Evolutionary insights into umami, sweet, and bitter taste receptors in amphibians

Abstract Umami and sweet sensations provide animals with important dietary information for detecting and consuming nutrients, whereas bitter sensation helps animals avoid potentially toxic or harmful substances. Enormous progress has been made toward animal sweet/umami taste receptor (Tas1r) and bitter taste receptor (Tas2r). However, information about amphibians is mainly scarce. This study attempted to delineate the repertoire of Tas1r/Tas2r genes by searching for currently available genome sequences in 14 amphibian species. This study identified 16 Tas1r1, 9 Tas1r2, and 9 Tas1r3 genes to be intact and another 17 Tas1r genes to be pseudogenes or absent in the 14 amphibians. According to the functional prediction of Tas1r genes, two species have lost sweet sensation and seven species have lost both umami and sweet sensations. Anurans possessed a large number of intact Tas2rs, ranging from 39 to 178. In contrast, caecilians possessed a contractive bitter taste repertoire, ranging from 4 to 19. Phylogenetic and reconciling analysis revealed that the repertoire of amphibian Tas1rs and Tas2rs was shaped by massive gene duplications and losses. No correlation was found between feeding preferences and the evolution of Tas1rs in amphibians. However, the expansion of Tas2rs may help amphibians adapt to both aquatic and terrestrial habitats. Bitter detection may have played an important role in the evolutionary adaptation of vertebrates in the transition from water to land.

Tas1r3 is co-expressed with either Tas1r1 or Tas1r2, which form into heterodimers: a Tas1r1-Tas1r3 heterodimer functions as an umami taste receptor, whereas a Tas1r2-Tas1r3 heterodimer senses sweet compound (Li et al., 2002;Nelson et al., 2001). Although most vertebrates have three complete Tas1r genes, the repertoires of Tas1rs in other species may differ in both sequence and numbers. Three main processes have shaped the different numbers of Tas1rs in different species: gene loss, pseudogenization, and duplication (Liu et al., 2014). For example, Tas1r1 is a pseudogene in the giant panda (Ailuropoda melanoleuca; Zhao, Yang, et al., 2010) and six pinniped species (Sato & Wolsan, 2012). Tas1r1 is absent, unamplified, or pseudogenized in 31 species of bats examined .
Tas1r2 is lost in the genome of the zebra finch (Taeniopygia guttata) and chicken (Gallus gallus) and pseudogenized in some carnivore species (Jiang et al., 2012;Li et al., 2005Li et al., , 2009Zhao, Zhou, et al., 2010). In reptiles, Tas1r genes possibly have been lost or pseudogenized in snakes; in testudines and crocodilians, Tas1r genes are either intact or partial (Feng & Liang, 2018). These mutations are thought to result in the disfunction of Tas1r and then affect sweet/umami recognition in animals. Nevertheless, a lineage-specific increase in the number of Tas1rs has been described in fish. In 15 fish species examined, the number of Tas1r2 genes differs widely, ranging from 1 to 4 (Dong et al., 2018). To date, several studies have investigated the ecological factors that drive the evolution history of Tas1rs in vertebrates. The evolution of Tas1r genes is sometimes explained by feeding ecology (Zhao, Zhou, et al., 2010), sometimes inconsistent with dietary differences (Feng & Liang, 2018;Feng & Zhao, 2013;. Bitter substances are recognized by taste 2 receptors encoded by the Tas2r gene family (Adler et al., 2000). Tas2r