Bat Distribution Size or Shape as Determinant of Viral Richness in African Bats

The rising incidence of emerging infectious diseases (EID) is mostly linked to biodiversity loss, changes in habitat use and increasing habitat fragmentation. Bats are linked to a growing number of EID but few studies have explored the factors of viral richness in bats. These may have implications for role of bats as potential reservoirs. We investigated the determinants of viral richness in 15 species of African bats (8 Pteropodidae and 7 microchiroptera) in Central and West Africa for which we provide new information on virus infection and bat phylogeny. We performed the first comparative analysis testing the correlation of the fragmented geographical distribution (defined as the perimeter to area ratio) with viral richness in bats. Because of their potential effect, sampling effort, host body weight, ecological and behavioural traits such as roosting behaviour, migration and geographical range, were included into the analysis as variables. The results showed that the geographical distribution size, shape and host body weight have significant effects on viral richness in bats. Viral richness was higher in large-bodied bats which had larger and more fragmented distribution areas. Accumulation of viruses may be related to the historical expansion and contraction of bat species distribution range, with potentially strong effects of distribution edges on virus transmission. Two potential explanations may explain these results. A positive distribution edge effect on the abundance or distribution of some bat species could have facilitated host switches. Alternatively, parasitism could play a direct role in shaping the distribution range of hosts through host local extinction by virulent parasites. This study highlights the importance of considering the fragmentation of bat species geographical distribution in order to understand their role in the circulation of viruses in Africa.

Few studies have investigated parasite species richness in bats [9][10][11]. However, Turmelle and Olival [12] showed viral richness in bats correlates with IUCN status and population genetic structure. The distribution range of hosts has been often considered as a potential determinant of parasite species richness [13][14][15]. Hosts distributed over large areas are more likely to encounter new parasites that may infect them [14,16]. However, the shape of the distribution has received little attention [12,13] but may have implications on the role of bats as pathogen reservoirs. Distribution shape and habitat fragmentation were observed at two different scales and Fahrig [17] suggested that the processes affecting changes in distribution and habitat preference of a species are independent. The shape of the distribution being mostly the products of speciation, extinction and range expansion [18]. Area shape is an important aspect of the distribution of animals and plants, which is strongly linked to population demographics and the subsequent contraction and expansion of their distribution [19,20]. Therefore, area shape must be taken into account together with phylogenetic information in any comparative analysis of parasite diversity. Two alternative explanations can be proposed on the potential link between host distribution shape and parasite species richness: a longer border, due to fragmentation, may entail higher habitat diversity which would intensify contacts with various sources of parasites leading an overall increase in parasite diversity. Alternatively, a longer border may increase host species vulnerability due to area fragmentation and reduced host population size, hence pathogen transmission.
The first comparative analysis was performed to test the hypothesis that distribution shape and more specifically the fragmentation of the distribution area, correlates with viral richness in bats. We investigate the determinants of viral richness in 15 species of African bats, on which we found new information on virus infection and bat phylogeny. Body weight, roosting behaviour and migration [10,21] were also included in our analysis because of their potential influences on parasite or viral species richness.

Ethic statements
All the capture events, animal handling, euthanasia and transfer of samples across country borders were performed in accordance with the guidelines of the American Society of Mammalogists (http://www.mammalsociety.org/committees/ animal-care-and-use) [22]: Bats were captured following recommendations by Kunz and Parsons [23]. Captured bats were removed carefully from nets as soon as possible to minimize injury, drowning, strangulation, or stress. Safe and humane euthanasia was achieved through the use of inhalant anaesthetic (halothane) prior to autopsy.
All work (capture, euthanasia and autopsy) was carried out with authorization from the respective wildlife authorities of each  [4,24,25].

