Introduction

Virulence is generally defined as the degree to which a parasite reduces the fitness of its host [1], commonly measured as the cost in host survival. It is the consequence of complex interactions between host and parasite strategies [2]. On one hand, the host can influence virulence through its ability to clear the parasite (its resistance) or to limit its effects and associated costs (its tolerance) [3,4]. On the other hand, the parasite’s influence on virulence can be decomposed into processes that are linked to using its host’s resources and growing (exploitation) or mechanisms that damage the host independently of the parasite’s growth, such as toxins (per-parasite pathogenicity) [2,5,6].

Despite the complexity of this interplay [7], most ideas about the evolution of virulence are based on simple notions: that virulence is governed by the parasite and that its evolution is constrained by its trade-off with the rate of transmission. Thus, to maximize a parasite’s total transmission (that is, the rate of transmission times the period of infection), evolutionary pressure should favor a higher rate of transmission until the advantage of faster transmission is balanced by a shorter infection period due to greater virulence. Therefore, evolution should lead to an intermediate level of virulence [8,9]. While this and variations of this idea have been extensively discussed and studied for the last 50 years [8–10], empirical studies have given mixed support [11,12]. This is partly due to an overly simplistic description of transmission, the key parasite fitness measure and a vital trait for infection development and spread. As a parasite fitness trait, transmission is rather complex, since it encompasses stages within and among hosts, each with their own set of factors that can interact and trade-off against each other [13]. Nevertheless, several of these factors have been studied, and indeed whether transmission is vertical or horizontal [14], continuous or not [15], or whether a parasite is directly transmitted, has environmental stages [13], or is vector-borne [16] strongly affects predictions about the evolution of virulence. However, a strongly overlooked factor is the timing of transmission, that is, the time it takes for the parasite to grow, develop, and potentially evolve within its host before being transmitted.

According to the trade-off model, earlier transmission should lead to higher virulence [17], though details of the timing of transmission and pathogenicity can lead to more complex predictions. One study that measured the effect of time to transmission was performed by Cooper and colleagues on nuclear polyhedrosis virus (NPV) infections in the moth Lymantria dispar [18]. They selected the virus for early or late transmission and tracked over generations the virulence and viral growth at two-time points of the moth development. Selection for early transmission led to greater virulence but lower viral burden than selection for late transmission. While this result appears to agree with the virulence-transmission trade-off theory, it is difficult to interpret, for the experiment selected not only for early or late transmission but also inadvertently for rapid growth. This situation is made more complex by the fact that a longer time within the host before transmission might provide an opportunity for competing parasites to evolve [19,20]. Such competition is thought to be responsible for the evolution of higher virulence of the microsporidian Glugoides intestinalis when its host, Daphnia magna, is allowed to live longer [21]. Additionally, part of the parasite’s impact may be due to the phenotypic responses of the host, which are governed by trade-offs between early and late reproduction.

Lastly, most studies that aspire to understand parasite evolution tend to focus on one transmission stage, either within or among hosts, and neglect the link between the two. This raises major concerns for epidemiology and the understanding of parasite and virulence evolution [13]. As expected from life history theory, a high fitness in one environment is likely to trade-off with success in other environments. This is particularly crucial for parasites, as their fitness is dependent on the success in the first host, survival in the environment they are transmitted to (e.g., water, soil, surfaces), and the ability to infect a new host. Hence, it is important to relate the diverse environments a parasite is exposed to fully understand its fitness and evolution. Focus on merely one of them might greatly bias our predictions regarding parasite evolution and which constraints between traits might exist. For instance, a parasite with high fitness within one host might not be as well equipped to survive the outside host environment, and vice-versa. Such trade-offs have been hinted at by some authors but in poorly controlled conditions.

Taking this into account, we first aimed to test whether selecting for the parasite’s time to transmission affects its virulence. Next, in a common garden experiment, we examined if changes in virulence due to parasite evolution influenced infection dynamics and the response of non-evolved hosts. We then integrated our results into the modern virulence decomposition framework, which allowed us to separate the parasite’s impact on the host into growth-dependent (exploitation) and growth-independent (per-parasite pathogenicity) costs. Finally, we connected these findings to a parallel study with the same parasite lines, which assessed differences in parasite endurance and infection ability after exposure to various temperatures and durations outside of the host.

