The ability to predict and prevent viral epidemics has become a major objective in
the public health disciplines. Knowledge of viral hosts enables the identification
of maintenance populations from which epidemics may emerge1
2
3
4. From an ecological perspective, the natural host of a given virus may be regarded
as its habitat5
6. Unfortunately, unbiased assessment of habitat is impossible, as it requires the
investigation of all host species in all possible geographical locations. Moreover,
the delimitations of viral species and populations are often not well defined. However,
viral phylogenies can identify whether extant viruses encountered in a given host
are less or more directly linked to common ancestors, providing clues to virus origins4
5
6.
Hosts of relevant infectious disease agents share distinctive properties that can
be summarised into a cumulative definition of animal reservoirs (AR) as of interest
in public health. We define as ARs those taxa of extant animals that contain higher
genetic virus diversity than target taxa, harbour the virus continuously on the level
of social groups (with or without disease)4
7, and are naturally infected beyond the geographical limits of communicating social
groups3
4. These cumulative parameters are amenable to field investigations, without the requirement
to manipulate wild animals and their habitats extensively. Nevertheless, logistical,
ethical and ecological limits in field investigations still require a focussing on
appropriate candidate taxa. Here, it is helpful to consider that the potential of
animals to act as AR can be linked to social and behavioural parameters, which in
turn are inherent to taxa. Virus transmission and maintenance is generally favoured
by large social groups, close social interaction, high population densities, as well
as spatial mobility and fast population turnover2
4
8. It is also helpful to consider that genetic proximity between species is favourable
of cross-species pathogen transmission6
9
10
11. For instance, all cumulative criteria are met by the best studied of relationships
between a virus genus and its AR, the influenza A virus in waterfowl3
9
10.
In applying the above considerations to larger taxa of mammalian viruses, such as
the members of the mammalian paramyxoviruses (PVs), mammals come into focus as potential
AR. In mammals, the criterion of large natural group sizes is particularly met by
species within the orders Chiroptera (bats) and Rodentia (rodents)1
8. Indeed, sporadic detection of PV in members of both orders has been previously
reported by us and other groups12
13
14
15
16. Here we tested 119 bat and rodent species worldwide, and identified a large range
of novel PV related to major pathogens. These included very large diversities of henipa-
and rubulaviruses, as well as relatives of respiratory syncytial virus (RSV), mouse
pneumonia- and canine distemper virus in bats. Multiple morbilli-related viruses were
found in bats and rodents. These data will change our conception of PV host associations
and influence future attempts to assess and predict epidemic risks.
Results
PV detection in small mammals
To assess the spectrum of PV in bats and rodents, 86 species of bats (4,954 individuals)
and 33 species of rodents (4,324 individuals) were sampled in 15 locations worldwide
(Fig. 1, Table 1 and Supplementary Table S1). All samples were screened for Paramyxoviridae
by various reverse transcription (RT)–PCR assays. Pooled serum samples were additionally
screened by random amplification and deep sequencing of serum-derived cDNA. Although
this approach identified a range of viruses whose full genomes remain to be determined,
no PVs were detected (Supplementary Table S2). This was presumably due to higher concentrations
of the identified viruses in blood, or because of properties that technically promote
detection by random amplification, such as circularity of DNA genomes. In contrast,
RT–PCR detected a large range of PV cDNA sequences as described below.
Phylogenies were inferred using maximum likelihood and Bayesian approaches (Fig. 2a–e
and Supplementary Fig. S1). Detection rates of viruses in bats and rodents were similar
at 3.3 and 3.1%, respectively, but bat viruses were dispersed across the phylogenetic
tree with pronounced genetic divergence, whereas rodent viruses belonged to three
clades with relatively low divergence that were not exclusively detected in rodents
(Supplementary Fig. S1). PVs were detected in six of ten families of bats tested,
representing all major chiropteran phylogenetic lineages (Fig. 3a,b). A simplified
criterion based on pairwise amino acid distance matrices was used to estimate how
many novel virus species might have been detected (refer to Methods section). In total,
66 novel PV taxa at the level of putative species were discernible. This number exceeded
the number of PV species currently known, as the International Committee on Taxonomy
of Viruses (ICTV) currently lists 36 species, and even upon inclusion of unclassified
sequences from GenBank, only an estimated 57 species can be discerned.
