Identification of Clinically Relevant Genes on Chromosome 11 in a Functional Model of Ovarian Cancer Tumor Suppression

Epithelial ovarian cancer (EOC) arises from the ovarian surface epithelium (OSE) by an as yet poorly defined genetic multistep process. As the leading cause of death from gynecological malignancy, EOC typically presents with advanced locoregional peritoneal dissemination, usually in the absence of visceral metastases. Improving the molecular understanding of ovarian cancer development and progression could contribute to novel strategies for the management of this poor prognosis neoplasm. To this end, it is necessary to elucidate the genetic changes that define ovarian cancer.

Inactivation of tumor suppressor genes (TSGs) is a frequent and important component of the process of neoplastic development. The genetic location of such genes can be inferred after the demonstration, at defined polymorphic loci, of frequent loss of heterozygosity (LOH) in tumor material in comparison to matched normal tissue. Frequent LOH has been demonstrated at defined regions of chromosome 11 (chr 11) in ovarian cancer, including 11p15 (1) and 11q23-q25 (2, 3). The large size of the regions identified and the high gene density, particularly of 11p15, limit the feasibility of a positional approach for the identification of TSGs. Furthermore, a number of issues such as inter- and intratumor heterogeneity may confound the analysis of LOH data (reviewed in Ref. 4) such that alternative or complementary approaches to the identification of TSGs are desirable.

Gene expression difference analysis aims to identify alterations in transcript abundance between RNA populations. Numerous techniques are available and have been widely used in the study of ovarian cancer (reviewed in Ref. 5). Differential display reverse transcriptase-PCR (DDRT-PCR) identified NOEY2 (ARHI; Ref. 6), DOC-1, and DOC-2 (7) as potential ovarian TSGs. cDNA-representational difference analysis (cDNA-RDA) has been coupled to high-density filter array (HDFA) hybridization in a comparison between primary cultures of normal human OSE and ovarian tumor-derived epithelial cells (8). Forty-four human OSE-specific and 16 ovarian cancer-specific genes were identified by this approach. Analysis of expression differences between grossly different RNA populations such as normal versus tumor tissue may not identify subsets of changes associated with particular cancer phenotypes resulting from alterations at specific chromosomal regions.

Microcell-mediated chromosome transfer (MMCT) is a controlled system in which to test the functional significance of defined chromosomal regions, under physiological regulation, generating resources for identifying the associated phenotypes and performing gene expression difference analysis.

Initial MMCT studies in ovarian cancer defined a reduction in growth in vitro associated with chr 11 transfer into the HEY cell line while not significantly affecting tumorigenesis in nude mice (9). Subsequently, transfer of chr 11 into the ovarian cancer cell line SKOV-3 efficiently suppressed tumorigenicity in mice (10). These studies lend support to the LOH data that suggests the existence of TSGs on chr 11, although neither study was able to define their subchromosomal location.

In this analysis, we describe a strategy that has identified several putative TSGs from chr 11 and a number of potentially related genes in a functionally defined ovarian cancer model. By MMCT, we have transferred a normal copy of human chr 11 into the OVCAR3 ovarian cancer cell line and characterized the resultant phenotypes of the hybrid cell lines. We then identified the mRNA alterations associated with these phenotypes by three expression difference analysis methods. Validation of identified difference products was performed by real-time quantitative RT-PCR. Importantly, this approach demonstrated that a high proportion of the products up-regulated upon MMCT localized to chr 11. The phenotypes associated with chr 11 transfer into OVCAR3 and the confirmed mRNA alterations identified, as well as their clinicopathological associations in two independent clinical series of ovarian cancers, are described.

