Introduction
Parasite infection is ubiquitous, but the extent to which individuals and populations are impacted varies depending on a suite of genetic, seasonal, population-level, community-level and environmental factors (Betts et al., Reference Betts, Rafaluk and King2016). Given this complexity, it is often not clear from where invading parasites originate and how they get a foothold in a population. Parasites could invade a host population through transmission from an environmental reservoir (Decaestecker et al., Reference Decaestecker, Lefever, De Meester and Ebert2004), from other infected populations (Penczykowski et al., Reference Penczykowski, Laine and Koskella2016) or spillover from other host species in the community (Craft et al., Reference Craft, Volz, Packer and Meyers2009). Most likely, parasites come from all these (and other) pathways, but the relative contributions of transmission from a reservoir, from physically distinct host populations or across species barriers are unclear. Improving our understanding of parasite origins could help to predict outbreaks and to identify transmission barriers that could be useful in their prevention.
Each of these transmission routes has been explored in the ecological literature, but often not concurrently with other pathways. For example, parasite transmission can occur from an environmental spore bank where transmission stages remain viable without a host and later infect a non-contemporary host population (Dragon and Rennie, Reference Dragon and Rennie1995). Such transmission through time could facilitate increased diversity and coexistence of competitors as has been demonstrated in other systems with storage effects (Cáceres, Reference Cáceres1997). Therefore, if transmission occurs primarily from environmental spore banks, we expect high parasite diversity within populations and genetic structure between them if founder parasites from spore banks vary over space.
Alternatively, parasite transmission could occur from one host population to another of the same host species across a landscape. In these cases, dispersal between host populations depends on the parasite's ability to reach the new population, colonize local host genotypes, survive in a potentially different microclimate and compete with local parasite genotypes (Penczykowski et al., Reference Penczykowski, Laine and Koskella2016; Ekholm et al., Reference Ekholm, Roslin, Pulkkinen and Tack2017). If parasite populations show strong spatial structure and low diversity, this could reflect challenges for parasites to overcome these barriers.
A third source of outbreaks is the transmission of parasites between host species within a community; indeed, these events have caused some of the most devastating epidemics because parasite virulence can be maladapted in novel hosts, and parasite production in reservoir host species is decoupled from novel host density (Daszak et al., Reference Daszak, Cunningham and Hyatt2000). Most parasites can infect multiple host species (Cleaveland et al., Reference Cleaveland, Laurenson and Taylor2001), and studying transmission across host species may provide insight into factors that constrain cross-species transmission. An obvious challenge for parasites is that hosts may vary in susceptibility, competence and defences, imposing different selective forces on parasites (Gandon, Reference Gandon2004). However, phylogenetic relatedness of hosts and similarity of within-host environments may lower barriers to transmission (Streicker et al., Reference Streicker, Turmelle, Vonhof, Kuzmin, McCracken and Rupprecht2010; Longdon et al., Reference Longdon, Hadfield, Webster, Obbard and Jiggins2011; Parker et al., Reference Parker, Saunders, Bontrager, Weitz, Hendricks, Magarey, Suiter and Gilbert2015; Shaw and Kennedy, Reference Shaw and Kennedy2022). The genetic structure of parasites across host species implicates patterns of multi-host parasite transmission in the wild: structure at the level of host species indicates barriers to transmission among hosts whereas lack of structure indicates frequent transmission between host species (Archie and Ezenwa, Reference Archie and Ezenwa2011).
