Parasites, by definition, exploit their hosts for survival and gain fitness by transmitting to new hosts (Poulin, 2007). To maximize their success, parasites possess many traits that have been selected specifically for infecting hosts, including sophisticated life history strategies to help them navigate the within-host environment (Thomas et al., 2002; Reece et al., 2009; Mideo and Reece, 2012). For example, parasites are capable of adjusting their growth and developmental schedules to mitigate the effect of drug treatment (Buckling et al., 1997; Paterson and Barber, 2007), better compete for limited host resources (Mideo, 2009; Leggett et al., 2013), and to take advantage of sparse transmission opportunities (Poulin, 2003; Crossan et al., 2007; Isaïa et al., 2020). Parasites with complex development that requires sequentially infecting more than one species, such as a vertebrate host and an arthropod vector, face the extra challenge of having to move through multiple environments that are likely to pose different selective pressures at various points of their life cycle. If parasite traits and strategies are genetically or physiologically correlated across life stages, then the selective environment in one host may constrain the expression of phenotypes in another (Gandon, 2004; Gilchrist and Coombs, 2006). This means that inferences about parasite life history evolution, and the forces that shape it, will be incomplete if only parts of their life cycle are considered at a time. As parasite traits influence disease transmission and severity, successful disease control relies on understanding how parasite life history evolution is shaped by processes operating across scales and environments (Reece et al., 2009; Williams, 2010; Mideo et al., 2013; Restif and Graham, 2015).

The malaria parasite (Plasmodium spp.) has a complex life cycle, spending part of its development in a vertebrate host and the other part in an arthropod vector. In the host, Plasmodium utilizes red blood cells (RBCs) as the main resource for growth and development. Upon invading an RBC, a Plasmodium parasite may enter the proliferation cycle where it replicates asexually, eventually bursting out of the RBC to release multiple progeny parasites (merozoites). Alternatively, Plasmodium merozoites may be committed to develop into gametocytes in the next infected RBC, which can be picked up by a mosquito vector during a bloodmeal. Inside the vector, gametocytes differentiate into microgametocytes and macrogametocytes (males and females, respectively), which fuse to form a diploid zygote that undergoes further developmental steps before the parasite is again infectious to its vertebrate host. The proportion of infected RBCs in a given cohort that produce gametocytes is termed the “conversion rate” (Bruce et al., 1990). From the perspective of the parasite, this is one of the most important life history “decisions” to make, representing a growth (i.e., asexual replication) versus reproduction (i.e., transmission) trade-off akin to those faced by free-living organisms (Stearns, 1992; Reece et al., 2009). On the one hand, there are clear benefits to producing many gametocytes: transmission to new hosts is a standard proxy for parasite fitness (Anderson and May, 1982; Ewald, 1983) and high gametocyte density within a host increases transmission success to mosquitoes (albeit in a saturating manner; Paul et al., 2003; Huijben et al., 2010; Bell et al., 2012). On the other hand, replication is also important to parasite success because building up to high within-host densities can help parasites overcome innate host immune defenses (Metcalf et al., 2011) and may extend the duration of infections (Klein et al., 2014). One might therefore intuitively expect some intermediate level of investment in transmission. However, Plasmodium parasites tend to make few gametocytes (relative to merozoites) and have low conversion rates – in natural transmission settings, conversion rates have been estimated at only a few percent (Stewart et al., 2022) – a phenomenon known as reproductive restraint (Taylor and Read, 1997).

A number of hypotheses have been put forth to explain the evolutionary puzzle of reproductive restraint in Plasmodium (e.g., Taylor and Read, 1997). Given that conversion rate is a parasite trait expressed in a vertebrate host, theory has largely interrogated the evolutionary influence of processes acting within the host (e.g., McKenzie and Bossert, 1998; Mideo and Day, 2008; Greischar et al., 2016; Birget et al., 2018). For example, multi-genotype infections can generate selection for low investment in transmission relative to replication due to within-host resource competition (Mideo and Day, 2008) or immune-mediated apparent competition (McKenzie and Bossert, 1998), results that have been supported by experimental data (Pollitt et al., 2011). The evolutionary influence of within-vector processes on transmission investment has received comparatively less attention. It has long been hypothesized that parasites exhibit reproductive restraint because high parasite density in the vector results in increased mortality (virulence), which decreases the probability of the parasite infecting a new host (Taylor and Read, 1997; Koella, 1999). While this is supported by some empirical evidence (Ferguson et al., 2003; Dawes et al., 2009; Pollitt et al., 2013), conditions under which these effects would become apparent and relevant (e.g., parasite load, parasite-vector species pair) remain open questions (Ferguson and Read, 2002). Furthermore, theory suggests that this effect on its own cannot plausibly generate selection for low transmission investment; instead, it generates selection for low gametocyte numbers, which can be achieved via a multitude of investment and growth strategies (Mideo and Day, 2008). Focusing solely on vector mortality also neglects the complicated series of developmental steps that parasites must undergo within the vector.

