Introduction

The reconstruction of species relationships is crucial to understanding evolutionary history, delimiting species boundaries, and characterizing the extent of biodiversity. Molecular phylogenetics, which estimates species relationships using genetic data, has revolutionized the understanding of species evolution with high analytical power and robustness for identifying morphological convergence (e.g., Teeling et al. 2005). However, molecular phylogenetics often reveals discordant tree topologies when generated using different genes or genomes (e.g., mitochondrial versus nuclear loci and genomes; Degnan and Rosenberg 2009; Ge et al. 2022). This is especially true in recently diverged lineages (e.g., Eaton and Ree 2013) and at ancestral nodes of rapid radiations (e.g., Alda et al. 2019). Such discordances may arise from relationships more complicated than a bifurcating tree and can indicate complex evolutionary processes that have shaped the observed species divergence (Ge et al. 2022). For example, a recent study of bryophyte plants found that phylogenetic discordances were mainly caused by incomplete lineage sorting (ILS) rather than post-speciation gene flow, suggesting extensive ancestral admixture during early divergence which might have facilitated radiation into diverse habitats (Meleshko et al. 2021). In another example, the discordance between mitochondrial and nuclear phylogenies of bears demonstrated historical introgression of the brown bear mitochondria in the extant polar bear lineage and fixation of this mitochondrial haplotype, possibly under natural selection (Hailer et al. 2012). These studies show the power as well as the importance of comparing phylogenies generated from different molecular markers and interpreting tree discordances in the species phylogeny from an evolutionary perspective. It is thus clear that genomic analyses using multiple loci from both nuclear and mitochondrial genomes are required to fully understand species relationships and their evolutionary history.

Despite the accumulation of molecular phylogenies, multilocus analyses and comparisons remain limited, especially in non-model taxa, which mostly rely on a few widely sequenced genes. For example, in bats, the mitochondrial gene cytochrome oxidase subunit 1 (COI) has been widely used for species barcoding (Lim et al. 2017; Chornelia et al. 2022) and the cytochrome b (CYTB) gene is one of the most analyzed markers in bat phylogeny on various scales (Amador et al. 2018; Velazco et al. 2021). Given limited resources (e.g., budget and samples), it is beneficial to focus on these widely used markers so that the results can be compared with published data and reused in future research. Some bat studies addressed the importance of multilocus phylogeny by including select nuclear sequences, such as introns (Demos et al. 2019) and/or exons (Roehrs et al. 2010; Juste et al. 2013), and at times observed mito-nuclear discordances that indicated complex evolutionary histories such as historical introgression and rapid radiation. However, the power of datasets that survey variation in only a few nuclear regions may be limited within certain taxa (e.g., those with low diversity or low mutation rates) and they might not fully represent genome-wide diversity or the species phylogeny.

The development of ultra-conserved elements (UCEs; Faircloth et al. 2012) provides a great opportunity to use genome-wide nuclear markers for multilocus phylogenies in non-model taxa, including bats (Platt et al. 2018; Yi and Latch 2022). The UCE libraries represent hundreds to thousands of loci across the nuclear genome and allow high-resolution phylogenetic reconstruction as well as a better approximation of the species tree. Because UCE libraries target nuclear regions, the comparison with a mitochondrial phylogeny would require additional sequencing of all or part of the mitogenome (such as in Platt et al. 2018). To partially overcome this limitation, the MitoFinder pipeline has been recently developed to efficiently extract mitogenomic data from off-target fragments in UCE libraries (Allio et al. 2020). Therefore, the UCE sequencing combined with MitoFinder annotation could provide highly efficient use of data to construct multilocus species phylogenies as well as detect mito-nuclear discordances. Such multilocus phylogenetic analyses are particularly important for taxa characterized by rapid radiation and recent divergence (such as in bats, Shi and Rabosky 2015) which tend to generate genealogical discordances and complex species relationships. In addition, both UCE markers and the annotated mitochondrial genes can be readily combined with published datasets to make the most use of available sequencing data, which would further benefit phylogenetic studies in non-model systems.

Here we applied MitoFinder to the previously published UCE libraries of the serotine bat genus Eptesicus (Yi and Latch 2022) to gain insights into their mitochondrial phylogeny and potential mito-nuclear discordances. Specifically, we focused on the clade of Eptesicus (Histiotus), a poorly studied taxon endemic to South America. Histiotus was initially classified as a morphological genus based on their enlarged ears that differ from the short ears of all the other Eptesicus species globally (Nowak and Walker 1994; Díaz et al. 2019; Rodríguez-Posada et al. 2021). Nevertheless, molecular studies repeatedly found Histiotus clustered within the genus Eptesicus (Hoofer and Van Den Bussche 2003; Roehrs et al. 2010; Yi and Latch 2022), leading it to be classified as a subgenus of Eptesicus (Giménez et al. 2019). A recent study proposed splitting the Eptesicus clades into four genera including Eptesicus (restricted to E. fuscus and E. guadeloupensis), Cnephaeus (Old World species), Histiotus (big-eared Neotropical species), and Neoeptesicus (other Neotropical species; Cláudio et al. 2023). However, additional research is needed to better understand the evolutionary history and phylogenetic arrangements in this group to assess the (sub)generic classification in Eptesicus.

