Histone Deacetylase Inhibitors Inhibit Rhabdomyosarcoma by Reactive Oxygen Species–Dependent Targeting of Specificity Protein Transcription Factors

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma that is primarily observed in children and adolescents and accounts for 5% of all pediatric cancers and 50% of soft tissue sarcomas in children (1–3). Embryonal RMS (ERMS) and alveolar RMS (ARMS) are the two major classes of RMS in children and adolescents and differ with respect to their histology, genetics, treatment, and prognosis (4, 5). ERMS accounts for over 60% of RMS patients and is associated with loss of heterozygosity at the 11p15 locus and mutation in Ras signaling (6, 7), whereas ARMS occurs in approximately 20% of RMS patients and is associated with translocations resulting in formation of pro-oncogenic gene products resulting from the fusion of PAX3 or PAX7 with the Forkhead gene FOXO1A (8, 9). ARMS patients have a poor diagnosis, and patient survival is <20% for metastatic ARMS. Treatments include radiotherapy, surgery, and chemotherapy with cytotoxic drugs (10, 11); however, RMS patients that survive current therapies have a dramatic increased incidence of several health problems as adults (12). Thus, there is a critical need for development of new therapeutic regimens for treating childhood RMS.

Specificity protein 1 (Sp1) transcription factor is overexpressed in human RMS tumors, and other Sp family members, including Sp3 and Sp4, are also overexpressed in RMS cell lines (13). The importance of Sp transcription factors (TF) in RMS is primarily due to pro-oncogenic Sp-regulated genes that are themselves drug targets for RMS, and these include CXCR4, hepatocyte growth factor receptor (c-MET), insulin-like growth factor 1 receptor (IGF-1R), and platelet-derived growth-factor receptor α (PDGFRα; refs. 14–17). Clinical studies using drugs that specifically target Sp TFs and Sp-regulated genes for treatment of RMS have not yet been reported; however, there is an open phase I/II trial (NCT01610570) evaluating the efficacy of mithramcyin in solid tumors, including RMS. Mithramycin acts in part by binding to GC-rich sequences and regulating chromatin accessibility, including the ability to displace Sp1 from oncogenic promoters. Thus, the therapeutic potential of Sp TFs in RMS is gaining traction.

Genomic analysis of RMS from several patients indicated that “skeletal muscle (rhabdomyosarcoma) may have even higher levels of ROS than other cancer cells and may be particularly sensitive to therapeutics that induce oxidative stress” (18). This sensitivity is thought to occur because with such a high baseline burden of ROS, there is little tolerance for further oxidative stress and this was confirmed by showing that ROS inducers were highly effective inhibitors of RMS tumor growth using patient-derived xenografts in mouse models (18). Recent studies in our laboratory (19) demonstrate that ROS inducers also inhibit pancreatic cancer cell and tumor growth, and this is due, in part, to a novel epigenetic pathway (20) in which ROS-mediated repression of cMyc results in downregulation of Sp TFs and pro-oncogenic Sp-regulated genes. In this study, we demonstrate that ROS-inducing histone deacetylase (HDAC) inhibitors block RMS cell and tumor growth by initially targeting cMyc, which results in downregulation of miRNAs and induction of ZBTB transcriptional repressors, which, in turn, downregulate Sp TFs.

RD, Rh30, and SMS-CTR rhabdomyosarcoma cell lines were purchased from the American Type Culture Collection, and cells were maintained as previously described (13, 19). Cells were authenticated in 2014 (Promega Powerplex 18D) at the Duke University DNA Analysis Laboratory. Various reagents (including antibodies) are summarized in Supplementary Materials and Methods.

Proliferation of RD and Rh30 rhabdomyosarcoma cells (1.0 × 10⁵ per well) in the presence or absence of transfected siRNAs and after treatment with panobinostat and vorinostat (dimethyl sulfoxide, DMSO, as empty vehicle; ±GSH, 3 hours prior to treatment) was essentially carried out as previously described (13, 19). Primary human myoblasts (HSMM; Lonza), Rh30, or RD cells were plated in 96-well plates at a density of 10,000 cells per well. The next day, cells were treated with vehicle (DMSO) or increasing doses of panobinostat. Twenty-four hours after treatment, cells were analyzed by the MTT assay.

