Increased Risk of Local Recurrence Is Associated with Allelic Loss in Normal Lobules of Breast Cancer Patients¹

Advances in mammographic detection of breast tumors have allowed earlier detection and better localization, thereby offering the possibility of minimal surgical resection. Despite the increasing acceptance of breast conservation in the treatment of early-stage breast cancer, concerns regarding excision width have not been fully addressed (1). The effectiveness of wide resection margins in reducing the rate of local recurrence in early-stage patients (2, 3, 4) remains to be established. In randomized trials designed to test the efficacy of postlumpectomy adjuvant therapies, margin status did not affect clinical outcome (5, 6). This suggests that some ipsilateral recurrences arise de novo from clonal precursors within a localized “field” of increased susceptibility. This possibility is further supported by high rates of local recurrence observed in younger patients with histologically clear surgical margins (7). Molecular indicators that enable early detection of such high-susceptibility, histologically normal regions would be an important clinical adjunct. LOH³ in morphologically normal lobules adjacent to breast cancer at 3p22–25 has been suggested as one event that could increase susceptibility to breast cancer (8). We demonstrate here that allele-specific LOH within this region in morphologically normal TDLUs adjacent to carcinoma (3pLOHn) is a high-risk marker of local recurrence in early-stage breast cancer patients. In addition, we have narrowed the region of LOH to 3p24.3, a chromosomal region not reported previously to harbor genes causally associated with the suppression of malignancy. This pilot study provides a strong rationale for examining 3pLOHn in a larger patient cohort for further validation and clinical application.

Because the primary goal of this study was to investigate the role of morphologically normal TDLUs adjacent to cancerous breast tissue in the development of locally recurrent breast disease, the most important criteria for sample selection were: (a) early-stage patients treated with breast-conserving surgery; and (b) the presence of TDLUs, composed of ≥500 cells, within the tumor sample (for optimal DNA yield). Under Institutional Review Board-approved guidelines, we identified 64 cases of stage I and II patients diagnosed with invasive ductal breast carcinoma at the California Pacific Medical Center, San Francisco, and the University of California, San Francisco Medical Center between 1986 and 1993 for which archived samples and information regarding local recurrence or disease-free status were available. The median age of the patients was 53 years (range, 25–87 years), and the median follow-up was 76 months (range, 11–134 months).

Serial 5-μm sections of paraffin-embedded tumor and adjacent normal tissue were used for manual microdissection. To maximize the DNA yield, the same TDLUs or group of tumor cells was microdissected and pooled from two to five sequential sections. The criterion for selection of morphologically normal TDLUs, reviewed by two pathologists (B. M. L. and Z. H. M.), was the presence of several acini displaying prominent lumens. Each acinus within microdissected TDLUs consisted of a single continuous myoepithelial or basal layer surrounding a single layer of luminal epithelial cells with uniformly small nuclei containing evenly distributed chromatin.

Nonmalignant skin DNA from each case was isolated and assessed for the presence of heterozygosity at the following RFLP and microsatellite loci between 3p11–26: D3S2397, 71G12, D3S2405, D3S1597, D3S2414, D3S1244, D3S3038, D3S1255, TRβ1, D3S2423, D3S2396, D3S1768, D3S1766, 2C06, and D3S2438. Informative cases were microdissected and analyzed for LOH. Nonmalignant skin and tumor DNAs were compared. Samples of tumors that displayed allelic loss at any locus were further tested for LOH in morphologically normal TDLUs adjacent to tumor.

We used a robust PCR-based assay for LOH analysis to effectively use the small amounts of DNA obtained by microdissection of individual TDLUs. Microdissected cells were incubated in lysis buffer (10 mm Tris-HCl, 1 mm EDTA, 1% Tween 20, and 400 mg/ml proteinase K) at 50°C until the sample was clear. DNA was amplified in a 10-μl reaction mixture for a total of 40 cycles, as described previously (8). To minimize the chance of over interpretation of PCR results because of technical artifacts often associated with LOH analysis, the following measures were adopted: (a) the number of cells was determined before microdissection and solubilization in a standard volume:cell ratio; (b) a minimum DNA template volume containing the equivalent of 50 cells was amplified by PCR; and (c) allelic intensity was compared within multiple runs of a sample in a single PCR assay and between independent assays. Depending on the test locus, the expected constitutional allelic ratio of 1:1 varied by an average of 5% in control samples of nonmalignant skin from different individuals. LOH was recorded only if we observed a >30% reduction in the allelic ratio of test sample compared with control and a <5% difference between parallel duplicate reactions of test samples showing reduced allelic ratios.

