The role of lipolysis in human orosensory fat perception.

Taste perception elicited by food constituents and facilitated by sensory cells in the oral cavity is important for the survival of organisms. In addition to the five basic taste modalities, sweet, umami, bitter, sour, and salty, orosensory perception of stimuli such as fat constituents is intensely investigated. Experiments in rodents and humans suggest that free fatty acids represent a major stimulus for the perception of fat-containing food. However, the lipid fraction of foods mainly consists of triglycerides in which fatty acids are esterified with glycerol. Whereas effective lipolysis by secreted lipases (LIPs) liberating fatty acids from triglycerides in the rodent oral cavity is well established, a similar mechanism in humans is disputed. By psychophysical analyses of humans, we demonstrate responses upon stimulation with triglycerides which are attenuated by concomitant LIP inhibitor administration. Moreover, lipolytic activities detected in minor salivary gland secretions directly supplying gustatory papillae were correlated to individual sensitivities for triglycerides, suggesting that differential LIP levels may contribute to variant fat perception. Intriguingly, we found that the LIPF gene coding for lingual/gastric LIP is not expressed in human lingual tissue. Instead, we identified the expression of other LIPs, which may compensate for the absence of LIPF.

The perception of taste elicited by countless chemicals present in food plays an important role for the survival of organisms. The gustatory system monitors not only the caloric content and the appropriate electrolyte supply, but also the presence of putatively harmful substances in food items ( 1 ). Detection of food constituents within the oral cavity is achieved by taste receptor molecules expressed by sensory cells specifi cally devoted to the detection of one of LIP activity in human orosensory fat perception ( 19 ). The latter would argue that only the free fatty acids already present in food may be perceived by human subjects or, perhaps at higher threshold concentrations, triglycerides may be directly detected within the oral cavity [compare ( 6,24 )].
In order to clarify this important question, we set out to investigate whether triglycerides represent an adequate "fatty" stimulus in vivo and in vitro and to determine if lipolytic enzymes are present in the human oral cavity. We performed sensory experiments on human panelists using either triolein or oleic acid as a stimulus in the absence or presence of orlistat. The results of these experiments were compared with data obtained from heterologously expressed fatty acid-sensitive G protein-coupled receptors, GPR40 and GPR120, two main candidates for human and rodent orosensory fat perception. Further, lipolytic activity was measured in human salivary samples. By RT-PCR and in situ hybridization studies, we determined expression of LIPs in minor salivary glands [von Ebner's gland (VEG)] of human tongue tissue.
To minimize potential textural cues induced by the stimuli, human sensory experiments on oleic acid, oleyl alcohol, and glyceryl trioleate were performed in an aqueous triacylglyceridefree lipid (TFL) matrix containing 5.0% (w/v) highly refi ned mineral oil (white oil, mouse embryo tested; Sigma-Aldrich) as a control for lubricity, 5.0% (w/v) Gum Arabic (Ph Eur, from acacia tree) to minimize viscosity cues, 0.3% (w/v) whey protein (100% Natural Whey Protein Isolate; Olimp) as an emulsifi er, and 0.1% (w/v) EDTA (bioUltra) to prevent fatty acid oxidation ( 18 ). Prior to sensory analysis, the stimuli were purifi ed by anaerobic low temperature recrystallization from acetonitrile and chromatography, respectively, to afford a purity of each stimulus of >99% (HPLC-evaporative light scattering detector, high resolution gas chromatography-MS).

