Introduction

Breast and ovarian cancer are the most frequent cancer types affecting women worldwide (Bray et al. 2018). Beyond sporadic cancer, there is a critical population of individuals who carry mutations in some cancer-predisposing genes thus being at higher risk of develo** those malignancies. Particularly, from 5 to 10% of all breast cancer (BC) and up to 25% of all ovarian cancer patients (OC) have a genetic predisposition, constituting a medical entity named Hereditary breast and ovarian cancer syndrome (HBOC) with mainly autosomal inheritance, incomplete penetrance and variable expressivity (Balmaña et al. 2011). More than two decades ago, the BRCA1 and BRCA2 susceptibility genes were discovered and cumulative risks of develo** cancer were estimated. In female patients who carry a deleterious variant in BRCA1, the risk of develo** BC at 70 years is 60% and for OC of 59%. Regarding BRCA2, the cumulative risk for BC is 55% and for ovarian cancer is 17% at the age of 70 (Mavaddat et al. 2013). Presence of pathogenic/likely pathogenic variants in these genes makes it possible to improve the follow-up of carrier patients through two strategies: early detection using imaging tests (MRI and mammograms) or the use of preventive surgery (mastectomy or oophorectomy) to reduce the risk of cancer development (González-Santiago et al. 2019). Yet, BRCA1/2 was found mutated in only 25% of the HBOC cases, remaining the genetic predisposition of a great proportion of high-risk individuals undisclosed (Schubert et al. 2019).

In our Hospital until mid-2017, molecular diagnosis of HBOC was solely based on the identification of variants in BRCA1 and BRCA2 using Sanger sequencing and MLPA analysis. In recent years, the development of second-generation sequencing (NGS) allowed to include other genes related to a higher risk of develo** HBOC beyond BRCA1/2, improving performance and response time (Pinto et al. 2016). NGS multigene panels are being implemented in carrier screening programs (Castéra et al. 2014; Coppa et al. 2018) having both advantages and disadvantages: the benefits of multiplexing capabilities, reduction of costs, and identification of novel variants need to be balanced with the challenges of identifying an increased number of variants of uncertain significance (VUS), especially related with racial bias of current population studies (Kurian et al. 2018; Li et al. 2020).

It was recently suggested that multigene panels might benefit not only high-risk but also individuals not meeting referral criteria for testing (Beitsch et al. 2019; LaDuca et al. 2020). In addition, the identification of actionable variants in genes predisposing to HBOC could well lead to an improved implementation in preventive approaches, screening, surveillance and treatment as we move towards Precision Medicine and Cancer Interception (Blackburn 2011; Serrano et al. 2020).

Multiple prevalence studies on genetic variants in BRCA1/2 related to HBOC were carried out in different regions of Spain (Infante et al. 2006; Miramar et al. 2008; Blay et al. 2013; Juan et al. 2013; Gabaldó Barrios et al. 2017; Ruiz De Sabando et al. 2019; Pajares et al. 2018). However, the south part of Spain is slightly under-represented with only one study (Pajares et al. 2018) on BRCA1/2 which may be relevant given the epidemiological history of the country. Furthermore, none of the above-mentioned studies uses NGS to assess the variant status of other genes beyond BRCA1/2.

The main goals of this study were to evaluate our cancer registry to assess the prevalence of germline variants in cancer-predisposing genes besides BRCA1/2 in Granada’s high-risk HBOC population and to follow-up patients who had a pathogenic (class V) or likely pathogenic (class IV) variant to assess clinical impact regarding prophylactic surgery (either mastectomy, oophorectomy, or both) and periodic follow-up.

Patients and methods

Patient recruitment and study design

This retrospective cohort study includes 824 high-risk patients for hereditary breast and/or ovarian cancer (HBOC) referred to the Genetic Counselling Units at two University Hospitals in Granada. The inclusion criteria were defined according to the Spanish Society of Medical Oncology (SEOM) (Graña et al. 2011) (Supplementary table 1). Fulfilling at least one criterion was sufficient to be included in the study, independently on whether more criteria were also fulfilled. Additionally, 294 patients not fulfilling any criteria defined above (mainly due to lack of family information) were also included. The whole cohort consisted of breast cancer (n = 650) and ovarian cancer (n = 66) patients as well as high-risk cancer-free individuals (n = 108). Median age of onset for breast cancer was 45.6 ± 10.9 (range 21–90) and 49.2 ± 11.7 (range 25–70) for ovarian cancer (Table 1). This study was approved by the ethical committee and informed consent was obtained for all patients. Histological cancer subtypes were assessed at the Pathology Unit after solid biopsy followed by tissue immunostaining. Molecular subtypes for BC were referred as luminal A (ER+ and/or PR+, HER2−), luminal B (ER+ and/or PR+, HER2+), HER2 positive (ER− and PR−, HER2+) and basal (ER−, PR− and HER2−). Regarding OC, serous, endometrioid, mucinous and epithelial subtypes were characterized in 69.7% of the population. Full histological characterization of the whole cohort could not be assessed in 11.8% of the samples referred as unk (unknown). Patients were followed up from the date of testing. For patients with a negative result (either class I, II or III), no further action was taken. Patients with a positive result (class IV or V) were divided into those undergoing risk-reducing preventive surgery (mastectomy, oophorectomy or both) and those who decided only to attend specific check-ups and controls.

