Relative migration rates and local adaptation in a mosquito-protozoan interaction

Authors


Holly H. Ganz, Department of Environmental Science, Policy and Management, University of California, 137 Mulford Hall, Berkeley, CA 94720-3112, USA.
Tel.: 510-643-1227; fax: 510-643-5438;
e-mail: hganz@nature.berkeley.edu

Abstract

Theory predicts that the direction of local adaptation depends on the relative migration rates of hosts and parasites. Here we measured relative migration rates and tested for local adaptation in the interaction between a tree hole mosquito (Ochlerotatus sierrensis) and a protozoan parasite (Lambornella clarki). We found strong support for the hypothesis that the host migrates more than its parasite. Hosts colonized artificial tree holes in the field at a much higher rate than the parasite. Field releases of the parasite demonstrated that it colonizes and persists in natural tree holes where it was previously absent, suggesting that parasite distribution is limited by its migratory ability. Although the host migrates more than its parasite, we found no evidence for local adaptation by hosts and some evidence for local adaptation by parasites. Other life history traits of the host and parasite may also influence patterns in local adaptation, particularly parasite virulence and host dormancy.

Introduction

The geographic mosaic theory of coevolution describes the patchwork pattern observed in traits involved in species interactions across a broad geographic range (Thompson, 1994,1999). Such geographic patterns arise from coevolutionary processes between interacting species composed of structured populations that experience local differences in gene flow and selection (Thompson, 1994,1999). Coevolutionary processes produce spatial genetic differentiation of host and parasite populations, often resulting in parasites that exhibit greater fitness on sympatric host populations compared with allopatric host populations (Parker, 1985; Lively, 1989; Gandon & Van Zandt, 1998; review: Kaltz & Shykoff, 1998; Lively et al., 2004). Local adaptation in parasites is usually defined as arising when a parasite population exhibits higher mean fitness on sympatric hosts compared with allopatric hosts. However, an alternative definition of local adaptation is when the parasite is more fit in its local environment than foreign parasites (Ebert et al., 1998; Kawecki & Ebert, 2004). A decrease in infectivity with increasing geographic distance is consistent with local adaptation by the parasite under the first definition, assuming that distance correlates with the genetic relatedness among sites (Ebert, 1994).

Local adaptation of parasites to their hosts following the former definition (i.e. higher fitness in local hosts compared with foreign hosts) is commonly observed in host-parasite interactions, including plants and herbivores, plants and pathogens, invertebrates and microparasites and vertebrates and parasites (review in Kaltz & Shykoff, 1998). Though, whether the parasite or the host exhibits local adaptation depends on where the system is in the coevolutionary process (Gandon et al., 1996; Morand et al., 1996; Lively, 1999). Hosts may be temporally more resistant to sympatric parasites such that parasites are more infectious in novel hosts than local hosts. For example, Kaltz et al. (1999) found that a fungal pathogen of the plant, Silene latifolia, was more infective of foreign hosts than local hosts. Nevertheless, a number of studies have found no evidence for local adaptation by either host or parasite (Kaltz & Shykoff, 1998; Mutikainen et al., 2000). This variation in the outcome of local adaptation studies may be explained by differences between the host and parasite in migration rates, evolutionary potentials and parasite virulence (Gandon et al., 1996,1998; Lively, 1999; Gandon & Michalakis, 2002).

