Abstract

A 3-month ecologic investigation was done to identify the reservoir of Ebola virus following the 1995 outbreak in Kikwit, Democratic Republic of the Congo. Efforts focused on the fields where the putative primary case had worked but included other habitats near Kikwit. Samples were collected from 3066 vertebrates and tested for the presence of antibodies to Ebola (subtype Zaire) virus: All tests were negative, and attempts to isolate Ebola virus were unsuccessful. The investigation was hampered by a lack of information beyond the daily activities of the primary case, a lack of information on Ebola virus ecology, which precluded the detailed study of select groups of animals, and sample-size limitations for rare species. The epidemiology of Ebola hemorrhagic fever suggests that humans have only intermittent contact with the virus, which complicates selection of target species. Further study of the epidemiology of human outbreaks to further define the environmental contact of primary cases would be of great value.

A thorough understanding of the epidemiology of transmission of a pathogen from its wild reservoir host to humans is essential to design effective surveillance and control schemes for zoonotic diseases. Intensive surveys have followed previous outbreaks of Ebola (EBO) hemorrhagic fever (EHF), but the number of outbreaks has been small and the reservoir for EBO virus remains unknown [1–4]. Although all previous searches have yielded negative results, composite data generated in the laboratory (e.g., cell culture susceptibility, limited experimental infection of animals) and theoretical considerations suggest that small mammals are the most likely reservoirs [5–7].

After previous EHF outbreaks, an extended delay hampered searches for the virus reservoir, often with ⩾1 year passing between the outbreak and the organization and start of ecologic surveys. In most cases, the primary case (i.e., the person who became infected from the wild reservoir and began the human-to-human transmission chain) could not be identified with sufficient certainty or had traveled large distances before the onset of illness. Often this person was deceased, and information about his or her activities was fragmentary. This limited the effective focus of the surveys [1–4]. During the 1995 outbreak in Kikwit, Democratic Republic of the Congo (DRC), an international team tentatively identified a local resident as the primary case. This individual had not traveled before becoming ill, and a likely transmission locality could be established. Although the primary case had become infected in December 1994 and the ecologic study team arrived in June 1995, it was reasonably likely that an intensive study might identify the reservoir.

From 10 June to 28 August 1995, we conducted an extensive survey under World Health Organization (Geneva) auspices and the sponsorship of the Centers for Disease Control and Prevention (CDC, Atlanta). Here we report on the collection of vertebrates gathered during this survey and present the results of the virologic studies of these materials. We discuss the problems that are inherent to the search for the reservoir of a rare zoonosis and make some suggestions for future research. The report on the collection of invertebrates is presented separately [8].

Materials and Methods

Selection of collection sites

The putative primary case was a farmer who lived in Kikwit but worked several maize and cassava fields in the forest at Mbwambala, about 8 km southeast of Kikwit [8, 9]. He had recently (December 1994) also excavated a charcoal pit near his fields. He had not traveled beyond the environs of Kikwit during the months before becoming ill, but he did visit his fields and charcoal pit daily. Since he was the only probable EHF case identified with no definitive link to a prior human case, it seemed unlikely that he had become infected in or around his home in the densely populated town of Kikwit. Therefore, the search for the EBO virus reservoir focused primarily on the forest biotopes near his fields, which could be exactly located (figure 1), and limited small-mammal trapping and livestock sampling were also conducted near his home.

Figure 1

Location of trapping sites around Kikwit, Democratic Republic of the Congo, where vertebrates were collected in 1995. Trap lines are identified within parentheses.

Figure 1

Location of trapping sites around Kikwit, Democratic Republic of the Congo, where vertebrates were collected in 1995. Trap lines are identified within parentheses.

Because of the remaining uncertainty about the identification of the primary case and the lack of information on this person's activities or behavior that could have influenced his contact with a variety of vertebrates from areas surrounding Kikwit, additional collection sites were selected in order to maximize the number of sampled habitats (figure 1). In addition, small game animals (e.g., large rodents, small carnivores, antelopes, pangolins, monkeys) were purchased from hunters or at “bush meat” markets. The exact origin of these specimens could not be established with certainty; often they came from considerable distances (>50 km) since most game has disappeared from the Kikwit area due to hunting.

To investigate the possibility of virus transmission from EHF patients to commensal rodents, we also placed traps in and around the pavilion that housed EHF patients at the Kikwit General Hospital and the hospital morgue, where deceased EHF patients were held until burial. When we made these collections, there were still acute EHF cases in the pavilion, but the last patient death at the hospital occurred the day small-mammal trapping began there (24 June 1995). At the peak of the EHF epidemic, when basic hygiene was low, small mammals had access to patient excrement, medical wastes, and cadavers on the pavilion floor, in the semi-attached privies, and in the area surrounding the pavilion where excrement and medical wastes were discarded.

