Sildenafil Suppresses Inflammation-Driven Colorectal Cancer in Mice

Inflammation is a driver of carcinogenesis, and patients with inflammatory bowel disease (IBD) have increased risk for developing colorectal cancer (1, 2). This association provides a rationale for the utility of anti-inflammatory therapies for colorectal cancer prevention (3). Indeed, epidemiologic evidence has shown that long-term non-steroidal anti-inflammatory drug (NSAID) use correlates with reduced incidence of colorectal cancer including in patients without IBD (4, 5). NSAIDs are widely known to prevent adenoma formation, but might also be an effective treatment for some tumors (6, 7). Although the effects of NSAIDs are presumed to be due to inhibition of cyclooxygenases (COX) and prostaglandin synthesis, COX-independent antitumor effects have also been observed. Exisulind/Aptosyn is a sulfone derivative of the NSAID sulindac with no NSAID activity, but it can block colon cancer cell growth by inhibiting phosphodiesterases (PDE; ref. 8). Exisulind has been reported to induce apoptosis of colon cancer cells in vitro by increasing 3′, 5′-cyclic guanosine monophosphate (cGMP) and activation of type-1 cGMP-dependent protein kinase (PKG1; ref. 9). Clinical trials in human familial adenomatous polyposis (FAP) patients and also sporadic adenomas, demonstrated that exisulind could induce regression of colorectal polyps by promoting apoptosis and mucus differentiation in the glandular epithelium (10, 11). Exisulind faltered in a phase III clinical trial and ultimately was not approved due to undisclosed safety issues (12). This outcome might be attributed to the fact that exisulind is a weak phosphodiesterase-5 (PDE5) inhibitor, and that therapeutic doses are associated with hepatotoxicity (13).

Interest in PDE5 as a cancer target waned with exisulind, but compelling evidence has accumulated that indicates a critical role for cGMP in the regulation of epithelial homeostasis and protection against colitis and colon cancer in mice. The endogenous peptide hormones guanylin and uroguanylin generate cGMP in the intestinal epithelium by activating receptor guanylyl-cyclase C (GCC; ref. 14). Knockout mice that are defective in cGMP signaling components exhibit crypt hyperplasia, increased luminal apoptosis, and reduced differentiation, particularly of goblet cells (15–17). Treatment of wild-type mice with the PDE5 inhibitor vardenafil suppressed proliferation and apoptosis, and increased secretory cell density, indicating that homeostatic regulation of the intestinal epithelium by cGMP is a normal physiologic process (18). The antiproliferative effect of cGMP suggested a tumor suppressive role, and this idea was strengthened by the observation that GCC knockout mice are more susceptible to tumorigenesis in both carcinogen-induced, and in Apc^min mouse models (19, 20). Lending further support for a tumor suppressive role of endogenous cGMP, treatment with exogenous uroguanylin was able to reduce tumorigenesis in Apc^min mice (21). The mechanisms underlying the antitumor effects of cGMP in the intestine are not fully understood. It is possible that direct inhibition of proliferation by cGMP could suppress tumor progression. However, recent work from several laboratories has highlighted barrier-protective functions of cGMP, which when compromised in GCC-deficient mice lead to increased susceptibility to colitis (22–24). Moreover, increasing intestinal cGMP with either GCC agonists or with PDE5 inhibitors can protect against experimental colitis in mice (18, 25). Because inflammation is a central driver of many cancers, it is reasonable that promoting epithelial barrier function could also contribute to the prevention of intestinal tumorigenesis by cGMP.

The molecular mechanisms underlying the effects of cGMP in the intestine are currently an area of active investigation. The upstream cGMP-generating machinery includes guanylin/uroguanylin that bind and activate GCC in the epithelial plasma-membrane. The homeostatic effects of cGMP have been shown to be mediated by PKG2, but the downstream effector pathways are less clear (17, 18). In gastric and colon cancer cell lines, cGMP has been reported to inhibit Ras/ERK (26, 27) and β-catenin signaling (9, 28), either of which could suppress proliferation and promote differentiation. In both cell lines and in mice, cGMP has been shown to inhibit the AKT and JNK pathways (18, 20), which regulate homeostasis in the intestinal epithelium. More recently, it was reported that by suppressing AKT, cGMP/PKG2 activates Forkhead Box O (FoxO) transcription factors in both colon cancer cell lines and in vivo (29). As tumor suppressors, FoxO can inhibit proliferation in dividing cells and provide antioxidant protection to quiescent cells (30, 31). It is therefore likely that FoxO is an important downstream effector of cGMP on proliferation and barrier maintenance in the intestinal epithelium.

