A bioinspired in vitro bioelectronic tongue with human T2R38 receptor for high-specificity detection of N-C=S-containing compounds

Talanta 199 (2019) 131–139

Contents lists available at ScienceDirect

Talanta journal homepage: www.elsevier.com/locate/talanta

A bioinspired in vitro bioelectronic tongue with human T2R38 receptor for high-specificity detection of N-C=S-containing compounds

T

Chunlian Qina,b,1, Zhen Qina,f,1, Dongxiao Zhaoc,1, Yuxiang Pana, Liujing Zhuanga, Hao Wana,b, ⁎ ⁎⁎ Antonella Di Piziod,e, Einav Malachd, Masha Y. Nivd, Liquan Huangc, , Ning Hua,b, , ⁎⁎ Ping Wanga,b, a

Biosensor National Special Laboratory, Key Laboratory for Biomedical Engineering of Education Ministry, Department of Biomedical Engineering, Zhejiang University, Hangzhou 310027, China b State Key Laboratory of Transducer Technology, Chinese Academy of Sciences, Shanghai 200050, China c College of Life Sciences, Zhejiang University, Hangzhou 310058, China d The Institute of Biochemistry, Food Science and Nutrition, The Robert H. Smith Faculty of Agriculture, Food and Environment, The Hebrew University of Jerusalem, Rehovot 76100, Israel e Leibniz-Institute for Food Systems Biology at the Technical University of Munich, 85354 Freising, Germany f Key Laboratory of Healthy & Intelligent Kitchen System Integration of Zhejiang Province, No. 218 Binhai 2nd Road, Ningbo 315336, China

ARTICLE INFO

ABSTRACT

Keywords: Bioinspired in vitro bioelectronic tongue Human T2R38 receptor Caco-2 cells N‐C=S‐containing compounds Electric cell-substrate impedance sensing

Detection and identification of bitter compounds draw great attention in pharmaceutical and food industry. Several well-known agonists of specific bitter taste receptors have been found to exhibit anti-cancer effects. For example, N-C=S-containing compounds, such as allyl-isothiocyanates, have shown cancer chemo-preventive effects. It is worth noting that human T2R38 receptor is specific for compounds containing N-C=S moiety. Here, a bioinspired cell-based bioelctronic tongue (BioET) is developed for the high-specificity isothiocyanate-induced bitter detection, utilizing human Caco-2 cells as a primary sensing element and interdigitated impedance sensor as a secondary transducer. As an intestinal carcinoma cell line, Caco-2 endogenously expresses human bitter receptor T2R38, and the activation of T2R38 induces the changes of cellular morphology which can be detected by electric cell-substrate impedance sensing (ECIS). After configuration and optimization of parameters including timing of compound administration and cell density, quantitative bitter evaluation models were built for two well-known bitter compounds, phenylthiocarbamide (PTC) and propylthiouracil (PROP). The bitter specific detection of this BioET is inhibited by probenecid and U-73122, and is not elicited by other taste modalities or bitter ligands that do not activate T2R38. Moreover, by combining different computational tools, we designed a ligand-based virtual screening (LBVS) protocol to select ligands that are likely to activate T2R38 receptor. Three computationally predicted agonists of T2R38 were selected using the LBVS protocol, and the BioET presented response to the predicted agonists, validating the capability of the LBVS protocol. This study suggests this unique cell-based BioET paves a general and promising way to specifically detect N-C=S-containing compounds that can be used for pharmaceutical study and drug development.

1. Introduction Taste, also called gustation, constitutes mammalian chemosensory together with olfaction, and acts as the detector of nutriment and toxins in behaviors such as feeding and breeding [1,2]. Among the five basic taste qualities, the bitter taste is considered to identify toxins, though

no strong relation between bitterness and toxicity was found in a recent quantitative study [3]. In mammalian taste system, bitter compounds activate a subset of taste buds type II cells which express bitter taste receptors [4,5]. These receptors belong to a large family of G-protein coupled receptors (GPCRs), and are called TAS2Rs or T2Rs [6]. The activation of T2Rs results in the raise of intracellular Ca2+ via classic

Corresponding author. Corresponding authors at: Biosensor National Special Laboratory, Key Laboratory for Biomedical Engineering of Education Ministry, Department of Biomedical Engineering, Zhejiang University, Hangzhou 310027, China. E-mail address: [email protected] (P. Wang). 1 These authors contribute equally to this work. ⁎

⁎⁎

https://doi.org/10.1016/j.talanta.2019.02.021 Received 1 October 2018; Received in revised form 29 January 2019; Accepted 4 February 2019 Available online 06 February 2019 0039-9140/ © 2019 Elsevier B.V. All rights reserved.