Description of principal field sites

Below, we list the six principal trapping sites together with a general description of the habitats. At these sites, trapping took place daily throughout the indicated periods, although individual trap lines may not have been operational for the entire period. Additional collection sites included the Kinzambi Mission (04°58.631′ S, 18°46.470′ E), the Sacred Heart Mission (05°01.719′ S, 18°50.783′ E), and the Trappist Mission (05°02.262′ S, 18°51.192′ E), where bats were netted and trapped. At Luano, Kimbinga, Kimputu Nseke, and Mbalaka markets, species used as bush meat and other assorted species were purchased.

The first trapping site was Mbwambala (field 1 and charcoal pit: 05°03.471′ S, 18°54.552′ E; field 2: 05°03.958′ S, 18°54.838′ E).

Trap lines A–G, H, K, N, SC1, SC2, S1, S2, and Q were set from 10 June to 1 August 1995. The site had a highly disturbed secondary forest on hilly terrain with generally very steep slopes. The area was a patchwork of four vegetation types, one of which was an immature secondary forest that occurred in small patches and had a canopy height of 15–20 m. The overstory generally provided much less than 100% cover and was dominated by Musanga cecropioides. The shrub layer (as high as 5 m) was also generally thin and was dominated by Caloncoba welwitschii and Trema orientalis. A dense herbaceous layer (100% cover) of ∼2 m in height was dominated by Haumania liebrechtsiana and Palisota ambigua. This dense herbaceous vegetation provided for a relatively cool and moist microenvironment at the soil surface.

Another vegetation type in Mbwambala was the large patches of highly disturbed fallow land (probably abandoned maize and cassava fields) covered by thick, low brush, which usually was a monoculture of 1- to 2-m-high Chromolaena (Eupatorium) odoratum. These areas were dense and nearly impenetrable but hot and relatively dry at the soil surface because of the lack of an herbaceous layer.

The third type of vegetation in Mbwambala was cultivated patches of cassava, maize, or, rarely, bananas. These patches were generally cleared of other vegetation and were hot and dry at the soil surface.

The last type of vegetation in Mbwambala was linear patches of dense mesic vegetation along watercourses. These habitats were dominated by dense herbaceous vegetation covering soil that was moist to wet. Plant species diversity was high; typical species included Aframomum species, Costus species, and Ataenidia conferta.

The second trapping site was Kakoi (05°05.884′ S, 18°57.598′ E), where trap lines KK1–KK4 were set between 11 July and 5 August 1995. The area included a rare patch of primary forest and a secondary forest. Both patches were on nearly level ground.

Trap lines KK3 and KK4 were set in the primary forest, a relatively undisturbed patch characterized by a multilayered canopy, high diversity of plant species, and little light penetration to the soil level. The area had five vegetative strata: (1) a principal overstory canopy dominated by Brachystegia laurentii and Goss-weilerodendron balsamiferum and providing about 90% canopy closure at a height of 40–45 m; (2) a secondary overstory canopy at 25–30 m composed principally of Strombosia pustulata; (3) a primary understory layer (10–15 m high) dominated by Diospyros bipidensis and providing ∼70% closure; (4) a secondary understory layer dominated by 2- to 8-m-tall Crotonogyne poggei; and (5) an herbaceous layer (only ∼0.2 to 1 m high), which was less dense than in the secondary forest. It was dominated by Leptaspis cochleata and Palisota species.

Trap lines KK1 and KK2 were set in the secondary forest. The canopy in the secondary forest site was much more open and had greater light penetration than that found in the primary forest. The overstory canopy layer (15–18 m high) consisted primarily of M. cecropioides and provided only about 30% canopy closure. The understory layer was 5–6 m tall, provided about 75% cover, and was dominated by C. welwitschii and T. orientalis. The herbaceous stratum, which reached 2–3 m in height and provided 90% cover, was dominated by H. liebrechtsiana and Palisota ambigua.

The third trapping site was Kilombo Savanna, where trap lines SA1–SA3 were set between 26 June and 1 July 1995; SA1 was set at 05°09.325′ S, 18°48.466′ E, and SA2 and SA3 were set at 05°09.116′ S,18°46.828′ E. The savannas in the Kikwit area were anthropogenically derived and artificially maintained by frequent burning. There were two vegetative strata: an upper stratum of scattered bushes or small trees that reached only 2–6 m in height and an herbaceous stratum consisting of annual grasses and forbs, which provided nearly 100% ground cover. The arborescent vegetation in the Kilombo Savanna area consisted of Erythropheum africanum, Pterocarpus angolensis, and Hymenocardia acida. The herbaceous vegetation was < 1 m high and dominated by Pteridium aquilinum, Loudetia arundinacea, and Loudetia demeusei. Trap line SA1 was located in a savanna—secondary forest ecotone.

