A Novel Pharmacologic Activity of Ketorolac for Therapeutic Benefit in Ovarian Cancer Patients

Ovarian cancer is the leading cause of death from gynecologic malignancies and the second most common gynecologic cancer (1). Five-year patient survival remains less than 50% and the mortality rate has not changed appreciably in two decades (1). The majority of women are diagnosed with metastatic disease, and although a substantial proportion of women respond to initial treatment, recurrence is common (2). Despite concerted efforts, identification of effective targeted therapies has remained elusive in this disease (3). There remains a great need to identify new strategies to treat and manage ovarian cancer.

The Ras-homologous (Rho) family of small GTPases (Rac, Cdc42, and Rho) are key regulators of cancer-relevant cellular functions, including actin reorganization, cell motility, cell–cell, and cell–extracellular matrix (ECM) adhesion and invasion (4–6). In many human tumors (including colon and breast), there is clinical and experimental evidence that aberrant Rho-family signaling contributes to tumor growth, survival, invasion, and metastasis (4, 7–10). On the basis of these functions, Rac1 and Cdc42 have been recognized as attractive therapeutic targets (11, 12) and inhibitors are effective in experimental systems (13–19), but specific inhibitors of Rac1 or Cdc42 have not been translated to clinical use.

A Cdc42-selective inhibitor effectively blocked migration of two ovarian tumor cell lines (19, 20), suggesting that Rho-family GTPases may be potential therapeutic targets in ovarian cancer. Using findings obtained from a high-throughput screen of the Prestwick library of off patent, FDA-approved drugs and cheminformatics approaches, we identified the R-enantiomers of a limited number of non-steroidal anti-inflammatory drugs (NSAIDs) as inhibitors of Rac1 and Cdc42. The corresponding S-enantiomers are considered the active component in racemic drug formulations acting as NSAIDs with selective activity against cyclooxygenase (COX). One candidate, R-ketorolac, inhibited ovarian tumor cell migration and adhesion without causing cytotoxicity (21–24). Our data indicate that the clinically administered racemic ketorolac (Toradol) has two distinct pharmacologic activities; the well-established inhibition of COX 1 and 2 by S-ketorolac serving as the basis of the FDA-approved indication for pain management, and a previously unrecognized property of Rac1 and Cdc42 inhibition conferred by the R-enantiomer. Cell-based measurement of GTPase activity demonstrated that R-ketorolac specifically inhibits EGF-stimulated Rac1 and Cdc42 activation at low micromolar concentrations. The GTPase inhibitory effects of R-ketorolac in cells mimic those of established Rac1 (NSC23766) and Cdc42 (CID2950007/ML141)-specific inhibitors (19, 24).

In this study, we report that R-ketorolac achieved an effective concentration in peritoneal fluids and inhibited Rac1 and Cdc42 activity in cells retrieved from the peritoneal compartment of post-surgical ovarian cancer patients following administration of the racemic drug for postoperative pain management. A medical record review to compare the ovarian cancer–specific survival of ovarian cancer patients who did or did not receive ketorolac for postoperative analgesia revealed increased survival of patients receiving ketorolac. This observation is in keeping with reports for improved clinical outcomes associated with ketorolac usage, as compared with other NSAIDs, in breast cancer patients (25–27). Although it has been long recognized that R-enantiomers of NSAIDs are poor inhibitors of COX activity (28–30), potential pharmacologic activities or benefits of the R-enantiomers have remained largely unexplored. Our findings show that Rac1 and Cdc42 are unrealized therapeutic targets in ovarian cancer and use of ketorolac may benefit ovarian cancer patients.

Immunohistochemical staining was performed using standard procedures (see Supplementary Methods). Rac1 was stained with mAb (clone 102; BD Biosciences, 10155-1-AP) and Cdc42 was stained with rabbit pAb (Protein Tech Group). A Vectastain Ready-to-Use (RTU) ABC-peroxidase kit and ImmPact DAB (SK-4105) were used to visualize primary antibody labeling with hematoxylin nuclear counterstain (H-3401) for tissue staining and samples were mounted in VectaMount (H-5000; all from Vector Laboratories).

