Astringent Effects of Red Wine Associated with Responses of Aquaporins Found in Human Tongue and Salivary Tissues

Abstract

Aquaporin (AQP) water channels facilitate fluid transport across cell membranes, are implicated in rodent oral water sensing, and were examined in the human tongue for their modulation by mouthfeel compounds and ensuing effects on perception and saliva tribology. Immunohistochemistry demonstrated abundant AQP1 and AQP2 and moderate AQP5 in human tongue and high AQP5 in salivary gland. Trained human tasting panels evaluated astringency intensities; results correlated with wine tannin measures and inhibition of AQP1 water flux (R ² ≥ 0.9). Wine tannin extract additions of 1 g/L reduced cell swelling −125 ± 20% (SE) (P < 0.0001), increased saliva-tannin friction coefficient 16.0 ± 1.9% (P < 0.0001) and astringency scores 17.3 ± 5.5% (P < 0.006). Osmotic swelling assays identified wines, red wine polyphenols, alum sulfate, and tannic acidall archetypal astringentsas inhibitors of AQP1 and to a lesser extent AQP5. Astringent block of AQPs present in tongue and salivary gland suggests a mechanistic role of water flux in drying sensations, beyond the established changes in saliva resulting from tannin and proline-rich-protein interactions. Insights into AQPs as molecular components of mouthfeel could inform fundamental debates on how astringent phenomena arise and increase understanding of nutrient sensing and uptake as found within the digestive tract and throughout the body.

Introduction

Water is crucial for survival and regulated by thirst mechanisms, signaled by a complex system of internal monitors to generate behavioral responses to maintain homeostasis. Basic tastes (sweet, salty, sour, bitter, umami) elicited by foods are well understood and important for driving feeding behaviors, while mouthfeels such as astringency are less understood mechanistically. Nearly all foods and beverages contain some amount of water; however, the “taste of water” as a salient precept remains absent from most descriptions of foods we eat or drink. Water evokes responses in taste afferent nerves, but the initial mechanisms that transduce a water sensory stimulus into a neural-conducted signal remain to be defined.

The basis for the sensation of astringencya complex of sensations due to shrinking, drawing, or puckering of the epithelium as a result of exposure to substances such as alums or tannins, as defined by the American Section of the International Association for Testing Materialsis not well understood. Pathways for astringent signaling have been vigorously debated for more than 50 years but have not achieved a consensus, with some lines of evidence supporting taste , or tactile signaling pathways; conversely, other substantial lines of evidence support salivary protein aggregation and/or precipitation with loss of lubrication of oral surfaces as the major effects of most astringents. Loss of lubrication is a phenomenon that has been measured extensively using physical chemical techniques such as tribology, a method which quantifies friction at interacting surfaces for wines − and for other foods. Conversely, others have speculated receptor involvement, − epithelial cell interactions, or a combination to evoke perceptions of astringency. The complexity of this issue is evident from findings that astringency is not a single modality, but can be perceived as distinct subqualities such as drying, puckering, roughness, and others. , Adding acidity often elicits puckering usually described as a tightening reflex or drawing sensation of the cheeks and facial muscles. No evidence is available to discern whether perception of puckering results from isoelectric precipitation of salivary proteins at low pH, involuntary pouting of orbicularis oris muscle as a response to sourness, or both. Similarly, data are lacking on whether aggregated salivary protein particulates on oral surfaces cause roughness. The mechanisms of perception for these subqualities remain to be clarified.

Detection of water by taste receptor cells in rat tongue was proposed as a gustatory modality mediated by aquaporin (AQP) water channels, in addition to the recognized gustatory modalities for sweet, sour, salty, bitter, and umami tastants mediated by separate pathways. Patch clamp recordings showed that the inhibition of AQP water pores by tetraethylammonium − in rat taste bud cells reduced the current responses to hypotonic stimuli, supporting the proposal that AQPs could be involved in sensing the taste of water. − AQP1, AQP2, and AQP5 were found in rat tastebud cells, with AQP5 and likely AQP1 in the apical membrane, a position consistent with possible roles as the starting point for sensory detection; AQP2 in the basolateral membrane is consistent with a role in maintaining homeostasis. The G-protein coupled receptors and ion channels are known to signal basic tastes; however, the presence of AQPs in human oral tissues and their sensitivity to modulation by gustatory stimuli remained unknown.

Aquaporins are members of a large family of membrane intrinsic channel proteins originally demonstrated to facilitate water fluxes, logically associated with critical roles in maintaining fluid homeostasis in diverse tissues. Roles of AQPs in sensory signaling, development, and cell motility involve permeabilities to an expanding range of solutes (such as glycerol, hydrogen peroxide, ammonia, urea, ions, gases, and more) in addition to water, with new substrates continuing to be discovered. , These advances have greatly expanded the understanding of AQPs as sophisticated multifunctional channels with diverse roles in physiological and pathological processes. ,

AQP channels show tissue-specific distributions throughout the human body. AQP1 was first discovered in red blood cells and is abundantly expressed in tissues that move high volumes of fluid across barrier membranes, including for example kidney proximal tubule, choroid plexus, and ciliary body in the eye. Nonetheless, AQP1 also provides for modulation of functions at the cellular level in a surprisingly diverse array of cells including photoreceptors, erythrocytes, retinal pigmented epithelium, dorsal root ganglia, activated glia, trabecular meshwork, endothelial cells, and more, as reviewed recently. , AQP2 is a key component in the regulation of fluid in the kidney collecting duct, where it serves as the target of action for the antidiuretic hormone vasopressin which induces translocation of the AQP2 channels to the kidney cell membrane to increase water reabsorption back into the body. AQP5 is expressed in salivary glands, lacrimal glands, and lungs. Mistargeting of AQP5 to basal rather than apical membranes causes Sjögren’s syndrome, a disease characterized by deficient saliva production due to autoimmune destruction of the salivary glands.

