Infection prevalence, intensity, and tissue damage caused by the parasitic atworm, Bdelloura candida, in the American horseshoe crab (Limulus polyphemus)

Parasite infection dynamics can have profound implications on a host’s tness; yet, there is a dearth of information on parasites in the American horseshoe crab (Limulus polyphemus) (Linnaeus 1758), a species that has experienced population declines in recent decades. Therefore, we aimed to quantify the prevalence, intensity, and gill surface area coverage of the ectoparasitic atworm (cocoon and adult stages), Bdelloura candida in adult (n = 29), sub-adult (n = 7) and juvenile (n = 32) horseshoe crabs collected from Moriches Bay, NY (40.7810° N, 72.7171° W) in 2019 and 2020. Subsamples of horseshoe crab gill tissue (10%) were collected from live specimen, then B. candida cocoons were enumerated across the gill subsamples using microscopy while the extent of tissue damage was quantied with histology. B. candida was present in all adult and sub-adult crabs (100%), whereas juveniles exhibited 6.2% prevalence. Cocoon intensities per sample ranged from 28 to 805 cocoons, with 4.0–94.0% of gill lamellae harboring cocoons. In infected individuals, the total cocoon surface area coverage on gill tissues ranged from 0.06–14.51%, with higher cocoon intensities observed in the ventral-most gill quartiles relative to the dorsal-most gill regions. Sex was strongly supported as a primary driver behind B. candida infection intensities with adult females harboring higher intensities. Among infected gill lamellae, cocoon intensity was lower in mitochondrial-rich regions relative to mitochondrial-poor regions. These results provide novel insight into B. candida infection dynamics across horseshoe crab demographics, but further research is necessary to quantify the physiological impacts of the infection on L. polyphemus.


Introduction
The American horseshoe crab (Limulus polyphemus) is an iconic marine arthropod species that has persisted ~ 445 million years (Rudkin et al. 2008) and plays critical ecological roles in coastal marine systems from Maine, USA to the Yucatan peninsula in Mexico (Botton 2009;Smith et al. 2017). These ecological roles include bioturbation (sediment irrigation), structural habitat for epibionts, predation on marine bivalves and benthic macrofauna (Botton and Ropes 1987;Botton et al. 2003), and these animals provide a critical food source for migratory shorebirds, such as the endangered red knot (Calidris canutus rufa) (Tsipoura and Burger 1999;Smith et al. 2017). In addition to their numerous ecological roles, horseshoe crabs are an important human health and economic resource. The blood of horseshoe crabs serves to extract a unique compound, Limulus amebocyte lysate, (referred to as LAL) that is used to detect the presence of endotoxins from gram-negative bacteria in human medical supplies, and this practice has resulted in a threefold increase in biomedical harvest since the 1980s (Eyler et al. 2015; Smith et al. 2017). Moreover, roughly 1-2 million crabs are exploited annually for bait in the whelk and eel shery throughout the US East Coast (ASMFC 2019). Exploitation from these industries has been perceived to be the primary contributor towards their recent coastwide declines, and thus, led to their "Vulnerable" conservation status in the US (Smith et al. 2016). Despite the current management interest, the impacts of biotic stressors (e.g. parasite infections) on horseshoe crab tness and population dynamics have rarely been addressed, as management strategies have primarily focused on direct Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js anthropogenic exploitation. In particular, there is a dearth of information in regard to the role host-parasite relationships play on horseshoe crab tness.