For large-scale ovarian tumor profiling, tissue microarrays were purchased from U.S. BioMax, Inc. (cat# OV1005 061 and OV8010 009). In total, 180 unique tissue samples were included in the evaluation; ranging from stage I–IV and grades 1–3 (see Supplementary Table S1). All tumor types were validated and staining was scored by a pathologist with gynecologic pathology specialty (Lesley Lomo, MD, Dept. of Pathology, University of New Mexico) and evaluated for location (nuclear and cytoplasmic), as well as intensity of positive staining. Scoring was based on the product of the percentage of cells stained and the intensity of the staining in each localization (3+: strong, 2+: intermediate, 1+: weak and 0: no staining), resulting in a minimum of 0 (100% cells × 0) and a maximum of 300 (100% cells × 3+).

Quantitative PCR (qPCR) analysis of Rho family GTPases was performed using Tissuescan Ovarian Cancer cDNA microarrays from Origene (cat# HORT301, HORT302, HORT303) and standard techniques (see Supplementary Methods and Supplementary Table S2). qPCR amplification used Qiagen Quantitect primers for Cdc42 (QT01674442), Rac1 (QT00065856), and RhoA (QT00044723), and β-actin (Origene), and custom Rac1b primers from Invitrogen. Primer information for Rac1b, methods for PCR product validation and additional information on the Ovarian Cancer cDNA arrays is presented in Supplementary Methods and Supplementary Table S2. Analysis of serous cancer only is shown in Supplementary Fig. S1. C_t values were obtained using iCycler software. Relative expression levels were determined using ΔΔC_t values.

A phase 0 trial investigating the use of postoperative ketorolac was reviewed and approved by the University of New Mexico Health Sciences Center Human Research Review Committee (clinicaltrials.gov, NCT01670799). Patients presenting with a new diagnosis of ovarian, fallopian tube or primary peritoneal cancer were screened for eligibility. Eligible women were at least 18-years-old, had an ECOG Performance Status <2 and had consented to a planned debulking surgery—Consent was obtained before surgery if primary eligibility was met. Secondary eligibility after surgery included confirmed histologic diagnosis of epithelial ovarian, fallopian tube, or primary peritoneal cancer; optimal cytoreduction and placement of an intraperitoneal port for planned chemotherapy; adequate renal function and no postoperative complications prohibiting ketorolac use. Patients with known bleeding disorders or other contraindications to NSAID use were excluded.

Subjects received a single IV dose of Toradol (15 or 30 mg based on the patient age and creatinine clearance) within the first 72 hours of surgery when all clinical safety parameters were met. Use of other NSAIDs during the trial were not permitted; however, narcotic regimens were allowed for postoperative pain management. All study protocols were reviewed by an independent data and safety monitoring board. Baseline ascites samples were obtained at surgery. Subsequent peritoneal fluid samples were collected from the intraperitoneal port before dosing and at 1, 6, and 24 hours after single-dose ketorolac administration. Peripheral blood was collected at the same time points. Serum and peritoneal fluid were separated from cellular material via low speed centrifugation. Tumor cells were further purified on Ficoll gradients to remove red blood cells and negative selection with anti-CD45 beads to remove lymphocytes. The resulting tumor cell fractions were analyzed by flow cytometry for EpCAM and MUC16/CA125 (see Supplementary Fig. S2 for representative purity).

R- and S-enantiomers of ketorolac were analyzed by high-performance liquid chromatography (HPLC) using published procedures (31, 32). See Supplementary Methods and Supplementary Fig. S3 for additional information. The R value for the standard curve of total ketorolac was 0.9997 and represented a concentration range that spanned established human serum concentrations (0.092–6.0 μg/mL).