Despite the emphasis on relative migration rates in theory on local adaptation in host-parasite interactions, few empirical studies have attempted to measure relative rates of migration and tested for local adaptation. This study was designed to characterize the relative migration rates of a host and parasite and to test for evidence of local adaptation. We studied the interaction between a tree hole mosquito, Ochlerotatus sierrensis Ludlow (Insecta: Diptera: Culicidae) and a protozoan parasite, Lambornella clarki Corliss and Coats (Ciliophora: Hymenostomida: Tetrahymenina). These two species co-occur in ephemeral water-filled cavities in trees (tree holes) in forested habitats of Western North America. Larvae and protozoa hatch from drought-resistant stages when tree holes are inundated at the onset of the rainy season, which typically starts in November and continues into April (US, NOAA, 1993). Lambornella clarki occurs commonly in water-filled tree holes where O. sierrensis larvae develop (Corliss & Coats, 1976; Washburn et al., 1988), and it is the most important natural enemy of immature O. sierrensis (Washburn & Anderson, 1986; Washburn et al., 1991). In tree holes lacking mosquitoes, L. clarki is found in a free-living, nonparasitic form (the trophont), but it will transform into a parasitic form, the theront, in response to chemical cues excreted by mosquito larvae (Washburn et al., 1988). The parasitic form of L. clarki invades the host by attaching to the larval cuticle, producing a cuticular cyst, and subsequently forming a hole in the cuticle by which it enters the insect's haemocoel (Washburn et al., 1988). The host, O. sierrensis can defend against attack by L. clarki by encapsulating and melanizing invading parasites in the haemocoel (Corliss & Coats, 1976; Washburn et al., 1988). In addition, O. sierrensis reduces the risk of infection by preying upon free-living L. clarki (Washburn et al., 1988). Lambornella clarki exhibits high virulence within O. sierrensis; parasitized hosts usually die within 4 weeks in the field (Egerter et al., 1986; Washburn et al., 1988,1991). Hosts infected in later larval stages may survive infection to emerge as infected adults (Egerter et al., 1986; Washburn et al., 1988). Infected hosts of both sexes are parasitically castrated; L. clarki invades the ovaries of infected females shortly after eclosion (Egerter et al., 1986; Washburn et al., 1988,1991). Subsequently, blood-seeking behaviour is blocked, and infected females assume oviposition behaviour and deposit parasites instead of eggs in neighbouring tree hole habitats (Egerter et al., 1986; Washburn et al., 1988,1991).

Some parasites have the ability to modify the behaviour of their hosts to increase their dispersal to other susceptible hosts (Moore, 2002). Lambornella clarki clearly manipulates the behaviour of its host because infected females seek out oviposition sites rather than bloodmeals (Egerter & Anderson, 1989); and, while at an oviposition site, infected O. sierrensis (hosts) spend less time engaged in behaviours, such as grooming, that are not beneficial to the parasite (Yee & Anderson, 1995). The natural history of L. clarki indicates that this parasite relies on its host for dispersal and empirical observations suggest that uninfected hosts are capable of greater rates of movement than infected hosts. Infected hosts also do not live as long as uninfected hosts (Egerter et al., 1986), and infected female hosts exhibit shorter flight duration compared to uninfected female hosts (Yee & Anderson, 1995) in the laboratory. In the field, it is common to observe tree holes with O. sierrensis (hosts) and not L. clarki (parasites), while it is rare to find tree holes with L. clarki and no O. sierrensis (Washburn & Anderson, 1986; Yee & Anderson, 1995; Holly H. Ganz, unpublished data).

For this study, we hypothesized that the parasite, L. clarki exhibits lower rates of migration than its host. We present field data that support this hypothesis, and we used these results to develop hypotheses regarding local adaptation in the L. clarkiO. sierrensis interaction. Theory predicts that local adaptation by parasites arises when a parasite migrates at rates greater than or equivalent to that of its host (Ladle et al., 1993; Judson, 1995; Gandon et al., 1996,1998). We tested for local adaptation by performing a cross-infection study between two host and parasite populations. In addition, we tested infectivity and performance (parasite load) by these two parasite populations in two novel host populations. In a second experiment, we tested the infectivity of a single parasite population in host populations that were collected at varying geographic distances from the parasite source population. Although not as powerful an approach as cross-infection studies, this geographic comparison provides a second measure of local adaptation assuming that geographic distance correlates with genetic distance from the source population (Parker, 1985; Ebert, 1994; Morand et al., 1996).

Materials and methods

Relative migration rates

Migration rates of the host and parasite were characterized by measuring colonization of artificial tree hole habitats in the field, the distribution of the host and parasite in the field, and persistence of the parasite when introduced into natural tree hole habitats.

Colonization of artificial tree holes in the field

Studies of colonization of artificial tree holes were conducted at three sites in Northern California: the University of California Hopland Research and Extension Center (Hopland), the University of California Sierra Foothill Research and Extension Center (Sierra Foothill), and a site in Marin County (Marin). Both host and parasite occur in approximately 50% of tree holes surveyed at Hopland, 25% of tree holes at Sierra Foothill and around 50% of tree holes at Marin (unpublished data; Washburn & Anderson, 1986). Artificial tree holes (ovitraps) were placed in areas known to support both O. sierrensis and L. clarki, where they were filled with filtered tree hole water and lined with paper towels, following the methods of Woodward et al. (1998). At Hopland, eight artificial tree holes were placed in each of four transects and monitored for 18 weeks (April–August 1990) and a total of 576 samples were obtained. At Sierra Foothill, six artificial tree holes were placed in each of nine sites and monitored for 5 weeks during peak mosquito emergence (May–June 1999) and 252 samples were obtained. At Marin, 25 artificial tree holes were monitored for 16 weeks (April–August 1990) and 400 samples were obtained. Presence of O. sierrensis eggs and/or L. clarki was recorded and artificial tree holes were replaced on a weekly basis. Water samples were filtered in the laboratory and examined for the presence of L. clarki under a dissecting microscope (10–40 times).

