Biology and targetsColorectal cancer

Benefit restricted to RAS/BRAF wild-type (~40%)

Only 0.4% EpCAM-negative; 92.1% in grade 3 tumours

cBioPortal: 5,400 of 7,237, 74.6%, in crc_msk_2026; 867 of 1,134, 76.5%, in crc_msk_2017; 1,145 of 1,516, 75.5%, in crc_eo_2020; 387 of 534, 72.5%, in coadread_tcga_pan_can_atlas_2018; 168 of 224, 75.0%, in coadread_tcga_pub; 361 of 619, 58.3%, in coadread_dfci_2016; 451 of 1,015, 44.4%, in crc_sysucc_2022. Deep deletion adds 40 of 7,237 (crc_msk_2026) and 18 of 592 (TCGA), and structural variants disrupt APC in a further 54 of 7,237 samples. The TCGA analysis found WNT signalling altered in 93% of tumours (Cancer Genome Atlas Network 2012); adding intronic APC splice alterations and large in-frame CTNNB1 deletions took oncogenic WNT alterations to 96% of 1,134 prospectively sequenced cancers (Yaeger 2018).

cBioPortal: 5,295 of 7,237, 73.2%, in crc_msk_2026; 824 of 1,134, 72.7%, in crc_msk_2017; 1,107 of 1,516, 73.0%, in crc_eo_2020; 314 of 534, 58.8%, in coadread_tcga_pan_can_atlas_2018; 121 of 224, 54.0%, in coadread_tcga_pub; 320 of 619, 51.7%, in coadread_dfci_2016. Loss of chromosome 17p sequences was found in 75% of carcinomas but only rarely in early adenomas in the 172 specimens that built the genetic model (Vogelstein 1988).

G12C ~3-4%

cBioPortal: 3,096 of 7,237, 42.8%, in crc_msk_2026; 495 of 1,134, 43.7%, in crc_msk_2017; 666 of 1,516, 43.9%, in crc_eo_2020; 218 of 534, 40.8%, in coadread_tcga_pan_can_atlas_2018; 94 of 224, 42.0%, in coadread_tcga_pub; 173 of 619, 27.9%, in coadread_dfci_2016; 239 of 1,015, 23.5%, in crc_sysucc_2022. In CO.17, 42.3% of 394 evaluable tumours carried a K-ras exon 2 mutation (Karapetis 2008); ras-gene mutations were present in 58% of adenomas larger than 1 cm but only 9% of adenomas under 1 cm (Vogelstein 1988).

cBioPortal, mutation records: 908 of 3,169, 28.7%, in crc_msk_2026; 137 of 504, 27.2%, in crc_msk_2017; 191 of 680, 28.1%, in crc_eo_2020; 58 of 223, 26.0%, in coadread_tcga_pan_can_atlas_2018; 45 of 177, 25.4%, in coadread_dfci_2016; 73 of 243, 30.0%, in crc_sysucc_2022.

cBioPortal: 1,486 of 7,237, 20.5%, in crc_msk_2026 (E545K 331, H1047R 232, E542K 216 samples); 229 of 1,134, 20.2%, in crc_msk_2017; 298 of 1,516, 19.7%, in crc_eo_2020; 147 of 534, 27.5%, in coadread_tcga_pan_can_atlas_2018; 45 of 224, 20.1%, in coadread_tcga_pub; 132 of 619, 21.3%, in coadread_dfci_2016; 130 of 1,015, 12.8%, in crc_sysucc_2022.

cBioPortal, mutation records: 569 of 3,169, 18.0%, in crc_msk_2026; 90 of 504, 17.9%, in crc_msk_2017; 118 of 680, 17.4%, in crc_eo_2020; 37 of 223, 16.6%, in coadread_tcga_pan_can_atlas_2018; 43 of 177, 24.3%, in coadread_dfci_2016; 55 of 243, 22.6%, in crc_sysucc_2022.

cBioPortal, mutation records: 622 of 3,169, 19.6%, in crc_msk_2026; 101 of 504, 20.0%, in crc_msk_2017; 140 of 680, 20.6%, in crc_eo_2020; 49 of 223, 22.0%, in coadread_tcga_pan_can_atlas_2018; 24 of 177, 13.6%, in coadread_dfci_2016; 38 of 243, 15.6%, in crc_sysucc_2022.

cBioPortal, missense records: 491 of 3,155, 15.6%, in crc_msk_2026 (codons 59 and 61, 152; codons 117 and 146, 270; other, 69); 85 of 503, 16.9%, in crc_msk_2017; 17 of 93, 18.3%, in coadread_tcga_pub; 104 of 676, 15.4%, in crc_eo_2020. A146T alone is 185 of 3,169 KRAS records in crc_msk_2026 and 33 of 504 in crc_msk_2017.

