Expression of Mitochondrial Regulators PGC1α and TFAM as Putative Markers of Subtype and Chemoresistance in Epithelial Ovarian Carcinoma

Epithelial ovarian carcinoma (EOC), the major cause of gynaecological cancer death, is a heterogeneous disease classified into five subtypes. Each subtype has distinct clinical characteristics and is associated with different genetic risk factors and molecular events, but all are treated with surgery and platinum/taxane regimes. Tumour progression and chemoresistance is generally associated with major metabolic alterations, notably altered mitochondrial function(s). Here, we report for the first time that the expression of the mitochondrial regulators PGC1α and TFAM varies between EOC subtypes; furthermore, we have identified a profile in clear-cell carcinoma consisting of undetectability of PGC1α/TFAM, and low ERα/Ki-67. By contrast, high-grade serous carcinomas were characterised by a converse state of PGC1α/TFAM, ERα positivity and a high Ki-67 index. Interestingly, loss of PGC1α/TFAM and ERα was found also in a non-clear cell EOC cell line made highly resistant to platinum in vitro. Similar to clear-cell carcinomas, these resistant cells also showed accumulation of glycogen. Altogether, our data provide mechanistic insights into the chemoresistant nature of ovarian clear-cell carcinomas. Furthermore, these findings corroborate the need to take into account the diversity of EOC and to develop subtype specific treatment strategies.


Introduction
Epithelial ovarian carcinoma (EOC) is the major cause of gynaecological cancer death in women [1]. It is treated with surgery and platinum/taxane regimens as first-line therapy, but despite initial responsiveness, .75% of patients relapse into metastatic and essentially incurable chemoresistant disease [1,2]. EOC is a heterogeneous disease that is broadly classified as serous (high-and low-grade serous carcinoma (HGSC and LGSC, respectively)), or non-serous (endometrioid, EC; mucinous, MC; clear-cell, CCC) histology. Each subtype has distinct clinical characteristics and is associated with different genetic risk factors and molecular events [3,4]. Sensitivity to chemotherapy also varies between them, and survival correlates with subtype [5]. In CCC, both response rates and five-year survival are lower than in HGSC [5][6][7] and CCC has been generally accepted as an aggressive subtype that is poorly responsive to existing chemotherapy regimes.
To support the increased demand for energy and building blocks during tumour progression, cancer cells undergo major metabolic alterations [8,9]. Chemoresistance and metastasis are associated with such alterations, notably altered mitochondrial function(s) [8]. Whereas normal cells depend on mitochondrial oxidative phosphorylation (OxPhos) for energy, tumour cell metabolism includes, among other pathways, decreased OxPhos and increased glycolysis (the Warburg effect) and glutaminolysis, both of which require mitochondria [9,10]. Tumour cells also show survival-promoting alterations e.g. in regulation of autophagy in response to nutritional and other conditions of cellular/ metabolic stress; these processes involve the AMP-activated protein kinase AMPK [11].
In normal cells, the transcriptional coactivator peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1a) is a major coordinator of metabolism and mitochondrial function [12] and together with SIRT1 and AMPK constitutes a sensitive network responding to metabolic stress [13]. Together with SIRT1, PGC1a has been reported to reside inside mitochondria with direct functions on mitochondrial biofunction [14]. PGC1a also regulates mitochondrial biogenesis through regulation of NRF1-2 [15], which in turn regulates mitochondrial transcription factor A (TFAM) [16]. TFAM plays a critical role in maintaining copy number and structure of mtDNA [17,18], and is hence crucial for efficient transcription of mtDNA genes such as cytochrome c oxidase subunit 2 (MT-CO2) and other OxPhos proteins [17]. TFAM binding to mtDNA is regulated via phosphorylation [19]; unbound TFAM may be degraded by the Lon protease [19,20]. Although well studied in normal cells, for instance regarding aging and neurodegeneration [21,22], diabetes and obesity [23], it remains unknown what roles PGC1a/TFAM play in tumour progression. Mitochondrial biogenesis has, in addition to PGC1a, in several studies been linked to expression and localisation of oestrogen (E 2 ) and oestrogen receptor alpha (ERa) [24,25]. The EOC subtypes express ERa at different levels, and expression is lowest in CCC [26].
The aim of this study was to examine the role of mitochondrial regulation in EOC by examining and correlating the expression of the mitochondrial and metabolic markers PGC1a, TFAM and ERa in clinical samples, and to relate the expression to EOC subtypes. Furthermore, we examined these markers in an EOC cell line and a multiresistant subline thereof.

