Breast cancer is a heterogeneous disease encompassing a wide variety of pathological entities and a range of clinical behaviour. These are underpinned at the molecular level by a complex array of genetic alterations that affect cellular processes 1–3. In seeking to characterize these aberrations, the study of cancer genetics has had a major impact in our understanding of the development and progression of breast neoplasms.
Historically, breast cancer progression was seen as a multi-step process, akin to the Volgenstein's model for colon carcinogenesis 4, encompassing progressive changes from normal, to hyperplasia with and without atypia, carcinoma in situ, invasive carcinoma, and metastasis (Figure 1) 5.
Whilst most of the concepts regarding the morphologically defined breast cancer precursor lesions remain valid, immunohistochemistry and molecular genetics have changed the way that the breast cancer multi-step model is seen 2, 3, 6–13. No longer do we perceive it as a single pathway, but as a complex series of stochastic genetic events leading to distinct and divergent pathways towards invasive breast cancer. The distinction between ‘lobular’ and ‘ductal’ pathways has now been blurred. In addition, some of the lesions along the pathway have been repositioned and the role of others has been questioned 2, 3, 5–7, 11, 14–17.
Although some aspects of the molecular genetics of invasive breast carcinomas have been known for a long time, the inability to study its putative precursors precluded any direct correlation. Since the advent of reliable methods of tissue microdissection 18 (ie laser capture microdissection), DNA amplification 19–21, and genome and transcriptome analysis [ie cytogenetics, comparative genomic hybridization (CGH), loss of heterozygosity (LOH), gene expression analysis, microarray CGH], it has become possible to study the molecular aspects of benign proliferative and pre-invasive breast lesions and their relationship to invasive breast carcinoma 10, 15–17, 22.
Invasive and in situ breast carcinomas
In the past 15 years, molecular pathology of invasive breast cancer has received great attention. Attempts to redefine breast cancer taxonomy 23–25, refine the prognostic indicators, and predict recurrences and metastases have been carried out 24–32.
Classification of breast cancer into prognostically meaningful groups has traditionally been done by histological (sub)type; however, it has become increasingly clear that grade (degree of differentiation) is a better predictor of outcome than type 2, 12, 13. Therefore, it is not surprising that molecular methods have demonstrated that grade, more than any other clinico-pathological parameter, strongly reflects the extent, complexity, and type of genomic aberrations 2, 9, 12, 13.
The genetic aspects of low- and high-grade breast cancer not only differ in quantitative terms, but are also segregated by the type of aberrations: grade I and tubular breast carcinomas show a low number of genomic alterations with highly recurrent losses of 16q, whereas grade III breast carcinomas show complex genotypes frequently harbouring loss of 11q, 14q, 8p, 13q; gain of 17q, 8q, 5p; and high-level gains (amplifications) on 17q12, 17q22–24, 6q22, 8q22, 11q13, and 20q13 2, 12, 13. Since the molecular profiles of grade I and grade III tumours are so different, they suggest that for the most part, progression from low- to high-grade breast carcinoma is rare 2, 12, 13. For instance, physical loss of 16q, the most frequent genomic change observed in grade I tumours, is exceedingly infrequent in high-grade breast carcinomas 2, 12, 13, 33. Although loss of 16q is observed in a subset of grade III carcinomas, it has recently been shown to occur by a distinct mechanism (LOH in combination with mitotic recombination) 34.
CGH and conventional cytogenetic studies have shown that there is some degree of variation in the pattern of genetic alterations between different histological types of invasive breast cancer 2, 10, 35–46. Although differences between histological subtypes do exist, this association is not as strong as it happens to be with grade 2, 9, 12. Comparative analyses between invasive ductal (IDCs) and lobular breast carcinomas (ILCs) have demonstrated that overall a lower number of genetic changes are found in ILCs relative to IDCs 10, 35, 36, 44–46. Although some specific chromosomal abnormalities are found at a significantly different frequency in each histological type 10, 35, 36, 44–46, this may only highlight the fact that most ILCs are of lower nuclear grade. Interestingly, several recurrent unbalanced changes, including physical loss of 16q, were common to both types, indicating that ILCs and low-grade IDCs may arise via common pathways of tumourigenesis 35, 36, 44–46.