The developmental success of Plasmodium within a vector, following ingestion of gametocytes from an infected vertebrate, is highly density-dependent and non-linear (Baton and Ranford-Cartwright, 2005; Sinden et al., 2007). When the density of gametocytes circulating in the host is low, the starting inoculum in a vector will likely also be low (Sinden et al., 2007; Bell et al., 2012). Importantly, this reduces the odds of a vector picking up both microgametocytes and macrogametocytes during its bloodmeal, hampering the parasite’s fertilization success and subsequent ookinete production (Sinden et al., 2007; Bell et al., 2012). Low ookinete density in turn reduces the success of invading the vector’s midgut epithelium (Sinden et al., 2007). Ookinetes that manage to establish in the midgut will incubate as oocysts which subsequently burst to release sporozoites, which migrate to the salivary glands of the vector from which they may have an opportunity to infect a host when the vector takes another bloodmeal (Baton and Ranford-Cartwright, 2005). Initially, the production of sporozoites increases linearly with oocyst density, however, at higher densities, “crowding” results in fewer sporozoites ultimately making it to the salivary glands (Pollitt et al., 2013). These conflicting effects of load on parasite developmental success in the vector, coupled with potentially increased vector mortality resulting from high parasite load (Ferguson et al., 2003; Dawes et al., 2009; Pollitt et al., 2011), point to additional trade-offs faced by the parasite in allocating resources towards transmission stages, and make it non-intuitive to predict how processes within the vector will influence parasite life history decisions made within the host. Moreover, Plasmodium transmission potential (a proxy for parasite fitness) is typically calculated using infectivity functions that link within-host gametocyte density to the probability of infecting a vector, usually assessed from the proportion of exposed mosquitoes that survive and subsequently develop oocysts (Paul et al., 2003; Huijben et al., 2010; Bell et al., 2012). These infectivity functions increase with gametocyte abundance, implying that more gametocytes entering a mosquito means more fitness for parasites and, in particular, more parasites successfully exiting the vector – a strong assumption given the density-dependent and non-linear trajectory of parasite development within the vector. Failure to account for these processes may therefore impede accurate estimation of parasite transmission potential and resulting selection on parasite life history.

In this study we sought to understand whether explicit consideration of parasite within-vector life stages alters the resolution to the growth-reproduction trade-off experienced by malaria parasites within the host, thus altering the optimal conversion rate. We developed a multiscale model of rodent malaria to achieve this. At the within-host level, we capitalized on a previously-published model of within-host parasite development and infection dynamics (Greischar et al., 2016; Figure 1 ). To capture processes within the vector, we propagated the gametocytes produced from the within-host model through a series of empirically-derived functions that quantitatively capture parasite growth in the vector (Sinden et al., 2007; Dawes et al., 2009; Churcher et al., 2010; Pollitt et al., 2013; Figure 2 ). This integration of within-host and within-vector models allowed us to investigate the joint effect of multiple density-dependencies and trade-offs in malaria parasite development and determine whether constraints of life inside a vector can influence the evolution of parasite traits that are expressed within a host. It also allowed us to refine estimates of parasite transmission potential by incorporating processes that occur in the vector and more holistically define parasite fitness as transmission to new hosts after passage through mosquito.

