1. Individuals of free-living organisms are commonly infected by multiple parasite species. Under such circumstances, positive or negative associations between the species are possible because of direct or indirect interactions, details in parasite transmission ecology and host-mediated factors. One possible mechanism underlying these processes is host immunity, but its role in shaping these associations has rarely been tackled experimentally.
2. In this study, we tested the effect of host immunization on associations between trematode parasites infecting eyes of fish. We first analysed the associations between three species (Diplostomum spathaceum, Diplostomum gasterostei and Tylodelphys clavata) in wild hosts, roach (Rutilus rutilus) and perch (Perca fluviatilis). Second, using rainbow trout (Oncorhynchus mykiss) as a model fish species, we experimentally investigated how sequential immunization of the host (i.e. one parasite species infects and immunizes the host first) could affect the associations between two of the species.
3. The results indicated that most of the associations were positive in wild hosts, which supports between-individual variation in host susceptibility, rather than competitive exclusion between the parasite species. However, positive associations were more common in roach than in perch, possibly reflecting differences in ecological conditions of exposure between the host species. The experimental data showed that positive associations between two of the species were eroded by host immunization against one of the parasite species.
4. We conclude that sequential immunization of hosts has a marked effect on interspecific parasite associations and basically can determine if positive associations are detected or not. This implies that correlative results suggesting non-interactive community structure in general may be obscured by the sequence of previous parasite exposure and corresponding dynamics of host immunization.
In general, most of the previous studies have focused on immunization as a mediator of competitive interactions while its effect on positive, or neutral, associations has received very little attention. Especially, effects of sequential host immunization (i.e. one parasite species infects and immunizes the host first) on interspecific associations between parasites have remained largely unexplored (but see Jackson et al. 2006). In this study, we explored associations between three related species of trematode parasites infecting eyes of freshwater fish: Diplostomum spathaceum (Rudolphi) found in the lens, and Diplostomum gasterostei (Williams) and Tylodelphys clavata (von Nordmann) infecting the humour. These species are ubiquitous parasites of freshwater fish and have very similar three-host life cycles including an avian definitive host, and snail and fish intermediate hosts. Co-infections of these parasites in same fish individuals are typical in nature, although parasite numbers can vary to some extent between populations and host species (Valtonen, Holmes & Koskivaara 1997; Valtonen et al. 2003; Karvonen et al. 2006). Fish are known to mount an immune response against D. spathaceum (probably also against the other species), which reduces the establishment in subsequent exposures (see Chappell, Hardie & Secombes 1994 for review; Karvonen et al. 2005). More specifically, exposure of fish to D. spathaceum elicits responses in both non-specific (e.g. activation of macrophages) and specific (e.g. T-cell activation) branches of the immune system. In general, specific responses develop in few weeks after the first exposure as a result of parasite antigens, which has been demonstrated experimentally using both live and attenuated parasite infective stages (Chappell et al. 1994). This system is particular, however, as the parasites (metacercariae) of all three species are basically protected from host immunity in the eye. This is why host responses have to tackle the parasites within 24 hours from exposure while they are migrating in host tissues towards the eye. In general, heavy infections of D. spathaceum in the eye can lead to impaired fish growth (Karvonen & Seppälä 2008) and increased susceptibility to predation (Seppälä, Karvonen & Valtonen 2005). However, no such effects have been described in the other species, D. gasterostei and T. clavata, infecting the eye humour.
In the first part of this study, we explored interspecific associations between D. spathaceum, D. gasterostei and T. clavata in nature by analysing published data sets on the parasite infections in two main fish host species, roach (Rutilus rutilus L.) and perch (Perca fluviatilis L.) in four Finnish lakes. We expected to find positive associations between the species as they have similar ecological features such as strong temporal overlap in transmission from snails to fish (Karvonen et al. 2006), as well as similar routes, mechanisms and interests of transmission to avian definitive hosts. They also infect different parts of fish eye (except for D. gasterostei and T. clavata both found in the humour), which reduces the likelihood of direct competitive interactions. Despite the temporal synchrony in the parasite transmission to fish, however, spatial differences in distribution of snail intermediate hosts and fish host species (habitat specialization) are still likely. This could generate differences in the sequence of exposure and subsequent immunization to the parasites, especially between fish species, and have an effect on interspecific associations among the parasites (Jackson et al. 2006). In the second part of this study, we explored this experimentally by investigating how previous immunization of rainbow trout (Oncorhynchus mykiss Walbaum) against one of the species, D. spathaceum, mediated associations between D. spathaceum and D. gasterostei in subsequent co-exposures taking place in natural conditions. Essentially, in this set-up, we placed two types of fish groups to a lake at the same time: groups which had been previously exposed and immunized to D. spathaceum, and corresponding control groups. Previous immunization of fish manipulated the sequence of parasite establishment (one species infecting and immunizing the host first), thus corresponding to a situation where the previous exposure to parasites would have been spatially heterogeneous (see above). On the other hand, groups of immunologically naïve control fish (without any previous immunization to parasites) were exposed simultaneously to both parasite species, thus corresponding to the situation of spatial and temporal synchrony in parasite transmission (see Karvonen et al. 2006).
