The novel oncogene KIF14 (kinesin family member 14) shows genomic gain and overexpression in many cancers including OvCa (ovarian cancer). We discovered that expression of the mitotic kinesin KIF14 is predictive of poor outcome in breast and lung cancers. We now determine the prognostic significance of KIF14 expression in primary OvCa tumors, and evaluate KIF14 action on OvCa cell tumorigenicity in vitro. Gene-specific multiplex PCR and real-time QPCR were used to measure KIF14 genomic (109 samples) and mRNA levels (122 samples) in OvCa tumors. Association of KIF14 with clinical variables was studied using Kaplan–Meier survival and Cox regression analyses. Cellular effects of KIF14 overexpression (stable transfection) and inhibition (stable shRNA knockdown) were studied by proliferation (cell counts), survival (Annexin V immunocytochemistry) and colony formation (soft-agar growth). KIF14 genomic gain (>2.6 copies) was present in 30% of serous OvCas, and KIF14 mRNA was elevated in 91% of tumors versus normal epithelium. High KIF14 in tumors independently predicted for worse outcome (p = 0.03) with loss of correlation with proliferation marker expression and increased rates of recurrence. Overexpression of KIF14 in OvCa cell lines increased proliferation and colony formation (p < 0.01), whereas KIF14 knockdown induced apoptosis and dramatically reduced colony formation (p < 0.05). Our findings indicate that KIF14 mRNA is an independent prognostic marker in serous OvCa. Dependence of OvCa cells on KIF14 for maintenance of in vitro colony formation suggests a role of KIF14 in promoting a tumorigenic phenotype, beyond its known role in proliferation.
Ovarian Cancer (OvCa) is the leading cause of death from gynecological malignancies in the Western World.1 The majority of women who present with epithelial OvCa will die of their disease because of diagnosis at advanced stage. Treatment of OvCa has been virtually unchanged in over 40 years and includes cytoreductive surgery and cytotoxic chemotherapy. Improved surgical techniques, new chemotherapeutic agents, altered modes of chemotherapy delivery and emerging molecular targets have improved life expectancy but have done little to impact cure rates. There is a desperate need for improved early detection, diagnostic and prognostic indicators, and the identification of molecular pathways that drive and maintain OvCa cells, to achieve novel therapies that can cure.
Our studies of the pediatric eye cancer retinoblastoma revealed some fundamental mechanisms of cancer initiation and development.2–8 The most prevalent karyotypic abnormality in retinoblastoma following RB1 loss is gain of the long arm of chromosome 1.4, 5 Subsequently, 1q gain was shown to be common in breast, lung, liver, papillary renal cell and OvCas.9–19 Expression analysis of genes within the minimal 1q region of gain (1q32.1) revealed the mitotic kinesin family member 14 (KIF14) as the potential oncogene.13 We showed that high KIF14 expression in breast and lung cancers could predict for poor outcome,20, 21 suggesting that the oncogenic role of KIF14 has wide relevance to cancer.
Human KIF14 was first identified as a 6,586 base-pair kinesin family 10 cDNA clone.22 Like all kinesins, KIF14 is a microtubule-dependent molecular motor, containing a kinesin motor domain and a forkhead-associated domain.23–25KIF14 also contains an C-terminal citron kinase (CIT) binding region and a N-terminal protein regulating cytokinesis (PRC1) binding region.23, 24 KIF14 function has been only partially characterized, showing that it is essential for the final phase of cytokinesis,24 where interaction with PRC1 and CIT is required.23 Knockdown of KIF14 in HeLa cells results in the generation of binucleate cells, polyploidy and apoptosis, depending on the degree of knockdown.24 We showed that transient KIF14 knockdown by siRNA in H1299 and HeLa cells significantly decreased proliferation and colony formation, suggesting that KIF14 may have an important oncogenic role in cancer cells.21
In this study, we show for the first time that KIF14 is highly overexpressed, corresponding to genomic gain in multiple subtypes of primary OvCa tumors, suggesting a potential initiating event. High KIF14 mRNA expression in serous OvCa tumors independently predicts worse outcome. Unlike low expression, high KIF14 overexpression in OvCa tumors does not correlate with expression of proliferation markers (Ki67, PCNA) or KIF14 cytokinesis binding proteins (CIT, PRC1), suggesting a proliferation-independent function for high levels of KIF14 in OvCa tumorigenesis. We identify KIF14 as essential in enhancing anchorage-independent colony growth and the in vitro tumorigenic phenotype in OvCa cell lines. We show KIF14 expression to be a novel prognostic marker with potential importance in initiation and maintenance of serous OvCa. Consequently, targeting KIF14 could prove curative in OvCa.