We surmised that astringent compounds, such as tannins, might modify cellular water detection by directly inhibiting AQP water channels in the gustatory tissues involved in sensory signaling. This study is the first to show that the same classes of AQP channels found in rat taste buds (AQP1, AQP2, AQP5) also are expressed in the human tongue and salivary gland and that effects of compounds which are known to impart astringency to foods and beverages differentially alter AQP water channel activities, supporting their role as part of a transduction mechanism for water sensing as a possible sixth gustatory modality, in combination with parallel effects on mouthfeel.

Materials and Methods

Chemicals

Sensory standards tannic acid (ACS reagent, ≥99%) and aluminum potassium sulfate dodecahydrate (≥99%) were purchased from Merck (NSW Australia). The reference red-wine tannin extract was prepared in-house from Shiraz wine; no commercial tannin was used. Unless otherwise stated, reagents were of analytical grade and were employed without further purification.

Immunohistochemistry of Human Tissue

Slides with formaldehyde-fixed paraffin-embedded (FFPE) 5 μm sectioned tongue and salivary gland tissues from untreated human subjects were purchased from OriGene (supplied via Australian Biosearch, WA Australia; #CS814593 tongue, >98% normal; #CS712355 salivary gland, normal). Mouse kidney FFPE samples were prepared locally and sectioned as positive controls. AQP1 proteins in mouse and human have high amino acid sequence identity (94%) and show comparable immunolabeling by antibodies directed against the conserved epitopes. Slides were held at 60 °C for 1 h, dewaxed with 2 rinses in xylene (5 min each), and dehydrated in a graded ethanol series (twice for 3 min each), followed by two washes in 1% phosphate-buffered saline (PBS; 5 min each) to remove residual solvents. Antigen retrieval in a citric acid buffer (10 mM, pH 6) bath used a steam microwave at 100 °C (Sixth Sense, Whirlpool, VIC, Australia) for 10 min to unmask epitopes possibly affected by fixation, and slides were cooled (30 min, room temperature). Two 1% PBS washes (5 min each) prepared tissue sections for blocking with 5% goat serum (100 μL per sample; 30 min) to minimize nonspecific antibody binding.

Slides were incubated with primary anti-AQP antibodies (4 °C overnight) and rinsed in PBS (two washes, 5 min each) to remove unbound antibodies. Rabbit polyclonal anti-AQP antibodies purchased from Merck (NSW Australia; #A5560 AQP1, #SAB4501547 AQP2, and #A4979 AQP5) label mammalian AQPs including human and mouse. Secondary antibody (1:500 dilution; Merck #SAB5600287 biotin-tagged goat antirabbit antibody) was applied at room temperature for 1 h and rinsed in PBS (two washes, 5 min each). Signal detection used streptavidin-HRP conjugated antibody (1:500 dilution, Dako; P0397) at room temperature for 1 h, visualized by a chromogenic reaction with diaminobenzidine (DAB) and hydrogen peroxide (Merck) which produced a brown precipitate, marking the presence of target proteins. Counterstaining cell nuclei with 10% hematoxylin (Merck) provided a histological context. Slides were coverslipped with Pertex mounting medium (Medite Medizintechnik, Burgdorf Niedersachsen, Germany) for imaging. High-resolution images were captured using a NanoZoomer Digital Pathology System (Hamamatsu Photonics K.K.; Shizuoka, Japan). Digital images were viewed using NDP.view2 software (Hamamatsu Photonics K.K.; Shizuoka, Japan).

Osmotic Water Permeability Assays

To assess aquaporin function, water permeability assays were conducted using the oocyte expression system, as published previously. Oocytes were harvested by partial ovariectomy from adult female frogs completely anesthetized by cold Tricaine methanesulfonate bath immersion.

Enzymatic treatment of isolated lobes with collagenase type I (2.5–3 h; Merck Company NSW Australia) was used to dissociate single oocytes and remove surrounding follicular cells.

Selected stage 5 oocytes were individually injected with 40 nL of sterile water containing 1–5 ng of copy RNA synthesized in vitro (Invitrogen mMessage mMachine kit, Fisher Scientific, SA AUS) from linearized plasmid DNA for human AQP1 (P29972), AQP2 (NP_000477), or AQP5 (NP_001642), and oocytes were maintained at 18 °C for 48 h in Frog Ringers solution to allow AQP protein expression. Frog Ringers consisted of isotonic Na⁺ saline (in mM: 96 NaCl, 2 KCl, 5 MgCl₂, 5 HEPES) pH 7.6, supplemented with 0.6 mM CaCl₂, 0.5 mg/mL tetracycline, 100 U/mL penicillin, 100 U/mL streptomycin, 5% (v/v) horse serum, and 2.5 mM Na pyruvate.