Sensory analyses
Twenty assessors, who gave informed consent to participate in the sensory tests of the present investigation and had no history of known taste disorders, participated for at least 2 years in sensory training sessions and were asked to recognize and quantify the taste of aqueous solutions (5 ml each) of the following purifi ed reference compounds dissolved in bottled water (Evian, low mineralization: 500 mg/l) by means of duo and triangle tests, respectively, using a sip-and-spit procedure ( 26 ): sucrose (50 mmol/l) for sweet taste, lactic acid (20 mmol/l) for sour taste, NaCl (12 mmol/l) for salty taste, caffeine (1 mmol/l) for bitter taste, monosodium glutamate (8 mmol/l, pH 5.7) for umami taste, an aqueous emulsion of oleic acid (1 mmol/l, containing 0.02% Emultop) for fatty taste, a water/sunfl ower oil emulsion (3/6 w/w, containing 0.5% Span 65) for oily mouthfeel, and an emulsion of stearic acid (1 mmol/l, containing 0.02% Emultop) for a grainy powdery mouthfeel. diminished responses upon acute stimulation with several fatty acids, demonstrated by gustatory nerve recordings ( 11 ). However, one of the prerequisites for a bona-fi de additional "fat" taste quality would be the identifi cation of a separate dedicated taste cell population expressing these candidate receptors. While all of the above-mentioned fat-taste receptor candidates have been detected within taste buds on the tongue of rodents, evidence on the uniqueness of the corresponding taste cell population is still lacking. Although evidence is also mounting in humans (14)(15)(16)(17)(18)(19), much less is known about the orosensory perception of fat constituents relative to rodent models. However, it appears conceivable that the same group of candidate receptors shown in rodents or a subset thereof, such as CD36 ( 20 ) and GPR120 ( 18 ), may contribute to human fat perception as well.
One of the many open questions concerning human fat perception is what might be the predominant form of the fatty stimulus presented to the oral cavity? Similar to the other two macronutrients, carbohydrate and protein, which are not directly detected by taste receptors, but rather in form of their building blocks, the main dietary lipids are triglycerides and not free fatty acids. In the case of carbohydrates, the contribution of ␣ -amylase, a polysaccharidedegrading enzyme secreted by minor salivary glands directly into the trenches of foliate and vallate papillae, to the slowly developing sweet taste of otherwise tasteless starch is undisputed. Moreover, a recent study reported the presence of this enzyme in a subset of rodent taste receptor cells ( 21 ). An almost identical system for the degradation of triglycerides into free fatty acids exists in rodents. In this case, lingual lipase (LIP), a triglyceride degrading enzyme, is produced by minor salivary gland cells and secreted into the trenches of gustatory papillae in the back of the rodent tongue. In fact, it was demonstrated that lingual LIP secreted in saliva of minor salivary glands effectively liberated free fatty acids from trioleate in rats. Moreover, this enzymatic activity was required and suffi cient for the development of triglyceride preference in rats, as demonstrated by the blocking activity of the LIP inhibitor, orlistat ( 22 ).
For a role of lipolytic activity in human orosensory fat perception, contrasting evidence has been presented. While a recent report demonstrated that human subjects responded to oral stimulation with triolein with thresholds similar to those observed after stimulation with the corresponding free fatty acid, oleic acid, in the absence of orlistat, the thresholds increased, however, indicating an involvement of lipolytic enzymes in stimulus presentation ( 23 ). Another recent study performed in human subjects who chewed and expectorated different triglyceride-containing food items for a constant time period before quantifying the amount and nature of free fatty acids concluded that the concentration of free fatty acids in saliva after mastication would reach the required detection threshold. However, the contribution of oral LIP activity was not directly addressed ( 24 ). In contrast to these observations, however, other studies demonstrated the absence of lingual LIP by Western blotting, detected only traces of LIP activity in human saliva ( 25 ), and questioned the involvement of C18 column with a fl ow rate of 250 l/min starting with 60% acetonitrile and 40% aqueous formic acid (0.1% in water) for 6 min, then increasing the acetonitrile to 100% within 10 min, followed by isocratic elution for an additional 10 min. Analysis of glyceryl trioleate was done on a 150 × 2.0 mm, 4 m, Hydro-RP column (Phenomenex, Aschaffenburg, Germany) by using a binary gradient with 2-propanol/formic acid (0.1%)/ammonium formate (10 mmol) as solvent A and acetonitrile/formic acid (0.1%)/ammonium formate (10 mmol) as solvent B. Using a fl ow rate of 250 l/min, chromatography was done by increasing solvent A from 20 to 30% within 5 min, then to 60% in 5 min, followed by isocratic elution for 8 min. For quantitation of the oleic acid and glyceryl trioleate, a fi ve point calibration curve was recorded using 13 C 1 -palmitic acid as stable isotope labeled internal standard for the free oleic acid and glyceryl 13 C 3 -trioleate as the internal standard for glyceryl trioleate, and by plotting the ratios of peak areas of analyte/internal standard versus concentration ratios of analyte/internal standard.

Correlation of individual lipolytic activity with sensory sensitivity
For comparison of individual lipolytic activity at foliate papillae with sensory data for fatty perception, two groups of assessors were selected from the 20 subjects, group A (5 subjects) showing high sensitivity for glyceryl trioleate with detection thresholds of less than <0.65 mmol/l, and group B (5 subjects) showing lower triglyceride sensitivity with detection thresholds >1.0 mmol/l. To validate the different triglyceride sensitivity of both groups, a three-alternative forced-choice test was performed with glyceryl trioleate (0.7 mmol/l) in TFL matrix (2 ml) using a sip-and-spit procedure (two replicates). Whereas the sensitive group A was able to correctly pick out the right sample ( P < 0.001; ISO 5495:2005), there was no signifi cant outcome for group B, thus demonstrating a clear separation of both subject groups regarding their orosensory triglyceride sensitivity. For determination of the individual lipolytic activity of both subject groups, the assessors rinsed their oral cavity for 30 s with bottled water (5 ml) and were then challenged with fi lter paper discs (1 cm 2 ) loaded with glyceryl trioleate (5 l, 5.15 mol/cm 2 ) and placed on the foliate papillae for 150 s, respectively. After removal, lipolytic activity was terminated by transferring the paper discs collected from each individual subject into glass vials containing diethyl ether (2 ml; Merck, Darmstadt, Germany), followed by quantitation of free oleic acid using the stable isotope dilution assay detailed above (two biological replicates). A Welch's t -test ( 27 ) was applied to demonstrate signifi cant difference in lipolytic activity between group A and B.

