Activation of the Bile Acid Pathway and No Observed Antimicrobial Peptide Sequences in the Skin of a Poison Frog

The skin secretions of many frogs have genetically-encoded, endogenous antimicrobial peptides (AMPs). Other species, especially aposematic poison frogs, secrete exogenously derived alkaloids that serve as potent defense molecules. The origins of these defense systems are not clear, but a novel bile-acid derived metabolite, tauromantellic acid, was recently discovered and shown to be endogenous in poison frogs (Mantella, Dendrobates, and Epipedobates). These observations raise questions about the evolutionary history of AMP genetic elements, the mechanism and function of tauromatellic acid production, and links between these systems. To understand the diversity and expression of AMPs among frogs, we assembled skin transcriptomes of 13 species across the anuran phylogeny. Our analyses revealed a diversity of AMPs and AMP expression levels across the phylogenetic history of frogs, but no observations of AMPs in Mantella. We examined genes expressed in the bile-acid metabolic pathway and found that CYP7A1 (Cytochrome P450), BAAT (bile acid-CoA: amino acid N-acyltransferase), and AMACR (alpha-methylacyl-CoA racemase) were highly expressed in the skin of M. betsileo and either lowly expressed or absent in other frog species. In particular, CYP7A1 catalyzes the first reaction in the cholesterol catabolic pathway and is the rate-limiting step in regulation of bile acid synthesis, suggesting unique activation of the bile acid pathway in Mantella skin. The activation of the bile acid pathway in the skin of Mantella and the lack of observed AMPs fuel new questions about the evolution of defense compounds and the ectopic expression of the bile-acid pathway.

AMPs have been one of the most well-studied and important components of the innate immune system in amphibians and have been a major focus for the development of therapeutics for humans (Zasloff 2002a(Zasloff , 2002b. These peptides have specific cationic and hydrophobic regions that allow them to target bacteria, fungi, enveloped viruses, and some cancerous cells for lysis (Reddy et al. 2004). Several amphibian taxa have independently acquired the genes to produce skin secreted peptide arsenals and these arsenals are diverse across the phylogeny of frogs (Roelants et al. 2013;König et al. 2015;Xu and Lai 2015). The question remains as to how AMPs originate and the functional diversification among defense arsenals in multiple taxa; though at least in Xenopus, some AMPs appear to be derived from a gene cluster that originally had a gastrointestinal hormone function (Roelants et al. 2013).
Another open question is how interactions between different classes of compounds and chemical communities in amphibian skin accomplish diverse protective functions. However, investigating this question is made difficult by the rate at which new compounds are still being discovered, observations of AMPs in some but not other taxa, and the small fraction of species that have been studied (König et al. 2015). Recently, a new bile acid derived compound, named tauromantellic acid (TMA), was discovered from the skin of Madagascar poison frogs in the genus Mantella and poison dart frogs in the genera Dendrobates and Epipedobates (Clark et al. 2012). This was the first example of a bile acid found within the skin secretion of any frog (Clark et al. 2012). Since bile acids normally act as surfactants to aid intestinal digestion, Clark et al. (2012) hypothesized that tauromantellic acid could potentially shield frogs from their own toxicity via the creation of micelles that sequester or aid in the uptake of alkaloids (Clark et al. 2012). However, the ring structure of TMA shares striking similarity to that of the broad spectrum antimicrobial aminosterol, squalamine, previously isolated from many tissues of the dogfish shark (Squalus acanthias), including the liver and gallbladder (Moore et al. 1993;Rao et al. 2000) and the white blood cells of the sea lamprey (Petromyzon marinus; Yun and Li 2007), raising questions as to whether TMA could play a role in antimicrobial functions.
Squalamine is a potent, broad spectrum antimicrobial in which a polyamine (spermidine) is coupled to the steroid ring scaffold of a bile acid (Moore et al., 1993). Although TMA has not been assayed for antimicrobial activity, bile acids of similar structure are known to exhibit weak activity (Sannasiddappa et al. 2017), about 1000-fold lower than frog skin AMPs, and we expect TMA itself will not exhibit potency comparable to frog skin AMPs. However, the union of an anionic bile acid with a cationic polyamine is sufficient to create an amphipathic compound with potent antibacterial activity (Jones et al. 1996;Savage et al. 2002;Shu et al. 2002;Tessema et al. 2013).
The diversity of AMPs and other compounds in frog skin raises new questions about the evolutionary history of AMPs, their relationship with other compounds, and their collective functions. To begin to answer these questions, we compared gene expression in the skin of thirteen frog species, including a newly generated skin transcriptome for Mantella betsileo. We used the Database of Anuran Defense Peptides (Novković et al. 2012) to identify the presence of AMPs in frog taxa spanning the anuran phylogeny. Additionally, we analyzed expression of genes in the primary bile acid biosynthesis pathway in order to look for the functional signals of tauromantellic acid in other frogs and we tested the antimicrobial activity of MA coupled with spermine. Collectively, we show that AMPs are widespread in the frogs we studied, with a likely absence in Mantella, that genes needed for the bile acid pathway are uniquely expressed in Mantella skin, and that there is significant anti-microbial activity for MA when it is coupled to a spermine. Taken together, our results show a correlation between a possible absence of AMPs and activation of the bile acid pathway in the skin of Mantella.

