Allelic Loss and Microsatellite Alterations of Chromosome 3p14.2 Are More Frequent in Recurrent Cervical Dysplasias¹

Whereas evidence has been mounting in recent years establishing HPV³ as the etiological agent in cervical cancer, it is clear that other factors are involved because the majority of patients with HPV infections do not develop invasive lesions. As is well known, the natural history of cervical dysplasia is characterized by regression in the majority of mild and moderate dysplasias and progression only in a minority of mild, moderate, and even severe dysplasias (1).

It is widely accepted that there are multistep molecular changes leading to malignant transformation from preneoplastic lesions to invasive tumors. However, the sequence of molecular events responsible for cervical carcinogenesis has not yet been elucidated. To date, there are no reliable clinical or molecular biomarkers to predict an individual’s cervical dysplasia progression risk. The current management strategies are based on outcome statistics and patient compliance issues. This current system often leads to expensive and uncomfortable evaluations and overtreatment for millions of women.

Numerous studies in recent years report a high LOH rate in different chromosomal regions, indicating hot spots for potential TSGs in cervical carcinogenesis (2, 3, 4, 5). Some of these losses were also demonstrated in the dysplastic lesions associated with the invasive tumors, suggesting that the LOH in these critical regions may be an early event (2, 3, 6, 7). Our group and others have demonstrated that the FHIT gene, a putative TSG that overlaps FRA3B in chromosome 3p14.2, is frequently altered in cervical cancer, whereas LOH, homozygous deletions, and aberrant transcripts are frequently detected in these invasive cancerous lesions (8). Despite these interesting molecular findings, the lack of protein-altering mutations and tumor heterogeneity brings into question the exact role of the FHIT gene as a TSG in cervical carcinogenesis.

The objective of this study was to conduct molecular analyses on a cohort of recurrent and nonrecurrent cervical dysplasia patients in an attempt to define molecular predictors. Specifically, we performed HPV and allelotyping analyses within these two cohorts, using 16 polymorphic markers covering previously defined regions of loss on chromosomes 3p, 5p, 6p, 9p, 11q, and 17p. Our goal was to improve our understanding of dysplasia recurrence with the hope of defining molecular biomarkers to predict progression risk.

Paraffin-embedded histology slides cut from the cervical cone biopsy specimens of 19 subsequent biopsy-proven recurrent cervical dysplasia cases and 18 appropriately followed nonrecurrent cases were retrieved from the Departments of Pathology of Tulane University Hospital and the Parkland Health and Hospital System from 1993–1995. All patients were followed regularly with Pap smears and colposcopy, when indicated, for 3.5–5 years from the time of their initial cone biopsy. Cases were reviewed by two skilled gynecological pathologists (N. D. and R. A.),and most severe areas of dysplastic epithelium were identified for precision microdissection.

Precision microdissection was performed on the targeted dysplastic regions above the basement membrane to obtain a pure population of dysplastic cells without normal stromal or infiltrating lymphocyte contamination. Approximately 1000 cells were microdissected for each case. Matched normal stroma was dissected and used as constitutional normal DNA for each case. DNA was extracted using proteinase K digestion as described previously (9).

The presence of HPV DNA in the dysplasia specimens was determined by PCR analysis using general and type-specific primers for HPV. The general primer pairs (GP5+/GP6+) target the HPV L1 open reading frame,and they detect a broad range of genital HPVs (10). The type-specific primer pairs (TS-16 and TS-18) were used to identify the oncogenic HPV-16 and HPV-18, respectively (11). HPV analysis of the microdissected specimens was performed using the modified two-round PCR from the method described previously (2), in which the product of first-round PCR was used as template and amplified using the same primers. DNA made from near confluent cultures of the human cervical carcinoma cell lines CaSki(HPV-16) and HeLa (HPV-18; American Type Culture Collection, Manassas,VA) was used as positive controls.

To evaluate LOH and MAs within both cohorts of cervical dysplasia, a panel of 16 microsatellite markers was used covering the following chromosomal regions: (a) 3p21.2–24.2 (D3S1351,ITIH, and D3S1478); (b) 3p14.2(D3S1234 and D3S1600); (c)5p15.1–15.2 (D5S406, D5S208, and D5S807);(d) 6p21.2–24 (D6S277, D6S105, and D6S291); (e) 9p21 (IFNA and D9S1748); (f) 11q23.3 (D11S490 and CD3D); and (g) 17p13 (p53CA). These primers were selected for their reported high LOH rate (2, 3, 4, 5). The sequences were obtained from the Genome Database⁴ or the respective reference and synthesized by Life Technologies, Inc.(Gaithersburg, MD).

