Seroprevalence and Associated Risk Factors of Bovine Brucellosis at the Wildlife-Livestock-Human Interface in Rwanda

Bovine brucellosis is endemic in Rwanda; however, little information is available on seroprevalence and risk factors. Therefore, a cross-sectional study was conducted among cattle farmed at the wildlife-livestock-human interface (n = 1691) in five districts and one peri-urban district (n = 216). Cattle were screened using the Rose Bengal test, then the results were confirmed by indirect enzyme-linked immunesorbent assay. Potential risk factors were determined with a questionnaire and analyzed for their association with seropositivity. In all districts, the animal and herd-level seroprevalence was 7.4% (141/1907) and 28.9% (61/212), respectively, 8.3% (141/1691) and 30.9% (61/198) at the interface, and 0.0% (0/216) in peri-urban areas. Among the potential risk factors, old age (≥5 years), cattle farmed close to wildlife, herds of cattle and small ruminants, history of abortions, and replacement animals were significantly associated with brucellosis (p < 0.05). Low awareness of zoonotic brucellosis, assisting calving without biosafety protection, drinking raw milk, and manual milking were each observed in more than 21.7% of cattle keepers whose herds were seropositive. This study confirmed brucellosis endemicity in cattle farmed close to wildlife in Rwanda, suggesting the need to focus control efforts in these areas. Educated farmers with a high awareness of zoonotic brucellosis had low bovine brucellosis seropositivity, which emphasizes the importance of education.


Introduction
Brucellosis is a neglected and contagious bacterial disease of veterinary and public health importance that remains endemic in developing countries including Rwanda [1]. Brucellosis affects farm and marine animals, wildlife, and humans [2,3]. The disease is caused by bacteria belonging to the genus Brucella. Each Brucella species has a preferential host, for instance, Brucella abortus has a preference for cattle [4], B. melitensis for goats and sheep [5], B. suis for pigs [6], B. ovis for sheep [7], and B. canis for dogs [8]. Among these species, B. melitensis, B. abortus, B. suis, B. canis cause severe disease in humans [3]. the national level and several herds that are located in districts of the Eastern Province are in remote areas where access to veterinary services is limited. The climate in the Eastern Province is warmer and drier, characterized by annual average rainfalls ranging between 700-950 mm, and annual average temperatures ranging between 20 • C and 21 • C. The vegetation is grassland with low inclined hills with an average altitude of 1513.5 m. In Northern and Western Provinces, the climate is the coolest and wettest and is characterized by annual rainfalls ranging from 1400-1600 mm and annual average temperatures ranging from 15-17 • C. The topology is mountainous with volcanoes, and the average altitude ranges between 2000 and 3000 m [27]. The Akagera and Virunga National Parks are home to buffaloes (Syncerus caffer) [28]. In this study, the wildlife-livestock-human interface was comprised of cattle farms in five districts that border the national parks. Further information on the study area is shown in Figure 1.
Microorganisms 2020, 8, x FOR PEER REVIEW 3 of 16 systematic at the national level and several herds that are located in districts of the Eastern Province are in remote areas where access to veterinary services is limited. The climate in the Eastern Province is warmer and drier, characterized by annual average rainfalls ranging between 700-950 mm, and annual average temperatures ranging between 20 °C and 21 °C. The vegetation is grassland with low inclined hills with an average altitude of 1513.5 m. In Northern and Western Provinces, the climate is the coolest and wettest and is characterized by annual rainfalls ranging from 1400-1600 mm and annual average temperatures ranging from 15-17 °C. The topology is mountainous with volcanoes, and the average altitude ranges between 2000 and 3000 m [27]. The Akagera and Virunga National Parks are home to buffaloes (Syncerus caffer) [28]. In this study, the wildlife-livestock-human interface was comprised of cattle farms in five districts that border the national parks. Further information on the study area is shown in Figure 1.