Talanta 199 (2019) 131–139

C. Qin, et al.

inositol triphosphate/diacylglycerol (IP3/DAG) pathway and the release of neurotransmitters such as ATP [1,7]. Subsequently, taste afferent fibers are modulated by ATP, and thus transmit taste information to the central neural system, where the perception of bitterness eventually forms [8]. To mimic the mammalian taste system, conventional electronic tongues (ETs) were proposed to detect bitter compounds with various structures [9,10]. The sensing arrays are usually composed of chemical sensors such as ion selective electrodes, whereas ETs have the limitation for bitter specific detection [11–13]. Considering the merits of functional components in mammalian taste system, bioelectronic tongues (BioETs) were developed using taste epithelium of rats or mice germ cells which can endogenously express T2Rs [14–16]. Meanwhile, the concept of in vivo BioET was also proposed based on the whole rat taste system and electrophysiological recording in gustatory cortex [17]. However, the sensing elements of such BioETs are derived from rodents, and they actually mimic the taste perception of rodents. Differences do exist in the sense of taste between human and rodents, for example, rodents rarely identify the taste of aspartame, which is a common artificial sweetener that has bitter off-taste for humans [18]. To bridge the taste gap between species, novel BioETs are developed based on human taste receptors. Heterologous expression systems, such as human embryonic kidney cells (HEK-293), are commonly engineered into taste cells [19,20]. After transfection, HEK-293 cells express human taste receptors and act as type II cells in taste buds. Once activated by bitter compounds, morphological changes of cells occur such as membrane ruffles, and these changes can be detected by electric cellsubstrate impedance sensing (ECIS). However, HEK-293 cells have weak attachment on sensor surface. To improve the attachment, sensor

pre-treatments such as poly-lysine and laminin modification are generally employed, which results in loss of sensitivity. Furthermore, bitter response of HEK-293 cells requires both functional T2Rs, as well as the integrity of the whole IP3/DAG pathway [20,21]. Due to this reason, Gα16gust44 should be co-expressed in HEK-293 cells, which significantly lowers the transfection efficiency. To overcome the difficulties of conventional expressing systems mentioned above, it is necessary to find a human cell line which endogenously expresses T2Rs. Caco-2 cell, a human intestinal cell line, was reported to express the T2Rs [22]. Compared with cells expressing the whole repertoire of T2Rs, Caco-2 cells only express one known bitter receptor, T2R38. Moreover, due to the properties of carcinoma cells, the confluence and strong attachment properties make them suitable for ECIS, and no surface treatment is required [23,24]. Therefore, Caco-2 cell is a good candidate to detect the bitter compounds targeting at T2R38. Here, we developed a bioinspired BioET using Caco-2 cells and ECIS to specifically detect T2R38 agonists (Fig. 1a). Among the 25 known human bitter receptors, T2R38 is selectively activated by bitter compounds with isothiocyanate moiety (NCS), including the well-known bitter compounds phenylthiocarbamide (PTC) and propylthiouracil (PROP) [6]. Once the receptor is activated, a series of intracellular cascade reaction occurs, resulting in morphological changes of cells such as membrane ruffle. These changes can be detected by electric cellsubstrate impedance sensing (ECIS) (Fig. 1b and c). In order to test the BioET, PTC and PROP were screened on it, and the results were also verified by calcium imaging. Moreover, the specificity of the BioET was further investigated by other typical bitter compounds and other taste qualities that cannot activate T2R38.

Fig. 1. High-specificity bitter detection using cell-based bioinspired bioelectronic tongue. (a) Schematic of bioinspired bioelectronic tongue for detection of isothiocyanate bitterness. (b) Activation of IP3/DAG pathway induces increased intracellular Ca2+ and membrane truffles. (c) Application of bitter compounds induces morphological and attachment changes and increases cell-electrode impedance. 132

Talanta 199 (2019) 131–139

C. Qin, et al.

Finally, we use chemoinformatics methods to predict new potential T2R38 agonists and validated the predictions with BioET [25,26].

solution was added to cover the cells and incubated at 37 °C for 30 min (4) Working solution was removed and rinsed with HBSS. Then HBSS was added to cover the cells and incubate for 20 min (5) Taste substances were added and then it was observed under a fluorescence microscope. Caco-2 cells in 96-well plate were loaded with a mixture of Calcium sensitive dye Fluo-4/AM (maximum absorption: 494 nm, maximum emission: 506 nm) and Pluronic F-127 when confluence reaches 80%. The calcium signal was recorded by fluorescence microscopy and images are sampled 0.5 frame/s. The recorded videos were further analyzed by MATLAB (MathWorks, MA). All the procedures were protected from bleaching effects in the dark room.

2. Materials and methods 2.1. Reagents Minimum essential medium (MEM) and Penicillin-Streptomycin (PS) for cell culture were purchased from Genom Biosciense (Hangzhou, China). Fetal bovine serum (FBS) was purchased from Sijiqing Bioscience (Hangzhou, China). Dimethyl sulfoxide (DMSO), phenylthiocarbamide (PTC), propylthiouracil (PROP), thiosinamine, αNaphthylthiourea (ANTU), acetylthiourea, and phenyl isothiocyanate were purchased from Sigma-Aldrich (St. Louis, MO, USA). Other tastants including sucrose, NaCl, HCl, and monosodium glutamate (MSG) were purchased from Sinopharm Chemical Reagents Co. Ltd (Shanghai, China).

2.5. PTC and PROP analogues prediction by computational methods To select ligands that are likely to activate T2R38 receptor, we designed a ligand-based virtual screening (LBVS) protocol, combining different computational tools: the simple 1D molecular descriptors filter, 2D fingerprints-based molecular similarity analysis, and the more sophisticated 3D ligand-based pharmacophore modeling. The protocol has been applied to an external database made of molecules from BitterDB that were not tested against T2R38 [35].

