Introduction

Understanding the evolution of organ reduction, atrophy or indeed complete loss is a fascinating quest, dating back to the seminal work of Charles Darwin, On the Origin of Species (Darwin 1859). Yet, to identify a structure as vestigial—described as a trait with no function, operating sub-optimally, or even with a modified function from that originally served—is no easy undertaking (Werth 2014; Allmon and Ross 2018), often yielding contradictory anatomical descriptions (e.g., Jacob et al. 2000; Nweeia et al. 2012). The increasing availability of whole-genome sequences, on the other hand, provides novel tools to untangle genomic signatures impacting organ reduction or loss (e.g., Zoonomia Consortium 2020). A key question is, thus, to understand how genomic changes impact these processes. Among such signatures we find, more commonly than initially anticipated, gene loss episodes: such as in the morphological simplification of the urochordate Oikopleura dioica, the eye regression observed in cave-dwelling populations of the teleost Astyanax mexicanus, the loss of gastric glands in disparate vertebrate species, dietary switch in Cetacea, or the loss of sebaceous glands in some mammalian lineages (e.g., Olson 1999; Castro et al. 2014; Albalat and Cañestro 2016; Cronk 2009, Guijarro-Clarke et al. 2020; Li et al., 2020; McGaugh et al. 2014; Emerling et al. 2017; Springer and Gatesy 2018; Lopes-Marques et al. 2019a; Themudo et al. 2020; Springer et al. 2021).

A remarkable example of inconsistent observations in the functionality versus vestigiality of an organ can be found in the anatomical observations of the pineal gland in mammals (Ralph 1975). The pineal gland is a small endocrine organ present in the brain and plays a central role in the development of entrainment behaviors through the action of melatonin (circadian rhythmicity). From a physiological standpoint, melatonin synthesis occurs in a specialized cell type, the pinealocyte, through an enzymatic cascade involving the arylalkylamine N-acetyltransferase (Aanat) and N-acetylserotonin methyltransferase (Asmt) enzymes (Klein et al. 1997; Simonneaux and Ribelayga 2003); subsequent signaling uses a set of high-affinity receptors, Mtnr1A and Mtnr1B, involved in the response of the clock machinery to melatonin stimulation, leading to local and overt phase shifts (Fig. 1a) (Axelrod et al. 1964; Lewy et al. 1980; Reppart et al. 1996). Although anatomical studies clearly support a well-defined pineal gland in most mammals, in lineages such as cetaceans, mole rats, and sirenians, a true pineal gland seems to be absent, yet some equivocal observations exist, ranging from complete absence to detectable presence of this gland in some species or in individuals within a species (Ralph 1975; Ralph et al. 1985; Kim et al. 2011; Panin et al. 2012). Conflicting evidence reporting measurable levels of circulating melatonin (i.e., bottlenose dolphin) shed further doubt (Panin et al. 2012). Interestingly, gene loss signatures were identified in these lineages, supporting the loss of function of melatonin synthesis, a hallmark of pineal function, and/or signaling (Fang et al. 2014; Huelsmann et al. 2019; Lopes-Marques et al. 2019b), further demonstrating the power of genome analysis toward the clarification of organ function. The presence of a functional pineal gland is also contentious in xenarthrans (armadillos, anteaters, and sloths), a relatively understudied taxonomic group characterized by its intriguing nature (Fig. 1b; Oksche 1965; Benítez et al. 1994; Superina and Loughry 2015; Freitas et al. 2019; Santos et al. 2019) and representing one of the earliest branching clades of placental mammals (Murphy et al. 2007; O’Leary et al. 2013; Gibb et al. 2016). Xenarthrans are considered imperfect homeotherms, given their poor ability to adjust body temperature (Mc Nab 1979, 1980, 1985). This inaptitude for thermal regulation, possibly related with their low metabolic rate and low energetic content diet, makes xenarthrans’ activity patterns highly affected by air temperature, with potential effects in their circadian cycles (Chiarello 1998; Giné et al. 2015; Maccarini et al. 2015; Di Blanco et al. 2017). While a recent report clearly identified pineal glands in the six-banded armadillo (Euphractus sexcintus), Linnaeus’s two-toed sloth (Choloepus didactylus), and in the southern tamandua (Tamandua tetradactyla), a distinct pineal was not found or was reported missing in species such as southern long-nosed armadillo (Dasypus hybridus), pale-throated sloth (Bradypus tridactylus), giant anteater (Myrmecophaga tridactyla), or big hairy armadillo (Chaetophractus villosus) (Benítez et al. 1994; Ferrari et al. 1998; Freitas et al. 2019). However, in the nine-banded armadillo (Dasypus novemcinctus) inconsistent reports advocate for either the presence or absence of a genuine pineal gland (Harlow et al. 1981; Freitas et al. 2019). Also, variable concentrations of circulating serum melatonin during the 24 h day–night cycle have been detected in this species (Fig. 1b), raising the hypothesis of an extrapineal source for melatonin production (Harlow et al. 1981, 1982). With the emergence of various whole-genome sequences from Pilosa (sloths and anteaters) (e.g., Uliano-Silva et al. 2019) and Cingulata (armadillos) (e.g., Lindblad-Toh et al. 2011), Yin et al. (2021) have recently reported the molecular erosion of Aanat in Xenarthra; yet no attempt was made to expand this analysis to the full melatonin-related gene hub. Thus, we are now able to interrogate whether the gene repertoire of circadian rhythmicity is modified in this lineage and clarify the physiological status of the pineal gland within this group.