Quantitative real-time PCR
RNA was extracted using Qiagen RNeasy mini kit (Qiagen) and 500 ng/sample was used for the generation of cDNA using High Capacity cDNA Reverse Transcription Kit (Life Technologies) according to manufacturer's instructions. Quantitative real-time PCR (qRT-PCR) was performed on a 7900 HT Fast Real-Time system using TaqMan Gene expression assays (PPARC1A: Hs01016719_m1, TFAM: Hs01082775_m1, ESR1: Hs0017486 0_m1) and TaqMan Fast Universal PCR Master Mix (all from Applied Biosystems/Life Technologies) with the following cycling conditions: 95uC for 20 sec, 40 cycles of 95uC for 1 sec and 60uC for 20 sec. Expression of ACTB (Hs99999903_m1) was used as an endogenous control and the fold gene expression was calculated using the 2 2DDCT -method [28]. Raw data are shown in file Raw Data S1.

Study population and ethical considerations
The study set (n = 53 cases) was drawn from cases originally diagnosed at the Karolinska University Hospital (Stockholm, Sweden). Participants provided written informed consent and the study was carried out with the approval of the Regional Ethics Committee of Stockholm (EPN 2013/318-31/2). Two pathologists reviewed all cases. The diagnoses included in the study set included 21 HGSC (40%), 14 CCC (26%), ten EC (19%), 7 MC (13%) and one LGSC (2%). Clinicopathological features of the study population can be found in Table 1. A tissue microarray (TMA) was constructed using paired 1 mm cores from representative tumour material. Representativity was confirmed using a hematoxylin and eosin control section.

Immunohistochemistry and antibodies
Immunohistochemical staining (IHC) was performed on whole TMA using the EnVision + System (Dako). Antigen retrieval was achieved by immersing the sections in citrate buffer at pH 6.0 and boiling for 20 minutes in a water bath. The sections were then incubated with primary rabbit polyclonal antibodies against human voltage-dependent anion channel (VDAC) and PGC1a (both from Abcam, 1:800 and 1:100, respectively), TFAM and Lon (both from Sigma-Aldrich, 1:10 and 1:800, respectively) and rabbit monoclonal antibodies against human MT-CO2 (1:100, Abcam) for 30 min at room temperature. A positive reaction was detected using the EnVision +System (Dako) and slides were counterstained with hematoxylin (Dako). Staining of Ki-67, ERa and progesterone receptor (PR) was performed at the routine laboratory at the Department of Pathology-Cytology, Karolinska University Hospital (Stockholm, Sweden), using rabbit monoclonal antibodies from Ventana Medical Systems.  Table 2. Expression status of PGC1a and TFAM in the EOC study population.

Evaluation of immunohistochemistry
For evaluation of immunoreactivity and histological appearance, all TMA slides were examined independently by two investigators (MG and MS). Positive cytoplasmic expression of VDAC, ERa, PGC1a, TFAM, MT-CO2 and Lon was defined as immunoreactivity in .1% of epithelial cells, while immunoreactivity in ,1% of epithelial cells defined negative protein expression. The scoring for each individual sample can be seen in the Raw Data S1. Ki-67 index was assessed as a continuous variable as percentage of positive tumour cells.

Glycogen assay
Cells were grown to confluency prior to harvesting. Intracellular levels of glycogen were then detected in 1610 6 cells using the colorimetric Glycogen Assay Kit II (Abcam) according to manufacturer's instructions.

