Introduction

As members of the order Chiroptera, the second largest group (~25%) of mammalian species [1], more than 1,240 bat species are currently found worldwide, everywhere except the South and North Poles. Recent studies have successfully identified novel bat viruses on the basis of consensus primers or sequence-independent PCR amplification [2,3,4]. Over 80 virus species have been detected in bats, including several emerging human pathogens (i.e., severe acute respiratory syndrome coronaviruses, lyssaviruses, henipaviruses, Marburg virus and Ebola virus) [5]. Furthermore, some bat species live in close proximity to human habitation, highlighting the potential role of bats as reservoirs of zoonotic diseases [6].

Using polymerase chain reaction (PCR) screening with consensus primers, closely related viral species in bats have been detected [2,3,4]. However, the high assay specificity of these methods has limited the detection of divergent or unknown viruses. The advent of next-generation sequencing (NGS) technology has facilitated metagenomic analyses to simultaneously screen for all viral families. NGS has been successfully applied to detect novel viruses in different types of samples (including lungs, liver, brain, pharyngeal swabs and anal swabs) [7] and is expected to provide more in-depth information on previously unidentified viruses and viral taxonomy. Studies from the US [8, 9], France [10], New Zealand [11] and China [12,13,14,15] have analyzed the virome profiles of bats with NGS technology and have identified various unknown mammalian viruses. Most of these studies analyzed viral contigs by homogenizing different organs and tissues of bats; however, this approach is restricted by the unclear tissue distribution and route of transmission of bat viruses. The paucity of knowledge on bat-derived viruses warrants more extensive investigation and surveillance of different bat species in various geographic areas. The digestive tract is connected to the outer environment through ingestion of foods and liquids, providing a niche for virus transmission from bats to humans and other animals. Investigation of the virome in fecal samples from bats, and subsequent epidemiological and phylogenetic characterization, of any viruses, would significantly advance our understanding, especially with regards to prevention and management of zoonotic diseases in different geographic regions.

Papillomaviruses (PVs), rotavirus A (RVA), caliciviruses (CVs) and picornaviruses (PiVs) are widespread pathogens causing human diseases and have been detected in fecal samples of some bat species [16,17,18,19]. However, it is still unknown whether bats play a role as reservoirs in transmission of these viruses. More extensive studies of the prevalence and characteristics of these viruses isolated from different bat species and in different geographic areas is needed. To conduct this analysis, fecal specimens from six bat species, Cynopterus sphinx, Miniopterus schreibersii, Rousettus leschenaultii, Hipposideros larvatus, Rhinolophus blythi and Scotophilus kuhlii, were collected.

In this study, we report on the virome profile of fecal samples from six bat species in four cities in southern China, using metagenomic analysis, sequence-independent amplification and high-throughput sequencing (Solexa, Illumina).

Materials and methods

Ethics statement

The study protocol was reviewed and approved by the Animal Ethics and Welfare Committee of the School of Public Health and Tropical Medicine, Southern Medical University, China. All animals were treated in strict accordance with the guidelines for Laboratory Animal Use and Care from Southern Medical University and the Rules for the Implementation of Laboratory Animal Medicine (1998) from the Ministry of Health, China.

Sample collection

Fecal sample collection from bats was performed in Hainan province [Haikou city (latitude: 20.02° N; longitude: 110.35° E)], Guangdong province [Huizhou (latitude: 23.11° N; longitude: 114.42° E), Guangzhou (latitude: 23.13° N; longitude: 113.26° E) and Yunfu (latitude: 22.92° N; longitude: 112.04° E) cities], all in southern China. Briefly, fecal samples from bats were immersed into maintenance medium in a virus sampling tube and temporarily stored at -20 °C. After completion of sampling, fecal samples were transported to the laboratory and stored at -80 °C. Bat species identification was confirmed by amplification and sequencing of the cytochrome B (cytB) gene, which has been commonly applied in archaeology [20].

