Our objectives were to observe natural transmission of SNV in P. maniculatus
populations in a semi-controlled outdoor setting, to empirically test the influence of seasonality and density on the frequency of transmission in a closed population, and to clarify the relationships between seasonality, density, host aggression, and reproductive physiology. According to theoretical models and mark-recapture data, all of these factors influence disease dynamics, but we are the first to examine these variables by experimentally manipulating host population densities across seasons in a field setting. The high-density group had many more SNV transmission events than the low-density group (11 and 2, respectively). Transmission of SNV in low-density enclosures occurred in only 1 of the 3 density experiments, and the overall transmission incidence in the high-density treatment was 2-fold higher than in the low-density treatment. Nevertheless, this difference between the density treatments was not statistically significant. Sample sizes were low and any effect of density on transmission frequency might have been obscured by the number of observed zeros. Alternatively, SNV transmission may be frequency dependent. Six years of cowpox virus dynamics in bank voles showed that transmission appeared to be density dependent during the winter, but frequency dependent in the summer–emphasizing the importance of seasonal variation in host behavior and susceptibility on disease processes 
. As we found a strong effect of seasonality in our experiments, it is possible that underlying SNV transmission processes may exhibit similar variations. A larger sample size and more iterations of the experiment over a longer range of seasons may be needed to more reliably quantify these apparent differences. Unfortunately, large-scale enclosure experiments are very time consuming and labor intensive, and require a large area and much construction material.
The incidence of SNV transmission decreased significantly as experiments were conducted later in the breeding season. This observation is consistent with previous mark-recapture studies, which have indicated strong seasonal trends in seroconversion and increased prevalence of infection during the breeding season 
. Douglass et al.
reported that the incidence of seroconversions remained relatively high but constant throughout the breeding season, while we detected a decreasing incidence from June to October. However, that study reported seroconversions detected at monthly sampling intervals in free-roaming populations across Montana; we detected transmission events weekly or biweekly and were able to assign a tighter temporal window to the events.
Initiation and cessation of the breeding season for P. maniculatus
populations are highly variable and depend on photoperiodic cues, temperature, and food availability 
. These influential factors vary geographically and annually, and may trigger differential effects among individuals in the same population 
. At our site, the proportion of adult males in breeding condition captured in the enclosures was significantly greater in experiments conducted during the mid-breeding season than in the early and late breeding season. Also, fewer scrotal males were captured outside the enclosures in the early breeding season than in the mid-breeding season, although this trend was not statistically significant (). This pattern differs from previous reports from longitudinal data in southwestern and central Montana, where the percentage of scrotal males often peaked at 80% during May or June, and decreased linearly over the course of the breeding season to approximately 2% in October 
. Our analyses included only adult males, but in the open population studied by Douglass et al.
, the proportion abdominal would have continuously increased throughout the breeding season through the recruitment of young of the year.
Studies of caged albino and wild-type house mice, free-roaming vole populations (Microtus montanus
and Microtus pennsylvanicus
), and P. maniculatus bairdii
have shown a strong and significant effect of high population densities in suppressing reproduction in both males and females 
. In all of these species, in animals living in densely populated areas, there was an increased investment in adrenocortical-related glands, but little or no gonadal development or function 
. The adrenocortical response assists in survival when individuals are faced with extreme environmental changes or physiological stress 
. Although the deermouse population density in our high-density enclosures (80 mice/ha) was similar to naturally observed high population densities in Montana, this density appears to be sufficient to affect the reproductive function of these mice. At most trapping sessions, the proportion of reproductive males in the low-density group was similar to that in the outside population. The population density of male and female mice outside the enclosures ranged from 28–46 mice/ha in August 10–28, 2008, which was similar to our low-density treatment (40 mice/ha). Although the majority of comparisons were not statistically significant between density treatments, in 2 of the 3 density experiments, the high-density enclosures consistently had lower proportions of reproductive adult males than low-density enclosures (). In Experiment 2, the percentage of adult males in breeding condition in high-density enclosures decreased from 40% to 5% during the third week of the experiment, and remained at that low level, while in the low-density enclosures, that percentage remained consistently around 30% (). Also, mice from the low-density group were in reproductive condition significantly longer than the mice from the high-density group in Experiment 2. When data were pooled across experiments 1–3, the proportion of adult males with scrotal testes was significantly lower in the high-density group than in outside mice (FET: outside vs. high-density: p
0.01, outside vs. low-density: p
0.57). This suggests that the decrease in reproductive condition was primarily a result of high population density. Although our experiment does not provide sufficient data to test such a hypothesis, we speculate that the decrease in sexual preparedness associated with high density conditions may result in decreased aggression, improved immune system function, and potentially decreased incidence of transmission. This might help explain some of the difficulty in demonstrating a clear positive relationship between population density and SNV transmission.