Sulforhodamine B assay
Growth rates in SKOV-3 and SKOV-3-R cells were evaluated based on total cellular protein in the samples, in turn assessed using the sulforhodamine B-based TOX6 kit (SRB assay, Sigma-Aldrich) according to manufacturer's instructions. Briefly, SKOV-3 and SKOV-3-R cells were seeded in 96-well plates at 4610 3 cells/well in quadruplicates for each time point. After overnight incubation the wells for t = 0 h were fixed while the remaining wells were incubated additional 72 h before fixation. Resulting absorbances (i.e. total cellular protein) were compared with that at t = 0 h. siRNA transfection SKOV-3 cells seeded at 7.5610 4 cells/well (12-well plates) or 15610 4 cells/well (6-well plates) were transfected with 50 nM Silencer siRNA specific for PPARGC1A or with 50 nM Silencer Negative Control No. 1 siRNA (Life Technologies, id: 114747 and AM4611, respectively) after overnight incubation. siRNA, diluted in Opti-MEM (Life Technologies), was mixed by vortexing with 3 ml HiPerFect Transfection Reagent (Qiagen) per 100 ml Opti-MEM/siRNA mixture followed by 10 min incubation at RT. Subsequently, transfection complexes were added to the cells and incubated for 5 h under normal growth conditions, whereafter these complexes were replaced by antibiotic-free growth medium. At 72 h post-transfection, mRNA and protein levels were analysed by qRT-PCR and western blot, respectively. In parallel experiments, at 24 h post-transfected cells were treated with cisplatin (Hospira) for 48 h. Response to treatment was evaluated using the SRB assay.

Statistical analysis
To assess the relationship between protein expression levels and the variables in the clinical material, a non-parametric Chi 2 test or Fisher's exact test was used as indicated in the tables. Comparisons between groups were assessed using the independent t-test (in vitro data), Mann-Whitney U test or Kruskal-Wallis test as indicated (IHC data). A two-tailed P value ,0.05 was considered significant, and p values ,0.05, ,0.01 or ,0.001 are represented with one, two or three asterisks, respectively, in the figures. IBM SPSS 20.0 statistical software (SPSS Inc.) was used for all analyses. For IHC analysis of MT-CO2 and Lon, two and five samples, respectively, were unavailable for scoring, which was taken into account in the statistical analysis.

PGC1a and TFAM expression in multiresistant EOC cells
The EOC cell line SKOV-3 was subjected to repeated longterm treatment with cisplatin generating the multiresistant subline SKOV-3-R, as previously described in [27]. Since mitochondrial content was previously found to be higher in the resistant subline [27], expression of mitochondrial regulators PGC1a and TFAM was investigated here. mRNA expression of the genes coding for PGC1a and TFAM (PPARGC1A and TFAM, respectively) was significantly decreased in SKOV-3-R compared to parental SKOV-3 (p,0.001 for both) (Fig. 1A). Accordingly, protein expression of PGC1a and TFAM was reduced in SKOV-3-R cells (Fig. 1B).
The mitochondrial outer membrane VDAC was used as a positive control for presence of mitochondria. Expression of VDAC was found in all samples and was equal across all EOC subtypes (not shown).
Across the EOC subtypes, there was no significant correlation between TFAM, PR, age or histological stage. However, tumours expressing TFAM displayed a significantly higher Ki-67 index compared to TFAM negative tumours (median score 33% and 20%, respectively) (p = 0.014) (Fig. 2). There was also a significant positive correlation between TFAM and expression of ERa (p = 0.014). In order to investigate if lack of TFAM expression affected the expression of mtDNA-encoded genes, protein expression of MT-CO2 was assessed. Of 51 tumours stained for MT-CO2, 47 (92%) were positive and four (8%) were negative. Three of the four MT-CO2 negative tumours were also TFAM negative; however, there was no significant relation between expression of MT-CO2 and TFAM (Table S1 in File S1).
TFAM protein levels are in part regulated through degradation by the Lon protease [19,20]. To investigate if increased levels of Lon might explain the reduced levels of TFAM found in CCC, expression of Lon was assessed. Lon was localised to the cytoplasm, and was found in 38 tumours (79%) which were differently distributed across all EOC subtypes (p = 0.032). There was no significant correlation between expression of TFAM and Lon (Table S2 in File S1).

