Molecular prevalence, genetic characterization and patterns of Toxoplasma gondii infection in domestic small mammals from Cotonou, Benin

Toxoplasmosis, one of the most prevalent parasitic infections in humans and animals, is caused by the intracellular protozoan parasite Toxoplasma gondii. Small mammals play a key role as intermediate reservoir hosts in the maintenance of the T. gondii life cycle. In this study, we estimated the molecular prevalence and provide genetic diversity data for T. gondii in 632 small mammals sampled in four areas of Cotonou city, Benin. Both the brain and heart of each individual were screened through T. gondii-targeting qPCR, and positive samples were then genotyped using a set of 15 T. gondii-specific microsatellites. Prevalence data were statistically analyzed in order to assess the relative impact of individual host characteristics, spatial distribution, composition of small mammal community, and urban landscape features. An overall T. gondii molecular prevalence of 15.2% was found and seven genotypes, all belonging to the Africa 1 lineage, could be retrieved from the invasive black rat Rattus rattus and the native African giant shrew Crocidura olivieri. Statistical analyses did not suggest any significant influence of the environmental parameters used in this study. Rather, depending on the local context, T. gondii prevalence appeared to be associated either with black rat, shrew, or mouse abundance or with the trapping period. Overall, our results highlight the intricate relationships between biotic and abiotic factors involved in T. gondii epidemiology and suggest that R. rattus and C. olivieri are two competent reservoirs for the Africa 1 lineage, a widespread lineage in tropical Africa and the predominant lineage in Benin.


Introduction
Small mammals, including rodents, have an important epidemiological role in the evolutionary ecology and transmission cycle of Toxoplasma gondii, for which they act as natural reservoirs. They are one of the preferred prey groups of the definitive felid hosts (e.g., cats [14,35]), and therefore play a key role in the circulation of Toxoplasma strains within local and novel environments [30,42,62,67]. This is particularly true in the increasingly changing global context that favors the ongoing introduction and spread of invasive rodents [40,41]. Accordingly, rodents are generally considered relevant hosts to assess environmental contamination by toxoplasmosis, and to estimate the infection risk for definitive hosts [1,6,67]. Nevertheless, considering rodents a meaningful indicator to predict environmental contamination by T. gondii requires us to differentiate between potential biological and ecological differences within and between reservoir rodent species. As an illustration, a positive relationship between rodent body mass and probability to be infected by T. gondii has been found both within (i.e., older individuals were more often infected [20,67]) as well as between species (i.e., larger species had a longer life span, hence more opportunities to encounter T. gondii [2,20,46]). Species-specific ecological requirements have also been found to influence T. gondii prevalence, essentially due to variation in oocyst-rodent contact patterns [68,81]. For example, fossorial species that live in burrows are always in contact with soil and likely include T. gondii paratenic hosts (such as earthworms, [2]) in their diet. In addition to biological and ecological species-level features, the local environment is expected to influence the persistence of oocysts within the external environment, thus greatly impacting parasite ecoepidemiology [81,84].
Despite this burgeoning knowledge on the reservoir-based ecology of T. gondii, the effects of environmental, ecological, and host-individual factors on the circulation of T. gondii in natural populations of rodent reservoirs have been scarcely investigated in a concomitant way. Moreover, studies on the prevalence of toxoplasmosis in small mammals sampled in various parts of the world show great variability following region and habitat, host species, and detection method [30]. In Africa, most studies are based on serological detection of T. gondii and seroprevalence was found to be relatively low: e.g., 1.6% and 1.2% in Mus musculus and Mastomys natalensis, respectively in Niamey, Niger [58]; 2.3% and 4.2% in Rattus rattus from rural Gabon [57] and across Senegal [11], respectively and 3.8% in Rattus norvegicus in Egypt [59]. However, such seroprevalence-based results have largely been questioned from several points of view, such as the intrinsic differences in host immune response [40] and the variable sensitivity/specificity of serological tests used [17,30]. These limitations have therefore led to recognize qPCR-based approaches as more sensitive and reliable, since this method allows direct targeting of the parasite DNA in examined tissues [40].
Furthermore, while rodents, and specifically laboratory mice, are widely used to study toxoplasmosis in laboratory conditions (e.g., host immune response, parasite virulence and/or parasite amplification for subsequent genetic analyses), studies using the direct molecular approach for investigations in natural populations are largely lacking, especially in the African context as far as we are aware (but see e.g., [30,40,41]). However, investigations in natural populations are key to unravel the intricate eco-evolutionary processes underlying the circulation of T. gondii in various habitats. As an illustration, the first Types I, II and III isolates of this parasite were historically classified according to their virulence in the laboratory mouse [73]. However, such a virulence in laboratory mice is clearly not generalizable to all rodent species, and distinct resistance profiles have been observed in different subspecies or strains of the same rodent host species [49,61]. These resistance profiles result from a complex evolutionary balance between the host immune response and the genetic diversity of T. gondii in each region of the world [22,48,69,72].
The circulation of T. gondii in domestic animals has been demonstrated in several parts of Benin (West Africa), including Cotonou city [47,82,83], but no data are available for small mammals hosts. Nevertheless, this city has a major international seaport that has been shown to be responsible for the introduction of invasive rodent species and associated zoonotic pathogens into the country [13], thus questioning the genetic diversity and ecology of T. gondii strains that may circulate locally through these reservoir species. In this study, we explore the role of natural populations of small mammals (rodents and shrews) from Cotonou city in the epidemiology of T. gondii. To do so, we focus on the Cotonou small mammal community in order to (i) assess the molecular prevalence of T. gondii within each species-specific population, (ii) explore the impact of certain intrinsic and extrinsic factors on individual infection by T. gondii, and (iii) further document the genotypes circulating among small mammals within this large West African seaport city.

