PSMB7 is associated with anthracycline resistance and is a prognostic biomarker in breast cancer

Background: To date individual markers have failed to correctly predict resistance against anticancer agents in breast cancer. We used gene expression patterns attributable to chemotherapy-resistant cells to detect potential new biomarkers related to anthracycline resistance. One of the genes, PSMB7, was selected for further functional studies and clinical validation. Methods: We contrasted the expression profiles of four pairs of different human tumour cell lines and of their counterparts resistant to doxorubicin. Observed overexpression of PSMB7 in resistant cell lines was validated by immunohistochemistry. To examine its function in chemoresistance, we silenced the gene by RNA interference (RNAi) in doxorubicin-resistant MCF-7 breast cancer cells, then cell vitality was measured after doxorubicin treatment. Microarray gene expression from GEO raw microarray samples with available progression-free survival data was downloaded, and expression of PSMB7 was used for grouping samples. Results: After doxorubicin treatment, 79.8±13.3% of resistant cells survived. Silencing of PSMB7 in resistant cells decreased survival to 31.8±6.4% (P>0.001). A similar effect was observed after paclitaxel treatment. In 1592 microarray samples, the patients with high PSMB7 expression had a significantly shorter survival than the patients with low expression (P<0.001). Conclusion: Our findings suggest that high PSMB7 expression is an unfavourable prognostic marker in breast cancer.

The major cause of cancer therapy failure is either primary drug resistance or the development of secondary resistance against currently available antineoplastic agents. Many different mechanisms of drug resistance have been identified, including the overexpression of members of the family of adenosine triphosphate-binding cassette (ABC)-transporters such as P-glycoprotein (ABCB1), the multidrug resistance-associated proteins MRP1 and MRP2 and the mitoxantrone resistance protein/breast cancer resistance protein (Gottesman et al, 2002). Drug resistance is also mediated by defects in intrinsic cellular safeguard mechanisms capable of activating apoptosis. The loss of programmed cell death induced by antitumour drugs can be achieved through disruption of regulators of DNA damage signalling, such as p53 and bcl-2. Moreover, alterations in cell cycle checkpoints, repair of damaged cellular targets and various additional more or less wellcharacterised mechanisms can contribute to the resistance.
The anthracycline antibiotic doxorubicin (adriamycin) is widely used for treatment of breast, ovarian, bronchogenic and gastric solid tumours, lymphomas and leukaemias (Martin et al, 2003).
Triple negative (ER-, PR-, HER2-) breast patients receive anthracycline-based chemotherapeutic agents in monotherapy or in combination protocols. Although a number of different mechanisms have been proposed for the cytotoxic effect of anthracyclines, the primary mechanism of drug action is likely to be the inhibition of DNA biosynthesis through topoisomerase II binding and consequently conferring an S/G2 cell cycle arrest (Gewirtz, 1999). As a topoisomerase inhibitor, it potentially induces apoptosis in normal cells as well (Sabourin and Osheroff, 2000).
In several reports, authors applied DNA array technologies to find genes correlated with doxorubicin resistance (Ayers et al, 2004;Kang et al, 2004;Folgueira et al, 2005;Gianni et al, 2005;Cleator et al, 2006;Gyorffy et al, 2006;Hess et al, 2006). In our earlier study, we used gene expression patterns of closely related chemotherapy-resistant and -sensitive parental cancer cell lines to identify discriminatory genes associated with drug resistance. The use of a set of different cell lines for the identification of discriminating genes allows a tissue-independent application for predicting therapy response. We applied our model to predict sensitivity in a set of 44 breast cancer samples. The patient group characterised by the gene expression profile similar to that of drug-sensitive cell lines exhibited over 50% longer survival than the group exhibiting the profile characteristic of resistant cells (Gyorffy et al, 2005). Overall, in these gene expression signatures, a considerable number of genes were associated with doxorubicin resistance. However, it is desirable to discriminate markers associated with treatment response and markers responsible for resistance.
The proteasome is a multicatalytic proteinase complex with a highly ordered ring-shaped 20S core structure. The core structure is composed of 4 rings of 28 non-identical subunits; 2 rings are composed of 7 a subunits and 2 rings are composed of 7 b subunits. Proteasomes are distributed throughout eukaryotic cells at a high concentration and cleave damaged and needless peptides in an ATP/ubiquitin-dependent process in a non-lysosomal pathway. Proteasome inhibitors are drugs blocking the action of cellular complexes that break down proteins. They have an antitumour activity in cell culture, and induce apoptosis by disturbing regulation of cell cycle proteins (Adams et al, 1999). Bortezomib was the first proteasome inhibitor, which is already used in clinics as a chemotherapy agent (Kane et al, 2006). Bortezomib effectively eradicates myelome multiplex by activating the stress apoptosis signal of endoplasmic reticulum (Schewe and guirre-Ghiso, 2009). So far, several studies examined its effect on proteasomes in different circumstances (Ludwig et al, 2005;Montagut et al, 2005;Montagut et al, 2006;Sterz et al, 2008). Proteasome inhibitors might also increase chemosensitivity. Cotreatment of doxorubicin and bortezomib can conversely increase the efficiency of each other (Ciolli et al, 2008). However, to date the direct association with doxorubicin resistance as well as the potential prognostic or predictive value of selected proteasome subunits has not been investigated.
In this study, our intention was to identify gene expression signatures obtained from resistant cells in long-term culture, untreated sensitive parental cells and parental cells after short-term drug administration. On the basis of relevance in clinical samples, we selected one proteasome subunit, the proteasome unit b-type 7 (PSMB7) for further examination using RNA interference (RNAi) to validate its function in doxorubicin resistance. Our aim was also to identify effects of paclitaxel crossresistance on the resistant cell line. Finally, microarray data from 1592 patients were used to examine the clinical relevance of PSMB7.

