Host-Parasite Interactions and Population Dynamics of Rock Ptarmigan

Populations of rock ptarmigan (Lagopus muta) in Iceland fluctuate in multiannual cycles with peak numbers c. every 10 years. We studied the ptarmigan-parasite community and how parasites relate to ptarmigan age, body condition, and population density. We collected 632 ptarmigan in northeast Iceland in early October from 2006 to 2012; 630 (99.7%) were infected with at least one parasite species, 616 (98%) with ectoparasites, and 536 (85%) with endoparasites. We analysed indices for the combined parasite community (16 species) and known pathogenic parasites, two coccidian protozoans Eimeria muta and Eimeria rjupa, two nematodes Capillaria caudinflata and Trichostrongylus tenuis, one chewing louse Amyrsidea lagopi, and one skin mite Metamicrolichus islandicus. Juveniles overall had more ectoparasites than adults, but endoparasite levels were similar in both groups. Ptarmigan population density was associated with endoparasites, and in particular prevalence of the coccidian parasite Eimeria muta. Annual aggregation level of this eimerid fluctuated inversely with prevalence, with lows at prevalence peak and vice versa. Both prevalence and aggregation of E. muta tracked ptarmigan population density with a 1.5 year time lag. The time lag could be explained by the host specificity of this eimerid, host density dependent shedding of oocysts, and their persistence in the environment from one year to the next. Ptarmigan body condition was negatively associated with E. muta prevalence, an indication of their pathogenicity, and this eimerid was also positively associated with ptarmigan mortality and marginally inversely with fecundity. There were also significant associations between fecundity and chewing louse Amyrsidea lagopi prevalence (negative), excess juvenile mortality and nematode Capillaria caudinflata prevalence (positive), and adult mortality and skin mite Metamicrolichus islandicus prevalence (negative). Though this study is correlational, it provides strong evidence that E. muta through time-lag in prevalence with respect to host population size and by showing significant relations with host body condition, mortality, and fecundity could destabilize ptarmigan population dynamics in Iceland.


Introduction
Parasite communities of wildlife species have rarely been studied over an extended time period [1]. Detailed studies can provide insight into many aspects of parasite ecology, the relationships between host and parasites, and adaptations resulting from this intimate network [1] [2] [3]. Parasites can influence population dynamics by affecting host body condition, fecundity, and survival, and host-parasite interactions are known to be one of the driving forces of multiannual cycles in animal populations (e.g., [4] [5] [6] [7] [8] [9]). Cycles have been reported in a variety of herbivorous species in northern latitudes including moths, hares, lemmings, voles, and grouse [9] [10], and are well documented for species in the genus Lagopus. The parasitic nematode Trichostrongylus tenuis is thought to be one of the agents driving the red grouse Lagopus lagopus scoticus cycle in Scotland [11]. Parasites of willow ptarmigan (L. l. lagopus) in Norway, have been reported to affect demographic parameters including body condition and breeding success of the host, and were negatively associated with changes in numbers of the ptarmigan population [12] [13].
Some rock ptarmigan (Lagopus muta) populations show multiannual cycles [13] [14] [15] [16]. Traditionally, the ptarmigan population in Iceland has cycled with a 10-12 year period [15], but recently this has changed and the periods have become shorter [17]. The gyrfalcon (Falco rusticolus) and the rock ptarmigan in Iceland have a coupled predator-prey cycle and the falcon is thought to be an agent driving the ptarmigan cycle [18] [19]. However, other researchers have demonstrated the importance of studying the effects of several different factors on the cyclic behaviour of a population (e.g., [20] [21] [22]). Parasites have the potential to effect host survival by making birds more prone to predation, weakening body condition, or changing host behaviour [12] [23] [24].
The parasite fauna of rock ptarmigan in Iceland has been described and 17 species have been reported [25] [26]. In this paper, we want to examine the potential role of the parasites on the population dynamics of rock ptarmigan. We do this by investigating the relationship between ptarmigan density, body condition, mortality, fecundity, and measures of parasite abundance and aggregation over a period of 7 years (2006-2012). We look at the combined parasite community and focus on six known pathogenic parasite species, namely Eimeria muta and Eimeria rjupa, intestinal microparasites that can cause coccidiosis [27], Capillaria caudinflata, a helminth known to cause capillariasis in Icelandic ptarmigan [28], Trichostrongylus tenuis, a helminth known to cause trichostrongylosis in grouse (also known as grouse disease) [3] [29], Metamicrolichus islandicus, a skin mite that can cause mange (pers. obs.), and Amyrsidea lagopi, an amblyceran chewing louse that can cause feather damage [30]. The Anderson and May model [4] [31] suggests that time delays in parasite abundance in relation to host population size, reduced aggregation of parasites during periods of high abundance, and the influence of parasites on host mortality and fecundity are the three destabilizing qualities of parasite-host dynamics. So, if the parasite community or any of the parasites of the rock ptarmigan are of importance, we expect the prevalence of a species to track host density, but with a time-lag, be least aggregated during the parasite's peak of prevalence, and show a direct relation with the host's body condition, fecundity, and/or mortality.

