Frequency of pathogenic germline variants in BRCA1, BRCA2, PALB2, CHEK2 and TP53 in ductal carcinoma in situ diagnosed in women under the age of 50 years

Ductal carcinoma in situ (DCIS) is considered a non-obligate precursor of invasive breast cancer of ductal/no special type (IDC) as many IDCs have evidence of associated DCIS at presentation [1, 2] and the two components have similar genetic changes, suggesting that in the majority of cases the invasive component has arisen from the DCIS [3]. Synchronous DCIS is more frequent in luminal and HER2-positive IDC (53% and 63%, respectively) than invasive basal breast cancer (33%) [4]. Since the introduction of screening mammography, there has been an increase in the reported incidence of pure DCIS with no invasive component [5], with about 20% of screen-detected tumours being pure DCIS [6].

Most non-genetic risk factors for breast cancer have similar associations with DCIS and IDC, again supporting the notion that DCIS is a precursor of invasive cancer [7, 8]. Epidemiological studies have shown there is an inherited predisposition to DCIS, with women with DCIS being 2.4 times more likely to have an affected mother and sister with breast cancer than controls [9]. One study of almost 40,000 women suggested that the familial relative risk of DCIS may be greater than that of invasive breast cancer [10], but this was not confirmed in the Million Women Study, which showed a similar association with family history for DCIS and IDC [8].

The familial risk associated with invasive breast cancer is in part explained by both high-risk rare variants and low-risk susceptibility loci. We have shown that the majority of low-risk invasive breast cancer predisposition loci also predispose to DCIS and, as for invasive disease, different loci predispose to ER-positive and ER-negative DCIS and high and low grade DCIS [11]. However, the frequency of pathogenic high- and moderate-risk variants in DCIS is not clear. Claus et al. studied 369 women (mean age 53.8 years) with pure DCIS selected from a case-control study of carcinoma in situ and found that 2.4% had pathogenic variants in BRCA2 and 0.8% in BRCA1 [12]. Hall et al. analysed a highly selected cohort of women with carcinoma in situ (LCIS and DCIS) that were referred to Myriad for genetic testing. They found that 5.2% of women with pure carcinoma in situ (CIS) had BRCA1/2 mutations (2.3% if women with a family history of breast cancer were excluded) and like Claus et al. that BRCA2 mutations were more common than BRCA1 [13]. Both these studies were performed before gene panel genetic testing was available, and therefore, other breast cancer predisposition genes such as PALB2, CHEK2 and TP53 were not assessed.

In this study, we report the frequency of rare variants in five known breast cancer predisposition genes (BRCA2, BRCA1, TP53, CHEK2 and PALB2) in 655 cases of pure DCIS with no invasive disease in women diagnosed before the age of 50. These cases were included in our previous study of low-risk susceptibility loci in DCIS [11].

The analysis was performed on 655 cases of pure DCIS with no invasive disease diagnosed in women aged under 50, together with 1611 controls. The median age of cases was 45 years (interquartile range 6) and of controls was 52 (interquartile range 12). Data on grade and oestrogen receptor (ER) status were available for 98% and 73% of the cases in the study, respectively, Table 1.

The mean coverage of our target region was 800 reads across all samples, with an average of at least 40 reads for more than 98% of the target region per sample. Of the 321 amplicons analysed, seven failed to amplify consistently across the majority of the samples; however, even for these seven the majority of samples had at least 10 reads for 90% of the amplicon, and there was no difference in amplification between cases and controls, Additional file 4. No exon-level CNVs were identified using CNVkit version 0.9.5. Thirteen copy number variants across eight samples were detected by ONCOCNV; however, none were confirmed using MLPA, Additional file 5.

We found an association with DCIS and pathogenic variants in BRCA2, CHEK2, PALB2, BRCA1 and TP53, Table 2 (individual raw data, Additional file 2).