Material and methods
Ninety fresh frozen OvCa tumor samples, ten fresh frozen normal non-neoplastic ovaries from patients undergoing oopherectomy for non-oncological conditions and four normal tubal epithelium tissues were obtained from the University Health Network (UHN) BioBank. Twelve RNA samples from short-term cultures (less than five passages) of primary ovarian surface epithelium (pOSE) were a kind gift from Dr. Mark Nachtigal (University of Manitoba, Winnipeg, MB).26 Thirty-two fresh frozen OvCa tumor samples were obtained from the Ontario Tumor Bank (OTB). The UHN and OTB Research Ethics Boards and respective tissue access committees approved the study, and all tissues were banked with written informed consent. Clinical data associated with the UHN BioBank samples were reviewed by a gynecological oncologist and gynecologic pathologist (MQB and PAS) to ensure data integrity and quality, and all UHN Biobank tissues (adjacent H and E stained slides) were reviewed by a gynecologic pathologist (PAS) to ensure that samples contained >80% tumor cells. Tissues and data obtained from the OTB were reviewed by pathologists and medical oncologists using similar criteria at the respective collection centers.
DNA extraction and quantitative multiplex PCR
Genomic DNA was extracted using an isolation kit (Qiagen, Mississauga, ON) as per manufacturer's instructions. Quantitative multiplex PCR (QM-PCR) was performed by Solutions by Sequence (Toronto Western Hospital Research Institute, Toronto, ON) with two sets of KIF14 gene-specific primers (KIF14 A: Forward: TGGAGGATTATGCTTACTGTGGA, Reverse: TGCCATGTCTCTCTCCTGTTGA; KIF14 B: Forward: GACAAATCAAGCACTATTTACTC, Reverse: GCTTACAGTTATCTTGACTTTC), where the forward primer was labeled with a 5′- Cy5 dye (Integrated DNA technologies, Toronto, ON). Five internal control genes represented normal gene copy number (two copies): WI-7221 (10q21), ALK198 (12q13), ALK8 (12q13), C4 (15q26) and VHL (3p25), regions rarely gained in ovarian cancer.27–29 Separate reactions (multiplexed with endogenous control gene primers) were performed for each KIF14 gene-specific primer set, and PCR performed as described.13 Copy number was calculated using OpenGene Visible Genetics™ System Software (Siemens, Mississauga, ON) as described.13 Minimum value for gain was the mean copy number + three standard deviations (SD) of five normal ovary tissue samples. Copy number >2.6 was considered as gain, whereas copy number <1.65 was considered as loss.
RNA extraction and reverse transcription
Total RNA was extracted from tissues and cells by homogenizing in TRIzol reagent (Invitrogen, Mississauga, ON), followed by chloroform extraction and isopropanol and ethanol precipitations. Total RNA from primary OvCa (pOvCa) cells26 were a kind gift from Dr. Mark Nachtigal. One microgram of total RNA was used in reverse transcription (RT) reactions, as previously described.13
To confirm RT, 1 μl of each reaction was tested in end-point PCR for KIF14 and the housekeeping gene HPRT (hypoxanthine phosphoribosyl transferase) as described.20
End-point and real-time RT-PCR
For end-point PCR, 1 μl of the RT reaction was added to a 25 μl PCR reaction containing 0.5U Hot Start Taq polymerase (Fermentas, Burlington, ON), 0.2 mM dNTPs, 1.5 mM MgCl2 and a 1q gene primer pair; cycling conditions were as previously described.13 TBP was used as an endogenous control, and products were visualized by gel electrophoresis and ethidium bromide staining. For real-time PCR, RT reaction products were diluted 10-fold, and 1.5 μl was added to 1X TaqMan® PCR master mix (Applied Biosystems, ABI, Carlsbad, CA) and 1X TaqMan® Gene Expression Assay primer-probe mix for KIF14 (Hs00978216_m1). Mean expression of three housekeeping genes was used as an endogenous control: TBP (Tata-box binding protein, Hs_99999910_m1), HPRT (Hypoxanthine phosphoribosyl transferase, Hs99999909_m1) and GAPDH (glyceraldehyde-3-phosphate dehydrogenase, Hs99999905_m1). Proliferation markers Ki67 (Hs00606991_m1) and PCNA (proliferating cell nuclear antigen, Hs00427214_g1), and KIF14 binding proteins CIT (Citron kinase, Hs00392339_m1) and PRC1 (protein regulating cytokinesis 1, Hs00187740_m1), were similarly assayed in separate reactions. Triplicate reactions were conducted for each gene and each tissue sample, and PCR performed using the SDS 7900HT system (ABI) as described.21 SDS 2.1 software (ABI) was used to calculate ΔΔCt relative expression values, normalized to endogenous control genes and calibrated to one tubal epithelium sample (sample #60923, showing highest KIF14 normal sample expression). Relative quantity (RQ) was calculated according to the formula RQ ± SD = 2[ΔΔCt±(ΔCtSD)] using Microsoft Excel.