Cell lines were maintained in DMEM/10% FCS/penicillin/streptomycin with selective media (HygromycinB and G418 as appropriate). MCH556.1.5 [obtained from Eric Stanbridge via A. G. Jochemsen (Leiden, the Netherlands)] is a mouse-human monochromosome somatic cell hybrid containing normal human chr 11, with neomycin insertion at 11q14-q22. OH1 is a clonal, hygromycin-resistant (Hyg^r) OVCAR3 cell line derived by transfection of tgCMV/HgTK (11) into the ovarian adenocarcinoma cell line OVCAR3 (12). OVCAR3 and OH1 have a hypertriploid karyotype with rearrangement of chr 11. The hygromycin/neomycin(neo)-resistant clonal control cell line OHN was derived by transfecting pMC1neoPolyA (13) into the clonal OH1 cell line. OVCAR3 was transfected with pMC1neoPolyA, and the clonal line ON1 was derived. ON1 was then used to transfer a neo-tagged chromosome to OH1 by MMCT to derive the control microcell fusion clonal line ONOH. Empirically, the control cell lines OHN and ONOH behaved identically to OH1 across the range of assays used in this study and demonstrate that the effects observed are unlikely to be caused by clonal selection.

To minimize cell line heterogeneity and artifacts of clonal selection, the hygromycin-transfected clonal cell line, OH1, was used as the recipient for microcell fusion. MMCT of chr 11 from donor MCH556.1.5 cells to OH1 recipient cells was performed using a protocol adapted from published methods (14, 15).

Metaphase spreads of microcell hybrid (MCH) cell lines were prepared for in situ hybridization using standard methods (16). Slides were then hybridized with biotinylated FITC-labeled, chr 11-specific paint probes (STAR*FISH; Cambio, Cambridge, United Kingdom). After overnight hybridization, the probe was visualized under a fluorescence microscope using an avidin-based detection system (Vector Labs, Burlingame, CA) and 4′,6-diamidino-2-phenylindole counterstain following standard protocols.

DNA was extracted from control and MCH cell lines for microsatellite analysis typically at passages 3–4, using a QIAamp kit (Qiagen, Crawley, United Kingdom). Twenty-three chr 11 microsatellite markers were selected from a high-resolution radiation hybrid map (17) and amplified using a standard PCR program. For detection of microsatellite PCR products on the automated laser fluorescence system (Pharmacia, Peapack, NJ), one primer from each marker pair was fluorescently labeled. For detection of microsatellites by nonfluorescent labeling, PCR products were separated on 6 or 9% denaturing polyacrylamide/urea gels, transferred onto Hybond N membrane and probed with a γ-³²P-end labeled (CA)₃₅ oligonucleotide (18).

Log-phase cultures of control and MCH cell lines were harvested and 1 × 10⁴ cells seeded in quadruplicate in 24 well trays. Cells were then harvested at 4-day intervals and counted using a Coulter counter.

A total of 1 × 10⁷ cells/cell line was harvested, washed, pelleted, and resuspended in 250 μl of serum-free media, mixed with 250 μl of Matrigel (Becton Dickinson, High Wycombe, United Kingdom) and s.c. injected into SCID mice in quadruplicate. Tumor volumes were calculated weekly based on bidimensional tumor diameter measurements.

Single-cell suspensions of cell lines and MCH clones for DNA analysis were prepared by standard protocols (19). Nuclei were extracted, fixed, and stained with propidium iodide and analyzed using a FACScan (Becton Dickinson). Relative DNA content and distribution of cells with respect to the cell cycle were analyzed using MODFIT software (Verity Software, Topsham, ME). As a measure of apoptosis, flow-cytometric detection of phosphatidylserine expression in cell lines was performed using the fluorescein-labeled Annexin-V Apoptosis Detection kit (R&D Systems, Abingdon, United Kingdom).

A total of 5 × 10⁴ cells (100 μl)/cell line prepared in serum-free RPMI was added to 400 μl of serum-free RPMI in the top compartment of 8-μm transwell culture inserts (Corning Costar, High Wycombe, United Kingdom), the undersides of which were coated with fibronectin (10 μg/ml) or laminin (10 μg/ml) and then incubated for 24 h. The numbers of cells present in the top and bottom compartments was then determined using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (20). Statistical analysis was performed using the Kruskal-Wallis nonparametric ANOVA with Dunn’s multiple comparison adjustment using the In-Stat program (Graphpad software).

