Filovirus-reactive antibodies in humans and bats in Northeast India imply zoonotic spillover

Bats are reservoirs for several zoonotic pathogens, including filoviruses. Recent work highlights the diversity of bat borne filoviruses in Asia. High risk activities at the bat-human interface pose the threat of zoonotic virus transmission. We present evidence for prior exposure of bat harvesters and two resident fruit bat species to filovirus surface glycoproteins by screening sera in a multiplexed serological assay. Antibodies reactive to two antigenically distinct filoviruses were detected in human sera and to three individual filoviruses in bats in remote Northeast India. Sera obtained from Eonycteris spelaea bats showed similar patterns of cross-reactivity as human samples, suggesting them as the species responsible for the spillover. In contrast, sera from Rousettus leschenaultii bats reacted to two different virus glycoproteins. Our results indicate circulation of several filoviruses in bats and the possibility for filovirus transmission from bats to humans.

Bats are the proposed natural reservoir of filoviruses, involved in enzootic virus maintenance and zoonotic virus transmission to susceptible hosts [4]. The majority of described filoviruses are endemic in the African continent, although filovirus-specific antibodies were detected in bats from Bangladesh [5], the Philippines [6], and Singapore [7]. The genome of a novel filovirus, Měnglà virus, was detected in bats from China [8] and is the second Asiatic filovirus described after Reston virus [9]. Lloviu virus was discovered in Spain in 2011 and detected in Hungary in 2016 [10,11]. Bats are hunted by humans across Africa and Asia, and at least 167 bat species are consumed [12]. High-risk activities, such as bat hunting and mining in bat-dwelling caves, pose a threat of cross-species filovirus transmission [13].
In the Northeast Indian state of Nagaland, local ethnic groups have conducted bat harvests for at least seven generations as a source of food and traditional medicine. These bat hunters are exposed to saliva, blood, and excreta from the bat species Rousettus leschenaultii and Eonycteris spelaea. We conducted a serological survey of both hunted bat species and human hunters to study if humans have been exposed to filoviruses potentially originating from bats.

Ethics statement
All study participants provided written informed consent by signing a form in their native language Nagamese. All human samples and surveys were collected under National Centre of Biological Sciences (NCBS) IEC permit 7/001 and National University of Singapore (NUS) IRB permit N-17-034E. Negative control sera were collected under the NUS IRB permit number H-18-029. All bats were sampled under the NCBS, Tata Institute of Fundamental Research IACUC permit #UR-6/2014, which adheres to provisions of the Prevention of Cruelty to Animals Act (1960) and the Breeding of and Experiments on Animals Rules (1998) and the NUS IACUC permit B16-0159 under the National Advisory Committee for Laboratory Animal Research (NACLAR) guidelines in Singapore.

Sample and data collection
In 2017, 85 individuals participating in an annual bat harvest in Mimi village (Fig 1) were provided a paper-based survey to record their gender, age, occupation and number of times involved in the bat harvest. Blood of consenting volunteers was collected in a serum separation tube (Vacutainer, Becton Dickinson, New Jersey, USA). Bat blood from E. spelaea (n = 16) and R. leschenaultii (n = 30) was collected by cardiac puncture after being sacrificed by the harvesters.
Blood samples were centrifuged at 1,500 x g for 10 min and sera were stored at 4˚C until transport to NCBS, where they were held at -80˚C. All sera were gamma-irradiated at 20KGy with the Blood Irradiator 2000 (BRIT, Mumbai, India) and heat-inactivated at 56˚C for 60 min prior to screening. Pooled kidney, lung and spleen samples obtained from E. spelaea (n = 34) and R. leschenaultii (n = 69) were collected and stored individually in RNALater (Sigma-Aldrich). Personnel handling potentially infectious material in the field wore N95 particulate respirators, surgical gowns, face-shields and were double-gloved. Surfaces were disinfected with 3% Virkon solution. Needles and scalpel blades were single use and disposed in sharps containers. Sealed containers were autoclaved at the Healthcare and Research Center of the Naga Hospital Authority Kohima, Nagaland, India. Small amounts of tissue (lung, spleen, kidney) was excised, combined by individual bat and homogenized in AVL Buffer (Qiagen) at NCBS. An aliquot of homogenate was used for downstream PCR analysis, while another aliquot of homogenate was used for NGS analysis.

