Within-host adaptive speciation of commensal yoyo clams leads to ecological exclusion, not co-existence

Sampling sites

From June 14th to July 25th, 2017, the first author (after obtaining a Florida state collecting permit) performed low tide field sampling of 86 L. scabricauda burrows at five adjacent shallow water sandflat study sites within the IRL’s Ft. Pierce Inlet (Fig. 2). The three sites (1, 4 & 5) on the northern margin of the Inlet (Fig. 2) collectively contain the type localities of five commensal clam species: Divariscintilla octotentaculata, D. luteocrinita, D. troglodytes, D. yoyo, and D. cordiformis (Mikkelsen & Bieler, 1989; Mikkelsen & Bieler, 1992).

Host burrow identification and host capture

During each field session, visible host burrows covered by ∼5–10 cm of seawater within the targeted study site were flagged for sampling. These typically represented <20% of the Lysiosquilla scabricauda burrows (characterized by their irregularly square openings, about one cm²in area, often covered by a sand cap that was differentially textured from the surrounding sand flat) visible within each intertidal site. L. scabricauda specimens were collected manually using a bait-and-capture technique (Goto, Harrison & Ó Foighil, 2018; Fig. 3). A bait fish was placed directly over a submerged burrow opening and held for a 3-minute trial period to elicit an attack by a resident stomatopod. If no host response occurred during that interval—approximately 50% of the burrows investigated—a new host burrow was then attempted. The raptorial appendages of lysiosquillid stomatopods are barbed, razor sharp, and designed to impale soft-bodied prey (DeVries, Murphy & Patek, 2012). Thus, thick fishing gloves were worn for protection and once a resident stomatopod impaled the bait or otherwise presented at the mouth of the burrow, its raptorial appendages were sequentially grasped by hand (Fig. 3) and held firmly as it attempted to pull itself downward into its burrow. As the restrained stomatopod tired, it was slowly pulled upward out of the burrow. A large majority of collected host specimens lacked ectocommensal clams and these were released within 5 min of capture.

Collecting burrow-wall commensal clams

Once a host L. scabricauda had been collected, its burrow was then sampled for yoyo clam burrow wall commensals using a stainless-steel bait pump (“yabby pump”). As emphasized by Mikkelsen & Bieler (1989) and Mikkelsen & Bieler (1992), this method effectively samples only its own length (0.5–1.0 m) of the vertical parts of the stomatopod’s U-shaped burrow, leaving the deeper horizontal section unsampled. The contents of single pulls of the yabby pump were expelled into a 2 mm sieve and this process was repeated until three pulls failed to return any observable clams, or until the vertical arm of the host burrow collapsed from the repeated suctioning. Regardless of species, yoyo clams were readily recognized by their characteristic off-white, mucoid appearance against the mesh and the residual sediment particles retained in the sieve. Individual clams were carefully picked up using a feather weight forceps and placed into 50 ml tubes of seawater. Any ectocommensal clams detected on the stomatopod host were similarly detached from the base of the host pleopods and placed in seawater tubes. Back in the laboratory, all live commensal clams sampled from individual host burrows were maintained together in burrow-specific, labelled finger bowls containing filtered sea water with slight aeration. All clams were then identified to species using a dissecting microscope and their mantle lengths measured. Species counts and the number of individuals per species were recorded for each burrow sample.

Statistical analyses of co-occurrence data

Several statistical tests were performed on the commensal species frequency data. These included over-dispersion tests for the two most frequent species (“P”. squillina and D. octotentaculata) using the “overdispersion.test” function in R 4.3.1, to determine if the observed individual distributions in host burrows were more clustered than expected by chance.

