TKTL1 Is Activated by Promoter Hypomethylation and Contributes to Head and Neck Squamous Cell Carcinoma Carcinogenesis through Increased Aerobic Glycolysis and HIF1α Stabilization

There are >40,000 new cases of head and neck squamous cell carcinoma (HNSCC) in the United States each year, with a mortality rate of 12,000 annually. HNSCC develops through a prolonged multistage process involving the accumulation of genetic and epigenetic alterations. Gene silencing by promoter hypermethylation of CpG islands has been linked to the inactivation of tumor suppressor genes, such as DCC, p16, and RASSF1 in HNSCC (1, 2). In contrast, less is known about the effect in tumor development or progression of DNA hypomethylation. DNA hypomethylation has recently been proposed to play a role in cancer gene expression activation and genomic instability (3). Rare examples of hypomethylation leading to the activation of genes that are important in cancer have resulted in the overexpression of Bcl-2, R-ras, and Cyclin D2 (4–6). In HNSCC as in other solid tumor, DNA global hypomethylation is found and is associated with smoking, alcohol consumption, and stage (7). Using pharmacologic demethylation in normal oral keratinocyte cells combined with expression analysis in primary tumor tissues as a whole genome discovery approach, we recently identified a set of novel candidate oncogenes that undergo hypomethylation and increased expression in primary human HNSCC cancer, including TKTL1 (8).

As in most solid tumors, HNSCC displays dramatically altered glucose metabolism. Over 75 years ago, Warburg noted that cancer cells typically depend more on increased rates of glycolysis even in the presence of available oxygen, a phenomenon known as aerobic glycolysis or the Warburg effect (9). It is suggested that the Warburg effect confers a selective advantage for survival and proliferation of tumor cells. Recently, the relation between HIF1α and metabolic reprogramming has been explored in HNSCC and other solid tumors. HIF1α participates in the transcriptional activation of genes that are involved in crucial aspects of cancer biology including angiogenesis, cell survival, glucose metabolism, and invasion (10). It is proposed that HIF1α represents one principal mechanistic axis for the regulation of aerobic glycolysis (11). Further, recent studies suggested that several intermediary glucose metabolites could promote HIF1α activation by interacting with the HIF prolyl hydroxylases (PHD 1-3). The expression of a mutant mitochondrial-ND2 mutation contributes to the development of a HNSCC malignant phenotype through increased reactive oxygen species, upregulation of PDK2, attenuating PDH activity, elevating pyruvate, and lactate production, and thereby HIF1α stabilization (12). Moreover, we found that the inhibition of PDH activity through the enhanced expression of PDK1 is associated with the normoxic stabilization of HIF1α by glycolytic metabolites. Out study suggests that the buildup of glycolytic metabolites, resulting from high PDK1 expression, may in turn promote HIF1α activation, thus sustaining a feed-forward loop for HNSCC malignant progression (13).

In this study, we describe TKTL1, a gene-encoding enzyme responsible for the transketolase reaction in pentose phosphate pathway (PPP; ref. 14). The PPP, which facilitates glucose conversion for nucleic acid synthesis, glucose degradation to lactate, and regeneration of redox equivalents, connects with glycolysis through metabolites such as fructose-6-phosphate and glyceraldehyde-3-phosphate (15). In our preliminary study, we have found that TKTL1 promoter demethylation and overexpression occurs in HNSCC (8). Here, we further explore the functional significance of TKTL1 in HNSCC tumorigenesis and show an oncogenic function of TKTL1 in promoting HNSCC tumor growth through the enhancement of aerobic glycolysis and increased normoxic HIF1α expression, facilitating HNSCC tumor growth.

The tumor samples from patients with HNSCC and normal samples were obtained from patients surgically treated in the Department of Otolaryngology-Head and Neck Surgery at Johns Hopkins Medical Institutions, Baltimore, using appropriate informed consent obtained after institutional review board approval. Microdissection of frozen tumor tissue was done to assure that >80% of tissue contained HNSCC. The normal tissues consisted of tissues obtained from patients that underwent uvulopalatopharyngoplasty from sleep apnea. UM22A and UM22B cells were provided by Ajay Verma (Merck & Co., North Wales, PA). FaDu cells were from the American Type Culture Collection. Human HNSCC cell line JHU-O11, JHU-O22, and JHU-O28 were from the Johns Hopkins University Department of Otolaryngology-Head and Neck Surgery (16, 17).