The fourth trapping site was in Ngome Savanna (05°10.554′ S, 19°06.997′ E), where trap lines SA4–SA6 and SA7 were set between 1 and 21 July 1995. The physiognomy of the derived savanna near Ngome was similar to that at Kilombo. The dominant bush species were H. acida and Albizia adianthifolia; the herbaceous layer was dominated by the grass L. demeusei.

The fifth trapping site was Wamba, where trap lines WA1 (05°06.644′ S, 18°55.202′ E), trap line WA2 (05°06.742′ S, 18°54.96′ E), and trap line KB1 (05°03.630′ S, 18°52.428′ E) were set between 28 July and 28 August. Trap lines WA1 and WA2 were in “bush,” with a thin overstory of scattered palms reaching 10–15 m in height, a shrub understory with a height of 2–3 m providing 50%–75% cover, and a thick herbaceous layer with a height of 30–40 cm providing 70%–90% cover. Trap line KB1 was in a savanna habitat, with scattered small trees and shrubs (3–12 m tall) and a 70%–80% herbaceous cover. Plant species identifications are not available.

The sixth trapping site was Kwanga-Ngamzi (05°09.480′ S, 18°56.366′ E), where trap lines MN1–MN4 were set between 21 July and 6 August. Trap lines MN-1 and MN-2 were located in secondary forest covering abandoned agricultural fields. The overstory consisted of Elaeis guineensis (∼6 m in height), which provided ∼20%–30% cover. A 3-m-high understory characterized by Costus lucanusianus and C. odoratum provided 70%−80% cover. The herbaceous layer was dominated by Hyparrhenia diplandra, which provided 70%−80% ground cover.

Trap line MN-3 was also in regenerating secondary forest. The overstory provided 50%−60% cover and included 15-m-high E. guineensis and Pycnanthus marchalianus. The understory was ∼4 m high, provided 50%–60% cover, and consisted of Megaphrynium macrostachyum. The herbaceous layer was sparse (20%−30% cover), ∼40 cm high, and consisted primarily of H. liebrechtsiana.

Trap line MN-4 was located in a hilltop cassava field that contained cultivated cassava and scattered palms and C. odoratum.

Small vertebrate collection

Trapping focused on small mammals and birds, while larger mammals (>2 kg) were mostly purchased. Occasional captures of other vertebrates were included in the collection. About 300–350 live-capture Sherman traps (8 × 9 × 23 cm; H.B. Sherman Trap, Tallahassee, FL) were laid out each night in lines, with ∼5–10 m between traps. The bait consisted of peanut butter mixed with rolled oats and dried fish or, in some cases, oil palm nuts. Tomahawk wire mesh live-capture traps (Tomahawk Live Trap, Tomahawk, WI) were laid out in groups or scattered and baited with cassava pieces smeared with peanut butter or, for the large traps, fish and chicken viscera to attract carnivores. Forty of the traps were Tomahawk model no. 102 (13 × 13 × 41 cm), and 20 were model no. 108 (25 × 30 × 81 cm). Where appropriate, some traps were attached to horizontal branches. At Mbwambala, Kwanga-Ngamzi, and the Ngome Savanna, we placed 2 1 80 m plastic sheet drift fences (50 cm high) with an unbaited pitfall trap (a buried 15 L bucket, without water) every 10 m. About 15 Victor gopher traps (Woodstream, Lititz, PA) were set each night for 2 weeks in active rodent burrow excavations found at a single savanna site.

Mist nets were placed at locations where bats or birds had been observed or were deemed likely to occur and were set at ground level or elevated 2–2.7 m, depending on the site characteristics. Most nets were operated 24 h per day and checked regularly; however, near buildings they were used only shortly before dusk. A handmade Tuttle bat trap was operated at mission sites and collected molossid bats only.

All traps were checked daily in the early morning. Live-capture traps that contained animals were placed in double plastic bags and tied closed before being transported to an isolated, outdoor central processing area. Animals were sampled following standardized procedures [10, 11]. Before opening plastic bags containing animals, dissectors donned disposable surgeons' gowns, double latex gloves, and powered air-purifying respirators fitted with HEPA filters. After animals were anesthetized with methoxyflurane, weight, sex, and body measurements were recorded, and blood samples were obtained from the retroorbital sinus with capillary tubes or by cardiac puncture in small mammals and from the brachial or jugular vein in birds.