This finding has blurred the boundary between ductal and lobular lesions and has raised the question as to whether the designations ‘ductal’ and ‘lobular’ are appropriate. In fact, the majority of neoplastic breast diseases arise from the terminal duct-lobular unit (TDLU) and so this terminology is not intended to reflect the micro-anatomical site of origin, but a difference in cell morphology 2, 47. It should also be stressed that although loss of 16q is observed in both grade I IDC and ILC, the culprit genes might differ in these two lesions 47–51. The likeliest candidate gene involved in loss of 16q in atypical lobular hyperplasia (ALH)/lobular carcinoma in situ (LCIS) (see below) and ILC is CDH1 (E-cadherin), which maps to 16q22.1 2, 47, 50, 52, 53. It is accepted that most of ALH/LCIS and ILC harbour loss of 16q, followed by gene mutation, promoter methylation or further loss of CDH1. Loss of E-cadherin, a critical cell adhesion molecule 2, 47, 52, 53, is reflected at the morphological level by the characteristic discohesive nature of individual cells and overall growth pattern of lobular carcinomas. However, CDH1 is almost certainly not the target gene in grade I IDCs, as loss of E-cadherin expression and CDH1 gene mutations are exceedingly rare in these tumours 48. The hunt for the gene actually involved in grade I ductal cancers is ongoing and proving exceedingly difficult 48, 50, 51.
Over the last few years, a pleomorphic variant of lobular carcinoma (PLC) has been described 47, 53–57. Briefly, in pleomorphic LCIS and ILC, neoplastic cells show the typical discohesiveness of lobular neoplasms; however, they are of high grade and show features of apocrine differentiation. Although molecular data on the PLC are scant 47, 53–57, these tumours have overlapping genetic changes with both classic ILC and grade III invasive ductal breast carcinomas, harbouring recurrent loss of 16q and lack of E-cadherin expression, but also showing overexpression of Her-2 47, 53–57. In addition, anecdotal evidence suggests that PLC may have a more aggressive biological behaviour than ILC 53, 55–57.
Studies using expression arrays have also provided new insights into breast cancer taxonomy. Breast carcinomas can be classified into four categories according to their transcriptome: oestrogen receptor (ER)-positive, HER-2-positive, basal, and normal breast-like carcinomas 23. It is interesting that even this separation is fundamentally into ER-positive and -negative groups, which for the most part fall into low- and high-grade carcinomas, respectively. In following publications, it was shown that these groups have important prognostic implications; namely, patients with HER-2-positive or basal carcinomas (high-grade) do worse than patients with ER-positive or normal breast-like tumours (low-grade) 24, 25. However, recent data also suggest that each of these groups may be heterogeneous at the genetic and clinical levels 26, 58, 59. In addition, it is still unclear whether there is any correlation between the expression profile, molecular genetic features, and histopathological parameters of these tumours.
The frequent association and morphological similarities between invasive and some forms of proliferative breast diseases have led pathologists to assume that certain entities would be biologically related (eg LCIS and ILC, DCIS and IDC) 2, 5. The complexity of these relationships has been thoroughly explored using the advancement in molecular pathology. For example, immunohistochemistry, LOH, CGH, and gene expression analysis have recapitulated the genotypic/phenotypic patterns observed in invasive ductal and lobular carcinomas in ADH/DCIS and ALH/LCIS 2, 5, 9, 11, 16, 44, 60–62. The distinct molecular genetic features found in different grades of invasive carcinomas are also mirrored in pre-invasive lesions of comparable morphology 2, 5, 11.