Purification of fecal samples and extraction of viral nucleic acids

Tubes with fecal samples in maintenance medium were vigorously vortexed to homogenize the samples. Samples of each species from the same site were then pooled by adding 1 ml from each maintenance medium sample into a fresh sample tube. Six pooled samples, classified by species and collection site, were then centrifuged twice at 13,000×g for 20 min at 4 °C. The supernatant was then filtered through a 0.22 μm filter (Millipore Inc., USA) twice to remove eukaryotic cell- and bacterium-sized particles. Filtrates were concentrated in a 100-kDa Pellicon II filter (Millipore Inc., USA). To remove the remaining extra-cellular nucleic acids, the filtrates were treated with DNaseI (NEB, USA) and RNaseIf (NEB). 375 μl of the filtrate from each pooled sample was then digested in a mixture of 3 U DNase and 25 U RNase at 37 °C for 90 min in 10× DNase buffer (NEB). The viral DNA and RNA were simultaneously extracted using a TaKaRa MiniBEST Viral RNA/DNA Extraction Kit Ver.5.0 (TaKaRa Inc., Japan).

Reverse transcription (RT) and sequence-independent PCR amplification of viral nucleic acids

Reverse transcription was performed using a Transcriptor First-Strand cDNA Synthesis Kit 6.0 (Roche Inc., USA) and 100 pmol of primer K-8N (GACCATCTAGCGACCTCCACNNNNNNNN) [21]. The first-strand cDNA of each sample was converted into double-stranded cDNA in the presence of 5 U of Klenow fragment (NEB) in 10× NEB buffer 2, which was incubated at 37 °C for 1 h and inactivated at 75 °C for 10 min.

Sequence-independent PCR amplification was conducted with 5 μl of double-stranded cDNA template in a final reaction volume of 50 μl, which contained 2× Gflex PCR Buffer, 200 μM deoxynucleoside triphosphate (dNTP), 1 μM primer K (GACCATCTAGCGACCTCCAC) and 1.25 U Tks Gflex DNA polymerase (TaKaRa). The PCR cycles were set as follows: 94 °C for 1 min, followed by 40 cycles of 98 °C for 10 s, 55 °C for 30 s and 68 °C for 1 min.

Library construction and sequencing

A total of 1 μg of DNA per sample was used as input material for DNA sample preparation. Sequencing libraries were generated using a NEBNext® Ultra™DNA Library Prep Kit for Illumina (NEB) following the manufacturer’s recommendations. Briefly, the DNA sample was fragmented by sonication to a size of ~300 bp, and fragments were end-polished, poly-A tailed and ligated with full-length adaptors for Illumina sequencing by further PCR amplification. Finally, PCR products were purified (AMPure XP system, USA), and libraries were analyzed for size distribution using an Agilent 2100 Bioanalyzer (USA) and quantified using real-time PCR.

Pre-processing sequence filter

A pre-processing workflow was applied to ensure that high-quality reads were obtained. The following reads were removed: (1) reads containing low-quality bases (Sanger quality value ≦5 and low-quality bases exceeding 40% of the length of the reads); (2) reads with excessive N bases (≧10% of given length of the reads); (3) reads of less than 50% of a given length; (4) reads containing an overlap of 15 bp with sequences on a user-defined primer/adaptor sequence list; (5) homopolymer-containing reads; and (6) duplicated reads.

The most closely related host genome sequences in this study were scanned and discarded with SOAP2 mapper software [22], by comparison with the available genome sequences of Myotis brandtii, Myotis davidii, Myotis lucifugus and Eptesicus fuscus (whole-genome shotgun sequencing project accession numbers: ANKR00000000.1, ANKR00000000.1, AAPE00000000.2 and ALEH00000000.1). Sequences originating from the host genome (≧90% consistency) were deleted.