Study Area
The study area is in northeast Iceland centred on Lake Mývatn (65˚40' N 17˚00' W). The general topography is flat with rolling hills rising from the coast to 400-500 m above sea level at the southern border, 70 km inland. This relief is broken by isolated mountains, the highest (anatomical terms follow [37]). These six body measures (a-f) were highly correlated with each other. Factor 1 from a principle component analysis (PCA) was used as an index of body size. This Factor explained 61.4% of the variance in the original variables and was highly related to them (loadings: wing = 0.831; head + bill = 0.833; tarsus = 0.528; tarsus + midtoe = 0.647; sternum = 0.891; and sternum-coracoid = 0.899).
Scoring was used to quantify the abundance of the quill mite M. lagopus. We examined quills of seven feathers-the upper-wing primary coverts 4 and 5 and secondary flight feathers 3-7 (numbered distal to proximal). Each feather was scored: 0 = no mites, 1 10 mites present, and 2 > 10 mites present. Scores were summed to derive a value for each individual.
Collection and quantification of endoparasites. We removed the lower gastrointestinal tract. The small intestine and large intestine were separated from the ceca, placed in a zip lock plastic bag, and stored at -20˚C. Fecal material (1-2 g) was taken from the large intestine prior to freezing, or, if the large intestine was empty, the posterior part of the small intestine. The modified McMaster procedure described in [25] was applied to obtain quantitative values of oocysts of Eimeria muta and E. rjupa, and their identification was based on [44]. Blastocystis sp. were found, but not quantified.
Parasite measures. We examined the parasite community (i.e., all parasite species known for the Icelandic rock ptarmigan except Blastocystis sp. and Mesocestoides canislagopodis) for adult and juvenile hosts in all years of the study. For this, we used a comparative value for each bird consisting of the ranked parasite data where 1 was allotted to the lowest positive finding and midranks were used for ties [12]. The ranked values of each parasite species were summed depicting the abundance for endoparasites, ectoparasites, and all parasites for each bird. We further used parasite richness, the mean of the total number of parasite species of an individual host in each age group [48].
For the individual parasite species we used parasite prevalence and aggregation. Parasite prevalence was defined as the proportion of birds infected by a particular parasite species [48]. Parasite aggregation was illustrated by a measure that applies to the right skewed parasite distributions, the Discrepancy index D [49]. This is a dimensionless index that ranges between 0 and 1; the distribution becomes more aggregated as the value approaches 1.