For this, we used the mosquito Anopheles gambiae and its microsporidian parasite Vavraia culicis, which is a common generalist parasite of mosquitoes with low costs to the host [22–25] (Stock infection leads to a mean time to death of 18 days, compared to uninfected with 24 days) and whose life cycle allows us to easily control and manipulate both time-to and mode-of transmission. With this approach, we hoped to shed light on the role and importance of considering time to transmission, as well as within and among hosts’ evolution in evolutionary and epidemiological studies. Results from this study are of universal interest for the understanding of parasite evolution and disease spread, but are also particularly relevant in vector-borne disease evolution, where most vector-transmitted parasites require long-lived hosts or vectors to complete their life cycle.

Results

In this study, we selected the microsporidian parasite V. culicis for early or late transmission (and shorter or longer time within the host) in the host Anopheles gambiae for six generations (Fig 1a). Then, we assessed changes in parasite development within the host, as well as the host response to the parasite, when infected by the early or late-selected parasite, or upon infection with the reference stock parasite that did not undergo selection. Virulence of the infection by the different parasites was also measured and decomposed into exploitation (cost dependent of parasite growth) and per-parasite pathogenicity (cost independent of parasite growth). It is noteworthy to emphasize that virulence in this study is measured as a cost in host survival.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Longevity, fecundity and virulence evolution.

(a) Experimental design, describing direct and indirect forms of parasite selection in this study. (b) Host survival after infection with evolved or reference parasites. Early and Late curves were generated with 487 and 519 females split throughout five replicates. The stock was composed of 96 females. (c) Virulence, as the natural log of the maximum hazard for each of the replicates of each selection regime (n = 5). The dashed line represents the virulence of the parasite stock. See Table A in S1 Text for a complete statistical analysis. (d) Cost of infection in fecundity, as the percentage difference in egg numbers of infected mosquitoes comparatively to the uninfected ones, for days 10 and 15 post adult emergence. The dashed line corresponds to the reference fecundity or uninfected individuals. The sample size was the following: Early with 39 and 45, Late with 56 and 50, Stock with 40 and 38 for days 10 and 15, respectively.

https://doi.org/10.1371/journal.ppat.1013294.g001

Concerning the parasite, we quantified its: i) cost in host survival (Fig 1b and 1c); ii) cost in host fecundity (Fig 1d); iii); costs in developmental traits (Fig 2); spore production rate and spore load dynamics (Fig 3); iv) compared virulence using different host harm metrics (i.e., host larval, pupal and adult stage mortality (Fig 2), and fecundity (Fig 1d) to understand the infection fully (Fig 4); and v) using host mortality as the main proxy for virulence, we decomposed the parasite contribution for this virulence metric through the quantification of cost dependent and independent of growth in host survival - i.e., host exploitation and per-parasite pathogenicity, respectively (Fig 4). Regarding the host, we measured: i) its fecundity shifts in response to the different parasites (Fig 1d); and ii) developmental time and mortality during the larval and pupal stage upon infection (Fig 2). Altogether, with this set of results, we aimed to understand the drivers of parasite and virulence evolution within the host stage and how they contribute to host infectivity.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Host developmental traits.

Selected regimes (i.e., early and late) were analyzed together with and separately from thetheirock. (a) Larvae mortality across regimes. The regimes were composed of 2068, 2011 and 2037 individuals for Early, Late and Stock. (b) Pupae mortality with each regime consisting of 1955, 1873 and 1936 in the same regime order. (c) Early pupation, as the proportion of individuals that pupated until day eight post eclosion. In the same regime order, the sample size was 1934, 1867 and 1929 individuals. (d) Body size at adult emergence using wing length of five-day-old females as a proxy. Each regime consisted of 50 individuals. Different letters denote means that they are significantly different from one another. See Table B in S1 Text for complete statistical analysis and Fig A in S1 Text for mean and standard error for each of the replicates.

https://doi.org/10.1371/journal.ppat.1013294.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Infection dynamics.

(a) Spore production rate, as the daily proportion of females with detectable spores for each treatment during the first ten days of adulthood. (b) Spore dynamics, as the mean number of spores per female for each of the first ten days, as well as days 15 and 18 (i.e., chronic load). For each regime and time-point, 50 females were assayed. See Table C in S1 Text for complete statistical analysis.

https://doi.org/10.1371/journal.ppat.1013294.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Virulence metrics and decomposition.

