Grape Seed Extract Efficacy against Azoxymethane-Induced Colon Tumorigenesis in A/J Mice: Interlinking miRNA with Cytokine Signaling and Inflammation

See related commentaries by Katiyar and Athar, p. 617 and by Kong et al., p. 622

Colorectal cancer (CRC) is the second leading (both genders combined) cause of cancer-related deaths and the third most frequently diagnosed malignancy in the United States (1). Adoption of Western and sedentary lifestyles increase CRC incidence worldwide, with most (∼75%) CRCs developing sporadically (2). Despite efforts, compliance with screening recommendations is low and not enough to halt cancer growth/progression and prevent associated mortality (2). Traditional CRC chemopreventive agents, such as nonsteroidal anti-inflammatory drugs (NSAID), exert gastrointestinal side effects, suggesting that additional approaches are needed to manage CRC (3, 4). Use of natural products with limited or no toxicity is one such approach to prevent, suppress, or reverse the carcinogenesis process (5). In this regard, several studies have indicated that one third of all cancer deaths in United States could be prevented through diet modification and that high consumption of fruits and vegetables or their bioactive components could significantly decrease CRC (5–7). Grape seed extract (GSE) is one such natural agent, rich in proanthocyanidins, which has shown cancer chemopreventive and anti-cancer efficacy in various preclinical cancer models including CRC (6, 8, 9). Importantly, clinical studies have shown that active components of GSE are bioavailable and well-tolerated in humans (10). However, to our knowledge, GSE has not been examined in mouse model of sporadic CRC. Considering that the sporadic CRC patient population represents the largest clinical population of diagnosed CRCs, we aimed to evaluate GSE efficacy against a chemical carcinogen azoxymethane (AOM)-induced colon tumorigenesis in A/J mouse model that mimics generation of sporadic neoplastic lesions as observed in the human colon (11). Furthermore, we also conducted mechanistic studies interlinking miRNA expression with cytokine signaling and inflammation in overall efficacy of GSE against colon tumorigenesis.

GSE-standardized preparation was a gift from Kikkoman Corp. As reported earlier (12), the analyzed composition of GSE is 89.3% procyanidins, 6.6% monomeric flavanols, 2.24% moisture content, 1.06% of protein, and 0.8% of ash. AOM was purchased from Sigma and dissolved in saline. Primary antibodies include, anti-PCNA (proliferating cell nuclear antigen) from Dako; anti-iNOS, anti-VEGF, anti-Ago2 from Abcam; anti-COX-2, anti-C-myc, anti-β-catenin from Santa Cruz Biotechnology; anti-phospho-ERK1/2 and anti-NF-κB/p65 from Cell Signaling; and anti-mouse and anti-rabbit horseradish peroxidase (HRP) secondary antibodies from Invitrogen and Cell Signaling, respectively. Mouse cytokine antibody array was from RayBiotech. miRNA was isolated using Qiagen miRNeasy Kit, amplified by Qiagen RT² miRNA qPCR kit, and quantified using miRNA Cancer PCR array (Qiagen).

Animal experiment was done under institutional guidelines using approved IACUC protocol. Animals were maintained under standard conditions with free access to water and food. GSE was mixed in AIN-76A powder diet (Dyets Inc.) at 0.25% and 0.5% (w/w). Five-week-old male A/J mice (from Jackson Laboratory) were fed AIN-76A powder diet for 2 weeks (to acclimatize them to new diet) and then divided into 5 groups: (i) negative control group (n = 20) on AIN-76A diet; (i) positive control group (AOM, n = 35), injected with 5 mg/kg AOM intraperitoneally (i.p.) once a week for 6 weeks, and on AIN-76A diet; (iii) AOM + 0.25% GSE group (n = 35), GSE supplemented diet was started 2 weeks after last AOM injection and continued for 18 weeks (n = 25) or 28 weeks (n = 10); (iv) AOM + 0.5% GSE group (n = 35), GSE supplemented diet was started 2 weeks after last AOM injection and continued for 18 weeks (n = 25) or 28 weeks (n = 10); and (v) 0.5% GSE group, GSE supplemented diet was started at 5 weeks of mice age and continued for remainder of study. The selection of 2 GSE doses for current study was based on our published studies wherein these GSE doses showed a dose-dependent chemopreventive effect against AOM-induced aberrant crypt foci formation in F344 rats and strong efficacy against small intestine tumorigenesis in APC min/+ mouse models (8, 9). Similarly, the selection of AOM-induced colon tumorigenesis experimental protocol in A/J mice in present study was based on our and others recent studies showing measurable to strong colon tumor numbers for agent chemopreventive efficacy studies (5, 13). Body weight and diet consumption were recorded weekly. At the end of the study, at 33 and 43 weeks of age, mice were sacrificed, entire colon excised starting from ileocecal junction to anal verge and cut open longitudinally along main axis and gently flushed with ice-cold PBS, divided into 3 equal sections (proximal, medial, and distal), tumors counted, and tumor diameters measured with digital calipers under dissecting microscope. Colon tissues and/or tumors were either fixed flat in formalin and embedded in paraffin, snap-frozen in liquid nitrogen, or stored in Qiagen RNAlater.

