Whole genome sequences of M. rufoniger

The genomic DNA from the wild carcass of M. rufoniger found in Gosudonggul cave, Danyang, in South Korea, was sequenced using Illumina HiSeq2000 platform. A total of 124 Gb of paired-end short DNA sequences were produced with a read length of 100 bp, and target insert sizes of 566 bp and 574 bp from two genomic libraries. After reducing low sequencing quality reads and possible microbial contaminated reads, we acquired a total of 115 Gb of DNA sequences (Table 1; S1 Table). To confirm the species identification of the sample, a phylogenetic analysis was performed using a multiple sequence alignment of mitochondrial cytochrome b sequences, and our sample was verified to be the closest to M. rufoniger (S1 Fig). We performed a K-mer analysis (K = 17) using the M. rufoniger whole genome sequences, and its genome size was predicted to be approximately 1.88 Gb (S2 Fig; S2 Table). This is similar to those of other bat species and smaller than those of other mammals [3,4,12]. Since there is no de novo assembled M. rufoniger genome, the DNA reads were aligned to all three available Myotis bat reference genomes (M. brandtii, M. davidii, and M. lucifugus; Table 1; S3 Table). The consensus DNA sequences of M. rufoniger were generated by substituting M. rufoniger single nucleotide variants (SNVs) detected from the whole genome sequencing data against the three reference genomes. The genome coverage (≥10×, 87.63%) was the highest when the M. lucifugus assembly was the reference, while the coding sequence (CDS) coverage (≥10×, 96.60%) was the highest when the M. davidii genome was the reference. The number of SNVs (58,660,193) was the lowest when the M. brandtii was the reference, while the number of small insertions and deletions was the lowest when the M. daividii was the reference. However, when the M. brandtii assembly was the reference, the mapping rate (96.50%) and the number of consensus genes (17,247) was the highest. Therefore, we used M. brandtii reference-guided M. rufoniger consensus sequences for following analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sequencing and mapping statistics of M. rufoniger genome.

https://doi.org/10.1371/journal.pone.0180418.t001

We examined the base composition of M. rufoniger genome (S4 Table) and found that the ratio of GC content was higher in the CDS region (52.15 ~ 52.95%) than that of the whole genome (41.49 ~ 42.28%), and this is concordant with the other mammalian species (the GC content ratios of the other species were 48.98 ~ 54.35% in the CDS region and 37.82 ~ 42.93% in the whole genome; S5 Table). The proportion of known transposon-derived repeats in the M. rufoniger genome was 19.05%, which is comparable with those of other bats (19.34 ~ 33.32%), but significantly lower than those of other mammalian genomes (25.27 ~ 51.62%; S6 Table).

The body of the M. rufoniger carcass used in this study was damaged and its morphological sex determination was not possible. We therefore compared its heterozygosity in X-chromosomal scaffolds with that of the other male Myotis bat (S7 Table). Since mammalian males have both X and Y sex chromosomes while females have two X chromosomes, a male individual has to show a lower heterozygosity in X chromosome than that of a female individual. Our sample showed an even lower heterozygosity in X-chromosomal scaffolds than that of the male M. davidii individual, indicating that it is clearly a male.

Relationship of M. rufoniger to other mammalian species

We identified orthologous gene clusters from 13 mammalian genomes (seven bat genomes: Myotis brandtii, Myotis lucifugus, Myotis davidii, Eptesicus fuscus, Pteropus alecto, Pteropus vampyrus, and Rousettus aegyptiacus, and six other mammalian genomes: Homo sapiens, Mus musculus, Bos taurus, Equus caballus, Heterocephalus glaber, and Monodelphis domestica; S2 Table) using OrthoMCL software [13]. The M. rufoniger genes were matched and added to the orthologous clusters of M. brandtii’s genes, and 6,782 single-copy gene families among the 14 species were identified [7].

The divergence times of M. rufoniger and the 13 mammals were estimated using 1,258,141 four-fold degenerate sites from the 6,782 single-copy genes (Fig 1). The closest clade to bats was E. caballus (horse) and B. taurus (cow) clade (diverged at 79.45 million years ago [MYA]). The Myotis bats diverged into its current clade approximately 11.52 MYA. M. davidii was estimated to be the closest to the M. rufoniger among the three Myotis bats used in this study. The divergence time between M. rufoniger and M. davidii was estimated to be 10.41 million years, indicating a fairly recent divergence which is suitable for our close species comparative genomics analyses.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Phylogenetic relationships and divergence times in bats and mammalian species.