Fig. 1
figure 1

(Illustrations used elements from Servier Medical Art: https://smart.servier.com/)

Melatonin synthesis and signaling. a Melatonin, generally described as a phase marker of the circadian clock, is initially synthetized from tryptophan which is converted in serotonin (Pévet 2002). The final steps of this synthetic pathway include a two-step metabolization of the intermediate serotonin into melatonin, a process catalyzed by Aanat and Asmt (Klein et al. 1997; Simonneaux and Ribelayga 2003). In mammals, Mtnr1a and Mtnr1b receptors are involved in the response of the clock machinery to melatonin stimulation (Reppart et al. 1996). b Summary of the available information regarding xenarthran’s melatonin levels, pineal gland presence, and habits. *Evidence obtained only from D. novemcinctus.

Full size image

Materials and Methods

Sequence Collection

To clarify the functional status of Aanat, Asmt, Mtnr1a, and Mtnr1b in eight xenarthran species (Online Resource 1), the genomic loci were retrieved for gene annotation using three strategies (following Alves et al. (2019) and Lopes-Marques et al. (2019b)): (a) for species with annotated genes, the genomic sequence of the target gene (ranging from the upstream to the downstream flanking genes) was collected directly from National Center for Biotechnology Information (NCBI); (b) for species with annotated genomes but lacking the annotation of the target gene, the genomic region between two conserved flanking genes (downstream and upstream) was directly collected; and (c) for unannotated genomes, blastn searches were performed, using as query a set of three genes, including Homo sapiens (human) target gene coding sequence (CDS), as well as those of the flanking genes in the same species. From the blast results, the best matching genome scaffold corresponding to the consensus hit across those obtained per each query sequence was retrieved. When no consensual blast hit was obtained, all hits corresponding to the H. sapiens CDS query were inspected, the aligning regions submitted to a back-blast search against the nucleotide database of NCBI, with the matching genomic sequence corresponding to the gene of interest being the one selected (when existing). When several matchings were found, the best genomic scaffold (yielding the highest query coverage and identity value) was collected for annotation.

For the 156 non-xenarthran mammals with annotated genomes (Online Resource 1), the first two strategies described above were adopted to obtain the genomic region corresponding to the target gene. In Dugong (Dugong dugon), since no annotation is currently available, the genomic sequence containing the target gene was retrieved via blastn searches.

