Antibodies against medically relevant arthropod-borne viruses in the ubiquitous African rodent Mastomys natalensis

Over the past decades, the number of arthropod-borne virus (arbovirus) outbreaks has increased worldwide. Knowledge regarding the sylvatic cycle (i.e., non-human hosts/environment) of arboviruses is limited, particularly in Africa, and the main hosts for virus maintenance are unknown. Previous studies have shown the presence of antibodies against certain arboviruses (i.e., chikungunya-, dengue- and zika virus) in African non-human primates and bats. We hypothesize that small mammals, specifically rodents, may function as amplifying hosts in anthropogenic environments. The detection of RNA of most arboviruses is complicated by the virus’s short viremic period within their hosts. An alternative to determine arbovirus hosts is by detecting antibodies, which can persist several months. We developed a high-throughput multiplex immunoassay to detect antibodies against 15 medically relevant arboviruses. We used this assay to assess almost 1,300 blood samples of the multimammate mouse, Mastomys natalensis from Tanzania. In 24% of the samples, we detected antibodies against at least one of the tested arboviruses, with high seroprevalences of antibodies reacting against dengue virus serotype one (7.6%) and two (8.4%) and chikungunya virus (6%). Seroprevalence was higher in females and increased with age, which could be explained by inherent immunity and behavioral differences between sexes and the increased chance of exposure to an arbovirus with age. We evaluated whether antibodies against multiple arboviruses co-occur more often than randomly and found that this may be true for some members of the Flaviviridae and Togaviridae. In conclusion, the development of an assay against a wide diversity of medically relevant arboviruses enabled the analysis of a large sample collection of one of the most abundant African small mammals. Our findings suggest a role in the transmission of multiple arboviruses by this ubiquitous rodent and provide a solid foundation for future molecular screening to elucidate the role in the arbovirus transmission cycle. Author summary One of the main causes of zoonotic related human morbidity and mortality is the transmission of arthropod-borne viruses such as dengue virus, Yellow Fever virus, and chikungunya virus. These viruses cannot only infect humans but also livestock, pets, and wildlife, though our understanding of their non-human hosts remains limited. Rodents are thought to be an interesting host for these viruses because they can be abundant, often live near humans and some are already known to be viral hosts. However, research has focused on non-human primates, neglecting other potential hosts. To address this gap, we have developed a high-throughput antibody test to screen rodent blood against 15 different arboviruses. Our findings reveal that a proportion of Mastomys natalensis, a common African rodent species, carry antibodies that (cross-)react against these viruses. We hypothesize that immunologically naïve juveniles may drive transmission, particularly during population outbreaks. These outbreaks coincide with environmental conditions that are favorable for mosquitoes, the vectors of these viruses. Thus, increasing the risk of spillover to humans, livestock, and wildlife. Understanding the role of rodents in arbovirus transmission dynamics is crucial for mitigating zoonotic disease risks.

provide a solid foundation for future molecular screening to elucidate the role in the arbovirus transmission cycle.

Author summary
One of the main causes of zoonotic related human morbidity and mortality is the transmission of arthropod-borne viruses such as dengue virus, Yellow Fever virus, and chikungunya virus.These viruses cannot only infect humans but also livestock, pets, and wildlife, though our understanding of their non-human hosts remains limited.Rodents are thought to be an interesting host for these viruses because they can be abundant, often live near humans and some are already known to be viral hosts.However, research has focused on non-human primates, neglecting other potential hosts.To address this gap, we have developed a high-throughput antibody test to screen rodent blood against 15 different arboviruses.Our findings reveal that a proportion of Mastomys natalensis, a common African rodent species, carry antibodies that (cross-)react against these viruses.We hypothesize that immunologically naïve juveniles may drive transmission, particularly during population outbreaks.These outbreaks coincide with environmental conditions that are favorable for mosquitoes, the vectors of these viruses.Thus, increasing the risk of spillover to humans, livestock, and wildlife.Understanding the role of rodents in arbovirus transmission dynamics is crucial for mitigating zoonotic disease risks.

