Competition in the deep sea: phylogeny determines destructive impact of wood-boring xylophagaids (Mollusca: Bivalvia)

How biotic interactions contribute to structuring deep-sea communities remains poorly known. An example of exploitation competition, in which over time one species dominates a habitat to preclude its use by competitors, is highlighted here. Sunken wood is the obligate habitat of deep-sea wood borers of the Xylophagaidae Purchon, 1941 which eat wood and, with symbiotic bacteria, digest it. Enigmatically, some wood falls remain robust despite xylophagaid boring. Xylophagaids bore other wood falls so heavily that they crumble after just a few months. We perform a meta-analysis of experimental wood deployments and view the results from a phylogenetic perspective. Of 15 deployments recovered after 5 to 25 months from 200 to 3232 m deep, seven were so heavily bored to be crushable by hand. Eight were bored but remained robust. The groups did not significantly differ in wood size, type, temperature, depth, or duration. Members of the Xylophaga dorsalis (Turton, 1819) clade, reported in two studies to recruit and bore faster than confamilials, bored all seven crushable deployments; seven of the eight intact deployments were bored by other xylophagaids. Fecal chimneys line and thus narrow the boreholes of this clade; fluid flow is impeded, a clear liability. The lignin-rich fecal chimneys may, however, cue larval settlement and at resultant high population densities, lower oxygen availability. Members of the X. dorsalis clade, being hypoxia-tolerant, thrive but other xylophagaids appear to suffer, perhaps due to interference competition. The shared derived characters that unite this clade allow them to exploit low-oxygen wood that is intolerable for competitors.


Introduction
Competitive interactions in the deep sea remain little-known but are likely to be important in structuring communities. We present a study that reveals, we argue, exploitation competition, in which individuals of one species monopolize limiting resources. Our focus is on sunken wood that forms deep-sea wood falls. Although ephemeral and stochastically distributed, wood falls can sustain high levels of alpha biodiversity if colonized by obligate wood-boring bivalves of the Xylophagaidae Purchon, 1941, with seven named genera and about 61 named species.
Xylophagaids bore into wood and generate slightly elongate, non-overlapping boreholes that differ subtly by species (Amon et al. 2015a). They eat the wood they remove, digesting cellulose with the aid of symbiotic bacteria, which also fix nitrogen (Distel and Roberts 1997). These actions sustain biodiversity on deep-sea wood falls, in some cases over years (e.g., McClain and Barry 2014). If xylophagaids are absent, few other animals recruit to sunken wood (Gracia et al. 2019;Young et al. 2022). Because their boreholes generate space and the animals and their bacteria convert wood into digestible food available to non-xylophages, xylophagaids have been deemed ecosystem engineers (e.g., McClain and Barry 2014;Harbour et al. 2021).
The study of deep-sea wood falls and their borers has been accelerated by the use of experimental deployments that mimic natural wood falls. Turner (1973) used these to Communicated by C. Chen document the very existence of deep-sea wood borers; the deployments have since secured macrofauna and microbial specimens for study of wood fall biodiversity (Turner 1978;Sandström et al. 2005;Voight 2007;Romano et al. 2013;Yücel et al. 2013;Judge and Barry 2016;Kalenitchenko et al. 2018), and they have been used to return complete ecosystems for analysis (e.g., McClain et al. 2016). Short-term experimental deployments, however, reveal an enigma (Fig. 1). Xylophagaid-bored wood sometimes forms long-lived biodiversity hot spots, but other times, the xylophagaid borers reduce deployed wood to a minimal skeleton, crushable by hand, in as little as a few months (Haderlie 1983;Voight 2007;Romano et al. 2013;Amon et al. 2015b).
The heavy boring that results in the deployments being crushable by hand has been attributed to the proximity of unseen wood falls that serve as larval sources, to continual high larval availability, to an absence of predators, to long observation periods, to the bivalves' specialization to their ephemeral habitat, to warmer temperatures, and/or to properties of the wood (e.g., Metaxas and Kelly 2010; Amon et al. 2015a). Despite the fact that a given piece of wood is a finite resource that xylophagaids must consume to survive, biotic interactions among borers within a wood fall have been largely ignored. If one species, or a group of related species, can dominate a wood fall by means such as higher fecundity, increased larval availability, or mixotrophy, the contrast between long-lived, robust, and short-lived, heavily bored wood falls might be simply explained.