Field surveys for presence or absence of the host and parasite in natural tree holes

Thirteen locations in California were surveyed 2–3 times a year for 2–5 years (1984–1989) to establish the presence or absence of the host and parasite in natural tree holes. The field sites were distributed across California (Fig. 1 in Washburn & Anderson, 1986). Collections were made at least twice a year (fall and spring) with the timing depending on yearly rainfall patterns. Distribution of L. clarki in the field was determined by: (1) L. clarki found infecting O. sierrensis larvae in tree holes, (2) L. clarki trophonts observed in tree hole water and (3) L. clarki infections obtained after exposing freshly hatched first-instar larvae in the laboratory to tree hole water samples from the field. Suitability of tree hole water to support L. clarki was determined using water samples collected from the field. The parasite was placed in these water samples and these cultures were monitored for parasite growth and survivorship for 1 week.

Figure 1.

California populations included in this study: UC Hopland Research and Extension Center (Hopland; 39°1′15′′N, 117°4′31′′W), UC Berkeley Blodgett Forest Field Station (Blodgett; 38°54′43′′N, 120°39′53′′W), Sonoma (38°26′35′′N, 116°30′14′′W), Big Sur (36°15′1′′N, 115°46′32′′W), Fresno (36°45′33.5′′N, 119°48′46′′W), Santa Barbara (34°27′17′′N, 119°42′33′′W), LA-1 (34°8′14′′N, 118°17′4′′W), LA-2 (34°3′43′′N, 118°36′45′′W) and Julian (33°2′55′′N, 116°35′15′′W). Coastal range populations are indicated with a circle and populations from the Central Valley and foothills of the Sierra Nevada are indicated with a triangle.

Addition of the parasite to natural tree holes

After determining that L. clarki was absent in prior field surveys (during the preceding 2–5 years), the parasite was added to 72 natural tree holes in Mendocino, Marin, Fresno, Orange, Santa Barbara, Los Angeles and San Diego counties. Twice a year, in spring and fall, larval populations were sampled to evaluate if the parasite became established and whether it persisted in subsequent years. Larvae were inspected for infections under a dissecting microscope (10–40 times) and water samples were filtered and examined for the presence of L. clarki.

Local adaptation

In 2001 and 2002, we collected samples of nine different wild populations of O. sierrensis (tree hole mosquito) that spanned a broad latitudinal gradient in California (Fig. 1). For each population in this study, we collected more than 500 larvae (when possible) from a single tree hole. The Hopland population was sampled more intensively in order to characterize variation in host susceptibility within a population. Specifically, at the Hopland population, we collected larvae from eight sites that were widely distributed across the field station. Mosquito larvae were reared to adulthood in an environmental growth chamber at 15 °C with 14 : 10 (L : D) photoperiod in order to minimize fourth instar larval diapause (Jordan & Bradshaw, 1978). Emerging adults were used to establish laboratory colonies, which were maintained separately based on their tree hole of origin. Colonies were maintained in an environmental growth chamber at 21 °C with a 16 : 8 (L : D) photoperiod, offered a blood meal twice a week and provided with raisins ad libitum (as a carbohydrate source). Eggs were collected from artificial tree holes placed in the adult colony cages and incubated at 21 °C with 14 : 10 (L : D) photoperiod to promote embryonation. Upon completion of the embryonation period (21 days), eggs were stored at 4 °C on moist paper towels in sealed Petri dishes. Eggs were stored for less than 1 year and represent the first lab generation of these colonies.

