Reduced Fhit Expression in Sporadic and BRCA2-linked Breast Carcinomas¹

A complex set of genetic abnormalities has been detected in breast tumors, including amplification of oncogenes, mutation of the tumor suppressor gene TP53 and LOH³ from chromosomes 1, 3p, 6q, 7q, 8p, 11, 13q, 16q, 17, 18q, and 22q (reviewed in Ref. 1). The regions of LOH may target tumor suppressor genes. Furthermore, germ-line mutations of the BRCA1 and BRCA2 genes have been implicated in inherited breast cancer (2, 3). Somatic loss of the wild-type chromosome in tumors of BRCA1 and BRCA2 mutation carriers suggests that both alleles of the corresponding gene are inactivated in cancer, a pattern expected of a tumor suppressor gene (4). A germ-line mutation in the BRCA2 gene, termed 999del5, has been detected in Icelandic breast cancer families, and a common haplotype in the vicinity of the gene in these families suggests a founder effect (5, 6, 7). The BRCA2 999del5 germ-line mutation has been found in 8% of patients diagnosed with breast cancer in Iceland (8, 9). A higher frequency of aberrations at several chromosome arms in tumors of BRCA2 germ-line mutation carriers compared with sporadic tumors has been reported (10, 11, 12). These results suggested a specific or more aggressive tumor progression pathway in breast cancers due to a BRCA2 mutation.

In sporadic breast cancer, interstitial deletions of the short arm of chromosome 3, with involvement of several 3p subbands, have commonly been observed by studies of cytogenetics and loss of heterozygosity (reviewed in Ref. 1). Band 3p14 has been observed to be the minimal common deleted segment (13, 14, 15), and similar deletions have been observed in benign proliferative breast disease (16, 17), suggesting that 3p14 deletions could be an early event in a significant fraction of mammary carcinomas. After description of the FHIT gene, encompassing the common chromosomal fragile region (FRA3B) at band 3p14.2 (18), Panagopoulos et al. (19) showed that the FHIT gene was homozygously deleted in two cases of benign proliferative breast disease associated with 3p14 cytogenetic rearrangements and familial breast cancer. Additionally, homozygous deletions within the FHIT gene have been observed in some breast cancer-derived cell lines, and ∼20% of primary breast carcinomas were reported to exhibit altered Fhit transcripts by reverse transcription-PCR analysis (20, 21); altered reverse transcription-PCR products are frequently due to internal deletions within FHIT, which appears to be inactivated by deletions rather than point mutations (reviewed in Ref. 22).

In familial breast cancer kindreds, early studies by Bergthorsson et al. (23) showed a high frequency of allelic imbalance at 3p14 relative to the frequency of allelic imbalance in the same region in sporadic breast cancers. In a follow-up study, Bergthorsson et al. (10) mapped the 3p regions of imbalance in known BRCA2 carrier tumors relative to sporadic tumors. The degree of allelic imbalance for some 3p markers was significantly higher than in tumors from noncarriers; in both groups, elevated 3p14 imbalance was noted in patients with bilateral disease, and allelic imbalance was observed most frequently near marker D3S1217 (centromeric to FHIT) and in the FHIT gene. The authors concluded that genomic aberrations in 3p14 are especially frequent in tumors with BRCA2 gene defects and suggested that they were caused by regional loss of chromosome stability rather than selection; that is, loss of BRCA2 function may affect 3p14 stability in the tumors.

Because the BRCA2 gene is involved in DNA repair (24, 25, 26) and because the FHIT gene encompasses the common fragile region FRA3B, it seems likely that the frequent alterations at 3p14.2 in tumors of BRCA2 carriers could be initiated by faulty repair of fragile region double-strand breaks. This does not preclude the possibility that such alterations, which would frequently inactivate the FHIT gene, could impart a growth advantage to the affected cell. The very fact that these genetic alterations can be detected strongly suggests clonal (mono or oligoclonal) expansion of cells with FHIT locus alterations. To determine the effect of such 3p14.2 alterations on expression of the FHIT gene, we have studied expression of the Fhit protein by immunohistochemical detection in sporadic breast cancers and compared it with Fhit expression in cancers of BRCA2 mutation carriers.

The 50 sporadic breast cancers studied for tumor marker expression were paraffin-embedded normal and malignant tissues obtained for analysis from the Department of Pathology, Anatomy and Cell Biology at Thomas Jefferson University of Philadelphia; 10 cases were DCIS and were excluded from statistical analyses. The remaining 40 cases were invasive carcinomas. Clinical features of the invasive cases are listed in Table 1.