Functional expression of the GPRs and calcium imaging analyses
Constructs of human GPR40 or the short variant of human GPR120 in the vector pcDNA5FRT ( 18 ) were transiently transfected into human embryonic kidney cells HEK 293T-G ␣ 16Gust44. After 22 h, the cells were loaded for 1 h with 1 M of the Ca 2+sensitive dye Fluo-4-AM (Invitrogen) and 2.5 mM probenecid in A group of 12 assessors (4 males, 8 females, 26-40 years of age) was selected from these trained panelists in a preliminary screening test by verifying their oral fatty acid sensitivity using oleic acid (1.4 mmol/l) as the test stimulus ( 19 ). The sensory sessions were performed in an air-conditioned room (22°C) with separate booths in three independent sessions. The subjects were instructed to refrain from food, beverages, and oral care products for a minimum of 2 h before the sensory experiments. To prevent cross-modal interactions with odorants, the panelists used nose clips.
For the determination of detection threshold concentrations for the fatty and/or "scratchy" oral sensation of oleic acid, glyceryl trioleate, and oleyl alcohol, respectively, in TFL matrix, an aliquot (2 ml) of sample dilutions differing by 0.25 log units was then presented to the sensory panel in order of ascending concentrations and each dilution was evaluated by means of a twoalternative forced-choice test (ISO 5495:2005) of fi ve sample pairs using a sip-and-spit procedure. During the evaluation, the panelists were asked to rinse their oral cavity with water and wait for 5 min between each sample. The individual threshold concentration of each panelist was calculated as the geometric mean between the last incorrectly identifi ed sample and the fi rst correctly identifi ed sample solution (three repetitions). The threshold concentration of the panel was calculated as the geometric mean of all individual threshold concentrations ( P < 0.05).
In an additional set of experiments, the infl uence of the LIP inhibitor tetrahydrolipstatin (Sigma-Aldrich) on the fatty perception was investigated using glyceryl trioleate (10.0 mg/ml) and free oleic acid (3.0 mg/l), respectively, as stimuli. A solution of the respective stimulus in TFL matrix (2 ml) in the presence or absence of tetrahydrolipstatin (0.25%, w/w) was evaluated by a group of 12 assessors (4 males, 8 females, 26-40 years of age) against the blank TFL matrix (2 ml) by means of a two-alternative forced-choice test (ISO 5495:2005) using a sip-and-spit procedure and asking the question "which sample is more fatty" (three repetitions ).

Lipolytic activity at human foliate papillae
The experiments on oral lipolytic activity of human volunteers were approved by the ethics committee of the Technische Universität München (application #2311/09) and informed consent of participants was obtained. Twelve trained assessors (4 males, 8 females, 26-40 years of age), instructed to refrain from food, beverages, and oral care products for a minimum of 2 h before the experiment, were asked to rinse their oral cavity for 30 s with bottled water (5 ml). Thereafter, fi lter paper discs (1 cm 2 , fi lter paper classic, 94 mm; Melitta, Minden, Germany) loaded with triolein (5 l, 5.15 mol/cm 2 ) in the absence and presence of 0.1% (w/w) orlistat (Sigma-Aldrich) were placed on the foliate papillae of the volunteers for 10, 20, 30, 60, 90, and 150 s and, instantaneously after removal, were transferred into glass vials containing diethyl ether (2 ml; Merck, Darmstadt, Germany) for quenching LIP activity. After adding defi ned amounts of the internal standards 13 C 1 -palmitoleic acid and 13 C 3 -triolein, water (5 ml, Milli-Q water advantage A 10 system; Millipore, Schwalbach, Germany) was added and, after mixing and pooling the solutions of the individual subjects, lipids were extracted with chloroform/ methanol (2/1, v/v; 5 ml, three times) by vortexing the tubes for 30 s, followed by centrifugation for 10 min at 2,000 rpm. The organic phase was separated from solvent under a stream of nitrogen, the residue taken up in acetonitrile, and an aliquot (10 l) was analyzed by HPLC-MS/MS consisting of a Dionex UHPLC UltiMate ® HPLC-system (Dionex, Idstein, Germany) hyphenated by an API 4000 QTRAP mass spectrometer (AB SCIEX Instruments, Darmstadt, Germany). Operating in the negative atmospheric pressure chemical ionization mode, analysis of the fatty acids was performed on a 150 × 2.0 mm, 5 m, HyperClone ODS 45 s (LIPF, ␣ -amylase, VEGP, and pepsinogen) or 1 min (GAPDH) at 68°C ( Taq ) or 72°C ( Pfu ) for elongation with 40 cycles (30 cycles for GAPDH). Amplifi cation of the entire coding regions of mouse and human LIPF was performed using Pfu -polymerase under the following conditions: 30 s at 95°C for denaturation, 1 min at the respective annealing temperature, and 1-1.5 min at 72°C for elongation. For the amplifi cation of LIPJ and pancreatic LIP (PNLIP) 10× buffer advanced without MgCl 2 (5Prime) was used and MgCl 2 was added to a fi nal concentration of 1.5 and 1.2 mM, respectively. Amplifi cation products were visualized by agarose gel electrophoresis.

Cloning of cDNAs for the generation of riboprobes
For subsequent generation of riboprobes, all cDNAs were subcloned into the transcription vector pBluescript I KS+ (Stratagene). ␣ -Amylase: a 300 bp fragment of human ␣ -amylase cDNA was amplifi ed by PCR from human CV cDNA and subcloned into the EcoRV site of the vector. For mouse ␣ -amylase, a commercially available cDNA provided in the vector pCMV-Sport6 (open biosystem) was digested with Eco RI resulting in a 531 bp long fragment, which was subsequently cloned into the Eco RI site of pBluescript I KS+. LIPF (lingual/gastric LIP): human LIPF cDNA corresponding to the entire coding region was amplifi ed by RT-PCR using cDNA prepared from commercially available stomach total RNA (BD Biosciences Clontech). After the intermediate insertion into the expression vector pcDNA5FRT (Invitrogen) with BamHI and NotI, a 912 bp long fragment was obtained by BamHI and Hin dIII digestion and inserted into the transcription vector. The entire coding region of mouse LIPF was amplifi ed by PCR from vallate papillae cDNA of the C57BL/6 strain. Further subcloning of a 902 bp long fragment was performed exactly as described for human LIPF cDNA. Human LIPK, LIPM, and LIPN: amplicons specifi c for LIPK, LIPM, and LIPL were generated by PCR from human CV cDNA. After intermediate cloning into pTOPO-TAII or -blunt vectors (Invitrogen), Hin dIII/XhoI or Eco RI similar sized fragments (LIPK, 434 bp; LIPM, 500 bp; LIPK, 553 bp) were transferred into pBluescript I KS+ vector. The insert orientations and sequence identities of all cDNA clones were confi rmed by sequencing (Eurofi ns MWG Operon, Martinsried). For a list of oligonucleotides, serum-free medium. After washing with buffer C1 [130 mM NaCl, 5 mM KCl, 10 mM HEPES, 2 mM CaCl 2 , 10 mM glucose (pH 7.4)], cells were stimulated with agonists at concentrations referred to in Fig. 3 . Agonist stock solutions (C18:1, 10 mM; glyceryl trioleate, 50 mM) were freshly prepared in 100% ethanol before each experiment and then diluted to the working concentrations in buffer C1. The fi nal concentration of the solvent ethanol was kept at or below 0.3% (v/v) to avoid toxic effects on the cells. Receptor responses were measured using a fl uorometric imaging plate reader (FLIPRTETRA ® system, Molecular Devices) as before ( 18 ). Changes in fl uorescence after substance application were monitored and, after subtraction of background fl uorescence, shown as calcium traces.