Table 1 Clinic-pathological and genetic characteristics of the study cohort
Full size table

Genetic testing

As genetic testing became more widely implemented within the clinical routine, a transition between targeted sequencing to multigene panel testing was done. Thus, this cohort study includes 824 individuals that were either analyzed by targeted sequencing for BRCA1/2 (438/824; 53%) or by multigene panel analysis (386/824; 47%). Copy number variation (CNV) analyses to detect small insertion/deletions on BRCA1/2 were performed by multiplex ligation-dependent probe amplification (MLPA) following the laboratory standard operating procedures. Targeted sequencing for point mutations at BRCA1/2 was performed using standard Sanger sequencing. Next-generation sequencing (NGS) was performed using a 16 multigene panel (Hereditary cancer solution by SOPHiA GENETICS): ATM, BRCA1, BRCA2, BRIP1, CDH1, CHEK2, MLH1, MSH2, MSH6, PALB2, PMS2, PTEN, RAD51C, RAD51D, STK11and TP53 in a MiSeq System (Illumina) at the University Hospitals of Granada following the manufacturer's protocol. This panel covers the coding regions and splicing junctions (± 5 bp) of the genes understudy with a high-confidence calling of SNVs, Indels and CNVs in all genes of the panel. The analysis considers a minimum coverage of ≥ 50 × and an alternative allele with coverage greater than 20. Sequencing was done in a MiSeq (Illumina Inc) and bioinformatics analysis and variants annotation was performed using the SOPHiA DDM 5.8.0.3 software (human reference genome GRCh37/hg19).

Data analyses

Variant pathogenicity and clinical classification were carried out in accordance with a five-tier system of classifications for variants of the American College of Medical Genetics and Genomics (ACMG): class V pathogenic, class IV likely pathogenic, class III variant of uncertain significance, class II likely benign and class I benign (Richards et al. 2015) using the main databases: Human Gene Mutation Database Professional (HGMD); Leiden Open Variation Database (LOVD); Universal Mutation Database (UMD); Clinical Variation Database (ClinVar); Breast Cancer Information (BCI) and Exome Aggregation Consortium (ExAC). All pathogenic or likely pathogenic variants identified were confirmed by Sanger sequencing or MLPA. For pathogenicity evaluation of the variants of uncertain significance (VUS), different in silico prediction software based on supervised-learning methods were used: 1-SIFT; Poliphen-2; and Mutation Taster (all of which are integrated into the SOPHiA DDM 5.8.0.3 software). If a pathogenic/likely pathogenic variant was identified, no subsequent VUS was reported. In our study, we did not detect more than one pathogenic/likely pathogenic variant.

Descriptive analyses include absolute and relative percentages for categorical variables, mean and standard deviation for continuous normal variables, and median and interquartile range for continuous non-normal variables. For evaluating the association between categorical variables, Chi-square test with simulated p value (simulated p value because of low expected frequencies in cells in contingency tables) was used. The selected alpha is set to 0.05 (p values equal or lower than this one are considered significant). For continuous variables, the association was evaluated using t test or Mann–Whitney–Wilcoxon test, or one way or ANOVA Kruskall–Wallis test for more than two groups, depending on normality assumptions. When more than two groups were evaluated, the p value for pairs comparisons was corrected using the Holms method for avoiding increasing type I error due to multiple test. Univariate odds ratio with 95% confidence interval were estimated to measure association. When cell frequencies were below one, Haldane–Anscombe correction was applied for allowing the estimation of the confidence interval. For the estimation of adjusted OR, logistic regression models were used. Discrimination ability of logistic models was estimated through the concordance C index and Sommer’s D. The C index takes values between 0 and 1 (the closer to one the greater discrimination capacity). Supplementary table 2 gives a detailed analysis of OR calculations.