Total RNA from MCH clones and human ovarian cancer cell lines was extracted using TRI Reagent (Sigma, Poole, United Kingdom) and contaminating genomic DNA removed by DNaseI digestion. DNaseI-treated total RNA, extracted from a histologically reviewed panel of 18 primary EOCs, was provided by Dr. Simon Langdon (Cancer Research UK, Edinburgh, United Kingdom).

Oligodeoxythymidylic acid-primed first strand cDNAs were prepared from DNaseI-treated total RNAs using the First Strand cDNA Synthesis kit (Roche, Lewes, United Kingdom). Control reactions minus reverse transcriptase were also performed.

The three methods described below were applied to the control cell line OHN and the MCH clone 11OH2.1 containing transferred 11pter-q22.

DDRT-PCR (21) was performed according to described protocols (22, 23) using a combination of three single-base anchored antisense primers and seven arbitrary sense primers (primer sequences are available from the authors). Products identified as being differentially expressed were excised from 6% (w/v) polyacrylamide sequencing gels, reamplified, subcloned, and sequenced.

cDNA-RDA was performed according to the method of Hubank and Schatz (24). In summary, respective driver and tester populations were prepared from DpnII-digested double strand cDNA synthesized from OHN and 11OH2.1 mRNA. Reciprocal subtractive hybridizations were then performed with increasing degrees of stringency. Difference products generated from the second, third, and fourth rounds of subtraction were subcloned and sequenced. In a subsequent experiment, PCR products corresponding to highly represented cDNAs were added in excess to driver populations to remove them from tester populations. This allowed less abundantly represented cDNAs to be amplified. OHN and 11OH2.1 driver populations were supplemented (spiked) with excess Moloney murine leukemia virus (MMLV) and excess retinaldehyde dehydrogenase 2 (RALDH2) cDNA-RDA products, respectively, after the second round of subtraction (see “Results”). Difference products generated from subsequent third and fourth round subtractions were then subcloned and sequenced.

Products identified from both DDRT-PCR and cDNA-RDA were subcloned into pGEM-T Easy TA cloning vector and transformed into competent JM109 cells (Promega, Southampton, United Kingdom). Plasmid DNA was isolated using standard methods (Promega and Qiagen) and sequenced with appropriate vector primers using dRhodamine or Big Dye terminators (PE Biosystems, Warrington, United Kingdom). Basic Local Alignment Search Tool homology searches against GenBank database sequences identified the difference analysis products.

Three independent Atlas cDNA Expression filter arrays (Human 1.2, Human 1.2II, and Human Cancer 1.2; BD Biosciences Clontech, Palo Alto, CA), each containing ∼1200 cDNAs, were hybridized with probes derived from the control cell line OHN and the MCH clone 11OH2.1 (performed by BD Biosciences as a Custom Atlas Service). Products exhibiting a difference of >3-fold between OHN and 11OH2.1 and with an intensity difference of >50 units were selected for subsequent quantitative real-time PCR validation.

Products identified from DDRT-PCR, cDNA-RDA, and HDFA hybridization were validated for differential expression by quantitative RT-PCR (Light Cycler; Idaho Technologies, Salt Lake City, UT, and Roche) using cDNA from OHN and 11OH2.1. Selected validated products (defined as 2-fold validated difference or greater) were further characterized for expression using cDNAs prepared from a panel of 20 cancer cell lines and 18 primary EOCs. Quantitation of double strand DNA was via detection of incorporated SYBR Green I, as described previously (25).

IHC detection of CTSD in an independent series of 58 primary ovarian tumor sections was carried out using a monoclonal antibody directed to human liver CTSD (ab6313; Abcam, Cambridge, United Kingdom) using a biotin streptavidin detection kit (BioGenex) according to the manufacturer’s protocol. Sections were scored for presence and level of CTSD staining by an experienced pathologist (A. A-N.).

