Epibenthic predators control mobile macrofauna associated with a foundation species in a subarctic subtidal community

Abstract Foundation species (FS) are strong facilitators providing habitat for numerous dependent organisms. The communities shaped by FS are commonly structured by interplay of facilitation and consumer control. Predators or grazers often indirectly determine community structure eliminating either FS or their principal competitors. Alternatively, they can prey on the dependent taxa directly, which is generally buffered by FS via forming complex habitats with numerous refuges. The latter case has been never investigated at high latitudes, where consumer control is widely considered weak. We manipulated the presence of common epibenthic crustacean predators to assess their effect on mobile macrofauna of the clusters developed by a FS (barnacle Balanus crenatus and its empty tests) in the White Sea shallow subtidal (65° N). While predation pressure on the FS itself here is low, the direct effects of a spider crab Hyas araneus and a shrimp Spirontocaris phippsii on the associated assemblages were unexpectedly strong. Removing the predators did not change species diversity, but tripled total abundance and altered multivariate community structure specifically increasing the numbers of amphipods, isopods (only affected by shrimp), and bivalves. Consumer control in the communities shaped by FS may not strictly follow the latitudinal predation gradient rule.

dependent taxa. Yet, FS generally mitigate the impact of consumers on the associated organisms by sustaining habitat structure rich in refuges. Consequently, in the latter case predation is often buffered by habitat complexity, the outcome thus being system-specific and less predictable.
Barnacles are foundation species providing habitat for multiple dependent taxa on various hard substrates in temperate and polar waters (Barnes, 2000;Gil & Pfaller, 2016;Harley, 2006;Thompson, Wilson, Tobin, Hill, & Hawkins, 1996;Yakovis, Artemieva, Fokin, Grishankov, & Shunatova, 2005). Aggregated barnacles commonly form two distinct microhabitats types, composed, respectively, by live individuals and their empty tests remaining attached to a subtrate. Empty tests lack the activities of live barnacles such as filter-feeding and feces production, but contain more cavities and accumulate more sediment, often hosting the macrobenthic assemblages different from those associated with live barnacles (Barnes, 2000;Harley & O'Riley, 2011;Yakovis & Artemieva, 2015). The effect of predators on the suite of species facilitated by barnacles has never been experimentally investigated.
In the White Sea (65° N) shallow subtidal near Solovetsky Islands, a barnacle Balanus crenatus is a dominant FS developing clusters on mollusk shells and gravel scattered on muddy sand. Here, clustered barnacles and their comparably numerous empty tests cover all the small hard substrates and host a remarkably diverse assemblage of mobile and sessile macrobenthic taxa (Yakovis & Artemieva, 2017;Yakovis et al., 2005;, distinctly dissimilar to the fauna of the surrounding unstructured sediment (Yakovis et al., 2005). Equally complex patches of live barnacles and their empty tests develop strongly different assemblages of associated species, with live barnacles accommodating several times more juvenile bivalves Musculus discors and polychaetes Cirratulis cirratus and Pygospio elegans than empty tests (Yakovis & Artemieva, 2015). Also, field experiments with barnacle mimics show that species composition of mobile fauna inhabiting barnacle clusters is largely determined solely by the presence of structurally complex habitat (Yakovis, Artemieva, Fokin, Varfolomeeva, & Shunatova, 2007), potentially rich in refuges from predation.
While spatial distribution of barnacles themselves in tropical and temperate waters is often controlled by predators (Foster, 1987), this does not apparently happen in the White Sea subtidal. Although principal sources of barnacle mortality here are not completely clear , their abundance seems rather affected by hard substrate availability (Yakovis, Artemieva, Fokin, Varfolomeeva, & Shunatova, 2013) than controlled by relatively scarce and ineffective consumers (Yakovis & Artemieva, 2015). Weak predation pressure at 65° N is consistent with biotic interaction hypothesis (BIH) predicting low importance of biotic interactions (i.e., predation) in subpolar and polar regions (Freestone, Osman, Ruiz, & Torchin, 2011;Schemske, Mittelbach, Cornell, Sobel, & Roy, 2009). BIH has been developed in efforts to explain the latitudinal diversity gradient, which is the strongest global biogeographic pattern observed, and implies that primary selective pressures are coevolution of interacting species in tropics and abiotic factors closer to poles. Although most data available generally confirm BIH, there are multiple cases contradicting one, and predation in subpolar and polar regions is still nearly unexplored compared with temperate and tropical zones (reviewed by Schemske et al., 2009).