Anatomic gadolinium-enhanced T1-weighted MRI was also used to noninvasively assess colon tumor progression in mice. Bruker multi slice multi echo (MSME) T1-scans were conducted at a Bruker 4.7 Tesla PharmaScan (Bruker Medical) following a bolus injection of 0.1 mmol/kg MultiHance via a tail catheter on anesthetized mice (2% isoflurane). A mouse volume transmitter/receiver coil (36-mm diameter) was used for all MRI studies using flowing parameters: field of view (FOV) = 4cm, slice thickness = 1 mm, number of slices 16 (coronal) and 40 (axial), TR/TE = 725/11 ms, number of averages 2, matrix size = 256 × 256, flip angle = 180. Total acquisition time was 6.5 minutes for each plane. All imaging acquisition and analysis was conducted using Bruker ParaVision software (at the Animal Imaging Shared Resources, University of Colorado Anschutz Medical Campus).

Tissue lysates of colonic mucosa with tumors from randomly selected animals in different groups were applied to Mouse Cytokine Antibody Array. Expression of various cytokine molecules was analyzed in duplicate on the membranes, which were scanned and quantified by ImageJ, and densitometric data analyzed using antibody array analysis tool.

miRNA isolation was done using Qiagen miRNeasy Kit, starting with 20 mg of mouse colonic mucosa with tumors. Isolated miRNA was used for miRNA arrays using Qiagen RT² miRNA qPCR kit using Applied Biosciences Fast-Real time PCR 7500. Replicates were conducted and the average ΔC_t value of each miRNA assay was calculated for each group. The ΔΔCt was calculated for each miRNA between = ΔC_t (AOM + 0.5% GSE) − ΔC_t (AOM), then the fold change was calculated for each miRNA.

Paraffin-embedded sections (5 μm) of distal colon were subjected to standard immunohistochemical (IHC) procedures described previously (8). Dilutions of primary antibodies used were anti-PCNA (1:250), anti-iNOS (1:100), anti-COX-2 (1:50), anti-phosho-ERK1/2 (1:100), anti-Ago2 (1:100), anti-β-catenin (1:50), anti-NF-κB/p65 (1:50), anti-VEGF (1:100), and anti-C-myc (1:25). Negative staining controls were used for each protein. Apoptotic cells were detected by DeadEnd Colorimetric TUNEL system (Promega; ref. 8). Microscopic analyses were conducted using Zeiss Axioscope 2 microscope; photomicrographs captured by Carl Zeiss AxioCam MrC5 camera with Axiovision Rel 4.5 software. Quantification of IHC data includes counts from both colonic mucosa and tumor staining and is shown as mean ± SEM from 10 fields per section in each group. Statistically significant difference between AOM and AOM + GSE groups was analyzed for tumor multiplicity, tumor size, and protein expression using one-way ANOVA (Sigma Stat 2.03, Jandel Scientific). P ≤ 0.05 was considered significant.

GSE feeding did not show considerable difference in diet consumption and body weight gain profiles between AOM- and AOM + GSE–fed mice during entire study (data not shown). Real-time assessment of colon tumor formation was conducted in negative control, AOM, and AOM + 0.5% GSE mice at 32 weeks of age using CE-MRI (Fig. 1A). We did not detect any tumors/other lesions or abnormalities in negative control mice, which were further confirmed upon necropsy (conducted 1 week post-MRI at 33 weeks of mice age; Fig. 1A, left). In AOM group, 6 tumors were detected on gadolinium-enhanced T1-MRI images and 12 tumors were eventually found upon necropsy (Fig. 1A, middle). Mice in AOM + 0.50% GSE group showed 3 tumors on MRI whereas necropsy identified 4 tumors (Fig. 1A, right). Comparing the number of tumors between AOM and AOM + 0.5% GSE groups, GSE treatment resulted in 50% to 67% inhibition as identified by gadolinium-enhanced T1-MRI and necropsy, respectively.