Products selected for subsequent validation in two clinical series of EOCs, as described above, were investigated for associations with tumor stage, grade and histology by Fisher’s exact test (InStat; GraphPad software), and for associations with overall survival by Kaplan-Meier analysis with the log-rank test (SPSS version 10). The potential of CTSD as an independent prognostic variable in EOC was evaluated by multivariate Cox regression analysis using the variables found to be significant at the univariate level (tumor histology, grade, stage, and surgical debulk status) in combination with CTSD level as determined in 58 primary ovarian tumor sections. The forward stepwise method was used with the probability for entry being 0.05 and for removal 0.1 (SPSS version 10).

Neo-tagged, normal chr 11 was transferred by MMCT from the somatic cell hybrid MCH556.1.5 to OH1. Successful microcell transfer was confirmed by fluorescence in situ hybridization using a chr 11 paint probe (Fig. 1,A). Seven MCH clones were obtained from two independent experiments. The extent of murine cotransfer was investigated by mouse-specific chromosome paint and was confirmed to be minimal (data not shown). The extent of transferred chr 11 retained in the MCHs was determined by analysis of polymorphic microsatellite marker loci. In summary, whole chr 11 transfer was demonstrated in clones from the first experiment; clones from the second experiment lacked the 11q22-qter region. (Fig. 1 B).

Compared with the control cell lines OHN or ONOH, the introduction of whole chr 11 (MCH clone 11OH1.3) or chr 11 lacking 11q22-qter (MCH clones 11OH2.1, 11OH2.2, 11OH2.3, and 11OH2.4) resulted in significant and equivalent growth suppression in vitro, suggesting that the growth suppressor locus lay within the 11pter-q22 region (Fig. 2 A). A representative clone from each of the two independent MCH experiments (11OH1.3 and 11OH2.1) was used for additional functional analysis.

Transfer of chr 11 was associated with inhibition of s.c. tumor growth in SCID mice but not with complete suppression of tumorigenicity and was similar between the MCH clones 11OH1.3 and 11OH2.1 (Fig. 2 B), mirroring the in vitro growth findings.

MHC clones and control cells were assayed for alterations in cell cycle and apoptosis. Determination of propidium iodide uptake by fluorescence-activated cell sorting analysis showed that chr 11 transfer was not associated with obvious alteration of the cell cycle or ploidy compared with the control cell lines OHN and ONOH (data not shown). Annexin-V fluorescence-activated cell sorting analysis (26) was performed to assess if chr 11-mediated functional effects were caused by enhanced apoptosis. The percentage of apoptotic cells noted for the MCH clones 11OH1.3 and 11OH2.1 (3–5%) was the same as for control OHN (3%).

Cell migration assays were carried out to quantify the haptotactic migratory response of the MCH clones to purified extracellular matrix components fibronectin or laminin. 11OH1.3 and 11OH2.1 were significantly (P < 0.01) inhibited in their capacity to migrate toward a fibronectin haptotactic signal compared with parental control OHN, suggesting a locus affecting cell migration located within the 11pter-q22 region. Migration to laminin haptotactic signal was not significantly different from control cells for either 11OH1.3 or 11OH2.1 (Fig. 2 C).

Expression difference analysis was performed to identify RNAs altered in association with the defined phenotypes. The phenotypes described above were associated with transfer of the 11pter-q22 region, therefore, the MCH clone 11OH2.1 containing this region of introduced chr 11 but lacking 11q22-qter was used in comparisons with the OHN control cell line.

In total, 46 DDRT-PCR products were identified that appeared to be differentially expressed between the OHN control and 11OH2.1 MCH cell lines: 29 up- and 17 down-regulated in association with suppressed tumor growth and migration phenotypes after introduction of chr 11pter-11q22 into OH1.

Initial cDNA-RDA experiments identified MMLV as overexpressed in 11OH2.1 and RALDH2 as overexpressed in OHN. After spiking (see “Materials and Methods”), the total number of products identified through subsequent cDNA-RDA experiments consisted of 71 up- and 57 down-regulated products associated with introduction of chr 11pter-11q22 into OH1.

