The relationship between ductal and lobular breast carcinoma is highlighted in cases in which these morphologically divergent carcinomas coexist in proximity within a single patient. We hypothesized that such cases may result from the proliferation of a precursor lesion into a tumor containing areas of divergent morphologic features. In this study, we analyzed loss of heterozygosity (LOH) in 10 cases of coexistent ductal carcinoma in situ (DCIS), lobular carcinoma in situ (LCIS), and invasive carcinoma. DNA from the separate components of each lesion was subjected to LOH analysis using 13 markers on 7 chromosomes. In 7 cases, the DCIS and LCIS shared loss of a common allele, suggesting a clonal relationship. The invasive component shared loss of the same allele in 5 tumors. This finding indicates that coexistent lobular and ductal carcinomas exhibit shared genetic abnormalities, contradicting the conventional concept that these lesions represent separate, exclusive pathways of breast neoplasia. Instead, these traditionally segregated classes of breast cancer may, in fact, share common precursor lesions.
Ductal carcinoma in situ
Lobular carcinoma in situ
Loss of heterozygosity
Carcinoma of the breast has traditionally been classified according to a system that emphasizes the distinction between ductal and lobular subtypes.1 However, recent advances have suggested that the molecular characteristics of breast carcinomas correspond more closely with tumor grade as opposed to conventional histologic subtypes. Specifically, molecular profiling has revealed the existence of a so-called low-nuclear-grade breast neoplasia family, including columnar cell lesions, low-grade ductal carcinoma in situ (DCIS), classical lobular carcinoma in situ (LCIS), tubular carcinoma, tubulolobular carcinoma and classic invasive lobular carcinoma.2 Such lesions share a pattern of genetic abnormalities, marked by diploidy or near diploidy, recurrent loss of chromosome 16q, and gain of chromosome 1q. These anomalies are distinct from those typically seen in high-grade breast carcinomas, in which aneuploidy without 16q deletion is the rule.3
The proposed relationship among morphologically different members of the low-grade breast neoplasia family is highlighted in cases in which ductal and lobular preinvasive and invasive carcinomas coexist in proximity to one another, a scenario occasionally encountered in daily pathology practice. In theory, the separate components comprising these tumors could represent the proliferation of a single clone with different morphologic appearances, as proposed by Buerger et al.4 Such a finding would support the idea of a close molecular relationship between low-grade ductal and lobular carcinoma, despite their different histologic types as defined by the conventional classification scheme. To further explore this possibility, we used loss-of-heterozygosity (LOH) analysis to assess a series of tumors consisting of coexistent DCIS, LCIS, and invasive carcinoma for evidence of a clonal relationship between the morphologically distinct components.
Materials and Methods
After obtaining institutional review board approval, a search of the surgical pathology records of our institution for a period of 5 years (2002–2007) was used to identify breast carcinomas containing coexisting DCIS, LCIS, and invasive carcinoma in proximity. We found 10 such tumors with sufficient material for molecular analysis. The patient and tumor characteristics are summarized in Table 1. H&E-stained slides of each case were reviewed to verify the presence of DCIS, LCIS, and invasive carcinoma in each case, in accordance with conventional criteria for histologic diagnosis.5 The DCIS was low-grade in 6 cases, intermediate in 3 cases, and high-grade in 1 case. The LCIS was classical in 9 cases and pleomorphic in 1 case. The invasive component was of the ductal phenotype (IDC) in 9 cases, including well-differentiated ductal carcinoma (tubular carcinoma) in 6 tumors and moderately differentiated ductal carcinoma in the remaining 3 lesions. The invasive component was lobular carcinoma (ILC) in 1 case (case 10). Loss of E-cadherin immunostaining was confirmed in the LCIS component of each case. Coexisting DCIS and IDC components on the E-cadherin–stained sections were positive for E-cadherin membranous staining. Representative images are presented in Image 1.
Histologic and E-cadherin immunohistochemical images of coexistent ductal carcinoma in situ (DCIS), lobular carcinoma in situ (LCIS), and invasive carcinoma in proximity. A (Case 5) DCIS, LCIS, and invasive carcinoma (H&E, ×20). B, Loss of E-cadherin immunohistochemical staining in the LCIS component (lower half) relative to the E-cadherin-positive duct in the top half of the image (H&E, ×40).
DNA was isolated separately from each morphologic component of each lesion using laser-capture microdissection (Arcturus Automated Laser Capture Microdissection System, Molecular Devices, Sunnyvale, CA). Briefly, after histologic confirmation, areas of 180 to 2,950 mm2 of formalin-fixed, paraffin-embedded tissue were collected from each component of each tumor, by microdissection of 5-μm-thick, H&E-stained sections. DNA was purified from microdissected tissue using the PicoPure DNA extraction kit after digestion with Proteinase K (Molecular Devices).