genes possess a GPCR-like domain and a short extracellular N-terminus. They are ~900 bp and lack introns. Several studies have indicated that the repertoire of Tas2rs showed a very dynamic evolution among species. For instance, the number of intact (functional) Tas2rs is subject to intense variation: 0 in cetaceans and penguins (Feng et al., 2014;Zhu et al., 2014), 25 in humans (Go et al., 2005;Shi et al., 2003), 3-50 in lizards (Zhong et al., 2017(Zhong et al., , 2019, 1-5 in teleost (Dong et al., 2018), and 80 in the coelacanth (Latimeria chalumnae; Syed & Korsching, 2014). Not surprisingly, there are varying numbers of pseudogenes among species. Pseudogenization has occurred in almost all Tas2rs in cetaceans (Feng et al., 2014;Zhu et al., 2014).
Numerous studies have attempted to explain the evolutionary and ecological significance of Tas2rs. It is generally accepted that the taste receptor gene family has undergone a complex evolutionary process. They are susceptible to gene duplication, gene deletion, pseudogenization, positive selection, and other factors, resulting in the expansion or contraction of specific gene families among different evolution branches (Dong et al., 2009;Go, 2006;Hayakawa et al., 2014;Li & Zhang, 2014;Shi et al., 2003;Wang & Zhao, 2015). This complexity is assumed to reflect the evolutionary needs of the respective species. For instance, the relationship between the Tas2r numbers of vertebrates and their corresponding dietary habits has been addressed. Actually, some studies have uncovered the probable correlation between diets and Tas2r numbers: herbivores and insectivores who encounter bitter substances more frequently possess more Tas2rs than carnivores (Hu & Shi, 2013;Jiang et al., 2012;Liu et al., 2016;Wang & Zhao, 2015;Zhong et al., 2017). These findings have suggested that the dietary toxin content is one of the primary selective forces for the differences in Tas2rs repertoires among species.
Other than dietary habits, foraging patterns may also affect the repertoires of Tas2r genes. Previous studies have indicated that cetaceans (0-1; Zhu et al., 2014;Feng et al., 2014), snakes (1-2;Zhong et al., 2017), and penguins (0; Zhao et al., 2015) possess a dramatic contraction in the number of functional Tas2r genes. Their behavior of swallowing food whole without mastication reduces the contact of the TRCs with bitter stimuli, resulting in less contact with poisonous foods.
So far, studies on the function and adaptive evolution of Tas1r/Tas2r genes have been carried out extensively in many species.
However, data are quite limited in Tas1r/Tas2r families of amphibians, except for the western clawed frog (Behrens et al., 2014;Go, 2006;Shi & Zhang, 2006). Amphibian is the transition lineage from aquatic lifestyle to a terrestrial one in the history of vertebrate evolution and plays an important role in animal evolution. To gain extensive, systematic, and efficient evolution research on Tas1r/Tas2r, it was considered worthwhile to investigate more data and adaptive evolution of amphibian species. Therefore, this study aimed to examine the repertoires of Tas1r/Tas2r genes in amphibians and predict their functionality using available genome assemblies. This study recovered the phylogenetic relationship and determined the duplication and loss events of Tas2rs to understand their birth-and-death process in this evolutionary group. Furthermore, the selective pressure of Tas1rs in amphibians was estimated.