A two-round multiplex PCR method was used to amplify the microsatellite markers from microdissected cells as described previously (12). Multiplex (two to six primer sets in each reaction)PCR was done during the first amplification, followed by uniplex PCR incorporating [³²P]dCTP using the optimal annealing temperature for each individual marker. These results were confirmed in duplicate. The final products were separated on a 6%denaturing polyacrylamide gel and subjected to autoradiography. Areas of stroma that contained the highest density of infiltrating lymphocytes were used as constitutional normal DNA for each case. The sizes and relative intensities of alleles from the normal stroma and invasive tumors were compared directly. Markers that identified two distinguishable alleles of different sizes but similar intensity in the lane having constitutional normal DNA were termed “informative”(heterozygous). Markers that gave a single major band in the normal DNA were termed “noninformative” (homozygous). LOH was scored by visual detection of the complete absence of the upper or lower allele. Markers that gave shifted bands or expanded bands as compared with the normal control DNA were scored as having MAs.

FRL (FRL = the number of patients with LOH in that region/the number of patients informative in that region) was used to evaluate LOH. If any marker for a region was informative, the region was regarded as informative. If one or more markers showed LOH, we regarded the region as demonstrating loss. We used the RMA (RMA = the number of patients with MA in that region/the total number of patients)to evaluate the regional instability, and we used the MA index (number of markers with MA/total number of markers evaluated per patient) to evaluate the overall genomic instability. The differences in the FRL rate and the RMA rate between the recurrent and nonrecurrent groups were analyzed by a two-sided χ² test or by Fisher’s exact test. The difference in the MA index between the recurrent and nonrecurrent dysplasias was analyzed by Wilcoxon’s rank-sum test. P < 0.05 was considered significant.

The mean age of the patients was similar in both groups (Table 1). African Americans were predominantly represented in each group (n = 15), leaving an equal racial distribution. High-grade intraepithelial lesions (CIN II, CIN III, or carcinoma in situ) accounted for 18 cases in each group, with only one case of CIN I seen in the recurrent group. Of interest, there were four cases of recurrent dysplasias in women known to be infected with HIV, but none in the nonrecurrent group. However,the presence of HIV sequence was not determined molecularly in these specimens. The positive surgical margin in the conization specimens was also distributed equally between the two groups (Table 1).

Overall, HPV was detected in 18 of 19 cases (95%) in the recurrent group versus 15 of 18 (83%) cases in the nonrecurrent group(P = 0.264; Table 1). Seven of 19 (37%) recurrent cases were positive for HPV-16 and/or HPV-18, whereas 7 of 18 (39%)cases in the nonrecurrent group were positive for HPV-16 and/or HPV-18(P = 0.90). Three cases in the nonrecurrent group demonstrated coinfection with both oncogenic HPVs as compared with only one case in the recurrent group. Cases positive for HPV general primers but negative for HPV-16- or HPV-18-specific primers were not further classified. Fig. 1 demonstrates the PCR-based HPV subtyping results.

Of the seven chromosomal regions evaluated in all of the dysplasia cases, 5p15.1 demonstrated the highest regional loss rate of 42% (14 of 33 informative cases). With a trend toward significance, LOH at any 5p15.1 marker was greater than twice that in the nonrecurrent group(P = 0.062; Table 2). When comparing the regional loss rate between the recurrent and the nonrecurrent groups, only the FHIT region at 3p14.2 demonstrated a significantly higher loss rate in the the recurrent dysplasia group(P = 0.0048). Fig. 2shows examples of the allelotyping analysis with LOH and MAs.

Of the seven chromosomal regions evaluated, 5p15.1 demonstrated the highest RMA rate of 43% (16 of 37). When comparing the two groups, the FHIT region at 3p14.2 and the p53 region at 17p13 demonstrated a significantly higher RMA rate in the recurrent group(P = 0.009). When we examined the MA index of each individual case, overall, there is more MA in the recurrent group than in the nonrecurrent group (P < 0.0025; Fig. 3; Table 3).

Although it is a causative agent in most invasive cervical cancers and high-grade dysplastic lesions, HPV is only one of multiple somatic cell errors that leads to cervical cancer. The elucidation of the molecular pathogenesis can eventually lead to the development of clinically useful risk assessment biomarkers, which can help select those patients at risk for neoplastic progression and may allow individualized follow-up and treatment for those identified as being at risk. In addition, intermediate molecular markers will allow rapid advances in the field of chemoprevention for those at risk for cervical dysplasia.