Early data on women with BRCA1 and BRCA2 mutations suggested that presentation as pure DCIS was infrequent. However, Yang et al. showed that ~ 60% of BRCA1- and BRCA2-associated invasive tumours had associated DCIS of similar phenotype [23]. They also found that the number of pure DCIS cases was similar in BRCA1 and BRCA2 mutations carriers (21% vs 23%), in contrast to Krammer et al. who reported that pure DCIS was more frequent in BRCA2 mutation carriers compared to BRCA1 carriers (5%, 36/246, versus 9%, 23/250, P = 0.0026) [24]. In our study of sporadic pure DCIS, pathogenic BRCA2 variants were far more common than BRCA1 mutations (3.5% vs 0.6%). This is similar to the data of Claus et al. [12] who found that 2.4% had pathogenic variants in BRCA2 and 0.8% in BRCA1 in a slightly older group of DCIS patients (mean age 53.8 years). Similarly, Hall et al. found that 5.2% of women under 50 years of age with carcinoma in situ (LCIS and DCIS) had BRCA1/2 mutations, with BRCA2 mutations being more common than BRCA1 [13].

We found that BRCA2 mutations occurred in 2.4% of DCIS in women under the age 50 and in 9% under the age of 40. All but one of these variants had been previously described in invasive breast cancer. Of the genes studied, BRCA2 was the only gene where pathogenic mutations were associated with younger age. All the cases of DCIS in BRCA2 carriers were ER positive (where ER status was known), unlike invasive disease where only 77% are ER positive [25]. In contrast, BRCA1 pathogenic variants were infrequent, four in total (only one had been previously described), and associated with predominantly ER negative DCIS.

Pathogenic variants in CHEK2 were the second most common set of mutations after BRCA2 and occurred in 2.5% of pure DCIS under the age of 50. Unlike BRCA2, there was no association with age and the majority were the well-described c.1100delC variant. There was no evidence of an association with the rare missense variant p.I157T (c.T470C, rs17879961), which was found in three controls and no cases. This high frequency of CHEK2 variants in pure DCIS has not been previously described, although Schmidt et al. noted in their study of tumour characteristics in CHEK2 c.1100delC carriers that carriers from population- and hospital-based studies more often developed in situ tumours (LCIS and DCIS) compared to carriers from familial or clinical genetics center–based studies; this was interpreted as a bias estimate due to differential recruitment related to family history of breast cancer and screening [26]. Our findings are also supported by Couch et al. who reported data on the frequency of CHEK2 mutations in a series of breast cancer with and without pure DCIS allowing one to determine mutation rates in pure DCIS cases. In that study, 2.87% of DCIS cases had pathogenic CHEK2 variants compared to 1.43% in invasive disease [27].

The tumour phenotypes associated with PALB2 tumours are very similar to those associated with BRCA2 tumours, with 61% of invasive tumours having associated DCIS [28]. It is therefore not surprising that we have found PALB2 pathogenic mutations in women with pure DCIS and that they are more common in women with a first-degree relative with breast cancer. However, unlike BRCA2, they were not associated with age < 40 years, this is supported by the findings of Antoniou et al. who showed there was a constant relative risk, irrespective of age, for pathogenic variants in PALB2 [29]. The frequency of PALB2 variants in our data is supported by the study by Couch et al. where one can infer from the supplementary data that the frequency of pathogenic PALB2 variants in DCIS is 0.96% [27].

Pure DCIS (73% HER2 positive, 55% ER positive) and high-grade comedo DCIS have been described in Li-Fraumeni Syndrome [30, 31] but our data show that TP53 mutations are infrequent in sporadic DCIS. Although two of three pathogenic variants identified had been previously described, none of the women had a family history of cancer to suggest Li-Fraumeni Syndrome (LFS).This may be because they are de novo mutations or because two were loss of function mutations which often do not have such a typical LFS phenotype as dominant negative missense mutations [32]. The known missense mutation detected has been shown to be associated with Li-Fraumeni–like syndrome rather than true LFS [22].

