Targeted sequencing reveals the somatic mutation landscape in a Swedish breast cancer cohort

Abstract

Breast cancer (BC) is a genetically heterogeneous disease with high prevalence in Northern Europe. However, there has been no detailed investigation into the Scandinavian somatic landscape. Here, in a homogeneous Swedish cohort, we describe the somatic events underlying BC, leveraging a targeted next-generation sequencing approach. We designed a 20.5 Mb array targeting coding and regulatory regions of genes with a known role in BC (n = 765). The selected genes were either from human BC studies (n = 294) or from within canine mammary tumor associated regions (n = 471). A set of predominantly estrogen receptor positive tumors (ER + 85%) and their normal tissue counterparts (n = 61) were sequenced to ~ 140 × and 85 × mean target coverage, respectively. MuTect2 and VarScan2 were employed to detect single nucleotide variants (SNVs) and copy number aberrations (CNAs), while MutSigCV (SNVs) and GISTIC (CNAs) algorithms estimated the significance of recurrent somatic events. The significantly mutated genes (q ≤ 0.01) were PIK3CA (28% of patients), TP53 (21%) and CDH1 (11%). However, histone modifying genes contained the largest number of variants (KMT2C and ARID1A, together 28%). Mutations in KMT2C were mutually exclusive with PI3KCA mutations (p ≤ 0. 001) and half of these affect the formation of a functional PHD domain. The tumor suppressor CDK10 was deleted in 80% of the cohort while the oncogene MDM4 was amplified. Mutational signature analyses pointed towards APOBEC deaminase activity (COSMIC signature 2) and DNA mismatch repair (COSMIC signature 6). We noticed two significantly distinct patterns related to patient age; TP53 being more mutated in the younger group (29% vs 9% of patients) and CDH23 mutations were absent from the older group. The increased somatic mutation prevalence in the histone modifying genes KMT2C and ARID1A distinguishes the Swedish cohort from previous studies. KMT2C regulates enhancer activation and assists tumor proliferation in a hormone-rich environment, possibly pointing to a role in ER + BC, especially in older cases. Finally, age of onset appears to affect the mutational landscape suggesting that a larger age-diverse population incorporating more molecular subtypes should be studied to elucidate the underlying mechanisms.

Figure 1

Onco-plot of somatic SNVs in the Swedish breast cancer cohort. (A) Brick plot of somatic mutations in frequently and significantly mutated genes (SMGs). Each horizontal lane describes a single gene and the vertical lines represent different samples while different colors indicate the type of mutations (described in 1C). Age and hormone receptor status in the Swedish BC cohort are schematically described with patients ordered by age (under 70 years old shown in red, or above 70 years old shown in gray) and the receptor expression (ER, PR, HER2) is also shown (blue is positive expression, grey is negative). SMGs (PIK3CA, TP53 and CDH1) are marked with *. KMT2C contains mutations in 23% of the cohort. PIK3CA and KMT2C-mutated individuals are evenly distributed between the two age groups and 85% of TP53-mutated individuals were found in the younger group. In addition, CDH23 mutations were found exclusively in the younger age group. (B) Boxplot of the mutation variant allele fraction (VAF) per gene before (grey) and after VAF filtration 15% (light pink). The amount and clustering of variants is shown with red dots depicting the filtered set and the black dots the unfiltered set. The SMGs contain clusters of mutations with high VAF. (C) Breakdown of the different type of somatic SNVs and InDels. Missense mutations were most commonly observed (green, n = 637), followed by frameshift InDels (blue), and nonsense mutations (red). Notably, CDH1 contained more splice site mutations (orange).

Figure 2

Recurrent copy number aberrations in the Swedish breast cancer cohort. Regions of recurrent copy number amplifications (red) and deletions (blue) in the targeted array were identified with GISTIC 3.0 (q-value < 0.2). Cytobands, across all the chromosomes, containing the top 20 recurrently altered regions are annotated and marked with pink vertical lines.

Figure 3

Somatic nucleotide level mutations in PIK3CA, TP53, KMT2C and CDH1. Illustrations of the somatic mutation distribution at the protein level. Gene level information was lifted over based on the longest transcript for each of the frequently mutated genes to capture the longest space. The x-axis shows the amino acid position and the y-axis the number of mutations observed in our cohort. Transcript and protein names are included (A) In PIK3CA (NM_006218, mr 27.8%, NP_006209.2), H1047R, a well-known BC hotspot, is replicated in the Swedish cohort. (B) In TP53 (NM_000546, mr 21.3%, NP_000537.3), no hotspots were identified in the cohort but a large proportion of the mutations were in the P53 domain. (C) Lollipop plot of CDH1 (NM_004360, mr 9.9%, NP_004351.1) in which three mutations were found around the cadherin domain. (D) In KMT2C (NM_170606, mr 22.9%, NP_733751.2), the PHD domain is not affected but mutations upstream of the domain may lead to a truncated protein without this domain.

Figure 4

Mutational Signatures in the Swedish BC cohort. (A) The relative contribution of each sample to the three mutational signatures found in the Swedish cohort. The COSMIC signature 2 (green) is the APOBEC deaminase signature usually identified in BC. The COSMIC signature 5 (orange) is not clearly associated with a defined process but is found in many cancers. The COSMIC signature 6 (purple) is associated with a defective DNA mismatch repair. (B) The substitution motifs that define the three mutational signatures.