Conformational and Functional Effects Induced by D- and L-Amino Acid Epimerization on a Single Gene Encoded Peptide from the Skin Secretion of Hypsiboas punctatus

Skin secretion of Hypsiboas punctatus is the source of a complex mixture of bioactive compounds where peptides and small proteins prevail, similarly to many other amphibians. Among dozens of molecules isolated from H. punctatus in a proteomic based approach, we report here the structural and functional studies of a novel peptide named Phenylseptin (FFFDTLKNLAGKVIGALT-NH2) that was purified as two naturally occurring D- and L-Phes configurations. The amino acid epimerization and C-terminal amidation for both molecules were confirmed by a combination of techniques including reverse-phase UFLC, ion mobility mass spectrometry, high resolution MS/MS experiments, Edman degradation, cDNA sequencing and solid-phase peptide synthesis. RMSD analysis of the twenty lowest-energy 1H NMR structures of each peptide revealed a major 90° difference between the two backbones at the first four N-terminal residues and substantial orientation changes of their respective side chains. These structural divergences were considered to be the primary cause of the in vitro quantitative differences in antimicrobial activities between the two molecules. Finally, both molecules elicited equally aversive reactions in mice when delivered orally, an effect that depended entirely on peripheral gustatory pathways.


Introduction
According to Stahl [1], multifunctional drugs are agents with more than one therapeutic mechanism and it has been observed that almost all existing drugs have more than one known pharmacological target when higher doses exceeding the remedial recommendations are administrated. Fine-tuned with Paracelsus legendary quote, the multifunctional concept of a drug depends upon a given concentration to reveal its multiple pharmacological activities. Among many others, a classical example of this principle is Aspirin [2], the well-known drug used for more than a century as analgesic, antipyretic and anti-inflammatory, in the end 1960s was also confirmed experimentally to be an effective inhibitor of platelet aggregation [3], therefore an important agent for ischemic stroke prevention [4]. Pondering these facts and more recent ones [5], it is reasonable to ask: why the multifunctional drug principle should be confined only to small molecules, natural products and/ or their synthetic analogs? Would peptides and proteins also show similar properties under comparable experimental conditions? How many different structures, functional sites and therefore biological activities a single polypeptide chain would display after an extensive investigation?
A noteworthy indication towards a more comprehensive answer to these questions was given by Brandenburg and co-authors [6] in their recent review on antimicrobial peptides, when the multifunctional drug attributes of this class of molecules and a possible range of different applications were underlined.
The other end of the same spectrum show us that in principle, from the organism perspective, any exogenous substance alone may represent a potential hazardous to its integrity and/or survival. Living beings are equipped with numerous barriers, defense strategies and metabolic mechanisms to deal with physical, chemical and biological threats. In all that, small and macromolecules may be found acting in both ways, as belligerent and defense agents depending on the situation. It is interesting to note that animals may engender complex multimodal signals combined with volatile and toxic substances that play critical roles in defense strategies as well as in social interactions of a number of animal species [7][8][9][10]. In plants, it was demonstrated that homoterpene could also be released from damaged aerial tissues induced by herbivore attacks revealing a signaling mechanism similar to those in animals [11,12]. In other words, many molecules may have multiple functions and activities depending on the metabolic target and relative concentration.
As for amphibians, it is widely known that they protect themselves against various biotic hazards using an intricate collection of strategies including noxious secretions, aposematism, deimatic behavior and evasive tactics [13]. The granular glands in anurans skin store extraordinary quantities of toxic compounds [14][15][16][17], which are also the most evident molecular ingredients of some aversive chemosensory signals released against potential predators to reduce the chances of future attacks [18,19], as exemplified by the fact that some amphibian eggs and larvae are perceived as repulsive by some fish and birds [20][21][22]. Amongst the great variety of substances secreted by these glands, several of them are antimicrobial peptides [15,[23][24][25][26][27][28] typically characterized by basic, linear and amphiphilic primary structures of 30-35 residues in a single-chain [27,[29][30][31]. Although their protecting roles as microbicidal agentes have been thoroughly investigated during previous decades, little is known about other biological activities that antimicrobial peptides may display [32,33]. In the present work, we report the 1 H NMR structures and functional studies of a novel antimicrobial peptide that also displays aversive gustatory properties in mice models. The peptide naturally occurring in two enantiomeric configurations was isolated from Hypsiboas punctatus skin secretion and sequenced by a combination of ion mobility mass spectrometry, Edman degradation and cDNA methodologies.