Importantly, outbreaks might incorporate interacting aspects of transmission from the environment (and across time), over space, among host species, as well as evolutionary change. For example, in many emerging outbreaks, pathogens spill over from co-occurring host species and then evolve to exploit the new host (Fenner and Fantini, Reference Fenner and Fantini1999; Delaney et al., Reference Delaney, Balenger, Bonneaud, Marx, Hill, Ferguson-Noel, Tsai, Rodrigo and Edwards2012). In addition, pathogens may be introduced from other populations and then evolve to exploit hosts in a new environment (Burdon and Thrall, Reference Burdon and Thrall1999; Koskella, Reference Koskella2014). Moreover, multiple mechanisms might occur at once. For example, after its introduction to the United States, West Nile virus evolved to transmit more efficiently in new mosquito vectors and also spread in mosquito and vertebrate populations across the continent (Kilpatrick, Reference Kilpatrick2011). By analysing parasite population structures, we can learn about the relative influences of these different processes, which can provide us a better understanding of how parasites transmit in complex environments.
Here, we studied the population structure of Pasteuria ramosa, a wide-spread parasite of daphniid hosts, which are important planktonic grazers in freshwater systems (Ebert, Reference Ebert2008). Pasteuria ramosa is a genotypic specialist within host species due to specific interactions between the host and parasite that govern attachment of spores to cells in the host oesophagus after spores are consumed during feeding (Duneau et al., Reference Duneau, Luijckx, Ben-Ami, Laforsch and Ebert2011; Routtu and Ebert, Reference Routtu and Ebert2015; Bento et al., Reference Bento, Routtu, Fields, Bourgeois, Du Pasquier and Ebert2017) as well as additional within-host processes (Luijckx et al., Reference Luijckx, Duneau, Andras and Ebert2014). Infection of two host species by the same parasite ‘variety’ has been reported to be a rare event (Duneau et al., Reference Duneau, Luijckx, Ben-Ami, Laforsch and Ebert2011; Luijckx et al., Reference Luijckx, Duneau, Andras and Ebert2014) though one experiment passaged P. ramosa from one host species, through a second, then back to the first (Auld et al., Reference Auld, Searle and Duffy2017). After infection, the parasite castrates its host and propagates itself within the host haemolymph (Ebert et al., Reference Ebert, Rainey, Embley and Scholz1996). Pasteuria ramosa is an obligate killer, and spores are only released from decaying host corpses (Ebert et al., Reference Ebert, Rainey, Embley and Scholz1996). These spores can remain infective for many decades in lake sediments (Decaestecker et al., Reference Decaestecker, Lefever, De Meester and Ebert2004, Reference Decaestecker, Gaba, Raeymaekers, Stoks, Van Kerckhoven, Ebert and De Meester2007). It is possible that P. ramosa varieties specializing on different hosts are in fact distinct species. However, since the research community uses ‘Pasteuria ramosa’ to refer broadly to this parasite that infects across daphniid hosts, we have chosen to do so as well, and to use the term ‘genotype’ to refer to samples with distinct molecular signatures.
We investigated whether P. ramosa outbreaks were genetically distinct among lakes, if P. ramosa genotypes commonly moved between host species and if P. ramosa genetic structure changed over time using analyses of variable number tandem repeat (VNTR) data. We predicted that populations of this parasite would be differentiated by lake since we predicted that transmission from spore banks would be common, transmission between lakes would be low and selection within lakes would be strong. Similarly, we predicted that the parasite would not move readily between host species given previous findings of host genotype specificity. Finally, we predicted that parasite genetic structure would change over time given changing host population structure and environmental factors through an epidemic season.
Materials and methods
Field sampling
We sampled eight lakes (Table 1) in southeastern Michigan every two weeks from mid-July until mid-November 2015. On each date, we combined three vertical plankton tows of the whole water column (collected using a 12 cm diameter Wisconsin net with 153 μm mesh); the three tows were collected from at least 10 m apart within the deep basin of each lake. We then used random subsamples of these tows to assess the prevalence of P. ramosa infection; all daphniids (Daphnia dentifera, Daphnia retrocurva, Daphnia parvula, Daphnia pulicaria, Daphnia ambigua, Daphnia dubia, Ceriodaphnia dubia) were counted and diagnosed for P. ramosa infection. This was performed using a dissecting microscope (25–50× magnification), counting and diagnosing at least 200 hosts of each species or, when hosts were rare, until the entire sample was processed. The earliest stages of infection are not apparent (Ebert et al., Reference Ebert, Rainey, Embley and Scholz1996), so we were only able to diagnose later-stage infections. We assessed host density using a second sample that also combined three vertical tows from the same locations. The density sample was initially preserved in 90% ethanol and later subsampled to determine host density. We multiplied host density and prevalence of infection on a given date to determine infected host density. We did not perform statistical analyses on the field data, but epidemics were visualized using ggplot2 version 3.4.0 (Wickham, Reference Wickham2016) in R version 4.2.2 (R Core Team, 2020).