In addition, we conducted several simulations tests of the co-occurrence patterns of the different commensal species to determine if they fall within the expectation of a random larval settling process. IRL Divariscintilla species have “mixed” larval development in which early developmental stages are ctenidially-brooded, then released into the water column as early, straight-hinged “D” veligers to undergo an obligate period of planktotrophic larval development and dispersal, thereby greatly reducing the likelihood of resettlement in parental burrows (Mikkelsen & Bieler, 1989; Mikkelsen & Bieler, 1992). The details of “Parabornia” squillina’s early development are currently unknown but its prodissoconch structure is consistent with it also having an obligate planktototrophic larval dispersal phase (Fig. S1). (Similarly, presence of large numbers of small (<100 µm) brooded “ova” in the ctenidia of its Brazilian congener, “P”. palliopapillata, (Simone, 2001; Fig. 13 therein) is indicative of planktotrophic larval development in Galeommatoidea (Ó Foighil, 1988)).

We therefore assumed that each IRL commensal clam was independently recruited to its host burrow from a planktonic pool of metamorphosing veliger larvae. We were particularly interested in testing for settlement/survival effects among the ectocommensal “P”. squillina and the burrow-wall commensal Divariscintilla species. The competitive exclusion principle (Grinnell, 1904; Hardin, 1960; Chesson, 2003) predicts that IRL Divariscintilla species (which all share the same burrow-wall niche) will co-occupy burrows host less frequently with each other that with the niche-differentiated ectocommensal “P”. squillina. In contrast, the neutral theory of species assembly (Hubbel, 2001) predicts that each member of the IRL commensal vasconielline community will occupy host burrows irrespective of the presence or absence of any other member, and that the observed co-occurrence pattern will therefore match random larval settling expectations.

Expected co-occurrence distributions of different species pairs were generated by randomly allocating clams to burrows based on the observed proportions of each species across all burrows. For example, there was a total of 73 D. octotentaculata and 20 “P”. squillina in all burrows, meaning 78% (73/(73 + 20)) were D. octotentaculata and 22% were “P”. squillina. The two species were then randomly allocated to all burrows based on this probability, keeping the total clam count in each burrow equal to the observed value (i.e., if a burrow had five clam individuals, then simulation was done five times for that burrow). This process was repeated 1,000 times. After the simulation, numbers of burrows where two species co-occurred were counted and summarized in a histogram. The actual observed number of co-occurred burrows was also plotted on the histogram to compare with the theoretical distribution. P-values were calculated by the percentile of the actual observed value in the simulated distribution. This comparison was performed between “P”. squillina and all wall-commensal species (treated as the same type); “P”. squillina and D. octotentaculata; “P”. squillina and D. luteocrinita; and D. octotentaculata and D. leuteocrinita. The other wall commensal species had low occurrences therefore they were not compared with “P”. squillina individually.

Stable isotope analyses

An organism’s stable isotope composition is shaped by, and indicative of, its diet/trophic niche (Layman et al., 2007). To test if the IRL L. scabricauda vasconielline commensal species differ in their trophic niches, 18 burrow-wall commensals (11 D. octotentaculata, 4 D. luteocrinita, 2 D. yoyo and 1 D. troglodytes) and 20 “P”. squillina ectocommensals, together with samples of within-burrow potential basal trophic resources—tissue from 19 L. scabricauda specimens, suspended particulate organic matter from 35 burrows, and deposited organic matter from 34 burrows—were collected (Fig. 4) to measure their respective isotopic niche widths.

The clams were housed separately in petri dishes of filtered sea water for 12 h to allow for their gut contents to empty, after which their soft tissues were separated from their shells prior to further processing. Host stomatopod specimens were euthanized in an ice-water slurry. To sample burrow water particulate organic matter (POM) the flow of ambient water into burrows was first blocked with a cylindrical barrier (bucket with the bottom removed) enclosing the burrow opening. For each burrow, 1 liter of burrow water was field-collected with a large syringe (with a one mm filter attachment) and filtered in the laboratory onto 4.7 cm diameter Whatman GF/F glass microfiber filters using a six-manifold filtration system. Deposited organic matter was sampled by collecting the oxygenated layer of burrow wall sediment with a shallow teaspoon. All commensal tissue specimens and potential basal resource samples were lyophilized using a Labconco Freeze Dry System prior to further processing. This involved grinding the commensal samples, and the POM samples (first removed from the filter paper with a spatula), with disposable mortar and pestles, grinding the right merus of each host specimen for 4 min in a ball-mill grinder, and grinding each sediment sample for 4 min in a bead-mill grinder. Approximately 2.5 mg of each ground host and commensal species sample was weighed into individual 5 × nine mm pressed tin capsules. Approximately 5–7 mg of each POM and 35–40 mg of each sediment sample were weighed in 10.5 × 9 pressed, light-weight silver capsules and acidified with reagent grade HCL to remove any undetected shell fragments. All samples were analyzed at the Center for Applied Isotope Studies at the University of Georgia for carbon and nitrogen isotopic signatures. The isotopic niche width of each species was quantified as Standard Ellipse Areas (SEA_B) which estimate mean population-level isotopic niche spaces while accounting for variation in population size, in the R package SIAR (Jackson et al., 2011).

