Multi-Biomarker Analysis Uncovers High Spatio-Temporal Stability of a Subarctic Rhodolith (Lithothamnion glaciale) Bed Food Web

Abstract

We used lipid, fatty acid, and stable isotope analyses to investigate variation, over nine months, in the trophodynamics of 10 dominant cryptofaunal, macroalgal/algal, and environmental components from two sites within a rhodolith (Lithothamnion glaciale) bed in southeastern Newfoundland (Canada). There was an overall shift from a diatom-based food web following the spring phytoplankton bloom to a kelp/algae-based food web during fall, accompanied by preferred use of EPA (20:5ω3) over DHA (22:6ω3) in most cryptofauna. The food web contained three trophic levels that encompassed: (1) direct feeding relationships from primary producers (e.g., rhodoliths, macroalgae) to second-order consumers (e.g., sea stars, polychaetes); (2) trophic subsidy from within and outside the rhodolith bed via settlement, resuspension, and consumption of macroalgal fragments and other detrital organic matter; and (3) strong pelagic/benthic coupling. Riverine input did not affect cryptofaunal diets, as shown by the lack of terrestrial biomarkers at the study site nearest to the riverine input, and there were minor differences in trophodynamics between both study sites. The present study’s findings, applicable to relatively broad spatial and temporal domains, as well as those of complementary studies of the same rhodolith bed, uncover high spatio-temporal stability of the rhodolith bed framework and of resident cryptofaunal abundance, diversity, and trophodynamics.

1. Introduction

Dietary linkages in marine benthic ecosystems remain largely unresolved because of the broad range of diets of many species and the sometimes complex benthic–pelagic relationships, which can further vary spatially and temporally [1,2,3]. Fluctuations in environmental conditions, including sea temperature [4], salinity [5,6,7], and chlorophyll (an indicator of food availability) [8], can trigger significant changes in marine invertebrate behavior, abundance, biomass, and feeding patterns [9,10,11,12]. Invertebrates can also adapt physiologically to environmental variability by modifying their diet [3], which ultimately affects their lipid class [13], fatty acid [14], and stable isotope composition [15]. Therefore, studying temporal variation in organismal lipids, fatty acids, and stable isotopes has the potential to more fully elucidate feeding relationships and their connection with environmental variability across benthic communities.

Rhodoliths (free-living, non-geniculate red coralline algae growing as balls, branched twigs, or rosettes) often form dense aggregations, known as rhodolith beds, at depths of up to 150 m in tropical to polar seas [16,17]. The relatively complex morphology of rhodoliths creates suitable habitats for attachment [18,19,20], reproduction [6,19,21], and feeding [5,6,7] of highly diverse algal and faunal assemblages. The important contribution of rhodolith beds to marine biodiversity [6,22,23] and global calcium carbonate (CaCO₃) production [24,25,26] has contributed to the recent increase in studies on the factors and processes regulating their structure and function [27,28,29,30].

Knowledge about rhodolith ecosystem dynamics is largely inferred from a few studies on biodiversity patterns [31,32,33,34] and the pervasive effects of human disturbances on rhodoliths or rhodolith bed structure [22,23,34,35]. In general, studies broadly characterize changes in rhodolith bed macrofaunal assemblages without necessarily investigating the influences of environmental variability on such changes. However, because phytoplankton blooms, macroalgae, and bacterial components may contribute just as significantly to nutrition as the macroinvertebrates themselves [36], quantifying solely diversity and abundance may not offer a complete representation of the interconnections in rhodolith communities [3]. As environmental conditions shift seasonally in temperate seas, so may community dynamics, in part because of dietary shifts as resource availability changes [37]. Understanding the complexity of the trophodynamic system requires evaluating the effects of phytoplankton seasonality and seabed composition to support production for different trophic levels.

The present study characterizes spatial and temporal variation in trophic relationships of a large (>500 m²) rhodolith (Lithothamnion glaciale) bed in St. Philip’s, Newfoundland. It builds on previous studies that: (1) characterized rhodolith morphology and invertebrate diversity [6,38]; (2) addressed how the dominant invertebrate bioturbators prevent rhodolith burial [29]; (3) concluded that rhodoliths are resilient to changes in sea temperature [39,40]; and (4) identified bottom-up forcing and benthic–pelagic coupling as a mechanism and pathway for nutrient and energy flow [30] in the St. Phillip’s bed. The latter study was the first to focus on a robust trophodynamic analysis of a rhodolith bed, but it lacked spatial and temporal complexity. Specifically, we use chlorophyll levels from phytoplankton blooms to time three separate collections spanning nine months and analyze rhodoliths and their underlying substratum from two distant sites in the St. Philip’s rhodolith bed to characterize seasonal and spatial variation in lipid composition, nutrition, and trophodynamic relationships of dominant rhodolith cryptofauna. We test the hypotheses that: (1) seasonal fluctuations in sea temperature and food availability affect the lipid composition and diets of cryptofauna; and (2) riverine input affects cryptofaunal diets.

2. Materials and Methods

Unless specified otherwise in the following sections, we used the same field, analytical, and statistical approaches as Hacker Teper et al. [30]. Accordingly, we kept the methodological descriptions brief.

2.1. Study Site and Selection of Focal Species

Field work was completed during the spring, summer, and fall of 2017 in a rhodolith (Lithothamnion glaciale) bed which extends ~5 to 30 m in depth along the coast of St. Philip’s in southeastern Newfoundland, Canada. Consistent with our broader objective of characterizing spatial and temporal variability in rhodolith bed trophodynamics and differences in marine environmental conditions in this area, we chose to study two sites in the bed fringing Broad Cove, hereafter termed the “South” site (47°35′36.5″ N, 52°53′31.0″ W) and the “North” site (47°35′39.6″ N, 52° 53′ 24.6″ W). Both sites and their relative placement are shown in Hacker Teper et al. [30]’s Figure 1A. Adding the analysis of the “North” site to the present study allowed us to investigate trophic variability because of the expected greater environmental variability at this site resulting from its closer proximity to freshwater input from the marina.

We focused our food web analyses on the same six animal, two macroalgal/algal, and two environmental components of the bed: (1) common sea star, Asterias rubens; (2) wrinkled rock-borer, Hiatella arctica (bivalve); (3) juvenile Nereis spp. (polychaetes); (4) daisy brittle star, Ophiopholis aculeata; (5) green sea urchin, Strongylocentrotus droebachiensis; (6) red molted chiton, Tonicella marmorea, and Atlantic red chiton, T. rubra, which were too difficult to distinguish morphologically and, hence, were pooled into the genus Tonicella spp.; (7) pieces of Laminaria digitata [kelp] from nearby kelp beds; (8) Lithothamnion glaciale (rhodoliths); (9) seawater (containing seston) from a few meters above the rhodolith bed; and (10) sediment (containing infauna) underlying the rhodoliths. The present study addresses trophic interactions at both sites across three seasons (spring, summer, fall) and thereby expands knowledge of spatial and temporal dimensions in the Broad Cove rhodolith bed.

2.2. Timing of Sampling

To examine how benthic–pelagic coupling might change during different phytoplankton bloom events, we sampled the rhodolith bed: (1) during and at the end of the annual spring phytoplankton bloom in southeastern Newfoundland, when we anticipated peak diatom abundance in the water column [41,42]; and (2) at the end (early December 2017) of the annual fall phytoplankton bloom, when we expected peak dinoflagellate abundance [43]. The spring bloom began in the last few days of March and persisted through 23 April 2017, when we sampled the rhodolith community and collected rhodoliths for food web analyses. The bloom was declining when we sampled it on 10 July 2017 (see Appendix A for a detailed description of bloom characterization).