Study animals
Our study on the correlation of viral richness in bats was conducted using 15 bats species from Central and West Africa. We selected only the species for which we had enough samples and information on viral richness to carry out analysis. Bats were caught in the Republic of Congo, Gabon, Central African Republic (CAR) and Senegal [4] Figure 1).     Bats were captured using mist-nets or harp traps. Mist-nets (1262.4 m) were hoisted either in the tree canopy (defined as ''foliage'') or at the entrance of the small roosting caves (defined as ''cave'') just before twilight. Harp Traps were used at the entrance of big caves known to harbor large population of bats. Following capture, bats were identified on site by trained field biologists and individually euthanized under sedation in a field laboratory. Bats were weighed using a spring scale prior to autopsy and selected internal organs were collected during autopsy and stored at 2 80uC for future virological analysis. Data on the ecological traits of the 15 different bat species captured (i.e., roost type, body weight, migratory behaviour and colony size) was gathered from published literature ( Table 1, see Annex 1 for references).

Bat phylogeny
In order to improve the quality of the comparative analysis, a phylogenetic tree was built using 14 new molecular sequences of the bat mitochondrial cytochrome b gene ( Table 2). Total genomic DNA was extracted from ethanol-preserved tissue samples (muscle, liver or spleen) with Genomic DNA Tissue Mini Kit (Geneaid Biotech) according to the manufacturer's protocol. We amplified the mitochondrial gene for cytochrome b (cytb) using primer pairs F1 (modified; 59-CCACGACCAATGACAY-GAAAA-39) and R1 from Sakai et al. [26] in most microbats, L14724 and H15915 from Irwin et al. [27] in hipposiderids and fruit bats, LGL765F and LGL766R from Bickham et al. [28,29] in long-fingered bats (Miniopterus inflatus). The volume of PCR reaction was 25 ml, it contained 12.5 ml Combi PPP Master Mix (Top-Bio, Prague, Czech Republic), 200 mM of forward and reverse primers respectively, and 2.5 ml of extracted DNA. PCR protocol consisted in an initial denaturation at 94uC for 3 min, 35 cycles of denaturation for 40 s at 94uC, annealing for 40 s at 50uC, and extension for 90 s at 65uC, and a final extension at 65uC for 5 min. Resulting PCR products were inspected on 1.5% agarose gel and purified with Gel/PCR DNA Fragments Extraction Kit (Geneaid Biotech). If multiple bands appeared, the one of appropriate length was excised and purified from gel using the same purification kit. Purified PCR products were sequenced commercially (Macrogen, Seoul, Korea) with the respective forward primer using BigDye Terminator sequencing chemistry (Applied Biosystems, Foster City, CA, USA) on ABI 3730xl sequencer. Sequences were edited in Sequencher 4.6 (Gene Codes, Ann Arbor, MI, USA), manually checked for correct base reading and protein coding frame, and aligned by eye in BioEdit 7.0 [30]. Sequences of two artiodactyl taxa, Bos taurus (D34635) and Ovis ammon (AJ867276) were added to the alignment as outgroup taxa for rooting the bat phylogeny. Phylogenetic tree including branch lengths was inferred from aligned nucleotide sequences in PAUP*4.0b (Sinauer Associates, Sunderland, Massachusetts, USA) under maximum likelihood (ML) criterion and general time-reversible model of evolution with a portion of invariable sites and gamma distributed variation rates (GTR+I+C), which was suggested as the best evolutionary model and whose parameters were estimated in Modeltest 3.7. Topological constraints were set before computation of the ML tree, as corresponding to acknowledged phylogenetic relationships among genera, families and higher taxonomic ranks of bats as referred by Teeling et al. [31] and Almeida et al. [32]. Due to a priori definition of the tree topology, analysis of nodal support was not performed. The constrained ML tree was, however, compared to unconstrained ML tree using a Shimodaira-Hasegawa (SH) test, in order to assess possible significant difference, which might indicate unreliability of the constrained tree. Sequences generated in this study were

