Selection of resistant cell lines
In this study, a set of three isogenic drug-resistant ovarian cancer cell lines has been generated from the A2780 ovarian cancer cell line. The A2780 cell line has the advantage of being derived from a chemo-naïve patient, and is therefore sensitive to many chemotherapeutics [46–50] but has been demonstrated to be capable of developing resistance in vitro
[51–55]. In addition, the strategy of deriving isogenic drug resistant cell lines from one cell line eliminates variability due to intrinsic genetic differences between cell lines. Although numerous studies have investigated mechanisms of drug resistance to single agents, the standard of care for chemotherapy in ovarian cancer is a combined treatment with a platinating agent and a taxane. Therefore, in this study, we report the generation of dual drug resistance in vitro and characterization of cells selected for resistance to both classes of agents.
Characterization of levels of resistance
During the selection for single or dual drug resistance in our study, the gradual increase in drug concentration, beginning with a dose 1000-fold below the IC50 of the parental A2780 cell line, generated populations of resistant cells and avoided selection of a few drug resistant clones. This selection strategy may not seem to reflect the typical clinical approach of treating patients with high doses delivered in several cycles, but the dose administered to a patient is not likely reflective of the amount of drug that actually reaches a tumor. Studies of intratumoral drug distribution have shown that drug concentrations vary within a tumor, that not all tumor cells may experience a lethal dose, and that other factors such as intratumoral cell heterogeneity and tumor microenvironment interactions can interfere with consistent, high dose delivery of a drug in a tumor [56–58]. Although this situation is very difficult to imitate under in vitro conditions, we believe our approach beginning with a low concentration and gradually increasing the dose is more likely to mimic the variable and gradually increasing drug environment in a tumor and to select for a population of drug resistant cells representative of the cell heterogeneity present in tumors. Using this selection method, the A2780CBN cell line was acquired with an IC50 of 7.77 × 10-5 M carboplatin (Table 1), a concentration similar to the maximally tolerated plasma concentration of carboplatin (3.8 × 10-5 M) , indicating the A2780CBN cell line tolerates clinically detectable concentrations of carboplatin. The level of resistance in the A2780CBN line (13.56 fold) is comparable to resistance levels reported for cisplatin in vitro in ovarian tumor cells [10, 52].
The A2780DXL cell line had an IC50 of 3.61 × 10-7 M docetaxel, which was 4000 fold more resistant than the A2780CC. Although initially very toxic, once resistance had begun to develop, it was possible to increase the dose until this very high level of resistance occurred. Intraperitoneal delivery of docetaxel to patients was reported by Morgan et al. to result in mean peak plasma concentrations of 4.6-6.6 × 10-7 M docetaxel and 5.9-8.1 × 10-5 M mean peak intraperitoneal concentrations of docetaxel . Although the range between the plasma and intraperitoneal concentrations reported by Morgan et al. is more than 100-fold, depending on the compartment measured, the IC50 of our A2780DXL line falls just under the lower end of the range, indicating that the A2780DXL cell line tolerance also falls in a clinically relevant range. In vitro resistance to paclitaxel in ovarian cell lines has been reported in this range, as well [61, 62].
The selection of the dual resistant A2780CBNDXL cell line resulted in combined resistance, with an IC50 of 8.02 × 10-6 M for carboplatin and 8.02 × 10-9 M docetaxel (Figure 1). Compared to the A2780CC cell line, the fold change in resistance is about 13 for both drugs since the method we used increased the carboplatin and docetaxel doses at the same time and to the same extent. It is interesting that the increase in resistance is about 13-fold which is similar to the A2780CBN line. This may indicate that the carboplatin concentration was the limiting factor in this type of selection scheme. A role for carboplatin in determining the degree of resistance achieved in the A2780CBNDXL lines may be reflected by the principal component and hierarchical clustering analyses which both showed that the A2780CBN and A2780CBNDXL cell lines were more similar to each other than either was to the docetaxel resistant line. If we compare the IC50 values for the single agent resistant lines to the dual line (Table 1), there is about a 10-fold decrease in the amount of carboplatin tolerated by the dual line compared to the A2780CBN line and about a 45-fold decrease in the amount of docetaxel tolerated by the dual resistant line, indicating that dual drug treatment is effective at lower doses.
To ensure that the A2780CBNDXL cell line truly was resistant to both carboplatin and docetaxel, we exposed the dual line to each drug alone. Figure 2 shows that the A2780CBNDXL line is resistant to carboplatin (Figure 2A) and to docetaxel (Figure 2B), demonstrating that the A2780CBNDXL line is a dual drug resistant cell line. Compared to the A2780CC, the dual line is 10 fold more resistant to carboplatin and 8 fold more resistant to docetaxel. The degree of resistance to each drug appears to be less than when the dual line is exposed to both drugs simultaneously (13 fold), but this is likely due to the A2780CC line tolerating a higher concentration of drug when it is exposed to each drug alone compared to both drugs simultaneously.
