MYC Inhibition Depletes Cancer Stem-like Cells in Triple-Negative Breast Cancer

Breast cancer is the most frequently diagnosed form of cancer and the second leading cause of cancer-related death among women both in the United States and worldwide (1). Gene expression profiling has identified several distinct molecular subtypes of breast cancers including luminal A and B, HER 2–positive, basal-like and normal-like (2, 3). Approximately 70% of the triple-negative breast tumors are identified as the basal-like subtype (4). Despite recent advances in breast cancer treatment, targeted therapeutics for triple-negative breast cancer (TNBC) do not exist because this subtype lacks druggable targets such as hormone receptors and amplified HER 2 expressions (5). Anthracycline and taxane-based chemotherapy is a standard of care for TNBC treatment. However, sooner or later, patients with TNBC will experience drug resistance, tumor relapse, and/or metastasis after a transient response to initial rounds of therapies (5–8). Thus, there is an urgent need to develop innovative and more effective therapeutic approaches that achieve a more durable response to TNBC treatment.

Cancer stem-like cells (CSC), also known as tumor-initiating cells (TIC), have the unique ability to self-renew, differentiate, and generate the diverse cells that comprise the tumor. Previous studies have demonstrated that CSCs are highly tumorigenic, as few as 100 CSCs are able to form tumors when xenotransplanted into immunocompromised mice, whereas tens of thousands of nontumorigenic cancer cells do not (9, 10). Thus, CSCs are thought to be the underlying cause of tumor recurrence and metastasis by giving rise to new tumors (11, 12). Furthermore, it appears that CSCs are resistant to current anticancer therapies, as demonstrated by the fact that they are selectively enriched in residual tumors after chemotherapy and/or radiotherapy (7, 8, 13–15). These new observations strongly suggest that the inability of current cancer therapies to eliminate CSCs in heterogeneous tumors may account for relapse and treatment failure. Thus, targeting of CSCs may represent a novel therapeutic strategy for improving the efficacy of cancer treatment. However, there is currently no FDA-approved therapeutics that can effectively eradicate drug-resistant CSCs in TNBC treatment.

Dysregulation of the MYC oncogene has been implicated in the pathogenesis of a variety of human cancers, including TNBC (16–19). Interestingly, MYC overexpression is associated with poor outcomes in breast cancer (19). Evidence also exists that elevated MYC expression is particularly common in the triple-negative subtype of breast cancers (18, 20, 21). MYC is a transcriptional target of Wnt/β-catenin and activation of the Wnt/β-catenin signaling pathway has been linked to CSC self-renewal in basal-like breast cancer (22, 23). Notably, MYC has been shown to be a key factor required for stem cell reprogramming (24). Furthermore, recent studies have suggested that MYC is required for β-catenin–mediated mammary stem cell amplification and tumorigenesis (25). However, it is not known if targeting MYC is a valid therapeutic strategy to eradicate drug-resistant CSCs for breast cancer therapy.

C1572, also known as Triptolide, was originally isolated from the medicinal vine Tripterygium wilfordii Hook F, which has been used in traditional Chinese medicine for centuries (26), particularly for the treatment of a variety of autoimmune diseases and as an immunosuppressant in patients with organ and tissue transplantations (27–29). Minnelide is a water-soluble prodrug of C1572 that has been shown to exhibit promising tumor-suppression effects in pancreatic cancer, although the mechanism(s) of action is elusive (30). Interestingly, C1572 also can protect mice against cisplatin-induced acute kidney injury and alleviate autosomal dominant polycystic kidney disease via stimulating calcium (Ca²⁺) channel polycystin-2 mediated Ca²⁺ release (26, 31).

In the present study, through unbiased drug screen, we have identified C1572 as a promising lead compound that selectively depletes drug-resistant CSCs via targeting MYC in human TNBC cells. Strikingly, our results reveal that C1572 is 100-fold more potent than the commercially available small-molecule inhibitor of MYC, JQ1 (32), in inhibiting MYC in TNBC cells. Furthermore, our studies have demonstrated for the first time that C1572-mediated tumor growth suppression and CSC depletion correlate well with a marked inhibition of MYC expression in residual TNBC xenograft tumor tissues. Collectively, these results suggest that pharmacologic inhibition of MYC by C1572 may represent a novel and effective therapeutic approach for eliminating drug-resistant CSCs in TNBC.

