Comparing the two groups of people, one living near an ecosystem of bat in habited caves and forest reserves in western Uganda (exposed) and another in savanna rangeland in central Uganda (unexposed), we see that the risk of being filovirus seropositive is higher specifically among miners by at least five times compared to the unexposed group, but not higher among household members of miners or people living near the mines. Therefore, filovirus exposure risk is specifically associated with exposures occurring in the mines themselves, likely exposure to bats or their excrement or body fluids. As has been reported before, the decommissioned bat-occupied Kitaka mines where the exposed population is centred is inhabited by bats of species Rousettus aegyptiacus that are the known reservoirs for Marburg virus [3, 6, 17].
We expected a higher seroprevalence against Marburg virus than ebolaviruses, but the opposite was observed with Sudan virus seroprevalence being higher than Marburg virus. Whereas it has been confirmed during previous investigations that bats occupying the mines are actively infected with Marburg virus and had been associated with two MVD outbreaks [3, 18], no outbreak of EVD has been reported in this region. It was therefore surprising to find higher seroprevalence to Sudan virus instead of the expected Marburg virus. We cannot clearly explain why Marburg virus seroprevalence is lower than that of Sudan virus, but this is consistent with other studies where the two pathogens have been tested [10–16]. One of the explanations could be that the antibodies for Marburg virus are not as long-lasting compared to those of ebolaviruses. Studies in Egyptian rousette bats have shown that IgG antibodies against Marburg virus do not last more than 3 months . In the same study, bats were protected by secondary immunity when reinfected with Marburg virus. If a similar mechanism occurs in humans is yet to be fully understood, and further studies of the immune responses in MVD survivors in Uganda are needed. Another possibility for this finding is there could be a reservoir for Sudan virus or another closely related filovirus in Kitaka mines and/or inhabiting the area around the Kasyoho-Kitomi reserve ecosystems to which these individuals were exposed, especially the gold miners. This area is near Queen Elizabeth National Park, and so there is a possibility of having an unknown reservoir of ebolaviruses in the game reserve that has not been previously identified.
Another filovirus serological study by Nkoghe et al.(2011) in rural Cameroon and Gabonese populations where the prevalence of Ebola virus was higher in populations near forests . A second study in the Gabon found that pygmies, who are forest dwellers, had a higher percentage of ebolavirus seroprevalence than other populations at 7.02% compared to non-pygmies (4.2%) . This further indicates that communities that live in the forested areas, like the ones we studied in western Uganda are at higher risk of infection with filoviruses compared to those living in more developed or non-forested areas. Forested areas tend to have a greater abundance of fruiting trees that provide food to the fruit bats, the hypothesized reservoirs of ebolaviruses. However, in this study, going into the forest was not shown to be a risk factor for individuals being seropositive for filoviruses. While our investigation found that being a miner, but not necessarily living near the mines, is highly associated with being seropositive for filoviruses, it is likely that since the Kitaka caves are in forested ecosystem near a national park that is comparable to the central African forest, this makes the mines a more attractive location for bats to live. Gold mining has been previously described as a risk factor for Marburg virus infection in a study in DRC  with OR=13.9, 95% CI;3.1-62.1 but not for Ebola virus. We report artisanal mining and going inside the mines as risk factors for being seropositive for filoviruses, including Sudan virus, in Uganda (AOR=3.4; 95% CI 1.3-8.5). In fact, the very first cases of Ebola virus were reported in mining communities in DRC in 1976.
The four seropositive individuals we found in the “unexposed” Luweero district group could be due to travel and migration from high-risk areas but may also be due to the movement of reservoirs such as bats that are known to travel long distances hence spreading the infection. Ebola virus seropositivity has before been reported in a grassland savanna-like ecosystem in Nigeria similar to the grassland savannah ecosystem of Luweero where the four Sudan virus (3) seropositive cases were identified . However, frequent travels outside high-risk areas were protective (AOR=0.3; 95% CI 0.1-0.7). This may be because those who frequently travel away from risk areas are less likely to be exposed to the putative reservoir.
Being male was associated with a high risk of being seropositive (AOR=3.1; 95% CI 1.01-9.5) compared to being female, likely partly due to men being more likely to be miners and go inside the mines and the forests for manual work and become exposed and hence acquire infection. Cleaning a dead body was significantly associated with being seropositive for filoviruses. This has been widely reported in outbreaks of filoviruses as burials and funeral rites amplify these outbreaks. Contact with EVD/MVD suspect was a predictor of filovirus seropositivity and has been reported in a partial meta-analysis done on the risk of ebolaviruses transmissions 
Looking at the overall seroprevalence reported in this study, the findings suggest that there may be filovirus infections and outbreaks that occur in Sub- Saharan African countries and go undetected by the health care systems. This could possibly lead to large epidemics as was seen in West Africa . Additionally, our findings do not rule out the possibility that there could be cross-reactivity for filoviruses in our diagnostic assays caused by either another filovirus infection or a non-filovirus infection.
A similar unpublished study was carried out by CDC and the Ministry of Health following 2007 MVD outbreak in Kamwenge and Ibanda district in the same area. In that study, they found seroprevalence of Marburg virus at 1.2% (7/564) and Sudan virus at 1.2% (7/564). The seroprevalence of Marburg virus was slightly lower in our study whereas that of ebolaviruses was higher. We do not have a clear explanation for these differences in seropositivity between the two studies conducted in the same area eight years apart. Also, the seroprevalence in our study is lower than 8% and 3% for pooled seroprevalences reported in meta-analyses of seroprevalence of ebolaviruses performed in other parts of the world [1, 28]. Following the West Africa EVD outbreak, reports of asymptomatic infection in West African populations has been suggested in populations who had contact with EVD patients at 12%, and at 2.6% in non-contacts .
Our investigation also reported a lower seroprevalence of ebolaviruses (2.5%) than pooled seroprevalences reported in other studies in neighbouring Democratic Republic of Congo (DRC) at 10% [20, 30–33], Central African Republic and Gabon at 11% [34–38] , Sudan at 22% , Madagascar at 4% , Liberia at 13%  and Cameroon at 7% [41, 42] . However, our study showed higher seroprevalence than that reported in Nigeria at 2% , Germany at 1%  and Kenya at 1% . Only one Marburg virus seropositive person was confirmed in our study and this was much lower than has been reported in other studies [10–16]. However, further comparison of these studies is difficult due to differences in serological methods used and differences in filovirus species targeted.
These variations in seroprevalence could be due to differences in filovirus ELISA testing protocols and potential cross-reactivity caused by a non-filoviral infection. The test, developed by CDC that was used in this study has been shown to be more specific than other filovirus serological tests used in previous filovirus seroprevalence studies . This serological test was developed and validated by US Centres for Disease Control and Prevention (CDC) on known positive and negative human samples with a sensitivity >90% and specificity of >90%. However, we still see serological cross-reactivity within filovirus species even with this test. For example, one person in this investigation was seropositive for both Sudan virus and Bundibugyo virus, likely a cross-reactivity rather than representing previous exposure to Bundibugyo virus species, as has been described previously [26, 27]. Testing for filoviruses using serological tests can potentially overestimate the true level of seropositivity and therefore overestimate risk and exposure due to varying cross reactivity between differing and unvalidated serological assays used in previous studies. We are continuing to classify these serological results with validation assays including viral neutralization in order to confirm if our findings represent true undetected filovirus infections in these communities, or through cross-reaction with other viral infections, or variability in serologic assays performed.