Animals were euthanatized by cervical dislocation, overdose of anesthetic, or inhalation of CO2 and preliminarily identified to genus or species level before tissues (spleen, kidneys, liver, lungs) were removed by use of sterile scissors and forceps. Large mammals, which were usually dead when purchased, were sampled by obtaining blood from the heart and, when possible, a small amount of tissue from other organs. Unusual necropsy findings were recorded and, when appropriate, photographs were taken. All samples were placed in labeled cryovials and stored in liquid nitrogen until shipped on dry ice to CDC. In total, 3066 blood samples were collected, including samples from dogs, cattle, and pet primates.

Carcasses were labeled, fixed in formalin, and sent to the University of Antwerp, Belgium, where they were rinsed for 4 days and stored in alcohol until further identification and taxonomic study. Animals were identified at the Museum Alexander Koenig, Bonn, Germany (shrews); the Royal Museum for Central Africa, Tervuren, Belgium (birds and lizards); and the University of Antwerp (bats, snakes, rodents, other groups). Among the 2544 preserved carcasses, confirmed identification was obtained at least at the genus level for 2493 specimens; for the others, inadequate material or lost or damaged labels prevented unambiguous identification, and herein we will use the preliminary field identification for these specimens. The collected carcasses will be kept as voucher specimens in the Royal Museum for Central Africa and the Museum of Southwestern Biology, Albuquerque; a small reference collection of rodents will be deposited at the University of Kisangani, DRC. Tissues will be deposited at the Museum of Southwestern Biology.

Serology and virus isolation

After arriving at CDC (Atlanta) the blood samples were organized by species for those species for which commercial conjugates were available (rodents, insectivores, chiropterans, and ungulates, as determined by testing conjugate reactivity with blood adsorbed onto polyvinyl chloride [PVC] microtiter plates). Testing of other mammalian species, birds, and reptiles has been deferred until more suitable conjugates or indirect means of antibody determination are developed. An intermediate dilution of 1:25 was prepared from the blood samples, frozen, and then irradiated with 20,000 Gy (2 × 106) of gamma from a 60Co source. For those species tested, an antigen was prepared by basic buffer with detergent extraction from Vero E6 cells infected with the EBO (subtype Zaire) virus and adsorbed to wells of PVC microtiter plates. Sera were tested by an ELISA against this antigen as well as an antigen extracted from similarly prepared mock-infected cells. The sera were tested at dilutions of 1:100–1:6400 in 4-fold dilutions against both the positive and negative antigens. The optical density (OD410) of the mock antigen—coated wells was subtracted from that for the corresponding EBO virus antigen wells. A positive control was available from experimentally infected laboratory mice and was reactive with the conjugates used against rodent species. The tests for other species, such as ungulates, could only be indirectly controlled.

Virus isolation was attempted in a biosafety level 4 laboratory from the spleen (or liver if spleen was unavailable) of those animals from which the organ had been collected. The spleen was triturated in a volume of Hanks' balanced salt solution with 5% heat-inactivated fetal bovine serum to yield a 10% wt/vol suspension. After trituration, the suspension was divided into three aliquots: One was added to a small microtube for EBO virus antigen detection, another was frozen at below −70°C, and another was used for tissue culture isolation attempts.

A portion of the suspension, 0.2 mL, was inoculated, without refreezing, onto confluent monolayers of Vero E6 cells in a T25 flask and adsorbed with constant rocking for 1 h at 37°C. Eagle MEM with Earle's balanced salt solution with 2% heat-inactivated fetal bovine serum containing 20 mg gentamycin and 50 U nystatin/mL was added to the flask, and about every other day for 14 days, the cell monolayers were observed by use of a microscope. Maintenance medium was changed at day 7, and if no cytopathic effect was observed earlier, cells were removed from the flask with 3-mm glass beads, and a portion of the cells was centrifuged and resuspended in borate saline and applied to 4 wells of triplicate teflon-coated microscope slides.

Another portion of the culture was frozen and held at −70°C. Slides were air-dried and then irradiated with 20,000 Gy of gamma while refrigerated on dry ice. The slides were then fixed in acetone at room temperature and stained with a polyvalent anti-EBO hyperimmune rabbit serum made by immunizing rabbits with EBO (subtypes Zaire, Sudan, and Reston) viruses, followed by goat anti-rabbit conjugated to fluorescein isothiocyanate. Appropriate EBO virus control slides were used with each batch of slides, stained to ensure that the antiserum and fluorescein isothiocyanate conjugate were working.