In addition to the debate on nomenclature, phylogenetic discordances have been extensively reported among Eptesicus and Histiotus, leading to distinct evolutionary scenarios (Fig. 1). The mitochondrial tree based on three genes found sister relationships between E. (Histiotus) and Neotropical Eptesicus (Hoofer and Van Den Bussche 2003), whereas the nuclear tree using six genes found E. (Histiotus) sister to all the other New World Eptesicus (Roehrs et al. 2010), and the nuclear tree using hundreds to thousands of UCEs found E. (Histiotus) and the Nearctic E. fuscus as sister clades (Yi and Latch 2022; Fig. 1). However, the mitochondria-based studies had limited sampling of the Neotropical Eptesicus (e.g., Giménez et al. 2019; Rodríguez-Posada et al. 2021; Laine et al. 2023) and/or limited sampling of E. (Histiotus) (Hoofer and Van Den Bussche 2003; Roehrs et al. 2010), making it difficult to directly compare the mitochondrial and nuclear phylogenies. In addition, up to 11 species have been recognized as part of the Histiotus clade (Table 1; Simmons and Cirranello 2024, https://batnames.org/, accessed March 2024) but species diversity likely remains underestimated (Yi and Latch 2022). Despite this recent progress, systematic research on E. (Histiotus) has been limited by the difficulty of collecting these rare and cryptic bats (Carvalho et al. 2013) and getting access to various species in a single study. The limited data and lack of a taxa-complete phylogeny largely impede our understanding of their diversity and evolutionary history. To fill in these knowledge gaps, here we make the most of collected molecular datasets from field samples and museum collections to shed light on the Eptesicus evolutionary history.

Fig. 1
figure 1

Discordances in Eptesicus phylogenies reported in previous studies using different genomic markers

Table 1 The 11 recognized species in the Histiotus clade based on Simmons and Cirranello (2024). Species included in our analyses are labeled with asterisks. We kept the subgenus classification of Histiotus (see text for details)

Accordingly, we have two major goals in this study. First, we aim to assess mito-nuclear discordances in Eptesicus by building a mitochondrial phylogeny using MitoFinder annotations and comparing it to the UCE phylogeny (Yi and Latch 2022). Unlike previous studies, our broad coverage of geographic and taxonomic sampling and multiple mito-nuclear markers derived from the same set of individuals allow us to properly test contrasting evolutionary scenarios of serotine bats. Second, we aim to compile the published and newly generated mitochondrial sequences to produce the most taxa-complete mitochondrial phylogeny of E. (Histiotus). We focused on the mitochondrial phylogeny because of more available mitochondrial sequences on GenBank and the wide use of mitochondrial genes in bat species validation (Giménez et al. 2019). This study showcases a way to make the most of collected molecular datasets which might be especially beneficial for non-model systems that have limited resources. These results improve our understanding of the diversity and evolutionary history of serotine bats and specifically E. (Histiotus), which help inform their taxonomy as well as conservation in several South American biodiversity hotspots.

Material and methods

MitoFinder annotation of bat UCE libraries

The assembled contigs (including UCEs and off-target contigs) of 96 UCE libraries (Table S1; herein the Eptesicus UCE libraries) were generated in Yi and Latch (2022) and downloaded from the Dryad archive (https://doi.org/https://doi.org/10.5061/dryad.b8gtht7fn). MitoFinder v1.4.1 (Allio et al. 2020) was used to annotate mitochondrial genes from these contigs (command -a) based on user-provided mitochondrial reference genomes (see below) with the vertebrate mitochondrial code (-o 2). We set the minimum contig size as 50 bp (–min-contig-size 50) considering the potential degradation of museum dry skin specimens (Yi and Latch 2022). All the other parameters were kept as default. To obtain mitochondrial reference genomes for the MitoFinder annotation, we searched “vespertilionidae bat mitochondrion complete genome” on the NCBI nucleotide database (accessed June 2022) and selected all 44 items of the resultant genera in the UCE libraries, namely Myotis, Kerivoula, Lasionycteris, and Miniopterus. In addition, we included a near-complete partial mitochondrial genome of Eptesicus fuscus, 3 mitogenomes of the Old World species Eptesicus bottae, and 43 recently published (Laine et al. 2023) mitogenomes of E. (Histiotus) species. Complete records of these 91 reference genomes (Table S2) were downloaded as GenBank files and input in MitoFinder as the mitochondrial reference genomes. Because MitoFinder was developed using insect UCE datasets and has not yet been widely tested in vertebrates, we validated the accuracy of MitoFinder-annotated sequences in two ways. First, we blasted the results of annotated COI and CYTB sequences using the unix command blastn (-db nt -remote; accessed January 2023) with default settings to verify if they match sequences of the same or closely related species, which could serve as a rough evaluation of MitoFinder performance. The best match output by blastn was reported for each individual sample. Second, we ran MitoFinder on the UCE libraries of Myotis bats where the mitogenomes of corresponding samples were sequenced independently and are available on NCBI (Platt et al. 2018). Details of the related methods and results are available in supplementary materials.