2.3. Rhodolith Community

To broadly characterize the rhodolith community, scuba divers hand-collected, on 23 April 2017, all the rhodoliths from one 30 × 30 cm quadrat placed every 5 m along a 30 m long transect at a depth of ~15 m at the South site and North site (for a total of seven quadrats sampled at each site). We chose the North site for its greater proximity to freshwater input and its higher sediment abundance around rhodoliths (personal observations) than the South site. Briefly, we transported the rhodoliths to the Ocean Sciences Centre (OSC) of Memorial University of Newfoundland (MUN) and placed them in large (320 L) holding tanks supplied with running seawater pumped in from the adjacent Logy Bay. We measured, using calipers, the lengths of the longest, intermediate, and short axes, as well as the gross weight of the 247 rhodoliths from the South site and 163 rhodoliths from the North site. We calculated each rhodolith’s sphericity according to Graham and Midgley [44] and Sneed and Folk [45] and as described by Gagnon et al. [6] and Bélanger and Gagnon [38]. To calculate rhodolith abundance (biomass) and cryptofaunal abundance, we weighed each rhodolith, then extracted, identified, weighed, and counted all cryptofauna from the external surface and inside each rhodolith [30].

2.4. Rhodolith Percent Cover and Epiphyte Coverage

To investigate rhodolith epiphyte coverage differences between sites, scuba divers took photographs (with a Canon PowerShot D30, St. John’s, Canada) of rhodoliths from one 30 × 30 cm quadrat placed every 5 m along each of four 15 m long transects at a depth of ~15 m at the South site and North site (for a total of 24 photo quadrats each). From each image, we estimated the epiphyte coverage on rhodoliths and the percent cover of rhodolith, sediment, and macroalgae (kelp and algae) using the point intercept method. A 7 × 7 grid was overlaid onto each image in Adobe Photoshop, and the element present underneath each grid intercept was recorded (i.e., rhodolith, sediment, or macroalgal). We marked rhodolith intercepts for the presence or absence of epiphytes. We measured epiphyte coverage only using intercepts overlaying rhodoliths. The 7 × 7 grid yielded 49 intercepts on each image, each representing ~2.1% cover. The outline of the 30 × 30 quadrat frame provided a reference scale. We used Microsoft Excel to compute various descriptive statistics (see Section 2.10.4).

2.5. Collection and Preparation of Samples for Food Web Analyses

To prevent contamination from our rhodolith and cryptofaunal abundance investigation (see Section 2.3), we collected additional rhodoliths to use with our food web analyses. At both the South and North sites on 23 April, 10 July, and 2 December 2017, divers hand-collected in sealed plastic bags ~150 live rhodoliths measuring 8 to 10 cm along the longest axis. They also collected seawater a few centimeters above the rhodolith bed with two 12 L Niskin bottles deployed gently to prevent resuspension of sediment from the bed, and collected three sediment samples scooped from the top (~10 cm) layer of muddy sediment underneath rhodoliths with 15 mL centrifuge tubes. Briefly, we transported all samples to the OSC within 4 h of collection, where we transferred rhodoliths in their sealed plastic bags to large (320 L) holding tanks supplied with running seawater and kelp and sediment in their sealed plastic bags or centrifuge tubes to freezers at −80 °C. Seawater was processed immediately upon arrival. After 24 h, we removed and processed rhodoliths in preparation for food web analyses. We used a hammer and screwdriver to break open rhodoliths, and tweezers and forceps to extract cryptofauna inside each rhodolith. We scrubbed kelp and rhodolith samples and used a mortar and pestle to grind rhodoliths into a powder. Seawater was processed through a mechanized filtration system, which suctioned water with an aspirator through a 47 mm diameter GF/C filter (Whatman; General Electric, WHA1822047, Oakville, Canada) at the bottom of a Büchner funnel. We obtained separate cryptofauna, kelp, sediment, and seawater specimens for our lipid and FA analyses versus our stable isotope analyses [30].

2.6. Extraction and Characterization of Lipid Classes

The extraction of lipids followed the protocols of Folch et al. [46], with modifications by Parrish [47]. Briefly, we separated samples into two aqueous layers though centrifugation and sonication and used a double pipetting technique to remove only the bottom, organic layer [30]. We then used thin-layer chromatography with flame ionization detection (TLC-FID) (Mitsubishi Kagaku Iatron, Inc., Mark VI Iatroscan, Tokyo, Japan) to characterize lipid classes, using a 3-step method to separate lipid classes [48].

2.7. Preparation and Characterization of Fatty Acid Methyl Esters (FAME)

Fatty acid methyl esters (FAME) of lipids were prepared directly by transesterification from aliquots of lipid extract following a modified procedure described by Christie [49] and Hamilton [50]. This transesterification process created an organic layer of FAME. We analyzed these FAME samples using gas chromatography and flame ionization detection (GC-FID) (HP 6890 gas chromatograph, Agilent Technologies Canada, Inc., Mississauga, Canada), identified fatty acids by retention times on the GC column, and integrated the fatty acid peaks using Varian Galaxie Chromatography Data System V1.9.3.2 software [30].

2.8. Stable Isotope Preparation and Analysis

Samples in their scintillation vials were prepared separately for carbon and nitrogen analyses, transferred to desiccators, and subsequently taken to The Earth Resources Research and Analysis (TERRA) facility at MUN for analysis. Briefly, we analyzed bulk stable isotopes and complementary C and N elemental proportions (%) with the following quantities of solidified samples from each vial: 1 to 1.5 mg for animals; 4 to 4.5 mg for rhodoliths and kelp; 14 to 15 mg for sediment; 5 to 7 mg of filter for seawater δ¹³C and carbon content; and 8 to 10 mg of filter for seawater δ¹⁵N and nitrogen content [30].

2.9. Trophic Magnification of Fatty Acids

Stable isotope ratios are expressed in the conventional (δ) notation as parts per thousand (‰) per the equation of Minagawa and Wada [51]:

δ¹³C or δ¹⁵N (‰) = [(R_sample/R_standard) − 1] × 1000

where R_sample and R_standard are the ratios of ¹³C/¹²C or ¹⁵N/¹⁴N of a given sample and the corresponding standard, respectively. Results are reported relative to atmospheric N₂ for nitrogen stable isotopes, and Vienna Pee Dee Belemnite (VPDB) for carbon stable isotopes. We calculated species trophic position (TP_consumer) with the equation used by Gale et al. [52] and developed originally by Cabana and Rasmussen [53]:

TP_consumer = [δ¹⁵N_consumer − δ¹⁵N_base)/∆¹⁵N] + TP_base

where δ¹⁵N_consumer is the mean stable N isotope ratio of each species and Δ¹⁵N is the fractionation factor, which, to be consistent with rhodolith food web studies, is 3.4‰ [54]. δ¹⁵N_base and TP_base represent the nitrogen stable isotope composition and trophic positions from the base of the food web, respectively. We then calculated a trophic magnification factor (TMF) for fatty acids (FA) correlated with δ¹⁵N. This factor quantitatively represents the biomagnification of compounds along a food web [55,56]. TMF, the average change in a specific FA proportion or marker, was calculated from:

FA% = e^m×TP

or

log_e FA% = (m × TP) + b

and, therefore,

TMF = e^m

where FA% is the fatty acid (as a % of total identified fatty acids) or fatty acid marker of interest; TP is the trophic position; and m and b are the slope and intercept of the linear equation, respectively.