Here, we manipulated the presence of common crustacean predators (a shrimp Spirontocaris phippsii and a spider crab Hyas araneus) to assess their direct effect on mobile macrobenthic assemblages associated with subtidal barnacle clusters in the White Sea. BIH, consistently low predation pressure on the FS itself, and the apparent surplus of potential refuges within barnacle clusters predicted no pronounced effect. At the same time, consumers can strongly affect community structure even in subpolar waters at least indirectly via control of FS abundance (Estes & Palmisano, 1974) or directly in the absence of FS (Quijon & Snelgrove, 2005). In addition, the very presence of FS-shaped habitat seemingly suggests stronger biotic interactions (i.e., facilitation) than predicted by BIH for subpolar waters.
Given that the microhabitats constituted by live barnacles and their empty tests attract different associated fauna, we used experimental units of both types. This allowed the detection of interactive effects, in case either live barnacles or their empty tests would provide better protection from possible predation. This could, in turn, explain some difference in mobile assemblages between the two microhabitats. Revealing for the first time the effect of predators on FS-associated dependent assemblage in a subpolar sea would enrich our understanding of interspecific biotic interactions as drivers of community structure and functioning.

| MATERIAL AND ME THODS
To test the effect of predators on mobile macrofauna of barnacle clusters, we conducted a year-long field experiment at a 12-m-deep subtidal site in the White Sea near the Solovetsky Islands (the Onega Bay, 65°01.180′N, 35°39.721′E, see Site 1 in Yakovis & Artemieva, 2015). The exposure duration was selected according to the results of previous colonization and caging experiments on the same system (Yakovis & Artemieva, 2015;Yakovis et al., 2007). While caging experiments at lower latitudes typically last shorter, the rates of succession and predation in subtidal of the severely cold White Sea are very slow Yakovis & Artemieva, 2015;Yakovis et al., 2005) and require longer experiments to detect changes. Also, the accessibility of the study site in late fall and winter is poor due to storms and ice cover. We collected and defaunated (except barnacles and their empty tests with 4+ annual growth rings) empty Serripes groenlandicus shells (59 ± 1 mm long) with live barnacles Balanus crenatus (hereafter "LB" units), and similar shells with empty barnacle tests (hereafter "ET" units). These units were attached in the alternating order to the bottom of 300 × 375 × 70 mm plastic frames covered with 2.5-mm nylon mesh. Each frame (hereafter "block") contained 2 LB and 2 ET units. We randomly distributed these blocks between five treatments: (a) full cages (predator exclosures), (b) open (no mesh, subject to normal predation), (c) partial cages to control for the effect of caging (similar to full cages but with two side windows 175 × 50 mm each), (d) cages with spider crabs Hyas araneus added (crab enclosures), and (e) cages with shrimp Spirontocaris phippsii added (shrimp enclosures).
Upon retrieval of the blocks, we estimated the wet weight of barnacles in each unit from individual carino-rostral aperture length measurements according to the previously determined relationship (Yakovis & Artemieva, 2015). We also individually measured empty barnacle tests and calculated their equivalent weight using the same relationship as if they were live. The sum of calculated weights of live and dead barnacles (hereafter "equivalent barnacle weight," "EBW") was 40.75 ± 1.27 g per unit (n = 124). Neither the number of barnacles or empty tests per unit nor total unit weight could adequately represent habitat capacity because of the variation in size of individual barnacles and empty tests between the units, and lower weight of same sized empty tests compared to live barnacles. We thus selected EBW as a unit size measure to use in further analyses.
TA B L E 1 Abundances of crustacean predators by treatment in the field experiment. Zero initial abundances for Eualis gaimardi omitted for clarity Recruitment rate of barnacles at the study site is highly variable between the years (Yakovis et al., 2013 2015. Yet, the distribution was highly patchy, with most samples containing no shrimp. The higher estimate based on the number of individuals found in six auxiliary partial cages (each exposed for a year in 2009-2012) was 0.11 ± 0.02 per gram EBW. We opted to add 4 adult Spirontocaris individuals per shrimp cage, which resulted in the abundance of 0.020 ± 0.002 per gram EBW (compared with 0.040 ± 0.005 in partial cages retrieved in 2016, see Table 1).
The blocks were anchored to the bottom in July 2015 in a haphazard pattern (at least 0.5 m from each other within a square approximately 20 by 20 m) and exposed for 1 (Anderson, 2006), which indicated that group variances were heterogenous. Given that the largest group (namely predator exclosures) in our slightly unbalanced design had the smallest dispersion, this could potentially cause too liberal results of PERMANOVA test (Anderson & Walsh, 2013). Knowing that balanced PERMANOVA designs are insensitive to variance heterogeneity (Anderson & Walsh, 2013), we assessed the reliability of our results by running ten additional PERMANOVA tests, each on a separate nearly balanced subset of the data with two random exclosure blocks excluded. All the 10 tests produced the results consistent with PERMANOVA on the full data set, confirming the reliability of the latter.
