Predation governs the eulittoral distribution of a ubiquitous Mediterranean gastropod

Topshell gastropods are among the most important intertidal biofilm grazers. Present throughout the Mediterranean, this sea’s only exclusively eulittoral trochid, Phorcus turbinatus, is spatially separated from its infralittoral congeners, P. articulatus, P. richardi and P. mutabilis. Despite its ubiquity, data on its fundamental ecology are lacking. By a series of laboratory and field experiments, I show that predation fixes the lower boundary of the vertical range inhabited by P. turbinatus precisely at lowest astronomical tide (LAT). Predicted in situ predation risk rapidly increases from 30% d–1 at the mean position of P. turbinatus (+0.31 m) to 84% d–1 at 1 m below LAT. At least nine predators across all major phyla prey on P. turbinatus. Important in terms of abundance, zonation and frequency of predation events are the two muricids Stramonita haemastoma (+0.08 m) and Hexaplex trunculus (−1.73 m) on exposed and sheltered coasts, respectively, along with Thalassoma pavo (−0.56 m) and Hermodice carunculata (−1.24 m). Physiological constraints seem to be of secondary importance since P. turbinatus survives up to 1 week fully immersed in situ. Therefore, predation risk likely is the ultimate driver of vertical distribution and thus physiological adaptation of littoral Mediterranean trochids.


Introduction
While the upper boundary of eulittoral animals is logically set by the physical constraints of the atmosphere (desiccation, variable salinity, temperature fluctuation, low fluid density), drivers of the lower limit are of ecological nature (competition, predation, food availability) and therefore less straightforward (Ricketts et al., 1985;Raffaelli & Hawkins, 1999). Of course, physical factors play a role in governing the lower limit of fauna whose physiology is adapted to the high shore (Tait, 1918;Bannister et al., 1966;Micallef & Bannister, 1967;Houlihan & Innes, 1982), but they are often unsurprising given the animal's distribution and cannot be seen as the ultimate cause of zonation. Initiation of physiological adaptation in the evolutionary past must thus be of ecological nature (Raffaelli & Hawkins, 1999). The lower limit of sessile organisms, such as mussels and barnacles in the Atlantic and Pacific, is principally set by interspecific interference competition and predation (Connell, 1961;Dayton, 1971;Paine, 1974). Less well studied than sessile organisms, the case of mobile fauna is backed by almost no evidence suggesting that interspecific competition or food availability controls vertical distribution (Underwood, 1979;Raffaelli & Hawkins, 1999). Predation therefore seems to be the main vertical structuring factor for gastropods, the dominant intertidal grazers (Underwood, 1979). Their lower boundary is governed by predators including starfishes (Phillips, 1976;Fawcett, 1984;Watanabe, 1984a), crabs (Fawcett, 1984;Yamada & Boulding, 1996;Perez et al., 2009) and molluscs (Garrity & Levings, 1981;McQuaid, 1982;Fawcett, 1984).
Optimal model systems to study biological constraints on the lower boundary of eulittoral gastropods are sets of morphologically similar congeners with different vertical distributions. Due to their close phylogenetic relationship, they provide evolutionary snapshots of an upward vertical migration and allow us to tease apart factors affecting their current distribution. Some of the best-studied sister species are to be found among the topshells (Trochidae), the key herbivores aside from limpets (Patellidae) and winkles (Littorinidae). The oldest known case is that of Steromphala umbilicalis (da Costa, 1778) and S. cineraria (Linnaeus, 1758) in the Northeast Atlantic, the former occurring between 0.77 ± 0.26 and 3.83 ± 0.23 m (mean ± s.e.m.) above lowest astronomical tide (LAT), the latter below 1.44 ± 0.14 m above LAT (Colman, 1933;Evans, 1947a, b;Smith, 1969;Underwood, 1973). Yet, although the distribution of these trochids has been mapped repeatedly, underlying mechanistic drivers remain unclear. Interspecific competition has been suggested on a correlative basis (Smith, 1969) but not proven experimentally (Underwood, 1979). The diets of these trochids are unfortunately insufficiently studied, but while only S. cineraria lives on kelp, both seem to consume it (Pessarrodona et al., 2019b;Gilson et al., 2021), so food availability may not be important. Even the case of predation is unclear since the taller shell of infralittoral S. cineraria is more vulnerable to crushing than that of eulittoral S. umbilicalis but this is behaviourally compensated with faster evasion (Cotton et al., 2004). In the Northeast Pacific, Tegula funebralis (A. Adams, 1855), T. brunnea (Philippi, 1849), T. montereyi (Kiener, 1850 and T. pulligo (Gmelin, 1791) occur in succession from lower eulittoral to mid infralittoral (Riedman et al., 1981;Watanabe, 1984a). The last three prefer the same kelp food (Watanabe, 1984b) and as there is no evidence in support of interspecific competition, predation seems to be the most important structuring agent (Watanabe, 1984a). However, even this factor remains obscure since T. brunnea and T. pulligo, expected to have the highest and lowest mortalities, respectively, both have inferior anti-predator defences to T. montereyi (Watanabe, 1983) and are consequently consumed more frequently (Watanabe, 1984a).