Amphibian Skin Secretion
Adult specimens of H. punctatus were collected in Palmas, Tocantins, Brazil. All procedures in this study were approved and performed are under the license IBAMA 0637/91 AC. The frog skin glands were subjected to mild electrical stimulation (6 V during 1 min) and skin secretion was collected in ice-cold Milli-Q H 2 O, filtered, and freeze-dried and stored at 280uC, as described by Prates, et al 2004 [34,35].

Peptide Purification
The crude extract was dissolved in aqueous 0.1% trifluoroacetic acid (TFA), then fractionated by RP-HPLC (Class VP Shimadzu Co., Japan) using a VydacH 218TP510 semi-preparative C 18 column (Grace VydacH TP, USA). The fractions were eluted in a 120 minutes linear gradient of Solvent A (0.1% TFA in Milli-Q H 2 O) and Solvent B (0.1% TFA in acetonitrile) using a 2.5 mL min 21 flow rate. Chromatographic fractions containing peptides of interest were submitted to further purification steps on analytical columns VydacH 218TP54 C 18 (Grace VydacH TP, USA), Source TM 5 RPC ST (GE Healthcare, WI), and Shim-Pack-XR-ODS (30.0 mm62.0 mm) C 18 column on an Ultra-Fast-Liquid-Chromatographer (UFLC Prominence System Shimadzu Co., Japan). All the experiments were monitored at 216 and 280 nm (or 254 nm when phenylalanine residues were present), collected manually, frozen, lyophilized, and stored at 220uC.

Mass Spectrometry Analysis and Peptide Sequencing
The accurate molecular mass values of the peptides were determined by ESI-micrOTOF-Q II (Bruker Daltonics, Germany). The purity of each chromatographic fraction was evaluated by MALDI-TOF/MS (UltraFlex III, Bruker Daltonics, Germany) using close external calibration under reflector mode. Approximately 20 nM of lyophilized peptide was dissolved in Milli-Q H 2 O, mixed to a saturated solution of a-cyano-4-hydroxycinna-minic acid, spotted on a MALDI sample plate, and dried at room temperature. Peptide fragmentation was obtained by MALDI-TOF MS/MS experiments and the resulting data were analyzed manually using both Pepseq [36] and Flex Analysis 3.0 (Bruker Daltonics) softwares. The synthetic peptides were analyzed with a high-capacity ion trap LC/MSn system HCT Ultra ETD (Bruker Daltonics, Germany) equipped with a standard ESI ion source, with full scan data in UltraScan mode between m/z 100-3000, in positive mode, and with a fixed accumulation time of 30 ms. Additionally, amino acid sequencing was performed by the automated Edman degradation method on a PPSQ-23 protein peptide sequencer (Shimadzu Co., Japan).

Structural Studies on Ion Mobility Mass Spectrometry
Experiments were performed on a both MALDI Synapt G1 and ESI G2 HDMS instrument (Quadrupole Ion Mobility High-Definition mass spectrometry -Waters Co. MA, USA) equipped with nano-electrospray ionization source. All spectra were acquired by direct infusion of 1 mL?min 21 at a range of 300 up to 2000 m/z. Cross section calculations of Phenylseptin were conducted as described [37].