In hindsight, it seems clear that there are two major arms in the multi-pathway model of multi-step breast cancer progression (Figure 2): one comprising well-differentiated DCIS that progress to grade I IDC, and the other encompassing poorly differentiated DCIS that progress to grade III IDC. In the ‘low-grade arm’, tumours are of low nuclear grade, usually ER and PgR-positive, negative for Her-2 and basal markers, and harbour low genetic instability and recurrent 16q loss, whereas in the ‘high-grade arm’, the lesions show a higher degree of nuclear atypia, are more frequently hormone receptor-negative, frequently positive for either Her-2 or basal markers, and are genetically advanced lesions, showing a combination of recurrent genomic changes including loss of 8p, 11q, 13q, 14q; gain of 1q, 5p, 8q, 17q; and amplifications on 6q22, 8q22, 11q13, 17q12, 17q22–24, and 20q13 2, 5, 9, 12, 13. Based on their pathological and genetic features, classic LCIS and ILC are remarkably similar to those tumours in the ‘low-grade arm’ 2, 44, 47. However, in contrast to well-differentiated DCIS/grade I IDC, the vast majority of these tumours lack E-cadherin expression owing to genetic and/or epigenetic changes in the CDH1 gene 48, 50–52. On the other hand, the overlapping morphological features of PLC with both classic lobular and grade III carcinomas, and the combination of E-cadherin (16q) loss with occasional Her-2 positivity 53–57 add another level of complexity to these molecular pathways to breast cancer tumourigenesis.
Hyperplasias and columnar cell change
Apart from ADH and ALH, which bear remarkable morphological and molecular resemblance to well-differentiated DCIS and LCIS, respectively, the other non-obligate/premalignant lesions have proven more difficult to characterize and establish their actual position along the pathways 2, 16, 60–67. Interestingly, ADH and low-grade DCIS show identical immunoprofiles and low numbers of chromosomal abnormalities, comprising recurrent loss of 16q (see ref 2 and refs cited therein). The similarities between ALH and LCIS are not restricted to the morphological level, but also the immunohistochemical and genetic features 47. In fact, differentiating between them is arbitrary and subjective, being based on subtle quantitative morphological features. Hence, it is well accepted that both ADH and ALH are non-obligate precursors to the development of well-differentiated DCIS and LCIS, respectively. Alternatively, one could view them just as small DCIS or LCIS.
For a long time, hyperplasia of usual type (HUT) was seen as the precursor of ADH and DCIS (see Figure 1). However, its role in the multi-step model of breast carcinogenesis has been questioned (for a review see ref 2 and refs cited therein) 2, 5, 15, 16, 64–67. The morphological features and immunoprofile of HUTs are in stark contrast with those of the accepted precursors, since they are composed of a mixed population of cell types with a variable proportion of ER, PgR-positive luminal cells and myoepithelial/basal marker-positive cells. At the molecular level, no or only rare and fairly random chromosomal changes are observed 2, 15, 65. Despite the controversies, there is evidence to suggest that a small proportion of HUTs may be clonal, neoplastic proliferations (adenomas) that may putatively progress to ADH or DCIS, whereas the majority of them fail to show any evidence of a neoplastic/monoclonal nature using existing technology 2, 3, 16, 66.