In contrast to these well-studied Atlantic and Pacific systems, there is no knowledge on the ecology of gastropods living in the microtidal Mediterranean eulittoral. The only trochid that exclusively lives in this zone, where it can reach densities of up to 600 m -2 , is Phorcus turbinatus (Born, 1778) (Bannister et al., 1966;Micallef & Bannister, 1967;Crothers, 2001;Schembri et al., 2005). P. turbinatus has three morphologically similar Mediterranean endemic congeners, P. articulatus (Lamarck, 1822), P. richardi (Payraudeau, 1826) and P. mutabilis (Philippi, 1851) (Sousa et al., 2018). Their geographical ranges overlap throughout the Mediterranean except for the coasts of southern Sardinia and north-western Africa (Donald et al., 2012). P. turbinatus is 8 [4, 13] Ma (posterior mode [95% highest posterior density interval]) old and its ancestor diverged from the ancestral line of the younger P. articulatus, P. richardi and P. mutabilis 14 [9, 19] Ma ago (Donald et al., 2012). In contrast to P. turbinatus, P. articulatus usually inhabits the upper infralittoral (Crothers, 2001;Cheour et al., 2014;Boulajfene et al., 2019) where it lives among and likely grazes on seagrass and macroalgae (Cheour et al., 2014;Zaabar et al., 2018). Sometimes it is found in the eulittoral (Houlihan & Innes, 1982;Crothers, 2001) but where it coexists with P. turbinatus, it lives lower on the shore (Crothers, 2001). P. richardi is also found in the upper infralittoral (Houlihan & Innes, 1982) and P. mutabilis is said to inhabit a similar zone (Gofas & Jabaud, 1997). Notwithstanding this vague knowledge on the zonation of these Mediterranean trochids, data on their vertical distribution in relation to LAT are entirely lacking. Besides physiology (Bannister et al., 1966;Micallef & Bannister, 1967;Houlihan & Innes, 1982), published data are restricted to reproduction and anthropogenic impacts such as metal pollution and harvesting (Sousa et al., 2018). Therefore, severe knowledge gaps are present in our understanding of the fundamental ecology of Mediterranean littoral trochids, most notably the widespread P. turbinatus.
Here I aim to address these knowledge gaps by testing for mechanistic drivers limiting the distribution of P. turbinatus to the eulittoral. Due to the outlined evidence on other eulittoral gastropods, I hypothesised that the main driver is predation by an unknown number of unknown predators, which I sought to identify. Specifically, I conducted a field experiment to measure daily predation risk along a depth gradient. This experiment was designed to test if the probability of predation is affected by vertical position on the shore and therefore is important in fixing the lower boundary of the zone inhabited by P. turbinatus. This was followed up with a laboratory experiment on daily predation risk in confinement with confirmed predators where aquarium water level was varied to test if an increase in predator-free airspace affects predation probability. In a second field experiment, I measured the mortality of fully immersed P. turbinatus without predator access to account for mortality due to factors other than predation, such as oxygen and food limitation. To obtain knowledge of vertical and horizontal habitat preferences, I carried out field surveys alongside my experiments to determine the vertical distribution and density of P. turbinatus and various predators I had observed during the initial field experiment. The overarching goal of these experiments and observations was to achieve a better understanding of the fundamental ecology of P. turbinatus and propose a causal model for its evolution.