Major PV genera in bats
The genus Rubulavirus contains three human pathogens, the mumps virus, as well as
the respiratory viruses, Parainfluenzavirus 2 and 4. A large range of novel rubulaviruses
in fruit- and insect-eating African bats were identified in addition to those six
bat rubulaviruses previously known (Fig. 2a, Supplementary Table S1). The viruses
could be classified into 21 discernible taxonomic entities on the level of putative
species. No rubulaviruses were detected in rodents. A bat virus of high similarity
to mumps virus was fully sequenced from a bat spleen (15,378 nucleotides, (Supplementary
Fig. S2)). Amino acid similarity was well above 90% in all genes except P (89.5%),
suggesting this virus and human mumps to be conspecific. To determine antigenic relatedness
to human mumps virus, sera from 52 flying foxes and 78 insect-eating bats were tested
by immunofluorescence (n=26 Eidolon helvum, 5 Epomops franqueti, 5 Micropterus pusillus,
5 Myonycteris torquata, 11 Rousettus aegyptiacus, 14 Coleura afra, 21 Hipposideros
cf. caffer, 11 Hipposideros gigas, 17 Miniopterus inflatus, 11 Rhinolophus cf. alcyone).
Clear reactivity was observed in 41.5% of the 130 tested bat sera (Fig. 4a). Specificity
of these reactivities was confirmed by cross-testing against other PVs, including
the rubulavirus Parainfluenzavirus 2 (Fig. 4b). These data in total suggest that mumps
and related bat viruses may belong to one same serogroup.
Measles is one of the most important childhood diseases. Measles virus defines the
genus Morbillivirus that also contains the canine and phocine distemper virus, a dolphin
morbillivirus, as well as Rinderpest and Peste Des Petits Ruminants viruses. Novel
members of the genus Morbillivirus as currently defined by the ICTV were detected
for the first time in bats (neotropical vampire bats) but not in rodents (Fig. 2b).
A clade of morbillivirus-related rodent- and tupaia viruses, forming an unclassified
sister clade to the genus Morbillivirus, was extended by an estimated 25 novel bat
and rodent viruses, confirming the unclassified genus and the Morbillivirus genus
to form a major stem lineage of PV.
The genus Henipavirus comprises two known virus species causing fatal encephalitis
in humans. These viruses (Hendra virus (HeV) in Australia, Nipah virus (NiV) in Asia)
have been sporadically acquired from bats of the Pteropus genus by humans, swine and
horses. We have recently detected small sequence fragments of potentially related
viruses in a colony of E. helvum fruit bats in Ghana12. The present data identify
at least 23 distinct viral clades in phylogenetic relation to henipaviruses in six
bat species sampled in five different African countries (Fig. 2c). On the basis of
the minimal genetic distance between HeV and NiV (7% in the analysed L-gene fragment),
the African viruses were estimated to pertain to 19 novel virus species in the genus
Henipavirus (Figs 2c and 5). Full genome sequencing of a representative virus (GH-M74a)
from a bat spleen confirmed formal classification in the genus Henipavirus (18,530
nucleotides, (Supplementary Fig. S3)). It was noted that the RdRp gene of the analysed
African virus contained the catalytic motif GDNQ, as typical of the order Mononegavirales,
whereas HeV and NiV have an atypical GDNE motif. Sequencing of this motif in representatives
of all major African virus clades also yielded GDNQ, supporting the idea that African
rather than Asian viruses are identical to generic ancestors in this highly conserved
motif. In addition, the GDNE signature typical of HeV and NiV was found in a small
fraction of African viruses, and these were in closest phylogenetic relationship to
HeV and NiV (Fig. 2c). The most parsimonious explanation for the diversion of signatures
in this highly conserved motif was a single change in a common ancestor to the GDNE-containing
clade, as opposed to hypothetical convergent acquisitions of GDNQ in all but one of
the parallel lineages. The GDNE-ancestral virus would most likely have existed in
Africa, and would have been ancestral to HeV and NiV as well.
Close relatedness between NiV and these African viruses was also demonstrated by immunofluorescence
staining. To this end, Vero cells were infected in a biosafety level 4 laboratory
with NiV and incubated with serum from E. helvum before staining with a polyclonal
anti-bat antibody. Clear reactivity was seen (Fig. 4c).
In two different widely distributed species of New World bats, one frugivorous (Carollia
perspicillata, rate 0.5%) and one insectivorous (Pteronotus parnellii, rate 7.5%),
we identified virus sequences that were phylogenetically closely related to the genus
Henipavirus (area of distribution shown in Supplementary Fig. S4). These viruses formed
a sister clade to the genus Henipavirus, including its novel African representatives
(Figs 2c and 5a). It was impossible for ethical and logistical reasons to euthanise
Neotropical bats and sample internal organs that might have enabled the completion
of full genome sequences. Nevertheless, the N protein gene at the far opposite end
of the genome of one of these viruses was successfully sequenced from a faecal sample.