Ptarmigan Population and Demographics
We collected three sets of population data: (1) spring densities of territorial males, (2) age ratios in spring, and (3) age ratios in late summer. From those, measures of fecundity and mortality were derived.
Ptarmigan spring densities. Territorial male ptarmigan were counted on six plots each spring starting in 1981. The total size of these plots was 26.8 km 2 (range 2.4-8.0 km 2 ). Each plot was surveyed once during 10-24 May. The survey was conducted on foot by at least two observers in the early morning (05:00-10:00) or late afternoon (17:00-24:00). The locations of territorial males as well as ptarmigan kills were plotted on a map. A "kill" was the remains of a ptarmigan killed and eaten after arrival on the census plot in spring. The main cause of death was predation by gyrfalcon (84%) [50]. The "freshness" of the kill was based on the state of the feathers. The total number of males in spring is composed of the sum of the number of territorial males censused and killed. Not all kills could be sexed, so to estimate the proportion of males we used the sex ratio of ptarmigan killed by gyrfalcons in spring on the study area (73% males) [51]. [52] provides a detailed description of the census plots and methods. The ptarmigan population abundance index used was the annual mean density of males on these six plots and covers the years 2004-2013.
Age ratios in Spring. Fully grown ptarmigan were aged based on pigmentation of the primaries [36]. Two age classes were recognized, first year birds (juveniles = juv) and older birds (adults = ad). Spring samples for aging were birds found dead (mostly killed by gyrfalcons), birds trapped for banding or birds photographed while flying using high speed cameras.
Age ratios in late Summer (fecundity). In the last week of July and the first week of August we searched actively on foot for ptarmigan. We distinguished between adults (males and females) and chicks according to size, color, and sound. The age ratio was calculated using total number of females observed and assuming that half of the chicks were females. This ratio was used as a measure of fecundity.
Mortality rates. Mortality rates were calculated according to [35]. For these calculations, a year is defined starting on 1 st May and ending 30 th April. Two mortality rates are recognized: (1) Z 2 or apparent adult mortality rate; (2) Z X,W or juvenile excess mortality (mortality that juveniles suffer in excess to adults).
The population abundance index and spring age ratios were used to estimate the Z 2 mortality rate, assuming that spring abundance was proportional to the total number of birds in the study area. It was assumed that any bird alive at the end of winter was either in its second-year or older at the end of the following winter, provided it survived. So, adult mortality from spring to spring was calculated as: where Y t = spring abundance index year t Y t−1 = spring abundance index year t ─ 1 p t 2 = fraction of adult birds in spring year t The Z X,W mortality rate describes mortality that first year birds suffer from 1 August to 30 April in excess to adult mortality. The age ratios in late summer and at the end of the following winter were used to estimate excess juvenile mortality as: wherê p t;s 1 = fraction of juvenile birds late summer p t 1 = fraction of juvenile birds at end of winter p t;s 2 = fraction of adult birds late summer p t 2 = fraction of adult birds at end of winter Juveniles share Z 2 with the adults or at least a rate that shows the same trend [51]. Accordingly, the total mortality rate of juveniles was approximated as the sum of the Z 2 and the Z X,W rates.

Statistical Analysis
We calculated annual prevalence and confidence intervals and discrepancy indices for each parasite species using the software QP web [53]. All other statistical analyses were performed using the software package R [54]. Tests were two-tailed and statistical significance was set at p 0.05.
We used generalized linear models (GLMs) to assess the relationship between body condition (response variables) and parasite measures (explanatory variables) of individual birds. Gaussian family with identity link was specified to examine parasite richness and abundance of all parasites, ectoparasites, and endoparasites. Binomial family with logit link was specified to examine prevalence of the individual parasite species. Alpha levels (p 0.05) were adjusted using Holm-Bonferroni corrections.
We used linear regression models to test whether and how trajectories of ptarmigan mortality, fecundity, and population density (response variables) are related with parasite measures (explanatory variables). For density, parasite measures were related with current year ptarmigan densities. The time lag was estimated by fitting regressions with different time lags. The degrees of freedom, n-3, were calculated taking into account the estimation of the time lag in addition to the slope and the intercept. Because ptarmigan density measures are from spring and parasite measures from autumn, there is a 0.5 year time difference between current year parasite numbers and ptarmigan densities, a 1.5 year difference with densities 1 calendar year ago, and a 2.5 year difference with densities 2 calendar years ago. For all models, each age group was examined individually.

Ptarmigan Demographics
Population density. There were peaks in ptarmigan numbers in 2005 (8.1 cocks/ km) and 2010 (7.9 cocks/ km), and lows in 2007 (4.0 cocks/ km) and 2012 (3.5 cocks/ km) (Fig 1). During the tenure of our parasite study, the ptarmigan population increased for 4 years and decreased for 3 years (Fig 1).
Body condition. Adults were in a better condition than juveniles, but the trajectories showed the same pattern, rising to a peak in 2009, falling sharply in 2010, and improving slightly in 2011 and 2012 (Fig 1).
Mortality rates. The Z 2 mortality rate changed in a regular fashion; it was high at the start of the study in 2005/2006, decreased to 2007/2008, increased to a peak in 2009/2010, and decreased again ( Table 1). The Z X,W mortality rate fluctuated in an irregular fashion over the course of the study (Table 1).
The parasite species were highly aggregated, and the aggregation index for the different parasite species except for E. muta changed little over the years of this study (Fig 3C). The aggregation index of E. muta in juveniles peaked in 2007, reached a low in 2010, and then increased again (Fig 3C).

