LPA Induces Metabolic Reprogramming in Ovarian Cancer via a Pseudohypoxic Response

Cancer cells reprogram glucose metabolism to aerobic glycolysis to meet the increased anabolic demands of cell growth and proliferation. Association between cancer cells and aerobic glycolysis has been well recognized for many years now (1). However, the significance of this observation in relation to cancer progression and the adaptive mechanism(s) underlying this association is beginning to be understood only now (2–7). Metabolic reprogramming in cancer cells primarily involves a shift to aerobic glycolysis with or without an effect on mitochondrial oxidative phosphorylation (8). This is often accompanied by dysregulated lipogenic metabolism and adaptive mitochondrial reprogramming, both of which can contribute to aerobic glycolysis (3, 8–11). Studies focused on defining the mechanism underlying metabolic reprogramming in cancer cells have identified a critical role for oncogenes such as Ras and Myc and tumor suppressors such as p53 and pRB. In addition to the intrinsic genetic and epigenetic mechanisms regulated by the oncogenes and tumor suppressors, several extrinsic stimuli including those of growth factors and hypoxic stress have been shown to induce metabolic reprogramming in cancer cells (12–15). Of the different extrinsic factors, hypoxia-induced oxidative stress involving hypoxia-inducible factor 1 α-subunit (HIF1α) has been shown to play a major role in orchestrating the molecular events required to induce aerobic glycolysis in solid tumors (11, 16). However, the identity of extrinsic growth factors that can induce metabolic programming is largely unknown. In this context, the observations that ovarian cancer cells synthesize and release lysophosphatidic acid (LPA) into the internal milieu and LPA is present in a large quantity in the ascites and serum of ovarian cancer patients are highly significant. Taken together with our findings that LPA stimulates epithelial-to-mesenchymal transition of ovarian cancer cells via HIF1α even in normoxic conditions (17), it can be posited that LPA induces aerobic glycolysis in ovarian cancer through a pseudohypoxic response involving HIF1α. In our current study, we investigated the role of LPA in the glycolytic shift in ovarian cancer using patient-derived ovarian cancer cells and high-grade serous ovarian cancer cell lines in a metabolic flux analyzer. Our results indicate that LPA stimulates a pseudohypoxic response via a conduit involving Rac–NOX–ROS–HIF1α with the resultant induced expression of glucose transporter-1 (GLUT1) and the glycolytic enzyme hexokinase-2 (HKII). Consistent with the tumor-promoting role of this pseudohypoxic nexus, we demonstrate that the inhibition of HKII with 3-bromopyruvate (3-BP) can attenuate ovarian cancer xenograft tumor growth along with a concomitant survival advantage in an ovarian cancer xenograft mouse model.

Kuramochi, SNU119, OV90, TOV112D, OVCA429, OVCAR8, and SKOV3-ip cells have been previously described, and cell passaging was never exceeded 18 (18, 19). The cell lines obtained from NCI, ATCC, and Seoul National University were authenticated at IDEXX Bioresearch using nine human short tandem repeat profiles (20). Cells were monitored for mycoplasma contamination using previously published PCR-based protocol (21). Patient-derived cell lines ASC022315, ASC022415, and ASC031915 were isolated from the ascites samples of patients at the Stephenson Cancer Center, University of Oklahoma Health Science Center, Oklahoma City, Oklahoma. The study was approved by the OUHSC Office of Human Research Participant Protection (HRPP) Institutional Review Board and samples were collected with the informed consent from the patients. The ascites-derived ovarian cancer cells were maintained in MCDB:DMEM (1:1) supplemented with 10% FBS and 50 μg/mL streptomycin. For serum starvation, the above media without serum were supplemented with 0.1% BSA Fraction V, heat-shock, fatty acid ultra-free (Roche), 50 U/mL penicillin and 50 μg/mL streptomycin (Mediatech). Lysophosphatidic acid (1-oleoyl-2-hydroxy-sn-glycero-3-phosphate) was obtained from Avanti Polar Lipids and dissolved into 10 mmol/L stock solutions in PBS with 0.1% BSA and stored at −80°C until use. siRNAs against HIF1α, Gαi2, Rac1, Rac2, Rac3, Nox1, Nox2, Nox3, Nox4, Nox5, HKII, and the control ON-TARGETplus Non-targeting siRNA #1 were purchased from Dharmacon. shRNA targeting Gα12, Gα13, Gαq, Gαi2, and nonsense shRNA as well as immunoblot analyses and quantifications of immunoreactive bands have been previously described and used (22). N-acetyl-L-cysteine (NAC) and Bromopyruvic acid (3-Bromo-2-oxpropionic acid) was obtained from Sigma. Apocynin and Ki16425 inhibitor were procured from Tocris Bioscience. The Rac1 inhibitor NSC23766 was from EMD Millipore. The HIF1α inhibitor PX-478 was obtained from MedKoo Biosciences.