Sequencing and Transcriptome Assembly
We harvested total RNA from the dorsal and ventral skin of five captive bred Mantella betsileo that were anesthetized by carbon dioxide according to Georgetown Institutional Animal Care and Use Protocol No. 2016-1351. We transferred the skin collections to a tube of All-Protect reagent and extracted RNA using a Qiagen RNA kit using manufacturer's protocols (Qiagen, Germany). The mRNA from M. betsileo skin was enriched using oligo-dT primers and fragmented into approximately 200 nt fragments by alkaline hydrolysis. The cDNA library was generated via random hexamer priming and sequenced on an Illumina HiSeq 2000 sequencer to achieve paired-end 90 base pair sequences. We assessed read quality with a FastQC Report v.0.11.2, and trimmed reads using Sickle v.1.33 with a quality score of 30 and a minimum sequence length of 45 base pairs. Approximately 54 million raw reads were generated and 47 million reads passed quality filtering after trimming with Sickle v.1.33 (Joshi and Fass 2011). Following trimming, we assembled the transcriptome of M. betsileo using Trinity v.2.4.0 (Grabherr et al. 2011;Haas et al. 2013) using a minimum contig length of 300. We assessed the continuity of the assembly by the contig N50 statistic and measured read representation by aligning reads back to the transcriptome with Bowtie2 v.2.3.1 (Langmead and Salzberg 2012) to identify proper pairs and lone read alignments. We used BUSCO v.3 (Simão et al. 2015;Waterhouse et al. 2018) to estimate the completeness of the transcriptome based on a set of conservative Eukaryotic orthologs and Transrate v.1.0.3 (Smith-Unna et al. 2016) to evaluate the accuracy and completeness of the de novo assembly.

