INTRODUCTION
The co-introduction of exotic hosts and their parasites is contributing to the rise of infectious disease in native species of conservation concern (Daszak et al. Reference Daszak, Cunningham and Hyatt2000; Peeler et al. Reference Peeler, Oidtmann, Midtlyng, Miossec and Gozlan2010; Peeler and Feist, Reference Peeler and Feist2011). Spillover, or ‘enemy addition’ (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015), can occur if introduced hosts serve as sources of co-introduced parasites that can infect native species (Prenter et al. Reference Prenter, Macneil, Dick and Dunn2004; Peeler and Feist, Reference Peeler and Feist2011; Britton, Reference Britton2013; Sachman-Ruiz et al. Reference Sachman-Ruiz, Narváez-Padilla and Reynaud2015; Sheath et al. Reference Sheath, Williams, Reading and Britton2015). It remains unclear, however, whether parasitism of introduced hosts drives infection of native hosts following initial co-introduction. Parasitism of native species by co-introduced parasites has been attributed to the distribution and densities of introduced hosts (Gozlan et al. Reference Gozlan, St-Hilaire, Feist, Martin and Kent2005; Peeler and Feist, Reference Peeler and Feist2011), but an emerging literature suggests that, with time, infection of native species by co-introduced parasites may have little or nothing to do with introduced hosts (Kirk, Reference Kirk2003; Haddaway et al. Reference Haddaway, Wilcox, Heptonstall, Griffiths, Mortimer, Christmas and Dunn2012, Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). For example, shifts in host preference can result in greater prevalence of co-introduced parasites in novel host species (Frankel et al. Reference Frankel, Hendry, Rolshausen and Torchin2015). Additionally, co-introduced parasites sometimes achieve higher intensities (Kirk, Reference Kirk2003) and increased virulence (Haddaway et al. Reference Haddaway, Wilcox, Heptonstall, Griffiths, Mortimer, Christmas and Dunn2012) in native species. Co-introduced parasites also can become established in native hosts with the capacity to complete their life cycle independent of the introduced host (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015).
Streams across the Hawaiian archipelago have long served as a natural laboratory for examining conditions that influence infection of native and introduced hosts by co-introduced parasites. As a consequence of extensive species introductions and naturally depauperate communities of hosts and parasites (Font and Tate, Reference Font and Tate1994; Font, Reference Font2003), the non-native intestinal nematode Camallanus cotti (Fujita, 1927) has become the most widespread and abundant parasite of Hawaiian stream fishes (Font, Reference Font2003; Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). Camallanus cotti is a generalist parasite with a complex life cycle; a range of fishes can serve as definitive hosts from which mature female parasites release first stage larvae into the freshwater environment, after which larvae are consumed by cyclopoid copepods that serve as an intermediate host (Font and Tate, Reference Font and Tate1994; Font, Reference Font2003). There is some experimental evidence that suggests C. cotti larvae can be directly transmitted, although this has never been documented in the wild (Levsen and Jakobsen, Reference Levsen and Jakobsen2002). Although C. cotti is native to Asia, where it primarily infects cyprinid fishes (e.g. carps, true minnows, and relatives), poeciliids (e.g. guppies, swordtails) introduced from Texas (USA) are thought to have served as the primary vector of introduction to Hawai`i (Font and Tate, Reference Font and Tate1994; Gagne, Reference Gagne2015). Since being introduced, C. cotti has infected four of the five native stream fishes (Awaous stamineus, Lentipes concolor, Eleotris sandwicensis and Stenogobius hawaiiensis, but not Sicyopterus stimpsoni) (Font and Tate, Reference Font and Tate1994). The pathology of C. cotti in native fishes is not known, but C. cotti infection causes apathy, reduced libido, haemorrhage and erosion of the rectum mucosa in other species (Menezes et al. Reference Menezes, Tortelly, Tortelly-Neto, Noronha and Pinto2006). Thus it is considered a potential threat to native fishes in Hawaiian streams (Font, Reference Font2003).