Despite growing evidence for the anti-inflammatory role of cGMP signaling in the intestine, the effect of increasing cGMP on inflammation-driven carcinogenesis has not been tested. By increasing epithelial cGMP, PDE5 inhibitors can protect against dextran-sulfate sodium (DSS)-induced colitis in mice and can activate several antitumor effector pathways. The objective of the present study was therefore to test the ability of the PDE5 inhibitor sildenafil to suppress tumorigenesis in the azoxymethane/dextran sulfate sodium (AOM/DSS) model of colorectal cancer in mice.

Age-matched, male C57Bl/6J mice were purchased from The Jackson Laboratory at 6 weeks of age. Animals were acclimated for 2 weeks in the animal husbandry facility prior to experimentation. For all studies, the mice were maintained in a controlled environment 20°C and 14-hour light/10-hour dark cycle with free access to food and water. Pharmaceutical grade PDE5 inhibitor sildenafil (Revatio) was dissolved in water immediately prior to use. During various treatment courses, mice were allowed to drink freely from water bottles, with a calculated sildenafil dose of 5.7 mg/kg daily throughout the disease course. The Augusta University Institutional Animal Care and Use Committee approved all mouse procedures.

Inducing disease in this model was performed as described previously (18). Briefly, age-matched male mice were treated with 3% w/v dextran sulfate sodium (DSS) (36,000–50,000 Da; MP Biomedicals) in the drinking water for 5 days ad libitum, then switched to normal drinking water for 5 days during the recovery stage. Disease activity index (DAI) included measures of weight loss (0 = 0%, 1 = 1%–5%, 2 = 5%–10%, 3 = 10%–20%, and 4 = >20%), diarrhea (0 = normal stool, 1 = soft stool, 2 = very soft stool, 3 = diarrhea, 4 = dysenteric diarrhea) and rectal bleeding (0 = none, 1 = occult bleeding, 2 = blood on anus, 3 = blood on fur/tail, and 4 = gross bleeding). Maximal disease activity index and intestinal barrier permeability were observed on day 3 of the DSS recovery stage. To assess intestinal barrier permeability, age-matched male mice were starved overnight and given fluorescein isothiocyanate (FITC)-dextran (0.5 mg/g body weight, 4 kDa; Sigma-Aldrich) by oral gavage. Serum was collected 1.5 hours after oral gavage and the concentration of FITC-dextran was measured by quantitating fluorescence in a Cytation multiplate reader (Biotek).

Procedures followed here were carried out as described previously with the following modifications (32). Mice were given a single 7.5 mg/kg intraperitoneal (i.p.) dose of azoxymethane (13.4 mol/L 98%, Sigma Aldrich) and subsequently subjected to three consecutive cycles of DSS treatment (3% w/v DSS for 5 days) at weeks 9, 12, and 15 of age. Recovery periods after first and second cycles of DSS were for 14 days. After the third DSS cycle, mice were allowed to recover for 21 days prior to euthanasia. Throughout the study, body weight was measured to monitor disease progression. All mice were sacrificed at 18 weeks of age and the colorectums were processed for quantitation of polyp number, size, and histologic analyses. Briefly, colorectums were flushed with ice cold PBS and longitudinally cut open, flattened, highlighted with 1% Alcian blue dye, and placed on a light box for macroscopic high-resolution image captures. These images were enlarged for identification and marking of suspected polyps, and ImageJ software was used for enumeration and size quantification of the identified polyps. Size classification for polyps are the following: small (less than 4 mm²), medium (4–8 mm²), and large (greater than 8 mm²).

Mucosal tissues were scraped and polyps were excised with scissors and forceps and processed with lysis buffer containing 1% NP40, 0.1% SDS, protease, and phosphatase inhibitors (EMD Millipore). Polyps were dissociated with a short pulse in a tissue grinder. Tissue lysates were prepared and separated on PAGE gels as described previously (17). Protein was quantified using a BCA protein quantification kit (Thermo Scientific). Antibodies used for immunoblotting recognized β-actin (1:2000, Sigma), PKG1 (1:1000, gift from Robert Feil), PKG2 (1:1000, Santa Cruz Biotechnology), Phospho-VASP (Ser239; 1:1000, Cell Signaling Technology), VASP (1:1000, Cell Signaling Technology), PDE5 (1:1000, Cell Signaling Technology), β-catenin (1:500, BD Transduction), Cyclin D1 (1:1000, Cell Signaling Technology), Phospho-Akt (Ser473; 1:1000, Cell Signaling Technology), and Akt (1:1000, Cell Signaling Technology).