RD and Rh30 rhabdomyosarcoma cells (1.0 × 10⁵ per well) were seeded in 2-well Nunc Lab-Tek chambered B#1.0 Borosilicate coverglass slides from Thermo Scientific and were allowed to attach for 24 hours. After 24 or 72 hours (after Sp1 knockdown), Annexin V staining was determined as described (13, 19).

RD and Rh30 rhabdomyosarcoma cells (3.0 × 10⁵ cells per well) were seeded in DMEM/Ham's F-12 medium supplemented with 2.5% charcoal-stripped FBS and were allowed to attach for 24 hours. Cells were seeded and subsequently treated with varying concentrations of panobinostat or vorinostat for 24 hours (±GSH, 3 hours prior to treatment) or with 100 nm of siSp1, siSp3, siSp4 for 48 hours, and cells that migrated through the pores were then counted as described (19).

miRNA was isolated using the mirVana miRNA isolation Kit (Ambion) according to the manufacturer's protocol. Quantification of miRNA (RNU6B and miR-17, -20a, and -27a) was done using the TaqMan miRNA assay Kit (Life Technologies) according to the manufacturer's protocol with real-time PCR. U6 small nuclear RNA was used as a control to determine relative miRNA expression.

The chromatin immunoprecipitation (ChIP) assay was performed using the ChIP-IT Express Magnetic Chromatin Immunoprecipitation Kit (Active Motif) according to the manufacturer's protocol. RMS cells (5 × 10⁶ cells) were treated with panobinostat for 3 hours, and after crosslinking, lysing, and immunoprecipitation, the interactions with the cMyc and Sp1 promoters were determined as described (19).

RD and Rh30 rhabdomyosarcoma cells (3.0 × 10⁵ per well) were seeded in DMEM/Ham's F-12 medium supplemented with 2.5% charcoal-stripped FBS and were allowed to attach for 24 hours, and after various treatments, whole cell lysates were analyzed by Western blots essentially as described (13, 19).

RD and Rh30 rhabdomyosarcoma cells were seeded (1.2 × 10⁵ cells per well) in 6-well plates in DMEM/Ham's F-12 medium supplemented with 2.5% charcoal-stripped FBS and left to attach for 24 hours. Knockdown of NR4A1 was carried out using Lipofectamine 2000 reagent according to the manufactures protocol. Cellular ROS levels were ascertained using the cell-permeable probe CM-H₂DCFDA (5-(and-6)-chloromethyl-2′7′-dichlorodihydrofluorescein diacetate acetyl ester) from Invitrogen (19). RNAi studies were also carried out essentially as described (19).

Male SCID/beige mice were subcutaneously implanted with 5 × 10⁶ RD cells suspended in Matrigel (BD Biosciences). At 12 days after implantation, when tumors were palpable, mice were begun on a treatment regimen of 17.5 mg/kg panobinostat or vehicle (DMSO), dosed intraperitoneally. Mice were treated daily for 4 days, no treatment for 2 days, and then every other day for 10 days. Animals were examined every other day for tumor burden (approximated by external caliper measurements, where [(width)2 × length]/2), animal weight, and overall well-being. At study end, animals were humanely sacrificed and tumors were harvested for analysis.

Statistical significance of differences between the treatment groups was determined by the Student t test. The results are expressed as mean with error bars representing 95% confidence intervals for 3 experiments for each group unless otherwise indicated, and a P value less than 0.05 was considered statistically significant. All statistical tests were two-sided.

Figure 1A confirms that both panobinostat and vorinostat induced ROS in RD and Rh30 cells as previously reported in patient-derived RMS xenografts (18), and these responses were attenuated in cells cotreated with the HDAC inhibitors plus the antioxidant glutathione (GSH). Treatment of RD and Rh30 cells with different concentrations of both HDAC inhibitors also decreased proliferation of RD and Rh30 cell lines (Fig. 1B). For the more potent HDAC inhibitor (panobinostat), we also observed inhibition of metabolic activity in RD and Rh30 cells using the MTT assay, whereas in primary human skeletal muscle myoblasts (HSMM), significant inhibition was not observed at concentrations as high as 500 nmol/L, demonstrating specificity of panobinostat for the transformed cell lines (Fig. 1C). Finally, we also observed inhibition of RD and Rh30 cell growth by panobinostat, and vorinostat was significantly attenuated after cotreatment with GSH (Fig. 1D), indicating that induction of ROS by the HDAC inhibitors was important for their growth-inhibitory effects. However, the growth-inhibitory effects of panobinostat in Rh30 cells were also ROS-independent.