Recent work indicates that parasitism of native Hawaiian stream fishes by C. cotti is no longer associated with introduced hosts (Gagne and Blum, Reference Gagne and Blum2015; Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). For example, parasitism of the endemic goby A. stamineus by C. cotti extends beyond the distribution of introduced hosts and appears to reflect a complex set of factors rather than poeciliid densities (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). Although this suggests that native fishes are capable of sustaining C. cotti in Hawaiian streams, the extensive variation in parasitism across the archipelago indicates that native hosts may not provide stable transmission pathways (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). In contrast, parasitism in poeciliids does not appear to differ significantly across years (Vincent and Font, Reference Vincent and Font2003b ), suggesting that coincident introduced hosts provide stable transmission pathways that increase transmission and sustain infections of C. cotti in native fishes. If so, then parasitism of native hosts might be amplified by, or covary with, parasitism of coincident introduced hosts. Conversely, introduced hosts could reduce parasitism in a native host via the dilution effect, where increased diversity lowers parasite intensity in a focal species due to uptake of parasites by other available hosts (Keesing et al. Reference Keesing, Holt and Ostfeld2006).
In this study, we examined the abundance, intensity, and prevalence of C. cotti in populations of native and introduced hosts across the Hawaiian archipelago. We examined spatial variation to assess whether these parameters reflect host availability and other abiotic and biotic factors across watersheds. Specifically, we examined: (1) whether parasite abundance, intensity and prevalence differ between coincident and non-coincident native and introduced hosts, respectively; and (2) whether biotic and abiotic factors associated with parasitism are equivalent for native and introduced hosts. We also examined temporal variation because parasite abundance, intensity and prevalence can track shifts in conditions over time, which can sometimes manifest as stochastic outbreaks or cyclical variation in parasitism (Schall and Marghoob, Reference Schall and Marghoob1995; Glenn and Pugh, Reference Glenn and Pugh2006; Violante-González et al. Reference Violante-González, Aguirre-Macedo and Vidal-Martínez2008; Lachish et al. Reference Lachish, Knowles, Alves, Wood and Sheldon2011). Examining parasite abundance, intensity and prevalence, in space and time not only offered further perspective on the conditions that promote the success of co-introduced parasites, but it also provided a novel lens for understanding the potential value of remediation actions, such as the removal of introduced hosts, in stemming the rise of infectious disease in species of conservation concern like native Hawaiian stream fishes.
METHODS
Sample collection and laboratory methods
In 2009 and 2011, we sampled 42 watersheds on the five Hawaiian Islands containing perennial streams. Between two and four sites were sampled in each watershed. In accordance with permits reflecting conservation concerns, a total of 970 A. stamineus (2009, n = 562; 2011 n = 408) and 929 poeciliids (2009, n = 630; 2011, n = 323; Poecilia reticulata, n = 744; Poecilia sphenops, n = 123; Xiphophorus hellerii, n = 62) were collected from all watershed using hand nets (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). All specimens were necropsied for intestinal macroparasites using a dissecting microscope following Hoffman (Reference Hoffman1999).
Parallel datasets on parasite abundance, intensity and prevalence as well as potential abiotic and biotic factors were developed for the 2009 and 2011 samples following Gagne et al. (Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). Biotic factors measured included the population density of both introduced and native hosts, as well as host length. Abiotic factors included watershed land use and water chemistry (soluble reactive phosphate, ammonium, nitrate, total nitrogen, total suspended solids, total dissolved solids, water temperature and conductivity). Characterization of watershed land use was only assessed once, corresponding to the 2001 National Land Cover Dataset (Homer et al. Reference Homer, Dewitz, Fry, Coan, Hossain, Larson, Herold, McKerrow, VanDriel and Wickham2007). Estimates of genetic diversity were excluded from all analyses because genetic variation was not identified as an important indicator of C. cotti abundance, intensity or prevalence in A. stamineus and because genetic data were not available for introduced poeciliids (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). Watershed area and mean annual rainfall were considered in addition to the other factors examined by Gagne et al. (Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015) because both can influence surface flow, which in turn can influence C. cotti intensity and abundance in A. stamineus (Gagne and Blum, Reference Gagne and Blum2015). Data on watershed area were obtained from the Atlas of Hawaiian Watersheds and Their Aquatic Resources (Parham et al. Reference Parham, Higashi, Lapp, Kuamo'o, Nishimoto, Hau, Fitzsimons, Polhemus and Devick2008). Mean annual rainfall from 1987 to 2007 was calculated for each watershed using the Online Rainfall Atlas of Hawai`i (Giambelluca et al. Reference Giambelluca, Chen, Frazier, Price, Chen, Chu, Eischeid and Delparte2013).