Study Design and Sample Size
The study was a cross-sectional design, conducted between May 2018 and September 2019 that applied a multistage cluster sampling strategy to select herds in the selected districts and individual animals within herds. A herd was classified as the sampling unit and this was stratified by districts. The target population was all dairy herds present in the vicinity of national parks or the peri-urban areas of Gasabo district, Kigali city. Cattle aged 1-year-old -and above were selected for this study, and they were categorized as young (1 to 2 years old), adults of medium age (3 to 4 years old), and adults of old age (5 to 13 years old). The dental formula was used to determine the age of animals as previously described [29]. During sampling, a household that had dairy cattle on the same landsite, regardless of the size, was defined as a farm, whereas a farm owned by one or several people, regardless of the size, was considered as one herd. Within each district, households from all areas bordering a national park were randomly selected from sampling frames provided by the district and Maps of (A) Rwanda with different districts, (B) the Musanze and Nyabihu districts border the Virunga and Gishwati national parks, respectively, and (C) the Nyagatare, Gatsibo, and Kayonza districts border Akagera National Park, and Gasabo is an urban district with peri-urban areas. Red circles and blue triangles indicate seronegative and seropositive herds found in this study.

Study Design and Sample Size
The study was a cross-sectional design, conducted between May 2018 and September 2019 that applied a multistage cluster sampling strategy to select herds in the selected districts and individual animals within herds. A herd was classified as the sampling unit and this was stratified by districts. The target population was all dairy herds present in the vicinity of national parks or the peri-urban areas of Gasabo district, Kigali city. Cattle aged 1-year-old -and above were selected for this study, and they were categorized as young (1 to 2 years old), adults of medium age (3 to 4 years old), and adults of old age (5 to 13 years old). The dental formula was used to determine the age of animals as previously described [29]. During sampling, a household that had dairy cattle on the same landsite, regardless of the size, was defined as a farm, whereas a farm owned by one or several people, regardless of the size, was considered as one herd. Within each district, households from all areas bordering a national park were randomly selected from sampling frames provided by the district and sector veterinary officers. The study involved blood sampling and herd data collection. The sample size was determined using the formula previously described [30]: where Z 2 = 1.96 at the 95% confidence level; P is the expected prevalence estimated to be 10% based on a previous study [14] and d is a margin error of 5%. The total sample size per each district was adjusted for clustering using the following equation: N = n (1 + ρ (m − 1)), where N represents the new sample size, n stands for the original sample size, ρ (=0.2) for the intra-cluster correlation coefficient, and m (=4) represents the number of cattle sampled per herd [30]. The new sample size was 220 cattle per each district. To increase the precision and taking into consideration a large number of cattle in the Nyagatare and Kayonza districts, the sample size was increased by 3 for Nyagatare and 1.5-fold for Kayonza, and this led to 654 and 375 cattle being sampled from the two districts, respectively. The overall sample size was 1907, and these were selected from 212 herds. However, some households consented to participate in the study with the condition of testing all their animals. Therefore, a maximum of four, nine, 15, and more than 15 cattle were selected from 81, 68, 24, and 40 herds, respectively. The sample size at the wildlife-livestock-human interface was 1691 cattle that were selected from 198 herds, while that of peri-urban areas was 216 cattle that were selected from 14 herds.

Questionnaire Design and Data Collection on Individual Cattle and the Herds
Individual data, including the name of the owner, sample identification, age, sex, breed, and location were recorded in a separate list for all selected cattle in the 212 herds. A structured questionnaire comprising of open-ended and closed-ended questions was then administered in a face-to-face interview with one respondent, a cattle keeper/owner of each of the 212 herds to obtain information about potential herd-level risk factors that could be associated with exposure to Brucella infection in both cattle and humans. The interviews were conducted in the herd owner's language (Kinyarwanda) by the primary author or a research assistant who was provided with prior training on all aspects of questionnaire administration in rural and peri-urban areas. The questionnaire was pre-tested in two herds that were not included in the final data set and subsequently adjusted to ensure precision and good flow of the questions and responses. The questionnaire data comprised potential herd-level risk factors, including herd size, herd composition (presence of small ruminants and/or dogs), proximity or history of contact with wildlife, type of grazing system, access to veterinary services, disinfection of pastures and pens, and farmers' knowledge of the disease, among others. The questionnaire also included questions related to public health to assess the cattle keepers' knowledge of the zoonotic aspect of bovine brucellosis and predisposing practices in cattle husbandry. The geographical coordinates of each location were recorded using the Global Positioning System (GPS) device (Garmin etrex 10, Lexena, KS, USA) and were then used to generate a map of the study area using ArcGIS (ESRI ArcGIS, version 10.6).