RT-PCR
The collection of human biopsy material was approved by the ethics committee of the Charité, Campus Virchow Hospital, Germany (application # 240/2002) and informed consent of participants was obtained. Biopsy materials of human circumvallate papillae (CV), C57BL/6 mouse foliate papillae, and mouse stomach were subjected to total RNA extractions using TRIzol reagent (Invitrogen) according to the manufacturer's recommendations. Total RNA of human stomach was purchased from BD Biosciences Clontech. For removal of contaminating genomic DNA, RNAs were subjected to RNase-free DNase-I (Invitrogen) digestion for 30 min at room temperature. Synthesis of cDNA was done using SuperScript III RT and random hexamers (Invitrogen) following the recommendations of the manufacturer. Identical reactions omitting RT were performed to generate negative control templates. The cDNAs were amplifi ed with oligonucleotides specifi c for ␣ -amylase and von Ebner's gland protein (VEGP) to control for the presence of VEGs, pepsinogen to control for the presence of zymogenic cells of the fundus, and GAPDH to monitor the quality of the cDNA produced. Oligonucleotide sequences, annealing temperatures, and amplicon sizes are given in Table 1 . PCRs were performed using Taq-DNA-polymerase (Fermentas) or Pfu -polymerase (Clontech) and the following conditions: 30 s at 95°C for denaturation, 30 s at the respective annealing temperature, 30 s (LIPK, LIPM, and LIPN), is recognized by human subjects and activates a triacylglyceride-sensitive detection system or an effi cient lipolytic enzyme is in place liberating oleic acid from trioleate.
To test whether the long-chain fatty acid-sensitive GPR120, that we suspected to contribute to human orosensory perception of lipids, would equally respond to trioleate, we performed functional calcium imaging experiments. In addition to GPR120, we transiently transfected another long-chain fatty acid receptor, GPR40, implicated in rodent annealing temperatures, and thermostable DNA-polymerases used for the amplifi cations, see Table 1 .

In situ hybridization of human and mouse gustatory papillae
Linearization of cDNA constructs was done with BamHI (LIPF), Hin dIII (amylase), XhoI + NotI (LIPK), XhoI + XbaI (LIPM), and XhoI + SpeI (LIPN). Then, in vitro transcription was performed using T3 or T7 RNA polymerase (Roche Applied Science, Mannheim) for sense and antisense probes, respectively. The probes were labeled using Dig-RNA-labeling mix (Roche), treated with RNase-free DNase I (Invitrogen), and then purifi ed by precipitation. In situ hybridizations were performed mainly as before ( 28 ). Briefl y, cross-sections of human (10 m) or mouse (14 m) CV were processed and thaw-mounted onto positively-charged glass slides. Prior to hybridization, the sections were fi xed, permeabilized, and acetylated. Prehybridization was done at 50°C for 5 h, followed by hybridization overnight at 50°C. After hybridization the slides were washed several times at low stringency, followed by RNaseA treatment and high stringency washes using 0.4× saline sodium citrate buffer at 50°C . Hybridized riboprobes were detected using an anti-digoxigenin antibody conjugated with alkaline phosphatase and a colorimetric reaction. Images were taken with a microscope (Axioplan; Zeiss, Goettingen, Germany) connected to a CCD camera (RT slider; Diagnostic Instruments, Sterling Heights, MI) or with an automated microscope scanning system (MIRAX Midi; Carl Zeiss MicroImaging GmbH, Jena, Germany).