SIMPER procedure (Clarke, 1993) was used to identify the taxa principally responsible for the differences between the factor levels revealed by PERMANOVA. To visualize multivariate differences between the assemblages associated with combinations of Treatment Multivariate community analyses were conducted in PRIMER 6.0 software with PERMANOVA add-on (Anderson et al., 2008). GLM ANCOVAs were performed in STATISTICA 8.0 software package (Statsoft Inc., 2007). All mean values are given ± SE.  (Table 3 and Figure 1).

Treatment × Live interaction was insignificant. Number of taxa was
only affected by Treatment (Table 2), being significantly higher in predator exclosures than in any other treatment (Table 3). Species diversity was slightly (but significantly) higher in LB than ET with no significant effect of Treatment (Tables 2 and 3).
Assemblages were significantly affected both by Treatment and Live effects, but not their interaction (see Table 4 (Figure 4 and Table 3). Total abundance and TA B L E 3 Mean (± SE) abundances of 15 most abundant taxa, total abundance, number of taxa, and log-e species diversity (H′) by treatment in the field experiment  Table 2).
relative abundance of Dyopedos porrectus and juvenile Musculus discors were highest in the samples belonging to the first cluster (i.e., LB and ET from predator exclosures; Figure 4). There were relatively more polychaetes in the samples from the second and third clusters (containing, respectively, all other LB and ET), while total abundance in the third cluster was lowest (Figure 4).
Of 15 most abundant taxa, 10 responded to Treatment, Live, or their interaction (see Tables 2 and 3 Table 3 and Figure 5.

| D ISCUSS I ON
Consumer control is not generally expected to determine community structure at high latitudes (Freestone et al., 2011;Schemske et al., 2009 (Coles, 1980;Head et al., 2015;Leray, Beraud, Anker, Chancerelle, & Mills, 2012). Empty barnacle tests comprise more cavities and accumulate 2-3 times more sediment than live barnacles of the same size (Yakovis & Artemieva, 2015), which could provide better shelter from predation for at least some of the associated taxa. On the other hand, cirral activity of live barnacles could also interfere with predators resulting in lower predation pressure. Bleached and dead corals, for instance, provide poorer protection from predators for the associated damselfishes than live ones (Coker, Pratchett, & Munday, 2009). Despite the apparent presence of these potentially important mechanisms, the observed dissimilarity of mobile fauna associated with the two microhabitats does not noticeably result from predator-related processes. Also, the impact of predator removal on mobile assemblages is much stronger than this dissimilarity. The taxa that exhibit higher abundances in LB may benefit from direct trophic facilitation by live barnacles (which alter local water flow by cirral movements and produce feces), like obligatory coral-dwelling decapods consuming coral mucus (Head et al., 2015 and references therein).
All the taxa affected by predator presence manipulations were most abundant in exclosure cages. There are two ways predators can reduce the abundance of mobile prey: consumption and triggering predator-avoidance behavior. Behavioral responses to the cues hinting potentially higher predation risks often make prey's "landscape of fear" an equally important driver of its spatial distribution compared with lethal attacks (Laundré, Hernández, & Altendorf, 2001;Preisser, Bolnick, & Bernard, 2005 depending on their mobility and predator preferential diet. While some flourish since they directly avoid consumption, others might immigrate to minimize predation risks. Suggesting which mechanism is more important for a particular prey species is complicated, because the diet of Spirontocaris phippsii and Hyas araneus in the White Sea is totally unexplored. In fact, feeding habits of these two predators are generally obscure. According to stomach contents analysis, Spirontocaris spinus from an arctic fjord feeds on benthic foraminiferans and hydroids (Birkely & Gulliksen, 2003 H. araneus as 2.9 with δ 15 N of 11.9 ± 0.1‰ (Nadon & Himmelman, 2010), which identifies it as a predator or scavenger. In contrast to other crustacean macropredators, Hyas spp. has no effect on species composition and abundances of mobile benthic fauna from unstructured subtidal soft sediments in Bonne Bay, Newfoundland (Quijón & Snelgrove, 2005). Few other sources report Hyas araneus as a scavenger (Nickell & Moore, 1991, 1992 or a predator on live juvenile sea scallops (Nadeau, Barbeau, & Brêthes, 2009;Nadeau & Cliche, 1998 (Beukema, Honkoop, & Dekker, 1998;Eggleston, 1990).
