Detection of Oral Beta-Lactam Antibiotics Using a Taste Sensor with Surface-Modified Lipid/Polymer Membranes

Abstract

In our previous study, a taste sensor modified with 3-bromo-2,6-dihydroxybenzoic acid (3-Br-2,6-DHBA) exhibited significant responses to xanthine-based substances, suggesting an allosteric detection mechanism. This study investigates the potential of the 3-Br-2,6-DHBA-modified sensor membrane for detecting other drug classes. Eleven structurally diverse drugs—including caffeine, antibiotics, antivirals, analgesic-antipyretics from the WHO Model List of Essential Medicines for Children—were tested, as they were previously undetectable by a conventional bitterness sensor. Among them, amoxicillin, an oral broad-spectrum penicillin, and cefalexin, an oral cephalosporin, elicited significantly higher sensor responses when 3-Br-2,6-DHBA-modified membrane was used. To further examine this response, experiments were conducted using membranes modified with 3-Br-2,6-DHBA, 2,6-dihydroxybenzoic acid (2,6-DHBA), and benzoic acid. These tests confirmed that only 3-Br-2,6-DHBA-modified membrane produced significant responses to amoxicillin and cefalexin, suggesting that hydroxyl groups in 3-Br-2,6-DHBA contribute to allosteric effects via hydrogen bonding. Additional tests demonstrated higher responses for cefaclor and cefdinir, both oral cephalosporins. The interaction between 3-Br-2,6-DHBA and the beta-lactam ring, as well as adjacent five- or six-membered rings in amoxicillin and several oral cephalosporins, likely enables allosteric detection by stacking via π electron, hydrophobilc interaction, and hydrogen bonding. In conclusion, the 3-Br-2,6-DHBA-modified sensor membrane effectively detects amoxicillin and oral cephalosporins via allosteric mechanism.

1. Introduction

The taste sensor, initially developed by Toko, incorporates an “electronic tongue” consisting of sensor electrodes equipped with a multi-array of lipid/polymer membranes [1]. This sensor is capable of detecting and characterizing the five basic tastes, along with astringency, by monitoring changes in membrane potential caused by taste substances [2,3,4]. To improve patient adherence, pharmaceutical companies have increasingly developed patient-friendly oral medications with enhanced palatability. In addition to conventional tablets, capsules, and powders, new formulation types like orally disintegrating tablets [5,6,7] and orally disintegrating films [8,9] have been introduced worldwide, particularly for patients with swallowing difficulties. The taste sensor has become a valuable tool in pharmaceutical research and development and taste evaluation, including the quantitative assessment of drug bitterness, with its effectiveness reported in several academic studies [10,11,12,13,14]. Regarding a bitterness sensor, the so-called BT0 bitterness sensor has already been developed and is commercially available. It is an artificial lipid-based membrane sensor characterized by high selectivity and sensitivity to the bitterness of various drugs, showing a significant correlation with sensory evaluation scores [15]. Furthermore, its robustness and responsiveness have been previously validated [16]. The membrane is composed of phosphoric acid di-n-decyl ester (PADE) as the lipid component, along with bis(1-butylpentyl) adipate (BBPA) and tributyl o-acetylcitrate (TBAC) serving as plasticizers [16]. In addition, recent studies have demonstrated a significant correlation between the sensor’s output and human taste receptor responses [17].

Caffeine, known for its bitter taste, is a pharmaceutical substance listed in the Japanese Pharmacopoeia [18]. We selected caffeine as the bitterness control because it is a well-known bitter compound, as demonstrated in many previous studies. It is also present in high concentrations in commonly consumed beverages such as coffee and tea. In our previous study [19], nearly all human panelists were able to recognize its bitterness at a concentration of 3 mM caffeine. This level of bitterness is comparable to that of 0.03 mM quinine hydrochloride, one of the most bitter substances known. From a pharmacotherapeutic perspective, caffeine has been shown furthermore to elicit a high sensor response in a taste sensor equipped with surface-modified lipid/polymer membranes, as reported in our previous work [19], whereas no response in BT0 sensor.