Human orosensory perception of free and esterifi ed oleic acid
Previously we have shown that human subjects are able to detect free fatty acids of various chain lengths in sensory analyses performed in an aqueous TFL matrix to minimize potential textural cues induced by the stimuli ( 18 ). Specifi cally, long-chain fatty acids, if provided in texture masking fat-like matrix, induce a sensation described as fatty, whereas shorter chain fatty acids and fat alcohols or fatty acids when presented in aqueous solutions were rather described as scratchy ( 18 ). In order to test whether esterifi ed long-chain fatty acids were also recognizable by our panelists, and if so, whether identical or similar descriptors would be used, we performed additional sensory tests with the free long-chain fatty acid, oleic acid (C18:1), its esterifi ed form, glyceryl trioleate, and the corresponding fat alcohol, oleyl alcohol. As before, all stimuli were delivered in the aqueous TFL matrix to mask potential texture differences ( Fig. 1A ). In agreement with our previous observations, oleic acid was recognized and described as fatty at lower concentrations (threshold = 0.4 mM) and as scratchy at higher concentrations (threshold = 2.2 mM). The corresponding fat alcohol, oleyl alcohol, was perceived in a comparable concentration range and attributed exclusively as scratchy (threshold = 0.6 mM), again confi rming our previous results. Surprisingly, oral expectoration of triolein resulted in the sensitive recognition of this compound (threshold = 0.5 mM) and this time the only descriptor used was fatty, even at high concentrations. These results suggest that either the esterifi ed form of oleic acid differences in the peak enzymatic activities in gastric and oral tissues were reported in the two species [compare ( 29 )]. Hence, we also included stomach cDNA of both species in our analyses. In complete agreement with these previous reports, we failed to obtain a PCR product when amplifying cDNA from human CV ( Fig. 2A ). In contrast, a LIPF-specifi c amplicon was readily obtained from mouse vallate papillae cDNA. Performing the same analysis using human and mouse stomach cDNA samples gave the opposite result. A LIPF-specifi c PCR product was evident after amplifi cation of human but not mouse cDNA, corroborating again previous reports about the different expression patterns of this gene in both species. To rule out the possibility that we were unable to detect RNA present in mouse chief cells of the fundus, we amplifi ed mouse stomach cDNA with primers specifi c for pepsinogen. As pepsinogen is released from the same cell type as LIPF, the pepsinogen-specifi c product obtained from mouse stomach cDNA demonstrates the presence of this cell type in our sample. As LIPF in rodents is secreted by minor salivary glands into the trenches of foliate and vallate papillae, we performed additional control reactions to confi rm the presence of cDNAs originating from gland cell RNA in the investigated tissue samples. Indeed, PCR-products specifi c for human and mouse ␣ -amylase were obtained from both samples, confi rming the presence of minor salivary gland cells in these preparations. Serous minor salivary glands, also called VEGs, produce VEGP (lipocalin 1, tear prealbumin) ( 25,(30)(31)(32), a lipocalin-family member. This gene has been reported to be present in humans but absent in Mus musculus . The successful amplifi cation of a VEGPspecifi c PCR product only from human CV cDNA further confi rmed the complete representation of cell types in the fatty acid perception ( 11 ), which is, however, not expressed in human tongue tissue, in HEK 293T-G ␣ 16gust44 cells and stimulated the cells with high concentrations of oleic acid (3 M) and trioleate (3 M). As a negative control, cells transiently transfected with empty vector were treated accordingly ( Fig. 1B ). The prominent signals obtained after challenging GRP40 and GPR120 with oleic acid and the absence of calcium traces in receptor transfected cells stimulated with trioleate clearly demonstrate that both receptors respond exclusively to the free fatty acid. Hence, we concluded that humans possess either an effective lipolytic enzyme system within their oral cavity or trioleate is recognized by a yet unknown lipid sensing receptor.
In order to challenge the hypothesis of the involvement of lipolytic enzymes in fatty taste perception, we studied the infl uence of the LIP inhibitor, tetrahydrolipstatin, on the fatty perception induced by free oleic acid as well, as its ester, glyceryl trioleate ( Table 2 ). By means of a two-alternative forced-choice test (Which sample is more fatty?), randomized TFL solutions containing the respective fatty stimulus in the presence or absence of tetrahydrolipstatin were evaluated against the blank TFL matrix. Oleic acid solutions in TFL matrix were signifi cantly judged as the more fatty sample when compared with the TFL matrix alone independent on the absence ( ␣ -level 0.02) or the presence of the LIP inhibitor ( ␣ -level 0.01). Whereas trioleate could be clearly found to enhance fattiness of the TFL matrix ( ␣ -level 0.01), the presence of the LIP inhibitor diminished the sensory impact induced by the triglyceride ( ␣ -level >0.2). The data of these human sensory studies showed strong evidence for the active role of lipolytic enzymes in the orosensory perception of triglycerides.

Determination of lingual LIP gene expression in human taste tissue
To confi rm the presence or absence of lingual LIP in human taste tissue, we performed RT-PCR experiments with cDNA obtained from human gustatory tissue. Corresponding mouse tissue was used for control reactions as pronounced lingual LIP activity was reported in mice and rats ( 29 ). In fact, the proteins encoded by the orthologous LIPF genes were termed differently in humans ("gastric LIP") and mice ("lingual LIP") because pronounced a A solution of oleic acid (3.0 mg/l) and glyceryl trioleate (10 mg/l), respectively, in TFL matrix (2 ml) in the presence or absence of tetrahydrolipstatin (0.25%, w/w) was evaluated by a group of 12 assessors against the TFL matrix (blank) by means of a two-alternative forced-choice test using a sip-and-spit procedure and asking the question (Which sample is more fatty?).
b Number of positive answers obtained from a total of 36 responses obtained for 12 assessors in triplicate analysis. were quantitatively determined in pooled sample by means of HPLC-MS/MS using a stable isotope dilution assay . Compared with a blank vehicle (control), the absolute amounts of free oleic acid on the vehicle steadily increased from about 0.008 ± 0.003 nmol after 10 s to about 0.102 ± 0.012 nmol after 150 s, thus demonstrating a time-dependent release of the free fatty acid ( Fig. 4A ). Although the presence of the tetrahydrolipstatin could not completely abolish the release of free oleic acid from the triglyceride, the LIP inhibitor induced a signifi cant reduction of the free fatty acid release. Taking the 30 s incubation as the reference, the recovery of triglyceride in the glyceryl trioleate-loaded vehicle after an incubation time of 150 s was about 75% and close to 100% in the presence of tetrahydrolipstatin ( Fig. 4B ), thus clearly demonstrating the lipolytic activity of foliate papillae secretions resulting in the metabolization of about one-quarter of triglycerides.