Antigen-detection ELISAs were performed on all tissue homogenates after irradiation of the material with 20,000 Gy. The tissue homogenates were tested at dilutions of 1:4, 1:16, 1:64, and 1:256 as previously described, with slight modifications, for detection of antigen in infected primates [12]. Positive controls for the 4 known EBO virus subtypes (Zaire, Sudan, Reston, and Côte d'Ivoire) were run with each assay.

Results

Vertebrate collection

Most of the 3066 collected specimens were mammals (87%, 2663), followed by birds (9%, 265) and reptiles and amphibia (4%, 129) and 9 specimens from other taxa (figure 2). Among the mammals, most of the collection consisted of rodents (72%, 1914), bats (20%, 539), and insectivores (4%, 115). Small numbers of specimens from the orders Carnivora, Primates, Artiodactyla, Pholidota, and Macroscelida were also obtained.

Figure 2

Faunal composition of the vertebrate collection made around Kikwit, Democratic Republic of the Congo, 1995.

Figure 2

Faunal composition of the vertebrate collection made around Kikwit, Democratic Republic of the Congo, 1995.

Overall, 78 mammal species, 51 bird species, and 22 species of reptiles and amphibians could be differentiated. Among mammals, diversity was highest in Rodentia (29 species), followed by Chiroptera (18 species) and Insectivora (10 species: table 1). The distribution of the number of specimens collected from each mammal species was skewed (figure 3). Few specimens were obtained from the majority of the species, with <10 specimens available from 48 mammalian species and 22 species represented by a single specimen. There were 6 species, accounting for 66% of all collected mammals, for which >100 specimens were collected.

Table 1

Number of vertebrates collected per locality (specimens from which blood samples were tested by ELISA for antibodies to the Zaire subtype of Ebola virus and from which virus isolation was attempted from the spleen).

Table 1

Number of vertebrates collected per locality (specimens from which blood samples were tested by ELISA for antibodies to the Zaire subtype of Ebola virus and from which virus isolation was attempted from the spleen).

Figure 3

Distribution of abundance of species in the vertebrate collection made around Kikwit, Democratic Republic of the Congo, 1995.

Figure 3

Distribution of abundance of species in the vertebrate collection made around Kikwit, Democratic Republic of the Congo, 1995.

Serology and virus isolations

ELISA testing for EBO virus antibody has been completed on 2393 of the 2906 blood specimens available for testing (table 1). No antibody against the Zaire subtype of EBO was detected among specimens that could be tested.

Virus isolation has been completed on 2730 collected animals for which a spleen was available. For an additional 84 specimens, no spleen was available, so isolation attempts were made upon the liver. All attempts at isolation of EBO virus have been negative. However, a number of viruses preliminarily identified as arenaviruses were isolated and will be the topic of a future report. None of the 2814 tissue suspensions were positive for EBO virus antigen by ELISA.

Discussion

Despite extensive efforts in the field after this and other EHF outbreaks, the EBO virus reservoir remains unknown [4]. A number of potential reasons that are inherent to these kinds of studies may explain why field investigations have yielded disappointing results. For future work, it is important to explicitly acknowledge the existence of these problems.

Although our ecologic field team arrived soon after the outbreak was reported and confirmed, 6 months had already passed since the putative primary case became infected. Ecologic sampling was done during the dry season, whereas the potential primary transmission event happened during the wet season. Seasonal variation in the composition of small mammal species is small but does exist in tropical rain forests, and some species may display considerable fluctuations [13–16]. Such temporal variation would have resulted in the sampling of a faunal assemblage that differed in composition, abundance, and diversity from that present at the time of the infection of the primary case. The infection status of the reservoir species could also be significantly influenced by seasonal variation (e.g., if arthropod vectors are involved in the enzootic maintenance of EBO infection). Moreover, it is possible that animals infected at the time of the primary case's infection did not survive long enough to be represented in our sample. Similar doubts exist concerning the site selected for the field work. Although the identification of a relatively well-documented primary case enabled the implication of a likely infection site in which to focus our surveys, there inevitably remained some doubts that made the concentration of all efforts at a single site too risky.

These uncertainties were further aggravated by the logistical problems that will likely encumber any investigation in isolated areas, particularly in developing countries: transport and power supply problems, the cold-chain maintenance of biologic materials, lack of up-to-date maps of the area to help select sample collection sites, recruitment of specialist staff members, and biosafety concerns).