Bat mitochondrial phylogeny using MitoFinder annotations

We constructed a mitochondrial phylogeny using all MitoFinder annotated sequences from the Eptesicus UCE libraries. The nucleotide mitochondrial gene sequences (e.g., files final_genes_NT.fasta) were compiled into one fasta file and aligned in PHYLUCE v1.7.0 (Faircloth 2016). The command phyluce_align_seqcap_align was run without trimming (–no-trim), the MAFFT sequence alignment (–aligner mafft), output all genes (–taxa 1), and allowing taxa incompleteness (–incomplete-matrix). The output alignments were cleaned by removing locus names (command phyluce_align_remove_locus_name_from_files) and concatenated into the phylip format (command phyluce_align_concatenate_alignments). The concatenated alignments were used to estimate maximum likelihood (ML) phylogenetic trees in RAxML v8 (Stamatakis 2014) with the GTRGAMMA model, a rapid ML searching (-f a), and 1000 rapid bootstraps (-# 1000). RAxML was processed twice using no partitioning (i.e., one partitioning scheme) or the partitioning scheme by genes. We also estimated Bayesian informative (BI) phylogenetic trees in ExaBayes v1.5.1 (Aberer et al. 2014) with the GTR model (the only one available) and no partitioning or the partitioning scheme by genes. ExaBayes was carried out in two independent runs, each with two coupled chains for 5 million generations, 25% burnin, and default values of the other parameters. The summary stats of parameters were calculated using the command postProcParam, and convergence between the two chains within each run was checked by the effective sampling size (ESS > 200) of all parameters. The command consense was used to build an extended majority rule consensus tree using results of the two runs.

To provide statistical support for the detected mito-nuclear discordance, we conducted tree topology tests in IQ-TREE version 2.2 (Minh et al. 2020) using the above mitochondrial alignment and the UCE alignment from Yi and Latch (2022). For each alignment, we compared the respective ML tree to its modifications with different divergence sequences of Eptesicus fuscus, Histiotus, and Neotropical Eptesicus clades (illustrated in Fig. 1). Significant mito-nuclear discordances would be supported if the nuclear topologies were significantly rejected using the mitochondrial alignment and if the mitochondrial topology was significantly rejected using the nuclear alignment. Topologies were compared using log-likelihoods, the weighted Shimodaira-Hasegawa tests (SH; Shimodaira and Hasegawa 1999), and the approximately unbiased tests (AU; Shimodaira 2002), with significance estimated by 10,000 replicates. Tree topology tests were conducted using the best model determined by ModelFinder (Kalyaanamoorthy et al. 2017) based on the Bayesian information criterion (BIC) and model parameters were estimated based on the initial parsimony tree (-n 0). In addition, we also ran tree topology tests on the mitochondrial alignment to compare the topologies of E. (Histiotus) monophyly versus paraphyly regarding to the position of E. (H.) magellanicus (see results).

Sampling and sequencing of new E. (Histiotus) individuals

We also sampled thirteen new E. (Histiotus) individuals and sequenced their COI and CYTB genes in this study (Table S3). We focused on these two mitochondrial genes because of their wide use in bat systematic studies (Lim et al. 2017; Chornelia et al. 2022) and their abundance in GenBank. Four individuals of E. (H.) diaphanopterus were collected from its type locality (Table 1) and represented the first genetic material of this species. Two individuals of an unidentified E. (Histiotus) species were collected from the Brazilian Atlantic Forest in Iperó, São Paulo. Samples were collected following the Chico Mendes Institute for Biodiversity Conservation license number 83609–2. Individual bats were collected by mist nets and liver samples were taken in the field and stored in 95% ethanol. DNA was extracted using DNeasy Blood and Tissue Kit, following the manufacturer’s instructions. The COI gene was sequenced using the primers RepCOI-F (5’TNTTMTCAACANACCACAAAGA-3’) and RepCOI-R (5’-ACTTCTGGRTGKCCAAARAATCA-3’) in Nagy et al. (2012). The CYTB gene was sequenced using the primers Bat 05A_F (5’-CGACTAATGACATGAAAAATCACCGTTG-3’) and Bat 14A_R (5’-TATTCCCTTTGCCGGTTTACAAGACC-3’) in Martins et al. (2007). PCR cycling program for both genes used was 94 °C for 3 min, followed by 35 cycles of 94 °C for 30 s, 50 °C for 45 s, and 72 °C for 90 s, and concluding with a 10 min extension at 72 °C. PCR results were sequenced by ACTGene company (Nova Alvorada, Brazil). The forward and reverse raw sequences of each individual were aligned using Geneious and their consensus sequence was cleaned manually by correcting errors and trimming low-quality bases at the ends. Two individuals of E. (H.) diaphanopterus (BAT20, BAT21) only had forward sequences available for the COI gene which were also cleaned manually. The cleaned sequences were used in the downstream analyses.

Seven individuals of E. (H.) mochica, the most recently described species of the group, were collected from Peru (Table S3). Samples were collected following the permit Dirección de Gestión Sostenible del Patrimonio de Fauna Silvestre 02–2014-SERFOR-DGGSPFFS-DGSPFS. Individual bats were collected by mist nets and liver samples were taken from each individual in the field and stored in 95% ethanol. DNA was extracted using the phenol/chloroform protocol. The COI gene was sequenced using the VF1 (5’-TTCTCAACCAACCACAAAGACATTGG-3’) forward and VR1 (5’-TAGACTTCTGGGTGGCCAAAGAATCA-3’) reverse primers (Lim 2017) and the cycling program of 94 °C for 2 min, followed by 36 cycles of 94 °C for 30 s, 50 °C for 30 s, and 72 °C for 90 s, and concluding with a 3 min extension at 72 °C. The CYTB gene was sequenced using the LGL765 (5′-GAAAAACCAYCGTTGTWATTCAACT-3′) forward and LGL766 (5′-GTTTAATTAGAATYTYAGCTTTGG G-3′) reverse primers (Bickham et al 1995) and the cycling program of 94 °C for 2 min, 36 cycles of 94 °C for 45 s, 50 °C for 30 s, and 72 °C for 150 s, and concluding with a 3 min extension at 72 °C. PCR products were sequenced in the Laboratory of Molecular Systematics at the Royal Ontario Museum (Toronto, Canada). The forward and reverse raw sequences of each individual were aligned using Sequencher and their consensus sequence was cleaned manually by correcting errors and trimming low-quality bases at the ends. Only four of the newly collected E. (H.) mochica samples were sequenced for the CYTB gene because they were collected from the same geographic sites (Table S3) and had identical COI sequences.