It has various effects, such as reducing drowsiness, fatigue, and headaches, and acting as a vasodilator. Some over-the-counter cold medicines contain caffeine to counteract drowsiness caused by antihistamines and to potentially alleviate headaches. Although the BT0 sensor mentioned above has proven effective in quantifying the bitterness of many drugs, it was not capable of measuring the bitterness of caffeine, which possesses a xanthine skeleton without an intramolecular charge [12]. To overcome this limitation, Yoshimatsu from Toko’s group recently designed a novel sensor membrane based on the concept of allostery via hydrogen bonding [19]. This newly developed sensor successfully detected caffeine and related drugs, all belong to xanthines.

Maintaining high adherence to unpalatable medications is particularly challenging in pediatric patients as well as in many adults [20,21]. Many children cannot swallow solid dosage forms such as pills or tablets due to their unpleasant taste. Additionally, if some medications lack established pediatric dosages on the market, the formulation must be crushed, and the estimated required drug amount must be weighed for an individual child based on body weight. In such cases, crushing the oral formulation significantly increases its bitterness intensity, as described in previous studies [22,23]. Therefore, evaluating the bitterness of active pharmaceutical ingredients (APIs) used for both children and adults using taste sensors is crucial from an ethical standpoint.

Previous study evaluated the bitterness of various pharmaceutical substances, including those from the WHO Model List of Essential Medicines for Children (7th ed., 2019), using different taste sensors [24]. These studies identified several drugs, including caffeine, that exhibited none or minimal response to the BT0 membrane.

Building on these findings, we focused on the following 10 drugs, in addition to caffeine as a control: acyclovir, amoxicillin, cefalexin, isoniazid, linezolid, propylthiouracil, azathioprine, acetaminophen (paracetamol), carbamazepine, and folic acid. All 11 drugs, including caffeine, are listed in the WHO Model List of Essential Medicines for Children (7th ed., 2019), and their sensor output on the BT0 membrane was previously reported to be none or very low [24]. These drugs were selected based on the following criteria:

• The most important criterion was that, in the aforementioned study [24], the sensor output on the BT0 membrane for each 0.1 mM drug solution prepared using reference solution (30 mM KCl, 0.3 mM tartaric acid) was less than 1.0 mV, with no detectable response level.
• Other criteria are as follows: (1) The drug is an active pharmaceutical ingredient (API) widely used in clinical practice in Japan and worldwide, including for pediatric patients, and is administered orally. (2) The drug has the potential to cause palatability issues, including bitterness, when taken orally for repeated dosing. (3) The drug exhibits structural diversity, covering a wide range of chemical structures.

Recent advances in human taste perception research, particularly in the allosteric modulation of bitter taste receptors, have provided significant insights. A study revealed that cholesterol is essential for receptor activation and that taste molecules act allosterically on bitter taste receptors [25]. Human bitter taste receptors comprise 25 subtypes [26]. In the case of caffeine perception, multiple receptors, including TAS2R14 (also known as TR14), are identified to be involved [27,28]. Given these recent advancements, further clarification of allosteric modulation in caffeine bitterness perception is expected. Moreover, other bitter drugs may also interact with bitter taste receptors via allosteric modulation in vivo, even if they do not elicit a response from BT0 membrane.

Therefore, this study aimed to identify a new group of drugs that respond to a 3-Br-2,6-DHBA-modified sensor membrane. That sensor membrane was proved to show the high sensitivity for caffeine and related pharmaceuticals with xanthine skeleton, such as etofylline, proxyphylline, and diprophylline based on allostery [29]. To achieve the aim, we applied this modified membrane to various structurally diverse drugs to which the BT0 membranes had previously shown no response [24]. By leveraging the concept of allostery in the perception of caffeine and related substances, we conducted sensor output measurements to investigate their responsiveness and potential interactions between 3-Br-2,6-DHBA and these drugs.