Many studies have documented that allelic losses are common in various chromosomes in cervical cancer. Some chromosomal arms consistently have higher loss rates, indicating hot spots for potential TSGs (2, 3, 4, 5, 7, 13). Of all of the candidate loci, only 3p14.2 has been studied extensively. The FHIT gene, a putative TSG that overlaps the common fragile site FRA3B in chromosome 3p14.2, is frequently altered in cervical cancer. This is also the location where LOH, homozygous deletions, and aberrant transcripts are frequently detected in cancerous lesions. FRA3B, the most common of the constitutive aphidicolin-inducible fragile sites, is a specific chromosomal region prone to forming gaps or breaks and is associated with recurring abnormalities in tumors (8).

This study examined the HPV status and the allelic status of a cohort of patients with recurrent and nonrecurrent cervical dysplasia with an average of 3–5 years of follow-up. It is not surprising that the HPV rates are similar in both groups because HPV appears to be the earliest event in the pathway to cervical carcinogenesis. Numerous studies have shown that HPV infection is probably the initiating event in the development of cervical dysplasia and that other factors are responsible for persistence, recurrence, or progression. Although there is the possibility of a lack of statistical power in this cohort study to observe the difference in HPV rate between the two groups, the HPV findings here are consistent with the high rates of HPV infection in high-grade dysplasias reported previously.

It is interesting that in the LOH analysis, only the 3p14.2 FHIT region showed a significantly higher loss rate in the recurrent group than in the nonrecurrent group. It is not known whether this reflects the importance of the FHIT gene as a TSG in dysplasia destined to recur or whether it merely reflects a higher level of instability within the FRA3B fragile site. Chromosomal region 5p15.1 demonstrated the highest overall loss rate in both groups but was not statistically different between these small cohorts. The importance of the 5p15.1 locus is not ruled out by the lack of statistical significance that may reflect the small sample size or the relatively short follow-up period of 3–5 years. Of interest, LOH at 5p15.1 was more common than that seen within the 3p14.2 region, which is lower than we reported previously (2). It is possible that the lower rate of LOH seen within 3p14.2 could be attributed to the choice of markers used in this study (markers widely flanking FHIT versus the intragenic markers D3S1300 and D3S4103) or the geographic and racial differences in the cohorts. It is also possible that epigenetic or environmental differences in certain cohorts can elicit different molecular mechanisms that result in the same phenotype of cervical dysplasia.

Our study also examined the significance of MAs in these cohorts. The exact functional significance of MA has not been determined; however,there is evidence that HIV-positive patients have a higher rate of MA and greater genomic instability that may play a critical role in the development of HIV-associated cancer (12). The mean MA index in our study is significantly higher in the recurrent group than in the nonrecurrent group, reflecting the underlying genomic instability. Our recurrent group did include four HIV + patients, and this group also had a higher mean MA index of 0.218 as compared to the nonrecurrent group. One of four HIV + patients had both MA and LOH in chromosomal regions 3p14.2 and 5p15.1. It is possible that the HIV-related genomic instability contributed to the higher loss rate in the recurrent group; however, because the number of HIV+ cases is small, we did not attempt to stratify our data according to HIV status.

In the MA analysis, the FHIT locus at 3p14.2 again demonstrated a significantly higher MA rate in the recurrent group. Similar to the LOH rate, chromosomal region 5p15.1 demonstrated the highest overall rate of MA; however, the difference was not statistically significant. Of interest, the other significant regions of difference regarding MA in the recurrent cohort involved the markers within the p16 and p53 TSG regions. Combining the LOH and MA findings,we speculate that the two processes may be related and may act synergistically to inactivate critical TSGs. Conversely, regional instability reflected by the difference in MA may lead to allelic loss and subsequent development of neoplasia.

In summary, we have demonstrated that HPV status alone does not predict recurrence. A higher LOH rate in the FHIT region of the recurrent cases and a significant rate of MA at the same locus suggest an important role for the FHIT gene in cervical carcinogenesis. Furthermore, we demonstrated that the recurrent dysplasia group had higher RMAs and overall MAs, reflecting their underlying genomic instability, which contributes to allelic loss, thus increasing the risk for recurrence or progression. These findings appear to provide evidence for genomic characteristics and chromosomal locations implicated in HPV-associated cervical cancer and in particular to the dysplasias that are destined to recur.