Genetic predisposition to ductal carcinoma in situ of the breast

Background Ductal carcinoma in situ (DCIS) is a non-invasive form of breast cancer. It is often associated with invasive ductal carcinoma (IDC), and is considered to be a non-obligate precursor of IDC. It is not clear to what extent these two forms of cancer share low-risk susceptibility loci, or whether there are differences in the strength of association for shared loci. Methods To identify genetic polymorphisms that predispose to DCIS, we pooled data from 38 studies comprising 5,067 cases of DCIS, 24,584 cases of IDC and 37,467 controls, all genotyped using the iCOGS chip. Results Most (67 %) of the 76 known breast cancer predisposition loci showed an association with DCIS in the same direction as previously reported for invasive breast cancer. Case-only analysis showed no evidence for differences between associations for IDC and DCIS after considering multiple testing. Analysis by estrogen receptor (ER) status confirmed that loci associated with ER positive IDC were also associated with ER positive DCIS. Analysis of DCIS by grade suggested that two independent SNPs at 11q13.3 near CCND1 were specific to low/intermediate grade DCIS (rs75915166, rs554219). These associations with grade remained after adjusting for ER status and were also found in IDC. We found no novel DCIS-specific loci at a genome wide significance level of P < 5.0x10-8. Conclusion In conclusion, this study provides the strongest evidence to date of a shared genetic susceptibility for IDC and DCIS. Studies with larger numbers of DCIS are needed to determine if IDC or DCIS specific loci exist. Electronic supplementary material The online version of this article (doi:10.1186/s13058-016-0675-7) contains supplementary material, which is available to authorized users.

(Continued from previous page) were specific to low/intermediate grade DCIS (rs75915166, rs554219). These associations with grade remained after adjusting for ER status and were also found in IDC. We found no novel DCIS-specific loci at a genome wide significance level of P < 5.0x10 -8 . Conclusion: In conclusion, this study provides the strongest evidence to date of a shared genetic susceptibility for IDC and DCIS. Studies with larger numbers of DCIS are needed to determine if IDC or DCIS specific loci exist.
Keywords: Ductal carcinoma in situ, Association study, Genetic predisposition, Common variants Background Ductal carcinoma in situ (DCIS) is a non-obligate precursor of invasive breast cancer including invasive ductal/no special type carcinomas (IDC). Since the introduction of screening mammography there has been a 7-fold increase in reported DCIS incidence in the USA, primarily in postmenopausal women [1], with about 20 % of screen-detected tumors being DCIS [2]. Approximately 45-78 % of all invasive breast cancers are associated with DCIS [3,4]. It is hypothesized in the majority of these cases that the invasive component has arisen from the DCIS as they generally share the same somatic genetic changes. The proportion of IDC associated with DCIS varies depending on subtype, with luminal and human epidermal growth factor receptor 2 (HER2)-positive IDC having more frequent DCIS (53 % and 63 %, respectively) than invasive basal breast cancers (33 %) [5].
As most DCIS is treated surgically, the natural progression of untreated DCIS is not known. However, in one small study of patients with predominantly lowgrade DCIS misdiagnosed as benign breast disease and who received no surgical intervention, 6 out of 13 patients developed ipsilateral invasive carcinoma with mean time to the development of invasive carcinoma being 9.0 years [6]. In two specific DCIS trials in which DCIS was treated with breast-conserving surgery alone with no radiotherapy, long-term follow up shows that up to 30 % of women develop a recurrence (half of which will be DCIS and half invasive cancer) by 10 years [7].
Methods for accurately predicting the behavior of DCIS are poor [8]. Although grade has not been shown to be a good predictor of recurrence many clinicians use this classification to determine the use of radiotherapy following breast-conserving surgery. There is a strong correlation between the grade of the in situ and coexisting invasive components in IDC, suggesting that DCIS does not progress from low through to high grade before becoming invasive [9,10].
Most non-genetic risk factors for breast cancer have similar associations with DCIS and IDC, supporting the notion that DCIS is a precursor of invasive cancer [11,12]. There is also evidence from epidemiological studies that there is an inherited predisposition to DCIS. Women with DCIS have been shown to be 2.4 times (95 % CI 0. 8, 7.2) more likely to have an affected mother and sister with breast cancer than controls [13]. Furthermore, there is evidence from a study of almost 40,000 women that the familial relative risk of DCIS is greater than that of invasive breast cancer. For women aged 30-49 years with a family history of breast cancer the odds ratio (OR) for developing DCIS was calculated as 2.4 (95 % CI 1.1, 4.9) compared to 1.7 (95 % CI 0.9, 3.4) for invasive cancer. For women aged 50 years and above the risks were slightly reduced, but still higher for DCIS (OR = 2.2, 95 % CI 1.0, 4.2) than invasive disease (OR = 1.5, 95 % CI 1.0, 2.2) [14]. However, this was not confirmed in the Million Women Study, in which the association with family history was similar for DCIS and IDC [12].
A small part of this inherited predisposition is explained by BRCA1/2 mutations, as mutations in these genes are found in a similar proportion of DCIS and invasive breast cancer cases [15]. For low-risk common breast cancer predisposition alleles most of the initial breast cancer association studies have not been powered to identify associations with DCIS, so it is not clear whether all the low-risk susceptibility loci that have been identified are associated with DCIS and what the strength of any associations are.
It is now evident that some low-risk susceptibility loci are associated with different pathological subtypes of breast cancer and support the hypothesis that breast tumor subtypes arise through distinct molecular pathways [16][17][18]. In order to identify further low-risk susceptibility loci, it will be necessary to look at specific morphological subtypes including DCIS and the cytonuclear grade and estrogen receptor (ER) status of the disease. In this study we analyzed 3,078 cases of pure DCIS collected through the ICICLE study (a study to Investigate the genetics of In situ Carcinoma of the ductaL subtype) and performed a meta-analysis with 2,352 in situ cases collected through the Breast Cancer Association Consortium (BCAC). Our aims were to assess whether any of the known low-risk breast susceptibility alleles have different associations for DCIS and IDC, and to identify if there are any DCIS-specific low-risk alleles.