Nucler Magnetic Resonance (NMR)
The sample was prepared dissolving the peptide at 2 mM in a 60% TFE/D 2 O (v/v) solution. The NMR experiments were conducted at 20uC on a Bruker Avance III spectrometer (Bruker DRX-800) operating at 800 MHz for the 1H frequency. The Total Correlation Spectroscopy (TOCSY) spectra were acquired using the MLEV-17 pulse sequence. The spectral width was determined as 6,900 Hz, and the 512 t 1 increments were collected with eight transients of 4,096 points. NOESY spectra were acquired using mixing times of 80, 100, 120, 140 and 160 ms. For this experiment, the spectral width was 6,900 Hz and the 512 t 1 increments were collected with 16 transients of 4,096 points for each F 1 . The 1 H-13 C HSQC spectra were acquired with F 1 and F 2 spectral widths of 27,160 and 8,993 Hz respectively. The 400 t 1 increments were collected with 56 transients of 1,024 points. The acquisitions were carried out in an edited mode allowing CH and CH 3 correlations could show a positive phase and CH 2 correlations, a negative phase. The 1 H-15 N HSQC spectra were acquired with F 1 and F 2 spectral widths of 27,160 and 8,993 Hz, respectively. Eighty t 1 increments were collected (with 400 transients of 1,024 points), for each free induction decay. NMR spectra were analyzed using NMRVIEW, version 5.0.3. NOE intensities obtained at 120 ms mixing times were semi-quantitative converted into distance restraints using the calibration described by [38]. The highest intensity peaks corresponded to restraints with upper limit of 2.8 Å , the medium intensity with 3.4 Å , and the weakest peaks, with 5.0 Å . The backbone atoms chemical shift values were used to generate the dihedral angle restraints, using the TALOS+ software [39]. Structures were calculated (200) using these restraints on the Xplor-NIH v.2.17 software [40] by simulated annealing protocol. The structure calculation started with an extended model, with 14000 steps at high temperature, 6000 steps during cooling and a time step of 0.005 ps. The generated structures were then submitted to a refinement procedure, using again a simulated annealing calculation, using the Internal Variable Module [41] available in Xplor-NIH, with the Hydrogen-Bonding Database, Ramachandran Torsion-Angle Database and the Carbon Chemical Shift potentials. Among the refined structures, the 20 lowest-energy ones were selected to represent the model ensemble and then validated on Procheck online software [42].

Gene Cloning, cDNA Sequencing and Analysis
Total RNA from H. punctatus was extracted using the Trizol reagent according to a previously described protocol [23]. Reverse transcription of total RNA (1 mg) was performed with d(T)-anchor primer and the Superscript Reverse Transcriptase kit TM (Invitrogen), according to the manufacturer's instructions. PCR reactions were performed using degenerate primers designed based on conserved signal peptide [24]. The amplified cDNA fragments were cloned into pGEM-T easyH (Promega, Maldison, WI, USA) and sequenced in both strands using an ABI Prism 3700 DNA Analyzer system with BigDye TM terminator and POP-5 TM Polymer (Applied Biosystem Perkin Elmer). Comparisons of the cloned cDNA sequence with other GenBank TM database sequences were performed using the BLASTp software [43] from the NCBI databank (http://www.ncbi.nlm.nih.gov). The Conserved Domain Database search (CDD-Search) from the NCBI site was used to compare motif identity and similarity with known conserved domains [44]. Sequence alignments were obtained by CLUSTAL W software [45] and were edited using BIOEDIT [46]. The calculated molecular mass of deduced proteins and sequenced peptides were determined by the Protein Machine software available at the Expasy website (http://us.expasy.org/ tools/).

Solid Phase Peptide Chemical Synthesis
The peptides L-Phes, D-Phes and one synthetic short-analog L-Phes.1 were manually synthesized by solid phase using the Fmoc/ t-butyl strategy. An Fmoc-PAL-PEG-polystyrene resin was used for synthesis of the amidated C-terminal segment. Cleavage and final deprotection were conducted with a TFA:thioanisole:ethanedithiol:triisopropylsilane (91.5:5:2.5:1, v:v:v:v) solution for 1 hour at room temperature. Peptide purification was performed through RP-HPLC with a Vydac 218TP1022 preparative column and purity was assessed by MALDI-TOF/MS.