Correlation of individual triglyceride sensitivities with lipolytic activities in saliva
For comparison of individual lipolytic activity at foliate papillae with sensory data for fatty perception, subjects were screened for the sensitivity for glyceryl trioleate on the basis of detection thresholds by means of a two-step procedure. First, subjects were evaluated in their detection thresholds for glyceryl trioleate in TFL matrix resulting in a group A of fi ve glyceryl oleate-sensitive subjects showing detection thresholds of <0.65 mmol/l and a group B of fi ve less sensitive subjects with detection thresholds >1.0 mmol/l ( Fig. 5A ). Second, a three-alternative forcedchoice test was performed with glyceryl trioleate in TFL matrix at a level of 0.7 mmol/l. Whereas the sensitive group A was able to correctly pick the right sample ( P < 0.001), there was no signifi cant outcome for group B, thus demonstrating a clear separation of both subject groups regarding their orosensory triglyceride sensitivity. For determination of the individual lipolytic activity of both subject groups, glyceryl trioleate loaded vehicles were placed on the foliate papillae of the subjects for 150 s, followed by tissue samples subjected to RNA preparation and cDNA synthesis.
In order to directly visualize LIPF expression in minor salivary gland tissue of humans and mice, we performed in situ hybridization experiments in cross-sections of human CV and mouse vallate papillae. Using antisense probes specifi c for human and mouse LIPF, respectively, we did not observe staining above background level in human tissue sections ( Fig. 3 ), however, we obtained intense staining in corresponding mouse cross-sections ( Fig. 3 ). The absence of specifi c signals in control sections hybridized with the corresponding sense probes demonstrated the specifi city of the procedure. On the contrary, amylase gene expression was detectable in tissue sections of both species, although the obtained signal intensities indicate that in mouse VEG, ␣ -amylase mRNA is more abundant compared with human VEG. Because the human ␣ -amylase gene exhibits a substantial copy number variation ( 33 ), the difference in staining intensity between human and mouse tissue sections may, to some degree, depend on the status of the individual biopsy material donor.
In conclusion, the lack of gene expression in minor salivary gland tissue demonstrates that LIPF cannot account for the liberation of free fatty acids from triglycerides in humans and thus does not explain the discrepancy between the detection of triglycerides in our sensory experiments and the observed specifi city for long-chain fatty acids of the candidate taste receptors.

Determination of lipolytic activity in human minor salivary gland secretions
To quantitatively monitor the lipolytic activity in human foliate papillae, the release of oleic acid from glyceryl trioleate was measured by placing fi lter paper discs loaded with glyceryl trioleate and/or without the LIP inhibitor tetrahydrolipstatin onto the foliate papillae of the 12 triglyceride-sensitive volunteers from 10 up to 150 s. Instantaneously after removal and sample work-up, the amounts of free oleic acid as well, as the recovery of glyceryl trioleate, 10q23.31 ). Of these, LIPK, LIPM, and LIPN, of which the genes are located downstream in close proximity to the LIPF gene and share the identical direction of transcription, possess a putative leader peptide at their amino termini and thus may be secreted enzymes similar to LIPF ( Fig. 6A ). Hence, we focused on the latter three LIP family members in our expression analyses. By RT-PCR of cDNA from human lingual epithelium using primers specifi c for putative secretory LIPs and ␣ -amylase, we detected LIPK gene expression in nongustatory lingual tissue ["taste cellfree" (TCF)], as well as in preparations of human CV from two individuals ( Fig. 6B ). A faint PCR product specifi c for LIPM was evident in only one of the two CV samples.
LC-MS/MS quantitation of free oleic acid using the stable isotope dilution assay. Comparing the amounts of free oleic acid released upon contact with foliate papillae secretions for the subjects in both groups A and B revealed a clear separation and showed higher amounts of oleic acid, that means higher lipolytic activity, in the samples collected from the fat-sensitive group A ( Fig. 5B ). On average, concentration of free oleic acid detected for participants from group A (mean = 5.77, SEM = 0.52) was higher compared with samples from group B (mean = 2.27, SEM = 0.49 ), application of Welch's t -test revealed a signifi cant difference between both groups (t(17.44) = Ϫ 4.87, P < 0.001). These data clearly indicate that individual differences in lipolytic activity might play a role in the different release patterns of free fatty acids from triglycerides and in orosensory fat sensitivity.