Panobinostat and vorinostat also induced Annexin V staining, a marker of apoptosis in RD and Rh30 cells (Fig. 2A), and cotreatment with GSH attenuated this response. Using a similar treatment protocol, we also show that both HDAC inhibitors induced cleavage of PARP and caspase 3 (markers of apoptosis) in RD and Rh30 cells (Supplementary Fig. S1A and S1B), and these effects were attenuated after cotreatment with GSH. Panobinostat and vorinostat inhibited invasion in RD and Rh30 cells in a Boyden chamber assay (Fig. 2B), and this response was also attenuated in cells cotreated with GSH. Thus, ROS induction by both HDAC inhibitors resulted in the induction of apoptosis and inhibition of cell growth and invasion in RD and Rh30 cells. Hydrogen peroxide and other ROS-inducing anticancer agents decrease expression of Sp1, Sp3, and Sp4 transcription factors in pancreatic, colon, and bladder cancer cells, and Sp1, Sp3, and Sp4 are also highly expressed in RMS and tumors (19, 21–27). Panobinostat also decreased expression of Sp1, Sp3 (high and low molecular weight forms), and Sp4 in RD and Rh30 cells, and cotreatment with GSH attenuated this response (Fig. 2C). We also observed similar effects in RD and Rh30 cells treated with vorinostat alone or in combination with GSH (Fig. 2D), and both HDAC inhibitors increased histone-3 acetylation in RD and Rh30 cells which was unaffected by cotreatment with GSH. This suggests that the effects of panobinostat and vorinostat in this system are due to induction of ROS and are independent of their activity as inhibitors of histone deacetylation. Supplementary Fig. S1C also shows that both HDAC inhibitors downregulate Sp1, Sp3, and Sp4 expression in a third RMS cell line, SMS-CTR cells.

Figure 3A and B shows that 75 to 150 μmol/L hydrogen peroxide and 100 to 200 μmol/L t-butylhydroperoxide, respectively, inhibited growth of RD and Rh30 cells and also decreased expression of Sp1, Sp3, and Sp4 proteins and cMyc (Fig. 3C and D). These results were consistent with the effects observed for panobinostat in RMS cells (Fig. 2C and D). Supplementary Fig. S2 shows that after knockdown of Sp1, Sp3, and Sp4 by RNA interference, there is inhibition of Rh30 and RD cell growth (Supplementary Fig. S2A), induction of Annexin V staining (Supplementary Fig. S2B), and decreased invasion in a Boyden chamber assay (Supplementary Fig. S2C), demonstrating the important role of Sp TFs in the growth, survival, and invasion of RMS cells, and this was consistent with previous studies on Sp TFs in RMS cells (13).

Results illustrated in Fig. 4A show that within 3 hours after treatment of RD and Rh30 cells with panobinostat, there was a significant decrease in expression of cMyc, and after 6 hours, Myc protein levels were not detected for the 24-hour duration of the experiment. Interestingly, we also observed a rapid decrease in expression of Sp1 within 3 to 6 hours, whereas decreased expression of Sp3 and Sp4 proteins was observed only at longer time points. This suggests that like cMyc, Sp1 expression may also be reduced by rapid ROS-dependent chromatin shifts that have previously been observed in colon and pancreatic cancer cells after treatment with hydrogen peroxide and phenethylisothiocyanate (PEITC; an ROS inducer), respectively (19, 20). In addition, cMyc downregulation may be due, in part, to loss of Sp1 which also regulates cMyc as illustrated in Fig. 4B, which shows that Sp1 knockdown decreases cMyc expression in RD and Rh30 cells (Fig. 2B). In contrast, the rate of degradation of Sp-regulated gene products, including EGFR, bcl2, and survivin (Fig. 4C), was similar to that observed for Sp3 and Sp4 proteins and different from either cMyc or Sp1. ROS-induced changes in histone marks in RD and Rh30 cells were investigated in a ChIP assay after treatment with panobinostat for 3 and 6 hours. In RD cells (Fig. 4D), there was a decrease in Pol II and the H3K4me3 activation mark on the cMyc and Sp1 promoters and an increase in H4K16Ac. In Rh30 cells (Fig. 4E), there was also a decrease in Pol II and H3K4me3 and an increase in H4K16Ac on the cMyc and Sp1 promoters as observed in RD cells; however, we also observed an increase in the H3K27me3 deactivation mark on the cMyc promoter. These results are consistent with previous studies on ROS-induced changes in histone marks (20) and with the exception of the increase in H4K16Ac (activation mark), the epigenetic changes correlate with the observed rapid decreases in cMyc and Sp1 protein expression (Fig. 4A).