The estimated divergence time (million years ago; MYA) is given at the nodes, with the 95% confidence intervals in parentheses. The calibration times of M. brandtii—H. sapiens (97.5 MYA), and M. brandtii—P. alecto (62.6 MYA) were derived from the TimeTree database. Colored branches and circles represent bat groups (blue: insect-eating microbat, red: fruits-eating mega bat). M. domestica, used as an outgroup species, was excluded in this figure.

https://doi.org/10.1371/journal.pone.0180418.g001

M. rufoniger specific amino acid sequences

From previous studies [3,4], bat-specific amino acid sequences within FSHB, GHR, IGF1R, TP53, MDM2, SLC45A2, RGS7BP, RHO, OPN1SW, and CNGB3 genes were reported. They represent some general characteristics of Myotis bats: delayed ovulation (FHSB), long lifespan (GHR and IGF1R), powered flight (TP53 and MDM2), echolocation (SLC45A2 and RGS7B), and low vision (RHO, OPN1SW, and CNGB3). The same amino acid sequences within the above listed genes were also identified in the M. rufoniger genome (S3 Fig).

To further identify bat-specific amino acid sequences linked to environmental adaptations and unique evolutionary features, we investigated M. rufoniger specific amino acid changes compared to other Myotis bats. A total of 3,366 unique amino acid changes (uAACs) from 2,125 genes were identified in M. rufoniger. Among them, 1,696 uAACs from 1,228 genes were predicted to be function altering by the protein variation effect analyzer (PROVEAN) software (variant score ≤ -2.5; S8 Table) [14]. These function-altering variants and associated genes can provide insights onto M. rufoniger specific evolutionary adaptation. When gap-containing genes were excluded, 1,239 uAACs from 929 genes were predicted to be function altering (S9 Table). A functional enrichment analysis of the 929 genes having function-altering uAACs was performed using DAVID (Database for Annotation Visualization and Integrated Discovery) tool [15]. The probably function-altered genes were significantly enriched in reproduction related terms (Gene Ontology [GO] analysis with a P–value ≤ 0.05 by EASE scores [modified Fisher’s exact test] and with a 10% of false discovery rate [FDR]) including reproductive processes in a multicellular organism (P-value: 0.00026, 39 genes, GO: 0048609), ovulation cycle (P-value: 0.00052, 11 genes, GO: 0042698), and gamete generation (P-value: 0.0017, 31 genes, GO: 0007276; S10 Table). In M. rufoniger, pigment related terms were significantly enriched in the function-altered genes as in the melanin biosynthetic process (P-value: 0.0037, 4 genes, GO: 0042438; S10 Table). The genes are DCT (Dopachrome tautomerase), SLC45A2 (Solute carrier family 45), TYRP1 (Tyrosinase-related protein 1), and OCA2 (Oculocutaneous albinism II). We could identify some uAACs in DCT, SLC45A2, TYRP1, and OCA2 genes in the other Myotis bats as function-altering using PROVEAN software (Table 2; S11 Table) [14] indicating those genes are not only specific to M. rufoniger’s red fur color. The multiple sequence alignments and specific amino acid sequences are presented in S4 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Myotis bats’ uAACs within melanin associated genes.

https://doi.org/10.1371/journal.pone.0180418.t002

As we were interested in the variants’ whole gene function, all the PROVEAN variant scores of each gene were summed up. We then ranked the sums of the function-altered gene candidates (S12 Table). For the top 20 genes, the same analysis was carried out in the other Myotis bat species for comparison (Table 3; S13 and S14 Tables). We ranked M. rufoniger’s genes according to the numbers of variants and summed variant scores, and compared them to those of other Myotis bat genes. From the result, N6AMT1 (N-6 Adenine-Specific DNA Methyltransferase 1) showed the fourth-lowest sum of variants scores (sum of PROVEAN variant scores: -40.285 and number of variants: six in M. rufoniger; no uAAC in the other Myotis bats), indicating a high degree of function alteration of N6AMT1 in M. rufoniger. The multiple sequence alignments and specific amino acid sequences are presented in S5 Fig.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. The summed PROVEAN variant score for top 20 ranked genes. The lower the more significant.