| Taxonomic sampling of genome data
Class Amphibia is generally classified into three orders: Anura (anurans), Caudata (urodeles), and Gymnophiona (caecilians). This study focused on currently available genomes of 14 species from the National Center for Biotechnology Information (NCBI) databases. of the genomes ranged from 2.9 kb to 20.7 Mb, implying high-quality assemblies.

| Gene annotation
As Tas2r genes contain no introns, and most have a similar gene length of ~900 bp, gene annotation was performed by sequence alignment with TBlastN (Altschul et al., 1990), which was implanted in TBtools (Chen et al., 2020). First, previously known Tas2r protein sequences were retrieved from the GenBank or literature and used as initial queries (25 from human, 35 from mouse, 3 from chicken, and 80 from coelacanth). Second, the queries were used to blast against a genome assembly by TBlastN (Altschul et al., 1990), with an e-value of 1 × 10 −10 . BLAST hits of <100 bp were discarded, and the overlapping hits were merged. The remaining records were prolonged for 500 bp in both 5′ and 3′ directions, which were regarded as the genomic locations of the homologous genes. Third, genomic nucleotide sequences were extracted as candidate Tas2r genes. The outputs were divided into three categories: intact genes, partial genes, and pseudogenes according to a previous study (Li & Zhang, 2014). Intact genes refer to sequences with >270 amino acids with both start and stop codons. Partial genes refer to sequences that lack either a start or a stop codon. They may be complete genes, but their open reading frames (ORFs) are truncated due to incomplete genome sequencing or assembling. The homology of partial genes in their corresponding genome was analyzed through alignment. If multiple fragment sequences of the same species can be aligned with overlapping regions, they are considered to be from different gene loci. If there are no overlapping regions during alignment, they are considered to be from the same gene site, which may be caused by sequence spacing due to sequencing or assembly. Sequences with premature stop codons and/or ORF-disrupting mutations were regarded as pseudogenes.
As Tas1r genes contain introns, a more complex bioinformatic pipeline was employed. First, previously known Tas1r1, Tas1r2, and Tas1r3 protein sequences were used as queries to identify the genomic locations of homologous genes in a genome. Second, genomic DNA sequences were extracted and used to perform pairwise alignments with query protein sequences by Genewise (Madeira et al., 2019), which provided the exon/intron structures and frameshifting errors. When receiving negative BLAST results, synteny analysis was performed to examine Tas1r genes with closely related species as the reference. If neighboring genes flanking Tas1r genes could be found, Tas1r genes were regarded as absent.
The obtained protein sequences of each Tas1r/Tas2r were verified by the TMHMM method (Krogh et al., 2001) for the presence of seven transmembrane domains. A Tas1r/Tas2r sequence was regarded as a pseudogene if it did not have seven transmembrane domains. In addition, annotated sequences were examined by reciprocal SmartBLAST (https://blast.ncbi.nlm.nih.gov/blast/ smart blast/) as well as phylogenetic analyses to ensure that the best hits are known Tas1r/Tas2r genes. The gene nomenclature was named with a four-letter prefix corresponding to the species names as well as a numerical suffix consecutively. For example, the Tas2r1 gene of L. leishanense is referred to as Lele_Tas2r1.

| Phylogeny of taste receptor genes in amphibians
To explore the evolutionary relationship among Tas1r/Tas2r genes in amphibians, a phylogenetic analysis of intact genes was performed.
Partial genes or pseudogenes were not included in the phylogenetic analysis due to a large number of gap sites after alignments. Multiple sequence alignments of amino acid sequences were performed by MAFFT with the L-INS-I strategy (Katoh et al., 2002). GBLOCKS (Castresana, 2000) was then used to optimize the quality of alignment results. The selected conserved region was used in the following analysis. The phylogenetic relationship of the genes was inferred by the maximum likelihood (ML; Dempster, 1977)  The branch support analysis was evaluated with 1000 ultrafast bootstraps (Minh et al., 2013). The tree was rooted with vertebrate V1R/V2R vomeronasal receptor. The procedures processed in ML phylogenies were all implemented in PhyloSuite .

| Reconstruction of Tas2r repertoire evolution
Large-scale gene births and deaths are the major forces of functional genetic innovation. To infer the history of births (duplication) and deaths (deletion) of Tas2r genes across the amphibian phylogeny, a reconciliation analysis was performed with NOTUNG 2.6 (Chen et al., 2000). This method estimates the history of gene duplication and deletion times by comparing gene tree with species tree. The species tree was estimated from the TimeTree database, which provided the generally accepted phylogenetic tree (Hedges et al., 2006).
All birth-and-death events of Tas2rs were placed in each branch of the species tree to show the evolutionary trajectories of Tas2r repertoires in Amphibia.

| Adaptive evolution analysis of Tas1r genes
The selective pressure of Tas1r genes was tested by two steps. First, the number of nonsynonymous substitutions per nonsynonymous site (dN) and the number of synonymous substitutions per synonymous site (dS) were used to compute overall ω (dN/dS) values. ω = 1, ω < 1, and ω > 1 represent neutral, purifying, and positive selection, respectively. The mean ω for each Tas1r was calculated by the CodeML method (Yang, 2007) with EasyCodeML (Gao et al., 2019). The generally accepted phylogenetic tree was inferred from TimeTree (Hedges et al., 2006). Moreover, positive selection could act on individual amino acid residue. Therefore, in the second step, codon-based analyses with CodeML and FUBAR (Murrell et al., 2013) were performed to detect potential positive selection sites.
In CodeML, the models between M7 (purifying selection) with M8 (positive selection) with EasyCodeML were compared. A likelihood ratio test was used to estimate whether there was a significant difference between the models. FUBAR analysis on the Datamonkey server (http://class ic.datam onkey.org/; Pond et al., 2005) was used to find evidence of episodic positive/diversifying selection with a posterior probability of 0.9.