Experimental design

We examined the geographic patterns in host susceptibility to infection by the parasite, L. clarki by performing a common garden experiment in the laboratory. Host eggs were hatched synchronously by immersion in 0.1% sodium sulphite solution for three hours (Schwan & Anderson, 1980) and freshly hatched (<24-hour old) first-instar larvae were exposed to parasites. Parasites were released from 25 infected larvae about 48 h prior to the experiments and pooled in 2 L of dilute tree hole water (15% autoclaved tree hole water and 85% deionized water) in order to obtain a density of the parasitic form of L. clarki sufficient to produce a moderate infection rate in the hosts.

In the first experiment, we exposed larval populations of O. sierrensis hosts from four discrete geographic populations (Hopland, Sonoma, Los Angeles (LA-1) and Fresno; Fig. 1) to two parasite populations (Hopland and Sonoma, Fig. 1) to test for variation in host resistance and local adaptation. Each parasite population was derived from infected larvae collected from a single tree hole. In addition, hosts collected from four different tree holes within the Hopland population were exposed to the two parasite populations to quantify local variation in host resistance. Freshly hatched first instar larvae were placed in groups of three in small Petri dishes containing 15 mL of dilute tree hole water and 1 mL of the L. clarki culture was added to the parasite treatments and we added1 mL of dilute tree hole water to the controls.

In the second experiment, we increased the resolution of the geographic survey by including eight geographically discrete host populations (six populations from the coastal range: Hopland, Sonoma, Big Sur, Santa Barbara, Los Angeles (LA-2), Julian and two populations from the Sierra Nevada foothills or eastern Central Valley: Blodgett and Fresno, Fig. 1) and hosts from eight sites from within the Hopland population. We exposed larvae from the eight geographically discrete populations to L. clarki collected from a single tree hole at Hopland. We placed individual freshly hatched first instar larvae in plastic dishes containing 4.5 mL rearing fluid, eliminating the potential for horizontal transmission of L. clarki infection. For each host treatment, 30 larvae were exposed to parasites and 30 larvae were maintained as a control. We added 0.5 mL L. clarki culture to the parasite treatments, and we added 0.5 mL of dilute tree hole water to the controls. We distributed the dishes randomly among 30 trays, each containing one parasite treatment and one control dish for each host population.

In both experiments, trays were placed in an environmental growth chamber and maintained at 12 °C in total darkness for 30 days. This temperature is conducive to L. clarki infections and is within the range of winter temperatures observed in natural tree holes in California (Holly H. Ganz, unpublished data). Three times weekly, larvae were fed ad libitum a mixture of ground guinea pig chow, Tetramin flakes, liver powder and yeast powder. Using a dissecting microscope (10–40 times), we examined each larva for signs of infection three times per week. For each infected larva (live and dead), we recorded the presence and number of L. clarki inside the body cavity and the number of days survived after initial treatment. After 30 days, a measure of parasite load (the number of parasites within each infected host) within infected larvae was determined by assigning infected larvae to one of five categories: 0 (uninfected), 1 (1–25 cells), 2 (26–50 cells), 3 (51–75 cells), 4 (76–100 cells) and 5 (>100 cells). Dead larvae were stained for 30 min with black amide dye to ascertain if L. clarki invasion cysts were present on the host's cuticle (Soldo & Merlin, 1972; Washburn et al., 1988). Because the parasite cycle is typically completed within 30 days at 12 °C, each larva was scored for infection and the experiment was terminated after 30 days.

Statistical analysis

In the first experiment, each group of larvae was a replicate and we scored groups for the number of infected individuals and we determined the proportion infected. In the second experiment, a larva represented a replicate because individuals were placed in separate dishes. In both experiments, we used logistic regression to test for differences in parasite infectivity (proportion infected) and parasite load among host populations and among sites within the Hopland population (using the Fit Model function in jmp, 2004). In the first experiment, we performed a multinomial logistic regression, which distils the data for three nonindependent larvae into a single independent observation (i.e. 0, 1, 2 or 3 infected larvae per dish). The model for this logistic regression included host population, tree holes nested within host population, parasite population, and interactions among them as independent variables and proportion of infected larvae and parasite load as dependent variables. In this model, a significant parasite-by-host interaction effect in the logistic regression would indicate local adaptation by either the host or parasite. In the second experiment, we tested for differences among host populations and among tree holes within the Hopland population by binomial logistic regression, including host population and sites nested within the Hopland host population in the model. The dependent variable was whether a larva was infected by L. clarki or not. In the second study on geographic patterns in host susceptibility, we used Spearman rank correlation to test whether parasite infectivity or parasite load correlated with geographic distance of the host population from the parasite source.