https://doi.org/10.1371/journal.pone.0180418.t003

Demographic history and the genetic diversity of Myotis bats

Deeply sequenced genomes allow the estimation of population structure history [10,11]. To investigate the demographic history of Myotis bats, a pairwise sequentially Markovian coalescent (PSMC) model inference analysis was conducted [11]. M. rufoniger was estimated as the most flourished species around 30 ~ 300 k years ago compared to the other Myotis bats, and its demographic history showed that its population peak was 50 k years ago (Fig 2). However, its effective population size was dramatically decreased during the last glacial period (10 ~ 50 k years ago), and it was estimated to be the lowest in present. Also, a consistent decline in the effective population size of Myotis bats since ~30 k years ago was found (Fig 2). We further investigated the genomic diversity (which can be affected by population size) of the M. rufoniger and compared it to those of the other Myotis bats. The M. rufoniger genetic diversity (0.00391), based on the heterozygous SNV rate, was lower than those of the M. davidii (0.00471) and the M. brandtii (0.00614), confirming the M. rufoniger’s low effective population size (S13 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Demographic history of Myotis bats.

T_surf, atmospheric surface air temperature (T indicates temperature, surf indicates surface); RSL, relative sea level; 10 m.s.l.e., 10 m sea level equivalent; g, generation time (years); μ, mutation rate per base pair per year.

https://doi.org/10.1371/journal.pone.0180418.g002

Sample and genome sequencing

The partially decomposed carcass of M. rufoniger was found in August 2012 at Gosudonggul cave in Danyang, South Korea (coordinates: 36° 59' 18.67'' N, 128° 22' 53.26'' E; elevation: 180.65 m) by Gosudonggul cave management office staff. The M. rufoniger was moved to the Natural Heritage Center of the Cultural Heritage Administration (South Korea) and kept frozen (Keeping temperature: -70°C). We acquired the M. rufoniger sample from the Natural Heritage Center under the Cultural Heritage Administration permit in June 2013. The wing membrane tissue, which seemed not decayed, was acquired for genomic DNA sequencing (S1 Picture).

Its genomic DNA was sequenced using Illumina HiSeq2000 platform, with a read length of 100 bp and insert sizes of 566 and 574 bp from two genomic libraries. The DNA was extracted using a Tiangen Micro DNA Kit. The amount of DNA was quantified by fluorometry using a UV spectrophotometer (Tecan F200). A partial DNA degradation was identified by electrophoresis (S2 Picture), but it was judged good resulting in no significant sequencing quality problems. The genomic DNA was sheared to approximately 566 bp and 574 bp using a Covaris S2 Ultrasonicator, and then used in the preparation of whole-genome shotgun libraries according to Illumina’s library preparation protocols. The efficacy of each step of the library construction was ascertained using a 2100 Bioanalyzer (S6 Fig). The final dilutions of the two libraries were then sequenced using a HiSeq2000 sequencer with the TruSeq PE Cluster kit v3-cbot HS and TruSeq SBS kit v3-HS for 200 cycles.

Sequencing read filtering

A DNA read was filtered out when the read’ Q20 base content was lower than 70%, using IlluQCPRLL.pl script of NGSQCToolkit (ver 2.3.3; S1 Table) [29]. We additionally downloaded genome data of the other bats (Myotis brandtii, Myotis davidii, Myotis lucifugus, Pteropus Alecto, and Pteropus vampyrus) from the NCBI database. Low quality DNA reads from the other bats were also filtered out using the same method (S1 Table). The possible microbial contaminated DNA reads were filtered out when the reads’ alignment scores to microbial (bacteria and fungi) genomes were higher than that to the Myotis bat genomes (M. brandtii, M. davidii, and M. lucifugus; S3 Table). We used microbial genomes that were downloaded from Ensembl database and the three Myotis bat genome assemblies were from NCBI. The DNA reads were mapped to the microbial and Myotis bat genomes using BWA-MEM algorithm of BWA (version 0.7.15) with Mark shorter split hits as a secondary option (-M) [30]. The rmdup command of SAMtools (ver 0.1.19) was used to remove PCR duplicates [31]. The possible microbial contaminated reads in the other Myotis bats were also filtered out by using the same method but mapping was to their own genome assemblies (S1 Table).