2. Materials and Methods

2.1. Reagents

In related to drugs, caffeine (used as a control), acyclovir, amoxicillin, cefalexin, isoniazid, linezolid, azathioprine, carbamazepine, folic acid, cefaclor monohydrate, and cefdinir were purchased from Tokyo Chemical Industry (TCI). Propylthiouracil and acetaminophen (paracetamol) were obtained from FUJIFILM Wako Pure Chemical Corporation. All drugs were in powder form.

Sample solutions or suspensions of the 13 drugs were prepared with a reference solution (30 mM KCl, 0.3 mM tartaric acid) as described [19]. It was also confirmed that a 30 mM caffeine solution prepared with this reference exhibited a sufficient output of approximately 40 mV in previously. Based on these findings, drug solutions were primarily prepared at a concentration of 30 mM using the same reference solution. For poorly soluble drugs, the concentration was adjusted to 10 mM. The drug preparations were as follows:

• 30 mM solutions: caffeine, cefalexin, isoniazid, acetaminophen (paracetamol)
• 10 mM suspensions: acyclovir, amoxicillin, linezolid, propylthiouracil, azathioprine, carbamazepine, folic acid, cefaclor monohydrate, and cefdinir (resulting in slightly turbid suspensions)

Additionally, a 0.5 mol/L potassium hydroxide (KOH) solution and 99.5% ethanol were used to clean the electrodes, as described in Section 2.4, “Measurement Procedure of the Taste Sensor”. The KOH solution and ethanol were obtained from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan) and Japan Synthetic Alcohol (Kawasaki, Japan), respectively. The chemical structures of the 11 drugs from the WHO Model List of Essential Medicines for Children (7th ed., 2019) and two oral cephalosporins (cefaclor monohydrate and cefdinir) are shown in Figure 1.

Table 1. Clinical information on 11 drugs from the WHO Model List of Essential Medicines for Children (7th edn., 2019), and two oral cephalosporins (cefaclor monohydrate* and cefinir* used in additional experiments).

Drug Name	Medical Effects/Indications	Formulation	Reference No. **
Caffeine	CNS stimulant, analgesic; prevents drowsiness, relieves headaches	Powder	30
Acyclovir	Antiviral for herpes simplex, shingles	40% granules jelly	31
Amoxicillin	Broad-spectrum penicillin for infections, including H. pylori	Capsules, fine granules	32
Cefalexin	Oral cephalosporin antibiotic for infections	Fine granules	33
Isoniazid	First-line oral antituberculosis drug	Tablets	34
Acetaminophen (Paracetamol)	Antipyretic, analgesic for pain relief	Powder	35
Linezolid	Synthetic antibacterial for MRSA, sepsis	Tablets	36
Propylthiouracil	Antithyroid agent	Tablets	37
Azathioprine	Immunosuppressant for organ transplant rejection prevention	Tablets	38
Carbamazepine	Antiepileptic, antimanic for seizures, psychiatric conditions	Tablets, 50% fine granules	39
Folic Acid	Treats/prevents folic acid deficiency, supplementation	Tablets, granules	40
Cefaclor Monohydrate *	Oral cephalosporin for infections (Staph, Strep, E. coli)	10% fine granules	41
Cefdinir *	Oral cephalosporin for infections (Staph, Strep)	10% granules	42

* Used in additional experiments. ** The package inserts corresponding to the 11 drugs shown in Figure 1 are numbered sequentially from 30 to 42, following the order in which the drugs appear.

summarizes clinical information, including medicinal effects, indications, and available formulations, for the 11 drugs from the WHO Model List of Essential Medicines for Children (7th ed., 2019), as well as cefaclor monohydrate and cefdinir, which were used in an additional study. The corresponding reference numbers (Nos. 30–42) are also listed in Table 1 [30,31,32,33,34,35,36,37,38,39,40,41,42].