Gene Annotation and Mutational Validation

The open reading frames (ORFs) of the mammalian orthologues of Aanat, Asmt, Mtnr1a, and Mtnr1b were investigated using PseudoChecker (pseudochecker.ciimar.up.pt), an online platform suitable for gene inactivation inference (Ranwez et al. 2018; Alves et al. 2020). For this purpose, the human gene orthologue was used as a comparative coding sequence input (NCBI Accession ID regarding human Aanat: NM_001088.3; Asmt: NM_001171039.1; Mtnr1a: NM_005958.4; Mtnr1b: NM_005959.5) to deduce the coding status of a given candidate gene in each target species. By making use of PseudoIndex—a user assistant metric built into the PseudoChecker pipeline that rapidly estimates the erosion condition of the tested genes—putative ORFs of the orthologous gene from each target species were assigned a discrete value from 0 to 5, with 0 suggesting a fully functional gene and 5 complete inactivation (Alves et al. 2020). When PseudoIndex was higher than 2, we proceeded to manual annotation and validation of possible disrupting mutations as previously described by Lopes-Marques et al. (2019a, b, c) and Alves et al. (2021). Briefly, by using H. sapiens CDS for each target gene as reference, each exon was isolated and mapped to the genomic region of the candidate pseudogenes using Geneious Prime (2019.2.3) “map to reference” tool. The aligned regions were individually screened for ORF-disrupting mutations (exon deletions, absence of start codon, sequence frameshifts and premature stop codons) and identified mutations were annotated. Mutational validation was performed through retrieval of raw sequencing reads in (at least) two independent Sequence Read Archive (SRA) projects (when available).

RNA-seq Analysis

Transcriptomic analysis was performed as previously described by Lopes-Marques et al. (2019c). Succinctly, RNA-seq datasets of multiple tissues were obtained from SRA projects to inspect the functional condition of each target gene in xenarthran species (when available) and Human (H. sapiens) (Online Resource 2). Transcriptomic reads recovered through blastn, were mapped to corresponding references genomes using the “map to reference” tool from Geneious Prime (2019.2.3) and manually removed if presenting poor alignment. Finally, reads were classified as spliced reads (spanning over two exons), exon–intron reads or exonic reads depending on the genomic region they mapped.

Results

Erosion of Melatonin-Related Genes in Xenarthra

Analysis of the melatonin synthesis genes in armadillos, anteaters, and sloths using PseudoChecker (Alves et al. 2020), showed that all analyzed species presented a PseudoIndex equal to five (Online Resource 3)—thus suggesting that the ORF of Aanat and Asmt includes inactivating mutations. Subsequent manual annotation of all collected xenarthran genomic sequences revealed Aanat and Asmt gene erosion across all analyzed species (Online Resource 4). Regarding Aanat, in agreement with Yin et al. (2021), we found numerous ORF-disrupting mutations, including a conserved 1-nucleotide deletion in exon 1 in Armadillos and a conserved 2-nucleotide deletion in exon 3 in Sloths (Fig. 2; Online Resource 4). On the other hand, although Yin et al. (2021) were not able to recover the genomic sequence containing the Aanat CDS in Anteaters, we uncovered, among other disruptive mutations, a conserved in-frame premature stop codon in exon 2. The identified mutations were next validated by searching at least one ORF-disruptive mutation per species in the corresponding SRAs; reads corroborating the identified mutations were systematically found (Online Resource 5). The analysis of Asmt in Xenarthra also revealed variable disruption patterns. In cingulatans, we found similar mutational events, however, not conserved within members of this group. Specifically, in D. novemcinctus exons 1–4 and 7 were not found, possibly due to poor genome coverage or complete exon deletion (Fig. 2). Moreover, in southern three-banded armadillo (Tolypeutes matacus) several insertions/deletions (indels) have been identified in exon 6, contrasting with D. novemcinctus where a validated in-frame premature stop codon in the same exon was detected (Fig. 2; Online Resources 4 and 6). In Vermilingua (anteaters), we were only able to recover exons 4–5 in M. tridactyla which provide a range of mutations with predicted disruptive effects (Fig. 2; Online Resource 4). In T. tetradactyla, despite the completeness of the assembly in the genomic region likely to contain exon 2 of Asmt, this was not found (Fig. 2; Online Resource 4). For Folivora (sloths), across several identified ORF-disrupting mutations, a trans-species conserved 4-nucleotide insertion in exon 6 was revealed and further validated by SRA searches (Online Resources 4 and 6). RNA-Seq analysis in Linnaeus’s two-toed sloth (Choloepus didactylus) Aanat yielded a high proportion of exon–intron reads versus spliced reads, in clear contrast with the pattern found in H. sapiens (Online Resource 7). In the case of Asmt, no transcriptomic reads were recovered for C. didactylus. Similarly, SRA transcriptome searches were unable to retrieve reads of D. novemcinctus for both genes.