In this study, we investigated whether parasite development in the vector influences the evolution of transmission investment (conversion rate), a key life history trait malaria parasites express in the host (Taylor and Read, 1997; Reece et al., 2009). We considered two aspects of parasite-vector interactions that may influence this trait: non-linear and non-monotonic density-dependent developmental transitions (Sinden et al., 2007), and parasite-induced vector mortality (Dawes et al., 2009; Churcher et al., 2010). These represent conflicting selective pressures as both high and low gametocyte densities are likely detrimental to parasite fitness. While infection of the vector fails at low inoculum (Bell et al., 2012), high parasite density hampers sporozoite production (Pollitt et al., 2013) and increases vector mortality (Dawes et al., 2009; Churcher et al., 2010), limiting the number of new infections generated. To assess the combined effects of these processes on parasite life history evolution, we constructed a multiscale model including a stochastic simulation model of parasite within-vector developmental transitions. Parasite fitness was defined as the number of new infections achieved in hosts (with or without parasite-induced mortality to the vector), an advance over previous proxies of parasite fitness which assume infecting the vector is directly proportional to parasite fitness (Greischar et al., 2016; Birget et al., 2018). This approach also allowed us to derive new infectivity functions describing the probability of infecting a new host given a particular within-host gametocyte density. By deriving these new infectivity functions (specifically, C and D in Table 3 ), we have provided a ready tool for translating within-host dynamics into the probability of infecting new hosts, whether those within-host dynamics are model-simulated or experimentally observed, and without having to explicitly model within-vector processes. Unlike previous infectivity functions (Bell et al., 2012), ours completes the transmission chain, from host to vector back to host, and therefore provides a more holistic fitness metric for studying the evolution of other parasite traits and the evolutionary influence of particular environmental factors.

Our results revealed that within-vector development had little effect on the optimal parasite conversion rate. Regardless of how parasite fitness was defined and even with parasite-induced mortality increased up to 50% over empirical estimates, the optimal conversion rate was estimated to be around 42–44% ( Table 3 ; Table A1 in Appendix A ). These estimates are high compared with empirical observations (Taylor and Read, 1997) likely due to the lack of within-host competition and immunity in this model, known determinants of reproductive restraint in Plasmodium parasites (McKenzie and Bossert, 1998; Mideo and Day, 2008; Pollitt et al., 2011; Greischar et al., 2016). We note that while adding these complexities would likely decrease our baseline optimal conversion rate, they are unlikely to qualitatively alter the effects of within-vector processes that we have investigated here (except in some specific cases, e.g., transmission-blocking immunity). Despite the optimal strategy being largely unchanged when within-vector processes are considered, the transmission potential of the parasite decreased substantially following passage through the vector. While this drop was expected – it cannot be more likely to leave the vector than it is to enter – it was surprising within-vector processes did not impact optimal conversion rates. One explanation could be that our simulated parasite populations never generated high enough oocyst numbers to hinder sporozoite production (maximum oocyst density in our simulation model was 115; Appendix C ). Thus, the density-dependent developmental transitions, and conflicting selective pressures, that precluded easy a priori predictions about trait evolution did not end up influencing conversion rate in our simulations. High oocyst loads close to the level that generated the crowding phenomenon in lab settings (i.e., 150–200) have been observed in natural populations, but are very rare (Bompard et al., 2020). Most wild-caught mosquitoes carry 1–10 oocysts, with the median being greater than 5 (Pringle, 1966; Paul et al., 2003; Bompard et al., 2020). Understanding the factors that drive this right-skewed oocyst load distribution, especially the conditions that give rise to the long tail, would be important for studying parasite-vector interactions and parasite evolution in natural settings.

We found that the effect of increasing parasite load on vector mortality was small relative to gains in transmission potential. The near-flat slope of the trade-off curve (except at very low densities; Figure 7 ) further suggests that natural, not parasite-induced, mortality is the more significant source of mortality for vectors. This was also apparent from our sensitivity analysis of increasing parasite-induced mortality beyond empirical observations, which still did not alter the optimal parasite conversion strategy. Our findings therefore provide further evidence against the long-held hypothesis that Plasmodium parasites suppress transmission investment to reduce harm to their vector (reviewed in Taylor and Read, 1997; see also Mideo and Day, 2008), since the effect of parasite load on vector survival appears to be weak beyond an inevitable cost. In contrast, the huge gains in transmission potential may incentivize production of gametocytes to achieve high transmission probability after paying the price of vector mortality. This corroborates the modeling study of Churcher et al. (2010), which finds that high gametocyte density should be favored: even though increasing gametocyte density results in fewer vectors being alive to take bloodmeals, each bloodmeal is more infectious due to the high sporozoite load at the salivary gland, therefore resulting in high number of total new infections generated.

Other details of parasite-vector interactions may impact parasite life history evolution and could provide fruitful avenues for future work. For instance, the extrinsic incubation period (EIP, the time it takes the parasite to develop in the vector, from inoculation to becoming infectious), is a non-monotonic function of ookinete load with intermediate numbers (200–2,000) resulting in the shortest EIP (Childs and Prosper, 2020). Short EIPs are favored due to the reduced cumulative risk of vector mortality, however, how high ookinete numbers could translate to high sporozoite load (and therefore per-bite transmissibility). How this non-monotonic and density-dependent relationship between vector survival and parasite transmission may influence optimal conversion rate, in the context of other biological complexities we considered here, is unclear. Furthermore, EIP is known to be influenced by various biotic and abiotic factors such as temperature and vector condition (Paaijmans et al., 2012; Hien et al., 2016; Ohm et al., 2018). It could therefore be useful to investigate how these factors and parasite within-vector growth and development interact to influence parasite life-history evolution.