Parasites and Ptarmigan Demographics
Population density. Endoparasites and in particular E. muta prevalence showed a significant positive relationship with ptarmigan population density in both age groups. E. muta aggregation in juveniles showed a significant inverse relationship with density. The best fit provided the model considering a time lag of 1 year (Table 3). That is, endoparasite abundance and E. muta prevalence peaked and E. muta aggregation was lowest 1.5 years after the peak in ptarmigan numbers (Figs 2 and 3). Surprisingly, T. tenuis prevalence in juveniles showed a significant negative relationship with ptarmigan population density, a 2 year time lag fitted best, but prevalence for this species was very low (Table 3).
Body condition. Body condition of adults was significantly negatively related with parasite richness, parasite abundance, and prevalence of E. muta and T. tenuis, but positively with C. caudinflata prevalence (Table 4). Body condition of juveniles was significantly related with parasite richness, parasite abundance, and all selected pathogenic parasite species except T. tenuis (Table 5).
Mortality rates. Annual adult mortality was significantly positively related with prevalence of E. muta and negatively with M. islandicus (Table 4, Fig 4). Annual juvenile mortality was significantly positively related with endoparasites and C. caudinflata prevalence, and marginally with E. muta prevalence (Table 5). Excess juvenile winter mortality was significantly positively related with C. caudinflata prevalence ( Table 5, Fig 5).
Fecundity. Fecundity was significantly positively related with prevalence of A. lagopi in both age groups and marginally with E. muta prevalence in juvenile birds (Tables 4 and 5, Fig 6).