| Tas1r and Tas2r repertoires
This study annotated 16 Tas1r1, 9 Tas1r2, and 9 Tas1r3 intact sequences that appear to be functional genes (Table 1). One truncated Tas1r1 and Tas1r2 in common frog and one truncated Tas1r3 in Tibetan Plateau frog were also identified as pseudogenes (see Appendix 2 for the genomic location). We failed to identify the Tas1r1 gene from genome assemblies of the western clawed frog and the African clawed frog. Thus, synteny analysis was performed to examine whether Tas1r is lost or not. Tas1r1 is flanked by NOL9 and ZBTB48. This linearity is conserved across human, mouse, and twolined caecilian. The presence of NOL9 and ZBTB48 next to Tas1r1 was confirmed, providing evidence of whole Tas1r1 deletion in the two taxa (Appendix 3). Genes that flank Tas1r2 in mice and most species surveyed (MIB2, GOLIM4) were located on the same contig in American bullfrog, Tibetan Plateau frog, African clawed frog, western clawed frog, and Gaboon caecilian. In the genome of the twolined caecilian, Tas1r3 is flanked by DVL1 and CPTP. DVL1 and CPTP were adjacent to each other on the same contig in Leishan spiny toad, African bullfrog, American bullfrog, strawberry poison frog, African clawed frog, and western clawed frog (Appendix 3). Thus, it was speculated that perhaps the Tas1r2/Tas1r3 gene is lost in the respective species. Hence, the absence could lead to the inactivation of both umami and sweet taste functions in Leishan spiny toad, African bullfrog, American bullfrog, strawberry poison frog, Asiatic toad, western clawed frog, and African clawed frog and the loss of sweet taste function in Tibetan Plateau frog and tiny cayenne caecilian (Table 1). In addition, multiple copies of Tas1r1 genes in five amphibian genomes were found. Nevertheless, duplication of Tas1r2 only occurred in the common frog, and duplication of Tas1r3 only occurred in common frog, Tibetan Plateau frog, and tiny cayenne caecilian.
This study annotated 1400 Tas2r genes in amphibian genome assemblies, including 1156 intact, 30 partial, and 214 pseudo Tas2rs  To detect whether tandem duplication happens in amphibians, the genomic location of Tas2r genes was examined. Indeed, some Tas2rs were organized into clusters on specific chromosomes or scaffolds (Appendix 4). For example, Tas2r

| Phylogeny of Tas1r and Tas2r genes
To delineate the evolutionary history and relationships among amphibian Tas1r/Tas2r genes, phylogenetic analysis by the ML method based on all intact genes was performed. Figure 2 shows the phylogenetic relationship of intact Tas1r genes. Most branches in the tree showed high bootstrap support, indicating the reliability of phylogenetic relationships among Tas1rs. Clusters of Tas1r1, Tas1r2, and Tas1r3 could separate from each other. Each gene cluster formed into two clades, including anurans and caecilians.
Based on the phylogenetic tree, intact Tas2r genes were categorized into five large and eight small clades ( Figure 3). Some lineages showed a cluster of Tas2r genes from the same species (marked with one color), suggesting that these lineages are enriched with speciesspecific gene duplications. In contrast, other lineages showed genes from distantly related species.

| Lineage-specific gene births and deaths of Tas2rs
The phylogeny of Tas2r genes implies that extensive gene expan- were observed in the branch of Gymnophiona. Because the increase occurred in Anura (43 intact Tas2rs) compared to Gymnophiona (15 intact Tas2rs), data suggested that the reduction of Tas2rs may have occurred before the divergence between Anura and Gymnophiona.
As shown in the evolutionary trajectory tree, extensive speciesspecific gene duplications may be responsible for the considerably larger Tas2r repertoires in some anurans, for instance, 92 gains in Leishan spiny toad and 81 gains in the Asiatic toad (Figure 4).

| Purifying selection in the Tas1r gene family
To understand the selective pressure of the Tas1r gene family, the ratio (dN/dS) of nonsynonymous mutation rate (dN) to synonymous mutation rate (dS) was calculated. All Tas1r genes were under strong purifying selection (0.169-0.210; Table 2). This result indicated that Tas1rs were evolving-constrained, and the function was conserved in the species, which remained as the relevant Tas1r genes.
A small number of positively selected sites were found in each gene interspecies.