2.2. Cell culture Caco-2 cells are cultured in MEM containing 20% FBS and 1% PS. Digestion is performed after cells reach high confluence (> 80%) in T25 flask. Then, cell suspension is centrifuged at 1100 rpm for 5 min and the supernatant is removed. Suspensions of different cell density are prepared and directly planted into wells of the E-plates (ACEA Biosciences Inc., Hangzhou, China). It is noteworthy that no pre-treatment of the sensor surface and the Caco-2 cells is needed. Additional MEM is used to replenish the 200 μL wells.

(1) Substructure search with the moiety NCS as query. (2) Similarity filter. Molecules were filtered according to the molecular properties calculated for known T2R38 agonists [35]: Molecular Weight (MW), Rotatable Bonds (RBs), Hydrogen Bond Donors (HBDs), lipophilicity as atomic log P (AlogP), number of heteroaromatic rings (NumHetRing). Applying this filter (89 ≤ MW ≤ 470, RBs ≤ 7, HBDs < 5, −2 < AlogP ≤ 4.4, NumHetRing ≤ 1). (3) Fingerprints: Using MOLPRINT2D fingerprints we calculated the nearest neighbor similarities of filtered compounds from steps 1 and 2 towards known actives. (4) 3D similarity pharmacophore model: 3D similarity was evaluated by ligand-based pharmacophore models, i.e. 3D representations of the essential chemical features necessary to exert optimal interactions with the biological target, and to trigger its biological response. The performance of generated pharmacophore models was evaluated by assessing their ability to successfully identify all known agonists, while rejecting true inactive molecules. The pharmacophore model that showed the highest performance in terms of specificity (0.65), has the following features: two hydrogen bond donors, one H-bond acceptor (mapping the thiourea moiety) and one general hydrophobic feature (S-Fig. 1).

2.3. Cellular impedance recording The detection principle of ECIS has been introduced in former literatures [27–33]. In brief, cells on interdigitated electrodes of ECIS can be simplified as RC networks and thus increase the cell-electrode impedance by impeding the ion current. The cell-electrode impedance is positively correlated with the number and attachment of cells. Besides, the cellular morphology changes will also result in the impedance changes. To quantify and normalize the impedance across channels, cell index is calculated as:

CI =

Zcell (fi ) Zbaseline (fi )

1

where fi represents the scanning voltage frequency of the impedance detection, Zbaseline(fi) and Zcell(fi) are the frequency-dependent impedance without and with cells, respectively. A sinusoidal voltage with amplitude of 20 mV and frequency of 10 kHz is applied on the interdigitated electrodes (IDEs) of ECIS [34]. Normalization is performed based on the CI value right before taste stimulation. Besides, to characterize the intensity of the impedance response, the area under curve (AUC) enveloped by the CI response peak and baseline is calculated. The baseline is defined as the connection between the start and end point of the response peak.

Compounds were then ranked by normalizing and summing the scores from the two 2D (fingerprints) and 3D (pharmacophore) similarity. Thiosinamine, ANTU and phenyl isothiocyanate were selected to validate the function of the BioET. 2.6. Statistical analysis All results were presented as mean and standard deviation. Data were performed with an unpaired Student's t-test using Graphpad Prism 6 (GraphPad Software Inc., USA) or Excel 2013 (Microsoft, USA). P < 0.05 was considered statistically significant.

2.4. Calcium imaging

3. Results and discussion

In this study, Fluo 4-AM was used as a fluorescent probe for calcium ion. The cells were examined with a fluorescence microscope at an excitation wavelength of 494 nm and an emission wavelength of 516 nm. The experimental steps are as follows: (1) 5 μM Fluo-AM working solution was prepared with HBSS. To prevent Fluo 4-AM from aggregated in buffer and promote its entry into cells, 20% (w/v) Pluronic F-127 dissolved in DMSO was added to the working solution with a final concentration of 0.04% (w/v). (2) The pre-cultured Caco-2 cells were removed from the MEM and washed with HBSS to prevent AM body from degradation by the esterase. (3) 1 mL of working

3.1. Construction and optimization of cell-based bioelectronic tongue To construct the cell-based BioET, the optimization should be performed. To unify the conditions, DMSO is necessary to dissolve the PTC and other analytes in water-based MEM. To determine if DMSO will affect the viability of Caco-2 cells, 0.1% and 1% DMSO were used for toxicity test. As shown in Fig. 2a, DMSO at such concentrations has no effect on the cell growth. CI values at typical time (10, 40, 80, 250 min 133

Talanta 199 (2019) 131–139

C. Qin, et al.

Fig. 2. (a) Effects of different concentrations of DMSO on CI values. (b) Comparison of CI values of different DMSO concentrations at 4 representative time points. Results are given as the mean ± SD (n = 3). Student t-test, P > 0.2 for all pairs. (c) Fluorescence of a single cell before and after administration of 1 mM PTC. (d) Fluorescence intensity of single cells in response to 1 mM PTC. (e) The effect of bitter compounds and receptor inhibitor on intracellular Ca2+ fluorescence intensity. Results are given as the mean ± SD (n = 4). Student t-test, ***p < 0.001 compared with control.