Ethics statement
Sampling was conducted under the research agreement signed between the Republic of Benin and the French Research Institute for Sustainable Development (IRD) renewed on April 6, 2017, as well as the scientific collaboration agreement between IRD and Abomey-Calavi University signed on the September 30, 2010 and renewed on July 3, 2019. All trapping sessions were conducted following the explicit oral authorization of local authorities (Cotonou City Hall and local heads of urban districts) as well as the agreement of each household's owners and inhabitants. Field campaigns within the Cotonou Autonomous Seaport were implemented following official written authorization issued by its General Management Board. None of the species captured in this study have protection status (see CITES list, https://checklist.cites.org/). All animals were handled and processed in a humane manner, with respect to animal welfare following the guidelines of the American Society of Mammologists [74]. Our study and the uses of associated samples and genetic data were implemented following Nagoya protocol recommendations (Access and Benefit Sharing agreement number 608/DGEFC/DCPRN/PF-APA/SA).

Study sites
Our study was performed across two major types of areas: (i) three socio-environmentally contrasted urban districts of Cotonou city between 2017 and 2018, namely Ladji, Agla and Saint-Jean; and (ii) Cotonou seaport (Autonomous Port of Cotonou, or APC) (Fig. 1). Though of rather recent origin, Agla is currently one of Cotonou's most populous neighbourhoods. At the time of the present study, it was a still a partially swampy area undergoing rapid urbanization. It is located within a large shallow land and was thus characterized by wide flooded areas at the early stages of the rainy season, following accumulation of rainfall and elevation of the water table [52]. Ladji is located on the shore as well as upon Lake Nokoué (i.e., stilt-built houses). With over 10,000 inhabitants, this informal settlement is socio-environmentally very disadvantaged and suffers from insufficient basic services. It is partly flooded at the end of the rainy season following overflow of Lake Nokoué. Saint-Jean is a formal residential area of colonial origin. Even though the population is working-class, many services are formally available, and the area does not flood per se during the rainy season. However, wide ponds may persist in some places for days to a few weeks after heavy rains. APC is located in the central administrative highly humanmade part of the city. Two traditional and seven industrial port zones were targeted: the former gather areas where traditional fishing products are disembarked, prepared, stored and sold, while the latter correspond to the canteens (where workers take lunch) as well as areas under restricted access (e.g., administrative services, docks, storehouses, workshops, garages, and parking lots) (for details about these APC sampling sites).