Data analysis
Three measures of parasitism were calculated for each site and year following Bush et al. (Reference Bush, Lafferty, Lotz and Shostak1997): abundance (mean number of parasites per fish); intensity (mean number of parasites per infected fish); and prevalence (percentage of fish infected with C. cotti) for each site.
We first assessed spatial variation and temporal stability of parasite abundance, intensity and prevalence in A. stamineus and all poeciliids, respectively, in 2009 and 2011. We also examined interspecific variation by comparing each measure between A. stamineus and all poeciliids across sites and years. Because the majority of poeciliids collected were guppies (P. reticulata) we also analysed guppies separate from other poeciliid species. Although we found C. cotti in A. stamineus from 36 of 38 watersheds and C. cotti in poeciliids from 19 of 22 watersheds (Table 1), site level analyses only included sites where ⩾3 individuals were scored for parasites (A. stamineus = 849 individuals, 35 watersheds, 73 sites; Poeciliids = 929 individuals, 39 watersheds, 44 sites). Finally, we performed comparisons between A. stamineus and poeciliids when co-occurring and when the other was not present. The reciprocal comparison also only included sites where ⩾3 individuals were collected and that had corresponding snorkel survey data (A. stamineus without poeciliids = 378 individuals from 36 sites; A. stamineus with poeciliids = 314 individuals from 37 sites; Poeciliids without A. stamineus = 165 individuals from nine sites; Poeciliids with A. stamineus = 757 poeciliids from 36 sites). Brunner and Munzel test, which has been shown to be a robust non-parametric approach for comparing parasite abundance and intensity between groups (Brunner and Munzel, Reference Brunner and Munzel2000; Neuhäuser and Poulin, Reference Neuhäuser and Poulin2004), was performed to compare parasite abundance and intensity between species and years. Chi-squared tests were performed to compare prevalence. Because poeciliid population densities were often much greater than those of A. stamineus, we also assessed the relative abundance of C. cotti in native and introduced hosts by multiplying mean abundance by host density, respectively. Thus relative abundance is a measure of the total number of parasites within a species (or categorical group of species) at a given site. We compared relative abundances between species using the Brunner and Munzel test.
Table 1. The abundance, intensity and prevalence of Camallanus cotti infecting Awaous stamineus (Awaous), poeciliids and Poecilia reticulata in 2009 and 2011 listed by watershed and grouped by island. The density of the host fish is also provided
Generalized linear mixed models (GLMM) were constructed using site level data to examine the influence of host body length (i.e. the length of individual fish), native (A. stamineus) and introduced (poeciliid) host population densities, watershed characteristics [area, % agricultural and urban land use (hereafter %ag-urb), mean annual rainfall] and water chemistry (PC1 and PC2 derived from a Principle Components Analysis of soluble reactive phosphorus, ammonium, nitrate, total nitrogen, total suspended solids, water temperature, conductivity and total dissolved solids) on each metric of parasitism in A. stamineus and poeciliids, respectively. To recognize non-independence among sites and years, models included random effects of year, sites within watersheds and watersheds within islands. Values of %ag-urb were arc-sine transformed. Poisson error distributions were used for models examining the effects of site-level attributes on C. cotti intensity and prevalence, and a negative binomial distribution was used for models of parasite abundance. GLMM analyses only included sites where ⩾3 A. stamineus or poeciliids were scored for parasites as well as where corresponding populations and ecological data was available. Consequently, parasite abundance and prevalence models for A. stamineus included 692 individuals from 29 watersheds containing 65 sites, and parasite intensity models (which consider only infected fish) included 59 sites. Models for poeciliids included 504 individuals from 15 watersheds and 26 sites. All analyses were conducted in R (R Core Team, 2012) using the glmmADMB package (GLMMs) (Skaug et al. Reference Skaug, Fournier, Nielsen and Magnusson2006) and the lawstat package (Brunner and Munzel test) (Gastwirth et al. Reference Gastwirth, Gel, Hui, Lyubchich, Miao and Noguchi2013).