Fig. 2
figure 2

Right panel: Schematic representation of the identified Aanat and Asmt genes ORF abolishing mutations in Xenarthra orders (Cingulata and Pilosa), Human (Homo sapiens), and Elephant (Loxodonta africana). Each box represents an exon and lines represent intronic regions. Phylogenetic trees were calculated in www.timetree.org; last accessed March 13, 2021 using species list. Silhouettes were sourced from Phylopic (http://phylopic.org). Left panel: Sequence alignments of the identified conserved disruptive mutations in both Aanat and Asmt genes of Xenarthra

Full size image

Regarding the examination of the Mtnr1a gene, no genome scaffold containing the target gene was possible to be obtained for T. matacus, M. tridactyla, Screaming hairy armadillo (Chaetophractus vellerosus), and T. tetradactyla (Online Resource 3). Moreover, in D. novemcinctus, the Mtnr1a coding status could not be accessed likely due to fragmentation of the respective genomic region [presence of sequencing gaps (Ns)]. On the other hand, for both species comprising the two-toed sloth group (Choloepus sp.), we were able to identify a validated 8-nucleotide deletion and a 20-nucleotide deletion in exon 2 together with a 4-nucleotide insertion in exon 1 (Fig. 3; Online Resources 4 and 8). In contrast, the current assembly of B. variegatus is incomplete, thus only exon 1 of Mtnr1a was recovered, presenting no ORF-disrupting mutations (Fig. 3; Online Resource 4). Curiously, in elephant (Loxodonta africana) exon 2 was not found despite the completeness of the assembly in the Mtnr1a region. The analysis of Mtnr1b CDS in T. matacus and C. vellerosus uncovered several inactivating mutations, including a conserved premature stop codon that truncates exon 2 (Online Resource 4). Pilosa species (sloths and anteaters) Mtnr1b gene annotation revealed the presence of several ORF-disrupting mutations, of note a single transversal mutation present in all analyzed species, namely a 2-nucleotide deletion in exon 2 (Fig. 3; Online Resource 4). This mutation was investigated and validated in both Choloepus species (Online Resource 9). Searches for transcriptomic evidence for Mtnr1a and Mtnr1b in C. didactylus retrieved a low number of reads, mostly corresponding to immaturely mRNA (Online Resource 7).

Fig. 3
figure 3

Right panel: Schematic representation of the identified Mtnr1a and Mtnr1b genes ORF abolishing mutations in Xenarthra orders (Cingulata and Pilosa), Human (Homo sapiens), and Elephant (Loxodonta africana). Each box represents an exon and lines represent intronic regions. Phylogenetic trees were calculated in www.timetree.org; last accessed March 13, 2021 using species list. Silhouettes were sourced from Phylopic (http://phylopic.org). Left panel: Sequence alignments of the identified conserved disruptive mutations in both Mtnr1a and Mtnr1b genes of Xenarthra

