Analysis of phototoxin taste closely correlates nucleophilicity to type 1 phototoxicity

Results and Discussion

To examine whether riboflavin is phototoxic to the genetically tractable model organism Drosophila melanogaster, larvae grown in riboflavin-enriched (0.3 or 10 mM) media were illuminated with blue light (Materials and Methods). Indicative of riboflavin phototoxicity, survival of the larvae to adults was dramatically reduced by the irradiation, compared with equally illuminated larvae grown in normal food (Fig. 1B and SI Appendix, Fig. S1 A–C). Note that the larvae in riboflavin-rich food showed similar degrees of ingestion amounts when spectrophotometrically assessed (SI Appendix, Fig. S1 E–H), and irradiated larvae were no longer subject to riboflavin-enriched ingestion, as they were subsequently placed in normal food. When a single choice of food is offered as in our survival assay, the ingestion amount is estimated to be little affected by excessive riboflavin, as riboflavin in food did not lead to adult weight loss (SI Appendix, Fig. S1I), a marker of larval starvation (14). Next, we investigated whether riboflavin evoked gustatory aversion, using the capillary feeder assay (CAFE) (15). Although feeding avoidance to riboflavin alone at its water-saturating concentration of 0.3 mM was not observed, riboflavin was well avoided in the presence of 1 mM N-methyl maleimide (NMM), a concentration also singly subefficacious for robust feeding deterrence (Fig. 1C). NMM is a well-established electrophilic agonist activating TRPA1s from mammals to insects through covalent modification of cysteine sulfhydryl residues (10, 13, 16, 17). Such functional interaction between riboflavin and NMM implies that the dTrpA1 gene may be important for riboflavin gustation. Indeed, the dTrpA1^ins loss-of-function mutant (10, 17, 18) was impaired for NMM-aided riboflavin aversion (Fig. 1C). The defective feeding deterrence of the mutants was fully restored by dTrpA1(A), but not dTrpA1(B) cDNA with Gr89a-Gal4 (Fig. 1D) covering dTrpA1(A)-expressing bitter-sensing neurons (10, 19). The feeding deterrence was little affected by a typical indoor lighting condition (white light at ∼0.2 mW/cm²), despite involving the phototoxin (Fig. 1E). Riboflavin-evoked action potentials were readily recorded in the dTrpA1-positive gustatory receptor neurons (GRNs) of labella i-a bristles (10, 13, 16, 19) in the presence of 1 mM NMM, regardless of lighting condition (Fig. 1 F and H). The GRN responses also required dTrpA1 (Fig. 1 F and G), revealing coherence between the behavioral and cellular riboflavin-tasting mechanisms. Despite its chemical reactivity, NMM does not seem to covalently modify riboflavin and generate new chemical species, as spectral absorbance of riboflavin was unchanged after 30 min incubation with NMM (SI Appendix, Fig. S1J).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Riboflavin, phototoxic to Drosophila, causes gustatory aversion through dTrpA1(A). (A) Riboflavin structure. (B) Phototoxicity of riboflavin to Drosophila (n = 4–11). ***P < 0.001, unpaired t test. (C) Riboflavin gustatory aversion assessed by NMM-aided CAFE assays with wild-type (WT, Canton S) and dTrpA1^ins animals (n = 5–7). (D) Rescue of feeding aversion in dTrpA1^ins by dTrpA1(A), but not by dTrpA1(B) (n = 6–7). (E) Riboflavin CAFE with or without ambient light of ∼0.2 mW/cm² (n = 7–12). (F) Riboflavin responses in dTrpA1-expressing bitter taste neurons in i-a bristle sensilla of indicated genotypes (n = 7–15). (G) Rescue of impaired spiking in dTrpA1^ins by dTrpA1(A) (n = 16–17). (H) GRN responses in indicated lighting conditions (n = 6–9). On: ∼0.25 mW/cm². n.s., not significant, paired t test. *P < 0.05; **P < 0.01, ANOVA Tukey or Welch’s ANOVA Games-Howell test; ^#P < 0.05; ^##P < 0.01, ANOVA Dunn’s rank test.

