TMEM16A Induces MAPK and Contributes Directly to Tumorigenesis and Cancer Progression

Amplification of the chromosomal band 11q13 is frequently seen in malignancies arising from breast, bladder, head and neck (SCCHN), and esophagus (1). Detailed molecular analysis of the 11q13 region led to identification of TMEM16A (also known as TAOS2, DOG1, and ANO1; refs. 2–4). This gene belongs to the TMEM16 family, characterized by the presence of 8 transmembrane domains and a highly conserved domain of unknown function (DUF590).

Recently, TMEM16A was shown to be a calcium-activated chloride channel (CaCC; refs. 5–7). The biologic importance of Tmem16a is underscored by the lethal phenotype in the knockout mouse because of abnormal tracheal development (8–10). TMEM16A contributes to many important physiologic functions, including the control of epithelial fluid transport, saliva production, and gastrointestinal tract motility (10–14). The third extracellular loop of TMEM16A (between transmembrane regions 5 and 6) has been suggested to be the putative pore-forming domain (6). Using mutagenesis approaches, the lysine residue at amino acid 610 (in the putative pore region) was found to be critical for its channel function (15). Specifically, mutation of this residue from lysine to alanine (TMEM16A-K610A) was shown to be hypomorphic, with greatly reduced chloride conductance.

CaCCs can be proapoptotic and suppress tumor formation in breast epithelia (16). However, there are exceptions, the putative CaCC, bestrophin (BEST1) is upregulated in colon cancer cells, and can promote proliferation (17). In addition, several reports have shown that TMEM16A is overexpressed in many tumor types including esophageal cancers, gastrointestinal stromal tumors, and SCCHN (18–20). These seemingly discordant findings prompted us to investigate the role of TMEM16A in epithelial malignancies.

We initially determined the effects of TMEM16A on tumor proliferation (in vitro and in vivo) through gain- and loss-of-function experiments. Our results show that TMEM16A induces potent and specific stimulation of the extracellular signal–regulated kinase (ERK)1/2 (as determined by phospho-ERK1/2), and contributes to the growth of cancer cell lines. These studies provide, to the best of our knowledge, the first mechanistic description of the role of TMEM16A in human malignancies and suggest that this protein may play an important role in facilitating tumor growth.

A rabbit polyclonal serum was obtained by immunizing rabbits with an epitope (CARVLEKSLKKESRNKEKR) from exon 14 of TMEM16A. This epitope was chosen on the basis of a BLAST search that defined a unique 21 amino-acid sequence. For immunoblotting, equal amounts of protein were separated on SDS-PAGE and transferred to nitrocellulose membranes. The membranes were then probed with anti-TMEM16A serum, phospho-ERK1/2, pERK1/2, p-ERK5, ERK5, B-Raf, C-Raf, phospho-AKT Ser 472, AKT (Cell Signaling), and Cyclin D1 M-20 (Santa Cruz). β-Tubulin or actin was used as a loading control. Ras activation kit was used according to instructions (Millipore, EMD). All immunoblots were scanned at 600 dpi and postprocessed using Photoshop software (Adobe Systems). Any manipulation was applied to the entire image to preserve image integrity.

HEK-293T cells were obtained from American Type Culture Collection (ATCC; Manassas, VA). UM-SCC1 and T24 cells were obtained from the University of Michigan (a gift from Dr Thomas Carey). EPC1 cells were a gift from Dr H. Nakagawa. All cell lines were genotyped to establish identity within 6 months of experimentation. Stable overexpressing clones were made using DNA transfection or retroviral infection. Cells were selected after transduction and viable cells were pooled. Individual clones were identified by the method of limiting dilutions. Each clone was kept for 10 passages, after which an early passage sample was thawed and used.

