We collected serum samples from a high-risk occupational group and their household contacts in Macenta Prefecture located within the forested region of Guinea. Samples were screened for immunological responses to EBOV using a rational, stepwise laboratory assessment incorporating a GP-specific screening ELISA, multi-target western blot analysis (GP-NP-VP40) and neutralisation assays using live EBOV. A notable proportion of this population (4.0%) demonstrated responses to multiple EBOV antigens (group A). This group included five individuals exhibiting clear EBOV Makona neutralisation and high IgG antibody titres against GP. Both these strong neutralising and other non-neutralising group A responses (n = 15) were observed in communities affected and unaffected by the 2013-16 EBOV outbreak. Further epidemiological analysis failed to demonstrate overt demographic or spatial patterns in outcomes, though we observed a consistent inverse association between forest fragmentation and these distinct EBOV serological phenotypes.
We believe the multi-faceted immune signatures observed likely occurred from exposure to temporally and phylogenetically distinct pathogens. Among the five individuals exhibiting strong neutralisation responses, we observed strikingly similar serological phenotypes to those of PCR-confirmed Makona EVD survivors. We previously showed that 1–3 years post-infection, over 95% of EVD survivors exhibited persistent neutralisation, a finding corroborated by other longitudinal studies18,25. Further, our data indicate that GP-ELISA responses were closely correlated with neutralisation titre (Fig. 3), with concentration of the former comparable to titres exhibited by confirmed EVD survivors. Other studies using sera from EBOV-Makona survivors have shown high proportions (> 99%) exhibiting multi-antigen binding (GP, NP or VP40) as seen in all five individuals in our study19. These multi-antigen responses typically persist for several years after infection, reported as high as 96% in 2017 in a cross-sectional study, with greater waning to 66% during a 5-year longitudinal investigation. Interestingly, the five individuals reported here were distributed across communities previously affected (n = 3) and unaffected (n = 2) by EBOV. However, resistance and under-reporting of cases was common in Guinée Forestière during the 2013-16 outbreak and it is plausible that unreported cases occurred in villages considered unaffected26. These participants may, therefore, represent previously undetected survivors of the 2013-16 outbreak. Given a prevalence of approximately 1% of this phenotype, this highlights the potential magnitude of under-reporting during the 2013-16 outbreak in rural areas of Guinée Forestière.
The other major sub-group we observed were individuals exhibiting antibody binding across multiple EBOV antigens in the absence of neutralisation, with these responses diverse in their combination of antigen targets and magnitude of binding. The aetiology of exposure among this group, therefore, appears less clear. Firstly, despite a low prevalence outcome, we do not consider assay specificity as influential given repeated binding of each sample across assays and different EBOV antigenic targets. Second, this group is unlikely to represent survivors of the 2013-16 outbreak. Whilst not all EBOV survivors generate neutralising antibody responses27, the vast majority do18, the responses are persistent and, our sample collection occurred in 2017 when survivors would be expected to exhibit EBOV neutralisation. A third option is asymptomatic or mildly symptomatic survivors with different immunological phenotypes than those followed-up in large-scale cohorts in west Africa, largely recruited from Ebola Treatment Units (ETUs). Research in west Africa has established that asymptomatic EBOV infection can generate IgG antibody responses yet neither neutralisation nor the degree of protective immunity has been well-characterised. The incidence of asymptomatic EBOV-Makona infection resulting in detectable antibody responses is likely to range between 3–10% among contacts of cases 2,28. If we putatively accept that all neutralisation responses we observed (5) arose from symptomatic EBOV infection during the 2013-16 outbreak, the 15 non-neutralising multi-target responses occurred at an unfeasible ratio (5:15 or 75%) even after accounting for censoring due to EBOV mortality (typical case fatality rates of 56–78%8,29 equating to 11–21:15 or 42–57%). Thus, neither symptomatic nor asymptomatic infection during the 2013-16 outbreak can explain this phenomenon in totality.