PCR, next generation sequencing and serology
RNA was extracted with PureLink RNA Mini Kit (Invitrogen), and cDNA was synthesized using SuperScript III Reverse Transcriptase (Invitrogen). Samples were tested with a nested pan-filovirus PCR assay targeting the L gene as reported previously [14]. The positive control for the PCR was synthesized by Integrated DNA Technologies based on a region of the L gene of Bundibugyo virus (Genbank Accession: KU182911). The plasmid was amplified and extracted after transformation into competent cells using a NucleoBond Xtra Maxi Plus EF kit (Macherey-Nagel, Düren, Germany) and following the manufacturer's instruction. Purified plasmid was quantified, 10-fold serially diluted and validated with primers prior to use in the assay. Aliquots of homogenate for NGS analysis were further pooled by species (E. spelaea n = 34 and R. leschenaultii n = 34). NGS libraries were made (S1 Appendix) and validated by bioanalyzer, then sequenced on a HiSeqX Illumina machine with 2 x 150 bp reads by Medgenome Labs Ltd. (Bangalore, India) [15,16].
Human and bat sera samples were screened in a filovirus multiplex microsphere immunoassay as previously described [17].  Table 1) were expressed in a mammalian cell-culture system [18,19]. In 2018, purified, oligomeric GPe antigens (minus MLAV) were coupled to MagPlex microspheres (Luminex, Austin, TX, USA) and bat and human samples were diluted at 1:100 in PBS and run on a Bio-Plex 200 system (Bio-Rad, Hercules, California, USA) in duplicate. After sera incubation with GPe-coupled microspheres, samples were washed, incubated with biotinylated-Protein A and biotinylated Protein G (1:1 ratio) (Thermo Fisher Scientific, Waltham, MA, USA), washed and then finally incubated with streptavidin-phycoerythrin (PE) (Bio-Rad). After the discovery of Měnglà virus, the serum samples were re-run in 2019 with all GPe antigens and individual serum samples were diluted 1:100 for human and 1:250 for bat sera in PBS. Median fluorescence intensities (MFI) were measured using a MAGPIX machine (Bio-Rad) (S1, S2 and S3 Tables). Cell culture supernatant from a GPe untransfected cell line, was prepared and included in the multiplex immunoassay as a mock antigen sample to normalize non-specific antisera reactivity. Due to the absence of negative sera from the study site, we obtained seven presumptively negative human sera samples from a sample bank at Duke-NUS Medical School, Singapore. These were tested using a MAGPIX machine following the technical details described above in eight technical replicates to determine the variation of individual samples in repeat measurements to individual GPe (S4 Table).

Phylogenetic and statistical analysis
Next generation sequencing data was analysed as previously described [20]. Briefly, FASTQ files were trimmed for quality at a PHRED score of 20 and were then analysed in DIAMOND using the NCBI nr reference database [21]. DIAMOND outputs were analysed in MEGAN to determine the sequence similarly [22]. In the absence of negative serum controls from Nagaland, two independent methods were employed to define positive and negative cut-offs. The MFI values of the mock antigen were subtracted from each GPe MFI and the values were transformed to be positive, with the lowest number being 1. A log-normal model was fitted to the MFI data and the parameters of the fit were estimated in order to calculate the 95 th percentile of the log-normal distribution (S1 Fig). To control for ebolavirus cross-reactivity [23], a separate cutoff was implemented at the three-fold change above the arithmetic mean of the mock-adjusted scaled MFI. Positive samples were defined as exceeding both thresholds. The 2018 Bio-Plex data was analysed with the same two statistical methods, but not mock-adjusted (S2 Fig). The spread of MFI values for duplicate measurements of each individual sample was plotted (S3 and S4 Figs). Mean negative human sera MFI values were plotted and the standard deviation for each sample was presented (S5 Fig). All analyses and visualizations were implemented in R 3.5.1 [24]; the R code can be retrieved from the authors upon request.