Plasmid transfections were conducted using Fugene (DNA) or Lipofectamine 2000 (siRNA) according to the manufacturer's instructions. TMEM16A and TMEM16A-K610A mutant were subcloned into pBabe-puro vectors. Retroviruses were generated by transfecting PhoenixAmpho cells with these plasmids. Lentiviral shRNA and retroviral particles were used to transduce cells. Antibiotic selection was carried out. All transduction experiments were repeated at least 3 separate times to ensure reproducibility. GFP-tagged dominant negative ERK2 construct was previously described (21). ERK1/2 siRNA was obtained from Cell Signaling.

Whole-cell patch clamping was conducted as previously described (22). All pipette solutions were previously described (22). All experiments used a standard protocol that alternated a current–voltage (I/V) step measurement. The I/V measurement stepped the holding potential from −100 to +100 mV in 20 mV steps. All patch clamp results were normalized by the cell capacitance recorded at the start of the experiment.

MQAE chloride efflux assays were conducted on cells plated onto optical Petri dishes (Matek) precoated with poly-l-lysine as previously described (23). MQAE was introduced into the cells using a hypotonic shock followed by recovery for 10 minutes before the start of the experiment. The MQAE-loaded cells were then mounted on the stage of an IX-81 Olympus microscope and perfused to 37°C. MQAE fluorescence intensity was captured every 15 seconds at the 445-nm wavelength in response to excitation at 340 nm.

The magnitude of fluorescence in each cell in the field was quantified from a circular region of interest (ROI) drawn within the cell, and the time course of fluorescence change was plotted as the average ± SEM of all ROIs in the field (typically 10–50 cells, “n”) for a single coverslip. The rate of change in fluorescence upon the switch from high chloride to low chloride was determined for each ROI. All ROIs exhibiting a positive, linear rate of change (R² ≥ 0.75) from at least 3 separate, identical experiments (coverslips, “N”) were pooled, and statistical significance was assessed for all ROIs using Student's t test.

For proliferation and viability analysis, cells were plated in 96-well optical plates at 5 × 10³ cells/well. The plates were treated the following day, as indicated. One to 3 days after treatment, the CellTiter-Glo Assay (Promega) was used according to the manufacturer's directions.

Soft agar assays were conducted as previously described (24). Briefly, 5 × 10⁴ cells suspended in 0.7% agar solution were plated in a 35-mm dish on top of 1.4% agar. Colonies were counted 3 weeks after plating. Colonies with a diameter greater than 100 μm were counted using crystal violet. Anchorage-independent viability was determined by plating cells (5 × 10³) in poly-HEMA coated plates as previously described (25). Cell viability was assessed 3 days after plating.

All animal studies were conducted under approval from the University of Pittsburgh and were conducted in accordance with established guidelines. Nude mice were injected on each flank and subsequent tumor volumes were measured when a palpable tumor was noticed. Measurements of length and width were recorded and used to determine the volume of each tumor. At the conclusion of the experiment, tumors were harvested and processed for further evaluation.

Paired normal and tumor tissues were collected after obtaining informed consent and approval from our Institutional Review Board. Normal adjacent mucosa is defined as histologically benign appearing mucosa (as judged by an experienced pathologist) acquired from the margins of the tumor resection. Tissue arrays containing replicate cores were created from patients who underwent curative surgery for SCCHN at our institution. Staining was conducted with anti-TMEM16A antisera and scored using a semiquantitative system (H-score) and the results correlated with survival. H-score was defined as the relative intensity, scored on a 0 to 3 scale, multiplied by the percentage of positively stained cells. The H-scores for the population were analyzed to determine the median score. High and low expressors were categorized as having H-scores above or below the median.

Data were abstracted from the Oncomine database, and used to assess the relative expression of TMEM16A in tumors versus normal adjacent mucosa. We specifically evaluated the expression the RefSeq version of TMEM16A: NM_018043.5. A fold-change of at least 3 was used as a cut-off value.