Identifi cation of LIP family members in human gustatory tissue
The discrepancy between the absence of LIPF (lingual/ gastric LIP) gene expression in human gustatory tissue and detection of lipolytic activity in saliva of minor salivary glands, forced us to investigate whether other secreted LIPs may account for the observed lipolytic activity. Indeed, several years ago a cluster of LIP genes on human chromosome 10 was identifi ed upon a screening for genes predominantly expressed in human keratinocytes ( 34 ). In addition to the well-characterized LIPF [lingual/gastric lipase ( 35 )] and LIPA [lysosomal cholesterol ester hydrolase ( 36 )], four LIP family members, LIPK, LIPM, LIPN, and LIPJ are present in this locus (human chromosome  small deletion, which may have resulted from alternative splicing events. As already shown in Fig. 2, before we could not detect LIPF-specifi c PCR products, whereas the PCR reactions specifi c for ␣ -amylase cDNA were positive in all samples analyzed .
Performing an extensive literature search for other LIPs that may account for the observed oral triglyceride-cleaving activity revealed no further likely candidate enzymes: Appropriately sized PCR products specifi c for LIPN were obtained from nongustatory lingual epithelium and both CV preparations, although at different levels. In addition to these amplicons, which corresponded to the predicted length of the amplifi ed cDNA region, smaller fragments were obtained. Cloning and sequencing of the amplifi ed PCR products revealed that these smaller amplicons also corresponded to the LIPN cDNA sequence except for a Data from rodent studies suggest that the main stimulus for the orosensory perception of fat constituents is free long-chain fatty acids ( 7,8,22,39 ). Although long-chain fatty acids are naturally present in fat-containing food, they predominantly occur esterifi ed with a glycerol core in the form of triglycerides ( 40 ). In rodent studies, it was demonstrated that triglycerides do not represent an Numerous LIPs do not utilize triglycerides as substrate (e.g., phospholipases such as LIPH, LIPI, phospholipases A-D; mono-/diglyceride LIPs such as DAGLA, DAGLB, MGLL), require coactivators like bile salts (BSSL, CEL) or hormonal stimulation for full enzymatic activity, occur only in intracellular compartments (e.g., LIPA, the lysosomal LIP or ATGL, the adipocyte triglyceride LIP), have been fi rmly associated with specifi c tissues or functions (such as the lipoprotein LIPs LIPC, LIPD, LIPG) or share multiple of the above mentioned features disqualifying them as secreted oral triglyceride LIPs. We assessed however, the gustatory expression of two further LIPs, PNLIP ( 37 ) and LIPJ ( 34,38 ), because PNLIP is secreted and hydrolyzes triglycerides and LIPJ, even though it lacks a signal peptide required for secretion, because the corresponding gene is located within the same cluster of genes as LIPA, LIPF, LIPK, LIPM, and LIPN ( 34 ), and therefore may be subject to similar gene regulation. As shown in Fig. 6C , both cDNAs were not detectable in tongue tissues, thus precluding a role in oral triglyceride breakdown.
Because expression of LIPK, LIPM, and LIPN genes were fi rst reported in human keratinocytes ( 34 ), we performed in situ hybridization experiments of human CV sections to determine whether transcripts of these genes are present in VEGs ( Fig. 7B ). Indeed, in situ hybridizations with all three antisense probes specifi c for LIPK, LIPM, and LIPN ( Fig. 7 ) resulted in the clear labeling of VEG cells in addition to keratinocytes located in the basal layers of the tongue epithelium ( Fig. 7A ). The absence of signals when using the corresponding sense probes ( Fig. 7 ) demonstrates the specifi city of the hybridization procedure. The identifi cation of putatively secreted lipolytic enzymes expressed in VEG cells indicates that these LIPs can surrogate for LIPF activity and may account for the LIP activity observed in the previous experiment.