Gene Identification and Expression Levels
To identify candidate-coding regions, we used TransDecoder v.3.0.0 (Haas and Papanicolaou 2016) with default parameters along with BlastP v.2.2.29+ and Pfam v.31.0 (Finn et al. 2016). We used Blastx v.2.2.29+ to search for matches for each transcript against the Swissprot database (The UniProt Consortium 2017) and used an e-value of 1e-20 to identify matches. In this manner, we were able to search the peptides identified by TransDecoder as well as the entire set of transcripts. To determine if transcripts were antimicrobial peptide precursors, we conducted BLAST searches to the Database of Anuran Defense Peptides (DADP; Novković et al. 2012). The DADP is a manually created database of all known antimicrobial amphibian peptides and when identified, contains both the more conserved signal and the more divergent, bioactive domain sequences. We also repeated the BLAST using a less stringent value of 1e-5 to search for more loosely matching AMPs that might exist but could be difficult to detect in M. betsileo due to sequence divergence. Additionally, we used HMMER v. 3.2.1 (Eddy 2009) with an e-value of 1e-5 to search assembled transcripts for matches to the Pfam database (Finn et al. 2016). The Pfam database is a collection of protein families, sourced from UniProtKB sequences, represented by multiple sequence alignments and hidden Markov models (HMMs), which we used as another method to detect the presence of antimicrobial peptide domains in the frog skin transcriptomes. We used RSEM v.1.3.0 (Li and Dewey 2011) to identify gene expression levels and assigned functional annotations using Trinotate v.3.0.2 (Haas et al. 2013) with default parameters. Additionally, we extracted Gene Ontology assignments using GOseq with Trinotate, which categorically identified genes with a standardized molecular, biological, or cellular function. We used R v.3.3.1 (R Core Team 2017) to generate a normalized TPM Expression Histogram of genes based on their relative categorical ranking of high, middle, or low expression value, which we determined by calculating the 25% and 75% quartiles of the entire gene data set and separating the genes into their respective expression ranges.

Comparison of AMP Expression
We downloaded RNA-Seq data from the NCBI Short Read Archive database and selected 12 frog skin datasets that span the anuran phylogeny (Table S1). We used the raw reads for each dataset and ran Sickle, Trinity, Blast, RSEM, and HMMER using the same parameters as the M. betsileo dataset.
To compare expression levels between species, we used two methods of normalization. First, we created a rank order for genes in each transcriptome based on raw counts such that the lowest expressed gene was ranked 1. We then identified the location of AMPs within the rank ordered list and calculated the AMP expression percentile by dividing the ranked order number of each AMP by the total number of genes within the transcriptomes, multiplied by 100 to obtain a percentage. Thus, the highest expressed gene had an expression percentile of 100% because its ranked order number within the list was equal to the total number of genes in the transcriptome. Second, we identified common gene names among all 13 species and created a matrix of these genes and relative raw counts in each frog species. We then used edgeR v.3.6 to create a normalized expression matrix of genes across all species by correcting for library size. We converted the normalized expression counts to percentages and calculated the percentile expression of each gene within the bile acid pathway. Additionally, we generated a phylogeny of all 13 species using the Time Tree of Life (Hedges et al. 2015) to test for a phylogenetic signal of AMP expression (Ives et al. 2007). Two species were not available in the Time Tree of Life database (Atelopus glyphus and Craugastor fitzingeri), so we substituted the two most closely related species (Atelopus zeteki and Craugastor longirostris, respectively).

Comparison of Primary Bile Acid Biosynthesis Pathway
We used the KEGG Automatic Annotation Server (KAAS; Moriya et al. 2007) to annotate genes found in the Primary Bile Acid Pathway in order to identify potential differences between species that may reflect functional changes. We used the identified genes from our transcriptome assembly in conjunction with the normalized percentile expression values to map genes in the primary bile acid pathway proportionalized by expression level using Cytoscape v. 3.6.0 (Shannon et al. 2003).

Anti-microbial Assays
Antimicrobial activity assays of Sm dihydro-MA, ampicillin, squalamine, and MSI-1436 were carried according to procedures in Moore et al. (1993). The minimal inhibitory concentration (MIC) was determined by incubating logarithmic-phase organisms (about 10 6 colony forming units/ml) in 0.5 strength trypticase soy broth at 37°for 18-24 hr in the presence of various concentrations of antimicrobial. The MIC is the concentration of antimicrobial where visible growth was inhibited.

Data Availability
The skin transcriptome assembly and raw sequence data for Mantella betsileo can be found at the Gene Expression Omnibus under accession GSE61355. This Transcriptome Shotgun Assembly project has been deposited at DDBJ/EMBL/GenBank under the accession GGTL00000000. The version described in this paper is the first version, GGTL01000000. Supplemental material available at Figshare: https://doi.org/10.25387/ g3.7543436.