Blood Collection
Samples were taken without causing damage to the animals and respecting their welfare. Blood samples were collected aseptically in a 4-mL plain vacutainer tube from the jugular or tail vein of each selected animal. The vacutainer tubes labelled with each animal identification were transported to the nearest campus of the University of Rwanda where they were stored overnight at room temperature to allow clotting. The following day, sera were collected in a sterile microcentrifuge tube and kept at −20 • C while waiting for serological testing at the Rwanda Agriculture Board, Department of Veterinary Services.

Serological Tests
Animal sera were screened for the presence of Brucella antibodies using the Rose Bengal test (RBT, Onderstepoort Biological Products, Pretoria, South Africa) according to the protocol previously described by Alton et al. [31]. Briefly, equal volumes (30 µL) of serum and antigen were mixed for four minutes. A Brucella positive and one negative reference samples served as controls. An obvious, clear, and complete agglutination was recorded as a strong (+++) result, while a clear but not complete agglutination was recorded as a medium (++) result. An agglutination that was only visible at the margins was recorded as a weak (+) result. Indirect ELISA was used to confirm RBT positive results in series according to the manufacturer's instructions (IDvet Diagnostics, Grabels, France). For each test microplate, samples were tested as singles while the positive and negative controls were tested in duplicates. The optical densities (ODs) of samples were determined at 450 nm using an ELISA reader (original multiscan Ex, Thermo Fisher Scientific, Waltham, MA, USA). The sera samples with 120% seropositivity and greater were confirmed positive. In this study, the sera samples showing seropositivity above 119.4% were rounded to 120% and considered positive.

Data Analysis
Individual or herd-level seroprevalence for each district and the entire study were calculated by dividing the total number of animals or herds that were simultaneously positive to RBT and i-ELISA by the total number of animals or herds sampled, multiplied by 100. A herd was considered positive if at least one animal tested positive. Data were recorded and analyzed in Microsoft Excel spreadsheets. Each potential risk factor from the individual-and herd-level data was assessed for significant statistical association with the serological status (considered as a binary outcome: positive or negative), using the Chi-square or Fisher's Exact tests of association. Variables that were significantly associated with brucellosis seropositivity (p < 0.05) at univariate analysis were selected and tested for collinearity using the chi-square test. If a pair of variables was found to be collinear, then only one variable considered to be more biologically associated with brucellosis was considered for multivariable analysis. The screened-in variables were then included in initial multivariable logistic regression models, separately for the individual-and herd-level data. The regression was performed by a generalized linear model (GLM) function, considering a binomial distribution. Subsequently, a stepwise elimination procedure was conducted to arrive at the most adequate model that minimized the Alkaike information criteria (AIC). Univariate and multivariate analyses were performed using R Console version 3.5 (R Core Team, 2017) at a 5% level of significance. The selected model was then subjected to the goodness-of-fit test, by the Hosmer-Lemeshow (χ 2 ) test, followed by the determination of odds ratios (OR) for each variable in the final model [32].
