Relationship of XIST Expression and Responses of Ovarian Cancer to Chemotherapy ¹

Currently, clinical management of epithelial ovarian cancers is surgical cytoreduction followed by consolidative chemotherapy. First-line chemotherapy in ovarian cancer consists of a platinum analogue in combination with Taxol (1). Despite a significant initial response rate of advanced ovarian carcinoma to the first-line chemotherapy, many patients relapse, and fewer than 15% of the patients will become long-term survivors (2), suggesting an intrinsic or acquired resistance to first-line chemotherapy in ovarian cancer. However, the molecular basis for chemoresistance and the development of recurrence in ovarian carcinoma is still largely unknown.

Currently, cDNA microarray technology has been used as a new approach to characterize cancer pharmacology (3, 4) and to predict clinical outcome (5–7). We initiated a study of comparing the global transcription profiles between primary and recurrent ovarian tumors of the same patient. In one such analysis of a recurrent tumor (RC680) salvaged from ascitic fluid and the paired primary tumor (PC598C) of the same patient, the gene with the highest differential expression in these two RNA populations was XIST.³ Early in mammalian embryogenesis, X chromosome inactivation leads to the cis-limited transcriptional silencing of most of the genes on one of the two X chromosomes in female somatic cells, resulting in dosage compensation between males and females (8). Classical genetic studies have defined the X inactivation center at Xq13, from which X inactivation initiates and spreads along the X chromosome (9). The inactive X chromosome is heterochromatic, late-replicating in S-phase, underacetylated at histones H3 and H4, and highly methylated at some regions (10). The XIST gene resides within the X inactivation center and is the only transcript expressed exclusively from the inactive X chromosome (11). In female ES cells prior to X inactivation, low-level XIST expression can be detected from both active X chromosomes. After differentiation, XIST is expressed at high levels only from the inactive X chromosome through stabilization of XIST transcripts at the inactive X chromosome (12). Targeted mutation of the XIST homologue in mouse (Xist) indicates that the Xist RNA is required for female dosage compensation (13, 14) and XIST YAC transgenic experiments have also demonstrated that XIST is both necessary and sufficient for X inactivation (15). Clinically, the severe phenotype of Turner syndrome has been associated with small ring X chromosome, where the XSIT locus is either not present or not expressed. It has been hypothesized that the loss of XIST results in functional disomy for the genes contained in the ring (16).

The cell lines derived from RC680, OVCA680, and 15 other ovarian cancer cell lines were tested for sensitivity to anticancer drugs including carboplatin and Taxol, which the patient had received. Statistical analysis showed a significant association between Taxol sensitivity and XIST RNA level. Neither carboplatin nor cisplatin showed such correlation. Cytogenetic studies of the cell lines by spectral karyotyping, RNA FISH, and combined immunostaining and chromosome painting have suggested that loss of inactive X chromosome is one mechanism for the loss of XIST expression in some cell lines. To estimate the clinical significance of XIST expression, 21 cases of ovarian carcinoma tissues from the patients who were treated with Taxol and platinum drug were investigated, and XIST expression was found to be strongly associated with the patient responses to chemotherapy. Our data suggest that XIST is a potential marker for chemoresponse prediction in ovarian cancer.

Total RNA was extracted from the primary and recurrent tumor cells by Trizol reagent (Invitrogen Life Technologies, Inc.). Poly(A) RNA was purified from total RNA by Oligotex mRNA kit (Qiagen). RNA concentration was determined by a RiboGreen fluorescence method (Molecular Probe). cDNA synthesis and microarray hybridization was performed by GenomeSystems, Inc. through a subcontract to the Center for Genomics Research of Harvard University. Hybridization was performed on a UniGEM V microarray from Incyte Pharmaceuticals, Inc., which contains 7000 sequence-verified human genes and expressed sequence tag sequences mapped to National Center for Biotechnology Information’s UniGene database.