More fundamentally, there are several other problems that are inherent to the search for the reservoir of a pathogen such as EBO virus, and these problems require that we make certain assumptions. Our investigation was largely based on three key assumptions derived from our current understanding of the properties of the virus and epidemiology of the disease. The virus has several characteristics that suggest that mammals are the most likely host [5, 6, 17]. Therefore, assumption one was that the reservoir is a mammal. Second, documented EHF outbreaks in Africa have always been linked to rain forests, both for human outbreaks and for epizootics and occasional infections in chimpanzees [1, 2, 18–20]; therefore, assumption two was that the reservoir is at least a part-time forest species. Last, the number of reported EHF outbreaks is very low, suggesting that the probability for humans becoming infected from the wild reservoir is very low. Thus, assumption three was that such a pattern of disease could be explained if the reservoir is very rare, if the reservoir is not rare but occurs in habitats or has behavioral patterns that support only rare contact with humans, or if the virus in the reservoir population is inefficiently transmitted to other species yet effectively maintains itself in the reservoir species (e.g., sexual transmission), or a combination of these factors.

Our field work was set up according to assumptions one and two. Assumption 3 was difficult to take into consideration for practical and philosophical reasons. Indeed, the rapid initiation of field work was not compatible with the sampling of habitats where humans normally are not active (e.g., in the canopy, under the ground surface). More important, however, our investigation could not focus on a specific subgroup of rare species since there was no evidence implicating such a group or suggesting which field techniques should be used to target a specicific species; thus, we hoped that the reservoir could be discovered by undertaking a broad-based traditional collecting effort, with some assistance from good fortune and hard work.

The very skewed distribution of specimens among species is typical of natural habitats [21] and, tautologically, rare species are rarely caught. Moreover, unless the prevalence of infection is high, the small sample size from rare species makes it unlikely that infected animals would be trapped. This statistical problem necessitated the collection of large numbers of specimens from the common species whenever possible. In addition, there was a chance to detect a collateral infection even if these species were not the true reservoir. Inevitably, the amount of work involved in processing the large number of collected animals in the field decreased the effort we could invest in capturing rare species.

Furthermore, it was considered unacceptable not to sample all specimens that were collected (even if they belonged to common groups) because of current taxonomic difficulties regarding mammals, particularly in relatively isolated or unstudied areas, such as those where EHF outbreaks have occurred [22]. For many small mammal species, identification in the field is not possible. Often, it is not known which species of a certain genus occur in the area. Superficial identification or keeping a limited number of reference specimens may not be enough, since some species can only be distinguished by cranial morphology or even genetic techniques (e.g., karyotyping or DNA-sequencing). Finally, the alpha-taxonomy may not be complete, meaning that the specimens may belong to an unrecognized species. This latter problem is not uncommon for small mammals; for example, in the Kikwit collection, there were at least 5 species of the shrew genus Crocidura, while 4 other groups cannot yet be named and may eventually turn out to be separate species (table 1). Although the exact given name of a species may seem irrelevant in the present context, the recognition of specific taxa is important from an epidemiologic point of view. For example, within the genus Mastomys there are two cryptic species that can occur together and can be recognized only by the number of chromosomes; one of them, M. natalensis, is resistant to plague while the other one, M. coucha, is very susceptible to it [23]. There are many instances in medical entomology in which poor taxonomy has led to ineffective or even disastrous pest management strategies [24–25].

Moreover, several technical issues are of primary concern in studies to find the reservoir of a virus. It is not known whether EBO virus persistently infects the true reservoir species, such as occurs for hantaviruses [26] and arenaviruses [27] in their rodent hosts. If such persistence does occur, virus isolation would offer a reasonable opportunity for success and would allow identification of the reservoir. On the other hand, if the period of virus infection is relatively brief, the probability of recovering an EBO virus isolate would be much lower, and detection of antibody as evidence of past infection might be the best primary means of detection. It is also possible that the virus exists in some cryptic form that makes it difficult to isolate from tissues of the reservoir species, although this seems unlikely since EBO viruses are readily isolated from infected patients.

The quandary for those seeking the reservoir of EBO viruses is that insufficient information is available to choose a method that would limit the search to a smaller subset of animal species or to select a single technology for identifying the reservoir species. Other technical issues that complicate the laboratory testing for the evidence of the virus are the difficulties in testing for antibodies in a wide variety of taxa. Efforts are underway to remedy this.

Multiple outbreaks of EHF in humans and primates have occurred recently in Gabon, and genetic analyses of the viruses recovered confirmed that each outbreak was due to an independent introduction of the virus from a putative reservoir [19]. This relatively high level of virus activity suggests that Gabon is a promising site for future ecologic investigations.

