Introduction

BRCA1/BRCA2 germline mutations are found in up to 5% of breast cancer patients [1]. Germline mutations in BRCA1 confer a cumulative breast and ovarian cancer risk by age 80 of 72% and 44%, respectively, while BRCA2 mutations represent 69% and 17% risks of breast and ovarian cancer [2].

BRCA1/BRCA2 are tumor suppressor genes that encode large, ubiquitous and multifunctional proteins. They play central roles in DNA repair, cell-cycle control and chromosomal stability [3]. Carcinogenesis in carriers of BRCA germline mutations implies a somatic loss of the second allele either by loss of heterozygosity [4] or somatic mutation of the second allele [5]. Without functional BRCA1/BRCA2 proteins, cells are severely impaired in their ability to perform homologous recombination (HR) of double strand breaks (DSBs) of DNA [6, 7]. The resulting genomic instability makes BRCA mutated tumor cells susceptible to DNA damaging agents that induce DSBs, such as interstrand cross-linking agents (platinum derivatives and alkylating agents), or anthracyclines [8].

Recently, the question has arisen as to whether non-cancerous cells would be more sensitive to DNA damaging agents due to BRCA1/BRCA2 haploinsufficiency (mutation of a single allele). Preclinical data support this concept, revealing tissue and cell-type-specific genomic instability in non-tumoral tissues [9, 10]. Asymptomatic BRCA1 carriers have increased radiosensitivity [11]. DNA repair deficiency is further impacted as BRCA1/BRCA2 stores are depleted during genotoxic and replication stress [12]. In breast cancer patients, conflicting results have been reported regarding the correlation between BRCA1/BRCA2 germline mutations and chemotherapy-related acute hematological toxicity [13,14,15]. We hypothesized that BRCA1/BRCA2 haploinsufficiency in carriers of germline mutations could increase hematological toxicity of DNA damaging agent-based chemotherapy and that the location of mutations may play a role in this effect.

Materials and methods

Patient selection

Swiss cohort

We included all women screened for BRCA1/BRCA2 germline mutations at the Hôpitaux Universitaires de Genève, Geneva, Switzerland, and who received neoadjuvant or adjuvant chemotherapy for primary (non-metastatic) invasive breast cancer at the Hopitaux Universitaires de Genève or at seven participating clinics in Geneva between 1998 and 2016 (Swiss cohort). Exclusion criteria were a history of chemotherapy for a prior cancer, metastatic disease, primary prophylaxis with G-CSF, and treatment outside Geneva. All data were collected from medical records. Tumor characteristics and laboratory results were collected from pathology and laboratory reports. The study was approved by the local IRB in Geneva (CCER-15-158) and Lyon. Informed written consent was obtained from living patients.

French cohort

In addition to the Swiss cohort, we collected clinical data from a second cohort of patients (French cohort). Women were treated for primary breast cancer from 1990 to 2015 and screened for germline mutations of BRCA1/BRCA2 genes at the Centre Léon Bérard, Lyon, France, and at the Hospices Civils de Lyon, Lyon, France. All the patients in the French cohort received adjuvant or neoadjuvant chemotherapy. The study was approved by the local IRB. Informed written consent was obtained from all patients.

Data collection

For the Swiss cohort, we recorded tumor characteristics (TNM stage, grade, estrogen and progesterone receptors status, HER-2 status), hematological values (leukocyte count, neutrophil count, hemoglobin and platelet count) at baseline and at day 7–14 after the first chemotherapy cycle (C1), acute and long-term chemotherapy toxicity. All the patients had systematic check-up and blood counts at baseline (day 1) and day 7–14 after the C1. Hematological toxicity was graded according to the Common Terminology Criteria for Adverse Events (CTCAE) [16].

We collected chemotherapy regimen, timing (neoadjuvant or adjuvant), number of cycles and duration of the entire treatment. We recorded prophylactic G-CSF use following neutropenia in prior chemotherapy cycles, dose reductions, febrile neutropenia and hospitalizations.