Oocyte swelling was monitored optically using a dissecting microscope. Cross-sectional measurements of oocyte volume as a function of time after transfer into test salines were collected over 30 s with images taken at 1 Hz and analyzed using ImageJ software (National Institutes of Health; USA; version 1.54F) to assess swelling rates, calculated from linear slopes of cell volume as a function of time. The swelling rate was measured as the slope of the linear fit of oocyte volume plotted as a function of time during short (30 s) exposures to hypotonic salines with or without wine or wine components.

Human Sensory Panel Evaluations of Red Wines and Test Solutions

Perceived taste intensity scores for undiluted blended red wine samples (the same samples that were tested in oocyte swelling assays at 5-fold lower doses) were determined using the Quantitative Descriptive Analysis (QDA) method, as comprehensively described, including reference standard formulations. In brief, QDA sessions were conducted over 4 weeks following the generic method described by Heymann, King, et al. guided by international standards principles on sensory methodology. A total of 28 wines were evaluated by trained assessors for their appearance, aroma, mouthfeel, texture, and flavor. Assessors underwent six 2 h training sessions to develop and calibrate sensory attributes and reference standards. Formal assessments were performed in sensory booths under controlled conditions using a Williams Latin Square design to randomize wine presentation. Data collection was managed via Compusense20 software (Compusense V20, CAN), with intensity ratings recorded on a 15 cm unstructured line scale. A carefully curated subselection of six wine samples representing the edges and middle blends of the three-component experimental design was utilized for the current study. A separate QDA study was conducted using the same procedures as above investigating taste and mouthfeel compounds added to red wine; of particular interest for this study, the astringency scores of three of the 12 samples tested, comprising a red wine control with its own tannin extracted and back added at two levels, was compared to AQP1 response and saliva-sample tribology measures. Sensory testing at the AWRI was performed in accordance with ethics approval H-2022-192 with the informed consent of panel members.

Intensity Testing Using a Labeled Magnitude Scale

Psychophysical evaluation of nonwine samples was conducted using a labeled magnitude scale (LMS). The LMS allows participants to indicate their perceived intensity of a stimulus along a continuum of verbal descriptors that correspond to numerical values, ranging from “barely detectable” to “strongest imaginable (e.g.,) Sourness” This method provides a framework for capturing the subjective experience of taste and mouthfeel intensity across a wide range of stimuli and concentrations. A similar protocol was used by Galindo-Cuspinera, Winnig, et al. to investigate “water-taste” gustatory after-impressions of sweetness by water.

Participants for LMS ratings of archetypal astringent solution experiments were selected from the same cohort of trained panelists previously utilized for QDA ratings of wines. Given their prior training and familiarity with sensory evaluation protocols, no training was provided, except the instruction to rate the perceived intensity of the taste or mouthfeel qualities of a solution presented. Each assessment day was conducted with a group comprising 14–15 participants. The experimental design followed a Williams Latin Square arrangement managed through Compusense20 software.

Testing sessions were systematically organized over separate days. The initial session was dedicated to evaluating all basic taste compounds and mouthfeel attributes, providing a comprehensive platform for subsequent analyses. The remaining sessions focused specifically on varied concentrations of tannic acid and aluminum potassium sulfate, selected for further investigation based on their relevance to the modulation of aquaporin responses observed in preliminary assessments. Astringent solutions were presented in 30 mL aliquots, ISO wine glasses, monadically, with color masking lights. Participant breaks of at least 5 min between samples were enforced. Each solution was rated in duplicate by each participant with tannic acid and alum sulfate presented on separate days.

Saliva Collection and Tribology Measurements of Saliva-Red Wine Lubrication Changes

Whole-mouth saliva after acid stimulation from two donors was collected for tribology experiments according to methods used previously. The two saliva donors (1 male and 1 female) were nonsmokers who refrained from eating or drinking for 2 h before the collection. The donors rinsed their mouths with water for 10 s, followed by the administration of 100 μL of aqueous citric acid solution (2%) at the posterior part of the tongue to stimulate saliva production. The saliva secreted in the first 30 s was discarded; subsequent salivary output was collected for 2 min in a 15 mL centrifuge tube on ice. To minimize storage time, each saliva specimen was used within 30 min of collection. Saliva donors were vaccinated for Hepatitis B. Saliva was collected at a monitored collection location and handled by using sealed bags.

The tribometer was set up on an MCR302 rheometer (Anton Paar GmbH, Graz, Austria) by fitting it with a tribocell (SH-BC6-T-PTD200) and a ball holder (BC 12.7 D). Mouth oral friction was simulated by using a soft poly­(dimethylsiloxane)-poly­(dimethylsiloxane) (PDMS–PDMS) contact which consisted of 3-pins (Rq ≈ 0.1 μm) in the cell and a ball (Rq ≈ 0.022 μm) attached via the holder.