All preclinical animal studies were performed in compliance with the regulations and ethical guidelines for experimental animal studies of the Institutional Animal Care and Use Committee at the Medical University of South Carolina (Charleston, SC).

SUM149 and SUM159 human TNBC cell lines were developed by Dr. Stephen P. Ethier. We received these cell lines directly from Dr. Stephen P. Ethier's laboratory, and the cells were maintained as previously described (33, 34). The MDA-MB-231 and Hs578T human breast cancer cell line was purchased from the ATCC. The cells were cultured in DMEM medium containing 10% FBS, 2 l-glutamine, and 100 microgram/mL of penicillin-streptomycin (Invitrogen). Cell authentication was performed by short tandem repeat assays. DMEM, DMEM/F12 medium, recombinant human basic fibroblast growth factor (bFGF), and B27 supplement were obtained from Invitrogen.

Mammosphere formation assays were performed to determine the sphere-forming activity of CSCs as previously described (35–37). Briefly, single-cell suspensions prepared from human TNBC cell lines or TNBC xenograft tumors were cultured at 2,000 to 5,000 cells/mL per well in 24-well ultralow attachment plates (Corning) using serum-free DMEM/F-12 medium supplemented with 20 ng/mL basic FGF, 20 ng/mL EGF, 4 μg/mL insulin, 4 μg/mL heparin, 0.5 μg/mL hydrocortisone, and 0.4% BSA and B27 (Invitrogen). Culture medium was replaced every other day with 50% fresh medium. Tumor spheres were counted and photographed after 7 days of culture.

To knockdown MYC expression, human TNBC cells were transfected with MYC-specific siRNAs (Qiagen) using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's protocol. AllStars Negative Control siRNAs (Qiagen) were used as controls. At 48 hours after transfection, MYC expression levels in transfected cells were assessed by Western blot analyses.

In situ staining of senescence-associated β-galactosidase (SA-β-gal) was performed using a senescence β-galactosidase staining kit (Cell Signaling Technology) to determine senescent cells as we previously reported (38, 39).

Western blot analyses were performed as previously described (39). Briefly, protein samples were extracted using cell lysis buffer (Cell Signaling Technology) supplemented with a cocktail of proteinase inhibitors (Sigma-Aldrich). The protein concentrations were quantified using a DC Protein Assay Kit (Bio-Rad Laboratories). Fifty microgram protein samples were resolved on 4% to 20% Mini-PROTEAN TGX gels (Bio-Rad Laboratories) and transferred onto 0.2 μm PVDF membrane (Millipore). Blots were blocked with 5% nonfat milk for 1 to 2 hours at room temperature, and then probed with primary antibodies and incubated at 4°C overnight. After extensive washing with TBS-T, blots were incubated with appropriate horseradish peroxidase–conjugated secondary antibody for 1.5 hours at room temperature. Protein bands were detected using an ECL Plus Western Blotting Detection System (GE Healthcare Life Sciences).

SUM159 cells were treated with C1572 (0.2 μmol/L) or DMSO as vehicle control for 2 hours. Cell lysates were prepared using RIPA lysis buffer (Santa Cruz Biotechnology). For MYC immunoprecipitation (IP), cell lysates were pre-cleared using 5 μL normal rabbit IgG (sc-2027, Santa Cruz Biotechnology), together with 20 μL protein A/G PLUS-Agarose (Santa Cruz Biotechnology). The pre-cleared lysates were incubated with 5 μL rabbit anti-human c-Myc monoclonal antibodies (Cell Signaling Technology) for 1 hour and then with 20 μL Protein A/G PLUS-Agarose overnight at 4°C. Immunoprecipitates were collected by centrifugation and washed with RIPA buffer three times. After the final wash, IP products were resuspended in 50 μL SDS loading buffer and resolved on 4% to 20% Mini-PROTEAN TGX gels (Bio-Rad) and transferred onto PVDF membrane (Millipore). Western blots were performed using mouse anti-ubiquitin specific antibodies (Cell Signaling Technology).

