Survey of Ebola Viruses in Frugivorous and Insectivorous Bats in Guinea, Cameroon, and the Democratic Republic of the Congo, 2015–2017

To clarify the role of bats in the ecology of Ebola viruses, we assessed the prevalence of Ebola virus antibodies in a large-scale sample of bats collected during 2015–2017 from countries in Africa that have had previous Ebola outbreaks (Guinea, the Democratic Republic of the Congo) or are at high risk for outbreaks (Cameroon). We analyzed 4,022 blood samples of bats from >12 frugivorous and 27 insectivorous species; 2–37 (0.05%–0.92%) bats were seropositive for Zaire and 0–30 (0%–0.75%) bats for Sudan Ebola viruses. We observed Ebola virus antibodies in 1 insectivorous bat genus and 6 frugivorous bat species. Certain bat species widespread across Africa had serologic evidence of Zaire and Sudan Ebola viruses. No viral RNA was detected in the subset of samples tested (n = 665). Ongoing surveillance of bats and other potential animal reservoirs are required to predict and prepare for future outbreaks.

S ince the first outbreak of Ebola virus disease (EVD) in 1976 in the northern part of the Democratic Republic of the Congo (DRC), 26 recognized outbreaks have occurred in humans across Africa; fatality rates of outbreaks have been 25%-90% (1)(2)(3)(4). Each EVD outbreak most likely resulted from independent zoonotic events.
Bats are believed to play a role in the ecology of Ebola viruses as a reservoir species (5). Bats might infect humans directly or via intermediate amplifying hosts, like nonhuman primates or duikers (6,7). Bats might serve as a source of infection in certain areas where bats are hunted and eaten as bushmeat, but infection could also occur after consumption of fruits contaminated with saliva, urine, or feces from Ebola virus-infected bats (8,9). Ebola virus emergence through exposure to bats was suspected for at least 2 outbreaks: Luebo (the DRC) in 2007 and West Africa in 2013 (10,11).
Relatively few data are available to support the role of bats in the ecology of Ebola viruses. During the EVD outbreaks of 2003 in Gabon and the Congo, Zaire Ebola virus RNA and antibodies were detected in live-caught specimens from 3 fruit bat species (Epomops franqueti, Hypsignathus monstrosus, Myonycteris torquata); virus sequences were found in the livers or spleens of a few bats (6). In subsequent studies in Gabon, the Congo, Ghana, and Zambia, antibodies were detected in additional frugivorous bat species (Eidolon helvum, Epomophorus gambianus, Rousettus aegyptiacus, Micropteropus pusillus) and 1 insectivorous species (Mops condylurus) (12)(13)(14)(15)(16). The amplification and sequencing of viral RNA of other filoviruses in bats, such as Marburg virus in bats from Africa (17)(18)(19)(20), Lloviu virus in bats from Europe (21), and new filoviruses in bats from China (22), has provided additional evidence for a possible role of bats in Ebola virus ecology.
In general, EVD outbreaks have been limited in terms of their geographic spread and chains of human-to-human transmission (1). However, during the 2013-2016 outbreak, virus spread to the urban areas of 3 countries, infecting ≈30,000 persons in Guinea, Sierra Leone, and Liberia, and ≈11,000 deaths were recorded (23). This outbreak illustrated the potential for epidemic spread from a single zoonotic transmission, with severe public health and socioeconomic impact (24). Additional studies are urgently needed to identify the animal reservoir, predict EVD outbreak risks, and improve our capacity to control epidemics.
In previous modeling studies, areas were defined as at risk for EVD outbreaks on the basis of data collected from a limited number of wildlife bat species from a few geographic regions (5,25). Also, a wide variety of serologic assays and interpretation criteria have been used, making comparison of results challenging (12)(13)(14)(15)(16)26,27). For this study, we performed a large serosurvey with a highly specific and sensitive high-throughput assay to assess Ebola virus prevalence in bats from Africa (28). We studied bats from Guinea and the DRC, countries with previous EVD outbreaks, and Cameroon, a country considered at high risk for future EVD outbreaks (5,25).