RdRp- and N-gene fragments were phylogenetically congruent and no signs of genomic
recombination were seen (Fig. 5b).
RSV and the related human metapneumovirus, belonging to the Pneumovirinae subfamily,
are the leading causes of morbidity and mortality in children worldwide. We identified
a clade of novel bat viruses also in this subfamily, which formed a sister clade to
the human and bovine RSV (Fig. 2d). No bat viruses were detected in direct relationship
to the genus Metapneumovirus, whose only human member, the human metapneumovirus,
is most closely related to avian metapneumovirus C. These data support the notion
that the metapneumoviruses are of direct avian, rather than direct mammalian descent17.
Sendai virus, a member of the genus Respirovirus, was detected in wild rodents from
Thailand (Fig. 2e), confirming this virus not to be restricted to laboratory animals.
No viruses belonging to this genus were detected in bats, but bat sera were found
to react with respirovirus antigens (Fig. 4b), supporting previous assumptions that
bats should carry closely related viruses18.
Natural history of infection
To compare organ-specific compartments of PV replication, virus concentrations in
internal organs were determined in 22 African E. helvum bats infected with henipaviruses.
The spleens of 21 of 22 animals tested positive with high virus concentrations, whereas
all other internal organs as well as blood yielded virus at lower rates and concentrations
(Fig. 4d). Organ distribution of morbilli-related PV in 28 rodents was strikingly
different from that in bats, with broader and less organ-specific distribution of
virus. Highest viral loads and most frequent virus detections in rodents were seen
in the kidneys, and in bats in the spleens (Supplementary Fig. S5, Supplementary Tables
S3 and S4).
We next looked for evidence of continuous virus excretion on population level in bats7.
A colony of Myotis myotis in Germany known to carry members of the Morbillivirus-related
clade was monitored over 3 years for shedding of virus. Shedding rates and concentrations
were constant over three observation periods throughout 2008 to 2010 (Fig. 6). This
colony was a maternity colony, and parturition times fell in the middle of observation
periods but did not influence excretion, suggesting natural infection with little
differences in shedding between young and adult animals. It was noted that this pattern
was very different from that of measles virus in humans19
20.
Interestingly, the morbilli-related viruses from M. myotis in Germany were also observed
in genetically distant Coleura afra bats in Ghana. For several other PV clades, evidence
of their association with bats beyond the limits of communicating social groups was
obtained, as summarised in Fig. 7a and b. One pneumovirus, six henipaviruses and eight
rubulaviruses were detected throughout several sampling years and sites across four
different Western and Central African countries. Notably, widespread and constant
detection of several viruses was observed in cave-roosting Rousettus aegyptiacus that
do not participate in annual migrations and whose populations are isolated from each
other (Fig. 7c).
To identify signs of PV-dependent disease in bats, paramount serum chemistry parameters
were compared between infected and non-infected bats. No infection-dependent alterations
were seen in concentrations of lactate dehydrogenase, gamma glutamyltransferase, aspartate
aminotransferase, alanine transferase, bilirubin and albumin (Supplementary Table
S5). In summary, bats show an overwhelming PV diversity, continuous PV shedding without
evidence of efficient immune control, signs of organ-specific replication and absence
of PV-related pathogenicity. It was noted that these traits taken together are typical
of viral ARs.
Tracing hosts in PV phylogeny
To extract information on hosts during the more recent evolution of extant mammalian
PV, the most parsimonious hypothesis of historical host trait changes was reconstructed
along the PV tree. To take topological uncertainty into account, this analysis was
done on a large set of phylogenetic trees extracted from the terminal phase of an
optimised Bayesian inference of phylogeny. The numbers of deduced transitions between
ordinal categories of hosts including bats, rodents, primates, carnivores, ungulates
and birds were estimated and averaged over >10,000 trees. To achieve a proper representation
of PV from hosts other than bats and rodents, reference sequences were selected from
GenBank so as to provide a maximal genetic diversity per ordinal host category (Fig.
8a). In summary, these analyses determined host switches from bats into other host
categories to have occurred more often than switches originating from any other of
these host categories (Fig. 8b).