Previous studies in pancreatic cancers cells show that ROS-dependent downregulation of Myc also results in decreased expression of miR-27a and miR-20a/miR-17 and upregulation of the corresponding miR-regulated Sp repressors ZBTB34/ZBTB10 and ZBTB4, respectively (19). Results in Fig. 5A show that after treatment of RD and Rh30 cells with panobinostat, there was a decrease in cMyc expression and this was attenuated in cells cotreated with panobinostat plus GSH. Previous studies showed that miR-27a is expressed in RD and Rh30 cells (13), and treatment of these cell lines with 100 nmol/L panobinostat decreased expression of miR-27a, and this decrease was also attenuated in cells cotreated with panobinostat plus GSH (Fig. 5B). We also observed that miR-20a and miR-17 were expressed in both RMS cell lines, and downregulation of these miRNAs with panobinostat was also inhibited in cells cotreated with the HDAC inhibitor plus GSH. MiR-27a and miR20a/miR-17 are members of the miR-23a-27a-24-2 and miR-17-92 clusters, respectively, and there is evidence that both miR clusters are regulated by cMyc (19, 28–30), and this was confirmed by transfection with siMyc, which decreased miR-27a and miR-20a/miR17 expression in RMS cells (Fig. 5C). There is also some evidence that Sp1 regulates these same miR clusters, and knockdown of Sp1, Sp3, and Sp4 by RNAi decreased miR-27a and miR-20/miR-17 expression, and the most dramatic effects were observed for miR-27a in both RD and Rh30 cells (Fig. 5B). These results could also be due in part to downregulation of cMyc which can be regulated by Sp TFs (31).

Treatment of RD and Rh30 cells with panobinostat resulted in the induction of ZBTB10 and ZBTB34 (Fig. 6A), which are regulated by miR-27a, and ZBTB4 which is regulated by miR-20a/miR-17 (19, 22–26). In addition, transfection of RD and Rh30 cells with siMyc increased expression of ZBTB34, ZBTB10, and ZBTB4 and also decreased levels of Sp1, Sp3, and Sp4 and antagomir-27a also induced ZBTB10 and decreased expression of Sp proteins (Fig. 6B). Thus, cMyc and miR-27a knockdown decreases expression of miRNAs, induces ZBTBs, and decreases Sp TFs and thereby mimics the effects of ROS. In vivo studies with SCID mice bearing RD cell xenografts also showed that administration of panobinostat (17.5 mg/kg) decreased tumor volume and weight (Fig. 6C), and analysis of the tumors showed that panobinostat also decreased Sp1, Sp3, Sp4, and cMyc expression (Supplementary Fig. S3). These results are consistent with in vitro studies and demonstrate that the effectiveness of panobinostat as an anticancer agent in RMS cells is primarily due to targeting of Sp1, Sp3, and Sp4 transcription factors through an ROS-dependent mechanism involving downregulation of cMyc and Myc-regulated miRNAs (Fig. 6D).

Transformation of normal cells into cancer cells is a complex cell- and tissue-specific process that involves activation of oncogenes and inactivation of tumor suppressor genes (32, 33); however, many other genes also significantly contribute to the cancer cell phenotype. These genes are typically overexpressed in tumor versus nontumor tissues and have been termed as nononcogene addiction (NOA) genes and are excellent targets for mechanism-based antineoplastic agents (33). The transcription factor Sp1 plays a critical role in embryonic development, but there is evidence that levels of Sp1 decrease with age in rodents and humans (34–36). In contrast, Sp1 is highly expressed in tumors from RMS patients and in many other cancers, and high expression of Sp1 is a negative prognostic factor for patient survival, tumor recurrence, or tumor grade (13, 37–42). Stable transduction of human skeletal muscle myoblasts with PAX3-FOXO1, telomerase, and N-Myc resulted in formation of transformed cell lines similar to ARMS cells (43), and both the genetically transformed and ARMS cells expressed high levels of Sp1, Sp3, and Sp4 (13). Interestingly, transformation of the muscle myoblasts dramatically increased expression of Sp1 and Sp3 but not Sp4 proteins, which were highly expressed in the nontransformed cells (13); carcinogen/oncogene-induced transformation of human fibroblasts also dramatically increased Sp1 expression, but levels of Sp3 and Sp4 were not determined (44).