Introduction
The African continent harbors a diverse array of infectious diseases with profound impacts on public health, economic development, and general well-being [1,2].Diseases caused by arthropod-borne viruses, collectively known as arboviruses, are a growing threat for Africa and the rest of the world especially in relation to climate and environmental changes [3,4].Arboviruses are a polyphyletic clade that includes several viral families, of which the most important are Flaviviridae, Togaviridae, Bunyaviridae, and Reoviridae [5].Some well-known arboviruses, notorious for their devastating effects on human health, are dengue virus, Yellow Fever virus, Zika virus, and chikungunya virus.Mosquitoes, ticks, sandflies, and midges are the primary vectors responsible for arbovirus transmission as they engage in hematophagy.These vectors do not only affect humans and livestock, but also a wide range of wildlife hosts [6][7][8].Indeed, while for some arboviruses morbidity and mortality can be high in humans, similar impacts have been detected in other animals by arboviruses such as Rift Valley Fever virus in goats and sheep, West Nile virus in birds and horses and Japanese Encephalitis virus in birds and pigs [4,5,[9][10][11].The (re-)emergence of arboviruses appears, at least partially, to be caused by the increased urbanization and global connectivity, natural genetic evolution of viruses, and adaptations of the vectors to changing climate and environments [11,12].Emerging arboviruses pose a threat for humans, livestock, as well as wildlife, therefore it needs to be approached from a One health perspective (i.e., including human, animal, and environment health) [13,14].Nevertheless, our knowledge about the extent to which wild animals can serve as sylvatic hosts for human-infecting arboviruses and the natural diversity of arboviruses remains insufficient.This significantly limits our understanding of arbovirus transmission dynamics, which is required to develop effective control measurements.
For decades, efforts have been made to identify natural reservoirs of arboviruses to monitor, prevent, and control sources of infection that pose a threat to human health [15][16][17].Several studies have proposed non-human primates as significant potential reservoirs for arboviruses, as they have found arbovirus antibodies and viral RNA in this animal group [18][19][20].However, other animal groups such as small mammals have often been neglected as arbovirus hosts [21].Sporadic reports of arboviruses in small mammal species suggest that a more comprehensive investigation of their potential role as a host is needed [22][23][24].
Rodents have a number of characteristics that could make them potentially important hosts for several pathogens, including arboviruses [23].Particularly the high species diversity, the fact that many species can reach high population abundances and turnover rates.The risk of pathogen spillover to humans increases with the role of some rodents as a pest species, due to their proximity to humans [25,26].A notable example of such a pest species is the ubiquitous rodent Mastomys natalensis, commonly known as the multimammate mouse.This species inhabits many regions of sub-Saharan Africa, with a preference for crop fields, fallow land, and typically occurring within or at the fringes of urban settlements [27,28].In east Africa, and especially in Tanzania, the reproductive cycle of M. natalensis is strongly correlated with seasonal rainfall which leads to strong seasonal fluctuations in density (20 -500 individuals/hectare), and occasionally even severe population outbreaks (>1000 individuals/hectare) [29][30][31][32].This has large ecological and societal impacts due to crop damage and influences seasonal transmission dynamics of different pathogens [33][34][35].The multimammate mouse is a known host for several zoonotic pathogens such as Lassa mammarenavirus, Yersinia pestis, Leptospira interrogans, Leishmania major as well as different ecto-and endoparasites [31,[36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52].No studies have investigated or reported on arboviruses in M. natalensis, except Diagne et al. (2019) who have detected Usutu virus RNA in M. natalensis.However, other studies have reported on sporadic arbovirus detections in other rodent species in sub-Saharan Africa [22,24,53,54].These findings, along with the ecology of M. natalensis (i.e., high abundance during population outbreaks, proximity to humans, and its status as a proven pathogen host) may suggest that this species plays a role in the natural transmission cycle of arboviruses.Consequently, M. natalensis could thus pose a risk to humans in east Africa, particularly in Tanzania as an amplifying host.
Due to the symptomatic similarities between arbovirus and malaria infections, which has a prevalence of around 20% in Tanzania, it is probable that arbovirus cases are underreported [60,61].While these studies confirm that the local human population is indeed exposed to arboviruses, the specific dynamics of arbovirus transmission in this region remains unclear.
The goal of this study was to investigate the potential of wild M. natalensis to serve as a host for arboviruses in their natural environment.To achieve this, we first developed a multiplex immune assay to detect immunoglobulin G (IgG) antibodies against 15 different arboviruses causing disease in humans and subsequently conducted a comprehensive screening of almost 1,300 blood samples obtained from M. natalensis from Morogoro, Tanzania.