Parasites are ubiquitous in nature with most wild animals harboring at least a mild infection, with the prevalence and intensity frequently increasing with size, age, and density of the host (Zelmer et al 1998). As established stressors, parasites are capable of in uencing numerous aspects of their host's biology including survival, fecundity, population cycles, and behavior (Lehman 1993;Hudson et al. 1998; Tompkins and Bergon 1999;Ebert et al. 2000;Poulin 2010) with cascading impacts on the entire ecosystem. Although variable between species, mild parasite intensities are typically well tolerated by a host (e.g. pinworm in humans) (Lehman 1993; Stjernman et al. 2008); however, intense infections can exacerbate this relationship, leading to adverse outcomes for a host. For instance, infections by roundworm (Ascaris lumbricoides) in the human intestine are considered to be mildly symptomatic, but the mere presence of several worms can cause blockages leading to indirect intestinal damage (Despommier et al. 2020). Similar observations have been made in aquatic animals, as many Monogenea (commonly observed ectoparasites of sh and amphibians) have been observed to damage their host not through resource exploitation, but via their attachment organs that penetrate the epithelial layer leading to excess mucus production and in ammation (Whittington 2005). Moreover, parasites infecting gill tissue have been demonstrated to adversely impact survival of aquatic organisms exposed to oxygen-poor conditions (Molnar 1994), with oxygen carrying capacity reduced up to 50% in crustaceans and sh (Taylor et al. 1996). Although several internal parasitic symbionts have been recorded in horseshoe crabs (Leibovitz and Lewbart 2003), the basic information on external parasites (i.e. prevalence, intensity and associated physiological impacts) such as the atworm, Bdelloura candida, are not well established in the horseshoe crab literature.
B. candida is a 2 cm long triclad atworm that is an obligate parasite of the American horseshoe crab. Adult worms are primarily located on the walking appendages and book gills, while the cocoons are attached on the inner surface of the gill lamellae by means of an endplate (Sluys 1989 In this study, juvenile, sub-adult, and adult American horseshoe crabs were sampled from an estuary along the south shore of Long Island, New York, to better understand B. candida infection dynamics. The primary objectives of this work were to: 1) quantify B. candida intensity and prevalence across horseshoe Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js crab demographics (age groups and sex) and 2) investigate intrinsic factors that explain B. candida intensities in horseshoe crabs. Additionally, we examined pathology in icted by cocoons on horseshoe crab gill tissue via histological analysis and quanti ed the percentage of gill tissue occupied by B. candida cocoons. We also assessed the infection intensity across the vertical gill space to determine if infection intensity was random (homogenous) or aggregated (heterogenous) across horseshoe crab gill space (ventral most or dorsal most). Lastly, we enumerated the proportion of B. candida cocoons occupying the central-mitochondrial rich area (CMRA) vs. peripheral mitochondrial-poor area (PMPA) of each gill lamellae (Hans et al. 2018) to determine if there was a spatial preference for cocoon presence across these regions that provide different waste and respirational roles (Hans et al. 2018). This study provides novel insight into the infection dynamics of B. candida on horseshoe crabs and could serve as the basis of monitoring ectoparasite infection in wild horseshoe crab populations.

Sample collection
On 24 June, 2019, juvenile (instars 8-10; n = 30) and adult (n = 29) horseshoe crabs were randomly collected at Pikes Beach, Moriches Bay, Long Island, NY (40.77°N, -72.71°W; Fig. 1). Juvenile crabs between instars 8-10 (prosomal width range = 41.4mm -57.9mm) (Sekiguchi 1988;Carmichael et al. 2003) were hand collected haphazardly along a 500 meter transect at the water's edge (< 0.5m depth), then transported back to the lab where they were euthanized using an overdose of Tricaine-S (MS-222) and frozen at -20°C for long term storage. Adults (prosomal width range = 188.0 mm-283.0 mm) were collected from the intertidal zone (~ 0.5-1m deep) shortly before high tide and were temporarily placed in sh totes lled with ambient seawater. Sex, prosomal width, and weight were recorded for every individual. Sex was determined by the presence of modi ed pedipalps, weight was measured using a Pesola (Schindellegi, Switzerland) 10kg (± 0.3%) spring scale, and prosomal width was measured to the nearest millimeter using Vernalier calipers. The gills of adult crabs were subsampled by removing the upper-right portion of book gills, which constituted approximately 10% of their gills (Fig. 2). Both gill samples and B. candida samples were individually stored in 70% ethanol to be counted at a later time. After sample collection, all adult crabs were released immediately back to the water.
