INTRODUCTION

Intratumoral heterogeneity is the hallmark of human tumors. With tumor progression, subpopulations of neoplastic cells with different biological characteristics successively evolve within a tumor arising from a single progenitor cell (1). Although tumor cells are morphologically similar to each other, they may differ in biological features. Intratumoral heterogeneity has been investigated karyotypically and cytometrically. Data on intratumoral heterogeneity of genetic alterations are still limited. Generally, genetic changes in a tumor are divided into three types (2): (1) important and specific, (2) related to tumor progression, and (3) random and neutral. Analysis of intratumoral genetic heterogeneity will provide an important clue for exploration of the important genetic events for initiation and development of human cancers.

Genetic alterations have been suggested to be an essential etiologic factor contributing to the development of cervical cancer, either independently or in combination with infection of oncogenic types of human papillomavirus (HPV; 3, 4, 5, 6). Intensive cytogenetic and genetic studies have revealed several chromosome arms showing frequent aberrations in cervical cancer; in other words, deletions (5, 7), translocations, or amplifications (8). Among these chromosomal abnormalities, 3p deletions have been shown to be the most frequent event in cervical carcinogenesis, suggesting that this chromosome arm may harbor genes crucial for the development of cervical cancer (5, 7, 9). Different studies have suggested several regions on 3p to be frequently deleted in cervical cancer (10, 11, 12). However, it is still not clear whether single or multiple regions are related to the initiation of cervical cancer. Investigation of intratumoral heterogeneity of 3p deletions will dissect this genetic event in detail, providing a better understanding on roles of 3p deletions in development of cervical cancer and favoring the identification of tumor suppressor loci on 3p.

Investigation of intratumoral heterogeneity of genetic alterations provides not only an important insight into genetic processes but also clues to the clonal origin of human tumors (13, 14, 15). Knowledge of clonality of tumors is greatly relevant to the understanding of human carcinogenesis. Proponents of mutational theory of tumor development suggest that tumors arise from a series of mutations occurring in one cell and its progeny (1). This theory has been strongly supported by the majority of studies, which have shown that various human tumors are monoclonal. However, others have argued that tumors are not monoclonal in origin but require the interaction of multiple cells and that outgrowth of a dominant clone during the subsequent clonal competition or selection accounts for their apparent monoclonality (16, 17). This argument has been supported by the findings that hereditary forms of colonic adenomas are indeed polyclonal in origin (18).

Major evidence of clonality status of human tumors has been accumulated by analysis of X chromosome inactivation pattern in tumors compared with that in normal background tissue (19). This approach takes the advantage of mosaic distribution of either maternal or paternal X chromosome inactivation in normal female somatic cell, according to Lyon's hypothesis. A tumor showing one type of X-inactivation pattern would be implied to originate from a single transforming cell, whereas a tumor showing two types of X-inactivation pattern would be concluded to be biclonal or polyclonal in origin. Although the combination of innovative markers and sensitive PCR technique has greatly enhanced the sensitivity and heterozygotic populations for clonality analysis, one of the major drawback in previous clonality studies is that all X-inactivation approaches applied are only able to detect the dominant clones that give rise to the major signal in analysis (20, 21). The signal from less competitive clones would be covered by that from either major clones or contamination of normal tissues. Approaches to detect less dominant clones against the background of dominant clones would be essential in exploring the true nature of clonality of human tumors. A solution to this task is to analyze the X-inactivation patterns on different samples microdissected from different areas within the same tumor.

Cervical cancer is one of a few forms of human cancers in which exogeneous factors play an important role in the pathogenesis (3, 22). Clonality analysis would provide an important insight to the understanding of the role of HPV in cervical carcinogenesis. Several studies have investigated the clonality status of cervical cancer, and mainly monoclonal origin was found (23). Although polyclonality has been reported in a few cases of cervical cancers, possible contamination from normal tissues could not be excluded (24).

In this study, we analyzed intratumoral heterogeneity of 3p deletions in cervical invasive cancer. In parallel, by using genetic deletions of chromosome 3p and X-chromosome inactivation as markers, we analyzed the clonality patterns on multiple well-microdissected tumor loci from each case. Two of eight cervical cancers with 3p deletions were found to contain different clones according to different X-chromosome inactivation and genetic deletion patterns and thus provided the first clear evidence that cervical cancers could be of polyclonal origin.

