Bitter Taste Receptors Influence Glucose Homeostasis

TAS1R- and TAS2R-type taste receptors are expressed in the gustatory system, where they detect sweet- and bitter-tasting stimuli, respectively. These receptors are also expressed in subsets of cells within the mammalian gastrointestinal tract, where they mediate nutrient assimilation and endocrine responses. For example, sweeteners stimulate taste receptors on the surface of gut enteroendocrine L cells to elicit an increase in intracellular Ca2+ and secretion of the incretin hormone glucagon-like peptide-1 (GLP-1), an important modulator of insulin biosynthesis and secretion. Because of the importance of taste receptors in the regulation of food intake and the alimentary responses to chemostimuli, we hypothesized that differences in taste receptor efficacy may impact glucose homeostasis. To address this issue, we initiated a candidate gene study within the Amish Family Diabetes Study and assessed the association of taste receptor variants with indicators of glucose dysregulation, including a diagnosis of type 2 diabetes mellitus and high levels of blood glucose and insulin during an oral glucose tolerance test. We report that a TAS2R haplotype is associated with altered glucose and insulin homeostasis. We also found that one SNP within this haplotype disrupts normal responses of a single receptor, TAS2R9, to its cognate ligands ofloxacin, procainamide and pirenzapine. Together, these findings suggest that a functionally compromised TAS2R receptor negatively impacts glucose homeostasis, providing an important link between alimentary chemosensation and metabolic disease.

Because of the important role of TAS1R and TAS2R taste receptors in nutrient detection and response in the gustatory and digestive systems, we hypothesized that allelic variations affecting the function of individual TAS1Rs or TAS2Rs could significantly impact glucose homeostasis. We initiated a candidate gene study within the Amish Family Diabetes Study (AFDS) [28], followed by functional characterization of candidate receptor variants, to identify specific sequence variants in TAS1R genes and/or TAS2R genes that are associated with glucose homeostasis.