The incidence of secondary acute leukemia and myelodysplastic syndromes, and long-term consequences following chemotherapy (cardiac dysfunction and neuropathy) were collected in the Swiss and French cohorts.

Genetic analysis

BRCA1 and BRCA2 mutations were classified as pathogenic according to the ENIGMA BRCA1/BRCA2 Gene Variant Classification Criteria (http://www.enigmaconsortium.org/). Women with variants of uncertain significance were considered as non-carriers. Blood samples for germline DNA testing were obtained when the patients were referred to clinical genetics units. All participants were screened for BRCA1 and BRCA2 mutations. BRCA1 comprises 24 exons and encodes an 1863 amino-acid (AA) protein (https://www.ncbi.nlm.nih.gov/protein/AAC37594.1). The most important functional domains are the RING domain (amino acids 26–68) and the BRCT domain (amino acids 1650–1842). BRCA2 comprises 27 exons and encodes a 3418 amino acids protein (https://www.ncbi.nlm.nih.gov/protein/AAB07223.1). RAD51 binding domain (RAD51-BD) corresponds to the region covering amino acids 1002–2085 of BRCA2 (exon 11) (Fig. 1).

Fig. 1
figure 1

Location of pathogenic mutations in BRCA1 and BRCA2 genes

Full size image

Endpoints

The primary endpoint was the incidence of febrile neutropenia at day 7–14 of the C1 in the Swiss cohort. If a febrile neutropenia occurred before the check-up at day 7–14, this was considered as an event. Secondary endpoints were the incidence at day 7–14 of C1 of grade 4 neutropenia, corresponding to an absolute neutrophil count inferior to 0.5 × 109/L [17]; grade 3–4 neutropenia and hospitalization after C1; dose reductions and G-CSF use during the entire chemotherapy regimen; and long-term toxicities (acute leukemia/myelodysplastic syndromes, cardiac dysfunction and neuropathy). Long-term toxicities were assessed in both the Swiss and French cohorts.

Statistical analyses

Proportions were calculated for categorical data, whereas median and interquartile range (IQR) were calculated for continuous data. Statistical significance for categorical data was assessed using a χ2 test or a Fisher’s exact test, as appropriate. Continuous variables were compared using the Kruskall Wallis test. Patient’s characteristics and toxicity frequencies were compared pair by pair (BRCA1 carriers vs. non-carriers; BRCA2 carriers vs. non-carriers). Missing data or inapplicable responses were excluded when calculating p values. p values < 0.05 were considered as significant. Calculations were done with STATA software (version 14.0).

Results

Patients in the Swiss cohort

We reviewed the medical records of 666 patients and 231 met our inclusion criteria. Among eligible patients, ten refused to participate and 221 patients were included (Fig. S1): 23 were BRCA1 mutation carriers, 22 BRCA2 carriers and 176 non-carriers. The median age was 40 years (IQR 35–50), comparable in the non-carriers, BRCA1 and BRCA2 carriers (Table 1).

Table 1 Clinical, pathological and treatment characteristics at diagnosis by BRCA1/BRCA2 status
Full size table

Tumor characteristics and treatment

The proportion of triple negative (estrogen and progesterone receptors and HER-2 status negative) breast cancers (TNBC) was higher among BRCA1 carriers (77%, p < 0.001). Tumor staging according to TNM and chemotherapy regimens was similar in all groups, with the majority of patients receiving two DNA damaging-agents, an anthracycline and an alkylating agent (Table 1). Very few patients received platinum derivatives. The majority of patients received 3*FEC100/3*docetaxel or 6*FEC100 regimens. None received dose-dense regimen (cycle every 14 days). No differences were seen in baseline counts of leukocytes, platelets and neutrophils between BRCA carriers and non-carriers (Table S1).