Materials and methods
Associations between the parasite species in wild fish
Data on trematode infection in eyes of wild roach and perch were extracted from published work of Valtonen et al. (1997, 2003). These studies present seasonal (monthly) infection data of the parasites in four lakes in Central Finland: Lake Peurunka, Lake Vatia, Lake Saravesi and Lake Leppävesi. We excluded Lake Vatia from the data in Valtonen et al. (1997) as the lake was heavily polluted by that time, which had a marked effect especially on trematode communities. However, in the latter study (Valtonen et al. 2003), the lake had recovered and was in a state comparable with the other lakes. Numbers of D. spathaceum, D. gasterostei and T. clavata were variable between the lakes, sampling months and fish species (see Valtonen et al. 1997, 2003 for details), and therefore the data captured a wide range of natural variation in the parasite associations. It should be noted that the taxonomic identity and nomenclature of Diplostomum metacercariae, especially those infecting the eye lens, is notoriously complex and not completely resolved (e.g. Valtonen & Gibson 1997). This is because of close morphological resemblance of the metacercariae and unreliable species descriptions. In this study, we used the name D. spathaceum for the parasites found in the lens, a method commonly adopted in the previous literature.
We tested for positive pair-wise associations between D. spathaceum, D. gasterostei and T. clavata using one-tailed Pearson correlation analysis (tests were run one-tailed for the purpose of the subsequent meta-analysis, see below). Fish individuals where both parasite species in a species-pair were absent (double zeros) were excluded. In other words, we focused on cases where the numerical response of one species to the presence of other could be analysed. Analyses were conducted separately for each of the lakes and sampling months [8 months in Valtonen et al. (1997), 6 months in Valtonen et al. (2003), 15 fish individuals for each month] to avoid false correlations arising from pooling the data (Poulin, Steeper & Miller 2000; Bottomley, Isham & Basánez 2005). These parasites also accumulate in fish with age (e.g. Marcogliese et al. 2001). Thus, to control for the effect of fish age (determined from scales and operculum) on parasite associations, we analysed the relationship between fish age and parasite numbers for each parasite species–lake-sampling month combination using linear regression analysis and used the residuals from these regressions in the pair-wise correlations (Poulin et al. 2000; Behnke et al. 2005; Bottomley et al. 2005). To obtain a community-level analysis on the species interactions, we used Fisher’s meta-analysis (the inverse chi-squared method) to combine the result of the multiple correlations for each species-pair (Hedges & Olkin 1985; Sokal & Rohlf 1998). This is a widely used method for combining multiple test probabilities, which test for one hypothesis. Essentially, after calculating the one-tailed probabilities for positive associations for each parasite species–lake-sampling month combination (see above), natural logarithms of the probabilities were summed and multiplied by −2. This value of coefficient was then compared with chi-squared distribution with 2k degrees of freedom, where k is the number of individual correlations (Hedges & Olkin 1985; Sokal & Rohlf 1998; see also Karvonen et al. 2007; Tello, Stevens & Dick 2008). The combined P-value obtained from the chi-squared distribution indicated whether the positive associations between the species in a species-pair were significant.
Effect of sequential host immunization on the interactions between two parasite species, D. spathaceum and D. gasterostei, was investigated by exposing previously immunized fish and control fish to simultaneous infection from both parasites under natural conditions. We used rainbow trout (O. mykiss) in the experiment as it is susceptible to infection and known to elicit an immune response against D. spathaceum (Chappell et al. 1994; Karvonen et al. 2005). It should be emphasized, however, that the purpose of the experiment was not to explain the associations between the parasites in wild fish species, but use rainbow trout as a model fish species to explore the effects of sequential immunization.