Artificial burrow construction & observations

Two observable, artificial stomatopod burrows were constructed at the Smithsonian Marine Station at Fort Pierce. Each was positioned within a 110 L flow-through glass aquarium tank that was partitioned with a sheet of PVC to form a nine cm wide cross-section of sediment observable through the front facing wall of the tank. The PVC barrier was 10 cm shorter than the tank, allowing flow to pass across it, and it was kept in place by a series of 3.8 cm diameter PVC tubes (Fig. S2). To construct artificial stomatopod burrows, a 3.8 cm diameter electrical conduit tube was first planed in half and then taped to form a U shape. The inner surfaces of the tube were lightly coated with silicone rubber sealant, packed with dry sand, and allowed to set overnight. The unattached sand was then removed, and the sand-coated tube halves were individually placed into separate aquarium tanks with the cut edge held in place against the tank walls with additional dry sand and the tube openings flush with the sand surface (Fig. 5). The aquaria were then filled with aerated sea water and allowed to settle for 24 h.

A host Lysiosquilla scabricauda was introduced to one of the aquaria, was acclimated for a week, and offered shrimp and small fish as food. It entered the artificial burrow on the first day and remained in the burrow for the duration of the experiment (Fig. 5). During the acclimation period, the stomatopod used loose sand in the tank to form a cap for the artificial burrow and consistently maintained this cap by collecting excess sand in the burrow with its maxillipeds. The other aquarium was kept host-free.

An experimental community of four species (D. octotentaculata, D. luteocrinita, “P”. squillina, and D. troglodytes) were introduced, in proportions approximate to their natural IRL frequencies, to both tanks. The host-free aquarium housed 47 clams (34 D. octotentaculata, seven D. luteocrinita, five “P”. squillina, and one D. troglodytes). The host-containing aquarium housed 46 clams (33 D. octotentaculata, seven D. luteocrinita, five “P”. squillina, and one D. troglodytes). Clams were introduced in small cohorts of conspecifics placed between the burrow openings and observed for 15 min. Those that did not enter the burrow within 15 min (i.e., remained on the sand surface or climbed the aquarium glass) were manually transferred into the burrow after this observation period. A mix of cultured unicellular green and brown algae was added to the tanks once a day and each tank was observed for patterns of spatial use, grouping behavior, and mortality. To simulate the assumed light conditions of natural stomatopod burrows, the exposed side of the artificial burrow was covered with thick black plastic bags level with the burrow entrances when observations were not taking place.

Commensal occurrence and distribution

A total of 86 host burrows were sampled and 29 of these (33.7%) yielded ≥1 commensal vasconielline clam(s), collectively totaling 112 specimens from 6/7 of the known IRL vasconielline species (Goto, Harrison & Ó Foighil, 2018) and ranging in frequency from 1–21 commensals per burrow (Table S1). The burrow-wall commensal Divariscintilla octotentaculata was numerically dominant with 73 individuals sampled from 18 burrows (Fig. 6) distributed among 4/5 sampling sites (Table 1). Its ectocommensal sister species, “Parabornia” squillina, was the next most numerous with 20 individuals recovered from eight host burrows (Fig. 6), also distributed among four sampling sites (Table 1). Respective numbers for D. luteocrinita and D. troglodytes were 13 and three individuals from nine and two host burrows (Fig. 6) among three and two sites (Table 1). The products of non-adaptive within-host speciation, sister species D. yoyo and D. aff. yoyo, were the least numerous commensals recovered: respectively, two and one and both from single burrows (Fig. 6, Table 1). Sampled individuals of the four most common species ranged two-fold in mantle length (Table S2), likely representing different age classes. No specimens of the rarest IRL commensal vasconielline (Mikkelsen & Bieler, 1992), D. cordiformis, were recovered.