Table 1. Lake names and locations
Genotyping P. ramosa
We collected infected hosts from the field samples while processing the live sample to determine prevalence of infection. These were preserved individually in 90% ethanol at −20°C until DNA extraction. A mericon bacteria plus DNA extraction kit (Qiagen, Hilden, Germany) was used to extract DNA from 103 infected animals. Briefly, we removed the preserved infected animals from ethanol and placed them in sterile microcentrifuge tubes. We then vortexed them with a battery-powered pestle in 200 μL fast lysis buffer. After the sample was emulsified, we transferred it to bead tubes. These tubes were vortexed for 10 min, then centrifuged and then the DNA-containing supernatant was removed and preserved. The extracted samples were amplified at 11 VNTR loci (Mouton and Ebert, Reference Mouton and Ebert2008; Andras and Ebert, Reference Andras and Ebert2013; Table 2) by polymerase chain reactions (PCRs). For these reactions, we used 5 μL Qiagen multiplex mastermix (QIAGEN, Hilden, Germany), 10 nm forward primer with M13(−21) tail, 400 nm reverse primer and 400 nm M13(−21) 6FAM labelled forward primer or M13(−21) HEX-labelled forward primer. The labelled primers allowed all loci to be visualized in fragment analysis (Schuelke, Reference Schuelke2000). We used 2 μL of extracted DNA in total reaction volumes of 10 μL. We used the following amplification conditions: 94°C (15 min), then 42 cycles of 94°C (30 s)/50°C (30 s)/72°C (1 min) and a final extension time at 72°C for 10 min. After PCRs, we diluted 1 μL amplified product in 199 μL molecular-grade water. We then loaded 1 μL of this diluted product into prepared capillary electrophoresis loading plates containing 11 μL Hi-Di formamide and a LIZ500 (or a ROX500) size standard (University of Michigan DNA Sequencing Core). To visualize more samples on a single plate, both HEX and 6FAM dyes were used in some cases. In these cases, we diluted each of two distinctly labelled samples (1 μL each) in 198 μL molecular-grade water and then loaded 1 μL of the diluted combination into a well in the prepared capillary electrophoresis loading plates. The University of Michigan DNA Sequencing Core performed the fragment analysis; we used GeneMapper (Thermo Fisher Scientific, Waltham MA, USA) software to read fragment lengths.
Table 2. Forward and reverse primers used to genotype each locus
Forward primers all begin with the M13(−21) sequence (first 18 base pairs), allowing binding with a fluorescently labelled (6FAM or HEX) M13 primer (Schuelke, Reference Schuelke2000).
a Reported in Mouton et al. (Reference Mouton, Nong, Preston and Ebert2007).
b Reported in Andras and Ebert (Reference Andras and Ebert2013).
