For women with early-onset breast cancer who neither carry a mutation in a known breast cancer susceptibility gene nor have BRCA1 promoter region methylation in peripheral blood DNA, features of their tumour morphology predict wide variation in the risk of breast cancer for their first-degree female relatives, far more than could be attributed to chance. For example, female relatives of the approximately one-third of early-onset breast cancer cases who had tumours that lacked any morphological feature associated with familial risk were at no increased risk of breast cancer. Female relatives of the approximately one-sixth of early-onset breast cancer cases who had tumours that had two or more of the four features associated with familial risk were at a substantially increased risk of breast cancer. The latter finding suggests there are undiscovered strong familial risk factors for early-onset breast cancer. The former suggests there is a substantial proportion of early-onset breast cancer that does not have a familial, let alone genetic, cause.
This work has built on previous work by us and others that identified specific morphological features which predict cases with a germline BRCA1
]. We recently extended this to show that those features also predict cases with methylation of the promoter region of BRCA1
in peripheral blood DNA [13
]. Here, we have widened the scope of what we are predicting to risks for relatives and, at the same time, included all the other morphological features assessed by a standardised and validated pathology review [18
]. We also assessed morphological predictors both including and excluding cases with a known genetic cause or methylation of the BRCA1
When first considering all index cases, we found that absence of extensive sclerosis, extensive intraductal carcinoma, and trabecular and lobular growth patterns and absence of acinar and glandular growth patterns were all independent predictors of breast cancer for their first-degree female relatives. When considered together, for each of these six features their presence was associated with an approximate doubling of risk for relatives. This means there is a substantial gradient of risk for relatives across a number of these features.
We have previously shown that the presence of trabecular growth pattern was the major predictor of BRCA1
mutation status, as well as of methylation of the BRCA1
promoter region [12
]. When we excluded index cases with known genetic causes or methylation of the BRCA1
promoter region, absence of extensive sclerosis, circumscribed growth, extensive intraductal carcinoma and lobular growth pattern were each independent predictors of breast cancer risk for the first-degree female relatives of the remaining index cases. Note from Figure that these factors are not highly associated with one another. Again, the strengths of prediction were each an approximate doubling or more of risk, and there was a substantial gradient in risk for the relatives associated with the number of these four features in the tumour of the index case. Therefore, this study has identified characteristics of breast cancers in young women, other than basal-like carcinoma (which is likely rare in our subgroup analyses due to exclusion of the index cases with either known BRCA1
germline mutations or methylation of the BRCA1
promoter region), that are associated with increased risks of breast cancer for first-degree relatives.
There are several strengths of our study. First, it is population-based, so inference can readily be made at least to the populations from which these index cases were sampled, if not more generally. Second, because the index cases were not selected on the basis of family history, this study gives insights across the whole spectrum of index cases. Third, we have used data from multiple pathologists conducting a standardised and validated tumour review and any systematic variation across the pathologists would mean that our study has under-estimated the predictive value of these morphology features.
The study also has some weaknesses. It is possible that, despite extensive testing, some index cases carrying mutations in BRCA1
have not yet been identified, although this is likely to be, at most, a few percent [3
]. There may be some false positive statistical results due to the vagaries of chance. The study has used only early-onset index cases, so we cannot infer that the same results will apply to index cases of later onset. Our index cases were predominantly Caucasian and we cannot infer that these results apply to other populations. Therefore, there is a need for replication using independent samples, and using samples of index cases with different ages at onset and different ethnic backgrounds. Finally, although we have identified a small number of features that independently predict risks for relatives in our data set, there was still substantial unexplained variation in risk. Therefore, larger and pooled studies might identify a bigger and better set of predictors. These might also clarify which factors are the better independent predictors, given that some features are highly associated with one another (see Figure ).
While there are several studies of breast cancer in relatives as a function of the ER and PR tumour status of an affected woman (for example, [33
]), few have studied histological features (for example, [34
]). Several studies have shown that stratification of breast tumours by ER and PR status might be useful in partitioning breast cancer families into more homogeneous subsets [35
], but we have found that these traditional immunohistochemical features are not as important as morphological features for predicting BRCA1
mutation carriers [12
] or, as in this study, predicting breast cancer risks for relatives. Specific morphological types of breast cancer might be associated with familial cancers in general (or example, [36
Being able to quantify a woman's risk of breast cancer is important for genetic counselling, prevention and screening, as well as for aetiological research. Currently, risk estimates associated with having a family history are based on the age at diagnosis of cancer, not on the cancer's morphological features. We have shown here that, at least for first-degree relatives of women with early-onset breast cancer, this is likely leading to a misassignment of familial risks for about one-half of index cases, being too high for one-third and too low for one-sixth, by factors of two-fold or more. Assignment of women with a family history of breast cancer to different screening strategies based on their absolute risk could be substantially improved were the morphology features of the tumours of the relatives taken into account. The role of morphological features to predict BRAC1
mutation status has been established for some time [31
]. For early-onset breast cancer, we showed that only two key features could better predict BRCA1
mutation carriers than family history and/or ER and PR tumour status [12
]. Now we have shown that just four other morphological features predict familial risks after excluding the known breast susceptibility genes, and again better than ER and PR tumour status.
It is highly relevant to genetic research and counselling that one-third of early-onset breast cancers might not even have a heritable genetic basis, as evidenced by our observation that, for the group of index cases with none of the four predictive features, their relatives were not at increased risk. Most linkage studies and the recent genome-wide association studies have generally pooled all cases when trying to find predisposing genes. It is also of relevance to know if the type of breast cancers, as defined by the morphology features identified here, 'runs in the family', as this would greatly assist efforts at susceptibility gene discovery.
For the affected women, their germline status, as well as changes (both somatic genetic and epigenetic) to tumour DNA, might influence gene expression that manifests in nuances of cell growth and neoplastic tissue organisation that are able to be observed by microscope. For example, breast tumours arising in women with germline BRCA1
mutations and TP53
mutations have characteristic morphology and acquired genetic and epigenetic mutation profiles [37
]. For unaffected relatives, any familial risk associated with the breast tumour morphology features of an affected relative must be related to similar genetic or epigenetic variants that increase cancer susceptibility. From this, it follows that DNA from affected women whose breast tumour morphology features are associated with the highest familial risks should be prioritised for comprehensive genetic and epigenetic screening as they provide the best prospect for the discovery of novel variants that are associated with risk.
On their own, the ER and PR status of the index cases were both associated with risk for relatives overall, with about a 50% increased risk if the tumour of the index case was negative for either of these immunohistochemical markers. After the exclusions, the associations were of the same magnitude and direction but no longer statistically significant. However, neither of these immunohistochemical markers were independent predictors of risks for relatives once the more strongly predictive morphological features were taken into account. That is, we present here an advance on using ER and PR alone to assess the role of pathology in predicting increased familial and possibly genetic risks.