Sixteen ovarian cancer cell lines were used in this study. ALST, CAOV3, OVCA3, OVCA420, OVCA429, OVCA432, OVCA433, OVCA633, OVCA680, OVCA702, and OVCA810 were established in our laboratory from tumor tissues obtained from different patients. OVCA680 and OVCA702 were derived from recurrent ovarian carcinomas, whereas others were all derived from primary ovarian carcinoma tissues. SKOV3, ES-2, and TOV21G were purchased from American Tissue Culture Collection. RMG-1 and RMUG-L were purchased from Japanese Collection of Research Bioresources (JCRB Cell Bank). ES-2 was maintained in McCoy’s SA medium supplemented with 10% FCS, whereas RMG-1 and RMUG-L were cultured with DMEM-F12 HAM medium plus 10% FCS. Other ovarian cancer cell lines were grown in medium 199 and MCDB 105 (1:1) supplemented with 10% FCS. Five breast cancer cell lines used in our study were MDA-MB231, MDA-MB361, MDA-MB453, MDA-MB468, and MCF-7. All breast cancer cell lines were obtained from American Tissue Culture Collection and maintained in DMEM supplemented with 10% FCS. Ovarian tumors were obtained from patients at Brigham and Women’s Hospital in Boston. All of the surgical specimens were collected with patient consents and were confirmed by histological examination.

TaqMan Reverse Transcription reagents (ABI) were applied for cDNA synthesis. The SYBR Green reagents kit (ABI) was used for quantitative real-time PCR analysis and performed according to the manufacturer’s recommendation. During PCR, reactions were continuously monitored by an ABI Prism 5700 Sequence Detector (ABI). Cyclophilin messages were used as the internal control. The sequences of the primers used were listed as follows: XIST-11302F, 5′-GCAGTTTGCCCTACTAGCTCCT-3′; XIST-11456R, 5′-TCCTCAGGTCTCACATGCTCA-3′; Cyc-531F, 5′-GATGCTCCTGCGTTCTGATGT-3′; and Cyc-631R, 5′-GGTGGAAGCTGCTTCCCTTAA-3′. All primers were designed with PrimerExpress 1.5 software (ABI). All of the forward and reverse primers were designed on different exons separated by long introns to prevent noise signals generated from contaminated genomic DNA during the PCR reactions.

Drug cytotoxicity was determined by MTT dye-based methodology. Briefly, cells were seeded in 100 μl of medium (5000 cells) per well in 96-well plates. After overnight incubation, 100 μl of medium with different dilutions of Taxol were prepared and added to treat the cells for 48 h. Cell viability was quantified with the MTT cell proliferation assay kit (Roche) as described by the manufacturer’s manual. The IC₅₀ is the concentration of drug resulting in a 50% reduction in absorbance at 550 nm to that of untreated cells. Every assay was performed in triplicates, and the drug IC₅₀ of each cell line was the average of three to six independent experiments.

Correlation analysis of the relationship between XIST RNA level and the drug IC₅₀ (or relapse time) was carried out using the MINITAB statistical package. The Pearson correlation coefficient calculated between any two variables provides a measure of the strength of their linear relationship. Its significance is tested using a two-tailed t test. Both the calculation of the Pearson correlation coefficient and its test are based on two assumptions: (a) each pair of observations is independent of any other pair; and (b) each variable is normally distributed. In view of the second assumption, log transformation is therefore applied for each variable for the statistical analysis. For comparison, we provide Ps obtained using Efron’s (17) nonparametric bootstrap resampling method. To examine the clinical relevance of XIST expression in responses to cancer chemotherapy, multiple regression analysis was carried out using the MINITAB package. The three subtypes (serous, mucinous, and endometrioid) in the histology information was transformed into two dummy variables, whereas a binary (0, 1) variable was adopted for the stage by combining stages 3, 3C, and 4 into one group and stages 1A and 2A into the other. In addition, the relapse time, CA125, and age were transformed into log form. We adopted backwards elimination steps to select predictor variables. The method started with all variables in the regression equation. Then the variable was removed one by one, based on the Fs associated with the variables. The procedure was terminated when no variable in the regression equation had an F that was <4. Residual plots were used to check the adequacy of the final model.