Analysis of sequence reads and de novo metagenomic assembly

Valid sequence reads were aligned with sequences in the NCBI non-redundant nucleotide database (NT), the non-redundant protein database (NR) and viral Refseq databases (downloaded from the NCBI FTP server in November 2014) using BLASTn, BLASTx and tBlastx, respectively. The BLAST hits were defined as significant when e-value ≤10−5 [14]. A rapid, highly restrictive BLASTn homology search (with word size 40) against a non-redundant nucleotide (nt) database was performed to eliminate additional host reads and eukaryotic contaminants. The virus-like clean data was then subjected to further analysis.

Clean data from the above filter process were assembled using SOAP de novo assembling software [23] using different K-mer references (45, 55, 59). Contigs with the largest N50 were selected. Assembled sequence contigs greater than 150 bp were compared with sequences in the GenBank non-redundant nucleotide database using BLASTn. The reference sequence of the source organism with the best contig alignment, with an e-value cut-off of 10−5, was retrieved for further phylogenetic analysis. More than one hit was sometimes obtained for each unassembled read and contig, at different taxonomic levels. To guarantee biological significance, the eligible sequences were parsed and exported with MetaGenome Analyzer (MEGAN 4) using the LCA algorithm to assign each sequence to the appropriate taxon using the NCBI taxonomic database [24]. The first and best hit was regarded as the taxonomic annotation of the contigs or unassembled reads.

Molecular epidemiology of bat PVs, RVA, CVs and PiVs

To delineate the molecular epidemiology of bat PVs, RVA, CVs and PiVs from different bat species in four geographic regions of Guangdong and Hainan provinces in southern China, previously described PCR primers were used to amplify the conserved regions of the Late 1 (L1) gene of PVs (450 bp), the major capsid region (VP7) gene of RVA (847 bp), the RdRp gene of CVs (319-331 bp) and the 3Dpol gene of PiVs (571 bp), respectively.

Phylogenetic analysis

Sequence editing and identity calculations were conducted by using BioEdit, version 7.0.4 (http://www.mbio.ncsu.edu/BioEdit/bioedit.html). Nucleotide and amino acid sequence alignments were constructed for different open reading frames (ORFs) and the corresponding amino acid alignment using Clustal W version 2.0 [25]. Phylogenetic trees were constructed using Clustal W version 2.0 and Mega 6 [26] according to the neighbor-joining and maximum-likelihood algorithms. Bootstrap values of the constructed phylogenetic trees were generated by iteration using 1,000 replicates.

Nucleotide sequence accession numbers

All sequences in this study have been submitted to GenBank. The accession numbers for the pol gene of retrovirus, IVa2 gene of adenovirus (AdV), Vp7 gene of rotavirus and L1 gene of PV are KY321921, KY321918, KU746889-KU746892 and KU727225-KU727235, respectively.

Results

Bat sampling

Bat samples were collected between August 2011 and October 2014 in four cities (Huizhou, Guangzhou, Yunfu and Haikou) in southern China. We sampled the fecal specimens from 500 bats (6 species). Eighty-six fecal samples, classified by species and collection site, were pooled as six samples (E-Table 2). These samples were processed as follows: nucleic acid purification, RNA/DNA extraction and sequence-independent RT-PCR for NGS. Molecular epidemiological analysis of of RVA, PV, CV and PiV in the 500 bats was also conducted (E-Table 3).

Virome analysis

Viral metagenomic overview

After removing the contaminating reads, 108,443,256 sequence reads with an average length of 150 bp had been generated from the extracted nucleic acids (Table 1). The fecal samples from bats revealed significant microbial diversity. A total of 90,181,623 sequence reads were classified as cellular organisms, including bacteria, archaea, eukaryota and other unassigned reads. There were 18,261,633 reads that best matched to viral proteins. The sequences from each pool were assembled de novo into contigs of variable lengths. A total of 13,466,976 sequences were assembled into viral contigs (average length: 182 bp, e-value ≦0.0001), among which 5,496 were viral contigs longer than 500 bp (average length: 776 bp).