Cell culture, shRNA lentivirus construction and transductions
SKOV3 and OvCa 429 cells (Dr. Mark Nachtigal) were grown in DMEM H16 minimal medium (SKOV3) or alpha-MEM (OvCa 429) supplemented with penicillin–streptomycin and 10% fetal bovine serum at 37°C, 5% CO2. Immortalized OSE (IOSE) 144 and 386 cells were a kind gift from Dr. Nelly Auersperg (University of British Columbia, Vancouver, BC) and grown as described.30, 31 Packaging cells (293FT, Invitrogen) were grown as SKOV3 cells. All parental and derived stable cell lines were authenticated (December, 2010) using STR (short tandem repeat) fingerprinting (The Centre for Applied Genomics, The Hospital for Sick Children, Toronto, ON). shRNA-expressing lentiviruses were generated by co-transfection of pLKO.1 containing shRNAs and expressing puromycin resistance, targeting KIF14 (5 separate constructs, #816–819 plus a scrambled (S) control, Sigma-Aldrich) with packaging constructs pPAX2 and pMD2G (a kind gift from Dr. Jason Moffat, University of Toronto, Toronto, ON) in 293FT cells using Gene Juice (EMD Biosciences, Gibbstown, NJ). Virus supernatant was harvested 48 hr post-transduction and concentrated 10× using the LentiX viral concentrator (Clontech Laboratories, Mountain View, CA). KIF14 knockdown in OvCa cell lines was confirmed 72 hr post-transduction using end-point RT-PCR and western blot (WB) analyses. Selection of shRNA-expressing cells was conducted by treatment with puromycin hydrochloride (Sigma-Aldrich) initially at 1 μg/ml for SKOV3 cells, and 0.5 μl/ml for OvCa 429 cells for 14 days, followed by maintenance at 0.1 μg/ml for both cultures. Stable OvCa cell lines overexpressing KIF14 were generated by transfection of either KIF14-Myc, KIF14-EGFP (gift from Dr. Francis Barr, Max Planck Institute of Biochemistry, Martinsreid, Germany) or pcDNA 3.1 expressing neomycin resistance (empty vector control, EV). At 72 hr post-transfection, cells were selected with 400 μg/ml of G418 (Sigma-Aldrich) for 14 days, then maintained in 100 μg/ml. End point RT-PCR was performed as described.20, 21 WB analysis was performed by probing with an anti-KIF14 antibody (Bethyl Laboratories, Montgomery, TX), anti-Myc or anti-EGFP antibodies (Abcam, Cambridge, MA) normalized to β-tubulin (Sigma-Aldrich). Horseradish peroxidase-labeled secondary antibodies (Chemicon, Billerica, MA) were detected using a chemiluminescence reagent (Denville Scientific, Metuchen, NJ) and incubated with photographic film (Denville). Signal intensity measurements of KIF14 and β-tubulin were calculated via Photoshop CS3.
Cells were stained with an anti-KIF14 antibody (Bethyl Laboratories), followed by a mouse anti-rabbit Alexa 488 secondary antibody (Molecular Probes, Mississauga, ON), and nuclei visualized by DAPI staining (Sigma-Aldrich). Rabbit polyclonal IgG solution (Invitrogen) was used in control slides. Stained cells were visualized at a magnification of up to 400× with an epifluorescence microscope (Leica, Wetzlar, Germany).