The extracellular acidification rate (ECAR) was determined in XFe96 Extracellular Flux analyzer (Seahorse Bioscience) using the XF Glycolysis Stress Kit (Seahorse Bioscience) by following the manufacturer's instructions. Briefly, based on cell density titration experiment carried out in our lab, the ovarian cancer cell lines were seeded at a density of 30,000 cells/well in a XF96 cell culture microplate. The cells were then either serum starved, stimulated with LPA or treated with inhibitor based on the experimental protocol till the desired time. Before the start of the assay, the medium from these cells was removed and washed with freshly prepared nonbuffered XF assay medium supplemented with 2 mmol/L glutamine. About 175 μL of the same XF assay media was added to the cells and incubated in a CO₂-free incubator for a minimum of 1 hour. Glycolysis stress test was carried out by adding 25 μL of the stock of the following reagents, from the kit at recommended intervals, to reflect the indicated final concentration, viz., glucose (Glu; 10 mmol/L), oligomycin (Om; 1 μmol/L), and 2-deoxyglucose (2-DG; 50 mmol/L). XFe96 assays consisted of sequential mixing, pausing, and measurement cycles, and assays were performed three times in triplicates. The data were analyzed and exported using the Seahorse Wave software (Seahorse) to the GraphPad Prism to obtain the graphs and bar charts and also carry out statistical analyses.

Nu/Nu nude female mice (5–6 weeks old) were purchased from Taconic Labs Inc. and were housed in a barrier facility under 12-hour light/dark cycle under pathogen-free conditions, with food and water ad libitum. All experiments were performed with the approval of the University of Oklahoma Health Science Center Institutional Animal Care and Use Committee. SKOV3-ip and OVCAR8 cells at log phase (5 × 10⁶ cells/100 μL) were injected on the right flank of nude female mice subcutaneously to generate tumors. When the tumors reached ∼50 mm³, the mice were randomized into control and treatment groups (SKOV3-ip; n = 10, OVCAR8; n = 5). The 3-BP (10 mg/kg/day) was diluted in sterile normal saline and injected intraperitoneally into the mice. The injections were carried out every day for 3 days on the first week and third weeks. Tumor volume was measured twice a week and recorded. At the study termination, tumor volumes were measured, animals were sacrificed, necropsy carried out to collect tumor specimens for IHC staining.

Tumors dissected out were fixed with 10% neutral formalin, paraffin-embedded, cut into 4-μm sections, mounted on positively charged slides and processed for IHC staining. The IHC was performed according to the manufacturer's protocol using Leica Bond Polymer Refine Detection System (DS 9800). The slides were incubated at 60°C for 45 minutes followed by deparaffinization and rehydration in an automated Multistainer (Leica ST5020). Subsequently, these slides were transferred to the Leica Bond-IIITM, treated for target retrieval at 100°C for 20 minutes with a retrieval buffer, pH 6.0. Endogenous peroxidase was blocked using peroxidase-blocking reagent, followed by the primary antibody, rabbit monoclonal anti-phosphohistone H3 (PHH3, 1:300, Sigma #369A) incubation. For the secondary antibody, post-primary IgG-linker and/or poly-AP IgG reagents were used. The substrate chromogen, 3,3′-diaminobenzidine tetrahydrochloride (DAB) detects the complex as brown precipitate, while hematoxylin counterstain the cell nuclei (blue). The slides were then dehydrated (Leica ST5020) and mounted (Leica MM24). Antibody-specific positive and negative (omission of primary antibody) controls were parallel stained. HIF1α antibodies were purchased from BD Biosciences, whereas antibodies to GLUT1, HKII, GAPDH, Rac1, and β-actin were obtained from Cell Signaling Technologies, Rac3, Nox1, Nox2, Nox3, Nox4, and Nox5 were obtained from Abcam, Rac2 is from ThermoFisher, and Gα12_, Gα13_, Gαq, and Gαi2 antibodies were purchased from Santa Cruz Biotechnology.