Figure 1B summarizes the effect of GSE feeding on AOM-induced tumor formation in both colon and small intestine following necropsy. Whereas AOM group showed 8.56 ± 0.58 colon tumors/mouse at 33 weeks of age, GSE treatment post-AOM exposure significantly reduced colon tumorigenesis in a dose-dependent manner; specifically, there was a 46% (P < 0.001) and 55% (P < 0.001) inhibition in the number of colon tumors in AOM + 0.25% GSE and AOM + 0.5% GSE groups (Fig. 1B), respectively. Significant inhibition in colon tumorigenesis by GSE feeding was also observed at 43 weeks of mouse age; however, the 2 doses showed comparable efficacy (Fig. 1B). GSE feeding also significantly decreased the number of small intestine tumors compared with AOM group (Fig. 1B); however, again there was no GSE dose–response effect.

As tumor incidence is greater in distal region of the colon in AOM mouse model, GSE effect on tumor formation in different regions of the colon was also analyzed. At 33 weeks of age, AOM group showed 2.97 ± 0.3 colon tumors in the distal region/mouse, whereas AOM + 0.5% GSE group had 1.66 ± 0.2 tumors, indicating a 44% (P < 0.001) reduction in tumor formation in the distal region by GSE feeding (data not shown). This trend was also observed in other regions (proximal and medial) of the colon and also at 43 weeks of mouse age (data not shown); 0.25% and 0.5% GSE doses showed comparable efficacy. Additional analysis of colon tumorigenesis study data showed that GSE feeding also reduces the overall tumor size (Fig. 1C). Importantly, while there was a significant dose-dependent decrease in the number of tumors in the size range of 2 to 3 mm, there were absolutely no tumors larger than 3 mm in AOM + GSE–fed groups compared with AOM alone (Fig. 1C).

CRC growth and progression to advanced stages is associated with enhanced cell proliferation and evasion of apoptosis (11), and therefore the agents which could inhibit proliferation and induce apoptosis could play an important role in controlling CRC growth and progression (5). As we found strong efficacy of GSE in preventing AOM-induced colon tumorigenesis, tissue samples were next analyzed for proliferation and apoptotic indices. Qualitative examination of PCNA-stained sections showed a decrease in PCNA-positive cells in AOM + GSE–fed groups compared with AOM alone (Fig. 2A). Further quantification showed 17% ± 2% PCNA-positive cells in AOM + 0.25% and AOM + 0.5% GSE–fed groups of mice compared with 47% ± 5%, in AOM alone (Fig. 2A), accounting for 65% (P < 0.001) decrease in proliferation. Regarding in vivo apoptosis, GSE feeding on AOM-induced colon tumorigenesis, qualitative microscopic examination showed increased number of TUNEL-positive cells in AOM + GSE–fed groups compared with AOM alone (Fig. 2B). The number of apoptotic cells was about 5% (Fig. 2B, P < 0.001) in both AOM + GSE–fed groups compared with 1.7% ± 0.43% in AOM alone, accounting for a 3-fold (P < 0.001) increase (Fig. 2B).

The mixture of host-derived cytokines produced in the tumor microenvironment plays an important role in cancer progression (14) where cancer cells use cytokines to promote growth, diminish apoptosis, and stimulate both invasion and metastasis (14). Thus, management of cytokine equilibrium could be an important strategy for both prevention and treatment of various malignancies including CRC (14). To examine whether GSE efficacy against AOM-induced colon tumorigenesis involves altered expression of various cytokines, we used inflammatory cytokine antibody array analysis on colonic mucosa with tumors from AOM and AOM + 0.5% GSE groups (Fig. 3A). Our results indicated that GSE feeding causes an alteration in the expression of various cytokines/interleukins involved in innate immunity, proliferation, and apoptotic cell death. Specifically, GSE up regulated the levels of B-lymphocyte chemoattractant (BLC; ∼8-fold), granulocyte macrophage colony-stimulating factor (GM-CSF; ∼10-fold), chemokine CCL1 (I-309; ∼9-fold), interleukin (IL)-1α (∼10-fold), IL-23 (∼20-fold), and M-CSF (∼9-fold) but downregulated IL-1ra (∼26-fold), IL-2 (∼9-fold), and IL-27 (∼8-fold) levels compared with AOM alone (Fig. 3A).