Opportunistic samples of L. polyphemus were obtained from a trawl survey in September-October 2020 from muddy habitat (~ 2m depth) and intertidal beaches of Moriches Bay (40.79°N, -72.71°W). This sample was comprised of 2 juveniles (n = 2; instars 9 & 10; prosomal width range = 40.5 mm − 57.2 mm) and a sub-adult cohort (n = 7; instars 14-18; prosomal width range = 113.0mm -179.0mm). Gill samples from these crabs were not removed, but instead were carefully examined for the presence of B. candida cocoons and adult worms.
To obtain B. candida adult intensity in a standardized format, forceps were used to remove B. candida worms from the book gills and legs over a timed 5-minute period immediately following the physical measurements of the adult crabs. This served as a proxy for catch-per-unit-effort (CPUE). This time Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js approach was implemented for two reasons: 1) every adult atworm could not be accurately removed from the adults by hand or by brief freshwater rinses and 2) we wanted to minimize the desiccation stress to horseshoe crabs during sampling. All adult worms collected during each 5-minute sampling period were stored in 30ml of 70% ethanol in 50mL Falcon conical centrifuge tubes and were processed later in the lab. Prevalence of B. candida was determined by tallying the presence of either adult worms or cocoons on horseshoe crab gill tissues or other appendages. Population prevalence was determined as the percentage of individuals that were infected by B. candida.

B. candida intensity measurements
To measure adult atworm intensity, collected adult atworms were counted and recounted under a dissecting microscope by three readers. Samples were randomly chosen, and readers were blinded from previous recorded intensities in the same sample to minimize bias. The nal intensity count was determined if two or more observers had the same intensity counts. If all counts differed between readers in a sample, a fourth reader counted the sample to nalize intensity based on agreement with another reader. Intensity was de ned as the total count of adult B. candida.
To measure the intensity of B. candida cocoons, each individual lamella was removed from the book gill subset. B. candida cocoons were enumerated under the dissecting microscope for each gill lamella in each gill sample subset. The intensity of cocoons was also enumerated separately on the central mitochondria-rich area (CMRA) and the peripheral mitochondria-poor area (PMPA) of the lamellae. Figure 2 visually illustrates CMRA and PMPA delineations.

Gill and cocoon surface area preparations and measurements
After measuring cocoon intensity, each lamella was placed on a lightbox with a ruler and photographed using a digital Panasonic LUMIX DMC-T380 waterproof camera. Every gill subset was sampled from the ventral to the dorsal side. However, lamella measuring less than 1 cm in diameter (typically the rst few in the book gill) at its widest point were not photographed as they were not observed to harbor any infections and their small size contributed little to total respiratory surface area. One gill sample deteriorated during processing and could not be analyzed.
To measure the proportion of gill surface area covered by the cocoons, cocoon area and gill lamella surface area were measured using ImageJ (version 1.8.0) software (Schneider et al. 2012). The local threshold tool was used to automatically detect and measure the surface area of the cocoons against the lamella, which limited human error. In cases where the color threshold inadequately distinguished cocoons or lamella from each other, manual measurements were made in ImageJ. Average cocoon size was determined by randomly sampling 100 cocoon measurements across all individual adult crabs.

Histological analysis
Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js A small gill sample (~ 2cm x 1cm x.5cm) from a horseshoe crab, not used in the other analyses, was removed and placed in a histo-cassette then xed in 10% buffered formalin, and embedded in para n wax. Heavily infected sections of lamellae were selected to ensure the detection of B. candida cocoons. Sections (~ 5 µm) were mounted on slides and stained with Harris's haematoxylin and Eosin. Multiple slides were cut from the same para n block, and slides were visually inspected for any signs of pathology, such as in ammation, necrosis or encapsulation.