The friction coefficient μ upon contacting red wines was measured by tribology following a previously studied method for wines with minor modifications. μ is defined as μ = F _f/F _N, where F _f is the force opposing the motion and F _N is the normal force perpendicular to the plane of contact. The measurements were carried out at a controlled temperature of 23.0 ± 0.2 °C. The results obtained from a testing protocol are illustrated in Supplementary Figure 1, showing a plot of μ values as a function of time. The PDMS pins and ball were cleaned following established protocols and attached to the instrument. The shaft was lowered onto the cell and stopped at normal force F _N = 1 N, and the rotational speed was set at a linear speed v = 5 mm/s. The friction coefficient was captured every 0.5 s. Upon starting the test, 500 μL of freshly collected saliva was added to 500 μL of wine and mixed by dispensing and withdrawing fluid with a 1 mL pipet for 5 times. The choice of the 1:1 mixing ratio was based on previous literature. At 90 s, the whole mixture was gently poured into the tribo-cell. The measurement concluded at 400 s, and the averaged μ during the last 100 s was taken as the result. Experiments were conducted over three consecutive days, with each of the three samples of red wine measured at least twice per day, to mitigate any day-to-day saliva variation, and a total of 28 measurements were made on the wine sample set.

Preparation of Samples for Sensory Panel Testing

All testing solutions for human sensory panels were prepared 14–24 h in advance of the evaluation sessions for consistency and reliability. Isolated compounds used for some taste and mouthfeel solutions were analytical grade, sourced from Merck (NSW, AUS) except for red wine tannin extract. Red wine tannin extract was isolated from Shiraz wine (Coonawarra, Australia, 2022, Alcohol 13.6%, pH 3.64, Titratable acid 6.4 g/L at pH 8.2, Volatile Acidity 0.86 g/L, Tannin 2.38 g/L) by the method of Kennedy and Jones, with modifications described in McRae, Schulkin, et al. and added back to the same baseline wine for dose–response testing. Dealcoholized wine samples were prepared by placing aliquots (100 mL) of Cabernet Sauvignon (Margaret River, Australia, 2019, Alcohol 14.3%, pH 3.62, Titratable Acid 6.2 g/L at pH 8.2, Volatile Acidity 0.64 g/L, Tannin 1.9 g/L) and Chardonnay (Adelaide Hills, Australia, 2022, Alcohol 12.0%, pH 3.21, Titratable Acid 7.2 g/L at pH 8.2, Malic acid 1.63 g/L, Volatile Acidity 0.58 g/L) in a rotary evaporator with diagonal condenser at 45 °C under modest vacuum pressure, adjusted manually to maintain a gentle boil for ∼30 min or until 30% of the volume was evaporated, in order to remove the majority of the alcohol; then original volumes were restored with distilled water before testing. Samples for testing in AQP-expressing oocytes were taken from the same human panel tasting solutions and assessed at 1/5 dilution in osmotic swelling assays. The standard swelling assay saline was 50% hypotonic saline (50% isotonic saline and 50% nanopure filtered water) as per published protocols. Wine solutions were diluted to 20%, and saline and water were added to achieve osmotic similarity to standard 50% hypotonic. Single compounds were added to standard 50% hypotonic at the final concentrations as indicated in the results. Osmotic pressures were checked with an Osmo-1 osmometer (John Morris Scientific, NSW AUS), calibrated with Clinitrol Reference Solution 290 (Advanced Instruments, VIC AUS).

Statistical Methods and Correlation Analysis

Distributions of the swelling data were evaluated for outliers and normality by Grubb’s test and Shapiro-Wilk, respectively. If results lacked normality, statistical significance was analyzed with the nonparametric Kruskal–Wallis test, with Bonferroni correction or Mann–Whitney U tests to identify differences between samples using XLSTAT (V 2024.3, Lumivero, Denver, CO).

Sensory data were analyzed using a combination of Minitab 21.1 (Minitab Inc. Sydney, NSW) and XLSTAT. Following convention, QDA data for red wine samples were assessed by ANOVA for sample, assessor, and presentation replicate effects as well as for two-way interactions, treating assessors as a random effect. Posthoc mean comparison on the sensory score of samples was conducted by Tukey’s honest significant difference (HSD) test at a 95% confidence level. Psychophysical sensory intensity data collected for astringent solutions on the labeled magnitude scale did not show normality and were analyzed by Kruskal–Wallis with Bonferroni correction and scaled 0–10 for visualization purposes.

Tribology data measures were analyzed using XL STAT. Average friction coefficient values of the final 100 s of each run were checked for outliers and normality; a subsequent two-way ANOVA was conducted for the effects of sample, day, and their interaction, followed by a protected HSD posthoc means comparison test at a 95% confidence level.

Box plots show the compiled experimental data. The box contains data points falling in the first to third quartiles; the horizontal bar inside the box is the median value; the red plus symbol (+) is the mean; solid dots show maximum and minimum values; whiskers depict 1.5× the interquartile range after removal of outliers as detected by the Grubbs test.

Human and Animal Ethics Approvals

Experimental procedures involving were conducted in accordance with protocols approved by the University of Adelaide Animal Research Ethics Committee (Approval ID: M-2022–050). Saliva samples and tribology were conducted in accordance with CSIRO Human Ethics approved protocols (Approval ID: 2024_002_HERC). Sensory testing involving human participants was performed following ethical standards approved by the University of Adelaide Human Research Ethics Committee (Approval ID: H-2022-192).

Results

Detection of AQP1, AQP2, and AQP5 Proteins in Human Tissues

Immunohistochemical staining of paraffin-embedded sections revealed AQP1, AQP2, and AQP5 were present in human tongue (Figure ). As expected, kidneys were strongly positive for AQP1 (Figure A) and AQP2 (Figure B). Positive signals for AQP1 and AQP2 in the tongue were evident (Figure C,D). In contrast, AQP5 showed a comparatively low signal level in tongue tissues (Figure E), but it was strongly expressed in salivary gland regions (Figure F).