miRNAs are small noncoding RNAs which are implicated in cell-cycle regulation, cell growth and differentiation, stress response, and apoptosis (15). In addition, miRNAs also play an essential role during carcinogenesis and their expression patterns have been now correlated with disease prognosis in patients with cancer (15). Furthermore, miRNA knockdown or re-expression induces drug sensitivity, impedes proliferation, and reduces invasion of cancer cells (15). In view of above summarized critical roles of miRNAs, next we assessed whether their altered expression also contributes to GSE efficacy against AOM-induced colon tumorigenesis. A comparison of miRNA expression profiles of the tissue from AOM and AOM + 0.5% GSE groups showed that GSE feeding causes an up regulation of Snord 68 (∼6,500-fold), miR-19a (∼2,400-fold), miR-20a (∼1,000-fold), and miR-let7a (∼1,200-fold; Fig. 3B) butdown regulates miR-205 (∼400-fold), miR-135b (∼100-fold), miR-196a (∼24-fold), miR-21 (∼25-fold), miR-148a (∼21-fold), and miR-103 (∼9-fold; Fig. 3B). Furthermore, IHC analysis revealed about 6.5-fold induction in the level of Ago2, which is involved in miRNA processing (16), in AOM + 0.5% GSE group compared with AOM alone, indicating an increased RNA-induced gene silencing in response to GSE feeding (Fig. 3C).

On the basis of an extensive literature search for the possible role of above identified various cytokines/interleukins and miRNAs with altered expression following GSE treatment and the associated cross-talks among them, we were able to identify a pattern in their modulation as discussed in detail in the Discussion (14, 17, 18). Specifically, we found that GSE mediates its efficacy against AOM-induced colon tumorigenesis by possibly modulating NF-κB and its downstream targets related to inflammatory signaling, β-catenin signaling, and mitogen-activated protein kinase (MAPK) pathway. Accordingly, next we investigated their possible roles in GSE efficacy against AOM-induced colon tumorigenesis.

Transcription factor NF-κB mediates the transcription of various genes which are associated with tumor initiation, promotion, progression, metastasis, as well as inflammation, including those for COX-2, iNOS, and VEGF (19). In addition, inflammation is associated with CRC development, and chronic inflammation increases CRC risk (20). In addition, experimental and epidemiologic evidences indicate an upregulation of COX-2, iNOS, and VEGF (downstream target genes of NF-κB) in various malignancies, including human CRCs (21). Together, on the basis of this information and our array data analyses suggesting the involvement of NF-κB and its downstream inflammatory molecules in GSE efficacy, IHC analyses were done to examine their expression. Our results showed that GSE feeding indeed interferes with NF-κB activity, as observed by a significant decrease in nuclear expression of NF-κB/p65 in the distal colon/tumor tissues by both GSE doses (Fig. 4A). Next we examined COX-2 and iNOS levels (Fig. 4B and C). AOM alone treated mice showed a significant increase in COX-2 and iNOS immunoreactivity scores [∼10-fold (P < 0.001) and ∼2.5-fold (P < 0.001), respectively] in the distal colon/tumor tissues, compared with untreated negative controls. Importantly, mice in the AOM + GSE–fed groups exhibited a significant downregulation of COX-2 and iNOS protein levels in the distal colon/tumor tissues (Fig. 4B and C). Similarly, a decrease in VEGF immunoreactivity scores (Fig. 4D) was also observed in the distal colon/tumor tissues from the mice in AOM + GSE–fed groups compared with AOM alone.