Sensitivity of cells to doxorubicin and paclitaxel
Cell proliferation assay (Cell Proliferation Kit I (MTT) Roche, IN, USA,) was used to monitor sensitivity of doxorubicin-resistant cell lines. Protocol was carried out as directed by the manufacturer. Briefly, 5000 cells per well in a microplate were plated, and drug was added after an overnight incubation in increasing grade of 0.0001x -100x of the clinically administered dose (doxorubicin: 0.02 mg ml -1 , paclitaxel: 0.025 mg ml -1 ) in triplicates. Six days later, MTT (3-[4, 5-dimethylthiazolyl-2]-2, 5-diphenyltetrazolium bromide) was added to each well for 4 h. After this incubation period, the formed purple formazan salt crystals were solubilised and quantified 1 day later with the use a BioTek PowerWave XS Microplate Spectrophotometer.

Autofluorescence of doxorubicin
A total of 50 000 sensitive MCF-7 cells per well were plated in triplicate in a 96-well plate for four different conditions: (1) cells treated by doxorubicin, (2) untreated cells, (3) doxorubicin in L15 medium without cells and (4) L15 medium alone. Doxorubicin treatment (0.2 mg ml -1 ) was carried out on the second day for 1.5 h. MTT staining and reading the results were performed as described above. Differences were assessed by using a t-test.

RNA isolation and cDNA synthesis
RNA was isolated from 1 Â 10 7 cells in logarithmic growth phase using the Qiagen RNeasy Mini Kit following the manufacturer's protocol (Qiagen GmbH, Hilden, Germany). Isolated total RNA was quantified by UV-spectroscopy and quality checked by analysis on a LabChip (BioAnalyzer, Agilent Technologies, Santa Clara, CA, USA). RNA samples were stored at À801C. cDNA was synthesised from 5 mg total RNA by firstly annealing to 5 pmol ml -1 HPLC purified T7-(dT)24 primer (MWG-Biotech, Ebersberg, Germany) at 701C for 10 min. Second, reverse transcription, second-strand synthesis and cleanup of double-stranded cDNA was performed according to the protocols provided by Affymetrix (http://www.affymetrix.com/index.affx). Synthesis of biotin-labelled cRNA was performed using the BioArray High Yield RNA Transcription kit (Enzo Diagnostics, Farmingdale, NY, USA). cRNA concentration was determined by UV-spectroscopy and the distribution of cRNA fragment size was checked on a LabChip (BioAnalyzer, Agilent Technologies).

Hybridization protocol
The fragmented cRNA was hybridised to the HG-U133A arrays (Affymetrix, Santa Clara, CA, USA) in a hybridization oven at 451C for 16 h. Subsequent washing and staining of arrays was performed using the GeneChip fluidics station protocol EukGE-WS2. Finally, probe arrays were scanned using the GeneChip System confocal scanner (Hewlett-Packard, Santa Clara, CA, USA). For each cell line triplicates (sensitive, resistance and treated), hybridization was made once, so altogether 12 hybridizations were made. The microarray data can be downloaded from GEO (accession no. GSE3926).