Discussion
The Anderson and May model [4] [31] identified low aggregation of parasites within the host population, parasite-induced host mortality and reduction in host reproductive potential, respectively, as well as time delays in parasite reproduction and transmission as the major regulating or destabilizing qualities that parasites can have on host population dynamics. Our results indicate that the eimerid E. muta parasitizing rock ptarmigan in Iceland fulfils all of these conditions. Aggregation levels of E. muta fluctuate inversely with its prevalence, with low aggregation at peak prevalence and vice versa, and both prevalence and aggregation track ptarmigan density with a 1.5 year time lag, respectively. Further, E. muta is associated with poorer body condition, increased mortality, and reduced fecundity (Fig 2). For red grouse in Scotland, coupled parasite-host cycles have been described and tested experimentally [55] [56]. In this grouse population exhibiting 4-8 year cycles, Trichostrongylus tenuis reduced fecundity and survival and there was a density dependent relationship between grouse numbers in one year and worm burdens in the subsequent year [55]. The fairly low levels of parasite aggregation observed in this system increased the tendency of the system to oscillate [55]. Alternatively, ptarmigan body condition after the peak in abundance nosedives for other reasons than parasites making the birds more susceptible to parasitism. Here, the chicken versus the egg question arises: are ptarmigan in poor body condition more susceptible to Eimeria or does the parasite cause ptarmigan to be in poor body condition? The Anderson-May model suggests for parasites to have a regulating effect on host numbers, it needs the parasite to affect the host which is what our data implies for E. muta. However, our study is correlational and so we do not know if E. muta is sufficiently virulent in itself or if it becomes virulent only or increasingly when the system is already enfeebled due to other factors, or possibly due to the combined parasite community. Parasite richness, the overall parasite community, and majority of chosen pathogenic parasites were all correlated with ptarmigan condition directly (Tables 4 and 5). Whatever the initial trigger may be, our data implies that there is a strong relation between ptarmigan condition, population density, and particularly the eimerid E. muta. Ptarmigan eimerids have a direct life cycle with oocysts shed in feces, sporulation in the environment, and infection through ingestion. The oocysts are present in the feces year round, but prevalence of E. muta peak between October and January [33]. Eimerids generally are host specific [57] and we assume E. muta (and also Eimeria rjupa) will not persist in species other than rock ptarmigan. We know the prepatent period of Eimeria varies between 4 and 6 days Table 3. Results from linear regression models between spring densities (2004-2012) and selected pathogenic parasites (2006-2012) of adult and juvenile rock ptarmigan in northeast Iceland. Density and parasite trajectories were regressed without (t) and with time lag, shifting the parasite trajectories 1 (t-1) and 2 (t-2) years back in time. Best fits are shown.  [58]. Thus, oocysts shed by ptarmigan in the first week of October are caused by infections occurring by mid September when the ptarmigan are moving to autumn habitats in alpine areas. Host density-dependent shedding of oocysts and their subsequent persistence in the environment from one year to the next could be the reason for the observed time-lag between ptarmigan numbers and E. muta prevalence. Ptarmigan hatched in the two years succeeding the peak in their numbers should be exposed to the maximum number of infective oocysts in the environment. Environmental persistence is well known for Eimeria, infective oocysts of some species can survive up to 602 days in soil [59], stand repeated freeze and thaw cycles [60], and this life history characteristic is sufficient to maintain eimerid populations [61]. Life history characteristics of parasites including the time needed for helminth larval stages to grow to maturity after infection and arrested development of the helminth larvae, have been given as explanations for time lags between host and parasite populations in similar systems [55] [62]. In addition to the severalfold relationship between population parameters and E. muta, we found a close relationship between juvenile excess winter mortality and annual juvenile mortality (a measure that partly consists of excess winter mortality) as well as C. caudinflata prevalence (Fig 5). Capillaria species are known to cause severe symptoms such as diarrhoea, weakness, weight loss, and a drop in egg production (capillariasis) [3]. C. caudinflata has an indirect life-cycle with earthworms as intermediate host [63]. Ptarmigan chicks have a mixed diet of plants and invertebrates such as earthworms, but adult birds eat mainly plants [64]. It is of interest that adult birds in our sample show C. caudinflata infections. The life span of C. caudinflata is c. 10 months [65]. This suggests that some of the adults could be carrying infections acquired as juveniles, but high prevalence among adults suggests that also members of this cohort get infected. In Icelandic ptarmigan, prevalence of C. caudinflata eggs in feces peaks in October and January [33], suggesting peaks of intestinal worm burden and/or worm reproduction during these times of the year. There is an approximate four-month time difference between ptarmigan hatch and the first peak in C. caudinflata egg shedding in October. The data suggest that juveniles suffer more from C. caudinflata infections than adults, i.e. the relationship with the mortality rates; C. caudinflata infections seem to be one of the drivers of the Z X,W rate (mortality in August through April). That juveniles suffer more from C. caudinflata infections than adults may have to do with varying levels of immune function to resist or treat infections [66]. The reverse relation between fecundity and the amblyceran chewing louse Amyrsidea lagopi prevalence is of interest (Fig 6). It is well known that feather lice can be severely damaging to their host (summaries of various studies in [67]), including by reducing fecundity [68] [69] [70]. The amblyceran mallophagan Menacanthus stramineus in high intensities, for instance, was related with reduced egg production [71]]. Contrary in other studies, controlled experiments did not show any effects of lice on reproductive success of swifts Apus apus and rock doves Columba livia [72] [73]. In Icelandic ptarmigan, A. lagopi is strongly associated with the creation of feather holes [30]. If this association happens through trade-offs between host reproduction and self-maintenance or for other reasons needs to be tested experimentally. A study on collared flycatchers Ficedula albicollis showed that feather wear of parents are tradedoff by parental activity, and that the degree of feather wear was associated with survival of the flycatchers [74].

Adult ptarmigan
The reverse relation between adult mortality and prevalence of Metamicrolichus islandicus is peculiar. When mortality is high, then fewer birds in the population are infested with this parasite that can cause mange. We do not have an explanation for this.
The helminth Trichostrongylus tenuis is a pathogenic nematode that was associated with reduced body condition in willow ptarmigan in Norway [12] and has been shown to be the determinant of the red grouse cycle [56]. In our study, this nematode occurred in such low numbers ( 28 worms) that it seems improbable that this parasite could have caused serious harm [29]. Yet, this parasite did show an inverse relation with body condition in adult ptarmigan and the 1.5 year time lag is near significant for both age groups.
The combined parasite community correlated with ptarmigan condition, but only endoparasites were also positively correlated with annual mortality in juveniles, probably mostly attributable to E. muta. Comparatively, though none of the parasites in Norwegian willow ptarmigan had a significant impact on their own, the parasite community was negatively related with host fitness, and this in turn was suggested to promote effects on host body mass and breeding mortality [12].
Population size of the main resident predator, the gyrfalcon, changes also in a delayed density dependent manner with the ptarmigan population which has been suggested to be the main driver of the ptarmigan population cycles [18]. So, parasitism in the ptarmigan population may act directly and/or sublethal parasitism may act synergistically with predation by acting upon different population parameters [63] or by making the ptarmigan more prone to gyrfalcon predation [24] [75] [76].