Statistical analyses
A generalized linear model (GLM) was employed to determine which intrinsic factors explained the most variance behind cocoon intensity in adult horseshoe crab gill tissues. Only crabs sampled in 2019 were included in the statistical analyses because only prevalence was evaluated in the opportunistically sampled crabs in 2020. The response variable was cocoon intensity, and the explanatory variables were adult atworm intensities, total gill surface area, and sex in the GLM. Prior to constructing the model, cocoon intensity data were t to Poisson and negative binomial error structures and each distribution was compared by Akaike's information criterion (AIC) in the tdistrplus R package (Delignette-Muller and Dutang 2015; R version 4.0.2, R Core Development Team 2020) to determine the most appropriate error structure given the data. In all GLMs, multicollinearity between the explanatory variables was assessed by calculating the variance in ation factor (VIF) in the Performance R package (Lüdecke et al. 2020). Variables with a VIF greater than 5 were removed from the GLM, as VIFs above 5 are considered to be moderately or strongly correlated with each other and strong correlations between one or more predictor The t of all possible GLM model variants was assessed with the dredge function in the MuMin R package (Barton and Barton 2015). Inference was drawn from models within each GLM candidate set using the small sample size corrected Akaike Information Criteria (AICc) and AIC weights (Burnham and Anderson 2002). Model variants with ΔAICc < 2 and AIC weights > 0.10 were considered to have moderate support (Burnham and Anderson 2002).
A separate GLM was constructed to examine the relationship between adult atworm intensity and the explanatory variables total gill surface area, sex, and horseshoe crab size. Similar to the other GLM, negative binomial and Poisson error distributions were t to adult atworm intensity data and were assessed with AIC. As aforementioned, the multicollinearity analysis was repeated, eliminating size, and model selection was again carried out using the dredge function.
To determine if intensity was homogenous or aggregated across gill tissue space (ventral to dorsal sections), each individual crab's gill subsets were split into four quartile groups and the total cocoon intensity for each gill quartile was summed. A Friedman rank-sum test was used to determine if intensities differed between quartile groups (Friedman 1939). Individual horseshoe crabs were treated as Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js unreplicated blocks, while gill quartiles were considered groups in the Friedman model. If intensities differed between gill quartile groups in the Friedman test, multiple pairwise comparisons between the quartile groups were examined using a post-hoc Conover-Iman test (Conover and Iman 1979). A Bonferroni correction was used to control for multiple comparisons with an assumed familywise error rate of =0.05. The Conover-Iman test was conducted in the PMCMR R package (Pohlert 2014). A Spearman rank correlation test was also employed between the quartile group with the highest cocoon intensity (4th quartile) and the total cocoon intensity across gill samples to determine if the most infected gill quartile could serve as a relative proxy of overall cocoon infection intensity.
To determine if B. candida cocoons were spaced randomly or preferentially across mitochondrial regions of the lamellae in infected individuals, we calculated the average proportion of cocoons on CMRA regions vs. PMPA regions. Each of these regions can be assessed visually given their contrasting pigmentation on gill lamellae ( Fig. 2C; Hans et al. 2018). Additionally, in order to determine if cocoon placement was proportional to the coverage of CMRA and PMPA regions within gill lamellae, CMRA regions of 100 random lamellae were randomly measured using ImageJ. Moreover, we employed a beta regression model (Ferrari and Cribari-Neto 2004) to determine if total cocoon intensities in uenced the proportions of cocoons on the CMRA regions. A beta regression accounts for the fact that proportions are bound between 0 and 1, and error variances are not normally distributed. In the model, cocoon CMRA proportion was the dependent variable, and the total cocoon intensity was the independent variable. Beta regression models were t using maximum likelihood estimation in the betareg package in R (Cribari-Neto and Zeilas 2010).