2.2. Components of the Modified Membrane and Solvent

Tetradodecylammonium bromide (TDAB), which served as the lipid component of the taste sensor, was obtained from Sigma-Aldrich (St. Louis, MO, USA). Dioctyl phenyl phosphonate (DOPP), used as a plasticizer, was sourced from Dojindo Laboratories (Kumamoto, Japan). Polyvinyl chloride (PVC), functioning as the supporting material, was supplied by FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan). Tetrahydrofuran (THF), employed as the solvent for membrane preparation, was also purchased from Sigma-Aldrich.

3-Bromo-2,6-dihydroxybenzoic acid (3-Br-2,6-DHBA), 2,6-dihydroxybenzoic acid (2,6-DHBA), and benzoic acid (BA) were obtained from FUJIFILM Wako Pure Chemical Corporation and used as surface modification reagents. A reference solution (pH 3.5) was provided by Intelligent Sensor Technology, Inc. (Atsugi, Kanagawa, Japan). The chemical structures of TDAB, 3-Br-2,6-DHBA, 2,6-DHBA, and BA, which were employed for sensor surface modification, are shown in Figure 2.

2.3. Sensor Preparation: Lipid/Polymer Membrane and Surface Modification

As described in previous studies [19,43], the TDAB lipid membrane was prepared by mixing 10 mL of 3 mM TDAB in THF, 1.5 mL of DOPP, and 800 mg of PVC to achieve a homogeneous solution. This mixture was poured into a petri dish and allowed to dry naturally at room temperature over three days. A portion of the dried TDAB membrane was subsequently attached to the sensor electrodes.

According to a previous study [44], modified membranes were prepared by immersing TDAB-based sensor electrodes in 0.03 wt% solutions of 3-Br-2,6-DHBA, 2,6-DHBA, or BA for 72 h, allowing the compounds to adsorb onto the membrane surface. This adsorption-based surface modification relies on electrostatic interactions and hydrophobic affinity between the modifying agents and the lipid/polymer membrane. Specifically, in the case of 3-Br-2,6-DHBA modification, as previously reported [45,46], TDAB dissociates to acquire a positive charge, while the carboxyl groups of 3-Br-2,6-DHBA ionize to become negatively charged. This charge interaction facilitates adsorption, as the TDAB membrane selectively interacts with negatively charged ions, including Br⁻ from TDAB and ionized 3-Br-2,6-DHBA. Furthermore, the hydrophobic nature of 3-Br-2,6-DHBA enhances its incorporation into the membrane surface, complementing the electrostatic interaction and contributing to stable adsorption.

2.4. Measurement Procedure of the Taste Sensor

All taste measurements in this study were performed using a commercially available taste-sensing system (TS-5000Z, Intelligent Sensor Technology, Inc., Kanagawa, Japan). The detection unit of the sensor system was connected to a reference electrode, composed of an AgCl-coated silver wire, and a sensor electrode incorporating the TDAB membrane. The internal solution used for both the sensor and reference electrodes consisted of a mixture of 3.33 M KCl and saturated AgCl.

The measurement procedure was as follows:

(1)

The sensor electrodes were first immersed in a reference solution containing 30 mM KCl and 0.3 mM tartaric acid, and the electric potential of this solution (V_r) was measured for 30 s. The reference solution simulates human saliva and has minimal inherent taste [47].

(2)

Subsequently, the electric potential of the sample solution (V_s) was measured for 30 s while the electrodes were immersed in the sample.

(3)

The relative response value was calculated as the difference between V_s and V_r.

(4)

Finally, the membrane surface was cleaned with an aqueous cleaning solution composed of 10 mM KOH, 100 mM KCl, and 30% ethanol (v/v).

Taste sensor measurements were performed five consecutive times within a single day. Data from the third to fifth measurements were used for analysis.