Of the 29 burrows with commensals, 22 (75%) were monospecific (either Divariscintilla octotentaculata, D. luteocrinita, or “Parabornia” squillina), four had 2-species assemblages, and three had 3-species assemblages (Fig. 7). Burrows with multispecies assemblages (N = 7) shared two characteristics: they were exclusively comprised of burrow-wall commensals, and all contained D. octotentaculata individuals. The ectocommensal “P”. squillina was not recovered from any host burrow that also yielded burrow-wall commensals (Divariscintilla spp.).

The over-dispersion tests of “P”. squillina and D. octotentaculata rejected the null hypothesis, meaning that for each species, the observed frequency distribution was significantly (P < 0.001) more clustered than that expected by chance alone. For the comparative recruitment simulation tests (Fig. 8), when the five burrow-wall commensal species were collectively treated as one group in comparison to the ectocommensal “P”. squillina, the null hypothesis of random assembly was strongly rejected (P < 0.001). “P”. squillina was never observed to co-occur with burrow-wall commensals in the field, whereas in the simulated random recruitment scenarios there were always at least four burrows where the two groups co-occurred. Similar results were found for individual ectocommensal/burrow-wall commensal simulation tests: “P”. squillina and D. octotentaculata, as well as “P”. squillina and D. luteocrinita (Fig. 8). In stark contrast, when evaluating co-occurrence among the burrow-wall commensals D. octotentaculata and D. luteocrinita (Fig. 8), the observed field value fell well within the simulated random recruitment distribution (P = 0.405), indicating these two burrow-wall commensal species likely co-recruited to IRL host burrows following a random process.

Stable isotope analyses of dietary niche

Individuals of the six IRL commensal species sampled, together with samples of their potential basal resources (host tissue, burrow-water POM, and burrow-wall sediment), were analyzed to determine their respective carbon and nitrogen isotopic signatures (Table S3). Only three of the six commensals—Divarscintilla octotentaculata, “Parabornia” squillina and D. luteocrinita—were recovered in sufficient numbers to generate isotopic analysis Bayesian-estimated Standard Ellipse Areas (SEA_B) plots, and their respective SEAc values were 0.184, 1.128, and 0.255. The corresponding SEAc values for host tissue, burrow-water POM, and burrow-wall sediment samples were respectively 0.826, 3.354, and 7.336.

Divariscintilla octotentaculata’s inferred isotopic niche space (its Bayesian-estimated ellipse area) was distinct from that of D. luteocrinita (Fig. 9), a fellow burrow-wall commensal, but it overlapped substantially with that of “Parabornia” squillina (Fig. 9), its ectocommensal IRL sister species. All three commensal species did not overlap in isotopic niche space with any of their three potential basal resources but they placed closest to burrow-water POM, and furthest away from the host Lysiosquilla scabricauda (Fig. 9).

Artificial burrow observations

Individuals of all four commensal species (Divarscintilla octotentaculata, “Parabornia” squillina, D. luteocrinita and D. troglodytes) introduced into aquaria containing artificial host burrows (with and without a mantis shrimp host) preferred firmer surfaces to the loosely packed aquarium surface sand. Most clams that encountered the edge of an artificial burrow opening crawled down that burrow and most that encountered the aquarium glass wall crawled up that surface (prior to being manually relocated into the artificial host burrow).

In all three observations of the host-free aquarium (Figs. 10A–10C), commensal clam species were partially intermixed throughout the artificial burrow. The most numerous species, Divarscintilla octotentaculata, exhibited the clearest spatial aggregation with most individuals dominating the left 1/3rd of the burrow, leaving the rest of the burrow occupied primarily by a mixture of Parabornia squillina and D. luteocrinita (Figs. 10A–10C). Most individuals, regardless of species, were located within the horizontal segment of the artificial burrow, primarily attached to the lateral and upper burrow walls. Commensals were typically sedentary during observation periods, though there were changes in individual positioning between observations and two D. octotentaculata specimens left the burrow and attached to the aquarium walls.

