Activation of airway epithelial bitter taste receptors by Pseudomonas aeruginosa quinolones modulates calcium, cyclic-AMP, and nitric oxide signaling

Bitter taste receptors (taste family 2 bitter receptor proteins; T2Rs), discovered in many tissues outside the tongue, have recently become potential therapeutic targets. We have shown previously that airway epithelial cells express several T2Rs that activate innate immune responses that may be important for treatment of airway diseases such as chronic rhinosinusitis. It is imperative to more clearly understand what compounds activate airway T2Rs as well as their full range of functions. T2R isoforms in airway motile cilia (T2R4, -14, -16, and -38) produce bactericidal levels of nitric oxide (NO) that also increase ciliary beating, promoting clearance of mucus and trapped pathogens. Bacterial quorum-sensing acyl-homoserine lactones activate T2Rs and stimulate these responses in primary airway cells. Quinolones are another type of quorum-sensing molecule used by Pseudomonas aeruginosa. To elucidate whether bacterial quinolones activate airway T2Rs, we analyzed calcium, cAMP, and NO dynamics using a combination of fluorescent indicator dyes and FRET-based protein biosensors. T2R-transfected HEK293T cells, several lung epithelial cell lines, and primary sinonasal cells grown and differentiated at the air–liquid interface were tested with 2-heptyl-3-hydroxy-4-quinolone (known as Pseudomonas quinolone signal; PQS), 2,4-dihydroxyquinolone, and 4-hydroxy-2-heptylquinolone (HHQ). In HEK293T cells, PQS activated T2R4, -16, and -38, whereas HHQ activated T2R14. 2,4-Dihydroxyquinolone had no effect. PQS and HHQ increased calcium and decreased both baseline and stimulated cAMP levels in cultured and primary airway cells. In primary cells, PQS and HHQ activated levels of NO synthesis previously shown to be bactericidal. This study suggests that airway T2R-mediated immune responses are activated by bacterial quinolones as well as acyl-homoserine lactones.

The immune system has been referred to as the "sixth sense" (1)(2)(3)(4)(5) because it detects invading pathogens similarly to the way the other five senses (touch, taste, smell, sight, and hearing) perceive the external environment. Fitting this viewpoint, we and others have demonstrated that the immune and taste systems both utilize some of the same chemosensory receptors. Bitter taste receptors (also known as a taste family 2 receptors, or T2Rs) 2 are G protein-coupled receptors (GPCRs) used by the tongue to detect compounds that are perceived as bitter (6,7). There are 25 T2R isoforms on the human tongue, and many isoforms are also expressed in organs such as the brain, nose and sinuses, lung, and reproductive tract (6 -10). The functions of these so-called "extraoral" taste receptors and their physiological ligands in many tissues are largely unknown, although they may be responsible for some off-target effects of many commonly used medications that taste bitter (e.g. aspirin) (11).
T2R38 is expressed in the motile cilia of cells of the upper airway (nose and sinuses), and it is activated in response to acyl-homoserine lactone (AHL) quorum-sensing molecules secreted by Gram-negative bacteria (12), including the common airway pathogen Pseudomonas aeruginosa. T2R38 activation causes an increase in intracellular free Ca 2ϩ concentration ([Ca 2ϩ ] i ) that activates nitric-oxide synthase (NOS) to produce nitric oxide (NO) (9,(12)(13)(14)(15)(16)(17). NO activates production of cGMP to activate protein kinase G, which increases ciliary beating and enhances mucociliary clearance, the major physical defense of the airway. The T2R-activated NO also directly diffuses into the airway surface liquid, where it has antibacterial effects (12,18), probably through damage to bacterial cell walls and/or DNA (19,20).
The relevance of T2R38 to immunity is supported by clinical data linking polymorphisms in the TAS2R38 gene to sinonasal disease (16,21). Patients homozygous for the AVI TAS2R38 polymorphism, which results in nonfunctional T2R38 protein (22), are more susceptible to Gram-negative bacterial infection (12), have higher levels of sinonasal bacteria (23,24) and biofilms (25), are at higher risk for chronic rhinosinusitis (26 -29), and may have worse outcomes after endoscopic sinus surgery (30) compared with patients homozygous for the functional (PAV) allele of TAS2R38.
We subsequently identified other T2R isoforms, T2R4, T2R14, and T2R16, in sinonasal cilia (31,32). These isoforms are also in bronchial cilia (10). Activation of these T2Rs in sinonasal cells by known bitter compounds produces similar Ca 2ϩ and NO responses. We hypothesize that yet-unknown pathogen-produced products activate these other T2Rs in addition to AHL activation of T2R38. One such class of compounds may be quinolones, which are involved in quorum sensing in bacteria.
Activation of host T2Rs by bacterial quinolones may play a role in innate immune detection of these bacteria. Here, we tested responses of human T2Rs to three bacterial quinolones by live-cell imaging of HEK293T cells heterologously expressing known airway T2Rs and bronchial epithelial cell lines that endogenously express T2Rs. We also tested responses in primary human sinonasal epithelial cells cultured at the air-liquid interface (ALI), a standard model for recapitulating a differentiated airway epithelium in vitro (61)(62)(63).
We screened PQS, DHQ, and HHQ against HEK293T cells heterologously expressing several human T2Rs plus a G-protein chimera, G␣ 16 -gust44, that couples the receptors to robust [Ca 2ϩ ] i signals. This is a common model system used for determining whether compounds activate specific T2Rs (45,46) and is frequently used in other GPCR drug discovery assays (65,66). We focused primarily on the T2Rs expressed in sinonasal cilia (10,12,31,32). Cells transfected with G␣ 16 -gust44 alone were used as a negative control. PQS (100 M) activated several T2Rs ( Fig. 1 (b and c) and Fig. S1). Responses were larger when only responsive cells (probably those that were transfected) were selected (Fig. S1), but we chose to determine the statistical significance of the response based on the entire population ( Fig. 1) to reduce any potential biasing. The most robust Ca 2ϩ responses were with T2R4 and T2R16, but some activation of T2R38 was also observed. HHQ significantly activated only T2R14, whereas DHQ activated no T2Rs examined here (Fig. 1,  d and e). No effects were observed with the highest concentration of vehicle used in these experiments (0.1% DMSO) (Fig. S2) (31).
We tested the dose dependence of PQS against T2R4 (Fig. 2,  a and b) and found a statistically significant elevation of Ca 2ϩ as low as 10 M (10 Ϫ5 M). Whereas [Ca 2ϩ ] i responses to PQS were lower than those observed with ATP or quinine (Fig. 2b), we did not reach receptor saturation at the maximal concentrations (100 -200 M) that could be achieved in aqueous solution. This is common in studies of T2R agonists (45)(46)(47)67), which are often highly hydrophobic. PQS solubility may be enhanced during infections by P. aeruginosa production of rhamnolipid surfactants (68). Nonetheless, these data sup-Bitter taste receptors detect bacterial quinolones port detection of PQS by T2R4-transfected cells at concentrations reported to be produced in P. aeruginosa cultures (20 -30 M) (48).
To confirm that T2R4 itself is a PQS receptor and that PQS is not sensed by another receptor produced by HEK293T cells in response to T2R4 transfection, we tested the effects of PQS in the presence of GABA, a T2R4 blocker (69). Responses to PQS, but not the purinergic receptor agonist ATP, were diminished in T2R4-expressing cells in the presence of GABA (Fig. 2c). We similarly tested the multidrug resistance transporter inhibitor probenecid, an allosteric inhibitor of T2R16 (70). Probenecid pretreatment (1 mM for 1 h, as described (70)) reduced T2R16dependent [Ca 2ϩ ] i responses to both PQS and T2R16-agonist salicin, but not purinergic receptor agonist ATP (Fig. 2d).
Ca 2ϩ elevation in response to PQS stimulation depended on ER Ca 2ϩ release, as it was absent in BEAS-2B cells after Ca 2ϩ store depletion by either treatment with the SR/ER Ca 2ϩ ATPase inhibitor thapsigargin in Ca 2ϩ -free (0-Ca 2ϩ ) extracellular solution (no added Ca 2ϩ ϩ 2 mM EGTA) (Fig. S3a) or purinergic stimulation with ATP in Ca 2ϩ -free extracellular solution (Fig. S3b). Pretreatment of BEAS-2B cells with the IP 3 receptor inhibitor xestospongin C (74) likewise blocked the PQS [Ca 2ϩ ] i response (Fig. S3c). We also treated H292 cells with the G␤␥ inhibitor gallein (75) (Fig. S3d), which was shown to block bitter agonist signaling in smooth muscle in some studies (76, 77) but not another (78). Gallein significantly reduced