To achieve a statistical exclusion of alternative hypotheses regarding the host switching
process, maximum likelihood analyses of ancestral state reconstruction were done under
different model restrictions as summarised in Fig. 8c. All of these analyses suggested
that bats but not rodents, primates or birds should be preferred donors of PV along
the tree. Moreover, although assumptions of fossilised bat hosts at the root points
of both PV subfamilies did not significantly reduce the model likelihood, all other
fossilised host assumptions did.
Discussion
In this study, we have gathered evidence of bats being in close evolutionary and ecological
relationship with several genera of mammalian PVs. The investigation of viral host
associations can be a lengthy and controversial process that depends on targeted and
ecologically informed sampling21
22
23
24
25
26
27
28
29. We believe that we have assembled the largest and most diversified single-sample
set for investigating viruses in mammalian hosts. We focused on bats and rodents because
of their unique properties among the mammals in terms of large social group sizes,
intense social interaction and high population densities. Further criteria included
high spatial mobility in the case of bats, and high population turnover in the case
of rodents1
2
4
30. However, our data were not limited to these taxa, as an essential part of our
analysis involved database-derived PV from a large range of other mammals as well
as birds. The evolutionary distance between the analysed host taxa considerably exceeded
that of either the bats or the rodents studied. It should be mentioned that PV entries
in GenBank with and without our new data were not over-emphasising chiropteran or
rodent hosts. As shown in Supplementary Fig. S6, the majority of PV entries were from
primates, birds, carnivores and ungulates. Even after addition of our novel data to
GenBank data sets, the number of PV sequences from bats was just even with that from
ungulates.
Like almost all PVs currently contained in GenBank, our novel viruses were identified
by RT–PCR and sequencing. PCR primers have an inherent bias due to their sequence
specificity. We have applied a large range of published and own primer sets to compensate
for this bias. Some of these have been thoroughly validated on clinical panels from
a large range of hosts, confirming high sensitivity over a large genetic range of
viruses31. More recent studies have suggested to improve the detection of novel viruses
by hypothesis-free approaches such as random cDNA amplification from serum followed
by deep sequencing32. Inclusion of such an approach in our study indeed revealed novel
viral sequences that should be investigated further, but did not yield any PV. Low
serum concentrations of PV RNA may have prevented detection, reminding us that hypothesis-free
virus testing may not be sufficiently sensitive to enable comprehensive detection
of viral flora32
33.
ICTV currently lists 36 accepted PV species. Gene databases comprise more viruses,
many of which are only partially sequenced and not classified into defined species
(Supplementary Tables S6 and S7). Although 83 clearly distinct PV taxa could be discriminated
in our data set based on phylogeny, a classification criterion using a distance threshold
comparable to that between HeV and NiV (7.0–7.5% in the L-gene fragment used) identified
66 independent novel taxonomic entities on the rank of tentative species to be represented
in our data. Nevertheless, the number of novel taxa added to the PV phylogeny is unlikely
to cause a bias toward bat- and rodent viruses in subsequent analyses of host associations.
This is foremostly because reference sequences in these analyses were selected to
maximise the patristic distance within each genus of PV (Fig. 8a). The algorithm underlying
the parsimony-based ancestral state reconstruction only counts host switches once
per resolved phylogenetic root point, irrespective of the number of depending leaves
carrying identical traits. In general, probabilistic approaches such as ML- and Bayesian
methods are considered more powerful than parsimony models for the reconstruction
of character states evolving along phylogenetic branches34. However, this is mainly
because probabilistic models take branch lengths into account, whereas parsimony-based
methods only consider tree topology34. For the particular task of reconstructing viral
host associations, we favour parsimony approaches out of theoretical considerations.
First, we had to assume that host transition happens rarely and is unlikely to take
place in a bidirectional manner, because of the fitness valley effect that will prevent
host changes to be reversed easily4
10. In particular, viruses conquering a new host will leave behind population immunity
in their original AR, preventing re-introduction, and making back and forward transitions
(as well as host switching as a whole) a rare process4
10. Another argument was the uncertainty of deeper branch lengths in viral phylogenies35.
More recent studies on non-retroviral RNA viruses invading mammalian germlines demonstrated
tremendous discrepancies between apparent evolutionary rates of extant RNA viruses
versus those of phylogenetic stem lineages36
37
38
39. Rate differences in deep branches will have great influence on probabilistic models,
but will not affect parsimony assumptions. Even though recent findings of viral germline
fossils comply with the idea that ancient PV may have existed in mammals36
37
38
39, such data are so far unavailable for PV, and we have not attempted in our analysis
to determine the evolutionary origins of PV. Such an analysis would require a different
approach to sampling, as viruses from taxa other than mammals and birds are currently
underrepresented in databases. To attenuate the contribution of deep phylogenetic
nodes, we have limited our analysis to an estimation of trait switches along trees,
rather than inferring host associations for deep phylogenetic nodes. With this limitation
in mind, we can conclude that bats have most often been the donors of those viruses
currently encountered in other mammals.