Acute hematological toxicity

Overall, BRCA1 carriers were more likely to develop acute hematological toxicity than BRCA2 carriers, or non-carriers. At day 7–14 of C1, the incidence of grade 3–4 febrile neutropenia was significantly higher among BRCA1 carriers compared to non-carriers (35% vs. 10%, p = 0.002). Grade 3–4 neutropenia (67% vs. 36%, p = 0.01) and grade 4 neutropenia (57% (12/21) vs. 18% (28/160), p < 0.001) were more frequent among BRCA1 carriers compared to non-carriers (Table 2). Unplanned hospitalizations were more frequent among BRCA1 carriers (22% vs. 8%, p = 0.043) than non-carriers. Incidence of anemia and thrombocytopenia was similar in all groups (Table S1). Chemotherapy dose-reductions were more frequent among BRCA1 carriers (14% vs. 3%, p = 0.03). G-CSF use at any time throughout the treatment was more frequent among BRCA1 carriers (86% vs. 51% for non-carriers, p = 0.005). For BRCA2 carriers, they did not show increased acute hematological toxicity when compared to non-carriers (Table 2).

Table 2 Acute chemotherapy-related hematological toxicity by BRCA1/BRCA2 status
Full size table

Among patients with TNBC, grade 3–4 febrile neutropenia (6/17, 35% vs. 6/50, 12%, p = 0.038), grade 3–4 neutropenia (11/15, 73% vs. 13/47, 28%, p = 0.003) and grade 4 neutropenia (9/15, 60% vs. 6/47, 13%, p = 0.001) were significantly more frequent in BRCA1 carriers compared to non-carriers (Table 3).

Table 3 Acute chemotherapy-related hematological toxicity in patients with triple negative breast cancer
Full size table

Location of mutations in BRCA1/BRCA2 genes and neutropenia

Compared to non-carriers, subgroup analyses by location of mutations revealed that among BRCA1 carriers, the majority of patients were likely to have grade 3–4 neutropenia (14/16, 88%; p < 0.001), but none of those having mutations located in the RING domain (0%, p = 0.165) (Table 4).

Table 4 Location of pathogenic mutations in BRCA1/BRCA2 and incidence of grade 3–4 neutropenia
Full size table

Long-term toxicities

Data from the Swiss and French cohorts were pooled to analyze long-term chemotoxicity among 898 patients (167 BRCA1, 91 BRCA2 mutation carriers and 640 non-carriers). The clinical characteristics of this cohort are described in Table S3. Median follow-up for the entire cohort was 5.8 years. We observed a myelodysplastic syndrome in one BRCA2 mutation carrier and an acute leukemia in one non-carrier. Regarding cardiovascular complications, we observed heart failure in one BRCA2 carrier and 3 non-carriers. Sequelar neuropathy was observed in 3 non-carriers.

Discussion

The data from this multicenter cohort study show a significant increase in acute hematological toxicity with chemotherapy in breast cancer patients who carry BRCA1 germline mutations. To our knowledge, this is the first study investigating the link between BRCA germline mutation status and hematological toxicity with systematic check-up at day 7–14 after the first cycle of chemotherapy. This method eliminates reporting bias present when patients consult only if they have fever after chemotherapy. We mainly focused on toxicity after the first cycle because patients who receive G-CSF as secondary prophylaxis after febrile neutropenia have significant decrease in the incidence of febrile neutropenia in the subsequent cycles of chemotherapy, as it was reported in the landmark G-CSF registration trial [18]. The evaluation of hematological toxicity after the first cycle seemed to us the least biased.

Here, we reported a significant increase in febrile neutropenia in BRCA1 mutation carriers only, but not in BRCA2 carriers. Febrile neutropenia is a frequent and life-threatening consequence of chemotherapy. Cancer patients who experience febrile neutropenia have 15% additional mortality compared to similar patients undergoing myelosuppressive chemotherapy [19]. Risk factors for febrile neutropenia are older age, performance status, comorbidities, female gender, treatment regimens, advanced disease and genetic factors such as TP53 genotype [20]. Our cohort was homogenous in terms of age, gender, extension of the disease and chemotherapy regimen.