Table 1 Overview of Solexa sequencing
Full size table

The total viral contigs could be assigned into five viral clades, including 54 families and 219 genera: Reverse-Transcribing Viruses, dsDNA Viruses, dsRNA Viruses, ssDNA Viruses and ssRNA Viruses. Vertebrate viruses comprised the largest population and included viruses classifiable as members of the Alloherpesviridae, Herpesviridae, Adenoviridae, Orthomyxoviridae, Papillomaviridae, Retroviridae, Hepadnaviridae, Flaviviridae, Polyomaviridae, Paramyxoviridae, Rhabdoviridae, Coronaviridae, Picornaviridae, Arenaviridae, Astroviridae, Bunyaviridae, Iridoviridae, Poxviridae and Circoviridae families (Figure 1, Table 2; please refer to Table 2 for details of other viral contigs, plant, bacterial, fungal, archaeal and algal in nature).

Fig. 1
figure 1

Taxonomy information for the six bat groups using the MEGAN software. GZ.CS: Cynopterus sphinx collected in Guangzhou; HN.MS: Miniopterus schreibersi collected in Hainan; HN.RL: Rousettus leschenaulti collected in Hainan; HZ.HL: Hipposideros larvatus collected in Huizhou; YF.RB: Rhinolophus blythi collected in Yunfu; YF.SK: Scotophilus kuhlii collected in Yunfu

Full size image
Table 2 Overview of the viral contigs identified in this study
Full size table

There were 1424 deltaretrovirus-related contigs, accounting for most of the virus-related contigs. These had the highest amino acid identity with the Pol protein of Pteropus vampyrus endogenous retrovirus group K (90%). Sequences of contigs from members classifiable within the genera Phlebovirus, Tospovirus, Rotavirus A, Mastarenavirus, Picobirnavirus, Simplexvirus, Varicellovirus, Mastadenovirus, Circovirus, Mamastrovirus, Coronavirus, Orthohepadnavirus, Reoviridae, Iflavirus, Bracovirus, Orthopoxvirus, Vesiculovirus and Chrysovirus (as well as all phage-like contigs) shared an identity of greater than 80% with sequences in the existing database. In addition, the contigs representing influenza A viruses (with identities of 89.6–100% to sequences in the database, average: 92.8%; 13.2% of the sequence identities were greater than 95%) and coronaviruses (identities of 95–100%, average: 99.8%) found in bats shared high sequence identity with those from humans. Other contigs including those classifiable as members of the genera Alphapapillomavirus (sequence identities of 82.8–92.5%, average: 83.4%), Betaretrovirus (61.9–84.3%, average: 69.7%), Alpharetrovirus (46.5–86.1%, average: 67.5%), Varicellovirus (28.7–100%, average: 49.7%), Cyprinivirus (35.6–100%, average: 58.5%), Chlorovirus (33.3–77.8%, average: 53.4%) and Cucumovirus (46.9–68.8%, average: 57.1%) had low sequence identity to viruses in the existing databases (data not shown).

Bat ReVs (retroviruses)

A total of 2274 contigs were identified as belonging to the five genera of retroviruses (alpha-, beta-, gamma-, delta- and unclassified retroviruses), with e-values of less than 10−5 (Table 2). The retroviral sequences identified herein represented two genes, encoding the protease/polymerase (Pol) and envelope glycoproteins (Env). According to the BLASTx analysis, some of these genes were very closely related to those of Pteropus vampyrus endogenous retroviruses. All retroviral contigs that encoded the Env protein also contained stop codons within the translated region, according to BLASTx analysis. One of the longest retrovirus coding sequences encoding a Pol protein was used to construct a phylogenetic tree (Figure 2). The phylogenetic trees indicated that the longest Pol (HN.RL, accession number: KY321921) sequence identified clustered with those of Pteropus vampyrus endogenous retroviruses. HN.RL shared the highest amino acid sequence identity (85.7%) with Pteropus vampyrus endogenous retroviruses, when compared to other retroviruses detected in bats (56.7–59.8% sequence identity), duck (57.4%), fowl (56.7%), galidia (62.1%), murine (57.7–60.1%), baboon (58.1%), koala (58.1%) and porcine (57.4%) sources (Table 3).

