Determinants of Crimean–Congo haemorrhagic fever virus exposure dynamics in Mediterranean environments

Abstract Crimean–Congo haemorrhagic fever (CCHF) is an emerging tick‐borne human disease in Spain. Understanding the spatiotemporal dynamics and exposure risk determinants of CCHF virus (CCHFV) in animal models is essential to predict the time and areas of highest transmission risk. With this goal, we designed a longitudinal survey of two wild ungulate species, the red deer (Cervus elaphus) and the Eurasian wild boar (Sus scrofa), in Doñana National Park, a protected Mediterranean biodiversity hotspot with high ungulate and CCHFV vector abundance, and which is also one of the main stopover sites for migratory birds between Africa and western Europe. Both ungulates are hosts to the principal CCHFV vector in Spain, Hyalomma lusitanicum. We sampled wild ungulates annually from 2005 to 2020 and analysed the frequency of exposure to CCHFV by a double‐antigen ELISA. The annual exposure risk was modelled as a function of environmental traits in an approach to understanding exposure risk determinants that allow us to predict the most likely places and years for CCHFV transmission. The main findings show that H. lusitanicum abundance is a fundamental driver of the fine‐scale spatial CCHFV transmission risk, while inter‐annual risk variation is conditioned by virus/vector hosts, host community structure and weather variations. The most relevant conclusion of the study is that the emergence of CCHF in Spain might be associated with recent wild ungulate population changes promoting higher vector abundance. This work provides relevant insights into the transmission dynamics of CCHFV in enzootic scenarios that would allow deepening the understanding of the ecology of CCHFV and its major determinants.


INTRODUCTION
Crimean-Congo haemorrhagic fever (CCHF) is one of the human viral diseases of greatest concern to the World Health Organization because of its high lethality, the absence of efficient prophylactic measures (Keshtkar-Jahromi et al., 2011), its capacity for human-to-human transmission (Tsergouli et al., 2020) and the enormous mutational capacity of the causative virus (CCHF virus [CCHFV]; Bente et al., 2013). The virus is mainly transmitted by tick bites and a high proportion of infected people can cope with the infection without clinical manifestations, leading to an important but currently limited impact.
Even so, CCHFV is widely distributed in the African, European and Asian continents (Bente et al., 2013), and certain conditions seem to be favouring the emergence of human cases. In the Iberian Peninsula, the virus appears to be endemically established Moraga-Fernández et al., 2021). Nevertheless, it is not until the end of the 20th century (1985) that exposure to CCHFVfirst described in 1944 (Hoogstraal, 1979)-was reported in Portugal (Filipe et al., 1985). In 2010, CCHFV was detected in Hyalomma lusitanicum ticks collected on red deer (Cervus elaphus) in western Spain. In 2016, the first human case of CCHF was notified (Cajimat et al., 2017;Negredo et al., 2017). Ten cases with three deaths were notified until 2021 (Sánchez-Seco et al., 2022) and two additional cases (one lethal) were documented in summer 2022 (ECDC, 2022a). Attention paid to Hyalomma spp. ticks in Europe was anecdotal before CCHF human cases were notified (Barandika et al., 2011;Ruiz-Fons et al., 2006).
From then on, Hyalomma spp. dynamics was studied in greater depth (Spengler & Estrada-Peña 2018;Valcárcel et al., 2020). ECDC regularly updates the distribution of European tick species, currently including H. lusitanicum (ECDC, 2022b). To date the relationship between H.
lusitanicum abundance at fine spatiotemporal scales and virus transmission remains unexplored, leaving an important gap in the knowledge of CCHFV ecology. There is no effective vaccine to prevent the potentially serious consequences of CCHFV infection for humans, and thus prevention is the only approach to avoid new human cases.
An essential parameter in ecology and biodiversity conservation is population size of living organisms in ecosystems. Estimating population size is required to estimate their contribution to the ecological communities in which they exist and assess the status of their populations for conservation (e.g. Nunney & Elam, 1994) or population control purposes (e.g. Hagen et al., 2018;Shea, 1998). In epidemiology, knowing the size of the epidemiological population under study is also essential to make inferences about diseases or pathogens in the population (Walton et al., 2016). Species abundance data are generally limited in space and time, especially invertebrates. Only in cases where invertebrates are of conservation interest (Micó et al., 2013), or relevant as pests (Ørsted et al., 2021), or if they are vectors of health-threatening pathogens (Jaenson et al., 2012), is there interest in promoting research on their spatial distribution and population dynamics.