Ethics statement
All studies were performed with ethical committee approval (listed in acknowledgements) and subjects participated in the studies after providing informed consent.

Study populations
Cases came from ICICLE (MREC 08/H0502/4), a UK study of DCIS, and from 37 studies forming part of the BCAC included in the Collaborative Oncological Gene-Environment Study (COGS) [19] (Additional file 1). The ICICLE study recruited patients from participating centers throughout the UK with the aim of identifying predisposition genes for DCIS. Patients aged 60 years or less at the time of diagnosis, with a current or past history of DCIS (without invasive disease of any histological subtype) were eligible. A total of 3,078 subjects were recruited following identification from local pathology reports in 97 UK hospitals. All cases were genotyped with the iCOGS chip and compared to 5,000 UK controls selected from four UK studies (BBCS 1,231 controls, SBCS 704 controls, UKBGS 370 controls, SEARCH 2,695 controls) participating in BCAC (Additional file 2) and already typed on the iCOGS chip. Controls were randomly selected prior to analysis, and were excluded from casecontrol comparisons with BCAC cases from the originating study. After excluding individuals based on genotyping quality (see subsection "Genotyping and analysis") and non-European ancestry, data for the ICICLE study available for analysis included 2,715 subjects with DCIS (cases) and 4,813 controls.
Women with all types of breast cancer were recruited into the BCAC studies. Pathological information in BCAC was collected in the individual studies but was also combined and checked through standardized data control in a central database. A total of 2,352 subjects with DCIS were identified in the central BCAC pathology database (see Additional file 3 for number of cases by study). Controls came from the 37 BCAC studies (37,654 in total).