Full size image

Melatonin-Related Genes are Inactivated in Other Non-xenarthran Mammals

Sequence search and analysis for Aanat in 157 non-xenarthran mammal genomes returned a total of 11 species with no annotation of a Aanat-like sequence: Bison bison bison (American bison), Bos indicus (Zebu), Bos mutus (Wild yak), Bubalus bubalis (Water buffalo), Camelus ferus (Wild bactrian camel), Odocoileus virginianus texanus (White-tailed deer), Pantholops hodgsonii (Tibetan antelope), Sus scrofa (Wild boar), Myotis davidii (David’s myotis), and Myotis lucifugus (little brown bat). For the latter, we were not able to retrieve the genomic locus containing the target gene, given that both upstream and downstream flanking genes are also not annotated. Analysis using PseudoChecker (Alves et al. 2020), showed that 32 species non-xenarthran mammals presented a PseudoIndex higher than 2 (Online Resource 3). From these species, members of Cetacea (cetaceans) and Pholidota (pangolins) presented among their members, a conserved (and validated) in-frame premature stop codon in Exon 1 (Online Resource 4 and 10). Moreover, we also found ORF-disrupting mutations in Exon 1 of velvety free-tailed bat (Molossus molossus), Kuhl's pipistrelle (Pipistrellus kuhlii), and Sunda flying lemur (Galeopterus variegatus) with the latter being validated through SRA genomic reads (Online Resources 4 and 10). In D. dugon, several disruptive mutations were identified, such as an eight-nucleotide insertion in exon 2 and the presence of a stop codon in exon 3 (Online Resources 4 and 10).

Regarding Asmt, in 17 species the genomic fragments containing Asmt-like nucleotide sequences were not recovered given the lack of annotation for both target and flanking genes (Online Resource 3). For this gene, 74 species displayed a PseudoIndex higher than 2 (Online Resource 3), the majority due to fragmentation of the genomic region (presence of Ns), true absence of exons, poor alignment identity or incompleteness of the scaffold in the Asmt genomic region (Online Resource 4). Gene lesion events were found and validated mostly in Cetaceans, G. variegatus, Manis sp. (pangolins) and some Rodentia, with the latter showing poor alignment identity with the reference (Fig. 4a; Online Resources 4 and 11). Other examples of species presenting disruptive mutations but with no SRA validation (given that no genomic independent SRAs projects are available) include Brandt’s bat (Myotis brandtii), the Prairie vole (Microtus ochrogaster) with 2-nucleotide deletion in exon 1, and the Nancy Ma’s night monkey (Aotus nancymaae) with single nucleotide deletions in exon 2 (Online Resource 4). In the case of D. dugon, no ORF-disrupting mutations were found for Asmt; however, not all the exons were recovered due to incompleteness of the scaffold (Online Resource 4).

Fig. 4
figure 4

a Mutational landscape of melatonin synthesis and signaling genes along the mammalian tree. For each gene, we represented in green the orders where no ORF-disrupting mutations (frameshift mutations, in-frame premature stop codons, loss of canonical splicing site or exon deletions) were found across all members. On the other hand, orders where all members presented ORF-disrupting mutations are highlighted in red. In orders (and in genes) with no consensual disruption pattern, number of species presenting a coding/non-coding sequence were depicted, respectively. Species where no SRA validation was possible were not included in this figure. *Indicates the presence of contradictory reports in Mtnr1a for Trichechus manatus latirostris. Silhouettes were sourced from Phylopic (http://phylopic.org). Phylogenetic relationships were adapted from Vazquez et al. (2018). b Summary characterization of mammalian lineages presenting complete molecular erosion of melatonin synthesis and signaling genes, regarding their sleep type, habitat, and lifestyle (Color figure online)

Full size image

We next expanded our search to understand whether melatonin signaling genes would be compromised in non-xenarthran mammals. For Mtnr1a, a total of 44 species exhibited a PseudoIndex higher than 2 (Online Resource 3), yet manual curation revealed ORF-disrupting mutations only in pangolins (validated through SRA Projects; Online Resource 12) (Huelsmann et al. 2019), Hawaian monk seal (Neomonachus schauinslandi) and D. dugon (Online Resource 4). In cetaceans a different scenario emerged, with several exons completely absent (Online Resource 4) (Huelsmann et al. 2019; Lopes-Marques et al. 2019b).