All of the three variables that were significantly (p < 0.05) associated with brucellosis seropositivity in the univariate analysis were included in the final multivariable logistic regression model. Cattle from Gatsibo (OR = 22.2), Nyagatare (OR = 9.7), Kayonza (OR = 7.8), Musanze (OR = 4.2), and Gasabo (OR = 10.0 × 10 −7 ) were associated with higher odds of brucellosis seropositivity compared withthe Nyabihu district, although the odds were not statistically significant (p > 0.05) for Musanze and Gasabo districts. Cattle of medium age (3 to 4 years old) (OR = 2.4, p = 0.03) or older (≥5 years) (OR = 3.0, p = 0.005) were associated with significantly higher odds of brucellosis seropositivity (p < 0.05) than young cattle (1 to 2 years). The indigenous cattle breed, "Ankole", was associated with a higher likelihood of seropositivity (OR = 1.8) than the crossbreeds. "Exotic' breeds were not included in the final logistic regression due to the relatively small number of cattle (43) available for sampling. The Hosmer and Lemeshow goodness of fit test was not statistically significant (χ 2 = 3.04, p = 0.93), showing that the model fitted the data well, with the observed data matching the values expected in theory ( Table 2).

Univariate and Multivariate Analyses of Potential Herd Risk Factors
Of the 20 variables considered in the univariate analysis, only 10 showed a significant association (p < 0.05) with herd-level seropositivity, and these included herd composition, grazing system, presence of endemic diseases, sharing watering points, history of abortion, good knowledge of bovine brucellosis, access to veterinary services, introduction of new cattle into the herds, and feeding abortion tissues to dogs (Table S1). Although other herd factors were not significantly associated with brucellosis (p > 0.05), high proportions of seropositive animals were observed between levels of variables and these data are available in the Supplementary Materials (Table S1).
Among the 10 variables that were significantly (p < 0.05) associated with brucellosis in the univariate analysis, only six comprised the final multivariate logistic model analysis (Table 3). Herd owners without any level of education (OR = 7.2, p <0.05) and those with primary education (OR = 6.7, p <0.05) were more likely to have seropositive herds than those with tertiary and secondary education, and the odds were statistically significant. Another important significant predictor for herd-level seropositivity included herd composition with herds that had both cattle and small ruminants being more significantly associated with brucellosis seropositivity (OR = 2.8, p < 0.05) compared to herds with cattle only. Good knowledge of animal brucellosis among herd owners was more likely to be associated with brucellosis seropositivity (OR = 5.5; p < 0.05). The history of abortions and the introduction of new animals into the herd were also significant predictors (p < 0.05) of brucellosis. Cattle reared under free-grazing were associated with higher odds of seropositivity (OR = 1.9) than those under zero-grazing, although the odds were not statistically significant (p > 0.05). The Hosmer and Lemeshow goodness of fit test was not statistically significant (χ 2 = 3.9, p = 0.87) showing that the model fitted the data well, with the observed data matching the values expected in theory (Table 3).   Table 4 shows the univariate associations between six risk factors and cattle keepers having seropositive herds. Brucellosis seropositive herds were significantly associated with cattle keepers with insufficient education (p < 0.05). Low awareness of zoonotic brucellosis was common in most cattle keepers, 85.9% (182/212) and among them, 26.9% had seropositive animals. Although calving was not significantly associated with herd seroprevalence, most respondents, 76.9% (163/212) assist cattle during parturition without personal protective equipment (PPE), and 31.9% of them had seropositive herds. The number of cattle keepers who drink raw milk was 39.2% (83/212) and of them, 21.7% had seropositive herds. Manual milking was commonly observed in 98.6% of the herds and of them, 28.6% had seropositive herds (Table 4). There was a significant correlation between awareness of zoonotic brucellosis and boiling milk and between education level and boiling milk (p < 0.05). Most of the cattle keepers, 80.0% (24/30) that were aware of brucellosis being zoonotic also boiled their home milk before consumption. Of 182 cattle keepers that were not aware of brucellosis as a zoonotic disease, 42.3% (77/182) drank un-boiled milk. Educated cattle keepers, 79.5% (35/44) were more likely to boil their home milk before consumption compared to uneducated, 58.8% (50/85), and those with primary education, 53.0% (44/83) (Figure 2). and between education level and boiling milk (p < 0.05). Most of the cattle keepers, 80.0% (24/30) that were aware of brucellosis being zoonotic also boiled their home milk before consumption. Of 182 cattle keepers that were not aware of brucellosis as a zoonotic disease, 42.3% (77/182) drank un-boiled milk. Educated cattle keepers, 79.5% (35/44) were more likely to boil their home milk before consumption compared to uneducated, 58.8% (50/85), and those with primary education, 53.0% (44/83) (Figure 2).

