oesophageal adenocarcinoma (Weaver et al., 2014), while in breast cancer and ovarian cancer, mutations in TP53 were found to occur as clonal events (Shah et al., 2012, Bashashati et al., 2013). In prostate cancer and type one low-grade glioma tumours, mutations to TP53 were often found to be subclonal (Hong et al., 2015, Suzuki et al., 2015). In combination, this implies that in many cancers mutations in TP53 may be one of the founder events during tumour evolution, while in other cancers it may play a role in maintenance and progression, occurring after the emergence of the founding cancer clone.

Analysis of paired primary tumour and metastasis in colorectal cancer consistently identified driver mutations in KRAS, NRAS and BRAF in both contexts, suggesting these somatic events are generally early and are maintained through the metastatic dissemination process in this cancer type (Brannon et al., 2014). Conversely, in myelodysplastic syndrome (MDS), clonal analysis of single biopsies revealed that mutations in NRAS were often a later event while mutations in genes involved in splicing, such as U2AF1, were often the earliest (Papaemmanuil et al., 2013).

Table 1-1 Summary of truncal and branched driver events across cancer types

Tumour type Trunk Drivers1 Branch Drivers References

Acute myeloid leukemia (AML) DNMT3A, TET2, t(15;17),t(8;21), t(16;16), inv(16) WT1, KRAS, NRAS, KIT (Welch, 2014)

Breast TP53, PIK3CA BRCA2 {Martins,2012; Nik-

Zainal ,2012.; Shah, 2012; Yates, 2015) Chronic lymphocytic

leukaemia (CLL)

MYD88 SF3B1, TP53 (Landau et al., 2013)

Colorectal2 KRAS, NRAS, BRAF

TP53, PIK3CA (Brannon et al., 2014, Vakiani et al., 2012) Ewing Sarcoma EWSR1–ETS

fusion STAG2 {Tirode, 2014) Follicular lymphoma BCl2-IGH (14;18), MLL2, CREBBP, EZH2 MYD88, TNFAIP3, MYC, TP53 (Okosun et al., 2014)

Glioma IDH1 SMARCA4, BRAF,

TP53, ATRX (Johnson et al., 2014) Myelodysplastic syndrome (MDS) SF3B1, SRSF2, U2AF1, DNMT3A

NRAS (Papaemmanuil et al.,

2013)

Melanoma BRAF NRAS, MEK1 (Van Allen et al., 2014)

Myeloma IgH rearrangement KRAS, NRAS, BRAF, FAM46C

(Bolli et al., 2014, Lohr et al., 2014, Melchor et al., 2014)

Non-small cell lung cancer (NSCLC) EGFR, KRAS, TP53 HGF (Govindan et al., 2012) Oesophageal adenocarcinoma

TP53, SMAD4 MYO18B, TRIM58, C

NTNAP5, ABCB1, PCDH9, UN C13C, SEMA5A , CCDC102B (Weaver et al., 2014) Ovarian TP53 PIK3CA, CTNNB1, NF1 (Bashashati et al., 2013, Schwarz et al., 2015) Prostate ERG rearrangements, 21q22 deletion, NKX3-1 deletion FOXP1, SPOP CDKN1B, AR amplification (Baca et al., 2013, Haffner et al., 2013, Gundem et al., 2015)

Pancreatic KRAS, CDKN2A, TP53, SMAD4

OVCH1 (Yachida and

Iacobuzio-Donahue, 2013) Renal VHL, PBRM1*, 3p LOH SETD2, BAP1, KDM5C, MTOR, TSC1, TSC2, TP53 (Gerlinger et al., 2014a, Gerlinger et al., 2012)

1 _{Genes with * have also been found to be subclonal in multi-region samples. LOH refers to Loss of}

Heterozygosity. 2 Comparative sequencing analysis was used between matched primary and metastatic colorectal lesions to define potential branched status.

In multiple myeloma (MM) known driver events, such as BRAF mutations, were found to be clonal in some tumours and subclonal in others, suggesting these alterations can contribute to either tumour initiation or maintenance and

progression (Lohr et al., 2014, Bolli et al., 2014). Similarly, in chronic lymphocytic leukaemia (CLL), mutations in TP53 were identified as subclonal in approximately 50% of tumours, and only mutations to MYD88 were almost always present in the founding cancer clones when identified (Landau et al., 2013).

Taken together these data suggest rules for tumour evolution can be deciphered but are often specific within tumour types and cannot necessarily be applied across different cancers.