Phylogeographic analysis of long-legged bats, Macrophyllum macrophyllum, with notes on roosting behavior and natural history

Specimen sampling and literature review

Bats were captured from their daytime roosts near rivers or creeks at three sites located in the municipalities of Cuiabá in the Cerrado and Santo Antônio do Leverger in the Pantanal, both in the Brazilian state of Mato Grosso (Table 1). Sampling occurred between February 2021 and December 2023. To capture the bats, mist nets were placed around the entrances of their shelters, and as the animals were disturbed by human presence, they flew and became entangled in the nets. Collecting sites are in the Cerrado and Pantanal ecoregions sensu Olson et al. (2001) from elevations below 150 m above sea level. Climate in both areas is classified as “Aw”, or “Tropical with dry winter”, following Köppen’s classification (Alvares et al., 2013). Annual variations in temperature and rainfall are characterized by being markedly seasonal, with a dry season from April to September and a rainy season from September to March.

We reviewed all known published records of the species by searching for the term “Macrophyllum macrophyllum” on Google Scholar and examining the literature cited in the species accounts published by Harrison (1975) and Williams & Genoways (2008). We extracted information on the types of roosts used, colony sizes, and other species inhabiting the same roosts as M. macrophyllum. Roost descriptions that were too vague or general were excluded from our compilation. For instance, Wied (1826) mentioned that the species “spends the day sitting on the rocks and old trunks of the forest” but gave no additional detail. We also ensured that duplicate records were not included. For example, the roosting record reported by Voss et al. (2016) and Velazco et al. (2021) refers to the same observation, and we cited only the earliest reference.

Permits to collect the animals

The capture and handling of specimens complied with the current guidelines for the use of wild mammals in research of the American Society of Mammalogists (Sikes & ASM Animal Care and Use Committee of the American Society of Mammalogists, 2016). In this context, we adhered to the 3R principle—Replacement, Reduction, and Refinement—to optimize ethical and methodological standards (Field et al., 2019). This research was approved by the Animal Care and Use Committee of the Universidade Federal de Viçosa (CEUA-UFV process number 06/2025). Permits for specimen collection were granted by the Instituto Chico Mendes de Conservação (SISBIO permits 76825, 77787) under the Ministry of the Environment, Brazil. The specimens were initially fixed in formalin and preserved in 70% ethanol and are now housed in the Brazilian collections of the Museu de Zoologia João Moojen (MZUFV) in Viçosa and the Coleção de Mamíferos da Universidade Federal de Mato Grosso (UFMT) in Cuiabá. Tissue samples (liver and pectoral muscles) were also collected and preserved in 90% ethanol.

DNA extraction, amplification, and sequencing

Fresh tissue of a total of 32 specimens preserved in ethanol and liquid nitrogen were included in the molecular sampling (Table 1). Voucher specimens are deposited in the following collections: Carnegie Museum of Natural History (CMNH), USA; Centro de Coleções Taxonômicas da Universidade Federal de Minas Gerais (UFMG), Brazil; Escola Superior de Agricultura Luiz de Queiroz, Universidade de São Paulo (LMUSP), Brazil; Instituto de Desenvolvimento Sustentável Mamirauá (IDSM), Brazil; Instituto Nacional de Pesquisas da Amazônia (INPA), Brazil; MZUFV; Royal Ontario Museum (ROM), Canada; UFMT; and Universidade Federal da Paraíba (UFPB). Specimen sampling covers most of the distribution of the genus (Fig. 1).

Genomic DNA was extracted from 21 specimens from Brazil according to the protocol described by Aljanabi & Martinez (1997), with only slight modifications. Polymerase chain reactions (PCR) used MVZ05 and MVZ16 primers (Smith & Patton, 1993) for amplification of approximately 800 basepairs (bp) of the mitochondrial Cytochrome b gene (cytb), following the protocol of Saldanha et al. (2019). The PCR purification, preparation and sequencing were outsourced to the “Biotecnologia, Pesquisa e Inovação - BPI, São Paulo, Brazil”. Sequences of both forward and reverse directions were aligned to assemble the consensus in the software Geneious v. 7.1.3 (Kearse et al., 2012) and deposited in GenBank (Table 1). Additionally, 11 specimens at ROM from Ecuador, Guyana, Panama, and Suriname were sequenced for 1,140 bp of cytb following the protocol of Lim et al. (2008). DNA extraction was done using a phenol-chloroform procedure, PCR amplification used primers LGL765 and LGL766 Bickham, Wood & Patton (1995), and nucleotide sequencing was performed on an Applied Biosystems 3730 analyzer in the Laboratory of Molecular Systematics at ROM. Sequences were verified and aligned using Sequencher v. 4.8 (Gene Codes Corporation, Ann Arbor, MI, USA).

