Parasites can alter the traits or densities of mutualistic partners, potentially destabilizing mutualistic associations that underpin the structure, functioning, and stability of entire ecosystems. Despite the potentially wide-ranging consequences of such disruptions, no studies have directly manipulated parasite prevalence and/or intensity in a mutualistic partner, nor quantified the resulting community-level effects. Here, we investigated the effects of a common trematode parasite (Cercaria opaca) on the strength of a keystone facultative mutualism in western Atlantic salt marshes between the foundational marsh cordgrass, Spartina alterniflora, and the ribbed mussel, Geukensia demissa. Cordgrass increases mussel survivorship and growth through shading, while mussels enhance cordgrass growth by producing nutrient-rich biodeposits. This mutualistic association also creates conditions that enhance biodiversity and ecosystem functioning, and mediates the ability of foundational plants to resist and recover from extreme drought. We used lab and field assays to show how increasing infection with trematode metacercariae negatively influenced mussel biodeposit production, as well as the strength of mussel shells and byssal attachments. By conducting a field manipulation using experimentally infected mussels, we demonstrated that the mutualistic benefits of mussels to cordgrass growth decreased with increasing trematode infection intensity — a pattern likely generated by reduced mussel biodeposition and enhanced mortality. Additionally, increasing parasite loads in mussels led to predictable decreases in the abundances of benthic invertebrates, as well as in key ecosystem characteristics and process rates (i.e., redox potential and sediment accretion). Finally, a survey of five North Carolina salt marshes demonstrated that infection with C. opaca was most common in mussels in areas experiencing cordgrass die-off due to drought, and that infection intensity decreased with distance from die-off areas. Because the mussel-cordgrass mutualism underpins marsh ecosystem resilience to drought-associated die-off, our results suggest that parasitism may depress recovery from these disturbances. Although this is the first experimental demonstration of parasites indirectly altering community structure and functioning by undermining an ecologically influential mutualism, this type of relationship could be common in nature, given that parasites frequently infect influential mutualists.

Experimental manipulation of parasite infection intensity in the field

Acquiring infected marsh periwinkles

 From June to July 2017, adult Littoraria irrorata were collected at low tide from Gunning Hammock Island, NC (34.672454 N, -76.501743 W), where infection with Cercaria opaca was known to be common. Collected periwinkles were gently washed in filtered seawater and placed in small petri dishes. Each dish was filled entirely with filtered seawater (~30 mL) and sealed with a plastic lid such that the periwinkle was completely immersed and unable to escape. Petri dishes with periwinkles were allowed to sit at room temperature for 24 h. After this period, the contents of each dish were inspected under a microscope to determine the presence of shed cercariae (Morton 2018). Average incidence of trematode infection from five collections (4501 snails in total) was 39.4% ± 1.8 (mean ± SE). For each infected individual, cercariae were identified to species (Stunkard & Cable 1932, Holliman 1961, Coil & Heard III 1966, Heard III 1968, Epstein 1972, Wardle 1974). Infection with C. opaca was most prevalent (36.7% ± 1.5). We only retained periwinkles that were either infected by C. opaca, or that did not shed any cercariae and were assumed to be uninfected. To distinguish between these two groups of periwinkles, individuals infected with C. opaca were marked with a small dot of red paint, while uninfected individuals were marked with a dot of blue paint. Infected and uninfected periwinkles were kept in separate aquaria and provisioned with moistened cordgrass wrack until used to experimentally infect mussels. Following their use in experimental infection of mussels, the infection status of all periwinkles was confirmed via dissection.

Experimental infection of mussels

 We collected Geukensia demissa from an intertidal marsh on Radio Island, NC (34.723292 N, -76.680375 W), where adult gastropod first intermediate hosts were not present and where previous dissections of mussels (>150) revealed no metacercariae in the tissues (Morton pers. obs.). We selected mussels that fell within a narrow size range (7 – 8.5 cm shell length) for use in experiments. Mussels were transported back to the lab where they were gently washed in filtered seawater to remove any excess mud from the exterior. Any debris attached to the byssus was clipped off with scissors. Mussels were then placed in circular glass dishes (diameter = 20 cm, height = 6 cm) in groups of eight. Each dish was filled with filtered seawater and equipped with a bubbler stone for aeration. A circular lid crafted from 0.635 cm vexar mesh was placed on top of each dish and held firmly in place with a small stone weight. Mussels were provisioned with Shellfish Diet 1800® every day (8 mL per dish per day) and the water within each dish was changed twice weekly. Dishes containing mussels were divided into the following treatments: (1) High exposure to infected periwinkles (two infected snails per individual), (2) Medium exposure to infected gastropods (one infected snail per individual, and 1 uninfected snail), (3) No exposure, with mussels exposed to 2 uninfected periwinkles per individual. If mussels died, they were removed from their dishes along with the appropriate number of periwinkles such that the periwinkle to mussel ratio never changed. If any periwinkles died, they were immediately replaced.