Antimicrobial Assays
The antimicrobial activity of the peptides L-Phes and D-Phes were investigated using the bacterial strains Pseudomonas aeruginosa ATCC 27853, Staphylococcus aureus ATCC 43300, and Escherichia coli ATCC 25992. In addition, we investigated the antimicrobial activity of the peptides towards a phytopathogenic bacterium Xanthomonas. axonopodis pv glycines ISBF 327 obtained from Embrapàs microorganism collection. All microorganisms were grown in stationary culture at 37uC and after that were transferred to Mueller-Hinton liquid medium, according to the National Committee for Clinical and Laboratory Standards (NCCLS) to perform the bioassays (NCCLS Institute protocols). The peptide was dissolved up to 8-fold in Mueller-Hinton liquid broth. The highest peptide concentration used in the assay was 131 mM in an initial inoculum of 2.5.10 5 cfu?mL 21 (colony-forming units/mL). The final volume was 100 mL per well, 50 mL of the peptide and 50 mL of the inoculum. The experiment was carried out in stationary culture at 37uC, and the spectrophotometer readings were performed 12 h after incubation. The minimal inhibitory concentration (MIC) was determined based on three independent measurements, using the optical density parameter (A 595 nm). Conventional antibiotics (ampicillin and chloramphenicol) had their minimum inhibitory concentrations determined against the three experimental bacterial strains.

Behavioral Gustatory Assays
Male, adult (8-10 weeks old) mice on a C57BL/6 background were used for behavioral assays. Animals used in this study included littermates generated from mice heterozygous for a partial deletion of the Trpm5 gene, as previously described [47], originally bred from mice generously donated by C. S. Zuker (UCSD, San Diego, CA). Genotype was confirmed by PCR amplification using specific primers for Trpm5 gene (forward 59-ATTCTAGAGCCCACCCGCCCCATC-39 and reverse 59-TTCACCTGCCCAGCCCTCATCTAC-39) and for PGK-neor cassette that replaced exons 15 to 19 encoding the first five transmembrane domains of Trpm5 in mutant animals (forward 59-TTGCACGCAGGTTCTCCGGC-39 and reverse: 59-TA-GAAGGCGATGCGCTGCGA-39). Heterozygous mice were identified by amplification of both the Trpm5 gene and PGKneor cassette; knockout mice were identified by the absence of Trpm5 gene but amplification of PGK-neor cassette; and wild-type animals by absence of PGK-neor cassette but positive Trpm5 gene amplification. Behavioral experiments were conducted in two mouse behavior chambers enclosed in a ventilated and soundattenuating cubicle (Med Associates Inc., St. Albans, VT). Each chamber was equipped with two slots for sipper tubing placement in symmetrical locations of one of the walls. All slots were equipped with licking detection devices (contact lickometers, Med Associates Inc., St. Albans, VT, USA) performing with a 10 ms resolution. All experiments were conducted in accordance with the J.B. Pierce Laboratory and Yale University regulations on usage of animals in research. All procedures performed in this study were approved by the Animal Care and Use Committee of The J. B. Pierce Laboratory.
9.1 Controls and stimulus concentrations. As a positive control, well known plant bitter alkaloids were used: caffeine, theophylline, and theobromine. Stimulus concentrations were based on previous studies, which reported, in rats, that the threshold for integrated cranial nerve responses is 10 mM for the glossopharyngeal nerve and approximately 10-30 mM for the chorda tympani [48]. Accordingly, we used 10-20 mM (as well as the sub-threshold concentration of 5 mM) in our studies to evaluate the role of Trpm5 in generating behavioral responses to caffeine. Finally, to allow for comparisons between behavioral responses across controls on a molar basis, we have used these same concentrations in theophylline and theobromine experiments. As a negative control we used free amino acid mixtures corresponding to the peptide primary structures. Stimulus concentrations were selected based on experiments conducted by Maehashi and Huang, 2009, where it was shown that a 10-20 mM threshold of free-amino acid is able to display bitter response in mice and rats. For the peptides L-Phes, D-Phes, and L-Phes.1 we considered the same concentrations as starting points.
9.2. Short-term two-bottle preference tests. Once habituated to the behavioral chamber, each animal was presented with two bottles, contained either water or the stimulus solution prepared in distilled water. Animals were given free access to both bottles during the experiment (10 minutes). The number of licks for each sipper was recorded and used to calculate the preference ratios (see below). To reduce confounds associated with side-biases, mice were tested in each condition for four consecutive days with daily inversion of bottle positions. The average preference ratio across testing days was then calculated for each animal. Animals had been water-deprived for 16 hours previous to the beginning of each session.
9.3. Data analysis. Results from data analyses were expressed as mean 6 SD. Analyses of behavioral data were performed with the custom software Origin Professional (v8, OriginLab, Northampton, MA) using 2-way or 1-way ANOVAs followed by two-sample or independent one-sample t-tests.
Bonferroni corrections for multiple comparisons were performed whenever appropriate.
9.4. Lick response and preference measures. All twobottle preference tests were analyzed by calculating the preference ratios as: Where n(.) denotes the total number of licks counted on a given sipper during a session. These values were subjected to a 2-way ANOVA genotype 6 stimulus concentration model and tested against 0.5, which is the reference value meaning indifference with respect to water.