after stimulation) were derived and compared. CI of the first three time points can describe the CI peaks generated by activation of T2Rs and CI of the last can measure the effect on cell vitality. Cell viability and statistical analysis indicate that DMSO has no effect on cell vitality and GPCR of Caco-2 cells membrane (Fig. 2b, n = 3, Student t-test, P > 0.2 for all pairs). And compared with 1% DMSO, 0.1% DMSO was unable to dissolve the required PTC and PROP completely. Therefore, 1% DMSO is biocompatible for cells to establish the cell-based BioET. Genes encoding the T2R38 receptor are diverse, and the three most common polymorphisms observed in T2R38 occur at amino acid position 49, where either a proline or an alanine is encoded, at position 262, where either an alanine or a valine is encoded, and at position 296, where either a valine or an isoleucine is encoded, giving rise to two frequent haplotypes, PAV and AVI [36]. The gene polymorphism of T2R38 results in tasters (T2R38-PAV, which can respond to bitter compounds like PTC and PROP) and non-tasters (T2R38-AVI, which cannot respond to PTC or PROP). To ensure the expression of T2R38-PAV in Caco-2 cells, calcium imaging was utilized [37]. Fluorescence of a single Caco-2 cell changes with PTC treatment as shown in Fig. 2c. The intracellular Ca2+ increases dramatically due to the release of Ca2+ from endoplasmic reticulum following the activation of T2R38-PAV (Fig. 2d and S-Movie 1). Comparing with the control group in Fig. 2e, fluorescence intensity significantly increases under PTC or PROP treatment (Student t-test, n = 4, P < 0.001) and lasts over 60 s. To make sure the cellular response is induced by activation of T2R38-PAV, 1 mM receptor inhibitor

p-(dipropylsulfamoyl) benzoic acid (Probenecid, Prob) is applied to Caco-2 cells before bitter stimulation [38]. The results demonstrate that calcium response cannot be triggered by PTC or PROP after probenecid inhibition, proving that the activation involves T2R38-PAV. Thus the expressing and functionalization of T2R38-PAV is guaranteed. Supplementary material related to this article can be found online at doi:10.1016/j.talanta.2019.02.021. Subsequently, the key parameters of the BioET are optimized including the timing of stimulation and the density of Caco-2 cells. Two phases: (I) attachment phase, (II) growth and proliferation phase, were observed after cells were seeded on the interdigitated electrodes (SFig. 2a). Theoretically, it is better to apply the stimuli in phase (II) to exclude the instability caused by cell attachment in phase (I). By analyzing the growth rate of cell growth, it is obvious that the growth is stable (after 6 h) (S-Fig. 2b). Previous studies have demonstrated that cell density influences the performance of cell-based BioET [34]. Thus Caco-2 cell density in the range of 10–100 k cells/well was tested by 1 mM PTC at 6 h. 1 mM PTC induces the specific peaks of growth curves (S-Fig. 2c). Such phenomenon is the consequence of increased intracellular Ca2+ through IP3/ DAG pathway following the activation of T2R38. Other than that, translocation of Paxillin was observed as well as truffles of cell membrane [19,29]. As a result, adhesive and morphological changes occur in cell-electrode interface which leads to impedance and CI changes. Considering the peak time and baseline, area under curve (AUC) was defined and calculated to evaluate the PTC-induced response. 134

Talanta 199 (2019) 131–139

C. Qin, et al.

Fig. 3. (a, b) CI responses to different concentrations of PTC and PROP. (c, d) Dose-dependent response between AUCs and concentrations of PTC and PROP. (e, f) CI response inhibited by 0.5 mM probenecid and 1 μM U73122. Results are given as the mean ± SD (n = 4).

stimulation. The detection ranges of the cell-based BioET are 1 μM1 mM for both PTC and PROP. The limits of detection (LODs) were also calculated using triple-standard-deviation method, and the results are 0.09352 and 0.8404 μM for PTC and PROP, respectively. According to data from the BitterDB, the effective concentrations of calcium imaging are 0.04 and 0.11 μM for these two compounds [6,39]. Compared with calcium imaging of transfected HEK cells, the data collected from the cell-based BioET is within the same order of magnitude, and the sensing capability of the BioET is verified. Therefore, the BioET can mimic human perception of PTC and PROP, and has the potential to quantify the bitter intensity of related compounds. To validate the PTC and PROP interaction with T2R38, 0.5 mM receptor inhibitor probenecid was used to investigate the response of the BioET [38]. After probenecid treatment, T2R38 on Caco-2 cells membrane is inhibited, and no response is observed compared with the control groups (Fig. 3e and f). Besides, 1 μM PLCβ inhibitor U73122 was applied to block pathway of GPCR and ablate the bitter response [40], and no response is observed compared with the control group. The result proves that the bitterness response of the impedance sensor is exactly caused by the activation of T2R38. Thus, probenecid and U73122 inhibition assay validates the capability of the BioET to act as conventional calcium imaging to indicate the activation and inhibition of T2R38.