RESULTS
Parasitism of A. stamineus
Taking both years into account, 43·8% of the sampled A. stamineus were infected by C. cotti, exhibiting an average abundance (i.e. worms/fish) of 2·77 and an intensity (i.e. worms/infected fish) of 6·33 (Table 1). Parasite abundance, intensity and prevalence were significantly higher in 2009 than in 2011 (abundance, P = 0·0048; intensity, P < 0·0001; prevalence, P < 0·0001; Fig. 1, Table 1). Notably, evidence of parasitism was found in Hanakapiai in 2011, but not in 2009 or any prior surveys (Font and Tate, Reference Font and Tate1994; Font, Reference Font2003; Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). Alelele watershed on Maui was the only watershed with no infected A. stamineus throughout both years of the study (n = 29 individuals; Table 1).
Fig. 1. (A) The abundance, intensity, and prevalence of co-introduced Camallanus cotti infecting native Awaous stamineus in 2009 and 2011. Whiskers are drawn following the Tukey method. (B) The abundance, intensity, and prevalence of co-introduced Camallanus cotti infecting introduced poeciliids in 2009 and 2011. Whiskers are drawn following the Tukey method.
Although some shifts occurred, the watersheds with the highest abundance, intensity and prevalence of C. cotti infecting A. stamineus remained relatively consistent between sampling years (Fig. S1, Table 1). Notably, in 2009 three watersheds on Molokai (Pelekunu, Wailau, and Waikolu) registered abundance and intensity values that were three to five times greater than any other watershed sampled in either year (Fig. S1, Table 1). Despite large reductions in abundance and intensity, Pelekunu, Wailau and Waikalu watersheds remained among the five watersheds harbouring the most heavily parasitized individuals in 2011 (Fig. S1, Table 1). A similar, but less asymmetric, pattern was observed for estimates of prevalence (Fig. S1, Table 1).
Parasitism of Poeciliids
Across all watersheds and years, 31·7% of poeciliids were infected by C. cotti, with an average abundance of 0·90 and an intensity of 2·85. The abundance and prevalence of C. cotti infecting poeciliids were significantly higher in 2009 than in 2011 (P < 0·001, Fig. 1, Table 1). The watersheds that contained poeciliids without C. cotti infections were (1) Kamalo Gulch on Molokai, where C. cotti also were not found in co-occurring A. stamineus; (2) Waimanalo on Oahu, where A. stamineus were infected; and (3) Waiaha on Hawai`i, where A. stamineus were absent. The highest abundance, intensity and prevalence of C. cotti occurred in poeciliids sampled from a diversion canal adjacent to Waihee stream on Maui (Table 1). Waiulaula, which sometimes exhibits discontinuous surface flow, had the highest abundance, intensity and prevalence of C. cotti in poeciliids sampled from a stream channel (Table 1). Consistent with previous studies (Vincent and Font, Reference Vincent and Font2003a ), we found that P. reticulata had the highest abundance, intensity and prevalence of C. cotti although the levels were generally representative of parasitism of the full complement of introduced poeciliids sampled in the study (Fig. 2).
Fig. 2. (A) The abundance and relative abundance of co-introduced Camallanus cotti infecting native Awaous stamineus, all poeciliids and Poecilia reticulata, with error bars representing 95% confidence intervals (B) The intensity and prevalence of co-introduced C. cotti infecting native A. stamineus, all poeciliids, and P. reticulata, with error bars representing 95% confidence intervals.
