Bitterness is innately aversive and generally associates with reduced liking and consumption of foods and beverages (Hayes, 2020; May-Wilson et al., 2022). While this relationship is widely studied in foods (e.g., vegetables, beer, dark chocolate) (Mattes, 1994; Keller et al., 2002; Dinehart et al., 2006) less quantitative evidence regarding aversive tastes and medication compliance exists for adults, but presumably, the same pattern exists (Shahiwala, 2011; Schiffman, 2015) especially given relevant aphorisms from diverse cultures. For example, there is a Chinese idiom (“liánɡ yào kǔ kǒu”) which claims, “good medicine tastes bitter,” while in English, Mary Poppins famously sang “a spoonful of sugar helps the medicine go down.” In a retrospective study with children, taste-related complaints were reported as the number one reason for rejection of liquid oral medication (Mennella et al., 2015). Collectively then, it seems reasonable to conclude that bitterness from medications may result in reduced compliance. Thus, we examined the sensory profile of two antibiotics, chloramphenicol and ofloxacin.

Despite long understood associations between bitterness and pharmacological activity [e.g., phlorizin from apple tree bark; see (Ehrenkranz et al., 2005)], quantitative data on the sensory profile of various medications is surprisingly sparse. Work by Schiffman and colleagues reported recognition thresholds of 42 medications, with all but one being bitter once the concentration was detectable (Schiffman et al., 2002; Schiffman, 2015). Although chloramphenicol was not assessed, the detection threshold of ofloxacin was estimated to be 0.38 mM and was reported to be bitter (Schiffman et al., 2002; Schiffman, 2015). Another study by Schiffman et al. (1999) reported the sensory profile for three protease inhibitors, ritonavir, indinavir, and saquinavir. Healthy controls and HIV-patients reported intensity ratings on a 7-point intensity scale for suprathreshold concentrations. Bitterness was the highest rated sensation, rated between “moderate” and “moderate strong” for all three drugs, with other sensations reported as having a “weak” intensity or lower, including metallic, medicinal, and astringent. Other drugs that have undergone formal sensory testing include the antiretroviral drug tenofovir alafenamide (TAF): when given to a small panel of healthy adults in a liquid formulation, the mean bitterness was rated near “moderate” on a general Labeled Magnitude Scale (gLMS), albeit with variable ratings ranging from “weak” to “very strong” (Schwiebert et al., 2021). Although individual differences were not explicitly investigated by those authors (perhaps due to the small sample size; n = 16), the wide range in intensity ratings suggests there may be substantial variability in the bitterness of this drug. Likewise, another antiretroviral cocktail (lopinavir/ritonavir) varied in bitterness ratings, with a downstream impact on hedonic ratings (Mennella et al., 2017). Two notable exceptions to the overall lack of quantitative sensory data on pharmaceuticals include the antimalarial drug quinine and the thyroid drug propylthiouracil (PROP)—in part because quinine is used outside a therapeutic context as a food ingredient (FEMA# 2,976) and PROP is commonly used as a probe of variation in taste (Smith and Davies, 1973; Lawless, 1980; Bartoshuk et al., 1992; Bartoshuk et al., 1994; Delwiche et al., 2001; Duffy et al., 2004; Chang et al., 2006; Hansen et al., 2006; Hayes et al., 2008; Wooding et al., 2010; Mennella et al., 2011; Feeney et al., 2014; Boxer and Garneau, 2015; Hayes et al., 2015; Nolden et al., 2020). Collectively, the limited data that exists suggests many commercial pharmaceuticals are perceived as bitter, and this bitterness may vary across individuals.

Differential bitterness has been repeatedly shown to associate with normal variation in bitter taste receptor genes (TAS2Rs). This gene family has 25 members which encode receptors (hT2Rs) tuned to a wide range of ligands (Meyerhof et al., 2010). Critically, the TAS2Rs are highly polymorphic relative to the rest of the human genome (Kim et al., 2005). Kim and colleagues were the first to identify a relationship between single nucleotide polymorphisms (SNPs) in the PTC gene (which was quickly renamed TAS2R38), and individual differences in perception of phenylthiocarbamide (PTC), a structurally similar chemical to PROP (Kim et al., 2003; Bufe et al., 2005; Kim et al., 2005). Subsequently, a robust relationship between TAS2R38 genotypes and perception of PROP and PTC emerged e.g., (Duffy et al., 2004; Hayes et al., 2008; Allen et al., 2013). Such individual differences extend beyond PTC and PROP, with variation in the perception of diverse compounds that associates with polymorphisms in multiple TAS2Rs [e.g., (Allen et al., 2013; Allen et al., 2014; Risso et al., 2014; Hayes et al., 2015; Roura et al., 2015)].

