Comparative behavioural observations demonstrate the ‘cleaner’ shrimp Periclimenes yucatanicus engages in true symbiotic cleaning interactions

Cleaner shrimps are ecologically important members of coral reef communities, but for many species, cleaner status (i.e. dedicated, facultative and mimic), clientele and ecological role remain unverified or described. On Caribbean coral reefs, the spotted ‘cleaner’ shrimp Periclimenes yucatanicus forms symbioses with sea anemones that may serve as cleaning stations for reef fishes. The status of this species as a cleaner is ambiguous: only a single in situ cleaning interaction has been reported, and in the only test of its efficacy as a cleaner, it did not effectively reduce parasite loads from surgeonfish. It has subsequently been hypothesized by other authors to be a cleaner mimic. We conduct a comparative investigation of cleaning behaviour between P. yucatanicus and the ecologically similar, closely related, dedicated cleaner shrimp Ancylomenes pedersoni in Curacao, Netherlands Antilles. We provide the first detailed field observations on cleaning behaviour for P. yucatanicus and test multiple behavioural expectations surrounding mimicry in cleaning symbioses. We found that P. yucatanicus regularly signals its availability to clean, client fishes visit regularly and the shrimp does engage in true symbiotic cleaning interactions, but these are brief and our video reflects a species that appears hesitant to engage posing clients. In comparison to A. pedersoni, P. yucatanicus stations had significantly fewer total visits and cleans, and 50% of all cleaning interactions at P. yucatanicus stations were shorter than 10 s in total duration. Our behavioural observations confirm that P. yucatanicus is a true cleaner shrimp; we reject the hypothesis of mimicry. However, investigation is needed to confirm whether this species is a dedicated or facultative cleaner. We hypothesize that P. yucatanicus has a specialized ecological role as a cleaner species, compared to A. pedersoni.

BMT, 0000-0002-0401-1570 Cleaner shrimps are ecologically important members of coral reef communities, but for many species, cleaner status (i.e. dedicated, facultative and mimic), clientele and ecological role remain unverified or described. On Caribbean coral reefs, the spotted 'cleaner' shrimp Periclimenes yucatanicus forms symbioses with sea anemones that may serve as cleaning stations for reef fishes. The status of this species as a cleaner is ambiguous: only a single in situ cleaning interaction has been reported, and in the only test of its efficacy as a cleaner, it did not effectively reduce parasite loads from surgeonfish. It has subsequently been hypothesized by other authors to be a cleaner mimic. We conduct a comparative investigation of cleaning behaviour between P. yucatanicus and the ecologically similar, closely related, dedicated cleaner shrimp Ancylomenes pedersoni in Curacao, Netherlands Antilles. We provide the first detailed field observations on cleaning behaviour for P. yucatanicus and test multiple behavioural expectations surrounding mimicry in cleaning symbioses. We found that P. yucatanicus regularly signals its availability to clean, client fishes visit regularly and the shrimp does engage in true symbiotic cleaning interactions, but these are brief and our video reflects a species that appears hesitant to engage posing clients. In comparison to A. pedersoni, P. yucatanicus stations had significantly fewer total visits and cleans, and 50% of all cleaning interactions at P. yucatanicus stations were shorter than 10 s in total duration. Our behavioural observations confirm that P. yucatanicus is a true cleaner shrimp; we reject the 2017 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.
Cleaner species are broadly categorized as either dedicated, recently defined by Vaughan et al. [11] as a species committed to a cleaning lifestyle for all of their non-larval ontogeny, or facultative, defined as a species committed to a cleaning lifestyle for only part of their non-larval ontogeny [3,11]. However, many species are designated as cleaners through anecdotal data, the grey literature and oneoff interactions (reviewed by Vaughan et al. [11]). These often come with little evidence to support that these observations are not the result of poor taxonomic identification or incidental cleaning, or whether they enhance client fitness [2,3,11,12]. Defining a species as a cleaner also implies that their interactions with clients are symbiotic. Vaughan et al. [11] have now extended the definition of a cleaning symbiosis to include communicative behaviours (i.e. signalling) that precede cleaning interactions. These include both the signals demonstrating the willingness and availability of the cleaner, and the subsequent pose and submission of the client [11]. Thus, without systematic study, the cleaner label is likely to be overextended in many cases, misattributing the ecological role and importance of many species and larger taxonomic groups.
