Invasion of the body snatchers: the role of parasite introduction in host distribution and response to salinity in invaded estuaries

Abstract

In dynamic systems, organisms are faced with variable selective forces that may impose trade-offs. In estuaries, salinity is a strong driver of organismal diversity, while parasites shape species distributions and demography. We tested for trade-offs between low-salinity stress and parasitism in an invasive castrating parasite and its mud crab host along salinity gradients of two North Carolina rivers. We performed field surveys every six to eight weeks over 3 years to determine factors influencing parasite prevalence, host abundance, and associated taxa diversity. We also looked for signatures of low-salinity stress in the host by examining its response (time-to-right and gene expression) to salinity. We found salinity and temperature significantly affected parasite prevalence, with low-salinity sites (less than 10 practical salinity units (PSU)) lacking infection, and populations in moderate salinities at warmer temperatures reaching prevalence as high as 60%. Host abundance was negatively associated with parasite prevalence. Host gene expression was plastic to acclimation salinity, but several osmoregulatory and immune-related genes demonstrated source-dependent salinity response. We identified a genetic marker that was strongly associated with salinity against a backdrop of no neutral genetic structure, suggesting possible selection on standing variation. Our study illuminates how selective trade-offs in naturally dynamic systems may shape host evolutionary ecology.

1. Introduction

A complex network of biotic and abiotic forces drives the evolutionary trajectories of species [1]. The strength of these selective pressures can vary across distributional ranges, environmental gradients, ecological communities and, recently, with human-mediated global change [2]. Anthropogenic introductions have shifted range boundaries globally [3], creating novel selective forces that alter population demographics, community interactions, and ecosystem dynamics [4]. Such changes are particularly evident when invaders, like predators and parasites, strongly interact with other community members [5]. Parasites can mediate many aspects of their hosts' biotic interactions, and both losses and gains of parasites due to species invasions can alter community dynamics [6]. Indeed, parasites are strong drivers of evolutionary and ecological change, with their influences ultimately scaling up to the ecosystem level due to distribution and abundance shifts [6,7].

Often conceptualized as an arms race, hosts and parasites are under constant bidirectional pressure to evolve defences or offences [8]. For hosts, a typical adaptive response is the development of physiological and/or behavioural resistance to infection [9]. While potentially effective, such evolved defences can be energetically costly and take time [10]. In the meantime, infection can elicit harmful fitness effects, including reduced reproductive output and lessened competitive and predator-avoidance capabilities [11,12]. Yet, what if a host could avoid costly arms races altogether by eluding the parasite? For example, hosts could migrate (spatially or temporally) outside a parasite's influence or evolve broader abiotic tolerances to exploit environmental refugia intolerable to the parasite [1,13]. Indeed, adapting to exploit parasite-free refugia could be a more rapid and less costly approach to combatting strong parasite pressures when compared with the evolution of an enhanced or targeted immune response [14]. Such refuge exploitation may be especially likely in hosts pre-adapted to variable conditions—permitting a rapid response to changing environments [15] and ultimately allowing hosts to trade abiotic stress for the benefits of low- or no-parasite environments [1,13].

Currently, much of what we understand about evolved host response is based on controlled laboratory studies; yet in nature, hosts exist in rich and variable environments. Indeed, highly dynamic systems like estuaries have fluctuating biotic and abiotic forces (e.g. salinity, temperature, predators, parasites, competitors) across time and space [1,2]. Within estuaries, salinity is a particularly dominant physical driver of species distributions, and estuarine organisms have evolved salinity tolerances that reduce physiological (osmotic) stress and enhance survival and reproduction [16]. Estuarine species often possess highly plastic osmoregulatory capabilities [17], which may manifest in the short term as regulation of osmoregulatory genes responding to environmental salinity, and over evolutionary timescales as changes to protein structure and function to more effectively regulate ion concentrations [18]. Species in low and variable salinity environments may experience balancing selection on genes associated with osmoregulation, thereby maintaining a pool of standing genetic variation that may fuel future expansion into environments outside average salinity distributions [19].

While estuarine hosts may be capable of exploiting parasite-free refugia, any associated physiological costs could represent an adaptive trade-off [20]. A prominent example is the eastern oyster (Crassostrea virginica) exposed to two novel diseases, ‘Dermo’ (Perkinsus marinus) and ‘MSX’ (Haplosporidium nelsoni) in Chesapeake Bay [21]. There, parasite prevalence and intensity are positively correlated with salinity [1,21], and low-salinity refugia provide oysters a respite from mortality and sublethal impacts of disease but can also impede reproduction and evolution of disease resistance [1]. Studies like these are rare, in which biotic and abiotic selective pressures are simultaneously examined in the natural world; yet, this kind of contextualized research is pivotal for understanding the relationship between evolutionary change and ecological ramifications.