Ethics statement

Although minimally invasive sampling methods (such as wing biopsies) can achieve the aims of the current study, the thirteen bats used in this study were sampled for museum collections under authorized research permits for broader objectives of collaboration on the International Barcode of Life initiative to document the world's biodiversity (www.ibol.org). This includes the establishment of data standards such as permanent voucher specimens that correspond to DNA sequences (Ratnasingham and Hebert 2007) and facilitate a more biologically integrative approach to research. The sampled bats have been housed in scientific collections in the Royal Ontario Museum (Ontario, Canada) and the Universidade Federal da Paraíba (Paraíba, Brazil; Table S3).

Mitochondrial gene trees

We compiled the MitoFinder-annotated COI and CYTB sequences with the newly sequenced samples and available NCBI sequences to construct the most taxa-complete mitochondrial gene trees of Eptesicus. For the COI gene tree, we searched “cytochrome oxidase subunit 1 COI” on the NCBI nucleotide database in the genus Eptesicus and the outgroup species included in our annotated dataset and blast results. The COI gene sequences found on NCBI were downloaded as fasta files. In addition, 57 recently published COI gene sequences (Laine et al. 2023) were downloaded from https://github.com/nvlain/histiotus2022. The compiled sequences were aligned using MAFFT v7 (Katoh and Standley 2013) with the –auto command. Alignments were output in the phylip format and analyzed in RAxML v8 to estimate the ML tree using no partitioning, the GTRGAMMA model, and a rapid ML searching (-f a) with 5000 bootstraps (-# 5000). For the CYTB tree, we searched "eptesicus mitochondrial cytochrome b", “histiotus mitochondrial cytochrome b", and CYTB sequences of outgroup species on the NCBI nucleotide database and downloaded results as fasta files. Preliminary results showed extremely long tip branches of two samples (accession MT674673, 255 bp; our annotated sample, accession OP157118, 397 bp), probably biased by their short sequences and resulting poor alignment (Kimball et al. 2021). Therefore, these two sequences were removed from the analyses. In addition, two sequences (BBMIH015-19, BBMIH012-19) of E. (H.) colombiae were obtained from Rodríguez-Posada et al. (2021) and 57 CYTB sequences of Eptesicus or E. (Histiotus) species were obtained from Laine et al. (2023; https://github.com/nvlain/histiotus2022). All compiled sequences were aligned using MAFFT and estimated for the ML phylogeny as described above.

Lastly, we also concatenated COI and CYTB genes using the 120 individuals that had both sequences available, including MitoFinder annotations, our newly sequenced individuals, and the sequences recently published by Laine et al. (2023). All gene sequences were compiled, aligned, and concatenated using PHYLUCE commands in the same way described above. The ML phylogenetic trees were constructed in RAxML v8 using the concatenated alignments with the GTRGAMMA model, a rapid ML searching (-f a), 5000 rapid bootstraps (-# 5000), and no partitioning or the partitioning scheme by gene. All trees generated in this study were rooted by the outgroup taxon Miniopterus and visualized in FigTree (http://tree.bio.ed.ac.uk/software/figtree).

Results

MitoFinder annotation of bat UCE libraries

A total of 15 mitochondrial genes were annotated from the bat UCE libraries using MitoFinder, namely ATP6, ATP8, COX1 (i.e., COI), COX2, COX3, CYTB, ND1, ND2, ND3, ND4, ND4L, ND5, ND6, rrnL (i.e., 16S), and rrnS (i.e., 12S). At least one mitochondrial gene was annotated in 82 of the 96 Eptesicus UCE libraries (Table S1). Not all samples used in the UCE phylogeny were annotated by MitoFinder, and 10 samples that failed the UCE enrichment (i.e., < 500 UCEs in Yi and Latch 2022) were annotated with at least one mitochondrial gene in this study. The annotated COI and CYTB sequences were submitted to NCBI for additional verification and publication. A total of 64 samples were annotated with COI sequences (60 passed NCBI verification, GenBank accession OP137035-OP137094) and 62 with CYTB sequences (61 passed the NCBI verification, GenBank accession OP157085-OP157145). It was not clear why five of the annotated sequences did not pass NCBI verification although they had reasonable blast results (some even blasted to the same species; Table S1); we excluded them from gene tree constructions for clarity. All of the annotated E. (Histiotus) samples were blasted to the same subgenus and even species for both COI and CYTB (Table S1). One sample, MVZ:Mamm:224,796 (OP137037 and OP157088), was originally identified as Eptesicus but blasted to Myotis, supporting previous findings based on the UCE phylogeny (Yi and Latch 2022). Importantly, the MitoFinder-annotated CYTB sequence of AMNH:M-278524 (E. (Histiotus) mochica) was blasted to the CYTB sequence (MK429705) of the same specimen sampled and sequenced by another study (Giménez et al. 2019). Both CYTB sequences are 1140 bp and they only differ at five nucleotide bases, four of which are incompletely specified bases (N, Y, W) in MK429705 whereas our annotation (OP157128) had specified bases at all sites. Similarly, MitoFinder analyses of Myotis UCE libraries (Platt et al. 2018) annotated CYTB sequences in 24 samples and COI sequences in 22 samples, almost all of which were blasted to the mitogenome reference of the same specimen (supplementary materials, Table S4). Accordingly, the BLAST results indicated high accuracy of our MitoFinder annotations on bat UCE libraries.