RESULTS
To understand the variation in expression of AMPs and other genes expressed in the skin of frogs, we downloaded publicly available RNA-Seq data for 12 anuran species and generated a new transcriptome from Mantella betsileo. The publicly available data were chosen because they represented a deep sample of anuran phylogenetic history that we could use to answer questions about the evolution and diversification of AMPs and other genes expressed in frog skin compared to Mantella. We chose to produce RNA-Seq data for Mantella betsileo because this species was shown to produce TMA regardless of environmental factors (Clark et al. 2012). It is important to note that the publicly available data were produced under a variety of protocols and sequencing methods, which could produce non-biological variation. However, our questions focused on the presence of AMPs and genetic pathways related to phenotype, which are less likely to be impacted by the technical variation that could arise from extraction, library, and sequencing methods. We attempted to further guard against these potential confounding variables by assembling the skin transcriptome for each species using the same computational pipeline.

Quality of Transcriptome Assembly for Mantella
Of the 23,892,703 total sequences, 77.5% (18,515,394) were retained as paired end reads after trimming with Sickle. The contig N50 value of the assembly was 910 and more than 82% of reads aligned back to the assembled transcriptome reference. To estimate the completeness of our Mantella transcriptome, we used BUSCO to look for 303 genes assumed to be ubiquitous in all eukaryotes. Of the 303 core BUSCO groups searched, 198 (65.3%) were identified as complete and singlecopy BUSCOs. Only 8 BUSCO genes out of the 303 genes searched could not be found within the transcriptome, further validating the high level of completeness of the assembly. Transrate analyses identified the accuracy with which the complete set of transcripts was represented in the read data in the absence of a reference genome. Of the 18 million total fragments that successfully mapped back, 76.6% of reads aligned in the correct orientation on the same contig, which was comparable to other frog transcriptomes we analyzed (average = 71% (6 4.5 SE).

No AMPs observed in Mantella betsileo
We searched for AMPs across 200 million years of frog evolutionary history and tested the hypothesis that there were changes in AMP abundance and expression among lineages. We retrieved a diversity of AMP matches by BLAST, with 74 unique AMPs in total and an average of 8.67 (6 2.96 SE) unique AMPs per lineage. Odorrana had the most unique AMPs (37) and Craugastor, Atelopus, and Fejervarya had the least (1), suggesting that there was substantial variation in the number of unique AMPs found in each lineage (Figure 1; Figure S1). However, despite this extensive variation, there was no phylogenetic signal in either the number of unique (K = 0.52, P = 0.47) or expression rank (K = 0.54, P = 0.338) for these AMPs.
AMP expression level was on average at the 92.0 (6 1.11 SE) rank percentile, revealing that AMP precursors were among the most highly expressed genes in frog skin (Figure 1). The two highest numbers of AMPs were found in Pelophylax and Odorrana, which are relatively closely related; however, there were several AMPs expressed at low levels in Ranid lineages as well as the two species most closely related to Mantella.
In contrast to all other frogs, we did not observe AMP sequences in the skin transcriptome of Mantella betsileo using BLAST against the DADP or Swissprot, or HMMER against the Pfam protein domain database (Figure 1; Figure S1; Table S2). AMPs are not annotated in the DADP or Pfam for Mantella; thus, the lack of annotation could explain the lack of observed AMPs even though they may be present in Mantella. AMP sequences are typically divergent (König et al. 2015) and there is not a single shared amino acid among all AMP sequences for either the more conserved signal sequence or the more rapidly evolving bioactive region ( Figure S2). However, we did observe AMPs in the transcriptomes for six other species that were not known to previously have AMPs (Table 1;  Table S2). For these six, three species (Atelopus glyphus, Craugastor fitzingeri, and Fejervarya limnocharis) had highly expressed transcripts that matched to Buforin-II, a sequence that is derived from Histone 2A, and experimentally demonstrated to have antimicrobial activity (Park et al. 1996;. Furthermore, all observed BLAST hits from the DADP to Buforin-II contained the AMP variant sequence "TRSSRAGLQFPVGRVHRLLRK", while the remaining BLAST hits from the Swissprot Database to Histone 2A did not contain the antimicrobial bioactive sequence referenced above. The other three species (Paa boulengeri, Polypedates megacephalus, Rhacophorus dennysi), had sequences that matched a diversity of AMP classes ( Figure S1).
Second, since AMPs are among the most highly expressed genes in frog skin, we examined the most highly expressed genes across all 13 species. This approach allowed us to determine if any of the unidentified transcripts in Mantella with high levels of expression were related to AMP or defense function. Across all 13 species, we found that the most highly expressed genes matched to keratin and ribosomal proteins. The important distinction was the presence of AMPs in the upper expression threshold for all other species except M. betsileo. An alternative explanation for the lack of AMPs in Mantella is that we may not have had enough depth of sequencing to detect AMPs in the skin transcriptome. However, as demonstrated, AMPs are among the most highly expressed sequences and thus are more easily identified in RNA-Seq data compared to lower expressed genes (Mortazavi et al. 2008;Malone and Oliver 2011;Conlon 2011;Conesa et al. 2016). Taken together, we observed a diversity of AMPs across nearly 200 million years of anuran evolution, but AMPs were not observed in M. betsileo. Our lack of observations of AMP sequences in Mantella suggests either that Mantella does not have AMPs, or if present, Mantella have AMPs that are so divergent that our methods cannot detect them.

