Introduction

Germline pathogenic variants in BRCA1 and BRCA2 account for approximately 5% of all breast cancers and up to 30% of hereditary breast cancer. Germline BRCA1/2 mutation carriers have a high incidence of breast cancer in their lifetime [1]; the lifetime probability is approximately 57–65% for those with BRCA1 mutations and 45–49% for those with BRCA2 mutations [2]. BRCA1/2 mutations can undergo loss of heterozygosity (LOH), wherein the non-mutated allele loses function in parallel with the original identified mutation [3]. Women with germline BRCA mutations are more prone to developing breast cancer at a younger age and with more aggressive disease than those with somatic mutations [2].

Approximately 70% of breast tumors containing germline BRCA1 mutations are diagnosed as triple-negative breast cancer (TNBC), a highly aggressive subtype of breast cancer associated with poor outcomes due to limited treatment options [2, 4]. A standard treatment approach for patients with early-stage TNBC is neoadjuvant chemotherapy, including patients with BRCA1/2 mutations [4, 5]. The FDA has approved pembrolizumab plus neoadjuvant chemotherapy as standard of care for patients with high-risk early-stage TNBC [6]. In addition, several further trials have demonstrated that platinum-containing regimens were effective for TNBC in the neoadjuvant setting; however, the efficacy of neoadjuvant platinum for patients with BRCA mutations remains unclear [5, 7,8,9,10].

BRCA1/2 mediate the repair of DNA double-strand breaks (DSBs) by homologous recombination repair (HRR) [11]. Cancer cells with germline BRCA1/2 mutations rely on poly(ADP-ribose) polymerase (PARP) enzymes 1 and 2 for DNA repair [12]. Inhibition of single-strand DNA repair mechanisms mediated by PARP1/2 results in increased single-strand DNA breaks, which eventually culminate in the accumulation of DNA DSBs that are normally repaired via HRR [11, 13]. Hence, PARP inhibition is synthetically lethal in BRCA-mutated cancer cells, which are deficient in HRR [11]. Moreover, PARP inhibitors trap PARP to sites of DNA damage and prevent replication fork progression, resulting in irreparable DNA DSBs [11, 14]. Synthetic lethality therefore provides a mechanistic rationale for the treatment of BRCA-mutated breast cancer, including TNBC, with PARP inhibitors [15, 16]. In addition, pre-clinical models suggest that BRCA LOH may increase sensitivity to PARP inhibition [17].

Talazoparib is an oral PARP inhibitor approved for use in the US, EU, and multiple other countries as a monotherapy for the treatment of patients with deleterious or suspected deleterious germline BRCA-mutated human epidermal growth factor receptor 2 (HER2)-negative locally advanced or metastatic breast cancer [18, 19]. In the phase II NEOTALA study, talazoparib monotherapy in patients with germline BRCA-mutated, HER2-negative early breast cancer (Stage I–III; disease confined to the breast) including TNBC was active in the neoadjuvant setting (pathologic complete response [pCR] rate by independent central review [ICR] was 45.8% [95% confidence interval (CI), 32.0–60.6%] in the population evaluable for primary analysis on pCR) and showed a safety profile consistent with previous findings [20, 21].

The aim of the biomarker analyses reported here was to assess the mutational landscape of patients in the NEOTALA study and to explore potential associations of these factors with patient outcomes [22]. This was achieved by evaluating tumor tissue as well as blood and saliva samples from patients enrolled in the NEOTALA study with exploration of tumor BRCA mutations including germline-somatic concordance and zygosity; the non-BRCA germline DNA damage response (DDR) mutational landscape; homologous recombination deficiency (HRD) as assessed via genomic LOH; and the non-BRCA tumor mutational landscape. In addition, TP53, MYC, and RAD21 genes were selected for exploratory analysis on pCR based on mutational prevalence, potential for crosstalk with DDR pathways [23, 24], and potential associations with outcome previously detected in the phase III EMBRACA study [25, 26].