Results
Glucose dysregulation, including elevated plasma glucose, increased hepatic gluconeogenesis, and decreased insulin mediated glucose transport, is a hallmark of type 2 diabetes mellitus (T2DM) [29]. We first asked whether any variants in taste receptor genes are associated with T2DM in the Amish. We genotyped haplotype-tagging, single nucleotide polymorphisms (SNPs) in or around all TAS1R and TAS2R genes in all T2DM cases (n = 145) and a subset of controls (n = 358) from the Amish Family Diabetes Study (AFDS [28]; Tables 1-4). Only four SNPs, all on chromosome 12, showed significant associations with T2DM (Tables 1-3). Of these, the non-coding SNP rs2588350 showed the greatest significance (P = 0.0007; p_ACT value = 0.025 after correction for multiple comparisons). No SNPs on chromosomes 5 or 7 were associated with T2DM ( Table 2). Surprisingly, although both the TAS1R2 and TAS1R3 proteins are required for normal glucose sensing in both the gustatory and gastrointestinal systems (e.g., [16,17,[30][31][32]), we observed no significant associations with TAS1R haplotype-related SNPs (Table 3). Indeed, all TAS1R3 SNPs were monomorphic in the Amish (Table 3).
To confirm that the T2DM association of these three SNPs reflects an underlying dysregulation of glucose and insulin homeostasis, we performed association analyses with glucose and insulin traits obtained in 693 non-diabetic AFDS subjects who had been given a standard 3-hour OGTT. The minor alleles of all three SNPs were significantly associated with several measures of glucose and insulin homeostasis ( Table 5), including glucose area-under-the-curve (AUC) and insulin response. Insulin AUC was higher for individuals with the minor allele of any of the three SNPs, but these differences were only statistically significant for rs3741845 and rs2588350 ( Table 5).
Estimates of insulin resistance based on homeostatic model assessment were also significantly affected in subjects with the rs3741845 T allele. Thus, the minor alleles of rs2588350, rs619381 and rs3741845 display similar phenotypic associations. As these three SNPs display significant LD (Figure 1), we conclude that the minor alleles of these SNPs comprise a single haplotype associated with dysregulated postprandial glucose homeostasis.
Though any of the three SNPs within this glucose dysregulation haplotype could potentially affect receptor expression or function, and thus glucose and insulin homeostasis, the rs3741845 T allele is a particularly attractive candidate risk allele: this SNP alters an amino acid within a region of TAS2R9 that is predicted to influence ligand binding and response of other GPCRs, including TAS2Rs [33]. Therefore, we asked whether the Ala to Val change alters the ligand response of TAS2R9. Since TAS2R9 was an orphan receptor, we utilized a high-throughput screening strategy to identify bitter-tasting stimuli that activate TAS2R9. Three of the 64 bitter-tasting compounds screened (Supplemental data, Table S1) activated TAS2R9 Ala187-expressing cells: the fluoroquinolone antibiotic ofloxacin (Figure 2A and D), the tricyclic gastric acid inhibitor pirenzapine ( Figure 2B and E) and the antiarrhythmic drug procainamide ( Figure 2C and F). They did so with an EC50 of 0.2, 1.8 and 2.8 mM, respectively. The Val187 variant of TAS2R9 showed a dramatic loss of function, with no responses to any of the bitter stimuli, even at high concentrations ( Figure 2D, E and F and data not shown). This functional decrement is not due to differences in surface expression ( Figure 2G). Thus, the rs3741845 minor allele (T) causes a major functional deficit in ligand response of TAS2R9.
Though the mechanism by which this taste receptor-associated haplotype affects glucose and insulin homeostasis remains unclear, these receptors could be involved in the modulation of GLP-1 secretion from gut enteroendocrine L cells [14]. We used reverse transcription-polymerase chain reaction (RT-PCR) to determine if TAS2R9 is expressed in these cells. We amplified both TAS2R9 and TAS1R3 (a subunit of the sweet and umami taste receptors previously reported to be expressed in enteroendocrine L cells [15,23,25]) from cDNA obtained from NCI-H716 cells (a human enteroendocrine L cell line; Figure 3A), from human cecum ( Figure 3A), and from human tongue (data not shown). We were unable to amplify TAS2R7 from any of these cDNA pools ( Figure 3A and data not shown), though we could amplify a product from human genomic DNA (data not shown). The TAS2R9 products were amplified from cDNA and not genomic DNA contaminants: PCR from control samples that were not reverse transcribed gave no TAS2R9 product (data not shown), and oligos that recognize coding sequences in exons 4 and 6 of taste receptor TAS1R3 amplify a product lacking the two intervening introns ( Figure 3A). Independent clones of the TAS2R9 product amplified from the NCI-H716 cells had either an A or T at bp 560, indicating that this cell line is heterozygous for this allele. Next, we tested whether a TAS2R9 ligand can promote GLP-1 secretion from enteroendocrine L cells. Stimulation of NCI-H716 cells with ofloxacin elicited a concentrationdependent secretion of GLP-1 from this cell line ( Figure 3B). siRNA knockdown of the G protein a-gustducin ( Figure 3C), which mediates bitter taste responses in the tongue [34] and which has been implicated in taste receptor-mediated GLP-1 secretion in the gut [15], reduced ofloxacin-stimulated GLP-1 secretion ( Figure 3B). Together, these results are consistent with a role of TAS2R9 in the regulation of nutrient-dependent GLP-1 secretion from L cells.