 After 3 months, 20 mussels from each treatment group were selected at random and dissected to determine average infection intensity. At this time, mussels from the “high exposure” treatment had reached an average infection intensity comparable to that observed in field-collected specimens from local marsh die-off areas (1138 ± 111 metacercariae per individual). The medium exposure treatment had an average infection intensity that was roughly half of mussels in the high exposure group (632 ± 43 metacercariae per individual). No metacercariae were found in the tissues of mussels not exposed to infected periwinkles.

Field manipulation

 In October 2017, we established 32 permanent 0.5 m² plots in a structurally homogenous swath of intermediate cordgrass marsh at Hoop Pole Creek, Atlantic Beach, North Carolina (34.705288 N, -76.749650 W). The study site was characterized by very low periwinkle densities (<1 snails/m²) and scattered mussel mounds. Dissections of mussels (n = 30) and periwinkles (n = 100) from the study area before the beginning of the experiment revealed no trematode infections. Plots were marked at two corners with 2 cm diameter PVC posts labeled with the individual number of each plot, and we selected plot areas without pre-existing mussels or mud crab burrows. After the establishment of experimental plots and before the application of treatments, we enumerated and measured the heights of all live cordgrass stems in each plot. Additionally, we counted the number of mud crab (Panopeus herbstii) burrows, marsh crab (Sesarma reticulatum) burrows, and juvenile and adult fiddler crab (Uca spp.) burrows (adult burrows were 5 to 10 times wider than juvenile burrows) as a proxy for the abundance of these organisms (Macia et al. 2001, Angelini et al. 2015, Martínez-Soto & Johnson 2020). Additionally, we counted the number of adult and juvenile periwinkles in each plot (Angelini et al. 2015). There were no statistically significant differences in any of these initially measured variables between treatment groups (ANOVA, p > 0.23, all cases).

Plots were assigned to the following infection intensity treatments, corresponding to the aforementioned mussel exposure treatments (n = 8 replicates per treatment): (1) highly infected mussels, (2) medium infected mussels, and (3) uninfected mussels. We also established 8 plots with no mussels to control for the effect of mussel addition. Mussels from corresponding treatment groups were transported to the field in aerated, seawater-filled buckets, with mussels in each group placed in their own separately labeled bucket to avoid confusion. Thirty mussels from the same exposure treatment were arranged in circular mounds at the center of corresponding treatment plots. All mussels were inserted such that they protruded no more than 2 cm from the sediment. The length, width, and depth of each mussel was measured prior to being inserted into the sediment. A week after experimental setup, plots with mussels were examined to ensure that all mussels were alive and were firmly rooted in the sediment. Any dislodged mussels were re-inserted into the sediment and any dead mussels were replaced with new mussels of the appropriate treatment group, and the dimensions of these new mussels were recorded.

To examine how differing infection intensity could modify the role of mussels in binding sediment and influencing vertical accretion of the marsh, three level rods were installed in each plot (Bertness 1984). Level rods consisted of 0.5 cm diameter PVC dowels driven vertically into the marsh until firmly embedded in the underlying clay (~0.75 m below the marsh surface). Each rod was cut so that it stuck out 15 cm above the marsh surface and was labeled with a permanent marker. The height of each level rod over the marsh surface was taken one week after their installation and every month afterward for the year-long duration of the experiment (October 2017 – October 2018).

 To determine whether parasite load could indirectly alter marsh sediment oxygen availability, we measured sediment redox potential (mV) within each plot. Redox measurements were taken using a platinum electrode probe and a Fisher Scientific Accument double junction Ag/AgCl electrode (+200 mV correction added to the measured value), connected through a Fisher Scientific Accument PH/mV/C meter, model AP71 (Thermo Fisher Scientific, Waltham, Massachusetts, USA) (Gittman & Keller 2013). At absolute low tide, each electrode was placed 10cm into the mussel mound sediment near the center of each plot and left for 30 minutes to allow the sediment surrounding the probe to stabilize. We measured redox in all plots before the sediments were disturbed by the application of various treatments. We measured redox again one week after mussels had been added to assigned plots and then subsequently every month following installation.