Statistical analyses of events associated with recurrence were limited to 48 patients with stage I and II breast cancer whose tumors displayed 3pLOH. Clinical data in the correlative analyses included: patient status (alive, with local recurrence, or no evidence of disease), age, lymph node positivity, tumor grade, tumor size, tumor margins (negative or positive), chemotherapy (yes or no), and hormonal therapy (yes or no). Information regarding patient outcome was obtained by chart review. Molecular analyses were blinded to patient outcome until allelic ratios were determined. Multiple retrospective statistical analyses were carried out.

We first tested whether the prevalence of 3pLOHn was the same in recurrent and nonrecurrent cases by Fisher’s exact test. In a second analysis, a Cox proportional hazards model was used to assess the risk of recurrence in the presence or absence of 3pLOHn. Patients with distant recurrence but without detectable local recurrence were censored at the time of distant recurrence. Factors such as grade, stage, nodal status, age, and treatment modality were included in the model. Finally, we compared time to local recurrence in cases with and without 3pLOHn using a Wilcoxon rank sum test.

In addition, a χ² test was used to determine whether LOH occurred more frequently in tumors at the TRβ1 locus compared with any other locus and whether adjacent normal TDLUs were more likely to display LOH at locus TRβ1 in such tumors than in tumors showing LOH at other loci. Each table compared the number of tumors (or adjacent normal TDLUs) with LOH at that locus versus the number at all other loci combined. Two-sided tests of significance were used for all analyses.

Stage I and II tumors were selected for this LOH study to ensure that normal adjacent TDLUs would not be overgrown by aggressively proliferating malignant populations, as is often the case in late-stage tumors. Up to 15 loci were analyzed for LOH to define the most commonly deleted region. The TRβ1 locus was included because it was reported to be deleted in an earlier study of normal TDLUs adjacent to carcinoma (8). LOH involving at least one locus on 3p was present in 48 of 64 of the tumors, and the prevalence ranged from 46 to 67% at loci between 3p24.2 and 3p25. Importantly, 3pLOHn was present at ≥1 locus between 3p11–26 in adjacent morphologically normal TDLUs in 13 of the 48 breast tumors that displayed 3pLOH (Fig. 1). Variations observed in the degree of LOH between microdissected lobular units reflected the ratio of intralobular fibroblasts to epithelial component of TDLUs. Similarly, in some tumors, intratumor fibroblasts comprised a noticeable fraction, whereas in other cases such contamination was minimal.

We analyzed matched sets of primary tumor and normal adjacent TDLUs to determine the most common region of early allelic loss. As shown cumulatively in Fig. 2, LOH was detected in 113 of 172 allelic ratios of 48 informative tumors. The TRβ1 locus at 3p24.3 was most commonly involved in allelic loss in normal adjacent TDLUs. Seven of 18 LOH events at TRβ1 in contrast to only 14 of 95 LOH events at other loci were displayed by normal adjacent TDLUs (P = 0.02, two-sided Fisher’s exact test). Interestingly, in the tumor itself the frequency of loss at TRβ1 was not significantly different from the frequency of LOH at other loci, 18 of 24 at TRβ1 versus 95 of 138 at other loci.

Several tests demonstrated an association between the presence of 3pLOHn and local tumor recurrence in 48 patients whose tumors displayed 3pLOH. The prevalence of 3pLOHn, which showed that 6 of 11 (55%) locally recurrent cases, displayed 3pLOHn compared with 7 of 37 (27%) cases that did not recur locally. This difference is statistically significant (P = 0.0475 by two-sided Fisher’s exact test).