Sample origin
The samples used in this study were collected during previous published and During these studies, M. natalensis were live caught using Sherman traps (H.B.Sherman Traps, Tallahassee, USA) in a heterogeneous landscape (e.g., woodlands, maize fields, and fallow land) on the premises of the Sokoine University of Agriculture in Morogoro, Tanzania.Blood was collected from the retro-orbital plexus using a 50µL hematocrit capillary tube and preserved on filter paper (Serobuvard; LDA 22; Zoopole, France).The filter paper was dried for 12 hours at room temperature and archived at -20°C in envelopes with desiccant.Additional data related to characteristics such as sex, reproductive status, weight, and body measurements were recorded.More detailed Information pertaining to the trapping procedures and sampling methodology can be found in the primary research documents associated with these studies [31,[62][63][64].
To obtain positive control samples we inoculated captive M. natalensis individuals (age: 5-12 months) from our breeding colony at UAntwerp with recombinant virus-derived proteins (Table 1) [66][67][68].We subcutaneously injected 4µg of the respective virus protein and 1µL of vaccine adjuvant (Quil-A® adjuvant; InvivoGen; Toulouse, France), dissolved in autoclaved phosphate buffered saline (PBS) to achieve a final volume of 1mL.This inoculum was evenly divided, with 0.5mL administered into the scruff and 0.5mL into the hindlimb of the animal, using a 25-gauge, 12.5mm needle and a 0.5mL syringe.This inoculation was duplicated for each viral protein (i.e., performed in two mice) and repeated twice for each mouse (i.e., inoculation on day 0 and day 20).We collected blood, according to the same method as in the previous studies, every 10 days from day zero until day 30, at day 30 we also collected whole blood from which serum was extracted.Serum from day 30 from individuals were the antibody response increased over time were considered as positive samples.Day 30 had the highest antibody titer in our tests and is also a time point at which IgG antibody development is anticipated to have reached a peak [69,70].

Data analysis and statistics
All data preparation, analysis and statistical procedures were conducted using R Statistical Software (R version 4.3.3)[73].We used the body weight of the wild-caught M.
natalensis individuals at the time of sample collection as a rough proxy for age, which we subdivided into three categories based on the 1/3 quantiles of weight; juvenile (5 -26.7g), subadult (>26.7 -42g) and adult (>42 -91g).These weight classes coincide to the expected sexual maturity, with sexual maturity estimated to occur between 30 -40g [29,32].
To control for variation between different assay plates and testing days, the MFI results were transformed to relative antibody units using the positive dilution series as a standard curve.The MFI result of the positive control starting dilution (i.e., 1:200) was equalized to 3,125 units and each following dilution step was adjusted proportionally (i.e., the final dilution step 1:625,000 corresponded to 1 unit).The results of the two sessions were combined by linear alignment adjustment.This alignment was based on 86 duplicate samples encompassing the measurable range, allowing the adjustment of the results from the first session.
Finally, each sample was categorized as a binary value (i.e., 1= positive, 0= negative) for each of the tested arboviruses.This was done based on whether the unit value exceeded the mean cutoff value for that specific arbovirus antigen.Five cutoff values were determined for each arbovirus antigen: I) the mean plus three times the standard deviation of the negative controls (i.e.'NegCtrl') [18,65]; the change-point analysis, using R package 'changepoint' (version: 2.2.4), calculated at most one changepoint based on the II) mean (i.e.'CHP.m'),III) variance (i.e.'CHP.v') and IV) a combination of mean and variance (i.e.'CHP.mv') of wildcaught samples [74,75] and V) the maximum value of an average antibody curve (i.e.'Recap').
This curve was based on wild-caught individuals that were recaptured at least three times and showed seroconversion.Seroconversion of an individual was considered when the individual's maximum unit value was at least four-fold the minimum unit value.This four-fold increase is a standard seroconversion confirmation measure in human antibody studies [76].
An average antibody curve, with days as the explanatory variable, was created for each antigen by aligning the maximum unit value of each recaptured seroconverted individual to the same day.The binary results were used to calculate the seroprevalence for each arbovirus along with a 95% confidence interval (CI), using the 'binom.exact'from the package 'binom' support is reserved for p-values lower than 0.05 (< 0.001, symbol: ***).This representation in terms of statistical support is based on current statistical reporting practices [81].