Pre-processing of microarray data
Quality control analysis was performed according to the suggestions of The Tumor Analysis Best Practices Working Group (Hoffman et al, 2004). Scanned images were inspected for artefacts. All RNA targets included in the analysis exhibited present calls of 425% and were not degraded. We have applied RMA (Irizarry et al, 2003) for the normalization of hybridization intensities. The pre-processing was performed using Bioconductor packages in the R statistical environment. RMA is one of the most common array normalization methods supplying cross-project normalization with good specificity and excellent sensitivity.

Feature selection
We arranged the complete data set into three classes according to the resistance and treatment properties of the cell lines. We compared drug-sensitive parental, doxorubicin/daunorubicintreated parental and resistant cell lines. For identifying discriminating genes, the Prediction Analysis for Mircorarrays (PAM v.1.23) package was used as described earlier (Tibshirani et al, 2002). PAM uses soft thresholding for producing a shrunken centroid, which allows the selection of genes with high discriminative potential. An overview of the applied statistical approach is presented in Figure 1. We decided to pick the top 100 genes from each comparison for the selection of the best discriminatory group of genes. Then genes were identified, which discriminate between parental and treated parental cell lines (reflecting effect of the drug treatment) and resistant and parental cell lines (reflecting genes of resistance).

Additional analyses
Hierarchical clustering was performed using the Genesis software (Sturn et al, 2002). Before clustering, we performed a second normalization at the gene level to set the average expression of each transcript to 0 to present gene inductions or repressions of identical magnitude as numerically equal. Gene annotations were performed using the Netaffx Analysis Centre (http://www.affymetrix. com/analysis/index.affx).

Oligos for RNAi
Target of siRNA oligos for PSMB7 were designed using siRNA Target Finder (http://www.ambion.com) and siDESIGN Center (http://www.dharmacon.com) softwares. DNA oligonucleotide templates for three target mRNAs were ordered (Csertex Kft, Budapest, Hungary), then were synthesised by Silencer siRNA Construction Kit (Applied Biosystems, Darmstadt, Germany) using the manufacturer's instruction. Synthesised siRNA duplex concentrations were measured by NanoDrop ND-1000 spectrophotometer (BCM, Houston, TX, USA) then diluted as needed. PCR primers were designed by Primer3 software (http://frodo.wi.mit. edu/) and their binding site was verified by NCBI BLAST (http:// blast.ncbi.nlm.nih.gov/Blast.cgi). Out of three synthesised siRNA oligos, one silenced effectively PSMB7 better than 80%. Figure 3 shows binding location of every synthesised oligos and PCR primers (effective siRNA is framed by continuous line). Effectiveness of siRNA oligos was validated by RT -PCR ( Figure 3B).

SiRNA transfection of MCF-7 cell lines
Transfection was carried out with 5 ml SiPORT NeoFX Transfection Reagent (Applied Biosystems, Budapest, Hungary) in six-well plates. A total of 230 000 cells were plated into each well. The final concentration of siRNA was 10 nM in 2.5 ml serum and antibioticfree medium. After 24 h of incubation, medium was washed twice with sterile PBS, then normal growth medium was added. Total RNA was extracted in the 48th h after transfection with Qiagen RNeasy Mini Kit as described earlier, cDNA synthesis and DNA amplification were carried out by OneStep RT -PCR Kit (Qiagen GmbH) with gene-specific PCR primers. GAPDH was used as internal control. We used 24 amplification cycles for amplification. Amplification products were separated on 2% agarose gel stained with ethidium-bromide. The effect of silencing was calculated by Adobe Photoshop CS2 software.

Combination of RNAi and chemotherapy treatment
To investigate the function of selected gene in chemoresistance, we combined RNAi and drug treatment. PSMB7 overexpressed in resistant cells was silenced in doxorubicin-resistant MCF-7 cell lines (MCF-7-RAdr); 24 h after transfection, the reaction was stopped by replacing the medium with normal growth medium containing drug in concentration of 0.2 mg ml -1 . Cells were trypsinised on the 72nd h and counted by CASY DT Cell Counter (Innovatis AG, Reutlingen, Germany). We used negative siRNAtreated (AllStars Negative Control siRNA, Qiagen GmbH) MCF-7-RAdr cells, siRNA-untreated MCF-7-RAdr cells and siRNAuntreated MCF-7 cells for controls with their doxorubicin-treated analogue. For a statistical analysis, the number of living cells was counted in each well. Experiments were carried out three times, and each well was measured three times in each experiment. The number of drug-treated cells was normalised to untreated cells in siRNA-treated, negative siRNA-treated and siRNA-untreated wells. The t-test was used for analysis of difference between groups. Significance level was set at P ¼ 0.05.