Sampling
In each of the three districts, 9-11 households (hereafter designated "district sites") were investigated (see details in [24,26]), while nine sites were sampled in APC (hereafter designated "APC sites"). Each area was investigated twice: in October 2017 and June 2018 for Agla, Ladji, and Saint-Jean on the one hand, and in September-November 2017 and March 2018 for APC on the other. Standardized longitudinal trapping sessions were conducted for three consecutive nights using both Sherman and locally made wire mesh traps. A fish/peanut butter mixture was used as bait. Captured small mammals were brought alive into the laboratory where they were euthanized within the same day following diethyl-ether exposure and subsequent cervical dislocation, then sexed, weighed and measured. Immediately after euthanasia, presence of ectoparasitic fleas was checked by combing the animal. Particular attention was paid to unambiguous species-specific identification of all individuals through morphology and external measurements, cytochrome b DNA sequencing, PCR-RFLP analyses and microsatellite genotyping (see details in [24,26]). The age of individuals (i.e., adult vs. juvenile stage) was assessed according to the body mass following Granjon & Duplantier [45] and signs of sexual maturity (i.e., external testicles and/or active seminal vesicles in males; developed mammae and uterus, presence of embryos and/or embryo scars in females). Among other samples, ethanol-preserved or directly frozen hearts and brains were collected on each small mammal specimen for subsequent molecular processing.

Molecular detection of T. gondii
Toxoplasma gondii presence was investigated on both heart and brain samples of each small mammal collected. To do so, brains were ground separately in 1 mL of physiological water (0.9% NaCl) and 30 mg of heart cut at the apex for DNA extraction and qPCR. We relied on DNA extraction and qPCR-based screening as previously described [5,40]. Briefly, DNA was extracted using a commercial kit as recommended by the manufacturer (Qiagen QIAamp DNA Mini Kit, Courtaboeuf, France) from 200 lL of liquid brain homogenate, and from 30 mg of the heart apex. Real-time PCR targeting the noncoding 529 bp repeat region was used to assess the presence of T. gondii DNA as described by Ajzenberg et al. [5]. Each PCR contained 5 lL of extracted DNA, mixed with 15 lL of a PCR reagent mix composed of 1X LightCycler FastStart DNA Master Hybridisation Probes kit (Roche Diagnostics, Mannheim, Germany), 0.5 U of UNG (Roche Diagnostics, Mannheim, Germany), 5 mMol/L of MgCl 2 , 0.5 lMol/L of each primer, and 0.1 lMol/L of TaqMan probe (Eurofins, Ebersberg, Germany) which is labeled with a fluorescent dye (6-carboxyfluorescein, 6-FAM) at the 5 0 end and a dark quencher (Black Hole Quencher, BHQ1) at the 3 0 end. A cycling protocol was run on a Rotor-Gene 6000 thermocycler (Corbett Life Science, Sydney, NSW, Australia) as follows: initial decontamination by UNG at 50°C/2 min and denaturation at 95°C/10 min, followed by 50 cycles at 95°C/20 s and 60°C/40 s. The results obtained were expressed in cycle threshold (Ct) values. Each amplification curve was visually checked to verify its sigmoidal appearance in order to avoid the inclusion of false positives. Genomic DNA isolated from Type I (RH strain, http://toxodb.org) T. gondii isolate cultures and sterile water were used as positive and negative qPCR controls, respectively. Each heart and brain extraction sample was investigated twice by qPCR, and individuals that provided at least one positive test were considered positive.