Cultures were synchronized by 260 nm demecolcine (Sigma) for 18–24 h, and metaphase cells were differentially collected by mitotic shake-off method. Metaphase spreads were prepared as described before (18). After removing excess liquid, the slide was always kept in a humidified chamber for both immunolabeling and DNA FISH processes. For immunostaining of acetylated histone H4, the slide was blocked with 40 μl of 3% donkey serum in KCM buffer (18) for 30 min and followed by 40 μl of anti-acetylhistone H4 (Lys12; Upstate Biotechnology) antibody (1:90) incubation for 1 h. Then the slide was washed twice at room temperature for 5 min each with KCM buffer. Biotin-conjugated donkey anti-rabbit secondary antibody (Pierce) was used at ×200 dilution as described for the primary antibody. After the final wash, the slide was fixed in 3:1 methanol/acetic acid (v/v) for 10 min at room temperature and allowed to air dry. The slide was then processed for DNA FISH. The chromosomes on the slide were denatured by applying 100 μl of denaturing solution (0.5 n NaOH, 1.5 m NaCl) for 5 min at room temperature, followed by neutralization solution [0.5 m Tris-HCl (pH 7.4), and 1.5 m NaCl] for 5 min at room temperature twice. The slide then was prepared, hybridized to FITC-labeled X chromosome paint probe (Vysis, Inc.), and washed as described in the manufacturer’s manual. The biotinylated anti-rabbit antibody was detected using Texas Red conjugated Streptavidin (Pierce) diluted 200 times in 4 × SSC (pH 7.2)/1% BSA at room temperature for 1 h in the dark. The slide was washed twice at room temperature for 5 min each with 4× SSC (pH 7.2)/0.1% Tween 20, followed by standard mounting protocol with Vectashield mounting medium, which contains DAPI as the counterstain (Vector Lab). The image was captured by an Olympus BX60 fluorescence microscope equipped with an Optronics CCD camera.

Interphase nuclei were prepared, and RNA FISH was performed as described before (15). The human XIST probe was a pool of ∼3.5 kb of sequence derived from exon 1 and 1 kb derived from exon 6 (15).

Metaphase spreads were prepared and hybridized with a mixture of 24 differentially labeled chromosome painting probes according to the manufacturer’s recommendation (Applied Spectral Imaging, Inc.). The images were captured with an ASI camera and analyzed with SkyView 1.2 software (Applied Spectral Imaging, Inc.).

A cDNA microarray analysis of RNA samples prepared from paired primary and recurrent ovarian tumor cells was performed as described in “Materials and Methods” (Fig. 1). The gene XIST was found to be the most differentially expressed and down-regulated gene in the recurrent tumor (6.7-fold by microarray analysis and 51.3-fold by real-time quantitative PCR). We established a cell line, named OVCA680, from the recurrent tumor. This and 15 other ovarian carcinoma cell lines were subjected to drug sensitivity assays with a panel of antineoplastic agents. Because the patient from whom the tumor specimens were obtained has been treated with Taxol and platinum drug during the first-line chemotherapy, we focused on these two types of anticancer agents (Table 1). Compared with other cell lines, OVCA680 demonstrated a high level of Taxol resistance and moderate resistance to carboplatin and cisplatin. We also determined the relative levels of XIST transcripts in the cell lines by quantitative real-time PCR (Table 1). Similar to other microarray studies (3, 4), a natural logarithm was applied to normalize the data as described in “Materials and Methods.” Significant correlation was found between the levels of XIST transcript and Taxol IC₅₀s (r = −0.584; Fig. 2,a, solid regression line). Using a two-tailed t test, the P is 0.018. For comparison, we calculated the P from 1000 bootstrap samples for the null hypothesis of zero correlation, P_boot = 0.016. However, the levels of XIST transcript is not significantly correlated with carboplatin (r = 0.017; P = 0.949, P_boot = 0.946; Fig. 2,b) and cisplatin (r = 0.116; P = 0.668, P_boot = 0.670; Fig. 2,c). Because carboplatin and cisplatin involve similar mechanisms of cell killing, the IC₅₀s of these two drugs exhibit strong correlation (r = 0.892; P < 0.001, P_boot < 0.001). The regression analysis of XIST expression to Taxol IC₅₀s showed an inverse correlation, whereas there was no significant correlation to both carboplatin IC₅₀ and cisplatin IC₅₀s. When we included the results of five breast cancer cell lines in the correlation analysis of Taxol IC₅₀ and XIST expression level, we obtained an even more significant correlation (r = −0.601; P = 0.004, P_boot = 0.004; Fig. 2 a, dotted regression line).