Primary Bile Acid Biosynthesis Pathway Functionality in Frog Skin
The lack of AMPs observed in Mantella is notable given that AMPs contribute to immunity in amphibians, are considered prominent components of most frog skin secretions, and are evolutionarily widespread (Conlon 2011;König et al. 2015; Figure 1; Figure S1). Recently, Mantella and other poison frogs were discovered to secrete a novel metabolite named tauromantellic acid. TMA has structural similarity to a bile acid, a metabolite synthesized by the liver and employed to function in the emulsification of fats and oils. Given that tauromantellic acid is endogenous and likely derived from cholesterol, we searched for the presence of genes active in the bile acid synthesis pathway and examined expression variation among species by normalizing expression data for these genes across all 13 frog skin transcriptomes.
We found evidence of expression for an average of 70.4% (6 3.64 SE) of the 18 total pathway genes across the other 12 frog species; however, M. betsileo was the only species to have high expression for CYP7A1, a gene critical to the function of the bile acid pathway (Figure 2). CYP7A1 had a rank percentile expression of 96% in M. betsileo vs. 4% in Odorrana margaretae, the only other species to have any evidence of expression for this gene. CYP7A1 catalyzes the first reaction in the cholesterol catabolic pathway in the liver and is the rate-limiting step in regulation of bile acid synthesis (Hubacek and Bobkova 2006). The presence of this gene determines the functionality of the primary bile acid biosynthesis pathway and thus suggests activity in the skin of M. betsileo.
We observed higher expression in Mantella for two other genes normally expressed in liver and involved in bile acid synthesis. BAAT was highly expressed exclusively in M. betsileo. The average BAAT expression percentile in all other 12 species was 5.82% (6 3.60 SE) as compared with the 30.2% expression percentile observed in Mantella. Additionally, another gene (AMACR) had an average expression percentile of 5.89% (6 1.57 SE) across all other 12 species as compared to an expression percentile of 29.4% found in Mantella. This gene encodes a racemase essential for bile acid synthesis and conversion of pristanoyl-CoA and C27-bile acyl-CoAs to their (S)-stereoisomers for degradation of methyl branched fatty acids (Savolainen et al. 2004).