Validation on clinical samples
For in silico validation of the genes, a database containing processed GEO microarray samples was established as described earlier (Gyorffy and Schafer, 2009  First, genes were filtered to include only subunits of the proteasome (n ¼ 64). Then a survival analysis for each of these probe sets were performed using BRB Arraytools 3.8.0-b_1 package (developed by Dr Richard Simon and Amy Peng Lam, available at http://linus.nci.nih.gov/BRB-ArrayTools.html). The false discovery rate (FDR) was set below 0.05. Finally, Kaplan -Meier plots were drawn to illustrate the effect on survival in these patients.

Identification of differentially regulated genes
To identify genes associated with the resistance to doxorubicin/ daunorubicin, we contrasted gene expression profiles of closely related drug-resistant and -sensitive cell lines derived from breast, pancreatic, colon and gastric cancer. For microarray analysis, RNA was prepared from non-treated parental cells, from parental cells treated with doxorubicin for 24 h and from resistant derivatives cultured continuously in the presence of doxorubicin at a concentration that completely eliminated the sensitive cells. The complete data set comprising raw data and Affymetrix .CEL files is available in the GEO database (http://www.ncbi.nlm.nih.gov/geo/) using the GEO accession number GSE3926. The prediction analysis of microarrays was performed to pick the top 100 genes differentially expressed in each sample set independent of tissue origin. To visualise discriminatory expression changes, we clustered the top genes (see Figure 2A and B). In these, three probe sets representing proteasome subunits were expressed stronger in doxorubicin-resistant cell lines, but not in doxorubicin-treated parental cells (see arrows in Figure 2).

Autofluorescence of doxorubicin
Examining the effect of doxorubicin autofluorescence on MTT results, there was no difference between doxorubicin-treated anduntreated cells (P ¼ 0.38). Fluorescence of doxorubicin supplemented L15 medium differed significantly from doxorubicintreated cells (Po0.01), but did not differ from untreated L15 medium (P ¼ 0.065).

Prognostic potential of 64 proteasome subunits
The normalised gene expression of the GEO-downloaded 1592 samples was filtered to include only the proteasome subunits present on the HG-U133A microarrays. The prognostic potential for all 64 probe sets was computed by dividing the patients into over-and underexpressed group as compared with the median (thus, 531 patients were in each group). The analysis resulted in 12 significant genes below an FDR of 0.05. Of these, only one was also detected by the microarray analysis (PSMB7), which was further investigated. Immunocytochemistry confirmed the overexpression of PSMB7 as observed on the microarrays ( Figure 2C and D).

Survival of cells after gene silencing and treatment with doxorubicin and paclitaxel
To examine the function of the PSMB7 gene in doxorubicin resistance, we combined gene silencing with drug treatment to assess its effect on cell survival. Viability of cells without doxorubicin treatment was better in both siRNA-untreated and siRNA-treated wells compared with cells with doxorubicin treatment. A specific cytotoxic effect of negative control siRNA was negligible as compared with the effect of gene-specific siRNA.
Data after normalization are presented in Figure 4; 79.8±13.3% of resistance cells survived after doxorubicin treatment. Combined with gene silencing, only 31.8 ± 6.4% of the MCF-7-RAdr cells survived. The significance between siRNA-treated and siRNAuntreated MCF-7-RAdr cells after doxorubicin treatment was P40.001. After doxorubicin treatment, 48.3±8.1% of sensitive cells survived. A total of 73.3% of negative control siRNA-treated cells survived. After combination of paclitaxel treatment and gene silencing, 22.6±4% of the MCF-7-RAdr cells survived compared with siRNA-untreated cells, whereas 43.8±6% of sensitive cells survived (data not shown). Relative cell vitality of siRNA-treated and siRNA-untreated MCF-7-RAdr cells after paclitaxel treatment differed significantly (P ¼ 0.03).