To investigate the mechanism(s) by which XIST expression is regulated, several cytogenetic experiments were performed. SKY analysis of the cell lines showed extensive translocation and only one X chromosome in the OVCA680 cells (Fig. 3a). Because histone underacetylation is one of the hallmarks unique to inactive X chromosome, we performed a combined immunostaining and X chromosome painting experiment on the metaphase spreads of OVCA680 cells to determine the histone acetylation status of the single X chromosome. A strong signal for acetylated histone H4 was observed on the X chromosome (Fig. 3, b and c), indicating that the X chromosome is an active X chromosome. In addition, we also performed RNA FISH using an XIST-specific probe on the cell lines, because XIST is exclusively expressed from and remains associated with the inactive X chromosome. We did not see any XIST signal in the RNA FISH analysis of OVCA680, ALST, CAOV3, and OVCA3 (data not shown). On the other hand, SKY and X chromosome painting analysis of SKOV3, one of the cell lines with relatively high levels of XIST expression, showed that there are five X chromosome in the cells (Fig. 3, d and e). RNA FISH with the XIST-specific probe demonstrated that this cell line is composed of cells with either one or two inactive X chromosome(s) (Fig. 3f). Hence, duplication of inactive X chromosome might contribute to the higher level of XIST expression in the SKOV3 cell line. Because both OVCA680 and CAOV3 contain only one X chromosome (Fig. 3a and data not shown), loss of inactive X chromosome may be one mechanism for the loss of XIST expression. In addition to loss of active X chromosome, some other mechanisms may be responsible for down-regulation of XIST expression. For example, some cell lines that exhibit low levels of XIST expression, such as OVCA3 and ALST, still maintain two X chromosomes (Fig. 3, g and h). There is an X chromosomal translocation observed in the OVCA3 cell line, whereas no obvious cytogenetic changes were identified in the ALST cells. These data suggest that multiple mechanisms might be involved in the changes of XIST expression in the ovarian carcinoma cell lines.

We then looked into the clinical relevance of XIST expression by analyzing patient responses to cancer chemotherapy. We chose primary tumor specimens from patients who have been treated with at least six cycles of carboplatin and Taxol after surgical cytoreduction and with medical history for at least 1 year after completion of chemotherapy. Two patients who were treated with triple-doublet protocol (three cycles of carboplatin/Taxol, two cycles of cisplatin/gemcitabine, and two cycles of Adriamycin/Topotecan) and relapsed within 6 months after completion of treatment were also included. All together, 21 cases collected from December 1995 to March 2000 were selected. Total RNAs were prepared from tumor tissues, and the XIST RNA levels were determined by quantitative real-time PCR. We chose relapse time (the time from completion of first-line chemotherapy to diagnosis of recurrent disease) as the parameter to measure the clinical outcome of the first-line chemotherapy. Correlation analysis was performed, and it was found that there was a strong correlation between relapse time and XIST expression level (r = 0.653, P = 0.001, P_boot < 0.001; Fig. 4). To account for other variables such as histology information, CA125, and patient age, as given in Table 2 multiple regression analysis was also performed. We adopted the backwards elimination approach to select the predictor variables and chose the two-variable regression as the final model based on the sum of squares error and the degree of freedom. It was observed from the final model that XIST expression was significantly related to the relapse time and marginally to patient age (Table 3). For assessing the adequacy of the final model, residual plots were drawn, and they did not show any pattern, implying that the final model is adequate. Hence, significant positive correlation between XIST expression and relapse time still exists in the multivariate analysis.