Genotyping and analysis
After DNA extraction from peripheral blood, ICICLE samples were genotyped on the iCOGS custom Illumina iSelect array (Illumina, San Diego, CA), which contains 211,155 single nucleotide polymorphisms (SNPs), at King's College London. The remaining cases and controls were genotyped as part of the COGS project described in detail elsewhere [19]. The ICICLE cases were analyzed using the same quality control (QC) criteria as the COGS project. Briefly, genotypes were called using Illumina's proprietary GenCall algorithm and 10,000 SNPs were manually inspected to verify the algorithm calls. Individuals were excluded if genotypically non-European or not female, or had an overall call rate <95 %. SNPs were excluded with a Gen-Train score <0.4, call rate <95 % (call rate <99 % if minor allele frequency (MAF) was <0.1) and Hardy Weinberg equilibrium (HWE) value of P <10 -7 or evidence of poor clustering on inspection of cluster plots. All SNPs with MAF <0.01 were excluded. A cryptic relatedness analysis of the whole dataset was performed using 46,789 uncorrelated SNPs and led to the exclusion of 28 cases and 18 controls due to relatedness between the ICICLE and BCAC samples (PIHAT >0.1875).
For ICICLE cases and controls, principal component analysis (PCA) was carried out on a subset of 46,789 uncorrelated SNPs and individuals or groups distinct from the main cluster (327 cases and 164 controls) were excluded using the first five principal components (PCs) (Additional file 4). Following removal of outliers, the PCA was repeated and the first five PCs were included as covariates in the analysis.
The adequacy of the case-control matching was evaluated using quantile-quantile plots of test statistics and the inflation factor (λ) calculated using 37,289 uncorrelated SNPs that were not selected by BCAC and were not within one of the four common fine-mapping regions, to minimize selection for SNPs associated with breast cancer (Additional file 5). As the majority of the SNPs on the iCOGS array are associated with breast, ovarian or prostate cancer, the SNPs selected for this analysis were taken from the set of prostate cancer SNPs, with the assumption that these SNPs were more likely to be representative of common SNPs in terms of population structure in our study.
For each SNP, we estimated a per-allele OR and reported corresponding 95 % CI using logistic regression analysis, including the five PCs as covariates, using PLINK v1.07 (http://pngu.mgh.harvard.edu/~purcell/plink/). Genotyping and analysis of BCAC studies have been described in detail elsewhere [19]. In brief, data were analyzed using the Genotype Library and Utilities (GLU) package to estimate per-allele ORs for each SNP using unconditional logistic regression. All analyses were performed in subjects of European ancestry (determined by PC analyses) and adjusted for study and seven principal components.
Case-control ORs for DCIS cases vs controls from BCAC and ICICLE were combined using inverse varianceweighted fixed-effects meta-analysis, as implemented in METAL [20]. Case-only analyses were also carried out to compare genotype frequencies for (1) ER-positive (ER+) vs ER-negative (ER-) DCIS, (2) high grade DCIS vs low and intermediate grade DCIS, and (3) DCIS vs IDC (see Additional file 3 for number of cases by study), (4) DCIS diagnosis in patients <50 years of age vs DCIS diagnosis in patients ≥50 years, and were used as a test for heterogeneity of ORs by tumor subtype/age (see Additional file 6 for number of cases by group). Only studies with data on both subtypes contributed to case-only analysis comparing these subtypes. Similar case-only analyses were performed for the IDC cases in these studies to assess whether any heterogeneity evident in DCIS also occurred in IDC.
Novel SNPs showing the strongest evidence of association with DCIS (P <6 × 10 -6 ) in the meta-analysis (after excluding previously reported loci) were genotyped in a phase II analysis at LGC Genomics (LGC, Teddington, UK). The phase II samples consisted of 653 DCIS cases from the ICICLE and Breakthrough Generation Studies and 1,882 controls from the ICICLE study not previously genotyped on the iCOGS chip. All individuals included in the analysis were of European ancestry (self-reported).
For the known breast cancer predisposition loci P <0.00066 was considered statistically significant (with Bonferroni correction for multiple testing on 76 known loci). All of the known breast cancer susceptibility loci were included in the iCOGS chip with the exception of rs2284378 (20q11), which was identified as an ERbreast cancer predisposition SNP after the iCOGS chip was developed [21].