2.10. Statistical Analysis

2.10.1. Lipid Classes

We used two separate two-way permutational MANOVAs (PERMANOVA) (Euclidean distance matrices with 9999 permutations) to test total lipid (N = 139) and lipid classes (N = 147) for temporal and spatial differences with the factors ‘month’ (April, July, and December) and ‘site’ (North and South). We also used separate one-way PERMANOVAs with the factor ‘component’ (each of the 10 components of the food web studied (six animal species, two macroalgal species, and two environmental components)) to examine differences in proportions of total lipids (and overall lipid classes among samples, accounting for accidental loss of samples during the analyses). To limit extraneous data variability while focusing on the most significant lipid classes, only lipid classes present in over 50% of the samples were included in the analysis. For consistency with the companion study [30], and for comparison purposes, we pooled the data into the following five functional groups reflecting the three dominant feeding strategies of the six animal species, two macroalgal species, and two environmental components: (1) suspension/filter feeders (two species); (2) grazers (two species); (3) predators (two species); (4) kelp and rhodolith; and (5) seawater/sediment (samples combined because of expected benthic–pelagic coupling) (see Section 2.1 for details). Using these five groups, we then ran a one-way PERMANOVA with the factor ‘functional group’. We ran one-way ANOVAs to test for relationships between individual components and total lipid or individual lipid classes. We examined relationships between total lipids and each of the major lipid classes with conventional Spearman rank-order correlation tests [57].

2.10.2. Fatty Acids

We used the same statistical approach (one two-way PERMANOVA with the factors ‘month’ and ‘site’ and two one-way PERMANOVAs, one with the factor ‘component’ followed by one with the factor ‘functional group’), with data exclusion to account for samples lost during analyses, to examine differences in the proportions of fatty acids and their proportional sums among samples (N = 142). We then used a one-way SIMPER analysis (run on untransformed data with a Bray–Curtis similarity matrix) with the factor ‘component’ (each of the 10 components of the food web studied (six animal species, two macroalgal species, and two environmental components)) to identify potential food sources and the main fatty acids contributing to the multivariate lipid structure of each component [3,32]. To limit extraneous data variability while focusing on the most significant fatty acids, only fatty acids contributing to over 70% of the similarities were included in the SIMPER analysis. We used a follow-up principal coordinates analysis (PCO; also run on untransformed data with a Bray–Curtis similarity matrix) with the factor ‘component’ (same as above), mainly for visualization of the feeding relationships among specific groups of organisms [58,59]. To increase clarity on the PCO, only samples with a Pearson coefficient of correlation >65% were displayed. We ran two-way ANOVAs to test if individual fatty acid composition among all components changed by month or by site. We also ran individual two-way ANOVAs to test if the proportions of the 8 fatty acids (16:0, 16:1ω7, 18:0, 18:1ω9, 18:1ω7, 20:1ω11, 20:4ω6 [ARA], and 20:5ω3 [EPA]) exhibiting at least 70% correlation changed significantly either seasonally or temporally according to specific components.

2.10.3. Stable Isotopes

We examined differences in carbon (δ¹³C) and nitrogen (δ¹⁵N) isotope ratios with two one-way PERMANOVAs (one for each type of ratio; both were types based on Euclidean distance matrices with 9999 permutations) with the factor ‘component’ and two two-way PERMANOVAs with the factors ‘month and site’. All taxa (six animal species, kelp, and rhodoliths) and both environmental components (seawater and sediment) were included in the δ¹³C isotope ratio analysis (N = 128, accounting for accidental loss of samples during the analyses). All components were also included in the δ¹⁵N isotope ratio analysis (N = 161), accounting for accidental loss of samples during the analyses. We also carried out a cluster analysis (with the “Group Average” clustering method) on δ¹³C and δ¹⁵N isotope ratios simultaneously, and a complementary SIMPROF test (Euclidian distance matrix with 9999 permutations) (N = 120 because of a few unmatched pairs of δ¹³C and δ¹⁵N ratios) to group and map, in the form of a dendrogram, statistically different components of the food web [32,54]. Seven main isotopic groups emerged from the SIMPROF test. We therefore ran a follow-up one-way PERMANOVA with both isotopic ratios combined and complementary one-way ANOVAs and Tukey HSD post hoc tests with the factor ‘group’ to examine differences among these seven main trophic groups.

2.10.4. General Aspects of Statistical Tests

In all PERMANOVAs, data were untransformed and computed on Bray–Curtis similarity or Euclidian distance matrices (9999 permutations). All fatty acid multivariate data were computed using Bray–Curtis similarity matrices, while lipid and stable isotope multivariate data were computed using Euclidean distances matrices due to their better ability to handle missing data. We used pairwise comparisons to further explore PERMANOVA tests with multiple measured variables. We used PERMDISP (9999 permutations) to inform our decision (p = 0.3236); we tested for homogeneity of multivariate variances and confirmed that all variances were homogenous. We used PCO (principal coordinate analysis) instead of PCA (principal component analysis) to more efficiently account for missing data [60]. We combined the North site and the South site in all figures because of the absence of statistical differences (t = 0.723, P (perm) = 0.695). Multivariate analyses were also conducted on fatty acid, isotope, lipid class, fluorescence, and temperature data together using Bray–Curtis similarity. When all the chemical and environmental variables were combined, lipids were the major drivers of significant differences (see Appendix B for additional justification of the general statistical approach). In all ANOVAs, we verified the homogeneity of variance and normality of residuals by examining the distribution of the residuals and the normal probability plot of the residuals, respectively [61]. We used a significance level of 0.05 in all analyses and reported all means with standard deviations (mean ± SD) unless stated otherwise. We used standard error where applicable to maintain consistency with corresponding literature [6,56,62]. We used PRIMER v7 with PERMANOVA+ for multivariate statistical analyses, Minitab 18 for univariate statistical analyses, and Microsoft Excel for descriptive statistics.

3. Results

3.1. Rhodolith Community

Rhodolith biomass was similar between the South (19.5 ± 0.1 (SE) kg m⁻²) and North (20.2 ± 0.2 kg m⁻²) sites. However, the significantly fewer rhodoliths (N = 163, p < 0.001) at the North site were significantly larger (114.1 ± 9.9 cm³, p < 0.001) than the 247 rhodoliths at the South site (70.0 ± 4.5 cm³). Rhodolith shape did not vary significantly between sites; together, they were predominantly spheroidal and compact (~62%), but otherwise, they were platy (~11%), bladed (~16%), or elongated (~6%) (Figure 1). Sediment (p = 0.264) and rhodolith (p = 0.074) percent cover did not vary significantly between sites, but epiphyte coverage per rhodolith was significantly higher (p < 0.001) at the North site (33.0 ± 3.0%) than the South site (14.1 ± 2.4%) (Figure 2). The South site’s macrofaunal biomass (34.5 ± 4.3 g kg⁻¹ rhodoliths) nearly doubled that of the North site (18.8 ± 8.4 g kg⁻¹ rhodoliths) (p = 0.024). In total, the 2918 animals extracted from the rhodoliths from the North (N = 1727) and South (N = 1191) sites did not differ significantly (p = 0.226) in terms of total abundance, but included at least 21 species representing six phyla. Echinoderms (528 ± 61 individuals kg⁻¹ rhodoliths) and molluscs (516 ± 53 individuals kg⁻¹ rhodoliths) numerically dominated both sites (

Table 1. Taxonomical breakdown and abundance of invertebrate macrofauna associated with rhodoliths (Lithothamnion glaciale) collected in April 2017 at the South and North sites. Each phylum’s total abundance (bolded values) includes macrofauna that could not be identified to the genus level.