Phylogenetic analyses

All sequences were aligned using the MUSCLE algorithm’s default settings (Edgar, 2004). The final character matrix used in the phylogenetic analyses comprised 40 taxa. In addition to sequences from 33 individuals of Macrophyllum (Table 1), we included seven sequences of the following closely related species of phyllostomid bats (GenBank accession numbers in parentheses): Trachops cirrhosus (FJ155483, MH102398, MH102399), Lophostoma silvicola (JF923851), Phyllostomus hastatus (FJ155479), Chrotopterus auritus (FJ155481), and Vampyrum spectrum (FJ155482).

Phylogenetic trees for the cytb gene were inferred using maximum likelihood (ML) and Bayesian inference (BI) approaches. The ML analysis done in IQ-TREE version 1.6.12 (Nguyen et al., 2015), with ultrafast bootstrap values calculated on the consensus tree based on 1,000 replicates (Hoang et al., 2017). The BI analysis was conducted in MrBayes 3.2.7a via the online CIPRES platform (Miller, Pfeiffer & Schwartz, 2010; Ronquist et al., 2012). Four Markov Chain Monte Carlo (MCMC) chains were run for 40,000,000 generations, sampling every 4,000 generations. A burn-in of 30% was discarded. Convergence was verified using Tracer 1.7.1 (Rambaut et al., 2018), by evaluating effective sample sizes (ESS > 200) and trace plots to ensure proper mixing and stationarity of MCMC chains. Support was assessed using posterior probabilities. In both BI and ML analyses, we divided our data into three partitions (1st, 2nd, and 3rd codon positions) and estimated the best substitution scheme using IQ-TREE. We used the “-mset mrbayes” command in IQ-TREE, so the model chosen would be appliable to MrBayes. The chosen models for each partition were SYM+G4 (1st codon), HKY+F+G4 (2nd codon), GTR+F+I+G4 (3rd codon).

Population analyses

We examined genetic variation in the sample of 33 specimens of M. macrophyllum. Due to the missing data in some shorter sequences, we trimmed them to 801 base pairs for the population analyses. To verify if there is geographic structuring among the main haplogroups, we divided our samples using three distinct approaches (Table 1). One geographic approach considered hydrographic basins, following the classification proposed in the HydroSheds database (Lehner & Grill, 2013), and resulted in the following five groups: Amazon, La Plata, Northeast South America, Pacific, and São Francisco. In the second approach, individuals were grouped according to the ecosystem predominant for the locality, resulting in five groups: Amazonia, Central American forests, Cerrado, Guianas Shield, and Pantanal. For the third approach, we used a Bayesian hierarchical clustering (BHC) analysis to determine the most likely number of genetic clusters (K) in our dataset (Tonkin-Hill et al., 2019). The BHC analysis was done using the “fastbaps” package in R, using k.init = 10 to obtain the best partition under a Dirichlet Process Mixture model (R Core Team, 2020). Four populations were defined in this method (Table 1), as follows: Pop1: Brazil (Cuiabá, Mato Grosso); Pop2: Brazil (Minas Gerais, Pará, Santo Antonio do Leverger. Mato Grosso), Guyana, Suriname, and Venezuela; Pop3: Brazil (Amazonas), Ecuador (Napo); and Pop4: Panama.

Molecular variance (AMOVA) was calculated to evaluate the distribution of genetic variability within and among the defined populations and groups, based on the three proposed approaches: hydrographic basins, ecosystems, and BHC. Additionally, overall and pairwise fixation indices (F_st) were computed for each of the three approaches, with statistical significance (P- values) determined through 1,000 permutations. Both AMOVA and F_st were performed using Arlequin 3.5 (Excoffier & Lischer, 2010). Pairwise sequence divergence among the populations recognized in the BHC were estimated with a Kimura 2-parameter model using the dist.dna function in “ape” R package (Paradis & Schliep, 2019).

A minimum spanning haplotype network was constructed using PopArt v.17, with epsilon set to zero (Bandelt, Forster & Röhl, 1999; Leigh & Bryant, 2015). To verify correlation between pairwise genetic distances and geographic distances between populations we performed a Mantel test (10,000 permutations) in R using the “ape”, “geodist” (Padgham & Sumner, 2021) and “vegan” packages (Oksanen et al., 2022).

Natural history

The literature review, together with our new data, yielded 73 publications and 134 records of the species that account for 32 observations of day roosts used by M. macrophyllum (Table S1). We divided the roosts into eight types (Table 2); most of the roosts (43.75%) were in drainage culverts, followed by caves (15.63%). Also, species of the genera Carollia and Glossophaga were typically found roosting with M. macrophyllum. Groups usually were composed of 3-8 individuals (Hill & Bown, 1963; Peracchi & Albuquerque, 1971; Simmons & Voss, 1998; Tavares & Anciães, 1998; Stutz et al., 2004; Voss et al., 2016) but larger colonies of approximately 50 were also recorded (Seymour & Dickerman, 1982; Peracchi, Raimundo & Tannure, 1984).