Discussion
By combining human genetic approaches with high-throughput receptor screening, we have identified an important link between taste receptor function and the modulation of glucose homeostasis. Our study provides genetic and biological validation of an association between a TAS2R haplotype on human chromosome 12 with the regulation of glucose and insulin levels. The novel role of TAS2Rs in the maintenance of glucose homeostasis should help elucidate the relative contributions of taste receptor-mediated chemoreception in the gustatory and digestive systems and suggests new lines of investigation for ameliorating risk of metabolic disease and for developing novel avenues for treatment.
Our conclusions are foremost based on the genetic association of a TAS2R haplotype, including the TAS2R9 T560 allele, with measures of glucose and insulin dysregulation in non-diabetic Amish individuals ( Table 5) and with increased incidence of  1-4). The observed concordance of several related but independent phenotypic measures (e.g., measures of glucose response during an OGTT, measures of insulin response during an OGTT, and T2DM diagnosis) with the same allele provides important internal replication of the association. The few T2DM genome-wide association studies (GWASs) that included rs619381 and rs3741845 (but not rs2588350) [35][36][37] do not report significant associations between T2DM and these SNPs. However, the conservative cutoffs necessary for GWAS (,5610 27 ) would likely exclude these associations from consideration in most studies. Indeed, one heavily replicated T2DM gene, PPARG, would not have met the criteria for a novel T2DM gene in at least one GWAS [38]. The AFDS, though a relatively small study in a homogeneous population, has exhibited sufficient power to identify or replicate at least two T2DM risk alleles that are replicated in other populations [39,40]. However, association studies of rs3741845 in genetically heterogeneous populations may be further complicated by population stratification, as the frequency of the rs3741845 alleles varies greatly across human populations (Supplementary data, Table S2). Further genetic and mechanistic analyses will be needed to determine the extent to which contributions of specific TAS2R variants to glucose dysregulation are found in other populations. We also provide important biological validation of the association data: a physiological consequence of the TAS2R9 polymorphism (i.e., a loss of response to several ligands). The rs3741845 SNP predicts an amino acid change in the second extracellular loop or fifth transmembrane domain of TAS2R9, a region suggested to form part of the ligand binding pocket and to be important for receptor activation [33,41,42]. In contrast, rs619381 affects an amino acid in the C-terminal domain of TAS2R7, a region unlikely to directly impact ligand interactions, and the rs2588350 SNP is a non-coding polymorphism. While any of these SNPs could potentially impact TAS2R expression or function, and therefore glucose and insulin homeostasis, we reasoned that the TAS2R9 variant was the most likely to significantly alter receptor function. The observation that TAS2R9, but not TAS2R7, is expressed in human enteroendocrine cells (Figure 3) further supports a key role for TAS2R9. The single amino acid change from Ala to Val at position 189 has a profound effect on TAS2R9 function, abolishing responses to three different ligands ( Figure 2). No systematic structurefunction analyses have been performed for TAS2Rs, and studies that can differentiate ligand binding from other aspects of receptor activation have been limited to TAS1Rs (e.g., [22,43,44]). Though we cannot resolve whether the Ala to Val change specifically impacts ligand binding, binding-induced conformational changes, or effective G protein coupling, the inability of the Val187 variant to respond to any of three different ligands provides compelling evidence that this variant is incapable of transducing stimuli.
The observation that TAS2R9 is expressed in enteroendocrine cells and that a TAS2R9 ligand can elicit GLP-1 secretion suggests a possible mechanism, regulation of incretin response in the gut, by which variation in taste receptor function could impact glucose and insulin regulation. Gut TAS2Rs could be stimulated by a number of compounds, including ingested toxins or bitter-tasting peptides that result from the fermentation of proteins such as casein [45]. Gut flora, which can vary dramatically between obese and lean individuals [46,47], could also serve as a source of TAS2R stimuli in normal or pathogenic states. Many TAS2Rs are broadly tuned to multiple stimuli, and some bitter stimuli activate more than one TAS2R [12,20,41,42,[48][49][50]. The three TAS2R9 ligands identified in this study, ofloxacin, pirenzapine and procainamide, are not natural ligands for this receptor, some of which would be expected to activate TAS2R9 with a higher efficacy. Even so, they do serve as effective tools to assay the consequences of the Ala187Val mutation ( Figure 2). Interestingly, some fluoroquinolones, particularly gatifloxacin and levofloxacin (the L-isomer of ofloxacin), have been associated with dysglycemia in diabetic and non-diabetic patients [51]. TAS2R9 does not respond to the three other fluoroquinolones we tested (gatifloxacin, ciprofloxacin and enoxacin; Supplementary data, Table S1), but it is intriguing to consider whether some bitter-tasting pharmaceuticals may affect glucose homeostasis, at least in part, through actions on TAS2Rs.
However, we cannot rule out alternative physiological mechanisms that link TAS2R function to the modulation of glucose homeostasis. For example, taste receptors could affect glucose homeostasis through a gustatory mechanism by altering the perceived qualities of food and impacting food preference and intake [6,30]. Indeed, taste receptor polymorphisms affect the ability to recognize taste stimuli by altering the perceived qualities of food and impacting food preference and intake [1,6]. Intragastric infusion of sweet-and bitter-tasting compounds also impacts taste preference [52,53]. Therefore, blindness to particular bitter-tasting compounds could lead to increased ingestion of toxins [3]; alternatively, hypersensitivity could result in avoidance of otherwise beneficial foods (for example, individuals with the phenylthiocarbamide-sensitive version of TAS2R38 are more sensitive to the bitterness of certain vegetables [6]). It is also unclear to what extent the unique lifestyle and history of the Amish impacts the contribution of TAS2R variants to manifestations of dysregulated glucose and insulin homeostasis, including the development of insulin resistance and T2DM. In any case, our studies reveal that bitter taste receptors can influence glucose and insulin homeostasis. The novel role of TAS2Rs in maintenance of glucose homeostasis should help elucidate the relative contributions of taste receptor-