Formalin-fixed colorectal tissues were embedded in paraffin and sectioned at a 5-μm thickness by the Augusta University Histology core. Sections were subsequently deparaffinized in xylene, rehydrated with decreasing concentrations of ethanol, and boiled in citrate buffer for antigen-retrieval. Sections were either stained directly with hematoxylin and eosin (H&E) or Alcian-Blue/Periodic acid-Schiff (AB/PAS), or immunohistologically with antibodies: Ki67 (1:100, Dako Cytomation) and cleaved caspase-3 (1:500, Cell Signaling Technology). Antibody visualization was with an anti-rat/anti-rabbit ABC staining kit (Santa Cruz Biotechnology) for DAB amplification. Proliferation and apoptosis indices were standardized by counterstaining with hematoxylin and analyzed using ImageJ software. AB/PAS quantitation was done by expressing area of stained tissue as a function of total area. All image analyses used at least three animals and at least five random fields encompassing a minimum of 50 crypts per mouse. For immunofluorescence, double labeling of paraffin embedded sections, CD11b (1:100 dilution, Novus Biologicals) and GR-1 (1:100 dilution, eBioscience) were used, followed by anti-rabbit/anti-rat Alexa Fluor 647 and 488 (1:200 dilution; ThermoFisher Scientific). Sections were mounted in aqueous mounting media with DAPI (ThermoFisher Scientific) and visualized via confocal microscopy (Zeiss LSM 780 Upright Confocal). Quantitation of CD11b and GR-1 double positive staining cells was absolute cell counts from the distal colon per mouse.

Mucosal tissues were scraped and polyps were excised with scissors and forceps and flash-frozen in TRIzol reagent (Life technologies). RNA was DNAse treated (TURBO DNA-free kit, Life Technologies) and converted to cDNA using M-MLV reverse transcriptase (Invitrogen). Quantitative PCR (qRT-PCR) analysis of the cDNA was performed using SYBR Green PCR Master Mix (Applied Biosystems). Relative expression levels were calculated using the 2^–ΔΔCT method with β-actin (ACTB) as a reference. Amplifications were performed in triplicate wells, and melt curve analysis was done to confirm the specificity of the primers used. Primers were designed using Primer Blast Software (NCBI; Table 1).

All data were expressed as mean ± SEM, unless otherwise stated. A Student t test was used for the comparison of two means, unless otherwise stated. The statistical significance was set at P < 0.05.

We have previously reported that IP injection of the PDE5 inhibitor vardenafil can activate cGMP signaling in the intestinal epithelium of mice, leading to changes in epithelial homeostasis and reduced sensitivity to DSS (18). To determine whether orally administered sildenafil could also activate cGMP signaling in the colon epithelium, phosphorylation of the endogenous PKG substrate vasodilator-stimulated phosphoprotein (VASP), and homeostasis were assessed. An acute dose of sildenafil (1.4 mg/kg i.p., 4 hours) caused VASP phosphorylation (Fig. 1A), and after 7 days led to increased goblet cell density and reduced proliferation in the colon (Fig. 1B and C). When challenged with DSS, mice treated with sildenafil exhibited increased barrier function and reduced measures of colitis, including weight loss, bleeding, and diarrhea compared with controls (Fig. 1D and E). Taken together, these results demonstrate that different PDE5 inhibitors have similar effects on epithelial homeostasis and that oral delivery of sildenafil can protect the barrier from DSS-induced damage.