 At the end of the experiment, we determined benthic chlorophyll a (chl-a) concentration as a proxy for algal biomass in each plot by taking five replicate cores (1.75 cm diameter) from the center of plots at low tide (Parsons et al. 1984, Howes & Teal 1994). Cores were placed in opaque vials, put on ice, and transported back to the lab to be processed immediately. The top 2 mm of sediment were removed from each core and placed in labeled vials of 100% methanol for 48 h to extract pigments. Vials were wrapped in aluminum foil and stored in a -20°C freezer until processed. Prior to fluorometric analysis, vials were warmed to room temperature. Each vial was decanted into a 5 mL glass tube which was then placed in a calibrated Tuner 10-AU fluorometer, and the fluorescence was read (Parsons et al. 1984). Chlorophyll concentration was calculated as mg chl-a per 0.5 m².               

At the end of the experiment, we enumerated all macroinvertebrates as before, as well as all flowering cordgrass in each plot. Following this, all live cordgrass within each plot was harvested, washed, and dried for 3 weeks at 70 °C, before biomass was determined to the nearest 0.01 g. Cordgrass belowground biomass was determined by taking 25 cm deep sediment cores using a 6 cm diameter coring device (Bertness 1984). Cores were washed and sieved through a 2 mm mesh sieve. Live roots and rhizomes were separated out on the basis of rigidity and coloration (Bertness 1984), dried at 70 °C for three weeks, and weighted to the nearest 0.01 g. All mussels used in the experiment were collected, measured, and dissected to determine parasite infection intensity. Mussel meats were removed from the shell, gently torn apart with a scalpel, and then sandwiched between two panes of gridded plexiglass so that metacercariae within all mussel tissues could be easily counted under a dissecting microscope.

Differences in treatment means for different response variables (final mussel density, aboveground biomass, belowground biomass, sediment elevation, macroinvertebrate abundances, benthic chl-a, sediment redox potential, flowering, and biodeposit production rate) were evaluated using one-way ANOVA, followed by Tukey HSD tests for pairwise comparisons. Because the data did not meet the normality assumptions of ANOVA, differences in treatment means for mud crab burrow abundance were evaluated using a Kruskal-Wallace test, followed by Dunn’s test of multiple comparisons with p-values corrected using Holm’s method to control for the family-wise error rate. All analyses were performed in the R statistical computing environment (v4.1.2; R Core Team 2021).

Mechanisms underlying parasite effects on the mussel-cordgrass facultative mutualism

Effects of parasites on mussel mortality

In order to determine if parasite infection drove patterns of mussel mortality in the field, we enumerated all live and dead mussels in each of our experimental plot every month for the duration of our field experiment. Because shells of dead mussels typically remained in situ, we were able to infer the putative cause of death of mussels during monitoring. Mussel mortality was either attributed to consumptive (shells showed obvious signs of crab predation) or non-consumptive (empty, articulated shells without obvious signs of crab predation) effects.

Treatment differences in total mussel survivorship and mortality rate from both consumptive and non-consumptive sources were evaluated using one-way ANOVA, followed by Tukey HSD tests for pairwise comparisons.

Effects of parasites on mussel biodeposit production (field assay)

Immediately prior to the collection of all mussels, forty live individuals were selected from each treatment group and subjected to a biodeposit production assay using methods adopted from Smith and Frey (1985). Mussels were inserted into the sediment of a marsh area adjacent to the site of the field manipulation. Each mussel was surrounded with an apron made from a circular section of white sheet plastic (diameter = 20 cm). Aprons were pinned in place with galvanized lawn staples and marked with an identifying number using permanent marker that corresponded to each mussel’s mound number and treatment group. Mussels at the apron’s center began filter feeding and producing biodeposits as the high tide covered them. During the last ~30 minutes of tidal inundation when mussels were still covered with water, biodeposits were collected from each apron with a 50 mL pipet, returned to the laboratory, dried at 70°C for two weeks, and then weighted to the nearest 0.01 g. Biodeposit production rate was calculated as the net dry weight of collected biodeposits divided by tidal inundation time (5 h). All mussels used in this assay were dissected to determine parasite intensity along with the others used in our experiment.