Study site
The Maltese Islands are located almost exactly in the centre of the Mediterranean and are dominated by a rocky coastline due to their limestone karst nature (Pedley et al., 1976). Aside from P. turbinatus (Bannister et al., 1966;Micallef & Bannister, 1967;Crothers, 2001;Schembri et al., 2005; personal observation), the coastal waters of these islands are inhabited by other topshells including P. articulatus (Crothers, 2001; personal observation), Steromphala divaricata (Linnaeus, 1758) (Bannister et al., 1966) and the Maltese endemic S. nivosa (A. Adams, 1853) (Evans et al., 2011). There are no published observations on P. richardi or P. mutabilis living around the Maltese Islands and I have never encountered them on frequent dives. That being said, they may live less abundantly among the islands' seagrass meadows, consisting of Posidonia oceanica (Linnaeus) Delile and Cymodocea nodosa (Ucria) Asch., or shallow macroalgal forests, dominated by Cystoseira compressa (Esper) Gerloff et Nizamuddin and C. amentacea var. stricta Montagne (Cormaci et al., 1997), and I could easily have missed them. This study was conducted along the north-western coastline of Gozo (also Għawdex) ( Fig. 1a) which is the second largest island of Malta and located in its north-west.

Field surveys
Between 29th May and 30th June 2021, I carried out field surveys at three Gozitan rocky shore sites with contrasting exposure (Fig. 1a). The prevailing winds in Malta are north-westerlies and westerlies (Galdies, 2011). Consequently, I chose the sheltered side of Il-Bajja tax-Xwejni on the north coast (36.0795°N, 14.2475°E) as my most sheltered, inside Il-Qala tad-Dwejra on the west coast (36. 0482°N, 14.1919°E) as my intermediate and the outside of Id-Dejjaq tar-Ras on the west coast (36.0483°N, 14.1893°E) as my most exposed site (Fig. 1a). Apart from these differences in wind and thus wave exposure, all sites likely experience similar irradiance and temperature due to their geographical proximity (max. 6.28 km distance, Fig. 1a). Il-Qala tad-Dwejra and Id-Dejjaq tar-Ras have a similar rock face slope (~45º) and are composed of lower coralline limestone (also Żonqor, 28.1-23.03 Ma) (Pedley et al., 1976). Il-Bajja tax-Xwejni is made of more shallowly sloping lower Globigerina d'Orbigny, 1826 limestone (also Franka, 23.03-20.43 Ma) (Pedley et al., 1976) and has an abundance of rock pools. The vertical section of limestone inhabited by P. turbinatus is devoid of macrophytes at all sites but borders onto Cystoseira C. Agardh forest below. It is possible that the sites differ in other minor factors that are difficult to observe but therefore likely have little effect. Within a 100-m rocky shore section at each site ( Fig. 1a), I measured P. turbinatus density with 100 haphazardly placed 40 × 40-cm quadrats (cf. Underwood, 1973;Schembri et al., 2005;Perez et al., 2009) and vertical position in relation to seawater level with a folding rule (± 1 cm). Position measurements were then standardised to lowest astronomical tide (LAT). Gastropod individuals were collected from their eulittoral zone at all sites for subsequent field and laboratory experimentation. In the case of in situ experiments, I always deployed individuals at their home site.

Field experiments
Over the survey period, I performed two field experiments across all surveyed sites, one to assess survival with, the other without predator access. During the first, I tethered 100 P. turbinatus individuals across sites (36 at Il-Bajja tax-Xwejni, 33 at Il-Qala tad-Dwejra and 31 at Id-Dejjaq tar-Ras) to weights with thin cotton thread by drilling a 1-mm diameter hole in the last shell whorl (Fig. S1, cf. McQuaid, 1982;Watanabe, 1984a;Perez et al., 2009). After measuring shell length (± 0.01 mm, digital callipers, ACCUD CO., LTD., Vienna, Austria) and total mass (± 0.01 g, TGC 150-2, KERN & SOHN GmbH, Balingen, Germany), I randomly deployed each individual on the rocky shore. Position in relation to LAT, ranging from 0.84 m above LAT and 8.97 m below LAT, was measured with a folding rule as described above or with the depth sensors of a dive computer (Chromis, SCUBAPRO, Nürnberg, Germany) and underwater camera (Tough TG-5, Olympus SE & Co. KG, Hamburg, Germany). After 24 h, I recorded whether the snail had been eaten (1) or not (0) and by what predator, if present upon retrieval. Predation events were recorded with certainty since snails were either completely removed including the operculum, leaving behind an empty shell, or partially consumed in cases where predators were still present. Snails that succumb due to causes other than predation, such as suffocation or starvation, usually have intact opercula that they clamp shut. I therefore recorded 0 if the gastropod including operculum was present, regardless of dead or alive. I mapped the vertical distribution of the four most commonly observed predators using ruler, dive computer and camera as described above and measured the densities of the two most abundant predators with a 40 × 40-cm quadrat or 1 × 10-m belt transect (cf. Garrity & Levings, 1981;Fawcett, 1984;Rilov et al., 2001). Quadrats