Microsatellite genotyping
Genotyping relied on 15 microsatellite (MS) marker multiplex PCR corresponding to 15 loci located on 11 different T. gondii chromosomes [4]. Some of them were described as poorly polymorphic but still valuable for strain typing (TUB2, W35, TgM-A, B18, B17, M33, IV.1, and XI.1), while others are highly polymorphic and allow us to distinguish between strains of a same type (M48, M102, N60, N82, AA, N6, and N83) [4,57]. The Beninese multi-locus genotypes (MLGs) were compared to reference strains described in previous studies (see Table 1) in order to reach strain-specific assignment. For genotyping, DNA isolated from Type II (ME49 strain, http://toxodb.org) T. gondii isolates cultures and sterile water were used as positive and negative controls, respectively.

Statistical analyses
We used Chi-square tests on the whole dataset to compare T. gondii prevalence between host species, sampling localities (Agla, Ladji, Saint-Jean, and APC), and screened organs. Then, analyses of APC data (hereafter referred to as "APC-based analysis") were performed separately from the three urban districts (hereafter referred to as "District-based analysis") since (i) they display distinct socio-economic, historical and environmental characteristics, (ii) no landscape/GIS data were available for the APC, and (iii) trapping campaigns were not performed at the same period in both types of area. After determining prevalence levels of T. gondii (i.e., the percentage of infected hosts calculated within a 95% confidence interval (CI) estimated with Wald's method) for each small mammal population, we carried out a three-way analysis using specific packages in RStudio v.4.1.2 software [79].
We first used Chi-square and Fisher's exact tests to investigate possible relationships between T. gondii infection and certain host individual variables (sex, age, and presence of ectoparasitic fleas), period of capture (trapping session) and organs screened for T. gondii detection (heart or brain).
Second, we carried out generalized linear mixed models (GLMMs) on specific datasets to test whether T. gondii prevalence in small mammals was significantly associated with specific intrinsic and extrinsic parameters, namely the relative abundance of each species within the small mammal community (estimated by species-specific trapping success at each sampling site), host individual characteristics (sex, age, and presence/absence of fleas), period of capture (trapping session) as well as environmental features (defined by trapping site coordinates on the first four axes of PCA carried out on set landscape metrics). For the latter parameters, we took advantage of a previous study relying on the same experimental design (i.e., same sampling campaigns, hence same sampling sites) for which the dataset encompassed our own samples in the same districts of Cotonou ( [27]). In this study, a wide range of landscape descriptors including landcover data, social uses associated with buildings, as well as surface water occurrences were mapped and quantified using a dedicated Geographic Information System [27]. These landscape data were integrated and explored through a principal component analysis (PCA; see Supplementary figure and Dossou et al. [27]). For the purpose of our study, site-specific PCA coordinates along the four first components (representing 56% of total inertia of all axes) were extracted and used as explanatory socio-environmental variables in GLMMs (only for "district-based analysis" given that this information was not collected for APC). Using the Akaike Information Criterion (AICc), we carried out a model selection procedure from a full starting model containing all predictors. Toxoplasma gondii prevalence was treated as a binary response (therefore assuming a binomial distribution) and either districts (for "district-based analysis") or APC sites (for "APC-based analysis") were considered random factors to account for possible spatial variation. Models with all possible combinations of the terms included in the starting model were generated, and the most parsimonious model (i.e., the one explaining the highest variance level with the fewest explanatory variables) was chosen among those selected within two AIC units of the best model retrieved [12]. The significance of explanatory variables and their interactions was determined by deletion testing and log-likelihood ratio tests. The final model was validated by the graphical checking of normality, independence, and variance homogeneity of residuals.
Lastly, as implemented in Mariën et al. [55] and Sirdar et al. [76], possible spatial/site clustering of T. gondii-carrying reservoirs was investigated in the sampling localities through Cuzick-Edwards (nearest-neighbour) tests [19,70]. To do so, Monte-Carlo permutations (n = 10,000) were run to infer whether clustering of the field data was significantly higher (p 0.05) than under random simulations [36]. After confirming that T. gondii-positive animals were significantly clustered, we used Kulldorff's spatial scan statistics in order to characterize the local clusters, which make it possible to assess the number and geographical position of local clusters, as well as the number of rodents involved in the potential clusters identified [53].