cBioPortal: 1,127 of 7,237, 15.6%, in crc_msk_2026; 145 of 1,134, 12.8%, in crc_msk_2017; 205 of 1,516, 13.5%, in crc_eo_2020; 90 of 534, 16.9%, in coadread_tcga_pan_can_atlas_2018; 37 of 224, 16.5%, in coadread_tcga_pub; 86 of 619, 13.9%, in coadread_dfci_2016.

cBioPortal: mutation in 1,078 of 7,237, 14.9%, plus deep deletion in 205, in crc_msk_2026; 173 of 1,134, 15.3%, plus 42 deletions, in crc_msk_2017; 238 of 1,516, 15.7%, plus 46 deletions, in crc_eo_2020; 68 of 534, 12.7%, plus 28 deletions of 592, in coadread_tcga_pan_can_atlas_2018; 73 of 619, 11.8%, in coadread_dfci_2016. SMAD2 and SMAD3 are deleted alongside it (74 and 49 deep deletions in crc_msk_2026). A specific region of chromosome 18 was lost in 73% of carcinomas and 47% of advanced adenomas but only 11 to 13% of early adenomas, which is how 18q entered the model (Vogelstein 1988).

cBioPortal: 575 of 7,237, 7.9%, in crc_msk_2026 (575 of 859 BRAF mutation records); 85 of 1,134, 7.5%, in crc_msk_2017; 88 of 1,516, 5.8%, in crc_eo_2020; 48 of 534, 9.0%, in coadread_tcga_pan_can_atlas_2018; 20 of 224, 8.9%, in coadread_tcga_pub; 111 of 619, 17.9%, in coadread_dfci_2016 (a prospective population cohort, so it captures the older, right-sided, sporadic MSI cases that trial cohorts lose); 23 of 1,015, 2.3%, in crc_sysucc_2022. BRAF mutation was found in 250 of 3,063, 8.2%, of first-line metastatic patients pooled from CAIRO, CAIRO2, COIN and FOCUS (Venderbosch 2014).

cBioPortal: 912 of 7,237, 12.6%, in crc_msk_2026; 104 of 1,134, 9.2%, in crc_msk_2017; 136 of 1,516, 9.0%, in crc_eo_2020; 58 of 534, 10.9%, in coadread_tcga_pan_can_atlas_2018; 67 of 619, 10.8%, in coadread_dfci_2016. KMT2D reads 805 of 7,237 and KMT2C 605 of 7,237 in the same study.

cBioPortal: 869 of 7,237, 12.0%, in crc_msk_2026; 64 of 534, 12.0%, in coadread_tcga_pan_can_atlas_2018; 62 of 619, 10.0%, in coadread_dfci_2016; 154 of 1,516, 10.2%, in crc_eo_2020; 107 of 1,134, 9.4%, in crc_msk_2017. Named as a new significantly mutated gene by the TCGA analysis (Cancer Genome Atlas Network 2012).

cBioPortal: 1,088 of 7,237, 15.0%, in crc_msk_2026; 124 of 1,134, 10.9%, in crc_msk_2017; 58 of 534, 10.9%, in coadread_tcga_pan_can_atlas_2018; 44 of 619, 7.1%, in coadread_dfci_2016. Structural variants involving TCF7L2 appear in 4 of 594 TCGA samples (VTI1A-TCF7L2 in 3) and 12 of 7,237 in crc_msk_2026. Recurrent TCF7L2 mutations were among the new findings of the Genentech exome and transcriptome series, and the TCGA analysis reported a NAV2-TCF7L1 fusion (Seshagiri 2012, Cancer Genome Atlas Network 2012).

Stage matters more than anything else. In unselected surgical series: 38 of 276, 13.8%, MSI-high in the TCGA cohort with a further 44 MSI-low (Cancer Genome Atlas Network 2012); 91 of 529, 17.2%, in the 619-exome prospective cohort (Giannakis 2016); 770 of 7,217, 10.7%, in crc_msk_2026, which splits into 710 of 5,138 primaries (13.8%) and 59 of 2,045 metastases (2.9%); 125 of 1,513, 8.3%, in crc_eo_2020; 76 of 1,015, 7.5%, in crc_sysucc_2022 (cBioPortal). In first-line metastatic trial populations it collapses: 153 of 3,063, 5.0%, pooled across CAIRO, CAIRO2, COIN and FOCUS (Venderbosch 2014) and 18 of 515, 3.5%, in a phase 3 advanced cohort, 13 of them from MLH1 promoter hypermethylation (Koopman 2009).

cBioPortal: 550 of 7,237, 7.6%, in crc_msk_2026; 67 of 534, 12.5%, in coadread_tcga_pan_can_atlas_2018; 25 of 224, 11.2%, in coadread_tcga_pub; 52 of 619, 8.4%, in coadread_dfci_2016; 65 of 1,134, 5.7%, in crc_msk_2017. FAM123B was one of the three genes the TCGA analysis added to the expected driver list, with ARID1A and SOX9 (Cancer Genome Atlas Network 2012).