Phenylseptins on H. punctatus Skin Secretion
The crude skin secretion was fractionated by semi-preparative reverse-phase HPLC and mass analyzed by both MALDI TOF-TOF/MS and ESI Q-TOF/MS systems yielding about 20 major fractions ( Figure 1A) from which, components ''a'' and ''b'' eluted at 61.0 and 64.3 minutes, respectively, showed identical molecular masses and MS/MS spectra. Each of the fractions was submitted to further purification steps on analytical scale to attain homogeneous peptide samples ( Figure 1B-1E), confirmation of their distinct retention times, MALDI-TOF/MS monoisotopic molecular mass analyses (M+H + = 1954.20 Da, Figure 2A) and primary structure elucidation from their corresponding MS/MS data ( Figure S1). For both components (a and b), de novo sequencing based on the resulting sets of undistinguishable daughter ion series revealed 100% similarity between the amino acid sequences of the two components (see Table S1) bearing the following 18 residue polypeptide chain: FFFDTLKNLAGKVI-GALT-NH 2 . Edman degradation of the intact components and their respective tryptic fragments excluded typical MS/MS data ambiguities regarding leucine, isoleucine, lysine and glutamine residues (Table S2). Moreover, the final sequence confirmation of both peptide and Thr18 C-terminal amidation was achieved by cDNA sequencing as shown in Figure 3 and ESI Q-TOF/MS exact mass of M+2H + = 977.561 m/z (1.0 ppm) determinations ( Figure S2).

One Single Gene Encoded Peptide
RT-PCR amplification produced various 320-bp fragments that revealed one single nucleotide sequence encoding a canonical antimicrobial pre-pro-peptide precursor (18) containing a deduced 18 residues peptide sequence ( Figure 3A) identical to the previously obtained by mass spectrometry and Edman degradation. The pre-pro-peptide precursor also showed primary structure identity with antimicrobial sequences previously deposited in GenBank TM from three different Ranidae species, Rana rugosa (37% gaegurin), Rana temporaria (48% brevinin-2Ta and 46% 2Tb), Rana esculenta (50% brevinin-2Ef), and from one Hylidae specie, Litorea aurea (54% aurein 3.1; 56% aurein 2.5 and aurein 2.3; 57% aurein 2.2) ( Figure 3B). The unusual triple phenylalanine residues at the N-terminal of the two putative antimicrobial peptides, their discrepancies in retention times by RP-HPLC and the confirmed complete sequence identity brought us to the hypothesis of a possible epimerization of at least one amino acid residue along the polypeptide chain, hence the name Phenylseptins for the two peptide analogs. The gene sequence was deposited under the genebank (gb|) accession number HQ012497.