3. Results

3.1. Sensor Outputs of 11 Drugs Measured Using BT0 Membranes and 3-Br-2,6-DHBA-Modified Membranes

The concentrations of the drug sample solutions (some in suspension) were prepared as described in the Materials and Methods, using a reference solution to achieve final concentrations of 30 mM or 10 mM (for some suspensions). The measurement results for 11 drugs from the WHO Model List of Essential Medicines for Children, obtained using a BT0, are shown in Figure 3a. For all drugs, including caffeine, the sensor output was either negligible or very low, even at high concentrations such as 10 mM or 30 mM. These findings are consistent with our previous study, which used much lower drug concentrations (below 0.1 mM for each drug) and similarly showed no response with BT0 [24].

The measurement results using 3-Br-2,6-DHBA-modified membrane are shown in Figure 3b. Caffeine (30 mM), used as a control, exhibited the highest sensor output, approximately 40 mV. Amoxicillin (10 mM) and cefalexin (30 mM) induced relatively high sensor outputs, exceeding 20 mV, while the other drugs exhibited little or no response in the 3-Br-2,6-DHBA-modified membrane.

3.2. Sensor Output of Amoxicillin, Cefalexin, and Acyclovir with 3-Br-2,6-DHBA-, 2,6-DHBA, or BA-Modified Membranes

First of all, amoxicillin and cefalexin were selected owing to their positive sensor output in 3-Br-2,6-DHBA modified membrane. In addition, acyclovir was selected as control, since it shares structural similarity with caffeine and also has no electrical charge. For amoxicillin and cefalexin, high sensor outputs were observed with membranes modified with 3-Br-2,6-DHBA, whereas the membranes modified with 2,6-DHBA or BA showed no response to these drugs. In contrast, acyclovir, used as a control, produced no sensor output with any of the three modified membranes (Figure 4).

3.3. Concentration Dependence of Amoxicillin and Cefalexin on Sensor Output in Modified Membranes

Since only amoxicillin and cefalexin showed positive responses with the 3-Br-2,6-DHBA-modified membranes, the concentration dependence of these two drugs on sensor output was evaluated using the same modified membranes. For amoxicillin, a broad-spectrum penicillin containing a beta-lactam ring, the sample concentrations were set at 0.1, 0.3, 1, 3, and 10 mM. As shown in Figure 5a, the sensor response increased with increasing concentration. For cefalexin, a cephalosporin antibiotic also containing a beta-lactam ring, sample concentrations were set at 0.1, 0.3, 1, 3, 10, and 30 mM. Similarly, the sensor response increased with increasing concentration, as shown in Figure 5b.

3.4. Extended Response Analysis of 3-Br-2,6-DHBA-Modified Membranes for Two Oral Cephalosporins

To investigate whether 3-Br-2,6-DHBA could detect beta-lactam antibiotics beyond the two initially tested drugs, additional tests were conducted with two oral cephalosporins, cefaclor and cefdinir. In Japan, no oral monobactam products are available, as only aztreonam is available as an injectable formulation [48]. Similarly, most carbapenem antibiotics are administered via injection for patients with severe infections [49,50].

Although cefaclor and cefdinir (listed in Table 1) are not included in the WHO Model List of Essential Medicines, they were selected due to their frequent use in treating infections in both children and adults, as indicated in the corresponding references in Table 1. For both cefaclor and cefdinir, the sample concentration was fixed at 10 mM and prepared using a reference solution. The prepared samples were slightly suspended. Measurements were performed using 3-Br-2,6-DHBA-modified membranes, following the same procedure as for the other 11 drugs. The results are shown in Figure 6. Cefaclor and cefdinir (both 10 mM suspensions) exhibited sensor outputs at approximately 60–70% of the level observed for caffeine (30 mM).

4. Discussion

Yoshimatsu et al. suggested that caffeine and related substances, which were previously not detected using the conventional sensor BT0, caused a response in the 2,6-DHBA-modified membrane via molecular interactions, specifically allosteric effects mediated by hydrogen bonding [19]. Their study investigated the detection mechanism, with a particular focus on the roles of both intramolecular and intermolecular hydrogen bonding in contributing to allosteric behavior. It was proposed that the hydrogen bond between caffeine and the hydroxyl group of an aromatic carboxylic acid affects the dissociation of the carboxyl group, thereby modulating subsequent hydrogen bonding interactions. This process exemplifies an allosteric effect, wherein caffeine binding at one site on the aromatic carboxylic acid facilitates hydrogen bonding at a remote site. The insights gained from that study provided critical foundational knowledge for the present research.