A total of 208 fish species and 51 shrimp species are classified as cleaners [11]. Cleaner fishes have received far more attention in the literature, although shrimps have been shown to be equally important cleaner species (reviewed by Vaughan et al. [11]). Furthermore, the cleaner definition appears to be more broadly applied to shrimps than to fishes, with entire families and genera incorrectly defined as cleaners [11]. The genus Periclimenes (Palaemonidae) has over 175 described species, all of which have been referred to as cleaner shrimps [13]. However, after the designation of the genus Ancylomenes [14], Vaughan et al. [11] note that only one species remaining in Periclimenes (P. yucatanicus) has actually been observed engaging in a cleaning interaction [15]. Interestingly, the evidence supporting the designation of P. yucatanicus as a true symbiotic cleaner shrimp is weak [15], and it has even been hypothesized to be a cleaner mimic [4,16], possibly using similar coloration to increase foraging ability and/or decrease predation pressure [17][18][19][20][21][22][23]. Thus, there is little support that any species within Periclimenes are dedicated or facultative cleaners.
Here, we use standardized observations of in situ cleaning to conduct the first field-based assessment of the cleaning behaviour and clientele of the spotted cleaner shrimp P. yucatanicus in Curacao, Netherlands Antilles. Additionally, we test a series of behavioural expectations (see Material and methods) surrounding mimicry in cleaning symbioses for P. yucatanicus to determine if previous speculation (e.g. [4,16]) surrounding mimicry in this species is warranted. We take a comparative approach by incorporating observations from a well-studied, closely related, dedicated cleaner species, Pederson's cleaner shrimp Ancylomenes pedersoni (e.g. [4,24]).
The evidence that P. yucatanicus is a true cleaner is limited to a single in situ observation by Spotte et al. [15] from the Turks and Caicos. Only one study has been conducted on the cleaning efficacy of P. yucatanicus, and in this mesocosm experiment [4] it was ineffective in reducing monogenean flatworm loads on surgeonfish. However, Titus et al. [25] hypothesized that cleaner species may specialize on different ectoparasites, which would reduce competition with other cleaners and allow for broadly overlapping client diversity. Over 20 families of reef fishes are documented as clients of A. pedersoni, and the signalling behaviour, visitation rate and interaction lengths across a number of client species are well characterized [1,4,15,16,[24][25][26][27][28][29][30]. However, none of this information is known for P. yucatanicus, despite both shrimp being sea anemone symbionts on Caribbean coral reefs, having similar patterns of host use (e.g. [31][32][33]), co-occurring on the same individual anemone host [32,34], superficially resembling each other in terms of size, pattern, coloration (figure 1) and being closely related [ These similarities are central to the hypothesis that P. yucatanicus is a mimic of A. pedersoni. While closely related, they are not sister taxa [14,34,35], and the classic cleaner patterns and coloration both species possess could be convergent or a shared primitive feature inherited from their common ancestor. In most localities, A. pedersoni is more abundant than P. yucatanicus [32,36], satisfying the requirement for negative frequency-dependent selection expected of protective mimicry scenarios (e.g. [37]). To date, mimicry has only been described between the juvenile bluestripe cleaner wrasse (Labroides dimidiatus: true cleaner) and the fangtooth blenny (Plagiotremus rhinorhynchos: cleaner mimic; e.g. [19,22,23,38,39]): the fangtooth blenny is an aggressive mimic, increasing foraging success by duping posing clients and then 'attacking', removing scales and live tissue [19,22]. No mimics have been described among cleaner shrimp.

Comparative behavioural observations
Accurately delineating the status of a species as a cleaner can be challenging. Standardized methods revolve around in-depth field observations to assess the frequency at which a prospective cleaner species is visited (proxy for ecological importance) [40], the length of the services provided (e.g. [40]), gut content analyses and food preference trials (e.g. [26,[41][42][43][44]), manipulative experiments to determine the efficacy of parasite removal (e.g. [4]) and signalling and behavioural observations. Here, we focus solely on in situ client visitation rate and clean length and provide observations of cleaner signalling.