D-phenylalanine is Present in Native D-Phes
Reverse-phase UFLC on a Shim-Pack-XR-ODS (2.0 mm i.d. 6 30.0 mm) C 18 column analyses determined that Phes 'a' was associated with a retention time of 11 minutes, and Phes 'b' with a retention time of 12 minutes ( Figure 2B -black line). After enzymatic digestion with immobilized trypsin, UFLC analyses were performed to determine the retention times of all fragments from of these peptides ( Figure S3 and Table S2). These analyses demonstrated that the fragments 1-7, corresponding to Phes 'b' Nterminal region (FFFDTLK; M+H + = 917.55 Da) showed almost one minute difference in retention time when compared with the same fragment in Phes 'a'. In order to investigate the presence of a D-phenylalanine residue at the second N-terminal position as suggested by previous related works [49][50][51][52], analogs containing

Conformational Studies of Phenylseptin Analogs by Ion Mobility Mass Spectrometry (IM-MS)
In order to investigate possible conformational attributes of both Phenylseptins, these molecules were submitted to IM-MS analyses revealing at least two conformers (M+3H + = 652.04 m/z) with comparable intensities for L-Phes and a major one for D-Phes with determined drift times of 10.45, 12.89 and 10.80 ms, respectively ( Figure 4A). These results suggest higher degrees of freedom for the L-Phes structure in the gas phase in comparison to the D-Phes molecule.

NMR Structures of the Phenylseptins
Both NMR structures of L-and D-Phenylseptins at 60% TFE were produced using total sets of 361 and 373 NOEs, respectively (see Table 1 for details). NOEs correlations at the peptides three phenylalanine N-terminal region indicated significant structural differences between L-and D-Phes most likely due to Phe2 enantiomerization ( Figure 4). Nevertheless strong NOEs correlations of the sequential dNN (i,i+1) and daN (i,i+3) obtained for the rest of the molecule (residues 4-18) indicated classical a-helical conformations at that region for both peptides ( Figure 4F-4G). The Root Mean Square Deviation (RMSD) analyses of all backbone atoms for individual structures were 0.5360.21 for L-Phes and 0.3860.18 for D-Phes, and no distance violations greater than 11 was observed. The RMSD values for segment 4-17 were similar between L-Phes and D-Phes (0.2260.09 and 0.2260.08, respectively), while L-Phes showed 0.4060.22 for the 1 Phe-3 Phe segment and D-Phes showed 0.0860.06. In the calculated D-Phes structure it was observed that Phe1 and Phe3 aromatic rings have significant interactions whilst Phe2 share NOEs with the peptide side chain with Leu6 producing a more structured N-terminal than in L-Phes ( Figure S5).

Phenylseptins Display Different Antimicrobial Activities
In vitro antimicrobial tests revealed that both peptides exhibited distinct levels of activities against pathogenic S. aureus (ATCC 29313), E. coli (ATCC 25922) and P. aeruginosa (ATCC 27853).
Overall, D-Phes was more effective than L-Phes and Magainin (classical positive control) [15] against these pathogenic bacteria but less active than DS01 [53], which was used as the most potent antimicrobial peptide (our groups positive control) ( Table 2). When tested against the soybean phytopathogen, Xanthomonas axonopodis pv glycines, D-Phes was found to be 8-fold more effective than L-Phes (4.1 and 32.7 mM, respectively).

Bitterness Motif Hypothesis
Analyses of Phenylseptin sequences revealed that their Phe-rich N-terminals share sequence similarities with the bitter Phe-Phe-  Phe motifs shown in Figure 5 [54]. Therefore, we hypothesized that the triple-phenylalanine N-terminal sequences found in the two peptides would also produce similar aversive behavioral responses to those produced upon stimulation of oral bitter taste sensors.