Among the six densities, Caco-2 cells with density of 40 k/well have the largest AUC value, which implies the highest response after stimulation (S-Fig. 2d). Optical image also demonstrates the similar results for optimization of cell density (S-Fig. 3). Cells only cover a small surface of ECIS when the cell density is lower than 20 k/well, and fewer cells result in the lower responses due to the limited sensing elements. The contact inhibition is observed where CI value decreases when the cell density is higher than 60 k/well. Therefore, Caco-2 cells with a density of 40 k/well were cultured on the sensor surface, and bitter stimuli were applied after 6 h. 3.2. Quantitative detection of known T2R38 ligands To determine the sensitivity of the BioET, two known T2R38 bitter ligands PTC and PROP were applied, and Caco-2-cell-based BioET was used to quantitatively detect known ligands PTC and PROP from 1 μM to 1 mM. Normalized cell growth curves are shown in Fig. 3a and b. Compared with the control group, PTC and PROP with different concentrations induce the dose-dependent responses by observing CI peaks. By calculating the extreme points, the response phase and area were determined. The average delays of CI peaks are 3.07 ± 0.82 min and 1.50 ± 0.55 min (mean ± standard deviation) for PTC and PROP respectively. However, no significant changes in the response duration of the two ligands were observed. Such delays should be the result of no stir being exerted to bitter application in order to prevent turbulence which may alter the attachment of Caco-2 cells and further impact on CIs. CI peaks are difficult to quantify, so AUCs of CI peaks were selected to evaluate the compound response instead of raw peak magnitudes. As shown in Fig. 3c and d, the response of AUC has a positive correlation with the stimulation concentration of PTC and PROP. The results prove AUC to be an appropriate indicator to characterize the intensity of

3.3. Specificity analysis of cell-based bioelectronic tongue The specificity of this cell-based BioET was verified by other taste compounds and other bitter compounds. Tastants of different taste qualities were tested: 3 mM HCl (sour), 0.1 M sucrose (sweet), 0.1 M NaCl (salty) and 3 mM MSG (umami). 0.1 mM PTC is taken as bitter stimulation and the concentration is lower than the other four tastes considering the lower taste threshold of bitterness. As shown in Fig. 4a, 135

Talanta 199 (2019) 131–139

C. Qin, et al.

no response is induced by taste compound stimulation other than 0.1 mM PTC. This result demonstrates the specificity of the BioET for bitterness detection. More quantitative study shows that the response peaks induced by 0.1 mM PTC are significantly higher than other taste compound within the same periods (Fig. 4b, Student t-test, PbitterPbitter-sweet < 0.0001, Pbitter-salty < 0.0001, Pbittersour < 0.0001, umami < 0.0001, n = 4). In terms of the 25 human bitter receptors, T2R38 mainly detects compounds with isothiocyanate moiety (N-C=S), which only occupies a small part of the bitter space [4]. Typical bitter compounds such as quinine and denatonium benzoate (dena) activate receptors including T2R4, T2R10, T2R39 and T2R47 rather than T2R38 [6]. Considering their versatility and broad response domain, 1 mM quinine and denatonium benzoate were applied as analytes for specificity tests compared with 0.1 mM PTC. The results demonstrate that no response is induced by these two control compounds and no effect on cell proliferation and vitality is observed (Fig. 4c). Similarly, quantitative study is performed by calculation of AUCs, and significant difference exists when comparing PTC-induced AUC with other bitterness compounds (Fig. 4d, Student t-test, PPTC-quinine < 0.0001, PPTC-dena < 0.0001, n = 5). It is reasonable to conclude that the BioET can also be used in the exclusion of other non-T2R38 activated compounds (such as caffeine which is the same with quinine to activate T2R10 and T2R14) [6]. This BioET has very good specificity with negligible response to other taste qualities and bitterness compounds without T2R38 activation. In contrast to previous studies using either germ cell-based or in vivo BioETs, this cell-based BioET has high specificity on receptor level, and explains the sophisticated bindings between bitter receptors and ligands. The superiority of this BioET is due to sole receptor expressed Caco-2 cells. Meanwhile, comparing with BioETs using other heterologous expression systems such as bioengineered HEK-293 cells, this expression system avoids the pretreatment of both cell and sensor surface, such as cell transfection and sensor surface modification.