Genetic analyses
We identified P. ramosa genotypes and analysed population structure with the Poppr package version 2.9.3 (Kamvar et al., Reference Kamvar, Tabima and Grünwald2014) in R version 4.2.2 (R Core Team, 2020). We excluded 1 locus (Pr17) from analyses because it was uniform across hosts. We filtered our dataset to include samples that amplified at eight or more out of 10 loci. This resulted in 93 samples that were used for the analyses (and between one and eight samples from a given species/lake/sampling date; Table 3). Multilocus genotypes (MLGs) were identified as samples that were identical at all loci (loci that did not amplify were assumed to be null alleles since several of these reactions were redone and yielded the same result). A distance matrix between the MLGs was constructed using Prevosti distance, which is the fraction of allelic differences between two samples out of all loci (Wright, Reference Wright1978). To show relationships among genotypes, we built a dendrogram from the distance matrix using the unweighted pair group method with arithmetic mean (Sokal and Michener, Reference Sokal and Michener1958). We used bootstrapping, sampling 100 times to evaluate support for tree topology. Nine samples amplified two or more alleles for at least one locus. This most likely indicates that a single host was coinfected by multiple P. ramosa genotypes. As a result, we analysed two datasets: in one, we included the alleles with the highest amplification in each sample (i.e. ignoring coinfection, but using the dominant genotype within a host) and, in the other, we included all alleles found within an animal (i.e. including all coinfecting genotypes). We conducted the same analyses on both datasets. Results of the analyses on the two datasets were qualitatively similar. Here, we report results of analyses from the dataset without coinfections.
Table 3. Sample sizes for each lake, host and date
We analysed our data with Mantel tests and partial Mantel tests with R package ‘vegan’ version 2.6.4 (Oksanen et al., Reference Oksanen, Simpson, Blanchet, Kindt, Legendre, Minchin, O'Hara, Solymos, Stevens, Szoecs, Wagner, Barbour, Bedward, Bolker, Borcard, Carvalho, Chirico, De Caceres, Durand, Evangelista, FitzJohn, Friendly, Furneaux, Hannigan, Hill, Lahti, McGlinn, Ouellette, Ribeiro Cunha, Smith, Stier, Ter Braak and Weedon2023) to understand how the Prevosti distance between every pair of parasite samples correlated with collection from the same vs different lake, the same vs different host species and number of days between collection.
We also carried out two hierarchical analyses of molecular variance (AMOVA) to quantify the extent to which parasite genotypes clustered by lake, host species and sample date (Excoffier et al., Reference Excoffier, Smouse and Quattro1992). In the first AMOVA, host species was nested within lake; in the second, lake was nested within host species. We carried out the analyses both ways because neither lake nor host species is clearly a higher hierarchical level. Sample date was the third hierarchical level in both cases. This level was included to incorporate potentially important temporal information. However, small sample sizes at each sampling date make results from this level difficult to interpret. AMOVA partitions variation in Prevosti distances between samples into the hierarchical groups (Excoffier et al., Reference Excoffier, Smouse and Quattro1992). We then randomly permuted the distance matrix 1000 times, each time calculating variance assigned to hierarchical groups to create a null distribution with which to test significance of population structure (Excoffier et al., Reference Excoffier, Smouse and Quattro1992).
Finally, we calculated the fixation index (F ST) between the groups of P. ramosa samples from D. dentifera and D. retrocurva hosts to better understand structure within host species. F ST was calculated using R package ‘hierfstat’ version 0.5.11 (Goudet, Reference Goudet2005). To test for isolation by distance we used linear models to determine if F ST/(1 − F ST) correlated with the natural log of geographic distance between lakes (kilometers) plus 1 or the amount of time between sampling dates (Rousset, Reference Rousset1997).
Results
Pasteuria ramosa infected multiple host species, but often at different times
Lakes differed in host community composition during the epidemic season (July–October; Fig. 1A). Pasteuria ramosa-infected hosts were documented in all eight lakes and in four host species, D. dentifera, D. retrocurva, D. parvula and C. dubia. Daphnia dentifera and/or D. retrocurva were the most common species in seven of the eight lakes (Fig. 1A) and were the most parasitized (Fig. 1B and C). However, these hosts tended not to be infected at the same time in a given lake (Fig. 2).