Here, we examined the interplay of abiotic and biotic selective pressures on host eco-evolutionary response using a host–parasite relationship in southeastern US estuaries. The host Rhithropanopeus harrisii (white-fingered mud crab), hereafter ‘RH’, is native to the Northwest Atlantic and Gulf of Mexico, where it is a generalist predator of small biota, prey to numerous vertebrate and invertebrate predators, and host to several parasites [22]. Its rhizocephalan parasite Loxothylacus panopaei, hereafter ‘LP’, is native to the Gulf of Mexico and southeast Florida but was absent from much of the Atlantic Coast until it invaded Chesapeake Bay in the 1960s and then quickly spread throughout the southeast [23]. This ‘body-snatching’ parasite is a highly modified barnacle that infects panopeid crabs (figure 1a) and causes permanent castration, eliminating individual fitness and exerting strong selective pressures on hosts [24]. Essentially, infected crab ‘zombies’ serve as vessels for the barnacle's reproduction, providing the adult parasite with an effective means of reproducing using a mobile host [12]. In contrast with the coevolved range where LP prevalence is often less than 10% [25], the impact of LP in recently invaded areas is striking: prevalence is upwards of 91% [23], and infection induces hosts to spend more time hiding and less time feeding, leading to significantly lower prey consumption [26,27]. On an ecosystem level, LP introduction causes broad shifts in community structure, driven by abundance declines in hosts [6]. Salinity is also important: in laboratory studies, LP larvae (infective stage) have a narrower salinity tolerance (8–30 practical salinity units (PSU)) than RH hosts (adults: 1–41 PSU; larvae: 2–40 PSU) [28,29]. However, prior to our study, there had been no explicit test of the relationship between host and parasite distribution along fine-scale salinity gradients, nor of the host's physiological response to low-salinity waters that provide a refuge from parasitism but may incur greater osmotic stress.

Figure 1.

(a) Infection cycle of Loxothylacus panopaei (LP) with hosts Rhithropanopeus harrisii (RH) and Eurypanopeus depressus depicted. In an infected crab, a mature (fertilized) externa of the parasite releases male and female larvae. Females infect the host, extend nutrient rhizoids inside the body and produce a virgin externa (unfertilized sac) on the crab's abdomen. The male fertilizes the female and the cycle begins anew [24] (diagram: Houghton, Blakeslee, Fowler). (b) Map of sample sites and estuary reaches in the Pamlico (triangles) and Neuse (squares) Rivers. Colours designate salinity levels (see Methods), and numbers correspond to sites in (c). The inset depicts the USA with a red box for study location. (c) Boxplot of mean (dashed horizontal) and median (solid horizontal) salinity levels by site. The vertical dashed line separates sites by parasite absence (1–6) and presence (7–11).

We combined field ecology, laboratory experiments, and transcriptomics in the extensive estuarine/lagoonal system of eastern North Carolina (NC) (figure 1b) to answer the following questions: (i) what drives parasite prevalence, host abundance, and associated taxa diversity from upper (near fresh) estuary populations to lower estuary reaches (moderate salinities)? (ii) What factors explain an individual crab's probability of infection throughout the estuary? (iii) How does salinity acclimation influence host physiological response to salinity, and does this response vary by source population? (iv) Is there genetic evidence for selection or population structure in host populations along salinity gradients? Ultimately, our study integrates evidence from a key host species up to the level of its ecosystem, illuminating the selective trade-offs that shape host evolutionary ecology in nature. By examining the parasite's invaded range, our study can investigate ‘evolution in action’ as the parasite shapes host response in a dynamic system.

2. Methods

(a) . Field surveys, processing, and analysis

From July 2016 to July 2019, we performed extensive surveys along salinity gradients in two NC estuaries (Pamlico = 6 sites; Neuse = 5 sites) (figure 1b). Sites were categorized using natural breaks in salinity distributions (ANOVA F_11,189 = 26.37; p < 0.0001). These salinity breaks included ‘Near Fresh’, ‘Very Low’, ‘Low’, ‘Low–Moderate’, and ‘Moderate’ (figure 1b,c). We surveyed parasite prevalence, host abundance, and community diversity by deploying two passive collectors per location. Collectors are small plastic crates (19 × 22 × 16 cm) filled with autoclaved oyster shell (replaced per sampling event) that provide standardized habitat to attract RH and other mobile and sessile organisms [25]. Collectors were attached to docks or the benthos in the near shore subtidal zone (approx. 1 m water depth) and left to passively recruit for six to eight weeks. All panopeids were collected for laboratory processing, while other organisms (electronic supplementary material, table S1) were counted, identified to lowest taxonomic level, and released. Per sampling event, we took temperature and salinity using a hand-held YSI (Yellow Springs, OH, USA) and deployed HOBO loggers (Onset Corp., Bourne, MA, USA) to record temperature once per hour per site (electronic supplementary material, figures S1 and S2) and salinity once per ½ hour at a subset of sites (electronic supplementary material, figure S3). Because salinity loggers could not be placed at all sites, we used YSI-recorded salinity in statistical analyses (see electronic supplementary material, Methods).