The pro-oncogenic functions of Sp1 have been reported in many different cell lines (reviewed in 42); however, it is also important to determine the role of Sp1, Sp3, and Sp4 because all three proteins are overexpressed in RMS and other cancer cell lines. Supplementary Fig. S2 shows that after knockdown of Sp1, Sp3, and Sp4 in Rh30 and RD cells, there was a significant decrease in cell proliferation, induction of apoptosis (Annexin V staining), and inhibition of invasion, indicating that all three Sp proteins are NOA genes in RMS cells. In previous studies, we demonstrated that tolfenamic acid–induced downregulation of Sp1, Sp3, and Sp4 inhibited growth, induced apoptosis, and inhibited invasion in RMS cells, and this was accompanied by decreased expression of several pro-oncogenic Sp-regulated genes (13).

Based on a recent report that ERMS patient-derived xenografts were highly sensitive to drugs that induce oxidative stress (18), we hypothesized that the efficacy of the ROS-inducing agents, such as panobinostat, was also due, in part, to repression of Sp1, Sp3, and Sp4. This hypothesis was based on previous studies showing that several ROS-inducing anticancer drugs, including curcumin, phenethylisothiocyanate (PEITC), methyl 2-cyano-3,12-dioxooleana-1,9-dien-28-oate (CDDO-Me), GT-094 (a nitro-aspirin analog), celastrol, and betulinic acid, also repressed Sp1, Sp3, and Sp4 expression (ROS-dependent) in pancreatic, bladder, and colon cancer cells (19, 21–26).

HDAC inhibitors typically enhance histone acetylation, and this response contributes to the anticancer activities of this class of compounds; however, these inhibitors also modulate several pathways in cancer cells and this includes induction of ROS. Therefore, we used the HDAC inhibitors panobinostat and vorinostat that are also ROS inducers (45) to investigate downregulation of Sp1, Sp3, and Sp4 in RMS cells in culture and in an in vivo model. Both panobinostat and vorinostat induced ROS, inhibited growth, induced apoptosis, and inhibited invasion of RD and Rh30 cells, and all of these responses were attenuated after cotreatment with the antioxidant GSH (Figs. 1 and 2; Supplementary Fig. S1). Moreover, panobinostat also inhibited tumor growth in SCID mice bearing RD cells as a xenograft (Fig. 6B). In parallel studies, it was observed that panobinostat and vorinostat also decreased expression of Sp1, Sp3, and Sp4 in Rh30 and RD cells (Fig. 2C and D), and this response was attenuated after cotreatment with GSH. Moreover, treatment with hydrogen peroxide and t-butylhydroperoxide also decreased expression of Sp1, Sp3, and Sp4 as previously observed in other cell lines (23, 27, 46). We also observed that panobinostat decreased cMyc, Sp1, Sp3, and Sp4 proteins levels in tumors (Fig. 6B) and this paralleled the effects observed in cell culture (Fig. 2C). Thus, both knockdown of Sp TFs by RNAi (Supplementary Fig. S1) and treatment with ROS inducers result in decreased expression of Sp1, Sp3, and Sp4, decreased growth and invasion, and induction of apoptosis in vitro and tumor growth in vivo. This suggests that the ROS-mediated repression of Sp TFs plays an important role in the antineoplastic effects of these HDAC inhibitors in RMS cells and is consistent with the in vitro and in vivo efficacy of panobinostat in RMS patient-derived xenografts (18).