Figure 1. Lakes differed in host density and community composition (A) as well as prevalence (B) and (C) density of P. ramosa-infected hosts during the 2015 epidemic season (July–October). In (B) and (C), lines that terminate before the end of the sampling period indicate that too few hosts were counted (<20) to accurately assess prevalence of infection.
Figure 2. Pasteuria ramosa did not tend to infect D. dentifera and D. retrocurva in the same lakes at the same time. Each point is one sample date in one lake. For comparison purposes, this figure has the same scale on x and y axes. However, one datapoint had to be removed with this scaling: Cedar (10/09/2020) with 13 820 infected D. dentifera per m2 and 0 infected D. retrocurva per m2.
Genetic distance between P. ramosa samples tended to be smaller for samples from the same lake, the same host species and closer in time
In 93 P. ramosa samples, we detected between 3 and 14 alleles at each of 10 loci (mean: 9 alleles per locus). We identified 42 distinct MLGs based on the combination of alleles at the 10 loci.
There was clear structuring of P. ramosa at the level of lake and host species (Fig. 3), and both Mantel tests and AMOVA supported lake, host species and sample date as important factors explaining P. ramosa genetic structure. Moreover, Mantel tests identified significantly higher Prevosti distances between P. ramosa samples from different lakes (r = 0.396, P = 0.001, Fig. 4A) and different host species (r = 0.237, P = 0.001; Fig. 4B). Pasteuria ramosa samples were also increasingly distinct with more time between sampling dates (r = 0.168, P = 0.001; Fig. 4C). Partial Mantel tests controlling for the effect of time between sampling dates gave similar impacts of same vs different lake (r = 0.372, P = 0.001) and same vs different host species (r = 0.248, P = 0.001) on Prevosti distances.
Figure 3. In natural outbreaks, P. ramosa strains clustered by lake, host species and by sampling date. (A) Dendrogram of P. ramosa isolates coloured by lake. (B) Dendrogram of P. ramosa isolates coloured by host species (purple: D. retrocurva*, blue: D. dentifera, green: Daphnia parvula*, red: Ceriodaphnia). Samples are named with the scheme: LakeCode.SpeciesCodeSampleNumber.SampleDate. Lake codes are M, Mill Lake; CW, Crooked Lake (Waterloo); B, Bishop Lake; L, Little Appleton Lake; G, Gosling Lake; Ce, Cedar Lake; N, North Lake; W, Walsh Lake. Species codes are R, D. retrocurva; D, D. dentifera; P, D. parvula. *D. retrocurva and D. parvula are sister species (Colbourne and Hebert, Reference Colbourne and Hebert1996). Bootstrap support above 30% is shown on nodes.
Figure 4. Lake, host species and time between sample collections are correlated with Prevosti distance between P. ramosa samples. (A) Prevosti distance between samples is greater if samples were collected from different lakes or (B) different host species. (C) Prevosti distance increased as samples were collected farther apart in time.
Our first AMOVA indicated marginally significant differences between P. ramosa genotypes in different lakes (variation explained = 11.9%, P = 0.085; Table 4), significant differences between genotypes infecting different host species within lakes (variation explained = 24.0%, P = 0.025; Table 4), and significant differences in P. ramosa genetic structure over time during outbreaks (variation explained = 15.2%, P = 0.002; Table 4). The same genotypes were identified in the nearby lakes Crooked (W) and Cedar as well as in Walsh and Mill (Fig. 3A), perhaps explaining the marginal effect of lake as the highest level of hierarchy in this analysis. Our second AMOVA treated host species as the highest level of hierarchy and lake and sample date as lower levels. In this analysis, host species was a marginally significant factor explaining the variation of P. ramosa genotypes (variation explained = 7.01%, P = 0.067; Table 4); within species, lakes were a significant factor explaining variation (variation explained = 29.97%, P = 0.002; Table 4); sampling date within species/lake was also a significant factor explaining variation (variation explained = 14.88%, P = 0.001; Table 4). The lack of significance of host species on parasite genetic structure in the second AMOVA can potentially be explained by finding the same parasite genotypes infecting sister species, D. retrocurva and D. parvula, in Mill Lake (Fig. 3B).