Genomic DNA was extracted using the phenol chloroform, and bisulfate modification of genomic DNA was done as described in ref. (16). Quantitative unmethylation-specific PCR (QUMSP) was carried out as previously described (8). In our previous cohort, we have shown that the QUMSP method is highly correlated with bisulfite sequencing method in the detection of TKTL1 hypomethylation.

Real-time reverse transcription-PCR (RT-PCR) was done on the ABI PRISM 7000 Sequence Detection System using the QuantiFast SYBR Green PCR kit (Qiagen). The primers used for real-time RT-PCR were available upon request. β-Actin was used for the normalization of the expression data of each gene. Each sample was run in triplicate to ensure quantitative accuracy, and the threshold cycle numbers (Ct) were averaged. The results were reported as fold changes and were calculated using the 2⁻δδ^(Ct) method (18, 19).

A human cDNA clone of TKTL1 was purchased from the American Type Culture Collection. HindIII and XbaI restriction sites flanking the whole open reading frame of TKTL1 were constructed by PCR. The PCR products were digested with HindIII and XbaI restrictases and gel, which were extracted and subcloned between into the corresponding sites of pcDNA3.1 expression vector. The sequence and direction of the insert was confirmed during each step by sequencing.

The HNSCC cell lines were transfected with TKTL1 expression plasmids at 40% to 60% confluence by using FuGENE 6 (Roche) as previously described (12). Validated plasmids encoding shRNA for TKTL1 and negative control were introduced into the FaDu cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions (8). Cell growth rate was measured by MTT assay (Sigma; ref. 12). The HIF1α inhibitor, NSC134754, was obtained from National Cancer Institute through their Developmental Therapeutic Program. HIF1α small interfering RNA was from Ambion. Ascorbate and 3-bromopyruvate was purchased from Sigma. The soft agar assay was carried out as previously described (20).

Nonobese diabetic/severe combined immunodeficient mice (5-wk-old; Jackson Laboratories) were injected s.c. with 5 × 10⁶ TKTL1- or vector-expressing O11 cells. One week after injection, tumor volume was assessed every 2 to 3 d using a digital caliper and calculated using the following formula: volume = (a × b²)/2, in which a is the widest diameter of the tumor and b is the diameter perpendicular to a. At 5 wk after injection, the tumors were dissected and weighed.

Metabolic extracts were prepared and measured by liquid chromotography-mass spectrometry/mass spectrometry (LC-MS/MS) as previously described (13, 21). Before extraction, the frozen cell samples were suspended in 2 mL of 100% methanol and were mixed vigorously. The cells were sonicated before the resulting suspension was centrifuged at 9,500 × g for 10 min. The supernatant was then dried under a nitrogen flow and reconstituted in 500 μL of methanol/water (10:90, v/v). Chromatographic analysis was done using an Agilent 1100 Series high performance liquid chromatography. The column effluent was monitored using an API 3000 triple-quadruple mass-spectometric detector (Applied Biosystems). The spectrometer was programmed to monitor the ion of glyceraldehyde-3-p at m/z 168.7, ribose-5-p at m/z 228.7, and that of fructose-6-phosphate at m/z 258.7. For each measurement, at least triplicates were done.

Pyruvate and lactate production was measured by the enzymetic method using a commercially available fluorescence-based assay kit or using the CMA 600 Analyzer as previously described (13, 22). Glucose in the medium were quantitated using the amplex red glucose/glucose oxidase kit (Invitrogen; ref. 23). ATP levels were assessed using an ATP bioluminescence assay kit (Roche; refs. 22, 24). Cell number was determined using a Coulter particle analyzer.

Approximately 30 μg of whole protein extract were electrophoresed and transferred to a polyvinylidene difluoride membrane, probed overnight at 4°C with the antibody against TKTL1 (R-Biopharm, kindly provided by Dr. Coy, University of Wuerzberg, Wuerzberg, Germany), HIF1α (BD Biosciences), and actin (Sigma). Quantification of band intensity was done using the Image J software (NIH).