In Mtnr1b, a total of 69 non-xenarthran mammals displayed a PseudoIndex higher than 2 (Online Resource 3). However, and contrary to the pattern found for Mtnr1a, annotation of the collected sequences revealed Mtnr1b gene erosion across multiple species, mostly affecting the Carnivora and Cetaceans but also Pholidota, Sirenia and some Primates (Fig. 4a; Online Resource 4). Examples of conserved inactivating mutations were found in bears (Ursus sp.) with an in-frame premature stop codon in exon 1, weasels (Mustela sp.) sharing several indels in exon 2 and pangolins with a single nucleotide deletion and an in-frame premature stop codon in exon 1 (Online Resources 4 and 13). Other species with ORF-disruptive mutations include Nannospalax galili (northern Israeli blind subterranean mole rat), exhibiting a single nucleotide deletion in exon 1, D. dugon with a conserved fourteen-nucleotide deletion in exon 2 or G. variegatus with an indel also in exon 1 (Online Resources 4 and 13). Detailed characterization of each target gene in mammals is available in Online Resource 4 and the minutiae of SRA validation can be found in Online Resources 10, 11, 12, and 13.

Discussion

Here, we set out to investigate how evolutionary genomic signatures might untangle the physiological status of controversial vestigial structures, using the pineal gland as a case study (Pévet 2002). For this, we addressed the evolution of the melatonin-related gene hub, encompassing melatonin synthesis and signaling genes, in Xenarthra and other mammals. Our results strongly suggest a complete landscape of gene loss in Xenarthra, which further reinforce reports suggesting the lack of a pineal gland in several members of this superorder (Quay 1965; Harlow et al. 1981; Benítez et al. 1994; Ferrari et al. 1998; Freitas et al. 2019). Interestingly, the mutational profile of melatonin synthesis and signaling genes within Xenarthra, supports the occurrence of independent inactivation events among Xenarthran orders (Cingulata and Pilosa). However, this analysis is highly dependent of the quality of genome sequencing projects and their assembly, thus additional near error-free Xenarthra reference genomes (e.g., Uliano-Silva et al. 2019) is needed to strengthen our conclusions. The present data also suggest that, in species in which a pineal gland was described (e.g., Freitas et al. 2019), the canonical pineal gland physiology leading to melatonin secretion is likely disrupted. Nevertheless, similarly to what was described for Tursiops truncatus (bottlenose dolphin) (Panin et al. 2012), previous radioimmunoassay methods have reported the presence of melatonin circulating in D. novemcinctus (Harlow et al. 1981), implying either the existence of independent pathways for melatonin synthesis and signaling (Slominski et al. 2003; Tan et al. 2016) or possible acquisition of melatonin from food sources (Tan et al. 2010).

Understanding whether gene loss events are the result of neutral or adaptive evolution is a complex process. In a scenario of a 'neutral' gene loss under a situation of regressive evolution, the loss of the full melatonin-related gene hub, should be considered as a trigger consequence in the process of occupying a novel ecological niche or driven by changes in metabolism (e.g., Moreau and Dabrowski 1998; Protas et al. 2006). This does not seem to be the case, given that the ancestral of Xenarthra likely was a myrmecophagous digger and/or burrower, with some climbing capabilities and it was found in the equatorial latitudes of South America (Gaudin and Croft 2015). Therefore, given there is no perceptive changes in terms of diet or habitat in Xenarthra as a whole, the hypothesis of a neutral fixation seems improbable.