MATERIALS AND METHODS

Cases

Fourteen cases of invasive cervical cancers were included in this study. All specimens were collected from Blokhin Cancer Center, Moscow, from 1993–1997. Clinical diagnosis of selected cases were further confirmed before their inclusion into the study. All tumors were squamous cell carcinoma. In five cases, single or multiple synchronous CIN lesions were found and analyzed. Clinical data of each individual case are listed in Table 1. HPV status of the collected cases was analyzed in our previous study (25). Tumor specimens were collected as punch biopsies from surgical extirpated cervical cancers. Fresh tissues were snap-frozen in liquid nitrogen immediately after sampling from surgical specimens and then transferred to a −70° C environment for long-term storage.

TABLE 1 Clinical, Virological, and Pathological Data Of Patients With Invasive Cervical Squamous Cell Cancer
Full size table

Microdissection and DNA Preparation

A 10-μm cryostat section was cut from each tissue block. Each section was stained with one or two drops of methylene blue for 7 seconds, washed with running deionized water for a few minutes, and then dried at room temperature. Areas of lesions, morphological normal squamous epithelia, and surrounding normal stroma/epithelia were dissected directly from the stained slides under the microscope with a fine surgical scalpel. To ensure the accuracy of the test, two normal control tissues were microdissected from each case and analyzed in parallel. Number of lesional samples from each tumor varied according to the size of individual tumor. Precautions were taken to prevent cross-contamination, including change of the surgical instruments from section to section and individual rinsing of the slides.

Microdissected lesions and normal control tissue were transferred into Eppendorf tubes containing 50 μL 1× PCR buffer and proteinase-K (500 ug/mL, Beri Mahamnn, Germany). The samples were incubated overnight at 56° C. Reaction was stopped by heating the samples at 95° C for 10 min.

Loss of Heterozygosity Analysis

Twelve microsatellite markers flanking the frequently deleted regions on 3p reported in cervical cancer were selected for analysis. The genetic information of selected markers was obtained from the Genome Database (John Hopkins University, http://gdbwww.gdb.org/gdb/). Two microliters of each DNA sample were amplified in 10 μL of PCR reaction mixture containing 1× standard PCR buffer, 1.5 mm MgCl2, 200 um each of deoxynucleotide, 1 unit of Taq gold DNA polymerase, and 0.05 μL fluorescence-labeled dUTP. Amplification was performed for 35 cycles (45 s at 95° C, 30 s at 55° C, and 1 min at 72° C), with a initial denaturation step of 10 minutes at 95° C and a final step of 7 minutes at 72° C. Reagents for PCR amplification in addition to the primer sets were purchased from Perkin-Elmer (Perkin-Elmer Cetus, Norwalk, CT). Each microsatellite marker was amplified individually by independent amplification. Efficiency and specificity of PCR amplification were checked by agarose gel electrophoresis and ethidium-bromide staining.

One and a half microliters of amplification products were mixed with 2.5 μL formamide, 0.5 μL Rox internal size standard, and 0.5-μL loading buffer. The mixture was heated at 95° C for 5 min and then immediately chilled on ice. One or two microliters of mixture was loaded and electrophoresed in an ABI 377 DNA sequencer (Perkin-Elmer Cetus) for 2 hours. Collected data were analyzed with Genscan software. In heterozygotic cases, signal intensity of major peak was compared between normal control DNA and analyzed tumor DNA, and loss of heterozygosity (LOH) was determined by more than 70% of signal reduction from the affected allele in tumor DNA compared with normal control.

PCR amplification and Genscan analysis were repeated on the samples showing LOH for confirming the LOH patterns.

X-Chromosome Inactivation Analysis

By using a microsatellite within the human androgen receptor gene as a marker, X-chromosome inactivation analysis was performed on case M3, M21 and M23 that showed different patterns of genetic deletions on multiple microsatellite markers on 3p. Ten microliters of sample DNA were submitted to protein precipitation by the procedure introduced in our previous study (26). DNA from the supernatant was precipitated by adding 2×-concentrated alcohol and was incubated at −70° C for at least 2 hours. After centrifugation at 10,000 × g for half an hour, the invisible pellet was redissolved in 10 μL of dual distilled water. Before PCR amplification, the samples (10 μL) were submitted for methylation-sensitive restriction enzyme treatment in 20 μL containing 20 units of HpaII and 1× reaction buffer (New England Biolabs Inc., Beverly, MA). The mixtures were incubated at 37° C overnight with continuous shaking. Reaction was stopped by inactivation of HpaII at 70° C for 45 minutes.