After our imposed selection criteria, 45 products up- and 28 products down-regulated in association with introduction of chr 11pter-11q22 were identified by HDFA analysis.

All products identified by DDRT-PCR, cDNA-RDA, and HDFA (247) were analyzed by real-time quantitative RT-PCR. A reliable result was obtained for the majority (179). Only 16 products (8.9%) showed a validated expression difference of ≥2-fold between OHN and 11OH2.1. In summary, 12 of the 16 products were identified as being up-regulated in the 11OH2.1 MCH clone in association with chr11-mediated growth and migration suppression. Four of these 12 products localized to chromosome 11: CTSD (11p15.5); CRYAB (11q22); RPL27A (11p15.5); and PSMD13 (11p15.5). Four products were identified as being down-regulated in association with chr11-mediated growth and migration suppression (Table 1).

The expression of CTSD, CRYAB, PSMD13, RPL27A, IGFBP2, and RALDH2 were investigated in primary normal OSE, a panel of 20 cancer cell lines (ovarian, colon, and breast) and a panel of 18 primary EOCs by quantitative RT-PCR (CTSD, Fig. 3,A; CRYAB, Fig. 3B; others not shown). In the panel of 20 cancer cell lines, both CTSD and CRYAB were shown to be greatly reduced in expression when compared with normal OSE.

The relationship between the transcriptional expression level of CTSD, CRYAB, PSMD13, RPL27A, IGFBP2, and RALDH2 in a panel of 18 EOCs and the clinicopathological parameters of tumor stage, grade, histology, and overall survival was examined. Survival analysis using the log-rank test revealed a reduction in median survival from 2.3 to 1.3 years for patients with CTSD expression < 6% of normal OSE (P = 0.0107, log-rank test; Fig. 4,A). In addition, expression of CTSD < 20% of normal OSE level was associated with a high tumor grade (P = 0.0179, Fisher’s exact test). Median survival was reduced from 2.1 years to 9 months for patients with CRYAB expression < 20% of normal OSE (P = 0.0126, log-rank test; Fig. 4 B) and expression of CRYAB < 50% of normal OSE level was associated with nonserous tumor histology (P = 0.0023, Fisher’s exact test). No other significant associations between transcriptional expression level and tumor stage, grade, histology, or overall survival were found (data not shown).

The expression of CTSD was also detected by IHC in an independent panel of 58 primary EOCs. Survival analysis using the log-rank test revealed a significant survival association. Median survival was ∼6 and 1.5 years, respectively, for patients with high levels (1+, 2+, and 3+ staining; n = 46) and low levels (0 and +/− staining; n = 12) of CTSD (P = 0.0184, log-rank test; Fig. 4,C). The association between CTSD levels, tumor histology, and International Federation of Gynecologists and Obstetricians stage were investigated and significant associations detected (summarized in Table 2). No other associations between CTSD level and clinical parameters were found.

Cox regression analyses revealed the following established factors to be significant prognostic variables at the univariate level: tumor histology (P < 0.001); grade (P = 0.012); International Federation of Gynecologists and Obstetricians stage (P < 0.001); and surgical debulk status (P = 0.001). Furthermore, at the univariate level CTSD, as determined by IHC in our clinical series of 58 primary EOCs, is a significant prognostic variable (P = 0.023). Multivariate Cox regression analysis of CTSD level in combination with the other significant univariate prognostic factors indicated that the association between CTSD level and overall survival was not an independent factor in EOC (P = 0.196).