LOH analysis was then performed on each component using 13 commonly informative and deleted markers.6 These included 6 markers on chromosome 16q and 1 each on chromosomes 1, 8, 9, 11, 13, and 17 Table 2. Germline allelotypes for each patient were derived from DNA isolated from normal lymph node tissue removed at the time of surgery. Dye-labeled polymerase chain reaction primers for each marker were obtained commercially (Applied Biosystems, Foster City, CA) or designed based on the information obtained from the UniSTS database (http://www.ncbi.nlm.nih.gov/sites/entrez?db=unists) and synthesized commercially (Operon, Huntsville, AL) (Table 2). Amplification was carried out for 45 cycles under conditions recommended by the manufacturer.
Electrophoretograms were generated using the ABI PRISM 3100 Genetic Analyzer (Applied Biosystems), and analyzed with GeneMapper, version 4.0 (Applied Biosystems), and GeneMarker, version 1.6 (Softgenetics, State College, PA), with LOH defined as a more than 30% change in the ratio of informative markers. Sample electrophoretograms are provided in Image 2.
Sample electrophoretograms illustrating patterns of loss-of-heterozygosity (LOH). For each case, results for normal lymph node germline DNA and for each tumor component are shown in parallel. Heterozygous alleles found in germline DNA are denoted by arrows; arrowheads correspond to sites of allelic loss in tumor components. A (Case 8), No LOH in any component at locus D9S171 on 9p21.3. B (Case 9), Unshared LOH with loss of an allele only in the invasive carcinoma component at locus D16S752 at 16q22. C (Case 3), Shared LOH involving loss of the same allele at D16S752 in the LCIS and invasive components; the DCIS component shows no loss of this allele. D (Case 1), Discordant LOH showing loss of opposite alleles in the invasive and LCIS components at locus D1S407 on 1p36.21; the DCIS component shows no loss of either allele.
When comparing LOH results in 3 components of a given tumor, a myriad of results are possible, including noninformative locus; no LOH in any component (LCIS, DCIS, or invasive carcinoma); LOH in 1 component only (ie, unshared LOH); concordant LOH of the same allele in 2 or more components (ie, shared LOH); or discordant LOH (ie, loss of the opposite allele in the separate components).
Results of LOH analysis revealed several patterns. In cases 1 through 5, evidence was found to imply a clonal relationship among all 3 components, DCIS, LCIS, and invasive carcinoma Table 3. In these cases, shared LOH in all 3 morphologic components was found for at least 1 locus.
A second pattern, found in 3 cases, exhibited shared LOH between 2 of the 3 tumor constituents. In one of these tumors, LCIS and DCIS shared LOH at 1 locus, whereas DCIS and IDC shared LOH at a separate locus (case 6; Table 3). Another lesion showed shared LOH between LCIS and DCIS, but not IDC, whereas a third case showed shared LOH between DCIS and IDC, but not LCIS (cases 7 and 8, respectively; Table 3).
A third pattern, found in a single tumor, showed no shared LOH among any of the components (case 9; Table 3).
Finally, in the 1 case of coexisting DCIS, LCIS, and ILC, shared LOH was found between ILC and LCIS and also between ILC and DCIS, but not between DCIS and LCIS (case 10; Table 3).
In addition to the shared LOH, unshared LOH was also found in every case. Moreover, examples of discordant LOH were seen in 5 cases (cases 1, 3, 7, 8, and 10), indicating loss of opposite alleles at a given locus in 2 separate components of a tumor.
Recent advances in the molecular characterization of breast cancer have led to the understanding that tumor progression involves a complex series of genetic events. These events can be grouped into at least 2 distinct multistep pathways, terminating in low-grade DCIS/well-differentiated IDC or high-grade DCIS/poorly differentiated invasive carcinoma.3 These separate pathways are distinguished by quantitative and qualitative differences in genetic aberrations: low-grade carcinomas are associated with relatively few such events, with loss of 16q and gain of 1q being the most common abnormalities. In contrast, high-grade carcinomas typically harbor multiple alterations—most frequently loss of 8p, 11q, 13q, and 14q; gain of 1q, 5p, 8q, and 17q; and amplification of 6q22, 8q22, 11q13, 17q12, 17q22–24, and 20q13—but only rarely exhibit loss of 16q and gain of 1q. Tumors in the latter group also share more aggressive clinicopathologic features, including greater nuclear atypia, negative estrogen and progesterone hormone receptor status, and protein over-expression of HER2. Thus, histologic grade bears a strong correlation with the underlying genomic aberrations in breast cancer, and this finding forms the basis for separating breast cancers into low- and high-grade subgroups.