Viral richness
Two methods were used to document viral richness of the studied bat species. First, we tested our bat samples for viruses. We Determinant of Viral Richness of African Bats PLOS ONE | www.plosone.org used (i) nested Reverse-Transcription polymerase chain reaction (RT-PCR) assay targeting the RNA-dependent RNA polymerase gene using generic consensus primers for the genus Coronavirus [33]; (ii) hemi-nested RT-PCR targeting the N terminal end of the NS5 gene by using degenerate primers for the genus Flavivirus [34,35]; and (iii) filoviruses (Marburg virus and Ebola virus) as previously described [4,36] (Table 3). Then, additional virological data were drawn from literature. In published papers, the methods used to detect viruses directly were mouse inoculation, cell culture, electron microscopy and PCR; indirect methods utilised to detect markers of replication and viral infection in bats from organs, tissues or blood were direct fluorescent antibody, indirect fluorescence antibodies, radio immuno assay, rapid fluorescent focus inhibition test, fluorescent antibody test, and seroneutralization. The serological detection of arbovirus antibodies alone (particularly genus Flavivirus and Alphavirus) was not considered as evidence of a viral association because of some degree of crossreaction within the virus family, rendering it difficult to differentiate viruses. Viruses forming distinct clusters within the same genus were recorded as a unique viral species. For example, in Rousettus aegyptiacus, bat gammaherpes viruses (Bat GHV) 1, 2, 4, 5, 6 and 7 were recorded as one unique viral species and Bat GHV 3 as another viral species [37]. For Ebola virus, different viral species of this genus were considered as a single virus. For each bat species, we calculated the viral richness as the total number of different viruses described for the given bat species.

Geographical distribution size and shape
To test the impact of the fragmentation of the distribution area on viral richness in bats, we used the geographic range maps of each studied bat species provided by the 'IUCN Red List of Threatened Species' web site, one of the biggest databases available on mammalian distribution, based on international experts' knowledge. The maps were imported in a GIS using MapInfo professional V 5.5. We then drew polygons following species distribution to obtain area and perimeter measures for all drawn polygons. The shape of the geographic range was estimated using the ratio of the total perimeter to the total surface area following the approach used by Kauffman cited in Fortin et al. [38]. The higher the ratio, the greater is the fragmentation of the distribution (Figure 2).  Comparative analyses of the determinants of viral richness Using information on bat phylogeny described above, we calculated the independent contrasts for each of the investigated variables with the package APE [39] implemented in R (R Development Core Team 2013). To confirm the proper standardization of contrasts, we regressed the absolute values of standardized contrasts against their standard deviations. Contrasts were then analysed using standard multiple regressions, with all intercepts forced through the origin [40]. We tested the importance of the phylogenetic signal on each variable using the parameter K (which is the ratio of observed phylogenetic covariance divided by the expected covariance under Brownian motion), with the package picante [41] implemented in R (R Development Core Team 2013).
As in previous studies [12,13], we performed standard multiple regressions using independent contrasts, with the intercept forced at zero and viral richness as the dependent variable. Independent variables were geographical range, fragmentation of the distribution, roost type (foliage vs cave), average body weight and migratory behaviour (yes vs no) ( Table 1). We did not include colony size as variable as information was missing for two species. Number of sampled hosts or sampling effort (number of samples we tested added to the number of samples reported in published papers) ware also considered as an independent variable. The analysis was conducted on 14 of the 15 captured species for which sample size was considered sufficient (.30). We then selected the best subset selection of variables using AIC criteria.

Bat Phylogeny
We reconstructed the phylogenetic tree of the bat species investigated in this analysis using 15 sequences under the constraint of acknowledged taxonomic relationships (Figure 3). The constrained tree (2lnL = 6439.91045) did not differ significantly from the unconstrained tree (SH test: diff. lnL = 7.89267, Table 5. Levels of phylogenetic signal in the variables investigated using the parameter K and the parameter lambda.  Figure 4. Partial relationship between viral richness and distribution fragmentation, assessed by a measure of distribution shape using (A) phylogenetic independent contrasts, or (B) raw values (and using residuals from the general regression modelling in Table 7). doi:10.1371/journal.pone.0100172.g004 P = 0.126), and was thus considered as a reasonable depiction of bat phylogeny.

Determinant of the viral richness
Only viral richness showed statistically significant level of phylogenetic signal using estimates of K among all the traits investigated (Table 5). However, distribution shape showed a level of phylogenetic close to significance (Table 5).
Four variables were retained in the preferred model, which was back-selected, based on the AIC criterion, and using the raw data (non corrected for phylogeny) ( Table 6). Using the independent contrasts (variables controlled for phylogeny), the best model had the same four independent variables (Table 6). Taking into account host sampling, we found that viral richness in bats was greater in large-bodied and widely distributed bats and when their geographical distribution was fragmented (Tables 5 & 6). There were no significant relationships between viral richness and migratory behaviour or roosting behaviour. Finally, greater fragmentation of the geographic distribution was highly associated with increased viral richness ( Table 7, Figures 4A & 4B).