The odds ratios presented in this study are higher than those previously reported for these genes in invasive disease. This is likely due to the size of the study which is too small to yield stable estimates of associations with DCIS, but does give useful estimates of prevalence and, of note, is twice as large as the study by Claus et al., the only other study documenting the prevalence of BRCA1/2 mutations in sporadic DCIS. The other reason is due to the use of older controls. When we compare our cases to 21,384 non-Finnish European controls from gnomAD (gnomAD controls v2.1, http://​gnomad.​broadinstitute.​org), we see that the odds ratios fall to similar levels previously reported for invasive cancer with the exception of CHEK2 (OR = 3.69, 95%CI.19–6.23) which still remains higher than that reported in studies of invasive breast cancer (OR~ 2), Table 2. Schmidt et al. also found this in their large series from the Breast Cancer Association Consortium (invasive: OR = 2.4, 95%CI 2.04–2.82, in situ: OR = 3.53, 95%CI 2.38–5.23) [26]). We also found a similar finding in lobular cancer where LCIS had a stronger association with CHEK2 mutations than ILC (ILC OR = 4.29, 95%CI 1.60–11.51, P = 0.0017; LCIS OR = 9.95, 95%CI 3.44–28.82, P = 5 × 10⁻⁵, Petridis et al. accepted Cancer Epidemiology, Biomarkers & Prevention). This suggests that CHEK2 pathogenic variants may be predisposing to the in situ stage of breast cancer with some not progressing to the invasive state.

In a study of 6478 patients with invasive breast cancer under the age of 50, Schmidt et al. [33] found a higher frequency of BRCA1 mutations compared to our study of DCIS (3.2% versus 0.6%). We believe that this difference in frequency of BRCA1 mutations stems from the fact that the vast majority of the samples in our study are ER+ and only 13% of the samples are ER−, compared to 25% in the invasive study of Schmidt et al.

There is currently a debate as to the need for mutation screening in women with DCIS. In this study, 7.2% of women with DCIS (irrespective of ER status) under the age of 50 had pathogenic variants in one of five known breast cancer predisposition genes. This level does not reach the current UK threshold for genetic testing (https://​www.​nice.​org.​uk/​guidance/​cg164/​chapter/​Recommendations#genetic-testing); however, women under 40 years of age had a 13% (11% excluding CHEK2 variants) probability of having a germline mutation which does reach the UK threshold of 10% for routine testing. For women under 40 years of age with a family history of breast cancer, the frequency of germline mutations increases to 21%.

There has been particular focus on ER-negative DCIS and whether these women should undergo HER2 testing and, if this is also negative, be offered BRCA1 and BRCA2 testing, as is recommended for those with triple-negative invasive breast cancer. Unfortunately, in our series, we did not have data on HER2 receptor status as it is not routinely assessed in cases of DCIS in present clinical practice. However, looking solely at ER-negative DCIS, only 9% under the age of 50 and 8% under the age of 40 had pathogenic variants and these were in BRCA1, TP53 and CHEK2. These figures rise to 16% and 11%, respectively, if only women with a family history of breast cancer are considered, Table 5. In contrast, the frequency of pathogenic variants in ER-positive DCIS under the age of 40 was much higher; 9% of women under the age of 50 had pathogenic variants rising to 29% under the age of 40 (14% and 42%, respectively, if only women with a family history of breast cancer are considered) Table 6.

This study has shown that a DCIS-associated malignant pathway can occur in patients who have pathogenic variants in BRCA2, CHEK2, PALB2, BRCA1 and TP53. We also show that the focus of genetic testing should be on ER-positive, intermediate-, and high-grade DCIS from patients under the age of 40, rather than ER-negative DCIS, although restricting such testing to those age under 40 would fail to identify the majority of CHEK2 and PALB2 mutation carriers. Once mutations are identified in these women, chemoprevention with tamoxifen and surveillance is a potential alternative to prophylactic mastectomy, particularly in CHEK2 carriers where the risk of invasive disease is less. Further studies with long-term follow-up data are required to ascertain whether these germline pathogenic variants identify a subgroup of DCIS that are more likely to progress to invasive disease or whether somatic changes in the DCIS are a more important predictor of recurrence.