Ishida et al. explored the molecular interactions between caffeine and various hydroxybenzoic acids using NOESY and ¹H NMR spectroscopy [43]. Their findings suggested that the interaction between caffeine and 2,6-DHBA involves hydrogen bonding between the hydroxyl groups of 2,6-DHBA and either the carbonyl group or the imidazole nitrogen of caffeine, in addition to π–π stacking interactions between their aromatic rings, which further stabilizes the complex.

Thus, both intramolecular and intermolecular hydrogen bonds, as well as stacking interactions between the structures of modified agents and substrate drugs, appear to be essential for detection via allostery using the modified membrane.

Gibson and Fowler reported that the imidazole moiety in caffeine exhibits high aromaticity, with π-electrons delocalized within its structure. The lone pairs on the nitrogen and oxygen atoms primarily affect the six-membered ring, contributing to the overall aromatic character of the caffeine molecule [51]. Consequently, the π-electrons in caffeine’s two ring systems, together with the aromatic ring of the modified membrane’s 2,6-DHBA, can potentially engage in π–π stacking interactions in various orientations, promoting stacking between caffeine and 2,6-DHBA.

The stacking appears to contribute effectively to the interaction between 2,6-DHBA and caffeine in allostery. Regarding π–π stacking interactions, notable progress has been made in various fields and successfully applied in sensor development [52,53]. In the present study, four drugs—amoxicillin, cefalexin, cefaclor, and cefdinir—exhibited positive responses in the 3-Br-2,6-DHBA-modified membrane, as shown in Figure 3 and Figure 6, respectively. Regarding their structure, these four drugs all contain a beta-lactam ring, as shown in Figure 1.

Zhao et al. assessed the bitterness of caffeine and xanthine-based substances using the novel allosteric membrane with 3-Br-2,6-DHBA, which has high sensitivity to caffeine and related pharmaceuticals. They demonstrated that the sensor outputs for these substances with the new membrane closely aligned with the results of human gustatory tests [44]. Their study suggested two critical structural factors that likely determine the responsiveness in the newly developed allosteric membrane:

(1)

The molecule contains two spatially separated C=O groups.

(2)

Between these two C=O groups, there is a moiety capable of stacking (via π–π interactions or hydrophobic interactions) with aromatic carboxylic acids containing hydroxyl groups.

Among the 13 drugs, the chemical structures of which are shown in Figure 1, amoxicillin, as an oral broad-spectrum penicillin, and three oral cephalosporins appear to meet above criteria. Amoxicillin contains a beta-lactam ring and an adjacent five-membered thiazolidine ring (containing both N and S atoms), while the three cephalosporins contain a beta-lactam ring and an adjacent six-membered dihydrothiazine ring (also containing N and S atoms). These structures seem critical in determining their detectability by the newly developed allosteric membrane.

Regarding aromaticity, as discussed above in related to caffeine, aromaticity in the drug substrates seems advantageous for stacking with the aromatic ring of 3-Br-2,6-DHBA via π-π interactions or interactions assumed to involve π-electrons. Nevertheless, the beta-lactam ring in all four drugs is a four-membered cyclic amide that does not satisfy the conditions for aromaticity by itself, and it is generally not considered an aromatic compound, as previously discussed [54]. This is in contrast to caffeine.

However, considering the possibility of interactions, the π-electrons accompanying the oxygen atom in the C=O bonds within each antibiotic structure are expected to contribute to stacking interactions with 3-Br-2,6-DHBA via π-electrons associated with C=O. Additionally, 3-Br-2,6-DHBA contains a C=O group as part of its carboxyl group. Furthermore, hydrophobic interactions between 3-Br-2,6-DHBA and the beta-lactam ring, thiazolidine ring (in amoxicillin), and dihydrothiazine ring (in the three oral cephalosporins) are also expected.