A stock group of 1100 immunologically naïve rainbow trout was obtained from a commercial fish farm. The farm used ground water supply, which ensured that the fish had no prior experience with the parasites. A group of 400 randomly chosen fish [some of the remaining fish were used in keeping the fish density constant in the caging experiment (see below); rest of the fish were used in other experiments] was divided into eight and put into eight tanks, each with 50 fish and 65 L of water, after which randomly chosen four tanks received a dose of 30 D. spathaceum cercariae per fish (total 1500 cercariae tank−1) for 30 min. A pooled cercarial suspension had been extracted from five naturally infected Lymnaea stagnalis snails (see Karvonen et al. 2003), which ensured that the fish were exposed to freshly emerged parasite cercariae of multiple genotypes (Rauch, Kalbe & Reusch 2005). The infected snails were collected from the area of the caging experiment (see below). Other tanks received sham exposure with water without cercariae. Water temperature was brought up to 15 °C for the time of exposure and subsequent 48 hours to ensure parasite infection success. After this, fish were maintained in the tanks in 250 L for 26 days during which time the water temperature corresponded to natural lake temperature being, on average, 11·8 °C. The water was filtered to ensure that the fish gained no further infections. After 26 days, the mean number of parasites in the exposed fish was 2·17 ± 0·38 (SE) whereas the control fish had no parasites.
In the beginning of July, a total of 300 randomly chosen fish (150 immunized fish and 150 control fish; remaining fish were used for other purposes) were taken evenly from all eight tanks and transferred to Lake Konnevesi. Fish were placed in three cages so that each cage received groups of 50 immunized fish and 50 control fish. Fish were marked by clipping the adipose fin from one, randomly selected group, under anaesthesia [0·01% MS-222 (Sigma Chemical Co., St Louis, MO, USA) as an anaesthetic]. Cages were 120 × 80 × 100 cm and had a mesh size of 10 mm that allowed parasite cercariae to pass (Karvonen, Seppälä & Valtonen 2004). The size of the cages was large enough not to restrict the movement of fish individuals but small enough to ensure homogeneous exposure to the parasites within a cage. In other words, exposure of the treatment groups (immunized/control) was similar within each cage and followed natural parasite dynamics (see below). Cages were anchored to shallow littoral zone of the lake so that they floated at the surface c. 50 cm off the bottom. Distance between the cages was c. 10 m. Fish were fed with commercial fish pellets every second day.
After 4 weeks, 10 immunized and 10 control fish were taken from each cage. A corresponding number of new fish from the original stock group, marked with a different type of fin clip, was introduced to the cages to keep the fish density constant. Note that these fish did not affect the infection of the treatment groups (immunized/control) because (i) the exposure was homogeneous between all fish individuals within a cage (see above); and (ii) the parasite cercariae readily penetrate into fish once they become in contact with a host after which they cannot re-infect another fish. Fish taken from the cages were brought to laboratory, euthanized with an overdose of 0·01% MS-222 and examined for metacercariae from eye lens (D. spathaceum) and humour (D. gasterostei). The experiment was terminated after another 4 weeks (8 weeks from the beginning of the experiment) when 20 immunized and 20 control fish were examined from each cage (the remaining fish were used for other purposes). Overall, this set-up was designed to capture natural small-scale spatial variation in parasite exposure, if any, as well as the temporal variation in parasite seasonality (Karvonen et al. 2004), by using replicated cages and two sampling times. This ensured that the infection of the treatment groups within the cages corresponded to natural parasite dynamics.
The data were analysed using mancova with numbers of D. spathaceum and D. gasterostei as response variables, and immunization status (immunized/control), sampling time and cage as fixed factors (cage was not treated as a random factor because this would have required more replication). Fish length was used as a covariate (all fish belonged to the same age cohort, 1-year-old, but there were slight differences in size). Parasite numbers were transformed [log(n + 1)] to meet the assumptions of the analysis. mancova was then followed by ancovas separately on the log-transformed parasite numbers of each species. Note that the fish immunized with D. spathaceum had few parasites already when introduced to cages, but as these numbers were very low (mean 2·17 ± 0·38), this is unlikely to affect the observed result (see below). We also recognize that parasite numbers determined from fish individuals taken from the same cage do not represent completely independent observations in a statistical sense. However, this is unlikely to have an effect in this study because (i) parasite cercariae are produced continuously from the snails so that no dilution in the level of exposure could have occurred; (ii) fish were evenly distributed within the cages and therefore encountered the cercariae in a random manner; and (iii) the infection cannot be transmitted directly between fish individuals. Associations between the parasite species in immunized and control fish were analysed using Pearson correlation analysis separately for both sampling times.