Fig. 2
figure 2

Phylogenetic analysis of the pol gene of ReVs detected in bats from southern China, based on a partial 171 amino acid sequence. The tree was generated using the neighbor-joining algorithm and the p-distance model. A bootstrap test of 1000 replicates was used. Numbers above the branches indicate NJ bootstrap values. Bold triangles indicate retroviruses detected in the present study. RL: Rousettus leschenaultii; HN: Hainan province

Full size image
Table 3 Amino acid sequence identities (%) comparing the partial POL gene between bat ReVs and other established ReVs
Full size table

Bat AdVs (adenoviruses)

We identified 41 contigs representing AdV genomes in Cynopterus sphinx (GZ.CS), Miniopterus schreibersii (HN.MS), Hipposideros larvatus (HZ.HL) and Rousettus leschenaultii (HN.RL). No AdV contigs were found in Scotophilus kuhlii (YF.SK) or Rhinolophus blythi (YF.RB). These AdV contigs represented 25 E4orf1 protein-encoding and 16 IVa2 mature protein-encoding genes. A phylogenetic tree based on the IVa2 gene sequences (Figure 3) revealed that the contig representing AdV in HN.MS fell within the lineage of human adenovirus C, sharing a sequence identity of 100%. Additionally, HN.MS shared a sequence identity of 87.2–100% with human AdVs, compared with only 60.7–62.7% sequence identity with bat AdVs (Table 4). This indicates the putative evolution of a novel viral strain.

Fig. 3
figure 3

Phylogenetic analysis of the IVa2 gene of AdVs detected in bats form southern China, based on a 102 amino acid sequence. The tree was generated using the neighbor-joining algorithm and the p-distance model. A bootstrap test of 1000 replicates was used. Numbers above the branches indicate NJ bootstrap values. Bold triangles indicate adenoviruses detected in the present study. MS: Miniopterus schreibersi; HN: Hainan province

Full size image
Table 4 Amino acid sequence identities (%) for the partial IVa2 gene between bat AdVs and other established AdVs
Full size table

Prevalence and phylogenetic analysis of bat PVs (papillomaviruses)

In total, 2.2% (11/500) of fecal samples tested positive for PVs, of which two were found in Scotophilus kuhlii and nine in Rhinolophus blythi (E-Table 3).

A phylogenetic tree was constructed based on an L1 gene nucleotide sequence alignment of 47 PV types (representing multiple PV species/genera) in combination with the bat PVs detected herein. The phylogenetic tree constructed by the neighbor-joining algorithm clustered the established PVs into 17 distinct, previously defined genera, ranging from Alphapapillomavirus to Sigmapapillomavirus [27, 28]. Within this phylogenetic tree, the 11 bat PVs from Scotophilus kuhlii and Rhinolophus blythi clustered in the same clade, constituting a monophyletic clade with viruses classified within the genus Betapapillomavirus. However, the bat PVs reported here actually appear to belong to an unassigned genus because the bootstrap support value for the closest node was less than 80% (Figure 4). Eleven bat PVs shared high sequence similarities in the L1 gene region. Notably, the similarities of the PVs from Scotophilus kuhlii and Rhinolophus blythi ranged from 94.5% to 100%. Likewise, low sequence similarity was found between the PVs reported here and previously detected bat PVs Ms-PVs-1 (42.1–43.6% similarity), MrPVs1 (45.8–47.3%) and RaPVs1 (47.7–50.0%). Even when compared to the most closely related clade (Betapapillomavirus), the bat PVs shared a sequence similarity of only 55.0–58.9%. The similarities, calculated by pairwise alignment of the corresponding amino acids, are provided in Table 5.