The Icelandic sporadic and familial (BRCA2 999del5) breast cancer tissues have been described previously (8). Twenty-nine sporadic tumors (average age at diagnosis, 47 years; range, 29–79) and 29 familial tumors (average age at diagnosis, 51 years; range, 33–78) were studied for 3p14.2 LOH within the FHIT gene. Fifty-eight sporadic and 34 familial cases were analyzed for expression of Fhit.

Fhit expression was analyzed by standard staining protocols, as described previously (27, 28). Briefly, deparaffinization from xylene to 95% alcohol and rehydration prior to microwave antigen recovery were carried out on a Leica Autostainer (Leica, Inc., Deerfield, IL). The deparaffinization process included a 30-min methanolic peroxide block for endogenous peroxidase activity. The antigen recovery step was carried out in an 800-W microwave oven in 200 ml of citrate buffer (0.01 m, pH 5.5–5.7). The immunostaining was performed on a Techmate 1000 (Ventana Medical Systems, Tucson, AZ) using capillary gap technology. Tissue sections were exposed for 10 min to normal horse serum (1:50), followed by overnight primary antibody incubation (1:2000). The primary antibody was a polyclonal rabbit immunoglobulin raised against gstFhit protein (29). The next day, slides were washed, incubated for 30 min with a biotinylated goat anti-rabbit IgG secondary antibody (1:200), and exposed to avidin-biotin complex for 45 min. After a wash, slides were subjected to three consecutive 8-min applications of DAB/peroxide solution. Treated slides were counterstained with hematoxylin and coverslipped prior to evaluation.

Immunohistochemical detection of p53, ER, PR, proliferation antigen Ki67, and ErbB2 receptor expression was by the same method using the following antisera and dilutions: ER, mouse monoclonal antibody ID5 from Dako Corp. diluted 1:80; PR, monoclonal antibody IA6 from Novocastra, 1:80; Ki67, antibody clone MIB1 from Coulter-Immunotech at 1:100; p53, clone 1801 from Novocastra, 1:80; and ErbB2, clone CB11 from Signet Laboratories, 1:100.

Expression of Fhit in sporadic and familial breast cancers from Iceland was studied using a tissue microarray method (30). This technique has the advantage of minimal tissue requirements and is therefore especially useful for studying rare and valuable tissue specimens, such as tumors from patients carrying mutations in cancer susceptibility genes. Briefly, small tissue cores (diameter, 0.6 mm; length, 2–3 mm) were retrieved from morphologically representative sites in the tumors and subsequently arrayed to a blank paraffin block at a high density using a custom-built instrument. Sections from this block were cut with an adhesive-coated tape sectioning system (Instrumedics, Inc., Hackensack, NJ) and stained with the Fhit antibody. Sections were deparaffinized, rehydrated, and rinsed in tap water before antigen retrieval by heating in a 0.01 m citrate buffer (pH 6.0) twice for 5 min at 850 W. Sections were incubated with the gstFhit antibody overnight at room temperature. Immunohistochemical staining was visualized using the StreptABComplex/HRP Duet (Mouse/rabbit from DAKO) according to the manufacturer’s instructions. Fhit expression in normal breast epithelial cells and stromal cells, respectively, served as positive and negative controls. Tumors were graded by intensity of staining as negative (−), weakly positive (+), moderately positive (++), and strongly positive (+++). The standard tissue sections of the 50 sporadic Philadelphia cases frequently showed heterogeneity for Fhit expression and were thus scored as heterogeneous (++/−, +/−, −/++, −/+, etc.; if the plus sign occurs first, >50% of the cancer cells were positive for Fhit expression; if the minus sign occurs first, the majority of cancer cells were Fhit negative). Examples of the scoring system were exchanged between laboratories to assure consistency in scoring for the study.

The PCR primers used for the microsatellite marker analysis were localized at 3p14.2 within the FHIT gene: D3S4260 (intron 3), D3S1461 (intron 4), D3S2757 (intron 4), D3S1300 (intron 5), and D3S1234 (intron 5). DNA samples (25 ng) were subjected to PCR analysis in a total volume of 25 μl. DynaZyme™ polymerase (Finnzymes Oy, Espoo, Finland) was used in the buffer solution provided by the manufacturer with 100 mm of each deoxynucleotide triphosphate and 0.25 mm of primers. After 5 min denaturation at 94°C, samples were subjected to 35 cycles of amplification, consisting of 30 s at 94°C, 30 s at 55°C, and 1 min at 72°C, followed by a final extension for 10 min at 72°C. The PCR products were denatured in formamide buffer, separated on 6.5% polyacrylamide sequencing gels, and transferred to a Hybond-N+ nylon membrane (Amersham, Aylesbury, United Kingdom). Hybridization and visualization of the PCR products were based on a peroxidase-labeled probe (ECL kit; Amersham). LOH was evaluated visually by comparing the intensity of alleles from normal and tumor DNA. Absence or significant decrease in the intensity of one allele relative to the other was considered an allelic imbalance.