Inflammation resulting from DSS-induced damage to the colon epithelium can promote tumor growth following treatment with the mutagen azoxymethane (AOM). AOM/DSS treatment is widely used because the resulting adenomas show similar genetic mutations and dysplasia-to-carcinoma sequence observed in human colitis-associated colorectal cancer (33). Our demonstration that PDE5 inhibitors can protect mice from DSS prompted us to determine whether sildenafil could also suppress tumorigenesis in the AOM/DSS model. In these experiments, mice were treated with AOM and following a recovery period they were exposed to three successive cycles of DSS (Fig. 2A). Sildenafil administered in the drinking water throughout the treatment period conferred protection from body weight loss during the first cycle of DSS, but was not effective during the subsequent DSS cycles (Fig. 2B). Colons from both control and sildenafil treated animals developed numerous polyps in the distal-rectal area with fewer forming proximally (Fig. 2C). Enumeration of the polyps revealed a large range in multiplicity with a median of 10 polyps per mouse in control animals. Sildenafil treatment significantly inhibited polyp formation with a median of 5 polyps per mouse (P = 0.0073; Fig. 2D).

Although a large range of polyp size was observed in the colons of mice treated with AOM/DSS in our studies, histologic analysis showed that the majority of the lesions were adenomatous polyps with leukocyte infiltration and varying degrees of dysplasia (Fig. 3A). The suppressive effect of sildenafil on polyp number was more pronounced in the smaller polyps, but the treatment did not affect mean polyp size (Fig. 3B and C). Although there was no difference in polyp size between groups, further analysis showed that there was reduced infiltration of myeloid cells, and reduced expression of inflammatory mediators in polyps from mice treated with sildenafil compared with those from control animals (Fig. 3D and E). This was most significant for iNOS (Nos2) and IFNγ (Ifng) (P = 0.012 and 0.038, respectively).

Histologic analysis of polyps from AOM/DSS-treated mice showed that the sildenafil treated animals harbored polyps that were less proliferative, and whereas there was a small increase in apoptotic index in treated mice, the difference was not significant (Fig. 4A and B). There was a notable increase in mucus content in polyps from sildenafil treated animals compared with those from controls (Fig. 4C). Much of the mucus was observed in glandular cavities that were exaggerated in many of the sildenafil treated polyps. These cavities also contained abundant apoptotic cells, indicating a more differentiated epithelium.

Components of the cGMP signaling axis have been reported to exhibit altered expression patterns in human colorectal cancer compared with normal mucosa, but this has never been examined in murine colorectal cancer models (34–36). The expression of cGMP signaling components in polyps from AOM/DSS treated mice was measured using qPCR and immunoblotting. The mRNA levels for GCC (Gucy2c) and guanylin (Guca2a) were significantly downregulated in polyps compared with normal colon epithelium (P = 0.021 and 0.003, respectively; Fig. 5A). Although uroguanylin did not change, the loss of guanylin expression was dramatic in both control and sildenafil polyps, with a 53 and 35-fold loss, respectively (Fig. 5A). Sildenafil treatment did not significantly affect the expression of these components in the polyps. Signaling downstream of cGMP typically involves PKG1, PKG2, and PDE5 (37). Analysis of mRNA expression for these components showed a small but insignificant increase in (Prkg1) and a similarly small decrease in (Pde5a), but a dramatic increase in (Prkg2) (Fig. 5B) compared with colon mucosa. These changes were not affected by treatment of the mice with sildenafil at either the mRNA or the protein levels (Fig. 5C).

Administration of sildenafil to mice throughout the AOM/DSS treatment phases resulted in 50% inhibition of polyp multiplicity, but it remains unclear what process of tumorigenesis was affected. To gain additional insight into the mechanism of action of sildenafil, the drug was administered to mice early during the AOM phase prior to exposure to DSS (Fig. 6A). In these experiments, the control animals produced a median of 10.5 polyps per mouse, but surprisingly, the sildenafil treatment significantly inhibited polyp formation by 38% with a median of 6.5 polyps per mouse (P = 0.039; Fig. 6B). This result suggested that the protracted presence of sildenafil during the later promotion phases did not affect polyp formation despite the marked effect on polyp phenotype. To test this possibility, sildenafil was administered subsequent to the early carcinogenesis phases of AOM treatment and the first DSS cycle. When administered at this later stage, there was a small but insignificant reduction in the number of polyps per mouse (Fig. 6C).

The polyps formed in mice treated with sildenafil during later stages had a similar but less pronounced phenotype as those with sildenafil treatment throughout. They were similar to those with sildenafil treatment throughout, in that they showed increased mucus, and reduced proliferation and pro-inflammatory cytokine expression (Fig. 6D–F). The reduced inflammatory leukocytes and cytokines observed in the treated polyps in the present study are consistent with the idea that sildenafil can suppress myeloid-derived suppressor cells, as has been reported using non-colon cancer models (38, 39). However, the dramatic effect of sildenafil when added early, and removed before inducing inflammation by DSS treatment suggests that immune regulation is unlikely to play a major role in the suppression of tumorigenesis. This idea is further supported by the lack of effect when sildenafil was administered during the second and third DSS cycles.