known T2R38 ligands have been found, including PTC and PROP, and most of them contain the isothiocyanate moiety. Considering the structural similarity to known T2R38 agonists, three PTC and PROP analogues were chosen by computational methods as candidate ligands of T2R38 whose taste qualities are not yet determined, including thiosinamine (also called allylthiourea), α-Naphthylthiourea (ANTU), and phenyl isothiocyanate. It is reported that above-mentioned compounds can cause the decomposition of liver glycogen in rats, and ANTU has been used as a rodenticide [42]. Therefore, taking safety issues into account, biological assays and BioETs should be used to discriminate their taste qualities and corresponding receptors instead of human taste panel. We therefore checked the response of the cell-based BioET assays to thiosinamine, ANTU, and phenyl isothiocyanate (predicted T2R38 ligands) and acetylthiourea (known T2R38 ligand) (Fig. 5a). Since no prior knowledge is received about their putative effective concentrations, the test concentration range was set to 1 μM–1 mM taking PTC as the reference. The screening strategy aims at comparing compound-evoked responses to the control group of mixture containing 0.1 mM compounds as well as 1 mM probenecid, and a candidate can be determined as T2R38 ligands only if it induces receptor-specific CI peaks and the peaks can be inhibited by 1 mM probenecid at the same time. Such conditions can eliminate the interference of other GPCRs and attribute the CI peaks to the activation of T2R38. The results are shown in Fig. 5b-e, and the three predicted T2R38 ligands (thiosinamine, ANTU and phenyl isothiocyanate) are capable of inducing CI peaks with the same response as acetylthiourea in spite of the difference in curve shapes. It is noteworthy that CI peaks evoked by 0.1 mM of these candidates can be significantly inhibited by 1 mM probenecid, which indicates that all three analogues can activate T2R38 on Caco-2 cells membrane. Therefore, we can conclude that thiosinamine, ANTU and phenyl isothiocyanate are undiscovered bitter compounds, and T2R38 is one of the receptors that they activate. In addition, different concentrations of compounds can induce different magnitudes and AUCs of CI curves. With the help of chemoinformatics tools, the efficiency of identification and screening of undiscovered ligands can be improved dramatically. In fact, computational methods were recently developed and successfully used both to predict the bitterness of compounds, i.e. BitterPredict, and to predict compounds that activate specifically one or more bitter taste receptors [43,44]. Apart from BioETs, biological methods including calcium imaging and IP1 assay have been used in the validation of predictions. However, these methods require complex

3.4. Screening of T2R38 agonists The screening of bitter compounds has long been a notorious job in research and industry especially when facing toxic compounds which cannot be tested by human taste, not to mention the pairing of receptors and ligands. Fortunately, the development of chemoinformatics and molecular modeling provides new predictive methods based on the theory “similar structure derives similar taste quality” [25,41]. However, such predictions need experimental validation. Currently, 21

Fig. 4. The specificity test of cell-based BioET. (a) CI responses and (b) AUC analysis under the administration of different taste modalities. Results are given as the mean ± SD (n = 4). Student t-test, Pbitter-sour < 0.0001, PbitterPbitter-salty < 0.0001, Pbittersweet < 0.0001, umami < 0.0001. (c) CI responses and (d) AUC analysis under the bitter compounds with and without isothiocyanate group. Results are given as the mean ± SD (n = 5). Student ttest, PPTC-quinine < 0.0001, PPTC-dena < 0.0001.

136

Talanta 199 (2019) 131–139

C. Qin, et al.

Fig. 5. Bitter prediction of the cell-based BioET. (a) Chemical structures of PTC and PROP and their analogues. CI responses and dose-dependent AUC under the (b) acetylthiourea, (c) thiosinamine, (d) ANTU and (e) phenyl isothiocyanate. Among them, acetylthiourea is a known T2R38 ligand. Dash line is normalized time point.

137

Talanta 199 (2019) 131–139

C. Qin, et al.

manipulation in pre-treatment, and the use of fluorescent probes (e.g. Fluo-4/AM) or monoclonal antibodies increase the cost of detection. Since the BioET can mimic human bitter perception and function as calcium imaging in the manner of impedance detection, the cell-based system provides a new and efficient platform in the identification and screening of receptor-specific ligands. As for the high throughput of the methods mentioned above utilizing 96-well plates, this BioET can load three16-well E-plates and is promising to be equipped with higher throughput plates.