Id-Dejjaq tar-Ras
Il-Qala tad-Dwejra Il-Bajja tax-Xwejni *** *** were haphazardly placed and transects laid parallel to shore, all individuals within counted and the count divided by the area in square metres. During the second experiment, I deployed a total of 200 P. turbinatus individuals across sites (80 at Il-Bajja tax-Xwejni, 40 at Il-Qala tad-Dwejra and 80 at Id-Dejjaq tar-Ras) below LAT under the exclusion of predators. To achieve this, groups of 10 or 20 gastropods were placed in one to four (total of 13) 20 × 20-cm mesh bags with 1-mm 2 mesh aperture across five experimental repeats (three at Il-Bajja tax-Xwejni and one each at the other sites). Each snail density was evenly represented across experiments (6 × n = 10 and 7 × n = 20). As a weight, each bag contained a pebble of less than 10-cm diameter from the supralittoral of Il-Bajja tax-Xwejni and therefore free of biofilm. Bags were deployed between 0.5 and 4.37 m below LAT in the afternoon, retrieved precisely after 0.64-7 d (calculated to the minute) and survivals (1) and mortalities (0) recorded. Survivals were recorded with certainty by observing each gastropod for a few minutes by which time it usually started moving. In case it did not, I tapped against its closed operculum to look for a retraction response. I chose to vary the experimental duration to be able to model survival probability as a function of the explanatory variable immersion time. In addition, with the instrumentation described above I measured total mass, if the gastropod was still present in the shell, and shell length. For empty shells in cases of early mortality and progressed tissue decay, I derived total mass from shell length using the power equation y = 0.001x 2.75 (t test of slope k, t 290 = 15.36, P < 0.001) which I obtained from 291 paired measurements by generalised nonlinear least squares (Fig. S2).

Laboratory experiments
During laboratory experiments, I kept organisms in two 39 × 24 × 24-cm, 22.46-l glass aquaria (MARINA, HAGEN GmbH & Co. KG, Holm, Germany) filled with fresh seawater from Il-Bajja tax-Xwejni and aerated with an air pump equipped with air stones (air 275 R plus, sera GmbH, Heinsberg, Germany). Each aquarium was separated into two compartments of equal dimensions and covered with 1-mm mesh (P. turbinatus will readily crawl about the laboratory!). At the beginning of each 24-h experiment, I cleaned and refilled the tank with fresh seawater and placed one confirmed muricid predator of known mass and shell length in each compartment with a single P. turbinatus for a total of 41 independent predator-prey pairs. Seawater temperature was monitored throughout the experiment using a digital thermometer (± 0.1 °C, DigiScan, JBL GmbH & Co. KG, Neuhofen, Germany) and kept at 0.71 ± 1.92 °C (mean ± s.d.) above ambient sea surface temperature (Met Office, 2021). Thus, aquaria conditions were representative of the warmer and less stable eulittoral temperature and no specimens succumbed to in situ conditions. As with the first field experiment, I recorded whether the snail had been eaten (1) or not (0) after 24 h. Between experiments, water level was randomly varied at 4-cm intervals between 4 and 24 cm (3.74-22.46 l) to test its effect on predation risk. Each water level treatment was repeated four to eight times (mean = 5.67). In a single preliminary behavioural experiment, I compared the vertical movement response of five P. turbinatus individuals separated from one confirmed muricid predator by 1-mm mesh to five control individuals over a period of 15 min.

Data analysis and visualisation
I analysed and visualised data in R v4.1.2 (R Core Team, 2022) within the integrated development environment RStudio v2021.09.2 (RStudio Team, 2022). My primary tool of statistical inference in this study are frequentist linear models and their generalised and multi-level derivatives. Model optimality was determined based on graphical scrutiny of residuals for homogeneity of variance and Gaussian distribution along with common information criteria (Zuur et al., 2009).