The mitochondrial phylogeny and mito-nuclear discordances of genus Eptesicus

All mitochondrial genes annotated from the 82 Eptesicus UCE libraries (Yi and Latch 2022) were concatenated and aligned (14,088 bp, 5987 parsimony-informative sites) to construct both ML and BI phylogenetic trees. The ML trees showed almost identical topologies, with relatively higher bootstrap supports when using no partitioning schemes compared to partitioning by gene (Fig. S1A, B). Four major clades were identified, representing E. fuscus, E. (Histiotus), Neotropical Eptesicus, and the Old World Eptesicus. A sister relationship between E. (Histiotus) and the Neotropical Eptesicus was recovered, consistent with previous mitochondrial studies (Fig. 1; Fig. 2A). The BI trees using no partitioning and partitioning scheme by gene recovered almost identical topologies but placed three samples of E. (H.) magellanicus from Chile (sample AMNH:M-93314 was reidentified as E. (H.) magellanicus; Velazco et al. 2021) as the basal lineage of the Neotropical group, rendering E. (Histiotus) paraphyletic (Fig. S1C, D). This topological difference between ML and BI trees might be caused by the different evolutionary models (GTRGAMMA versus GTR) or a lack of phylogenetic signal leading to suboptimal performance of Bayesian methods. Topology tests were conducted in IQ-TREE using the mitochondrial alignment on these two topologies, i.e., E. (H.) magellanicus included in E. (Histiotus) or diverged before the divergence of the other E. (Histiotus) species and Neotropical Eptesicus. The ML topology had a slightly higher log-likelihood (delta 1.8559) but neither topology was significantly rejected by the weighted SH tests (p values 0.599 and 0.412) or the AU tests (p values 0.609 and 0.391) using 10,000 replicates. Therefore, our data did not have the power to distinguish between these two topologies. For clarity, here we focused on the ML phylogeny as it is most consistent with previous studies (Hoofer and Van Den Bussche 2003; Roehrs et al. 2010) and the morphological prediction of E. (Histiotus) monophyly. Interestingly, an E. (Histiotus) colombiae, FMNH 72165, was clustered within the Neotropical Eptesicus clade in both ML and BI trees (Fig. 2A; Fig. S1), similar to the NCBI sequences of E. (Histiotus) colombiae in gene trees (see below). It should be noted that this museum skin-derived sample was only annotated with two mitochondrial genes (COX3, ND2) in this study and it failed the UCE enrichment in Yi and Latch (2022). Therefore, MitoFinder annotation can be a good complementary source of phylogenetic data especially when using samples of low DNA quality or quantity (e.g., museum specimens).

Fig. 2
figure 2

The Maximum Likelihood tree topologies using A) the mitochondrial annotations (same in Fig. S1A) and B) the nuclear UCEs (modified Fig. 2 in Yi and Latch 2022). Both trees were constructed using no partitioning scheme. Bootstrap supports are shown on the nodes. Branch lengths are removed for the purpose of illustration (see Fig. S1A for the phylogeny with branch lengths). Each tip represents one individual sample named by species identity (updated identity in the E. (H.) clade, consistent with Fig. 3) followed by sampling locality (Dominican Republic: DR). Full names are given in Table S1 corresponding to sample catalogues. The same individuals are connected by lines between trees. Outgroups are collapsed and potentially mislabeled sample names are in red (but see alternative explanations for E. (H.) colombiae and E. sp_Peru in discussion). Rectangles highlight the major clades and subclades. Mônico and Soto-Centeno (2024) raised the Caribbean Eptesicus fuscus to full species named E. dutertreus while maintaining the species status of E. guadeloupensis. The Neotropical Eptesicus Clade A was monophyletic in the UCE tree (Yi and Latch 2022) but paraphyletic in the mitogenome phylogeny

Both ML and BI mitochondrial trees differed from the nuclear UCE phylogeny (Yi and Latch 2022) in several ways (Fig. 2). First, E. (Histiotus) was more related to the Neotropical Eptesicus in the mitochondrial phylogeny, whereas the UCE phylogeny grouped E. (Histiotus) and E. fuscus as sister clades. Importantly, tree topology tests rejected nuclear topologies using the mitochondrial alignment and rejected the mitochondrial and nuclear gene topologies using the UCE alignment (Fig. 1, Table 2), supporting significant mito-nuclear discordances. Second, the Old World Eptesicus was paraphyletic in the UCE phylogeny but monophyletic in the mitogenome tree. Third, the Neotropical Eptesicus species split into two well-supported UCE clades but were less distinguishable in the mitochondrial tree. Fourth, the Caribbean Eptesicus fuscus (recently raised to full species E. dutertreus, Mônico and Soto-Centeno 2024) was clustered inside the continental E. fuscus in the mitochondrial tree but clearly separated from the continental E. fuscus in the UCE tree (Fig. 2). These mito-nuclear discordances echoed findings in previous studies (Hoofer and Van Den Bussche 2003; Roehrs et al. 2010; Turmelle et al. 2011; Laine et al. 2023) and thus likely reflected real evolutionary differences between the two genomes rather than methodological artifacts. In addition, the mitochondrial phylogeny also supported some individual misplacements (highlighted in Fig. S1) reported previously (Yi and Latch 2022), including the E. brasiliensis specimen FMNH 174918 that is nested within E. (Histiotus) (E.sp_Peru in Fig. 2).