Here we use a meta-analysis including multiple xylophagaidbored deployments from around the world to compare the taxonomic composition of borers in wood that remained largely intact to those in wood that was crushable by hand (or heavily bored) on recovery. We assign the xylophagaids removed from wood of contrasting condition to their respective clades defined by anterior adductor cover and siphon morphology and molecular characters (Voight et al. 2019). We find members of a single clade are associated with every heavily bored deployment in our analysis. We then consider aspects of their biology that may contribute to their competitive superiority.

Material and methods
To assess the importance of taxon membership in determining the fate of wood falls, we undertook a two-part study. First we performed a meta-analysis based on literature reports of experimental wood deployments, including ten deployments from the northeast Pacific, two from both the Indian Ocean and Mediterranean Sea, and one from a Norwegian fjord; the depths are from 200 to 3232 m. Inclusion of diverse datasets is among the primary advantages of a meta-analysis, especially given the costs of deep-sea research. Other advantages are that the increased sample size allows more powerful statistical analyses and that the odds of a given random event impacting every study included and generating a bias are greatly reduced; the deployments considered here were made from 1979 to 2017, during 7 months of the year. We excluded deployments less than 2 cm in smallest dimension as overly vulnerable.
Deployments considered to be heavily bored were those the authors reported were crushable by hand at recovery. Deployments considered to be largely intact were those the authors noted to be solid or intact or to require use of a hammer and chisel to sample the xylophagaids. We quote the published description of the recovered wood in Table 1 with the page reference; we also re-examined the preserved wood (deposited at the Field Museum collections) from which the Fig. 1 Two deployments of Douglas fir recovered after 24 months at depth a from 2701 m depth, bored by Xylonora corona (Voight, 2007). The distance from the top to the bottom of the wood is 10.16 cm; b from 2211 m depth bored by Xylophaga oregona Bartsch, 1921, a member of the X. dorsalis (Turton, 1819) clade (from Voight 2007) Table 1 Recovered wood deployments listed by their condition at recovery with the dominant xylophagaid borer, locality, depth, sea temperature, soft or hardwood, deployment month year, and duration in months, size, and surface area, condition when recovered and reference. At the end of each section are the W statistics and probability from the Mann-Whitney U test between the groups for each variable.
Months are numbered with the summer solstice month being 1. An asterisk after the species name indicates JRV verified the species determination Taxon species Voight (2007) reported were removed, terming it crushable or intact (Fig. 1). The locality, depth, temperature, wood type and treatment history, deployment month and duration, surface area, and the most common xylophagaid species that bored each of the deployments considered are reported on Table 1 with the group ranges and medians. Temperatures were reported in the original reports, although an equipment malfunction forced the 10-month deployments reported by Voight (2007) from over 2200 m to be estimated at 2°C. Because the data include Northern and Southern Hemispheres, for statistical analysis we assign 1 to the summer solstice month (July in the North, December in the South) and number other months sequentially. In two instances, deployment weights rather than surface areas were reported (Amon et al. 2015a), forcing us to calculate and compare the medians of the weights in an analysis separate from the surface area comparison. We compared environmental variables between the heavily bored and largely intact deployment groups with Mann-Whitney U tests. If temperatures were reported as a range (Romano et al. 2013), we used the mean value in the test. Test statistics and probabilities are reported in Table 1, as is whether softwood (pine or fir) or hardwood (mango) formed the deployment. These types of wood differ in their density, anatomy, and chemical composition (Pereira et al. 2003).
Xylophagaid species names cited in the original works were accepted; JRV examined specimens from the studies indicated in Table 1. Amon et al. (2015b) accessed the types at the Natural History Museum (D. Amon, pers. comm.); the species cited by Haderlie (1983) and Harbour et al. (2021) are the only xylophagaids known in their study areas. Because the genus Xylophaga Turton, 1822 is paraphyletic (Voight et al., 2019), species assigned to that genus but not members of the clade containing X. dorsalis (Turton, 1819), the type species of the genus, are indicated by "s.l." To test whether heavily bored and largely intact deployments were bored by different species groups, we compared three groups. First was the species of Xylophaga s.l. and Xylonora Romano, 2020 in Romano et al. (2020). The calcareous mesoplax plates of these species serve as a putative synapomorphy; they are currently separated only by whether or not they have been sequenced (Romano et al. 2020). We assume they are a clade, but acknowledge they may be paraphyletic. Second, we include species of Abditoconus Voight, 2019, which have unusual nitrogen isotopes that are comparable to those of species of the clade of Xylophaga dorsalis (Voight et al. 2020). Third were species in the clade of Xylophaga dorsalis. In these species, the excurrent siphon is shorter than the incurrent siphon; it opens and thus releases most fecal pellets, inside the borehole. The pellets are compacted and accumulate around the siphon, typically extending beyond the wood-water interface to form a "chimney" (Purchon, 1941;Turner, 2002;Romano et al. 2014). The clade includes, among others, X. oregona Voight, 2007; X. washingtona Bartsch, 1921; and X. indica E. A. Smith, 1904, following Turner (2002. We tested whether the above three groups were more frequent in the heavily bored compared to the largely intact deployments using Fisher's exact test. We also compiled published rates of boring and population densities of xylophagaids, focusing on heavily boring taxa.