In addition to riboflavin/NMM-dependent potentiation of dTRPA1(A) in vivo, we examined whether the riboflavin is able to sensitize bitter gustation, as dTrpA1 is expressed in bitter GRNs (10). Caffeine and berberine at suboptimal concentrations of 0.2 and 0.001 mM, respectively, failed to induce appreciable avoidance in the CAFE assays (SI Appendix, Fig. S2 A and E). However, when the low concentrations of bitter compounds were offered with riboflavin, the avoidance was significantly increased. Such synergistic feeding aversion was absent in dTrpA1^ins (SI Appendix, Fig. S2 B and F). Consistent with the behavioral outcome, we obtained very similar results in GRN responses (SI Appendix, Fig.S2 C, D, G, and H). Again, dTrpA1(A), but not dTrpA1(B), was necessary for the riboflavin-dependent GRN and feeding responses to the low concentrations of bitter compounds. Thus, ingestion of riboflavin that is phototoxic to flies can be suppressed by a dTrpA1-dependent mechanism, as the dTrpA1(A)-dependent riboflavin gustation can enhance the responses to diverse chemicals stimulating bitter GRNs. Dramatic riboflavin accumulation was reported in yeast mitochondria up to a millimolar range when moderate concentrations of riboflavin were provided from the environment (20). Moreover, there are flavogenic yeasts that can produce riboflavin several-hundred-fold in certain conditions (21), together implying that there may be natural situations in which high doses of riboflavin are available for feeding, and that flies may encounter the high riboflavin concentration used in our experiments.

Heterologously expressed in Xenopus oocytes, dTRPA1(A), but not dTRPA1(B), showed dose-dependent current responses (Fig. 2A and SI Appendix, Fig. S3A), as the riboflavin current was normalized with respect to the current acquired by 0.1 mM NMM. This result indicates that dTRPA1(A) serves as a direct receptor for riboflavin in an isoform-dependent manner. Moreover, agTRPA1(A) from malaria-transmitting mosquitoes, Anopheles gambiae, yielded much larger current responses to riboflavin than dTRPA1(A), whereas isoform dependence persisted (Fig. 2A and SI Appendix, Fig. S3C). The current acquired with 0.1 mM NMM is close to the maximum for both dTRPA1(A) and agTRPA1(A), as the apparent conductance at −50 mV after NMM perfusion is saturating to the maximum similarly for both channels (SI Appendix, Fig. S3F). Therefore, data normalization with 0.1 mM NMM semiquantitatively compares highly divergent current responses to riboflavin between species, regardless of possibly varying NMM sensitivities of TRPA1(A)s. In contrast to the conservation of riboflavin-sensing function in the two insect TRPA1s, human TRPA1 (hTRPA1), deficient for nucleophile sensitivity (13), failed to respond to riboflavin (Fig. 2B and SI Appendix, Fig. S3E). Ambient light (0.5 mW/cm²) is dispensable for riboflavin-induced TRPA1 current responses despite riboflavin phototoxicity, as the response of heterologous agTRPA1(A) insignificantly varies in darkness and light (Fig. 2 C and D). Together with observation on hTRPA1 lacking riboflavin responses, the isoform and species dependences are reminiscent of previously described nucleophile sensing by TRPA1(A) (13), suggesting that the nucleophile detection mechanism may pertain to riboflavin sensing by TRPA1(A). In the previous study, nucleophile sensitivity of dTRPA1(A) was drastically reduced by substitutions of the conserved residues (C105A and R113A/R116A) in the amino-terminal isoform-specific region, whereas electrophile-dependent activation was preserved. We here show that riboflavin responses of dTRPA1(A) were also severely decreased by the mutations (Fig. 2E and SI Appendix, Fig. S3B). To extend implication of this result to agTRPA1(A), C85A and R93A/R96A substitutions corresponding to the dTRPA1(A) mutations above were introduced. The mutant channels agTRPA1(A)C85A and agTRPA1(A)R93A/R96A in oocytes showed severe impairment in the capability of detecting nucleophilic reductant DTT, which potently activated agTRPA1(A)WT (Fig. 2F). Similarly, riboflavin responses of these mutant agTRPA1(A)s were dramatically reduced, reflecting intimate relationship between the mechanisms sensing riboflavin and nucleophilic DTT (SI Appendix, Fig. S3D). The abrogation of riboflavin responsiveness by the mutation was also recapitulated in in vivo GRNs and feeding behavior (SI Appendix, Fig. S3G); expression of dTrpA1(A)R113A/R116A cDNA dominant-negatively suppressed GRN and feeding responses to riboflavin/NMM mixture. Unlike a previous report of successful expression of hTRPA1 in Drosophila (22), we failed to observe consistent functional expression of hTRPA1 in fly GRNs (SI Appendix, Fig. S3H), hindering us from examining its riboflavin responsiveness in flies. To corroborate that a common activation pathway exists for riboflavin and nucleophiles, heterologous dTRPA1(A) fully activated by 0.3 mM riboflavin was further stimulated by applying 10 mM DTT premixed with 0.3 mM riboflavin. The current induced by 0.3 mM riboflavin was insignificantly increased by an additional 10 mM DTT, although dTRPA1(A) was able to show larger currents in response to electrophilic NMM at 0.1 mM (Fig. 2 G and H). Thus, we present four observations as evidence of intimate functional correlation between riboflavin and nucleophile sensitivities: both similar isoform and species dependences, covarying responses of loss-of-function dTRPA1(A) and agTRPA1(A) mutants, and saturation of nucleophile-induced activation by riboflavin. These results strongly advocate that riboflavin detection is a physiological function of TRPA1(A)-mediated nucleophile sensing. In addition, riboflavin analogs, lumiflavin, lumichrome, and FMN were also sensed by TRPA1(A) in vitro and in vivo, and induced gustatory aversion in manners parallel with the riboflavin response (SI Appendix, Figs. S4 and S5), implying the importance of the core structure that accommodates the light-absorbing conjugation system consisting of delocalized electrons in connected π orbitals. This is consistent with the Grutthuss-Draper law that photoabsorption is requisite for photochemical reactions.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Riboflavin stimulates TRPA1(A)s through the nucleophile-dependent activation pathway. (A) Riboflavin dose dependence of TRPA1 isoforms from Drosophila and Anopheles expressed in Xenopus oocytes (n = 4–5). (B) Lack of riboflavin responses of hTRPA1 (n = 4–5). (C and D) Riboflavin-induced currents from agTRPA1(A) with or without ambient light (∼0.5 mW/cm²; n = 5). (E and F) Riboflavin and DTT responses of nucleophile-insensitive TRPA1(A) mutants (n = 4–5). (G and H) Saturation of nucleophilic DTT-evoked activation of dTRPA1(A) by riboflavin (n = 6). NMM, electrophilic agonist, N-methylmaleimide; n.s., not significant, paired t test. *P < 0.05; **P < 0.01; ***P < 0.001, Tukey or unpaired t test.