Phylum/Species	Mean (±SE) Density (Individuals kg−1 Rhodoliths)
	South	North
Annelida	75.9 (8.3)	88.1 (21.5)
Myxicola spp.	8.9 (4.7)	6.7 (3.0)
Nerididae (including Nereis spp.)	21.7 (4.7)	17.8 (4.4)
Potamilla reniformis	30.0 (10.3)	41.0 (13.7)
Arthropoda	72.8 (13.1)	125.8 (19.2)
Amphipoda	34.1 (12.4)	29.7 (9.1)
Cancer irroratus	1.8 (1.0)	4.6 (2.2)
Pandalus borealis	36.9 (7.5)	91.5 (16.3)
Echinodermata	452.7 (47.0)	603.1 (97.4)
Asterias rubens	19.7 (4.1)	10.8 (4.2)
Ophiopholis aculeata	336.7 (30.8)	381.7 (68.9)
Ophiura robusta	72.7 (13.8)	193.5 (37.7)
Strongylocentrotus droebachiensis	22.8 (3.7)	17.2 (6.2)
Mollusca	427.6 (39.6)	603.7 (78.2)
Hiatella arctica	152.8 (23.2)	171.0 (26.6)
Lacuna vincta	1.8 (1.0)	2.6 (1.5)
Margarites costalis	15.8 (3.6)	33.9 (4.8)
Modiolus modiolus	14.7 (3.5)	22.1 (5.4)
Moelleria costulata	8.6 (3.0)	11.0 (3.6)
Puncturella noachina	21.2 (5.4)	51.3 (9.1)
Tonicella marmorea / T. rubra	191.6 (27.2)	260.9 (36.2)
Turbonilla spp.	3.4 (2.3)	5.6 (2.9)
Velutina velutina	2.4 (1.5)	3.5 (1.5)
Nemertea	19.3 (5.5)	19.5 (4.7)
Sipuncula	5.8 (5.3)	6.9 (3.6)

). Species included in the biochemical analyses were particularly abundant at both sites, including Ophiopholis aculeata (359 ± 58 individuals kg⁻¹ rhodoliths), Tonicella marmorea/T. rubra (226 ± 25 individuals kg⁻¹ rhodoliths), and Hiatella arctica (161.9 ± 31.5 individuals kg⁻¹ rhodoliths). A few species not included in the analyses were also relatively abundant, including the brittle star Ophiura robusta (133 ± 27 individuals kg⁻¹ rhodoliths), the caridean shrimp Pandalus borealis (64 ± 12 individuals kg⁻¹ rhodoliths), and the polychaete Potamilla reniformis (35.5 ± 8.8 individuals kg⁻¹ rhodoliths).

3.2. Total Lipid Content and Lipid Classes

Of the ten food web components included in the lipid analysis, the brittle star O. aculeata contained the highest average concentrations of total lipids (9.8 ± 4.1 mg g⁻¹ [ww]), ranging from 8.9 ± 1.6 mg g⁻¹ at the North site in December to 15.0 ± 5.5 mg g⁻¹ at the South site in July (Table S1). Both the macroalgal and environmental groups had significantly lower total lipid concentrations (0.4 ± 0.1 to 2.7 ± 3.0 mg g⁻¹) than the animal group (t > 4.29, p < 0.01 in all cases). Total lipid varied significantly by component (PERMANOVA; Pseudo-F_9,129 = 12.225, P (perm) < 0.01), but not between sites (Pseudo-F_1,138 = 0.321, P (perm) = 0.55) nor among months (Pseudo-F_2,138 = 0.813, P (perm) = 0.16). The interaction was also not significant.

The ten food web components contained nine lipid classes (PL, TAG, FFA, ST, AMPL, HC, SE, KET, and ALC, see Appendix C), with six (PL, TAG, FFA, ST, AMPL, and HC) present in >50% of all samples. PL was present and the dominant lipid class in every component, except for O. aculeata (North site, 35.9 ± 3.4%) and Tonicella spp. (South, 36.8 ± 4.5%; North, 33.4 ± 2.4%) in July, with an average proportional contribution to total lipid concentration of 44% in O. aculeata to 70% in seawater (Table S1). Proportions of PL of all combined components did not vary by site nor month; however, Tonicella spp. (ANOVA, F_2.14 = 10.84, p < 0.01), L. digitata (F_2.7 = 35.98, p < 0.01), and seawater (F_2.16 = 22.80, p < 0.01) each significantly varied individually by month. High average proportions of TAG were most common in animal species, with 30% in O. aculeata, 29% in H. arctica, and 28.2% in Tonicella spp. TAG proportions varied significantly among some individual components; Tonicella spp. (F_2,14 = 57.47, p < 0.01), Nerididae (F_2,16 = 4.25, p = 0.038), and S. droebachiensis (F_2,15 = 4.61, p = 0.033). ST was most prevalent in animal species, with average high proportions of 20% in S. droebachiensis and 18% in Nereis spp. ST proportions varied significantly among some individual components, such as Tonicella spp. (F_2.14 = 8.26, p = 0.006) and seawater (F_2,16 = 5.48, p = 0.019). PL to ST ratios in animal species significantly varied from April to December (pairwise PERMANOVA, t = 2.6363, P (perm) = 0.005) and nearly significantly from April to July (t = 1.8437, P (perm) = 0.055), but not between July and December (t = 0.9321, P (perm) = 0.373). (Table S1).

Overall lipid class composition varied significantly among components (PERMANOVA; Pseudo-F_9,137 = 40.385, P (perm) < 0.01) and five functional groups (suspension/filter feeder, grazer, predator, macroalgal, and environmental components) studied (Pseudo-F_4,142 = 19.363, P (perm) < 0.01), but did not vary monthly (Pseudo-F_2,146 = 0.568, P (perm) = 0.68), nor between sites (Pseudo-F_1,146 = 0.121, P (perm) = 0.94). The interaction was also not significant. Total lipid concentration significantly correlated with PL (r = 0.094, p = 0.036, N = 138), TAG (r = 0.409, p < 0.01, N = 138), FFA (r = −0.115, p < 0.01, N = 138), and ST (r = −0.095, p = 0.02, N = 138).

3.3. Fatty Acid Profiles

The ten food web components included in the fatty acid (FA) analysis contained 63 FAs, with 50 present in >50% of all samples. Each component exhibited a distinct set of dominant fatty acids and biomarkers (Table S2, Figure 3). With a proportional contribution to FA profiles ranging from 14% in H. arctica in December to 32% in A. rubens in April, EPA (eicosapentaenoic acid, 20:5ω3; a typical diatom biomarker) was typically the dominant FA within each of the six animal species sampled (Table S2). In December, ARA (arachidonic acid, 20:4ω6; a kelp and amphipod biomarker) and DHA (docosahexaenoic acid, 22:6ω3; a dinoflagellate biomarker) were, respectively, the most abundant FA in S. droebachiensis (~25%) and H. arctica (~20 to 24%). Across all months, ARA, vaccenic acid (18:1ω7; an aerobe, bacteria, and marine vegetation biomarker), palmitoleic acid (16:1ω7; a diatom biomarker), and palmitic acid (16:0; a common flagellate and marine vegetation biomarker [3,63,64,65]) were the next most prominent FA among the animals, with contributions peaking between 8 and 25%.