| D ISCUSS I ON
Studies on umami, sweet, and bitter taste receptors have made enormous progress in recent years. Despite the special evolutionary status of amphibians, their taste receptor families have been rarely described except for the western clawed frog (Behrens et al., 2014;Go, 2006;Shi & Zhang, 2006). In this study, the repertoire of Tas1r and Tas2r genes from a wide collection of amphibian species was presented for the first time. Unlike the conservation of only one copy of Tas1r1/Tas1r2/Tas1r3 in numerous vertebrates (Shi & Zhang, 2006), two copies of the Tas1r gene in several amphibian species were found. Similarly, duplication events of the Tas1r gene (n = 1-4) were also reported in several teleost (Dong et al., 2018).
The result may support the fact that amphibians diverge from the ancestral fish-tetrapod stock during the evolution of animals from strictly aquatic forms to terrestrial types. This study confirmed previous findings and showed that all three Tas1rs were absent in the western clawed frog (Shi & Zhang, 2006). Note that the annotations of the Tas1r gene in the genome database are sometimes incorrect.
For example, V2R genes of the African clawed frog and the western clawed frog were annotated as Tas1r1 in the NCBI probably due to the sequential similarity of the two GPCR families. To ensure our prediction accuracy, annotated genes were verified in the genome database with reciprocal BLAST, synteny analysis, and phylogenetic analyses.
Sweet and umami tastes help animals to recognize dietary information for nutritious carbohydrates and proteins, respectively, and thus are pivotal for the survival of animals. However, given the similarity of feeding preferences in amphibians and their distinct phylogenetic positions, this study failed to discover a correlation between feeding ecology and Tas1r evolution. Most of the dietary preference of anurans are similar to each other, but their Tas1r genes can be intact, pseudogenized, or absent, suggesting that no correlation exists between Tas1r functionality and feeding ecology. It seemed that loss of umami/sweet tastes could occur in any species. This study found that Tas1r1 genes of the African clawed frog are absent from its genome assembly. Surprisingly, robust glossopharyngeal nerve responses have been recorded in amphibians when various amino acids are applied to taste organs on the tongue (Feder & Burggren, 1992;Gordon & Caprio, 1985;McPheeters & Roper, 1985;Yoshii et al., 1982). The African clawed frog has been reported to have high gustatory sensitivity to amino acid, such as arginine (0.1-1.0 μM; Yoshii et al., 1982). The above contradiction between the absence of Tas1r1 genes and amino acid sensitivity could be explained by the following evidence. Although

F I G U R E 3 Evolutionary relationships of 1156 intact
Tas2r genes in amphibians. The phylogenetic tree was constructed using the ML method. The vomeronasal 1 receptor 3 gene (V1R3; NCBI accession no. AB670529) of East African cichlids (Lithochromis xanthopteryx) was used to root the tree because V1R genes are relatively close to Tas2r genes among GPCRs. Genes from different species are indicated by different colors of branches Tas1r1+3 functions as the main umami receptor in mammals, non-Tas1r1 genes responsible for detecting amino acids likely exist. For example, an odorant receptor, preferentially tuned to recognize basic amino acids, was identified in goldfish (Speca et al., 1999).
The odorant receptor shares sequence similarities with calcium sensing, metabotropic glutamate, and V2R vomeronasal receptors (Speca et al., 1999). To the authors' knowledge, the western clawed frog has the largest V2R repertoire in 14 species investigated (Urszula Brykczynska et al., 2013;Chen et al., 2019;Shi & Zhang, 2007). It was supposed that a mass of V2Rs may be involved in detecting amino acids. Analysis of gustatory nerve responses in metabotropic glutamate receptor 4 (mGluR4) knockout mice provided functional evidence for the involvement of mGluR4 in umami taste responses (Yasumatsu et al., 2014). As for sweet taste, research on gustatory transduction in taste cells demonstrated that the ability to detect sweet substances is present in frogs (Kusano & Sato, 1958;Toshihide et al., 1995). The cAMP or cGMP cascade may be involved in the transduction of sweet stimuli in bullfrog TRCs (Kolesnikov & Margolskee, 1995). It would be interesting to examine whether there are other transduction mechanisms involved in the umami/sweet sensation of amphibians.
This study reported the largest Tas2rs family for any species so far. Meanwhile, the number of Tas2rs genes (especially intact genes) varies greatly among different amphibian species. Although the African clawed frog, derived from the diploid species western clawed frog, has undergone a whole-genome duplication (WGD) event to be a tetraploid species, it possesses a moderate size of Tas1r/Tas2r repertoire compared to other amphibians. This finding suggests that WGD may not have played a major role in the evolution of the amphibian Tas1r/Tas2r repertoire.
A large repertoire of Tas2rs in amphibians likely reflects their adaptation to variable lifestyles and environments. Most frogs and