Immunization to D. spathaceum was induced using an experimental exposure of fish to the parasite cercariae, the dose of which was based on our earlier experience of the system (e.g. Karvonen et al. 2003). The level of infection in the experimental fish after the caging ranged from 8 to 96 corresponding to infection intensities observed in the wild fish populations, where intensities typically ranged from zero to hundreds (Valtonen & Gibson 1997; Marcogliese et al. 2001). For example, Chappell (1969) recorded over 200 metacercariae from an individual three-spined stickleback (Gasterosteus aculeatus), and Wootten (1974) found 550 metacercariae from an individual rainbow trout. There was no mortality of fish during the experiment. All fish appeared healthy and no alterations in behaviour, feeding or growth were observed. The caging experiment used a total of 300 immunized and control fish (as well as fish from the original stock group used in keeping the fish density constant), but more fish were immunized for purposes of other experiments. All fish were killed with an overdose of 0·01% MS-222, a commonly used fish anaesthetic. The experiment was carried out with a permission of Lab-Animal Care and Use Committee of the University of Jyväskylä (license number 8/27.3.2006).
In accordance with our hypothesis, a majority of the coefficients among the individual correlation analyses were positive (57·8% overall; Table 1), indicating that positive associations among the three parasite species were more common than negative ones. However, at the level of fish species, positive associations were still more common in roach (67·3%) compared with perch (48·1%). Combination P-values calculated for roach indicated that all pair-wise associations among the three parasite species were significantly positive and this pattern was consistent between the data sets (Table 1). On the other hand, in perch, only one of the six pair-wise comparisons (D. gasterostei–T. clavata in the first data set) was significantly positive (Table 1).
Table 1. Result of the Fisher’s meta-analysis (the inverse chi-squared method) on individual Pearson correlations calculated for species-pairs of the three trematode species (Diplostomum spathaceum, Diplostomum gasterostei and Tylodelphys clavata) infecting eyes of roach (Rutilus rutilus) and perch (Perca fluviatilis)
D. spathaceum–D. gasterostei
D. spathaceum–T. clavata
D. gasterostei–T. clavata
Data were extracted from Valtonen et al. (1997, 2003). Number of fish indicates the total number of fish included in the N individual correlations. Positive % indicates the percentage of positive correlations (remaining correlations were negative). Significance of the coefficient was obtained from chi-squared distribution with 2k degrees of freedom, where k is the number of individual correlations (see statistical analysis for details).
In the caging experiment, five fish were excluded from the data because parasite numbers could not be determined from one of the eyes. Numbers of D. spathaceum in the eyes of fish increased with time (Table 2, Fig. 1), being 50·71 ± 2·54 (SE) and 42·32 ± 1·91 in the control and immunized fish respectively, at the end of the experiment. Corresponding coefficients of variation (measuring variation in parasite numbers, calculated as variance divided by mean) were 7·39 and 5·16 (Levene’s test for equality of variances in untransformed data: F = 3·369, P = 0·069). Numbers of D. spathaceum did not differ between the treatment groups (main effect of status), but there was a significant interaction with time because the parasite numbers were higher in controls at the end of the experiment (t-test: t116 = 2·188, P = 0·031) (Table 2, Fig. 1). Numbers of D. gasterostei also increased with time and were 26·34 ± 1·29 and 24·57 ± 1·42 in the control and immunized fish respectively, at the end of the experiment. Corresponding coefficients of variation were 3·69 and 4·92 (Levene’s test for equality of variances in untransformed data: F = 0·108, P = 0·743). Again, immunization with D. spathaceum had no main effect on the numbers of D. gasterostei, but the interaction with time was marginally significant probably because parasite numbers were higher in immunized fish after 4 weeks of exposure (Table 2, Fig. 1). However, the parasite numbers did not differ between immunized and control fish at the end of the experiment (t-test: t116 = 1·244, P = 0·216). The pattern of D. spathaceum infection did not differ between the cages, but it was different in D. gasterostei so that the parasite numbers were higher in one of the cages at the end of the experiment (significant main effect of cage; Table 2). However, this pattern did not differ between the treatment groups (treatment group–cage interaction; Table 2), and therefore pooled data from the cages were used to analyse differences between the groups. Effect of fish length was not significant in any of the analyses.