Two methods were used to assess GTPase activity in cells based on effector binding. Commercial GLISA kits from Cytoskeleton, Inc., analyzed Rac1 (cat# BK-128), Cdc42 (BK-127), or RhoA (BK-124) per the manufacturer's instructions. Alternatively, Rac1, Cdc42, and RhoA activities were measured using a flow cytometric effector–binding assay (24, 33). Briefly, active Rac1 and Cdc42 GTPases in prepared cell lysates were quantified individually based on binding to GST-PAK1-PBD from Millipore (cat#14-864) immobilized on GSH beads and use of specific antibodies for bound GTPase detection. Antibodies used to quantify the amount of active (GTP-bound) GTPases captured on the beads are listed in Supplementary Methods. Fluorescence intensity (mean channel fluorescence, MCF) was measured by flow cytometry (Accuri C6). GTPase activity was calculated by the (MCF of sample group—MCF of unstimulated group)/MCF of stimulated group. Equal amounts of protein were used for each assay.

A medical record review was conducted under Institutional Review Board approval with a waiver of patient consent. Ovarian cancer patients were identified from the New Mexico Tumor Registry (NMTR), a member of the population-based Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute (34). Inclusion criteria were as follows: invasive, epithelial ovarian cancer (any histology), ages 40 to 79 years at diagnosis, years of diagnosis 2004 to 2006, and receipt of surgery at an Albuquerque, New Mexico hospital (only three hospitals in the metropolitan area provide this level of surgery). Diagnosis years of 2004 to 2006 ensured at least 6 years follow-up (mortality followed through Dec 31, 2012) for each patient. We abstracted the surgical medical records for all analgesics and anesthesia medications used before hospital admission, during surgery and hospital stay, and given at discharge. Of the 138 potential cases, 6 women did not undergo surgery because of advanced disease/severe comorbidities or desired palliative care only, 1 woman had her surgery in another state, 2 women died before surgery, and medical records were not located for 6 women, leaving 123 women in the final analysis.

qPCR findings for GTPase expression levels were analyzed using one-way ANOVA followed by Dunnett multiple comparisons test to determine differences between ovarian cancer grade. IHC data were analyzed using one-way ANOVA followed by the Tukey post-test to determine significant differences between groups. Data obtained from patient fluid and cell samples were analyzed as a repeated measure ANOVA followed by the Dunnett multiple comparisons test to determine significant differences between groups. For the retrospective medical record review, information from the medical record was merged with information from the NMTR for final analysis. Clinical and treatment characteristics of patients who did and did not receive perioperative ketorolac were compared with χ² tests and t tests. In a preliminary, crude analysis the Kaplan–Meier method was used to estimate the survival probabilities. The difference in survival based on receipt of perioperative ketorolac was examined using the stratified log-rank test to adjust the effect of a single categorical factor such as age group, AJCC stage, etc. Because this was an observational study and not a randomized controlled trial, the final analysis was based on a Cox proportional hazards model to adjust for clinical and treatment characteristics that may have differed between those who did and did not receive perioperative ketorolac. We estimated the hazard ratio (HR) for ovarian cancer–specific mortality comparing those who did and did not receive perioperative ketorolac while adjusting for age at diagnosis (<50, 50–64, ≥65) AJCC stage (I, II, III, and IV), completion of chemotherapy as planned (yes and no), and receipt of neoadjuvant chemotherapy (yes and no). On the basis of the Cox proportional hazards regression, an example survival plot is presented in Fig. 5. Additional survival plots are presented in Supplementary Figs. S4–S7.