In the laboratory, all panopeids were identified, measured (carapace width (CW) to nearest 0.1 mm), and sexed (juvenile [less than 4 mm CW], male, female, ovigerous female). Panopeids included RH, Eurypanopeus depressus (flatback mud crab) and Panopeus herbstii (Atlantic mud crab). Of the 10 236 panopeids surveyed, the majority were RH (90.7%), followed by E. depressus (9.1%) and P. herbstii (0.2%). The latter two crabs are more common in higher salinity waters outside our survey region [24]. However, ED overlaps with RH in sites greater than 10 PSU; thus E. depressus was included in some analyses since it competes with RH and is an additional competent host for LP [25]. LP infection was determined by examining the crab's abdomen for visible virgin or mature parasite externae (figure 1a). Infection prevalence per species per location was calculated as the number of infected crabs divided by all crabs above 4 mm CW (i.e. the smallest size reported with a visible externa [25]). While a conservative measure, this is a standard approach for this system [23,25]. We separately examined infection prevalence of virgin and mature externae because they represent distinct developmental stages in the parasite's life cycle and were expected to respond differently to environmental factors.

Field data were analysed using R Studio with the following generalized linear mixed models (GLMMs): (i) binomial (logit link) model predicting infection prevalence of (a) mature and (b) virgin LP externa; (ii) negative binomial (log link) model predicting adult RH abundance; (iii) binomial (logit link) model predicting individual RH infection status. Correlations were evaluated (cut-off = 0.6), and explanatory variables were standardized. Random terms of site (n = 11) and salinity level (n = 5) were made factors. For each GLMM, the model with the highest AICc (corrected Akaike information criterion) weight (relative likelihood of candidate model among all tested models) was selected as the best model. Fit was determined, and model assumptions were checked. Detailed explanations of statistical analyses are in the electronic supplementary material, Methods.

(b) . Salinity exposure experiment

Uninfected adult RH were collected in January 2019 from three populations in the Pamlico River representing ‘Near Fresh’, ‘Very Low’, and ‘Low’ salinities (figure 2; electronic supplementary material, table S2). Crabs were acclimated for 3 days at their collection salinities at 20°C. Twenty crabs from ‘Very Low’ and ‘Low’ source populations were randomly divided into two groups and acclimated to 3 or 15 PSU. Crabs were ramped ±2.5 PSU every 24 h until reaching final experimental salinities. For ‘Near Fresh’, 10 crabs each were ramped to 3 and 15 PSU, with a third group started at 0.0 PSU and transitioned to 0.8 PSU after 5 days due to significant mortality (6/10 crabs died). After reaching experimental salinities, crabs were maintained for 7 days in individual chambers of an 18-well polycarbonate parts-box. Crab position was randomized across treatments, and boxes were rotated daily. Crabs were fed every 3 days (carnivore pellets), and water was changed 24 h post-feeding. Mortality was recorded daily (electronic supplementary material, table S3). Following the acclimation period, each crab was subjected to three replicate righting response trials to document possible sublethal impacts of acute salinity exposure [30]. For each trial, a crab was placed on its back and its righting time recorded. Immediately following the third trial, crabs were dissected, and the second posterior (primarily osmoregulatory) gills from the crab's left and right sides were preserved in RNALater and stored at −80°C until processing [31]. Time-to-right data were log transformed for normality and analysed using an ANOVA with sex, size, and treatment as fixed effects and time-to-right as response variable.

Figure 2.

Heatmap of salinity and temperature in predicting (a) mature and (b) virgin LP infection prevalence in RH populations. Observed prevalences are indicated with dots (larger sizes = higher prevalence). Infection primarily occurs greater than 10 PSU (vertical dashed line). (c) Prediction plots (15 PSU, 22°C) of E. depressus abundance on mature and virgin externae prevalences in RH populations. Data are grouped by day-of-year (DOY), another significant driver of prevalence in both models. Lines represent modelled output; dots are observed values. (Online version in colour.)