Numerous LOH studies have indicated that chr 11 is likely to contain TSGs involved in EOC. We used the functional approach of MMCT to introduce human chr 11 into OVCAR3 and observed two phenotypes: suppressed tumor growth and suppressed migration to a fibronectin haptotactic signal. Phenotypic comparison of MMCT clones containing either whole chr 11 or chr 11 lacking 11q22-qter indicated that the putative TSG(s) responsible for these phenotypes is located outside the 11q22-qter. We then examined MMCT clones and parental controls using a gene expression difference analysis strategy, including quantitative RT-PCR-based validation, and identified potential chr 11 suppressor genes and components of chr 11 suppressor pathways. The clinical significance of two identified chr 11 products, CTSD and CRYAB, was established by demonstrating associations between transcript or protein levels and relevant clinical parameters in independent series.

MMCT permits functional complementation of a gene defect(s) without previously having defined the gene(s) of interest. Furthermore, this complementation is more relevant than is the case with overexpression studies because the genes are under the physiological control of endogenous promoters and enhancers. A common feature of MMCT is that transfer of the donor chromosome is not always complete (27). This breakage aids refinement of the region responsible for conferring functionality, typically achieved by microsatellite mapping of the parental line and MMCT derivatives. Ultimately, the cell lines derived from MMCT provide a defined and isogenic resource with which to perform gene expression difference analyses to identify the gene(s) underlying the observed phenotype(s). In our approach, we compared gene expression in control cells to clones containing transferred chr 11 lacking the 11q22-qter region because the functional suppression we describe here was ascribed to the introduction of this reduced region.

By using three independent difference analysis techniques, we have comprehensively characterized the differences in gene expression following chr 11 introduction into OVCAR3. These techniques were essentially nonoverlapping in the gene products that they identified, which highlights the importance of combining multiple difference analysis strategies for the detection of mRNA alterations. The notable exception to this nonredundancy was the identification of RALDH2 by both DDRT-PCR and cDNA-RDA.

The three expression difference analysis methods described here represent a screening approach to the identification of altered mRNA levels. Because of a high false positive rate, it is not sufficient to assign genes as being differentially expressed based solely on their identification in any of the expression difference analysis methods. Validation of each product by another method is a vital step in the determination of a reliable profile of gene expression changes.

cDNA-RDA is reportedly biased toward the detection of products with large expression differences (28). We detected 16 validated expression differences upon introduction of chr 11 to OVCAR3, however, only 4 (RALDH2, CB1, CB2, and MMLV) showed > 6-fold induction or repression. The high proportion of small expression differences coupled with the modest overall number of validated changes detected suggests that the introduction of chr 11 into OVCAR3 by MMCT does not profoundly affect global gene expression.

Of the genes down-regulated upon chr 11 transfer, two uncharacterized sequences (CB1 and CB2) were identified by cDNA-RDA, highlighting the usage of this technique in identifying sequences that may have been overlooked using microarrays with predefined content. CB1 and CB2 were not analyzed further in this study. chr 11 transfer suppressed RALDH2 expression ∼300-fold, as determined by quantitative RT-PCR. Subsequent analysis revealed a highly variable pattern of RALDH2 expression with no clear distinction between cancer cell lines EOCs and normal OSE (data not shown). RALDH2 was therefore regarded as having only contextual significance and was not considered further. HDFA identified suppression of IGFBP2 expression after chr 11 transfer and has been previously reported to be 10-fold up-regulated in ovarian cancer (29). IGFBP2 expression has been correlated with that of CA125 and is potentially useful as a tumor marker in ovarian cancer (30).

Of the 12 products identified as up-regulated upon chr 11 transfer, MMLV represents an artifactual expression difference associated with inevitable minor murine chromatin cotransfer during MMCT. MMLV was not considered further; however, its identification is evidence of the robustness of the expression difference analyses. Importantly, of the 11 remaining up-regulated human genes, 4 (36%) localized to chr 11 itself: 3 to 11p15.5 (CTSD, RPL27A, and PSMD13) and 1 to 11q22 (CRYAB). Significant associations were observed between expression of two of these genes (CTSD and CRYAB) and clinicopathological features of EOC.