Differences between experimental variables were estimated using Student's t test or Mann-Whitney tests as appropriate. A probability level of 0.05 was chosen for statistical significance. All computations were done in R 2.8.1.⁵

Our laboratory has previously described the coordinated activation of candidate proto-oncogenes through promoter hypomethylation in HNSCC and lung cancer (8). We found that promoter hypomethylation and overexpression of TKTL1, a candidate oncogene, occurred in HNSCC. To examine the frequency of TKTL1 promoter hypomethylation in HNSCC, we further tested tissues from 58 HNSCC cancers and 11 normal mucosas in noncancer control subjects, using a method we devised. This method, previously termed TKTL1 QUMSP, is a rapid, quantitative assay for specifically measuring the nonmethylated TKTL1 promoter. We found that the TKTL1 promoter was significantly demethylated in primary HNSCC (log₁₀ median, 0.45; log₁₀ range, −1.04 to 3.01) compared with that in normal mucosa (log₁₀ median, −0.44; log₁₀ range, −0.97 to 0.02; P < 0.001), confirming that TKTL1 is hypomethylated in HNSCC (Fig. 1A). To validate the overexpression of TKTL1 at the transcriptional level in HNSCC, we further compared TKTL1 expression patterns in 36 HNSCC tumor tissues to those seen in 16 normal mucosas using quantitative RT-PCR. Our analysis revealed the mean expression levels for TKTL1 in HNSCC tumor tissues was roughly 7-fold higher than that in the normal mucosa (P = 0.047; Fig. 1B). Further, Western blot was done to confirm TKTL1 overexpression in HNSCC tumor samples and different HNSCC cell lines. Of the 16 HNSCC tumor samples we examined, 6 were found to have increased expression of TKTL1 protein compared with the normal mucosas. Meanwhile, among the HNSCC cell lines blotted, we found that TKTL1 was relatively overexpessed in FaDu and UM22B cell lines compared with levels in nomal mucosal samples (Fig. 1C).

To understand the function of TKTL1 in HNSCC, we first explored the effect of ectopic TKTL1 expression on cellular growth in vitro in JHU-O11 cell line, which expresses relatively lower endogenous levels of TKTL1. As shown in Fig. 2A (left), following transfection with a TKTL1 cDNA expression construct, JHU-O11 cells showed significant increases in cellular proliferation (P < 0.05). To further functionally characterize the oncogenic property of TKTL1, we established stable cell lines overexpressing TKTL1 in O11 cells and performed anchorage-independent growth assays in cells stably expressing TKTL1. Forced expression of TKTL1 in the stable clone was verified by immunobloting with the monoclonal anti-mouse TKTL1 antibody (Fig. 2A, middle). Two weeks after seeding on soft agar, we noted a dramatic increase in size and colonies in TKTL1-expressing cells compared with the vehicle-expressing cells (417 ± 22 versus 274 ± 25; n = 3; P < 0.05; Fig. 2A, right). Similarly, the JHU-O28 cells transiently transfected with the TKTL1 expression construct grew faster than that transfected with vector (Fig. 2B). Further, stable expression of TKTL1 in O28 cells exhibited markedly increased colony formation relative to vector controls (n = 3; P < 0.05; Fig. 2B). Taken together, these data suggest that overexpression of TKTL1 provides a growth advantage for HNSCC tumor cells in vitro.

To examine the effects of TKTL1 overexpression on HNSCC tumor growth in vivo, we s.c. injected the TKTL1- or vehicle-expressing O11 cells into nonobese diabetic/severe combined immunodeficient mice. The tumor growth was monitored over a 5-week period. As shown in Fig. 2C, mice injected with the TKTL1-expressing cells showed a significant increase in tumor development compared with those injected with the vector control cells (P < 0.01). The tumors developed from the TKTL1-expressing cells were larger in size compared with that from the vehicle expressing cells (Fig. 2D, left). The TKTL1-expresssing cells caused larger tumors than the vector-expressing cells [TKTL1-expressing cells (mean, 132 mg; 95% confidence interval, 68.4-180.1 mg) versus vector-expressing cells (mean, 59.1 mg; 95% confidence interval, 49.8-84.2 mg); P = 0.016; Fig. 2D, right]. These data support the role of TKTL1 in promoting HNSCC tumor growth.