F I G U R E 4 Evolutionary trajectories of amphibian Tas2r gene repertoires. The numbers in circles and boxes denote the number of intact
Tas2rs. The numbers on branches denote gene increases (+; caused by gene duplication) and decreases (−; caused by gene deletion). For example, Leishan spiny toad gained 92 Tas2rs and lost 11 Tas2rs after branching off from its common ancestor with the Mexican spadefoot toad. The phylogenetic relationships and divergence times of these species were referred to TimeTree (Hedges et al., 2006)

TA B L E 2 Selection Analysis of Amphibian Tas1rs
toads inhabit both aquatic and terrestrial habitats, which necessarily contain a larger variety of toxic substances. Accordingly, their ecological needs should encompass vastly different requirements for their taste system. For instance, aquatic factors, such as pH were proven to have influenced the divergence of taste receptor genes (Caprio et al., 2014;Lin et al., 2004). A great number of Tas2rs could also fulfill their dietary needs. Although some larger amphibian species eat vertebrates, most frogs feed on worms, insects, and other small arthropods that contain more potentially toxic substances.
As anurans are primarily visual, opportunistic predators (sit-and-wait foraging), their selection of prey is limited by the gape of the predator. As a result, the biological importance of bitter taste likely resides in the ability to detect and reject unpalatable, potentially dangerous prey once it is captured, rather than detecting prey (Barlow, 1998).
Only after visually selected prey have reached the mouth, does the taste system function as a toxin detector such that unpalatable and potentially poisonous food is spat out. Hence, we speculate that the lifestyle, ecological, and dietary complexity of amphibians elevate their evolutionary pressure for a wide variety of Tas2r genes. The large number of Tas2rs is consistent with anatomical evidence showing more taste buds and a larger number of taste receptors in amphibians than other vertebrates (Kinnamon & Cummings, 1992;Kinnamon & Margolskee, 1996;Lindemann, 1996). Terrestrial and aquatic anurans have enlarged, specialized organized taste disks that often are found atop large epithelial papillae (Barlow, 1998;Reutter & Witt, 1993), perhaps contributing to specific adaptations for tasting in air and water. Moreover, large numbers of Tas2rs  Caecilians are highly adapted for a burrowing existence. They primarily dwell in highly organic, friable surface layers of the soil, where they maintain tunnel systems. As far as is known, all caecilians are carnivores. Free-ranging diet includes earthworms, platyhelminths, arthropods, frog eggs, tadpoles, and anoline lizards (Bogert, 1970;Daniel, 1998;Wake, 1994). Because animal tissues contain fewer toxic chemicals, it implies reduced importance of bitter taste in caecilians compared with anurans.
In general, this study characterized Tas1r/Tas2r genes and investigated their evolution. It will not only provide abundant raw data but also further recover the evolution dynamics of Tas1r/Tas2r genes.
In particular, studying these genes in the large-scale evolutionary unit can reflect the evolutionary process more comprehensively and systematically. It will also help provide accurate data support for research on function and feeding behavior.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no competing interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The genomic locations or GenBank accessions of Tas2r genes are available at Zenodo (https://doi.org/10.5281/zenodo.5642517).

R E FE R E N C E S A PPE N D I X 2
The GenBank accession numbers or genomic locations of the Tas1r genes in this study

A PPE N D I X 3
Scaffold and gene location of the flanking genes. The genome of two-lined caecilian was used as a reference, in which the neighboring genes of