Cross resistance to completely different drugs or compounds in cell lines selected for resistance to a specific drug is a recognized phenomenon [53, 63, 64]. In contrast, cross resistance between platinating agents and taxanes is not very common . In a review of more than 100 models of acquired drug resistance, approximately 70% of cisplatin resistant and paclitaxel resistant cells remained sensitive to paclitaxel and cisplatin, respectively . Since cross resistance could conceivably contribute to a phenotype of dual drug resistance, the sensitivity of the single drug resistant cell lines to the opposite drug was tested. In this study, both the A2780CBN and A2780DXL lines were shown to lack cross resistance to docetaxel and carboplatin, respectively (Figure 3). Interestingly, the A2780DXL cell line showed a trend towards hypersensitivity towards carboplatin, although this was not statistically significant (Figure 3B). Hypersensitivity occurs when a resistant cell line is more sensitive to a drug than the parental cell line it was derived from [8, 44], and was observed in almost 30% of the models of acquired drug resistance surveyed by Stordahl et al. While the lack of cross resistance in the single agent resistant A2780 cell lines does not prove that the dual agent resistant line developed without cross resistance, it seems more likely that a genuine dual resistance was generated in the A2780CBNDXL line and not just a single agent resistance with cross resistance to the opposite drug.
Proliferation of resistant cell lines
The rates of proliferation determined for each of the resistant cells lines and the co-cultured control line show that all the resistant lines have a reduced rate of proliferation compared to the A2780 parental line (Figure 4). While it is well known that malignant cells exhibit a higher rate of proliferation than normal cells [66–68], it is not as well-established that drug resistant cells may also demonstrate an altered rate of proliferation. Gene expression leading to increased cell proliferation and drug resistance has been reported [69, 70]. However, reports of reduced cell proliferation associated with increased drug resistance have also been made and an association between multi-drug resistance and decreased proliferation exists, which supports our observation of decreased proliferation in not only the single agent resistant but the dual agent resistant cell line [13, 71, 72]. Moreover, reduced proliferation in drug resistance may not be so surprising when one considers that most cytotoxic chemotherapy agents are designed to target rapidly proliferating cells; reduction of proliferation could be one way to promote a drug resistant phenotype.
Microarray analysis of gene expression patterns in the resistant A2780 cell lines
The number of unique changes in gene expression detected in each cell line was similar (Figure 5). Considering the different mechanisms of action of carboplatin and docetaxel, it is expected that the carboplatin and docetaxel resistant cell lines should not have many changes in gene expression in common. However, the relatively low amount of common gene expression changes between the dual line and each of the single agent resistant lines indicates that the majority of the changes in the dual line are unique and not a simple combination of the patterns present in each single agent resistant line. Furthermore, the separation of the three resistant cell lines by principal component analysis of all the genes with altered expression supports our claim of a distinct pattern of gene expression in the dual resistant cell (Figure 6). Additional evidence for the unique pattern of gene expression induced by simultaneous exposure of the cells to both carboplatin and docetaxel is present in the hierarchical cluster analysis which shows a different pattern of gene expression in all three resistant cell lines (Figure 7). Based on these results, we can state that development of resistance to more than one chemotherapy agent has the potential to induce novel changes not associated with resistance to each single agent.
Validation of microarray results
QPCR amplification of validation gene set transcripts confirmed the results of the microarray analysis, except for the GSTO1 gene, which was not confirmed by QPCR as significantly upregulated in the A2780CBN line, although expression was detected by microarray hybridization (Table 4, Additional file 2: Table S2). The QPCR results were more sensitive in detecting changes in gene expression not found by microarray analysis. For example, 11 additional instances of altered gene expression were detected by QPCR for ANXA1, CDH11, CDH7, CYP1B1, FLRT3, GSTO2, LGI1, MT2A, and PARP9 (Table 4). Fold changes were in the same direction but the QPCR results often showed a much greater change, e.g. the ABCB1 and ABCB4 gene expression detected by QPCR was around 10–1000 greater than the microarray results (Table 4). The improved accuracy of detecting gene expression by QPCR in our study may be due to the design of the QPCR primers, which were based on transcript specific sequences from the protein coding transcript for each gene whereas the oligonucleotides used in the microarray are designed to detect all possible transcripts of a gene, including non-coding transcripts. Therefore, our QPCR primers are more accurate in detecting gene expression that is more likely to be associated with protein expression and represent true genetic response to drug selection.