Nucleophiles act as electron donors in chemical reactions. Similarly, type 1 photochemical reactions transfer photoexcited electrons from phototoxins to other molecules, thereby generating free radicals and exerting phototoxicity (1). In addition to the intimately linked riboflavin and nucleophile sensitivities of TRPA1(A), the conceptual agreement between nucleophilicity and type 1 phototoxicity prompted us to investigate whether the nucleophile receptor TRPA1(A) indeed selectively detects type 1 photochemical potential of pigment molecules. To this end, we first set out to identify more pigments activating TRPA1(A) in darkness. Of 11 color-expressing dyes and four cyclic chemicals with the π conjugation system (SI Appendix, Fig. S7 and Table S1), five pigments, such as rose bengal, eosin Y, porphine (meso-tetraphenylporphine-4,4′,4″,4‴-tetrasulfonic acid), methylene blue, and merocyanine 540, were found to stimulate agTRPA1(A) in addition to riboflavin and its analogs (Fig. 3 and SI Appendix, Figs. S6 and S7). Isoform and species dependences were observed as with riboflavin. Moreover, agTRPA1(A)C85A and agTRPA1(A)R93A/R96A mutants, defective for both nucleophile and riboflavin sensing, showed severely reduced responses to the dyes activating wild-type agTRPA1(A). Compared with other pigment agonists of TRPA1(A), rose bengal provoked relatively large currents from dTRPA1(A) and agTRPA1(A). This is probably a result of dual stimulation of TRPA1(A)s via both electrophile- and nucleophile-dependent activation pathways, given that dTRPA1(B) and agTRPA1(B), which sense electrophilic but not nucleophilic reactivity (13), were moderately stimulated by rose bengal. TRPA1(A) can be activated by either nucleophiles or electrophiles, and responds more robustly to amphiphilic chemicals, as other polymodal TRP channels are synergistically stimulated by two different activation modalities (23, 24). For example, TRPA1(A) and TRPA1(B) yielded rose bengal-like activation profiles in response to hydrogen peroxide (H₂O₂) bearing both electrophilicity and nucleophilicity (13). Porphine shows inhibition of agTRPA1(A) at a high concentration (0.2 mM; SI Appendix, Fig. S6). As porphine is a structural backbone of heme in oxygen-delivering red blood cells, hematophagous mosquitoes may have evolved to accept rather than feel averse to porphine by repressing TRPA1(A) activity at the high concentrations of porphine. Detection of these pigments by agTRPA1(A) is independent of ambient light, as observed with riboflavin (SI Appendix, Fig. S8).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Identification of pigments activating TRPA1(A) with isoform and species dependences as observed with nucleophiles. (A and B) Maximum current amplitudes by pigments and chemicals of indicated TRPA1 variants. If current increases are not saturated in dose dependence experiments, the maximally soluble concentrations in Barth’s solution were used. Colors of box plots simulate those of pigments. Gray boxes for Drosophila TRPA1 isoforms (A), and black boxes for chemicals without color expression (B). *P < 0.05, Games-Howell test between four agTRPA1 variants; ^§P < 0.05, ^§§P < 0.01, and ^§§§P < 0.001, unpaired t test between dTRPA1 isoforms; ^††P < 0.01 and ^†††P < 0.001, unpaired t test between TRPA1(A)s. n = 4–6 (A) and 3–10 (B). Lowercase letters indicate statistically distinct groups (P < 0.05, Tukey).

Next, to test our hypothesis that TRPA1(A) selectively detects radical-generating phototoxins among pigments, we examined whether TRPA1(A)-activating pigments were type 1 phototoxins by adopting polymerization of acrylamide as a reporter of free radical production. Beyond its biochemical utility in gel preparation for polyacrylamide electrophoresis (PAGE), radical-initiated polymerization is a well-characterized phenomenon in the field of polymer science. Indeed, riboflavin and FMN have been used to initiate PAGE gel polymerization in combination with light illumination (25). This was reproduced in our hands with not only riboflavin/FMN but also most of TRPA1(A)-activating pigments at 5 μM (Fig. 4A). Typical acrylamide polymerization for PAGE by ammonium persulfate, which chemically, not photochemically, generates free radicals, requires tetramethyl-ethylenediamine (6.67 mM). Consistently, tetramethyl-ethylenediamine was necessary for light-induced