TNBC orthotopic xenografts were established as previously described (8). Briefly, 2 million SUM159 or MDA-MB-231 cells were mixed with Matrigel (1:1) and transplanted into the fourth mammary fat pad of NOD-SCID-IL₂R gamma^null (NSG) mice (The Jackson Laboratory). Four weeks after tumor cell implantation, when tumor volume reached approximately 100 mm³, mice were randomized into three groups and treated with C1572 (0.3 or 0.6 mg/kg) or vehicle control, respectively, via intraperitoneal injection three times per week for 3 weeks. Tumor volume was measured weekly using calipers and calculated using the following formula: tumor volume = length × width²/2. At the end of drug treatment study, animals were euthanized and tumors were surgically removed, weighted, and subjected to tissue section preparations and histopathologic analyses.

Cancer-initiating cells (CIC) in residual tumor cells after drug treatment were assessed using limited dilution assay (LDA) as previously described (37). Briefly, different doses/dilutions (100,000, 10,000, 1,000, and 100 cells) of SUM159 cells were mixed with Matrigel (1:1) and implanted into the fourth mammary fat pad of NSG mice. The formation of tumors after tumor cell transplantation was examined twice per week for 10 weeks. The frequency of CICs was calculated using L-Calc software (Stemcell Technologies).

Comparisons between groups were carried out using the Student t test. Differences were considered statistically significant at P < 0.05. The error bars indicate SEM. All analyses were carried out with the GraphPad Prism program (GraphPad Software).

The mammosphere formation assay (MFA) is a useful technique to measure the sphere-forming ability of CSCs, and the sphere-initiating cell (SIC) is a surrogate of CSCs in culture (37). To develop novel therapeutic agents that target CSCs, we performed an unbiased screen of 165 small-molecule compounds using MFA. Through the screen, we identified 18 compounds that inhibit the formation of mammospheres by 50% or more (Fig. 1A). Among the 18 hits in the screen, C1572 was the top hit and completely depleted SICs in culture (Fig. 1A and C). To confirm the effect of this compound on CSCs, we investigated the response of sorted ESA⁺/CD44⁺/CD24⁻ cells to C1572 treatment. The result showed that at nanomolar concentrations, C1572 significantly inhibited the formation of mammospheres by ESA⁺/CD44⁺/CD24⁻ cells in a dose-dependent manner (Fig. 1D; Supplementary Fig. S1).

To determine whether a transient treatment with C1572 is sufficient to inhibit CSCs in TNBC, we incubated human TNBC cells with C1572 overnight (∼16 hours) and then removed the drug and performed MFA to detect CSCs in residual tumor cells. As shown in Fig. 1E and F, even a short period of incubation with C1572 dramatically reduces the number of SICs in both SUM159 and Hs578T residual tumor cells. Similar results were observed in MDA-MB-231 cells (Supplementary Fig. S2A). Notably, C1572 shows no significant effects on human hematopoietic stem and progenitor cells (HSPC) at doses lethal for CSCs and exhibits only modest inhibitory activity at higher doses (Supplementary Fig. S2B). Taken together, these results suggest that treatment with C1572 selectively depletes TNBC CSCs while sparing normal tissue stem cells such as HSPCs.

Next, we investigated the mechanisms whereby C1572 depletes TNBC CSCs. Previous studies have shown that MYC, STAT3, and β-catenin are involved in regulating CSC survival and self-renewal functions (40–42). We determined whether C1572 treatment affected the expression of these proteins in TNBC cells. Surprisingly, C1572 dramatically decreased MYC protein levels in SUM159 cells in a dose-dependent manner but had no significant effects on STAT3, β-catenin, and NF-kB p65 expression (Fig. 2A). Similar results were observed in MDA-MB-231 and Hs578T cells (Supplementary Fig. S3A–S3C). The time course studies indicated that incubation with C1572 led to a reduction in MYC protein levels as early as 2 hours, reaching a maximum effect at 4 to 6 hours after treatment (Fig. 2B). It is worth noting that MYC protein levels gradually increased and were almost back to normal levels 4 hours after drug withdrawal, suggesting that the inhibition effect of C1572 is likely reversible (Supplementary Fig. S3D). However, MYC protein levels decreased again at 24 hours after drug removal (Supplementary Fig. S3D). These results suggest that C1572 may induce both acute and long-term MYC inhibition.

To gain insight into the mechanisms by which C1572 inhibits MYC in TNBC cells, we measured MYC mRNA expression levels. The real-time RT-PCR data indicate that C1572 has no significant impact on MYC mRNA levels at doses that deplete MYC protein (Fig. 2C). In contrast, this compound markedly shortened the half-life of MYC protein, suggesting that C1572 may stimulate protein degradation (Fig. 2D; Supplementary Fig. S3E). In agreement with this, we showed that C1572 increased the levels of both phosphorylated and ubiquitinated MYC (Fig. 2E and F). Furthermore, we found that co-treatment with a proteasome inhibitor, MG-132, could rescue the C1572-induced decline in MYC protein (Fig. 2G). These data support the hypothesis that C1572 inhibits MYC via stimulating its posttranslational modifications and subsequent degradation.