On the basis of the established functions of Rho-family GTPases, ovarian cancer metastasis is predicted to be strongly dependent on Rac1/Cdc42–regulated pathways for exfoliation, formation of multicellular aggregates, mesothelial adhesion, and localized invasion into the interstitial collagen-rich submesothelial matrix (2, 3, 35). To test whether these GTPases are dysregulated in ovarian cancer, we examined grade-dependent expression of Cdc42 and Rac1 protein by immunohistochemical staining of human tumor samples (Fig. 1) and GTPase mRNA using quantitative PCR analysis of ovarian cancer tissue cDNA arrays (Fig. 2 and Supplementary Fig. S1). Cdc42 protein overexpression levels were highly significant for malignant, high-grade tumors (P < 0.001) compared with lower-grade tumors without an apparent increase in mRNA levels. In contrast, there was little evidence of increased expression of Rac1 protein with increasing grade (Fig. 1). However, significantly elevated expression of a constitutively active splice variant Rac1b (36–38) was detected in ovarian tumors (Fig. 2). These findings provide evidence for aberrant expression of Rac1 and Cdc42 in ovarian cancer and suggest that they may represent therapeutic targets for this disease.

In previous work, we identified R-ketorolac as an inhibitor of Rac1 and Cdc42 at low micromolar concentrations (21–24). R-ketorolac is inactive against the enzyme targets of S-ketorolac, COX-1, and COX-2 (28–30), and the inhibitory concentration (IC)₅₀ values for inhibition of Rac1 and Cdc42 by S-ketorolac were more than 100-fold greater than R-ketorolac (24). Thus, the racemic (R,S) ketorolac possesses two distinct pharmacologic activities and our findings identify R-ketorolac as a novel inhibitor of Rac1 and Cdc42.

In this “phase 0” feasibility study, ketorolac was administered for its FDA-approved indication for postoperative pain management. Eligible patients had suspected advanced-stage ovarian, fallopian tube, or primary peritoneal cancer with planned optimal cytoreductive efforts. Secondary eligibility was met if the patient had confirmed ovarian cancer, was optimally cytoreduced and had an intraperitoneal port placed for planned peritoneal chemotherapy. The patients had no active postoperative bleeding, did not require therapeutic anticoagulation, and had good postoperative organ function. Forty-two patients met primary eligibility, and considering secondary eligibility requirements, 20 patient samples were collected at surgery. Samples of blood and peritoneal fluids after ketorolac administration were obtained from 13 patients. Pathologic analysis of the 20 surgical samples confirmed fifteen patients (75%) had stage III or IV disease. Histologically, 100% had high-grade carcinoma; one was a carcinosarcoma, 16 of 19 were pure serous carcinomas, one was primary peritoneal and one fallopian tube primary, and the rest ovarian primary tumors. The average age was 60.8 (range, 33–81), race and ethnic distribution was 85% Caucasian (29.4% Hispanic), 5% American Indian, and 10% Black or African American. Ketorolac dosages administered on the basis of clinical indications were 15 mg (33% of patients) or 30 mg (77% of patients). As illustrated in Fig. 3A, blood and peritoneal fluid were obtained at T = 0, 1, 6, and 24 hours after administration of the recommended dose of ketorolac.

Serum and peritoneal fluid samples were analyzed by HPLC to resolve and quantify R- and S-ketorolac enantiomers to determine enantiomeric ratios and distribution over time (Fig. 3B and C). Clinical-grade ketorolac tromethamine is a 1:1 mixture of R- and S-enantiomers (Supplementary Fig. S3); however, the racemic distribution favors the R-form in both serum and peritoneal fluids at each time point. This is in keeping with the established shorter half-life of S-ketorolac in human serum based on differences in pharmacokinetic parameters for each enantiomer (28, 30, 39). Ketorolac distributes to the peritoneum within 1 hour after IV administration, and ketorolac levels in the peritoneal fluids are nearly equivalent to those present in the serum at 6 hours and decline dramatically by 24 hours in both serum and peritoneal fluids. Our results represent the first evidence of ketorolac distribution to peritoneal fluids.