Multiple studies have systematically investigated which hT2Rs are activated in vitro by specific bitter compounds, including drugs (Meyerhof et al., 2010; Thalmann et al., 2013; Levit et al., 2014) [see review (Clark et al., 2012)]. Such data initially suggested chloramphenicol is an agonist for six different hT2Rs, specifically, −1, −8, −10, −39, −43, and −46 (Meyerhof et al., 2010). Later, Thalmann and colleagues (2013) extended this set, by identifying hT2R41 as a potential receptor for chloramphenicol. In a series of functional experiments, they identified a genetic variant, L127, which stimulated an ∼10-fold reduction in response relative to cells that expressed the P127 variant of the receptor (Thalmann et al., 2013). The hT2R41 receptor appears to be narrowly tuned, as the only other ligand identified to date is the non-nutritive sweetener sucralose (Lossow et al., 2016), which is known to evoke bitter sensations at some concentrations (Antenucci and Hayes, 2015). The earlier work of Meyerhof and colleagues (2010) likely failed to identify chloramphenicol as an agonist of hT2R41 because the variant used in the original study was the non-functional variant, at least, according to Thalmann and colleagues (2013).

To our knowledge, only a single study has tested ofloxacin as a hT2R agonist: Dotson and colleagues (2008b) found hT2R9 responded to ofloxacin in vitro (along with two other drugs, procainamide and pirenzepine). They also reported functional variation in TAS2R9 (V187A), with the A187 variant being the active form for all three drugs tested.

Collectively, these studies indicate that genetic differences in TAS2Rs are associated with bitterness perception of medications, with the implication that individuals expressing specific functional alleles for some TAS2Rs may have a greater risk for non-compliance. Mennella and colleagues (2015) provided evidence of this in children: in a retrospective interview, children with one or two copies of the functional TAS2R38 variant (P49) reported greater rejection of liquid formulations of medications (type unspecified), relative to children with two copies of the non-functional variant (A49) of TAS2R38. To explore these types of relationships, we 1) obtained sensory profiles of two antibiotics (chloramphenicol and ofloxacin) in sample of nominally healthy adults, and 2) tested to see if variation in bitterness ratings could be explained by functional polymorphisms in TAS2Rs. We also 3) included PROP as a positive control, given the robust association between PROP bitterness and TAS2R38 diplotype (e.g., Nolden et al., 2020).

Our working hypothesis is that there are individual differences in the bitterness from medications which are mediated by genetic variability in TAS2Rs; we reasoned such variation may put some individuals at greater risk for lower compliance due to heightened bitterness, so we asked whether such differences might manifest phenotypically. Here, we report the first systematic investigation of the bitterness of chloramphenicol and ofloxacin. Our data suggest individual differences in the bitterness perception extend to antibiotics, and these differences are systematically related to TAS2R variants previously shown to be functional. We also observed a positive correlation between the bitterness ratings of PROP and both chloramphenicol and ofloxacin, with the strongest correlation reported between chloramphenicol and ofloxacin. While PROP bitterness is associated with genetic polymorphisms in TAS2R38 (Duffy et al., 2004; Kim et al., 2005), not all variability in PROP is explained by TAS2R38 (Hayes et al., 2008). This is demonstrated by the relationship between ofloxacin and PROP but not with the TAS2R38 diplotype. This finding aligns with prior work showing bitterness (Nolden et al., 2020) and sweetness (Allen et al., 2013) are each associated with PROP phenotype but not with TAS2R38 genotype. That is, PROP bitterness is a phenotype that appears to confound narrowly tuned genetic variation (from TAS2R38) with a broader factor that predicts differences in orosensation.

Present data reinforce the consistent findings that TAS2R38 variants can explain individual variability in the bitterness perception of PROP. Here, TAS2R38 genotypes were included primarily as a positive control due to the many studies demonstrating a robust relationship with PROP bitterness (Duffy et al., 2004; Kim et al., 2005; Hayes et al., 2008; Wooding et al., 2010; Allen et al., 2013; Garneau et al., 2014; Nolden et al., 2020), which were confirmed yet again in the present study (Figure 2). Also, we note that Figure 2 indicates that some AVI homozygotes still reported moderate to strong bitterness from PROP, despite having two copies of the AVI variant of hT2R38, which is thought to be non-functional. Data in Figure 2 are consistent with the “second receptor” hypothesis, wherein another hT2R receptor that ligates PROP may recover function in some individuals who carry two copies of the non-functional TAS2R38 allele see discussion in (Hayes et al., 2008; Nolden et al., 2020).

