Genomic profile of breast cancer brain metastases
The landscape of genomic alterations in BCBM (n=822) as compared to unpaired local BCs (n=11,988) and N-CNS (n=15,516) metastases is presented in Supplementary file 1 (prevalence all). The 30 most frequently altered genes in BCBM including the types of alterations are presented in Figure 1A whereas the frequent genes alterations in BCBMs compared to local BCs (Figure 1B) and N-CNS metastases (Figure 1C) are highlighted in the volcano plots using the log2 odds ratio (OR) and the adjusted p-values. Ascompared to local BCs, 31 genes with prevalence >3% were significantly different in their alteration prevalence in BCBMs with an FDR<0.1 and 26 of them (Figure 2A), were also enriched between BCBM and N-CNS. The most significantly enriched genes in BCBM were: TP53 (71.8%), MYC (25.9%), ERBB2 (24.6%), PTEN (16.7%), CDKN2A (10.3%), BRCA1 and CDKN2B (7.8% each), [Supplementary file 1 (FDR prev all), Figure 2A]. ESR1 alterations were more prevalent in BCBM compared to local BCs but both were significantly less than in N-CNS metastases (6.4% vs 3.7%, FDR 0.003 and 20.7%, FDR 6.7x10-27 respectively). There was no significant difference in PIK3CA prevalence between BCBMs (30.2%) and BCs (31.8%) (Figure 2B). PIK3CA alterations were also significantly more prevalent in N-CNS metastases (38.9%) as compared to both BCs (FDR 5.8x10-33) and BCBMs (FDR 3.8x10-6). Overall, 28 genes were altered in BCBM with prevalence >3% but not significantly different to local BCs (FDR>0.1, Figure 2B). Several rarer, but potentially actionable alterations were also enriched in the BCBM cohort including alterations in ROS1, KIT, NTRK1, RICTOR1, and JAK2, each representing ~2-3% of BCBM [Figure 1B, Supplementary file 1 (FDR prev all)].
When examining pathogenic short variants (SVs), the most significantly enriched genes in BCBMs relative to local BCs with an FDR<0.1 were TP53 (70.7%), PIK3CA (26.5%), BRCA1 (6.7%), ARID1A (6.4%), NF1 (6.1%), BRCA2 (5.1%) (Table 1); whereas, CDH1 (3.8%) and AKT1 (1.8%) had significantly reduced prevalence in BCBM in relation to local BCs (Table 1). When examining copy number (CN) alterations, 19 genes had a CN alteration prevalence of at least 3% in BCBM and an FDR<0.1 compared to local BCs (Table 1). The reported CN alterations represent amplifications (CN>6) for oncogenes and deep deletions (CN=0) for tumor suppressors. The top significantly amplified and deleted genes were MYC (25.9%), ERBB2 (22.5%), PIK3CA (5.23%) and PTEN (9.4%), CDKN2A (8.4%), CDKN2B (7.4%) respectively. In addition, PDL1, PDCD1LG2 (PDL2) and the stemness marker SOX2 showed a higher amplification prevalence in BCBM at 3.4% compared to 1.6%-0.96% in BC and 1.1%-1.15% in N-CNS for the 3 genes (Table 1). Finally, REs (rearrangements, large structural aberrations) were uncommon genomic alterations in BCBMs. The only gene with an RE prevalence >3% in BCBMs and FDR<0.1 was CDK12 (3.5%) (Table 1). All the data of genes showing SVs, CNs and REs in the 3 cohorts is presented in Supplementary file 2 (detailed prevalence).
Genomic profile of breast cancer brain metastases according to receptor status
ESR1 and ERBB2 alterations were as expected, more prevalent within ER-positive and HER2-positive tumours respectively, with ESR1 alterations (point mutations and amplifications) significantly enriched in ER-positive /HER2-negative BCBMs (FDR<0.001). The most frequently altered genes in BCBMs such as TP53, PIK3CA, MYC, were present within all the BCBM subtypes. The FGF/FGFR (FGF3/4/19 and FGFR1/2) pathway was frequently altered in ER-positive/HER2-negative (21.7%-22.2%), ER-positive/HER2-positive (7.6%-19.7%) and ER-negative/HER2-positive (8.2%-10.6%) but not in ER-negative/HER2-negative BCBMs. Frequently altered genes by BCBM subtype in ER-positive /HER2-negative cases were CDH1 (8%) and BRCA2 (7%); in ER-positive/HER2-positive cases, PIK3C2B (11%), MDM4 (11%), TBX3 (9%) and AKT2 (8%); in ER-negative/HER2-positive cases, LYN (9%). CDK12 was enriched in HER2-positive BCBMs (15%) compared to local BCs (8.6%) and N-CNS (8.5%) but without reaching significance (p<0.01, FDR>0.1). Significantly enriched genes FDR<0.1 by BCBM subtype in ER-negative/HER2-negative were BRCA1 (14%), CCND3 (9%), JAK2 (8%) and the immune checkpoint inhibition (ICI) marker CD274 (PDL1) (7%). The 30 most frequently altered genes according to receptor status are visualized in Figure 3 and Supplementary Figure 1 and all the data is available in Supplementary file 1.