DISCUSSION
Here, we investigated whether human probands orally perceive triglycerides and whether this may occur directly or via the recognition of long-chain fatty acids liberated from triglycerides by lipolytic activity. We demonstrate that humans perceive triolein similarly to its constituent, oleic acid, as fatty. Moreover, we show that oleic acid is effi ciently liberated from trioleyl glycerol upon exposure to saliva secreted from foliate papillae and that blocking of lipolytic activity by tetrahydrolipstatin (orlistat) reduces the generation of oleic acid. Most notably, the subjects' sensitivity to oleic acid was correlated with their LIP activity. Even though these data argue for an important role of LIP activity in human saliva, we demonstrated that the LIPF gene encoding the human ortholog of lingual LIP (called gastric LIP in humans) is not active in human gustatory tissue. A broader search for enzymes which could compensate for the absence of the LIPF gene product in humans revealed the expression of alternative LIP genes expressed, in particular, in VEGs secreting saliva directly into the trenches of foliate papillae. with those of a previous report ( 42 ) substantiates the existence of oral LIP activity. Hence, there is good agreement about the presence of signifi cant lipolytic activity in the salivary secretions of human minor salivary glands paralleling the observations in rodents ( 7 ). In order to represent a potent stimulus for the orosensory perception of free fatty acids, the necessary stimulus concentration at the receptor side must be reached. In that respect, the production of LIPs in minor salivary glands which directly secrete their saliva into the trenches of foliate papillae and CV housing the vast majority of oral taste buds, places the generation of the stimulus in the immediate vicinity of the receptive cells. This should allow effi cient stimulus generation even with small overall lipolytic activity. Our fi nding that humans produce LIPs in minor salivary glands ( Figs. 5,6 ) and that lipolytic activity is detected in minor salivary gland secretions ( Fig. 4 ) attests to a potential role of lipolysis in the processing of fatty stimuli. In light of the rather low overall LIP activity in human saliva, which appears neither suffi cient nor necessary for the bulk digestion of triglycerides, a specifi c function in the generation of orosensory perceived stimuli appears even more likely. Whether this enzymatic activity is indeed required for the orosensory perception of food may depend on the nature and processing grade of food items consumed. In the case of some high fat food items which contain already suffi cient amounts of free long-chain fatty acids, the small additional amount of fatty acids liberated by orally secreted LIPs may not be required ( 24,43 ). However, if the liberation of fatty acids from triglycerides is necessary to reach or exceed threshold concentrations, the activity of LIPs may become crucial for orosensory perception of food lipids ( 23 ).
Our data demonstrate that not LIPF (lingual/gastric LIP) ( Fig. 2 ), but other LIPs ( Figs. 5, 6 ) likely facilitate triglyceride degradation in human minor salivary gland secretions. Intriguingly, after many speculations about whether the enzymatic activity of human lingual LIP may suffi ce to effectively degrade triglycerides, it turned out that the LIPF gene is not even expressed in human VEGs, but rather related genes coding for alternative LIPs may replace the function of lingual LIP. Interestingly, DeNigris et al. ( 29 ) reported in 1988 that lingual/gastric LIP shows substrate specifi city for the medium-chain fatty acid containing tricaprylin (8:0) over the long-chain fatty acid containing triolein (18:1) by a factor of severalfold. This was also true for human gastric LIP, which exhibited about 4-fold higher enzymatic activity when offered tri[ 3 H]caprylin compared with tri[ 3 H]olein. However, this was not observed for lingual tissue. Instead, lipolytic activity for tricaprylin was not detectable in contrast to lipolytic activity for triolein. The observed preference of lingual/gastric LIP for mediumchain fatty acid-containing triglycerides over long-chain fatty acid-containing triglycerides is in perfect agreement with data obtained for partially purifi ed rat lingual LIP, which exhibited a 5-10 times higher specifi c activity for tributyrin compared with triolein ( 44 ). It is tempting to speculate that this fact may indeed point to the involvement of different enzymes in humans, such as LIPK, LIPM, and LIPN, as reported here. Moreover, as we have adequate gustatory stimulus and that the liberation of free fatty acids by the enzyme lingual LIP is required for the orosensory perception of triglycerides ( 22 ). Humans detect the free long-chain fatty acid, oleic acid, at lower concentrations than the triglyceride, triolein. Whereas the presence of the LIP inhibitor, tetrahydrolipstatin (orlistat), significantly elevated the orosensory detection threshold for trioleate, detection of oleic acid was not signifi cantly changed ( 23 ), suggesting that a similar mechanism for fat detection requiring the liberation of free long-chain fatty acids via lipolytic enzymes may also exist in humans. Our own experiments with human subjects demonstrate that triolein is judged by the panelists as fatty at a concentration range similar to oleic acid ( Fig. 1A ). However, unlike oleic acid, trioleate is not an appropriate stimulus for GPR40 and GPR120 ( Fig. 1B ), two prime candidates for orosensory fat perception in humans and rodents (11)(12)(13)18 ). Because other candidate molecules presumably involved in the perception of fat constituents such as CD36/FAT ( 9,10,23,41 ) and delayed-rectifying potassium channels ( 3 ) also respond to free fatty acids, another, so far undiscovered, triglyceridesensitive detection mechanism has to be assumed or, alternatively, lipolytic activity resulting in the generation of appropriate stimuli for fatty acid responsive detection mechanisms should be effective in humans as well .
The question of whether humans possess suffi cient lipolytic activity in their oral cavity to release free fatty acids effi ciently from triglycerides has resulted in inconclusive answers. Clearly, human saliva exhibits lower LIP activity than rodent saliva ( 29 ). However, lipolysis of triglycerides is an important step in the digestive process of this macronutrient. It was shown that the peak production site for LIP activity varies considerably among animal species. Whereas the main production site for LIP in rats and mice is the oral cavity and not the stomach, the opposite is found in humans and other vertebrates ( 29 ). It is therefore conceivable that bulk digestion of triglycerides in rats and mice starts at a more proximal point along the alimentary tract compared with humans, however, the question if the minute amount of lipolytic activity found in human saliva is suffi cient to liberate sensorially relevant concentrations of free fatty acids remains to be answered.
Previous studies investigating lipolytic activity present in the human oral cavity demonstrated that whole saliva collected from human probands after stimulation with cream contained an average lipolytic activity of 2 mol fatty acids/min/l ( 19 ). This is in good agreement with previous data by Spielman et al. ( 25 ), who reported identical whole saliva values for lipolytic activity in humans . The same study also collected saliva secreted from VEGs into the trenches of the foliate papillae and reported a 5-to 7.5fold higher LIP activity, suggesting that minor salivary glands may represent the main source for the secretion of this enzyme. Also, by direct extraction from human VEG tissue, lipolytic activity was obtained (0.3 mol/min/g wet weight), although at levels several hundredfold below similarly processed gastric tissue, which might in part be due to the fact that lingual gland tissue was obtained post mortem ( 29 ). However, the good agreement of these values previously shown that long-chain fatty acids rather than medium-chain fatty acids represent the main orosensory stimulus for fat detection in humans, the preferential liberation of this type of fatty acids would make perfect sense. An important step toward the understanding of human orosensory fat perception would be a careful biochemical characterization of these alternative LIPs to evaluate their individual contribution for stimulus generation.
Intriguingly, we observed that individual variation of triolein detection thresholds correlated with lipolytic activities measured in minor salivary gland secretions of human probands ( Fig. 5A, B ). While several studies reported considerable interindividual variations of fatty acid sensitivity among humans ( 15 ), and some of these studies even correlated the observed variations with nutritional status ( 19,45 ) and/or single nucleotide polymorphisms in candidate "fat taste receptors" ( 23 ), none of the previous reports investigated variances in lingual LIP activities. Because the perireceptor milieu constituted through salivary secretions is an important contributor for the perception of ingested substances ( 46 ) and salivary gland function is regulated by autonomic innervation ( 47 ), lipid sensitivities may even be infl uenced by the acute or chronic nutritional status of individuals and may involve modifi cations of the amounts of secreted LIPs.
Our work demonstrates the oral perception of triglycerides by human probands as well as the impact of differential LIP activities on individual threshold concentrations. Moreover, we have shown that the expression of LIPs different from LIPF in human minor salivary gland tissue may account for the release of the stimulating free fatty acids from triglycerides. Together, our results provide evidence for an important role of lingual lipolysis in the generation of appropriate stimuli for the oral perception of fat components in humans. Future work needs to be targeted toward a careful biochemical and functional characterization of the specifi c enzymes found in this study which seem to underly this mechanism.