A Cox proportional hazards model was used to determine the association between 3pLOHn and risk of local recurrence. Table 1 shows the ratios of the risk of recurrence for cases with 3pLOHn to risk of recurrence for cases without 3pLOHn (defined as the hazards ratio), taking into account several other factors associated previously with recurrence. The hazard ratios for local recurrence in the presence of 3pLOHn ranged from 3.9 to 5.2 and were statistically significant (Table 1), even in the presence of the other confounding factors, e.g., grade, tumor margin, and others. The increased freedom from recurrence in patients without 3pLOHn is summarized in the Kaplan-Meier plot in Fig. 3.

We also compared the time to local recurrence for patients with and without 3pLOHn. The mean time to recurrence for those with 3pLOHn was 62.2 months, whereas without 3pLOHn was 38.6 months (P = 0.068 based on Wilcoxon rank sum test, two-sided). This suggests de novo tumor development in recurrences of patients with 3pLOHn, which would ostensibly require a greater time interval. Because both curves were of cases known to have recurred, the average recurrence-free survival estimates will be less than those of unselected stage I and II patients.

Increased risk of tumor recurrence has been associated with dissemination of tumor cells beyond the surgical margins, tumor size and grade, and age at diagnosis (reviewed in Ref. 9). Deng et al. (8) have suggested that the existence of localized LOH involving chromosome 3p22–26 in morphologically normal breast TDLUs surrounding cancers might predispose individuals to local recurrence. Evidence reported in this paper supports that hypothesis and narrows the region of most frequent loss to 3p24.3. Patients with 3pLOHn in this region were at 4–5-fold higher risk of recurrence than in those patients where 3pLOHn was not detected. Furthermore, 3pLOHn remained a significant predictor of local recurrence even when considered with other risk factors such as age, grade, positive nodes, tumor margins, and radiation therapy often found to be significantly associated with outcome (10). This observation is consistent with the hypothesis that breast epithelial cells carrying LOH at 3p24.3 have increased susceptibility to additional “hits” that are necessary for tumor development. This model also explains our observation that the mean time to recurrence for patients with 3pLOHn was significantly longer than in patients without 3pLOHn (62.2 months versus 38.6 months) because tumors arising de novo would be expected to take longer to develop than those resulting from regrowth of residual tumor cells. Interestingly, in close agreement with our findings but using the approach of analyzing the local recurrences themselves, Haffty et al. (11) found that the mean time to recurrence in patients thought to have a new primary tumor was 67 months versus 42 months in those representing residual disease. More recent studies have shown a better overall survival for patients with new primary tumors than those with true recurrences (12).

These findings implicate a tumor suppressor gene in the region of LOH at 3p24.3 defined in this study. Known tumor suppressor genes on 3p include FHIT at 3p14.2 (13, 14) and VHL at 3p25 (15). However, neither of these genes is in the narrowly defined region of common deletion at 3p24.3 and thus unlikely to be involved in 3pLOHn-related risk. Candidate genes in the 3p24.3 region include TRβ1 and RARβ. Notably, all cases of local recurrence associated with 3pLOHn displayed LOH encompassing the TRβ1 locus.

We speculate that the 3pLOHn observed in this and a previous study (8) occurs early in breast development and is propagated throughout one or more branches of the mammary tree during breast development and maturation. The extent of breast tissue harboring LOH will depend on the timing of the initial genetic insult. For example, the acquisition of LOH before puberty could result in wide distribution throughout the breast tissue. In contrast, LOH occurring in the adult breast might be expanded during limited cycles of breast proliferation such as pregnancy and thus remain relatively confined. This is likely to vary among patients and remains to be determined. In this model, patients with 3pLOHn would be at increased risk of developing breast cancer and would be at increased risk of local recurrence. Propagation of genetic aberrations conferring increased susceptibility during breast development could partially explain the increased incidence of breast cancer in atomic bomb survivors who were at a prepubertal stage when exposed (16). Although, the size of the present study is too small to immediately impact clinical practice, it provides a strong rationale for validation in larger independent patient populations. Simultaneously, technical improvements, which enable rapid detection of such deletions, for example in ductal lavage or nipple aspirates, must be sought. Concurrently, an understanding of the biological characteristics of tumors, which originate from the pathway involving allelic loss at chromosome 3p24.3, could provide improvements in prevention and control for this patient subset.

We thank Drs. Debu Tripathy, Mark Rounsaville, Kathleen Lamborn, and Gary Clark for critical reading of the manuscript.