In spite of our preference for parsimony-based analysis, we have challenged this result
by repetition in a probabilistic approach, testing the influence of alternative hypotheses
by either restricting the trait change model or imposing fossilised ancestral state
assumptions40. These analyses confirmed the results of the parsimony model and formally
excluded other host hypotheses within the available data set. However, it should be
mentioned that ancestral state analyses have not been systematically applied to the
theoretical problem of viral host switching before. Many open questions remain, including
the essential issue of developing significance criteria in host hypothesis testing.
Only if complex data sets like ours become available for several other virus taxa,
it will be possible to approach this major theoretical task. The current limitation,
however, is in the sparse and incomplete biological sampling of habitat.
Beyond phylogenetic analysis, we have identified traits of the natural history of
infection that suggest a specific connection between PV and bats. The epidemiology
of a morbilli-related virus was fundamentally different from that of measles virus
in humans, or that of Rinderpest in cattle. Human measles is the prototype of viruses
depending on steady transmission in sufficiently large social groups, potentially
absent from isolated and remote populations41. Black41 has defined this pattern of
pathogen prevalence as the 'introduced disease' pattern, based on observations in
isolated human tribes. Strikingly, in bats, the morbilli-related virus was excreted
by adult animals at similar rates as by young animals, which is very untypical of
morbilliviruses in other mammals. This pattern of prevalence was classified by Black41
as 'endemic—high incidence, low morbidity', as exemplified in humans by hepatitis
B virus or herpes viruses. Some researchers have argued that bats in general might
deal with viral infections differently than other mammals42. However, we have recently
described a variance of viral persistence patterns in bats that is congruent with
observations in other mammals, with a typical 'introduced disease' pattern for astro-
and coronaviruses, although a bat adenovirus in the same group of animals showed Black's
'endemic' pattern41
43. Adenoviruses provide a good template to explain the shedding pattern of the morbilli-related
virus found in the present study, as they are known to persist in tissue and to be
shed without signs beyond the acute phase of disease—a property determining the ability
of viruses to persist in small populations41
44. Accordingly, the morbilli-related virus in our study was detected in a species
of bats forming small- to medium-sized social groups (M. myotis), possibly requiring
long-term excretion for virus maintenance on group level. This is rather untypical
for morbilliviruses in other mammals that depend heavily on efficient transmission
and sufficient group size to be maintained30
41
45. In this light, the difference in organ association between bats and rodents was
quite interesting. Although PVs in rodents were associated with the kidney, favouring
excretion, their highest concentrations and prevalences in bats were seen in the spleen.
Although we have no further direct proof, this matches the concept that bat-borne
PV might not as much depend on highly efficient transmission, but might routinely
employ mechanisms of persistence to follow Black's 'endemic' pattern of prevalence41.
This anomality might indeed identify bats as AR of these viruses. Moreover, the morbilli-related
bat virus was detected in Europe, but also in an unrelated species forming rather
small social groups in sub-Saharan Africa. Even though we have not been able to conduct
longitudinal investigations of excretion in other PV genera, detection in groups without
social connection as well as re-detection in subsequent years was seen also for rubula-
and henipaviruses in this study. Detection of these viruses was not associated with
changes in serum chemistry parameters, suggesting symptomless infection despite virus
replication in internal organs, which may be regarded as typical for a virus in its
natural host context that is not dependent on efficient horizontal transmission4
7
46.