Macroalgal components were rich in palmitic acid (18 to 31%) and EPA (9 to 23%). Additionally, large proportions of oleic acid (18:1ω9, 8 to 18%, a common crustacean, detritus, dinoflagellate, and brown seaweed biomarker [3,66,67,68,69]) and linoleic acid (18:2ω6, 4 to 14%, a vegetation biomarker) were present in kelp, while 16:1ω7 (5 to 8%) was abundant in rhodoliths. Seston FAs were dominated by stearic acid (18:0, 18 to 33%; a common detritus biomarker [69,70]), 16:0 (15 to 34%), and 18:1ω9 (8 to 33%) (Table S2). Overall, FA composition differed significantly among the ten food web components (PERMANOVA; Pseudo-F_9,132 = 53.596, P (perm) < 0.01) and five functional groups (Pseudo-F_4,137 = 27.898, P (perm) < 0.01) studied.

FA composition changed significantly monthly (Pseudo-F_2,141 = 2.09, P (perm) = 0.025), but a subsequent pairwise test revealed that it changed specifically only between April and December (PERMANOVA, t = 1.769, P (perm) = 0.0123). FAs at each site were not significantly different (PERMANOVA, Pseudo-F_1,141 = 1.575, P (perm) = 0.14), nor was the interaction effect. In addition to the slight significant changes between months, nine individual FA profiles significantly differed seasonally (

Table 2. Significant seasonal changes in all fatty acids in the six animal species, two macroalgal species, and two environmental components sampled in April, July, and December 2017 at the South and North sites.

FA	Factor	F Value	p Value
14:1	Month	3.24	0.036
i15:0	Month	3.46	0.035
15:0	Month	6.84	<0.010
ai17:0	Month	9.02	<0.010
17:0	Month	3.90	0.023
17:1	Month	12.87	<0.010
16:4ω3	Month	18.33	<0.010
20:3ω6	Month	4.80	0.010
20:4ω3	Month	3.30	0.032

). Of the eight FA which exhibited at least 70% correlation in the ten components (Figure 3), seven (16:1ω7, 18:0, 18:1ω9, 18:1ω7, 20:1ω11, 20:5ω3 [EPA], and 20:4ω6 [ARA]) showed significant changes among individual components (

Table 3. Significant seasonal changes in specific components within the eight fatty acids exhibiting at least 70% correlation in the six animal species, two macroalgal species, and two environmental components sampled in April, July, and December 2017 at the South and North sites.

FA	Component	Factor	F Value	p Value
16:1ω7	Tonicella spp.	Month	51.96	<0.010
16:1ω7	H. arctica	Month	50.22	<0.010
16:1ω7	S. droebachiensis	Month	10.60	0.003
16:1ω7	A. rubens	Month	6.31	0.015
18:0	Tonicella spp.	Month	8.60	<0.010
18:0	H. arctica	Month	49.64	<0.010
18:0	A. rubens	Month	12.84	<0.010
18:0	Seawater	Month	12.82	<0.010
18:1ω9	S. droebachiensis	Month	4.77	0.035
18:1ω7	H. arctica	Month	19.62	<0.010
18:1ω7	A. rubens	Month	9.75	<0.010
18:1ω7	Sediment	Month	17.27	<0.010
20:1ω11	Tonicella spp.	Month	10.31	<0.010
20:1ω11	H. arctica	Month	12.26	<0.010
20:1ω11	Nereis spp.	Month	14.10	<0.010
20:1ω11	S. droebachiensis	Month	35.5	<0.010
20:1ω11	Sediment	Month	12.55	<0.010
20:5ω3	Tonicella spp.	Month	12.21	<0.010
20:5ω3	H. arctica	Month	16.80	<0.010
20:5ω3	Nereis spp.	Month	5.15	0.029
20:5ω3	S. droebachiensis	Month	12.99	<0.010
20:5ω3	A. rubens	Month	18.68	<0.010
20:5ω3	Seawater	Month	35.86	<0.010
20:5ω3	Sediment	Month	8.64	0.010
20:4ω6	Tonicella spp.	Month	13.55	<0.010
20:4ω6	H. arctica	Month	12.41	<0.010
20:4ω6	Nereis spp.	Month	7.44	0.010
20:4ω6	S. droebachiensis	Month	29.14	<0.010
20:4ω6	L. digitata	Month	13.31	0.017

).

Figure 3. (A) PCO plot (based on Bray–Curtis similarity matrices) of the 8 fatty acids exhibiting at least 70% correlation and DHA, which had 21% correlation in the six animal species, two macroalgal species, and two environmental components sampled in April, July, and December 2017 at the South and North sites. (B) Typical fatty acid trophic biomarkers for fatty acids included in the analysis (adapted from Parrish [10] and Legeżyńska et al. [71]) (Dinos = Dinoflagellates).

Figure 3. (A) PCO plot (based on Bray–Curtis similarity matrices) of the 8 fatty acids exhibiting at least 70% correlation and DHA, which had 21% correlation in the six animal species, two macroalgal species, and two environmental components sampled in April, July, and December 2017 at the South and North sites. (B) Typical fatty acid trophic biomarkers for fatty acids included in the analysis (adapted from Parrish [10] and Legeżyńska et al. [71]) (Dinos = Dinoflagellates).

Of the three essential fatty acids (EPA, DHA, and ARA), EPA was the most abundant, being present in all food web components, but particularly among the six animal species (Table S2). ARA was present in all components except for trace amounts in seston, peaking in December in S. droebachiensis (25%) at higher proportions than EPA. All components had DHA, which was typically less abundant than EPA and ARA, except in H. arctica, which peaked at both sites in December with proportions higher than EPA (20–24%) (Table S2). Animal FA profiles were richest in polyunsaturated FA (PUFA), which ranged from 38% in Tonicella spp. in July to 67% in A. rubens in December (Table S3). Monounsaturated FAs (MUFAs) were lower, varying from 17% in S. droebachiensis in April to 35% in Tonicella spp. in July, but were nevertheless still generally more abundant than saturated FAs (SFAs), ranging from 8% in A. rubens in December to 27% in Nereis spp. in April.

Environmental components and L. glaciale presented lower levels of PUFA than the animals, ranging from 15% in seston in December to 39% in L. glaciale in July, and typically contained higher proportions of MUFA (35–43% in infauna) and SFA (17–65% in seston). The average DHA/EPA ratio was significantly higher in December than in April (pairwise PERMANOVA, t = 2.780, P (perm) = 0.01) and July (t = 2.785, P (perm) = 0.01) and was highest in H. arctica and seston (0.8 and 0.6, respectively) with both peaking in December at the South site. Conversely, the DHA/EPA ratio showed low values in sediment, L. digitata, Tonicella spp., A. rubens, S. droebachiensis, and L. glaciale (0.1 to 0.3), and were below detection (0) in Nereis spp. and O. aculeata (Table S3).

Proportional sums of FA varied significantly among the 10 components (PERMANOVA, Pseudo-F_9,132 = 37.054, P (perm) < 0.01) and five functional groups studied (Pseudo-F_4,137 = 45.993, P (perm) < 0.01), but not by site (Pseudo-F_1,141 = 1.759, P (perm) = 0.18) nor month (Pseudo-F_2,141 = 0.41, P (perm) = 0.77). The interaction was also not significant.

3.4. Stable Isotopes and Trophic Magnification

Stable carbon and nitrogen isotope ratios (δ¹³C and δ¹⁵N) showed no significant seasonal or temporal changes, except in kelp, which was more carbon-δ¹³-enriched in July and December (−15.2‰) than in April (−21.0‰) (δ¹³C, PERMANOVA, Pseudo-F_2,6 = 332.97, P (perm) = 0.013) and increased its nitrogen-δ¹⁵ from 1 to ~1.98 during the same period (δ¹⁵N, Pseudo-F_2,6 = 300.97, P (perm) = 0.005).