Due to their heterogeneity and right skew and the suboptimality of data transformation (Zuur et al., 2009), density data were analysed with a generalised linear model based on the gamma distribution and logarithmic link function. Vertical distribution data were also heterogenous but normal. I therefore weighted the residual variance by species with generalised least squares using nlme v3.1-153 (Pinheiro et al., 2021). The relationship between total gastropod mass and shell length is nonlinear, so I approximated it using a generalised nonlinear model with a power equation in nlme v3.1-153.
Due to their binomial nature (0, 1), in situ and in vitro predation risk along with survival data required modelling using generalised linear models based on the binomial distribution with cauchit, complementary log-log (cloglog) and probit link functions, respectively. In the case of survival data, groups of 10 or 20 gastropods were contained in the same mesh bag (see above) and are therefore not statistically independent, introducing additional uncertainty. This is a case of hierarchical design that requires hierarchical (multi-level) modelling. Therefore, I modelled these data using a generalised linear mixed effects model with a random intercept for each bag, allowing me to account for the additional uncertainty (variance of the random intercept). As different snail densities (10 or 20 per bag) could potentially affect the response of survival to immersion time, I tested for an interaction between time and density during model fitting. There was no interaction (likelihood ratio chi-square test, X 2 = 2.19, P = 0.14), meaning the effect of time on survival probability is not different between densities, so the interaction term was removed in the optimal model. Since binomial data are not symmetrically distributed around the mean and normal approximation (Wald) is consequently suboptimal, asymmetric Wilson score s.e.m. were calculated with binom v1.1-1 (Dorai-Raj, 2015).
Across sites, the vertical distribution of P. turbinatus is centred at 0.31 ± 0.01 m above LAT (n = 720) with a lower boundary at exactly 0 m above LAT. Here, I propose that this pattern is ultimately caused by predation. The relationship between vertical position on the shore (v) and the probability of predation (p) is best described by the equation p = tan −1 (−1.9×v−0.14) + 0.5 (Fig. 2a, Fig. 2b), where P. turbinatus is encountered with highest probability. Interestingly, daily predation risk does not change with gastropod mass (likelihood ratio chi-square test, X 2 1,100 = 2.16, P = 0.14, n = 100), which was 2.53 ± 0.12 g on average for tethered individuals.
The muricids S. haemastoma and H. trunculus live on exposed and sheltered coasts, respectively, but both coexist at Il-Qala tad-Dwejra, my site of intermediate exposure (Fig. 1). Mesh bags baited with live P. turbinatus readily attracted large groups of S. haemastoma (cf. Brown & Alexander Jr, 1994) and H. trunculus (cf. Peharda & Morton, 2006) at exposed and sheltered sites, respectively. While I was unable to clarify whether these muricids were attracted to chemical cues of live trochids or those that had succumbed inside the bags, this indicates they can actively hunt P. turbinatus via chemoreception and are not reliant on chance encounters. This is supported by laboratory observations of H. trunculus chasing and attacking P. turbinatus (Fig. S3c, d). T. Fig. 2 Effect of predation on the vertical distribution of Phorcus turbinatus above lowest astronomical tide (LAT). a Probability of predation on P. turbinatus in relation to vertical position across sites. The line and shaded regions are the cauchit binomial generalised linear model fit and 95% confidence interval. Note that two outlying datapoints at 6.57 m (1) and 8.97 m (0) below LAT are not plotted here but included in the model. b Vertical distribution of P. turbinatus and its most fre-quent predators pooled across sites. Curves, solid and dashed lines are kernel density estimates, medians and means. Shaded regions are the central 95% of probability density. The black line indicates the position in relation to LAT above which no predation events were recorded. Datapoint size is inversely scaled to sample size. Asterisks indicate statistical significance of pairwise contrasts at ɑ = 0.05: *P < 0.05, ***P < 0.001 pavo and H. carunculata are ubiquitous although the latter only appears in summer. While the former is obviously an active hunter, I did not observe the latter aggregating as abundantly as the muricids so it may rely on chance encounters (cf. Simonini et al., 2017Simonini et al., , 2018. I estimated the densities of the most abundant predators, S. haemastoma and T. pavo, to be 8.63 ± 0.77 m -2 (n = 100) and 0.8 ± 0.06 m -2 (n = 40), respectively. H. trunculus was generally rare except when occurring in feeding aggregations and H. carunculata only became more abundant as seawater temperature increased above 26 °C towards the end of my experimental period.