cBioPortal: 734 of 7,237, 10.1%, in crc_msk_2026; 95 of 1,134, 8.4%, in crc_msk_2017; 105 of 1,516, 6.9%, in crc_eo_2020; 46 of 534, 8.6%, in coadread_tcga_pan_can_atlas_2018; 72 of 619, 11.6%, in coadread_dfci_2016.

cBioPortal, mutation records: 219 of 3,169, 6.9%, in crc_msk_2026; 32 of 504, 6.3%, in crc_msk_2017; 42 of 680, 6.2%, in crc_eo_2020; 15 of 223, 6.7%, in coadread_tcga_pan_can_atlas_2018; 16 of 177, 9.0%, in coadread_dfci_2016; 13 of 243, 5.3%, in crc_sysucc_2022. That is about 3% of all colorectal cancers. Among 839 KRAS-mutant metastatic patients, 145, 17%, carried G12C, more often men and with lung and liver metastases, and their overall survival was shorter than with other KRAS alleles (hazard ratio 1.32) (Schirripa 2020).

cBioPortal: mutation in 446 of 7,237, 6.2%, plus deep deletion in 70, in crc_msk_2026; 68 of 1,134, 6.0%, plus 19 deletions, in crc_msk_2017; 34 of 534, 6.4%, plus 17 deletions of 592, in coadread_tcga_pan_can_atlas_2018; 51 of 619, 8.2%, in coadread_dfci_2016. PTEN mutation prevalence differs by side and by precise location (Loree 2018).

cBioPortal: 298 of 7,237, 4.1%, in crc_msk_2026; 46 of 1,134, 4.1%, in crc_msk_2017; 66 of 1,516, 4.4%, in crc_eo_2020; 33 of 534, 6.2%, in coadread_tcga_pan_can_atlas_2018; 20 of 224, 8.9%, in coadread_tcga_pub; 27 of 619, 4.4%, in coadread_dfci_2016; 24 of 1,015, 2.4%, in crc_sysucc_2022. NRAS runs the other way from KRAS on codon usage: in crc_msk_2026 the missense records split 137 at codons 59 and 61 against 123 at codons 12 and 13, with Q61K the single commonest allele (63 records), then G12D (54), Q61R (40) and Q61L (20) (cBioPortal).

cBioPortal: 416 of 7,237, 5.7%, in crc_msk_2026; 47 of 1,134, 4.1%, in crc_msk_2017; 59 of 1,516, 3.9%, in crc_eo_2020; 23 of 224, 10.3%, in coadread_tcga_pub; 19 of 534, 3.6%, in coadread_tcga_pan_can_atlas_2018; 29 of 619, 4.7%, in coadread_dfci_2016. Eight of the first colon cancer cell lines with high microsatellite instability carried TGFBR2 mutations clustered in short repeated sequences, with no surface receptor and little transcript (Markowitz 1995).

cBioPortal: 503 of 7,237, 7.0%, in crc_msk_2026; 84 of 1,134, 7.4%, in crc_msk_2017; 106 of 1,516, 7.0%, in crc_eo_2020; 33 of 534, 6.2%, in coadread_tcga_pan_can_atlas_2018; 36 of 619, 5.8%, in coadread_dfci_2016. Structural variants disrupt CTNNB1 in 29 of 7,237 crc_msk_2026 samples. Large in-frame CTNNB1 deletions, missed by standard variant calling, were part of what took WNT alteration to 96% of tumours (Yaeger 2018).

Recurrent RSPO2 and RSPO3 fusions occurred together in 10% of more than 70 colon tumours analysed by exome, transcriptome and copy number, and were mutually exclusive with APC mutation (Seshagiri 2012; coadread_genentech). Targeted panels find far fewer because they bait the RSPO3 intron but not RSPO2: cBioPortal structural variants give PTPRK-RSPO3 in 32 of 7,237 samples, 0.44%, in crc_msk_2026 (RSPO2 3 samples) and RSPO3-PTPRK in 3 of 1,516 in crc_eo_2020.

cBioPortal: 335 of 7,237, 4.6%, in crc_msk_2026; 39 of 1,134, 3.4%, in crc_msk_2017; 61 of 1,516, 4.0%, in crc_eo_2020; 24 of 534, 4.5%, plus 13 deep deletions of 592, in coadread_tcga_pan_can_atlas_2018; 44 of 619, 7.1%, in coadread_dfci_2016.

cBioPortal: 378 of 7,237, 5.2%, in crc_msk_2026; 53 of 1,134, 4.7%, in crc_msk_2017; 64 of 1,516, 4.2%, in crc_eo_2020; 20 of 534, 3.7%, in coadread_tcga_pan_can_atlas_2018; 38 of 619, 6.1%, in coadread_dfci_2016. ERBB3 mutations were among the recurrent receptor kinase events of the Genentech exome series (Seshagiri 2012).