Discussion
Bovine brucellosis is a contagious bacterial disease of veterinary and public health importance and the disease is endemic in sub-Sahara African countries including Rwanda. This study, which was carried out in six districts, is the first to report on the seroprevalence of brucellosis and associated risk factors in cattle farmed at wildlife-livestock-human interfaces in Rwanda. The findings of the present study confirmed that brucellosis determined with serological tests (RBT and i-ELISA) is endemic in cattle farmed close to the national parks, especially those harboring several buffalos, and the occurrence therein was significantly higher than that in peri-urban areas in the Gasabo district, Kigali city. The overall adjusted animal and herd seroprevalence rates (7.4% and 28.9%) obtained in cattle from six districts in this study as well as the previous rates (9.9-30.2%) obtained in the Nyagatare district of Rwanda using RBT alone [14], the 7.4% rate reported in the Huye district of Rwanda using RBT alone [17], and the rate of 18.9% reported in the Nyagatare district using only RBT [16], confirm that brucellosis is endemic in Rwanda.
Of the 260 (13.6%) sera that were detected as positive for brucellosis using RBT, 118 (45.4%) were strong positive, 32 (12.3%) were medium while 110 (42.3%) were weak positive. Of the 110 RBT-weak positives, 3 (2.7%) were confirmed seropositive using i-ELISA. Most veterinary laboratories in developing countries diagnose brucellosis by detecting only RBT strong positives (complete and clear agglutination) and medium positives (clear agglutination) due to the lack of expertise in detecting weak positives as RBT is a subjective test. Additionally, the confirmation test is not always performed due to either the lack of confirmatory test reagents or the limited number of personnel. Therefore, if the weak positive animals are undetected and then approved for trade, this could contribute to the spread of brucellosis to the naïve herds at the destination.
The animal-level seroprevalence (8.3%) observed in cattle at the wildlife-livestock-human interface is in line with the respective results (8.3% and 9.6%) reported in cattle at the wildlife-livestock-human interface in Zimbabwe using RBT and c-ELISA [33,34], and (9.7%) in Ethiopia using RBT, and i-ELISA [35]. The herd-level seroprevalence of 30.9% observed at the interface in this study is comparable to that obtained in cattle at the wildlife-livestock-human interface in Kenya (26.7%) using i-ELISA [22], and in Ethiopia (32%) using RBT and i-ELISA [35]. On the other hand, our finding was lower compared to the results obtained in cattle farmed at the wildlife-livestock-human interface: in Kenya (68.7%) using i-ELISA [23], and in Zambia (58.1%) using RBT and c-ELISA [36]. This difference was explained by the absence of vaccination programs in the study area in Kenya [23] while in Zambia, the high seroprevalence was associated with abortions and cattle shared grazing pastures and watering points with wildlife [36]. Moreover, the seroprevalence of brucellosis was reported in cattle, buffaloes, and humans at the interface in Tanzania [20,37] and in Zimbabwe [24,25], and this suggests the spillover of brucellosis between these species. The current study together with the above studies confirmed that bovine brucellosis is prevalent in cattle farmed at the wildlife-livestock-human interface, and higher incidences of brucellosis occur in herds with increased interactions between livestock and wildlife. The animal and herd-level seroprevalence rates observed in this study differed significantly among districts (p < 0.05). Cattle from districts that border national parks had higher animal and herd seroprevalences compared to those from peri-urban areas of Gasabo district where no animal was found positive. This difference can be ascribed to the relatively large size of herds, and the free grazing system observed in the Eastern and Western Provinces as compared to the zero-grazing system among cattle farms in Gasabo district, and in which animal health is managed better by the easily accessible veterinary services and readily available animal scientists. Zero-grazing system minimizes contacts between animals and thus reduces the risk of disease transmission. Districts bordering the Akagera National Park in the Eastern Province were more likely to have seropositive cattle (p < 0.05) compared to the Musanze and Nyabihu districts that border the Virunga and Gishwati-Mukura National Parks, respectively. In addition, the animal-level seroprevalence recorded in Musanze was high compared to that recorded