Bitter taste receptors detect bacterial quinolones
the [Ca 2ϩ ] i elevation with PQS ( Fig. S3d), supporting a role for G␤␥ signaling in this response.
To confirm the Ca 2ϩ responses to quinolones observed with small-molecule Ca 2ϩ indicators, we transfected BEAS-2B cells with a genetically encoded green Ca 2ϩ -sensitive protein, GCaMP6s (79). BEAS-2B cells exhibited an increase in GCaMP6s fluorescence, signaling an increase in Ca 2ϩ , in response to PQS, HHQ, and the T2R10/14 agonist thujone (45) (Fig. S4, a-c). The PQS Ca 2ϩ response observed with GCaMP6s was likewise inhibited by U73122 (Fig. S4d), supporting a requirement for receptor-mediated activation of PLC.
Because HHQ appeared to activate only one known airwayexpressed T2R isoform, T2R14, we examined the HHQ response in T2R14-expressing (73) H292 cells co-transfected with GCaMP6s (to identify transfected cells and measure Ca 2ϩ responses) and two different sets of two shRNA constructs directed against T2R14. HHQ stimulated robust Ca 2ϩ responses in H292 cells transfected with GCaMP6s only (Fig.  4a) or GCaMP6s ϩ scramble shRNA (Fig. 4b). Both sets of distinct T2R14-directed shRNAs markedly reduced the response (Fig. 4, c and d), whereas the purinergic or proteaseactivated receptor 2 (PAR-2) GPCR responses were not significantly affected by these shRNA vectors (Fig. 4e). These results support a role for endogenous T2R14 in detection of HHQ by lung epithelial cells.