Our array data analyses also suggested a possible involvement of β-catenin and MAPK signaling in GSE efficacy against AOM-induced colon tumorigenesis, and accordingly, next we focused on these 2 pathways. β-Catenin is important in the development of CRC and inappropriate stabilization, translocation and activation of β-catenin occur in sporadic and familial CRC resulting in downstream modulation of oncogenic target genes such as c-myc and VEGF (22). IHC results for β-catenin staining revealed a significant dose-dependent reduction (P < 0.001) in the nuclear expression of β-catenin in the distal colon/tumor tissues of AOM + GSE–fed groups compared to AOM alone (Fig. 5A). Following the above results and further examining the downstream β-catenin pathway targets, we assessed the expression of c-myc (22). The quantification of IHC staining for c-myc in distal colon/tumor tissues revealed a significant (P < 0.001) dose-dependent reduction in AOM + GSE groups compared with AOM alone (Fig. 5B). The RAS/RAF/MEK/MAPK signaling cascade plays an important role in tumorigenesis including CRC (23), and accordingly, we also investigated the role of this pathway in GSE efficacy against AOM-induced colon tumorigenesis by assessing the expression of extracellular signal–regulated kinase 1/2 (ERK1/2), which is phosphorylated in response to RAS activation (23). Microscopic examination of distal colon/tumor tissue sections from all groups revealed a significant reduction (94%–97%, P < 0.001) in the phosphorylation of ERK1/2 (Fig. 5C) in AOM + GSE–fed groups compared with AOM alone.

This is the first study reporting that long-term dietary feeding of GSE (0.25% and 0.5% w/w in AIN-76A diet) for 18 and 28 weeks following exposure to AOM results in a significant decrease in AOM-induced colon tumor multiplicity. More importantly, GSE feeding caused a significant reduction in colon tumor size in a dose-dependent manner. Furthermore, GSE feeding did not show any effect on food consumption and body weight gain profiles, which is consistent with previous anti-cancer efficacy studies with GSE in other cancer models (6). In the present study, we also established feasibility of noninvasive real-time examination of colon lesions by gadolinium-enhanced MRI, which allowed the visualization of colon tumors in AOM mouse model with 75% to 81% sensitivity. This, somewhat low, sensitivity of MRI (as compared with necropsy results) can be attributed to nonoptimal experimental conditions during tumor visualization (such as small size lesions and bowel movements), which could possibly be improved in future by altering image contrast as reported for other studies (24). Biomarker analyses in colonic mucosa with tumors indicated that GSE feeding decreases proliferation but induces apoptosis in its chemopreventive efficacy. We also focused our efforts on identifying the targets and pathways modulated by GSE in its colon cancer chemopreventive efficacy by using cytokine and miRNA arrays. Our results and their analyses showed that a wide number of cytokines/ILs and miRNAs were significantly modulated by GSE which were associated with alterations in NF-κB, β-catenin, and MAPK signaling, suggesting that inflammation, proliferation, and apoptosis are targeted by GSE in its chemopreventive efficacy. However, unlike previous studies with GSE by us and others showing a dose-dependent cancer chemopreventive and anti-cancer efficacy in various cancer models including AOM-induced ACF in rats (8), in present study, we observed comparable efficacy of GSE on colon tumor numbers and most of the associated mechanistic pathways. However, GSE did affect colon tumor sizes as well as β-catenin and C-myc expression in colon tumor tissues in a dose-response manner. Together, these findings suggest that possibly a lower dose of GSE is optimum for its efficacy in AOM-induced colon tumorigenesis in A/J mice and that additional proliferation markers should be examined in future to further support β-catenin and C-myc findings.

Further analysis of array data showed that GSE decreases IL-2, which is an important finding because activation of NF-κB, a transcription factor associated with inflammation, leads to IL-2 production (25). GSE also significantly downregulated the levels of IL-27 and IL-1α, which are known to modify NF-κB signaling (26–28). Similarly, miRNA array analysis showed that GSE upregulates miR-19a, which is linked to NF-κB regulation (29). miR-20a is known to inhibit hypoxia-inducible factor (HIF)-1α pathway, which plays a major role in the survival of cancer cells in tumor microenvironment; importantly, GSE upregulated miR-20a suggesting its activity in inhibiting HIF-1α pathway and its downstream target VEGF (30, 31). Similarly, miRNA array results showed that GSE causes a dramatic downregulation of miR-205, which targets VEGF (32) that is also transcriptionally regulated by NF-κB. Consistent with these observations, our results showed that GSE decreases the expression of nuclear NF-κB p65 subunit as well as those of COX-2, iNOS, and VEGF. These results are important as overexpression of proinflammatory molecules, such as iNOS and COX-2, is observed in AOM-induced CRCs, and selective COX-2 inhibitors have been shown to prevent disease progression (33, 34). In addition, decreased expression of the pro-angiogenic factor VEGF suggests that GSE might also interference in tumor angiogenesis; a strong dose-dependent decrease in colon tumor size by GSE supports this notion. Together, these results suggested that GSE exerts its efficacy against colon tumorigenesis, at least in part, by modulating NF-κB signaling and its downstream transcriptional targets COX-2, iNOS, and VEGF, which are implicated in inflammation, tumor promotion, progression, and metastasis of CRCs (19, 35, 36).