Amplification of androgen receptor gene was performed on HpaII-treated DNA samples in a nest-PCR reaction. In the outer PCR run, a total of 20-μL DNA sample were amplified in 50-μL of reaction mixture containing 1× PCR buffer, 1.5 mm MgCl2, 200 um of each deoxynucleotide, 1 unit Taq gold polymerase, and 20 pm of each outer primer. Amplification was carried out on a Perkin-Elmer 9600 Thermocycler for 25 cycles with cycling parameters of 45 seconds at 95° C, 45 seconds at 60° C, and 1 minute at 72° C. The cycles were supplemented with an initial 10 minutes at 95° C for denaturation and with an ending step of 7 min at 72° C for full extension. An inner PCR run was performed on a 25-μL reaction containing 5 μL of PCR products from outer PCR, 10 pm of each inner primer (forward primer was labeled with 6-Fam), and other PCR reagents in the same concentration as in the outer PCR reaction mixture. The mixture was amplified for 35 cycles with the same cycling parameters as the outer PCR reaction, with the modification of annealing temperature from 60° C to 55° C. Efficiency of amplification was checked by agarose gel electrophoresis.

One and a half microliters of PCR products were mixed with 2.5 μL of formamide, 0.5 Rox size standard, and 0.5 μL of loading buffer. The mixture was heated at 95° C for 5 min and then immediately chilled on ice. One microliter of mixture was loaded and electrophoresed in an ABI 377 DNA sequencer (Perkin-Elmer Cetus), and data were analyzed by Genscan software.

To ensure the reproducibility of the results, the analyses were repeated from microdissection on case M3 and M21, which are heterozygotic for the androgen receptor gene microsatellite marker.

Interpretation of the Results for Clonality Analysis

In the cases heterozygotic for androgen receptor gene locus, a nonrandom X-chromosome inactivation pattern was recorded in the sample when 70% reduction of signal intensity from one allele compared with the signal from the other allele was observed after HpaII digestion. Nonrandom X-chromosome inactivation patterns in analyzed lesions were viewed monoclonally when matched normal cervical tissue showed a random X-chromosome inactivation pattern. Clonal composition of each individual tumor was assessed by comparing the X-chromosome inactivation patterns among different samples from the tested tumor.

RESULTS

Chromosome 3p Deletions in Cervical Cancer and Synchronous CIN

Genetic deletions of 3p were analyzed in multiple samples from individual tumors in 14 cases of cervical invasive cancer. Samples were microdissected from separate nests showing no direct contact with each other. Number of microdissected lesions varied according to the size of specimen and the number of cells available for microdissection. To ensure the accuracy of analysis, two normal cervical tissues from different areas of the specimen were microdissected and analyzed in parallel. In some cases, normal cervical squamous epithelia, if available, were obtained for analysis. Totally, 120 samples, including 78 invasive lesions, 10 adjacent CINs, 4 normal squamous epithelia, and 28 normal cervical tissues were prepared from 14 cases.

Losses of heterozygosity (LOH) involving one or more microsatellite markers on 3p were found in 8 of 14 cases (57%). Allelic losses were recorded as homogeneous (clonal) when all lesional samples from individual tumor showed losses of identical allele and as heterogeneous (subclonal) when LOH was observed in part of tested lesional samples or when losses of different alleles were detected among different samples from the same tumor. Table 2 shows the frequency of heterogeneous and homogeneous LOH at different genetic loci on 3p. No LOH was found in normal epithelial samples. All but three markers (D3S1285, D3S1276, and D3S1271) showed heterogeneous and homogeneous deletions in different percentage of tumor samples. D3S1285 had only heterogeneous deletions in the four LOH-positive tumors, and D3S1276 and D3S1271 showed no LOH in any tumor. Overall LOH (heterogeneous plus homogeneous deletions) as well as homogeneous LOH were most frequently detected at three regions: FHIT gene (D3S4103 and D3S1300), 3p31.2–21.1 (D3S1478), and 3p24.2–22 (D3S1283). The frequencies of overall LOH at these four markers were 60% at D3S1300, 54.6% at D3S4103, 45.5% at D3S1478, and 50% at D3S1283, and the frequencies of homogeneous deletions were 40% at D3S1300, 27.3% at D3S4103, 27.3% at D3S1478, and 30% at D3S1283. No tested marker on 3p showed homogeneous deletions in all cases. However, except for case M23, in which all positive markers showed heterogeneous patterns, all cases with LOH had homogeneous deletion at least at one of the four markers indicated above.