In subsequent in vitro experiments, I did not observe any of seven S. haemastoma eating P. turbinatus within a 24-h period but 35.29 [21.49, 52.09] % d -1 (n = 34) succumbed to H. trunculus. Predation risk did not change with snail mass (likelihood ratio chi-square test, X 2 1,34 = 0.45, P = 0.5, n = 34) or aquarium water level (likelihood ratio chi-square test, X 2 1,34 = 1.82, P = 0.18, n = 34) but the visualised data suggest a slight decrease in risk with increasing water level (Fig. S4). This is contrary to my expectation that vertical air space helps P. turbinatus evade predation. Instead, H. trunculus may be a more successful hunter when the water volume is lower and the concentration of prey and its kairomones consequently higher. Conversely, P. turbinatus does not seem to respond to the mere presence of H. trunculus, since neither kairomone-treated nor control group individuals moved above water level. During attacks of H. trunculus, however, direct contact does not seem necessary to elicit chemically mediated avoidance behaviour either (Fig. S3c). This suggests that P. turbinatus may be unable to detect muricid kairomones from a distance and/or can only detect kairomones exuded during attack. Both results certainly need empirical validation considering their low sample size, statistical ambiguity and preliminary nature.
Aside from these primary predators in terms of abundance and frequency of predation events, I observed the starfish Coscinasterias tenuispina (Lamarck, 1816) preying on P. turbinatus (Fig. S3e) and groups of the small mitrid Episcomitra cornicula (Linnaeus, 1758) feeding inside the shells of dead individuals. The latter species is probably too small to kill P. turbinatus and may only be a scavenger. I could not determine their vertical range due to the former's rarity and the latter's crypsis. Over three months of regular visits to the infralittoral, I only observed C. tenuispina twice. Interestingly, despite the presence of the shell-crushing crab Eriphia verrucosa (Forskål, 1775), not a single tethered P. turbinatus shell was broken after 24 h in situ. Nevertheless, I did observe E. verrucosa capturing dislodged and immersed P. turbinatus. This apparent paradox may be explained by my observation that E. verrucosa carries its prey into hiding to consume it, which is not possible with tethered snails. Tethering may underestimate predation risk in this regard which may in turn compensate potential overestimation due to restrained movement. Besides not being crushed, none of the tethered P. turbinatus shells displayed drill holes. While this at first seemed to rule out other molluscan predators such as Octopus vulgaris Cuvier, 1797, the drilling muricids H. trunculus and S. haemastoma surprisingly never drill P. turbinatus but instead force open its weaker operculum, so other molluscs could be acting similarly. The shells that were emptied (65%) were in fact in such good condition that 6% of them were occupied by the hermit crabs Clibanarius erythropus (Latreille, 1818) and Pagurus anachoretus Risso, 1827 (Fig. S3f).
Immersed P. turbinatus usually seek to position themselves at the air-water interface (cf. Underwood, 1979;Thain et al., 1985) but these snails nonetheless survive up to 7 d completely immersed in the infralittoral when predators are excluded. When submerged, the decline in survival probability (p) with time (t) is best described by the equation p = (−1.08 × t + 4.09) (Fig. 3, chi-square test, X 2 1,200 = 17.32, P < 0.001, n = 200). Therefore P. turbinatus has an immersed halflife ( t 1 2 ) of 3.79 d or 91 h. Interestingly, submerged survival does not change with gastropod mass (chi-square test, X 2 1,200 = 0.34, P = 0.56, n = 200), which was 2.12 ± 0.09 g on average for bagged individuals, indicating that body mass may not influence the ability to incur oxygen debt. Survival of immersed P. turbinatus may also have been influenced by factors unrelated to physiology, which would cause the single week reported here (Fig. 3) to be an underestimate of maximal immersed survival. First, I observed muricids clinging to most retrieved bags, so some P. turbinatus may have been attacked and eaten through the mesh. Second, P. turbinatus may have starved since neither the mesh nor the pebble weight presumably had sufficient biofilm to sustain 10-20 individuals.