Results

Relative migration rates

Colonization of artificial tree holes in the field

At Hopland and Sierra Foothill, 50% (287/576) and 62% (156/252) of artificial tree holes were colonized by O. sierrensis (hosts), respectively, and none were colonized by the parasite, L. clarki. At Marin, 71% (284/400) of artificial tree holes were colonized by O. sierrensis and <1% (3/400) were colonized by L. clarki.

Field surveys for presence or absence of the host and parasite in natural tree holes

The mosquito, O. sierrensis was found in 97% (147/152) of tree holes holding water. The parasite, L. clarki occurred in 25% (37/147) of tree holes with mosquitoes. In repeated surveys of tree holes that had the parasite, 98% (36/37) retained the parasite in subsequent years. In the laboratory, L. clarki survived and reproduced for at least 7 days in 82% (64/78) of water samples collected from tree holes where they were absent.

Field release of the parasite

The parasite, L. clarki established and persisted for first year in 41% (32/78) of tree holes where they were released and persisted for a second year in 94% (30/32) of tree holes where they had become established.

Local adaptation

Experiment 1

Infectivity: invasion cysts of L. clarki were observed within 48 h in the parasite treatment. The two parasite treatments did not differ significantly in the number of invasion cysts observed on larvae sampled during the first week (χ2 = 0.69, d.f. = 1, P = 0.4) and no invasion cysts were found in controls. The four O. sierrensis (host) populations exhibited significant variation in susceptibility to infection (Table 1, Fig. 2). Hosts from the four tree holes within Hopland were all susceptible to infection (Table 1, Fig. 2). Hopland L. clarki (parasites) were not significantly more infectious than Sonoma parasites overall (Table 1). Notably, O. sierrensis (host) populations differed in their response to two parasite populations (host-by-parasite interaction effect, Table 1; Fig. 2); Hopland hosts were more susceptible to Hopland parasites, and Sonoma hosts were more susceptible to Sonoma parasites, which is consistent with local adaptation by parasites. However, one distant host population (Fresno) exhibited high susceptibility to infection by both parasites, while the other distant host population (LA-1) was highly resistant to both parasite populations. Intriguingly, Hopland L. clarki infected both their local hosts and Fresno hosts at a similar rate and Sonoma L. clarki infected their local hosts and Fresno hosts at a similar rate.

Table 1.  Logistic regression on number of infected larvae in experiment 1.
Sourced.f.Wald χ2P > χ2
  1. There were two L. clarki (parasite) populations (Hopland and Sonoma).

Host population215.60.0014
Within Hopland32.110.55
Parasite population13.690.0548
Host × parasite population27.840.0495
Figure 2.

Mean number of infected larvae (in groups of three) from four California populations exposed to two parasite populations (Hopland and Sonoma). Vertical bars represent 1 SE of the mean.

Parasite load: parasite load did not differ among infected larvae from different O. sierrensis populations (Table 2). Within Hopland, infected hosts from four different tree holes did not differ significantly in parasite load (Table 2) and parasite load did not differ between the two parasite populations (Table 2). The host-by-parasite interaction effect was not significant for parasite load (Table 2). Thus, once a host is infected, L. clarki are able to reproduce inside hosts from different populations at the same rate.

Table 2.  Logistic regression on parasite load of infected larvae in experiment 1.
Sourced.f.Wald χ2P > χ2
Host population20.00140.999
Within Hopland30.360.95
Parasite population10.3390.56
Host × parasite population20.4080.94

Experiment 2

Infectivity: As before, we observed invasion cysts of L. clarki within 48 h in the parasite treatment. Overall, levels of infection were lower in the second study but were sufficient to allow for comparisons in host susceptibility to a Hopland parasite among O. sierrensis populations that ranged in distance from the parasite source. No infections were observed in control host populations. When both local (Hopland) and distant O. sierrensis populations are included in the logistic regression, host populations differed significantly in susceptibility to Hopland parasites (Table 3). Hosts from eight different tree holes within Hopland were all susceptible to infection by Hopland parasites (Table 3). Distant host populations did not differ significantly in susceptibility (χ2 =10.04, d.f. = 6, P = 0.12). Overall, host susceptibility (as measured by proportion of hosts infected) decreased with increasing geographic distance from parasite source population (Fig. 3, rs = −0.52, P = 0.0475), which supports the conclusion that L. clarki is locally adapted to its host. Still, as we found in the prior study, Fresno hosts were highly susceptible to infection and a greater proportion of hosts were infected than the local hosts. Another distant host population from the foothills of the Sierra Nevada (Blodgett) was also highly susceptible. Nonetheless, all five distant host populations from the coastal range were not very susceptible to infection by Hopland parasites. If we include only those populations from the coastal range in the Spearman rank correlation, the relationship between host susceptibility and geographic distance from the parasite source is highly significant (Fig. 3, rs = −0.795, P = 0.0012).