Stable carbon isotope ratios (δ¹³C) differed significantly among the 10 food web components included in the carbon isotope analysis (Pseudo-F_9,118 = 50.929, P (perm) < 0.001), ranging from most depleted in seawater in December at the North site (−27.5‰) to most enriched in L. digitata in July at the South site (−15.1‰) (Table S4).

Stable nitrogen isotope ratios (δ¹⁵N), which were lowest in L. digitata in April (3.4‰) and highest in A. rubens in April and December (11.4‰) (Table S4), also differed significantly among the ten food web components included in the nitrogen isotope analysis (Pseudo-F_9,151 = 135.15, P (perm) < 0.001), indicating distinct trophic levels (see below).

Hierarchical clustering analysis of δ¹³C and δ¹⁵N separated the ten components in seven distinct groups (Psuedo-F_6,113 = 127.29, P (perm) < 0.001; Figure 4 and

Table 4. Tukey HSD test of bulk carbon (δ¹³C) and nitrogen (δ¹⁵N) stable isotope ratio groups based on agglomerative hierarchical cluster analysis. The six animal species, two macroalgal species, and two environmental components sampled in April, July, and December 2017 at the South and North sites were included in the analysis.

Component	Group	Sub-Group	(δ13C) Tukey HSD	(δ15N) Tukey HSD
L. digitata	1	-	C	D
Seawater	2	a	A	B
Seawater		b
Seawater, Nereis spp.	3	a	B	C
Seawater, Sediment		b
Seawater, Nereis spp.		c
Sediment		d
H. arctica	4	a	C	B, C
L. glaciale		b
A. rubens, Nereis spp., O. aculeata, Tonicella spp.	5	a	C	A
Nereis spp., O. aculeata		b
A. rubens, S. droebachiensis		c
Tonicella spp., S. droebachiensis		d
A. rubens	6	-	D	A
Tonicella spp.	7	a	D	B, C
L. digitata		b

). Two of these groups each contained all samples of a single food web component, namely, L. digitata (kelp) and A. rubens (sea star), hereafter termed Group 1 and Group 6, respectively (Figure 4 and Table 4).

Group 2 also had only a single food web component, seawater (seston), but was separated into two sub-groups (Groups 2a and 2b).

Group 3 contained four subgroups, of which only one was monospecific, but each included at least one of Nereis (polychaete) (Groups 3a and 3c), seawater (seston) (Groups 3a, 3b, and 3c), or sediment (Groups 3b and 3d) (Figure 4 and Table 4).

Group 5 also included four subgroups, with each containing at least one of A. rubens (Groups 5a and 5c), Nereis spp. (Groups 5a and 5b), O. aculeata (Groups 5a and 5b), S. droebachiensis (Groups 5c and 5d), or Tonicella spp. (Groups 5a and 5d) (Figure 4 and Table 4).

Groups 4 and 7 each had two distinct subgroups of monospecific components: H. arctica (Group 4a), L. glaciale (Group 4b), Tonicella spp. (Group 7a), and L. digitata (Group 7b) (Figure 4 and Table 4).

Group 1 contained L. digitata, sampled in April, while Group 7b contained L. digitata, sampled in July and December. 20:5ω3 [EPA], and 22:4ω6 exhibited a TMF > 1, and hence were biomagnified through trophic levels, whereas 21 had a TMF < 1, indicating biodilution (

Table 5. Trophic multiplication factor (TMF) of 33 fatty acids (FA) as calculated from the slope (m) of corresponding linear relationship between FA concentration and bulk nitrogen (δ¹⁵N) stable isotope ratio [30]. Six animal species, two macroalgal species, and two environmental components sampled in April, July, and December 2017 at the South and North sites were included in the analysis. Only FAs with statistically significant correlation coefficients (r) are shown.

FA	TMF	m (±SE)	b (±SE)	r	p-Value
20:3ω3	2.76	1.0 (0.3)	8.1 (0.2)	0.3	<0.001
20:1ω9	2.65	1.0 (0.1)	7.3 (0.2)	0.6	<0.001
18:3ω6	2.58	0.9 (0.4)	8.0 (0.2)	0.3	0.015
20:2ω6	2.50	0.9 (0.2)	7.3 (0.2)	0.4	<0.001
20:2a	2.11	0.7 (0.1)	7.7 (0.2)	0.6	<0.001
16:4ω3	1.41	0.3 (0)	7.6 (0.2)	0.5	<0.001
17:1	1.33	0.3 (0.1)	8.1 (0.2)	0.1	<0.001
20:1ω11	1.32	0.3 (0)	7.4 (0.1)	0.7	<0.001
18:1ω11	1.24	0.2 (0.1)	8.2 (0)	0.4	0.023
22:4ω6	1.21	0.2 (0.1)	8.2 (0.2)	0.1	0.020
20:4ω6	1.17	0.2 (0)	7.4 (0.2)	0.6	<0.001
20:5ω3	1.14	0.1 (0)	6.2 (0.2)	0.6	<0.001
18:1ω9	0.95	−0.1 (0)	8.7 (0.2)	−0.4	0.002
22:6ω3	0.94	−0.1 (0)	8.6 (0.2)	0.0	0.044
16:0	0.87	−0.1 (0)	10.2 (0.2)	−0.7	<0.001
16:1ω7	0.87	−0.1 (0)	9.2 (0.2)	−0.4	<0.001
14:0	0.85	−0.2 (0.1)	8.9 (0.3)	−0.2	0.032
i15:0	0.85	−0.2 (0.2)	9.1 (0.2)	−0.5	<0.001
18:4ω3	0.76	−0.3 (0.1)	8.9 (0.2)	−0.3	<0.001
21:5ω3	0.71	−0.3 (0.1)	8.6 (0.2)	−0.2	0.011
16:1ω11	0.64	−0.5 (0.1)	8.6 (0.2)	−0.4	0.001
18:4ω1	0.58	−0.6 (0.2)	8.6 (0.2)	−0.2	0.028
18:3ω3	0.57	−0.6 (0.1)	8.8 (0.2)	−0.3	<0.001
ai15:0	0.51	−0.7 (0.1)	8.8 (0.2)	−0.3	<0.001
16:3ω3	0.50	−0.7 (0.3)	8.5 (0.2)	−0.4	0.035
16:3ω4	0.48	−0.7 (0.2)	8.7 (0.2)	−0.3	0.001
i16:0	0.34	−1.1 (0.4)	8.7 (0.2)	−0.2	0.005
15:0	0.32	−1.1 (0.4)	9.0 (0.2)	−0.1	0.004
16:2ω4	0.22	−1.5 (0.4)	9.1 (0.3)	−0.3	0.001
16:1ω5	0.16	−1.8 (0.3)	9.1 (0.2)	−0.4	<0.001
16:1ω9	0.13	−2.1 (0.5)	8.9 (0.2)	−0.4	<0.001
14:1	0.07	−2.6 (1.0)	8.6 (0.2)	−0.2	0.011
i17:0	0.07	−2.7 (0.3)	9.4 (0.2)	−0.5	<0.001

).

4. Discussion

We used lipid, fatty acid, and stable isotope analyses to investigate the spatial and temporal variation in the trophodynamics of 10 animal and environmental components from two locations (sites) within a well-studied rhodolith (L. glaciale) bed in southeastern Newfoundland. In a previous study of the same rhodolith bed, we identified macroalgal-based detritus, but not kelp, as a key rhodolith community food source; identified three distinct trophic levels (producers, suspension/filter feeders and grazers, and predators); and suggested a direct link between an algae-based diet and carbon source in three cryptofaunal species (H. arctica, Tonicella spp., and O. aculeata) [30]. The present study’s broader spatial and temporal domains further strengthen the evidence for benthic–pelagic coupling, while unveiling generally high spatio-temporal stability in trophic linkages despite a species-specific ability to adjust diet based on food (including kelp) availability.