1.

Immunolabeling of tongue sensory epithelial tissue with anti-AQP1, -AQ2, and -AQP5 antibodies, with positive controls in mouse kidney and human salivary gland. (A) is mouse kidney, showing proximal tubule labeled (brown) with anti-AQP1, as a positive control. (B) is mouse kidney, showing AQP2 in collecting ducts as a positive control. Rows (C)–(F) show adjacent sections of normal human tongue and salivary gland, labeled with antibodies to AQP1 (C), AQP2 (D), or AQP5 (E,F), with (E) showing low levels in tongue, and (F) showing salivary gland as a positive control. Insets (top left) indicate the position of the field of view within the tissue slices. The left column shows 10× magnification images (scale bar 250 μm), and the right column shows 40× images from the same field of view (scale bar 50 μm).

In summary, both AQP1 and AQP2 were strongly expressed in the tongue epithelial cells. AQP5 showed low but distinct presence in sensory cells, in contrast to strong expression in salivary gland. In kidney, high expression of AQP1 in proximal tubule and descending thin limbs and AQP2 in collecting duct are well established, serving as positive controls. AQP5 as expected was strongly expressed in salivary gland, where it serves in salivary secretion.

AQP Activity Assessed by Osmotic Swelling Rates

Interactions of AQPs with components of wines and isolated compounds selected based on chemical composition, sensory properties, and consumer acceptance were tested for a range of taste/mouthfeel solutions and responses as detected by changes in oocyte swelling rates (Figure ). Possible AQP-modulating effects of tastants were assessed using swelling assays in frog oocytes expressing cRNA-encoded AQP1, AQP2, and AQP5, as compared with non-AQP-expressing control oocytes (Figure A). Saliva samples (Figure B and human panel evaluations (Figure C) provided data that were evaluated for correlations with the AQP data. Compiled swelling rate data in box plot histograms were used to compare responses of AQP-expressing oocytes under the wine (test) and no-wine (untreated) conditions (Figure D). Treatments were applied acutely, only during the short swelling assay itself. Agonist and antagonist effects of wine treatments were identified based on increases or decreases in measured swelling rates, respectively. Non-AQP-expressing control oocytes, as expected, exhibited minimal swelling responses. Low water permeability of native oocyte membranes is an attribute that has made this expression system a standard tool for AQP research for decades. Results were assessed for dose-dependent properties of responses, illustrated as one example from in vivo data for human perceptions of astringency intensity as a function of alum concentration (Figure E).

2.

Illustration of the experimental strategy. Diagrams of (A) the oocyte expression system for measuring water channel activity of cloned human aquporins. (B) the tribology method for assessing saliva friction. and (C) trained tasting panel evaluations of sensory intensity ratings. Example plots are shown for (D) oocyte volume as a function of time during swelling assays and (E) a representative summary of participants’ tasting responses to different concentrations of a standard astringent aluminum sulfate. Panels (A)–(C) were created in BioRender: Yool, A. (2025) https://BioRender.com/v41h537.

Identification of Key Tastants Influencing AQP Channel Activities

The results obtained for tastant effects on water channel activity (in vitro, above) in the oocyte expression system were compared with the intensities of human perception responses to wine and nonwine test solutions. Associations between the in vivo and in vitro data were analyzed by PLS-R. A strong link was demonstrated between the sensory panel’s scores for wine astringency and red wine tannin measures and the magnitude of inhibition measured for AQP1-mediated water fluxes in matched samples (Figure ). One interesting finding was that in particular, the phenolic levels (amounts of methylcellulose precipitable tannin (MCPT) and tannin molar mass (MM) constituents) showed direct associations with the magnitude of the AQP1 water channel block and the intensities of human perception of astringency in the beverage samples.

3.

Relationships between levels of phenolic constituents, astringent taste properties, and AQP1 channel activity. Methylcellulose precipitable tannin (MCPT) and tannin molar mass (MM) were tested with a range of doses. Comparisons were: (A) MCPT and in vivo astringency, (B) MM and in vivo astringency, (C) MCPT and AQP1 water channel activity, and (D) MM and AQP1 water channel activity. Data were fit by first order polynomial trend lines and yielded statistically significant (R ² ≥ 0.90) coefficients of determination values for all relationships.

Dose–Response Analyses of Astringent Effects

Inferred correlations between tannin-related astringent intensity and AQP1 channel activity were assessed for dose-dependent effects by adding isolated tannin compounds to a baseline red wine (Figure ). Calibrated doses of extracted tannins were added in 0.5 g/L increments.. Relationships between red wine tannin concentration, astringency intensity, and AQP channel function were assessed by sensory panel tastings and oocyte assays. Oocyte assays also included a hypotonic saline alone (without wine). Human taste panel assessments demonstrated significant (P < 0.001; ANOVA and posthoc Tukey’s test) increases in perceived astringent intensity as tannin dose increased (Figure A). These same three samples were also tested for friction coefficients of human saliva-wine mixtures (Figure B), also showing significantly (P = 0.005) increased friction (i.e., loss of lubrication) with increases in extracted tannin added to baseline wine. The same treatments also caused a dose-dependent block of AQP1 water fluxes (Figure C), showing progressive reduction in median values of AQP1 swelling rates with increasing tannin concentration. No differences in swelling rates were found between AQP1-expressing oocytes that were untreated versus those incubated in saline for 4 h post-tannin exposure, demonstrating the observed inhibition of AQP1 water flux was reversible, also showing that the inhibitory effect could not readily be attributed to an indirect phenomenon such as degradation or internalization of AQP1 membrane protein. The reversible nature of tannin block of AQP water channels is consistent with a model in which tannins could create a displaceable “lid” that allows water to exit better than enter the cells, possibly facilitating dehydration and shrinking of cells and oral tissues.