Table 2. Result of mancova and univariate ancovas on log(n + 1)-transformed numbers of Diplostomum spathaceum and Diplostomum gasterostei in immunized and control rainbow trout (Oncorhynchus mykiss) caged in a lake for a period of 8 weeks
Status = immunized/control, time = time under exposure (4 weeks/8 weeks) and cage = cages 1–3 were used as fixed factors. Fish length was used as a covariate.
S × T
S × C
T × C
S × T × C
Based on the results from the wild fish, we tested for positive associations between D. spathaceum and D. gasterostei in the caged fish using one-tailed Pearson correlation analysis. We found a statistically significant positive correlation between the numbers of the parasites in the control group in the second sampling time (r2 = 0·402, n = 58, P < 0·001; Fig. 2b). The correlation in the first sampling time also tended to be positive but was not significant at 5% level (r2 = 0·286, n = 28, P = 0·070; Fig. 2a). However, no such relationship was observed in the immunized treatment group in the first (r2 = 0·182, n = 29, P = 0·172; Fig. 2c) or the second sampling time (r2 = 0·123, n = 60, P = 0·174; Fig. 2d).
Interspecific associations in multispecies communities are an important determinant of community organization in free-living organisms, as well as in parasites exploiting their hosts as a resource (Poulin 2001; Lello et al. 2004). An important difference between these systems, however, is that host represents a rapidly responsive environment where associations between parasite species may be shaped by complex immunological responses of the host. To date, however, there is very little evidence on the role of host immunization in shaping these associations, particularly in others than competitive interactions (e.g. Lello et al. 2004; Råberg et al. 2006; Grech et al. 2008). In the present study, we explored interspecific associations between three trematode parasites infecting eyes of freshwater fish by analysing data sets on the infections in two host species in several populations with variation in levels of infection. We found significant positive associations between the species so that they were more common in one hosts species (R. rutilus) than in another (P. fluviatilis). This somewhat contradicts the earlier results, which have suggested negative associations between the parasite species particularly in cases of very high parasite numbers (Kennedy & Burrough 1977; Burrough 1978; Kennedy 1981, 2001). In the experimental part of the study using rainbow trout (O. mykiss) as a model fish species, we showed that positive associations between two of the parasite species were eroded by the priming of the host immune system against one of the species. These results suggest that sequential host immunization may shape associations between parasites, which may change interpretations based on correlative data sets.
Generally, positive associations in parasites may emerge through between-species similarities in transmission mechanisms as well as characteristics that render hosts to non-random patterns of multiple parasite species infections. For example, parasites may employ similar routes of transmission, or even same intermediate host individuals, leading to positive infection rates within individuals of the next host species (Poulin et al. 2000; Poulin 2001; Leung & Poulin 2007). Second, because of genetic, immunological or condition-dependent differences, some host individuals may be, or may become, more susceptible to infections from multiple parasites (Taylor et al. 1998; Jokela et al. 2000; Krasnov et al. 2005; see also Seppäläet al. 2009). Although we could not separate between these hypotheses in the present study, it could be expected that similar transmission dynamics and routes of these parasites are at least partly responsible for the observed positive associations. For instance, habitats of snails intermediate host species of D. spathaceum and D. gasterostei, as well as the timing of their cercarial release, are commonly overlapping (Karvonen et al. 2006) leading to simultaneous infection of fish from both parasite species. Despite of the overlap in the distribution of the snail intermediate hosts, however, typically there is still spatial variation in snail densities and community structure because of specific habitat requirements of different snail species (e.g. Wullschleger & Jokela 1999; Faltýnková, Valtonen & Karvonen 2008). Also, infection rates of parasites in snails are known to be spatially heterogeneous (e.g. Jokela & Lively 1995; Byers et al. 2008; Faltýnkováet al. 2008). This may generate spatial heterogeneity and differences in parasite exposure between host species if different areas are favoured by one host species over another. Interestingly, we observed only one positive association between the parasite species in perch whereas the positive associations were predominating in roach. However, at this point, it is unclear whether this result is related to differences in exposure between perch and roach. Resolution of this question would require experimental infections of the host species in different exposure conditions.
