We investigated the potential involvement of domestic dogs in the occurrence or dissemination of Ebola virus hemorrhagic fever in humans. Based on a large serologic survey of dogs in the 2001–2002 Ebola outbreak area in Gabon, we found evidence that dogs can be infected by Ebola virus, a finding that raises important human health issues. The ELISA method was based on the use of Ebola virus–Z antigens. Although cross-reactions can occur with antibodies to other subtypes, the presence of these subtypes in our samples is unlikely because only the Zaire subtype circulates in the study area: all patients and nonhuman primates tested in this part of central Africa were infected by the Zaire subtype alone. The 2 positive dogs in France, an apparently Ebola virus–exempt part of the world, could be attributed to false-positive reactions due to the calculation of the positivity cut-off and the 1:400 serum dilution step used in the tests.
We found that 40 of 159 dogs living in the 2001–2002 Ebola virus–epidemic area had detectable Ebola virus–specific IgG, indicating either true infection or simple antigenic stimulation. All the tests were standardized at the 1:400 serum dilution, and most serum specimens had high OD values even at higher dilutions, confirming the specificity of the reactions. These data are consistent with observations we made during the different Ebola outbreaks that occurred in Gabon and the Republic of Congo in recent years. We observed that some dogs ate fresh remains of Ebola virus–infected dead animals brought back to the villages, and that others licked vomit from Ebola virus–infected patients. Together, these findings strongly suggest that dogs can be infected by Ebola virus, and that some pet dogs living in affected villages were infected during the 2001–2002 human Ebola virus outbreak. No circulating Ebola antigens or viral DNA sequences (tested for by PCR) were detected in either positive or negative serum specimens, and attempts to isolate virus from these samples failed. These findings indicate either old, transient Ebola infection of the tested dogs, or antigenic stimulation.
Symptoms did not develop in any of these highly exposed animals during the outbreak, a finding that tends to support antigenic stimulation, asymptomatic, or very mild Ebola virus infection. Wild animals, especially gorillas and chimpanzees, can also be infected by Ebola virus, but the infection is highly lethal and causes huge outbreaks and massive population declines (5
). Other animals such as guinea pigs (15
), goats (16
), and horses (17
) remain asymptomatic or develop mild symptoms after experimental infection, but Ebola virus infection has never been observed in these species in the wild. Thus, dogs appear to be the first animal species shown to be naturally and asymptomatically infected by Ebola virus. Asymptomatic Ebola infection in humans has also been observed during outbreaks (18
) but is very rare. Although dogs can be asymptomatically infected, they may excrete infectious viral particles in urine, feces, and saliva for a short period before virus clearance, as observed experimentally in other animals. Given the frequency of contact between humans and domestic dogs, canine Ebola infection must be considered as a potential risk factor for human infection and virus spread. Human infection could occur through licking, biting, or grooming. Asymptomatically infected dogs could be a potential source of human Ebola outbreaks and of virus spread during human outbreaks, which could explain some epidemiologically unrelated human cases. Dogs might also be a source of human Ebola outbreaks, such as the 1976 Yambuku outbreaks in Democratic Republic of Congo (19
), the 1995 Kikwit outbreak, some outbreaks that occurred in 1996 and 2004 in Gabon and Republic of Congo (5
), and the 1976 (6
), 1979 (20
), and 2004 (21
) outbreaks in Sudan, the sources of which are still unknown. Together, these findings strongly suggest that dogs should be taken into consideration during the management of human Ebola outbreaks. To confirm the potential human risk of Ebola virus–infected dogs, the mechanisms of viral excretion (i.e. body fluids and virus kinetics of excretion) should be investigated during experimental canine infection. This research would also offer insights into the natural resistance of dogs.
The canine seroprevalence rates in Libreville and Port Gentil, the 2 main towns of Gabon, were significantly higher than that observed in France, which suggests antigenic stimulation in these towns where no cases of Ebola infection have been observed. Epidemiologic investigations showed that most seropositive dogs in Libreville and Port Gentil had probably never had contact with an infected source (dead animal or human case-patient), and that they had never visited the Ebola virus–epidemic area, in theory ruling out true infection. They may therefore have come into contact with free viral antigens, transmitted by aerosol or, to a lesser extent, experienced conjunctival exposure to virus-laden droplets of urine, feces, or blood of the unknown natural host. Ebola virus has been shown to be experimentally transmissible to rhesus monkeys by inhalation (22
) and conjunctival exposure (23
). Moreover, accidental transmission of Ebola virus to 2 rhesus monkeys that had no direct contact with experimentally infected monkeys was observed in a biocontainment laboratory, which also suggests aerosol, conjunctival, or oral transmission (24
The Ebola virus reservoir species appears to extend throughout central Africa, both in rural and urban areas and might therefore be a small terrestrial mammal or a bird. No good candidate species has yet been identified, despite extensive studies (25
). Epidemiologic observations during the 1976 outbreaks in Democratic Republic of Congo and Sudan identified bats as a potential reservoir (6
), and Ebola virus nucleotide sequences and Ebola virus–like virus capsids were detected in rodents in the Central African Republic (27
). The discovery of Ebola virus–positive pet dogs in undeclared affected areas suggests that these animals live in close contact with the Ebola virus reservoir, and this finding should help to narrow the search.
One striking result of this study is the significant increasing gradient of canine seroprevalence from France to the Ebola virus–epidemic area, including from villages with and without human cases in the area. The Cochran-Armitage test for trends in proportions showed that seroprevalence increased linearly from France (2%), to major towns (8.9%), then to Mekambo (15.2%), and then to villages in the Ebola virus–epidemic area (25.2%). This trend is supported by the increasing seroprevalence as the sampling area approached human cases and animal sources (Cochran-Armitage test, p < 0.0001). These findings suggest that canine seroprevalence could reflect contact with the virus and, thus, virus activity in a given area and also the risk for human infection.
The virus appears to jump from its natural host to humans only in specific, but unknown, conditions. Seroprevalence rates in dogs might serves as an indicator of Ebola virus in regions in which no animal deaths or human cases have been observed.
In conclusion, this study offers the first evidence that dogs might be asymptomatically infected by Ebola virus in the wild. This finding has potential implications for preventing and controlling human outbreaks. The increasing canine seroprevalence gradient from low-risk to at-risk Ebola virus–endemic areas indicates that this seroprevalence might be used as an epidemiologic indicator of virus circulation in regions where no other means of virus detection are available.