Of the 22 specimens obtained in Mato Grosso state, 14 were from a colony of approximately 30 recorded under a wooden fishing deck (WFD) on Rio São Lourenço during the rainy season in February 10th and 11th, 2021. The bats were hanging below the platform, a few centimeters above the water (Figs. 2A, 2B). Of these 14 specimens, 6 were males and 8 were pregnant females. Five specimens were sampled in a daytime roost in an expansion gap of a concrete bridge (GCB) on highway MT-040, crossing Rio Cuiabá Mirim, on October 20th, 2022 (Figs. 2C, 2D). Three specimens were caught in a drainage culvert made of concrete (DCC), on December 5th, 2022 (one specimen), and on December 26th, 2023 (two specimens). The culvert passed under a dirt road, and through it flowed a small creek (Figs. 2E, 2F). In both GCB and DCC, only males were captured. Seven of the eight females in the WFD roost were pregnant. Cohabitation was observed in two of the three roosts, with the species Rhynchonycteris naso and Carollia perspicillata recorded in GCB, and C. perspicillata and Glossophaga soricina in DCC (Table 2).

Phylogenetics

Mitochondrial DNA sequences of the cytb gene confirmed the monophyly of Macrophyllum relative to the outgroup taxa in other genera of phyllostomid bats, with maximum clade support values both in the BI and ML analyses. The BI phylogeny recovered five highly supported lineages within Macrophyllum, with posterior probability values ranging from 0.92 to 1 (Fig. 3A), although many of the more basal relationships were not well resolved. Specimens from Panamá were well supported (PP = 1) and sister to a clade from the Amazonian lowlands that had low support (PP = 0.38). Part of the western lowland Amazonian clade from Brazil is highly supported (PP = 0.92), as well as the eastern Ecuadorian clades (PP = 1). The clade including specimens from lowland Amazonia and Panamá is the sister group to a poorly supported (PP = 0.47) clade that consisted of all other South American samples including specimens from Cuiabá in the western Cerrado in a well-supported clade (PP = 1) sister to all remaining clades including the Brazilian Pantanal, Brazilian eastern Cerrado (Jequitaí), and the Guiana Shield. The Pantanal clade (specimens from Baia São João and Mimoso, Mato Grosso) was highly supported (PP = 0.96), as well as the Guianan clade (specimens from Guyana, Suriname, and Venezuela; PP = 0.99).

The ML-based phylogenetic inference tree had UltraFast bootstrap values between 63 and 100, but some of the more terminal relationships are unresolved (Fig. 3B). As in the BI tree, the Cuiabá clade had high support (BS = 100). The clade containing specimens from Pantanal was also highly supported (BS = 97), as well as the Guiana Shield clade (BS = 97). The western Amazonian clade, containing specimens from Ecuador and Amazonas, Brazil had lower support (BS = 88). Topology did not differ substantially between BI and ML, the primary difference being the position of the Cuiabá clade from western Cerrado, which is sister to the remaining specimens in the ML tree, whereas in the BI tree it is sister to a clade containing the eastern Cerrado, Pantanal, and Guianas (Fig. 3).

Genetic diversity and population structure

We identified 22 haplotypes from the aligned 801 base pair fragment of the cytb gene (Fig. 4). The haplotype network had a “reciprocally monophyletic” structure, in which multiple lineages are each connected by a long branch characterized by numerous mutations (Jenkins, Castilho & Stevens, 2018). No haplotypes were shared between different localities and the Mantel test indicated a significant correlation between geographic distance and genetic variation in the samples (R = 0.144; p = 0.047).

The geographic approach that considered four populations based on BHC had the highest F_st values, followed by the distinct ecosystems approach and the hydrographic basins had the lowest F_st (Table 3). In the BHC, the largest F_st values were observed between Pop1 (Cuiabá) and the other three populations (Fig. 5). The largest (but not statistically significant) difference was between Pop1 and Pop4 (Panama). Considering the distinct ecosystems, the largest F_st values were between Central America (Panama) and Pantanal specimens (Fig. 5). In the geographic approach that classified the populations into five hydrographic basins, the highest F_st values were observed between specimens from the Pacific coast (Panama) and São Francisco (Minas Gerais). Pair-wise genetic distances varied from 0.03, between Pop2 x Pop3 and Pop2 x Pop4, to 0.06, between Pop1 x Pop3 and Pop1 x Pop4 (Table 4).