Infection characteristics varied in complex ways among host species and with different temperature patterns. There were differences in infection probabilities among the four caddisfly host species observed in this study which changed with differing temperature patterns. In each of the treatments, U. seonum eggs had the lowest infection probability and U. rubiconum eggs had the greatest infection probability, though relative infection rates changed across species and treatments. Infection rates in T. evansi eggs and E. turbidum changed rank-order depending on the temperature pattern to which they were exposed. There were also differences between species in each of the treatments in the CV for maximum infection. The onset of infection was not different between species or treatments but there were species differences in the CV for day of first infection of egg masses within the treatments. These differences in infection have characteristics suggestive of both stabilising and equalising coexistence mechanisms.

The differences in infection probability between species seen here are suggestive of predators as a mechanism for stable coexistence within the environment [7]. Most striking is that the numerically dominant species, U. rubiconum, has consistently highest infection rates from Saprolegnia spp. These infection rates may contribute to equalising mechanisms if infection of U. rubiconum is consistently high, due for example to physiological differences, and so reduces its overall competitive dominance. It is also possible that these infections are stabilising if they are highest in U. rubiconum simply because it is numerically dominant and would switch to other species if they became dominant. Without testing the system across changes in abundance we are unable to separate these two effects but, in either case, increased infection in the most-abundant species makes persistence of the others more likely.

Infection probability is also mediated by changes in temperature, where increased temperatures for the duration of the egg period led to higher rates of infection. Though species vary, the significant main effect of treatment indicates an overall shift to higher infection rates, particularly in the High treatment. The result of this overall increased infection could be reductions in population densities if later life stages do not compensate for increased egg mortality [34, 43] This highlights that natural temperature changes, particularly increases in heatwaves and average temperatures as predicted with climate change, could exacerbate infection risk, possibly decreasing caddisfly populations. A decrease in population densities would leave populations more susceptible to other dangers, for example, due to low genetic diversity which can cause further declines in populations from inbreeding and mutation accumulation [18].

Species- and temperature-specific infection probabilities together show patterns consistent with an ability to support coexistence by shifting relative fitness outcomes for species depending on the environment. Infection rates were highest for all species in the High treatment but, in the Low and both spike treatments, differences between species were larger (c.f. Figures 2c and 4a). Most importantly, the relative infection rates differ between treatments. Ulmerochorema rubiconum had the highest infection rate in all treatments but the difference in infection rate between U. rubiconum and other species changed depending on the treatment. The relative mortality of T. evansi and E. turbidum also varied significantly with treatment as E. turbidum had a much lower infection probability in the Late spike treatment than T. evansi and this was reversed in the Low and Early spike treatments. This species by temperature interaction in infection probability means that in a fluctuating environment, there will be times when each species is more impacted by infection and others have a reprieve. While coexistence solely supported by these shifts would require each species to have conditions under which it is the least-infected, in reality many different processes are likely contributing to coexistence [44]. The shifts seen here in relative infection rates can contribute to stabilising coexistence by favouring different species in different conditions, even if they are not alone sufficient to maintain it.

Further differences between species could also alter a species’ ability to persist within a community. Saprolegnia spp. infection leads to mortality in caddisfly eggs [33] and, if the consequential decrease of hatchling numbers is not compensated for with lower mortality in later life stages [34, 43], it will lead to decreased population size and changes to relative abundances. As infection probability is much higher in U. rubiconum, T. evansi and E. turbidum than U. seonum in all treatments there is the potential for negative population consequences for those species and a shift in competitive relationships in larval life stages. Additionally, as T. evansi and E. turbidum lay many more eggs in a single egg mass than U. rubiconum [28], the number of hatchlings that survive from an infection which covers the same proportion of an egg mass would vary among these species. For example, if an egg mass was 50% infected and all other eggs hatched successfully, the number of hatchlings would be, on average, 586 for E. turbidum, 230 for T. evansi and only 97 for U. rubiconum. The egg numbers contribute to average fitness differences, where it appears U. rubiconum has an advantage based on abundance, despite its low eggs per egg mass. However, depending on larval density-dependence, this advantage is unlikely to scale linearly, providing an opportunity for the higher egg numbers in the other species.