Sampling
We analyzed a total of 632 small mammals belonging to seven species: Black rat, Rattus rattus (n = 234 individuals), African giant shrew, Crocidura olivieri (n = 179), House mouse, Mus musculus domesticus (n = 102), Brown rat, Rattus norvegicus (n = 77), Natal multimammate mouse, Mastomys natalensis (n = 28), Gambian pouched rat, Cricetomys gambianus (n = 7) and Deroo's mouse, Praomys derooi  The two species C. gambianus and P. derooi, which were poorly represented in the sampling (n = 7 and n = 5), had only one and two individuals infected, respectively and were not taken into account in the subsequent analyses. Regarding the targeted organ, no significant differences were found between the brain and heart, with prevalence reaching 9.9%

District-based analysis
For the three sampled districts, comparison of the prevalence showed a significant difference between species (v 2 = 9.9; p = 0.02), with pairwise comparisons indicating that shrews were significantly more infected than black rats (v 2 = 8.8; p = 0.003). No statistical pairwise difference was found between other species. Regarding the district sampled, no significant difference was found (v 2 = 2.9; p = 0.23). Regarding the trapping session, no difference was found between species nor between districts, except in Saint-Jean where prevalence was higher in June 2018 than in October 2017 (v 2 = 8.2; p = 0.004). Finally, regardless of the species considered, there was no significant relationship between T. gondii infection and respectively sex, age, or flea carriage (v 2 , all p > 0.2).

APC-based analysis
Our analyses excluded M. natalensis, for which only six individuals were trapped and none of them were infected. We found no significant difference in T. gondii infection between other species (v 2 = 1.8; p = 0.62). Toxoplasma gondii prevalences in the mouse population were significantly higher in September-November 2017 than in March 2018 (v 2 = 5.6; p = 0.02), while no differences were found in other species depending on the trapping period. Irrespective of the species considered, there was no sex, age, and fleas carriage-associated distribution of T. gondii infection (v 2 , all p > 0.3) except in R. norvegicus where juveniles seemed to be more commonly infected than adults (Fisher's test, p = 0.005).

Genotyping
Most of the qPCR-positive small mammals displayed DNA concentrations of T. gondii in their tissues that were too low to enable genotyping (Ct > 32). Samples from seven individuals (samples #803, 918, 954, 1000, 1491, 1589 and 1616) had sufficient T. gondii DNA for partially successful MS-markers amplification (10/15 to 14/15 amplified markers). Despite incomplete genotyping, enough markers were amplified for lineage identification: all seven samples belonged to the Africa 1 lineage genotypes (Table 1). These genotypes were retrieved from one black rat captured in the APC and two shrews in each of the three urban neighbourhoods.

Influence of intrinsic/extrinsic factors on T. gondii molecular prevalence
Several GLMMs were tested, either considering all small mammals species investigated or separately for each species with at least 50 individuals sampled, in both "district-based" and "APC-based" models (Table 2).

District-based GLMMs
We found a parsimonious model only when considering all small mammal species. A high T. gondii prevalence was (i) negatively associated with R. rattus relative abundance (v 2 = 7.99; p = 0.005), and (ii) positively associated with C. olivieri relative abundance (v 2 = 5.06; p = 0.024). Therefore, the socio-environmental proxies used this study failed to show any influence on T. gondii prevalence in small mammal communities.

APC-based GLMMs
Considering all small mammal species, T. gondii infection was positively correlated with the relative abundance of M. m. domesticus (v 2 = 7.22; p = 0.007). For this rodent species, the most parsimonious model best explaining T. gondii infection included the trapping session (v 2 = 7.45; p = 0.006) only; they were significantly more commonly infected in September-November 2017 than in March 2018.