In contrast, this strong genomic signature leading to the anatomical and/or physiological atrophy of this endocrine gland can also be viewed as an adaptive solution to overcome physiological limitations (Helsen et al. 2020). In agreement, most of the Xenarthra species are described as cathemeral (irregular daily activity pattern) and heterothermic species (Eisenberg and Redford 1999) with limited capacity to regulate their body temperature and with their movements heavily influenced by air temperature (Greegor 1985; Camilo-Alves and Mourão 2006; Giné et al. 2015; Attias et al. 2018). Thus, to reduce such energetic costs, xenarthrans may have suffered reductive episodes (i.e., genes losses), allowing behavioral strategies to overcome unfavorable environmental conditions and mitigate thermal limitations (Yin et al. 2021).

Accordingly, convergent evolutionary disruptive patterns in lineages also displaying labile body temperature (PholidotaMc Nab 1984; Heath and Hammel 1986; Weber et al. 1986; Yin et al. 2021) and suggestive bizarre sleeping patterns—the tree pangolin (Manis tricuspis) lacks of hypothalamic cholinergic neurons, associated to a normal sleep phenomenology (Imam et al. 2018)—make it plausible to hypothesize that, in these species, inactivation of melatonin-related genes can be associated with modifications in their circadian rhythmicity (Fig. 4b). To further support this scenario, species living in environments with specific thermal constraints (Cetacea and Trichechus manatus latirostris (Florida manatee); Huelsmann et al. 2019; Lopes-Marques et al. 2019b; Yin et al. 2021; Emerling et al. 2021) also display loss of melatonin synthesis and signaling. Therefore, the idea of evolutionary convergence to allow the emergence of unusual activity patterns should be strongly considered. In addition, pseudogenization of these genes possibly paralleled loss of other circadian rhythm related genes, namely Cortistatin gene, that encodes a pleiotropic neuropeptide with an important role in sleep physiology (Valente et al. 2021). Patterns of gene disruption observed in other mammalian lineages are challenging to disentangle since the absence of melatonin genes in many different species might represent true losses or can be artifacts due to poor quality of sequence projects. Notwithstanding, in Aanat no members of other mammalian orders presented disruption of this gene (Fig. 4a; Online Resource 14). Regarding Asmt, signs of pseudogenization/exon loss of Asmt found in Rodents are likely alignment artifacts, possibly due to the rapid evolution that this gene has experienced in this order (Kasahara et al. 2010). Moreover, lack of exons in other mammal species, do not allow us to conclude with certainty whether this gene has been partially deleted, unaligned due to sequence variability or constitute poor genome assemblies in Asmt coding regions. When looking to Mtrn1a gene, three species presented erosion of this receptor: the previous reported naked mole rat (Heterocephalus glaber) (Kim et al. 2011)—possibly consequence of the completely dark environment in which it inhabits, the Hawain monk seal and the elephant, whose pineal presence has been disputed (Dexler 1907; Haug 1972) and suggested to become involuted as individuals become adults (Shoshani et al. 2006). With a distinct pattern, Mtrn1b has been lost in several lineages, with members of Carnivora, Primata, Rodentia, Eulipotyphla, Afrosoricida, Diprotodontia, and Monotremata presenting disruptive mutations for this gene (Fig. 4a). Overall, we have detected the complete loss of melatonin signaling in naked mole rat and Hawaian monk seal which also lacks several exons in Asmt, leading us to propose that both melatonin synthesis and signaling modules are completely dismantled in this species—which goes in accordance with its adaptation to the marine environment.

Given that the evolution of melatonin-related genes should be directly linked with pineal gland function, by inferring their coding status we were able to deduce if the organ constitutes an evolutionary vestige, despite the conflicting anatomical reports. More importantly, the present study provides a clear case study on how genomic data can be used to disentangle whether a specific organ constitutes a functional or vestigial structure (Hiller et al. 2012).

Conclusion

To date, no unequivocal inferences on the functional status of pineal gland across mammals were provided, with anatomical observations in several species from different clades presenting conflicting conclusions. However, by making use of genomic data, our results provide solid evidence for pineal gland vestigiality not only in Xenarthra, but also in other mammalian lineages. Thus, we argue that analysis of genomic changes might constitute a powerful approach to gain insights into the vestigiality of specific organs.