PQS and HHQ decrease baseline and stimulated cAMP levels in lung epithelial cells
The other arm of the canonical T2R signaling pathway ( Fig.  3a) is G␣ gust , a G␣ protein of the transducin family (80 -83) which activates phosphodiesterase (PDE). Several airway cell lines have been reported to express G␣ gust (72,84), although G␣ i isoforms, which inhibit adenylyl cyclases, couple to T2Rs in airway smooth muscle (85). To test whether quinolones activated a G␣ gust /G␣ i pathway, we monitored cAMP dynamics in real time in living cells using a high-affinity fourth-generation Turquoise (CFP)/Venus (YFP) FRET-based EPAC cAMP indicator, EPAC-S H187 (86). This construct was previously used to image a decrease in baseline cAMP levels in HEK293T cells expressing G␣ i/o -coupled -opioid receptors (87). BEAS-2B cells transfected with EPAC-S H187 and stimulated with PQS or quinine exhibited decreases in baseline CFP/FRET emission ratio, signaling decreases in cAMP concentration (Fig. 5a). The adenylyl cyclase activator forskolin was used as a control. PQS also blunted cAMP increases during the simultaneous addition of the G␣ s -coupled ␤-adrenergic receptor agonist isoproterenol (100 M; Fig. 5b). Similar PQS-induced decreases in baseline cAMP were observed in A549 cells (Fig. 5c). When we tested HHQ and DHQ in H292 cells, we observed a decrease in baseline cAMP with HHQ but not DHQ (Fig. 5e), as would be expected if HHQ activates T2Rs and DHQ does not. PQS-mediated cAMP decreases were not observed when BEAS-2B cells

Bitter taste receptors detect bacterial quinolones
were treated with IBMX, a phosphodiesterase inhibitor ( Fig.  5f), suggesting G␣ gust activation of phosphodiesterase. A similar result was observed in H292 cells (Fig. S5).
To confirm that this change in EPAC construct fluorescence was due to activity of a G␣ gust /G␣ o /G␣ i protein, we treated BEAS-2B cells with pertussis toxin (PTX; 500 ng/ml; 18 h), which ADP-ribosylates and inactivates G␣ gust , G␣ o , and G␣ i subunits (76,88,89). PTX reduced the cAMP decrease observed with PQS and HHQ (Fig. 6, a and b), suggesting a role for a G␣ gust /G␣ o /G␣ i -linked receptor. We co-transfected BEAS-2B cells with EPAC-S H187 and the G␣ 16 /gust44 chimera used in HEK293T cells above. G␣ 16 is a G␣ q , which activates PLC instead of PDE, and the C-terminal G␣ gust portion will bind to T2Rs (45,(65)(66)(67). We hypothesized that overexpression of G␣ 16 /gust44 would out-compete endogenous G␣ gust or G␣ i proteins and reduce cAMP decreases. BEAS-2B cells cotransfected with WT G␣ gust and EPAC-S H187 exhibited PQSstimulated cAMP decreases similar to cells transfected with
To confirm the effects of quinolones on cAMP, we utilized another cAMP indicator, Flamindo2 (90). BEAS-2B cells transfected with Flamindo2 exhibited a PQS dose-dependent inhibition of isoproterenol-induced cAMP increases, evidenced by a decrease in Flamindo2 fluorescence (Fig. S6). The cAMP increase activated by 100 nM isoproterenol was inhibited dosedependently by PQS (1-100 M; Fig. S6, a and b), with the rate of Flamindo2 fluorescence change inhibited ϳ10-fold by 100 M PQS (Fig. S6c). Moreover, in H292 cells transfected with Flamindo2, HHQ (100 M) inhibited cAMP increases in response to 100 nM isoproterenol (Fig. S7). The effect of HHQ was substantially reduced when cells were co-transfected with shRNAs directed against T2R14 but not scrambled shRNA (Fig. S7).
The data above suggest that PQS and HHQ activate Ca 2ϩ through G␤␥ activation of PLC and decrease cAMP through a PTX-sensitive G␣, as would be expected during T2R stimulation and activation of G␣ gust . To test whether this change in baseline cAMP affects baseline protein kinase A (PKA) activity, we transfected BEAS-2B cells with a PKA-sensitive CFP/YFP construct, AKAR4 (91). BEAS-2B cells exhibit baseline PKA activity even in the absence of any stimulation, evidenced by the ability of the PKA inhibitor H89 to reduce AKAR4 FRET/CFP emission ratio (Fig. 7a), signifying a decrease in PKA activity. Even when cells were stimulated with forskolin to increase PKA activity, H89 was able to bring PKA activity back to a level below the initial baseline (Fig. 7a). PQS also caused a drop in baseline PKA activity (Fig. 7b). As a control, we expressed the protein kinase C-sensitive CFP/YFP construct CKAR (92) in BEAS-2B cells and observed no change in CKAR fluorescence despite a change with the diacylglycerol mimetic phorbol 12-myristate 13-acetate (PMA) (Fig. 7c). This suggests that PQS does not stimulate protein kinase C activation (not previously reported for taste receptors) but also that quinolones do not have nonspecific effects on the CFP or YFP variants in these fluorescent constructs.