FISH studies were carried out on the tissue microarray using a probe for the centromere of chromosome 11 (CEP11) labeled with SpectrumGreen (Abbott Molecular) and a probe prepared from a BAC clone (RP11-805J14; CHRI) and labeled by nick translation with SpectrumOrange. Slide processing and scoring were described previously (26).

Tmem16a knockout mice were generated as previously described (8, 9). Tissues were obtained by dissecting the oral cavity mucosa from mice after genotyping. Tissues were snap frozen and subsequently used for immunochemistry, RT-PCR, and immunoblotting as described earlier.

The reverse transcriptions were carried out with predesigned TaqMan primer and probe pairs as described earlier (2). Reverse transcriptase (RT) controls were carried out for each RNA input each time. Quantitative PCR (qPCR) was conducted for TMEM16A and GAPDH was used as an endogenous control. The primer and probe sequences, conditions, and concentration have been described previously (4).

All data are reported as mean ± SEM unless stated otherwise. Cell viability and tumor xenograft measurements were analyzed with a 2-tailed Student's t test. A paired t test was used to test for differences in TMEM16A expression between matched tumor and benign mucosa. A two-sided Mann–Whitney rank-sum test was used to test the association of overexpression of TMEM16A with 11q13 amplification and TMEM16A expression on IHC. Kaplan–Meier and log rank tests of equality of survivors were used to evaluate differences in overall survival or disease-specific survival by high versus low TMEM16A IHC scores, as defined by the median value. Multivariable Cox proportional hazards models were used to evaluate TMEM16A tumor protein levels as a predictor of overall survival after adjusting for age, sex, and disease stage. The assumption of proportional hazards was tested by evaluation of Schoenfeld residuals. P < 0.05 was considered as statistically significant.

To examine the protein expression profile, rabbit polyclonal antibody to TMEM16A, was generated and used for immunoblotting and IHC (Supplementary Fig. S1a and S1b). As expected, Tmem16a derived from murine tissues migrates at a higher molecular weight (∼150 kDa) than human TMEM16A (∼115 kDa). This is consistent with previously published data (6). We further validated this antibody's specificity in IHC by evaluating tissues obtained from wild-type and knockout mice.

TMEM16A protein is overexpressed by 5-fold in tumors when compared with paired adjacent normal mucosa (Fig. 1A and B; P < 0.001). Overexpression was noted in 14/17 (83%) samples. Next, we confirmed that TMEM16A mRNA is overexpressed by 5-fold in a separate cohort of primary tumors and paired adjacent normal mucosa (Fig. 1C; P < 0.05).

TMEM16A is endogenously expressed in benign secretory tissues such as salivary gland and breast tissue (14). We therefore wanted to confirm that breast malignancies also overexpress TMEM16A. We used the Oncomine database to determine the expression of TMEM16A in normal and malignant tissues from a variety of tumor types. We found that although TMEM16A may be expressed at a high level in normal breast tissue, its expression is even higher in neoplastic breast tissue (Fig. 1D). This suggests that although endogenous TMEM16A expression may be high in some normal tissues, malignant cells derived from these tissues further upregulate TMEM16A. This finding implicates TMEM16A as a potential target gene in malignant transformation.

We then determined the association between TMEM16A overexpression and clinical outcome of patients with SCCHN. TMEM16A expression was detected in approximately 85% of SCCHN tumors. Kaplan–Meier survival analysis showed that patients with high-level tumor expression of TMEM16A had decreased overall survival (P = 0.04, log rank test; Fig. 1E). High TMEM16A tumor levels tended to be associated with decreased overall survival in a univariable Cox proportional hazards model (HR = 3.04; 95% CI, 0.97–9.46) and in a multivariable Cox proportional hazards model adjusting for age, sex, and nodal stage (HR = 2.8; 95% CI, 0.86–9.16).