Proliferation, BrdU incorporation, apoptosis and soft-agar colony forming assays
Proliferation was measured by Trypan-Blue dye (Invitrogen) exclusion cell counts; puromycin-selected cells (14 days) were seeded at 2 × 104 cells/6-well dish (Day 0) and counted every 1, 2, 4 and 8 days via hemocytometer. BrdU incorporation assays were conducted by seeding puromycin-selected cells (14 days) at 4 × 104 cells/6-well dish the day before labeling. The following day, cells were labeled for 4 hr with 100 μM of BrdU reagent (Invitrogen) in complete media, followed by staining with an anti-BrdU antibody (Sigma-Aldrich), an anti-mouse Alexa 594 secondary antibody (Molecular Probes), and by DAPI staining (Sigma-Aldrich) to visualize nuclei. Positive BrdU cell counts were obtained through fluorescence microscopy (Leica) wherein each experiment replicate, five fields of view at 200× total magnification were counted for positive (red) and total cell numbers (blue DAPI-stained nuclei), expressed as percentage BrdU-positive cells.
Apoptosis assays were conducted by seeding puromycin-selected cells (14 days) at 4 × 104 cells/6-well dish the day before staining, and staining the next day with Annexin V-FITC antibody (Abcam). Positive apoptotic cell counts were obtained using fluorescence microscopy (Carl Zeiss, Toronto, ON); five replicate fields at 200× total magnification were counted for positive (green) and total cell numbers (number of cells in brightfield), and expressed as percentage apoptotic cells.
Colony assays were conducted by seeding puromycin or G418-selected cells (14 days) at 1 × 103 cells/6-well dish in 0.3% noble agar, atop a plug of 0.6% noble agar in growth medium (Biorad Laboratories, Mississauga, ON). After 14 days of growth, colonies were stained with crystal violet and counted with the 1.5× objective of a dissecting microscope, counting five fields of view for each experiment replicate. All experiments were conducted on three separate occasions in triplicate.
Unpaired t-tests were performed to determine statistical differences (where p < 0.05) between KIF14 relative quantity (RQ) in normal tissue samples and each OvCa histological subtype. Correlations between KIF14 RQ category (KIF14HIGHvs.KIF14LOW, dichotomized around the mean RQ of each tumor subtype) and stage, grade, Ki67, PCNA, CIT and PRC1 RQ were analyzed using Spearman's Correlation Coefficient (r). Progression-free survival (from date of surgery to date of recurrence) and overall survival (from date of surgery to date of death due to ovarian cancer) described the survival function for both Kaplan–Meier survival analyses and Cox proportional hazard univariate and multivariate regression analyses. Mean survival times (months) were calculated for patients with high versus low KIF14 overexpression, and early-stage (FIGO I–II) versus late-stage (FIGO III–IV) via Kaplan–Meier log-rank test. Univariate and multivariate analyses of KIF14 RQ values with clinical variables were conducted using Cox proportional hazard regression (SPSS v.19). Both the models include and account for censored data, and only samples with complete outcome data were analyzed (Supporting Information Fig. 2). Biological features of OvCa cell lines (proliferation, BrdU incorporation, apoptosis and colony formation) were assessed between controls and treated groups using paired t-tests. Unless otherwise specified, all statistical analyses were conducted using Graph Pad Prism 4.0.
KIF14 shows genomic gain and high overexpression in a subset of ovarian cancers
We measured genomic gain of KIF14 in ten non neoplastic ovary (NNO) and 109 primary tumors using multiplex QPCR32 with two different KIF14 primer sets. NNO samples showed normal KIF14 copy numbers. KIF14 genomic gain (>2.6 copies) was evident in 27.3% of serous, 18.2% of endometrioid, 10% of mucinous and 9% of clear cell tumors (Fig. 1a), but none of the samples showed amplification of KIF14.33 Genomic loss was seen in a small percentage of serous, mucinous and clear cell tumors. Importantly, all tumors with KIF14 gain (3/33 clear cell, 1/10 mucinous, 2/11 endometrioid and 15/55 of serous OvCas tested) showed high overexpression of KIF14 mRNA (above mean RQ, Fig. 1b), whereas all tumors with KIF14 loss (1/33 clear cell, 1/10 mucinous and 2/55 serous) showed very low KIF14 mRNA expression, close to normal levels. To confirm the importance of KIF14 as the gene of interest in OvCa within the 1q region of gain, we tested via end-point PCR the expression of 13 named genes on chromosome 1q flanking KIF14.13KIF14 was the only gene to be overexpressed in OvCas versus normal epithelium tissues (Supporting Information Fig. 1 and data not shown), demonstrating the biological relevance of KIF14 as a target gene for OvCa.