Moreover, the five-membered thiazolidine ring, which contains N and S atoms and is adjacent to the beta-lactam in amoxicillin, and the six-membered dihydrothiazine ring, also containing N and S atoms and adjacent to the beta-lactam in the three oral cephalosporins, are relevant for the interactions described.

As discussed in a previous review article [55], interactions between nonbonding electron pairs (lone pairs) and π-electrons are commonly known as “n → π interactions” or “lone pair–π interactions”, were explained including examples.

It has also been reported that in a certain thiazoline derivative, weak π-π interactions occur between the thiazolidine ring and the dichlorobenzene ring due to C-H⋯S and C-H⋯O hydrogen bonds within the molecule [56]. In light of such articles, regarding this study, the possibility of lone pair–π interactions (i.e., the lone pairs of nitrogen (N) and sulfur (S) atoms in the thiazolidine ring (amoxicillin) and dihydrothiazine ring (three oral cephalosporins) interacting with 3-Br-2,6-DHBA) cannot be denied, even though we do not have its direct proof or experimental results.

Regarding interaction via hydrogen bonding, four orally available beta-lactam antibiotics possess a carboxylic acid group capable of forming hydrogen bonds (via the C=O bond or O–H group), suggesting the potential for allosteric effects mediated through hydrogen bonding with 3-Br-2,6-DHBA.

As demonstrated in Figure 4, among the modified membranes, the 3-Br-2,6-DHBA-modified membrane was the most effective and significantly more sensitive than the 2,6-DHBA-modified membrane in detecting amoxicillin and cefalexin. These results are consistent with our previous study on caffeine and other related substances [19,44]. The pKa of the carboxyl group in 3-Br-2,6-DHBA is predicted to be 1.48, according to Marvin Sketch, which is lower than that of 2,6-DHBA (1.64). Under the pH conditions of the reference solution (pH 3.5), one possible explanation for this observation is that the higher degree of dissociation of the carboxyl group in 3-Br-2,6-DHBA may contribute to an allosteric effect via hydrogen bonding. Regarding BA, which contains only a single carboxyl group, almost no response was observed, as shown in Figure 4. This suggests that the hydroxyl group attached to the benzene ring of 3-Br-2,6-DHBA is essential for achieving the allosteric effect through hydrogen bonding between 3-Br-2,6-DHBA and amoxicillin or cefalexin. In the case of acyclovir, its structure resembles that of caffeine, which interacts strongly with the 3-Br-2,6-DHBA-modified membrane. Nevertheless, acyclovir did not induce any response in any of the modified membranes. Moreover, it does not meet two critical structural requirements mentioned above, as reported in the previous article. [44].

Overall, in the interaction between the 3-Br-2,6-DHBA-modified membrane and amoxicillin or each of the oral cephalosporins, which are beta-lactam antibiotics, stacking through π electron, hydrophobic interaction is expected. It can also be considered that hydrogen bonding between each drug and the hydroxyl groups of 3-Br-2,6-DHBA, an aromatic carboxylic acid, affects the dissociation state of the carboxy group involved in binding. Thus, the stacking and hydrogen bonding enable allosteric detection in amoxicillin or each of the oral cephalosporins.

However, the extent to which π–π stacking and/or hydrophobic interactions contribute may differ from that observed in caffeine and related substances. Unlike compounds such as caffeine that exhibit aromatic character due to electron delocalization, the beta-lactam ring (a four-membered cyclic amide) is generally considered non-aromatic. This isdue to the significant ring strain and limited conjugation within the beta-actam structure, which prevent effective π-electron delocalization necessary for aromaticity. Therefore, this non-aromatic characteristic in the four beta-lactam drugs might explain their lower sensor output compared to that of caffeine, which has aromaticity as a whole molecule, as shown in Figure 3 and Figure 6.