Prevalence and intensity
From the horseshoe crabs collected in June 2019, B. candida infections (both adult worms and cocoons) were present in 100% of adult gill samples (n = 29), while only 1 juvenile (instars 8-11) from this collection date (n = 30) was observed to harbor any signs of infection (3 cocoons). The opportunistic samples procured in October of 2020, showed a similar pattern with 100% of sub-adults (n = 7; instars 16-18) being observed to have either adult worms and/or cocoons. Of note, the two juveniles obtained in 2020 (i.e instars 9-11) were the least infected relative to all adults/sub-adults with one having no signs of infection and the other having a single cocoon. From the infected adult crabs, both cocoon intensity and adult worm counts varied considerably, as intensities of cocoons and atworms ranged from 37 to 805 (mean = 267 ± 41 SE) and 5 to 196 (58 ± 9 SE), respectively. Adult female horseshoe crabs had an average of 306 ± 42 SE cocoons; whereas, males had a lower average cocoon intensity (85 ± 5 SE cocoons). Mean infection count was 3.2 ± 0.50 SE cocoons per lamella; however, total intensity varied across two orders of magnitude, ranging from 0.4 to 12.2 cocoons per lamella. Total number of gill lamellae (> 1cm diameter) also varied between crabs with counts ranging from 53 to 125 with an average of 91 ± 3 SE. Cocoon surface area differed by four orders of magnitude among adult crabs, covering between 0.10%-14.5% of total gill surface area (based on gill subsamples), with 2.3% ±0.6% SE Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js representing the average (Fig. 3). The proportion of cocoons occupying the CMRA regions ranged from 0.05 to 0.53 (x _ = 0.21 ± 0.02 SE) in infected adult horseshoe crabs (Fig. 4).

Microscopic analysis
B. candida cocoons were elliptical in shape with an anchor shape protrusion extending from one side, and they were attached to the lamellae epithelium via a cement-like excretion that surrounded the cocoon body and the anchor (Figs. 5C-D). After randomly sampling 100 single cocoons, the average cocoon area was 3.37 ± 0.07 SE mm 2 . Histological observations showed healthy horseshoe crab lea ets consisted of parallel lamellae connected via pillar cells, with space between the pillars believed to be vascular channels (Fig. 5A). The dorsal tips of the lea ets were covered in a thickened proteinaceous matrix (Fig.   S1), which also covered the epithelium of the lea ets, albeit not as thick as the blunted dorsal tips (Fig. 5A). Mats of bacteria and algae also frequently covered sections of gill epithelium but were super cial and not observed to cause any in ammation (Fig. 5A). Surface defects that impacted the organization of epithelial tissue and disrupted the typical gill structure were observed in histology sections of infected lamellae with melanization frequently observed on the outer layer of these lesions (Fig. 5E). Granulomas were frequently observed in the vascular lumen of the lamellae causing hemocyte aggregation, encapsulation and in ammation (Figs. 5B, E-F).

Generalized linear model: cocoon intensity
The negative binomial distribution was chosen for the cocoon intensity GLM, as the AIC was lowest for the negative binomial distribution relative to the Poisson distribution (ΔAIC 4673.7). Therefore, a negative binomial was used as the family error structure in the candidate model set. The GLM model selection process revealed that three out of 8 model variants had moderate support (cumulative AIC weight > 0.88) according to ΔAIC values and AIC weights (Table 1). Adult atworm intensity was supported in all three models, but the top model had support for gill surface area as well (Table 1). There was also some support for sex in uencing cocoon intensities (model 3, Table 1). The evidence ratio of the top model relative to model three was ~ 1.9 (Table 1), suggesting that the top model is nearly 2 times more probable than model 3. However, model selection uncertainty was present between the top three models given ΔAIC < 2 (  Table 1 Candidate model results for generalized linear model (GLM) for the cocoon intensity of adult horseshoe crabs after model tting with the dredge package in R. A total of 8 possible model combinations were assessed, but only models with moderate support (Akaike weight > 0.10 and ΔAICc < 2) are shown here (see Methods) and are ranked by AICc. The cumulative Akaike weight between these models comprised 0.88. The global model included sex, adult worm intensity, and gill surface area, as explanatory variables with cocoon intensity set as the response variable. NP represents the number of parameters. The negative binomial distribution was the family error structure used in the GLM, as AIC was lower for the negative binomial distribution relative to the Poisson distribution (ΔAIC 4673.7). The negative binomial distribution was used as the family error structure in the GLM given it was a better t (AIC = 293.9) compared to the Poisson distribution (AIC = 1166.2). Model selection revealed that 3 out of 4 possible model variants relating adult atworm intensity to covariates had moderate to strong support (Table 2) with model selection uncertainty present. The top model only included sex and had an evidence ratio of ~ 3 relative to models 2 and 3, indicating this model is nearly 3 times more probable compared to the other models. Model support for gill surface area (models 2 and 3) was weaker compared to sex in terms of affecting adult atworm intensities ( Table 2). Speci cally, cocoon intensities were different between all quartile group pairwise comparisons (Table 3). Gill quartile 4 (ventral most) had higher cocoon intensities relative to all other gill quartiles (Fig. 6). Gill quartile 4 exhibited a strong positive correlation (Spearman; ρ = 0.96, p = < 0.001) with overall infection intensity in the gill subsample (Fig. 7).  reducing space competition among conspeci c ectoparasites. Additionally, the body size argument may also explain why horseshoe crab sex was a contributing factor in B. candida infection rates because adult females are larger than males (Loveland and Botton 1992) and thus, females have more available surface area or "habitat" for B. candida to reside. However, host size has not been found to be a limiting factor in ectoparasite infection intensities in some organisms; whereas, host whole body metabolism can be a more important determinant of ectoparasite intensities (Hechinger et al. 2019) because host energy can be more constraining to parasite infection loads.