In the host-containing aquarium, the initial disturbance associated with commensal introduction led the host mantis shrimp to attempt to cover up the light-exposed glass with a sand-mucus mixture. Following this, it rested within the burrow and the commensal clams gradually positioned themselves around it. By the first observation period, 15 h post-introduction (Fig. 10D), most commensals had formed a mixed species assemblage in the horizontal segment of the artificial burrow where the host primarily rested (although three Divarscintilla octotentaculata individuals had exited the burrow and were attached to the aquarium walls). Most burrow-wall commensals attached to the upper burrow wall where many engaged in characteristic “yo-yo” behavior in response to being touched by the host. Within-burrow positioning of the five ectocommensal “Parabornia” squillina individuals varied from observation to observation (Figs. 10D–10F). Although none immediately moved onto the host, at +15 h (Fig. 9D) three had attached to the base of the host pleopods and two were attached to the upper burrow wall. At +27 h (Fig. 10E) two P. squillina individuals remained attached to the host and three to the burrow wall (two upper, one lower), and at +63 h (Fig. 10F) one individual remained attached to the host.

Commensal survivorship in the experimental artificial burrows decreased with time and two quantitatively and qualitatively distinct patterns of commensal mortality were evident during the observation period (Figs. 10A–10F). One pattern was independent of host presence: in 5/6 burrow observations (Figs. 10A–10E), a “background” rate of mortality, ranging from 1-8 individuals per time increment, was characterized by the presence of dead clam bodies lying on the bottom of the burrows. In contrast, a greatly elevated mortality rate (34/35 commensals) was detected in the +63 h observation of the host-occupied burrow (Fig. 10F), characterized by the absence of observable dead clam bodies or clam tissue/shell fragments within or outside the burrow. The sole survivor was a single individual of the ectocommensal “Parabornia” squillina attached to the base of the host pleopods (Fig. 10F).

Synthesis

Collective consideration of the IRL field census, stable isotope, and captive behavioral data yields a heterogenous vista of symbiont coexistence and of symbiont exclusion. It may therefore be useful to view this study system as being composed of two distinct, superimposed patterns of commensal distribution: (1) all burrow-wall commensal species; (2) the ectocommensal species.

In this framing, the six burrow-wall commensals broadly adhere to neutral theory (Hubbel, 2001) expectations of species assembly in that they co-occur seamlessly in space and time, at least in the sections of IRL host burrows reached by yabby pump sampling (Mikkelsen & Bieler, 1992; this study). This is consistent with numerous studies that have found little evidence for competitive exclusion in marine benthic communities (Stanley, 2008; Shinen & Navarrete, 2014; Klompmaker & Finnegan, 2018). Such studies often emphasize the role of high rates of marine predation and disturbance in minimizing competition (Klompmaker & Finnegan, 2018) but another, possibly more apt, model for IRL burrow-wall commensal coexistence might be Laird & Schamp’s (2006) finding that coexistence of ≥3 competitors is possible if the competition is non-hierarchical. That important detail remains to be determined but at least one burrow-wall commensal (Divariscintilla luteocrinita) showed evidence of trophic differentiation (Fig. 9), and 4/6 appear to have fluctuated in relative frequency between 1987 and 2017 (Mikkelsen & Bieler, 1992; this study). Goto, Harrison & Ó Foighil’s (2018) vasconielline phylogeny established that the burrow-wall niche and its associated “hanging-foot” morphology is plesiomorphic among IRL commensals, implying that within-burrow coexistence may also be the ancestral condition. If so, it has proven to be remarkably stable and has survived the repeated addition of new burrow-wall commensals, mainly through within-host (ostensibly non-adaptive) speciation, but also through host switching (Goto, Harrison & Ó Foighil, 2018).