Table 4. Hierarchical analysis of variance organizing parasite samples by two hierarchical regimes. AMOVA 1 designates lake as the highest level followed by host species and sample date. AMOVA 2 designates host species as the highest level followed by lake and sample date
a The P values are calculated by 1000 random permutations of the distance matrix (composed of Prevosti distances) between genotyped parasites. Significance is attained if the observed ϕ-statistic (and variance component) is greater or smaller than it would be by chance (Excoffier et al., Reference Excoffier, Smouse and Quattro1992).
Isolation by distance for P. ramosa from one host species but not for another
When comparing P. ramosa from one host species across lake-days, F ST within D. dentifera hosts ranged from 0 to 0.323 and from 0 to 0.217 for samples from D. retrocurva. Within D. dentifera, F ST/(1 − F ST) did not reveal a pattern of isolation by geographic distance between lakes (F = 0.86, P = 0.36; Fig. 5A) or a correlation between the amount of time between sampling and genetic divergence (F = 1.75, P = 0.20). However, within D. retrocurva, F ST/(1 − F ST) correlated with geographic distance between lakes (F = 49.2, P < 0.001; Fig. 5B), indicating isolation by distance, but not with the amount of time between sampling (F = 1.08, P = 0.32).
Figure 5. F ST/(1 − F ST) values between P. ramosa groups infecting D. dentifera and D. retrocurva showed different relationships with the geographic distance between collection lakes. (A) F ST/(1 − F ST) values between P. ramosa groups from D. dentifera hosts did not show any relationship with geographic distance between the lakes, while (B) F ST/(1 − F ST) values between P. ramosa groups from D. retrocurva hosts showed a significant positive relationship with geographic distance between the lakes.
Lake and host species as important levels of P. ramosa structuring
Overall, there is support for both lake and host species as important levels of structuring of parasite populations. However, it is unclear which level (lake or host species) is more important. Few lakes had large numbers of P. ramosa samples from multiple host species, which could be the cause of this lingering uncertainty – a result which, on its own, suggests that host species is an important driver of patterns of infection (as also suggested in Fig. 2).
Discussion
In this study, we asked how the common parasite P. ramosa navigates complex environments with multiple host species distributed across lakes and where each lake has a spore bank. Importantly, these communities differ in host species composition. Mirroring this, we found that P. ramosa outbreaks differed in size and which host species were parasitized: while multiple species were infected, infections tended to be common in only one host species in a given lake. Consistent with this, we found that P. ramosa structured by host species, especially within lakes, and that lake also defined structure, especially when host species was treated as the highest level of hierarchy. We also found that P. ramosa structured by sampling date within lakes and host species, supporting other studies that suggest that P. ramosa evolves through epidemics.