It is nevertheless likely that the spatial heterogeneity in risk of infection in general causes differences not only in the level of exposure between host individuals, but also in the sequence how hosts are exposed, and immunized, to different parasite species. This again may affect the outcome of interspecific associations in subsequent co-infections from multiple parasite species. For example, in a recent study, Jackson et al. (2006) manipulated the dynamics of polystomatid monogenean co-infections in clawed toads and found that competition between the species-pairs was significantly affected by the sequence of host invasion by the parasites, among other factors. In the present experimental set-up using rainbow trout as a model fish species, we first manipulated the sequence of parasite establishment and subsequent immunization (D. spathaceum infecting first), and then allowed simultaneous infections from the two parasite species (D. spathaceum and D. gasterostei). However, we were unable to do reciprocal immunization with D. gasterostei because intermediate snail hosts of the parasite were unavailable at the time of the experiment, and therefore cannot address possible asymmetrical effects of immunization. The result indicated that the priming of the host immune system with D. spathaceum eroded the positive associations between D. spathaceum and D. gasterostei observed in the controls, which suggests that sequential host immunization may shape also other types of interactions in addition to competitive ones (see Jackson et al. 2006). The exact mechanism for this is unclear, but variation in numbers of D. spathaceum tended to be lower in the immunized fish compared with controls (P = 0·069). This suggests that immunization, although less-effective than anticipated (see Karvonen et al. 2005), decreased variation in susceptibility to D. spathaceum among the fish individuals. However, this is unlikely to provide a comprehensive explanation for the result because (i) the distributions of numbers of both parasite species were clearly overlapping between the treatments; and (ii) the association between the species in the control group was positive from low to high parasite numbers particularly at the end of the experiment (Fig. 2b).
It is also possible that closely related parasite species could elicit cross-reactive immune responses in host (e.g. Adams et al. 1989; Kurtz 2005), when responses against one parasite species could also decrease the infection success of others. Our results from the experimental exposure indicated that priming of host immunity with D. spathaceum did not affect the numbers of D. gasterostei, which suggests the absence of cross-resistance in this system. However, recent studies have demonstrated that species of this genus show interspecific and intraspecific associations even at the parasite genotype level, which may involve complex immunological interactions with the host immune system (Rauch, Kalbe & Reusch 2008; Seppäläet al. 2009). Furthermore, as our experiment was conducted in a natural lake system, the fish may have also been exposed to other parasite species of the same genus emerging from different snail species, which could lead to different types of immune responses within the host. It should be stressed, however, that the parasites used in the immunization of fish originated from the area of the caging experiment, and therefore it is likely that the fish were exposed to the same lens-infecting species also in the cages. Nevertheless, possibility that other parasite species were also present in the area cannot be excluded. Overall, it is clear that cross-reactive responses in this system require further work once the diversity and taxonomic identity of the eye trematode species have been resolved.
Regardless of the underlying mechanisms, we can conclude that the result is indicative of one important aspect in interspecies associations: associations between parasite species may, or may not, be detected in hosts depending on the sequential dynamics of host immunization. In other words, non-significant correlation analyses between numbers of two species, indicating non-interactive community structure in general, may be obscured by heterogeneous exposure and immunization to parasites. This is particularly important as it may lead to opposite conclusions on the nature of interactions within a community and have significance in the ecology and evolution of multiple species infections. Most of the few previous studies on the relationships between host immunization and interspecific associations in parasites have focused on competitive interactions. Furthermore, to our knowledge, only one previous study has shown that sequential parasite establishment may shape interspecific interactions, namely competitive, between macroparasites of a vertebrate host (Jackson et al. 2006). More research is obviously needed on the dynamics of sequential host immunization with multiple parasite species in facilitative and antagonistic parasite interactions.
We thank Henry Halonen for help in the experiment and Konnevesi Research Station for the facilities. Jukka Jokela gave valuable comments on the manuscript. The study was supported by grants from the Academy of Finland (AK, OS) and the Academy of Finland Centre of Excellence in Evolutionary Research (AK).