We compared KIF14 mRNA levels using real time QPCR of 122 primary ovarian tumors (Table 1) versus the mean KIF14 expression of four tubal epithelium (TE) and 12 pOSE short-term cultures [hereafter referred to as normal epithelium (NE)]. When the RQ of KIF14 mRNA of all tumors was compared to NE (mean RQ = 1.08 ± 1.02, Supporting Information Fig. 2a), significant overexpression was found (Fig. 1b, p = 0.004, one-way ANOVA). Clear cell and serous subtypes showed the highest expression (mean RQ = 28.85 ± 6.1 and 23.54 ± 3.5, respectively), followed by mucinous and endometrioid subtypes (mean RQ = 9.27 ± 2.4 and 8.97 ± 2.5, respectively). Comparing KIF14 expression between early stage (FIGO Stages I and II) and late stage tumors (FIGO Stages III and IV) revealed no significant differences (Supporting Information Fig. 2b, t-test p = 0.08). No correlations were seen between KIF14 expression and stage or grade in any histological subtype of OvCa (data not shown). Some samples were excluded from analyses due to lack of clinical information (Supporting Information Fig. 2c).
Table 1. Patient characteristics
High KIF14 expression in serous cancers does not correlate with proliferation markers or expression of KIF14 binding proteins
Samples in each OvCa tumor histological subtype were dichotomized around their respective KIF14 RQ means into low and high KIF14 overexpression (KIF14LOW and KIF14HIGH, respectively, Fig. 1b). The few mucinous and endometrioid cancers obtained were at early stage and were not included in further subgroup analyses due to insufficient numbers (Table 1).34–36
We correlated mRNA expression of proliferation markers Ki67 and PCNA, and KIF14 binding proteins CIT and PRC123 with KIF14 for clear cell and serous OvCas. A significant association was seen between KIF14 expression and the proliferation markers Ki67 and PCNA for both KIF14LOW and KIF14HIGH clear cell tumors. However, the only significant association between KIF14 expression and KIF14 binding proteins CIT and PRC1 was seen in KIF14LOW clear cell tumors for PRC1 (Supporting Information Table 1). KIF14 expression in all serous KIF14LOW tumors strongly correlated with all four genes, but in KIF14HIGH serous tumors, KIF14 did not correlate with either proliferation markers or KIF14 binding proteins (Supporting Information Table 1).
KIF14 expression predicts progression-free survival in serous ovarian cancers
Treatment and outcome data were available for 112 out of 122 patient samples, with mean follow-up time of 35 months. A larger proportion of patients with serous OvCa with KIF14HIGHversus KIF14LOW overexpression recurred (58.8% vs. 13.5%), and died of OvCa (70.6% vs. 29.7%, Supporting Information Table 2). Of the five patients with KIF14LOW tumors, four were at late stage. In contrast, in clear cell OvCa, fewer patients died in both KIF14HIGH and KIF14LOW groups, but recurrence was more frequent from KIF14HIGHversus KIF14LOW patients (55.5% vs. 10.7%, Supporting Information Table 2).
Table 2. Effects of KIF14 and other prognostic variables on multivariate Cox proportional hazards regression analyses
No significant difference was seen in overall survival (OS) of patients with KIF14HIGHversus KIF14LOW serous cancers (Fig. 2c), nor in progression-free survival (PFS) or OS of patients with clear cell cancers (Figs. 2b and 2d), although a trend to decreased OS in serous patients was seen [Hazards ratio (HR) = 2.26 (0.95–5.4), p = 0.06; Fig. 2c]. However, PFS was significantly lower for patients with serous KIF14HIGHversus KIF14LOW tumors [11 months vs. 38 months, respectively, Fig. 2a and Supporting Information Table 2, Kaplan–Meier log-rank test p = 0.03, HR = 3.2 (1.2–8.4)].
When KIF14 expression values were studied in a univariate Cox regression model for clear cell and serous subtypes, KIF14 expression levels were prognostic for PFS of patients with serous [Odds ratio (OR) = 1.23 (1.03–1.43)], p = 0.02, but not clear cell cancers. No OS prognostic value was shown for KIF14 expression in both clear cell and serous patients. Levels of Ki67, PCNA, CIT or PRC1 did not associate with OS or PFS within each tumor subtype.