Results
In total 1,280 DBS samples were tested of which 660 were female, consisting of 256 juveniles, 172 subadults and 232 adults, 620 samples were male with 199 juveniles, 313 subadults and 108 adults.Samples of recaptured individuals were considered as individual samples for all analysis.

Seroprevalence
The seroprevalences according to the different cutoff methods showed at least a weak statistical support for a different seroprevalence compared to the antibody curve seroprevalence for almost all arboviruses.Histograms of the data and seroprevalence for each tested arbovirus antigen according to the different cutoff methods is shown in S1 and S2 Figs.
The cutoff value according to the antibody curve based on the recaptured seroconverted individuals was used as the main cutoff value for all further calculations.

Pairwise arbovirus serostatus correlation
The correlations in serostatus of samples between the tested arboviruses are visualized in Fig 2 .Correlation between two arboviruses is depicted in color scale with the statistical symbol, lower triangle, and the number of positive samples in the upper triangle.The correlation in sample response between the tested arboviruses ranged from -2.44% for TBEV and ZIKV with no statistical support (p= 0.384) to almost 59% between WSLV and USUV with a strong statistical support (p< 0.001).The dendrogram based on the hierarchical clustering of the correlation showed that WSLV -USUV (correlation= 58.52%; p< 0.001), DENV2 -DENV4 (correlation= 56.03%; p< 0.001) and DENV1 -CHIKV (correlation= 50.31%; p< 0.001) are relatively closer to each other than to other tested arboviruses.
The analysis of the age variable showed a strong statistical support for a higher seroprevalence in subadults than in juveniles for DENV1 and DENV2, a moderate support for CHIKV and a weak support for YFV, ZIKV, TBEV, WSLV and CCHFV.There was no support for a difference in subadult and juvenile seroprevalence in the other tested arboviruses.A significantly higher seroprevalence in adults compared to juveniles was shown for ZIKV, DENV1, DENV2, WSLV and CHIKV with a strong support.A moderate support for a higher seroprevalence in adults than in juveniles was detected for YFV and CCHFV.Adults showed a weak statistical support for a higher seroprevalence in contrast to juveniles for WNV and TBEV.All other tested arboviruses showed no support for a statistical difference between adults and juveniles.The comparison between subadults and adults showed a strongly supported statistical difference for DENV1 and CHIKV with a higher seroprevalence in adults.
A moderate support for a higher seroprevalence in adults compared to subadults was detected for ZIKV and CCHFV.Yellow Fever virus, WNV, WSLV and ONNV showed a weak support for a statistically higher seroprevalence in adults than in subadults.The other arboviruses showed no statistically significant difference between adults and subadults.See Table 3 for estimates, standard errors, and p-values.In the case of DENV4, were there was a weak support for an interaction, the analysis showed that there was a strong statistical support for a higher seroprevalence in female adults compared to female juveniles (Est.female -adult ± SE= 1.844 ± 0.486; p< 0.001).A moderate support was shown for a higher seroprevalence in female subadults compared to female juveniles (Est.female -subadult ± SE= 1.379 ± 0.517; p= 0.008) and a weak support for a higher seroprevalence in female adults compared to male adults (Est.female -adult ± SE= 1.898 ± 0.736; p= 0.010).For USUV the model analysis with a weak interaction, showed that there was weak statistical support for a higher seroprevalence in female adults compared to female subadults (Est.female -adult ± SE= 0.868 ± 0.345; p= 0.012) and also a weak support for a higher seroprevalence in female adults compared to male adults (Est.female -adult ± SE= 1.515 ± 0.612; p= 0.013).samples from recaptures.The negative control-based cutoff (i.e., the mean plus three times the standard deviation of the negative control samples) gave unrealistic high seroprevalences.
This can be explained by the fact that the negative control samples originate from a breeding colony.These animals have thus never been exposed to a natural environment and the pathogens that occur in the environment.The statistical methods vary in their seroprevalence with some methods approximating the estimated seroprevalence according to the recaptured cutoff.This high degree of variation makes it difficult to decide on one method that works for all the tested arboviruses.The cutoff value for CCHFV seems extremely high compared to the other arboviruses, but the unit values for CCHFV are also much higher than for the other arboviruses (see S1 Fig) .The reason is that the unit values are calculated based on the positive dilution series and the positive controls for CCHFV were not of the same magnitude as for the other arboviruses.The value in determining a cutoff and the resulting seroprevalence is that it allows the comparison of results with previous and future studies on arbovirus seroprevalence in rodents or other wildlife.We are aware that the used cutoff and resulting seroprevalences could be an over-or underestimation and might not reflect the natural arbovirus seroprevalence.We therefore encourage future research to investigate and compare different cutoff methods for arbovirus (or pathogen) antibody detection in wildlife studies.