Species identification

To construct the phylogenetic tree and identify the sample’s species, the mitochondrial cytochrome b consensus sequence was generated by mapping the sequencing data to the M. rufoniger mitochondrial cytochrome b sequence (GenBank accession number: HQ184048.1) using BWA-MEM algorithm of BWA (version 0.7.15) with Mark shorter split hits as a secondary option (–M) [30]. The reads were realigned using the GATK (version 2.5–2) RealignerTargetCreator and IndelRealigner algorithms to minimize the read alignment mismatches [32]. The rmdup command of SAMtools (ver 0.1.19) was used to remove PCR duplicates [31]. The variants from the whole genome sequencing data against the reference were called using mpileup command of SAMtools (version 0.1.19) with–E option to minimize the noise, -A option to use regardless of insert size constraint, -q 20 to consider only high quality mapping reads, and–Q 30 to consider only high quality bases [31]. The called variants were filtered using vcfutils.pl varFilter command of SAMtools (ver 0.1.19) with -d 8 option (minimal depth of 8) [31]. SNVs detected from the whole genome sequencing data against the reference were substituted to construct mitochondrial cytochrome b consensus sequence of the M. rufoniger using vcfutils.pl vcf2fq command of SAMtools (ver 0.1.19) with -l 5 option (no indel present within a 5 bp window) [31]. A multiple sequence alignment of the mitochondrial cytochrome b sequences was conducted using MUSCLE (version 3.8.31) program with default options [2,33]. A phylogenetic analysis was conducted using MEGA 7.0 program [34]. The phylogenetic tree was inferred by using a Maximum Likelihood method based on the Tamura-Nei model with 1,000 replicates bootstrapping [35,36]

Genome size estimation

For a K-mer analysis, KmerFreq_HA command of SOAPec program in SOAPdenovo2 package was used with a K = 17 option [37]. The genome size was estimated from the number of K-mers (depth > 3) divided by the peak depth of the K-mer graph.

Mapping sequence data

The filtered sequenced reads were mapped to the three Myotis bat genome assemblies (M. brandtii, M. davidii, and M. lucifugus) using BWA-MEM algorithm of BWA (version 0.7.15) with Mark shorter split hits as a secondary option (-M) [30]. The reads were realigned using the GATK (version 2.5–2) RealignerTargetCreator and IndelRealigner algorithms to minimize the read alignment mismatches [32]. The rmdup command of SAMtools (ver 0.1.19) was used to remove PCR duplicates [31]. The variants of M. rufoniger were called using mpileup command of SAMtools with–E, -q 20, -A, and–Q 30 options [31]. The called variants were filtered using vcfutils.pl varFilter command of SAMtools (ver 0.1.19) with -d 8 and -D 250 options (minimal depth of 8 and maximal depth of 250) [31]. The DNA consensus sequences of M. rufoniger were generated by substituting the M. rufoniger SNVs detected from the whole genome sequencing data to each reference bat genome using the vcfutils.pl vcf2fq command of SAMtools (ver 0.1.19) with -l 5 option [31].

Repeat annotation

RepeatMasker (version 4.0.5) was used to identify transposable elements by aligning the M. rufoniger consensus genome sequence against a library of M. brandtii with default option [38]. For the comparison of the M. rufoniger’s repeat annotation with the other species, the same method was used for other genome references.

Sex determination

Male M. davidii reads were mapped to the M. brandtii (a male sample) genome assembly using BWA-MEM algorithm of BWA 0.7.15 version [30] with–M option. The reads were realigned using the GATK (version 2.5–2) RealignerTargetCreator and IndelRealigner algorithms to minimize the read mismatches [32]. The rmdup command of SAMtools (ver 0.1.19) was used to remove PCR duplicates [31]. Variants were called using mpileup command of SAMtools with–E, -q 20, -A, and–Q 30 options [31]. The called variants were filtered using vcfutils.pl varFilter command of SAMtools (ver 0.1.19) with -d 8 and -D 250 options [31]. To identify X-chromosomal scaffolds from the M. brandtii genome assembly, we conducted BLASTn (Blast 2.2.26) with an E-value cutoff of 1E-6 using X-chromosomal CDS sequences of human as queries [39]. Scaffolds having greater than or equal to 10 top hits were considered as X-chromosomal scaffolds.