Study Sites and Sample Collection
During November 2015-August 2017, we collected samples from free-ranging frugivorous and insectivorous bats in Guinea, Cameroon, and the DRC. We captured bats at night using ground mist nets or harp traps in roosting and foraging sites. We set up ground mist nets (12 × 3.2 m) of 30-mm and 60-mm mesh sizes at different heights (1-7 m) to maximize capture of different species. We opened nets or harp traps just before sunset and checked for bats every 1-2 hours. Captured bats were released the same night immediately after sampling. Using bat whole blood taken by venipuncture of the propatagial or brachial vein, we dropped blood samples directly onto Whatman 903 filter paper (GE Healthcare, Feasterville-Trevose, PA, USA). We air-dried and preserved samples individually in plastic bags containing silica desiccant and stored them in hermetic boxes; 2-3 weeks later, we transferred dried blood spots to -20°C until needed for analysis. Data recorded in the field included information on capture site (global positioning system coordinates, ecologic environment), capture method, morphology (body measurements, weight, color), sex, age class (adult, juvenile), and species (identified visually). We collected negative control samples (n = 145) from a captive-born insectivorous bat species (103 Carollia perspicillata bats) hosted at the Parc Zoologique de Montpellier (Montpellier, France) and 2 frugivorous bat species (19 Pteropus giganteus bats, 23 R. aegyptiacus bats) hosted at Wilhelma Zoo and Botanical Garden (Stuttgart, Germany). We collected and preserved samples the same way we did for free-ranging bats.

Determination of Cutoffs
In the absence of positive control samples, we used 4 different statistical methods to determine the MFI cutoff value for each antigen (29,30) (online Technical Appendix Table  1). First, we used a general formula that involved the MFI of the 145 negative control samples, and we assigned the cutoff as mean plus 4 times the SD (mean + 4×SD). Second, we used a change point analysis (31) to identify the value at which statistical properties of the underlying probability distribution changed. This value was used to identify outliers and classify them as reactive. We used the R package changepoint (32) to calculate a single shift in the arithmetic mean with the at-most-1-change method (33). Third, we fitted univariate distributions to our data and defined the cutoff as a 0.001 risk for error, as was used in other virus serology studies (13,34). We reduced the set of candidate distributions following a bootstrapped skewness-kurtosis analysis (35). We performed fitting by maximum-likelihood estimation and selected the best-fit distribution on the basis of the Akaike information criteria with the R library fitdistrplus (36). A negative binomial distribution best-fit the data; however, we also used the negative exponential distribution as in Pourrut et al. and Laing et al. (13,34). For every antigen, we computed bootstrap values using 10,000 replicates and averaged. We performed analyses with R version 3.3.2 software (https://www.r-project.org/). We considered a blood sample reactive if the MFI of the reaction was above the cutoff. We defined Ebola virus antibody positivity as reactivity to glycoprotein and nucleoprotein of the same lineage, as was done in our previous study (28).

Nucleic Acid Extraction and PCR Screening for Ebola Virus RNA
We extracted total DNA and RNA from dried blood spots as previously described using Nuclisens (bioMerieux, Marcyl'Etoile, France) or m2000sp methods (Abbott Molecular Inc., Des Plaines, IL, USA), which are known for a high performance recovering nucleic acids from dried blood spots (37,38). For bat species from Cameroon and Guinea, we screened for Zaire Ebola virus RNA by seminested reverse transcription PCR (RT-PCR) targeting the nucleoprotein region of the virus genome. We amplified a 126-bp fragment of Zaire Ebola virus using primers NP1F1 (forward, 5′-CGGACACACAAAAAGAAWGAA-3′) and NP1R-ZR (reverse, 5′-CTCTATCTTKGTGATRTGGCTCTGA-3′) in the first round of PCR and NP1F2 (forward, 5′-TTGT-GTGCGARTAACTAYGAGGAAG-3′) plus NP1R-ZR in the second round. For species from the DRC, we performed seminested RT-PCR targeting the viral protein 35 region of the genome using the protocol of He et al. with modifications (41). In the first round, we amplified a 217bp fragment with primers VP35-F (5′-ATYATGTATGAT-CACYTVCCWGG-3′) and VP35-R (52-AGCGRATGTG-GATSACRGGT-32) and, in the second round, a 184-bp product with primers VP35-R and VP35-in-F (5′-GCTT-TYCAYCAAYTAGTRCAAG-3′).