Conclusions

Despite considerable effort in the field and in the laboratory, no evidence was found of EBO virus infection in the fauna collected during and immediately after the 1995 EHF epidemic in Kikwit. Uncertainties about where and how the first human case might have contracted the EBO infection and the time that elapsed since the beginning of the epidemic led to difficulty in drawing solid conclusions from these negative data. Furthermore, even if these limitations were not at issue, sample size considerations would limit our ability to form exclusionary conclusions on all but a few commonly collected species of animals. If one uses 100 animals as a sample that allows such conclusions, very few species can be excluded for future considerations: 6 total species (bats: 1 genus, 2 species; rodents: 3 genera, 4 species).

Even so, the approach we used would be the most appropriate one were another outbreak of EHF to occur. The probability of detecting the virus from the reservoir is small, but the collection allows the description of the local fauna and suggests possible targets for laboratory-based experimental investigation. Ultimately, accumulating epidemiologic information combined with timely field and appropriate laboratory investigation will result in the answer to the EBO reservoir question.

Additional Study Participants

Additional study participants included D. S. Bressler, M. Curtis, M. L. Martin, L. Morgan, K D. Wagoner, and A. J. Williams, Special Pathogens Branch, J. A. Comer, J. Liz, G. O. Maupin, and J. G. Olson, Viral and Rickettsial Zoonoses Branch, DVRD, and R. McLean, DVBID, CDC, Atlanta; M. Colyn, University of Rennes, Rennes, France; F. De Vree, J. Hulselmans, V. Van Cakenberghe, E. Van der Straeten, W. Verheyen, and W. Wendelen, University of Antwerp, Antwerp, and M. Louette and D. Meirte, the Royal Museum for Central Africa, Tervuren, Belgium; R. Hutterer, Museum Alexander Koenig, Bonn, Germany; B. Ilenga and J.-B. Katshunga, Kikwit, and A. Lubini, Institut Supérieur Pédagogique de la Gombe, Département de Biologie, and M. A. Mandango, Institut Pédagogique National, Département de Biologie, Kinshasa, Democratic Republic of the Congo; K. Kargbo and J. Koniga, Guinea Field Station, Special Pathogens Branch, DVRD, CDC, Kindia, Republic of Guinea; and C. Merriman, York University, York, Canada.

Acknowledgments

During the field work, we appreciated very much the efforts of our local team of trappers, the support from the CDC logistics team, and the company of our Congolese and international colleagues in Kikwit.

References

1.
World Health Organization
Ebola haemorrhagic fever in Zaire, 1976
Bull World Health Organ
 , 
1978
, vol. 
56
 (pg. 
271
-
93
)
2.
World Health Organization
Ebola haemorrhagic fever in Sudan, 1976. Report of a WHO/International Study Team
Bull World Health Organ
 , 
1978
, vol. 
56
 (pg. 
247
-
70
)
3.
Heymann
DL
Weisfeld
JS
Webb
PA
Johnson
KM
Cairns
T
Berquist
H
Ebola hemorrhagic fever: Tandala, Zaire, 1977–1978
J Infect Dis
 , 
1980
, vol. 
142
 (pg. 
372
-
6
)
4.
Breman
JG
Johnson
KM
van der Groen
G
, et al.  . 
A search for Ebola virus in animals in the Democratic Republic of the Congo and Cameroon: ecologic, virologic, and serologic surveys, 1979–1980
J Infect Dis
 , 
1999
, vol. 
179
 
suppl 1
(pg. 
S139
-
47
)
5.
Swanepoel
R
Leman
PA
Burt
FJ
, et al.  . 
Experimental inoculation of plants and animals with Ebola virus
Emerg Infect Dis
 , 
1996
, vol. 
2
 (pg. 
321
-
5
)
6.
Turell
MJ
Bressler
DS
Rossi
CA
Lack of virus replication in arthropods after intrathoracic inoculation of Ebola Reston virus
Am J Trop Med Hyg
 , 
1996
, vol. 
55
 (pg. 
89
-
90
)
7.
Peters
CJ
Sanchez
A
Rollin
PE
Ksiazek
TG
Murphy
FA
Fields
BN
Knipe
DM
Howley
PM
Filoviridae: Marburg and Ebola viruses
Fields virology
 , 
1996
New York City
Raven Press
(pg. 
1161
-
76
)
8.
Reiter
P
Turell
M
Coleman
R
, et al.  . 
Field investigations of an outbreak of Ebola hemorrhagic fever, Kikwit, Democratic Republic of the Congo, 1995. Arthropod Studies
J Infect Dis
 , 
1999
, vol. 
179
 
