TNBC is an aggressive subtype of breast cancer that is characterized by histological and molecular heterogeneity. Transcriptome analysis during the past decade has identified four to seven subtypes within TNBC [16–17]. Other transcriptome analyses using RNA-Seq, microarray, and combined modality have reported similar findings [18–20]. On the other hand, PAM50-defined basal-like subtype does not always correspond to TNBC phenotypically [21–23]. Recent studies support immune-enriched environment, DNA repair deficiency, PI3K/AKT/mTOR activation, and luminal androgen receptor as critical pathways in TNBC that are capable of predicting response to innovative therapies [26]. Nevertheless, unlike in the cases of HR-positive or HER2-positive subtypes, gene expression-based signatures associated with TNBC do not show robust prognostic and/or predictive values and are not implemented into clinical applications [24–25]. For example, HER2-positivity is no longer a poor prognostic factor with modern anti-HER2 targeted therapy as opposed to TNBC [30]. TNBC patients also suffer from compromised treatment outcomes if only chemotherapy is given. With most novel therapeutics for TNBC being biomarker-driven, it is of the utmost importance to gain further understanding of molecular landscapes underlying TNBC through identifying alteration-therapeutic matches [10, 27–29]. We thus conducted this study with a large panel.
We begin with the four HER2-positive, none-TNBC breast cancers. TSO500 identified 4 ERBB2-amplified and 3 ERBB2-gained breast cancers. Two of the four clinically HER2-positive cases were also ERBB2-amplified. Currently, anti-HER2 targeted therapy is directed by IHC/FISH testing, although discordance between ERBB2 copy number variation and clinical HER2 status has not been resolved [31]. In our study, ERBB2/ERBB3 mutation distribution was independent of clinical HER2 status. However, none of the reported mutations in our study aligned with the hotspots recognized by U.S. Food and Drug Administration (FDA) for the use of neratinib, a pan-HER kinase inhibitor [32].
Next, we concentrate on actionable mutations. For synthetic lethality with PARP inhibitor, indicated for TNBC or HER2-negative patients, BRCA1 R1203*/S1286Rfs/S405*/S632Ifs and BRCA2 E1571Gfs/F2254Yfs/N2135Kfs/S521*/V2151Ffs truncating mutations are potential biomarkers [33–35]. Although most patients in this study were HR-negative, the 4 cases with the AKT1 E17K hotspot mutation might benefit from the oral AKT1 inhibitor capivasertib (AZD5363) [36–37]. Similarly, PIK3CA hotspot mutations from the SOLAR-1 trial were identified, suggesting the use of selective PI3K-alpha inhibitor alpelisib [38]: e.g., E542K/E545G/E545K/H1047L/H1047R /H1047Y. Several PTEN truncating mutations found (E150*/E150Gfs/E43Gfs*7/P38Kfs/Q245*/R130*/V290Sfs) would cause loss of PI3K/AKT pathway regulation, which is also indicated for AKT1 inhibitor [39].
In addition to single gene alterations, TSO500 in this study also detected multi-gene biomarkers that are often unidentifiable with smaller panels. Our analyses identified approximately 23% of Taiwanese breast cancers to be TMB-high. This proportion was even higher for late breast cancers compared to early cases (34.5% versus 19.5%, respectively), which aligns with the findings by Henry et al. [45]. In June 2020, FDA approved the tissue-agnostic indication for Merck’s (Merck & Co., Inc., Kenilworth, NJ) immune checkpoint inhibitor (ICI) pembrolizumab for TMB-high unresected or metastatic solid tumors as alternative treatments [40–41].
TMB-high is defined as at least 10 mutations/Mb assessed by the FDA-approved FoundationOne CDx (Foundation Medicine, Cambridge, MA). Notably, diagnostic accuracy of TMB assessed by a large CGP panel and the gold standard methods has been found to be comparable; a phase II study conducted by the Friends of Cancer Research TMB Harmonization Project further enhanced the concordance of TMB estimation across different panels, accounting for the variability in panel size, gene content, and bioinformatics pipelines [42]. For TMB evaluation, TSO500 has shown a high correlation with whole exome sequencing and FoundationOne CDx (γ2: 0.86 and 0.84, respectively, both P < 0.001) among a cohort of non-small cell lung cancer from the CheckMate 568 study [43]. Despite variation in targeted regions for TMB calculations (523 genes up to 1.3Mb for TSO500 versus 328 genes up to 0.8Mb for FoundationOne CDx), agreement of TMB status between these two large-sized panels was satisfactory high. Another study using pan-cancer samples from 8 KEYNOTE studies further augments the clinical validity and applicability of TSO500 by showing its Youden index of 10.17 mutations/Mb corresponds to the cut-off value by the standard FoundationOne CDx [44]. It must be noted that advanced breast cancers were under-represented in the KEYNOTE-158, the study the FDA’s approval came from [50]. Thus, the decision for the use of ICI for advanced breast cancers should be based on programmed death-ligand one (PD-L1) expression first; TMB-high status evaluation to identify more patients who might benefit from ICI therapy should be used only for cases without a positive PD-L1 testing [46–47].
In our analyses, the MSI scores were extremely low and multi-modal. Although MSI-H/deficient mismatch repair (MMR) status has also been recognized as a multi-gene biomarker for FDA-approved pembrolizumab and dostarlimab-gxly regardless of breast cancer subtype, clinical applicability of this biomarker remains ambiguous and is still far from being implemented into a routine testing [48]. While MMR-IHC, MSI-PCR, and MSI-NGS are possible methodologies, the diagnostic consistency across these platforms has not been evaluated with cut-off remaining undetermined.
There were some limitations in the study. First, targeted sequencing was performed on residual specimens from the VGH-TAPYLOR study. Although all samples were stored in ideal temperature and condition and the overall successful rate for targeted sequencing with QC metrics was approximately 90%, nucleic acid degradation remains a possibility. Second, this study was conducted with a version of TSO500 without HRD score that has been shown to widen the potential population eligible for PARP inhibition [49]. Third, this study was observational in nature. The impact of CGP on patients’ treatment outcomes could not be evaluated because no intervention was made.