Patients and methods

Study design and patients

NEOTALA (NCT03499353) was a phase II, non-randomized, single-arm, open-label study evaluating the efficacy and safety of talazoparib in the neoadjuvant setting for patients with early-stage germline BRCA1/2-mutated TNBC. Details of the primary study have been published [20, 21].

Briefly, eligible patients were ≥18 years old with histologically confirmed early HER2-negative carcinoma of the breast, germline BRCA1/2 mutations, and no evidence of distant metastases. The primary endpoint was pCR by ICR after 24 weeks of neoadjuvant talazoparib treatment followed by surgery. Residual cancer burden (RCB) by ICR was a key secondary endpoint. RCB 0 refers to no residual invasive cancer, or pCR [27]. RCB I refers to minimal RCB, RCB II to moderate RCB, and RCB III to extensive RCB (progressive disease) [28].

Mutational analysis

To assess molecular eligibility for the NEOTALA study, peripheral blood was tested using BRACAnalysis CDx® (Myriad Genetics Laboratories, Inc., Salt Lake City, UT, USA) [20, 21]. In the analyses reported here, baseline tumor tissue was tested retrospectively using FoundationOne®CDx (Foundation Medicine, Inc., Cambridge, MA, USA), which is a qualitative next-generation sequencing (NGS)-based in vitro diagnostic test that uses targeted high throughput hybridization-based capture technology for detection of substitutions, insertion, and deletion alterations (indels), and copy number alterations (CNAs) in 324 genes and select gene rearrangements. BRCA1/2 zygosity was assessed using somatic-germline-zygosity (SGZ) [29], and germline versus tumor sequence comparisons were used to determine whether tumor BRCA mutations were of germline or somatic origin. Tumor genomic LOH (gLOH) was assessed as an exploratory metric of HRD as previously described [30,31,32].

Germline mutational status of 14 non-BRCA DDR genes (ATM, BAP1, BARD1, BRIP1, CHEK2, FANCC, MLH1, MRE11A, NBN, PALB2, RAD50, RAD51C, RAD51D, and XRCC2) was retrospectively assessed in baseline saliva samples using CustomNext®-Cancer (Ambry Genetics, CA, USA). For exploratory tumor and germline analyses, mutations were defined as known/likely pathogenic/deleterious single-nucleotide variants, insertions, deletions, rearrangements, or amplifications, unless otherwise stated. For analysis of prevalence of non-BRCA tumor mutations, CNAs were analyzed both collectively with other mutations and separately as these alterations often reflect larger genomic changes that are not necessarily associated with a specific gene [33, 34].

Patient population and biomarker analysis

The biomarker analysis populations included all patients from the NEOTALA study intent-to-treat (ITT) population with tumor samples suitable for analysis by NGS. The total number of patients evaluable for exploratory biomarker analyses varied depending on the assay (i.e., mutation detection, zygosity prediction for detected mutations [limited to short variants in tumors of sufficient quality and purity], or gLOH) and matrix (tumor or saliva). Agnostic heat-map visualizations of the FoundationOne®CDx tumor dataset were used to assess the overall mutational landscape and to identity candidate predictive biomarkers.

Efficacy endpoints

The association between the mutational status of TP53, MYC, RAD21, or RB1 and the primary endpoint pCR as per ICR was investigated using logistic regression. In addition, the association between the mutational status of these genes and RCB was explored. RCB index was assessed as a continuous variable [28] in this case rather than categorically, to facilitate this correlative analysis.

Results

Patients

A total of 61 female patients with TNBC and an average age of 44.6 years were enrolled in the ITT analysis and safety population.

Non-BRCA DDR germline landscape

Forty-nine patients were included in the population evaluable for non-BRCA germline DDR mutations. Variants of ambiguous or unknown significance were detected in the saliva of 3 (6.1%) patients, each for PALB2 and RAD50; 2 (4.1%) patients for ATM; and 1 (2.0%) patient, each for BAP1, BARD1, BRIP1, FANCC, and RAD51C (Supplementary Table S1). No patients had known/likely pathogenic non-BRCA germline DDR variants.

