We conducted this work in accordance with approved protocols by animal ethics committees of the National Health Laboratory Service (Johannesburg, South Africa; AEC 137/12) and the University of Pretoria (Pretoria, South Africa; EC054–14). During February–November 2017, a total of 1,674 Egyptian rousette bats (February, 107 bats; April, 600; May, 563; September, 214; November, 190) were captured, aged, and sampled at Matlapitsi Cave, Limpopo Province, South Africa, as described previously (13). We collected and processed oral and rectal swab samples from each bat and blood from a subset of 423 bats, as described previously (7). Swab samples were collected and pooled by aliquoting 4 × 25 μL of the media of each sample into a microcentrifuge tube, yielding a total of 416 pools containing 100 μL of pooled rectal swab samples and 4 pools containing 50 or 75 μL of pooled oral swab samples. We conducted serologic, virologic, and molecular tests as described previously (7). In addition, we conducted real-time quantitative reverse transcription PCR (qRT-PCR) for individual swab samples when the pool tested positive. We prepared sequencing libraries using the TruSeq RNA Access Kit (Illumina, https://www.illumina.com) with MARV-specific bait enrichment, followed by sequencing on an Illumina NextSeq, genomic alignment and phylogenetic analysis (13). We calculated the significance of differences in several positive swab samples and seropositivity using the Fisher exact test in Stata version 13 (StataCorp, https://www.stata.com).

Seven rectal swab pools (5 from April 2017 and 2 from September 2017) were qRT-PCR positive; the remaining rectal swab samples collected during February–November 2017 were all negative. The number of individual positive rectal swab samples ranged from 1 to 3 per positive pool, totaling 11 positive samples. Only 1 oral swab sample pool, from April 2017, yielded a positive qRT-PCR result, containing a single positive oral swab sample (Table 1). Of 600 rectal swab samples collected during 3 nights in April, 9 (1.5%) were positive; of 215 rectal swab samples collected during 2 nights in September, 2 (0.9%) were positive. We found no significant difference between the number of positive rectal swab samples collected in April and the number collected September. The number of positive rectal swab samples differed significantly from the number of positive oral swab samples collected in April (p = 0.02). Attempts to culture marburgvirus from qRT-PCR–positive swab samples were unsuccessful. Identical results from specimens with cycle threshold (C_t) >30 were obtained in other studies (6,13).

We obtained sufficient marburgvirus-specific sequence data only from 1 of the 12 individual positive swab samples for phylogenetic analysis: a rectal swab sample, collected from a juvenile female (bat 8095) in September 2017, from which we recovered 79.2% (15.1/19.1 kb) of the genome. We merged sequencing reads from replicate sequencing runs and mapped 2,472 reads to the MARV reference genome. Maximum coverage per base obtained was 291 reads; some regions had no coverage. The average coverage per base across the genome was 18.5 reads (when we included 0 coverage regions), and the average coverage when we excluded 0 coverage regions was 40 reads. We obtained near-complete coding sequences for the viral protein (VP) 35 (972/990 nt; 98.2%) and VP40 (898/912 nt; 98.5%) genes; coverage ranged from 49.7% (VP24) to 89.3% (glycoprotein) in other open reading frames of the genome.

Figure 1

Figure 1. Midpoint-rooted, maximum-likelihood phylogeny of complete and near-complete MARV and RAVV genomes. Phylogenetic tree shows evolutionary relationships of marburgvirus detected in a rectal swab sample from a subadult Egyptian rousette female...

The marburgviruses sequence (strain RSA-2017-bat8095) detected from the rectal swab sample of bat 8095 shared a common ancestor with all other RAVV complete or near-complete genome sequences, including 3 human isolates from Kenya (14), Uganda (15), and the Democratic Republic of Congo (3) and several bat isolates from Uganda (4) (Figure 1). The RSA-2017-bat8095 nt sequence shared »77% identity with the MARV SPU191–13bat2764 Mahlapitsi strain (GenBank accession no. MG725616) that was collected and characterized from Matlapitsi Cave 4 years earlier.

Figure 2

Figure 2. Marburgvirus seropositivity in adult and juvenile Egyptian rousette bats (Rousettus aegyptiacus), Matlapitsi Cave, Limpopo Province, South Africa, February–September 2017. Numbers in parentheses indicate the numbers of bats tested...

Of 423 bats tested, 143 (33.8%) were positive for antibodies against marburgviruses (73 adults and 70 juveniles). Lowest overall seroprevalence occurred in April 2017 (17.78%) and ranged from 8.3% in juveniles (9/108) to 55.6% in adults (15/27) (Figure 2). Overall seropositivity did not differ significantly between male and female bats, but the overall seropositivity differed significantly between juvenile (forearm <89 mm; <1-year-old) and adult bats (p = 0.03). We detected seroconversion in 6 (33.3%) of 18 recaptured bats (Table 2).

The period of the lowest seropositivity in juveniles (April–May) resulting in the highest number of potentially susceptible bats at Matlapitsi Cave was the same as previously identified (13). This finding coincides with demonstrable seroconversions and virus shedding and represents a period of increased exposure. The significantly higher number of marburgvirus-positive rectal than oral swab samples we detected contrasts with results from experimentally infected Egyptian rousette bats and field studies in Uganda (4,8,9). Experimental data on marburgvirus shedding were obtained from subcutaneously inoculated and colonized Egyptian rousette bats (7–9). Whether this mode of infection represents a natural portal of entry for marburgviruses in Egyptian rousette bats and to what extent viral shedding patterns in colonized bats can be extrapolated to natural settings are unknown.

Our findings highlight the risk for marburgvirus fecal environmental contamination and for Egyptian rousette bat roosting sites as a possible source of virus spillover. Roosting behavior enabling direct physical contact suggests that fecal–oral transmission of marburgviruses in bats can occur. Biting among animals or biting by hematophagous ectoparasites might result in inoculation of wounds with contaminated feces or exposure to contaminated fomites.

Our findings, combined with earlier detection of an Ozolin-like MARV in Egyptian rousette bats roosting at Matlapitsi Cave (13), suggest local co-circulation of multiple marburgviruses genetic variants. Detection of RAVV in South Africa, closely related to East African isolates, indicates that long-distance movement of Egyptian rousette bats contributes to widespread geographic dispersion of marburgviruses. Moreover, it implies that more virulent strains, such as the MARV Angolan strain (2), might be co-circulating. Entering caves and mining have been associated with MARV spillover (3–6,14,15) and detection of viral RNA in rectal swab samples, highlight the potential route of transmission. Confirmation of the period for the highest virus exposure risk further highlights the value of biosurveillance and demonstrates that marburgviruses continue endemic circulation in South Africa. This circulation represents a potential threat that needs to be communicated to at-risk communities as a part of evidence-based public health education and prevention of pathogen spillover.