Co-expression of PGC1a and TFAM in EOC tumours
As PGC1a is a major coordinator of mitochondrial functions including regulation of transcription of TFAM [12,15,16], we investigated the co-expression of these two proteins. The cohort was categorised into four different groups depending on their expression of PGC1a and TFAM (PGC1a-/TFAM-; PGC1a-/ TFAM+; PGC1a+/TFAM-and PGC1a+/TFAM+). The expression of PGC1a/TFAM varied significantly across the EOC subtypes (p = 0.001) ( Table 3, Fig. 3). There was also a significant difference in median Ki-67 index between the four groups, where positive expression of TFAM associated with increased Ki-67 index independently of PGC1a expression (p = 0.048) ( Table 3, and Fig. S3). Furthermore, there was a positive association between ERa expression and co-expression of PGC1a/TFAM (p = 0.002) ( Table 3), whereas no significant association with PR was observed.
Co-expression of PGC1a/TFAM was found in 33 samples (62% of the total cohort) of which 16 (48%) were HGSC. Within the HGSC subtype, 76% of the tumours were double-positive for PGC1a/TFAM (p,0.001). Lack of both PGCa and TFAM was observed in nine tumours (17% of total cohort). Of these nine, all but one were CCC (89% of all double negative) (p = 0.020). Within the CCC subtype, double-negative expression occurred in 57% of the tumours and all of these were ERa negative (Table 3).

CCC profile of multiresistant EOC cell line
SKOV-3 cells have been reported to have a 32-basepair deletion in exon 1 of ERa transcript and are thus ERa positive but oestrogen-insensitive [29,30]. To further investigate the link between regulation of ERa expression, as such, and PGC1a/ TFAM, levels of ERa were analysed on mRNA and protein levels. Compared to parental SKOV-3, the multiresistant SKOV-3-R cells show significantly decreased mRNA expression of the ESR1 gene (p,0.001) (Fig. 4A), and accordingly decreased ERa protein expression (Fig. 4B). Raw data can be found under S_PCR in the file Raw data S1. Accumulation of glycogen is characteristic of CCC [31], wherefore we investigated intracellular glycogen levels in SKOV-3 and SKOV-3-R cells. The SKOV-3-R were found to accumulate 25% more intracellular glycogen (p = 0.027) compared to parental cells (Fig. 4C).