(c) . Transcriptome sequencing and assembly

RNA was extracted from gills using a modified TRI (TRIzol), reagent protocol, and cDNA (complementary DNA) libraries were constructed using Illumina TruSeq stranded mRNA sequencing kits. Samples were sequenced as 50 bp single-reads and 150 bp paired-end reads on an Illumina HiSeq 4000. Forty-six samples were used in transcriptomic analyses, with 4–9 individuals per source-by-treatment group (electronic supplementary material, table S3). Raw reads were trimmed and discarded if they were less than 20 bp (50 bp single-end reads) or less than 50 bp (150 bp paired-end reads). We assembled a de novo transcriptome with Trinity v. 2.11.0 [32], using forward reads from all 46 samples, including forward reads only from 150 bp paired-end sequencing. The resulting transcriptome had 140 196 contigs and 214 737 isoforms; we retained only the longest isoform per gene for downstream analysis. The transcriptome was annotated using EnTAP v. 0.9.1 with Uniprot's swissprot and TrEMBL protein databases and NCBI's (National Center for Biotechnology Information) nucleotide database as references [33]. We annotated 22 507 contigs; 1467 contigs identified as contamination were removed, leaving a de novo RH gill transcriptome of 138 729 contigs. For downstream analyses, paired-end reads were converted to 50 bp single-end reads by taking only the forward reads and cutting them to 50 bp before trimming as above. Additional details in the electronic supplementary material, Methods.

(d) . Differential gene expression and population structure

For gene expression, we mapped individual reads back against the de novo transcriptome with Salmon v. 1.3.0 [34] and used the R package ‘DESeq2’ v. 1.28.1 to examine expression patterns [35]. Our variables were Source (collection location) and Acclimation (salinity treatment in the laboratory); we treated Source (Near Fresh < Very Low < Low) and Acclimation (0.8 < 3.0 < 15.0 PSU) as categorical. We tested for expression changes associated with acclimation salinity using a Wald test, taking into account source population but not testing for interaction. We also ran this analysis of salinity-associated expression changes using only crabs acclimated at 3 and 15 PSU, excluding the 0.8 PSU treatment which contained four crabs all derived from the ‘Near Fresh’ site. We used GO_MWU in R to test for enrichment of Gene Ontology (GO) categories in acclimation salinity-associated contigs [36]. We used log-fold-change values for each contig's expression difference between the highest and lowest acclimation salinity tested. For the 3 and 15 PSU acclimation salinities, we also tested for genes whose response to salinity depended on the source. We then examined the interaction between Source and Acclimation using a Wald test. Within selected contigs, we ran t-tests on normalized count data to compare the significance of pairwise expression differences between sample groups.

For population structure and selection, we mapped individual sample reads back to the transcriptome using default settings in Bowtie2 v. 2.4.1 [37]. After mapping, single nucleotide polymorphisms (SNPs) were identified and called using the Genome Analysis Tooklit v. 4.1.8.1 [38]. We retained only SNPs with high-quality (greater than or equal to 20 Phred), high-coverage (greater than or equal to 5 reads) individual genotypes in at least nine individuals in each of the three sources. We further screened for SNPs in strong linkage disequilibrium (R² ≥ 0.8). All but one SNP in each cluster was removed, leaving 16 592 high-quality, high-coverage, independent SNPs. We examined population structure using smartPCA v. 16000 in EIGENSOFT [39]. We tested for SNPs associated with site salinity using BayPass v. 2.2 [40]. Primers were successfully developed for three salinity-associated SNPs, and targeted Sanger sequencing was performed on crabs sourced from salinity gradients in the Pamlico and Neuse Rivers (electronic supplementary material, table S12). Annotated genes included cyclooxygenase and large proline-rich protein BAT3, plus one unannotated gene. See electronic supplementary material, Methods for additional details.

3. Results

(a) . Population-level infection prevalence, host abundance, and taxa diversity

The best model (conditional R² = 0.89; AICc weight = 1.00) for mature LP externae included the additive effects of salinity (z = 10.415; p < 0.001), temperature (z = 2.264, p = 0.024), E. depressus abundance (z = − 4.247; p < 0.001), and DOY (z = 6.518; p < 0.001) (electronic supplementary material, tables S4 and S5). For virgin externae, the best model (conditional R² = 0.85; AICc weight = 1.00) included the interaction of salinity and temperature (z = − 3.7848; p < 0.001) and the additive effects of salinity (z = 6.299; p < 0.001) and DOY (z = 2.284; p = 0.022) (electronic supplementary material, tables S6 and S7). Prevalence was negligibly less than 10 PSU. For mature LP externae, higher salinity and temperature correlated with higher prevalence (figure 2a). The pattern was nonlinear for virgin externae: the highest infection prevalences were found at higher salinities but lower temperatures (figure 2b). E. depressus abundance was another significant predictor of mature externae prevalence, with higher E. depressus abundances corresponding to lower prevalence in RH populations. This trend was not observed for virgin externae (figure 2c). For RH abundance, the best model (conditional R² = 0.24; AICc weight = 0.57) included LP prevalence only (z = −2.359; p = 0.018) (electronic supplementary material, tables S8 and S9), with increasing infection prevalence associated with lower host abundance (figure 3). For taxa diversity, there was a significant influence of season (F = 2.826; p = 0.014) and salinity nested in estuary (F = 7.012; p = 0.001). In an nMDS (Non-metric Multi-dimensional Scaling) plot, the Pearson correlation overlays (electronic supplementary material, figure S5) revealed specific taxa corresponding to salinity level, including dragonfly larvae (Near Fresh); American eels (Anguilla rostrata) (Near Fresh/Very Low); polychaetes, skilletfish (Gobiesox stromosus) and ED (Low–Moderate/Moderate); and grass shrimp (Paelomonetes spp.) and snapping shrimp (Alpheus heterochaelis) (Moderate). See electronic supplementary material, table S1 for taxa composition and counts.