The detection of antibodies against each of 15 tested arbovirus antigens indicates that these arboviruses, or closely related viruses, are present in M. natalensis.The overall arbovirus seroprevalence of 24% suggests that this rodent species is commonly infected with one or more arboviruses and that it could thus play a significant role in virus transmission and persistence.Our results corroborate previous studies, which detected USUV and WNV RNA in respectively M. natalensis and M. erythroleucus, in Senegal [24,54].Besides in this genus, arboviral RNA has also been found in other rodents in Africa, such as Rattus rattus for USUV and WSLV and Desmodillus auricularis for WSLV [22,24,53].The findings in our study thus further corroborate that arboviruses are likely present in rodents, and specifically in the ubiquitous M. natalensis.The demographic and ecological characteristics of M. natalensis may have particularly important implications for arbovirus transmission.The population densities of M. natalensis in Tanzania are strongly dependent on weather conditions.More specifically, early rainfall and elevated temperatures lead to an exponential growth in the population density, due to an increase influx of juveniles [29,30,32].The rainfall and increased temperatures are also beneficial for the breeding of mosquitoes and the multiplication of arboviruses within these vectors [82].Further, M. natalensis is highly abundant around houses and in the crop fields at the fringes of the villages.These factors increase the likelihood of arbovirus outbreaks in M. natalensis populations, with the possibility of spillover to humans.
Arboviruses that show the highest seroprevalence are DENV1, DENV2, DENV4 and CHIKV, with seroprevalences between five to nine percent.These seroprevalences could be caused by cross-reactivity due to antibodies of other dengue virus serotypes or other flaviviruses binding to the non-structural protein 1 (NS1 protein) of DENV1, DENV2 and DENV4.The same effect could also be true for alphaviruses binding to the envelope protein 2 (E2 protein) of CHIKV.Whether these seroprevalences are indeed due to the presence of the arbovirus specific antibodies or a related arbovirus remains to be investigated.Nonetheless, it indicates that a part of the sampled M. natalensis population in Morogoro is exposed to dengue virus and CHIKV or respectively to a related flavivirus and alphavirus.This hypothesis is supported by the fact that flavi-and alphaviruses are the most prevalent arboviral genera in humans, compared to other arbovirus genera, and potentially thus also in rodents involved in the sylvatic cycle [83,84].
A recent health survey has shown that, in the same region as where our rodent population was sampled, a high percentage of the human population is seropositive for CHIKV (9.83%) [55].Another study in the same region reported acute infection of CHIKV in 1.28% of patients with fever and malaria-like symptoms [85].Although these studies have not found any indication of dengue virus in humans, a large-scale cross-sectional study in Tanzania has found CHIKV and dengue virus antibodies in respectively 28.0% and 16.1% of the population [56].These studies clearly indicate that the human population in Tanzania is exposed to arboviruses and then specifically to CHIKV and DENV.
The cross-reactivity analysis via the correlation matrix and hierarchical clustering (Fig 2) showed an antibody response correlation between WSLV -USUV (59%), DENV2 -DENV4 (56%) and DENV1 -CHIKV (50%).We expected that phylogenetically related arboviruses would show elevated levels of correlation due to cross-reactivity [86].A remarkable result in this cross-reactivity analysis is that DENV1 -CHIKV cluster together with a correlation of 50%, based on the serostatus of the tested samples.The branch DENV1/CHIKV clusters also closer to ONNV than to the branches of WSLV/USUV and DENV2/DENV4.This is unexpected since CHIKV belongs to the Togaviridae and DENV1 to the Flaviviridae [87].The proteins used for the antibody detection are also two different proteins, with the E2 protein used for the Togaviridae and NS1 protein for the Flaviviridae, thus