[10] J. Han, M. Bender, K. Seehafer, U.H. Bunz, Identification of white wines by using two oppositely charged poly (p‐phenyleneethynylene) s individually and in complex, Angew. Chem. Int. Ed. 55 (2016) 7689–7692. [11] L. Escuder-Gilabert, M. Peris, Review: highlights in recent applications of electronic tongues in food analysis, Anal. Chim. Acta 665 (2010) 15–25. [12] X. Cetó, N.H. Voelcker, B. Prieto-Simón, Bioelectronic tongues: new trends and applications in water and food analysis, Biosens. Bioelectron. 79 (2016) 608–626. [13] D. Ha, Q. Sun, K. Su, H. Wan, H. Li, N. Xu, et al., Recent achievements in electronic tongue and bioelectronic tongue as taste sensors, Sens. Actuators B: Chem. 207 (2015) 1136–1146. [14] F. Zhang, Q. Zhang, D. Zhang, Y. Lu, Q. Liu, P. Wang, Biosensor analysis of natural and artificial sweeteners in intact taste epithelium, Biosens. Bioelectron. 54 (2014) 385–392. [15] L. Wei, L. Qiao, G. Pang, J. Xie, A kinetic study of bitter taste receptor sensing using immobilized porcine taste bud tissues, Biosens. Bioelectron. 92 (2017) 74–80. [16] L. Hu, J. Xu, Z. Qin, N. Hu, M. Zhou, L. Huang, et al., Detection of bitterness in vitro by a novel male mouse germ cell-based biosensor, Sens. Actuators B: Chem. 223 (2016) 461–469. [17] Z. Qin, B. Zhang, L. Hu, L. Zhuang, N. Hu, P. Wang, A novel bioelectronic tongue in vivo for highly sensitive bitterness detection with brain–machine interface, Biosens. Bioelectron. 78 (2016) 374–380. [18] A. Sclafani, M. Abrams, Rats show only a weak preference for the artificial sweetener aspartame, Physiol. Behav. 37 (1986) 253–256. [19] L. Hu, L. Zou, Z. Qin, J. Fang, L. Huang, P. Wang, A novel label-free bioengineered cell-based biosensor for salicin detection, Sens. Actuators B: Chem. 238 (2017) 1151–1158. [20] K. Riedel, D. Sombroek, B. Fiedler, K. Siems, M. Krohn, Human cell-based taste perception–a bittersweet job for industry, Nat. Product. Rep. (2017). [21] M. Behrens, K. Blank, W. Meyerhof, Blends of Non-caloric sweeteners saccharin and cyclamate show reduced off-taste due to TAS2R bitter receptor inhibition, Cell Chem. Biol. 24 (2017) 1199–1204 (e2). [22] K. Yano, S. Shimizu, T. Tomono, T. Ogihara, Gastrointestinal hormone cholecystokinin increases p-glycoprotein membrane localization and transport activity in caco-2 cells, J. Pharm. Sci. 106 (2017) 2650–2656. [23] M. Achour, S. Saguem, B. Sarriá, L. Bravo, R. Mateos, Bioavailability and metabolism of rosemary infusion polyphenols using Caco‐2 and HEPG2 cell model systems, J. Sci. Food Agric. (2018). [24] T. Wolf, S.R. Baier, J. Zempleni, The intestinal transport of bovine milk exosomes Is mediated by endocytosis in human colon carcinoma Caco-2 cells and rat small intestinal IEC-6 cells–3, J. Nutr. 145 (2015) 2201–2206. [25] A. Dagan-Wiener, I. Nissim, N.B. Abu, G. Borgonovo, A. Bassoli, M.Y. Niv, Bitter or not? bitter predict, a tool for predicting taste from chemical structure, Sci. Rep. 7 (2017) 12074. [26] S. Zheng, M. Jiang, C. Zhao, R. Zhu, Z. Hu, Y. Xu, et al., e-Bitter: bitterant prediction by the consensus voting from the machine-learning methods, Front. Chem. 6 (2018) 82. [27] L. Zou, Q. Wang, M. Tong, H. Li, J. Wang, N. Hu, et al., Detection of diarrhetic shellfish poisoning toxins using high-sensitivity human cancer cell-based impedance biosensor, Sens. Actuators B: Chem. 222 (2016) 205–212. [28] L. Zou, C. Wu, Q. Wang, J. Zhou, K. Su, H. Li, et al., An improved sensitive assay for the detection of PSP toxins with neuroblastoma cell-based impedance biosensor, Biosens. Bioelectron. 67 (2015) 458–464. [29] N. Yu, J.M. Atienza, J. Bernard, S. Blanc, J. Zhu, X. Wang, et al., Real-time monitoring of morphological changes in living cells by electronic cell sensor arrays: an approach to study G protein-coupled receptors, Anal. Chem. 78 (2006) 35–43. [30] Q. Liu, C. Wu, H. Cai, N. Hu, J. Zhou, P. Wang, Cell-based biosensors and their application in biomedicine, Chem. Rev. 114 (2014) 6423–6461. [31] N. Hu, J. Fang, L. Zou, H. Wan, Y. Pan, K. Su, et al., High-efficient and high-content cytotoxic recording via dynamic and continuous cell-based impedance biosensor technology, Biomed. Microdev. 18 (2016) 94. [32] N. Hu, T. Wang, Q. Wang, J. Zhou, L. Zou, K. Su, et al., High-performance beating pattern function of human induced pluripotent stem cell-derived cardiomyocytebased biosensors for hERG inhibition recognition, Biosens. Bioelectron. 67 (2015) 146–153. [33] N. Hu, T. Wang, H. Wan, L. Zhuang, R. Kettenhofen, X. Zhang, et al., Synchronized electromechanical integration recording of cardiomyocytes, Biosens. Bioelectron. 117 (2018) 354–365. [34] K. Solly, X. Wang, X. Xu, B. Strulovici, W. Zheng, Application of real-time cell electronic sensing (RT-CES) technology to cell-based assays, Assay Drug Dev. Technol. 2 (2004) 363–372. [35] A. Dagan-Wiener, A. Di Pizio, I. Nissim, M.S. Bahia, N. Dubovski, E. Margulis, et al., BitterDB: taste ligands and receptors database in 2019, Nucleic Acids Res. 47 (2018) D1179–D1185. [36] B. Bufe, P.A. Breslin, C. Kuhn, D.R. Reed, C.D. Tharp, J.P. Slack, et al., The molecular basis of individual differences in phenylthiocarbamide and propylthiouracil bitterness perception, Curr. Biol. 15 (2005) 322–327. [37] R.J. Lee, G. Xiong, J.M. Kofonow, B. Chen, A. Lysenko, P. Jiang, et al., T2R38 taste receptor polymorphisms underlie susceptibility to upper respiratory infection, J. Clin. Investig. 122 (2012) 4145–4159. [38] T.A. Greene, S. Alarcon, A. Thomas, E. Berdougo, B.J. Doranz, P.A. Breslin, et al., Probenecid inhibits the human bitter taste receptor TAS2R16 and suppresses bitter perception of salicin, PLoS One 6 (2011) e20123. [39] A. Wiener, M. Shudler, A. Levit, M.Y. Niv, BitterDB: a database of bitter compounds, Nucleic Acids Res. 40 (2011) D413–D419. [40] D.A. Deshpande, W.C.H. Wang, E.L. McIlmoyle, K.S. Robinett, R.M. Schillinger, S.S. An, et al., Bitter taste receptors on airway smooth muscle bronchodilate by