Treatment differences in biodeposit production rate were evaluated using one-way ANOVA followed by Tukey HSD tests for pairwise comparisons. We further explored the relationship between mussel biodeposit production rate and parasite infection intensity using multiple linear regression, with biodeposit production rate as the dependent variable and infection intensity and mussel length as predictors.

Effects of parasites on mussel shell thickness and strength

In order to determine the impact of trematode infection intensity on ribbed mussel shell strength, we conducted a laboratory experiment using field-collected mussels to determine the force required to cause structural failure of mussel shells. In October, 2018, we collected mussels (n = 50) from a marsh die-off area on Gunning Hammock Island. Mussels were stored in a flowthrough tank. To estimate the mass required to crush live mussels with different parasite loads, we constructed a crushing device based on a similar device used by Hughes (Hughes 2016) (Appendix S1:Figure S2). The device consisted of a pivoting wooden arm that came to rest on a stationary wooden platform that was firmly affixed to the edge of a tabletop with two bar clamps. A piece of 10 × 7 × 1.5 cm steel plate was affixed on one end of a 40 × 7 × 5 cm wooden platform such that the pivot-arm came to rest at its center. Another piece of steel plate was affixed to the edge of the pivot-arm that came to rest on the steel plate. A steel bolt affixed to the end of the pivot arm accommodated a free-hanging bucket to which weight (iron weights and water) could be added.

Mussels were retrieved from their holding tank one at a time to be tested. After determining each mussel’s wet weight, shell length, width, and depth, individual mussels were then positioned at the center of the crushing platform such that the steel edge of the pivoting arm made contact with the mussel at its widest point. A bucket was hung on the end of the pivot-arm, and weight was gradually added to the bucket until the shell of the mussel reached the point of structural failure. The force required to crush a mussel was calculated using the mass of the bucket and weights added to the weight of the pivot arm at rest. After being crushed, we dissected mussels and quantified infection intensity. We evaluated the relationship between parasite load and mussel shell strength using multiple linear regression, with crush force (N) as the dependent variable and infection intensity and mussel wet weight as predictors.

Because we were unable to take reliable measurements of mussel shell thickness after mussels were crushed, we used shell thickness metrics taken from mussels used in our preliminary lab biodeposition assay (Appendix S2: Section S1) to determine how mussel shell thickness varied with C. opaca infection intensity. Using a digital micrometer, we measured the thickness of clean mussel shell halves (n = 50 mussels) to the nearest 0.001 mm at three points (lip, ventral surface, and dorsal surface) following the methods of Goss-Custard (2002). Measurements from each mussel shell half were averaged together to generate a shell thickness metric for each mussel. We evaluated the relationship between mussel shell thickness and parasite load using multiple linear regression, with average shell thickness as the dependent variable and infection intensity and mussel size (wet weight) as predictors.

Effects of parasites on byssus formation and attachment strength

To determine how mussel byssus formation and byssal attachment strength varied with parasite infection intensity, we subjected field-collected mussels to a laboratory assay. Experimental units consisting of terra cotta flowerpots (diameter = 11 cm, height = 11 cm) were filled with 500 mL of marsh sediment. The sediment used was excavated from a salt marsh adjacent to Pivers Island, Beaufort NC (34.71823 N, -76.67121 W) and thoroughly homogenized prior to adding it to each pot to ensure uniformity of grain size. A 20 cm long segment of dry bamboo (diameter = 1 cm) was inserted into the sediment in the center of each pot. Mussels (n = 60) of an initially unknown infection status were collected from a marsh die-off area within the Rachel Carson Estuarine Research Reserve (34.709133 N, -76.58506 W) in September 2020. Mussels were gently washed with filter seawater and any byssal threads protruding from the shell were clipped off. A single mussel was inserted vertically into the sediment adjacent to the bamboo segment, such that it made contact with the interior margin of the mussel shell where the byssus emerges. Pots with mussels were placed in a flowthrough tank filled with filtered seawater. After a week, mussels were removed along with their corresponding bamboo, taking great care not to sever attached byssal threads. The length, width, and wet weight of each mussel was measured and the number of byssal threads attached to the bamboo post were counted under a dissecting microscope. Following this, the post connected to each mussel was secured to the edge of a table with clamps and a small harness comprised of two adjustable loops of wire were affixed to each mussel, with one loop of wire wound around the mussel on either side of the byssus. A polystyrene cup was hung from the harness with string and water was added to the cup until the byssal attachments broke, fully separating the mussel from the bamboo. The force required to disconnect a mussel was calculated from the mass of the cup of water added to the weight of the harness and mussel. Following detachment, each mussel was dissected and parasite load quantified.