Beyond virus evolution and ecology, these data might have important implications for
public health. HeV and NiV may be of African descent and have highly diversified relatives
in Africa. These viruses might be associated with unrecognised disease, given the
tremendous number of unresolved cases of encephalitis often ascribed to malaria in
Africa47. Observed patterns of viral loads suggest that virus could be acquired during
slaughtering of bats for alimentary purposes48, but possibly also through contact
with ubiquitous bat faeces (Supplementary Tables S3 and S8). It is for this same reason
that the significance of Henipavirus-related agents in widely distributed bats in
America deserves urgent further investigation. Moreover, the finding of agents serologically
related to eradicable viruses, such as mumps, distemper and measles virus, is highly
relevant in assessing perspectives and consequences of virus eradication19
49
50
51. Clearly, the bats investigated in this study carried viruses that were only similar
but not identical to those agents endemic in humans or livestock. These new data therefore
emphasise the importance of investigating possible transmission chains, as exemplified
by the case of severe acute respiratory syndrome, in which an agent derived from bats
was probably passed to humans by intermediate hosts such as carnivores52. In the case
of the mumps-related bat virus, a direct antigenetic relatedness between human and
bat viruses has been confirmed, and the close genetic proximity between both viruses
suggests that even cross-neutralisation might be possible. In light of the still narrow
representation of genetic diversity of bats covered in this study (ca. 7.5% of bat
species), further research might reveal further bat-borne PV in close relationship
to known pathogens of humans and livestock. If antigenic overlap exists, this could
become relevant for virus eradication concepts. Relevant antigenic overlap would be
defined by proof of cross-neutralisation between reservoir-borne and human or livestock
pathogens. In this latter case, elimination of circulating virus and the subsequent
cessation of vaccination might leave humans or livestock susceptible for reservoir-borne,
antigenetically related viruses.
Although these data identify a potential reservoir of important mammalian viruses,
we can only begin to understand their true significance by functional investigation.
Knowledge on the genetic range of pathogens carried by speciose small mammals may
enable early recognition of zoonotic epidemics and rapid decision-making in the public
health sector53
54. However, much more (and different) work needs to be done to actually assess and
ameliorate zoonotic risks. The most relevant provision in this field is that epidemic
risks emanating from wildlife virus reservoirs should trigger wildlife conservation
rather than interference with wild animal populations2.
Methods
Sampling and specimen preparation
For all capturing, sampling and exportation of small mammal specimens, permission
was obtained from the respective countries' authorities. Bats and rodents were identified
by trained field biologists. Fresh bat droppings were collected on plastic film below
roost sites12. Additionally, bats were caught with mist nets at roost or foraging
sites, kept separately in bags until individual examination. Sampling relied on faecal
pellets produced in bags, vein puncture for serum samples and mouth swabs. For organ
samples, bats were euthanised with ketamine and dissected immediately. Rodents were
caught with live traps or snap traps, euthanised and dissected. For faecal specimens,
ca. 100 mg of faeces was suspended in 500 μl of RNAlater solution (Qiagen, Hilden,
Germany) immediately after collection. Suspensions were homogenised by vortexing,
and 50 μl were suspended into 560 μl of buffer AVL from the Qiagen Viral RNA Mini
kit (Qiagen) and processed further according to the instructions of the manufacturer.
For blood or serum samples, up to 140 μl (depending on the available quantity) were
extracted. For solid organs, approximately 30 mg of tissue were homogenised in a TissueLyser
(Qiagen) or a ball-mill tissue grinder (Genogrinder 2000, Spex Centripep), followed
by extraction of RNA using the RNeasy Kit (Qiagen) or the ABI PRISM 6100 Nucleic Acid
PrepStation (Applied Biosystems, Foster City, CA, USA). Elution volumes were generally
50 μl for serum/blood and faecal specimens, and 100 μl for tissue specimens. RNA specimens
were subjected to molecular screening for PVs using a panel of oligonucleotides and
RT–PCR assays listed in Supplementary Tables S9 and S10.
General conditions for RT–PCR
About 100,000 RT–PCR reactions were done for this study, using ca. 100 different protocols.
The basic formulation for 25-μl RT–PCR reactions used the Invitrogen SuperscriptIII
OneStep RT–PCR kit (Invitrogen, Karlsruhe, Germany), with 800 nmol l−1 each of the
respective first-round primers, 2.0 mmol l−1 MgSO4, 200 μmol l−1 deoxynucleoside triphosphates
each, 1 μl enzyme mix, 1 μg bovine serum albumin, 10 U RNAseOut (Invitrogen) and 5
μl RNA extract. Amplification generally involved 30 min at 50 °C; 3 min at 95 °C;
a touchdown element of 10 cycles of 15 s at 94 °C, 20 s starting at 64 °C with a decrease
of 1 °C per cycle, and 40 s at 72 °C; and 35 cycles of 15 s at 94 °C, 20 s at 50 °C,
and 40 s at 72 °C, with a final elongation step of 2 min at 72 °C. A volume of 50
μl second-round PCR reactions used 1 μl of first-round PCR product, with 1×Platinum
Taq buffer (Invitrogen), 200 μmol l−1 deoxynucleoside triphosphates each, 2.0 mmol
l−1 MgCl2, 800 nmol l−1 each of the respective second-round primers and 1 U Platinum
Taq polymerase. All primers are listed in Supplementary Tables S9 and S10. Virus quantification
was done as described previously55. Briefly, amplicons from initial nested RT–PCR
screening assays were TA-cloned (Invitrogen), plasmids purified and re-amplified with
vector-specific oligonucleotides, and finally in vitro transcribed using the T7 promotor-based
Megascript kit (Applied Biosystems, Darmstadt, Germany). Further details are available
upon request from the authors.