Fig. 4
figure 4

Phylogenetic analysis of the L1 gene of papillomavirus detected in bats from southern China, based on a 392 nucleotide sequence. The phylogenetic tree was generated using the neighbor-joining algorithm with the p-distance or Maximum Composite Likelihood model. A bootstrap test of 1000 replicates was used. The numbers above the branches indicate the NJ bootstrap values. Bold triangles indicate the papillomavirus detected in this study. Bold squares indicated the papillomavirus detected in bats. RB: Rhinolophus blythi; SK: Scotophilus kuhlii; YF: Yunfu; HZ: Huizhou

Full size image
Table 5 Amino acid sequence identities (%) for the partial L1 gene between bat papillomaviruses identified in this study and other previously reported papillomaviruses
Full size table

Prevalence and phylogenetic analysis of bat RVA (rotavirus A)

In total, 0.8% (4/500) of fecal samples tested positive for RVA strains, all of which were found in Scotophilus kuhlii (E-Table 3).

A phylogenetic tree was constructed based on a VP7 gene nucleotide sequence alignment of established genotypes, calculated using the neighbor-joining algorithm with the p-distance or Maximum Composite Likelihood algorithm. The four RVA strains detected herein clustered with mammalian RVAs, constituting a monophyletic clade among the genotype 3 strains; however, the bootstrap value of the node was only 50%, which failed to confirm whether the bat RVAs detected herein belong to G3 (Figure 5).

Fig. 5
figure 5

Phylogenetic analysis of the major capsid region (vp7) of rotaviruses from bats in southern China, based on a 847 nucleotide sequence. The tree was generated using the neighbor-joining method with the p-distance or Maximum Composite Likelihood model. A bootstrap test of 1000 replicates was used. Numbers above the branches indicate NJ bootstrap values. Bold triangles indicate rotaviruses detected in the present study. Abbreviations of virus names are shown in Figure 2, footnote. YF: Yunfu; SK: Scotophilus kuhlii; RB: Rhinolophus blythi

Full size image

The putative VP7 genes of the four RVA strains detected in Scotophilus kuhlii were 847 bp in length and encoded a 282 amino acid protein. When compared with other mammalian rotaviruses, the VP7 nucleotide sequences of the four Scotophilus kuhlii isolates showed low levels of sequence identity to the established G3 genotypes (58.0–80.2%). Compared with human, bat, cat, dog and simian G3 RVA strains, the bat RVAs reported herein had sequence identities of 77.1–79.5%, 79.4–79.6%, 79.3–79.4%, 78.3–80.2% and 77.6–80.0%, respectively (Table 6).

Table 6 Nucleotide identities (%) of partial Vp7 gene sequences between bat rotaviruses and other established rotaviruses
Full size table

The bat VP7 sequences shared a maximum nucleotide identity of 80.0% with that of a simian RVA strain (RVA_simian-tc_USA_RRV_1975_G3P3), which is approaching the cut-off value (80.0%) for VP7 classification proposed by the Rotavirus Classification Working Group (RCWG). The amino acid identities of the VP7 antigenic regions: A (87–100), B (141–150) and C (208–224) within these bat RVAs, when compared with the G3 types was 85.7–92.8%, 80.0–100.0% and 82.3–88.2%, respectively, whereas these sequences exhibited identities of 50.5–85.7%, 50.0–100.0% and 64.7–88.2% to other G types, respectively (Table 6).

Prevalence and phylogenetic analysis of bat CVs and PiVs

No CVs or PiVs were found in the 500 fecal specimens studied.

Discussion

To determine the potential for zoonotic virus transmission from bats to humans, we collected bat samples from habitats mainly in residential areas, city parks, abandoned houses and mine caves, all of which were close to the living environment of humans. This study examined the virome composition and viral abundance in fecal and rectal samples from six bat species collected in four cities in southern China. This represents the first time, to our knowledge, a report on the viral diversity of Cynopterus sphinx, Rousettus leschenaultii, Hipposideros larvatus and Rhinolophus blythi in southern China.