Discussion
We have here shown that the expression of mitochondrial regulators PGC1a and TFAM varies significantly between HGSC and CCC. Furthermore, we identified a profile in CCC consisting of undetectability of PGC1a/TFAM, and low ERa/Ki-67. By contrast, HGSC samples were characterised by a converse state of PGC1a/TFAM and ERa positivity as well as a high Ki-67 index (summarized in Fig. 5). Interestingly, the CCC profile had developed also in an EOC cell line that is not of CCC origin and which had been made highly chemoresistant in vitro.
Here, out of the nine tumour samples that were double-negative for both PGC1a/TFAM, eight were CCC. In view of its characteristic histology, accumulation of intracellular glycogen and its highly chemoresistant nature, CCC represents a distinct clinicopathological subtype of requires new treatment strategies  PGC1a and TFAM as Biological Markers in Ovarian Carcinoma PLOS ONE | www.plosone.org [32]. We believe that our findings support this notion; furthermore, our data also provide mechanistic insights into where this chemoresistance comes from.
Very little is reported on the role of PGC1a in cancer in general, and even less in chemoresistance. Low levels of PGC1a have been reported in ten EOC samples compared with normal ovaries, and PGC1a overexpression in vitro led to upregulation of Bax, downregulation of Bcl-2 and to apoptosis [33]. PGC1a was reported to be low also in 15 out of 17 colon cancer biopsies compared to normal [34]. There were, however, no reported results on any association with data on treatment effects or chemoresistance. Nevertheless, the role of the PGC1a/oestrogenrelated receptors (ERR) axis is known to be instrumental in regulation of cancer cell metabolism (reviewed in [35]), which in turn is known to affect responses to chemotherapy [8,36].
Low expression of PGC1a might be expected to correlate with lower mitochondrial content. Mitochondria are required for the full effect of some chemotherapeutic drugs, e.g. cisplatin, due to their role in apoptosis and a requirement for mitochondrially produced ROS [37]. However, we did not observe any correlation between PGC1a expression and mitochondrial content, based on VDAC and MT-CO2 expression. PGC1a is involved also in regulation of antioxidant defence [21,38], and the observed low levels of PGC1a would thus be expected to not reduce ROS production and thereby the effect of platinum. Therefore, the role of PGC1a in EOC, particularly in notoriously resistant CCC [32], is likely not linked to antioxidant defence.
TFAM is a requirement for normal replication and transcription of mtDNA [17,18]; this is evidenced as reduced mtDNA content in colon carcinoma cell lines as well as in biopsies carrying truncating mutations in TFAM [39]. Here, of the four samples that did not express mitochondrially encoded MT-CO2, three did not show detectable TFAM. This suggests the possibility that a detectable proportion of EOC tumours do not have functional respiration, as MT-CO2 provides the substrate-binding site of Complex IV in the respiratory chain [40]. Based on the literature [18,41], we hypothesised that minute amounts of TFAM are sufficient for transcriptional activity and that the undetectability of TFAM protein might be due to Lon-mediated degradation [19,20]. However, as there was no significant correlation between the expression of TFAM and Lon, this was not the case.
CCC generally shows low Ki-67 index [32]. A correlation between TFAM and Ki-67 has, to our knowledge, not been studied. However, in analogy with the report that in highly proliferative transformed lymphoblastoid cells, increased expression of TFAM was suggested to be associated with protection from oxidative stress induced damage [42], we speculate that highly proliferative HGSC tumours, as indicated by Ki-67, upregulate TFAM as part of an antioxidant defence. Our in vitro results are also in accordance with this, since SKOV-3-R cells showed downregulation of TFAM and grow more slowly than the parental cells (Fig. S4). Moreover, TFAM expression correlated positively with ERa, which might reflect a potential for growth as well as for an efficient oxidative phenotype. In line with previous studies [26,43], expression of ERa varied across the EOC subtypes, with low levels in CCC.
In a mouse model, SKOV-3 tumour xenografts (non-CCC, EOC cells) treated with trastuzumab were at post-treatment found to be heterogeneous with large areas of ERa negative CCC morphology [44]. This finding is interesting in the present context, as the multiresistant SKOV-3-R cells had acquired the CCC profile, compared to parental SKOV-3 cells. Furthermore, SKOV-3-R showed accumulation of glycogen, again in accordance with a CCC phenotype. Knocking down PPARGC1A mRNA using siRNA in the SKOV-3 cells down-regulated PGC1a mRNA as well as protein levels seen at 72 h post transfection (Figs. S5A-B). TFAM protein was also reduced, although TFAM mRNA expression was restored at this time point (Figs. S5A-B).
Although not significant, the knockdown cells showed a clear tendency of reduced response to cisplatin treatment compared to the negative control cells (Fig. S5C). This tendency was not due to reduced growth, as knocking down PPARGC1A did not alter the growth rate or the expression of Ki-67 positive cells. We speculate that long-term downregulation of PGC1a and/or TFAM is necessary to decrease growth rate over time.
Although our SKOV-3 cell line model represents only itself, we suggest that downregulation of PGC1a activity as part of a chemoresistant phenotype profile might not be restricted to ovarian CCC. Low PGC1a activity would be expected in tumour contexts where loss of LKB1/AMPK represents loss of tumour suppressor function [45]. In keeping with a context-dependent tumour suppressor and oncogenic role for AMPK [45], the comparatively high levels of PGC1a expression that we observed in the highly proliferative, i.e., with high Ki-67 expression, HGSC tumours could then be suggested to belong to a context of AMPKmediated protection from proliferation-related oxidative stress.
In summary, we propose that the profile consisting of PGC1a/ TFAM downregulation and decreased ERa/Ki-67, and likely also glycogen accumulation, is representative of CCC and chemoresistance. We are currently investigating an underlying mechanistic explanation for this association. Fig. 5 summarises how HGSC and CCC may represent opposites with regard to the profile. Altogether, the findings corroborate the need to take into account the diversity in EOC and to develop specific treatment strategies for each subtype. File S1 Supporting tables. Table S1, Expression of MT-CO2 in EOC. Shown is the distribution of positive and negative expression of MT-CO2 in the total sample, in the five EOC subgroups and in TFAM-expressing and non-expressing samples, respectively. Table S2, Expression of Lon in EOC. Shown is the distribution of positive and negative expression of Lon in the total sample, in the five EOC subgroups and in TFAM-expressing and non-expressing samples, respectively. (DOCX) Raw Data S1 S_PCR: Gene expression raw data given as Ct, delta-Ct (DCt) and delta-delta-Ct (DDCt) values, respectively, for results shown in Figs