The observed relationship between chloramphenicol bitterness in vitro and TAS2R38 diplotype was wholly unexpected, since prior in-vivo data from heterologous expression systems suggest hT2R38 is not activated by chloramphenicol (Meyerhof et al., 2010). To the best of our knowledge, we are unaware of any prior investigation into whether ofloxacin activates hT2R38, and in the present study, ofloxacin bitterness did not associate with TAS2R38 diplotype. The AVI homozygotes tested here reported the highest bitterness, which would suggest that the AVI genotype might be the functional variant. We suspect this might be an artifact arising from linkage disequilibrium (LD) with a functional variant in another nearby TAS2R gene. This type of association has been reported previously for TAS2R19 and TAS2R31, where polymorphisms in both predict responses to quinine and grapefruit, but only TAS2R31 is thought to be causal see discussion in (Hayes et al., 2015). Here, we speculate one or more TAS2R38 SNPs may be in LD with an unmeasured and functional variant of another TAS2R. For example, Thalmann and colleagues (2013) reported the P127L variant of hT2R41 corresponded to differential activation by chloramphenicol in-vitro, but we cannot test this directly, as we do not have data on TAS2R41 variation in our participants. Thus, more work is needed to confirm the potential relationship between the bitterness of chloramphenicol and variability in TAS2R38 and/or TAS2R41, as well as exploring possible relationships with other TAS2Rs, since prior data suggests chloramphenicol activates 6 other hT2Rs in-vitro. Alternatively, we cannot rule out the possibility that current data are a Type I error or that prior in-vitro data reflect a Type II error.

Separately, we observed as significant the relationship between the TAS2R9 V187A polymorphism and the bitterness from ofloxacin, with the A187 form being the nominally-functional variant. These data align with prior work suggesting A187 responds to ofloxacin in vitro with reduced activation for V187 variant (Dotson et al., 2008). Previously, we reported V187A corresponds to variability in the bitterness from acesulfame-K (a widely used non-nutritive sweetener); however, in those data, V187 was associated with greater bitterness (Allen et al., 2013). More work is needed to clarify this discrepancy. Notably, Meyerhof and colleagues (2010) acknowledged that in their 2010 study, the hT2R9 amino acid sequence included the non-functional variant (V187); further, ofloxacin (nor procainamide or pirenzepine, which were used by Dotson et al. (2008)) was not included in the experiments by Meyerhof’s group (2010).

A substantial limitation of the present study is that our participants were not genotyped for the nominally functional variant of TAS2R41 previously shown to be activated by chloramphenicol in a heterologous expression system (Thalmann et al., 2013). Still, we can speculate that the relationship observed here between the TAS2R38 diplotype and the bitterness of chloramphenicol arises from LD between TAS2R38 and TAS2R41 variants. However, Meyerhoff and colleagues (2010) also identified 6 other hT2Rs that respond to chloramphenicol. It is possible that genetic variants in one or more of these receptors may be in LD with TAS2R38 SNPs. Another limitation of this study is the concentrations of chloramphenicol and ofloxacin that were used—specifically, the concentrations used were well below the therapeutic dose, to minimize potential risk to the participants that come with sampling pharmacologically active stimuli. Further, all stimuli were expectorated and not swallowed, to further reduce exposure and thus participant risk; however, this also means we might not have adequately stimulated receptors in the throat see (Bennett and Hayes, 2012; Running and Hayes, 2017). While the present data provide new insight into the bitterness arising from these two antibiotics, we cannot definitively predict what the sensory profile of these medications might be closer to their therapeutic dose. Moreover, it is unknown whether genetic variants would still be associated with the bitterness perception at higher concentrations and if this would translate to differential compliance. Dosing form matters as well, as tablets or capsules help reduce aversive sensations that may be present in liquid oral formulations. If perceived bitterness of therapeutics negatively and meaningfully impacts compliance, greater consideration for sensations arising from pharmaceuticals and their formulation seems warranted. It is plausible that in the future, patients could receive personalized formulations of medications, informed from their genetic profile for TAS2Rs or other taste receptors.

Here, we illustrate individual differences in bitterness from two antibiotic medications and show that such variation may be due to genetic variability in TAS2Rs. Strategies for minimizing bitterness perception should consider individual differences driven by genetic variants in TAS2Rs. This study also highlights the importance of considering a personalized approach to screening medications and identifying bitter blockers to increase acceptance of medications and drug formulations more effectively. Future studies should investigate personalized medications not only for their efficacy but also for their taste profile, minimizing the aversive sensations such as bitterness, that ultimately may lead to improved compliance.