Bacterial quinolones activate Ca 2؉ signals and decrease cAMP levels in primary sinonasal epithelial ALI cultures
We isolated primary sinonasal epithelial cells from residual surgical material and cultured them at the ALI, which causes differentiation into ciliated and goblet cells and mimics the in vivo airway epithelium (61). The ciliated cells in sinonasal and bronchial ALIs express T2R4, -14, -16, and -38 similarly to tissue explants (10, 12, 31, 32). We previously showed that T2R14 and -38 are co-localized closely enough to observe FRET when

Bitter taste receptors detect bacterial quinolones
primary antibodies were labeled with Alexa Fluor (AF)-conjugated Fab fragments (31). In contrast, T2R4 is localized to only the distal tips of motile cilia (10, 31) (Fig. 8, a and b).
We examined the colocalization of cilia T2R16 and T2R38 (Fig. S8a) using the same method used previously for T2R38 and T2R14 (31) with antibodies directed against the C terminus of both proteins. FRET was observed between T2R16 and T2R38 when both primary antibodies were labeled with either a donor (AF555) or acceptor (AF647) fluorophore (Fig. S8b). Acceptor bleaching increased donor fluorescence (Fig. S8c) 20 Ϯ 1% when T2R38 was used as the donor and T2R16 as the acceptor and 22 Ϯ 1% with donor T2R16 and acceptor T2R38directed antibodies (n.s.; Fig. S8d). Donor fluorescence increase with acceptor bleaching is roughly equivalent to FRET efficiency. When the two groups pooled, mean FRET efficiency was 21 Ϯ 1%, similar to the efficiency (16 Ϯ 2%) observed previously with T2R38 and T2R14 (31). The Förster radius (R o ) is the distance at which FRET efficiency is 50%, which is 51 Å for this FRET pair. Efficiency (E) decreases to the 6th power as distance (R) increases. Thus, R ϭ R o ⅐((1 Ϫ E)/E) 1 ⁄ 6 . With E ϭ 0.21, R ϭ 51 Å⅐((1 Ϫ 0.21)/0.21) 1 ⁄ 6 ϭ 64 Å. Thus, when bound to the C terminus of T2R38 and T2R16, our labeled antibodies are on average 6.4 nm apart. We cannot conclude from these or prior (31) data that T2R38, -16, or -14 forms heterodimers due to the sizes of the antibodies involved, but similar antibodybased measurements have been used to suggest a "close association" of endogenously expressed proteins (93).
Stimulation of T2Rs in sinonasal cilia activates small but sustained Ca 2ϩ responses (12,14,31). We observed similar Ca 2ϩ responses during PQS stimulation of sinonasal ALIs (Fig. 8, c  and d). As with T2R38, the sustained phase of the Ca 2ϩ response required Ca 2ϩ influx, as the signal decayed over the course of 3-4 min in the absence of extracellular Ca 2ϩ (0-Ca 2ϩ solution containing 2 mM EGTA) (Fig. 8e). Once differentiated, ALIs are difficult to transfect even with viral vectors, and thus we could not use EPAC-S H187 or AKAR4 in primary cells. However, we measured cAMP during forskolin and isoproterenol stimulation using a high-sensitivity cAMP ELISA and found that co-stimulation with PQS (10 -100 M) decreased stimulated cAMP (Fig. 9a), and this effect was blocked by PTX pretreatment (Fig. 9b), fitting with PQS activation of a G␣ gust or G␣ i . Because baseline cAMP levels were below the limit of detection of this ELISA, we could not quantify baseline cAMP changes using this method.

P. aeruginosa quinolones activate NO production in primary sinonasal ALI cultures
A hallmark of T2R stimulation in primary differentiated sinonasal ciliated cells is activation of Ca 2ϩ -dependent NOS (probably endothelial NOS) (12,14,31,32,87). NO can damage bacterial cell walls and DNA (19,20). We showed that airway T2R38-mediated NO production is bactericidal against P. aeruginosa (12,14,18). Apical application of PQS activated an increase in fluorescence of ALIs loaded with the fluorescent indicator DAF-FM (Fig. 10, a and b), which exhibits an increase in fluorescence quantum yield when it reacts with reactive nitrogen species (RNS) like NO and its reactive derivatives. There was an additive effect when PQS was combined with HHQ and DHQ (Fig. 10, a and b), and similar DAF-FM fluorescence increases were observed with quinine ( Fig. 10, a and b). RNS typically react to generate NO 2 Ϫ or NO 3 Ϫ products. We used a fluorometric assay based on the Greiss reaction to measure total NO 2 Ϫ ϩ NO 3 Ϫ in fluid overlaying the apical surface of sinonasal ALIs stimulated for 18 h with the T2R4/14 agonist quinine, T2R38 agonist PTC, T2R14 agonist flufenamic acid (FFA), or PQS. Cultures were also genotyped for TAS2R38 PAV (functional) versus AVI (nonfunctional) polymorphism status. NO 2 Ϫ /NO 3 Ϫ levels were increased in cultures homozygous for the PAV haplotype (PAV/PAV) stimulated with all T2R agonists (Fig. 10c). AVI/AVI cultures also exhibited NO increases to all agonists except PTC (Fig. 10c). Whereas the PQS response appeared to be slightly lower, it was not statistically significantly different from PAV/PAV cultures (Fig. 10c). However, when AVI/AVI cultures were stimulated with PQS in the presence of GABA (to block T2R4) and probenecid (to block T2R16), NO production was significantly reduced (Fig. 10d). In the presence of GABA and PQS, we noted a significant difference between NO production between PAV/PAV and AVI/ AVI cultures (Fig. 10e). Thus, under conditions of T2R4 and T2R16 blockade, the response is altered by T2R38 functionality, strongly suggesting a role for T2R receptors in this response.