TABLE 2 Frequency Of Homogeneous and Heterogeneous Deletions of 3p in Cervical Cancer
Full size table

Ten CIN lesions synchronous with invasive cancers were found in five cases (M2, M4, M13, M21 and M23). In two of these cases (M2 and M13), no LOH was detected in invasive lesions as well as coexisting CIN. Simultaneous presence of LOH in invasive lesions and synchronous CIN was observed in three other cases (M4, M21, and M23). Case M4 showed identical patterns of allelic losses between invasive and CIN lesions, whereas case M21 and M23 had different patterns of allelic losses between different CIN and invasive cancer foci (Figs. 1 and 2; also see detailed description in the following paragraph).

FIGURE 1
figure 1

Electrophoretograms illustrating the patterns of 3p allelic losses and X-chromosome inactivation from Cases M3, M21, and M23. X-chromosome inactivation analysis was performed using the human androgen receptor gene as a marker (see Materials and Methods). In case M3, HpaII treatment removed the shorter AR allele in all lesional samples as compared with two normal control samples, indicating that all samples contained tumor cells from a single clone. In Case M21, HpaII treatment removed the shorter AR allele in Samples II, T3, and T5, whereas the longer AR allele was removed in Samples T2 and T4, indicating polyclonal origin of this tumor. Case M23 was homozygons at androgen receptor gene microsatellite locus and thus uninformative for X-chromosome inactivation analysis. However, patterns of allelic losses at multiple 3p loci indicate the presence of multiple cell clones in this tumor (see detailed description in Results). The texts on the left denote the number of samples (N1 and N2, two normal control tissue; II, IIa and IIb, topographically different CIN II lesions; T1-T9, different samples from invasive cancer). Topographical locations of different samples and detailed information of 3p allelic losses are shown in Figure 2.

Full size image
FIGURE 2
figure 2

Topographical locations and patterns of 3p allelic losses and X-chromosome inactivation from Cases M3, M21, and M23. Sample numbers are the same as in Figure 1. □, uninformative; &OV0097;, informative; &OV0098; and &OV0103;, LOH with shorter and longer allele deleted; X1 and X2 denote X-chromosome inactivation pattern with longer and shorter allele inactivated (methylated).

Full size image

Patterns of 3p Deletions

Among eight cases with LOH, five (M1, M4, M12, M18, and M25) had allelic losses with identical patterns at all or at a majority of LOH-positive markers among different samples from each individual tumor. Different patterns of deletions at half or more of LOH-positive markers were seen in 3 cases (M3, M21 and M23). An exemplified electrophoretogram, patterns of allelic losses, and topographical locations of different samples of these three cases is shown in Figure 1 and Figure 2. Case M3 had homogeneous LOH patterns at D3S1293, D3S1478, and D3S1289 and heterogeneous patterns at the other three markers (D3S1307, D3S4103, and D3S1300). Case M21 had heterogeneous losses at five markers (D3S1283, D3S1298, D3S1478, D3S4103, and D3S1285). Sample II, a CIN II lesion, and T5 had identical losses at all seven markers positive for LOH in this case. Sample T3 had LOH patterns at four markers that were identical to sample II and T5, whereas LOH was absent at three markers (D3S1283, D3S4103, and D3S1285). Samples T2 and T4 showed identical patterns of allelic losses with the other three samples above at two (D3S1307 and D3S1300) and three (D3S1307, D3S1298, and D3S1300) out of seven markers with LOH, respectively. Case M23 showed the most clearly discrete patterns of allelic losses among different samples. Sample IIa (CIN II) and T4 had no LOH at any marker tested. Sample IIb (CIN II), T3 and T9 inevitably showed identical patterns at all 5 markers with allelic losses, whereas LOH patterns from other 4 samples (T5, T6, T7 and T8) were identical at six of eight LOH-positive markers. Among eight markers with LOH in this case, patterns of allelic losses at seven markers in Sample IIb, T3, and T9 are different with Samples T5, T6, T7, and T8 showing no deletions at three markers (D3S1293, D3S1263, and D3S1283) and LOH involving different alleles at three markers (D3S1478, D3S4103, D3S1300, and D3S1285).

X-Chromosome Inactivation Analysis

Using human androgen receptor gene as a marker, X-chromosome inactivation analysis was performed on Cases M3, M21, and M23 where heterogeneous LOH of 3p was found in a half or more LOH-positive markers. Case M3 and M21 were heterozygotic for androgen receptor locus and thus informative for analysis, whereas M23 was uninformative. All lesional samples from M3 and M21 showed nonrandom X-inactivation patterns. Comparing monoclonality among different tested samples from individual tumor, identical patterns were found in all seven samples from M3, whereas different patterns were observed in different samples from Case M21 (Figs. 1, 2). In M21, Sample II (CIN II), T3, and T5 had identical patterns with the longer AR (androgen receptor) allele inactivated, whereas Samples T2 and T4 showed the shorter AR allele inactivated.