4.

Boxplot summaries of dose-dependent effects of isolated tannins on human sensory perception, wine-saliva tribology. and AQP1 activity. Calibrated additions of an isolated red wine tannin extract to a baseline wine yielded dose-dependent effects on (A) increasing astringency ratings by sensory panelists, (B) increasing friction coefficients of wine-saliva mixtures, and (C) inhibition of AQP1 water channel swelling rates. Statistical comparisons above histogram bars shown as a and b were assessed by ANOVA and Tukey’s HSD grouping; c was evaluated by Kruskal–Wallis and post hoc Bonferroni tests for a corrected significance level of 0.005 (*), n = number of replicates.

Dissection of the Effects of Wine Components

To evaluate effects of tannins without the baseline red wine, doses of extracted tannin were dissolved directly in buffered hypotonic saline and tested in oocytes expressing AQP1, AQP2, and AQP5 channels (Figure ). Tannin treatment showed a trend toward decreased mean and median swelling rates for all three classes of AQPs, appearing to suggest the antagonist effect is not AQP-subtype specific. However, a statistically significant effect was seen only for AQP5, which could reflect variability due to outliers in the AQP1 and AQP2 treatment groups.

5.

Effect of red wine tannin extract dissolved in water on the AQP response. Analyses are from (A) AQP1, (B) AQP2,and (C) AQP5 expressing oocytes measured by optical swelling assays; data are swelling rates. Significance (*) was determined by Mann–Whitney U test, n = number of replicates.

AQP inhibition was not an indirect effect of the ethanol present in the baseline wine solutions. Normal and dealcoholized versions of red (Cabernet Sauvignon) and white (Chardonnay) wines were compared for their effects on osmotic swelling activities for the three classes of AQPs (Figure ). Cabernet Sauvignon inhibited swelling rates for the AQP1-, AQP2-, and AQP5-mediated responses as compared with hypotonic saline. Dealcoholized Cabernet Sauvignon also blocked all three classes of AQPs as compared with hypotonic saline, showing that AQP inhibition was not an indirect effect of ethanol. Effects of dealcoholized Cabernet Sauvignon were not significantly different from normal Cabernet Sauvignon for any of the AQP classes. Chardonnay showed a blocking effect on AQP1 and AQP5 responses that was not evident for AQP2. Effects of dealcoholized Chardonnay were not significantly different from those of normal Chardonnay for any of the AQP classes.

6.

Boxplot compilations of AQP swelling responses to Cabernet Sauvignon and Chardonnay wines with and without dealcoholization. Swelling rates for the wine nonwine (control) treatment groups are shown for (A) AQP1-, (B) AQP2-, and (C) AQP5-expressing oocytes. Statistically significant differences were determined by Kruskal–Wallis with Bonferroni tests using a corrected significance level of 0.005­(*), n = number of replicates.

Given that the effects of wines on AQP1 responses indicated a link to tannin levels (Figures –), concentration-dependent effects of single archetypal metal and organic astringents (alum sulfate and tannic acid, dissolved in water) were tested in vivo and in vitro (Figure ). Samples rated by a human sensory panel using a labeled magnitude scale showed that increasing alum concentrations resulted in dose-dependent increases in astringency intensity compared to water alone, at concentrations greater than 0.25 mM (Figure A). In AQP1-expressing oocytes, increasing the alum concentration showed a trend toward greater inhibition of water channel activity, with a significant antagonist effect evident at 2 mM (Figure B). Similarly, tannic acid caused a dose-dependent increase in rated human perception of astringent intensity (Figure C), and a trend to blocking AQP1 water channel activity that was statistically significant at 2 mM (Figure D).

7.

Effects of metal and organic astringent compounds on human astringency ratings and AQP1 swelling responses. (A) Scaled Labeled Magnitude Scale (LMS) intensity ratings for astringency of aluminum sulfate (“alum”) in solution. (B) AQP1 osmotic swelling effects with and without aluminum sulfate and recovery postexposure. (C) Scaled LMS intensity ratings for astringency of tannic acid. (D) Effect of tannic acid on AQP1 swelling responses. Significance determined by Kruskal–Wallis test with Bonferroni corrected significance (*) at a level of p < 0.005 in (A)–(C) and p < 0.008 in (D), n = number of replicates.