Table 2 Tree topology tests from IQ-TREE 2. The best topologies are labeled with asterisks. Negative signs (-) indicate that the tree topology is significantly rejected

The COI and CYTB gene trees

The newly collected samples generated 650-686 bp COI sequences (OQ921993-OQ921998, OQ935566-OQ935572) and 1055-1141 bp CYTB sequences (OQ923265-OQ923269, OQ935562-OQ935565). The compiled COI gene alignment had 1568 bp (602 parsimony-informative sites) and 363 sequences, including 60 MitoFinder-annotated sequences that passed NCBI verification and 13 new sequences of E. (Histiotus) species. Results showed good clustering on the genus level and divided the Eptesicus genus into the same four clades, namely E. fuscus, E. (Histiotus), Neotropical Eptesicus, and the Old World Eptesicus (Fig. S2). Divergence within clades was less supported, such as between E. serotinus and E. nilssonii and among the Neotropical Eptesicus species. Similar to mitogenome phylogenies, the Caribbean species was clustered inside the continental E. fuscus. Two sequences were potentially mislabeled: JF445351 originally identified as Eptesicus furinalis was indicated to be a Myotis species, and HM540266 originally identified as Eptesicus sp. was recently described as a new genus and species Cassitrellus yokdonensis (red tips in Fig. S2; Ruedi et al. 2018).

The CYTB alignment had 1164 bp (568 parsimony-informative sites) and 426 sequences, including 60 MitoFinder-annotated sequences and 9 newly collected sequences of E. (Histiotus) species. The Eptesicus genus was divided into the same four clades (Fig. S3) as those in the mitogenome and COI phylogenies. Four CYTB sequences identified as E. (Histiotus) colombiae (BBMIH012-19, BBMIH015-19, MT113464.1, MT113466.1; Rodríguez-Posada et al. 2021) were again clustered within the Neotropical Eptesicus clade, similar to the E. (Histiotus) colombiae sample in our mitogenome phylogeny (Fig. 2A).

The concatenated COI and CYTB alignment had 2690 bp (1063 parsimony-informative sites) and 120 sequences. The ML trees were highly similar except that the tree with no partitioning showed monophyly of E. (Histiotus), whereas the tree partitioned by gene indicated paraphyly of E. (Histiotus) by placing some of the magellanicus individuals at the base of the Neotropical group (Fig. S4).

Discussion

Our study provides an empirical test of MitoFinder utility in bat UCE libraries. We showed that UCE sequencing combined with MitoFinder annotation can be a useful option for molecular phylogenetic studies to simultaneously obtain multilocus nuclear and mitochondrial data. These complementary datasets allow the estimation of species phylogeny and the detection of mito-nuclear discordances that may indicate complex evolutionary histories. Comparing the annotated mitochondrial genes with published sequences can help to validate or update species taxonomy. Here we showed an example of using MitoFinder annotation to enrich the molecular phylogeny of E. (Histiotus), a group of morphologically diverged bats whose taxonomy and evolution remain far from being understood due in part to the lack of genetic data.

The Eptesicus phylogeny and mito-nuclear discordances

The mitogenome phylogeny constructed with MitoFinder-annotated sequences showed several major mito-nuclear discordances from the previously generated UCE phylogeny (Yi and Latch 2022), including monophyly of Old World Eptesicus, less resolved divergence of Neotropical Eptesicus, and the position of E. (Histiotus) (Fig. 2). The recovered mitochondrial phylogeny is overall consistent with previous studies (Hoofer and Van Den Bussche 2003; Roehrs et al. 2010; Juste et al. 2013), but our phylogeny included a much higher coverage of New World species and geographic lineages and thus provided a more complete understanding of the mitochondrial evolution in these bats. Furthermore, the comparison between mitochondrial and UCE phylogenies using almost the same set of individuals allowed a proper detection of mito-nuclear discordances, ruling out sampling bias or misidentifications, and indicates a convoluted evolutionary history in serotine bats with widespread genetic admixture both within and across subgenera.

Nuclear UCE supported sister relationships between E. fuscus and E. (Histiotus) whereas the mitochondrial phylogeny showed sister relationships between E. (Histiotus) and Neotropical Eptesicus, a mito-nuclear discordance significantly supported by our topology tests. Mito-nuclear discordances can arise from incomplete linage sorting or introgression, the former generates no geographic pattern in the ILS-impacted phylogeny while the latter generates discordant topologies with different geographic patterns (Toews and Brelsford 2012). In our results, both mitochondrial and nuclear phylogenies showed clear geographic patterns, indicating that this mito-nuclear discordance more likely reflects mitochondrial introgression, possibly from Neotropical Eptesicus to the common ancestor of E. (Histiotus) if assuming the UCE phylogeny as the species tree. This historical introgression might have occurred during the rapid radiation of these South American clades or following secondary contact between sympatric species complex (Laine et al. 2023), during which time incomplete species boundaries might be occasionally broken by hybridization. Similar mitochondrial introgression events have also been suggested in the evolution of Old World Eptesicus species (Juste et al. 2013; Artyushin et al. 2018).