The IC₅₀ values for Rac1 and Cdc42 by R-ketorolac are 0.57 and 1.07 μmol/L, respectively, whereas the IC₅₀ values for S-ketorolac for these targets was >100 μmol/L (24). The concentrations of R- and S-ketorolac in the peritoneal fluids were 0.98 and 0.32 μmol/L, respectively, 6 hours after IV ketorolac administration. Thus, R-ketorolac achieved concentrations in the peritoneal fluids at or above the IC₅₀ values for Rac1 and Cdc42 and is predicted to inhibit these GTPase targets in cells obtained from this compartment.

Tumor cell–enriched fractions were prepared from ascites samples obtained at the time of cytoreductive surgery from ovarian cancer patients and post-surgery immediately prior and 1, 6, and 24 hours post IV ketorolac administration (Fig. 3A). Both Rac1 and Cdc42 were highly activated in freshly isolated tumor cells from ascites and the activity level declined within 48 hours in culture medium (Fig. 4A), suggesting that the ovarian tumor environment fosters Rac1 and Cdc42 GTPase activation. Post-surgery, we observed a statistically significant decrease in Rac1 and Cdc42 activity with time after ketorolac administration (Fig. 4B and C). In contrast, RhoA activity was insensitive to ketorolac (Supplementary Fig. S8), further affirming the selectivity of the drug. R-ketorolac predominates in the peritoneal fluids and the S-enantiomer is virtually undetectable at 24 hours (Fig. 3C), indicating that the R-enantiomer is bioactive and accounts for the observed inhibition of the GTPases in vivo.

Perioperative ketorolac was used in 14% of the 123 women in the study. Younger women (<50 years) were more likely than older women to receive perioperative ketorolac (P < 0.05); all other clinical and treatment characteristics were similar between the two groups. At 60 months of follow-up, 3 of 17 ketorolac-treated patients (18%) and 40 of 92 nontreated patients (43%) had died of ovarian cancer. Stratified log-rank tests for categorical factors such as age group, AJCC stage, completion of chemotherapy as planned, and receipt of neoadjuvant chemotherapy as coded in Table 1, showed a consistent ketorolac survival benefit in each strata (see Supplementary Figs. S4–S7). The better survival in women treated with ketorolac consistently found in the stratified analysis was also evident in the proportional hazards analysis when we adjusted for age at diagnosis, AJCC stage, completion of chemotherapy as planned, and receipt of neoadjuvant chemotherapy. The adjusted HR for ovarian cancer–specific mortality associated with perioperative ketorolac (yes vs. no) was 0.30 [95% confidence interval (CI), 0.11–0.88; Table 1]. On the basis of the proportional hazards model, an example survival plot is shown in Fig. 5 for women who had AJCC stage III cancer, were 50 to 60 years at diagnosis, did not receive neoadjuvant therapy, and completed post-surgery chemotherapy as planned. Other survival plots are shown in Supplementary Figs. S4–S7, and although these plots highlight the results for women who completed their post-surgery chemotherapy as planned, all combinations of women defined by stage of disease, age at diagnosis, neoadjuvant therapy, and post-surgery chemotherapy showed a consistently better survival with ketorolac versus without. These preliminary findings suggest that perioperative ketorolac reduces ovarian cancer–specific mortality.