Discussion
Here, I present the first data on the fundamental ecology of Phorcus turbinatus and show unequivocally that predation is the primary mechanistic driver of its vertical zonation (Fig. 4). Predation risk for P. turbinatus predictably increases as it moves down the shore (Fig. 2a). This trend is further explained by the vertical ranges of confirmed predators (Fig. 2b). Interestingly, snail mass does not influence predation risk, suggesting that the probability of being eaten is so high for P. turbinatus in the infralittoral that it overrides any potential ontogenetic effect such as that described by Pessarrodona et al. (2019a) for the purple sea urchin Paracentrotus lividus (Lamarck, 1816). Nonetheless, factors unrelated to predation, such as oxygen and food limitation, could potentially be alternative causes of mortality among immersed gastropods. I rule out these confounders by showing that immersed gastropods without predator access survive far beyond the 24-h timescale in which they incur significant risk of predation (Fig. 3). While emersed P. turbinatus can apparently survive up to 2 d longer and has a 72% longer half-life ( t 1 2 = 6.5 d, Houlihan & Fig. 3 Effect of full immersion without predator access on the survival probability of Phorcus turbinatus across sites. The line and shaded region are the probit binomial generalised linear model fit and 95% confidence interval. Point-ranges are means ± binomial standard errors. Note that the 95% confidence interval is adjusted to account for the intercept variance between mesh bags (points) as an additional source of uncertainty Fig. 4 Proposed causal model for Phorcus turbinatus zonation, which may explain the evolution of its current eulittoral distribution in relation to lowest astronomical tide (LAT) and concomitantly its required physiological adaptation. The "present" probability distribution is equivalent to the empirical kernel density estimates shown in Fig. 2b while the "past" distribution is drawn according to the assumed zonation of the infralittoral ancestor of P. turbinatus which may have been similar to that of its contemporary Mediterranean congeners Innes, 1982), likely due to the accumulation of a larger oxygen debt when immersed (Micallef & Bannister, 1967;Houlihan & Innes, 1982), its ability to respire in both fluids simply points towards perfect adaptation to the eulittoral (Micallef & Bannister, 1967;Houlihan & Innes, 1982). Therefore oxygen limitation cannot be regarded as the proximate determinant of the lower distributional limit of P. turbinatus. Regarding food limitation, the biofilm zone grazed by P. turbinatus is mostly free of predators (Fig. 2b), so the case is not as clear as with the trochid Oxystele antoni D. G. Herbert, 2015, where greatest food availability spatially coincides with highest predation risk (McQuaid, 1982). Nonetheless, there was unlikely enough food in the mesh bags so immersed mortality may have been due to starvation, indicating that this factor is also of secondary importance. A similar case in point is Ligia oceanica (Linnaeus, 1767) which, albeit inhabiting the more terrestrial supralittoral, can tolerate immersion for months without food (Tait, 1918) but is consumed by Thalassoma pavo within seconds of touching the sea surface (personal observation). The presented evidence clearly demonstrates that the immediate cause of the current zonation of P. turbinatus is predation, providing an ecological and evolutionary analogy to the previously introduced, well-known Steromphala spp. and Tegula spp. model systems in the Atlantic and Pacific and a causal model for eulittoral trochids more generally (Fig. 4).
In contrast to Atlantic and Pacific gastropod model systems, starfish and crab predation do not seem to be important in determining the distribution of Mediterranean eulittoral gastropods. While I observed that P. turbinatus is rarely eaten by the starfish Coscinasterias tenuispina or the crab Eriphia verrucosa, 89% of tethered Littorina littorea (Linnaeus, 1758) in the Northwest Atlantic (Perez et al., 2009) and 10% of Tegula spp. in the Northeast Pacific (Watanabe, 1984a) are consumed by shell-crushing predators and 90% of Tegula spp. are consumed by starfish (Watanabe, 1984a). Instead, predatory gastropods seem of primary importance in the Mediterranean. Notably, several lines of evidence suggest that besides the two confirmed muricid predators (Fig. 2b, Fig. S3a, c, d), the eulittoral pisaniid Pisania striata (Gmelin, 1791) (Taylor, 1987;Schembri et al., 2005) and the infralittoral fasciolariid Tarantinaea lignaria (Linnaeus, 1758) (Taylor, 1987) also prey on P. turbinatus. First, when I placed these predatory snails in the ultimate vicinity of P. turbinatus, they readily attack it, whereupon the trochid performs a defence response like that described by Hoffman (1980), Garrity & Levings (1981), and Watanabe (1983), featuring no retraction into the shell (Fig. S3c), violent shell rotation (Fig.  S3d), escape and rapid ascent above the water surface. Second, trochid radula remains have been found in the digestive system of P. striata (Taylor, 1987). Third, T. lignaria preys on the cerithiid Cerithium vulgatum Bruguière, 1792 and the columbellid Columbella rustica (Linnaeus, 1758) (personal observation) and P. striata feeds on C. vulgatum (Taylor, 1987). Since C. vulgatum is often the main prey of Hexaplex trunculus (Taylor, 1987;personal observation) and this muricid also feeds on P. turbinatus, it seems likely that T. lignaria and P. striata with their similar diets would do so too. However, given the cooccurrence of P. striata and P. turbinatus in the lower eulittoral zone and the rarity of T. lignaria (Taylor, 1987;personal observation), it is unlikely that these predators have a strong additional influence on the vertical distribution of P. turbinatus. Taken together, there is ample evidence that predators across all major phyla (Annelida, Mollusca, Arthropoda, Echinodermata, Chordata) readily prey on P. turbinatus, but there are likely many more species which escaped my observation.