We sought to address the means by which overexpression of TKTL1 promotes HNSCC tumor growth. TKTL1 encodes a member of PPP, which has an important role in controlling the nonoxidative PPP (25). As shown in Fig. 3A, the PPP is directly connected to glycolysis, as fructose-6-phosphate and glyceraldehyde-3-phosphate are the intermediates in both pathways. We hypothesized that TKTL1 could enhance the increase in production of fructose-6-phosphate and glyceraldehydes-3-phosphate, increasing aerobic glycolysis. The fructose-6-phosphate and glyceraldehydes-3-phosphate levels in TKTL1-expressing O11 cells were determined by the LC-MS/MS–based analysis of metabolites. As shown in Fig. 3B and C, we found an increase in fructose-6-phosphate (8.8 ± 0.7 versus 5.6 ± 0.5; n = 3; P = 0.019) and glyceraldehydes-3-phosphate production (17.8 ± 2.3 versus 8.0 ± 0.7; n = 3; P = 0.039) in TKTL1-expressing O11 cells compared with vehicle-expressing cells. This suggests that TKTL1 may promote the aerobic glycolysis through enhanced metabolite production including fructose-6-phosphate and glyceraldehydes-3-phosphate.

Given the enhanced fructose-6-phosphate and glyceraldehydes-3-phosphate we observed, we were interested in examining the effects of TKTL1 overexpression on aerobic glycolysis. It is known that pyruvate and lactate, the end products of aerobic glycolysis, are elevated in cancer cells even in the presence of oxygen (26). We proposed that TKTL1 enhances the generation of fructose-6-phosphate and of glyceraldehyde-3-phosphate, resulting in elevated production of pyruvate and lactate. We found that significantly increased pyruvate and lactate production in TKTL1-expressing O11 cells compared with the corresponding vector controls (pyruvate: 23.4 ± 0.4 versus 17.2 ± 1.0, P < 0.01, Fig. 4A; lactate: 0.35 ± 0.03 versus 0.21 ± 0.03, P < 0.01, Fig. 4A). Similar results were obtained in O28 cells ectopically expressing TKTL1 (P < 0.05; Fig. 4B). As the Warburg effect described, tumor cells rapidly use glucose and convert the majority of it to lactate. To determine if the increased pyruvate and lactate production are accompanied by the increased glucose consumption, we monitored the glucose uptake in the TKTL1-expressing O11 cells. As shown in Fig. 4C, glucose uptake was significantly higher in TKTL1-expressing O11 cells than that in the vector control cells (6.3 ± 0.4 versus 5.3 ± 0.1; P = 0.012). Further, to evaluate the effect of the ectopic expression of TKTL1 on the cellular energy balance, ATP levels were measured in the TKTL1-expressing O11 cells and its corresponding vector controls. Consistently, we found that TKTL1-expressing O11 cells displayed increased ATP levels compared with the vehicle-expressing cells (8.1 ± 0.2 versus 6.4 ± 0.5; P = 0.037; Fig. 4D). Together, these data supported that TKTL1 increased aerobic glycolysis in HNSCC tumorigenesis.

We next explored possible downstream mediators of TKTL1's growth promoting effect. Recently, the cancer-specific aerobic glycolytic metabolism reported by Warburg has been linked to HIF1α degradation induced by pyruvate and lactate (27). We predicted that TKTL1 overexpression results in HIF1α accumulation through the increased pyruvate and lactate. We examined the expression level of HIF1α in TKTL1-expressing O11 cells. As shown in Fig. 5A, enhanced HIF1α accumulation was observed in the TKTL1-expressing O11 cells compared with the vehicle-expressing cells. Further, we found that two HIF1α downstream target genes, including GLUT1 and LOX, which are considered to be under the regulation of hypoxia, were upregulated in TKTL1-expresing O11 cells (Fig. 5A). Of note, the HIF1α accumulation and the upregulation of GLUT1 were also found in the tumor xenografts originated from the TKTL1-expressing O11 cells (Fig. 5B). Conversely, TKTL1 shRNA–mediated silencing of endogenous TKTL1 expression significantly decreased the HIF1α levels in FaDu cells (Fig. 5C). To address the contribution of HIF1α to growth, a HIF1α small-molecule inhibitor, NSC134754, was applied to the TKTL1-expressing O11 cells (12, 28). Our results found that treatment with 0.5 μmol/L of NSC134754 significantly inhibited the TKTL1-expressing O11 cell growth (Fig. 5D). We found that upon treatment with ascorbate, an agent which selectively reverses the HIF1α accumulation induced by pyruvate and lactate, the growth-promoting effect mediated by TKTL1 in O11 cells was abolished (Supplementary Fig. S1A and B; ref. 27). Furthermore, we observed that the introduction of HIF1α small interfering RNA resulted in growth inhibition that reversed the TKTL1-mediated growth stimulation in O11 cells (Supplementary Fig. S1C). Overall, these data indicate that the growth promoting effects of TKTL1are significantly dependent on HIF1α stabilization.