limiting the possibility of crossreactivity.Although we cannot exclude that there might be similar epitopes between the different proteins, but other studies have already indicated that cross-reactivity between the E2 protein of the Togaviridae and NS1 protein of the Flaviviridae is limited [88,89].Given that both Togaviridae and Flaviviridae viruses are circulating in humans in East Africa, we hypothesize that these viral families may also both be present in rodents [5,10].More specifically, it is plausible that both viral families could be found in M. natalensis, where pathogen co-infections are common [49].This hypothesis is further supported by the fact that some viruses in both families are transmitted by the same arthropod vectors, such as Aedes aegypti and Aedes albopictus for both dengue virus and CHIKV [4,82].
For some of the tested arboviruses we found statistical support for a higher seroprevalence in females than in males.This result is supported by previous studies where it is shown that female mice have a stronger innate immune response than male mice [90].
Other animals (e.g., birds, fish, insects) as well as humans also display stronger immune responses in females [91][92][93][94][95].The major driving forces behind these immune differences are genetic (i.e., X-chromosomes) and hormonal (i.e., different estrogen and testosterone levels) [96].In the case of M. natalensis, behavioral differences could also be the cause for this divergence in seroprevalence.Previous studies have already shown that home range, behavior and pathogen presence differ between male and female M. natalensis [48,97].
Besides the sex effects, we also found statistical support for a positive age effect on the presence of antibodies in some of the tested arboviruses.This increased seroprevalence with age corroborates previous findings for other pathogens (i.e., Bartonella sp., Anaplasma sp., helminths, and arenaviruses) [48,49,98].This age effect further supports our hypothesis that M. natalensis is exposed to arboviruses and that individuals develop antibodies and gain immunity via repeated exposures throughout their life.To maintain the arbovirus transmission among the M. natalensis population, there needs to be a proportion of the population that is either chronically infected or immunologically naïve.Chronic infections in M. natalensis have already been documented for mammarenaviruses [31,99].However, as far as we are aware naturally occurring chronic arbovirus infections have not been reported in humans or non-human vertebrates.Therefore, the presumable driving factor in sustained transmission is the presence of immunologically naïve individuals.During the breeding season, which coincides with increased rainfall and temperature, there is an influx of immunologically naïve juveniles.This influx can reach high proportions during population outbreak periods [30,64].We thus expect that it is mainly juveniles who are the major factor in sustaining the arbovirus transmission cycle.We predict that the prevalence of arboviral genetic material will be higher in juveniles than in adults, since juveniles do not possess the necessary antibodies to fight of the infection.
We conclude from our detected antibody responses that arboviruses, or related viruses, are present in M. natalensis in Morogoro, Tanzania.The higher seroprevalence we detect in females can be explained by genetic, hormonal, ecological and/or behavioral differences between sexes.Individuals are exposed to these viruses throughout their life and gain immunity as they age.We hypothesize that juvenile M. natalensis play an essential role in sustaining arbovirus transmission as they are immunologically naïve and can reach high densities in favorable climate conditions that coincide with optimal vector conditions.More extensive screening, such as virus neutralization tests and molecular screening of these viruses within M. natalensis are necessary to quantify the contribution of this rodent species in the arbovirus transmission cycle.
unpublished studies conducted by the University of Antwerp (UAntwerp) and Pest Management Center of the Sokoine University of Agriculture on M. natalensis in Morogoro, Tanzania, between 2010 and 2019 [31,62,63] (Fig 1).In this study, the samples were divided in two screening sessions.The first session consisted of approximately 500 dried blood spot (DBS) samples, from wild captured mice that were used in infection and behavioral experiments in six different years (i.e., 2010, 2011, 2015, 2017, 2018, and 2019) with an average of 80 samples per year.The second session consisted of 800 DBS samples from mice involved in capture mark recapture experiments in 2017 and 2019.All samples were randomly selected from the studies regardless of individual characteristics or trapping period.