However, Kaplan–Meier analysis showed no significant correlation between overall or disease-free survival and TMEM16A gene amplification (Supplementary Fig. S2a and S2b). Although patients whose tumors harbored TMEM16A amplification had a median survival of 50.7 months, whereas those not amplified for TMEM16A did not reach a median survival, suggesting that there may be a trend toward improved survival in patients without gene amplification. Gene amplification was not strongly correlated with protein expression.

Next, we investigated the effects of TMEM16A manipulation (through gain-of-function and loss-of-function) on the proliferation of cancer cells (Supplementary Fig. S3a and S3b). To evaluate the role of TMEM16A independently of 11q13 amplification, we chose to use the UM-SCC1 SCCHN cell line (that harbors 11q13 amplification) and the T24 bladder cancer cell line (that does not contain the 11q13 amplicon). Lentiviral shRNA was used to “knock-down” TMEM16A in both cell lines. We identified 2 independent shRNA sequences that caused a significant reduction (∼80%) in TMEM16A protein expression. TMEM16A knockdown led to a measurable change in whole-cell chloride conductance (∼50%), as assayed by MQAE fluorescence assays (Supplementary Fig. S3c–S3e; ref. 27). Forced overexpression of a mutant version of TMEM16A (TMEM16A-K610A) led to significantly smaller chloride conductance (Supplementary Fig. S3f). For further experiments, we used one of the sequences (shRNA#5).

To determine the consequences of TMEM16A knockdown on in vivo tumor growth, shRNA-treated UM-SCC1 cells were inoculated subcutaneously into nude mice. TMEM16A shRNA led to a significant decrease in xenograft growth (Fig. 2A and B). Furthermore, we found that TMEM16A shRNA-treated tumors exhibited decreased Ki-67 staining (Fig. 2C and D). Several reports have showed that the mitogen-activated protein kinase (MAPK)/ERK signaling pathway enhances the growth of epithelial cancer cells, in particular, bladder cancer, and SCCHN (28). We noted that pERK staining was less prominent in tumors derived from TMEM16A shRNA-treated cells versus control shRNA tumors (Fig. 2D).

Next, we exogenously overexpressed TMEM16A in several cell lines that do not harbor endogenous gene amplification. In T24 cells, TMEM16A was trafficked to the cell surface (Supplementary Fig. S1b), led to increased chloride fluxes, and resulted in enhanced xenograft growth (Fig. 3A–C). Tumor xenografts confirmed overexpression of TMEM16A and showed increased pERK1/2 and Ki-67 staining (Fig. 3D).

TMEM16A overexpression led to increased in vitro proliferation of HEK-293T cells, and increased anchorage-independent viability in immortalized keratinocytes (Supplementary Fig. S4a and S4b). TMEM16A knockdown in 2 other cell lines (Cal33 and PCI-15B) suppressed cell growth in vitro (Supplementary Fig. S4c). Furthermore, TMEM16A cooperated with constitutively active H-RAS to induce focus formation in immortalized MEFs (Supplementary Fig. S4d). Similarly, forced overexpression of TMEM16A conferred increased colonogenic survival in immortalized keratinocytes and tumor cell lines (Supplementary Fig. S4e). Taken together, these data suggest that TMEM16A overexpression can facilitate oncogenic transformation through cooperation with potent oncogenes.

We next determined the effect of TMEM16A knockdown on in vitro growth. Treatment with TMEM16A shRNA led to a significant retardation in proliferation when compared with control shRNA (Fig. 4A and B). TMEM16A shRNA abrogated the increase in both anchorage-dependent and -independent cell proliferation noted in overexpressing cells (Fig. 4C and D). TMEM16A knockdown induced an accumulation of cells in G₀/G₁ and a concomitant decrease in the S/G₂ phase, suggesting a block in cell-cycle progression (Supplementary Fig. S5a). Interestingly, TMEM16A overexpression impaired caspase-3/7 activation, suggesting that TMEM16A may in fact impair apoptotic cell death, along with promoting cell growth (Supplementary Fig. S5b).