A more likely precursor to ADH and well-differentiated DCIS are columnar cell lesions (CCLs) 8, 68–70, which are characterized by the presence of tightly packed columnar-shaped epithelial cells, with ovoid-to-elongated nuclei and prominent apical cytoplasmic snouts and intraluminal secretions. In fact, they comprise a spectrum of lesions with varying degrees of architectural and mild nuclear atypia. At the lower end of the spectrum are lesions composed of variably dilated acini lined by a single layer of the characteristic cells. At the higher end of the spectrum, lesions resemble ADH, with stratification, bridging, and punched out spaces. Throughout the spectrum, CCLs show an immunoprofile similar to that of ADH/well-differentiated DCIS 8. However, the degree of proliferation, architectural and cytological atypia are mirrored at the genetic level, with a stepwise increase in the number and complexity of chromosomal copy number changes as defined by CGH 70. Moreover, the hallmark genetic feature of ‘low-grade’ lesions, loss of 16q, is the most frequently detected recurrent change and in addition, there is some degree of overlap in the molecular genetic profile of CCL and associated more advanced lesions 69, 70. Interestingly, it is not infrequent to observe ALH/LCIS in the context of multifocal CCLs. Hence CCLs may be the link between normal breast and ADH, as well as between ‘ductal’ and ‘lobular’ neoplasia.
The precursor of poorly differentiated DCIS has been elusive. Based on morphological, immunohistochemical, and molecular findings, CCL, ADH, and well-differentiated DCIS would be unlikely candidates. Although apocrine change has long been considered a metaplastic process in breast tissues, usually associated with ageing, this concept has come into question with the application of molecular findings 2, 17, 71–73. At least a subset of lesions with apocrine morphology show molecular changes, including LOH/allelic imbalance at 1p (MYCL1), 11q (INT2), 13q, 16q and 17q 71, 72, and recurrent chromosomal changes as defined by CGH, including loss of 1p, 2p, 10q, 16q, 17q and 22q, and gain of 1p, 2q and 13q 17. These findings are more frequently observed in apocrine adenosis and apocrine hyperplasia compared with apocrine cysts. For the large part, these observations have been ignored by the diagnostic pathologist. It is remarkable that a proliferative lesion with the architecture of a micropapillary DCIS is considered benign because it happens to have abundant pink cytoplasm! The lesion regarded as papillary apocrine hyperplasia would be seriously considered DCIS by many if it was not apocrine in nature. Whether this prejudice is justified should be questioned. It may turn out to be wrong but we would suggest that there is compelling molecular data that at least some of these lesions may be the precursors of high-grade DCIS and invasive cancer. There is a real need for further work in this area.
Benign proliferative breast lesions and normal breast
Little is known about the presence of molecular genetic changes in benign proliferative breast lesions, such as adenosis, papillomas, and tubular adenomas 2, 65, 73. Although gross chromosomal aberrations are not observed in normal breast, most non-proliferative breast lesions, and papillomas, there are some data to suggest that non-recurrent LOH at some loci may be found 2, 65.
Although there are changes detected by LOH in some benign proliferative breast lesions, these show little overlap to the most prevalent genetic changes described in DCIS and invasive carcinomas and are thus insufficient to make them non-obligate precursors of DCIS or invasive breast cancer 2, 65, 73. In addition, whilst LOH in neoplastic lesions encompasses all informative markers mapping to a particular chromosome arm, LOH in adjacent normal-appearing epithelium involves only single, random markers across the genome 74, 75. Although the genetic abnormalities found in normal cells show little or no overlap with those of neoplastic lesions, these changes may indicate defective mechanisms for the maintenance of genomic integrity, which may contribute to the carcinogenic process and denote increased risk for developing breast cancer.
Conclusion and future prospects
Molecular analysis has been extensively utilized in the characterization of breast lesions and has been instrumental in unravelling a direct relationship between genotype and phenotype, identifying new subtypes, illustrating molecular pathways in the progression to invasive carcinoma, defining relationships between different histological variants, and providing invaluable prognostic parameters.
Now, the mission of pathologists and scientists alike is to translate our current understanding of the molecular evolution of breast cancer and the complexity of microarray data into practical methods that are suitable for diagnostic pathology and patients' management.
J Reis-Filho is the recipient of the Gordon Signy International Fellowship Award and is partially supported by PhD Grant SFRH/BD/5386/2001 from the Fundação para a Ciência e a Tecnologia, Portugal, and Programa Operacional Ciência, Tecnologia e Inovação POCTI/CBO/45157/2002.