in Nyahihu. This difference may be attributable to the presence and number of buffaloes within the various national parks. For instance, the Gishwati-Mukura National Park contains only monkeys, chimpanzees, and birds while the Akagera National Park contains many ruminants including buffaloes and the Virunga National Park alongside Musanze has buffaloes. Before the fencing of the Akagera and Virunga National Parks in 2014, cattle grazed and shared watering points with wild herbivores. Although these parks are now fenced, spotted hyenas cross the electric fence from the Akagera National Park to cattle farms (Field observation, 2019). We observed calves wounded around the anus and tail, and these wounds were caused by wild carnivores. These carnivores can move aborted tissue at the wildlife-livestock interface and a recent study isolated Brucella abortus and Brucella suis from lions and hyenas in the Serengeti National Park, Tanzania [38]. Brucella abortus was isolated from 14 dogs in 10 brucellosis positive cattle farms [39] and Brucella spp. were isolated from saliva, nasal discharges, and urine of dogs feeding on aborted tissue [40,41] and urine has been incriminated in the transmission of canine brucellosis [42]. In addition, occasional transmission of brucellosis through bites has been reported [43]. Thus, the movement of these carnivores feeding on aborted tissue and live calves and goats in both the park and cattle farms may play a role in the transmission of bovine brucellosis and other zoonotic diseases between the wildlife and livestock and vice-versa. Elsewhere in Africa, significantly higher brucellosis seroprevalences were reported in cattle in areas close to wildlife habitat compared to areas far from the home range of wildlife, i.e., in Uganda [26] and Tanzania [37,44]. The significantly higher occurrence of brucellosis observed in districts that border national parks can be attributed to previous interactions between wildlife and livestock, and indirect interactions by carnivores and rodents. It is therefore worth further investigating the occurrence of brucellosis in buffaloes and other wild animals in Rwanda.
This study showed that the age of the cattle was a significant predictor of brucellosis seropositivity, with the medium adult age category (3 to 4 years) and the old cattle (≥5 years) being more affected (OR = 5, p = 0.005) than young animals. This finding is in agreement with other studies carried out in Rwanda [14,16], and in Uganda [45]. Animals that are kept for a longer period in the herds have more chances of exposure and acquiring brucellosis, and this translates into increased brucellosis seropositivity with increasing age. It has also been reported that Brucella spp. have a tropism for reproductive organs of mature female animals, and the sex hormones and erythritol produced are responsible for the survival and multiplication of Brucella species [46]; this contributes to the overall higher seropositivity in sexually mature females.
In this study, herds in which cattle grazed together with small ruminants were significantly more likely to be seropositive than cattle-only herds, which is consistent with similar studies that found that mixing cattle and small ruminants was a significant predictor of brucellosis [47,48]. This suggests that small ruminants may play a role in the maintenance and persistence of brucellosis in cattle in Rwanda since the former are not vaccinated. This also indicates that there may be co-infection with B. melitensis and B. abortus in the same herd which is consistent with a recent study in South Africa that isolated both species in slaughtered cattle [49].
A history of abortions was a significant predictor for herd-level seroprevalence, and this is in agreement with previous reports from Uganda [26,45], and Tanzania [50]. Furthermore, this study also revealed that 98.6% of respondents did not dispose of abortuses properly and birth sites were not disinfected, which is consistent with a previous report in Nyagatare district [16]. Therefore, it is likely that there will be a continuous circulation of Brucella pathogens within and between herds. Various reproductive disorders that are associated with brucellosis have been reported in the cattle industry in Rwanda, including higher incidences of abortions, retained placenta, infertility of unknown origin, and longer calving intervals [51]. Such abortions can cause tremendous financial losses and wherever they occur in the herd, massive screening of the herd against brucellosis is very important and positive animals should be immediately slaughtered to stop the spread.