Study area
The study was carried out in Doñana National Park (DNP), in the southwest of peninsular Spain (37 • 01′N 6 • 26′W; Figure 1)

Sampling
Between 2005 and 2020, we attended wild ungulate population control events in DNP. The animals were shot by park rangers without considering animals' sex, age or health status but not for the purpose of this study, so according to current EU and Spanish animal protection regulations, an ethics committee report was not required for this survey. We surveyed red deer and wild boar per year with a homogeneous spatial distribution across DNP each survey year (Table S1). Ticks were counted and collected immediately after the animals were shot. We recorded animals' sex and age (from tooth eruption patterns; Saénz de Buruaga et al., 2001) and collected blood. Animals were grouped into (i) yearlings (6-12 months old), (ii) juveniles (12-24 months old) and (iii) adults (>24 months old). Samples were preserved refrigerated (4-8 • C) during transport to our laboratories where they were conserved until analyses. The location of each shot ungulate was recorded with portable GPS devices.
Ungulate sampling was performed mainly between October and November (

Serological analyses
No animals under 6 months of age sampled by park rangers were selected for serological study due to the potential interference of the presence of maternally derived antibodies with the serological diagnosis of virus exposure (see Casades-Martí et al., 2020). Sera were obtained after centrifuging blood samples at 10,000 × g for 10 min and preserved frozen at -20 • C. The presence of specific antibodies against CCHFV in sera was analysed using a commercial ELISA (ID Screen ® CCHF Double Antigen Multi-Species, IDvet, France; Sas et al., 2018) following the manufacturer's indications.
In this study, greater importance was given to yearling specimens as only in this age class does the presence of antibodies to CCHFV indicate that infection was recent (at least within the first year of the animal's life). In adult animals, the presence of antibodies may be either due to recent exposure or from previous infections because CCHFV antibodies are long-lasting  and the ELISA is highly sensitive (Comtet, 2021). Thus, sampling was slightly biased towards yearlings in relation to their proportion in the population. We

Environmental predictors
The ecology of CCHFV is intimately linked with the dynamics of hosts and vectors (Bente et al., 2013), and these are influenced by environ-   Barroso et al., 2020). Each LMU is allocated (at least) one transect (Barasona et al., 2014). Observations allowed estimating the annual kilometric abundance index (KAI) as the average number of individuals of a species per transect kilometre. We estimated the average KAI for each LMU and study year for red deer and wild boar ( Figure S1).

Vector abundance
Ticks were counted on 1965 wild ungulates sampled between 2010 and 2020, collected, and morphologically identified to species level . Tick developmental stage (larva, nymph or adult) and species were recorded, and these were preserved at were estimated ( Figure S2). Similarly, daily temperature data recorded by the weather station were used to estimate the average winter, spring and summer temperatures ( Figure S3).

Data analyses
The data were thoroughly checked to rule out any potential interference in statistical modelling (Zuur et al., 2010 The process to select the best-fit model was carried out by building all possible models, ranking them according to their Akaike information criterion (AIC), and calculating the average model with models showing an AIC difference with the best-fit model (ΔAICc) of less than two units (Table S2); we used the R package 'MuMIn' (Barton, 2009). For each model set, we analysed multicollinearity effects by estimating the variance inflation factor (vif) for model parameters (Fox & Weisberg, 2019). A final step in the analysis was to estimate the relative contribution of each model parameter to the variation in the probability of exposure to CCHFV. For this, we partitioned the coefficient of determination R 2 of the predictors included in the best model by estimating the semi-partial (part) R 2 and the structural coefficients with the R package 'partR2' (Stoffel et al., 2021). We thus studied the specific contribution of each model predictor to explain the variance of the model. The modelling process was repeated only for yearlings (N = 206) to estimate the determinants of the annual CCHFV incidence.