Because we observed differential expression of pERK1/2 in tumor xenografts, we examined whether TMEM16A expression influenced ERK1/2 activation. TMEM16A overexpression was associated with an increase in phosphorylated ERK1/2 (∼2-fold) and cyclin D1 (∼5-fold; Fig. 5A). TMEM16A siRNA led to a decrease in phospho-ERK1/2 and cyclin D1 (Fig. 5B). We noted that AKT and phospho-AKT levels were not significantly influenced by TMEM16A overexpression, suggesting specificity to the ERK1/2 pathway (Fig. 5B). Similarly, there were no changes in p-ERK5 or ERK5 (Supplementary Fig. S6c).

It is well known that RAS oncogenic signaling can activate ERK1/2. HRAS, in particular, has been associated with SCCHN development (29, 30). In fact, no mutations in KRAS or NRAS were observed in 2 recent genomic studies of SCCHN (30, 31). We therefore evaluated the presence of HRAS, KRAS, and NRAS mutations in the cell lines used in this study. T24 cells harbor the activating HRASG12V mutation, however, HEK-293T, EPC1, and UM-SCC1 cells did not harbor mutant HRAS, KRAS, or NRAS (data not shown). Taken together, these data suggest that the impact of TMEM16A on ERK1/2 signaling is independent of activating HRAS mutations.

Next, we explored the association between Tmem16a and cyclinD1 in oral cavity tissues obtained from wild-type and Tmem16a^−/− embryos. As expected, Tmem16a levels were reduced in knockout mouse tissues as compared with wild-type and heterozygous mice (Fig. 5C). Interestingly, cyclin D1 was also decreased in tissues derived from knockout mice, but not in their heterozygous or wild-type littermates. We observed intense nuclear cyclin D1 staining in the proliferative mesoderm of wild-type mice but not in the Tmem16a^−/− littermates (Supplementary Fig. S6a). We further explored this finding by measuring mRNA expression of Tmem16a and Ccnd1 in these tissues. As expected, knockout tissues had significantly lower levels of cyclin D1 (Supplementary Fig. S6b). This finding suggested that TMEM16A may be influencing cell proliferation on a fundamental level and occurs in vivo.

We postulated that if TMEM16A affects proliferation by activating MEK/ERK, inhibition of MEK/ERK should abrogate TMEM16A-induced growth. Indeed, treatment with either UO126 or dominant-negative ERK1/2 led to a complete abrogation of TMEM16A-induced growth (Fig. 5D and E). Similar data were observed with the specific ERK inhibitor AZD6244 and with siRNA against ERK1/2 (Supplementary Fig. S5c and S5d). Treatment with TMEM16A shRNA reversed the activation of MEK and ERK1/2 induced by TMEM16A overexpression (Fig. 5F). This observation suggests that TMEM16A overexpression directly impacts ERK1/2 activation.

Recent data suggests that ERK activation may impact the chloride conductance of TMEM16A (15). Similarly, mutations in the putative pore forming domain can impact the whole-cell chloride conductance in forced overexpression experiments (32). We sought to determine if the putative pore-forming region of TMEM16A was necessary for activation of ERK1/2. Forced overexpression of a mutant version of TMEM16A (TMEM16A-K610A) that has been described to display significantly abrogated chloride conductance (15) did not induce ERK1/2 or phospho-ERK1/2 (Fig. 6A). We verified that TMEM16A-K610A was trafficked to the cell membrane using biotinylation experiments (data not shown). Our data raise the intriguing possibility that TMEM16A affects ERK activation by modulating intracellular chloride levels. Unfortunately, manipulation of intracellular chloride levels can itself induce changes in the expression of ion channels. Therefore, further work is necessary to dissect the effect of intracellular chloride on ERK activation.