Ascites from which the patient-derived cells used in this study were obtained through Stephenson Cancer Center Biospecimen core with written informed consent from patients and the studies were conducted in accordance with the ethical principles and guidelines set forth in the “Belmont Report,” the Declaration of Helsinki, and the Nuremberg Code. In addition, the University of Oklahoma adheres to the Office for Human Research Protection requirements, as set forth in 45 CFR 46 and its subparts A, B, C, and D, and the FDA in 21 CFR 50 and 56.

Mitotic figures on PHH3 IHC slides were counted using ImageJ NIH software on 5 fields for each slide in triplicates at ×20 magnification. Only cells with metaphase or anaphase morphology were considered positive. The numerated PHH3 mitotic figures from control and treated xenograft tumor sections were tabulated and statistically analyzed for significance using t test.

All statistical analysis was performed using GraphPad Prism by the two-tailed Student t test with Welch correction.

To determine the role of LPA in inducing glycolytic shift in ovarian cancer cells, we monitored glycolysis and glycolysis capacity by monitoring the ECAR in ovarian cancer cell lines. To rule out the role for any signature oncogene or tumor suppressor gene in this regulation, our test panel consisted of ovarian cancer cell lines representing different genetic background as well as diverse ovarian cancer subtypes. A series of time course experiments with different doses of LPA were carried out to monitor ECAR in response to the addition of glucose, oligomycin, and 2-deoxyglucose. Glucose was added to initiate glycolysis and the ECAR flux observed following the addition of glucose was used to determine glycolytic rate. ECAR following the addition of oligomycin represented the overall glycolytic capacity of the cells, whereas the difference in ECAR before and after the addition of oligomycin defined the glycolytic reserve of the cells. As shown in Fig. 1, LPA induced an increase in both the rate of glycolysis and glycolytic capacity in all tested ovarian cancer cells irrespective of their genetic background or subtype categories. LPA is present in large concentration in the ascitic fluid that constantly baths the ovarian cancer. Thus, it can be postulated that the LPA present in the ascites stimulate such a pathological shift toward aerobic glycolysis in ovarian cancer cells. To test this possibility, we stimulated ovarian cancer cells derived from HGSOC patients with varying concentrations of LPA and determined the ECAR. As shown in Fig. 2, LPA stimulated an increase in aerobic glycolysis as well as the glycolytic capacity of these patient-derived ovarian cancer cells, thus establishing a causative role for LPA in metabolic reprogramming of ovarian cancer cells.

LPA is known to stimulate its biological responses through G protein–coupled LPA receptors (LPAR). Consistent with this view, treating the cells with LPA antagonists blocked LPA-stimulated ECAR in SKOV3-ip cells (Supplementary Fig. S1). Our previous studies have shown that LPA transmits diverse oncogenic signals through specific G protein α-subunits, namely, Gαi, Gαq, Gα12, and Gα13. To identify the Gα-subunit involved in transmitting the metabolic reprogramming stimuli from LPA, we carried out ECAR flux analyses using SKOV3-ip cells, in which the individual Gα-subunit was silenced using siRNA. Our results indicated that LPA-stimulated glycolytic shift primarily involves Gαi2 because the silencing of Gαi2 with specific siRNA led to the attenuation of both the rate of glycolysis and glycolytic capacity as indicated by the ECAR flux in OVCA429 (Fig. 3A) as well as in SKOV3-ip cells (Fig. 3B), while the silencing of other Gα-subunits failed to have such an effect (Supplementary Fig. S2).