Discussion
This is the first comparative analysis investigating the effect of distribution shape, i.e. geographical range fragmentation or edge range density, on viral richness in bats. Our first hypothesis was that bats living in caves in sympatry with other species with increased promiscuity and high population density of susceptible individuals, would generate opportunities for cross-species transmission of viruses and their rapid spread. However, our study does not support this hypothesis. Our results showed a significant influence of host body weight, distribution size and shape on viral richness; viral richness increases with larger distribution areas and fragmentation of bat distribution, according to the measure of their distribution shape. Before discussing this correlation, the difference between habitat fragmentation and habitat loss should be considered since Fahrig [17] suggested that the two processes are independent. An ecological explanation of the correlation between viral richness and distribution could be interpreted in the light of the historical biogeography of African bats, which falls within the domain of phylogeny and phylogeographic studies [31]. Range distributions and shapes are the product of speciation, extinction and historical displacements [18]. The accumulation of Table 6. Comparison of models used to test the effects of several independent variables (weight, size and shape of distribution, migration, roosting and sample size) on viral richness of bats (using the independent contrasts), using phylogenetic regression (Independent contrasts) or non-phylogenetic regression (raw values).

Analysis
Model ranks AIC  Table 7. Best model explaining viral richness in bats using independent contrasts (initial model is given in Table 6), using the phylogenetic regression (independent contrasts) and non-phylogenetic regression (raw values' and independent variables are ranked according to their contributions to the models using F values). parasite species, viruses in the present study, could be related to the historical expansion and contraction of bat species' distribution ranges, with potentially strong effects of distribution edges on virus transmission. Indeed, the marginal effect of phylogenetic signal on the distribution shape of the investigated bats (Table 5) suggests that both history and current ecological drivers may have shaped their distribution. For a given distribution area, the most fragmented distributions contain more edges than the less fragmented ones. Positive edge effects could be responsible for the positive effects of distribution shape on either the abundance or distribution of some bat species that may have facilitated virus host switches. However, critical information to explore this issue further is lacking due to the limits of current knowledge on African bats' phylogeography as well as the geographic distribution and phylogeny of their viruses (such as bats and rabies-related viruses [42]). Furthermore, it should be noted that the use of the distribution area obtained from ICUN Red List might not accurately describe the distribution shape of bat species. More accurate and precise distributions would definitively improve the robustness of the study. An alternative explanation produced by a theoretical study, attributes a direct role of parasitism in limiting the distribution range of hosts through the extinction of local hosts by virulent parasites [43]. However, this hypothesis has not been tested using empirical data.
As previously emphasized, we must differentiate the fragmentation of the distribution from habitat loss, as the consequences on bat species of the habitat loss are likely to be different to the consequences of the range fragmentation. Habitat loss following land use changes has been perceived as a major threat to biological diversity, whereas fragmentation may be positive or negative [42]. Habitat losses may increase species losses and, in turn, induce changes in ecosystem functions, including parasitism. Several studies have shown that parasites suffer more from habitat loss and isolation than their hosts, but other studies emphasize that habitat loss may increase the abundance of some hosts, and consequently their parasite loads, through an increase of host density-dependent transmission [13]. The consequences in terms of surveillance, spillover and emergence in human populations are then species specific, in relation to their historical biogeography, actual range size and shape, and on-going loss of habitat. As already emphasized by Turmelle and Olival [12], while biogeography can help to identify macro-ecological determinants of pathogen richness, and potentially epidemiological processes, control strategies need to be carried out at local geographic scales.
The number of viruses found in bats in our study added to the viruses described in bats in the literature is certainly an underestimation. Indeed, bats are reservoirs for many viruses and have the peculiarity to maintain viral replication at relatively low levels. Thus, chronicity of viral infections in bats requires the use of highly sensitive detection tools. However, in our study, samples were tested by Reverse-Transcription PCR assay using generic consensus primers, known to decrease sensitivity. The detection of these viruses may be improved by more sensitive methods, such as high-throughput sequencing and viral isolation yet much more expensive than PCR.