Moreover, E. (Histiotus) monophyly was supported in the UCE phylogeny but the mitochondrial phylogenies indicated either monophyly or paraphyly, regarding the position of E. (H.) magellanicus and E. (Histiotus) colombiae. This mito-nuclear discordance could also reflect historical mitochondrial introgression or ILS-caused polytomy. Future studies are needed to specifically test if and how potential ILS might have contributed to the detected mito-nuclear discordances and the species complex observed today. We acknowledge that our mitochondrial phylogeny might somewhat differ from a phylogeny constructed using complete mitogenome sequences, similar to what we found using the Myotis dataset (supplementary materials). Although mitochondrial regions and complete mitogenomes can generate different phylogenies (Meiklejohn et al. 2014; Kimball et al. 2021), such discordances should be small and not affect the major clade arrangements because the mitochondrial genes essentially behave like one linked locus (Platt et al. 2018; Laine et al. 2023), which is also why mitochondrial genes have been successfully used to estimate phylogeny for decades.

In addition to improving our understanding of the evolutionary history of serotine bats, our findings also help shed light on their classification. The interspecies genetic admixture and extensive mito-nuclear discordances involving New World Eptesicus and Histiotus clades resemble previous findings reported for bats and other mammals at the intrageneric level (e.g. Berthier et al. 2006; Larsen et al. 2010; Ge et al. 2022) and thus support a subgenus-level divergence in these groups rather than the full genus status as suggested in Cláudio et al. (2023). This highlights the need for broader genomic studies to elucidate the hybridization scenario in Eptesicus and to better inform their classification.

Our study largely enriched the sequence archive of E. (Histiotus) species by annotating the UCE libraries and sequencing new samples including for the first time the recently described E. (H.) diaphanopterus. Based on our most taxa-complete E. (Histiotus) phylogeny (Fig. 3), we provide important advances in the understanding of their evolutionary history and species classification. The updated species identities of all analyzed E. (Histiotus) samples are available in Table S5 (and Table S1 if from the Eptesicus UCE libraries).

Fig. 3
figure 3

Comparison of E. (Histiotus) in the maximum likelihood tree topologies generated using different molecular markers. Phylogenetic trees with branch lengths are shown in Fig. S4. The new samples collected in this study are labeled with asterisks. Tips are labeled based on our species assignment (Table S5). A) The nuclear UCE phylogeny from Yi and Latch (2022). B) The concatenated mitochondrial phylogeny using all annotated genes. C) The CYTB tree including NCBI sequences. D) The COI tree including NCBI sequences. E) Geographic distribution of the E. (Histiotus) samples analyzed in this study. Species ranges (not available for E. (H.) mochica, E. (H.) sp1, and E. (H.) sp2) follow Handley and Gardner (2008), Feijó et al. (2015), Rodríguez-Posada et al. (2021), Velazco et al. (2021), Marsh et al. (2022), and this study. Colors and taxonomic names are consistent with those in the phylogenetic trees. Note that E. (Histiotus) colombiae samples were not shown in these phylogenies because they were clustered within the Neotropical Eptesicus clade (Fig. S2, S4). The sample labeled as Eptesicus sp_Peru in the tree represents a Neotropical Eptesicus (see text for details) and is not included in the map

E. (Histiotus) colombiae

Individuals of E. (Histiotus) colombiae were nested in the Neotropical Eptesicus clade in all mitochondrial analyses (Fig. 2; tips labeled as E. (Histiotus) montanus colombiae in Figs. S2; S4). E. (Histiotus) colombiae inhabits the Andes of Colombia and Ecuador and was recently recognized as a valid species (Rodríguez-Posada et al. 2021). Misidentifications in all these cases seem unlikely because of the unequivocal morphological differences between E. (Histiotus) and Neotropical Eptesicus clades (i.e., enlarged ears in the former). Alternatively, assuming monophyly of E. (Histiotus) in the species tree, whole mitogenome replacement might explain this result, as observed in other mammalian groups (Ge et al. 2022). This scenario would indicate the first evidence of interspecific hybridization and introgression across subgenera in serotine bats. Future studies using nuclear data are required to fully delineate the phylogenetic position and evolutionary history of this species.

Magellanicus, montanus, and macrotus

Our mitogenome phylogeny supported the species status and the previously proposed early divergence of E. (H.) magellanicus (Díaz et al. 2019; Giménez et al. 2019; Laine et al. 2023). This species has the most southern distribution in the Andean region, which might have facilitated its early divergence and subsequent isolation from other E. (Histiotus) species. However, the gene trees of CYTB and COI clustered E. (H.) magellanicus within E. (H.) montanus (Fig. 3C, D), which is concordant with previous findings (Laine et al. 2023) and might indicate introgression between the two species that have largely overlapping distributions in southern South America.