Virome analysis

Similarly to previous studies [5,6,7,8], a considerable number of viral sequence reads could be classified into virus families: plant, fungal, bacterial, insect, archaeal and algal in nature. The identification of insect viruses, fungal viruses and plant viruses presumably corresponds to the habits and habitats of the bats, including insect-eating, cave-dwelling and social stucture. In addition, the identification of phages likely reflects the bacterial flora harbored inside the bats. Since we focused on the putative role of bats in zoonotic disease transmission, only mammalian virus contigs were analyzed further. Similarly to previous studies, most mammalian viruses were typical zoonotic viruses, including viruses classifiable as Influenza virus A, as well as members of the genera Mastadenovirus, Rotavirus, Mamastrovirus, Coronavirus, Hantavirus, Rhabdovirus and Parechovirus, and also members of the family Papillomaviridae. Metagenomic analyses in other studies have identified the full-length sequences of adenoviruses [8, 10, 12, 14], papillomaviruses [13, 14, 27, 28], parvoviruses [8, 13, 29], circoviruses [29], herpesviruses [11] and rotaviruses [12]. However, it is challenging to make specific or direct comparisons with our study due to significant methodological variations; for instance, the sample type (e.g., fresh guano, urine and roost guano), virus enrichment procedures (e.g., nuclease treatment, centrifugation and filtration) and amplification methods and/or high-throughput sequencing platforms [30]. The most important finding of our study was that the influenza A virus and coronavirus sequences we identified in bats shared their highest genetic identity with previous isolates from humans, suggesting possible cross-species transmission. However, contig sequences representing viruses classifiable within the genera Alphapapillomavirus, Betaretrovirus, Alpharetrovirus, Varicellovirus, Cyprinivirus, Chlorovirus and Cucumovirus were also identified that had identities of less than 70% when compared with viruses in existing databases, indicating the possible evolution of novel viral strains.

Similar to the findings of Yuan et al [15], retroviruses were one of the most commonly found viruses in bats. Rousettus leschenaultii from Hainan was found to harbor phylogenetically distant gammaretroviruses from those found in Myotis lucifugus, Megaderma lyra, Myotis davidii and Myotis ricketti, which suggested that bats might be suitable vectors for retrovirus transmission [31, 32]. Interestingly, we found that the AdV contigs from HN.RL (representing the IVa2 genes) were clustered with human adenovirus-C and shared higher amino acid sequence identity with human adenoviruses than previously reported bat AdVs. This might indicate the possible evolution of novel viral strains. Owing to the limited size and number of contigs generated by Illumina sequencing, conventional PCR assays are warranted to generate longer and more abundant sequences for further phylogenetic analysis [33, 34].

Comparable with previous studies [13, 15], except for retrovirus and AdV, analysis of more in-depth virus sequencing information (including phylogeny and sequence identity) was challenging for the following reasons: 1) many reads could not be assembled into longer scaffolds because the NGS reads were randomly amplified throughout the whole viral genome; 2) incomplete DNase/RNase enzyme digestion resulting in a lack of hits for a considerable number of sequence reads (including those from cellular organisms) during raw data analysis; 3) some viruses may have been present in the samples in relatively small amounts below the sensitivity of NGS detection; 4) host genome and other non-target nucleic acids could have lowered the sensitivity of high-throughput sequencing and introduced background noise into the results.

Papillomaviruses

The L1 ORF is characterized as the most conserved region in the PV genome and is used for PV classification. According to recent classification criteria: 1) PVs belong to different PV genera when they share less then 60% nucleotide sequence identity across the entire L1 ORF [28]; 2) the taxonomic status of PV types, subtypes and variants is based on the conventional criteria that the sequence of their L1 genes should be at least 10%, 2–10% and maximally 2% dissimilar from one another. Because the bat L1 sequences shared a maximum of 58.9% nucleotide identity with the L1 sequence of HPVs-92, these PVs cannot be placed in one of the existing genera. The bat PVs therefore represent tentative members of a novel, yet innominated, PV genus.