Overexpression of mutant TMEM16A-K610A did not promote anchorage-independent viability compared with vector controls (Fig. 6B–D). To confirm that the effects are indeed directly dependent on TMEM16A expression, we rescued TMEM16A shRNA-treated cells with shRNA-resistant versions of TMEM16A or TMEM16A-K610A (Supplementary Fig. S7a). The reduction in viability induced by TMEM16A shRNA was rescued by expression of an shRNA-resistant version of TMEM16A, but not resistant TMEM16A-K610A (Fig. 6E). This approach provides strong evidence to show that TMEM16A directly induces the proliferative phenotype.

The observation that TMEM16A induces activation of ERK1/2 led us to interrogate the RAS-RAF-MEK-ERK1/2 pathway. Indeed we observed robust activation of RAS-RAF-MEK-ERK pathway upon TMEM16A overexpression (Fig. 7A and B). Densitometric quantitation is shown in Supplementary Fig. S7b. Interestingly, this activation was independent of HRAS mutation status, because T24 cells harbor an activating HRAS mutation, but SCC1 cells do not (data not shown). We next sought to determine if Ras inhibition (using a dominant negative construct) could abrogate the observed phenotype. Forced overexpression of a dominant-negative mutant of H-Ras (H-RasN17) abrogated the observed activation of MEK, ERK1/2 and also the growth phenotype (Fig. 7C and D).

To validate TMEM16A as a potential target in epithelial malignancies, we wanted to determine if small-molecule inhibition of TMEM16A could inhibit tumor cell proliferation. Therefore, we treated UM-SCC1 and T24 cells with a novel TMEM16A inhibitor (T16A-inh01; refs. 33, 34). T16A-inh01 induced a dose-dependent reduction in cell viability (Supplementary Fig. S8a and S8B). To determine whether combined inhibition of TMEM16A and MAPK had an additive effect on cell viability, we treated cells with the MEK/ERK inhibitor UO126 alone or in combination with TMEM16A shRNA. TMEM16A knockdown resulted in a modest decrease in cell proliferation, however, we observed an additive effect with MAPK inhibition (Supplementary Fig. S8c and S8d). The chemical structure of T16A-inh01 is shown in Supplementary Fig. S8e. We did not observe additive inhibitory effects on ERK phosphorylation. These data show that TMEM16A and MAPK inhibition may act in an additive fashion to retard tumor cell proliferation, suggesting that ERK1/2 may not uniquely control TMEM16A-induced cell growth.

There is accumulating evidence that chloride channels influence tumor growth and progression (35–37), however the mechanism(s) by which this occurs remains unclear. Calcium-activated chloride channels (CaCCs) are a unique subset of chloride channels that play important roles in many fundamental physiologic processes (38, 39). Recently, ANO1/TMEM16A was described as a bona fide CaCC (5–7). However, it remains controversial whether TMEM16A is itself a functional CaCC, or forms a subunit within the protein complex that facilitates CaCC activity (7, 40, 41). Recent reports suggest that CaCCs can both promote and retard tumor cell proliferation (16, 17). These contradictory reports suggest that improved understanding of the impact of CaCC regulation and activation on cell proliferation may help to define whether these molecules can serve as a future therapeutic target.

TMEM16A is frequently overexpressed in several tumors including squamous cell carcinoma of the head and neck, esophageal cancer, and gastrointestinal stromal tumors (GIST; refs. 4, 18, 42, 43). To date, the exact role(s) that TMEM16A plays in tumor development/progression remains unclear. Ayoub and colleagues have recently reported that TMEM16A overexpression facilitates cell motility and may contribute to the development of metastases (44); however, no mechanism was proposed to explain this phenotype.

This report provides the first mechanistic link between TMEM16A expression and cell proliferation in human cancer. Our data show that TMEM16A expression directly impacts cellular proliferation. TMEM16A also cooperates with oncogenic H-RAS to induce focus formation in immortalized MEFs. Taken together, these data strongly suggest that TMEM16A may function as a proto-oncogene, and that its overexpression drive tumor growth.