Molecular Confirmation of Bat Species
We confirmed bat species identification recorded in the field on a subset of samples by using molecular tests. We amplified an ≈800-bp fragment of mitochondrial cytochrome b using primers cytb-L14724 (forward) and cytb-H15506 (reverse) (11,39,40). We substituted the cytb-L14724 primer with cytb-L140217 (5′-ATGACCAACATCC-GAAAATCNCAC-3′) to improve PCR performance for certain species. We purified PCR products through agarose gel (1%) and directly sequenced on an ABI 3500 sequencer (Applied Biosystems, Courtaboeuf, France). We performed BLAST analyses (https://blast.ncbi.nlm.nih.gov/Blast.cgi) to identify the most similar bat species. For samples with no or low similarity (<97%) hits with species in GenBank, we performed phylogenetic analyses with newly obtained sequences and reference sequences for different bat species using maximum-likelihood methods implemented with PhyML (http://www.atgc-montpellier.fr/phyml/) to determine genus.

Bats Antibodies against Different Ebola Virus Antigens
We tested all samples for Ebola virus antibodies. The number of samples reacting with >1 antigen was 734 (18.2%) by the mean + 4×SD method, 274 (6.8%) for the changepoint method, 175 (4.4%) for the binomial method, and 457 (11.4%) for the exponential method. Blood samples frequently reacted with glycoprotein antigens; samples reacted most with Zaire and Sudan Ebola virus antigens and least with Reston (Table 3). Simultaneous reactivity to >1 antigen (i.e., glycoprotein, nucleoprotein, viral protein 40) from the same virus lineage was rare. Simultaneous reactivity to the same antigen from different virus lineages was frequent; 32.3%-76.7% of blood samples were reactive to glycoprotein from >2 Ebola virus species, 18.4%-34.0% to viral protein 40, and 1.5%-4.4% to nucleoprotein (online Technical Appendix Table 2). When using the criterion simultaneous presence of antibodies to nucleoprotein and glycoprotein, the antibody positivity for Zaire or Sudan Ebola virus antibodies was generally <1% for all bats tested, regardless of cutoff method, and was lower among insectivorous than frugivorous bats: 0.05%-0.27% (insectivorous) and 0.06%-1.79% (frugivorous) for Zaire Ebola virus versus 0%-0.09% (insectivorous) and 0%-1.61% (frugivorous) for Sudan Ebola virus (Table 3; Figure 2). Three samples were positive for Zaire and Sudan Ebola viruses, but only by less stringent cutoff methods (i.e., mean + 4×SD).

Discussion
To clarify the role of bats in Ebola virus ecology and identify where the virus circulates between outbreaks, we tested >4,000 bats, almost doubling the total number of samples tested in all previous studies in Africa (5)(6)(7)42). We provided data on bats from Cameroon, added to the existing data on bats from Guinea and the DRC, and substantially increased the data available on insectivorous bats. We tested samples with the same assay, enabling comparison across species and countries. We used different statistical methods to determine positive sample numbers and expressed the proportion of reactive samples as a range on the basis of the different cutoff values proposed by those methods. As has been done in studies of human Zaire Ebola virus survivors (28,43), we defined Zaire and Sudan Ebola virus positivity as the presence of antibodies to both nucleoprotein and glycoprotein. As such, we estimated that 2-37 (0.05%-0.92%) bats were seropositive for Zaire Ebola virus and 0-30 (0%-0.75%) bats were seropositive for Sudan Ebola virus (  (44)(45)(46), experimental inoculation of R. aegyptiacus bats with Zaire Ebola virus leads to antibody development but infrequent or absent detection of viral RNA or shedding (44,47). R. aegyptiacus bats are therefore able to clear Zaire Ebola virus after a short infectious period without viral shedding and with little or no transmission. No antibodies or viral RNA were detected in noninoculated Miniopterus sp.      bats. The total number of frugivorous species shown to be Zaire Ebola virus seropositive has increased to 8, and 1 insectivorous bat species (Mops sp.) was confirmed to be seropositive. Zaire and Sudan Ebola viruses circulate in different species across Africa, with potential co-circulation of both viruses in some species. Although we have data on >8,000 bats from >40 species, this sample size is small, given the high numbers of bats that constitute colonies. This study illustrates the complexity of tracking the animal reservoir of Ebola viruses, not only because sampling of wild bats without performing euthanasia is difficult and time-consuming but also because of the absence of a reference standard for serologic tests.
To clarify the significance of Ebola virus antibodies, documenting the extent to which viral RNA and shedding can be detected in species with antibodies is crucial for predicting and controlling the risk for new outbreaks.
Efforts must continue not only to sample bats but also other animals to elucidate where the virus circulates in wildlife.