Previously, we have shown that LPA activates EMT in ovarian cancer cells via Gαi2-stimulated HIF1α (17). In addition, previous studies have shown that Gαi2 can lead to the activation of Rac (23) and Rac, once activated, can stimulate the ROS-generating NOX family of NADPH oxidases, resulting in an increase in the intracellular levels of ROS (24). Furthermore, ROS-mediated inhibition of prolyl hydroxylase domain-2 enzyme, which is involved in ubiquitin-mediated degradation of HIF1α, can lead to an increase in the levels of HIF1α (25–28), which plays a critical role in cellular hypoxic response and hypoxia-mediated metabolic reprogramming in cancer cells. Based on these studies, we reasoned that the LPA–LPAR–Gαi2 axis could induce a pseudohypoxic response involving the Rac–NOX–ROS–HIF1α pathway that ultimately leads to the metabolic reprogramming in ovarian cancer cells. To validate this rationale, we first tested whether Rac1 is required for LPA-mediated ECAR flux. Our results indicated that the inhibition of Rac1 with Rac1-specific inhibitor attenuated LPA-stimulated glycolytic shift in ovarian cancer cells represented by SNU119 (Fig. 4A) and SKOV3-ip cells (Supplementary Fig. S3). Consistent with these findings, siRNA-mediated silencing the expression of Rac1 in OVCA429 cells inhibited LPA-mediated glycolysis (Fig. 4B). In contrast, silencing of Rac2 or Rac3 failed to have such an effect (Supplementary Fig. S4). Next, we investigated the role of downstream NADPH oxidase in this signaling network. SNU119 and SKOV3-ip cells were treated with apocynin, an inhibitor of NADPH oxidase, and LPA-stimulated ECAR was monitored. Results indicated that inhibiting NOX led to the blunting of LPA-stimulated increase in glycolysis as well as glycolytic capacity in both the cell lines (Fig. 4C; Supplementary Fig. S3). Using siRNA specific to NOX2, it can be shown that the silencing of NOX2 in OVCA429 cells (Fig. 4D) attenuated LPA-stimulated aerobic glycolysis while silencing of NOX1, NOX3, NOX4, or NOX5 failed to do so (Supplementary Fig. S5). Similarly, treating the cells with ROS-scavenger N-acetyl-L-cysteine abrogated LPA-stimulated glycolysis (Fig. 4E; Supplementary Fig. S3).

With our previous findings that LPA stimulates an increase in HIF1α levels and activity via Gαi2 (17), along with the observations from several laboratories that ROS could stabilize and increase the levels of HIF1α (25–28), it can be reasoned that the LPA induces metabolic programming in ovarian cancer cells by eliciting a pseudohypoxic response that leads to an increase in HIF1α levels via Rac–NOX–ROS signaling conduit. It can also be surmised that HIF1α, thus stabilized, promotes metabolic reprogramming in ovarian cancer cells. To validate this paradigm, first we investigated whether the inhibition of Rac, NOX, or scavenging ROS could abrogate the increase in HIF1α levels. As shown in Fig. 5A, the inhibition of Rac or NOX or ROS abrogated LPA-stimulated increase in the levels of HIF1α in SNU119 cells (Fig. 5A). Next, we investigated whether the inhibition of HIF1α abrogated LPA-stimulated aerobic glycolysis in ovarian cancer cells. This was carried out using PX-478, an inhibitor that reduces the expression of HIF1α through multiple mechanisms (17, 29). Kuromochi or SNU119 cells were pretreated with PX-478, and LPA-stimulated glycolysis was monitored. As shown in Fig. 5, the inhibition of HIF1α by PX-478 (Fig. 5B and C) significantly reduced aerobic glycolysis in both Kuromochi and SNU119 cells. Thus, our findings establish that LPA stimulates aerobic glycolysis by eliciting a pseudohypoxic response leading to an increase in HIF1α levels, which, in turn, promotes aerobic glycolysis in ovarian cancer cells.

Generally, HIF1α expression and activation occur in hypoxic region of tumor cells, and it has been shown that HIF1α plays a determinant role in promoting a glycolytic shift in hypoxic tumor cells (30–32). During hypoxia, HIF1α has been shown to induce glycolytic shift in tumor cells by transactivating the expression of genes involved in glucose/glycolysis metabolism and mitochondrial functions (30). It can be observed that LPA stimulated an increase in levels of HIF1α by 2.5 minutes (Supplementary Fig. S6) along with the stimulation of its transcriptional activity by 20 minutes (17). Of the genes that are transcriptionally upregulated by HIF1α, increased expression of GLUT1 and HKII has been correlated with glycolytic shift in multiple cancer cells and tissues (8, 33–35). Therefore, we examined whether the observed LPA–Gαi2–HIF1α-mediated aerobic glycolysis involves the expression of GLUT1 and HKII. SKOV3-ip cells in which the expression of Gαi2 or HIF1α was knocked down using specific siRNAs were stimulated with LPA and the expression levels of GLUT1 and HKII were monitored. As shown in Fig. 6, LPA stimulated the expression of both GLUT1 and HKII, whereas silencing the expression of Gαi2 blunted LPA-stimulated expression of HIF1α, GLUT1, and HKII (Fig. 6A). Likewise, silencing the expression of HIF1α led to a decrease in LPA-stimulated expression of GLUT1 and HKII (Fig. 6B). Next, we investigated whether the increased HKII levels contribute to LPA-stimulated aerobic glycolysis. This was assessed using siRNA to HKII in OVCA429 cells. As shown in Fig. 6C, silencing of HKII drastically attenuated LPA-stimulated glycolysis in these cells.