JQ1 is a well-characterized and commercially available MYC small-molecule inhibitor. It was shown that JQ1 suppresses MYC transcription via interacting with one of the BET proteins, BRD4 (43). Notably, our studies demonstrate that C1572 is 100-fold more potent than JQ1 in inhibiting MYC (Supplementary Fig. S4). More importantly, the observation that C1572 selectively reduces MYC protein levels, but not the expression of several other nuclear proteins including STAT3, β-catenin, p65, and Bmi-1, suggests that this compound has an unusual degree of specificity for MYC (Fig. 2A and B; Supplementary Fig. S3B–S3C).

To determine whether the observed downregulation of MYC by C1572 was due to cell toxicity or cell-cycle arrest, we investigated the impact of several known cytotoxic and chemotherapeutic agents on MYC expression. We found that C1572 was the only compound that completely depleted MYC protein in human TNBC cells, whereas other agents, including paclitaxel, etoposide, SN-38, cisplatin, topotecan, curcumin, and AZD8055, did not cause any significant effects on MYC expression (Supplementary Fig. S5). Of note, treatment with C1572 had little effect on tumor cell viability, whereas paclitaxel (Taxol) treatment markedly induced apoptosis (Fig. 2H and J; Supplementary Fig. S6), suggesting that the MYC inhibition activity of C1572 is not due to cell death.

To test the hypothesis that C1572 depletes CSCs via targeting MYC, we performed a loss-of-function analysis of the MYC oncogene to determine whether silencing of MYC expression can phenocopy the effect of C1572 in targeting TNBC CSCs. As shown in Fig. 3A, the silencing of MYC by siRNA was confirmed by Western blot analyses. Interestingly, we found that silencing of MYC led to only a modest reduction in the number of viable TNBC cells at 48 hours after siRNA transfection (Fig. 3B). In contrast, MFAs revealed that knockdown of MYC expression markedly inhibited SICs (a surrogate of CSCs) in SUM159 cells (Fig. 3C and D). Similar results were observed in MDA-MB-231 cells (Fig. 3E). These results demonstrate that silencing of MYC expression phenocopies the CSC-depletion effect of C1572. Together with the data shown in Figs. 1 and 2, these new findings strongly support the hypothesis that C1572 eradicates TNBC CSCs via targeting the MYC oncoprotein.

Although silencing of MYC expression selectively depletes CSCs, it has little effect on the survival of TNBC cells (Fig. 3). These results support the concept that induction of apoptotic cell death is not the primary mechanism underlying MYC knockdown-induced depletion of CSCs. Given the fact that MYC plays a critical role in overriding oncogene RAS-induced senescence (44), we decided to investigate whether silencing of MYC induces senescence in human TNBC cells. SA-β-gal assays show that knockdown of MYC expression markedly increases the number of SA-β-gal–positive senescent cells (Fig. 4A and B), suggesting that knockdown of MYC induces senescence in TNBC cells. Notably, we found that treatment with C1572 can phenocopy MYC knockdown-induced senescence in SUM159 cells (Fig. 4C and D). Similar results were observed in MDA-MB-231 cells (Fig. 4E). Furthermore, the senescence phenotype was sustained even when C1572 was removed from the culture medium after a short period (14 hours) of incubation (Fig. 4F). Taken together, these new findings demonstrate for the first time that knockdown of MYC expression induces senescence in TNBC cells and suggest that targeted inhibition of MYC may represent a novel therapeutic strategy to eradicate CSCs via the induction of stem cell senescence.