Discussion
The data here support the hypothesis that T2Rs can detect bacterial quinolones and that this may play a role in innate immune signaling in airway epithelia. This supports a role for T2Rs in mediating interkingdom signaling through mammalian cell detection of bacterial metabolites. We used heterologous expression and multiple airway cell lines to quantitatively image PQS and HHQ responses using fluorescent ion indica-

Bitter taste receptors detect bacterial quinolones
tors and FRET-based protein biosensors in real time in living cells. This was combined with quantification of responses in primary cells to validate that the effects observed were not artifacts of one or more cell lines.
In contrast to the BitterX (64) prediction of HHQ as "not bitter," both PQS and HHQ activated different cilia-localized T2Rs, cautioning that experimental testing is always needed to confirm/refute results of computer predictions or modeling. Testing of PQS, HHQ, and DHQ against all 25 T2Rs in future higher-throughput studies will determine the receptors most sensitive for these compounds. We focused on known airway cilia T2Rs to test whether PQS and HHQ detection by T2Rs may be important for epithelial immune surveillance.
Quinine taste sensitivity varies with polymorphisms in a bitter receptor cluster on chromosome 12 (94); T2R14 is located on chromosome 12, whereas T2R4, -16, and -38 are located on chromosome 7. It remains to be determined whether these chromosome 12 polymorphisms affect sensitivities to HHQ or PQS. However, a clinical study recently reported that chronic rhinosinusitis patients with nasal polyps rated quinine as less bitter than control subjects in taste tests, suggesting that a reduced functionality of quinine-responsive T2Rs may play a role in susceptibility to chronic rhinosinusitis (95). Although further clinical investigation is needed, the results presented here suggest that a reduced response to bacterial quinolones may be a mechanism contributing to impaired immunity in patients less sensitive to quinine (95). Taste testing may someday have clinical diagnostic utility in chronic rhinosinusitis or other types of infectious diseases (96).
It is not surprising that bacterial quinolones activate both T2R14 and T2R4, as both are activated by quinine in vitro. T2R14 is a receptor with a broad range of activating compounds (97). Whereas it was somewhat surprising that PQS activates T2R4 but not T2R14 and HHQ activates T2R14 but not T2R4, the quinolone chloroquine activates T214 but not T2R4. Thus, there is precedent for differential quinolone recognition

Bitter taste receptors detect bacterial quinolones
between the two isoforms. It is most surprising that PQS activated T2R38 and T2R16, albeit to a lesser extent than T2R4. T2R38 and T2R16 are considered to be T2Rs with relatively specific types of activating molecules. T2R38 typically recognizes thiolactones (98), and T2R16 typically recognizes ␤-glycosides (99). However, exceptions to this specificity have been described. For example, T2R38 is also activated by the alkaloid yohimbine, and T2R16 is activated by the antimuscarinic diphenidol (45), so activation of these two receptors by structurally diverse compounds is not without precedent. Because of the wide range of cilia-localized T2Rs that can be stimulated with quinolone compounds (e.g. PQS), quinolones might be useful therapeutics to target this pathway in patients.
Importantly, quinolone T2R responses are not due to DMSO vehicle. A recent publication (100) erroneously reported that our previous studies of AHLs (12,14) used 3% DMSO. This study found that T2R38 expressed in HEK293T cells is a receptor for many small organic bacterial molecules at high concentrations (Ͼ1%), and concluded that prior AHL results were an artifact of 3% DMSO activation of T2R38 (100). However, in all studies, we have consistently utilized 1000ϫ DMSO stock solutions (100 mM) of AHLs and now quinolones, making final DMSO concentrations 0.1% or up to 0.2% if two DMSO-dissolved compounds are added together. We have repeatedly found no effects of these DMSO levels on airway cell Ca 2ϩ or NO signaling (12,14,15,31,101,102), including here. Moreover, other studies have supported a specific role for T2R38 in AHL detection (103-106) and also found no activation of endogenous T2R38 by 0.1% DMSO (107), as used here.  Asterisks represent significance compared with control (unstimulated) conditions), and number signs represent significance between the bracketed groups. Significance was determined by one-way ANOVA with Bonferroni post-test. d and e, bar graphs of PAV/PAV and AVI/AVI cultures stimulated with PQS in the presence or absence of GABA and probenecid (pretreatment as above) from separate sets of experiments; n ϭ 6 -10 ALIs/condition. Significance was determined by Student's t test. *, p Ͻ 0.05; **, p Ͻ 0.01; ##, p Ͻ 0.01.