On multivariate Cox regression analysis of clear cell and serous patients, including KIF14, stage, grade, Ki67, PCNA, CIT and PRC1 levels as covariates, KIF14 remained an independent prognostic factor for OS and PFS in serous patients (Table 2). In multivariate analysis, stage also remained as an independent prognostic factor for PFS in serous patients; we therefore sought to determine whether KIF14 expression could predict outcome in this subset based on stage. When serous patients were separated by individual stage, KIF14 expression levels had no predictive effect on OS or PFS in this subtype. When these same patients were dichotomized into early-stage versus late-stage groups, KIF14 levels were prognostic for PFS [OR = 1.09 (1.01–1.23), p = 0.03] in late-stage patients, but not for early-stage patients.
KIF14 overexpression enhances proliferation and tumorigenic potential in vitro
In pOvCa cell cultures, KIF14 mRNA was overexpressed 2.5-fold in comparison with pOSE cells (Supporting Information Fig. 3a, t-test p = 0.05), with a similar increase in OvCa versus IOSE cell lines. KIF14 protein was also overexpressed in pOvCa and IOSE cells, although IOSE cells showed much lower KIF14 expression than pOvCa cells (Supporting Information Fig. 3b, t-test p = 0.05). KIF14 overexpression in OvCa cell lines was confirmed via immunofluorescence (Supporting Information Fig. 3c). None of the cell lines exhibited genomic copy number gain or loss of KIF14 (Supporting Information Fig. 3d).
Manipulation of OvCa cell lines to express high KIF14 levels was performed to determine whether the association of KIF14HIGH expression with shorter PFS in primary OvCa tumors would translate into a functional effect on tumorigenicity in vitro. Assessments of in vitro tumorigenicity can be achieved through growth in soft agar, which will specifically measure growth in anchorage-independent conditions, resistance to apoptosis and clonogenicity, all important characteristics for the maintenance of cell transformation.37–39 OvCa 429 and SKOV3 cells were engineered to stably overexpress either a Myc-tagged KIF14, EGFP-tagged KIF1423 or an empty vector (EV) control (Supporting Information Fig. 4). Overexpression of KIF14 significantly increased proliferation (Fig. 3a and Supporting Information Fig. 5a, p = 0.01) and the number of colonies formed in soft agar after 14 days (Figs. 3b and 3c and Supporting Information Figs. 5b and 5c, p = 0.008 and p = 0.005, respectively) for both KIF14-Myc and KIF14-EGFP constructs. However, the KIF14-EGFP construct with lower KIF14 expression had less impact on both proliferation and colony formation (Fig. 3 and Supporting Information Fig. 5). In both SKOV3 and OvCa 429 cell lines, the increase in colony formation was higher than the increase in proliferation (5-fold vs. 3-fold, t-test p ≤ 0.05), suggesting that enhancement of anchorage-independent survival and tumorigenic capacity of OvCa cell lines is KIF14-dependent.
Knockdown of KIF14 significantly reduces proliferation and tumorigenicity in vitro
To study the effect of reduced levels of KIF14 on the in vitro transformed phenotype of OvCa cell lines, we generated five clones each of SKOV3 and OvCa 429 cells stably expressing shRNAs targeting KIF14 mRNA, plus a scrambled shRNA control clone (Supporting Information Fig. 6). Each SKOV3 anti-KIF14 clone showed reduced KIF14 protein 4 days post-transduction (Supporting Information Fig. 6a). Puromycin selection of SKOV3 cells for 30 days markedly decreased KIF14 in comparison to controls (Supporting Information Fig. 6b). Scrambled and untreated SKOV3 cells showed the same levels of protein expression. Similar results in KIF14 protein expression were observed with transduced OvCa 429 cells (data not shown).
All shRNA SKOV3 and OvCa 429 clones showed reduced proliferation and BrdU incorporation (Fig. 4 and Supporting Information Fig. 7), and significantly increased apoptosis as evidenced by Annexin V-positive cells. Most importantly, all shRNA clones showed significantly reduced numbers of colonies in comparison to scrambled and parental OvCa cell lines (t-test p ≤ 0.05; Figs. 4c and 4d, and Supporting Information Figs. 7b and 7c). Knockdown of KIF14 in both SKOV3 and OvCa 429 cells decreased colony formation to a greater extent than proliferation (5-fold decrease in colony formation vs. 2.5-fold decrease in proliferation, t-test p ≤ 0.05), suggesting that KIF14 is necessary to maintain tumor cell growth and survival in vitro.