Gustatory Preference Tests Involving Prototypical Bitters in Trpm5 KO and WT Mice
The methylxanthines caffeine, theophylline and theobromine are prototypical bitter tastants and were used as positive controls for bitter detection in WT animals and bitter-insensitivity in KO animals at 5, 10 and 20 mM concentrations. For the caffeine experiments, we designed a mixed-model two-way stimulus concentration 6 genotype ANOVA 6 which revealed, as expected, a robust effect of genotype (p,0.0003) on preference ratios. Similar analyses performed for theophylline revealed equally robust effects of genotype on preference ratios (p,0.001). In addition, these ANOVAs revealed significant stimulus concentration 6 genotype interactions (p,0.02 and p,0.008; caffeine and theophylline, respectively). The interaction effects prompted repeated measures one-way ANOVA models for each genotype separately (Table S3), which revealed a significant main effect of stimulus concentration on preference ratios in WT (p,0.02 and p,0.01, caffeine and theophylline, respectively), but not in KO mice (p = 0.4 and p = 0.95, caffeine and theophylline, respectively). For theobromine this analysis revealed neither significant main effect of genotype (p = 0.17) nor interaction (p = 0.4) on preference rations, although it was the only stimuli to show a significant effect of stimulus concentration (p,0.02). Oneway ANOVAs performed for each genotype separately confirmed a significant effect of theobromine concentrations in WT (p,0.04), but not in KO (p = 0.29). Post-hoc analyses of the individual preference ratios for each stimulus confirmed that, while WT mice displayed significantly higher preferences for water against 10-20 mM caffeine, 5-10-20 mM theophylline and 20 mM theobromine, KO mice were uniformly indifferent to all choices. Our findings so far show that, in 10 min-long two-bottle preference tests, Trpm5 KO mice are indifferent all concentrations of methylxanthines, while some concentrations (10-20 mM caffeine; 5-10-20 mM theophylline; 20 mM theobromine) are robustly avoided by WT mice. All control results are depicted in Figure 6A-6C and Table S3.

Gustatory Preference Tests Involving Phenylseptins in Trpm5 KO and WT Mice
After methodology validation, we tested both Phenylseptin peptides and one shorter analog. The analog L-Phes.1 was used as a control to investigate the importance of the full peptide  sequence. We reasoned that the peptide N-terminal might be a crucial modulator of bitterness potency in these peptides. The peptides were tested at 0.01 mM, 0.1 mM and 1 mM in behavioral tests identical to those previously described. WT (N = 7) and KO (N = 7) mice were exposed to 10-min two-bottle preference tests where choice was given between a Phes solution and (distilled) water. Analysis of two-way genotype 6 peptide concentration ANOVA designed as above revealed a robust effect of genotype for L-Phes (F 1,12 = 31.9, p,0.0002), D-Phes (F 1,12 = 21.2, p,0.0006) and L-Phes.1 (F 1,12 = 76.1, p,0.0005), on preference ratios. Conversely, no concentration effects were detected for L-Phes (F 1,12 = 1.12, p = 0.31), D-Phes (F 1,12 = 1.32, p = 0.27) or L-Phes.1 (F 1,12 = 1.23, p = 0.30). Also, no significant peptide concentration 6 genotype interactions for L-Phes (F 1,12 = 0.29, p = 0.75), D-Phes (F 1,12 = 1.32, p = 0.27) or L-Phes.1 (F 1,12 = 0.04, p = 0.84), on the preference ratios. Pairwise comparisons using post-hoc Tukey tests showed that even in low concentrations WT mice rejected the peptides, while KO mice remained indifferent. As observed for the control experiment with methylxanthines, post-hoc analyses on individual preference ratios of peptides confirmed that WT mice demonstrated high preference for water, while KO mice were indifferent to the choices given. Specifically, WT mice have shown significant higher aversion against all concentrations of L-Phes, D-Phes and L-Phes.1 (one-tailed t-tests, p,0.05; Figure 6D-6F). In summary, we attribute the aversive oral properties of the Phenylseptins tested entirely to their gustatory, possibly bitter-like, properties, given the strong impairments observed in mice lacking the taste ion channel TRPM5.