For comparison of Fhit expression with other tumor markers (ErbB2, ER, PR, Ki67, and p53) in sporadic breast cancers, the Fhit scores were correlated using Spearman correlations. The ranking of the Fhit expression codes was as in Table 1. Spearman (nonparametric) correlations assess monotinicity of relationship, not necessarily linearity. The Ps are two-tailed and used the asymptotic method but were checked by an approximate method for small sample sizes (StatXact 3.1; Cytel Corp., Cambridge, MA).

Wilcoxon rank sum tests were used to compare Fhit expression observed in the BRCA2 999del5 group to the control group of sporadic tumors and to compare Fhit expression in tumors with 3p14.2 LOH to those retaining LOH. Ps were calculated using a method appropriate for small sample sizes with numerous tied scores (StatXact).

To determine whether the FHIT gene might be a target of breast cancer-specific chromosome 3p alterations, we have examined formalin-fixed tissue sections of 50 consecutive breast cancers, 10 DCIS, and 40 invasive cancers, which were resected in 1997 at the Thomas Jefferson University Hospital in Philadelphia. Sections from each case were tested for expression of ER, PR, Ki67, p53, ErbB2, and Fhit by standard immunohistochemical detection using antisera specific for the individual proteins. Because normal breast tissue is included in most of these sections, it serves as internal control and showed that normal breast epithelial cells are uniformly strongly positive for Fhit expression (+++). Results for the 40 invasive carcinomas are summarized in Table 1, which lists each case individually with diagnosis, grade, and expression level of the six proteins tested. The cases have been listed in descending order of Fhit expression, with those expressing levels equal to that observed in normal breast ductal epithelium (+++) at the top and those expressing very little Fhit (−/++ or −/+) at the bottom. Four cases or 10%, shown at the bottom of Table 1, were mostly negative for Fhit expression [60, 70, 85, and 90% of cells were completely negative; the remaining cells were weakly (+) or moderately positive (++)]. Another 16 cases (40%), shown below the horizontal line in Table 1, were partially negative (10–40% negative, with the remaining cells weakly or moderately positive) for Fhit expression. Cases that were uniformly strongly positive (+++, 38%) are shown at the top of Table 1 and are cases in which the FHIT locus is unaltered. The uniformly moderately positive cases (++; Table 1) we have considered unaltered in our statistical analysis, although it is possible that one FHIT allele has been inactivated in some of these five cases. Examples of expression of Fhit in selected cases are shown in Fig. 1. For the 10 noninvasive DCIS cases (data not shown), only one was mostly negative (−/++, 80% negative), one case was +++/−, and the others were all +++ (five cases) or ++ (three cases).

To determine whether absence or decrease of Fhit expression in invasive breast cancers correlates with expression of other tumor marker proteins or with clinical features, we have compared Fhit expression, by Spearman correlation, with expression of each of the other proteins tested (shown in Table 1) and have then determined whether diagnosis or grade affected the correlation. As shown in Table 2, when Fhit expression was compared with expression of the five biological tumor markers, no significant correlations were observed, although there was a trend toward higher expression of Ki67, a proliferation marker. In addition, although not evident in the correlation analysis, four of the five invasive tumors with significant expression of p53 were tumors that were partially Fhit negative.

As an initial characterization of the regulation and function of the FHIT gene in tumors from individuals carrying the BRCA2 999del5 germ-line mutation, we have compared these tumors with Icelandic sporadic tumors. In all, 92 Icelandic primary invasive breast tumors, 58 sporadic and 34 BRCA2 999del5, were stained with the Fhit antibody. Fig. 2 shows examples of Fhit antibody staining, and Table 3 summarizes the results. The tumors in individuals carrying the BRCA2 germ-line mutation showed reduced Fhit expression (P = 0.002). They were more frequently negative or weakly positive with respect to Fhit antibody staining than sporadic tumors, 32 and 14%, respectively, and only 18% of the BRCA2 999del5 tumors but 48% of sporadic tumors show strongly positive Fhit expression.