Model
The difference in infection intensities among age groups may partially re ect contrasting foraging behaviors among life history stages and may make the juvenile cohort (< 12 instars) unsuitable for B. candida establishment. For example, B. candida is believed to indirectly consume food particle remnants from horseshoe crab feeding activities (Jennings 1977). Juvenile crabs (instars < 10) predominantly rely on sedimentary organic matter and meiofauna adjacent to salt-marsh habitats (Botton et al. 2003b); in contrast, older juvenile and adult crabs predominantly forage on larger-bodied prey, such as bivalves and polychaetes (i.e. Neries spp.) (Botton and Ropes 1987;Gaines et al. 2002). Therefore, the nutritional resources on juvenile horseshoe crabs may not be su cient or optimal for B. candida's nutritional requirements. However, the theory of B. candida foraging on remnants of horseshoe crab prey items remains controversial due to chemical analysis indicating that B. candida may obtain some nutritional energy directly from horseshoe crabs (Lauer and Fried 1977). The application of modern techniques to assess resource-use, such as stable isotope analysis (bulk or compound-speci c), could be used to resolve this controversy, as it could identify the nutritional resources adult B. candida predominantly relies on.
Given that this study emphasized one population of horseshoe crab hosts, we cannot state these infection intensity patterns apply to other populations, as other factors such as biogeographic differences in reproduction ratios, environmental conditions, migration patterns, abundance, and size may result in varying B. candida. For example, host population density is often positively correlated with ectoparasite intensity as a result of increased probability of direct transmission (e.g. contact, breeding, etc.) among conspeci cs within a population after controlling for other covariates (Arneberg et al. 1998). However, controversy surrounds the contribution of population density to increased parasite infection intensities. al. 1991). Therefore, assessing B. candida population dynamics between horseshoe crab populations with different characteristics (e.g. abundance, density, etc.) may be bene cial for elucidating mechanisms that regulate ectoparasite-host relationships, particularly in such an ancient and stable host species.
Within an individual host or even an organ, parasites are known to aggregate in regions that provide the best niche for them and in turn higher tness, this was observed in the cocoons of B. candida in this study as cocoons of were signi cantly more prevalent in the dorsal most quartile of gill lamellae (Fig. 6).