Discussion

Aquaporin Expression in Human Tongue Tissue

Our immunohistochemical analysis showed that AQP1, AQP2, and to a lesser degree AQP5 are expressed in the sensory domains of the human tongue. These results are generally similar to previous findings in rodent models, though in rat AQP5 was predominantly localized to the apical membranes of taste bud cells, along with lower amounts of AQP1, while AQP2 was mainly in the basolateral taste cell membranes. − Our results suggest that AQP1 and AQP2 are the major subtypes in humans, with low levels of AQP5; future work using confocal microscopy will be needed to define the subcellular localization patterns of the AQP classes in human tastebuds. Results in hand provide evidence that AQP classes are present in human tongue tissue and salivary glands and suggest these AQPs could contribute to both gustatory signaling and saliva release, expanding our understanding of AQPs as channels not involved just in fluid homeostasis, but in sensory perception. ,

A Role for Aquaporins in Astringency and Mouthfeel Perception

Our functional assays using oocytes expressing human AQPs demonstrated that AQPs are responsive to key beverage constituents, including various polyphenolics. AQP1 and AQP5 showed statistically significant evidence of sensitivity when exposed to wines, with AQP5 being more responsive to tannic acid and the red wine tannin extract. AQP1 also showed robust responses to wines, to tannin added to wine, and to salines with tannic acid and alum but did not respond to the extracted red wine tannin alone in water, possibly suggesting a role for other wine constituents such as organic acids as additional influences. Antagonism of water channels in both taste and nontaste cells could generate a tannin-induced restriction of water movement into oral cells and tissues, possibly facilitating the drying sensation that is archetypal for astringency. , No evidence emerged for the compounds tested here that implicate a role for AQP2 in mouthfeel signaling, but a broad array of astringents and tastants remains to be considered in future studies.

AQP1 and AQP5 showed strong inhibitory effects caused by tannins, which are well-known to contribute to the perception of astringency in red wines. , Effects on AQP1 responses were closely associated with measured levels of wine polyphenolic compounds such as methylcellulose precipitable tannin (MCPT) and tannin molar mass (MM), which also were strongly associated with astringency ratings by human tasting panelists (details of the sensory properties and chemical composition of the wine samples have been compiled previously). Levels of extracted tannins in red wine also were directly associated with the increased friction coefficient of saliva-wine mixtures, consistent with a dry sensation resulting from reduced lubrication. Gradient shrinking responses of AQP1 to alum sulfate and tannic acid further support the hypothesis that these water channels may play an important role in gustatory signaling by interacting with astringents to facilitate mouthfeel changes and possibly salivary flow. Together these findings highlight a potential molecular mechanism in which AQPs could mediate astringency perception in response to agents that block water entry into oral tissues, contributing to cell shrinkage and potentially explaining the rapidity of the onset of drying sensations that are typical of astringent stimuli. Notably, dealcoholized wine samples also blocked all three classes of AQPs as compared with hypotonic saline, showing that AQP inhibition was not an indirect effect of ethanol (such as altering membrane fluidity) at the levels tested, but instead was due to nonvolatile wine components, such as tannins and other compounds, which were not removed in the alcohol-removal process.

Tannins and AQP Function: A Molecular Mechanism for Astringency?

Although astringency is often characterized as a singular percept, it is thought to be composed of subqualities, “drying”, “roughing” mouthfeel, and “puckering”, the last overlapping with facial sensations of concomitant sourness. , The shrinking, drawing, or puckering of the epithelium is emphasized by the American Society for Testing Materials as standard terminology. Primarily, astringent phenomena have been attributed to the aggregation and precipitation of salivary proteins and have been thought to act mainly by reducing salivary lubrication due to the interaction of salivary proline-rich proteins with tannins aggregating to form complex precipitations altering lubricity, an interpretation recently supported using tribology of saliva exposed to wine. Although increased friction between oral surfaces can explain the “roughness” in mouthfeels after repeated astringent exposures, it does not account for the rapid sensations of shrinking and drawing, often encountered at first exposure when tasting “dry” red wines. The associations between AQP1 responses, red wine astringency ratings, and tannin levels appear to point to a role for AQP1 and AQP5 in mouthfeel signaling beyond saliva lubrication changes alone. The correlations between human taste perception and AQP1 responses further tested with defined reconstituted samples with tannin extract, tannic acid, and alum corroborate the potential role of AQP1 in mediating astringency perception. Evidence here supports the proposed role of AQPs in water sensing as a salient modality concomitant with astringency. , This novel sensory mechanism could act directly at the level of signal transduction in taste bud sensory cells and more broadly in the volume regulation of oral tissues.

Published work has suggested astringency is a tactile sensation, but receptor-based models for astringent detection such as the transmembrane mucin hypothesis have been suggested. A conundrum that remains unexplained for receptor-mediated pathways is the continually increasing intensity that is experienced with repeated astringent exposures, running counter to a typical decline characteristic of sensory receptor adaptation. Progressively increasing intensities of astringency experienced in wine tasting could be explained by an AQP-mediated mechanism; a testable hypothesis would be that the beverage imposes osmotic pressure to draw water out of sensory cells, and when coupled with extracellular tannins blocks water from re-entering the cell through AQPs, leading to cumulative cell shrinkage and epithelium constriction, in turn perturbing mechanosensitive trigeminal nerves below the oral epithelium. A direct role for taste receptor cells is supported by prior work demonstrating that astringents elicit taste signals in the chorda tympani in addition to tactile responses, , framing a concept that is compatible with AQP as key components in a putative transduction pathway as outlined by Gilbertson, Baquero, et al. Importantly, the interaction of tannins with AQP5, abundant in the human salivary gland, also could impair saliva production and composition, further compounding the perception of astringency due to increased ratio of tannin to saliva present in the mouth.