Spatial clustering of T. gondii-infected small mammals
Regardless of the trapping period considered, no significant clusters were observed in the four studied localities; however, when the two trapping sessions were grouped, a significant clustering of T. gondii-infected animals was observed in Saint-Jean district, with Monte Carlo p-values lower than 0.05 for q = 10 nearest neighbours (p-value = 0.03) based on the Cuzick-Edwards test (Table 3). Kulldorff's spatial statistic scan also indicated a single significant local cluster (p-value = 0.005) at one given Saint-Jean site (i.e., S6) where 5 out of 9 individuals were infected (Fig. 2).

Discussion
To our knowledge, the present study is the first to focus on interrelationships between several abiotic and rodent-based biotic factors and T. gondii circulation in African urban dwellings. Our results are consistent with previous findings from Europe and North America, showing that small mammals play an important role in the maintenance and circulation of T. gondii within the urban environment [11,30,[40][41][42]. Furthermore, our qPCR-based approach led to higher prevalence values (20.6% in M. m. domesticus, 19.6% in C. olivieri, 14.3% in M. natalensis, 14.3% in R. norvegicus and 9.9% in R. rattus) than those previously observed elsewhere using molecular methods, regardless of the T. gondii targeted gene, the methodology of screening, and the geographical scale considered (e.g., in R. rattus: 1.8% in São Paolo, Brazil [60]; 4.5% in Gran La Plata, Argentina [23]; 4.3% in Sichuan, China [85]). In the same manner, within the African context, the prevalence exhibited here was also higher than that found in Senegal with the same qPCR-based approach (i.e., 3.8% in black rats, 13.3% in house mice, and 15.5% in shrews; [40]). These differences in prevalence could result from the nature of the tissues analyzed. Importantly, the study in Senegal (and in most studies) was based only on brain samples, whereas our study relied on both brain and heart. Although no significant differences in molecular prevalence were found between organs, only 13.5% of the 96 positive animals displayed positive results in both organs. When focusing only on brain samples, our results show a prevalence of 6.4% in black rats, 14.7% in house mice, and 14.5% in shrews, which is much closer to the prevalence observed in Senegal [40]. This suggests that both organs (brain and heart) should be considered together when investigating molecular prevalence of T. gondii in future small mammal-centered surveys. Another non-exclusive hypothesis on the differences observed between the Senegalese and Beninese contexts could be the differences in socio-environmental conditions (e.g., climate, nature of soil, vegetation type) between both countries. As such, sampled localities in Senegal (Dakar, Rufisque and Joal-Fadiouth [40]) are characterized by a hot semi-arid climate [33], which could contribute to reduce the survival rate of T. gondii oocysts, while Cotonou is favorably more humid, with recurring floods [52,86]. For a clearer view, we advocate for study of different geographic contexts of T. gondii epidemiology within commensal small mammal communities showing contrasted eco-evolutionary trajectories.
Nevertheless, our analyses provide interesting insights on the tested factors. As previously described by Galal et al. [40], we showed that shrews were significantly more often infected than black rats over the three Cotonou districts. This could be due to differences in life traits: black rats are omnivorous and good climbers, while shrews forage only on the ground where they show marked digging and exploratory behavior for feeding [10,15,16]. The latter habits may likely favor contact with T. gondii oocysts excreted in the environment by stray and/or domestic cats. Moreover, no gender difference was observed for T. gondii infection for any of the reservoir species investigated here. This was congruent with findings outlined elsewhere [20,58,67], suggesting that gender-based differences in rodent behavior, immunity, and/or physiology do not influence T. gondii infection. Within-species comparisons also showed no difference in infection with age, except in APC where juveniles were more commonly infected than adults in R. norvegicus. Nonetheless, we should be cautious when interpreting this outcome given the low number of juveniles trapped for certain species, including R. norvegicus (e.g., in APC: 7 juveniles out of 61 individuals). Regardless of the locality and the small mammal species, no relationship was found between T. gondii prevalence and flea carriage. Although ectoparasites have been hypothesized as paratenic hosts and vectors of T. gondii [66,77,87], our results did not allow us to identify a significant relationship between T. gondii infection and the presence of ectoparasitic fleas, suggesting that flea parasitism does not appear to influence T. gondii infection in small mammals.