BRCA mutations and zygosity

Of 52 tumor samples from the population evaluable by FoundationOne®CDx, 39 (75.0%) and 13 (25.0%) patients exhibited BRCA1 and BRCA2 mutations, respectively; 1 (1.9%) patient exhibited mutations in both genes; and 1 (1.9%) patient had mutations in neither (Supplementary Table S2). Among the 45 patients evaluable for both germline and tumor analyses, 44 (97.8%) exhibited the same BRCA mutation in tumor as originally detected in germline. The remaining patient exhibited a germline BRCA1 rearrangement but lacked a tumor BRCA (tBRCA) mutation (Fig. 1). In the population evaluable for SGZ, 38/42 (90.5%) patients with tBRCA mutations were predicted to exhibit BRCA LOH (Table 1). Twenty-nine (69.0%) patients were predicted to have one or more BRCA1 mutation with LOH and 9 (21.4%) were predicted to have one or more BRCA2 mutation with LOH. Of the four patients not exhibiting BRCA LOH, two exhibited heterozygous tBRCA2 mutations, one exhibited a heterozygous tBRCA1 mutation, and one exhibited a heterozygous tBRCA2 mutation and a tBRCA1 rearrangement not amenable to SGZ prediction.

Fig. 1
figure 1

Percentage of tumors concordant with germline BRCA1 or BRCA2 mutational status. The proportion of patients with a known germline BRCA1 mutation per Myriad Genetics laboratory results who had a BRCA1 mutation (defined as known or likely pathogenic variant) detected in tumor using FoundationOne®CDx is shown, and likewise for BRCA2. One patient was discordant for BRCA mutational status (germline BRCA1 exon 13 ins 6 kb [large rearrangement]) but exhibited no BRCA1/2 mutation in tumor. All patients showing concordant BRCA1 or BRCA2 mutational status exhibited the same mutation in tumor as originally detected in germline

Full size image
Table 1 Patients with tumor BRCA mutations predicted to exhibit LOH
Full size table

Non-BRCA tumor mutational landscape

The overall tumor mutational landscape in the evaluable population is shown in Fig. 2. Based on variants with known or likely pathogenic impact, TP53 mutations were near universal across all patients with one exception (n = 51; 98.1%). Genes with mutations having ≥10% prevalence included BRCA1 (n = 39; 75.0%), MYC (n = 14; 26.9%), RAD21 (n = 14; 26.9%), BRCA2 (n = 13; 25.0%), RB1 (n = 11; 21.2%), PTEN (n = 10; 19.2%), and PIK3CA (n = 9; 17.3%) (Table 2). Excluding CNAs, non-BRCA genes with a mutational prevalence ≥10% included RB1 (n = 11; 21.2%), PIK3CA (n = 7; 13.5%), and PTEN (n = 7; 13.5%) (Table 2). In this TNBC subpopulation, MYC (14 [26.9%] patients) and RAD21 (14 [26.9%] patients) were the most frequently mutated genes based on copy number changes with known or likely pathogenic impact (Table 2). In this study, despite the higher prevalence of BRCA1 versus BRCA2 mutations, PIK3CA tended to be more frequently co-mutated with BRCA2 (n = 5/13) than with BRCA1 (n = 4/39).

Fig. 2
figure 2

Tumor mutational landscape – population evaluable by FoundationOne®CDx. The figure displays only those genes altered in two or more patients. For rearrangements, if a partner gene was present, both genes were labeled. If a patient has multiple alterations in a gene, this is indicated by the addition of a black square on the tile. The tile background color depends on the SV-LOH status determined in the following order: SV-LOH, SV-not under LOH, SV NA, or if neither of the alterations are SVs, the tile will be colored as a CN or RE. For example, (i) a patient with one SVLOH and one SV-not under LOH will be colored light blue, (ii) two SVs not under LOH will be colored light green, (iii) a patient with one SV and one CN will be colored as determined by its SV LOH status. CN copy number, LOH loss of heterozygosity, NA not available, RE rearrangement, SV short variant