Table 3.  Logistic regression on number of O. sierrensis larvae infected by L. clarki from Hopland in experiment 2.
Sourced.f.Wald χ2 > χ2
Host population714.60.04
Within Hopland79.810.20
Figure 3.

Proportion of hosts infected in the second experiment, which included fifteen sites that range in distance from parasite source population. Coastal range populations are indicated with a circle and populations from the Central Valley and foothills of the Sierra Nevada are indicated with a triangle.

Parasite load: parasite load did not differ among host populations (Table 4) or among tree hole sites within the Hopland population (Table 4).

Table 4.  Logistic regression on parasite load in larvae infected by L. clarki from Hopland in experiment 2.
Sourced.f.Wald χ2 > χ2
Host population40.00981.00
Within Hopland70.690.999

Discussion

Despite the emphasis on migration rates in predicting the outcome of local adaptation by hosts or parasites in theoretical models, few host-parasite interactions have been characterized for the relative migration rates of hosts and parasites and tested for local adaptation. In a well-studied interaction between a snail and its trematode parasite, genetic studies indicate that the parasites migrates more than its host (Dybdahl & Lively, 1996) and the parasite exhibits local adaptation (reviews: Lively, 1999; Lively et al., 2004). Lively (1999) pointed out the need for estimates of parasite migration and virulence in other interactions in nature because these are crucial parameters for determining when and if local adaptation is expected. In the present study, we found indications that a mosquito host exhibits higher relative migration rates than its protozoan parasite and, contrary to expectations based on theory, we found no support for local adaptation by hosts and some support for local adaptation by parasites.

Numerous empirical field observations lend support to the hypothesis that the host, O. sierrensis migrates more than its parasite, L. clarki. The life span of infected adult females (which disperse the parasite to new, unoccupied habitats) is shorter than uninfected females (Egerter et al., 1986), and they don't fly as long (and presumably as far) as uninfected females (Yee & Anderson, 1995). Results from field studies presented here show that the parasite colonized artificial tree holes at a much lower rate than its host in the field. Moreover, field surveys revealed that the host occupies most available tree hole habitats in the field, whereas the parasites occupy only a subset (Washburn & Anderson, 1986). Although some tree holes were not suitable for the parasite (because they contain water with high electrical conductivity and/or extreme pH values; Egerter & Anderson, 1985), water conditions within most tree hole habitats were amenable to survival of free-living L. clarki trophonts. In the laboratory, the parasite survived and reproduced in 95% of tree hole water samples collected from parasite-negative tree holes. Further, the parasite established successfully in 40% of parasite-negative tree holes and persisted in these tree holes for a second year at a high rate (94%).

If the host migrates more than its parasite, theory predicts that the host should be less susceptible to local parasites (Ladle et al., 1993; Judson, 1995; Gandon et al., 1996,1998). Although we found evidence that the host migrates more than its parasite, results of our infection studies provided no support for hosts exhibiting less susceptibility to local parasites and, instead, suggested that the parasite, L. clarki may be locally adapted to its host. Two parasite populations were better able to infect their local hosts than the other parasite population (Fig. 2). In addition, L. clarki from a single population (Hopland) was more infectious of geographically close host populations, which is consistent with local adaptation by parasites (Ebert, 1994). Host susceptibility among regional populations was more variable: two distant host populations from near the foothills of the Sierra Nevada (Fresno and Blodgett) were highly susceptible in both experiments and Fresno was at least as susceptible to infection as the local hosts. These results are consistent with theory, which predicts that novel combinations of hosts and parasites will produce variable results (Gandon et al., 1996; Morand et al., 1996). Furthermore, differences in the susceptibility among the distant host populations may reflect variation in host resistance (Thrall et al., 2002) and it may be that Fresno population is unconditionally susceptible to infection. Alternatively, five other distant host populations from the California coastal range were not very susceptible to infection by Hopland parasites (also from the coastal range). It should be noted that the Hopland parasite population could still be locally adapted even if there was no observed decrease in susceptibility with increasing geographic distance. Lively et al. (2004) found that neither host genetic distance nor geographic distance between populations affected the strength of local adaptation in a snail-trematode interaction.