A large body of evidence derived mostly from knockout animals has identified diverse regulatory roles for cGMP in the intestine, including secretion, epithelial barrier integrity, and proliferation differentiation along the crypt/villus axis. Activation of cGMP signaling with GCC agonists plecanatide and dolcanatide has recently been reported to reduce symptoms of colitis in mice (25). The ability of GCC agonists to reduce diarrhea in the context of colitis is intriguing as they are established secretagogues, but these observations highlight the translational potential of barrier-protection by elevating cGMP levels. PDE5 inhibitors amplify cGMP generated by the endogenous GCC agonists, and it was shown here that exposing mice to sildenafil in the drinking water elicited similar changes in colon epithelial homeostasis and protection from DSS-induced damage as reported previously using vardenafil (18). This strongly supports the endogenous regulatory role of cGMP in the intestine as the present study used a different mouse strain, a different PDE5 inhibitor, and a different mode of administration.

On the basis of the ability of PDE5 inhibitors to suppress DSS-induced colitis, we reasoned that increasing cGMP in this manner would also inhibit polyp formation in the AOM/DSS model of colon cancer in mice. This model is widely used because it recapitulates many aspects of human disease (33). The initial genotoxic AOM treatment initiates mutations that are promoted to develop into colorectal adenomas by chronic inflammation that is induced and maintained by cycles of DSS. It was shown here that treating animals with sildenafil throughout AOM/DSS treatment resulted in significantly fewer polyps per mouse relative to control animals on water alone. Surprisingly, the size of the polyps from animals treated with sildenafil did not differ from those from control animals. This result was not anticipated based on work with human colon cancer cell lines that predicts antitumor effects of cGMP signaling that include inhibition of proliferation, increased apoptosis, and inhibition of angiogenesis (9, 40, 41). Other in vitro studies have demonstrated that cGMP signaling suppresses proliferation without affecting apoptosis, and notably, that it can induce goblet cell differentiation (17, 42). Although sildenafil did not affect apoptosis, the polyps from treated mice were slightly less proliferative and more differentiated. Intriguingly, this latter finding supports previous observations in polyps from human patients treated with exisulind/Aptosyn (10, 11).

Another unexpected observation was reduced tumor inflammation in polyps from sildenafil treated mice as reflected by both cytokine mRNA expression and infiltration of myeloid leukocytes. Inflammatory leukocyte infiltration is a poor prognostic indicator for human colorectal cancer, where it is thought to promote tumor progression and metastasis (43, 44). Polyps in the AOM/DSS model typically do not become invasive or metastatic, so the possibility that sildenafil might promote a less aggressive tumor phenotype was not evaluated here. The phenotype of the polyps from sildenafil treated animals ostensibly paralleled that of normal tissue; both showed increased mucus differentiation, and reduced proliferation and inflammation. It is possible that barrier protection by elevating cGMP can also account for the reduced inflammation of the polyps, particularly smaller and more differentiated ones. The mechanisms underlying the effects of cGMP on epithelial homeostasis and barrier protection in the normal colon epithelium are not fully understood. Suppression of Akt, β-catenin/TCF, and JNK signaling, with subsequent activation of FoxO have all been suggested to occur downstream of cGMP in either normal mouse tissue, in some human colon cancer cell lines, or both (18, 20, 28, 29). In the present study, reduced levels of β-catenin were observed in the polyps from sildenafil treated animals, but this was more likely due to the lower abundance of proliferating cells and inflammation, because there was no evidence of PKG activity such as phosphorylation of vasodilator stimulated phosphoprotein (Supplementary Fig. S1).