In addition to establishing the Gαi2–Rac–NOX–ROS–HIF1α signaling axis in LPA-stimulated glycolytic response, these results also point to the metabolic nodes stimulated by LPA as potential therapeutic targets in ovarian cancer. This is further supported by the observations that the glycolytic pathway inhibitors could attenuate the proliferation of diverse cancer cells including ovarian cancer cell lines (36). Although the components of glycolytic pathway such as GLUT1, HKII, pyruvate kinases-M2, and lactate dehydrogenase-A have been evaluated for their efficacy as chemotherapeutic targets in many cancers (37, 38), there is a paucity of such information in the context of ovarian cancer. It is significant to note here that HKII is overexpressed in ovarian cancer tissues (39), and it has been identified as a negative prognostic factor in many cancers (40–42). Therefore, we interrogated whether HKII could serve as a therapeutic target in ovarian cancer. To test, we used two independent ovarian cancer cell-line–derived xenograft (CDX) mouse models in vivo using SKOV3-ip and OVCAR8 cell lines. SKOV3-ip or OVCAR8 cells were injected subcutaneously into the right flank, and the animals were examined every day for the appearance of palpable tumors. When the tumors reached 50 mm³ (day −11), 3-bromopyruvate (3-BP) was intraperitoneally administered in two cycles. Tumor growth in these animals was monitored and the studies were terminated at day 36, before the animal welfare was compromised. As shown in Fig. 7, the administration of 3-BP inhibited the growth of both SKOV3-ip and OVCAR8 xenograft tumors. A reduction in tumor volume was observed from day −14 following the administration of 3-BP in SKOV3-ip xenograft tumors (Fig. 7A and B) along with a significant regression in tumor growth (Fig. 7C). Tumors harvested from these animals were processed for IHC analysis using antiphosphohistone H3 antibodies (anti-PHH3) to determine the mitotic index of the cells. Results indicated that the mitotic figures labeled by anti-PHH3 were greatly reduced in cancer tissues derived from 3-BP–treated animals (Fig. 7D). Quantification of the data indicated that the residual tumor tissue derived from 3-BP–treated animals showed 75% reduction in mitotic cells, indicating the inhibitory effect of 3-BP on mitotic proliferation (Fig. 7E). A reduction in tumor volume could also be seen in OVCAR8 xenograft tumors from day −3 onward following 3-BP treatment (Fig. 7F and G). Anti–PHH3-staining of OVCAR8 xenograft tumor tissues from 3-BP–treated animals showed a similar decrease in the mitotic index (Fig. 7H and I). Together, these findings identify the signaling nodes in aerobic glycolysis modulated by LPA—especially that of HKII—as potential therapeutic targets in ovarian cancer.

Despite newer findings on the etiology and biology of ovarian cancer, it still remains the most lethal gynecologic cancer with the 5-year survival rate of 47% (43). While there has been a marginal increase in 5-year survival over the past 10-years, an effective targeted therapy for ovarian cancer is still lacking. Despite the observation that ovarian cancer patients respond very well to the initial platinum-based therapy, they often suffer from disease recurrence. The absence of an effective targeted therapy is one of the factors that has contributed to the relatively marginal improvement in patient survival seen with ovarian cancer. Newer regimens targeting angiogenesis and polyadenosine diphosphate (ADP)-ribose polymerases are being adapted as second-line therapies for relapsing ovarian cancer, and their effect on overall survival is rather modest (44). In addition, the etiological and genetic dichotomy underlying type I and II ovarian cancers as well as the subtypes within each of them poses significant challenges in the development of an effective targeted therapy regimen. Based on the observations that almost all of the recurrent ovarian cancer patients present with ascites (45, 46) and the ascitic fluid contain a high concentration of LPA, which promotes both mitogenic and motogenic signaling (47, 48), several laboratories including ours have focused on identifying a therapeutic target in LPA-signaling nodes. As previously discussed (17), targeting LPA receptors has proven to be a nonviable strategy due to the high concentration of LPA in the ascites as well as the presence of LPA-synthetic machinery in close proximity to LPARs in the membrane. Thus, it is of critical interest to identify the signaling nodes downstream of LPARs that can serve as potential therapeutic target in ovarian cancer. With this overarching goal, our previous studies have shown that LPA could stimulate the epithelial-to-mesenchymal transition in ovarian cancer cells through Gαi2-dependent increase in the levels of HIF1α (17). Our current study extends these findings further by demonstrating the role of LPA in orchestrating a pseudohypoxic response that induces metabolic reprogramming in ovarian cancer cells toward aerobic glycolysis. Our results also indicate that this particular LPA-stimulated signaling node involving Gαi2, Rac, NOX, ROS, and HIF1α is independent of ovarian cancer subtypes and the mutational status of the genes associated with these subtypes, including p53 and BRCA1/2. More significantly, we demonstrate here the therapeutic potential of targeting HKII, a downstream link in LPA-stimulated metabolic programming, in inhibiting ovarian cancer growth using a preclinical CDX-mouse model.