Work here showed the archetypal metal astringent alum sulfate antagonized AQP1. Alum at high concentrations has been shown to precipitate salivary protein. Effects of alum on salivary mixture physical properties are less clear. Vardhanabhuti, Cox, et al. showed increased friction over time using tribology; conversely measurable effects of alum on saliva friction or in human testing protocols were not found. Together these reports and our findings supporting a multimodal signaling pathway for astringency may help explain why prior investigators have suggested subtle qualitative differences between the perception of alum as compared to plant or wine tannins. ,

Strengths and Limitations

The major contributions of this study are the demonstration that known classes of AQPs expressed in tongue tissue of lower mammals also are present in human tongue and novel insights that link AQP-mediated responses to human mouthfeel perception, demonstrated by trends indicating dose-dependent sensitivity to archetypal beverage compounds. Interdisciplinary approaches spanned from a molecularly defined expression system to a human in vivo model. Understanding gustatory and somatosensory signaling has a direct relevance to food and beverage sensory science.

Limitations of this work include the use of a single normal human tongue tissue sample for histology, which could affect the generalizability of the findings. However, expression and cellular localization of the same classes of AQPs (AQP1, AQP2, and AQP5) have been well characterized previously in other mammals as reviewed in the Introduction; thus, the main contribution of work here was to show these findings could be extended to humans. Fresh-frozen specimens of human tongue tissue without substantial cancer pathology or any history of chemotherapy and radiation treatments were limited to two at the OriGene tissue bank, and only one of those samples was sectioned in an orientation that included sensory regions. Further work is warranted to expand sample sources and define the patterns of AQP expression and subcellular localization in human taste bud cells. Given the immunizations required for saliva donors and the specialized facility needed for saliva-wine tribology, only two samples assessed in this study appear to serve as useful pilot data. We did explore the use of frozen saliva samples to gain access to a larger population, but thawed saliva showed reduced viscosity and viscoelasticity and adhered poorly to PDMS surfaces, making it unsuitable for tribological experiments. Subtly distinct subqualities of astringency such as “drying” and “puckering” are difficult to dissect even with a highly trained sensory panel. The potential for introduced compounds to evoke side tastes constrains the isolation and interpretation of specific mouthfeel responses. Aquaporin channels expressed in the oocyte expression system, though an accepted model system, do not necessarily show the same pharmacological properties as channels expressed in mammalian cells, indicating that extension of the functional assays to a human cell model will be of interest. Additionally, more classes of AQPs, astringents, and tastants remain to be tested in order to represent a complete spectrum of interactions relevant to oral somatosensation and gustatory perception.

Work here demonstrated the expression of Aquaporin classes 1, 2, and 5 in human oral tissue, showing different levels of expression in the epithelium. These AQP classes tested for function in swelling assays revealed increased levels of tannins in red wine are linked to decreased AQP-mediated water fluxes and are associated with human evaluations of increased astringent intensity. We propose modulation of AQP-mediated water movement could be critical for mechanisms of mouthfeel perceptions of “drying” and “puckering” as characteristics of astringency. This discovery could help reconcile conflicting models of astringency perception that have been debated since the 1950s, providing a mechanism for both tactile- and taste-dependent pathways of signaling. Importantly, the proposed role for AQPs in taste perception does not contradict established roles of astringents in causing salivary protein aggregation and might add insight into mechanisms that reduce saliva secretion and lubrication via block of AQP5. Aquaporins appear to be important components of gustatory and oral somatosensation functions that are fundamental for maintaining health and wellness.

Glossary

Abbreviations used

AQP

aquaporin

FFPE

formaldehyde-fixed paraffin-embedded

HEPES

4-(2-hydroxyethyl)­piperazine-1-ethanesulfonic acid

LMS

labeled magnitude scale

MCPT

methylcellulose precipitable tannin

MM

tannin molar mass

μ

tribology friction coefficient

PBS

phosphate-buffered saline

PLS-R

partial least squares regression

QDA

quantitative descriptive analysis

Additional data that support the findings of this study are available from the corresponding authors upon reasonable request for processed data sets including sensory scores, tribological measures, and aquaporin responses across experimental conditions.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jafc.5c07124.

• Supplementary Figure 1 (PDF)

• Description of the Excel file (PDF)

• Excel data file that contains point estimates supporting Figures – , including data from wine chemical composition analyses, and means and n values for AQP swelling assays (XLSX)

D.E.N. provided curated wines and compound samples for experimental testing, designed, set up and led the sensory panel tasting evaluations with assistance from B.K., analyzed the tasting intensity results, compiled the sensory data and ran the cross-correlation statistical analyses. A.J.Y. designed and performed the oocyte swelling assays, and carried out aquaporin data analysis and compilation. S.K. carried out the immunohistochemistry experiments and imaging, with assistance from B.K. S.W. designed and carried out the tribology experiments, and analyzed data with assistance from D.E.N. E.P. supervised histology protocols and assisted with analysis and interpretation of immunolabeled images. D.E.N. and A.J.Y. contributed to the primary manuscript writing; all authors reviewed, edited and approved the final manuscript.

This research was funded by The University of Adelaide Academic support and the Australian Wine Research Institute (AWRI))-University of Adelaide Collaborative Research Investment Fund Support from Wine Australia was made possible with levies from Australia’s grapegrowers and winemakers, and matching funds from the Australian Government. The AWRI is a member of the Wine Innovation Cluster in Adelaide, SA.

The authors declare no competing financial interest.