Results
The majority of bat hunters were between 18 and 50 years of age, male, and participated at least eleven times in the harvest ( Table 2). All bat tissues tested were PCR-negative for filovirus-specific nucleic acid. There were a total of 13,993,300 reads from the R. leschenaultii NGS dataset and 7,975,905 reads from the E. spelaea NGS dataset and no filovirus sequences were identified.
In our 2019 serum screen (that included MLAV), we detected filovirus-reactive sera in 5.9% (5/85) of human samples, 6.2% (1/16) of E. spelaea samples, and 13.3% (4/30) of R. leschenaultii samples. The highest MFI values, corresponding to levels of filovirus-specific serum IgG, were detected for EBOV-GPe in human and E. spelaea sera, and for MLAV-and RAVV-GPe in R. leschenaultii sera (Fig 2). Our results suggest human exposure to two antigenically distinct filoviruses, the first group of sera (H10, H27, H30, H40) being reactive to EBOV-, BDBV-, and SUDV-GPe, and one individual serum (H45) singly reactive to MARV-GPe. An individual E. spelaea serum sample (E34) reacted to the EBOV-, SUDV-and TAFV-GPe, displaying a similar cross-reactivity pattern as seen for the first group of human sera and was previously reported for filovirus-positive E. spelaea samples from Singapore [7]. The 2019 screening results corroborated positive samples that were detected when screened in 2018 using a Bio-Plex 200 system (S4 Fig; H27, H30, H40, E34, R39), strengthening interpretation of positive sera samples screened in two different years with two different Luminex xMAP-based machines. There was minimal intra-and inter-individual variation in the negative human sera samples and these were all well below the MFI cutoff values for each GPe (S5 Fig).

Discussion
Despite the growing evidence that filoviruses are present in South and Southeast Asia, there has been a historical absence of outbreaks of filovirus haemorrhagic fever in this region. Reasons why clusters of human filovirus infections have not yet been detected in Asia, include (a) ecological barriers prevent zoonotic transmission, (b) viruses are unable to sustain transmission between humans or (c) an uncharacterized diversity of non-pathogenic, antigenicallyrelated filoviruses exist and cause asymptomatic infection in humans. Human populations with wildlife contact and no history of Ebola virus disease in Uganda [25] and the Democratic Republic of Congo [26] were reportedly ebolavirus seropositive. Similarly, here we report the presence of filovirus (e.g. ebolavirus, marburgvirus and dianlovirus) reactive antibodies in both human (e.g. bat hunters) and bat populations in Northeast India, a region with no historical record of Ebola virus disease. Cross-reactivity between EBOV, BDBV and SUDV of the tested samples is in agreement with a previous report [23]. The mammalian cell-culture expression system to produce oligomeric, native-like GPe capture antigens in this multiplex assay provides several benefits compared to peptide-based antigen assays, including the retention of post-translational modifications (i.e. glycosylation) and native quaternary structures allowing capture of conformational-dependent antibodies. The use of GPe from all presently described filoviruses-with the exception of the recently described Bombali virus-allows for simultaneous detection and antigenic differentiation of virus species-specific IgGs and the identification of cross-reactive IgG responses. Most ebolavirus serology surveillance studies are unable to address the complex biology of known and unknown filoviruses in terms of cross-reactivity of specific antibodies. This filovirus serological assay addresses many limitations of previously employed assays by using oligomeric, native-like virus antigens in a multiplex manner and represents an improved biosurveillance tool. Establishing thresholds for low sample number serum sets that lack both positive and negative controls is challenging, but our efforts to employ two independent statistical methods and two machine platforms yielded congruent results and is in agreement with prior approaches to estimate seropositivity [27,28].
Furthermore, this study describes serum reactivity to MLAV, incidentally in the same bat species (R. leschenaultii) mentioned in the initial report [2]. Reactivity to RESTV, which circulates endemically among bats, pigs and monkeys in the Philippines [29], and causes subclinical infection in animal care takers and slaughterhouse workers in the Philippines [9], was not detected in our study. Interestingly, MLAV and RAVV positive R. leschenaultii sera suggest circulation of two distinct filoviruses within the same species, which is serologically distinct from reactivity with E. spelaea and human samples. These results are concordant with previous findings of exposure to filoviruses antigenically closely related to EBOV, BDBV, and SUDV in E. spelaea in Singapore [17].
Two proposed mechanisms for sustained virus infection in the studied bat species are frequent co-roosting with other bats and the introduction of large numbers of susceptible juveniles into the population [30]. The two bat species sampled in this study, R. leschenaultii and E. spelaea, roost in large colonies in caves with rolling parturition patterns [31]. Though we have serological evidence of filovirus exposure, there was no genomic data detected. There are several reasons why this may be, including; small sample size, low virus copy numbers, uncertain epidemiological shedding periodicity, and high filovirus genetic diversity that is not captured by the primers employed here. Our results reinforce the need to select sentinel sites for virus surveillance at the human-animal interface and highlights some of the gaps in our understanding of filovirus transmission and ecology.
Supporting information S1 Appendix. Technical description of next generation sequencing sample preparation, library preparation, sequencing, and bioinformatic pipeline.