Full size image
Table 2 Most prevalent (≥5.0%) tumor mutations
Full size table

An exploratory analysis of the association between the tumor mutational status of TP53, MYC, and RAD21 and pCR (ICR: breast/axilla dataset) is presented in Supplementary Table S3. The mutational status of MYC or RAD21 had no observed association with pCR (for each MYC and RAD21: odds ratio = 0.52; 95% CI, 0.12–2.30; p = 0.3893). The mutational status of TP53 had no observed association with pCR (p = 0.9629), although this analysis was confounded by the numerical imbalance favoring TP53 mutations (Supplementary Table S3). Likewise, there was no apparent association between the mutational status of these genes and RCB scores (Supplementary Fig. S1).

A joint visualization of tumor gLOH status and pCR (ICR: breast/axilla dataset) is provided in Fig. 3 showing gLOH distribution ranges from 1.5 to 42.2% for evaluable patients, with no apparent relationship of gLOH score to response, and no signs of any potential threshold gLOH score distinguishing responders from non-responders. An analysis of response employing a gLOH threshold of ≥16% (which is predictive of biallelic BRCA loss in ovarian and breast cancer [30]) found that gLOH was elevated in 24/27 (88.9%) tumors evaluable for both gLOH and pCR. Consistent with the results displayed in Fig. 3, no association between tumor gLOH status and pCR was observed but this was confounded by the imbalance in prevalence favoring gLOH-high (odds ratio of gLOH-high vs. gLOH-low = not evaluable; p = 0.9447) (Supplementary Table S4).

Fig. 3
figure 3

gLOH distribution – population evaluable for gLOH. gLOH genomic loss of heterozygosity, ND not determined, NR no response, pCR pathologic complete response, pPR pathologic partial response

Full size image

Exploration of potential associations of tumor mutations in other genes with response

To explore additional potential associations of tumor mutational status with outcome, an agnostic heat-map visualization of response as a function of mutational status was performed in the evaluable biomarker population (Fig. 4). In general, despite a low number of patients, mutations in most genes appeared to exhibit a similar distribution across response categories. For example, no association was apparent between PIK3CA and/or PTEN mutational status and response category. One potential exception was RB1, in which 6/11 (54.5%) patients with alterations had pCRs by ICR versus 1/11 (9.1%) patients with a pathologic partial response by ICR (Fig. 4), although the mutational status of RB1 was not found to be significantly associated with pCR (odds ratio = 5.67; 95% CI, 0.62–52.09; p = 0.1254) (Supplementary Table S3). Similarly, RCB scores were typically lower for RB1-mutant patients in the population with tumors evaluable for FoundationOne®CDx and RCB analysis (Supplementary Fig. S1).

Fig. 4
figure 4

Tumor mutations by response category – evaluable biomarker analysis population. CN copy number, LOH loss of heterozygosity, NA not available, NR no response, pCR pathologic complete response, pPR pathologic partial response, RE rearrangement, SV short variant

Full size image

Discussion

In this biomarker analysis of the NEOTALA study, tumor BRCA mutations were evident in all but one patient. Strong concordance (97.8%) was observed between tumor BRCA and germline BRCA mutations analyzed by FoundationOne®CDx and BRACAnalysis CDx®, respectively, except for one patient who exhibited a germline BRCA1 mutation but lacked a tumor BRCA mutation. This exception could reflect genomic instability in that the mutation may have been lost [35], or alternatively, this germline rearrangement not being detectable in the tumor by FoundationOne®CDx. The strong concordance observed in this study is consistent with the findings in the phase III EMBRACA, phase II ABRAZO, and phase III OlympiAD studies in which 95.3%, 96.4%, and 99.3% of patients, respectively, exhibited concordance between tumor and germline BRCA mutational status [26, 36, 37].