Genomic signatures and PD-L1 expression of breast cancer brain metastases
We identified significantly higher homologous recombination deficiency as detected by genome-wide loss of heterozygosity (HRD-gLOH; cutoff 16%),18 in BCBMs compared to local BCs and N-CNS (52.0% vs 33.7% vs 31.5% respectively) with p=0.0001 for both comparisons (BCBM vs BC and BCBM vs N-CNS) (Table 2). HRD-gLOH was particularly high in the ER-positive/HER2-negative (43.2%, p=0.0682 vs BC and p=0.0306 vs N-CNS) and ER-negative/HER2-negative (70.5%, p=0.0269 vs BC and p=0.0039 vs N-CNS) BCBMs compared to both BCs and N-CNS (Table 2). Consistent with this, alterations in BRCA1, BRCA2 and PALB2 were more prevalent in BCBM (7.8%, 5.5% and 2.7%) relative to BC (4.3%, 4.2% and 1.3%) and N-CNS metastases (3.2%, 4.9% and 1.3%).
The prevalence of markers for immune checkpoint inhibition (ICI) in BCBM, tumour mutation burden (TMB; cutoff 10 mutations/Mb) and microsatellite instability (MSI), were also investigated.19,20 15.5% of BCBMs were TMB-High, 2.0% were MSI-High, 3.5% showed PD-L1 (CD274) alterations (3.4% were amplification) and 3.4% had PDCD1LG2(PD-L2, CD273) amplification. All were significantly higher in the BCBM versus local BC and N-CNS groups (p<0.0001 for all comparisons). TMB-High remained significant in ER-positive/HER2-negative (11.4%, p=0.0002) and ER-negative/HER2-negative (17.4%, p<0.0001) BCBMs compared to BCs but not in any other subtype.
PD-L1 positive expression, as identified by IHC using the Ventana PD‐L1 SP142 assay21,22 was lower in BCBMs (38.7%) compared to BCs (54.3%) but higher than N-CNS (26.2%). PD-L1 protein expression according to subtypes, was lower in BCBM than BC and higher than N-CNS (Table 2) with ER-positive/HER2-negative (23%), ER-positive/HER2-positive (27%), ER-negative/HER2-positive (57%). Of note, similar PD-L1 protein expression was observed in ER-negative/HER2-negative BCBMs (48.3%) compared to BCs (50.0%) but higher than in N-CNS (21.4%). Table 2 summarizes the percentage prevalence of the genomic signatures, of PD-L1/L2 alterations and PD-L1 positivity by IHC in the local BC, BCBM and N-CNS cohorts and within different receptor subtypes.
A subset of TMB-High BCBM samples (n=18) tested positive for PD-L1 expression (12/18, 66.7%). Although PDL1-positivity was lower in TMB-High BCBMs than the local BCs (83/113, 73.5%), it was still higher than the PDL1-positivity in TMB-High N-CNS metastatic sites (84/236, 35.6%).
Clinical correlations analysis and targetable genes
The majority of BCBMs had present at least one targetable gene (PIK3CA, ERBB2, BRCA1/2, ARID1A, PARP1, CDKN2A/2B), Figure 1 and Table 1. In addition, many of these samples tested positive for one or more ICI markers (TMB-High, MSI-High, PD-L1, PD-L2, Table 2). These data highlight that a significant proportion of patients with BCBM might be eligible for targeted therapy and/or immunotherapy either as a single agent or in combination if such somatic genetic data was available. Given this the actionability of the targetable markers was further evaluated in the OncoKB database (www.oncokb.org).23 For example, the United States Food and Drug Administration (FDA)-approved drugs (Level 1) are available for alterations in ERBB2, PIK3CA for breast cancer and other biomarkers such as MSI-High, TMB-High for all solid tumours. There is clinical evidence (Level 3) that mutations in AKT1, BRCA1/2 (somatic), IDH1 can be targeted by Capivasertib, Olaparib or Talazoparib, Ivosidenib respectively whereas biological evidence (Level 4) links CDK12 truncating mutations with Pembrolizumab-Nivolumab-Cemiplimab and CDKN2A oncogenic mutations with Palbociclib-Ribociclib-Abemaciclib. Gene alterations as identified in the BCBM cohort of this study with therapeutic drugs and level of evidence using the search terms ‘Breast cancer’, ‘CNS’ and ‘Glioma’ are summarized in Table 3. The percentage frequency of BCBM (by receptor status) where there is a targetable marker present and a list of current clinical trials as identified in www.clinicaltrials.org are summarized in Supplementary table 1.