Our study is unique in that we did not simply collect neutropenia and febrile neutropenia data based on medical records but rather used a methodology which mirrors that of G-CSF registration trials [21]. Indeed, our patients had systematic visits with the physician and blood tests including neutrophil counts on day 7–14 after their first chemotherapy cycle, allowing for a precise primary endpoint. While our Swiss cohort was multi-centered, the same follow-up protocol was used by all participating oncologists.

As neutrophils are abundant and have the shortest circulating half-life (6–8 h) of white blood cells [22], we suspected that neutropenia could be a surrogate marker for BRCA1/BRCA2 haploinsufficiency. We identified a significant increase in grade 3–4 and febrile neutropenia among BRCA1 carriers. There was a significant decrease in chemotherapy dose-intensity in BRCA1 carriers. Importantly, over the entire treatment period, BRCA1 carriers required G-CSF support in 86% of the cases compared to only 51% in non-carriers. Thus, it is very likely that BRCA1 carriers need G-CSF support to complete their (neo)adjuvant chemotherapy. Among patients who developed TNBC, we observed that BRCA1 carriers had significant increased risk to develop grade 4 neutropenia compared to non-carriers. This observation, which needs to be confirmed in larger and independent cohorts of patients receiving the same regimen of chemotherapy, would have implications for screening of BRCA1 germline mutations.

Our results are consistent with those reported by Huszno et al. who found an increased incidence in febrile neutropenia among BRCA carriers prior to the administration of the second cycle of anthracycline-based chemotherapy [14]. However, the authors did not detail the occurrence of febrile neutropenia in BRCA1 and BRCA2 separately, nor the incidence of hospitalizations or G-CSF use during chemotherapy. Additionally, they investigated Polish BRCA1/BRCA2 founder mutations only, which could induce a bias regarding the location of mutations in these genes [23].

A large, single-center, retrospective study among breast cancer patients showed no difference in dose-intensity or febrile neutropenia among BRCA1/BRCA2 carriers and non-carriers [13]. These diverging results with our study could have several explanations. First, the Dutch study considered “febrile neutropenia” only for patients who went to the hospital for fever. The patients did not have systematic counselling and neutrophil count at day 7–14, and this could lead to missing asymptomatic febrile neutropenia. Secondly, the authors did not consider the occurrence of febrile neutropenia in BRCA1 and BRCA2 separately. If we analyze BRCA1 and BRCA2 carriers together in our cohort, we would see less significant difference with non-carriers regarding the incidence of febrile neutropenia (24% in BRCA carriers vs. 10% in non-carriers, p = 0.0121).

A recent work by Kotsopoulos et al. evaluating hematological toxicity after chemotherapy in ovarian cancer, designed in a similar fashion to our study, compared BRCA carriers to non-carriers [24]. The authors analyzed hemoglobin, platelet and neutrophil counts before each cycle of chemotherapy. There was an increase in the incidence grade 3 or greater neutropenia among BRCA carriers, but no difference in dose delays, G-CSF use, anemia or thrombocytopenia rates. There was a trend toward an increase in chemotherapy dose reductions among BRCA carriers. The authors concluded that there were likely no clinical implications while treating ovarian cancer patients with BRCA mutations. There are several possible explanations for the differences observed between the results of Kotsopoulos’ study and our study: (a) in terms of methodology, Kotsopoulos et al. analyzed blood cell counts before each cycle of chemotherapy, but without intermediate blood tests 7–14 days after chemotherapy which corresponds to the nadir of neutrophils [18] and thus could have underestimated acute toxicity; (b) ovarian cancer patients generally received one DNA damage agent (platinum) whereas breast cancer patients usually received two (anthracycline and cyclophosphamide); (c) BRCA1 and BRCA2 carriers were analyzed together, not separately. It is possible that the authors would have observed significant differences between BRCA carriers and non-carriers had they analyzed BRCA1 and BRCA2 separately or looked specifically at the incidence of grade 4 neutropenia. Overall, despite these divergences, the neutropenia results mirror our own.