Furthermore, except for Saint-Jean where prevalence was higher in June 2018 than in October 2017 (32.5% vs. 6.4%), no differences in T. gondii prevalence were found between trapping sessions irrespective of the district or the species considered. This observation was also made for rodent-borne T. gondii seroprevalence in Niamey city (Niger) [58], suggesting that infection in small mammals may consistently occur throughout the year in West African urban regions. Nevertheless, we cannot exclude that the difference observed in Saint-Jean may be because both trapping sessions correspond to the rainy season. However, our data are restricted to only two trapping sessions, thus precluding any robust and definitive conclusion about potential seasonal variations in small mammal-borne T. gondii prevalence. Temporal surveys during each season and over at least two years would be required to capture such possible temporal dynamics.
The "district-based GLMMs" for explaining T. gondii infections did not show any clear influence of the local socioenvironmentat least as defined by the proxies used in this study. This finding may be interpreted in the light of the ubiquitous nature of T. gondii, which is adapted to diverse environmental matrices, including soil, vegetation, standing water, and the marine environment [71,86] and even in polar areas where felids are absent [31]. Rather than being impacted by socio-environmental determinants, T. gondii prevalence in Cotonou was shown to be related to the relative abundances of black rats and shrews. In fact, a high prevalence of the parasite appeared to be significantly associated with a low abundance of R. rattus, and a high abundance of C. olivieri. The social and feeding behavior of R. rattus could explain the low prevalence of T. gondii observed in this species. As explained above, the black rat is a good climber which often builds nests in roofs of houses [34], which might limit direct contact with the bare soil where T. gondii oocysts disseminate. Moreover, the large size of Rattus species [18,64] makes them less prone to predation by domestic cats which act as final hosts [14,35,65,78]. However, this interpretation, outlined elsewhere [1,40,67], should be taken with caution since the main  mode of T. gondii transmission in natural small mammal populations is still unknown [32]. Alternatively, shrews may appear to be an important host for the circulation of T. gondii within the urban environment. Altogether, it appears that species-specific susceptibility to Toxoplasma may be a key element of the ecology of this ubiquitous parasite. Therefore, we stress the need to further investigate the intricacies of host-T. gondii interactions at the species level.
Our "APC-based GLMMs" showed that the probability of T. gondii infection in the small mammal community was positively correlated with the abundance of M. m. domesticus. This rodent species is generally considered to be the preferred prey of domestic cats [35,65], which could lead to increasing rates of oocysts excretion in the environment and thus higher infection opportunities for other species, including small mammals. Of note, although probably not very frequent, cats were sometimes observed wandering inside the seaport storehouses during our field trapping campaigns. Moreover, in wild-derived mice, the hypothesis of the maintenance of T. gondii infection over several generations by congenital transmission has been formulated [50,56,62,63], and might at least partly explain the high prevalence observed in APC where all house mice were caught. Mice were found to be significantly more infected in September 2017 (rainy season) than in March 2018 (end of the dry season). Assuming environmental contamination, such a trend could be explained by the characteristics of sampled APC sites. Mice were essentially trapped in the industrial zones of APC that are characterized by large cemented surfaces and a total absence of vegetation. These conditions could be particularly unfavorable to the survival of oocysts during the dry season. Such a hypothesis is congruent with wet weather and water runoff as factors facilitating T. gondii oocyst survival and/or dissemination [67,80,84], and might explain the high prevalence observed in APC mice during the rainy season. From there, spatio-temporal surveys over different seasons in this very peculiar environment are needed to draw robust conclusions.