Our findings revealed that uneducated and less educated cattle keepers were significantly associated with higher herd-level seropositivity than herds whose owners had attained secondary and/or tertiary education. These findings are in agreement with those of Assenga et al. [44], who reported lower Brucella infection exposure in the herds of educated livestock farmers [44]. Illiterate or less educated farmers are likely to be less informed or to adopt slowly to innovations, and this may be matched with poor management practices such as the hygiene of cattle and their environment, and weaker implementation of recommended control measures such as the restriction of animal movements and vaccination. Indeed, we found that among the 26 cattle herders who vaccinated their animals, 88.5% (23/26) were educated. Nevertheless, education and learning are processes, and owners with less or no education can be helped, for example, through regular consultation with professionals in animal science. Furthermore, 78.8% (167/212) of cattle keepers had good knowledge of bovine brucellosis, which is known as "Amakore" in the Kinyarwanda language. The farmers knew that the disease is characterized by abortions and hygromas in the patellofemoral joint. However, herds belonging to such farmers had higher odds (5.5 times) of brucellosis seropositivity than herds with no awareness, which indicates that despite the knowledge, there is negligence in implementing recommended control measures such as restriction of movement and removal of seropositive animals from the herd. In contrast, several other studies reported poor knowledge of brucellosis among several cattle keepers [16,35], and this was associated with an increase of Brucella infection in herds.
The majority of cattle keepers (85.9%) did not know that brucellosis affects humans. It is therefore not surprising that 60.9% of cattle keepers mentioned that they drank boiled milk not to avoid brucellosis, but to prevent diarrhea or tuberculosis. Also, boiling milk was significantly associated with awareness of zoonotic brucellosis, and with education level. The low awareness of zoonotic brucellosis is further reflected in the observation that a high proportion (76.9%) of cattle keepers that assisted calving without wearing protective equipment or clothing-and given that manual milking was observed in almost all (98.8%) of the herds-this constitutes a high risk for cattle keepers. In congruence with our findings, low awareness of zoonotic brucellosis was also reported in more than 92.0% of cattle keepers in Ethiopia [35], of which, most farmers did not regard exposure to abortion tissues, drinking and eating raw animal products as risk factors. In this study, boiling milk was significantly associated with awareness of zoonotic brucellosis, and with education level (p < 0.05). Within cattle keepers, the increase in awareness of zoonotic brucellosis or in their education level influences an increase in the number of cattle keepers boiling milk before home consumption. This finding indicates that continuing education of cattle keepers and other exposed groups on the epidemiology of zoonotic brucellosis and other zoonotic diseases should contribute significantly to preventing zoonotic diseases in humans.

Conclusions
This study confirmed that brucellosis is endemic in cattle farmed at the wildlife-livestock-human interface and found that the history of abortions and introduction of new animals into herds are the major predictors of brucellosis. Therefore, aborting cattle, and cattle for replacement should be quarantined, tested, and the positives slaughtered. The interface should be more targeted by control programs such as vaccinations, testing and slaughter, and the requirement of an annual brucellosis-free certificate for national and international trade. Most cattle keepers had low awareness of zoonotic brucellosis and this was exemplified by them assisting calving without PPE and improper disposal of abortion tissues. This awareness should be raised among all stakeholders through education campaigns on zoonotic brucellosis. The One Health concept of involving veterinarians, environmentalists, and physicians could efficiently minimize zoonotic brucellosis, and the control of animal brucellosis would prevent the disease in humans since there is no vaccine for the latter. Further studies on brucellosis seroprevalence in wildlife, carnivores, and humans living at the interface are worthy of investigation in Rwanda.

Supplementary Materials:
The following are available online at http://www.mdpi.com/2076-2607/8/10/1553/s1, Table S1: Univariate associations of potential risk factors with herd-level seropositivity of Brucella spp. in cattle sampled at the wildlife-livestock-human interface in Rwanda. A questionnaire used to collect herd and public health risk factors. Excel document containing all data on individual seroprevalence and potential risk factors.