Comparison of parasitism between native and introduced hosts
Abundance, intensity and prevalence of C. cotti were all greater in A. stamineus than in co-occurring poeciliids (abundance and intensity, P < 0·001; prevalence, P = 0·002; Fig. 2). Parasite abundance, intensity and prevalence in A. stamineus were also significantly greater than in P. reticulata, the most common poeciliid species collected (abundance, intensity and prevalence, P < 0·001; Fig. 2). Relative abundance, which accounts for the density of the host species, was significantly greater for C. cotti in poeciliids (relative abundance = 0·827) than A. stamineus (relative abundance = 0·518, P = 0·038, Fig. 2).
Parasite abundance and prevalence in A. stamineus was significantly greater when poeciliids were not present (abundance without poeciliids = 4·23, s.e. = 0·58, abundance with poeciliids = 1·82, s.e. = 0·24, P = 0·006; prevalence without poeciliids = 50%, s.e. = 2·6%, prevalence with poeciliids = 41%, s.e. = 2·8%, P = 0·024, Fig. 3). Differences in parasite intensity approached significance (without poeciliids = 8·41, s.e. = 1·07; with poeciliids = 4·48, s.e. = 0·51, P = 0·059). Parasite abundance, intensity and prevalence in poeciliids also were significantly greater when A. stamineus were not present (abundance without A. stamineus = 1·43, s.e. = 0·25, abundance with A. stamineus = 0·73, s.e. = 0·07, P = 0·006; intensity without A. stamineus = 3·69, s.e. = 0·54,; intensity with A. stamineus = 2·65, s.e. = 0·18, P < 0·001; prevalence without A. stamineus = 39%, s.e. = 3·8%, prevalence with A. stamineus = 29%, s.e. = 1·6%, P = 0·012, Fig. 3).
Fig. 3. Abundance, intensity and prevalence within poeciliids when coincident with Awaous stamineus and non-coincident with A. stamineus, and the abundance, intensity and prevalence within A. stamineus when coincident with poeciliids and non-coincident with poeciliids.
Predictors of parasitism
The intensity and abundance of C. cotti in A. stamineus were positively related to host length (Table 2). Additionally, parasite intensity in A. stamineus was negatively associated with poeciliid density and positively associated with water chemistry PC1(Table 2), which explained 54·2% of the variation and contained strong loadings (>0·6) for all three nitrogen forms, conductivity, total dissolved solids and total suspended solids (only soluble reactive phosphate loaded strongly on to water chemistry PC2). No variables were significantly related to parasite prevalence in A. stamineus. GLMM for parasite abundance and prevalence in poeciliids recovered positive relationships with land use (% ag-urb; Table 2) and models for parasite intensity recovered a negative relationship with water chemistry PC1.
Table 2. GLMM model outcomes for the abundance, intensity and prevalence of Camallanus cotti infecting native Awaous stamineus and poeciliids according to: host length (Length), agricultural and urban land use (% Ag-urb), density of A. stamineus (Density Awaous), the density of poeciliids (Density Poeciliids), water chemistry principle components one (PC1) and two (PC2), watershed area (WS-area) and average watershed rainfall (Rainfall). Models were run with a combined 2009 and 2011 dataset with year as a random factor. Sites and watersheds were also included as random factors
Bold indicates significant at P < 0·05. Est, estimate + direction of effect; s.e., standard error; P, significance (P).
DISCUSSION
Following initial spillover, infection of native species by co-introduced parasites may or may not be influenced by parasitism of introduced hosts. In prior work (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015), we showed that A. stamineus have become infected with the co-introduced parasitic nematode C. cotti in remote watersheds that do not harbour introduced poeciliid hosts, suggesting that parasitism in native fish across the Hawaiian archipelago is no longer contingent on poeciliids serving as reservoirs. The current study provides further support for this inference. Not only did we find that C. cotti abundance, intensity and prevalence are greater in native than introduced hosts, we found that parasite intensity in native A. stamineus are inversely related to poeciliid density. Additionally, we found that infection levels in A. stamineus are largely consistent (i.e. sustained) in the absence of introduced poeciliids over time. Evidence that parasite abundance and intensity is lower in the presence of poeciliids also suggests that introduced poeciliids hosts reduce rather than elevate parasitism of native fish by C. cotti.