Having shown that inhibition of MYC by C1572 depletes TNBC CSCs in vitro (Figs. 1 and 2), we next sought to confirm the efficacy of C1572 in eradicating CSCs in culture using CSC transplantation analysis, also known as limited dilution assay (LDA). The number of CICs in residual TNBC cells after C1572 treatment was determined by LDAs as previously reported (37). After overnight (16 hours) treatment with C1572 there were no significant changes in the number of viable SUM159 cells (Fig. 2,H and 2J; Supplementary Fig. S6). However, these viable cells exhibited a significant defect in their ability to form tumors when transplanted into the fat pads of NOD/SCID IL₂R gamma^null (NSG) mice (Fig. 5A and B). More importantly, the results of LDAs demonstrate that the frequency of CICs is 0.033% versus 0.92% in SUM159 cells treated with either C1572 or DMSO, respectively (Fig. 5B and C). These results indicate that the frequency of CICs in cells treated with C1572 is 28-fold lower than that in SUM159 cells treated with DMSO as a vehicle control (Fig. 5C). In addition, we examined the weight of tumors generated from the two groups of cells. The data revealed that C1572-treated TNBC cells produced smaller tumors than cells treated with DMSO (Fig. 5A and D). To assess the number of CSCs in xenograft tumors, we performed MFAs using xenograft tumor-derived single-cell suspensions. The results showed that the number of SICs was markedly reduced in tumors derived from C1572-treated animals, suggesting that the number of CSCs was significantly lower in xenograft tumors derived from C1572-treated SUM159 cells than in those derived from cells treated with DMSO (Fig. 5E and F). Together, these findings demonstrate for the first time that even a transient inhibition of MYC by C1572 significantly depletes TNBC CSCs, suggesting that targeted inhibition of MYC is likely an effective therapeutic strategy to eradicate therapy-resistant CSCs for TNBC treatment.

To evaluate the ability of C1572 to suppress tumor growth and eradicate TNBC CSCs in tumor tissues in vivo, we established SUM159 orthotopic xenograft tumors in NSG mice and then treated tumor-bearing mice with a range of doses of C1572 or vehicle (DMSO). Tumor volume analyses indicated that C1572 suppressed the growth of TNBC xenograft tumors in a dose-dependent fashion (Fig. 6A), and the average tumor weight in C1572-treated mice was significantly lower than that of vehicle-treated mice (Fig. 6B and C). To assess the number of CSCs in residual xenograft tumors, xenograft-derived single cells were prepared and mammosphere formation assays were performed to determine the sphere-forming ability of CSCs in residual xenograft tumors after different treatments. We found that the number of SICs was markedly reduced in tumors from C1572-treated mice compared with those treated with vehicle control (Fig. 6D). Similar results were observed using an MDA-MB-231 orthotopic xenograft model (Fig. 6E and F). These findings confirmed the efficacy of C1572 in suppressing tumor growth in vivo and in eliminating CSCs in human TNBC xenografts. Of note, no significant changes in body weight were observed between C1572-treated and vehicle control mice, suggesting that C1572 is well tolerated in NSG mice (Supplementary Fig. S7).

To determine the causal correlation between C1572-mediated CSC depletion and MYC inhibition in tumor tissues in vivo, IHC analyses were performed to examine MYC expression levels in residual tumors after drug treatment. As shown in Fig. 6G, IHC showed that C1572-mediated CSC depletion correlated with decreased MYC expression in xenograft tumors, suggesting that C1572 suppresses tumor growth and depletes CSCs in tumor tissues via targeting the MYC oncoprotein.

In light of the observation that C1572 suppresses tumor growth in a dose-dependent fashion (Fig. 6A and C), we measured whether an increase in dosing frequency could improve the therapeutic outcome and whether therapy-induced tumor suppression would be sustained posttreatment. As shown in Fig. 6H, increased dosing frequency resulted in complete tumor regression. Equally importantly, C1572-induced tumor regression persists at least 3 weeks after the final dosing. Together, these results demonstrate for the first time the therapeutic potential of C1572 to induce human TNBC xenograft tumor regression and deplete drug-resistant CSCs in tumor tissues in a preclinical setting.

Therapy resistance and tumor recurrence is a major challenge limiting the success of breast cancer treatment, particularly for the TNBC subtype. Growing evidence indicates that CSCs are likely responsible for metastatic progression and tumor relapse via regeneration of new tumors (11). It has been shown that CSCs are selectively enriched in residual tumors following anticancer therapies (7, 8, 13–15), suggesting that the inability of traditional chemotherapy to kill CSCs may contribute to tumor regrowth and treatment failure. Thus, it is critical to develop new therapeutic agents that can effectively target drug-resistant CSCs and improve survival for breast cancer patients. In the present study, using unbiased drug screenings, we have identified C1572 as a potential therapeutic agent to target TNBC CSCs. Importantly, we show that treatment with C1572 suppresses TNBC xenograft growth in vivo in a dose-dependent fashion and markedly reduces the number of SICs in residual tumors. Thus, C1572 represents a new lead compound targeting drug-resistant CSCs in TNBC. Furthermore, it appears that there is no significant change in body weight between mice treated with C1572 and those treated with vehicle control, suggesting that this compound is well tolerated in mice and may have a favorable toxicity profile in possible clinical applications.