pigment-dependent acrylamide gelation. Whereas the polymerization was monitored for 60 min, the gel solution was continuously irradiated with white light at ∼90 mW/cm² from a xenon lamp, simulating natural sunlight in wavelength composition and irradiance on the ground (13, 26). Eight of nine TRPA1(A)-activating pigments completed gelation within 30 min (mostly <10 min except porphine), whereas 3.4 mM ammonium persulfate completed gelation within ∼14 min without light irradiation (Fig. 4A). Similar results were obtained with 0.1 mM of each pigment that failed to cause the polymerization at 5 μM (SI Appendix, Fig. S9A; 0.01 mM and 0.05 mM for hypericin and curcumin, respectively, because of solubility). Nevertheless, our photochemical gelation assays may not be sufficiently sensitive, leaving the possibility of false-negative results. Therefore, the phototoxic potential was reappraised with the fluorescent superoxide indicator, dihydroethidium, which is converted to 2-OH-ethidium by reacting with superoxide radicals (27) (Methods). Consistent with photochemical gelation, 1-min-long illumination of the white light to the nonpolymerizing pigments/chemicals led to little fluorescence increases of 2-OH-ethidium, whereas lumichrome, rose bengal, porphine, and methylene blue showed fluorescence increases (Fig. 4B and SI Appendix, Fig. S9B). Type 2 photochemical reactions excite triplet oxygen to singlet oxygen (¹O₂), which is also toxic because of its reactivity. To examine whether TRPA1(A)-activating pigments are associated with the type 2 reaction, we turn to anthracenediyl-bis(methylene)dimalonic acid (ABDA), which is bleached by ¹O₂ (28, 29). Pigment solutions containing ABDA were illuminated with the xenon light for 0–2 min. These experiments show that type 2 reactions do not clearly correlate with TRPA1(A) activation compared with type 1 reactions (Fig. 4C and SI Appendix, Fig. S9C). One-minute illumination to type 1-photoreactive lumiflavin and lumichrome bleached ABDA to degrees indistinguishable from ABDA-only control, although riboflavin, FMN, rose bengal, eosin Y, porphine, and methylene blue significantly increased ABDA bleaching. Moreover, ¹O₂ generation can be indirect, as ¹O₂ is often produced in the reaction of the type 1 reaction product superoxide with peroxides (30). Hypericin and α-terthienyl, previously reported to be natural phototoxins (31), were insignificantly distinct from controls, possibly because they are context-dependent phototoxins, as previously observed with berberine, which requires physical binding with DNA for phototoxicity (32). Correlation analysis between pigment-dependent agTRPA1(A) activation and gelation time reveals a negative monotonic relationship (Spearman ρ = −0.872; P = 0.0000002; Fig. 4D), indicating higher radical generation from the dyes more robustly activating agTRPA1(A). Supporting our in vitro results, feeding type 1 phototoxins severely decreased survival of larvae illuminated by white light at 90 mW/cm² (SI Appendix, Fig. S10). In contrast, pigments lacking apparent radical generation such as brilliant blue and sulforhodamine B did not aggravate the survival compared with normal food (SI Appendix, Fig. S10). Note that rose bengal was aphotically toxic, barring us from testing its in vivo phototoxicity even at low concentrations.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

TRPA1(A) activation by pigments is highly concordant to type 1 phototoxicity. (A and B) Light-induced radical generation probed by acrylamide polymerization (A, n = 3–5) and dihydroethidium transformed to 2-OH-ethidium on reaction with superoxide radicals (B, n = 4–12). Excitation/emission: 405/570 (B) or 405/540 nm (Inset). (C) Light-induced singlet oxygen production probed by ABDA bleaching (n = 3–4). Letters indicate statistical differences between groups. P < 0.05, Welch’s ANOVA Games-Howell tests. (D) Spearman correlation analysis showing negative monotonic relationship between agTRPA1(A) activation and acrylamide gelation time. (E–G) Concordances of agTRPA1(A) activation (E), molar photoabsorbance (F), and highest–lowest unoccupied molecular orbital energy gaps (G) to type 1 phototoxicity. Phototoxicity concordance represents percentage of data nonoverlapping between type 1 phototoxins and nonphototoxins (arrows). n.s., not significant. **P < 0.01, unpaired t test. Arrowhead: merocyanine 540.