Virulence was assessed using different metrics of host fitness/harm: (a) larval mortality, (b) pupae mortality, (c) adult mortality, and (d) reduction in fecundity (after conversion to a proportion, and the resulting value subtracted to 1). A linear model with regime as a fixed factor and replicate as a random factor was used to test for differences due to parasite treatment for each host fitness trait. (e) Exploitation was measured as the mean spore density at days 15 and 18 (i.e., spore load divided by wing size) for every replicate of each regime. A post hoc multiple comparison test was run afterward to determine pairwise differences between parasite regimes, illustrated by letters above the plot. If letters are missing, then no parasite regime differs from each other. (f) Per parasite pathogenicity, as the relationship between virulence and spore density. (g) Infectiousness-virulence relationship for this infection model. Given that in this infection model, the spore load is expected to decelerate in the measured chronic phase, the number of spores is not expected to increase and, therefore, provide a fair estimate of the total number of spores each of the regimes can transmit. There is a positive relationship between virulence and host infectiousness. See Table D in S1 Text for further details on the statistical analysis.

https://doi.org/10.1371/journal.ppat.1013294.g004

Discussion

We demonstrated that time to transmission affects the evolution of virulence. In particular, we showed that contrary to the predictions of a theoretical model [17], a longer time within the host (i.e., late transmission) was selected for greater virulence than a shorter time (i.e., early transmission).

Furthermore, we provided contrasting evidence to a study on nuclear polyhedrosis virus (NPV) infections in the moth Lymantria dispar [18], in which selection of the virus for late transmission led to lower virulence (and higher viral burden) than selection for early transmission. Below, we argue and explain why these two contrasting findings might reflect current limitations in the study of parasite evolution. We also demonstrate that our study serves as a suitable example to reflect on classical theories of parasite evolution and how the integration of parasite life history is vital for the understanding of its evolution and, consequently, virulence itself.

The different outcomes observed in these two studies emphasize the complexity of the evolution of virulence [7,12,13]. In particular, it may result from the different levels at which selection operates on parasites: i) competition between parasites within one host; ii) the ability to survive an environment outside the host; iii) or transmission among hosts [13]. The interlinking of these, and dependent traits, can either limit or facilitate their evolvability, which in turn has major consequences for the understanding of virulence evolution. Indeed, if we only consider selection at the level of transmission among hosts, as assumed by much of the theory of the evolution of virulence [8,9] and by Osnas and Dobson (2010), selection for early transmission would lead to a higher killing rate than selection for late transmission.

However, if we instead consider sources of selection within the host, we can expect a different evolutionary outcome, which in turn aligns with our results.

Since we selected for early and late transmission by letting the parasite be transmitted from mosquitoes that died early or that lived a long time, we indirectly selected for shorter and longer periods of infection. A longer period of infection, in turn, increases the chance of mutation (an unlikely explanation for our results) or the strength of density-dependence as the parasite load increases, and thus increases the possibility of competition among parasite genotypes [26,27], in a phenomenon described as clonal interference [19]. This selects for more transmissible or more rapidly replicating parasites in our study (Figs 3 and 4) and others [19,26]. This result is experimentally supported by a recent study from Sheen and colleagues (2024), which suggested that in the absence of turnover rates in poultry markets, and longer host lifespan, virulence is no longer expected to be maintained at intermediate levels but instead continuously increases as observed with circulating parasites such as H5N1 [28]. We expect this phenomenon to happen not only in poultry markets with lower turnover but also in society and the wild, and literature has reflected this trend [27]. Longer infection gives the immune response more time to act, giving greater selective pressure for the parasite to evade immunity and, consequently, replicate more rapidly and be more virulent [29], thus hindering the fixation of genotypes with low virulence [30]. In addition, the host’s immune response should increase at higher parasite burdens [6,31]. Therefore, early-transmitted parasites, which have had little time to replicate, are subjected to less selective pressure from the host’s immune response. Late-transmitted parasites, however, grow to a higher density and are subjected to a strong immune response, which then fades with the mosquito’s age [32,33], letting the selected genotypes replicate rapidly. Overall, thus, in contrast to selection due to transmission among hosts, within-host processes are likely to select for high virulence of parasites that can replicate for a long time and that transmit late.