Previous studies from several laboratories including ours have shown that LPA stimulates an increase in HIF1α levels in ovarian cancer cells. While the role of HIF1α in metabolic programming in tumor cell hypoxia is well known, its role in LPA-mediated metabolic reprogramming in normoxia is largely uncharacterized. Likewise, the mechanism by which LPA rapidly upregulates the levels and activity of HIF1α has not been elucidated. Our observation that LPA stimulated increase in HIF1α levels and activity could be seen by 2.5 minutes is the first report of its kind, and it indicates that the effect of LPA on metabolic reprogramming via this signaling nexus is an acute and rapid event. Furthermore, our findings that LPA stimulates an increase in HIF1α levels via Rac–NOX–ROS conduit and that this signaling nexus is critically involved in LPA-mediated metabolic reprogramming are quite novel and hitherto unreported. In addition, with the use of multiple ovarian cancer cell lines with varying genetic mutational signatures (Fig. 1), we demonstrate here that the metabolic reprogramming regulated by LPA is not dependent on p53, MYC, or BRCA1/2, PI3K, PTEN, or KRAS mutations. Such a similitude of response elicited by LPA in all of the cell lines identifies this signaling nexus as a potential candidate for targeted therapy in ovarian cancer. Finally, the importance of this pathway is further validated by our results using CDX-animal studies. Our observation that an HKII inhibitor significantly attenuated the growth of the ovarian cancer xenograft tumor growth highlights the overall importance of this pathway in ovarian cancer progression and its potential for specific targeting in ovarian cancer patients.

No potential conflicts of interest were disclosed.

Conception and design: J.H. Ha, R. Radhakrishnan, K. Moxley, C. Isidoro, P. Mukherjee, Y.S. Song, D.N. Dhanasekaran

Development of methodology: J.H. Ha, R. Radhakrishnan, M. Jayaraman, M. Yan, K.-M. Fung, D.N. Dhanasekaran

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.H. Ha, R. Radhakrishnan, M. Jayaraman, M. Yan, J.D. Ward, K.-M. Fung

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.H. Ha, R. Radhakrishnan, M. Jayaraman, M. Yan, J.D. Ward, K.-M. Fung, K. Moxley, A.K. Sood, C. Isidoro, P. Mukherjee, D.N. Dhanasekaran

Writing, review, and/or revision of the manuscript: J.H. Ha, R. Radhakrishnan, M. Jayaraman, J.D. Ward, K. Moxley, A.K. Sood, C. Isidoro, P. Mukherjee, D.N. Dhanasekaran

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.H. Ha, R. Radhakrishnan, M. Jayaraman, J.D. Ward, Y.S. Song

Study supervision: D.N. Dhanasekaran

This research was supported by National Institutes of Health grants CA116984, CA123233 (to D.N. Dhanasekaran), GM103639 (to D.N. Dhanasekaran and J.H. Ha) and Priority Research Centers Program grant 2009-0093820, the BK21 Plus Program grant 5256-20140100 through the National Research Foundation of Korea (to Y.S. Song). We also thank the Stephenson Cancer Center, OUHSC, Oklahoma City, OK; Mary K. Chapman Foundation Grant Award for Stephenson Cancer Center-MD Anderson Collaborative Research initiative (to D.N. Dhanasekaran, K. Moxley, and A.K. Sood) and an Institutional Development Award (IDeA) from the National Institute of General Medical Sciences of the National Institutes of Health under grant P20 GM103639 for the tissue pathology, IHC analysis, and fluorescence imaging services.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.