We hypothesized that enhancement of HIF1α constitutes a feed-forward mechanism through upregulation of glycolytic enzymes contributing to the TKTL1-mediated HNSCC tumorigenesis. To validate the involvement of HIF1α on glycolysis during TKTL1-mediated HNSCC tumorigenesis, we next performed quantitative RT-PCR of mRNAs corresponding to four well-established HIF1α target genes implicated in glycolysis, including HK2, ALDOC, PGK1, and PGM1, in the TKTL1-expressing O11 cells. As shown in Fig. 6A, elevated expression of HK2, ALDOC, PGK1, and PGM1 were found in TKTL1-expressing O11 cells in comparison with vector-expressing cells. We found that HK2, ALDOC, and PGK1 were also upregulated in the tumor xenograft that originated from TKTL1-expressing O11 cells (Supplementary Fig. S1D). Conversely, we found that upon treatment with 1 μmol/L of HIF1α small-molecule inhibitor for 48 hours, the upregulation of HK2, ALDOC, and PGK1 was eliminated (Fig. 6A). To examine whether this increased glycolytic phenotype is essential for the TKTL1-mediated tumor growth, 3-bromopyruvate, an agent that inhibits glycolysis through inhibition of hexokinase, was then applied to the TKTL1-expressing O11 cells (29). We found that treatment of 3-bromopyruvate significantly inhibited the TKTL1-expressing O11 cell growth (Fig. 6B). It is previously reported that the PPP including two distinctive branches, oxidative branch and nonoxidative branch, fulfill the requirement for ribose-5-phosphate for the synthesis of nucleotides and nucleic acids (15). It is thought that the nonoxidative branch of PPP is the main source for ribose-5-phosphate synthesis in the tumor cells (30, 31). To determine whether the production of ribose-5-phosphate is coordinated with the enhanced production of fructose-6-phosphate and glyceraldehydes-3-phosphate, we also measured the generation of ribose-5-phosphate in TKTL1-expressing O11 cells. Intriguingly, we noted that ribose-5-phosphate is significantly increased in the TKTL1-expressing O11 cells compared with the vehicle-expressing cells (15.2 ± 1.5 versus 8.6 ± 1.5; P = 0.035; Fig. 6C). This observation suggests that HIF1α promotes a feed-forward loop in TKTL1-mediated HNSCC tumorigenesis by upregulating glycolytic target genes, as well as indirectly provides increasing substrate for TKTL1-mediated supply of glycolytic metabolites (Fig. 6D).

TKTL1 has been reported to be overexpressed in several different cancer types, including lung, cervical, nasopharyngeal, breast, colon, and gastric cancers. Recent studies have correlated TKTL1 expression with poor survival in laryngeal carcinoma, colon, and urothelial cancers, as well as distant metastasis in ovarian carcinoma (25, 32). In our current study, QUMSP analysis revealed a dramatic increase of TKTL1 hypomethylation in the HNSCC tumors compared with the normal control mucosa. By evaluation of TKTL1 overexpression at transcriptional level and protein level, we observed that TKTL1 is significantly overexpressed in HNSCC cancers compared with that in normal tissues. It is likely that hypomethylation could be attributable to the overexpression of TKTL1. In support of this, we previously identified the HNSCC primary tissues showing both hypomethylation and overexpression of TKTL1. In the TKTL1-methylated human oral keratinocyte cell line OKF6, we reported transcriptional upregulation of TKTL1 after 5-aza/TSA (trichostatin A) treatment, supporting the concordance between methylation and expression (8).