Comparative evolutionary analysis

Orthologous gene families between M. rufoniger’s sequence and 13 mammalian genomes (Seven bat genomes: M. brandtii, M. lucifugus, M. davidii, E. fuscus, P. alecto, P. vampyrus, and R. aegyptiacus, and six other mammals: H. sapiens, M. musculus, B. taurus, E. caballus, H. glaber, and M. domestica; S2 Table) were identified using OrthoMCL software (version 2.0.9) by matching and adding the M. rufoniger genes to orthologous clusters of M. brandtii’s genes, and 6,782 single-copy gene families among the 14 species were collected [7,13]. M. brandtii, M. lucifugus, M. davidii, E. fuscus, P. alecto, P. vampyrus, R. aegyptiacus, H. sapiens, M. musculus, B. taurus, E. caballus, H. glaber, and M. domestica genomes and gene sets were downloaded from the NCBI database (S3 Table).

To construct a phylogenetic tree and estimate the divergence time of bats and other mammals, 1,258,141 four-fold degenerate sites from the 6,782 single-copy gene families were used to conduct the multiple sequence alignment (MSA) using MUSCLE (version 3.8.31) program [33]. The phylogenetic tree was constructed among the 14 species using a Randomized Axelerated Maximum Likelihood (RAxML) program (version 8.2) [40]. In RAxML program, GTR gamma model was used as the nucleotide substitution model. In order to check the branch reliability, 1,000 rapid bootstrapping was used. The M. domestica was used as an outgroup species. Based on this phylogenetic tree topology, the divergence time was estimated using Reltime-ML in MEGA 7.0 program [34]. In this process, the divergence time at the node between M. brandtii–H. sapiens was constrained to be 97.5 MYA, and M. brandtii–P. alecto was constrained to be 62.6 MYA based on the TimeTree database [41].

Amino acid changes were identified by constructing MSAs among the 6,782 single copy gene families using Clustal Omega (version 1.2.4) [42]. Function-altering amino acid changes were predicted using PROVEAN (version 1.1.5) [14]. The MSAs of the DCT, SLC45A2, TYRP1, OCA2, and N6AMT1 genes were manually checked, and uAACs within misaligned regions were excluded in the variant score analysis. Only homozygous amino acid variants were considered in the amino acid sequence comparison analysis to reduce bias from individual-specific variants.

Demographic history and genetic diversity

The demographic history of the Myotis bats was estimated using a pairwise sequential Markovian coalescent (PSMC) program [11]. We mapped the downloaded genome sequencing data of Myotis bats to the M. lucifugus genome assembly using BWA-MEM algorithm of BWA 0.7.15 version [30] with–M option. The reads were realigned using the GATK (version 2.5–2) RealignerTargetCreator and IndelRealigner algorithms to minimize the read mismatches [32]. The rmdup command of SAMtools (ver 0.1.19) was used to remove PCR duplicates [31]. SAMtools was used to extract diploid genome information from the BAM files [31]. Options of -N25 -t15 -r -p “4+25*2+4+6” were used for the PSMC analysis. The generation time for the Myotis bats was estimated as sum of average maturation and gestation time (S16 Table) [43,44]. The mutation rate of Myotis bats was estimated by multiplying reported neutral mutation rate for mammals (2.2 × 10⁻⁹ per base pair per year) by the generation time (1.2 year) [45]. Atmospheric surface air temperature and global relative sea level data of the past three million years were obtained and used for this analysis [46]. The genomic diversity was calculated by dividing the number of heterozygous SNVs by its genome size (bp) [8]. For the genomic diversity calculation, variants were called using mpileup command of SAMtools with–E, -q 20, -A, and–Q 30 options [31]. The called variants were filtered using vcfutils.pl varFilter command of SAMtools (ver 0.1.19) with -d 8 and -D 250 options [31].