β-Catenin signaling is extensively studied for its critical role in human CRC growth and progression (22), suggesting that the agents targeting this pathway would be useful in CRC control. In our study, we found that GSE causes an increase in M-CSF, which is linked to β-catenin signaling (37). In addition, GSE also significantly decreased miR-135b, which is also linked to β-catenin pathway (38). Moreover, GSE upregulated let-7a, which is implicated in inhibiting the expression of the c-myc oncogene, a downstream target of β-catenin (39). Consistent with these results, GSE significantly downregulated β-catenin protein expression, which was also associated with the decreased expression of its downstream target C-myc (22), suggesting their possible role in GSE efficacy against colon tumorigenesis.

Cytokine/IL array findings also showed upregulation of BLC, CCL1/I-309, and IL-1α but downregulation of IL-1rα and IL-27 by GSE; all these molecules are involved in regulating the MAPK pathway (40–43). Furthermore, in miRNA array analysis, GSE showed an upregulation of let-7a which is known to inhibit MAPK pathway (44). We also found that GSE decreases miR-205, which is upregulated in some malignancies and known to interact with both MAPK and NOTCH pathways (45). Notably, MAPK signaling cascade is known to play an important role in tumorigenesis by modulating cell growth, differentiation, proliferation, apoptosis, and migration (23). GSE also decreased miR-103, which is involved in oncogenic KRAS signaling and is upregulated in CRC cells (46). Our IHC results clearly showed that GSE causes a strong decrease in MAPK/ERK1/2 phosphorylation which is consistent with our cytokine/IL and miRNA array profiles. Together, these results suggested that GSE exerts its efficacy against colon tumorigenesis, at least in part, by also modulating KRAS/MAPK signaling.

In addition to the miRNAs described above and their association with signaling pathways, GSE also modulated the expression of several additional miRNAs, which are implicated in CRCs. For example, GSE downregulated miR-135b that is upregulated in CRC tissue compared with the normal surrounding tissue and has been specifically shown to target the APC gene involved in CRCs (47, 48). Similarly, GSE decreased miR-196a, the high levels of this miRNA exert a pro-oncogenic effect in CRC cells (49). Likewise, the ability of GSE to decrease miR-21 is of clinical significance because its increased levels correlate with increased stage of clinical CRCs (47, 50). Importantly, in this study, we also observed predominate global upregulation of miRNAs by GSE, coupled with downregulation of oncogenic miRNAs and upregulation of tumor-suppressive miRNAs. Accordingly, we also investigated a key protein involved in miRNA homeostasis (16), namely Ago2, which binds to the mRNA and represses translation or translocation. Colon tissue analysis revealed significant upregulation of Ago2 protein in colons of AOM + GSE–fed groups, which further supports our miRNA array results. Together, this is the first study showing alteration in miRNAs expression and cytokine signaling by GSE in its chemopreventive efficacy against AOM-induced colon tumorigenesis in A/J mouse sporadic CRC model.

No potential conflicts of interest were disclosed.

Conception and design: M. Derry, V. Balaiya, R. Agarwal, C. Agarwal

Development of methodology: M. Derry, V. Balaiya, A. Jain, S. Shrotriya, N.J. Serkova

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Derry, K. Huber, N.J. Serkova

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Derry, K. Raina, V. Balaiya, N.J. Serkova, R. Agarwal

Writing, review, and/or revision of the manuscript: M. Derry, K. Raina, V. Balaiya, N.J. Serkova, R. Agarwal

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Derry, R. Agarwal, C. Agarwal

Study supervision: R. Agarwal, C. Agarwal

This work was supported by NIH R01 AT003623, NCI Cancer Center P30 CA046934, and NCRR CTSA UL1 RR025780.

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.