Bitter taste receptors detect bacterial quinolones
We demonstrate for the first time, to our knowledge, that endogenous T2R activation can have direct effects on baseline cAMP levels and baseline PKA activity in intact airway epithelial cells. It was previously shown that the bitter compound denatonium lowers forskolin-induced cAMP levels in bronchial cells (84). Using state-of-the-art, high-dynamic range, high-affinity FRET indicators (86,90), we show here that T2R activation by PQS lowers basal unstimulated cAMP levels in addition to stimulated cAMP levels. This may have implications for ion transport in the airway, as the apical cystic fibrosis transmembrane conductance regulator (CFTR) chloride channel is activated by PKA phosphorylation in response to cAMP (108). Future work is needed to understand how T2R activation may affect CFTR function at baseline or during concomitant stimulation of cAMP-elevating pathways.
In addition to P. aeruginosa, Burkholderia species also produce and respond to quinolones (109). Coupled with previous data showing that T2Rs respond to AHLs (12,46,103,104), the data here support a role for T2Rs similar to other established immune pattern recognition receptors (PRRs), which recognize conserved pathogen-associated molecular patterns that signal the presence of infection. We hypothesize that at least some T2Rs exist in myriad extraoral tissues to detect bacterial and possibly fungal metabolites and activate defensive responses, making them a novel type of immune PRRs.
It is potentially important that both of the known bacterial products that activate airway cilia T2Rs (AHLs, PQS, and HHQ) are metabolites of Gram-negative bacteria. A recent study found that P. aeruginosa are more sensitive to the bactericidal effects of NO than Gram-positive Staphylococcus or the fungus C. albicans (18). Airway ciliated cells are probably the highest sites of NOS activity in the airway epithelium (110 -116), and thus ciliated cells may have evolved to express T2Rs that detect Gram-negative metabolites so as to activate this specific defensive pathway in response to the bacteria that are most sensitive to its effects.
PQS induces secretion of IL-8 and IL-6 in human mesenchymal stem cells (117) and stimulates chemotaxis of polymorphonuclear neutrophils (118). Longer-term (4 -24-h) stimulation with PQS was recently shown to induce oxidative stress and reduce heme oxygenase-1 expression in A549 cells and J774A.1 and THP-1 macrophage-like cell lines (55). These mechanisms may involve T2Rs. Neutrophils express T2R38 (103, 104) that would probably detect PQS. However, it will be difficult to tease out the relative importance of the PQS and AHL to infection and immune sensing in vivo, as the PQS and AHL pathways are intimately intertwined (119,120). At least one of the genes required for PQS synthesis is controlled by the N-3-oxododecanoyl-L-homserine lactone-activated LasR pathway in P. aeruginosa (121). Studies have also reported differential positive regulation of PQS by the LasR and negative regulation of PQS by the N-butyryl-L-homoserine lactone-activated RhlR pathways (35,122).
Further work is needed to define other bacterial and/or fungal compounds that activate T2Rs. We previously showed that functional T2R38 is necessary for airway cell NO responses to AHLs. A recent study reported that T2R10 and T2R14, not T2R38, respond to some AHLs when expressed in HEK293 cells (46). The reason for this discrepancy is not yet clear, as other studies have shown that T2R38 detects AHLs (103)(104)(105)(106). We hypothesize that cloned receptor polymorphisms and/or different methodologies (microscopy versus plate reader assays) may contribute to some differential results observed. Discrepancies in agonist activation extend beyond AHLs (e.g. epigallocatechin gallate for T2R14 and genistein for T2R39 (46)). Receptor interactions and/or dimerization may also occur in vivo and alter sensitivities. GPCR dimerization frequently has important functional consequences (123,124). Whereas studies of T2R-T2R interactions so far suggest minimal effect of T2R-T2R dimerization partners (125,126), these studies have been carried out in heterologous overexpression systems utilizing receptors with tags to help them traffic to the plasma membrane. We believe that, although heterologous expression studies are tremendously useful, such results must be combined with studies of endogenous receptors to better understand T2R signaling in vivo. An important example is the recent work demonstrating that T2R14 can heterodimerize with ␤-adrenergic receptors, important for trafficking to the plasma membrane in airway smooth muscle (73). Continued studies of endogenous T2R signaling are needed to shed light on the cell biology of these receptors and their role in innate immunity and other processes.

Reagents and solutions
Unless indicated, all reagents and solutions used were as described previously (12,14,15,101,127). Unless indicated below, all other reagents were from Sigma-Aldrich. Fura-2-acetoxymethyl ester (AM), fluo-4-AM, and DAF-FM diacetate were from Thermo Fisher Scientific (Waltham, MA). Anti-␤-tubulin IV (ab11315; mouse monoclonal), anti-T2R38 (ab130503; rabbit polyclonal), anti-T2R16 (ab75106; rabbit polyclonal), xestospongin C, thapsigargin, and 2-furoyl-LIGRLO-amide (2FLI) were from Abcam (Cambridge, MA). Anti-T2R4 (T13, sc169494; goat polyclonal) was from Santa Cruz Biotechnology, Inc. (Dallas, TX). Stock solutions of PQS, HHQ, and DHQ were made at 100 mM in DMSO (Ն1000ϫ). PQS has poor aqueous solubility that may be enhanced in vivo by bacterially produced biosurfactant rhamnolipids (68). We noted a precipitation and loss of activity of PQS and HHQ after ϳ30 min in aqueous solution, so working solutions were diluted immediately before use for each experiment with vigorous vortexing (Ͼ90 s) before the addition of a 2ϫ working solution (200 M PQS or HHQ for most experiments) of the compound into a well of a chambered coverglass (CellVis, Mountain View, CA) containing cells and an equal volume of Hanks' balanced salt solution (HBSS).

HEK293T cell culture and transfection
HEK293T cells (ATCC) were cultured in high-glucose DMEM (Gibco) plus 10% FBS. 24 h before transfection, cells were plated on glass coverslips coated with poly-D-lysine. Transfection was carried out with Lipofectamine 2000 according to the manufacturer's instructions. Each T2R was co-transfected with pcDNA3.1 containing G␣ 16 -gust44, a chimeric G␣ protein containing G␣ 16 (a G␣ q ) fused to the last 44 amino acids of gustducin, as described previously (65,66,128). Cells were used 24 -48 h after transfection.