We explored the use of high-throughput cDNA microarray technology to identify potential prognostic markers for ovarian cancer by mRNA expression profiling of paired primary and recurrent tumors. Further studies of selected genes in ovarian carcinoma cell lines and patient samples have revealed a significant association of XIST expression level with Taxol sensitivity and chemotherapeutic responses. The high correlation of XIST RNA level and Taxol sensitivity in ovarian cancer cell lines suggests that XIST might play a role in modulating Taxol cytotoxicity. XIST is expressed exclusively from and remains associated with the inactive X chromosomes in female cells. Insertion of a 480-kb YAC genomic clone containing XIST gene to somatic chromosomes in mouse ES cells showed localized inactivation of adjacent genes on somatic chromosome as well as the neomycin marker gene on YAC vector (19). This suggests that XIST RNA can either directly or indirectly modulate the chromosome it associates with. As much about the molecular basis for XIST involvement in X chromosome inactivation is still lacking, it is also not known whether XIST has any possible direct effect on drug resistance other than X chromosome inactivation. Several attempts at reconstituting XIST expression in the ovarian cancer cell lines by transfection of YAC genomic DNA containing the XIST gene have failed, either because of low transfection efficiency or silencing of the Neo selection marker on YAC vector. Cotransfection with a second marker DNA and selection of clones for the presence of second marker may be an alternative approach.

One other plausible mechanism that XIST RNA modulates Taxol sensitivity is the reactivation of resistance-specific genes on inactive X chromosome in the absence of XIST RNA. Previous studies using the mouse/human somatic cell hybrids and human leukemia cells have shown that XIST RNA is not necessary for the maintenance of X inactivation once the cells are differentiated (20, 21). A more recent report suggested, however, the synergy of XIST RNA, DNA methylation, and histone hypoacetylation in maintaining the inactive state of X chromosome (22). Our RNA FISH study on SKOV3 cells showed punctuated XIST expression signals localized to the XIST locus, unlike the intense signals covering the whole inactive X chromosome observed in differentiated mouse ES cells. More studies are under way to determine whether the situation is different in the differentiated ovarian cancer cell lines and whether the hypothesis that XIST modulates the expression of other genes is accountable for Taxol resistance, at least for those cells encompassing two or more X chromosomes. If it is true that XIST plays a role in regulating other resistance-specific gene(s) on the X chromosome, one possible down-stream candidate is XIAP, which is the most potent direct inhibitor of caspases and apoptosis among all human IAP family proteins (23). Down-regulated expression of XIAP has been shown to induce apoptosis in chemoresistant human ovarian cancer cells (24). Down-regulation of XIST might increase the expression level of XIAP and block drug-induced apoptosis to cause resistance phenotype.

It is also likely that loss of XIST expression is a marker for genetic instability associated with drug resistance, because loss of inactive X chromosome has been known to be a common event in cancer cells (25). However, the observed association between loss of XIST expression and drug resistance is only specific to Taxol, not to the alkylating agents (Fig. 1). The cell lines used in our study have not been pretreated with Taxol, and therefore the observed association is not attributable to drug-induced phenomenon. One possible explanation of this drug-specific association may lie with the action mechanism of Taxol. The action of Taxol is mediated through binding to microtubules and stabilizing them against depolymerization, thereby inhibiting cell replication by disrupting normal mitotic spindle formation (26). It has been observed that the action of Taxol on microtubules associates with mitotic arrest and induction of apoptosis, and mitotic arrest correlates with Taxol cytotoxicity (27). Treatments using Taxol together with other drugs that perturb cell cycle progression can dramatically interfere with the cell killing and apoptosis-inducing activity of Taxol (28). Mitotic checkpoint mechanisms regulated by MAD and BUB proteins control the proper segregation of chromosomes and play an important role in the genetic instability of cancers (29). It is therefore hypothesized that genetic instability in advanced cancer cells that have lost mitotic checkpoint function might cause the loss of XIST expression and the development of mechanism to escape Taxol-induced apoptosis as well.

In summary, our cDNA microarray analysis of paired primary and recurrent ovarian tumors has identified a high correlation between XIST RNA expression level and Taxol sensitivity in ovarian cancer cell lines. More interestingly, a strong correlation was also found between XIST RNA level and patient responses to chemotherapeutic treatment, suggesting that XIST expression might be used as a marker for prediction of first-line chemotherapy responses in ovarian cancer. This study raises an interesting perspective between X chromosome inactivation and drug resistance. Although the underlying mechanism for this correlation is unclear at present, more extensive studies on gene expression profiles of the paired primary and recurrent tumors may reveal more insights into the relationship between these two apparently unrelated processes.

We thank Dr. Bryan M. Turner for advice regarding combined immunostaining and chromosome painting.