Figure 3.

Prediction plot of the relationship between LP-infection prevalence and adult RH abundance. The line represents the predicted abundance; the dots are observed data.

(b) . Probability of individual infection

The best model (conditional R² = 0.89; AICc weight = 1.00) for infection probability included the interaction of salinity and temperature (z = −0.555; p = 0.579) and the additive effects of salinity (z = 13.266; p < 0.001), temperature (z = 2.258; p = 0.024), CW (z = 3.792; p < 0.001), and DOY (z = 7.276; p < 0.001) (electronic supplementary material, tables S10 and S11). Above 10 PSU, there was a significant positive relationship between salinity and infection probability; a similar but less pronounced pattern was observed for crab size (electronic supplementary material, figure S4).

(c) . Crab righting response

There was no effect of sex (F = 0.286; p = 0.595) or size (F = 0.366; p = 0.548) on crab righting response. There was a marginally significant effect of treatment (F = 2.13; p = 0.065), with a trend for ‘Near Fresh’ sourced crabs to have longer righting times, particularly when exposed to the lowest (0.8 PSU) and highest salinity treatments (15 PSU) (electronic supplementary material, figure S6).

(d) . Differential gene expression

Approximately 1.5% of contigs (1020/68 051) were differentially expressed (p < 0.05) with salinity (figure 5a,b)—roughly split between increased (N = 480) and decreased (N = 540) expression in higher acclimation salinities. GO-testing identified several Molecular Function categories upregulated at lower salinities, including two active transmembrane transporter terms (p < 0.001) and several oxidoreductase activity-associated terms (p < 0.001; electronic supplementary material, figure S7). Other oxidoreductase- and ion binding-related terms were associated with up- and downregulation in low salinities, including cation binding (p < 0.001), ion binding (p = 0.004), and ion channel inhibitor activity (p = 0.02 l; electronic supplementary material, figures S7 and S8). Several Cellular Compartment terms under the proton-transporting ATPase complex umbrella were significantly upregulated at lower salinities (p < 0.001; electronic supplementary material, figure S9). Results were similar when considering only the 3 and 15 PSU acclimation salinities (electronic supplementary material, figures S10–S12). A test for interactions between salinity and source identified 40 contigs whose expression differed significantly between 3 and 15 PSU acclimation depending on source (figure 5b). A majority of these showed the same relationship in crabs from ‘Very Low’ and ‘Low’ sources (either an increase or decrease with increased salinity), and a different pattern (usually no change) in crabs from the ‘Near Fresh’ source. When we plotted normalized data from ‘Near Fresh’ crabs after acclimation to 0.8 PSU, these crabs showed a similar pattern of expression to ‘Very Low’ and ‘Low’ source crabs after acclimation to 3 PSU (figure 5c,d). Of the 40 significantly interacting genes, 11 had informative annotations; of these, three were involved in immune response in Drosophila (NF-κ-B, Relish, Spätzle), and two were subunits of V-type proton ATPase.

Figure 5.

Population structure and selection across a salinity gradient. (a) PCA of 16 592 SNPs coloured by source in the Pamlico River. (b) Minor Allele Frequency (MAF) at a candidate SNP for selection across a riverine salinity gradient. MAF data derive from transcriptome sequencing (heavy black outlines) and subsequent targeted sequencing in additional Pamlico and Neuse River sites. (Online version in colour.)

(e) . Population structure and signatures of selection

No genetic structure was identified among RH collected from three salinity reaches in the Pamlico River (PC1: loading = 1.13%, p = 0.28; PC2: loading = 1.12%, p = 0.45; figure 5a). BayPass identified 15 SNPs associated with site salinity (Bayes > 25). When mRNA-Seq and targeted Sanger sequencing data were combined, an SNP in an unannotated gene showed a significant association with salinity across both rivers (R² = 0.63; p = 0.007; figure 5b) and the Pamlico River alone (R² = 0.58; p = 0.048). The remaining two SNPs with targeted genotyping data (annotated as cyclooxygenase and large proline-rich protein BAT3) showed non-significant trends with salinity in both rivers (electronic supplementary material, figure S13 and table S12).