The second part of our study considers characters shared among heavily boring xylophagaids to assess whether these characters may contribute to their more aggressive boring.
Although Voight et al. (2020) found no isotopic evidence of chemosynthetic input in xylophagaids, chemosynthetic bacteria could occur in the borehole, where sulfide is hypothesized to be oxidized (Kalenitchenko et al. 2018). If so, precipitated elements such as sulfur or calcium should be abundant in the borehole and especially in the fecal chimney. A chemical microanalysis with an Oxford Instruments XMax50 Energy Dispersive X-ray Spectroscopy (EDS) system attached to a Zeiss Evo 60 scanning electron microscope tested this prediction. To reveal their elemental composition, we analyzed fecal chimneys of X. dorsalis that had been preserved in ethanol after having been removed 6 months before from a wild wood fall that also had bacterial mats and pogonophoran tubes; we also analyzed fecal chimneys of X. oregona that had been preserved in situ in the wood of Douglas fir for 10 years. Considering the chimneys of two species allowed us to assess the generality of our results, despite the more anoxic substrate of the X. dorsalis. Micro-Raman spectroscopy with a WITec alpha 300R 532 nm laser Raman system was applied to the chimneys to confirm results. The chimneys had been longitudinally split and air-dried for at least 72 h prior to analysis.

Results
Our meta-analysis revealed that crushable deployments were significantly more often bored by species of the clade of Xylophaga dorsalis than were the intact deployments (p = 0.048). Of 15 deployments analyzed, seven were crushable by hand, all of which were bored by members of the Xylophaga dorsalis clade; seven of the eight intact deployments were bored by members of other xylophagaid clades. Mann-Whitney U tests found no significant differences in the environmental and deployment variables considered between the groups (Table 1), although temperature very nearly significantly differed (p = 0.0505). Most deployments in both groups were composed of softwoods; each group had one hardwood. Haderlie (1983) did not specify the wood deployed. All wood for which treatment history was reported was either green or untreated (Table 1). Amon et al.'s (2015b) deployments had essentially equal median weights (4.8 and 4.9 kg). The second part of our analysis focused on the fecal chimney, a character that unites the heavily boring members of the X. dorsalis clade. The chimney lines the borehole and extends beyond the wood-water interface. The inner lining of longitudinally sectioned, air-dried fecal chimneys had a distinct sheen, consistent with Purchon's (1941) report of a mucus lining. Our EDS analyses found organic matter composed over 90% of the fecal chimneys of X. dorsalis and X. oregona (Fig. 1); major components of seawater were also abundant ( Table 2, full data in Online Resource Table 1). The fecal chimney of X. dorsalis had considerably more sulfur, a major component of seawater, than did that of X. oregona, consistent with the evidence noted above that the wood of the former had previously supported chemosynthetic animals. The inner chimney of X. oregona had less sulfur than did the outer (Table 2).
Raman spectroscopy revealed the major components of the X. dorsalis chimney to be lignin, cellulose, and hemicellulose, specifically xylan (Fig. 2). It also identified the white spots on the siphon of X. dorsalis as calcite and the shell's beak and mesoplax to be aragonite (Online Resource Table 2).

Discussion
Among the reports considered here, regardless of deployment location, depth, temperature, or wood size or type, only members of the Xylophaga dorsalis clade bored wood so heavily that it could be crushed by hand; wood bored by other xylophagaids for the same duration remained largely intact. Although we do not assert that this pattern is invariant, no longer are ad hoc explanations, such as the proximity of an unseen larval source, required to explain heavily bored wood falls. Two reports that did not meet our minimal size deserve mention. A 20 × 8 × 2 cm deployment bored by Xylonora atlantica (Richards, 1942) crumbled after 8 to 10 months at 100 m depth in 17.5°C (Romey et al. 1994); Xylophaga washingtona bored a comparable 22.8 × 10 × 1.9 cm deployment at 200 m depth until it was spongy (with 255 boreholes/ cm 2 ) after 2.4 months at 7°C (Tipper 1968; temperature estimated from Meseca 1967). Because X. washingtona, a member of the X. dorsalis clade, bored for a shorter time in colder temperatures, these observations support the hypothesis of heavy boring by members of that clade. The one reported deployment that was bored by X. dorsalis but was not crushable by hand "would have degraded to the point of collapse" if it had spent longer than 10 months on the bottom (Harbour et al. 2021, p. 87).