QPCR confirmation of differences in gene expression among the three resistant A2780 cell lines
The one way ANOVA followed by Tukey’s post hoc test detected significant differences in expression among the resistant cell lines as determined by QPCR. Based on this analysis, four of the genes in the validation set of 16 genes, were found to be significantly different in the A2780CBNDXL line. Although also significant in the A2780CBN line, the AKR1C3 gene was expressed to a significantly different extent mainly in the dual resistant line. The role of aldoketoreductases in cisplatin and multidrug resistance has been described in several different types of cancer cells [43, 45, 73, 74]. Therefore, the discovery of a significant increase in AKR1C3 expression in the dual drug resistant line supports a role for aldoketoreductases in combined carboplatin and docetaxel resistance. The PARP9 gene was also mainly expressed in the dual drug resistant line. PARP proteins, in particular PARP 1, are involved in DNA repair and have become a therapeutic target in BRCA mutant cancers [75–77]. A direct role for PARP proteins has also been reported in cisplatin resistance [78, 79]. In this study we report a significant increase in expression of PARP9 in the dual resistant A2780CBNDXL line compared to the A2780CBN line, extending the impact of PARP proteins to combined carboplatin and docetaxel resistance. An additional two genes that were mainly expressed in the dual line were CDH11 and CDH7, with CDH11 being the most upregulated gene in the dual line (1022 fold upregulated). Cadherins, in particular CDH1 (E-cadherin), are known to contribute to invasiveness and stem cell like properties in ovarian cancer [80–83]. E-cadherin-mediated intercellular adhesion has also been shown to contribute to chemotherapy resistance . CDH11, however, is a classic type II cadherin, known to be involved with bone morphogenesis , and has been shown to play a role in epithelial to mesenchymal transition . As well, CDH11 mediates cell adhesion [87, 88] as does CDH7 [89, 90], another classic type II cadherin, also significantly over expressed in the dual line. Intercellular adhesion has been demonstrated as an important factor in multidrug resistance . Therefore, the distinct upregulation of CDH11 and CDH7 in the dual resistant A2780CBDXL cell line could indicate a role for type II cadherin mediated cell adhesion in this type of combined drug resistance.
The A2780CBN cell line contained most of the significant difference for two genes, GCLC and GSTO2. GCLC codes for γ-glutamylcysteine synthetase which controls the rate limiting step in the synthesis of glutathione while GSTO2 produces glutathione S-transferase omega 2. The combination of the two is known to play a role in anticancer drug resistance, including cisplatin resistance [11, 91, 92]. The increased expression of both genes in the A2780CBN line confirms that the importance of the glutathione pathway in carboplatin resistance, besides cisplatin resistance. A novel change with most of the difference in expression contained in the A2780DXL line is the upregulation of the LGI1 gene, which we have found in another docetaxel resistant line (MCF7txt) (A. Parissenti, unpublished data). The LGI1 gene was originally observed in glioma where increased expression of LGI 1 contributes to decreased proliferation of neuroblastoma cells [93, 94]. A decrease in proliferative capacity, as we observed for A2780DXL (Figure 3), could be promoted by changes in genes like LGI1. Overexpression of cytochrome enzymes, especially of the CYP 450 3A family [95, 96] are known to play a role in the metabolism of docetaxel. However, the A2780DXL line contained most of the significant difference for another cytochrome, CYP1B1, which does not play a role in metabolism of the drug although increased expression of CYP1B1 has been shown to be associated with resistance to docetaxel . However, an oxidized CYP1B1 estrogen metabolite has been reported to inhibit tubulin polymerization . Interestingly, expression levels of CYP1B1 are down regulated in our A2780DXL line (Table 4, Additional file 3: Figure S1), which contradicts the study by Martinez et al. , but seems to support the role of docetaxel in inhibiting tubulin polymerization reported by Sissung et al. .
Other genes found to be significantly different in the resistant lines compared to the parent line, were not mainly expressed in any one of the cell lines. The ABCB1 and ANXA1 genes, although previously shown to be associated with drug resistance [10, 14] were significantly different in all three cell lines, showing major changes in expression, but without any one line containing most of the difference. The remaining genes (Table 5) displayed a very similar change in expression across the cell lines without significant distribution of expression to one cell line.
Immunoblot confirmation of changes in protein expression
Immunoblots were performed to determine if changes in gene expression at the transcript level could be confirmed at the protein expression level. Of the five successful immunoblots, only the GCLC protein demonstrated a significantly different degree of expression in a cell line; upregulation in the A2780CBN cell line (Figures 8 and 9), confirming the glutathione pathway as a strong component of the resistance mechanisms in the A2780CBN cell line. However, the immunoblot data confirm the ANOVA results for both the ANXA1 and MT2A protein expression. As shown in Table 5, all three resistant cell lines display variable and quite different expression of ANXA1 transcripts and this is reflected by the immunoblot results (Figures 8 and 9). Although the MT2A blots seem to show a noticeable difference in the A2780DXL line, the fluctuating amounts of protein detected support the conclusion of no significant difference among cell lines displayed in Table 5. It is curious that both the ANXA1 and MT2A blots contradict the expectation from the microarray data, which indicated that expression of these two genes was specific to the carboplatin resistant line. The CYP1B1 blot follows the same trend of not supporting A2780DXL specific down regulation although this was demonstrated by both microarray and QPCR analysis. Finally, despite lack of statistical significance, expression of the AKR1C3 protein tends to be greatest in the dual resistant A2780CBNDXL line, which would support the microarray and QPCR results demonstrating a significant association of this gene with combined carboplatin and docetaxel resistance. The low concordance between the microarray, Q-PCR and protein expression data is not entirely surprising as this has been observed in other studies of gene and protein expression [99–101]. These studies show that there is not always a direct correlation between transcription levels and translation of a gene product, which indicates that caution should be observed in assuming that gene expression data can predict protein levels. Accurate knowledge of gene translation requires assessment of protein expression.