In contrast, “Parabornia” squillina’s apparent inability to coexist with other Lysiosquilla scabricauda commensals in IRL host burrows, despite its unique ectocommensal niche, is incongruent with both competitive exclusion principle and neutral theory expectations. Goto, Harrison & Ó Foighil’s (2018) vasconielline phylogeny shows that the ectocommensal niche is (1) a derived condition among IRL commensals; (2) a product of within-host adaptive speciation. In this case, within-host adaptive speciation involving clear ecological character displacement (Goto, Harrison & Ó Foighil, 2018) has apparently led to the introduction of strict ecological exclusion (and a truncation of realized niches) to a commensal community hitherto characterized by comprehensive co-existence: an inverse outcome of theoretical expectations.

The ecological factors regulating the observed IRL burrow-wall commensal/ectocommensal exclusion are currently obscure but potentially include differential recruitment to individual IRL host burrows and/or differential survival in “mixed-niche” burrow assemblages. Our field census data unfortunately could not distinguish among those possibilities because they did not include newly recruited juvenile commensals: based on prodissoconch sizes, they metamorphose out of the plankton at 350–390 µm in length (Mikkelsen & Bieler, 1989; Mikkelsen & Bieler, 1992; Fig. S1) and our smallest recovered specimen was 2.5 mm in mantle length. Resampling IRL host burrows during commensal recruitment peak periods using a sufficiently fine mesh sieve could address this deficiency. Replication of the adult exclusion pattern by juveniles, or detection of juvenile-specific “mixed-niche” burrow assemblages, would respectively support differential recruitment, or differential survival, exclusion mechanisms.

Current evidence for either potential exclusion mechanism is fragmentary at best. Regarding differential recruitment, the challenge is to explain the lack of “Parabornia” squillina recruitment to host burrows supporting multi-species burrow-wall commensal assemblages, and/or vice versa. Commensal galeommatoideans typically display positive chemotaxes to their respective hosts (Morton, 1962; Gage, 1968; Gage, 1979; Ockelmann & Muus, 1978), as apparently does “P”. squillina (Goto, Harrison & Ó Foighil, 2018). Preventing “mixed-niche” IRL recruitment of the seven IRL commensal species to the same exclusive host might require counteracting among-commensal negative chemotaxes/behaviors (burrow wall commensal species vs ectocommensal species or vice versa). Our laboratory behavior experiments (Fig. 10), show no clear evidence for such.

A variety of potential drivers, competitive and/or predatory, might contribute to differential burrow-wall commensal vs ectocommensal survival in “mixed” burrow assemblages. Note that the formal concept of competitive exclusion (Hardin, 1960) turns out to be inapplicable to this study system because it requires the species that cannot coexist—burrow-wall commensals and the ectocommensal (not subsets of the burrow-wall commensals as initially hypothesized)—to have identical niches, and this is clearly not the case (Fig. 1). As discussed above, evidence that a trophic competition driver is influential in regulating this system is mixed at best for the three commensal species with characterized trophic niches, with the sole member of the three to show trophic differentiation, Divariscintilla leuteocrinita (Fig. 9), co-occurring only with other burrow-wall commensals (Fig. 7).

A context-specific change in Lysiosquilla scabricauda predatory behavior is another potential driver of differential commensal survival i.e., selective host exclusion. Our artificial burrow behavioral experiment was unexpectedly terminated by the host-induced mass mortality of all non-host-attached commensals (Figs. 10E & 10F). That host behavioral change might have been triggered by starvation because captive mantis shrimp refused to feed on offered prey fishes (a response readily seen in the field). However, a host-starvation trigger does not explain the absence of ectocommensals in IRL burrows containing burrow-wall commensals (Fig. 7). An alternative trigger might be that the act of ectocommensal attachment itself induces a change in host predatory behavior leading to the eradication of co-occurring burrow wall commensals. This may seem far-fetched, but it is fully congruent with the observed field distribution data (Fig. 7) and mantis shrimp are behaviorally complex organisms with extraordinary visual systems (Thoen et al., 2014; Thoen et al., 2017; Franklin et al., 2017; Patel & Cronin, 2020). It could also be tested experimentally by tracking burrow-wall commensal survival in host burrows with (treatment) and without (control) added ectocommensals.