The MYC oncogene encodes for a transcription factor that is overexpressed in multiple human cancer types, including TNBC (16–21). Functional genomic studies have established a causal role for MYC in tumorigenesis and tumor maintenance, as evidenced by the fact that various tumors are addicted to MYC and that its inactivation leads to tumor regression in multiple preclinical tumor models (45–47). Important to the present study, MYC expression is selectively elevated in basal-like TNBC, elevating the activity of the MYC signaling pathway (20, 21, 48). Earlier efforts to target MYC with small molecules have been difficult, but our present study implies that this approach is a viable strategy to eradicate drug-resistant CSCs in TNBC. The present study provides a proof of principle that pharmacologic interference of MYC by C1572 is a new and effective therapeutic approach to target CSCs for TNBC treatment.

It is well documented that senescence induction is an important mechanism of tumor prevention in response to oncogene activation (47, 49). One possible mechanism underlying MYC's pro-tumor properties is likely through antagonizing oncogene activation-induced senescence (44, 47). We, and others, have shown that the induction of premature senescence is a crucial mechanism of action for some therapeutic agents and radiation to induce tumor suppression (38, 39, 50). New data presented here reveal that knockdown of MYC expression has only a modest impact on cell viability but markedly depletes CSCs in TNBC. Subsequent mechanistic studies indicate that MYC knockdown induces senescence in TNBC cells. These results suggest that inhibition of MYC may deplete CSCs via the induction of senescence in stem cell subpopulations. In agreement with this concept, it has been shown that inactivation of MYC can lead to tumor regression in vivo through the induction of senescence in hepatocellular carcinoma and lymphoma cells (47).

Although MYC has been considered an attractive target for cancer treatment (45–47), therapeutic agents that target MYC directly have been difficult to develop. Previous studies examining the therapeutic potential of MYC inactivation have been performed largely through genetic approaches (45–47). Recently, attempts have been made to block MYC's function indirectly, such as by inhibiting the key protein–protein interactions (e.g., BET bromodomain inhibition) involved in MYC transcription (43). As a result, JQ1 is available as a commercially available MYC small-molecule inhibitor. Surprisingly, our present study shows C1572 to be 100-fold more potent than JQ1 in inhibiting MYC. Moreover, based upon the fact it has no effects on several other transcription factors, including β-catenin, STAT3, p65, and Bmi-1, we believe that C1572 has fairly robust selectivity.

In summary, we demonstrate for the first time that C1572 selectively depletes drug-resistant CSCs in TNBC via targeting the MYC oncoprotein. Strikingly, C1572 is 100-fold more potent in inhibiting MYC than the commercially available MYC small-molecule inhibitor JQ1. Preclinical animal studies revealed that C1572 treatment depletes CSCs in TNBC xenograft tumors and markedly suppresses tumor growth in vivo. Furthermore, the CSC-depletion effects of C1572 correlate with a significant reduction of MYC protein levels in residual tumors. These new findings support the development of C1572 as a therapeutic agent to target CSCs for TNBC treatment, and also suggest that MYC may be exploited as a biomarker to evaluate the therapeutic response of this compound.

No potential conflicts of interest were disclosed.

Conception and design: A. Yang, K.D. Tew, G.Y. Wang

Development of methodology: A. Yang, G.Y. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Yang, S. Qin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Yang, S. Qin, S.P. Ethier, K.D. Tew, G.Y. Wang

Writing, review, and/or revision of the manuscript: A. Yang, B.A. Schulte, S.P. Ethier, K.D. Tew, G.Y. Wang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Yang, B.A. Schulte, S.P. Ethier

Study supervision: G.Y. Wang

This study was supported in part by NIH grants GM103542, HL106451, and UL1TR001450. This work was also partially supported by the Cell & Molecular Imaging Shared Resource of Hollings Cancer Center at the Medical University of South Carolina (P30 CA138313) and the Shared Instrumentation Grant S10 OD018113.

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.