Assessment of grade and ER status
For the ICICLE study, information on cytonuclear grade of DCIS was available for 2,578 cases, mostly from the local histopathology reports. In 200 cases where the grade data were missing from the report but the tumor block was available, an H&E section was cut and the DCIS was graded by the study histopathologist (SEP) according to UK and College of American Pathologists guidelines [22]. Data on grade of DCIS were available from histopathology reports for 828 BCAC cases.
A subset of 81 ICICLE cases, graded in the pathology report and with a tumor block available, were examined to assess the reliability of the cytonuclear grade provided by the pathology reports. In the majority of cases (86.5 %) grade was concordant with the pathology report. Nine cases were re-graded as low/intermediate grade and two cases as high grade. As the study pathologist re-graded the samples on a single H&E section, rather than all the blocks from an individual case, and in some cases on reexcision specimens with residual disease rather than the original excision specimen, the grade reported in the pathology report, if available, was used for the purposes of this study.
ER status from local histopathology reports was available for 1,086 ICICLE cases. For the remaining 781 ICICLE cases where the tumor block was available, immunohistochemistry was performed on 3-μM sections, which were incubated at 60°C for 1 h prior to automated staining using the VENTANA®. Estrogen receptor staining was carried out using CONFIRM™ anti-estrogen receptor (SP1) rabbit monoclonal primary antibody (Catalog number 790-4324) with no variation to the recommended protocol. ER staining was scored by three independent reviewers (CP, VS, DLe) using the Allred method, and any discrepancies were reviewed by the study histopathologist (SEP). DCIS with an Allred score ≥3 was considered ER+ and DCIS with scores of 0-2 (approximately equivalent to <1 % of nuclei) was regarded as ER-. ER status was available on 965 cases from BCAC (Additional file 6).

Assessment of known breast cancer susceptibility loci for association with DCIS
For the majority of known loci (n = 46) the risk allele for invasive breast cancer is the minor allele. For the ORs presented here the reference allele was set as the nonrisk allele to make it clear whether the association with DCIS was in the same direction as previously published for invasive breast cancer. Thus, ORs for DCIS will be >1 if in the same direction as invasive disease and <1 if in the opposite direction.
The case-only analysis (DCIS vs IDC) confirmed the shared genetic susceptibility between DCIS and IDC as none of the heterogeneity P values (P-Het) were significant after Bonferroni adjustment for 76 SNPs (Additional file 7). The case-only analysis (DCIS diagnosed at <50 years vs ≥50 years of age) revealed one SNP (rs527616, 18q11.2) that was significantly associated with DCIS in younger women (P-Het <50/≥50 = 0.0003) even though the overall P value for DCIS was not statistically significant after Bonferroni correction (OR 1.05, 95 % CI 1.01, 1.11; P = 0.020) (Additional file 8).

Assessment of known breast cancer susceptibility loci for association with DCIS by ER status
Following immunohistochemistry for ER in the ICICLE study samples, 1,484 cases (54 %) were classified as ER+ and 383 (14 %) as ER-. The ER data on BCAC DCIS were less complete with 664 (28 %) ER+, 301 (13 %) ER-and 1,387 cases (59 %) of unknown ER status (Additional file 6). Analysis by ER status confirmed that loci associated with ER+ IDC were also associated with ER+ DCIS (Fig. 2 and Additional file 9). These similarities were less clear for ER-DCIS and ER-IDC but this may be due to small numbers of ER-DCIS cases. A caseonly analysis of ER+ vs ER-DCIS was not performed due to the small numbers of ER-cases. Analysis of DCIS by grade revealed that although the majority of SNPs predispose to all grades of DCIS, some are grade-specific (Additional files 10 and 11). The two SNPs close to CCND1 were strongly associated with low/intermediate grade DCIS (rs75915166, OR 1.36, 95 % CI 1.17, 1.59; P = 7.2 × 10 -5 ; rs554219, OR 1.32, 95 % CI 1.18, 1.48; P = 8.2 × 10 -7 ) and there was no association with high grade DCIS (Table 2). Case-only analysis confirmed that these loci were low/intermediate grade-specific (rs75915166, P-Het low/highgrade = 0.00014; rs554219, P-Het low/highgrade = 0.00013) and this was independent of ER status (adjusted for ER status rs75915166, P = 0.0050; rs554219, P = 0.019).
rs10941679, 5p12 were borderline associated with low/intermediate grade DCIS (OR 1.26, P = 2.1 × 10 -7 , P-Het low/highgrade = 0.0033). This locus has previously been shown to be associated with low grade progesterone receptor (PR) + IDC [23]. There was no evidence of any high grade DCIS specific loci (Additional file 11).