Our findings argue against the mainstream theory that PVs are highly species specific and co-evolve with their host. Compatible with the results of Garcia-Perezstrict et al [35], this hypothesis of virus-host co-evolution is rejected by the existence of 11 closely related bat PVs in Scotophilus kuhlii and Rhinolophus blythi and by the lack of congruence between bats and bat PV phylogenies. Second, the possible inter-transmission of PVs to infect two different bat species (Scotophilus kuhlii and Rhinolophus blythi) argues against strict host specificity. However, the L1 sequence detected herein was only 450 bp in length, which does not represent the whole of the L1 gene. Further studies with additional genomic sequences are required to examine the potential pathogenic role of PVs in other bat species, which would provide insights into the ecology and evolution of PVs in bats. These findings might help address whether the prevailing paradigms regarding PV evolution should be reconsidered.

Rotavirus

The RVA VP7 glycoprotein (G genotypes) elicits neutralizing antibodies, and, to date, 27 G and 37 P genotypes from various hosts have been identified [36]. The nearly full-length VP7 sequences (847 nt) of four Scotophilus kuhlii rotavirus strains (SK/13YF128/129/200/212), encoding a 282-amino acid protein, were determined. These segments were most closely related to simian and dog RVA G3 strains (RVA_simian-tc_USA_RRV_1975_G3P3, RVA_dog-tc_ITA_RV52-96_1996_G3P3), sharing a similarity of 80.0% and 80.2%, respectively. Even though the similarity of these VP7 sequences did not exceed the appropriate cut-off (80.0%), the VP7 sequences detected herein represented more than 500 nt of the ORF (>50% of the ORF). Hence, the four bat RVA strains reported herein might be tentative G3 strains according to the RCWG classification [37].

G3P [3] and G3P [9] are genotype combinations commonly found in canines (RV198, RV52, A79-10, K9, CU-1, AUS and ITA) [38, 39]. Previous reports have identified canine RVA strains that infect humans, including the strains Ro1845, HCR3A, AU-1, E2451, L621, T152, CMH222, PA260, PAH136, PAI58, 6212, CMH120 and CU-1 [38,39,40,41]. In addition, some isolated animal RVA strains are believed to have a distant ancestor in common with canine RVA strains, when taking into account the majority of their gene segments [41]. Two distinct simian RVA strains (RRV and TUCH) have multiple genotypes more typical of canine-feline rotaviruses [40], suggesting potential transmission of canine-like RVA strains to simians. Although the genetic relationship of VP7 sequences between the four bat RVA strains and ITA/RRV is relatively distant, they might share a common ancestor. In light of the fact that the four bat RVA strains are not closely related to any known RVA strain, we speculate that these viruses are true bat RVA strains rather than viruses transmitted from other species.

The positive detection rate of RVA strains was low, which might have arisen for the following reasons: 1) we collected samples from adult bats which were immunocompetent and can clear viruses rapidly; 2) no neonates were involved, whereas RVA strains are known to infect mainly infants and young animals. To better understand the prevalence of RVAs in bats, further molecular and serological investigations with larger sample sizes and more species of adult and neonatal bats in different geographical regions are required.

Calicivirus and picornavirus

Notably, CVs and PiVs were not detected in the present study. The detection rates of CVs and PiVs in bats was also low in previous studies. In 2014, Kemenesi et al [17] found only three novel CVs in 447 bat fecal specimens in Europe, segregating with the Hipposideros pomona sapovirus genus found in 2012 [16]. In another study only 12 picornavirus-positive respiratory and alimentary specimens were identified from 1108 bats and 18 species [18]. The lack of CVs and PiVs detected in our study may be accounted for by low carrying rates for these two viruses. Other causes might be variations in duration of the virus shedding in feces, or geographical/seasonal/bat species differences.

Conclusion

In conclusion, this study describes the virome of six bat species in southern China and provides a more comprehensive understanding of virus ecology in common bat species found in human habitats. The detection of PVs in Scotophilus kuhlii and Rhinolophus blythi argues against the hypothesis that PVs have strict host specificity and co-evolution. The four bat RVA strains we isolated might be tentative G3 strains, according to the RCWG classification.