It has recently been shown that TMEM16A channel activity is linked to ERK1/2 activation and that the ERK inhibitor UO126 can inhibit TMEM16A channel function (15). Our data suggest that TMEM16A activates the RAS-RAF-MEK-ERK1/2 pathway, thereby influencing cellular proliferation. However, this activation does not occur when a hypomorphic mutant (TMEM16A-K610A) is expressed. These data raise the intriguing possibility that the chloride conductance of TMEM16A impacts ERK1/2 activation. However, the exact mechanism by which this occurs remains unclear.

The availability of small-molecule inhibitors against TMEM16A provides a novel potential method to inhibit tumor cell growth. The currently available small-molecule inhibitors of TMEM16A (niflumic acid and NPPB) exhibit off-target effects and have been shown to block other chloride channels. However, T16A-inh01 is a novel and potentially more specific TMEM16A inhibitor, and provides a method to inhibit channel activity (33, 34, 45). This developmental molecule likely has off-target effects, and therefore definitive conclusions cannot be defined at this time.

The ubiquitous expression of TMEM16A suggests that the endogenous protein has important physiologic roles that may be adversely impacted by pharmacologic inhibition. However, TMEM16A is known to undergo alternative splicing, and specific variants have been isolated from diseased tissues (such as diabetic gastroparesis; ref. 46–48). Recently, mutations in TMEM16A have been described in SCCHN. The existence of specific mutations in TMEM16A that have been identified from whole-exome sequencing of SCCHN, suggests that tumor-specific targeting may be possible in the future (30).

In conclusion, we have shown that TMEM16A expression: (1) is associated with decreased patient survival; (2) modulates proliferation in vitro and in vivo and induces ERK1/2 activation; (3) the pore-forming region of this molecule is necessary to facilitate these phenomena; and (4) pharmacologic inhibition (using a novel TMEM16A inhibitor) or siRNA-mediated knockdown of TMEM16A leads to a significant decrease in tumor cell viability by downregulating the TMEM16A-ERK1/2 signaling. In closing, these findings may describe a novel therapeutic role for TMEM16A in cancer treatment, because TMEM16A may act as a potential pharmacologic target.

No potential conflicts of interest were disclosed.

Conception and design: U. Duvvuri, D.J. Shiwarski, V.W. Lui, S.M. Gollin

Development of methodology: U. Duvvuri, D.J. Shiwarski, D. Xiao, X. Huang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): U. Duvvuri, D.J. Shiwarski, D. Xiao, C. Bertrand, R.S. Edinger, J.R. Rock, R.D. Harfe, B.J. Henson, R.S. Seethala, A.M. Egloff

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): U. Duvvuri, D.J. Shiwarski, D. Xiao, C. Bertrand, R.S. Edinger, R.S. Seethala, A.M. Egloff, X. Chen, J.R. Grandis

Writing, review, and/or revision of the manuscript: U. Duvvuri, D.J. Shiwarski, C. Bertrand, X. Huang, K. Kunzelmann, X. Chen, V.W. Lui, J.R. Grandis, S.M. Gollin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): U. Duvvuri, K. Kunzelmann, R. Schreiber, R.S. Seethala, X. Chen

Study supervision: U. Duvvuri, S.M. Gollin

The authors thank Dr. R.W. Sobol and the University of Pittsburgh Cancer Institute Lentiviral Facility for the generation of lentiviral particles.

This work was supported by the Head and Neck Cancer SPORE grant (NIH P50CA097190 J.R. Grandis), through a Veterans Administration CDA; CPPF (U. Duvvuri) and start-up funds from the Department of Otolaryngology (U. Duvvuri). The molecular cytogenetic studies were carried out in the University of Pittsburgh Cancer Institute Cytogenetics Facility, supported in part by P30CA047904 to N.E. Davidson/R.B. Herberman/S.M. Gollin. KK was supported by Deutsche Krebshilfe Projekt 109438.

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.