To assess how closely TRPA1(A) activation is associated with type 1 phototoxicity, we calculated phototoxicity concordance, the percentage of chemicals specifically concordant with the phototoxicity for each account of pigment properties such as TRPA1(A)-stimulating capability or photochemical properties; compounds with values out of the data range of the other phototoxicity category (either type 1 or non-type 1 phototoxins) were determined to be concordant to type 1 phototoxicity (Fig. 4 E–G and SI Appendix, Fig. S9D, arrows). Notably, type 1 phototoxicity is 89% concordant to amplitudes of agTRPA1(A) responses (Fig. 4E). Although light absorbance is requisite for photochemical reactions, molar attenuation coefficient (MAC) of pigments yielded low phototoxicity concordance of 42% among tested molecules (Fig. 4F and SI Appendix, Table S1). Maximum-conjugated π electron numbers (PENMC) (4), which determine wavelengths of absorbed light, also showed marginal phototoxicity concordance (21%; SI Appendix, Fig. S9D). The energy gap between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) has been proposed to be a predictor of phototoxicity, albeit with a mediocre level of prediction success (4, 33). Concordance of HOMO–LUMO gaps (HLG) to type 1 phototoxicity reached only 63% in our survey (Fig. 4G and SI Appendix, Table S1), which is lower than that of TRPA1(A) activation, at 89% (P = 0.0375, Berger’s exact unconditional test). Merocyanine 540, evoking the lowest current amplitude from agTRPA1(A), is the only pigment that failed to exhibit type 1 phototoxicity among those activating TRPA1(A) possibly because its MAC and HLG are suboptimal for robust photosensitization (Fig. 4 E–G, arrowheads, and SI Appendix, Table S1), suggesting that multiple photochemical descriptors such as HLG, MAC, and nucleophilicity must be concurrently fulfilled to constitute type 1 photoreactivity.

Our study on the molecular mechanism for gustatory aversion to phototoxins shows that TRPA1(A) selectively and aphotically recognizes pigments with intrinsic type 1 photochemical potential. Interestingly, agTRPA1(A) exhibited stronger responses to phototoxins than dTRPA1(A) in terms of current amplitudes and, in some cases, EC50s (SI Appendix, Table S1). Along with the recent finding on Anopheles’ sensitive biting cessation by light (34), which would require agTRPA1(A), given the previous study with Drosophila (13), natural variation of phototoxin responses between dTRPA1(A) and agTRPA1(A) suggests that malaria-transmitting Anopheles need to avoid light and phototoxins more acutely than Drosophila. This may be because of evolutionary pressure, possibly from their high susceptibility to phototoxicity, a potential window to controlling mosquitoes that transmit deadly diseases. Although human TRPA1 does not detect nucleophiles (13), mammalian TRPC5 responds to DTT and thioredoxin via reduction of an extracellular disulfide bond (35), implying that phototoxin sensing by nucleophile receptors may be a general phenomenon arising across a wide range of animal species. However, it awaits further investigation if TRPC5 is a nucleophile receptor sensing phototoxins rather than being narrowly tuned for specific reductants.

As nucleophile sensitivity of TRPA1(A) is necessary for its responsiveness to type 1 phototoxins, we propose that type 1 phototoxicity requires nucleophilicity. Association of nucleophilicity with type 1 phototoxicity is unprecedented to our knowledge, but conceptually befitting, as type 1 photoreaction, during which photoexcited electrons in pigments are transferred to other molecules/sites, is a likely manifestation of nucleophilicity that describes electron-donating chemical reactivity. Thus, a sensory function naturally invented through evolution discriminates phototoxins that have been imperfectly predicted by modern chemistry, using phototoxin descriptors such as HLG, MAC, and PENMC. Our findings position TRPA1(A) as the sensory nexus for photoprotection of insects; TRPA1(A) detects not only chemical causes of phototoxicity (phototoxins) but also consequences of photochemical reactions (free radicals and reactive oxygen species) (13, 36, 37), highlighting nucleophilicity as an underlying theme of type 1 phototoxicity.