RESULTS
The results showed a medium-to-high CCHFV exposure prevalence  (Table 2).
Despite the differences, the temporal pattern of change in annual population seroprevalence was similar for both species (Figure 2), and that displayed a slight negative trend within the time series. The ageassociated increasing exposure pattern was also evident in this study,  (Table 3), whereas the temporal pattern in adults was less pronounced and stable within a high seroprevalence (Figure 2). There was a marked variation in the interannual incidence of new cases of CCHFV infection according to observations in yearlings. We observed strong changes between contiguous seasons that in some cases reached a >30% decrease in just one year.
We found a very robust relationship between the spatial distribution of predicted H. lusitanicum abundance and the risk of exposure to CCHFV (Table 4; Figure S4). This spatial pattern was similar for red deer and wild boar, with the northern part of DNP as the area of highest exposure risk (Figure 3). We observed that the clear contrast in CCHFV seroprevalence between red deer and wild boar was evident and the difference was statistically significant (Table 5; Figure S5).
However, the yearling model showed no host effect on virus exposure  Figure S6). At the scale of the study, we observed a marked effect of wild ungulate abundance, with higher risk at higher deer abundance and with a limited but negative influence of wild boar abundance (Table 5; Figure S5). The results of the annual incidence modelling also indicated a significant positive effect of deer abundance on exposure  Table 1), their estimates and associated standard errors (SE), the statistic (z) and the p-value. *p < .05; **p < .01; ***p < .001. risk ( Figure S6), confirming it as an important risk factor. Weather predictors were overall less important in determining exposure risk, but the model results show their risk-modulating capacity. The influence of rainfall on exposure risk was contrasted, with an overall lower risk in rainy years, but with a positive effect of the accumulated precipitation during the summer drought. We observed a positive risk-modulating capacity of summer temperature which, although with a lower effect when compared to host factors, also appears as a relevant risk factor for the annual risk of exposure to CCHFV. The variance decomposition of the model into its predictor components corroborated the results of the prevalence and incidence models ( Figures S7 and S8

Methodological issues
The selection of an animal study model proved to be a good choice Further, antibodies can be transmitted as passive immunity from mother to offspring (see González-Barrio, Fernández-de-Mera, et al., 2015). To rule out the detection of maternally derived antibodies, no animals younger than 6 months old were included in the study.
We therefore modelled the probability of exposure to the virus in the general population and, specifically, in yearlings.
Modelling H. lusitanicum burdens on hosts on an annual basis could yield highly imprecise predictions due to limited sample size, so its abundance model was represented with spatial but not temporal accuracy. Therefore, we separately analysed the timeless influence of H.
lusitanicum spatial abundance and the temporal influence of additional environmental predictors. We were thus unable to explore whether the potential temporal variation in the spatial patterns of vector abundance could influence the exposure to CCHFV. However, we may assume that environmental favourability at the landscape level is sustained over time even when weather modulates vector annual abundance. We would expect that more environmentally favourable areas for H. lusitanicum would always yield higher abundance than less favourable areas, at least at the spatial scale of DNP, and that is what we observed (Peralbo-Moreno et al., 2022).

CONCLUSIONS
Despite the different measures of incidence and prevalence in yearlings and the total population, respectively, we note that determinants of inter-annual variation in virus exposure are also predominantly host related. This perhaps points out that although less appropriate, population seroprevalence may be a good temporal parameter over long time scales to study the determinants of virus transmission risk. This is especially relevant for studies to be carried out in areas where, unlike our specific case, sampling of wild ungulates is performed in hunting events where the number of yearlings shot is usually very low (see Casades-Martí et al., 2020 for details about specific age-class shooting in hunting events). However, to monitor temporal variations in pathogen exposure, it is still recommendable including yearlings in surveys González-Barrio, Fernández-de-Mera, et al., 2015).
This work provides relevant insights into the transmission dynamics of CCHFV in enzootic scenarios, not only providing important information to reduce the risk of virus transmission by taking actions according to our findings, but also to deepen the understanding of the ecology of CCHFV and its major determinants.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT
The study was conducted according to the guidelines of the Declaration of Helsinki.