Search for new DCIS predisposition loci
All SNPs that were genome-wide significant (P <5 × 10 -8 ) in the meta-analysis were correlated with one of the known breast cancer predisposition loci. There were three SNPs that were not correlated with known loci at P <6 × 10 -6 ( Table 3), all with very little evidence of an association with IDC.

Discussion
This study provides the strongest evidence to date for a shared genetic susceptibility between DCIS and IDC, based on 5,067 cases with pure DCIS (no invasive disease) and 24,670 cases with IDC. It differs from previous BCAC analyses of DCIS, as it has included an additional 3,078 DCIS cases, excluded all cases of pure LCIS and has also compared DCIS to IDC rather than all invasive disease.
An important finding of this study is the lack of DCIS/ IDC-specific loci among the known breast cancer predisposition loci. Of the five breast cancer predisposition alleles originally reported by Easton et al. [24], three were shown to be associated with in situ (998 cases of DCIS and LCIS) disease (rs2981582-FGFR2, rs3803662-TOX3, rs889312-MAP3K1) with rs889312 showing a stronger association with DCIS (P-trend 0.007, per allele OR 1.30 for DCIS, per allele OR 1.13 for invasive disease). However, this finding of potential DCIS-specific loci was not confirmed in the Million women study which found no differential association with DCIS vs IDC for twelve breast cancer susceptibility loci, including rs889312, although their sample size was smaller (873 DCIS and 4,959 IDC) [12]. In the recent BCAC , and rs3903072 (11q13.1), for which associations seemed to be restricted to invasive disease [19]; however, we found no evidence of an IDC-specific association with these loci after correcting for multiple testing. A recent study investigating the association between 39 of the known breast cancer predisposition loci and breast cancer in situ (BCIS) suggested that rs1011970 (9p21.3, CDKN2BAS) had a stronger association with BCIS than invasive breast cancer (BC), P-Het BCIS/BC = 0.0065. This trend remained in a DCIS vs BC analysis (P-Het DCIS/BC = 0.021) [25]. Our data, however, do not support this finding (DCIS OR 1.08, 95 % CI 1.02, 1.14; P = 0.011; IDC OR 1.05, 95 % CI 1.0, 1.09; P = 0.0025, P-Het DCIS/IDC = 0.33). We have also shown for the first time that seven of the known invasive breast cancer predisposition loci not previously shown to be associated with DCIS have comparable ORs for IDC and DCIS: rs4973768 (SLC4A7), rs3821902 (ATXN7) [26], rs109 95190 (ZNF365), rs554219 (CCND1), rs3757318 and rs2046210 (ESR1).
This lack of DCIS/IDC-specific loci is in contrast to our previous study of lobular cancer in which we showed that there are loci that are specific to invasive lobular cancer (ILC), showing no association with lobular carcinoma in situ (LCIS) and there was also a suggestion of LCISspecific loci [16]. When we compare the DCIS data presented here to our previous LCIS analyses it reveals that there is some overlap between loci that are associated with ER+ DCIS and LCIS ( Fig. 3 and Additional file 14). However, there are also some differences: rs6678914, LGR6 and rs865686, 9q31.2 are strongly associated with LCIS but there is little evidence of association with ER+ DCIS (P-Het DCIS/LCIS = 7.4 × 10 -5 and 6.6 × 10 -4 , respectively). We have also previously shown that rs11249433, 1p11.2 and rs11977670, 7q34 have a stronger association with invasive lobular cancer than IDC [16]. These loci were only weakly associated with LCIS and were not associated with ER+ DCIS in this analysis.