Of the 92 samples subjected to immunohistochemical analysis, DNA from 58 of them was available for LOH analysis. LOH was detected in 7 of 29 (24%) sporadic tumors and in 22 of 29 (76%) BRCA2 tumors. Among BRCA2 tumors, none of the seven tumors with retention of heterozygosity exhibited negative or weakly positive Fhit expression compared with 10 of 22 (45%) with LOH. A significant difference (P = 0.001) in levels of Fhit expression was observed in tumors with loss and retention of heterozygosity (summarized in Table 4). Forty-eight % of tumors with retention of heterozygosity at the FHIT locus showed the strongly positive staining pattern typical of normal breast epithelium, whereas only 14% of the tumors showing LOH were strongly positive. Similarly, only 8% of tumors with retention of both FHIT alleles exhibited negative or weakly positive Fhit expression, whereas 38% of tumors with LOH have this pattern of staining. Tumor DNA was necessarily isolated from a different population of tumor cells than were tested for Fhit expression. Because there is heterogeneity in expression of Fhit in subpopulations of tumor cells, as shown above for sporadic breast cancers, such heterogeneity probably accounts for the lack of concordance between LOH and Fhit expression results in individual tumors.

The microsatellite markers used, D3S1234, D3S1300, D3S2757, D3S4260, and D3S1481, are located at 3p14.2 within the FHIT gene. Fig. 3 is an example of LOH analysis showing results from tumors where breakpoints are detected within the FHIT gene, and Fig. 4 shows complete results of tumors showing LOH in comparison with Fhit expression. In most tumors showing LOH with intragenic FHIT microsatellite markers, there is concordance among markers in the sense that LOH is detected at all informative loci. Two tumors show retention of both D3S1481 alleles and LOH with other informative markers; two tumors show LOH with the D3S1481 marker and retention of both alleles at other informative loci within the FHIT gene, indicating breakpoints within the FHIT gene in these tumors.

Thus, we observed association of 3p14.2 LOH with reduced Fhit expression, as well as lower expression and higher 3p14.2 LOH in BRCA2 999del5 tumors than in sporadic tumors.

In this study, we report elevation in LOH at the FHIT locus at chromosome 3p14.2 and reduced Fhit antibody staining in tumors of BRCA2 999del5 carriers, compared with tumors from individuals without this mutation. We found reduced Fhit expression levels in both sporadic and BRCA2 999del5 tumors but significantly lower levels in the BRCA2 999del5 group. Elevation of 3p14.2 LOH in BRCA2 999del5 carriers is in line with earlier findings (1, 10, 11). Decreased Fhit expression was associated with LOH in the FHIT region. Our results suggest that FHIT could play an important role in hereditary breast cancer due to loss of gene copy number, with resulting loss of Fhit protein expression from deleted alleles.

Although the discovery of the BRCA1 and BRCA2 genes for familial breast cancer has led to exciting new clues into normal cell proliferation and double-strand break repair, it was a surprise that mutations in these two genes are not observed frequently in sporadic breast cancers (31), as might have been expected by parallels to previously described germ-line tumor suppressor gene mutants. Because of the association of these gene products with repair complexes (Refs. 24, 25, 26; reviewed in Ref. 32) and with each other (33), it is likely that these genes have a caretaker function, much as mismatch repair genes, MSH2 and MLH1, perform a caretaker function in prevention of HNPCC. Absence of Msh2 or Mlh1 protein in HNPCCs leads to genome-wide microsatellite instability that can be detected as shifts in mobility of microsatellite alleles on polyacrylamide gels, compared with constitutive alleles. Loss of Msh2 or Mlh1 protein thus causes HNPCC indirectly through unrepaired or misrepaired damage to genes involved in regulation of cellular growth or invasion. It is possible that loss of wild-type BRCA2 also contributes to cancer indirectly by failure to correctly repair double-strand chromosome breaks. Breast carcinomas from individuals carrying germ-line BRCA1 or BRCA2 mutations show increased frequencies of chromosomal abnormalities when compared with sporadic breast carcinomas. In a genome-wide survey by comparative genome hybridization, Tirkkonen et al. (12) showed distinct somatic alterations in tumors of BRCA1 and BRCA2 carriers compared with unselected sporadic breast cancers. There were about twice as many chromosomal changes in the hereditary cancers, but different chromosome arms were affected in the two types of familial cancers, with losses of 5q, 11q, 4p, 2q, and 12q occurring much more frequently in BRCA1 tumors than in sporadic and a higher frequency of 13q and 6q losses in BRCA2 tumors. BRCA2 tumors also showed elevated frequency of gains of 17q22-24 and 20q13. Small allelic deletions would not be detected by comparative genome hybridization analysis, but LOH analyses comparing BRCA2 cancers to sporadic cancers showed that the frequency of LOH at some chromosome regions was similar (1p, 16q) and at others significantly different (3p, 6q, 11p, 11q, 13q, and 17p; Ref. 11). These chromosome regions have been implicated in sporadic breast cancers at lower frequency (11). A majority of BRCA2 mutant carrier cancers show LOH at each of the loci. Interestingly, the chromosome arms 6q, 11p, 11q, and 13q each show three common fragile sites, whereas 3p has one in 3p14.2 within the region of highest loss on 3p; 17p does not exhibit a common fragile site. It is likely that each of these regions of high LOH contains at least one gene that is a target of loss; perhaps TP53 on 17p, BRCA2 at 13q, and FHIT at 3p14.2. Conversely, 16q exhibits two common fragile sites near a region of loss in sporadic breast cancers, but 16q LOH was not elevated in the BRCA2 mutant cases. Thus, presence of a common fragile site alone cannot explain elevated LOH frequencies.