4. Discussion

Environmental factors—both abiotic and biotic—delineate ecological niches, drive the evolution of species, and ultimately shape ecological landscapes [2,41]. Within estuaries, organisms contend with environmental conditions that rapidly change at both temporal and spatial scales—with salinity and temperature being two of the major physical drivers of species distributions [16]. Concurrently, biotic interactions with parasites, predators, competitors, and diseases strongly shape species distributions and demography across these same scales [16]. Selective forces can vary along environmental gradients, and organisms may ultimately face trade-offs where increasing stress from one pressure (e.g. salinity) alleviates stress from another more substantial pressure (e.g. parasitism) [1]. Ultimately, evolutionary responses to environmental stresses will be reflected at population and community levels, influencing ecosystem structure and function. In this study, we found a strong effect of biotic and abiotic forces on parasite prevalence and host abundance/response to salinity, suggestive of potential trade-offs between salinity tolerance and parasite infection.

(a) . Salinity and temperature drive rhizocephalan parasite prevalence in invaded estuaries

Across temperate regions, factors influencing parasite prevalence often include temperature, habitat, host abundance, and host diversity [42]. In estuaries, salinity is highly variable, and communities become structured along salinity gradients due to evolved tolerances [43]. These diversity gradients represent transition zones between estuary reaches, as in our system where certain species were bellwethers of their salinity category (electronic supplementary material, figure S5). Salinity can also strongly structure parasite diversity (e.g. estuarine fish [44], periwinkle snails and mussels [45]) and prevalence (e.g. blue crabs and Hematodinium infections [46]), with some parasites demonstrating narrower salinity tolerances than their hosts [47].

Here, we found strong support for salinity as a major driver of LP infection at the population (figure 2a,b) and individual (electronic supplementary material, figure S4) levels: the parasite was absent from sites with mean salinities less than 10 PSU (figure 1c). Interestingly, one site (Wrights Creek, figure 1b,c) on the cusp of this threshold demonstrated fluctuating LP presence/absence depending on weather conditions and associated salinity. Following back-to-back tropical storms (Florence, September 2018; Michael, October 2018), salinity and LP prevalence dropped significantly (electronic supplementary material, table S2A,B and figure S4) and taxa diversity altered (electronic supplementary material, figure S5) at some sites. This susceptibility to low salinity appears to occur primarily during the parasite's short-lived free-living larval stage. Once LP successfully infects a competent host and is protected internally, it tolerates salinities below this threshold; however, successful fertilization and maturation of the externa requires a second period of favourable salinity for larval survival and transmission success [27]. Indeed, we observed mature infections at locations with highly variable salinities over the study period (e.g. following rain events when salinities dropped less than 10 PSU).

Along with salinity, temperature was the other significant driver of infection prevalence: the highest infection prevalences tended to occur during warmer temperatures (i.e. summer and autumn) at sites greater than 10 PSU (figure 2a,b). These results are akin to the eastern oyster (C. virginica), which is more prone to the lethal effects of protozoan parasitism in higher salinity waters during warmer months [1,21,48]. While low-salinity refugia allow for rapid responses to the strong negative impacts of parasitism in oysters, there are associated costs [49]. For one, oysters in low-salinity waters are released from selective mortality and thus remain susceptible to disease (i.e. upstream, low-salinity populations are not under strong selective pressures to develop resistance). This may slow the evolution of resistance throughout the estuary [49], though transplant experiments show that evolved resistance can occur in oysters regularly exposed to parasites [50]. Second, low salinities may be associated with negative growth and body condition due to lower food availability, high turbidity, and freshwater runoff [51], with oyster fecundity reduced in upstream waters [1]. Thus, evidence suggests there may be trade-offs between disease avoidance and low-salinity stress in oysters. Similarly, RH receives a refuge from infection in low-salinity waters (less than 10 PSU) but may also incur costs (discussed below).

(b) . Infection prevalence and seasonal emergence vary based on parasite life stage