The indications of local adaptation by L. clarki despite its lower migration rate relative to that of its host, O. sierrensis suggests that relative migration rates do not always predict the direction of local adaptation in host-parasite interactions. As this host-parasite interaction is well characterized, in the following section, we consider other aspects of the life histories of the parasite and host that may contribute to the strength and direction of local adaptation. We focus in particular on parasite virulence, resource competition, and host dormancy. Although it is tempting to suggest that the shorter generation time and higher reproductive rate of L. clarki contribute to a higher evolutionary potential, leading to local adaptation by parasites (as predicted by Hamilton et al., 1990), theoretical explorations indicate that generation time does not always explain the tendency for parasites to be locally adapted (Lively, 1999; Gandon, 2002; Gandon & Michalakis, 2002).

Parasite virulence strongly influences local adaptation (Lively, 1999; Gandon, 2002) and this could be a factor here. The parasite L. clarki exhibits high virulence in O. sierrensis; most infected hosts die as larvae and those that do survive to adulthood are parasitically castrated (Egerter et al., 1986; Washburn et al., 1988,1991). Because L. clarki is facultatively parasitic and occurs in a free-living form in the absence of hosts (Washburn et al., 1988), the parasite can maintain a high level of virulence and survive local host extinction events, assuming that there is no trade-off between virulence and survival in the free-living stage. Facultative parasitism may be similar to ‘sit-and-wait’ pathogens, where the duration of pathogen survival outside the host is positively correlated with virulence (Ewald, 1987,1994). On the other hand, even highly virulent parasites may not provide strong enough local selection for host resistance traits because other ecological factors may be more important. Tree hole habitats in this region are resource-limited (Maciá & Bradshaw, 2000) and experience periodic drought and flooding. While the parasite, L. clarki lowers survivorship when larvae are raised with sufficient food, host mortality is compensatory under food limitation: emerging adults are at least as abundant with higher average fitness compared with uninfected controls (Washburn et al., 1991).

Aspects of the host's life history, particularly the existence of a dormant, host egg bank may affect the ability of the host to adapt to local parasites. Host eggs can remain dormant for several years and eggs from a single clutch can exhibit a wide span of hatching times (unpublished data). The hatching span of a clutch may be the best measure of female fitness in O. sierrensis, as was found in a tree hole breeding damselfly (Fincke & Hadrys, 2001), and warrants further study. Moreover, the dormant host egg bank will also contain eggs from females emerging from neighbouring tree holes, not all of which contain parasites.

In summary, we found no support for local adaptation by a host, O. sierrensis to its parasite, L. clarki despite providing support for the hypothesis that the host exhibits higher migration rates than its parasite. We suggest that, in addition to migration rates, other aspects of the life histories of the host and parasite may affect patterns in local adaptation, particularly parasite virulence and host dormancy.

Acknowledgments

Holly H. Ganz thanks H. Dingle, S.P. Lawler, C.M. Lively, and J.A. Rosenheim for their thoughtful advice and encouragement. Y. Michalakis, A.M. Bell, K.A. Copren and D. Ebert and an anonymous reviewer provided helpful comments on the manuscript. Thanks to N. Willits for statistical consulting. C.X. Osborne, A.M. Bell, S. Townsend and B. Larison helped with field collections. M. Moua, R. Yang, E. Wang and T. Yang assisted in the laboratory. D. Woodward (Lake County Vector Control District), D. Dritz and L. Styer gave advice on rearing mosquitoes. We are grateful to the staff of Hopland Research and Extension Center, Sierra Foothill Research and Extension Center and Blodgett Forest Research Station for allowing site access. Thanks to Paul O'Conner of the Greater LA County Vector Control District for collecting samples of O. sierrensis. Holly H. Ganz received funding from the University of California, Davis (Block Grant, Jastro Shields, Center for Population Biology and Humanities and Graduate Research), the University of California Mosquito Research Programme (Student Mini-Grant) and Sigma Xi. Support for studies on parasite and host migration was provided by a grant from the National Institutes of Health (J.R. Anderson, Principal Investigator).

Ancillary