suppl 1
(pg. 
S48
-
54
)
9.
Khan
AS
Tshioko
FK
Heymann
DL
, et al.  . 
Commission de Lutte contre les Epidémies à Kikwit
The reemergence of Ebola hemorrhagic fever, Democratic Republic of the Congo, 1995
J Infect Dis
 , 
1999
, vol. 
179
 
suppl 1
(pg. 
S76
-
86
)
10.
Mills
JN
Yates
TL
Childs
JE
, et al.  . 
Guidelines for working with rodents potentially infected with hantavirus
J Mammalogy
 , 
1995
, vol. 
76
 (pg. 
716
-
22
)
11.
Mills
JN
Childs
JE
Ksiazek
TG
Peters
CJ
Velleca
WM
Methods for trapping and sampling small mammals for virologic testing
 , 
1995
Atlanta
US Department of Health and Human Services, Public Health Service
12.
Ksiazek
TG
Rollin
PE
Jahrling
PB
Johnson
E
Dalgard
DW
Peters
CJ
Enzyme immunosorbent assay for Ebola virus antigens in tissues of infected primates
J Clin Microbiol
 , 
1992
, vol. 
30
 (pg. 
947
-
50
)
13.
Delany
MJ
The ecology of small rodents in tropical Africa
Mammal Review
 , 
1972
, vol. 
2
 (pg. 
1
-
42
)
14.
Dieterlen
F
Ökologische Populationsstudien an Muriden des Kivugebietes (Congo) Teil I
Zoologische Jahrbücher, Systematik
 , 
1967
, vol. 
94
 (pg. 
369
-
426
)
15.
Happold
DCD
A population study on small rodents in the tropical rain forest of Nigeria
Terre & Vie
 , 
1977
, vol. 
31
 (pg. 
385
-
458
)
16.
Okia
NO
Aspects of rodent ecology in Lunyo Forest, Uganda
J Trop Ecology
 , 
1992
, vol. 
8
 (pg. 
153
-
67
)
17.
Van der Groen
G
Webb
P
Johnson
K
Lange
J
Linsday
H
Eliott
L
Pattyn
SR
Growth of Lassa and Ebola viruses in different cell lines
Ebola virus haemorrhagic fever
 , 
1978
Amsterdam
Elsevier/North-Holland Biomedical Press
(pg. 
255
-
60
)
18.
Baron
RC
McCormick
JB
Zubeir
OA
Ebola virus disease in southern Sudan: hospital dissemination and intrafamilial spread
Bull World Health Organ
 , 
1983
, vol. 
61
 (pg. 
997
-
1003
)
19.
Georges Courbot
M
Sanchez
A
Lu
C
, et al.  . 
Isolation and phylogenetic characterization of Ebola viruses causing different outbreaks in Gabon
Emerg Infect Dis
 , 
1997
, vol. 
3
 (pg. 
59
-
62
)
20.
LeGuenno
B
Formenty
P
Wyers
M
Gounon
P
Walker
F
Boesch
C
Isolation and partial characterisation of a new strain of Ebola virus
Lancet
 , 
1995
, vol. 
345
 (pg. 
1271
-
4
)
21.
Krebs
CJ
Ecology: the experimental analysis of distribution and abundance
 , 
1994
4th ed
New York City
Harper Collins College Publishers
22.
Musser
GG
Carleton
MD
Wilson
DE
Reeder
DM
Order Rodentia, family Muridae
Mammal species of the world. A taxonomic and geographic reference
 , 
1993
2nd ed
Washington DC
Smithsonian Institution Press
(pg. 
501
-
755
)
23.
Isaacson
M
Arntzen
L
Taylor
P
Susceptibility of members of the Mastomys natalensis species complex to experimental infection with Yersinia pestis
J Infect Dis
 , 
1981
, vol. 
144
 pg. 
80
 
24.
Coluzzi
M
Applied field research in malaria in Africa. Heterogeneities of the malaria vectorial system in tropical Africa and their significance in malaria epidemiology and control
Bull World Health Organ
 , 
1984
, vol. 
62
 
suppl
(pg. 
107
-
13
)
25.
White
GB
Anopheles gambiae complex and disease transmission in Africa
Trans R Soc Trop Med Hyg
 , 
1974
, vol. 
68
 (pg. 
278
-
98
)
26.
Schmaljohn
C
Hjelle
B
Hantaviruses: a global disease problem
Emerg Infect Dis
 , 
1997
, vol. 
3
 (pg. 
95
-
104
)
27.
Childs
JE
Peters
CJ
Salvato
MS
Ecology and epidemiology of arenaviruses and their hosts. The Arenaviridae
 , 
1993
New York City
Plenum Press
(pg. 
331
-
84
)

Author notes

*
A list of additional study participants follows the Acknowledgments.