Orally administered antibiotics are associated with allergies, making sensory tests more ethically problematic than tests for other drugs. Among the selected drugs, amoxicillin and cefalexin are included in the WHO Essential Medicines List for Children. All of the antibiotics, including the additionally tested cefaclor [41] and cefdinir [42], are commonly used in clinical settings for pediatric as granule or for adult as tablet administration, as shown in Table 1. The output observed with 3-Br-2,6-DHBA-modified membranes for the oral cephalosporin antibiotics and amoxicillin, which were not sensitive to conventional BT0 membranes, is considered significant.

Based on the concentration dependency study of amoxicillin and cefalexin shown in Figure 5, the limits of detection using 3-Br-2,6-DHBA-modified membranes are estimated to be approximately 0.1 mM for amoxicillin and 1.0 mM for cefalexin. Assuming that a patient takes a 250 mg amoxicillin capsule or 500 mg cefalexin granules, and 1% of each drug dissolves in 2 mL of saliva, this would result in 2.5 mg of amoxicillin/2 mL and 5 mg of cefalexin/2 mL, corresponding to 1250 mg/L and 2500 mg/L, respectively. When converted to molar concentrations, this gives 1250/419.5 = 2.98 mM for amoxicillin and 2500/347.1 = 7.20 mM for cefalexin, respectively. These concentrations are both significantly higher than the detection limits of the two drugs mentioned above, indicating that quantification of drug concentrations under oral conditions poses no issue. This highlights a practical advantage for the application of the sensor.

In this study, we evaluated the bitterness of pharmaceuticals commonly used by patients, including children. Ensuring palatability is particularly crucial when developing medications for children. While adults generally tolerate unpleasant-tasting medications due to their understanding of the health benefits, children may struggle with aversion due to taste. A study conducted in the UK revealed that approximately one-third of children refuse to take medicine due to its taste or texture [57]. For antibiotics, taste has been identified as a common barrier to medication adherence [58].

Previous studies have reported differences in taste thresholds between children and adults, as well as variations in bitterness thresholds within these populations, often attributed to genetic polymorphisms [59,60,61]. However, to our knowledge, no studies have demonstrated a close match between the perceived taste intensity in children and adults. Notably, large-scale studies involving a significant number of child participants and panelists have reported inconsistencies in sensory evaluations between children and adults, particularly when assessing the suppressive effects of representative bitterness-masking agents such as calcium gluconate and monosodium glutamate on multiple bitter pharmaceuticals [62]. Large-scale sensory testing of pharmaceuticals, where toxicity risks are present, raises ethical concerns, especially in drug development.

A promising approach to addressing these challenges is the use of taste sensors. In this study, the ability to measure the bitterness of orally administered beta-lactam antibiotics—previously difficult to assess through taste evaluation—is considered a significant achievement. The concept of allostery, which underlies this sensor, is increasingly recognized as playing a role in various biological processes, including taste perception [63,64]. In particular, elucidating the role of allostery in taste perception is of great significance.

5. Conclusions

Based on the concept of allostery, we successfully detected sensor output for the broad-spectrum oral penicillin amoxicillin and three oral cephalosporins—cefalexin, cefaclor, and cefdinir—all beta-lactam antibiotics that were undetectable with the conventional BT0 membrane used for bitterness assessment. The high sensor output observed with 3-bromo-2,6-dihydroxybenzoic acid (3-Br-2,6-DHBA)-modified membranes is likely due to molecular interactions between the four beta-lactam antibiotics and 3-Br-2,6-DHBA, as well as allosteric effects mediated by intramolecular and intermolecular hydrogen bonding, which play a crucial role in detection. Hydrogen bond-mediated allosteric mechanisms in biological systems have recently attracted attention in taste perception. This study offers also valuable insights into the molecular recognition of bitter substances by bitterness receptors. This sensor is novel in that it enables molecular recognition of bitter compounds as made by multiple human bitter taste receptors, and it is considered to have potential applications in bitterness assessment in the pharmaceutical industry as well as in the search for agonists and antagonists of bitter taste receptors.