Most association studies of invasive breast cancer involve subgroup analyses based on ER status. In contrast to invasive breast cancer, ER status in DCIS is not routinely assessed in all centers despite evidence from the NSABP B-24 trial of benefit from endocrine therapy in ER+ DCIS [7]. A national audit of DCIS in the UK revealed that ER status was assessed in only 50 % of DCIS cases and ER positivity in low and intermediate grade DCIS was significantly more common than in high grade DCIS (P <0.001) (ER+ high grade 69 %, intermediate grade 94 %, low grade 99 %) [27]. In order to overcome this issue we performed ER immunohistochemistry on the samples from ICICLE for which ER status was unknown. However, there was still a large amount of Cytonuclear grade of DCIS is used by many clinicians to select those cases most likely to benefit from radiotherapy despite the fact that grade has not been shown to be a good predictor of recurrence. In the UK audit of DCIS, grade data were available for 99 % of DCIS cases, with 59 % classified as high grade, 29 % as intermediate and 11 % as low grade [27]. Similarly, in our study data on grade were available for 95 % of cases in ICICLE. In invasive disease only a minority of predisposition loci have been shown to be grade specific; rs2981582 (FGFR2) and rs13281615 (8q24) [28,29] and rs10941679 (5p12) [23]. We have shown that analysis of DCIS by grade reveals other known loci that are grade specific. The loci with the strongest association with grade were SNPs on 11q13, which had a stronger association with low/intermediate grade DCIS and IDC than high grade lesions. The finding of a strong association with low and intermediate grade ductal carcinomas that is independent of ER status in both DCIS and IDC for these loci is novel. rs614367 was the first locus on 11q13 shown to be associated with invasive breast cancer [30]. Fine mapping of the region subsequently identified two independent signals (rs554219 and rs78540526, r 2 = 0.38), which are the loci reported in this analysis. Functional analyses demonstrated that the risk variants modify enhancer and silencer elements, with the likely target gene being CCND1 [31].
A study of 150 cases of subsequent breast cancer (invasive and in situ) after DCIS observed significant association for both grade and ER status between the index DCIS and the subsequent breast cancer (whether ipsilateral or contralateral), suggesting that women with DCIS are at risk of developing subsequent breast cancers of a similar phenotype [32]. This finding supports the genetic predisposition data presented here, with ER and grade-specific loci in DCIS having similar specificity in IDC.
Although we did not identify any novel loci that reached genome wide significance, we did identify three potential novel DCIS predisposition loci, two of which were DCIS-specific (rs12631593, rs73179023), and therefore need further investigation in other cohorts of DCIS. As at least 45 % of patients with IDC have associated DCIS present at diagnosis consistent with direct precursor behavior, it may seem biologically implausible that an SNP predisposes to DCIS but is not associated with IDC. However, it is possible that there is a subset of patients with DCIS with very low probability of progression. If the finding of DCIS-specific predisposition loci were confirmed in other studies, identifying such a subset of patients with low-risk DCIS would be clinically valuable.

Conclusion
In conclusion this is the largest study to assess genetic predisposition in DCIS and shows that the majority of invasive breast cancer predisposition loci also predispose to DCIS. It highlights that, as for invasive disease, different SNPs predispose to ER+ and ER-DCIS. In addition it shows the importance of grade in both DCIS and IDC.

Competing interests
The authors declare that there are no conflicts of interest.

Authors' contributions
The study was conceived by ES and RR. Analysis and genotyping in ICICLE was performed by ES. Meta-analyses were performed by MGC.