Results

The majority of probands (650/824) had invasive breast cancer (BC) whereas only 8.0% (66/824) had ovarian cancer (OC) (Table 1). The remaining (108/824) were cancer-free individuals but referred as high-risk following the inclusion criteria (Supplementary Table 1). Most BC patients were females with similar distribution for the cancer-free cohort. Predominant histological subtype for BC was luminal A (324/650) whereas luminal B (99/650), and basal subtype were less prevalent (103/650), followed by HER2 positive (27/650). Regarding OC, the majority (31/66) had serous histological subtype followed by endometrioid disease (6/66).

Class IV/V variants were identified in 98/824 (11.9%) patients, being more frequent in OC (13/66) than in BC (78/650). Additionally, actionable pathogenic variants were detected in 8/108 unaffected patients. The most prevalent pathogenic variants, as half population was only tested for those two genes, were found in BRCA2 (47/98) and BRCA1 (24/98). In patients with breast cancer, 23.4% (18/77) of the pathogenic/likely pathogenic variants were found in BRCA1 whereas 48.1% (37/77) were located in BRCA2. Regarding ovarian cancer, pathogenic mutations were found in 38.5% (5/13) and 53.8% (7/13) for BRCA1 and BRCA2 respectively.

Furthermore, of the 51 class IV/V variants identified using the multigene panel, 25 were identified in other genes different from BRCA1/2: PALB2 (8/51), ATM (7/51), CHEK2 (4/51), MSH6 (2/51), TP53 (2/51) and RAD51C (2/51) (Fig. 1). Results from the multigene panel were compared with results obtained by Sanger sequencing to elucidate the clinical impact of using NGS panels instead of targeted sequencing. When comparing only BRCA1/2 outcomes, no significant differences were identified regarding cancer occurrence (OR 0.85 [0.21–2.85]) or variant pathogenicity (OR 1.05 [0.51–2.19]). However, when considering all genes included in the multigene panel, significant differences were observed regarding variant pathogenicity (p < 0.0001). Thus, the multigene panel was able to significantly resolve more high-risk patients (Supplementary Fig. 1.A) and this was solely dependent on the inclusion of additional genes as no significant differences for BRCA1/2 outcomes were observed between the two technologies (Supplementary Fig. 1.B). However, not only the number of patients with class IV/V variants increase when using NGS (from 10.7 to 13.2%), but also and more significantly, the number of patients with reported variants of uncertain significance (VUS) increased from 3.9 to 21.8% (for targeted and NGS respectively) (Fig. 1). All pathogenic variants identified in this study are shown in Table 2.

Fig. 1
figure 1

Pie charts describing study design and population distribution for a different type of genetic variants. Class I/II refer to benign and likely benign variants; Class III refers to variants of uncertain significance (VUS) and Class IV/V refer to likely pathogenic and pathogenic variants

Full size image
Table 2 Pathogenic variants identified in this study
Full size table

Interestingly, the age of onset of either BC or OC was significantly lower for BRCA1 mutated patients compared to those carrying either BRCA2 (p = 0.0013) or variants in other genes (p = 0.0012) (Fig. 2). In particular, the risk of presenting a pathogenic variant in the BRCA1 gene (with respect to other genes) was reduced by 9% with each unit increase in the age at diagnosis (p = 0.002), being present in 92% of patients with estrogen receptor positive (ER +) (p < 0.001), and 89% of patients with progesterone receptor positive (PR +) (p < 0.001) and this significant effect was maintained when adjusting by age.

Fig. 2
figure 2

Age of onset of the first tumor grouped by gene variant. Scatter plots showing the age of cancer onset distribution accounting for the gene carrying heterozygous germline variant. p value for Kruskal–Wallis test is **p = 0.001

Full size image

Regarding BRCA1 and BRCA2, there were six and nine pathogenic mutations respectively that were identified more than once, suggesting their frequency might be high in the study population. Predictably, some of them had previously been found to be common in Spanish populations (Table 2).

Our cohort also includes 101 VUS which pathogenicity was studied using different in-silico predictors. Novel variants (11 VUS and 12 pathogenic) all of them, except for one, frameshifts not previously described were identified in our cohort, characterized and submitted to ClinVar (Table 2; SUB6135852). Two of the novel pathogenic variants were located at BRCA2 and one at BRCA1. The remaining nine novel variants were located at ATM (4), PALB2 (1), CHEK2 (1), MSH6 (1) and RAD51C (2). All patients carrying either of these novel variants developed cancer (except for one cancer-free individual) with only one case showing ovarian while the remaining 9 cases had breast cancer.