A good example of this rationale is provided by the evolutionary models of obligate killers [12,34–36], which assume that the host is a limited resource, making the parasite’s growth density-dependent and decoupled from its transmission. This makes it possible for “selfish” genotypes to grow without counter-selection [34,37]. In our study, although V. culicis is not an obligate killer (see Materials and Methods), the experimental design made it behave like one by restricting its transmission route, and therefore, both assumptions are likely to hold in our experiment. In particular, our protocol eliminated many costs of transmission. Thus, for example, since the spores (infective stage of our parasite) were sampled shortly after their production, transmission was independent of their long-term quality or infectivity, so any trade-off between the replication rate and quality of spores played no role in transmission [12,38,39]. Indeed, the positive correlation between chronic spore load (i.e., a proxy for infectiousness) and virulence (when results of all selection regimes and replicates are pooled) (Fig 4c) corroborates this uncontrollable growth hypothesis. Several studies and models have supported this hypothesis [5,40,41].

This hypothesis introduces us to the third level at which selection can act on: the ability to endure the environment among hosts, a transmission stage common to most parasites [13,40,42]. This resting stage is crucial for V. culicis because its transmission cycle encompasses the death of an individual on water and the release of infective spores that will be ingested by young larvae. These spores can sit and wait for a few hours to several months [43], making their survival in the external environment essential for a successful transmission. In respect to this, a follow-up study of ours with these same evolved lines quantified the parasites’ ability to endure external conditions and their resilience to different temperatures over time [44]. Consistent with life history theory, we found that spore growth within the host was negatively correlated with survival in the external environment. This meant that spores selected for late transmission, which had the highest growth rate within a host, were the most susceptible to dying in the external environment. This effect was more pronounced with longer exposure to the outside environment and at higher temperatures. In contrast, early-selected and unselected reference spores remained highly viable for the whole duration of the experiment, i.e., 90 days. A reduction in parasite viability in late-transmitted spores led to lower infectivity in new hosts and less severe infections with lower parasite burdens compared to their counterparts. Hence, we demonstrated that selection within the host can inadvertently impact parasite fitness outside of the host. This is of particular significance from a social perspective, as increased host density due to scenarios with social crowding or aggregation behaviors might shorten the time parasites spend in the external environment. This favors parasites that prioritize evolution within the host stage and which invest in host exploitation and within-host growth.

Several other results in this study are noteworthy. First, adaptation to either early or late transmission increased the parasite’s prevalence of spores (Fig 3a), growth (Fig 3b) and exploitation of the host (Fig 4a). This is most likely because the parasite switched from being maintained on two mosquito species as hosts to a parasite infecting only on An. gambiae. The parasite could thus specialize on a single host without being impeded by the cost of having to also perform on a second one [25,45,46]. However, selection did not affect per-parasite pathogenicity for any of the regimes (i.e., damage independent of parasite growth such as production of toxins, Fig 4b). This is likely due to the limited microsporidian genome [47,48] and, to our knowledge, lack of virulence factors on which selection can act.

Second, our selection protocol had an impact on some aspects of the mosquitoes’ development. For instance, infection by late-selected parasites led to a shortening of the host life cycle through early pupation (Fig 2c). This result is likely due to the properties of this parasite treatment. Indeed it has been hypothesized infection by more virulent parasites induces strong stress responses that activate higher levels of immunity and shorten the host developmental time [6,49]. This hypothesis is supported by our results, as the most virulent parasite (late-selected) shortened the host developmental time (Fig 2c), while the less virulent parasite (early-selected) did not, and instead carried a cost in pupae mortality (Fig 2b). Nevertheless, neither of these changes was big enough to impact body size at adult emergence (Fig 4d).

Third, infection decreased fecundity, in particular later in life. This impact does not appear to be linked to the density of spores, for the average number of spores did not differ between the time points when fecundity was measured and infection by the unselected parasites gave the lowest fecundity despite having low spore loads (Fig 3). Rather the pattern may reflect the host’s response to infection. Indeed, as suggested by the life-history theory [50], the shorter the lifespan of infected mosquitoes, the greater the early investment in fecundity, while late fecundity was low and similar in each selection regime.