Parasitism of native vs introduced hosts
The higher abundance, intensity and prevalence of C. cotti in native A. stamineus relative to introduced poeciliid hosts possibly reflects differences in size and lifespan of host species. Native A. stamineus generally have longer lifespans (i.e. 3–4 years) and are larger (i.e. adults are ⩾100 mm) than introduced poeciliid hosts (i.e. ⩽1·5 years, adults are ⩽60 mm) (Reznick et al. Reference Reznick, Bryant and Holmes2005; Hogan et al. Reference Hogan, McIntyre, Blum, Gilliam and Bickford2014). A longer lifespan allows for the accumulation of higher parasite intensities as a result of greater exposure (Price and Clancy, Reference Price and Clancy1983). In addition, food consumption by host fish generally increases with body size, thereby raising the odds of consuming an infected intermediate host (Lo et al. Reference Lo, Morand and Galzin1998). Contrasting feeding preferences may also exacerbate differences in infection risk over time as poeciliids have ontogenetic dietary variation, which has not been observed in A. stamineus (Hutchings et al. Reference Hutchings, Athanasiadou, Kyriazakis and Gordon2003; Vincent and Font, Reference Vincent and Font2003a ; Blum et al. Reference Blum, Gilliam and McIntyre2014).
Abundance of C. cotti in A. stamineus is notably much higher than for hosts within the native range of C. cotti, which suggests that native Hawaiian stream fish are naïve hosts that lack defences against infection (Smith et al. Reference Smith, Acevedo-Whitehouse and Pedersen2009). Abundance is < 0·85 in all host species in the native range of C. cotti (Wu et al. Reference Wu, Wang, Gao, Xi, Yao and Liu2007) compared with 2·77 in A. stamineus and 0·9 in poeciliids in Hawaiian streams. Native Hawaiian stream fish, which are all endemic to the archipelago (Lindstrom et al. Reference Lindstrom, Blum, Walter, Gagne and Gilliam2012), have likely evolved without the pressures imposed by a complex parasite assemblage. This may translate into comparatively weak defensive responses to parasite infection (Font, Reference Font2003). For example, naïve hosts may not show dietary adjustments or other behavioural responses to reduce risk of exposure (Hutchings et al. Reference Hutchings, Athanasiadou, Kyriazakis and Gordon2003), such as ontogenetic shifts in feeding preferences like those exhibited by guppies, which progressively feed less at stream margins where intermediate copepod hosts are most plentiful (Vincent and Font, Reference Vincent and Font2003a ). Thus, native Hawaiian stream fish may feed more heavily upon cyclopoid copepods (the intermediate host of C. cotti) than introduced fishes. Native fish may also be more susceptible to infection due to lower immunocompetency, as has been found in other fishes with naturally low exposure to parasites (Scharsack et al. Reference Scharsack, Kalbe, Harrod and Rauch2007).
We found striking evidence of a mutual dilution effect. Co-occurrence and higher densities of alternative hosts appear to reciprocally reduce infection of native and introduced hosts by C. cotti. As in our prior research (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015), we found an inverse relationship between parasite intensity in A. stamineus and introduced host densities. We also found that parasite intensity, abundance and prevalence in A. stamineus was reduced when poeciliids were present and that parasite intensity, abundance and prevalence in poeciliids was reduced when A. stamineus were present. This suggests that co-occurrence and greater densities of native and introduced fishes dilute parasitism by increasing the overall availability of definitive hosts. Increased host diversity has been shown to reduce infection via the dilution effect, whereby increased diversity reduces infection due to the presence of dead end hosts (Keesing et al. Reference Keesing, Holt and Ostfeld2006). Our findings indicate that higher densities and diversity of viable definitive hosts also serve to reduce infection of each species present, with the dilution effect caused by a wider range and number of hosts able to take up parasites from a finite pool (Johnson and Thieltges, Reference Johnson and Thieltges2010). Thus, the relationship of reduced parasite abundance and intensity in native Hawaiian stream fish with increasing densities of poeciliids could be a result of poeciliids harbouring parasites that might otherwise infect native hosts. This inference is also supported by our finding that poeciliids cumulatively host a comparably larger number of parasites relative to A. stamineus. If so, then following spillover from an initial introduction of poeciliids into a watershed, the influence of introduced hosts on parasitism of native hosts by co-introduced parasites might shift from negative to positive.