This is the first report of any of my confirmed predators (Fig. 2b) preying on P. turbinatus. While muricids (Hoffman, 1980;Garrity & Levings, 1981) and wrasse (Evans et al., 2011) are known to be involved in limiting trochid populations, annelid predators have to my knowledge never been reported to play a structuring role in a littoral community. Nevertheless, there is considerable knowledge on the diets of these predators. Stramonita haemastoma is also called the oyster drill and known to prey mostly on bivalves, but also barnacles, bryozoans (Brown & Alexander Jr, 1994) and vermetid gastropods (Rilov et al., 2001). T. pavo consumes a plethora of species including several trochids (Kabasakal, 2001;Sinopoli et al., 2017). Hermodice carunculata is a true generalist that feeds on the sea urchin P. lividus (Simonini et al., 2017; personal observation), starfish, corals, anemones, ascidians, sponges and chitons (Simonini et al., 2018). H. trunculus eats P. lividus (Pessarrodona et al., 2019a; personal observation), C. vulgatum (Taylor, 1987; personal observation) and the mussel Mytilus galloprovincialis Lamarck, 1819 (Peharda & Morton, 2006). Although H. trunculus (Pessarrodona et al., 2019a) and other predatory snails (Hoffman, 1980;Garrity & Levings, 1981) are known to also have indirect, nonconsumptive effects on prey behaviour via chemical cues, I observed no evidence of this in the case of P. turbinatus. Only predation attempts involving direct contact or close proximity triggered the typical defensive response of shell spinning and simultaneous escape described above. While contact is expected to elicit a stronger response (Hoffman, 1980), this nevertheless indicates that P. turbinatus may be unable to detect passive-state and/or diluted muricid kairomones. A possible explanation is the rare immersion and consequently seldom encounter with predators underwater, potentially rendering early chemosensory detection inutile. However, a very similar experiment demonstrated that 67% of Nerita scabricosta Lamarck, 1822 escape after only 10 min in the same tide pool with their muricid predator Plicopurpura columellaris (Lamarck, 1816) (Garrity & Levings, 1981). Since many eulittoral gastropods readily respond to predator kairomones and defence elicitation depends on predator proximity and abundance (Underwood, 1979), additional experimentation is strictly necessary.
In summary, eulittoral P. turbinatus is preyed on by at least nine predators across all major phyla, causing over 98% mortality in the shallow Mediterranean infralittoral. When predators are excluded, P. turbinatus can survive in this zone for up to a week. Hence, the evidence presented here unequivocally points towards predation as the immediate and ultimate cause governing the lower distributional limit of this ubiquitous Mediterranean trochid. This study provides a first understanding of the fundamental ecology of P. turbinatus and hints at some evolutionary drivers that may have caused its ancestor to move up the rocky shore and diverge from the ancestor of its infralittoral congeners 14 million years ago (Fig. 4). Future studies should focus on comparing predation risk, defence mechanisms and dietary preferences between Mediterranean Phorcus spp. to further elucidate their ecology and finally reveal the interactions structuring a littoral that was already familiar to Aristotélēs.
Funding Open Access funding enabled and organized by CAUL and its Member Institutions. I received no funding for this work.
Data availability All data and code are available in the openaccess repository at github.com/lukaseamus/Phorcus. I place no restrictions on data and code availability.

Declarations
Conflict of interest I have no conflict of interest to disclose.
Research involving animal rights I followed Directive 2010/63/EU for the care and use of animals.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by 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:// creat iveco mmons. org/ licen ses/ by/4. 0/.