We evaluated the relationship between intensity of C. opaca infection and mussel byssal thread formation using multiple linear regression, with byssal thread count as the dependent variable and infection intensity and mussel wet weight as predictors. The relationship between mussel byssal attachment strength and parasite infection intensity was likewise evaluated using multiple linear regression, with the force required to dislodge mussels (N) as the dependent variable and infection intensity, number of attached byssal threads, and mussel wet weight as predictors. Prior to these analyses, infection intensity was log-transformed and detachment force was square root-transformed to meet the normality of residuals assumption of linear regression.

Survey of infection intensity within and around marsh die-off areas

In May 2018, we identified 5 marsh die-off areas along 40 km of the North Carolina coast (Gunning Hammock Island 34.6726073°N, -76.4992263°W; Harkers Island 34.6920993°N, -76.5265395°W; Town Creek 34.7238264°N, -76.6585029°W; Carrot Island 34.709038°N, -76.6589893°W; and Hoop Pole Creek 34.7054731°N, -76.7491902°W). Each of these die-off areas was characterized by a sizable, denuded mudflat (837 – 5187 m²) bordered on all sides by live cordgrass where dense grazing fronts of marsh periwinkles had accumulated. At each site, we haphazardly collected mussels along 50 m × 1 m transects along die-off borders, 20 m from the die-off border, and 40 m from the border (n = 25 mussels per transect). In order to determine the relationship between infection frequency in gastropod first intermediate hosts and infection intensity in mussels, we also haphazardly collected adult periwinkles along each transect (n = 100 snails per transect) and dissected them to determine the prevalence of infection with C. opaca within each transect. Mussels were weighed, measured, and dissected to determine infection intensity.

We used the lme4 package in R to conduct a linear mixed effects analysis to evaluate differences in infection intensity in mussels by distance from the die-off border (Bates et al. 2015). Die-off size and mussel size (length) were included as covariates and site was included as a random effect in our model. We approximated p-values for model factors using type II Wald χ² tests. We used the same approach to evaluate patterns of infection prevalence in periwinkles. Die-off size and distance from the die-off border were included as predictor variables while site was included as a random effect.

Supplemental methods: Effects of parasite infection on mussel filter feeding and biodeposit production 

To determine how mussel filtration rate varied with parasite infection intensity, we subjected field-collected mussels to a laboratory biofiltration assay. In July 2017, mussels of roughly the same size (8.0 ± 0.05 cm shell length) and initially unknown infection intensity were collected from mounds found in the vegetated marsh adjacent to a marsh die-off area on Gunning Hammock Island where infection with Cercaria opaca was common in periwinkle first intermediate hosts.

Mussels were gently washed under filtered seawater and any material attached to their byssus was removed with scissors. Mussels were stored in a flowthrough tank for 48 hours prior to the beginning of the filtration assay. Forty 1.8 L glass jars that served as experimental units were each filled with 500 mL of unfiltered seawater and aerated with an aquarium air stone. Immediately prior to the addition of mussels we took an initial 25 mL water subsample from each jar. Water subsamples were stored in numbered glass vials covered in foil and placed in a refrigerator prior to filtration. Following this, a single mussel was placed in each jar and allowed to filter feed. Eight aerated jars with no mussels acted as controls. We took subsequent 25 mL water subsamples each hour for the 3 hour duration of the experiment (Nagabushanam 1963, Jordan & Valiela 1982). At the end of the experiment, mussels were removed from jars and a final 250 mL water sample was taken. Mussels were weighed and measured as with previous experiments, placed in individually labeled bags and stored in a freezer before being dissected to determine infection intensity. Immediately following the end of the experiment, all water subsamples were vacuum filtered onto Whatman 47 mm microfiber filters which were then transferred to 25 ml vials of methanol and placed in a freezer to extract pigments. After 48 hours, samples were removed from the freezer and warmed to room temperature before Chl-a content was analyzed using a calibrated Tuner 10-AU fluorometer (Parsons et al. 1984). Average filtration rate was calculated from hourly rates and expressed in units of mg Chl-a per hour. Relationships between parasite load and filtration rate were evaluated using multiple linear regression with filtration rate as the dependent variable and infection intensity and mussel size (shell length) as predictor variables.