Abiotic and biotic influences on parasitism
Parasite burdens may vary if susceptibility to conditions that promote infection differ among hosts (Dunn and Dick, Reference Dunn and Dick1998; Plowright et al. Reference Plowright, Sokolow, Gorman, Daszak and Foley2008). Our results indicate that parasitism of native and introduced hosts by C. cotti reflects different biotic and abiotic factors known to influence susceptibility to infection. For example, a positive relationship between host length and parasite abundance and intensity were observed only in native A. stamineus, which likely reflects the accumulation of parasites over time in A. stamineus (Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015) as opposed to poeciliids (Vincent and Font, Reference Vincent and Font2003a ). The positive association detected between agricultural and urban land use and parasite abundance and prevalence in poeciliids could be an outcome of fewer co-occurring alternative hosts, because A. stamineus is less tolerant of stream conditions associated with intense watershed land use (Blum et al. Reference Blum, Gilliam and McIntyre2014).
Our results also indicate that variation in parasitism of native and introduced hosts can be a consequence of different responses to the same factor. Water chemistry PC1 was found to be associated with infection of both native and introduced hosts by C. cotti; however, it was positively associated with parasite intensity in A. stamineus and negatively associated with parasite intensity in poeciliids. Water chemistry PC1 largely corresponded to nitrogen variability. In other host–parasite systems, nutrients drive parasitism of definitive hosts by influencing the prevalence of intermediate hosts (Johnson et al. Reference Johnson, Carpenter, Ostfeld, Keesing and Eviner2008). In Hawai`i, nutrient-driven shifts in intermediate host populations would likely result in parallel rather than contrasting trends in parasitism of native and introduced definitive hosts (i.e. as the abundance of intermediate hosts increases, so does parasitism of all definitive hosts), which suggests that other outcomes of nitrogen variability are shaping parasitism of definitive fish hosts in Hawaiian streams. Stable isotope and gut content analyses indicate that the positive association between nutrients and parasite intensity in A. stamineus reflect changes in diet, where greater carnivory (i.e. consumption of copepods) in nutrient-enriched streams elevate risk of infection (Blum et al. Reference Blum, Gilliam and McIntyre2014). Not much is known about dietary preferences of introduced poeciliids under different nutrient regimes, but it is possible that greater competition with native fish for invertebrate prey could reduce parasite intensity in poeciliids in nutrient-rich streams (Holitzki et al. Reference Holitzki, MacKenzie, Wiegner and McDermid2013).
With some notable exceptions, the set of factors found to influence parasitism of A. stamineus by C. cotti were consistent with those identified in prior studies (Gagne and Blum, Reference Gagne and Blum2015; Gagne et al. Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015). Gagne et al. (Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015), for example, found that host size and conditions associated with nutrient loading were positively associated with parasite abundance and intensity in A. stamineus based on a single year of data. Gagne et al. (Reference Gagne, Hogan, Pracheil, McIntyre, Hain, Gilliam and Blum2015) also found, however, that parasite intensity and prevalence tracked A. stamineus population density. We did not find the same relationship in the current study, which might be due to temporal shifts in parasite abundance, intensity, and prevalence in watersheds harbouring some of the highest densities of A. stamineus across the archipelago. A comparison across a natural rainfall gradient and a broader comparison among islands also found that parasite abundance, intensity and prevalence increase with declining precipitation (Gagne and Blum, Reference Gagne and Blum2015). Average rainfall was lower and parasitism was higher in 2009, but we did not detect a statistically significant relationship with rainfall, possibly because our rainfall estimates were based on 30-year averages rather than annual measures, which are not readily available for each watershed (Giambelluca et al. Reference Giambelluca, Chen, Frazier, Price, Chen, Chu, Eischeid and Delparte2013).