The ventral most gills were preferentially used for cocoon placement, possibly due to the larger size of these lamellae as larger lamellae size not only provides more habitat, but also allows cocoon location to be away further from edge of the lamellae sheltering the cocoons from excessive ow, a critical concern for ectoparasites (Wootten 1974). Additionally, larger lamellae can pump more water, which may be necessary to meet the oxygen demand of the atworm cocoons which require oxygen to sclerotize (Huggins and Waite 1993). Unsurprisingly, the realized niche of a parasite is frequently smaller than the potential niche (Sukhdeo and Croll 1981), resulting from constraints on attachment, competition and nutrient acquisition, factors that can lead to hyper specialization within a larger organ. For example, the gills of sh are often segregated between parasites such as Monogenean atworms or parasitic copepods which will localize to particular gill arches in sh ( indicating that the placement of B. candida cocoons is nearly proportional to the CMRA. It is important to note that, the CMRA cocoon placement was not entirely avoided in this study and the likelihood of CMRA cocoon placement appeared to slightly increase with cocoon intensity, albeit, the beta regression results indicated a weak relationship between overall cocoon intensity and the proportion of cocoons in the CMRA region. Therefore, in other horseshoe crab populations it is important to determine if the spatial arrangement (random or clustered) of cocoons on gill lamella varies across B. candida intensity levels.
The extent of the deleterious impacts imposed by B. candida infections remains uncertain; however, horseshoe crab tness could be affected from the combination of anthropogenic and ambient environmental stressors coupled with B. candida infection. This study revealed no more than 15% of gill surface area in any adult was covered with B. candida cocoons, however this estimate is likely conservative as our analysis was unable to detect the cocoon cemented regions (Fig. 5C, D). Regardless of this potential underestimation, light infection intensities on gills from ectoparasites may have adverse impacts on horseshoe crab tness. For example, ectoparasite coverage on gills appears to be directly proportional to reduction on the velocity of oxygen uptake in aquatic organisms (Duthie and Hughes 1987) and may potentially affect horseshoe crab tness by reducing respiration e ciency, especially in hypoxic conditions. Hypoxic conditions are expected to become more chronic and frequent in coastal marine environments in the coming decades (Diaz and Rosenberg 2008), and the combined stress of B. candida and hypoxia may adversely impact horseshoe crab physiology. For example, when exposed to Loading [MathJax]/jax/output/CommonHTML/fonts/TeX/fontdata.js hypoxic conditions (< 2.0 O 2 mgL − 1 ) and parasitic nematode (Anguillicola crassus) infections over 4 days, eels with low swim bladder degenerative indexes (0-1) and the highest infection loads exhibited shorter time until death (10-25 hours shorter on average) than their uninfected counterparts (Lefebvre et al. 2007). Additionally, horseshoe crabs face a unique and direct anthropogenic stressor, in the form of blood extraction for biomedical purposes, that may make individuals with intense infections of B. candida more susceptible to sublethal effects (i.e. reduced oxygen uptake, increased respiration energy expenditure, etc.) or mortality (Smith et al. 2017;Owings et al. 2019Owings et al. , 2020. The sublethal impacts of biomedical blood extraction on horseshoe crab survivors (~ 70% survival) are numerous, such as a signi cant reduction in hemocyanin concentrations following typical blood volume extractions (30% blood volume) and reduced spawning frequency (Owings et al. 2019(Owings et al. , 2020, and it can take up to 4 months for amebocytes to fully recover to baseline levels (Novitsky 1984). Hemocyanin is essential for maintaining oxygen transport (Mangum 1980 Additionally, these amebocytes are involved in the immunological response to B. candida cocoons (Fig. 5), so the impairment of these cells could reduce the crabs' ability to respond to immunological insults and/or increase baseline level of stress of these animals and potentially increasing susceptibility to stressors. Therefore, understanding the simultaneous impacts of projected intensifying environmental conditions (i.e. temperature, ocean acidi cation, and hypoxia), biomedical blood harvest on hemocyanin levels, and B. candida infections are required, as their combined effect may engender more severe consequences for horseshoe crabs than when these effects are isolated.     B. candida cocoon intensity for four gill lamellae quartile groups. The rst quartile represents the dorsalmost gill lamellae; in contrast, the fourth quartile represents the ventral-most gill lamellae. Letters represent statistically signi cant differences between gill lamellae quartile groups from Conover post-hoc pairwise comparison tests shown in Table 3  Scatter plot of cocoon intensity from the 4th quartile of gill (ventral-most) lea ets (n=27) vs. total gill cocoon intensity. The Spearman rank correlation results were: ρ=0.96 and p= <0.001. Standard error is represented by grey shading.