Parasite life cycles can play a key role in host infection dynamics observed over time. Though LP has a direct life cycle using a single host, it has both free-living (larval) and host-associated (virgin and mature adult) life stages, and each is influenced by seasonal conditions (figure 1a). Moreover, rhizocephalans are sexually dimorphic, and males and females are under distinct environmental constraints for survival and reproduction [52]. In our study, we noted a seasonal trend in the emergence of mature versus virgin externae in RH host populations. Early in the year, few mature LP externae were observed. As temperatures warmed in the spring, higher rates of female virgin externae appeared (figure 2b) and by late spring/early summer, these virgin externae became fertilized, with mature externae increasing in prevalence over the summer (figure 2a). Because fertilized externae do not transition back to the virgin stage unless the mature externa is shed [53], the higher prevalence of mature infections in late summer/autumn represents the cumulative infections from spring to autumn. As temperatures declined over autumn and winter, the prevalence of mature externae also decreased. This reduction could be due to several factors, including enhanced host mortality from the combined stress of infection and colder winter temperatures; shed externae (i.e. host still infected but less detectable); and seasonal declines in the parasite's reproductive cycle and fertilization rates [54]. Interestingly, higher abundances of the alternate LP host, E. depressus, were associated with lower prevalence of mature externae in RH. This could be explained by the encounter dilution effect, in which per capita infection rates decrease as the number of available hosts increase [55]. An alternative explanation is greater susceptibility of E. depressus to LP infection where the two crabs overlap. While laboratory studies have confirmed that LP can infect either host species (AE Fowler 2019, personal observation), targeted trials to determine whether hosts differ in their susceptibility to infection are needed.

(c) . Parasite prevalence is negatively associated with host abundance

Not surprisingly, parasitic castrators impose strong population-level influences on hosts [56], particularly in high-prevalence scenarios. Although it seems counterintuitive for an obligate parasite—requiring its host for survival and reproduction—to cause a decline in host abundance, relationships like these often evolve in hosts with high fecundity and large population densities, like molluscs and crustaceans; they also depend on the scale of host and parasite recruitment [56]. Moreover, some population models have demonstrated oscillations in parasite and host population sizes, with parasite abundance lagging behind the host, analogous to Lotka–Volterra models. Because parasitic castrators limit a host population's reproductive output, parasite-induced host abundance declines are accordingly predicted to result in parasite abundance declines due to ensuing limitations in host resources [7,57]. In coevolved populations, these host–parasite oscillations typically reach an equilibrium, such that both associates can persist through time [58]. However, this equilibrium could be disrupted with the introduction of a new host or parasite.

In our system, LP is coevolved with RH in the Gulf of Mexico, where infection prevalences are typically less than 10% [25]. In the parasite's invasive range in Atlantic estuaries, prevalence is substantially higher (up to 60% here; 90% in past studies), supporting the host's greater susceptibility where it is naive [25]. Given the parasite's strong impact on host reproduction, we predicted that parasite prevalence would negatively correlate with host abundance, and our data supported this (figure 3). Other studies have similarly found negative associations between parasite prevalence and host abundance following parasite introduction, including abundance declines in E. depressus shortly after LP introduction in northeast Florida [7], and a crash in Upogebia pugettensis shrimp populations in the US Pacific following the introduction of a bopyrid isopod parasite (Orthione griffenis) [59]. Long-term investigation of the RH–LP system would be beneficial to track parasite prevalence and host abundance through time, particularly given the short-term coevolutionary history of host and parasite in invaded estuaries.

(d) . Evolutionary origins and potential consequences of exploiting low-salinity refugia

High infection prevalence, and associated reductions in host fitness, are expected to exert strong selective pressures on host populations. This response could take several forms, including the evolution of increased resistance to infection, avoidance behaviours that lower infection susceptibility, and exploitation of parasite-free refugia [9,13,14]. Where host and parasite share a long-term coevolutionary history, RH may have evolved the capacity to use all three defences. In the invaded range, where the host is only recently exposed to the parasite, environmental exploitation may be the fastest and most accessible response [60], particularly for a species like RH that has evolved in an environment where salinity fluctuates constantly.

Our gene expression data suggest that RH is highly plastic in its response to a wide range of estuarine salinities. Over 1000 genes showed significant variation in expression with salinity, regardless of source population (figure 4a,b). Genes upregulated in lower salinities were enriched for functions and processes surrounding osmoregulation, including ion binding, ion transport, and transmembrane transporter functions; structural genes associated with the ATPase complex, particularly V-type proton ATPase, were also significantly upregulated in lower salinities (electronic supplementary material, figures S7–S12). These global responses to salinity demonstrate that RH is capable of adjusting its physiology to tolerate a wide osmotic range [61]. Differential expression of these genes may partly explain RH's demonstrated ability to maintain hyperosmotic hemolymph below approximately 24 PSU [62]. More interesting is the smaller number of genes whose response to salinity was modulated by population source (figure 4c,d). While annotation was incomplete, they included two subunits of V-type proton ATPase, a proton pump which has been key to freshwater adaptation in other decapod crustaceans [31]. A canonical example of rapid crustacean evolution to freshwater involves the estuarine copepod Eurytemora affinis, in which freshwater invasion and laboratory selection both resulted in evolved changes in V-type proton ATPase activity and osmoregulatory ability at lower salinities in just a handful of generations [63]. Rapid freshwater transitions in this system have been facilitated by selection on standing genetic variation, including variation in or near ion transporter genes, maintained in estuarine source populations [19].

Figure 4.