Another important consideration of this study is the choice of the right metric of virulence. Depending on the infection setting, ecology, or host-parasite evolutionary history, it is often hard to decide which fitness or health trait to measure virulence and the harm caused by the infection to the host. A few authors have tackled this by measuring virulence across different fitness traits and calculating a somewhat cumulative virulence measure [51–53]. Here, using our host fitness measures, we did a similar comparison (Fig 4a–4d), which clearly showed that the two selection regimes selected for higher virulence in two different host metrics: while parasites selected for late-transmission invested in host mortality during the larval stage (Fig 4a) and predominantly adult stage (Fig 4c), early-selected parasites invested in sterilizing the host to their benefit, by reducing the number of offspring produced in a single clutch (Fig 4d). These findings fit nicely with the discussion points we just argued and illustrate the complexity of infection, and how focusing on the traditional metric does not always allow for a better understanding of the evolution of infection. Nevertheless, although we found differences in virulence depending on the fitness metric used, we do not believe it is appropriate to calculate a cumulative virulence measure, as we believe in a hierarchy of fitness and health traits, which are interlinked. Moreover, in addition to being interlinked, fitness traits often have case-specific impacts that depend on the environment and respective ecological constraints. This makes a cumulative measure challenging to calculate without such parameters. Nevertheless, it is evident to us that at least two host fitness traits are universally of the utmost importance for virulence evolution. Those are host survival and host reproduction, as although they are influenced by a variety of factors, they reflect the contribution to the next generation - which, in the end, is the most rigorous measure of fitness.

Altogether, our study highlights the importance of considering the whole parasite life cycle to understand its evolution. The findings and arguments presented here are intended to prompt the reader to reflect on classical theories and their applicability to modern society and disease spread strategies. We validate many of the concerns previously discussed in the literature concerning the current theory of virulence evolution and how they link to the oversimplification of parasite transmission, the main parasite fitness measure. Although usefully simplistic, the current theory of virulence disregards a considerable portion of parasite strategies, which encompass different transmission stages that evidently cannot be neglected, even when solely studying within-host aspects of infection. Importantly, we believe this study has major consequences for epidemiology, vector control, and disease spread strategies, regardless of the host-parasite model system in question. The latter is due to the nature of parasites’ transmission cycles, which, as discussed in detail by Silva et al. (2025) [13], present the same transmission stages regardless of the taxa, and therefore, evolutionary outcomes and interactions discussed here can be adapted and further explored in most parasite taxa. Moreover, one of the key take-home messages is the importance of the balance between within-host and among-host evolution. This has obvious consequences for disease evolution and spread, as it has become increasingly obvious and alarming in recent pandemics, such as the SARS-CoV2, with the rise of disproportionately high transmission associated with higher virulence in certain social settings [54–56]. Only through the understanding of this balance can we predict and design better disease control strategies. Undeniably, a continued short-sighted view of parasite evolution with within-host evolution as the sole focus of the field will hinder our understanding of parasite evolution and promulgate the controversy between virulence and transmission.

Lastly, we believe is imperative to make a small nod to the importance of these findings to the ecology of this parasite. Microsporidia, such as the one in this study, have been proposed as biological agents for malaria control [22,57] due to their ability to suppress Plasmodium infections in wild- and lab-grown mosquitoes [58,59]. In line with this, it is possible that the geographical distribution of mosquitoes (and microsporidia they carry) might influence microsporidia evolution and therefore their control potential. Surely, infection by our microsporidia is unlikely to affect mosquito population numbers, as our late-selected treatment with shorter lifespan has great earlier fecundity (Fig 1d), but the same cannot be expected regarding wild mosquito density. Although the mechanisms underpinning Plasmodium suppression are still unknown, it is fair to expect microsporidia evolution to trade off with its ability to interfere with Plasmodium development to a certain extent. Hence, this system seems a unique model to test microbe-driven vector-control mechanisms, while allowing to study of fundamental evolutionary questions regarding infection and microbe interactions.

Materials and methods

Supporting information

Acknowledgments

We thank Tiago G. Zeferino and Gwendoline Acerbi for their advice and technical support.