Anti-microbial Assays Reveal Anti-microbial Activity when MA is coupled to Spermine
The lack of observed AMPs, the discovery of tauromantellic acid (Clark et al. 2012), and the functionality of the primary bile acid biosynthesis pathway in the skin of Mantella betsileo suggests the possibility of a distinct immunological defense system that together may constitute a broader spectrum of antimicrobial compounds. Such an immune system would require conjugating polyamines to TMA to create structurally similar molecules to squalamine, or MSI-1436 (Figure 3). We synthesized a broadspectrum antimicrobial aminosterol (Sm dihydro-MA) by the coupling of spermine to a molecule of mantellic acid lacking the side chain unsaturation (Figure 3) and compared the activity of this molecule to other, similarly structured molecules. There was anti-microbial activity for Sm dihydro-MA, though it was not as potent as other molecules against Staphylococcus and Eschericia. However, Sm dihydro-MA was more potent against Pseudomonas bacteria and Candida fungus than ampicillin. Collectively, MSI-1436 was the most potent antibiotic against all organisms tested (Table 2).
A candidate 3-keto reductase enzyme necessary to couple a polyamine to TMA for potent antibiotic activity is expressed within our M. betsileo transcriptome, suggesting the plausible possibility of this defense system within the skin.

DISCUSSION
The presence of gene-encoded antimicrobial peptides in a variety of amphibian species has fueled the perception that most frogs contain an AMP defense arsenal in their skin. These peptides are thought to be essential for protection against infectious microorganisms (but see König et al. 2015 for an alternative hypothesis) and are found at high expression percentiles, yet in our study we did not observe AMPs in the skin of Mantella. No AMPs have been found in Mantella previously, and there are several cases where frog lineages were examined for AMPs, yet they were not found (König et al. 2015). Our computational methods are based on sequence similarity and rely on available sequences and it remains possible that an entirely new class of highly divergent AMPs could be present in Mantella that we cannot find using computational methods. Ultimately, more careful studies of skin extracts from Mantella, combined with genome sequencing could help resolve whether Mantella have either lost or have a highly divergent class of AMPs. Nevertheless, we think they probably do not have AMPs as suggested by our studies of skin n Table 1 AMP hits to number of annotations in the Database of Anuran Defense Peptides (DADP). The underlined species are those not present in the DADP, but with hits to AMPs. The right-most column indicates the number of AMPs listed in the DADP for each species and genus. We obtained hits for AMPs in every other species not found in the database except for Mantella extracts from Dendrobates, a group of poison frogs from South America that contain alkaloid compounds and TMA. Using the same successful procedures that worked to find AMPs in other frog species (Zasloff 1987;Clark et al. 1994), we could not find AMPs in Dendrobates, yet extracts from their skin had antimicrobial activity. Collectively, this raises questions about the function and mechanistic origin of amphibian defenses, especially in lineages where AMPs may not occur. Concomitant with the lack of observed AMPs in Mantella and Dendrobates is the presence of tauromantellic acid (TMA), which was recently discovered in poison frogs (Clark et al. 2012). The similarity of the ring structure of tauromantellic acid to that of the broad spectrum antimicrobial, squalamine, and the preliminary results of antimicrobial activity for mantellic acid and similar bile acids coupled to polyamines (Table 2) suggests a diversity of amphibian defenses. Tauromantellic acid is most likely synthesized from cholesterol within the skin of Mantella because the entire primary bile acid biosynthesis pathway is present, including the gene involved in the rate-limiting step for synthesis. The bile acid pathway members found in other frog species are most likely involved in other metabolic processes, which may account for the presence of genes in the pathway but lack of pathway functionality. Conversely, M. betsileo was the only species to abundantly express genes critical for functionality of the pathway including CYP7A1, BAAT, and AMACR.
Bile acids are detergents that play a fundamental role in both lipid and cholesterol processing mainly in the gastrointestinal and hepatobiliary systems (Smith et al. 2009). These chemicals are required for dietary lipid and fat-soluble vitamin absorption and maintenance of the balance between cholesterol synthesis and excretion. In general, bile acids are made primarily in the liver through the oxidation of cholesterol and their localization is tightly regulated due to their membrane and epithelial toxicity (Smith et al. 2009). Consequently, the fact that the bile acid pathway, which is predominantly found in the liver, appears expressed in the skin of Mantella is highly unusual. In humans, the presence of bile acid in the skin is associated with disease, due to high concentrations of hepatic bile acids in the blood stream.