In many human cancers, aberrant Rho-family GTPase activity or downstream signaling pathways are associated with increased aggressiveness and poor patient prognosis (4, 6, 8, 9, 11, 12). The specific mechanisms by which Rho-family GTPases modulate tumor development and progression remain under investigation (4, 5, 7–9, 14); however, experimental evidence places Rac1 and Cdc42 within the metastatic cascade. Little is known regarding Rac1 and Cdc42 expression in ovarian cancer. We demonstrated elevated expression of Rac1 and Cdc42 in human ovarian cancer specimens and high activity of these GTPases in freshly isolated tumor cells from ascites obtained at surgery (Figs. 1 and 4). In a recent study, high Rac1 protein expression in ovarian cancer was associated with early recurrence and poor prognosis (40). Furthermore, partial silencing of Rac1 by shRNA decreased tumor cell proliferation, migration and invasion in culture, and decreased growth of subcutaneous ovarian cancer xenografts in vivo (40). We find that R-ketorolac inhibits adhesion and invasion of primary human ovarian tumor cells from patient ascites (24), thereby indicating that pharmacologic inhibition of Rac1 and Cdc42 also blocks these tumor-relevant functions. These observations in conjunction with inhibition of ovarian tumor cell migration by a Cdc42-specific inhibitor (19) indicate the potential value of targeting Rac1 and Cdc42 in ovarian cancer.

In the present study, we show evidence that Rac1 and Cdc42 inhibition can be achieved in ovarian cancer patients following administration of racemic ketorolac (Toradol). Ketorolac is a 1:1 racemic mix of the R- and S-enantiomers. The S-form inhibits COX enzymes, which confers the drug's anti-inflammatory activities. The COX inhibitory action of S-ketorolac supports its indication for postoperative pain management, but also limits its long-term use due to COX-related toxicity (28–30). R-ketorolac has little activity against COX (28–30), and therefore is not functional as an NSAID, but is bioactive and inhibits Rac1 and Cdc42 (24). Importantly, the levels of R-ketorolac within the peritoneal fluids were sufficient to inhibit Rac1 and Cdc42 activity in cells obtained from the peritoneal cavity following ketorolac administration. The innovative phase 0 clinical trial design enabled real-time sampling of fluids and cells from the peritoneal cavity. Direct demonstration of the difference in racemic distribution of ketorolac enantiomers illustrates the value of a study design that allows direct testing of drug and cell activities within peritoneal fluids rather than extrapolation from serum drug levels.

Furthermore, the peritoneal bioactivity of ketorolac is shown to have benefit for ovarian cancer patient outcomes. We found that perioperative use of ketorolac reduces ovarian cancer–specific mortality (Fig. 5). There is precedence in the literature that ketorolac usage in the perioperative period is associated with improved cancer outcomes. The first observation was made for breast cancer patients in 2010 (25). In this study, ketorolac use was associated with a decrease risk of breast cancer relapse (HR, 0.37; 95% CI, 0.0–0.79). Follow-up articles noted that this relapse reduction was most pronounced in the first 24 months post-surgery (26, 27). No change in breast cancer recurrence was noted in patients who received sufentanil, clonidine, ketamine, or other intraoperative analgesics. Lung cancer patients receiving ketorolac displayed improved overall survival as well (41). The authors hypothesize that the benefit is due to the anti-inflammatory actions of ketorolac, particularly on the extravasation of circulating tumor cells in the transient inflammatory environment stimulated by surgery (42). Ketorolac appears to have more pronounced positive outcomes than other NSAIDs (42), and this may be based on the combined impact of anti-inflammatory activity by the S-enantiomer and R-enantiomer effects on Rac1 and Cdc42 leading to decreased adhesion and implantation of circulating or residual tumor cells. Ketorolac is not cytotoxic to ovarian tumor cells (24), but predicted decreases in establishment or further development of micrometastases due to Rac1 and Cdc42 inhibition would be expected to improve response to subsequent chemotherapy, which cannot be initiated until patients have recovered from cytoreductive surgery.