Representative differentially expressed contigs in response to salinity. (a,b) Example contigs up- and downregulated in response to acclimation salinity. (c,d) Example contigs for which response to acclimation salinity differs based on collection salinity. Circles indicate individuals, coloured by source. Boxplots show the median (centre line) and interquartile range (IQR) of values; whiskers extend to minimum values less than 1.5× the IQR from the IQR boundary; points with white dots indicate outliers. (Online version in colour.)

While the interaction of gene expression and salinity source we observed in V-type proton ATPase (and other genes) may be due to long-term acclimation to different estuarine reaches, they may also indicate within-estuary selection on standing genetic variation associated with osmoregulatory gene regulation. In the mid- to upper-estuary reaches RH inhabits, salinities are dynamic and can reach near-fresh salinity levels. In response to this strong but uneven pressure, RH may maintain standing genetic variation associated with tolerance of a range of salinities [19,64]. Our genetic data support this, finding a candidate SNP which showed salinity-associated changes in frequency along an estuary gradient, despite neutral genetic structure demonstrating a single well-mixed population (figure 5). Similar gradients in allele frequency across a backdrop of little neutral structure are characteristic of balanced polymorphisms which are under repeated selection in each settling generation [64]. While additional sequencing in more sites and estuaries is needed, we suggest that RH's evolution to dynamic estuarine conditions may have resulted in standing variation for salinity tolerance which has permitted it to exploit low-salinity refugia after LP's invasion. Escape into fresher waters may in turn cause shifts in allele distributions as the host adapts to consistently lower salinities in parasite-free refugia. In the coevolved range, this may have resulted in the crab's ability to invade several inland reservoirs and lakes in Texas where the species successfully reproduces in near-fresh conditions [65].

5. Conclusion

Parasitism is a potent selective force in estuarine communities; yet despite its ubiquity, little empirical work has explored its influence on host evolution in the context of a dynamic environment. Such investigations are of increasing importance in this period of human-mediated global change, where species ranges are changing in response to shifts in temperature, salinity, and precipitation, and from anthropogenic introductions around the world [47]. Such shifts expose hosts and parasites to new selective regimes that shape ecological communities via changes in host–parasite interactions, trophic dynamics, and overall ecosystem function. Consequently, species may face selective trade-offs in strong—and potentially conflicting—biotic and abiotic forces. Hosts may exploit low-salinity refugia to escape impactful parasites but may concurrently face fitness costs from osmotic stress [1]. Future work should focus on these potential costs, particularly related to differential survival and reproduction in low-salinity waters along a coevolutionary mosaic [25]. Future examinations of RH's non-native populations around the world would provide greater understanding of the role its broad salinity tolerance played in its invasion success [64]. Lastly, parasites are often ignored in studies of community and evolutionary ecology, but ours and other studies demonstrate the extensive impacts that parasites can have on the ecology and evolutionary trajectory of their hosts and the broader community. Concerted investigations are needed to further uncover parasite-driven influences on host evolutionary ecology in the natural world, particularly given their multiple downstream effects.

Ethics

Field collections were authorized by NC Division of Marine Fisheries (Scientific Permit no. 706671). Field access was approved by the NC Division of Parks and Recreation (Goose Creek State Park, permit 2019_0379), US Fish and Wildlife Service: Swan Quarter and Cedar Island Wildlife Refuges (permit R20-002), US Department of Agriculture: Croatan National Forest (permit CRO206902), the NC Estuarium, Matthews Point Marina, and the North Creek Landing Homeowners Association, Inc. Animal husbandry and dissection protocols were ECU-IACUC approved (AUP no. D346).'

Data accessibility

Datasets supporting this article have been uploaded as part of the electronic supplementary material. Raw transcriptome sequence data are archived in GenBank's SRA: BioProject PRJNA730785, BioSamples SAMN19241288–SAMN19241333. The processed transcriptome datasets and resources, including custom scripts, are available from the Dryad Digital Repository: https://dx.doi.org/10.5061/dryad.573n5tb7r [66].

Authors' contributions

A.M.H.B. and C.K.T.: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, supervision, validation, visualization, writing; D.L.P.: data curation, formal analysis, software, visualization, writing; A.E.F.: conceptualization, investigation, methodology, supervision, writing; C.S.M. and T.S.L.: investigation, methodology, supervision; R.B.B., K.M.S., L.C.L., and M.R.: investigation, methodology; M.E.T., A.W.M., and G.M.R.: conceptualization, supervision, writing. All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Competing interests

We declare we have no competing interests.

Funding

This work was supported by an ORAU Powe Junior Faculty Enhancement Award, an ECU TCHAS Research Initiation Grant, an ECU Coastal Scholar Award and an NC Sea Grant Community Collaborative Research Grant (17-CCRG-03).