Biomarker potential of PSMB7 in clinical samples
To assess the prognostic power of PSMB7 gene in resistance, clinical samples were used to validate the results of cell culture model. We grouped the 1592 breast cancer patients on the basis of the expression of PSMB7 using the Affymetrix HGU133A probe set 200786_at. Of these, 963 out of 1220 patients were ER positive and 187 out of 1156 lymph node positive. The average relapse-free survival was 6.42 years. Unfortunately, detailed treatment history was generally not available. Patients above the median had a significantly shorter survival than patients below the median (Po0.001) ( Figure 5).

DISCUSSION
In our study we examined the function of the proteasome subunit PSMB7 gene in drug resistance in breast cancer cell line and in breast cancer patients. So far, no earlier study has risen up concerning drug resistance neither for PSMB7 gene nor any of proteasome subunits. In a cell culture model with a combination of RNAi and drug treatment, we validated the causative function of this gene in chemoresistance. The examination of 1592 breast cancer patient showed that overexpression of this gene is associated with poor prognostic outcome.
At the preliminary phase of this study, we contrasted the expression profiles of four pairs of different human tumour cell lines of gastric, pancreatic, colon and breast origin, their counterparts resistant to the topoisomerase inhibitors daunorubicin or doxorubicin and the sensitive parental cells after a 24 hchemotherapy treatment using Affymetrix HG-U133A microarrays. We also performed additional analyses to exclude the influence of doxorubicin autofluorescence (Karukstis et al, 1998) on our results. Finally, we identified the top transcripts associated with doxorubicin/daunorubicin resistance, but not with treatment response. So far, the majority of microarray-based chemoresistance-associated gene sets concentrated on finding genes associated with resistance rather than on the difference between drug response and drug resistance. Surprisingly, we have found genes such as proteasome subunits coding PSMB7 and PSMD13 having function in the mechanism of resistance that earlier have not been described. Of these, only the expression of PSMB7 was able to predict survival in clinical samples and was, therefore, further investigated using RNAi.
Elevated expression of proteasome catalytic subunits led to increase in proteasomal peptidase activities in these tissues. Oral administration of D3T also exerted a pharmacodynamic action in some brain regions of these mice and proteasomal peptidase activities were significantly elevated in the cerebral cortexhippocampus. These results indicate that increased proteasome expression by inducers may have a function in protection/ attenuation of protein aggregate-mediated disorders (Kwak et al, 2007b).
After in vitro inspection, we further validated the discriminating function of differential PSMB7 expression in 1592 publicly available microarrays . By dividing the patients as having overor underexpressed PSMB7 as compared with the median, we achieved high significance. However, our approach is limited  because of inadequate treatment information. A future study for fine tuning of PSMB7 expression with RT -PCR in a large clinical sample collection with detailed clinical follow-up could identify exact cut offs for thresholds, which could validate and enhance its discriminative potential in clinical use. We might consider the idea to silence overexpressed PSMB7 as part of a chemotherapy. However, despite promising in vitro results, the establishment of effective RNAi conditions in vivo is still complicated. A major impediment to the clinical use of RNAi therapy is the need to deliver these macromolecules to each and every cancer cell to trigger a direct and specific effect. Several studies have shown in vivo efficacy in the delivery of siRNAs using various strategies such as complexing siRNAs with cationic lipids, nanoparticles, polyethyleneimine, cyclodextrin, chitosan and collagen (Nguyen et al, 2008). Xenograft mouse model was also effectively used in several studies (Xiao et al, 2008;Xie et al, 2008;Zhang et al, 2008Zhang et al, , 2009Shi et al, 2009). In human beings, RNAi is already in clinical use in therapy of macula degeneration and respiratory syncytial virus infection, but so far no effective RNAibased inhibition of tumour progression is available.
As it was examined in the last decades, overexpression of individual genes can be associated with resistance against given agents. ABCB1 gene strongly correlates to chemoresistance (Clarke et al 1992;Kamata et al, 2008;Larbcharoensub et al, 2008;Shi et al, 2008), TOP2A is a potential gene for predicting anthracyclin resistance (Molina et al, 2005;Tanner et al, 2006;Harris et al, 2007;Mano et al, 2007). Expression of metallothioneins is linked to tamoxifen resistance (Surowiak et al, 2004), and the gene Tau is a predictor of resistance against neoadjuvant paclitaxel therapy Andre et al, 2007). The favourable outcome of this study suggests that PSMB7 gene has an important function in predicting both doxorubicin and paclitaxel resistance.
In summary, our findings support the function of the proteasome in the development of chemotherapy resistance. High PSMB7 expression is an unfavourable prognostic marker in breast cancer.