Correlation between BRCA germline mutations and increased hematological toxicities has been recently reported in a phase I trial investigating the safety and efficacy of combining interstrand cross-linking agent carboplatin and PARP inhibitor talazoparib [25]. In this trial, the authors had systematic blood count analysis every week allowing detection of significant increased toxicity in BRCA carriers while the number of the patients was low (7 BRCA carriers out of 24 patients in total). This observation is consistent with our results and highlights the need of systematic neutrophil count 1 and/or 2 weeks after starting treatment with DNA damaging agents in order to detect increased toxicity in BRCA carriers.

Little is known regarding the correlation between genotype and response to chemotherapy in BRCA carriers. We observed increased hematological toxicity among BRCA1 carriers compared to non-carriers, consistent with preclinical data suggesting BRCA1 haploinsufficient cells are impaired in DNA repair and hypersensitive to genotoxic stress [9]. While BRCA1 mutations generally increase response to DNA damaging agents, mutations located in the RING domain exhibit resistance to platinum drugs and PARP inhibitors in mice [26, 27]. Similarly, we observed that patients carrying BRCA1 mutations located in the RING domain had low incidence of hematological toxicity. We did not observe an increase in acute hematological toxicity in BRCA2 carriers.

Regarding long-term complications, there is a known correlation between BRCA status and secondary cancers, especially hematological malignancies [28]. Alkylating agents, G-CSF and radiotherapy further significantly increase this risk [29]. A potential mechanism for increased hematological malignancies in BRCA carriers could be clonal hematopoiesis, i.e., the expansion of one clone of blood population at a rate disproportionately greater than other clones [30, 31]. We did not observe an increase in secondary myelodysplastic syndromes or acute leukemia. While a correlation exists in much larger cohorts [32], our study of 898 patients may not have allowed us to detect a difference, also because of a limited follow-up period. For long-term cardiovascular and neurological chemotoxicity, both platinum derivatives and taxanes are potentially neurotoxic, the latter being frequently used in early breast cancer therapy. Our results are consistent with the literature [33], showing an absence of increased long-term neurotoxicity among BRCA carriers. The only 3 cases detected in our cohort were among non-carriers. However, it must be noted that long-term peripheral neuropathy is likely under-reported in medical files [34].

Our study has several limitations. It is a retrospective study with selection bias. All patients had criteria for genetic screening, meaning our population is not representative of all breast cancer patients. For instance, our patients were young with a median age of 40. While our methodology was strict, our study remains a retrospective analysis and certain data are missing. Hematological data collected at day 7–14 may not always reflect the toxicity nadir. BRCA1/BRCA2 screening by next-generation sequencing introduced in our laboratory in 2011 is more sensitive than previous techniques used (Denaturing High-Performance Liquid Chromatography or High Resolution Melt pre-screening analysis followed by Sanger sequencing method) [35], and we did not retest all negative results with more modern techniques. The chemotherapy regimen was not uniform. We did not do a multivariate analysis. The small BRCA carriers sample size of the Swiss cohort prevented us from confirming the relevance of our study. Thus, our observations should be considered as exploratory and need to be validated by a large-scale prospective cohort with uniform genetic testing method and uniform treatment.

Conclusion

Among breast cancer patients, there is a significant correlation between chemotherapy-related acute hematological toxicity, febrile neutropenia, and BRCA1 germline mutation status. Our results, which need to be confirmed in an independent prospective cohort, suggest that care should be taken and primary prophylaxis with G-CSF offered when treating women with known BRCA1 mutations with neo(adjuvant) chemotherapy.