It has been reported that several components of the cGMP signaling axis are downregulated either in human colon tumors relative to normal tissue, or by human colon cancer cell lines relative to tumors (17, 34, 36, 45). The lack of detectable cGMP signaling in the polyps of sildenafil treated animals is not surprising considering the striking loss of the cGMP-generating machinery including both guanylin and GCC. The loss of guanylin by the polyps supports previous studies with human specimens, but the loss of GCC does not occur in human tumors (35). We speculated that polyps might increase the expression of alternative PDEs that might compensate for inhibition of PDE5. Although none of the PDE isoforms were affected by sildenafil treatment, we observed a striking increase in the expression of the dual specificity PDE10a (Supplementary Fig. S2). This observation supports a recent study using human specimens (46), indicating that PDE10 upregulation by colon tumors may be a universal phenomenon. Importantly, inhibitors of this isoform can suppress proliferation in colon cancer cell lines, which underscores the potential therapeutic importance of this isoform for colon cancer treatment (46, 47). Whether PDE10a upregulation contributed to the suppression of cGMP signaling in the sildenafil treated polyps remains to be determined. The loss of guanylin in the polyps most likely contributed to the lack of detectable PKG activation, and it is therefore difficult to assess the in vivo relevance of antitumor signaling pathways previously defined in vitro. The use of exogenous GCC agonists such as linaclotide and plecanatide might elicit more robust intratumoral cGMP signaling. However, future studies will need to determine whether these intraluminal peptide drugs that do not enter circulation are capable of penetrating more dysplastic tumors.

The utility of the AOM/DSS model as an experimental system is that the initiation and promotion phases of carcinogenesis are sufficiently separate to permit testing a drug's mechanism. The lack of marked cGMP signaling in established polyps suggested that the central antitumor effect of sildenafil is early during the initiation process. This idea was confirmed by the observation that sildenafil did not significantly affect polyp multiplicity when administered later during the promotion phase. It was still able to reduce proliferation and inflammation, and promote differentiation, but did not affect the number of polyps formed. In contrast, when sildenafil treatment was restricted to the AOM treatment phase, the polyp multiplicity was dramatically reduced. This unexpected result demonstrated that the important polyp-preventing effect of sildenafil treatment occurred prior to DSS-induced inflammation. Although it is presently unclear how increasing cGMP would affect the genotoxic effect of AOM, it is possible that the reduced proliferative compartment in sildenafil treated animals could limit the effect of mutagenic agents. These data support a previous study with exisulind in rats, where multiplicity was reduced following AOM treatment only if administered during the initiation phase (48). The profound effect of sildenafil early during carcinogenesis does not preclude the importance of its ability to protect against the effects of DSS. This is evident because there was a larger reduction in polyp multiplicity when sildenafil was present throughout all phases of the model compared with treatment during the AOM phase alone. In addition, the DSS dose used in the present study was relatively high, and it is possible that the effect on multiplicity could be even more profound if lower doses were used.

Taken together, our results demonstrate that targeting PDE5 with selective inhibitors can inhibit colon tumorigenesis in mice. This contrasts with work suggesting that sildenafil might promote melanoma aggressiveness by potentiating ERK signaling (49). This ERK promoting pathway is not present in colon and gastric cancer cells, where cGMP has been shown to inhibit the ERK pathway (26, 27). Although the molecular pathways mediating the preventative effect of sildenafil as shown here are not known, the mechanism appears to involve homeostatic changes in the epithelium that confer resistance to both genotoxic and barrier-damaging agents. Despite some clinical efficacy of the low-affinity inhibitor exisulind, efforts targeting PDE5 for colon cancer chemoprevention waned due to toxicity at effective doses. However, results shown here using the high-affinity and relatively safe PDE5 inhibitor sildenafil have shown similar effects in mice as were observed by higher levels of exisulind in human FAP patients. This work suggests that the utility of PDE5 as a target for chemoprevention of colorectal cancer might be revisited using currently available high-affinity inhibitors.

No potential conflicts of interest were disclosed.

Conception and design: B.N. Islam, D.D. Browning

Development of methodology: B.N. Islam, N. Singh, P.C. Rodriguez, D.D. Browning

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B.N. Islam, S.K. Sharman, Y. Hou, A.E. Bridges, F.G. Berger, S. Sridhar, D.D. Browning

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B.N. Islam, N. Singh, S. Kim, R. Kolhe, J. Trillo-Tinoco, D.D. Browning

Writing, review, and/or revision of the manuscript: B.N. Islam, S.K. Sharman, A.E. Bridges, J. Trillo-Tinoco, S. Sridhar, D.D. Browning

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.N. Islam, S. Sridhar

Study supervision: B.N. Islam, S. Sridhar, D.D. Browning

This work was supported by the NCI grant #CA172627‐01A1.

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.