Collectively, our findings support the potential repositioning of ketorolac as an addition to current ovarian cancer therapy. Our work demonstrates that the R-enantiomer of ketorolac acts as a first-in-class drug for inhibition of the cancer-relevant targets Rac1 and Cdc42 (23, 24) and provides the first evidence that these therapeutic targets can be inhibited in humans using an approved drug. There is precedence for pharmacologic activities dictated by R-enantiomers of specific NSAIDs against novel (non-COX) targets (43–45). For example, R-etodolac and its analogues SDX-301 and SDX-308 display antitumor activity in chronic lymphocytic leukemia and activity against multiple myeloma in cell and animal models (44, 46–49). R-etodolac also significantly suppressed tumors in a colitis-related mouse model colon cancer (43) and retarded tumor development and metastasis in a transgenic mouse model of prostate cancer (45). These examples and others demonstrate that R-enantiomers of NSAIDs can possess unanticipated anti-cancer activities based on interactions with non-COX targets. Further evidence that targeting Rac1 may provide therapeutic benefit in ovarian cancer was recently reported (50). Zoledronic acid is a nitrogen containing bisphosphonate that inhibits prenylation of small GTPases. Administration of this drug decreased growth of ovarian cancer peritoneal xenografts through inhibition of angiogenesis driven by a Rac1-mediated pathway (50). Collectively, our results suggest that racemic ketorolac may provide a survival benefit to ovarian cancer patients through inhibition of COX enzymes by the S-enantiomer and inhibition of the small GTPases Rac1 and Cdc42 by the R-enantiomer. Additional studies to determine whether clinical benefit can be observed in ovarian cancer patients through perioperative administration of ketorolac in a placebo-controlled clinical trial are in process.

T.I. Oprea is an employee of Sunset Molecular Discovery LLC and reports receiving commercial research grants from Givaudan Flavors Corporation. L. A. Sklar has ownership interest (including patents) in IntelliCyt. A. Wandinger-Ness, L.A. Sklar, T.I. Oprea, and L.G. Hudson are listed as co-inventors on a pending US patent application entitled “Modulators of GTPases and Their Use,” which is owned by the University of New Mexico and its technology transfer office, STC.unm, and is under an option to license. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Guo, L. Cook, T.I. Oprea, L.A. Sklar, C.Y. Muller, A. Wandinger-Ness, L.G. Hudson

Development of methodology: Y. Guo, S.R. Kenney, L.A. Sklar, E. Bedrick, C.Y. Muller, A. Wandinger-Ness, L.G. Hudson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Guo, S.R. Kenney, L. Cook, S.F. Adams, T. Rutledge, E. Romero, C.L. Wiggins, L. Lomo, C.Y. Muller, A. Wandinger-Ness, L.G. Hudson

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Guo, S.R. Kenney, L. Cook, T.I. Oprea, E. Bedrick, H. Kang, C.Y. Muller, A. Wandinger-Ness, L.G. Hudson

Writing, review, and/or revision of the manuscript: Y. Guo, S.R. Kenney, L. Cook, S.F. Adams, T. Rutledge, T.I. Oprea, E. Bedrick, C.L. Wiggins, H. Kang, L. Lomo, C.Y. Muller, A. Wandinger-Ness, L.G. Hudson

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Guo, E. Romero, T.I. Oprea, A. Wandinger-Ness

Study supervision: L. Cook, C.Y. Muller, A. Wandinger-Ness, L.G. Hudson

This work is dedicated to the memory of Austin Hudson-LaPore.

This study is supported by DOD OC110514 W81XWH-11-OCRP-TEA, UNM Science and Technology Corporation Gap Fund, Core Facility support (Flow Cytometry Shared Resource, Fluorescence Microscopy Shared Resource and Human Tissue Repository) from the University of New Mexico Cancer Research and Treatment Center (P30 CA118100). Focus Interactive Group grants from UNM Cancer Center (0990MD; 0990Q8 to A. Wandinger-Ness and L.G. Hudson; 1146 to L. Cook). Contract HHSN261201300010I from the National Cancer Institute. Center for Molecular Discovery grants from the NIH Molecular Libraries Program (MH074425 and MH084690 to L.A. Sklar). NCI R25CA153825 (predoctoral fellowship to Y. Guo); INBRE NCRR 5P20RR016480 (predoctoral fellowship to S.R. Kenney).

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.