cBioPortal high-level amplification: 354 of 7,237, 4.9%, in crc_msk_2026; 46 of 1,134, 4.1%, in crc_msk_2017; 71 of 1,516, 4.7%, in crc_eo_2020; 30 of 592, 5.1%, in coadread_tcga_pan_can_atlas_2018; 11 of 257 in coadread_tcga_pub. The TCGA integrative analysis concluded that MYC-directed transcriptional activation and repression is central to the disease even where the gene is not amplified, because APC loss switches it on (Cancer Genome Atlas Network 2012).

RAS wild-type enriched

208 of 9,643 sequenced metastatic colorectal cancers, 2.2%, and 22% of all BRAF mutations found (Jones 2017). cBioPortal, mutation records in crc_msk_2026: D594G 46, G469A 9, D594N 9, G466E 6, G469R 6, G469V 5, K601E 4, N581S 4, G466V 3, against 575 V600E; BRAF fusions appear as structural variants in 24 of 7,237 samples (TRIM24-BRAF 9, MKRN1-BRAF 4).

cBioPortal high-level amplification: 198 of 7,237, 2.7%, in crc_msk_2026; 35 of 1,134, 3.1%, in crc_msk_2017; 47 of 1,516, 3.1%, in crc_eo_2020; 20 of 592, 3.4%, in coadread_tcga_pan_can_atlas_2018; 8 of 257 in coadread_tcga_pub; 32 of 1,015 in crc_sysucc_2022. ERBB2 amplification was among the potentially drug-targetable recurrent copy-number events named in the TCGA analysis (Cancer Genome Atlas Network 2012). Immunohistochemical overexpression was 2.2% of 1,342 stage IV and 1.3% of 1,914 stage II-III patients, with 27 of 28 stage IV overexpressing cases amplified on fluorescence in situ hybridisation (Richman 2016).

cBioPortal high-level amplification: 101 of 7,237, 1.4%, in crc_msk_2026; 16 of 1,134, 1.4%, in crc_msk_2017; 19 of 1,516 in crc_eo_2020; 4 of 592 in coadread_tcga_pan_can_atlas_2018. Distal tumours are the ones that carry EGFR or HER2 amplification and overexpress epiregulin (Missiaglia 2014).

Pathogenic somatic POLE mutations in 66 of 6,517 colorectal cancers, 1.0%, across nine trials and cohorts (Domingo 2016). cBioPortal, counting only exonuclease hotspots (P286R, V411L, S297F, S459F and their neighbours): 52 of 7,237, 0.7%, in crc_msk_2026; 8 of 1,134 in crc_msk_2017; 11 of 534, 2.1%, in coadread_tcga_pan_can_atlas_2018; 6 of 224 in coadread_tcga_pub; 5 of 619 in coadread_dfci_2016; 12 of 1,015 in crc_sysucc_2022. The studies' own molecular subtype attributes agree: POLE in 8 of 1,134 (crc_msk_2017) and 13 of 1,468 (crc_eo_2020).

Enriched in MSI-high

cBioPortal structural variants: NTRK1 in 13 of 7,237, 0.18% (LMNA-NTRK1 6), and NTRK3 in 5, in crc_msk_2026; NTRK1 in 5 of 1,134 (LMNA-NTRK1 4) in crc_msk_2017; NTRK3 in 3 of 594 (ETV6-NTRK3 2) in coadread_tcga_pan_can_atlas_2018; NTRK1 in 4 of 1,516 in crc_eo_2020.

cBioPortal structural variants: 9 of 7,237, 0.12%, in crc_msk_2026; 1 of 594 (PPP4R3B-ALK) in coadread_tcga_pan_can_atlas_2018. ALK point mutations read 388 of 7,237 in crc_msk_2026 but are overwhelmingly passengers in hypermutated tumours, not the fusions that matter.

cBioPortal structural variants: 8 of 7,237, 0.11% (NCOA4-RET 4), in crc_msk_2026; 1 of 1,134 in crc_msk_2017; CCDC6-RET in 1 of 594 in coadread_tcga_pan_can_atlas_2018; 1 of 1,516 in crc_eo_2020.

cBioPortal, exonuclease hotspots: 2 of 7,237 in crc_msk_2026 and 1 of 619 in coadread_dfci_2016; any POLD1 mutation reads 366 of 7,237, 5.1%, in crc_msk_2026, almost all passengers in hypermutated tumours. Germline POLD1 p.Ser478Asn and POLE p.Leu424Val were identified as high-penetrance predisposition variants in families with multiple adenomas and early-onset colorectal cancer, with POLD1 also predisposing to endometrial cancer (Palles 2013).