Temporal changes in parasitism
Although we found higher levels of parasite abundance, intensity and prevalence in A. stamineus in 2009 compared with 2011, C. cotti infections were relatively stable in the majority of watersheds sampled over the study period. This is consistent with evidence of stable levels of parasite abundance and prevalence in poeciliids in one watershed on Oahu over 5 years (Vincent and Font, Reference Vincent and Font2003b ). Nonetheless, we observed notable shifts in parasite abundance, intensity and prevalence in three watersheds on the remote north shore of Molokai. The transient spike on Molokai may have corresponded to an epizootic event, which can increase host mortality (Kennedy et al. Reference Kennedy, Shears and Shears2001); periods of increased parasitism are often driven by environmental and host population dynamics and subside following a crash in host populations (Kennedy et al. Reference Kennedy, Shears and Shears2001; Briggs et al. Reference Briggs, Knapp and Vredenburg2010; Heins et al. Reference Heins, Baker and Green2011). Consistent with this, we found that densities of A. stamineus declined in these watersheds between 2009 and 2011, but we cannot say whether the declines are attributable to parasite-induced mortality. Parasite abundance and prevalence within poeciliids also was greater in 2009 than 2011, suggesting that yearly differences in parasitism could be indicative of regional rather than local shifts in environmental conditions (e.g. rainfall and surface flow) over time. Further work to better characterize the magnitude and footprint of temporal variation in parasitism is warranted, as transient spikes could be detrimental to populations of A. stamineus, particularly if coupled with other stressors, such as habitat degradation.
In conclusion, the greater abundance, intensity, and prevalence of C. cotti in native vs introduced fishes, as well as higher parasite abundance, intensity, and prevalence in watersheds that lack introduced fishes, show that the incidence of parasitism in native species is no longer strictly determined by the distribution and densities of the original introduced hosts in the Hawaiian archipelago. Evidence of a dilution effect also indicates that the co-occurrence and prevalence of introduced hosts can ameliorate parasitism of native fish by co-introduced parasites, which runs counter to the prevailing notion that the presence of introduced species is always detrimental to native fish (Font, Reference Font2007; Holitzki et al. Reference Holitzki, MacKenzie, Wiegner and McDermid2013). Indeed, our findings suggest that efforts to control (i.e. remove) introduced poeciliids could inadvertently increase parasitism in native fish. Importantly, other benefits of removing introduced poeciliids (e.g. reduced resource competition) could outweigh any increases in parasitism. Thus a better understanding of responses of native fishes to introduced fish is warranted to support management decisions. A more thorough pathological assessment, for example, would help address this concern and related concerns, like the possibility of parasite-induced mortality from epizootic events or interactions with other stressors that threaten amphidromous Hawaiian stream fishes (Brasher, Reference Brasher2003; Walter et al. Reference Walter, Hogan, Blum, Gagne, Hain, Gilliam and McIntyre2012).
SUPPLEMENTARY MATERIAL
The supplementary material for this article can be found at http://dx.doi.org/10.1017/S0031182016001001.
ACKNOWLEDGEMENTS
We thank E. Childress, J. Fenner, G. Glotzbecker, T. Haas, J.D. Hogan, B. Lamphere, D.P. Lindstrom, K. Moody, D. Oele, T. Rayner, J. Rossa and R.P. Walter for field assistance, as well as J. Thore, E. Hamann, A. Gulacheski and S. Piper for laboratory assistance. We also thank H. Bart, C. Criscione and W. Font for valuable guidance. All collections were approved and conducted in accordance with Special Activities Permits issued by the Hawaii Division of Aquatic Resources.
FINANCIAL SUPPORT
This study was funded by Tulane University (to M.B. and R. G.) and the US Department of Defense Strategic Environmental Research Development Program (SERDP) through project RC-1646 (to M. B., P. M. and J. G.).