In this TNBC population, a total of 38/42 (90.5%) patients with tumor BRCA mutations evaluable for SGZ were predicted to exhibit BRCA LOH. Previous studies support a high prevalence of BRCA1/2 mutational LOH in breast cancer [30, 38]. This high incidence of BRCA LOH coupled with a low overall population size precludes meaningful analysis of tumor BRCA zygosity on response. This finding contrasts with another study analyzing germline BRCA1/2-mutated breast cancer tumors [17]; a greater proportion of patients in this study had one or more BRCA1 mutation with LOH (69.0%) than patients with one or more BRCA2 mutation with LOH (21.4%). In this previous analysis, LOH was seen in 90.2% of BRCA1 carriers and 54.3% of BRCA2 carriers [17].

No patients had non-BRCA germline DDR gene variants with known/likely pathogenicity, which may reflect the small patient numbers in this study and indicate that patients were enrolled based on the presence of germline BRCA mutations. PALB2 and RAD50 were the most commonly occurring variants of ambiguous or unknown significance (each 6.1%). In this TNBC population, tumor PIK3CA mutations, excluding CNAs, had a prevalence of 13.5%, which aligns with previous findings that show PIK3CA is frequently mutated in breast cancer [39, 40] and is potentially an actionable target in the advanced setting [41].

In this study, tumor TP53 mutations, based on variants with known or likely pathogenic impact, excluding CNAs, were near-universal (n = 51; 98.1%). This finding is consistent with the anticipated high prevalence of TP53 mutations in TNBC [15, 42]. In the NEOTALA study, MYC and RAD21 both exhibited CNAs in 26.9% of tumors, with no association with pCR or RCB. This joint high prevalence is consistent with MYC and RAD21 being closely linked on amplicon 8q24 and frequently coamplified in BRCA1-mutated breast cancer [43]. In the best versus worst responder visualization followed by Cox regression analysis of the EMBRACA study, an association between MYC amplification and short overall survival (OS) was observed in patients with TNBC receiving talazoparib (hazard ratio [HR] = 1.88; 95% CI, 1.10–3.20); this association was not evident in patients with TNBC receiving chemotherapy (HR = 0.71; 95% CI, 0.30–1.64) [25]. This finding was of interest considering that MYC has previously been linked to DDR gene regulation [24, 44]. Carey et al. demonstrate that MYC expression directly activates homologous recombination in response to DNA damage and could be used to predict response of patients with TNBC to therapy with PARP inhibitor combinations [24]. No such association with OS was seen in EMBRACA for RAD21 [25], a structural component of the cohesin complex, despite being closely linked chromosomally with MYC [43].

In the NEOTALA study, an agnostic heat-map visualization of response as a function of mutational status revealed that mutations in most genes appeared to exhibit a similar distribution across response categories. A weak association of RB1 alterations with response was observed, although this did not reach statistical significance (p = 0.1254). This finding is interesting given that a previous study demonstrated that RB1 can localize to DNA DSBs, thus promoting DNA end resection and homologous recombination [45]. Notably, all four short variants of RB1 under LOH exhibited pCR, which is intriguing given a report demonstrating that RB1 loss is associated with tumors that are extremely sensitive to platinum retreatment in ovarian cancer [46]. Further exploration of the findings from this ad hoc exploratory analysis is warranted.

Genomic LOH (an exploratory metric of HRD), assessed continuously or categorically, was not associated with response. However, it should be noted that this lack of association is in a germline BRCA-mutated context, with all but three evaluable patients gLOH-high (i.e., gLOH ≥ 16%), and therefore does not address the free-standing predictive utility of gLOH in TNBC. For example, a composite HRD score comprising gLOH, telomeric allelic imbalance, and large-scale state transitions identified TNBC tumors, including BRCA1/2-non-mutated, as being more likely to respond to platinum-containing therapy [47].

The limitations of the NEOTALA study have previously been reported [21]; however, the limitations of the current study include the following: some of the retrospective exploratory analyses involved small sample sizes and certain analyses were confounded due to imbalances in the presence or absence of particular biomarkers.

Overall, the results from this exploratory biomarker analysis of the NEOTALA study support the central role of BRCA mutations in tumor pathobiology in patients with germline BRCA1/2-mutated TNBC and the value of assessing germline BRCA mutational status for molecular eligibility for talazoparib in this indication.