It is also noteworthy that the polyphyly found in both montanus and macrotus may indicate a still hidden diversity in these groups despite that they both underwent marked changes in taxonomy in the last decade. Interestingly, in the CYTB gene tree, the two museum specimens classified as E. (Histiotus) montanus montanus collected from Uruguay (USNM:548,682, AMNH:M-183876) did not cluster with the NCBI sequences of E. (Histiotus) montanus collected near the species type locality in Chile. This suggests that these specimens from Uruguay might represent an unrecognized new species that needs to be further evaluated using complementary data (e.g., nuclear phylogeny, morphology, ecology, echolocation signals). In addition, one specimen collected from Arequipa, southern Peru (FMNH 50780) classified as E. (H.) montanus inambarus (Handley and Gardner 2008) did not cluster with the NCBI sequences of E. (Histiotus) montanus, indicating another cryptic lineage. Rodríguez-Posada et al. (2021) mentioned a “particular” cranial morphology of the holotype of inambarus, but a thorough morphological comparison is required to confirm whether E. (H.) m. inambarus should be elevated to species rank.

Mochica, humboldti, and laephotis

The gene trees and mitogenome phylogenies consistently recovered a well-supported clade comprising the central and northern Andean E. (Histiotus) species (except E. (H.) colombiae discussed above). This might suggest a unique colonization event followed by marked diversification. The four taxa (E. (H.) mochica, E. (H.) humboldti, E. (H.) laephotis, and E. (H.) montanus inambarus discussed above) exhibit mostly non-overlapping distributions associated with distinct Andean habitats (Fig. 3E). It thus seems plausible that northward dispersal along the Andes allowed ancestral E. (Histiotus) populations to colonize heterogeneous environments, such as the lowland wooded savanna of Pacific coastal northern Peru, Paramo grasslands, and Andean montane forests (Rodríguez-Posada et al. 2021; Velazco et al. 2021), leading to divergent selection. Additional studies are needed to further explore the phylogeographic history of this group.

Among the northern Andean species, we report E. (H.) mochica from an elevation of 3,160 m in the Andes of Peru with identical or similar (12-basepair difference) DNA sequences for COI and CYTB, respectively, to specimens from the type locality in the Pacific coastal region of northwestern Peru. This reveals a much broader environmental niche for this recently described species (Velazco et al. 2021). The mitogenome phylogeny recovered E. (H.) mochica as a sister taxon to the E. (H.) montanus inambarus sample from Peru. The mitochondrial gene trees depict a different arrangement where E. (H.) mochica is sister to E. (H.) humboldti, both forming the sister clade to E. (H.) m. inambarus, which is in line with the CYTB phylogeny recovered in previous studies (Rodríguez-Posada et al. 2021). The nuclear UCE phylogeny showed another distinct arrangement by placing E. (H.) mochica as the sister to a clade comprising E. (H.) m. inambarus, E. (H.) laephotis, and E. (H.) humboldti. These tree discordances indicate more potential cases of introgression in the Histiotus clade or effects of ILS. Unlike the northern Andean species, E. (H.) laephotis, which occurs mostly on the eastern slope of the Andes from Peru to northern Argentina, has an early divergence within this group at both gene levels and mitogenome scales.

Velatus, diaphanopterus, and unnamed species

The last group of Histiotus bats comprises those with triangular-shaped ears distributed in central and eastern South America. E. (H.) velatus has received relatively little taxonomic attention although it has one of the widest distributions occurring from the Atlantic Forest of southeastern Brazil to the eastern flank of the Peruvian Andes (Fig. 3E). Surprisingly, the only available E. (H.) velatus individual from Peru was recovered as a sister clade to E. (H.) macrotus in mitochondrial phylogenies, and sister to E. (H.) mochica in the UCE phylogeny (Yi and Latch 2022; Fig. 3). It is thus an open question whether this Peruvian individual represents a true E. (H.) velatus whose type locality is in the Brazilian Atlantic Forest. Additional sampling throughout its distribution is essential to shed light on the species classification and evolutionary history. On the other hand, the newly generated sequences of E. (H.) diaphanopterus, with samples obtained from its type locality, provide strong molecular support for its species status (Feijó et al. 2015). The phylogenetic position of E. (H.) diaphanopterus, nevertheless, remains uncertain as CYTB and COI trees yielded distinct arrangements. Moreover, we uncovered a novel lineage of triangular-shaped ears from Brazil that might also represent a new species in the Histiotus clade. Future studies using complementary evidence (e.g., nuclear sequences, morphology, ecology, echolocation signals) are necessary to evaluate the classification of these cryptic lineages and characterize the hidden diversity in this group.

Conclusion

We showed that MitoFinder annotations of UCE libraries can provide a valuable resource for estimating mitochondrial phylogeny and detecting mito-nuclear discordances that help build testable hypotheses about species evolution. By re-analyzing previously published Eptesicus UCE libraries, we detected pervasive mito-nuclear discordances that suggest historical mitochondrial introgression in serotine bats both within and across subgenera, such as between the ancestor of E. (Histiotus) and Neotropical Eptesicus. In addition, by incorporating published and newly collected gene sequences, we generated the most taxa-complete phylogeny for E. (Histiotus), largely enriching its sequence archive, and uncovered cryptic lineages from Peru, Uruguay, and Brazil. Future studies integrating complementary data (e.g., nuclear sequences, morphology, ecology, echolocation signals) are required to properly evaluate their taxonomic status. Our study thus provides a foundation for future research on the diversity and evolution of E. (Histiotus) bats and highlights the importance of biodiversity conservation in South America.