For both groups we have described, recent or historical exposure to previously undetected outbreaks of EBOV or related filoviruses must be considered. This is particularly true for the non-neutralising cohort given a lack of feasible alternative hypotheses. In the region surrounding Macenta EBOV RNA fragments, BOMV complete genomes, and viable MARV were recently isolated from bats, representing multiple ecologically feasible pathways to spillover6,10,12. The first human Marburg virus case in west Africa was also isolated in a rural setting close to our study site, emphasising the unusually high-risk of zoonotic exposure to filoviruses faced by communities in this area. The epidemiological patterns we observed in our data also provide support for this zoonotic hypothesis. Firstly, the spatial occurrence of serological phenotypes was dispersed across the study area and is representative of expectations of zoonotic filovirus exposure under stuttering chain theory given the over-dispersed nature of EBOV transmission30. Our ecological modelling analysis also revealed a consistent inverse association of EBOV-directed immunological phenotypes with ecological fragmentation measures, specifically the proximity of intact closed canopy forest.
From an immunological perspective, the serological phenotypes we describe are also plausibly explained by historical exposure to EBOV or other non-EBOV filoviruses. In the few long-term follow-up studies of EBOV survivors, individuals exhibit persistent binding, even up to 40 years post-infection. Others have suggested that over a five-year period, persistence of GP-specific antibody responses is over 75% and 67% for at least two antigens19. Regarding the possibility of exposure to non-EBOV filoviruses, the GP of Ebolaviruses is highly cross-reactive between species although the magnitude of GP antibody binding varies between Ebolavirus species17,31. EBOV convalescent sera exhibits pan-species neutralisation though the proportion of individuals exhibiting cross-species neutralisation also varies dependent on infecting strain32,33. Cross-reactivity is not limited to the GP antigen but also occurs for both NP and VP40, including in Guinean survivors of Makona infection19. Within the wider filoviridae family, sera from MARV survivors does not typically neutralise Ebolaviruses yet it can cross-react with certain Ebolavirus antigens. Previous studies have identified a specific and temporally persistent affinity for EBOV-NP34 and a conserved, immunogenic GP domain shared by Marburg and Ebolaviruses35. Interestingly, in a sub-group of our cohort (group B) we observed multiple samples characterised by NP-only WB coupled to intermediate anti-GP titre.
We believe the rational, stepwise methodology we applied to sample analysis was a major strength of our study and reinforces the need for similar approaches in future filovirus research. Stepwise approaches are particularly important given the evident ecological overlap and cross-reactivity of endemic filoviruses in west Africa. We also used objective classification methods to remove reliance on arbitrary cut-offs and control group samples acquired from unsuitable settings during analysis. Our study also has several limitations to consider. Due to sub-sampling from intermediate GP titre samples following ELISA screening, individuals who may have responded to subsequent WB and neutralisation assays may have been missed. As high titre GP-ELISA correlated well with responses on downstream assays, we do not expect these numbers to be substantial but the true number of individuals in group A may be higher than we present. Given purposive sampling of bushmeat hunters and their contacts in a limited geographical region, our findings are not generalisable but reflect exposure among one of the highest-risk occupational groups in filovirus-endemic regions of west Africa. Given epidemiological links to similar groups in previous outbreaks, community engagement and surveillance among these populations must remain a priority. Disentangling exposure aetiology in Guinée Forestière where resistance to public health response was high and disease-related stigma persists is challenging. We have attempted to objectively consider plausible explanations and believe our results overcome some of these barriers to provide insight into filovirus exposure in these communities and identify targets for future surveillance.
This study was based on the hypothesis that, zoonotic spillover events of EBOV and related filoviruses, had occurred prior to and after the 2013–2016 outbreak in Guinea Forestière. Our comprehensive multi-staged serological analysis, ecological assessment and unbiased statistical analysis support this hypothesis. The various reports of filovirus detection in bats located in Guinea, Liberia and Sierra Leone, combined with the first MARV case close to our study site, provides logical support for our hypothesis. Numerous demographic, cultural and ecological factors need to align to result in a significant outbreak and most spillover events are likely restricted to a small number of cases which do not alert local health authorities yet leave an immunological footprint that can enable subsequent detection30,36. A major lesson learned from the ongoing COVID-19 pandemic is that the international community has paid little attention to the threat of emerging viruses with pathogen prioritisation exercises based on a lack of effective intelligence37. Our study further illustrates the need to combine serological, genomic and ecological evidence in development of risk-based approaches to identify areas most likely to give rise to outbreaks. It is not practical to keep a constant watch over the entire globe but focussing on high risk locations is certainly feasible.