Although infection probability is both temperature- and species-specific, the onset of infection was not affected by the same factors, suggesting the rate of infection spread caused the differences between groups rather than the initial infection timing. As infection onset was similar but total infection of eggs in treatments and species changed across groups, this suggests that a mechanism exists for avoiding infection or slowing the progress of infection throughout egg masses and to neighbouring egg masses. One possible mechanism driving varying infection spread rates could be species-specific oviposition preferences. For example, there is evidence that some Hydrobiosidae species prefer to oviposit on rocks that do not already have other egg masses present while other species congregate egg masses [31]. In this study, while rocks with egg masses were collected from a small number of riffles so that they were exposed to similar conditions for the first few hours after laying, we still saw large variations in the locations of egg masses on rocks and amount of congregating within and between species. Such preferences could lead to differences in infection due to the degree of hyphal spread able to occur between neighbouring egg masses. Another possible mechanism for decreased severity of infection could include an immune response within egg masses or individual eggs which are able to slow the progression of infections. Temperature can lead to a decreased immune response from heat stress [45] or increased pathogen replication and transmission [46] in other species, which could account for the result that constant high temperature had the greatest infection probability in this study, if applicable here. It is possible that different species have invested differently in defences against infection such as oviposition habits or immune response. Some may put more effort and resources into fighting the spread of infection, while others might invest in faster development of eggs to neonates in the hopes of hatching before infection takes hold.

In addition to differing infection probability across species, there were also differences in the variation of infection onset, as measured by CV, c.f. Figure 3b. Although the reason for this variation is unknown, one possible cause of changes in variation of day of infection could be that infection begins in eggs that are unviable and the timing is based on when the first egg within an egg mass dies. This infection mechanism occurs in fish eggs where infection by Saprolegnia spp. starts in unfertilised or unviable eggs and then quickly spreads to nearby healthy eggs [47]. If this opportunistic infection progression occurs in caddisfly egg masses, then the differences between species in the variation for timing of infection onset may be due to differences in the probability of eggs becoming compromised or unviable and therefore, more susceptible to infection.

Evidence is mounting that infection probability of Saprolegnia spp. is altered by another environmental factor in addition to temperature. Across three similar experiments ([34], results presented here and the single species in Appendix 2), total infection varied greatly and the impact of temperature varied depending on total infection rate. For example, in contrast to the current study, Taig et al. [34] found infection of U. rubiconum in the Early spike treatment to be significantly different from the Low treatment, however, the overall infection rate was also much different (36%; compared to 62% in the current study and 9% in the single species experiment). Thus, the effect of temperature across these three experiments was greatest at moderate overall infection rates. The effect of temperature is not simply due to the boundedness of infection, as each was analysed with logistic (beta-binomial) regression on the log-odds scale. Given that the experiment which had the greatest infection probability was the one which had additional infected egg solution added to each sample, the concentration of Saprolegnia spp. in the surrounding water, or perhaps even the number of hyphae specifically surrounding each rock, may alter baseline infection rates. The overall concentration of Saprolegnia spp. in the water is likely to be a result of a number of processes acting at larger spatial and temporal scales (e.g. other reservoir species or higher water temperatures over preceding weeks or months to promote growth). A similar outcome has been observed for fish species including carp (Cyprinus carpio), roach (Rutilus rutilus) and perch (Perca fluviatilis), where greater pathological changes were observed when exposed to increased levels of Streptomyces griseus spores under experimental conditions [48]. Although the cause of the changing rate of infection for caddisfly egg masses across experiments cannot be currently determined, our observations suggest that the relationship between temperature patterns and infection probability is likely interactive with other processes.

As this study was conducted under controlled laboratory conditions, it may not fully capture the complexity and variability of field environments. This is evidenced by the changed results between repeated experiments discussed above. Additionally, low sample sizes across some species also constrained some analyses. While low sample sizes limited statistical power for certain groups, the use of mixed-effects models and cautious interpretation helped minimise the risk of overstating these results. These limitations highlight the need for future studies that replicate these experiments under more natural field conditions and with greater sample sizes to better assess population-level consequences. Further, our study cannot directly demonstrate coexistence outcomes but can provide evidence of mechanisms which support coexistence (i.e. temperature and species specificity). To prove altered likelihood of coexistence, work is needed to establish whether these infection dynamics translate into long-term stabilising or equalising processes in natural systems.

Overall, we found that infection characteristics are complex and variable across species and temperature patterns. Our findings demonstrate that the host-parasite relationships between Saprolegnia spp. and the four caddisfly species in this study are both temperature- and species-specific. The species-specific response is suggestive of the host-predator interaction being a mechanism for maintaining coexistence through competitive dominance or frequency-dependent selection. Further, as species’ infection rates respond differently to temperature patterns, environmental fluctuations might provide the necessary separation of density-dependent feedbacks to promote coexistence by favouring different species in different conditions. Increased temperatures, as from climate change, may therefore yield overall directional shifts in community composition. Understanding these processes is essential for predicting how host-parasite interactions shape biodiversity in a changing climate by altering species densities and potentially ecosystem functioning if further interactions are interrupted.