Conversely, while AMPs are in the skin of most frogs, humans have liver-expressed antimicrobial peptides (LEAPs) that are believed to have a protective role against bacterial infection (Henriques et al. 2010). Thus, finding a liver associated pathway in the skin coupled with knowledge of AMP enrichment in most frogs, may suggest a deeper evolutionary connection between AMPs in the skin, genetic pathways in the liver, and potentially bile acid derived compounds used for defense that future research may uncover.
While we find moderate anti-microbial activity for MA when coupled to a spermine and the main reductase enzymes involved in conjugating spermine to MA are expressed in Mantella, recent Figure 2 Comparison of genes involved in the primary bile acid biosynthesis pathway in Mantella compared to Bombina. Gene pathway members were identified in each transcriptome using the KEGG Automatic Annotation Server (KAAS). The raw expression counts for each gene identified on the pathway were obtained from RSEM and used to generate a normalized expression matrix across all species. Pictured is a painted network of percentile gene expression values for a representative species (B. maxima) in comparison to M. betsileo generated using Cytoscape v. 3.6.0. The key difference is the large expression value for CYP7A1, a gene critical for the rate limiting step of bile acid production, in Mantella compared to all other frog species as represented by B. maxima. BAAT (a gene necessary for bile secretion) and AMACR (a gene necessary for the degradation of methyl-branched fatty acids) are also more highly expressed in Mantella compared to all other species. Together, the presence and abundance in expression of CYP7A1, BAAT, and AMACR may suggest functionality of the bile acid pathway in the skin secretions of M. betsileo. discoveries suggest that bile acids may have diverse functions. Many species of fish excrete intestinal bile acids into the surrounding waters as pheromones (Buchinger et al. 2014), which are detected by the olfactory system and influence mating and migratory behavior. Several of the lamprey pheromones include molecules that share striking structural similarity to tauromantellic acid (Buchinger et al. 2015). Therefore, tauromantellic acid may play an even larger role as a pheromone or an olfactory signal.
The presence of diet-derived alkaloids, functionality of the bile acid pathway, and the lack of observed AMPs in M. betsileo have expanded our understanding of the versatility in potential amphibian skin defense mechanisms. Yet, the intrinsic complexity of amphibian genomes and limitations on genomic resources restrict sequencing for whole-genome assembly projects that would aid in a better understanding of the evolution of genes involved in frog skin secretions. Further analyses of neotropical poison frogs in the genera Epipedobates and Dendrobates, which also secrete tauromantellic acid, could provide new understanding of the relationship between the bile acid pathway and the complex chemical community involved in defense within anuran skin (Rogers et al. 2018). We predict these genera lack AMPs and have expression of the bile acid pathway in the skin. Thus, a study of poison frogs in general would serve as an intriguing direction for future research. For Mantella and other poison frogs, if tauromantellic acid when coupled to a polyamine provides antimicrobial functions to amphibian skin, is it interacting with alkaloids, providing antimicrobial function, or a combination of both (Raaymakers et al. 2017)? Additionally, are AMPs present in taxa that have evolved metabolites like tauromantellic acid, which contribute to defense, or could there be another defensive mechanism?
The synthesis of tauromantellic acid would also aid in further testing of this molecule for antimicrobial activity. In addition, a thorough chemical analysis of the skin secretions from these species of poison frogs will be needed to test the hypothesis that antimicrobial bile acids are indeed produced in their skin. Finally, generalizing the presence of 3-keto bile acids found in Mantella may indicate the presence of this proposed cholesterol-based immune system in other organisms, including the neonatal human (Wahlén et al. 1989). The novel diversity in amphibian skin defense mechanisms described in these results offers new directions for genomic research and possible applications of these immunological compounds to human health. Figure 3 The structure of TMA, Sm dihydro-MA, squalamine, and MSI-1436 synthesized as described in Tessema et al. (2013). TMA was not synthesized for subsequent antimicrobial assays. n

ACKNOWLEDGMENTS
We thank the UConn Bioinformatics Core and NSF XSEDE research allocation TG-DMS140018 and TG-MCB141026 awarded to JHM for providing the computational resources needed for this project.