We predicted that P. ramosa populations would show structuring between lakes for two reasons. First, P. ramosa genotypes are host genotype specific due to variation in host and parasite proteins that allow for attachment of the bacterium to cells in the host oesophagus, the initial site of infection (Duneau et al., Reference Duneau, Luijckx, Ben-Ami, Laforsch and Ebert2011; Routtu and Ebert, Reference Routtu and Ebert2015; Bento et al., Reference Bento, Routtu, Fields, Bourgeois, Du Pasquier and Ebert2017). Theory predicts that host-specific parasites are more likely to be locally adapted and differentiated between populations (Barrett et al., Reference Barrett, Thrall, Burdon, Nicotra and Linde2008). Second, P. ramosa spores can survive for decades in sediments (Decaestecker et al., Reference Decaestecker, Gaba, Raeymaekers, Stoks, Van Kerckhoven, Ebert and De Meester2007), thus lakes may differ in P. ramosa standing diversity due to evolution during historical outbreaks (Andras et al., Reference Andras, Fields and Ebert2018) and stochasticity in parasite survival and sampling from the spore bank. Our prediction that P. ramosa would structure by lakes was generally supported by our Mantel tests and AMOVA analyses. It was interesting to us that the geographic distance between lakes was associated with increasing F ST/(1 − F ST) between groups within D. retrocurva but not within D. dentifera. Similar to our finding for D. dentifera, a previous study on European P. ramosa that analysed infection phenotypes of more than 50 isolates did not find correlations between infectivity and geographic distance (Fredericksen et al., Reference Fredericksen, Ameline, Krebs, Hüssy, Fields, Andras and Ebert2021). In our study, the same parasite genotypes were found in different lakes: genotypes were shared between Walsh and Mill lakes (in D. retrocurva and D. parvula) and in Cedar and Crooked (W) lakes (in D. dentifera). Walsh, Mill, Cedar and Crooked (W) lakes are all within 2.7 km of each other in the Waterloo State Recreation Area. We also found the same parasite genotypes across lakes in a different study which analysed samples collected in a different year (Shaw and Duffy, Reference Shaw and Duffy2023). It is possible that recreational use of these lakes as well as transport by waterfowl (Green and Figuerola, Reference Green and Figuerola2005) could move parasites between lakes. Alternatively, these samples could, in fact, be different from each other at loci beyond the ones used in this study. Sequencing P. ramosa samples at barcoding loci – such as the hypervariable region that has recently been found to govern infectivity (Andras et al., Reference Andras, Fields, Du Pasquier, Fredericksen and Ebert2020) – might allow for finer resolution of patterns of diversity within and across lakes.
We found that P. ramosa did not readily infect across certain species barriers but did across others. Previous research has shown that P. ramosa spores can attach to multiple host species' oesophagi (Duneau et al., Reference Duneau, Luijckx, Ben-Ami, Laforsch and Ebert2011; Luijckx et al., Reference Luijckx, Duneau, Andras and Ebert2014) despite genotype specificity of infection success within species, indicating that different host species may share resistance and susceptibility alleles. However, in previous studies, the same parasite genotype rarely infected two host species (despite attachment) indicating that additional steps in the hosts' resistance pathways operate differently in different host species (Luijckx et al., Reference Luijckx, Duneau, Andras and Ebert2014). Here, we found evidence that P. ramosa strains could move between closely related host species, D. retrocurva and D. parvula. These species are sister to each other and separated by less than 50 000 years of evolution (Costanzo and Taylor, Reference Costanzo and Taylor2010). If closely related hosts offer a more similar within-host environment or resistance pathway, this close phylogenetic relationship may explain how the same parasite genotype is able to exploit both; however, it is also clear that there can be substantial variation in interactions with parasites among genotypes of the same species (Duffy and Sivars-Becker, Reference Duffy and Sivars-Becker2007; Auld et al., Reference Auld, Edel and Little2012; Pfenning-Butterworth et al., Reference Pfenning-Butterworth, Vetter and Hite2023). In the lab, we have been successful at infecting D. parvula with spores from D. retrocurva, and we have moved P. ramosa between another closely related and hybridizing pair, D. dentifera and D. mendotae (C. D. Gowler and C. L. Shaw, unpublished data). In contrast, we did not observe the same P. ramosa genotypes in D. dentifera and D. retrocurva, and the genotypes infecting these two host species tend to differ at a number of loci. This pattern suggests that the parasite is less likely to move between these two hosts, at least under the conditions we observed. Again, more in-depth sequencing efforts would be useful to determine how distinct these parasite genotypes are and to estimate for how long they have been evolving separately.