The concept of separate low- and high-grade breast tumor families has also shed light on the relationship between ductal and lobular carcinomas. It is now clear that lobular carcinomas share genotypic similarity with tumors of the low-grade ductal carcinoma group, and, conversely, that low-grade ductal carcinoma may be more similar to lobular carcinoma than to high-grade ductal carcinoma. However, most evidence for this relationship relies on comparing genomic data obtained from variably sized populations of unrelated lesions.4,7–10 In the present study, we have chosen an alternative and more direct approach, namely, to study lesions that consist of morphologically distinct areas of LCIS, DCIS, and invasive carcinoma in proximity. The concept of a “family” of low-grade breast tumors, related genetically despite different morphologic appearances, would be supported if a clonal relationship could be demonstrated among multiple histologically distinct elements coexisting within individual lesions.
Indeed, in 5 of 9 tumors containing coexistent LCIS, DCIS, and IDC, shared LOH could be demonstrated among all 3 components, strongly suggesting a clonal relationship among the morphologically distinct components of these tumors. In these additional cases, shared LOH was found between 2 of the 3 components. Although prior studies have suggested this relationship on the basis of genetic characterization by comparative genomic hybridization and LOH analysis of a large number of tumors, our use of tumors containing lobular and ductal carcinoma allowed us to directly demonstrate a clonal relationship between these traditionally segregated, morphologically distinct subtypes of breast carcinoma within individual tumors. Our findings, in line with the prior literature, support and further strengthen the notion of a close relationship between low-grade ductal carcinoma and lobular carcinoma.
One model of breast cancer taxonomy holds that LCIS and well-differentiated DCIS could be derived from a common clonal precursor. Progression toward LCIS according to this model would be accompanied by loss of E-cadherin, through genetic or epigenetic mechanisms.11 Our results are consistent with this model in that coexistent LCIS and DCIS in most cases exhibit shared genetic abnormalities, while also showing divergent abnormalities (ie, unshared or discordant LOH). Thus, the morphologically distinct components in these mixed tumors apparently harbor changes of a common clonal ancestor, as well as distinct accumulated, unshared alterations. No relationship between grade or degree of differentiation and clonal relatedness could be discerned from our data, although a larger number of cases would likely be necessary to address this question.
Our approach mirrors earlier studies in which relatedness between an in situ tumor and its adjacent invasive component was assessed using microdissection of tissue from a single specimen.12,13 This method, however, has important limitations that constrain the interpretation of our findings. First, in most cases, at least 1 component was present only in scant quantities, thereby limiting the amount of tissue available for microdissection and genetic analysis and preventing us from performing more systematic techniques like array comparative genomic hybridization or single nucleotide polymorphism arrays.
Second, LOH analysis is also limited by the potential for contamination of target tissue with adjacent normal or morphologically distinct tumor tissue, a possibility we minimized by using laser-capture microdissection to separately acquire DNA from each tumor component.
Third, there are limitations in adapting LOH data to a quantitative assessment of the probability of clonality in the lesions studied here. Although the increasingly frequent use of LOH data has prompted the development of specific statistical tests for clonality,14,15 the currently available methods are designed to assess 2 populations of tumor cells, whereas our experimental design would require integration of 3 distinct genotypes (DCIS, LCIS, invasive carcinoma). Moreover, these methods rely on a foreknowledge of the general prevalence of LOH at each given locus in each separate tumor component. Thus, a more quantitative assessment of the probability of clonality awaits the development of statistical approaches that can incorporate more than 2 tumors or tumor components, as well as a more comprehensive genomic assessment of LOH prevalence determined separately for LCIS, DCIS, and invasive carcinoma. Nevertheless, although some cases of shared LOH in our study may be due to chance, multiple concordant losses in 3 different tumor components on separate chromosome arms, as seen in cases 3 and 5, are highly suggestive of a clonal relationship.
Finally, our findings may have clinical implications as well. Given that LCIS is clonally related to adjacent ductal carcinoma in some cases and the possibility that LCIS in these lesions could evolve into ductal carcinoma, the question of reporting of LCIS at surgical margins in the setting of adjacent low-grade ductal carcinoma is raised. However, further studies involving large numbers of patients would be needed to assess the hypothetical concern that concurrent LCIS at surgical margins in the setting of completely excised DCIS might be associated with increased local recurrence of low-grade DCIS.
High frequency of coexistence of columnar cell lesions, lobular neoplasia, and low grade ductal carcinoma in situ with invasive tubular carcinoma and invasive lobular carcinoma. Am J Surg Pathol. 2007;31:417–426.