We present for the first time evidence that expression of a mitotic kinesin is an independent prognostic indicator for progression-free survival in OvCa, providing support for the clinical utility of KIF14 as a marker of outcome in OvCa. Our data complements growing evidence supporting KIF14 as an oncogene, by demonstrating that KIF14 is not only a robust indicator of progression-free survival, but also an effector of tumorigenic potential in OvCa cell lines. These data point to KIF14 as an attractive therapeutic target in OvCa.
Of the 122 primary OvCa tumor samples measured for KIF14 mRNA, 91% (111 samples) showed KIF14 expression to be significantly above levels for all NE tissue samples, indicating the cancer-specific nature of KIF14 overexpression. We have previously shown that KIF14 expression is extremely low in almost all adult tissues,13 and now, in normal epithelial tissues (TE) and cells (pOSE), pointing to the tight regulation of KIF14 in normal cells with de-regulation in OvCa tumor cells. TE tissues and pOSE short-term cultures were selected as controls because they represent postulated cells of origin for OvCa,40, 41 and showed very similar KIF14 expression levels in contrast to non-neoplastic ovary (NNO) tissues which showed lower (but not statistically significant) KIF14 expression. Because of the likely loss of the OSE during NNO tissue manipulations (fixation, embedding and thus may represent stromal cell expression of KIF14), NNO KIF14 expression was excluded from our analyses. Although KIF14 mRNA was overexpressed in all OvCa subtypes versus NE tissues, expression was highest in clear cell and serous cancers. Prognostic significance of KIF14 expression was only demonstrated in patients with serous OvCas, indicative of the presence of subtype-specific molecular profiles in OvCa.35, 42 Subsequently, the similarity in KIF14 expression between early stage and late stage serous cancers suggests a unique serous OvCa molecular signature, where high overexpression of KIF14 may be an early event driving OvCa development.
When separated by stage, the prognostic significance of KIF14 was retained in late stage serous cancers. The low number of events resulting in low statistical power may explain the lack of prognostic utility in early-stage cancers. The significant survival advantage exhibited for KIF14LOW serous patients demonstrates a clinical benefit of KIF14 as a marker of “high risk” versus “low risk” for poor outcome in serous OvCa patients. Study of a larger number of early stage OvCas from each subtype will provide insight into the prognostic significance of KIF14 in this subgroup.
KIF14 genomic copy number in a subset of ovarian tumors (109/122) revealed that close to 30% of serous cancers (15/55) had low-level gain of KIF14, with similar gains in a number of early and late stage cancers, suggesting genomic gain could be a potential initiating event in the development of serous OvCa. The limited number of samples tested precludes significant statistical correlation between genomic gain and clinical features. However, the presence of genomic gain did correlate with KIF14 mRNA expression profile (all samples with KIF14 gain showed high KIF14 overexpression and were in the KIF14HIGH group, whereas samples with KIF14 loss showed very low KIF14 expression regardless of OvCa subtype), suggesting that genomic change is an important mechanism for driving KIF14 overexpression in KIF14HIGH OvCas. Combined with our expression data demonstrating KIF14 as the 1q gene of interest in OvCa, these data provide biological relevance to KIF14 as a potential oncogene. None of the KIF14LOW tumors showed KIF14 genomic gain. However, a few serous tumors exhibiting no KIF14 gain did show KIF14HIGH mRNA over expression, suggesting multiple mechanisms (genomic gain, transcriptional or epigenetic regulation of gene expression) through which KIF14 expression could be enforced in OvCa cells. Genomic characterization of a larger serous OvCa subset is warranted. Although recent advances in deep RNA sequencing have revealed important gene mutations and therapeutic targets in poor-prognosis OvCas,43, 44 to our knowledge, KIF14 has no known mutations in any disease. Genomic copy number gain remains an effective flag of potential oncogenes, whereas high mRNA expression characterizes more tumors than just those with genomic gain.
