Are filovirus spillovers ongoing in wildlife hunting communities in forested Guinea?

Antibodies as evidence of past exposure

The study asks whether people living at a heavily hunted forest edge in Guinea are being infected by filoviruses, a family that includes Ebola virus. Direct detection of the virus in humans would only capture people sampled while infected, so the team instead looked for immunological footprints: IgG antibodies that bind filovirus glycoproteins and can persist after infection.

Sera came from hunters and their close household contacts in 43 villages in Macenta, forested Guinea: 511 people in 2017 and 648 in 2023, with 52% of the 2017 group resampled six years later. Participants under 18, pregnant, with chronic disease, or visiting the region were excluded. Two comparison groups anchored interpretation: 63 UK blood donors from 2020 as a presumed unexposed, non-endemic population, and pooled sera from Guinean PCR-confirmed Ebola survivors as a positive control.

Binding tells exposure apart from protection

The screening assay was a multiplex microsphere-based immunoassay carrying glycoprotein antigens from nine filoviruses, including Ebola (EBOV), Bundibugyo (BDBV), Taï Forest (TAFV), Reston, Měnglà, Lloviu and Bombali viruses. It reports binding intensity, not whether the antibody blocks the virus. Because related filoviruses share antibody targets, a sample can bind several antigens after a single infection.

To separate broad cross-reactivity from exposure to a specific virus, the team used pseudotyped viruses: lentiviruses displaying a chosen filovirus glycoprotein and a luciferase reporter. Serum that neutralizes the pseudotype reduces luminescence, and the dilution giving 50% neutralization is recorded as the NT50. These assays were run in parallel with identical protocols and input titers, but the authors note that uncontrolled variables such as defective particle ratios and glycoprotein expression could still introduce variability between assays.

Clustering separates distinct antibody profiles

Rather than applying a separate positivity cutoff for each virus, the researchers used K-means clustering to group individuals by their whole binding pattern across the nine antigens, settling on 15 clusters. This avoids labeling every cross-reactive person as a true positive and instead identifies groups with shared exposure profiles.

Three clusters totaling 57 of 1159 individuals (4.9%) showed high anti-EBOV binding plus heterotypic reactivity that declined with phylogenetic distance, a pattern the authors treated as consistent with prior EBOV exposure because it resembled that of PCR-confirmed survivors. Other clusters showed more restricted binding: BDBV in 49/1159 (4.2%), TAFV in 13/1159 (1.1%), Reston in 4/1159 (0.3%), Měnglà in 8/1159 (0.7%), Lloviu in 8/1159 (0.7%) and Bombali in 1/1159 (0.1%). These are binding-defined groups, not confirmed infections.

Neutralization supports several specific exposures

Individuals with restricted binding profiles were prioritized for pseudotyped virus neutralization. Five of ten tested individuals neutralized BDBV more strongly than EBOV-Makona, which the authors interpret as exposure to a virus antigenically closer to BDBV than to EBOV. Two of these were marital partners in a village classified as unaffected by the 2013–2016 Ebola epidemic, which the authors argue makes a shared infection more plausible than two independent atypical post-EBOV responses.

For TAFV, five of eight individuals with specific TAFV binding and minimal EBOV binding neutralized TAFV more strongly, though NT50 confidence intervals overlapped. The paper also reports likely cross-neutralization involving Bombali, Lloviu and Měnglà viruses. The authors are explicit that cluster assignments are not definitive classifications of exposure, because multi-antigen binding can arise from cross-reactivity, sequential infections, or an antigenically intermediate virus.

Forest edges and primate meat mark higher risk

Questionnaire data from 2023 describe near-universal wildlife contact: 90% (581/648) consumed wildlife meat and 94% (583/619) helped handle, wash or prepare carcasses. Wildlife was eaten a median of 7 days per month, and hunters hunted a median of 6 days per month. Bat hunting was more specialized, with 32% (207/648) hunting bats or living with a bat hunter, and bat exposure was strongly seasonal.

Mixed-effects regression on satellite-derived land cover linked landscape patterns to serology. EBOV exposure was negatively associated with tree coverage, while forest fragmentation within 500 m of the village and living in built areas were positively associated with BDBV seroreactivity. TAFV and Lloviu reactivity were positively associated with proximity to intact forest at 5–10 km. Individual questionnaire variables such as age, sex, occupation and active hunting did not significantly improve these models, but nonhuman primate consumption was associated with higher odds of BDBV cluster membership in multivariable analysis (adjusted OR 5.03, 95% CI 1.70–14.90).

What serology alone cannot settle

The design deliberately sampled hunting families, which raises the chance of detecting exposures but cannot estimate risk for the wider population. Because wildlife meat exposure was nearly universal, the study could not isolate its effect from confounders such as socioeconomic status and remoteness. Participants under 18 were excluded, so early-life exposure was not evaluated.

Village classification as affected or unaffected by the 2013–2016 epidemic is vulnerable to underreporting and misdiagnosis, and individual infection history is unknown because participants were not PCR-tested while symptomatic. Pseudotyped viruses are surrogates for authentic viruses, and serum volume and assay throughput prevented neutralizing every sample against the full panel. These constraints qualify the interpretation of individual serological profiles rather than the population-level signal.

Landscape configuration, not just contact, drives risk

The paper's distinctive claim is that exposure risk is not explained simply by who hunts or how much bushmeat they eat. Different filoviruses showed different landscape associations: EBOV with lower tree coverage, BDBV with fragmentation near villages and built areas, and TAFV and Lloviu with proximity to intact forest at larger radii. The authors interpret this as evidence that landscape configuration shapes where human–animal contact occurs, and they suggest conservation could act as an ecological countermeasure against spillover.

That interpretation is a hypothesis about pathways of exposure, not a demonstrated causal chain. What the data support is a repeated cross-sectional association between serological profiles and satellite-derived land cover, combined with questionnaire evidence on hunting and carcass handling. The practical priority the paper establishes is to treat forest fragmentation and specific wildlife contact, especially nonhuman primate consumption, as tractable risk factors worth targeting in surveillance and intervention.

Sources used in this answer