Next-generation sequencing
Products of random cDNA amplification were loaded on 1.2% agarose gels. Primer dimers
and large fragments (>700 bp) were removed, and amplicons were extracted from agarose
gels. Fragments were end-repaired and a 454 sequencing library was constructed according
to the GS Junior Rapid Library Preparation protocol (Roche, Penzberg, Germany). Emulsion
PCR and sequencing reaction were performed as recommended by the manufacturer. Primer
sequences were trimmed from each read, and all reads were aligned against the NCBI
virus database using the tblastx local alignment algorithm in Geneious. All hits were
scored and alignments with lengths less than 50 amino acids and a bit-score less than
40 were excluded.
Serological assays
Reactivity of human mumps virus with sera from different insectivorous and frugivorous
bats was tested using mumps virus (strain Jones)-infected cells (Euroimmun EU38, Lubeck,
Germany). Prototype PVs used for comparison included measles virus (strain Edmonston),
RSV (RSV B Wash/18537/'62 (CH 18537)), parainfluenza virus 1 (PIV1, strain Sendai
CPJ-3 13) and PIV2 (EU 18/9, strain Greer). Bat sera were diluted 1/40 and detection
was done using first a goat anti-bat immunoglobulin (Ig) (Bethyl Laboratories, Montgomery,
TX, USA) followed by a donkey anti-goat Ig conjugated to cyanine 3 (Dianova, Hamburg,
Germany). Nuclei were stained with 4′,6-diamidino-2-phenylindole. Pictures were taken
with a Motic fluorescence microscope (Zeiss). NiV indirect immunofluorescence was
performed with NiV-infected Vero cells and a strain isolated from human brain tissue
(kindly provided to AM by Jane Cardosa, Malaysia). Bat serum was diluted 1/10 and
1/40, and detection was done with the goat anti-bat Ig (Bethyl Laboratories) used
for mumps immunofluorescence, followed by an Alexa Fluor 568 donkey anti-goat IgG
(Invitrogen, Karlsruhe, Germany). A guinea pig anti-NiV serum was used as a positive
control, followed by an Alexa Fluor 568 goat anti-guinea pig IgG.
Serum chemistry
A total of 119 native sera from E. helvum bats were analysed on a Dimension Vista
automated analyzer (Siemens, Munich, Germany). The number of analysed parameters was
limited by the amount of available sample material and included lactate dehydrogenase,
alanine transferase, aspartate aminotransferase, gamma glutamyltransferase, albumin
and total bilirubin.
Phylogenetic analyses
Nucleic acid alignments based on amino acid code were done in Mega4 (www.megasoftware.net).
Gap-free coding nucleotide sequence alignments were generated containing the novel
viruses as well as reference strains, using the complete deletion option in which
any site containing gaps was deleted from the data set (see Supplementary Table S3
for a list of all reference strains with isolation years, host and GenBank accession
number). The data set used for inference of Paramyxovviridae phylogenies comprised
559 nucleotide (nt). An additional analysis was done for only those rubulaviruses
for which only a shorter 217 nt fragment could be amplified (see Fig. 2a, note that
Supplementary Fig. S1 includes only rubulaviruses with complete 559 nt coverage).
Bayesian phylogenies were calculated with MrBayes V3.256 using amino acid sequences
(WAG+G model) and nucleotide sequences (GTR+I+G model). Both analyses yielded identical
topologies (Fig. 2 and Supplementary Fig. S1). Convergence of chains was confirmed
by the PSRF statistic in MrBayes57, as well as by visual inspection of individual
traces using TRACER from the BEAST package58. Outgroups were Rabies virus (GenBank,
NC_001542) for phylogenies, including the complete family Paramyxoviridae, Newcastle
Disease virus for phylogenies of the genus Rubulavirus (GenBank, NC_002617) and Human
Parainfluenzavirus 1 (GenBank, NC_003461) for the genera Henipa- and Morbillivirus.
In parallel to the Bayesian analyses, maximum likelihood algorithms (WAG+G substitution
model, 5 gamma categories and 1,000 bootstrap replicates) were applied using PhyML
V3.059. Trees were visualised with FigTree V1.3.1 from the BEAST package58.