Incomplete knowledge of the xylophagaids cautions us against making absolute statements, but available data indicate that the clade of X. dorsalis is exceptional. Member species of the X. dorsalis clade, termed Groups 5 and 6 by Turner (2002) and Voight (2008), have a pooled depth range of 10 to 4626 m, overlapping with but on average significantly shallower than that of other xylophagaids (Voight 2008). Their elevated nitrogen isotopic values are consistent with opportunistic filter feeding (Voight et al. 2020). The species X. dorsalis (as Xylophaga sp. A) recruits in greater numbers, and grows faster than do the xylophagaids Abditoconus brava (Romano, Pérez-Portela & Martin, 2014in Romano et al. 2014 reported as Xylophaga sp. B by Romano et al. (2013) and Xylonora atlantica reported by Romey et al. (1994). Xylophaga indica of the same clade (following Turner 2002) recruits faster and grows slightly faster than does X. s. l. murrayi Knudsen, 1967 in comparable deployments; these differences were attributed to warmer temperatures (4.3 versus 10.1°C), differences in predation, and/or larval supply that affected population densities (Amon et al. 2015b). Using a micro-CT scans, Amon et al. (2015a) calculated that Xylophaga depalmai Turner, 2002 and X. indica, members of the X. dorsalis clade, bore more (0.438 to 0.606 cm 3 /year) than does X. s. l. murrayi (0.235 cm 3 /year). We suggest phylogeny plays a significant role, but the lack of data concerning fecundity and the timing of larval competence from across the family limits our understanding of how and cannot support or refute the existence of a phylogenetic pattern. The higher rates of recruitment, boring, and growth demonstrated in two species of the X. dorsalis clade (Romano et al. 2013;Amon et al. 2015aAmon et al. , 2015b suggest that species of this clade act as exploitation competitors, consuming the limiting resource, wood, before competitors can do so. The fecal chimney may heighten their impact. This structure appears to us to impose two hydrodynamic disadvantages. First, the energy required for pumping the exhalant flow is apt to be increased because the chimney partially constricts the borehole and thus likely increases back pressure (Jørgensen and Riisgård 1988). Second, due to its poor separation from the inhalant flow, the excurrent flow risks refiltration of oxygen-depleted and dissolved CO 2enriched water (Du Clos and Jiang 2018), unless the inhalant siphon is fully extended. In situ photos show this to be the case in this clade (Purchon 1941;Bernardino et al. 2010, Fig. 2B;Romano et al. 2014, Fig. 2). Siphonal extension had been thought to increase the animals' access to well-oxygenated seawater (Purchon 1941) or to facilitate opportunistic filter feeding (Voight et al. 2020); its role in minimizing refiltration (Monismith et al. 1990) had not been considered.
Given these liabilities, the fecal chimney must convey some benefit. Potentially it relates to the high concentrations of sulfide emitted by the sunken wood, at least early in deployments (Sandström et al. 2005;Yücel et al. 2013;Kalenitchenko et al. 2018). The siphons in the X. dorsalis clade lack calcium carbonate or periostracal protection as is common in other xylophagaids (Voight et al. 2019); they may risk sulfide poisoning (Goffredi 2017). Kalenitchenko et al. (2018) found sulfide concentrations to be lower in boreholes than in the surrounding wood, leading them to suggest that sulfide is oxidized in the borehole. However, our elemental analysis found no significant concentrations of electron receptors in the chimneys which would be predicted if this hypothesis were supported. Instead, we suggest the excurrent flow, by ventilating the borehole, rids it of sulfide, especially near the animal's vulnerable siphon. The fecal chimney's mucus lining may also minimize sulfide exposure and enhance bivalve survival by minimizing the inward diffusion of sulfide as do mucus linings in infaunal burrows (Aller 1983;Hannides et al. 2005;Zorn et al. 2006).
Composed primarily of lignin, the fecal chimney, perhaps with metabolites added by the bivalve's digestive system, could serve as a larval settlement cue and accelerate recruitment. Fig 2 Typical Raman spectra on fecal matter of X. dorsalis where the presence of lignin (L), cellulose (C), and hemicelluloses (H), specifically xylan (X), was detected. Compounds were identified using Raman band assignments in Zeng et al. (2016) and Zhang et al. (2017). Spectra were acquired with a 532 nm laser at 10 mW with 30 s integration at 10× magnification and a 600 g/mm grating with a WITec alpha 300 R Raman system statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.