Our study presents evidence that P. ramosa could evolve over the course of outbreaks, which supports earlier findings (Gowler et al., Reference Gowler, Essington, O'Brien, Shaw, Bilich, Clay and Duffy2022; Shaw and Duffy, Reference Shaw and Duffy2023). However, due to the small sample sizes at different sampling dates for each lake and host species in our study, these results could also be due to sampling different genotypes at different sampling dates by chance. If our results do reflect evolutionary processes, selection on parasite populations could be driven by shifts or cycling in host community structure due to parasitism (Duncan and Little, Reference Duncan and Little2007) or other factors (Hu and Tessier, Reference Hu and Tessier1995; Geedey et al., Reference Geedey, Tessier and Machledt1996). Parasite evolution could also be in response to selection from non-host associated factors (e.g. abiotic conditions; Mitchell et al., Reference Mitchell, Rogers, Little and Read2005; Vale and Little, Reference Vale and Little2009; Rogalski and Duffy, Reference Rogalski and Duffy2020). Future studies that explore how host community structure influences parasite evolution would be particularly helpful for understanding the evolution of multi-host parasites.
The patterns we documented raise several questions about disease dynamics and feedbacks in this system. We are curious about how host species composition impacts epidemic size and if it is possible to predict which host species will be more severely impacted. P. ramosa appears to parasitize the more common host, but, in some cases, epidemics occur primarily in one host species even when multiple are comparable in density (e.g. North Lake, Bishop Lake and Cedar Lake in this study). It is possible that hosts destroy spores that are not capable of infecting them since spores become activated during contact with a potential host and then subsequently have a lifespan of only about 24 h (Ebert et al., Reference Ebert, Duneau, Hall, Luijckx, Andras, Du Pasquier, Ben-Ami, Rollinson and Stothard2016). Spore destruction by non-susceptible hosts could lead to the ‘friendly competition’ phenomenon that occurs between host species with a fungal parasite in Daphnia hosts (Hall et al., Reference Hall, Becker, Simonis and Duffy2009). A recent study found that different host species can reduce the infectivity of spores in a focal host species in the lab but did not find evidence of a dilution effect in the field (Fearon et al., Reference Fearon, Gowler and Duffy2023). Nonetheless, feedbacks across years could occur if spore banks are seeded with P. ramosa spores that are infective to one host but not its main competitor. Investigating host species structure, epidemic dynamics and which hosts are predominantly infected across years would be fruitful for understanding if epidemics can have consequences for community composition and parasitism in future years.
We quantified the genetic structure of the parasite, P. ramosa, in infected hosts during natural outbreaks across lakes, host species and over time within outbreaks. We found that parasites structure among lakes, host species and sample dates indicating that there are barriers to parasite establishment between lakes and certain host species in this system. However, transmission between lakes may occur due to transport of spores by wildlife or humans, and transmission between host species appears to occur when host species are closely related phylogenetically. We also found changes in parasite structure over time, supporting other studies that have documented evolution within outbreaks. Thus, analyses of parasite genetic structure can implicate ecological and evolutionary forces acting on parasites. Additional studies documenting parasite population structure across natural host–parasite systems could yield patterns that help predict parasite outbreaks in complex environments.
Data availability statement
The data and R code used for analyses presented here are available at https://github.com/clarashaw/Pasteuria2015/tree/main.
Acknowledgements
We thank M. Raquel Marchan Rivadeneira and Tim James for assistance through hurdles we faced during genotyping, and we thank Alliyah Lusuegro for her contributions to DNA extraction and genotyping. We also thank Tim James, Mark Hunter, Aaron King, Ed Ionides, Marcin Dziuba, Gideon Bradburd and 2 anonymous reviewers for their feedback on earlier drafts of this manuscript.
Author contributions
C. S. and M. A. D. conceived and designed the study. C. S. and R. B. gathered samples and performed molecular work. C. S. completed the data analysis. C. S. wrote the first draft of the manuscript. C. S. and M. A. D. edited the manuscript.
Financial support
This work was supported by the National Science Foundation (DEB-1305836), the Gordon and Betty Moore Foundation (GBMF9202 DOI: https://doi.org/10.37807/GBMF9202) and the University of Michigan EEB Block Grant.
Competing interests
None.
Open access