An amphipod Dyopedos porrectus, which also exhibited a strong negative response to predator presence, moves faster than bivalve juveniles, but builds flexible rods 4-6 cm long attached to surfaces of hard substrates, on which it stays hooked when feeding on suspended particles (Mattson & Cedhagen, 1989;Thiel, 1999). Consistent with our results, a sand shrimp Crangon septemspinosa eliminates a similar rod-building Dyopedos monacanthus in laboratory experiments (Thiel, 1999). Unlike Dyopedos, another abundant amphipod Crassicorophium bonellii is not affected by predator removals. Crassicorophium, however, is a tubicolous deposit-suspension feeder dwelling inside an open-ended tube it pumps water through to collect food particles (Foster-Smith & Shillaker, 1977) which apparently helps this and some other corophiids (Mook, 1983) to evade predators. An isopod Munna sp. positively responded to predator removals, being solely affected by shrimp and not crabs. Munna is also epibenthic, usually found on the surfaces of hard substrates, and not a tube-builder (Ambrose & Anderson, 1990;Hayward & Ryland, 2017). These isopods are extremely fast runners and look much more vigilant than Dyopedos in the laboratory (authors' personal observations), which may explain their invulnerability to relatively sluggish crabs.
In unstructured muddy sand and seagrass meadows, the removal of epibenthic predators causes the increase in abundance of infaunal predators which may lead to increased predation pressure on the rest of infauna (Ambrose, 1984 (Ambrose, 1993;Plyuscheva, Martin, & Britaev, 2010), according to post hoc tests, were not significantly affected by manipulations. The increase of Sphaerosyllis abundance in predator exclosures, however, was only marginally insignificant (Tables 2 and 3). In the field, Pholoe and Sphaerosyllis inhabit barnacle clusters rather than surrounding muddy sand (Yakovis et al., 2005), while all the three species significantly increase their abundance in response to adding structure (PVC tubes) to unstructured sediment (Yakovis et al., 2007). Given that some crab species can prey at least on Pholoe (Quijón & Snelgrove, 2005), it is likely that structural traits of barnacle clusters effectively protect these potentially vulnerable mesopredators from crabs and shrimp compared with less structured habitats.
At global geographical scale, strength of biotic interactions, including predation, is commonly supposed to reduce with latitude (Schemske et al., 2009). For instance, predator removals have almost no effect on species richness and abundances in temperate (41° N) compared with tropical (9° N) sessile epibenthic assemblages developing on PVC settlement panels (Freestone et al., 2011). This pattern supposedly results from impact on species' evolution of higher environmental stress level closer to the poles (Schemske et al., 2009). Unexpectedly, predation is increasingly found important at high latitudes (Giachetti, Battini, Bortolus, Tatián, & Schwindt, 2019 and references therein). Highest intensity of predation is generally linked to low abiotic stress, and so are associational defenses from predators backed by strong facilitators such as FS (Bruno et al., 2003). Subtidal habitats, however, are obviously much less affected by abiotic stress than commonly studied intertidal ones, especially in harsh environment of polar and subpolar seas with winter ice cover. At our research site, even most severe fall storms apparently do not largely influence benthic communities at 12-m-deep seafloor, since rather fragile barnacle-dominated epibenthic patches on empty bivalve shells can persist here for years Yakovis et al., 2005). This can cause local increase both in predation intensity and associational defenses by FS compared with the levels predicted solely from latitude. Consistently, many examples of strong trophic control in subpolar and polar waters come from subtidal habitats (Estes & Palmisano, 1974;Giachetti et al., 2019;Quijón & Snelgrove, 2005), although inter-and subtidal predation levels have been never specifically compared across latitudes.
It also appears that communities shaped by FS in general do not necessarily follow the latitudinal predation gradient rule. Seagrasses appear to be the only functional group of FS that span from tropical to polar waters studied extensively enough to allow a cross-latitude comparison. Direct consumer control of the fauna associated with seagrass meadows rather shows the reverse latitude gradient, if any. Being relatively strong at 58° N (Moksnes, Gullström, Tryman, & Baden, 2008), 37° N (Douglass et al., 2007), and 30° N (Leber, 1985), it is weak or absent at 35° N (Summerson & Peterson, 1984) and 26° N (Hammerschlag-Peyer et al., 2013). Although the areas covered by FS-generated habitats may globally decrease with latitude, the predation strength within these islands of relative stability may not be primarily controlled by environmental stress, which is ameliorated by FS. Here, we show that even at 65° N predators can severely affect abundances (but not the diversity) of the fauna associated with FS.
Relative strength of top-down versus bottom-up control in FS-driven communities is often switched by human disturbance (Bertness et al., 2014). It is thus important that the White Sea has faced a dramatic decline of fishery in recent decades (Berger, 2005); hence, our results are unlikely biased by anthropogenic influence on apex predators.

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
EY conceived and designed the experiments, performed the experiments, analyzed the data, contributed reagents/materials/analysis tools, wrote the paper, prepared figures and tables, and reviewed drafts of the paper. AA conceived and designed the experiments, performed the experiments, contributed reagents/materials/analysis tools, and reviewed drafts of the paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw abundances of all the taxa identified in the experimental units