The CRYAB gene lies on 11q23.1 in a region of frequent LOH in ovarian tumors (31). CRYAB is a major structural protein in vertebrate lens and also functions as a molecular chaperone of the small heat shock protein family (32). The mouse knockout of CRYAB displays hyperproliferative lens epithelial cells with notable genomic instability (33). Furthermore, the expression of CRYAB has been reported as decreased in testicular and breast tumors when compared with their normal counterparts (34, 35). The association described here between low CRYAB transcript expression and worsened survival suggests that loss of CRYAB expression may indeed be involved in promoting ovarian tumorigenesis or tumor progression and requires to be further studied in a larger independent clinical series.

CTSD is present in all cell types and functions, in the normal cell, as a lysosomal aspartyl protease. CTSD appears to function as a mediator of apoptosis induced by IFN-γ, Fas/APO-1, tumor necrosis factor α or oxidative stress and is released from lysosomes before cytochrome c relocalization (36, 37, 38, 39). CTSD may also be involved in p53-dependent tumor suppression and chemosensitivity (37). However, several studies have reported a high level of pro-CTSD secretion in breast cancer (reviewed in Ref. 40) and that expression of human pro-CTSD in the rat tumor cell line 3Y1-Ad12 increases the metastatic potential of xenografts in nude mice (41).

The prognostic significance of CTSD expression in ovarian cancer has been studied previously (42, 43). Scambia et al. (42) reported that primary ovarian cancer cases with low CTSD content showed a significantly lower percentage of complete response to chemotherapy and a shorter progression-free survival than cases with a higher CTSD content. The same group later reported a larger study of primary untreated ovarian cancer patients in which no significant change in 5-year survival was observed between CTSD-positive and CTSD-negative patients (44). Baekelandt et al. (43) studied a large group of untreated International Federation of Gynecologists and Obstetricians stage III ovarian cancer patients and reported a favorable prognostic significance associated with expression of CTSD.

The low expression of CTSD RNA observed here, relative to normal OSE, in a panel of EOCs and cancer cell lines, implies that loss of CTSD expression may be involved in ovarian cancer tumorigenesis or progression. Survival analyses of the data from both quantitative RT-PCR and a larger independent IHC study described here have shown that low levels of CTSD are significantly associated with adverse survival providing robust evidence for a link between CTSD level and overall survival in EOC. Data from the IHC study indicates a reduction in median survival from 6 years in the high expression group to 1.5 years in those tumors with low CTSD expression. Although this result is clear, there are significant associations between other clinicopathological factors and CTSD level as determined by either quantitative RT-PCR or IHC but not both. This may be because of sample size or an imperfect correlation between the cellular RNA and protein levels, although this possibility was not investigated here. Quantitative RT-PCR showed a significant association between CTSD and poor differentiation, which was not confirmed by CTSD IHC. IHC showed a clear association between low CTSD expression and advanced tumor stage, with the expected significant enrichment for serous histology.

The lack of multivariate significance of CTSD indicates its relationship with the known clinicopathological features of EOC. The significant association between low CTSD and the typical advanced stage, serous adenocarcinoma of the ovary (and thereby, with survival), is interesting and warrants mechanistic work into the relationship between level of CTSD expression and this clinical phenotype.

From the data presented both here and elsewhere in the literature, it is clear that the expression of CTSD is an important factor in ovarian cancer. Connolly et al. (45) recently reported a novel transgenic mouse model that develops ovarian cancer. This could be used in conjunction with existing CTSD-knockout mice (46) to assess the role of CTSD in the generation of particular phenotypes of ovarian cancer.

Detailed mutational and functional analyses of CTSD and CRYAB will additionally define their role in epithelial ovarian oncogenesis and may yield novel insights into the molecular basis of this disease.

We thank Alison Ritchie for animal husbandry, Mike Hubank and David Schatz for the cDNA-RDA protocol, David Fitzpatrick for use of the Roche Light Cycler, Robert Rush for performing multivariate analyses, Simon Langdon for primary ovarian tumor RNA samples, and Aart Jochemsen and Eric Stanbridge for the MCH556.1.5 mouse-human chr 11 somatic cell hybrid.