Subjects
The University of Maryland School of Medicine's Institutional Review Board approved all studies. The Amish Family Diabetes Study (AFDS) is an ongoing effort to identify genetic contributors to obesity, diabetes, cardiovascular disease and related disorders [28,39,54]. Detailed descriptions of the population (the Old Order Amish of Lancaster County, Pennsylvania, USA), study design, recruitment methods, phenotypic characterization, clinical characteristics of the subjects and statistical methods have been published previously [28]. Informed consent, including permission to contact relatives, was obtained before participation [28]. In brief, probands were defined as individuals with previously diagnosed diabetes (age of diagnosis between 35 and 65 years). First-and second-degree relatives of the probands were also recruited, as were first-and second-degree relatives of any additional diabetic individuals identified. Currently, the AFDS includes over 1300 subjects. Participants in the AFDS, the Old Order Amish of Lancaster, Pennsylvania, have a common lifestyle and socioeconomic status, and possess detailed genealogical records dating to the period of their early migration from Europe in the 1700's [28].

Genotyping
We identified candidate haplotype tagging SNPs (r 2 $0.8) from the HapMap [55] and additional SNPs in coding and regulatory regions from the Entrez SNP database [56] and from the literature [4,8,57,58]. In total, 70 TAS1Rand TAS2R-associated SNPs were genotyped in the AFDS. Forty-five of these SNPs were polymorphic in the AFDS and passed quality control filters and were subsequently analyzed (see below and Tables 1-3). All SNPs were genotyped using the TaqMan platform (Applied Biosystems) according to manufacturer's protocols. SNPs found to be monomorphic in the AFDS (n = 9) were not analyzed further. Genotypes were checked for Mendelian consistency; inconsistencies, which were detected in ,0.5% of genotypes, were removed from analysis. Genotype frequencies of all SNPs were tested for consistency with Hardy-Weinberg expectations by the x 2 test. Markers that showed extreme deviation from Hardy-Weinberg Equilibrium in controls (P,0.001) were eliminated from further analysis (n = 7), as were SNPs with call rates ,90% (n = 9).