We have examined LOH and expression of the FHIT gene in the BRCA2 999del5 familial versus sporadic cancers and have found that Fhit expression is reduced in a large fraction of sporadic breast cancers and in at least twice as many familial cancers. Our results may reflect the inability of mutated Brca2 to participate in the Rad51-mediated repair. Possibly due to infidelity in mitotic recombination and corresponding DNA repair, the BRCA2 999del5 cases follow a more aggressive pathway of chromosome damage, where fragile sites in the genome could be a hot-spot target. The numerous somatic FHIT gene deletions detected in BRCA2 carriers indicate that germ-line mutation of the BRCA2 gene results in accelerated accumulation of secondary somatic genetic changes at this locus in the tumors. These genetic aberrations result in reduced expression of a putative tumor suppressor gene, FHIT.

In sporadic breast cancer, ∼40% showed reduced expression of Fhit. In the study of familial versus sporadic cancer, Fhit expression was assessed on microarrays constructed using one sampling site per tumor. Therefore, it was not possible to evaluate the regional heterogeneity of Fhit expression. It is likely that some of these tumors are heterogeneous for expression of Fhit, as are the sporadic Philadelphia breast cancers, which could explain why the LOH data are not completely concordant with the Fhit expression data (shown in Fig. 4), i.e., one portion of a tumor could show LOH with concomitant Fhit reduction, whereas another portion might show very strong Fhit expression. In this study, in agreement with the previous LOH study, expression of Fhit was reduced in a significantly higher fraction of familial cancers relative to sporadic cancers analyzed in the same microarray experiments. Thus, it is likely that FHIT inactivation contributes to clonal expansion in both sporadic and familial breast cancer. Because there is a trend toward higher Ki67 and p53 values in breast cancers with reduced Fhit expression, it is possible that Fhit loss contributes to aggressivity of these cancers; however, studies of larger panels of breast cancers will be necessary to confirm these trends.

In a previous study of lung carcinoids (28), we have shown that normal bronchial epithelium is strongly positive for Fhit, typical (benign) carcinoids with minimal LOH on 3p were strongly positive, small cell lung cancers were completely negative, whereas atypical carcinoids (malignant) show extensive LOH in FHIT and moderate to weak Fhit expression. We concluded that loss of one FHIT allele leads to reduced Fhit expression. Similarly, weak/moderate Fhit expression may result from allelic loss at the FHIT locus in sporadic and familial breast cancers, suggesting that loss of one FHIT allele may contribute to clonal expansion in breast tissue.

Thus, our interpretation of molecular events at the FHIT locus in breast cancer is: one fragile FHIT allele is broken and misrepaired, resulting in deletion within one FHIT gene and down-regulation or loss of expression of that allele; reduction in Fhit expression allows a slight clonal expansion of Fhit ++ or + cells, increasing the opportunity for breakage and misrepair of the second FHIT allele and resulting in expansion of a subpopulation of Fhit negative cells. The tumor in this example would then be a ++/− or +/− tumor.

The defects in the FHIT gene described in this study are likely to be helpful in the understanding of the somatic genetic progression pathways that contribute to the development of malignancy in genetically predisposed individuals.

We thank professor Jonas Hallgrimsson of the Department of Pathology, University Hospital of Iceland, for his unfailing support and for help providing pathological material.