KIF14 may be involved in two ways in OvCa. First, it may promote disease progression, as KIF14 expression correlates with PFS and recurrence rate. The mechanism of action of KIF14 in tumor progression is currently unknown. The well-documented role of KIF14 in the control of cytokinesis, where presence of KIF14 binding partners CIT and PRC1 is required,23, 24 suggests that KIF14 overexpression would lead to increased cell proliferation and tumor growth. The tight correlation between KIF14 and Ki67, PCNA, CIT and PRC1 in low KIF14-expressing cancers supports this role. However, the lack of correlation between KIF14 expression and these four genes in KIF14HIGH serous tumors, combined with the lack of prognostic value of CIT and PRC1 in this subset, leads us to speculate on an additional action of high KIF14 supporting tumor cell survival independent of cytokinesis. Engineered overexpression of KIF14 in OvCa cell lines promoted an anchorage-independent phenotype in vitro, substantiating such a role for KIF14. The soft agar colony formation assay is an important in vitro measure of anchorage-independent growth, capacity to evade apoptosis and clonogenicity, all characteristics inherent to transformed cells, and sometimes required for maintenance tumorigenicity in vivo.37–39 Stable knockdown of KIF14 in OvCa cell lines decreased colony formation disproportionate to proliferation, suggesting that high KIF14 expression supports substrate-independent growth and survival of OvCa cells in vitro. KIF14 is a molecular motor;45 its abundance in poor prognosis tumors could facilitate shuttling of as yet unknown cargo(s) involved in growth factor signaling, evasion of apoptosis, or capacity for anchorage-independent growth, literally “driving” initiation, survival and maintenance of the cancer. Despite greatly reduced colony formation and increased apoptosis in response to KIF14 knockdown in OvCa cells, the presence of some colonies suggests a heterogeneous and incomplete knockdown in some OvCas, where KIF14 could remain at sufficient levels to induce substrate-independent cell survival and subsequent colony formation.
Our KIF14 knockdown results contrast with those of Carleton et al., who showed that siRNA knockdown of KIF14 in HeLa cells induced binucleation and apoptosis, consistent with a cytokinetic role.24 Although, in our hands, stable knockdown of OvCa cell lines induced significant apoptosis, stable shRNA-mediated knockdown of KIF14 did not induce binucleation in either SKOV3 or OvCa429 cells (data not shown). This discrepancy could be explained in that the type (stable vs. transient) and level of KIF14 knockdown may produce different cellular phenotypes, which could point to unique functions for KIF14 in different cell lines. Alternatively, our particular stable clones may be sensitive to multinucleation and induce apoptosis before any such phenotype can be characterized.
To our knowledge, this is the first linkage between kinesin expression and outcome in OvCa. Aurora kinases, which interact with kinesins and participate in mitotic spindle regulation, are well-documented oncogenes overexpressed in OvCa.46, 47 When tested in Phase I dose escalation studies for advanced solid tumors, pan- and selective Aurora kinase inhibitors showed partial responses,48 warranting more selective and less toxic drug targets. Recently, CIT has been implicated in hepatocellular carcinoma tumorigenesis, demonstrating overexpression in tumors, and siRNA knockdown decreases colony formation in SMMC-7721 cell lines and reduces xenograft growth.49 Although CIT shows similar levels of expression to KIF14 (data not shown) CIT expression did not possess any predictive value on OS or PFS in either univariate or multivariate Cox regression analyses of OvCas, in contrast to the clinical and therapeutic relevance of KIF14 expression as a prognostic marker in OvCa.
Our data strongly suggests that KIF14 overexpression can enhance a tumorigenic phenotype in OvCa cell lines; the mechanism of KIF14 oncogenic action in vivo remains to be determined. Nevertheless, we clearly demonstrate that KIF14 expression correlates with poor outcome in serous OvCa, and thus presents an attractive drug target for these patients. The fact that inhibition of KIF14 induces significant apoptosis in vitro suggests targeting of KIF14 in vivo could yield a favorable therapeutic result, as OvCas depend on KIF14 for maintenance of the tumorigenic phenotype. Human KIF14 protein has not been fully characterized in cancer cells; therefore, study of KIF14 protein interactions in OvCa may reveal important tumorigenic pathways that could be targets for selective inhibition.
The authors acknowledge the help of Ms. Diane Rushlow and Mr. Donco Matevski from Solutions by Sequence (Toronto Western Hospital Research Institute) for performing the DNA copy number analyses. They thank Dr. Sophia George for providing the tubal epithelium RNA, and the UHN BioBank for providing normal and OvCa tissue samples. The views expressed do not necessarily reflect those of the OMOHLTC.