Culture of human bronchial cell lines
A549, BEAS-2B, and NCI-H292 cells were obtained from ATCC and cultured in Ham's F12K (Gibco) plus 10% fetal bovine serum and 1% penicillin/streptomycin solution. All cells were used between passages 5 and 25 from receipt. Cells were seeded onto glass coverslips and transfected with EPAC-based cAMP indicator mTurq2⌬_Epac(CD,⌬DEP, Q270E)_td cp173 Ven (also known as EPAC-S H187 (86)), PKA indicator AKAR-4 (91), Ca 2ϩ indicator GCamP6S (79), or cAMP indicator Flamindo2 (90) using Lipofectamine 3000 according to the manufacturer's instructions. FRET constructs were imaged using a CFP/YFP ratiometric filter set (Chroma Technologies, Bellows Falls, VT) on an IX-83 microscope (Olympus Life Sciences, Tokyo, Japan) equipped with a 30ϫ (1.05 NA) silicone oil immersion objective as described previously (31,127). NCI-H292 cells co-transfected with GCaMP6s, Flamindo2, or EPAC-S H187 and shRNAs were imaged 72 h after transfection using Micro-Manger/ ImageJ (129) on an inverted microscope (Nikon, Tokyo Japan) with a 10ϫ (0.45 NA) objective, XCite 110 LED light source (Excelitas Technologies, Waltham, MA), GFP filter cube (Chroma), and Photometrics Cool Snap HQ cooled interline CCD camera (Roper Scientific, Tucson, AZ). Due to the low transfection efficiencies of more differentiated epithelial cells (Ͻ10 -15% for H292) and the fact that we were imaging responses in single transfected cells identified by fluorescent co-transfected proteins, we did not quantify the level of knockdown, as the population average would grossly underestimate knockdown due to the high percentage of untransfected cells.

Generation of primary sinonasal ALI cultures
All procedures were performed in accordance with the University of Pennsylvania School of Medicine guidelines regarding residual clinical material in research, the United States Department of Health and Human Services code of federal regulations Title 45 CFR 46.116, and the Declaration of Helsinki. Patients undergoing medically indicated sinonasal surgery were recruited from the Department of Otorhinolaryngology at the University of Pennsylvania with full institutional review board approval (approval no. 800614) and written informed consent. Inclusion criteria were patients Ն18 years of age undergoing sinonasal surgery for sinonasal disease (chronic rhinosinusitis) or other procedures (e.g. trans-nasal approaches to the skull base), in which tissue was classified as "control." Exclusion criteria included history of systemic inheritable disease (e.g. granulomatosis with polyangiitis, cystic fibrosis, systemic immunodeficiencies) or use of antibiotics, oral corticosteroids, or anti-biologics (e.g. Xolair) within 1 month of surgery. Members of vulnerable populations (e.g. individuals Յ18 years of age, pregnant women, cognitively impaired persons) were not included. Human sinonasal epithelial cells were enzymatically dissociated and grown to confluence in proliferation medium consisting of 50% Dulbecco's modified Eagle's medium/Ham's F-12 plus 50% bronchial epithelial basal medium (Lonza, Walkersville, MD) for 7 days (12,130). Cells were then dissociated and seeded on porous polyester membranes coated with BSA, type I bovine collagen, and fibronectin in cell culture inserts. Culture medium was removed from the upper compartment, and basolateral medium was changed to differentiation medium (1:1 Dulbecco's modified Eagle's medium/bronchial epithelial basal medium) containing human epidermal growth factor (0.5 ng/ml), epinephrine (5 ng/ml), bovine pituitary extract (0.13 mg/ml), hydrocortisone (0.5 ng/ml), insulin (5 ng/ml), triiodothyronine (6.5 ng/ml), and transferrin (0.5 ng/ml), supplemented with 100 units/ml penicillin, 100 g/ml streptomycin, 0.1 nM retinoic acid, and 2% NuSerum (BD Biosciences, San Jose, CA) as described previously (12,130).