4.1. Rhodolith Community

Macroalgal biomass and species composition commonly shift seasonally in rhodolith beds because of changing environmental conditions such as water temperature and nutrient availability [17,22]. In the Gulf of California, foliose red algae peaked in the winter [22], but were seen only in the fall (September and October) in the present study. This difference could reflect greater light availability in the Newfoundland summer [72] and similar temperature conditions in both studies, where annual summer water temperature in the St. Philip’s bed reaches up to 17 °C [73] and the annual winter water temperature in the Gulf of California reaches up to 18 °C [74]. Interestingly, foliose red algae cover was one of the few components that differed between the North (33%) and South (14%) sites in the present study. Smith et al. [75] attributed the increased biomass of marine epiphytic algae in coastal marine systems to eutrophication. We could hypothesize that the potential influence of freshwater feeding directly into the North site is due to its greater proximity to riverine input. However, the low abundance of terrestrial fatty acid biomarkers (18:2ω6 and 18:3ω3) and the larger rhodoliths at the North (~114 cm³) compared to the South (~70 cm³) site suggests otherwise. Bélanger and Gagnon [72] showed that eutrophication inhibits rhodolith growth. Based on preliminary diving observations of sediment patches between rhodoliths, we assumed that rhodoliths at the North site were larger because of a slightly greater sedimentary load from the riverine source stabilizing the rhodoliths [6]. However, our statistical analysis showed no difference in sediment cover between the North (26%) and South (31%) sites. Moreover, the larger rhodoliths at the North site contained a lower macrofaunal biomass (~18 g kg⁻¹ rhodoliths) than the smaller rhodoliths at the South site (~34 g kg⁻¹ rhodoliths). It is possible, therefore, that the larger and hollow, or nucleated but structurally less complex, rhodoliths at the North site provided less space or substratum for cryptofaunal recruitment than the smaller rhodoliths at the South site. Further studies of epiphytic cover and cryptofaunal biomass closer to the freshwater input are needed to better explain this pattern.

4.2. Lipid Content and Class Composition

Lipid content fluctuates in organisms through seasonal variability associated with changing environmental conditions, food availability, metabolic states, or reproductive status [76,77,78]. In the present study, although individual lipid classes of some individual components varied seasonally (TAG in Nereis spp., S. droebachiensis, L. digitata; ST in O. aculeata; AMPL in S. droebachiensis; PL in L. digitata; and HC in seawater), a significant overall lipid content change occurred only in Tonicella spp. We detected significantly higher total lipid (~13 mg g⁻¹) and lower PL levels (~35%) simultaneously with elevated levels of TAG (~45%) in July than in April or December. This change in lipid content suggests a seasonal diet shift to accommodate reproductive timing. A decrease in PL and an increase in short-term reserves (TAG) indicate that PL was expended and replaced with TAG to enable rapid turnover of the energy required for reproduction. Lee at al. [79] suggest that phospholipids have a storage function in cold-ocean invertebrates. Langer [80] observed peak gonad indices in April and May in Maine (USA), indicating reproductive timing of Tonicella marmorea, one of our sampled species. Based on the geographic difference between Maine and Newfoundland and lipid class changes in our study, the reproductive season of T. marmorea may begin later in Newfoundland.

In addition to organisms changing their lipid compositions in response to reproductive needs, we also expected to see lipid composition changes based on prevailing environmental conditions. In Hacker Teper et al. [30], we suggested that high proportions of phospholipids and unsaturated fatty acids, as well as low proportions of sterols, reflected the cold-water (~0.3 °C) conditions of the Northwest Atlantic in April [73]. The warmer waters of July (~6–8 °C) and December (~4 °C) could have resulted in the need for a less fluid lipid membrane through lower proportions of phospholipids and unsaturated fatty acids and higher proportions of sterols and saturated fatty acids [81]. We observed a decrease in phospholipid-to-sterol ratios from April (9.1) to July (6.3) to December (5.9), but not in polyunsaturated-to-saturated fatty acid ratios (2.6 vs. 2.2 vs. 2.7). This trend could suggest that the more rigid cell membrane lipid structure in July and December reflects the warming temperatures, whereas the unchanged fatty acid composition suggests strong benthic–pelagic coupling and a reliance on the polyunsaturated fatty acids available from major food sources (EPA from diatoms or DHA from dinoflagellates) [82].

4.3. Fatty Acids

As discussed in Hacker Teper et al. [30], our study sites presented high levels of EPA (~17%) and low levels of DHA (~3%) in April, probably because of the spring diatom phytoplankton bloom (~0.4 mg chlorophyll m⁻³). The present study showed similarly high levels of EPA (~18%) and low levels of DHA (~3%) at the end of the phytoplankton bloom in July (~0.07 mg m⁻³), likely because of the continuous sinking and accumulation of phytoplankton and organisms that utilize them on the seabed. In shallow waters, diatoms often reach the benthos intact following the spring bloom [81]. However, after the fall dinoflagellate phytoplankton bloom in December (nearly 0 mg m⁻³), we observed lower levels of EPA (~14%) and higher levels of DHA (~4%), which caused a significantly higher ratio of DHA compared to EPA (0.47 vs. 0.19 and 0.23). Combined with a lack of overall seasonal change between April and July, this change in diet suggests that organisms prioritize EPA and its availability over DHA. Although these trends were consistent across our entire sample set, a few species (H. arctica, O. aculeata, and A. rubens) exemplify this phytoplankton and fatty acid biomarker shift. For example, in H. arctica, the proportion of EPA decreased from April (~20%) and July (~22%) to December (~15%), whereas the proportion of DHA increased substantially from April (~10%) and July (~12%) to December (~22%). This change in a fatty acid biomarker provides clear evidence of a strong benthic–pelagic coupling link that resulted in filter- and suspension-feeding invertebrates utilizing the available phytoplankton; feeding is likely based on a “bloom” vs. “no bloom” scenario rather than on distinct seasons [83].

In Hacker Teper et al. [30], we suggested that A. rubens preys on H. arctica given their similar FA compositions and A. rubens’ feeding preference for mollusks [84]. Consistent levels of DHA in A. rubens from April through December compared to high levels of DHA observed in H. arctica support our hypothesis that H. arctica represents a major food source for A. rubens. Even though we cannot explain why there was little DHA in the diets of other invertebrates, the increase in the overall abundance of ARA from April (~6%) and July (~5%) to December (~8%) may explain their low levels of DHA. For example, in O. aculeata, the proportion of EPA decreased from April (~22%) and July (~26%) to December (~15%), whereas the proportion of ARA doubled from April (~1%) and July (~1%) to December (~2%). Similar increases in proportional ARA also occurred in Tonicella spp., Nereis spp., and S. droebachiensis. These increases in ARA rather than DHA suggest that, in the absence of abundant diatoms (EPA), organisms in the Newfoundland rhodolith community actively select an algal and kelp-derived (ARA) diet rather than feeding on abundant dinoflagellates (DHA). Kelp also becomes more abundant in the summer and early fall, providing an alternative food source for primary consumers [85].