Previous studies show that ROS-dependent repression of Sp1, Sp3, and Sp4 was due to downregulation of miR-27a and/or miR-20a/miR-17-5, resulting in increased expression of the transcriptional repressors ZBTB10/ZBTB34 and ZBTB4, respectively (19, 21–26). The ZBTB proteins are members of the POK family of transcriptional repressors (47) and competitively bind and displace Sp TFs from GC-rich sites on Sp promoters and Sp-regulated gene promoters. Using panobinostat as a model ROS inducer, we show that this compound also decreases miR-27a and miR-20a/miR-17 in RD and Rh30 cells (ROS-dependent; Fig. 5B), and this results in the induction of ZBTB10/ZBTB34 and ZBTB4 (Fig. 6A). These data also show that the high expression of Sp1, Sp3, and Sp4 in RMS cells is due, in part, to miR-dependent suppression of the ZBTB transcriptional repressors.

Treatment of colon cancer cells with hydrogen peroxide or pharmacologic doses of ascorbate, which induces hydrogen peroxide, decreases expression of Sp1, Sp3, and Sp4 in colon cancer cells (46), and hydrogen peroxide also induces genome-wide migration of chromatin-modifying complexes from non–GC-rich to GC-rich gene promoters and downregulates genes such as cMyc (20). Sp1, Sp3, Sp4, and cMyc all have GC-rich promoters, and treatment of RD and Rh30 cells with panobinostat rapidly decreased expression of cMyc and Sp1 proteins, whereas with the exception of the rapid decrease in expression of Sp3 (high MW band in RD cells), downregulation of Sp3 and Sp4 was primarily observed at later time points (Fig. 4A). Results of ChIP assays confirmed that treatment with panobinostat decreased pol II and the H3K4me3 activation mark in the GC-rich cMyc and Sp1 promoters in RD and Rh30 cells, and the H3K27me3 inactivation mark was also increased on the cMyc promoter only in Rh30 cells (Fig. 4D and E). Surprisingly, the H4K16Ac activation mark was increased by panobinostat in RD and Rh30 cells but decreased by PEITC-induced ROS in pancreatic cancer cells (19), suggesting that the epigenetic effects of ROS inducers are cell context-dependent and this is currently being investigated.

The importance of the ROS-dependent decrease in cMyc is that there are E-box and GC-rich elements in the miR-23a∼27a∼24-2 and mir-17-92 gene cluster promoters (28–30), and knockdown of cMyc or Sp TFs in RMS cells significantly decreased expression of miR-27a and miR-20a/miR-27 (Fig. 5C and D). It was also observed for the first time that Sp1, Sp3, and Sp4 play a role in expression of miR-27a, and this paralleled the highly effective downregulation of miR-27a in RMS cells treated with HDAC inhibitors (Fig. 5B).

In summary, results of this study show that HDAC inhibitors that induce ROS are highly effective inhibitors of RMS cell in culture and xenograft tumor growth in mice, and the effects in vitro were independent of HDAC inhibition. This complements results obtained with HDAC inhibitors using tumor-derived xenografts and lineage of origin studies (18, 48), thus confirming the potential efficacy of ROS inducers for RMS chemotherapy (18). This study demonstrates that panobinostat and vorinostat decrease expression of Sp1, Sp3, Sp4, and pro-oncogenic Sp-regulated genes by ROS-dependent epigenetic downregulation of cMyc and Sp1, which, in turn, decrease expression of miR-27a and miR-20a/miR-17, resulting in the induction of the ZBTB transcriptional repressors (Fig. 6D). The study demonstrates why ROS inducers are highly effective for treating RMS because this pathway leads to a cascade of events resulting in downregulation of Sp TFs and pro-oncogenic Sp-regulated genes. ROS inducers and other drugs targeting Sp TFs represent promising new approaches for RMS chemotherapy; moreover, because miRNAs that repress ZBTB repressors have been detected in serum (49, 50), this may also provide a biomarker for monitoring treatment efficacy.

No potential conflicts of interest were disclosed.

Conception and design: E. Hedrick, C.M. Linardic, S. Safe

Development of methodology: E. Hedrick, C.M. Linardic, S. Safe

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E. Hedrick, L. Crose, C.M. Linardic, S. Safe

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Hedrick, C.M. Linardic

Writing, review, and/or revision of the manuscript: E. Hedrick, C.M. Linardic, S. Safe

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Hedrick, S. Safe

Study supervision: S. Safe

This study was funded by the NIH [P30-ES023512 (to S. Safe) and R01-CA122706 (to C.M. Linardic)], the Sid Kyle Endowment (S. Safe), and the Hartwell Foundation (L. Crose).

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.