Estimation of known PV taxa and their GenBank representation
The known diversity of PVs and their hosts is shown in Supplementary Tables S6 and
S7. Supplementary Fig. S6 summarises the numbers of GenBank entries of PV by ordinal
host groups. Threshold amino acid distance values for classifying phylogenetic branches
were estimated by comparing the maximum amino acid distances within and between established
PV species in the corresponding sequence fragments. Measles virus, mumps virus and
RSV were selected to determine the maximum within-species distances per fragment,
based on their good coverage in sequence databases. Maximum amino acid diversity within
all publicly available mumps virus sequences was 1.6% in the translated 559 nt fragment,
5.4% for measles virus and 6.1% for RSV. For comparison, the amino acid divergence
between the species HeV and NiV ranged from 7.0 to 7.5% in this fragment. Only taxa
exceeding 7.0% amino acid distance were therefore counted as separate viruses.
Ancestral state reconstruction
Bayesian phylogenies were calculated using new PV sequences from this study, as well
as a set of reference sequences from ordinal mammalian host categories (bats, rodents,
primates, carnivores, ungulates), as well as birds. Reference sequences were selected
from GenBank to maximise the genetic distance per ordinal host category (Fig. 8).
A total of 10,294 trees were extracted from the Bayesian phylogenetic analysis and
ordinal mammalian host categories were assigned as noncontinuous state characters
in Mesquite. Numbers of reconstructed trait changes according to an unordered parsimony
assumption were summarised and averaged for each ordinal category over all trees.
Reconstructed character traits at root points were also summarised and recorded.
Hypothesis testing was done in Bayestraits40. This analysis was based on the ML phylogenetic
tree shown in Supplementary Fig. S1, including its 1,000 bootstrapped replicas. ML-based
reconstructions of trait changes were calculated for each of the 1,000 replicas in
Bayestraits, and the resulting median and mean values were listed in Excel. For hypothesis
testing, the whole analysis was repeated under different restrictions that either
synchronised the substitution rate assumptions of two different trait change processes,
or that assumed defined common ancestor nodes within the tree to be fossilised to
particular hosts. For each analysis, the median and mean log likelihoods were recorded.
Relative model likelihoods of alternative hypotheses were compared by subtracting
the respective median log likelihoods and linearising them.
Author contributions
J.F.D. and C.D. did field work in Gabon and Ghana, designed the study, did in silico
analyses and wrote the article. C.D. designed and conducted ancestral host association
analyses. V.M.C. did field work in Ghana and Germany, PCR experiments in Gabon and
Germany, and artwork for the article. M.A.M. did field work in Ghana, conducted IFA
and generated bat cell lines. G.D.M. did field work and PCR experiments in Gabon.
P.V. did field work in DRC. G.H. and A.R. designed and did field work in Costa Rica.
A.H. did field work in DRC, Gabon and RCA. C.P. did field work in DRC. T.B. did field
work in Ghana and Gabon, and PCR/cell culture experiments in Ghana F.G.R. organised
and conducted field work in Germany, Bulgaria, Romania and Ghana. S.Y. did field work
in Bulgaria. A.S. did field work in Germany, Ghana and Romania. S.O., Y.A.S. and T.K.
organised field work in Ghana. A.N.L. contributed to in silico analyses. J.S.C. conducted
field work in Thailand. A.S., A.B. C and C.R.F. conducted field work and PCR experiments
in Brazil. A.M. and S.E. performed immunofluorescence analyses for henipaviruses.
F.F. and R.B. performed immunohistochemical analyses. E.K.V.K. organised and conducted
field work in Panama and Costa Rica. R.K. did cell culture experiments (virus isolation
trials). E.N. collected bat samples in RCA. C.R. collected rodent samples in the Netherlands.
D.H.K. and S.M. collected rodent samples in SAR. R.G.U. collected rodent samples in
Germany. E.M.L. designed the study and collected bat samples in Gabon, DRC and Congo.
Additional information
Accession codes: All virus sequences reported in this study have been deposited in
the GenBank nucleotide database under accession numbers FJ971940 to FJ971960, HQ660085
to HQ660195 and JF828295 to JF828309.
How to cite this article: Drexler, J.F. et al. Bats host major mammalian paramyxoviruses.
Nat. Commun. 3:796 doi: 10.1038/ncomms1796 (2012).
Supplementary Material
Supplementary Information
Supplementary Figures S1-S6, Supplementary Tables S1-S11 and Supplementary References.