Independent of any treatment effects of density or season, the enclosure may have affected the length of time mice remained scrotal. When data from all 4 experiments were pooled, a significant decrease in the overall proportion of scrotal males emerged during the first trap session (Time 1: T1) after release in comparison to before they were released into the enclosures (Time Zero: T0; FET: p
0.047; data not shown). The proportion of scrotal adult males in the enclosures at T1 was also significantly lower than the proportion of scrotal adult males captured outside the enclosures (FET: p
0.003, data pooled across all 4 experiments). Additionally, in the second-longest running experiment (Exp. 3; 6 weeks), while approximately 30% of males in the high-density group and 10% of males in the low-density group had scrotal testes at T0 (data not shown), no males with scrotal testes were captured at T1 and at the next 2 trapping sessions, although breeding males were captured outside the enclosures at T1 (see ). One important factor may have been the absence of females inside the enclosures. Approximately 8–14 female mice were consistently captured outside, and most were pregnant or in breeding condition. However, despite the absence of females, 3 of the 4 experiments (except for the final fall experiment) always contained males in breeding condition, indicating the importance of seasonal cues in influencing breeding cycles. Additionally, enclosed males may have still received olfactory cues from nearby females outside the enclosure.
Although population density clearly affected the ability to maintain breeding condition, it had no statistically significant effect on the rate of weight gain. The supplementary food and water in the enclosures may have contributed to weight maintenance.
The average number of new wounds per mouse was significantly higher in the early summer than late summer and fall. As the breeding season begins, males often respond to seasonal cues and establish and defend territories 
, leading to increased wounding. The higher prevalence of wounds on males with scrotal testes supports the idea that breeding males are more likely to be aggressive and interactive than non-breeding males. At the end of the breeding season (late summer- early fall; 
), there are fewer breeding males, and, therefore, fewer fights.
The fact that incidence of transmission and average number of new wounds per mouse peaked at the beginning of the breeding season and decreased over time provides some support to the current view that direct contact may be the primary mode of transmission in wild deermouse populations, because the most transmission occurred during times where the mice were most aggressive. We cannot rule out the possibility that SNV may have been transmitted both directly and indirectly in the enclosures. Future studies could implement cameras, pit tag recorders, and fluorescent marking powder 
to gather a better understanding of the contact structure and dominance dynamics within enclosed populations and their relationship to transmission dynamics. Future manipulative experiments in enclosures will also allow testing hypotheses that environmental transmission may occur in nature.
A major limitation of our experiments was small sample size. We were able to maintain a limited number of mice per enclosure, and we observed 18 transmission events total in all 4 experiments. However, as natural transmission events are rare by nature, recording 18 events in a semi-controlled setting could be considered very successful. A previous laboratory study reported only 1 SNV transmission event out of 54 attempts 
. Nevertheless, larger experiments with greater numbers of mice per enclosure and increased numbers of replicate enclosures would have greater statistical power. Also, as we did not quarantine our susceptible mice after our experiments, we may have underestimated transmission rates. We conducted our experiments during only 2 seasons (summer and fall). To more completely understand seasonal effects on this system, subsequent studies should be run in winter and spring. Such studies may be challenging (especially in Montana) because of weather conditions and presumably decreased transmission during these seasons, although transmission during winter huddling in nest boxes could be examined. We also did not control or test for genetic variability in resistance to infection or dominance hierarchies, factors that may have influenced infection dynamics within the enclosures. Finally, in order to decrease the number of variables and keep our experiment simple and most likely to succeed, we used only male deermice. We do not know how this unnatural condition may have affected our results. Male-female mixed populations are a more natural arrangement of hosts, and therefore, to more fully understand natural SNV transmission, future experiments should also be conducted using mixed male and female populations. Comparing and contrasting the transmission and behavioral dynamics between same-sex and opposite-sex arrangements may help elucidate the relative roles of each type of interaction in disease transmission in the wild.
Our results, especially in the light of previous mark-recapture studies of effects of season and density on infection dynamics in wildlife populations, emphasize the importance of considering the strong effects of season as a confounder when making comparisons of density effects in natural populations. Seasonality, even when only evaluated within the timeframe of the breeding period (spring to autumn), may be more influential in disease dynamics than population density. Season influences host behavior, susceptibility, host reproduction, and other physiological processes, all of which are critical in maintaining disease transmission cycles in nature. Although there is a likely effect of host population density on disease transmission, density processes cannot be considered independently of seasonal factors when exploring natural host-pathogen systems.
We successfully conducted large-scale manipulative experiments that followed SNV transmission in deermice under controlled conditions. Our experiments provided further insight into the effect of seasonality and density on hantavirus transmission, reservoir host aggression, and host reproductive processes. Our successful methodologies might be used to address other questions in the field of wildlife disease ecology or in similar zoonotic host-pathogen systems.