Live cell imaging of [Ca 2؉ ] i and RNS production
All imaging experiments were carried out in HEPES-buffered HBSS. [Ca 2ϩ ] i and RNS were imaged using Fura-2 or Fluo-4 and DAF-FM, respectively, as described previously (12,14,15,31,131). For Ca 2ϩ imaging of submerged airway cell lines, cells seeded on coverslips were loaded with fura-2/AM (2 M) for 45 min at room temperature followed by a 20-min incubation in the dark. HEK293T cells were loaded with fura-2/AM (5 M) for 75 min at room temperature. Primary sinonasal ALIs were loaded with fura-2/AM (5 M applied apically) for 90 min, followed by washing and a 20-min incubation in the dark. ALIs were similarly loaded with 10 M DAF-FM diacetate for 90 min in the presence of 5 M carboxy-PTIO, followed by washing to Bitter taste receptors detect bacterial quinolones remove unloaded DAF-FM and cPTIO and incubation for 15 min before imaging. Because of potential toxicity issues related to AM-ester-based dyes (e.g. formaldehyde release from de-esterification, reactive oxygen species production from illumination, Ca 2ϩ chelation), care was taken to minimize dye loading as well as sample illumination using neutral density filters to minimize excitation light, increasing camera sensitivity with pixel binning, and minimizing sampling frequency (1 frame every 4 or 8 s for most experiments) using established loading and illumination protocols (12,31,101,108,(132)(133)(134).
Imaging of ALIs was performed using an Olympus IX-83 microscope (10ϫ, 0.4 NA PlanApo objective) equipped with a fluorescence xenon lamp (Sutter Lambda LS, Sutter Instruments, Novato, CA), excitation and emission filter wheels (Sutter Instruments), and a 16-bit Hamamatsu Orca Flash 4.0 sCMOS camera. Images were acquired and analyzed using MetaFluor (Molecular Devices, Sunnyvale, CA). For fura-2 [Ca 2ϩ ] i measurement in submerged cells, either a 20ϫ (0.8 NA; UPLNAPO) air or 30ϫ (1.05 NA; UPLSAPO) silicone immersion objective (Olympus) using MetaFluor. Due to potential errors and/or enhanced variability (e.g. (135)) introduced with calibration of ratios to [Ca 2ϩ ] i and the fact that the comparisons made here did not necessitate absolute measure of [Ca 2ϩ ] i , we report most data as 340/380 excitation ratio rather than calibrated [Ca 2ϩ ] i . Calibration of fura-2 fluorescence into estimates of global [Ca 2ϩ ] i was carried out in primary cells as described previously (31,108,133) to illustrate the magnitude of T2R versus purinergic [Ca 2ϩ ] i signals. However, fura-2 calibration represents global [Ca 2ϩ ] i and does not provide information about localized concentrations that may occur (136) (e.g. within cilia (137) or in the vicinity of NOS enzymes (138)).
Whereas different objectives were used to acquire different fura-2 data sets in this study, only data acquired with the same objective are compared. Data in Fig. 1 were imaged at 20ϫ, data in Figs. 2 and 3 were imaged at 30ϫ, and data in Fig. 8 were imaged at 10ϫ, as a longer working distance was required for imaging through transwell filters. DAF-FM measurements utilized raw fluorescence values to compare experiments performed under identical conditions with identical microscope settings.

/NO 3 ؊ and cAMP in primary sinonasal ALIs
A fluorometric NO 2 Ϫ /NO 3 Ϫ kit (Cayman Chemical, Ann Arbor, MI) was used, and measurements were carried out according to the manufacturer's instructions. ALIs were changed to basolateral medium lacking phenol red and serum and treated with agonist as indicated in 30 l of apical PBS. Apical solution was collected after 18 h and assayed for total NO 2 Ϫ /NO 3 Ϫ . For cAMP measurements, primary sinonasal ALIs were transferred to basolateral HBSS and stimulated with agonists as indicated in 30 l of apical HBSS. After 10 min, cells were lysed and assayed for cAMP using the Amersham Biosciences cAMP Biotrak Enzyme Immunoassay System (GE Healthcare) according to the manufacturer's instructions.

Immunofluorescence microscopy
Immunofluorescence microscopy was carried out as described previously (12), with modifications outlined below. ALI cultures were fixed in 4% formaldehyde for 20 min at room temperature, followed by blocking and permeabilization in Dulbecco's PBS containing 1% BSA, 5% normal donkey serum, 0.2% saponin, and 0.3% Triton X-100 for 1 h at 4°C. Primary antibody incubation (1:100 for anti-T2R antibodies, 1:250 for tubulin antibodies) was carried out at 4°C overnight. Alexa Fluor-labeled donkey anti-mouse or rabbit secondary antibody incubation (1:1000) was carried out for 2 h at 4°C. Transwell filters were removed from the plastic mounting ring and mounted with Fluoroshield with 4Ј,6-diamidino-2-phenylindole (Abcam). For direct labeling of primary antibodies to colocalize T2R38 and T2R16, Zenon antibody labeling kits (Thermo Scientific) were used according to the manufacturer's instructions. Images of ALIs were taken on an Olympus Fluoview confocal system with an IX-81 microscope and 60ϫ (1.4 NA) objective. Acceptor bleaching experiments were carried out using the tornado bleaching function in Fluoview software. Images were analyzed using Fluoview software and/or the FIJI (139) version of ImageJ (W. Rasband, Research Services Branch, NIMH, National Institutes of Health, Bethesda, MD).

Genotyping of patient samples and ALI cultures
Genomic DNA was extracted from cultured cells or sinonasal specimens following the directions of the manufacturer (QIAamp DNA Mini kit and Blood Mini kit; Qiagen). Following extraction, samples were quantified using a spectrophotometer (ND-1000, Nanodrop), using 1.5 l of extracted genomic DNA. Samples were diluted to 5 ng/l and genotyped using the ABI StepOne real-time PCR system. Alleles of the TAS2R38 gene were genotyped for a variant site using allele-specific probes and primers (Applied Biosystems).

Data analysis and statistics
One-way analysis of variance (ANOVA) was performed in GraphPad Prism with appropriate post-tests as indicated; p Ͻ 0.05 was considered statistically significant. For comparisons of all samples within a data set, Tukey-Kramer post-test was used. For preselected pairwise comparisons, Bonferroni post-test was performed. For comparisons with a single control group, Dunnett's post-test was used. All other data analysis was performed in Excel. For all figures, one asterisk or number sign (* or #) indicates p Ͻ 0.05, and two (** or ##) indicate p Ͻ 0.01, respectively; n.s. indicates no statistical significance. All data are presented as mean Ϯ S.E. Data values from bar graphs are given in the supporting information.