Based on our findings and on Hacker Teper et al. [30], we hypothesize that a resource-partitioning relationship may exist where multiple species utilize different forms of the same food (i.e., algae/kelp and particulate algae/kelp) through various feeding strategies (i.e., suspension feeding and grazing), where the organisms benefit from the residual food from their co-habitants (Figure 5). Accordingly, because suspension-feeding O. aculeata and grazing Tonicella spp. likely cannot utilize large pieces of drift kelp, they could potentially utilize residual feeding matter in the form of particulate kelp after S. droebachiensis or other larger organisms have grazed on it and broken it down [86,87]. Simultaneously, the complex physical structure of rhodoliths might trap algal particulates that subsequently become more enriched through microbial activity, resulting in increased nutritive value through increased nitrogen content [88,89]. Nereis spp. may not require large amounts of DHA [90], and, given the similarity between its fatty acid profile and that of O. aculeata and Tonicella spp., it likely receives ARA by preying or scavenging on organisms in the rhodolith community rather than feeding directly on a kelp source. However, depending on the specific Nereis species and population, we propose at least three different foraging strategies. Nereis spp. could: (1) prey on rhodolith biota [91]; (2) consume detrital and microbenthic algae in the uppermost portion of the sediment layer on which the rhodoliths sit [92]; and (3) filter feed on phytoplankton [93]. Some populations may also switch from one mode to another [92]. In fact, N. diversicolor can change its feeding strategy from predatory to suspension feeding [93], whereas juvenile N. virens feed on detritus [94] and plant matter, which promotes growth [95]. Regardless, the diverse fatty acid assemblage of Nereis spp. suggests that small, likely juvenile, individuals feed on a combination of prey, detritus, and phytoplankton.

We expected (marked) differences in cryptofaunal diets between sites given the proximity of the North site to riverine input and our observed between-site differences in rhodolith size, cryptofaunal abundance, and rhodolith epiphytic foliose coverage. However, this was not the case, with only a few, and rather small, differences in trophodynamics between sites and a lack of terrestrial biomarkers at the North site. This may be the outcome of potentially low-quality terrestrial sources, i.e., refractory carbon [96], compared to a high availability of nutritious phytoplankton. The few terrestrial biomarkers in our seawater samples could also indicate limited mixing between riverine and marine systems or the destruction of terrestrial matter upon entering the ocean [97]. A more comprehensive sampling design that includes water upstream of, at, and downstream of the mouth of the river would be needed to better speak to the abundance, fate, and contribution of terrestrial matter to this rhodolith bed’s trophydynamics.

4.4. Stable Isotopes

We showed that even though consumers like A. rubens, H. arctica, and O. aculeata altered their diets following the spring phytoplankton bloom, their overall δ¹³C and δ¹⁵N isotopic values did not change spatially or temporally. The lack of seasonal change in δ¹³C suggests a two-tier dietary preference whereby the overall components of an organisms’ food source (i.e., phytoplankton or bacteria) remain consistent throughout the year, but the specific components of the food source (i.e., diatom vs. dinoflagellate) change. The temporal stability of the δ¹⁵N values supports the notion that the consumers sampled did not alter their diets extensively, but rather adjusted to the available resources.

Although we found no major temporal change in δ¹³C or δ¹⁵N values, there was still some variation at the species-individual level, likely reflecting food obtained during different stages of the recycling pathway outlined above. Because POM degrades as it settles on the sea floor, organisms feeding during later stages of the pathway (i.e., consuming resuspended material or predation) typically exhibit greater enrichment in δ¹³C and δ¹⁵N than organisms feeding on sinking POM [83]. Furthermore, overlap in δ¹⁵N values indicates reliance on similar food sources and extensive competition for food [83]. For example, in the present study, Nereis spp. split into four different bulk isotope subgroups. We believe this variation in isotopic value reflects opportunistic feeding, where Nereis spp. is not a major predator in this system, but rather an opportunist. Likewise, the isotopic values of A. rubens varied greatly. Most notably, two A. rubens samples had more enriched δ¹³C values (~14‰) than the majority of A. rubens samples (~19‰). Asterias rubens samples with ~14‰ δ¹³C values likely obtained their food from sources outside the recyclable macroalgae food-partitioning pathway (outlined in Section 4.3) and, therefore, probably do not fall within the benthic–pelagic coupling link we proposed for other focal community members. Instead, given their enriched values, we propose that these individuals preyed on organisms in a kelp- rather than phytoplankton-based food web [98]. However, L. digitata, a species with seasonal growth patterns [99], had the largest seasonal isotopic difference, with δ¹³C values jumping from ~−21‰ in April to ~−15‰ in July and December, and its δ¹⁵N increased from ~3‰ in April to ~7‰ in July and December. Low values in April compared to both our seasonal samples and typical averages (~−12‰ to ~−15‰ for δ¹³C and ~7‰ for δ¹⁵N) [100,101] could reflect a March sea ice event inhibiting L. digitata photosynthesis and growth. Indeed, prolonged sea ice limits light availability and thus may inhibit metabolic processes, resulting in stunted growth, reduced carbon uptake, less enriched δ¹³C values, and lower δ¹⁵N values [102].

5. Conclusions and Future Research Directions

Our results support our first hypothesis that seasonal fluctuations in sea temperature and food availability affect the lipid composition and diets of rhodolith cryptofauna. This was in part demonstrated by the overall shift from a diatom-based food web following the spring phytoplankton bloom to a kelp/algae-based food web during fall, and the observation that EPA is of higher value than DHA to the rhodolith bed community studied. Our fatty acid and stable isotope analyses strongly support Hacker Teper et al. [30]’s conclusion that first- and second-order consumers share a common resource (diatoms or kelp) in this rhodolith food web. The lack of terrestrial biomarkers at the study site nearest to riverine inputs, and minor differences in trophodynamics between both study sites, refuted our second hypothesis that riverine inputs affect cryptofaunal diets, although a more comprehensive water sampling design that includes the river/ocean transition zone would be required in order to fully elucidate this aspect.

Based on our findings, we propose a partial food web (Figure 5) encompassing: (1) direct feeding relationships from primary producers (e.g., rhodoliths and macroalgae) to second-order consumers (e.g., sea stars and polychaetes); (2) trophic subsidy from within and outside the rhodolith bed via settlement, resuspension, and consumption of macroalgal fragments and other detrital organic matter; and (3) benthic–pelagic coupling, as proposed in Hacker Teper et al. [30] and unequivocally demonstrated with the present study. This food web appears spatially and temporally stable in part because of the ability of key consumers to adjust their diets to available resources sharing similar nutritional characteristics (e.g., diatoms vs. dinoflagellates during or after phytoplankton blooms) [34,103]. The high and stable abundance of a few key consumers, namely, chitons and brittle stars ([6], present study), along with their dominant feeding modes [38], i.e., grazing and suspension feeding, respectively, may be key to this stability by greatly limiting epiphytic growth on rhodoliths. Epiphyte-free rhodolith surfaces should facilitate light capture and gas/nutrient uptake, contributing to photosynthesis, growth, and maintenance of rhodolith shape and structural integrity. Complementary studies of the same rhodolith bed showed that: (1) rhodolith morphology (shape, size, hollowness) impacts rhodolith cryptofaunal abundance and diversity [6,38]; (2) rhodolith growth is adversely affected by coastal eutrophication [72], yet is resilient to changes in sea temperature within the 2–10 °C range [39]; and (3) bioturbation by a few dominant invertebrates within and outside rhodoliths is sufficient to prevent sediment accumulation on, and burial of, rhodoliths [29]. These and the present study, therefore, highlight the high spatio-temporal stability of the rhodolith bed framework and resident cryptofaunal abundance, diversity, and trophodynamics. Given a rapidly changing ocean climate, further laboratory and field studies should investigate the impacts and relative importance of physical, chemical, and biological forcing on rhodolith bed homeostasis.