Heterologous expression and functional assay
Receptor expression and functional assays were performed as previously described [20,49]. We used FLIPR (Molecular Devices) to screen the function of TAS2R9 and to establish dose-response curves for the tested compounds (Supplementary data, Table S1). We cloned the cDNAs encoding the TAS2R9 Ala 187 and Val 187 variants into a pEAK10-derived vector (Edge Biosystems, Gaithersburg, MD). The vector was engineered to generate translational fusion to the N-terminus of the rat somatostatin type 3 receptor (45 amino acids), and the C-terminus of the herpes simplex virus (HSV) glycoprotein D epitope, as described [49]. Immunocytochemistry was performed using antiserum against HSV glycoprotein D (Novagen, 1:10,000), as described [49], except the secondary antibody was a FITC-conjugated donkey antiserum against mouse IgG (Molecular Probes, 1:1,000).
Plasmids containing TAS2R9 cDNAs were transiently transfected into HEK293 cells stably expressing the chimeric G protein subunit GAE 16gust44 [59] using TransIT-293 (Mirus Corporation), according to the manufacturer's protocol. Cells were plated into 384-well plates and after 24-30 hr loaded for 1 h with the calcium-sensitive dye Fluo4-AM and stimulated with bitter compounds. Calcium signals were recorded simultaneously from each well after excitation at 488 nm. The obtained signals (F) were normalized to the fluorescence of cells before stimulation (Fo) and expressed as DF/F value: DF/F = (F2Fo) / Fo. Responses of four wells containing cells expressing the same receptor and receiving the same stimulus were averaged. Calculations were based on at least three independent transfection experiments.

Reverse transcription PCR
Total RNA was isolated from human enteroendocrine NCI-H716 cells with Trizol reagent, then reverse transcribed with random hexamer probes. A reaction without reverse transcriptase was included to control for genomic DNA contamination. Human cecum cDNA was obtained from Biochain Institute (Hayward, CA). TAS2R7 (GeneID: 50837) and TAS2R9 (GeneID: 50835) gene specific primers recognized the single coding exons of each gene. TAS1R3 (GeneID: 83756) gene specific primers were directed against exons 4 and 6. All PCR products were verified by sequencing.

GLP-1 assays
Human enteroendocrine NCI-H716 cells were maintained and assayed for GLP-1 secretion as described by Jang et al. [15] in the presence or absence of ofloxacin (Sigma Chemical) at 10, 50 or 100 mM in phosphate buffered saline (PBS), pH 7.2. Control samples received PBS only. GLP-1 was measured by ELISA and normalized to protein content. For siRNA knockdown experi- ments, an a-gustducin-specific siRNA (see [15]) was transfected into subconfluent NCI-H716 cells 48 hr prior to ofloxacin stimulation and GLP-1 secretion analysis. Reduction of agustducin message was verified by quantitative real time PCR. The efficacy of the stimulation was significantly reduced after knockdown of the G protein a-gustducin by RNA interference ( Figure S1B, C), indicating that ofloxacin-dependent GLP-1 secretion is mediated by a G protein-coupled receptor.

Statistical Analysis
Associations with SNP genotype and the various phenotypes were performed using pedigree-based analysis by regressing the effect of the marker genotype while accounting for residual familial correlations among related individuals using age, sex, and body mass index (BMI) as covariates (age and BMI are positively correlated with T2DM in the AFDS). To account for the relatedness among family members, we employed the measured genotype approach, in which we estimated the likelihood of specific genetic models given the pedigree structure. Parameter estimates were obtained by maximum likelihood methods and the significance of association was tested by likelihood ratio tests. When discrete outcome traits were analyzed, a threshold model was assumed. All analyses of the AFDS were carried out using the Sequential Oligogenic Linkage Analysis Routines (SOLAR) software program [60]. In the T2DM case/control analysis, a recessive genetic model was assumed. When analyzing data from non-diabetic AFDS subjects, a dominant model was assumed. To control for an inflation in the type I error rate due to the number of comparisons in our initial T2DM association analyses, we use the P ACT statistic [61], which attains the accuracy of permutation or simulation-based correction through the adjustment of correlated p-values. Unadjusted P values are reported in all tables.
Pairwise LD between the SNPs and haplotype block analysis was computed using Haploview 4.0 [62]. Haplotype blocks were defined by 95% confidence bounds on D' [63]. Concentrationresponse curves and EC50 values derived from the heterologous expression and functional assays were calculated in SigmaPlot by nonlinear regression.