Peptide Primary Structure and Gustatory Perception
1 mM of each amino acid free present in L-Phes and D-Phes sequences were mixed and the resulting solutions were offered to the mice in two-bottle preference tests as above. The two-way ANOVA free-amino acid solution X genotype model showed no significant effect for L-Phes (p = 0.20), D-Phes (p = 0.19) and L-Phes.1 (p = 0.23), or interactions (L-Phes p = 0.37, D-Phes p = 0.41, and L-Phes.1 p = 0.35) on preference ratios for both preparations. One-way ANOVAs were performed separately for each genotype using the amino acid preparation as factor. There were no effects on preference ratios in either WT or KO mice. In fact, as expected for the tested concentration, both WT and KO mice were equally indifferent to 1 mM free amino acid solution ( Figure 6G), indicating the importance of the primary structure for the activity.

Discussion
The present work reports a novel peptide named Phenylseptin (Phes), purified from the skin secretion of H. punctatus. This peptide naturally occurs in two different primary structures constituted by L-Phe2 and D-Phe2 (Figure 2 and S3) and C-terminal amidation at Thr18. Both sequences display distinct in vitro antimicrobial activities but similar sensorial aversive properties on mice. It was observed that the epimerization of the second N-terminal phenylalanine residue induced a 90u conformational change at the N-terminal of the D-Phes structure ( Figure 4) and that could be responsible for the significant differences on the antimicrobial activities obtained for two peptides (Table 1). However, the Nterminal structural differences demonstrated for both molecules appear to cause no detectable response regarding micès gustatory perception to bitterness. These findings are consistent with previous works concerning amino acid epimerization in frog peptides such as dermorphins and deltorphins, as well as predators  aversive taste behaviors to various amphibian skin secretions [16,51,[55][56][57][58].
The Phenylseptins reported in this study were found to adopt at least two major structural conformations and so far, according to our cDNA library, both forms should be considered a single gene product synthesized as a highly conserved pre-pro-protein that yields the two mature polypeptides (L2/D-Phes) after enzymatic processing and corresponding post-translational modifications. Electrospray and MALDI (data not shown) ion mobility mass spectrometry experiments demonstrated the presence of multiple peptide ion (M+H + ) conformers in the gas phase ( Figure 4A), which were confirmed by 1 H NMR in solution ( Figure 4D). The amount of structural data that was unveiled in this occasion for such a small polypeptide demonstrates how subtle, complex and  non-trivial peptide studies can be, and how easily one might be driven to ignore relevant structural and functional features, particularly when high throughput approaches tend to be overemphasized. Furthermore, the uncommon triple phenylalanine N-terminal of the Phenylseptins seems to share similar bitterness attributes observed in peptides released during food hydrolysis or aging process in fermented products [59]. The aversive gustative responses registered in normal mice caused by aqueous solutions with different molar concentrations of Phenylseptins are comparable to those constituted by the classical bitter alkaloids caffeine, theophylline and theobromine, when offered to the same animals ( Figure 6). Therefore, it is plausible to infer that similar aversive activity could also be part of a general warning strategy against natural predators of H. punctatus, especially if one may ponder what was uttered with remarkable precision in De Anima by Aristotle centuries ago: ''…taste also must be a kind of touch, because it is a sensation of that which is tangible and nutritive… Hence excess in tangible qualities destroys not only the sense-organ, but also the animal itself. For touch is the one sense that the animal cannot do without… It (the animal) has taste on account of what is pleasant and painful, to the end that it may perceive what is pleasant in food and feels desire and be impelled to movement.'' (Aristotle, III. 13) [60]. These results strongly suggest that Phenylseptins may be part of a multifunctional pool of peptides and proteins produced by the skin glands of H. punctatus with at least two experimentally tested defensive roles: against pathogenic microorganism and a potential predator warning agent. Table S1