Regarding the clinical impact of the genetic variants identified in this study, we found that the risk of having a pathogenic or VUS variant in MSH6 was reduced by 94% in patients ER + BC (p = 0.004) and 88% in patients PR + BC (p = 0.025) and this significant effect was maintained for age-adjustments. Furthermore, the risk of having a pathogenic or VUS variant in PALB2 was increased by 6% with each unit increase in the age at diagnosis (p = 0.023). The risk of develo** basal BC was increased 15.7 times in patients with BRCA1 pathogenic variants (p < 0.001) and was reduced by 81% in those with BRCA2 variants (p = 0.009). This effect was maintained when adjusting for age and also it was reproduced in multivariate logistic analyses including potentially confounding factors. The risk of develo** HER2 positive BC was increased 37.5 times in individuals with MSH6 pathogenic variants compared with variants in other genes (p = 0.031) and the risk of develo** luminal A BC was reduced by 77% (p = 0.089) in patients with BRCA1 pathogenic variants and this was maintained when adjusting by age. None of the pathogenic variants was significantly associated with luminal B BC. All these analyses were replicated including only women and no significant differences were observed (data not shown).

Considering clinical evolution, 6.1% (6/98) with a class IV/V variant died before the follow-up so they were excluded from this evaluation. Of them, three died due to cerebral metastases (two with BC before 40 and one with OC at 47) and other metastases for the remaining three.

More than a half of the patients (49/92) with a pathogenic or likely pathogenic variant underwent risk-reducing prophylactic surgery. We observed significant differences in the effect of the age at which cancer was diagnosed and the prophylactic clinical decisions undertaken (p = 0.008), with younger ages deciding to accept a mastectomy alone (median age 38.0) or in combination with oophorectomy (median age 35.5) compared to clinical follow-up alone (median age 44.5). The same effect was maintained if accounting for the age at which the genetic testing was performed with median ages of 43.1 and 39.9 for those undertaking mastectomy alone or in combination with oophorectomy respectively in comparison with those not deciding to undertake prophylaxis with median age of 55.0 (p < 0.001). However, no differences were observed among the age at which the prevention measures were taken (p = 0.342). Specifically, 47.0% underwent prophylactic oophorectomy, 36.7% underwent a prophylactic mastectomy, and 16.3% underwent both surgical procedures. Most of them carried a variant in BRCA2 (27/49), followed by BRCA1 (14/49). In addition, 8 patients opted for surgery carrying class IV/V variants in genes different from BRCA1 or BRCA2, specifically 4 patients with PALB2 variants (two with c.2964delA, and one with c.2257C>T or c.1535dupA) and 4 patients with a variant in either ATM (c.8122G>A), CHEK2 (c.349A>G), RAD51C (c.104dupC) or MSH6 (c.738_741delAAAA) respectively. Interestingly, only 14.3% of the patients carrying a pathogenic/likely pathogenic variant in ATM opted for prophylaxis (95% IC [1.97–58.1%] (p = 0.042).

Discussion

Analysis of additional genes besides BRCA1/2 determines the originality of this work, increasing the proportion of class IV/V variants identified in our cohort compared with previous studies in our region (Pajares et al. 2018). Detection rate of these variants was greater than the recommended 10% for hereditary cancer (Graña et al. 2011) highlighting efficiency of both cohort selection and genetic testing technology.

Frequency of pathogenic variants identified in this study (11.9%) is within the range of previous reports (Gabaldó Barrios et al. 2017; Pajares et al. 2018; del Manzanares Campillo et al. 2018; Blanco et al. 2013; Díez et al. 2003) although mutation frequencies for BRCA1/2 in Spain were reported to vary from 6.8 (Beristain et al. 2007) to 33.3% (Miramar et al. 2008). Discrepancies among studies might be due to several reasons, being the inclusion criteria one of the main sources. For example, one of the studies with greater mutation frequencies for BRCA1/2 (33.3%) included not only index patients but also their relatives, increasing the chances of detecting variants if we take into account that all these patients belonged to families with several breast/ovarian cancers (Miramar et al. 2008). Contrarily, a study showing very low mutation frequencies for BRCA1/2 included only unrelated index cases with or without a family history of breast and/or ovarian cancer (Beristain et al. 2007) reducing the likelihood of detecting variants in these genes. Also, founder mutations might affect differential mutation frequencies across regions as shown previously (Díez et al. 2003) and population size also impacts the ability of detecting cancer-predisposing mutations.