Our model assumed that the “reproductive value” of gametocytes is constant over time, meaning that a given within-host density of gametocytes always gives rise to the probability of transmission predicted by our infectivity function, regardless of which particular day those gametocytes are produced. This assumption will not hold if gametocyte infectivity declines over the course of an infection (e.g., due to a rise in transmission-blocking immune responses; Rosenberg, 2008) or if transmission opportunities are not uniformly distributed over the course of an infection. In natural settings, mosquitoes tend to bite at specific times of the day (Gatton et al., 2013) and their population sizes vary seasonally (e.g., Mordecai et al., 2019). Relaxing our model assumption of constant blood feeding at regular intervals to incorporate patterns in daily blood feeding rhythms should have little effect on our results, as the within-day variation in parasite density in our model is small compared to differences between days. Mosquito population dynamics, on the other hand, may have a stronger influence – previous work has shown that higher transmission investment is favored when the mosquito population is expanding (e.g., at the beginning of the rainy season), and vice versa (Greischar et al., 2019). Though we found no influence of parasite-induced vector mortality in our study, this could change if mortality affects the recruitment of adult mosquitoes (Beck-Johnson et al., 2013). Similarly, predictions could be altered if within-vector parasite load is associated with changes to the timing and frequency of vector bloodmeals (Schwartz and Koella, 2001; Handel and Rohani, 2015).

Our within-vector model was parameterized for the An. stephensi and P. berghei vector-parasite species pair. Given that parasite and vector species identities are known determinants of their interaction dynamics and outcome (e.g., Vaughan, 2007; Poudel et al., 2008), our prediction of no effect of within-vector dynamics may not hold for other species pairs, especially when extrapolating to those that cause human malaria in natural settings. In general, human malaria parasites tend to have higher transition efficiencies, particularly for the ookinete to oocyst and oocyst to sporozoite transitions, resulting in higher gland sporozoite loads compared with rodent malaria species (Vaughan, 2007). This suggests that the per-bite transmission probability could saturate at lower starting inoculum, alleviating the selection pressure for transmissibility and in turn, increasing the relative impact of parasite-induced vector mortality on parasite transmission potential. In this case, we speculate that the optimal conversion rate could be lower than what is predicted here. Moreover, malaria parasites in natural settings are likely to encounter many different vectors species in which they experience different developmental dynamics (Vaughan et al., 1994) and EIP (Ohm et al., 2018), resulting in different probabilities of the mosquito surviving long enough to transmit even if all else is equal. This raises the interesting possibility that parasites in diverse communities may have to be plastic or tune their strategies to the particularities of each community (Sibly, 1996). In different parts of the world where the diversity and seasonality of vector communities are drastically different from one another (Carter and Mendis, 2002; Baird, 2017; Hoi et al., 2022), vector diversity, especially that of seasonally varying community composition, may be an important driver of parasite local adaption.

Understanding parasite life history evolution is important because these traits, like conversion rate for Plasmodium, underlie key clinical and epidemiological outcomes (i.e., infection severity and infectiousness). Given that these same traits may help parasites evade our efforts to control them (Reece et al., 2010; Birget et al., 2018; Schneider et al., 2018) elucidating the constraints on parasite trait evolution may have applied relevance. For parasites that move between a vertebrate host and an invertebrate vector, the evolution of traits expressed during part of that life cycle may be constrained by processes and trade-offs that are hidden in another part. Though we find that life in a vector exerted little influence on the evolution of malaria parasite conversion rates expressed within a host, this is unlikely to be true for all traits and all parasite-vector combinations. Indeed, Spence et al. (2013) find that life in a vector can influence parasite antigenic expression and subsequent virulence in a host. Thus, further studies into parasite-vector interactions and vector ecology as potential evolutionary drivers of key parasite traits should remain a research priority.

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

AH: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing. MG: Conceptualization, Data curation, Investigation, Methodology, Software, Supervision, Writing – review & editing. NM: Conceptualization, Funding acquisition, Investigation, Methodology, Resources, Supervision, Writing – review & editing.