DISCUSSION

The investigation of intratumoral genetic heterogeneity provides an important clue to the genetic steps involved in tumor progression. More significantly, it makes it possible to distinguish the crucial genetic changes for tumor development from those incidental secondary alterations. Clonal genetic events are considered to be involved in the step of tumor initiation, whereas heterogenous genetic changes are likely incidental events or may be related to tumor progression. Cytogenetic and genetic analyses have suggested that deletions of chromosome 3p are the most frequent genetic events among the genetic alterations in cervical cancer (5, 7). Our study for the first time reported the intratumoral heterogeneity of chromosome 3p deletions in cervical cancer, describing not only the total prevalence of genetic deletions of 3p but also identifying the nature or patterns of the deletions. We observed allelic losses at one or more regions of 3p in 57% (8/14) of cases, supporting the previous finding that 3p deletions are a frequent genetic alterations in cervical cancer. More important, our results identified the clonal allelic losses at least at one tested marker in the majority of the cases (7/8) positive for LOH on 3p, suggesting the importance of 3p deletions in participation in clonal selection during the processes of cervical carcinogenesis. On the basis of analysis of the general incidence of allelic losses, previous studies have reported several regions on 3p with frequent allelic losses in cervical cancer. Although discrepancies still existed regarding the frequency of deletions and the common deleted loci, the most frequent deletions have been mapped to 3pter-p24.2, 3p24.2-p21.1, 3p21.2-p14.2 and 3p12.1–11 (10, 11, 27, 28, 29). In particular, the FHIT gene has been reported to be the genetic locus on 3p with the highest incidence of allelic loss (10, 30). However, it is not clear whether single or multiple loci on 3p are involved in an early stage of cervical malignant transformation. By a combination of multiple sampling and LOH deletion mapping, our current finding provided the first clue to this important genetic process. We detected three regions, 3p24.2-p22 (D3S1283, 30%), 3p21.3–21.2 (D3S1478, 27.3%), and the FHIT gene region (D3S4103 and D3S1300, 27.3% and 40%) on 3p with high incidence of clonal (homogenous) allelic losses. Moreover, allelic losses involving at least one of these three 3p loci were detected in the synchronous CIN lesions from all three cases showing LOH in invasive lesions. These results indicate that these regions, especially the FHIT gene, are most likely tumor suppressor loci that are essential for early onset of cervical cancer. Although clonal allelic losses were commonly detected at these 3p regions, none of these loci was exclusively homogenously deleted in all cases with LOH. However, seven of eight cases with LOH showed homogeneous deletion at least at one of these three regions. This finding proposes that more than one 3p loci is important for the early transformation in cervical carcinogenesis.

The most interesting finding in our current study was the clear genetic evidence for the polyclonal origin of cervical cancer. By analyzing the X-chromosome inactivation pattern on a single sample from each tumor, monoclonal origin of cervical cancer has been suggested by several previous studies (23, 31). Although a few cases of polyclonal cervical cancer have been reported (24), technical pitfalls such as contamination of normal tissue or failure of methylation-sensitive restriction enzyme cleavage cannot be excluded. By analyzing genetic deletions and X-chromosome inactivation patterns on multiple samples from each tumor, our study increased the chance of picking up a less dominant clone against the background of a more dominant clone. In our study, one case (M21) was clearly indicated to be of polyclonal origin by the evidence of heterogeneic 3p deletion patterns at multiple microsatellite markers as well as X-chromosome inactivation analysis. Another case (M23), which was uninformative at the androgen receptor gene locus, most probably contained at least two different clones because it is unlikely that such discrete patterns of genetic deletions occur coincidentally in tumor clones originating from a single cell. Therefore, our results indicate that cervical cancer could be polyclonal in origin.

Cervical cancer is considered to be one of a few forms of human cancer in which exogenous agents are important in their carcinogenesis. Previous studies have proven that cervical precancers could be polyclonal or multifocal in origin (32, 33). Polyclonality of cervical neoplasia strongly suggests “field defect” as an important etiological factor in cervical carcinogenesis, and thus, our results indirectly support the essential role of HPV in the pathogenesis of cervical cancer.