Who acquires infection from whom? Estimating herpesvirus transmission rates between wild rodent host groups

To date, few studies of parasite epidemiology have investigated ‘who acquires infection from whom’ in wildlife populations. Nonetheless, identifying routes of disease transmission within a population, and determining the key groups of individuals that drive parasite transmission and maintenance, are fundamental to understanding disease dynamics. Gammaherpesviruses are a widespread group of DNA viruses that infect many vertebrate species, and murine gammaherpesviruses (i.e. MuHV-4) are a standard lab model for studying human herpesviruses, for which much about the pathology and immune response elicited to infection is well understood. However, despite this extensive research effort, primarily in the lab, the transmission route of murine gammaherpesviruses within their natural host populations is not well understood. Here, we aimed to understand wood mouse herpesvirus (WMHV) transmission, by fitting a series of population dynamic models to field data on wood mice naturally infected with WMHV and then estimating transmission parameters within and between demographic groups of the host population. Different models accounted for different combinations of host sex (male/female), age (subadult/adult) and transmission functions (density/frequency-dependent). We found that a density-dependent transmission model incorporating explicit sex groups fitted the data better than all other proposed models. Male-to-male transmission was the highest among all demographic groups, suggesting that male behaviour is a key factor driving WMHV transmission. Our models also suggest that transmission between sexes, although important, wasn’t symmetrical, with infected males playing a significant role in infecting naïve females but not vice versa. Overall this work shows the power of coupling population dynamic models with long-term field data to elucidate otherwise unobservable transmission processes in wild disease systems.


Introduction
Nonetheless, previous laboratory studies in female mice infected with murid gammaherpesvirus-4 remained low (15). Thus, if male dominance behaviour is maintaining herpesvirus infection in wild rodent populations it is likely that transmission: i) depends mostly on male density, particularly 114 reproductive males, and ii) occurs before breeding season when reproductively active males are 115 establishing territories, and the number of contacts with other reproductive males increases. Overall We fit a series of mathematical models to the serological data, each capturing different potential aspects 148 of WMHV transmission (see below). All models used the same demographic framework to describe 149 the wood mouse population dynamics, against which WHMV transmission occurred (i.e., mouse 150 population dynamics were assumed to be independent of viral infection dynamics, implying a negligible 151 impact of infection on host survival or reproduction). Hence, we first fitted a demographic model to 152 the observed mouse population dynamic data, and in doing so estimated the key demographic 153 parameters (birth rates, survival rates and carrying capacities) for our specific system.    Table 1 for a description of all parameters in all models. assuming uniform priors (Table 1A), through fitting to data on mouse population abundance and 217 infection seroprevalence respectively. We ignored the first year (52 weeks) of predicted transient 218 dynamics of the simulation as burn-in time and fit the models over the subsequent 4 years of data.
As described above, model fitting was carried out in two stages. First, demographic parameters and yearly carrying capacities were estimated by fitting the simulated total number of weekly mice (lmice;

266
Overall during the four-year field sampling, 1,394 mice were captured a total of 4,316 times, with a 267 mean number of captures per mouse of 3.04 (range 1-28). A total of 1,343 mice were identified as 268 subadults or adults and were captured 4,197 times. The remaining 51 mice were identified as juveniles.

269
The number of subadult and adult mice considered in this study were 1,065, for a total of 1,817 270 samples. Captures with missing data or that did not have blood sample taken, were not considered.  Table 1A for 95% credible intervals from the posterior distributions of demographic 285 parameters.

287
The best-fitting transmission model by DIC (Table 2)

364
Our results, which quantify natural transmission rates in the wild, confirm the importance of males for 365 driving herpesvirus infection; the highest transmission rate related to males infecting other males, 366 followed by male-to-female transmission. This dominance of males in driving transmission may arise 367 from physiological (i.e. hormonal or immune-mediated) processes, behavioural differences, or both.

368
The observed sex bias in our results and other empirical studies (15-17), may be due to infected males 369 being more infectious than females. Suppression in immune responses are known to be a product of 370 testosterone increases in reproductive males (28-30) and corticosteroids have been suggested to reduce 371 the production of antibodies (31). In terms of behavioural explanations, previous studies on rodent viral 372 infections have shown a higher prevalence in males for hantavirus (32) and cowpox (33), and it has 373 been proposed that those differences are due to higher territorial aggression and travel distances in 374 males. Thus, male dominance behaviours, such as scent-marking and biting, may constitute important 375 transmission routes via urine and saliva, respectively. Laboratory experiments highlight the nose as an important point of viral entry (34), because among inoculation routes, intranasal is highly effective for season (16) and previous studies have concluded that scent marking, at least in voles, could also convey identity and the frequency of scent marking was not always related to mate choice (36,37); scent-380 marking behaviour could be the main transmission route(15). As such we note that the relatively high 381 rates of male to female transmission that we observe does not have to be due specifically to sexual 382 contact but also scent-marking behaviour, since the latter is used to advertise current reproductive

393
To support the suggestion that transmission in these natural populations is not driven by sexual contacts,

394
we found that a density-dependent transmission function best fit the observed seroprevalence patterns.

395
This implies that frequency-dependent transmission, typically assumed for sexually transmitted 396 pathogens (41), was not driving herpesvirus infection in this population. In addition, theory predicts 397 that sexually transmitted infections would be female biased due to a higher variance in mating success 398 among males (42), therefore exclusively sexual transmission is unlikely in polyandrous species, such 399 as the wood mouse (43). However, previous research has emphasized that WMHV in wild wood mice 400 shows no strong density-dependent patterns because prevalence was not significantly influenced by

415
Regarding limitations of our study, antibody detection is not the best test for determining herpesvirus 416 infection because maternal antibodies to the virus are inherited from mother to offspring and persist 417 into young adulthood (15). Therefore, we excluded juveniles from our study; however, we acknowledge 418 that using a PCR diagnostic method could refine our WMHV prevalence results. Furthermore, based 419 on previous findings on murine gamma herpesvirus pathogenesis, our model did not assume disease-420 induced mortality due to infection (7,46). Nonetheless, previous work has found some evidence for a

505
Dots represent data points and grey shaded regions cover the period in which fieldwork was conducted.