In exploring the oncogenic function of TKTL1, we found that HNSCC cells (O11 and O28 cells, as well as 22A cells)⁶

It has long been considered that aerobic glycolysis associated with the Warburg effect is an intrinsic component of a malignant phenotype. TKTL1 encodes transketolase of the PPP, which catalyzed the production of metabolites including fructose-6-phosphate and glyceraldehydes-3-phosphate. Interestingly, glyceraldehydes-3-phosphate has been shown to be significantly increased on disease progression from benign to clinically localized prostate cancer to metastatic prostate cancer (33). Here, we propose that TKTL1 increases the production of fructose-6-phosphate and glyceraldehydes-3-phosphate, resulting in an increased aerobic glycolysis. Several lines of evidence support this. First, enhanced production of fructose-6-phosphate and glyceraldehydes-3-phosphate was observed in TKTL1-expressing cells compared with that in the vector control cells. Second, production of pyruvate and lactate were significantly elevated in TKTL1-expressing cells in comparison with their vector control cells (in O11 cells and O28 cells, as well as 22A cells).⁶ Third, increased glucose consumption, accompanied by increased ATP yield, was shown in the TKTL1-expressing cells. In agreement to our study, Xu et al. (34) also reported that inhibition of TKTL1 resulted in a significantly slowed glucose consumption and lactate production.

Lu and colleagues (27) recently reported that the metabolites of Warburg effect, pyruvate and lactate, could dramatically increase HIF1α accumulation in cancer cells by inhibiting the prolyl hydroxylase enzyme activity, which could result in the degradation of HIF1α protein through hydroxylation. Given the increased production of pyruvate and lactate we observed in TKTL1-expressing HNSCC cells, we further investigate the effect of TKTL1 overexpression on HIF1α stabilization. We found the HIF1α accumulation, together with the upregulation of two HIF1α downstream target genes, GLUT1 and LOX, in TKTL1-expressing O11 cells and/or tumor xenografts. In addition, the use of a HIF1α small-molecule inhibitor, as well as HIF1α small interfering RNA, blocked TKTL1-mediated cancer cell growth. Conversely, shRNA knockdown of TKTL1 expression in FaDu cells decreased HIF1α accumulation, which is consistent with the decreased cellular growth after TKTL1 shRNA transfection as we reported previously. Of note, we also found that overexpression of TKTL1 in human immortal oral keratinocyte OKF6 resulted in HIF1α accumulation accompanied with increased lactate production.⁶ These suggest that TKTL1 may induce HNSCC development through constitutive activation of HIF1α.

We also propose that the enhanced HIF1α accumulation represents a feed-forward mechanism for the TKTL1-mediated tumorigenesis. In this respect, our study showed that four glycolytic genes directly targeted by HIF1α, including HK2, ALDOC, and PGK1, were upregulated in TKTL1-expressing O11 cells, as well as in tumor xenografts. The upregulation of HK2, ALDOC, and PGK1 in TKTL1-expressing O11 cells could be abolished after treatment with HIF1α small-molecule inhibitor, suggesting the specificity of pathway through HIF1α-mediated action. Meanwhile, treatment of 3-bromopyruvate, a glycolytic inhibitor, through the inhibition of hexokinase abolished TKTL1-mediated O11 tumor cell growth. These data indicate that the induction of HIF1α, coupled in turn with the HIF1α-promoted expression of glycolytic enzymes, could be an essential feed-forward loop for sustaining the glycolytic phenotype and malignant progression mediated by TKTL1.

The data in the present study support the following model for the mechanism by which TKTL1 functions to promote carcinogenesis in HNSCC cancer cells. As shown in Fig. 6D, in HNSCC cancer cells expressing high levels of TKTL1, TKTL1 enhances the production of fructose-6-phosphate and glycerealdehyde-3-phosphate (two essential metabolites required for glycolysis), thereby facilitating an increased production of pyruvate and lactate, and subsequent HIF1α accumulation and activation of its downstream target genes. Altogether, these studies define an oncogenic role of TKTL1 in human tumorigenesis and define a novel mechanism for increasing aerobic glycolysis and HIF1α accumulation in cancer. Although these data do not support TKTL1 as sufficient to induce carcinogenesis in an HNSCC model, they do support the role of TKTL1 as a significant contributor to a malignant phenotype.

No potential conflicts of interest were disclosed.

We thank M. Ochs for the statistical advice.

Grant Support: Specialized Program of Research Excellence grant P50 CA96784 and a Clinical Innovator Award from Flight Attendant Medical Research Institute.

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.