Fig 1 .
Fig 1. African continent with a focus on Tanzania.Samples were collected in the city of Morogoro (red triangle) which is located in the Morogoro region.

(version: 1 . 1 . 1 . 1 )
[77].The seroprevalence according to the different cutoff methods was compared to the seroprevalence of the antibody curve cutoff using the 'chisq.test'from the package 'stats' (version 4.3.3)[73]As an indication of cross-reactivity in antibody response between the tested arboviruses, pair-wise Pearson correlations were calculated on the binary results, according to the antibody curve cutoff, of all samples using the 'corr.test'function of R package 'psych' (version: 2.4.1)[78].The cross-reactivity in antibody response was visualized using the 'heatmap.2'function of the R package 'gplots' (version: 3.1.3)[79].A generalized linear model (logit link function and binomial error distribution) was constructed with the package 'stat' (version: 4.3.1),with the response variable being the binary serostatus of each sample[73].Age (juvenile, subadult and adult), sex and their interaction were included as explanatory variables.The analysis of variance was performed using a likelihood ratio test, with p-values calculated assuming a chi-squared distribution.Pairwise comparison of the seroprevalence was performed between the six combinations of the explanatory variables (two levels of sex and three levels of age), using the 'emmeans' package (version: 1.8.9)[80].To prevent reporting statistical findings based on the reliance of an arbitrary p-value of 0.05, we instead present significance in terms of levels of statistical support based on p-values.P-values exceeding 0.1 are labeled as "no" support and values around 0.05 (range 0.1 -≥ 0.01, symbol: *) as "weak" support."Moderate" support was assigned to p-values clearly below 0.05 (range < 0.01 -≥ 0.001, symbol: **), while "strong" The matrix is accompanied by a dendrogram based on the hierarchical clustering of the correlation coefficients.The branch lengths are a proxy for relative distance between arboviruses based on the serostatus response of the samples.

Fig 2 .
Fig 2. Correlation of the serostatus response between the tested arboviruses with a dendrogram of hierarchical clustering.Symbols in lower triangle represent significance of correlation, values in upper triangle, including diagonal line, represents the number of positive individuals for the corresponding arboviruses.

Fig 3
Fig 3 displays the seroprevalence for the six distinct levels (two levels of sex and three

Fig 3 .
Fig 3. Seroprevalence according to sex and age combinations with 95% confidence error bars for each arbovirus.Statistical support on seroprevalence difference is indicated by asterisks in the horizontal lines.Sample size: 660 females of which 256 juveniles, 172 subadults and 232 adult and 620 males with 199 juveniles, 313 subadults and 108 adults.Sample sizes for ONNV are 560 females: 150 juveniles, 150 subadults, 260 adults and 266 males: 94 juveniles, 92 subadults and 80 adults.

Table 1 . Recombinant arbovirus proteins for the inoculation of captive Mastomys natalensis.
Screening was done in 96 flat-bottom well plates, each plate contained DBS samples of 80 wild M. natalensis, two background controls, eight negative controls and a six step dilution series (1:200 -1:625,000) of a positive pool sample.Each well in the plate contained 50µL of the corresponding sample type.The samples of the wild M. natalensis were acquired [18,72]rf tube, maintained at a temperature of 4°C on a plate shaker.This elution was considered a 1:100 dilution and was diluted, with dilution buffer, to 1:200 prior to loading in the 96 well plate.This dilution gave the best signal to noise ratio in our preliminary tests and are in line with previous studies[18,72].The background control was reading buffer (1% bovine serum albumin, 0.05% NaN3, 100% phosphate buffered saline).The eight negative controls were four DBS, treated the same as the wild M. natalensis DBS, and four serum samples in a 1:200 dilution.All negative controls originated from the breeding colony at UAntwerp.The positive pool sample was compiled from serum from the positive individuals acquired through the arbovirus protein inoculation experiment.Serum from 15 positive individuals (i.e., one for each arbovirus antigen) was pooled to create the positive pool sample, each individual serum had a final dilution in the pool of 1:200.Plates, containing 50µL of sample and 25µL of bead mixture per well, were incubated for one hour at room temperature, in the dark and on a plate shaker (Heidolph Titrimax 100; Belgium).Results were quantified as the median fluorescent intensity (MFI) based on a minimum of 100 beads per antigen.

Table 3 . Difference in coefficient estimate on logit scale between the age levels with standard error (SE).
Data originates from the pairwise comparison of the age class variables of the generalized linear model.P-values marked in bold have at least a weak statistical support (p< 0.1).