No significant spatial cluster of T. gondii-infected animals was observed in Agla, Ladji and APC, but one significant cluster was retrieved in Saint-Jeanat site S6 where 56% of individuals were found to be infected. This site is adjacent to sites S7 and S8 where prevalence reached 27% (4/15) and 30% (3/10), respectively. This cluster can be interpreted as a local hotspot of parasite circulation. Based on our knowledge of the major routes of T. gondii transmission to intermediate hosts, we would expect the parasite infection to be strongly dependent on cat density [3,29,43,44,88]. Unfortunately, to our knowledge, robust data on cat abundances and activities in Cotonou are not available. During a previous survey (Houémènou et al., personal information) in the exact same households of Agla, Ladji and Saint-Jean conducted from September 2016 to June 2017 (i.e., just before the start of the sampling campaigns for the present study in October 2017 and June 2018), cats were observed or indicated by inhabitants at least once in 2/10 sites in Ladji, 3/10 sites in Agla, and up to 5/10 sites in Saint-Jean, thus allowing a rough indication of cat abundance in our districts. The apparently frequent occurrence of domestic cats in Saint-Jean households could explain the local hotspot identified by our clustering analyses. However, in the absence of reliable data, this hypothesis remains to be formally tested and it would therefore be critical in future studies on Toxoplasma eco-epidemiology in southern Benin to take cats (and other potentially impacting factors, such as climate data, co-infection, and human exposure) into account, since they should represent rarely investigated but obviously crucial factors.
The seven samples genotyped here were all classified as belonging to the Africa 1 lineage, a T. gondii lineage widespread in most West and Central African countries, including Benin [38,40,47,57]. They were retrieved in six shrews (two individuals in each urban district) and one black rat (originating from APC). Although the number of genotyped individuals is low, it nonetheless confirms the predominance of Africa 1 strain in domestic and peri-domestic environments of Southern Benin, including Cotonou, as previously shown in chickens [47] and, beyond, throughout tropical Africa on different hosts [38,57]. The Africa 1 strain is considered lethal in laboratory and wild mice (Mus musculus domesticus) [40,47,49,57] and the link between strain virulence and host availability in shaping genetic diversity of Toxoplasma worldwide is supported by an increasing number of studies [37,54,72]. In Senegal, the invasive house mouse Mus musculus domesticus is currently the most abundant rodent in seaport cities such as Dakar and Saint-Louis, as well as in several central and northern localities where most T. gondii isolates belong to non-mouse virulent lineages [21]. On the contrary, this invasive rodent species has not yet disseminated to certain inland regions where the Africa 1 mouse-virulent T. gondii lineage was shown to be predominant [39]. In Benin, house mice invasion is probably still in its earliest phases since this species has been described almost only within APC [25,51]. These differences in the respective history and dynamics of the house mouse invasion taking place in these two countries may explain the distinct T. gondii strains that circulate among local M. musculus populations. Testing such a hypothesis requires us to investigate the virulence of the autochthonous Africa 1 strain in invasive house mice from Senegal and Benin. Extended genomic analyses to better capture and understand the diversity of circulating T. gondii strains from wild small mammal are needed since infectivity and morbidity of toxoplasmosis was shown to be related to the protozoan genotype involved [48,72,75].
In conclusion, our study provides a new support for the ecological ubiquity of T. gondii in a West-African country. Especially, we (i) showed the need to investigate both brain and heart organs in molecular surveys of T. gondii to draw a more accurate picture of the parasite prevalence among small mammals, (ii) confirmed the dominance of Africa 1 lineage as the dominant strain within West-African urban areas, and (iii) highlighted the intricate interrelationships between several biotic and abiotic determinants that make the circulation of T. gondii a local context-dependent process in which the underlying factors seem to differeven at a very local scalefrom one socio-environmental situation to another. From all the foregoing, multidisciplinary studies are key to integrate socioeconomic, biological and environmental components when studying the ecology and transmission to humans of T. gondii in domestic environments.