We identified higher mutation frequencies in BRCA2 (48.0%) compared to BRCA1 (24.5%) in line with recent studies (Pajares et al. 2018; del Manzanares Campillo et al. 2018); however, other authors found greater frequencies in BRCA1 vs BRCA2 (Gabaldó Barrios et al. 2017; Díez et al. 2003). This differential prevalence might be affected by study population size, inclusion criteria or even region-bias (Fig. 3).

Fig. 3
figure 3

Graphical representation of the previous genetic testing studies in hereditary breast and ovarian cancer patients in Spain by regions. Number of individuals included in each cohort is shown with an "n". The number of pathogenic/likely pathogenic variants identified is also shown splitting by those studies including only BRCA1/2 or other genes and the number of novel variants identified is also shown for comparison with our study

Full size image

Furthermore, contrary to most studies in our region, we analyzed not only BRCA1/2 prevalence but also other HBOC risk-related genes. In our cohort, PALB2 was the most commonly mutated gene (2.1%) after BRCA1/2 validating previously reported frequencies (≈ 1%) (2013; Thompson et al. 2016), with two variants (c.1653T>A and c.2964delA) being found in three times each. Frequency of pathogenic variants in ATM (1.8%) also agreed with previously reported frequencies for Spanish populations (2%). However, frequencies of pathogenic variants in CHEK2 (1.6%) and TP53 (0.5%) in our cohort were slightly lower than previously reported (Walsh et al. 2006; Cipriano et al. 2018).

Data from NGS highlighted a reduction in the number of low-risk reported individuals (based on the presence of class I/II variants) from 85.4 to 65% (for targeted and NGS respectively) suggesting that the inclusion of additional genes will improve the detection of high-risk individuals. Importantly, this reduction came with an increase in VUS that need to be re-classified on a periodic basis. Importantly, 27 individuals in this cohort would have been misclassified as low-risk if a gene panel had not been carried out. In fact, it was recently demonstrated that a significant proportion of BC patients with germline variants do not meet classical NCCN testing criteria, suggesting that NGS might be used in the near future for screening less targeted populations as the costs fall.

Prophylactic mastectomy (bilateral and contralateral) is one of the most widely used options to reduce cancer risk in women who are carriers of a pathogenic/likely pathogenic variant in risk genes, being able to reduce more than 85% of the incidence of breast cancer. Oophorectomy is known as the surgical procedure of removing adnexal organs (ovary and fallopian tubes) unilaterally or bilaterally in women who are carriers of pathogenic BRCA1 and BRCA2 variants. Most of the published articles refer to risk-reducing surgeries based on pathogenic variants in these two genes, however, in our cohort, 8 of the women who underwent risk-reducing surgery carried a variant in genes such as PALB2, ATM, CHEK2, ATM and MSH6. Even though there are no many guidelines on prophylactic surgery beyond BRCA1 and BRCA2, the number of oncologists and gynecologists recognizing the risk of develo** HBOC associated with other high or intermediate penetrance genes is rising. Recently, the ACMG published a guideline for patients with germline variants in PALB2, in which risk-reducing mastectomy is an option to be considered to reduce BC risk as well as to include surveillance for pancreatic cancer; however, oophorectomy is not recommended for patients below 50 years. In our cohort, 3 out of 8 patients with class IV/V variants in this gene undertook prophylactic mastectomy as recommended by this recent guideline. Furthermore, we found that younger women at diagnosis (or genetic testing) were more likely to undergo prophylactic measures although the likelihood of selecting these preventive measures was reduced by 87% in patients with a pathogenic/likely pathogenic variant in ATM.

In addition, it is important to take into account other factors such as women's age at the time of clinical diagnosis of cancer and age at the time of the genetic diagnosis. This is of great importance since we demonstrated that women at younger ages preferably chose to undergo surgery. On the other hand, the family cancer history should also be considered as well as the psychological aspects such as fear of cancer recurrence or its association with the development of other cancers depending on the mutated gene (pancreas, endometrium, Cowden syndrome, stomach, etc.). To our knowledge, this is the first time that NGS including non-BRCA genes was performed in an Andalusian cohort to assess germline variants predisposing to HBOC. Our data agrees with previous studies for targeted BRCA1/2 variants in high-risk HBOC individuals and reduces the number of class I/II reported variants validating the use of NGS to increase the likelihood of variant identification.