4. Conclusion In this study, a high-specific bioinspired BioET was developed based on Caco-2 cells and interdigitated impedance sensor. This Caco-2 cell line endogenously expresses human bitter receptor T2R38 and combines with ECIS due to its strong attachment. After optimization of system parameters including timing of chemical application and cell density, quantitative detection was performed to measure known T2R38 ligands. Calcium imaging and inhibition experiments guarantee the feasibility of the cell-based BioET. Results indicate that this bioinspired system can mimic T2R38-related bitter perception. Owing to the sole expression of T2R38, receptor-specific detection was achieved. Neither tastants of other qualities nor bitterness of non-T2R38-ligands evokes receptor-specific CI peaks. Eventually, a strategy for identification and screening of undiscovered T2R38 ligands was proposed with the assistance of chemoinformatics tools. Indeed, the BioET confirmed the activation of three predicted T2R38 agonists. Based on the results, the novel BioET will be a promising platform for detection and identification of bitter compound targeting at certain bitter receptors, and can be used in the fields of pharmaceutical and food industry. Since isothiocyanates (agonists of T2R38) have shown cancer chemo-preventive properties [45], the BioET is promising in finding more agonists of T2R38. Acknowledgement This work was supported by Joint NSFC-ISF Research Fund (Grant No. 2463/16 and 31661143030) and Fundamental Research Funds for the Central Universities (Grant No. 2018FZA5018, 2018QNA5018), Major Research and Development Project of Zhejiang Province (Grant No.2017C03032, 2019C03066), China Postdoctoral ScienceFoundation Funded Project (Grant No.2017M621941). Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at doi:10.1016/j.talanta.2019.02.021. References [1] N. Chaudhari, S.D. Roper, The cell biology of taste, J. Cell Biol. 190 (2010) 285–296. [2] W. Meyerhof, M. Behrens, A. Brockhoff, B. Bufe, C. Kuhn, Human bitter taste perception, Chem. Senses 30 (2005) i14–i15. [3] I. Nissim, A. Dagan‐Wiener, M.Y. Niv, The taste of toxicity: a quantitative analysis of bitter and toxic molecules, IUBMB life 69 (2017) 938–946. [4] M. Behrens, W. Meyerhof, Vertebrate Bitter taste receptors: keys for survival in changing environments, J. Agric. Food Chem. (2017). [5] S.D. Roper, Taste Buds as Peripheral Chemosensory Processors, Semin. Cell Dev. Biol. 24 (2013) 71–79. [6] W. Meyerhof, C. Batram, C. Kuhn, A. Brockhoff, E. Chudoba, B. Bufe, et al., The molecular receptive ranges of human TAS2R bitter taste receptors, Chem. Senses 35 (2010) 157–170. [7] K.F. Medler, Calcium signaling in taste cells, Biochim. Biophys. Acta (BBA)-Mol. Cell Res. 1853 (2015) 2025–2032. [8] A. Carleton, R. Accolla, S.A. Simon, Coding in the mammalian gustatory system, Trends Neurosci. 33 (2010) 326–334. [9] Y. Hou, M. Genua, D. Tada Batista, R. Calemczuk, A. Buhot, P. Fornarelli, et al., Continuous evolution profiles for electronic‐tongue‐based analysis, Angew. Chem. 124 (2012) 10540–10544.

138

Talanta 199 (2019) 131–139

C. Qin, et al. localized calcium signaling and reverse obstruction, Nat. Med. 16 (2010) 1299. [41] W. Huang, Q. Shen, X. Su, M. Ji, X. Liu, Y. Chen, et al., BitterX: a tool for understanding bitter taste in humans, Sci. Rep. 6 (2016). [42] C.P. Richter, The development and use of alpha-naphthyl thiourea (ANTU) as a rat poison, J. Am. Med. Assoc. 129 (1945) 927–931. [43] A. Pizio, L.-M. Kruetzfeldt, S. Cheled-Shoval, W. Meyerhof, M. Behrens, M.Y. Niv, Ligand binding modes from low resolution GPCR models and mutagenesis: chicken

bitter taste receptor as a test-case, Sci. Rep. 7 (2017) 8223. [44] A. Levit, S. Nowak, M. Peters, A. Wiener, W. Meyerhof, M. Behrens, et al., The bitter pill: clinical drugs that activate the human bitter taste receptor TAS2R14, FASEB J. 28 (2014) 1181–1197. [45] Y. Zhang, Allyl isothiocyanate as a cancer chemopreventive phytochemical, Mol. Nutr. Food Res. 54 (2010) 127–135.

139