Comparative behavioural observations for both P. yucatanicus and A. pedersoni were conducted over 10 days between 24 May and 2 June 2015 on the same fringing reef adjacent to the Caribbean Marine Biological (CARMABI) Research Institute in Curacao, Netherlands Antilles (12°06 32.4 N; 68°56 0.7 W). The CARMABI house reef is approximately 50 m from shore; the reef crest is shallow (4-6 m) and quickly slopes to depths greater than 25 m. No major ecological data were quantified, but observationally the CARMABI reef is characterized by low scleractinian coral cover (qualitative estimate less than 10%) and large, crustacean-hosting sea anemones are common. Both P. yucatanicus and A. pedersoni were common at our study site when we conducted the trials and were readily discernible from one another through in-life colour patterns; elsewhere, P. yucatanicus is typically less abundant than A. pedersoni [32,36]. As on other Caribbean reefs, not all anemones hosted crustacean symbionts [32].
Before conducting behavioural observations, we measured (tentacle crown surface area (TCSA), in cm 2 ; after Huebner et al. [45]), mapped and tagged only those anemones hosting P. yucatanicus (n = 30) and A. pedersoni (n = 15) over an approximately 80 × 20 m area of reef (approx. 1600 m 2 ). All anemones used in this study were on reef crest and shallow fore reef habitats between 4 and 8 m depth. While we would have ideally controlled for anemone host, that was not feasible given the patterns of host use at this site. All P. yucatanicus hosted with their preferred host, the giant Caribbean anemone Condylactis gigantea [32], while all but one of the A. pedersoni stations were found in association with the corkscrew anemone Bartholomea annulata (one A. pedersoni was hosted by C. gigantea). While intraspecific group sizes generally do vary [31,32,36], all P. yucatanicus stations at our study site hosted only one shrimp. Shrimp group sizes for A. pedersoni stations varied between one and six shrimp, but only three of these hosted more than a single shrimp.
Each day, we deployed GoPro Hero3 video cameras at individual stations we had not previously recorded video from and exited the water following Titus et al. [25,30], to reduce the potential for diver presence to disrupt cleaning interactions (e.g. [30]). We could not deploy cameras at each cleaning station each day because we were limited to four video cameras. Anemones with shrimp were selected randomly following initial mapping. Owing to time constraints, we did not record every tagged station. Our final data set had n = 23 cleaning stations for P. yucatanicus and n = 12 stations for A. pedersoni. Cameras were mounted to lead dive weights and placed approximately 1 m from each anemone station. An even representation of video was recorded continuously at each station (median = 150 min; IQR: 127.5-150 min) for the duration of battery life (approx. 100-200 min), with only one station recording less than 100 min of video. The duration of each recording did not significantly correlate with client visitation at either P. yucatanicus (r 2 = 0.02, p = 0.55) or A. pedersoni (r 2 = 0.0005, p = 0.94) cleaning stations. Cameras were typically deployed at 8.00, but we did not control for time of day because Titus et al. [25] showed that time of day does not affect cleaning frequency for A. pedersoni. We did not video anemones lacking cleaner shrimp to control for whether anemone species differentially attract client fishes.
Videos were downloaded and cleaning interactions analysed after Huebner & Chadwick [28,29]. We quantified (i) client visitation rate (visits min −1 ), with a visit defined as a pause, or pose, directly in front of the cleaning station that lasted at least 2-3 s regardless of whether a cleaning bout immediately followed [28,46]. (ii) Clean rate (cleans min −1 ) and the length of the interaction (s), with a clean being defined as physical contact between the shrimp and posing client reef fish [25,28,29]. (iii) Client flinches, a proxy for cheating frequency, defined as a rapid flinch or jolt by the client reef fish during a cleaning interaction [47,48]. Flinches or jolts are correlates of cheating: both Bshary & Grutter [47] and Soares et al. [48] showed that non-parasitized fish flinch in the presence of cleaners more frequently than parasitized fish. This is a standard metric to quantify cheating in field-based cleaning research (e.g. [40,[47][48][49]). For all interactions, we identified client reef fishes to species using Humann & DeLoach [50].
Our data did not conform to a normal distribution. Comparative cleaning behaviours (visitation rate h −1 , clean rate h −1 , flinch rate min −1 and total time spent cleaning) were analysed using nonparametric Mann-Whitney U-tests in SPSS v.23 [51]. All data are presented as medians and 25th-75th interquartile ranges (IQR) unless noted otherwise.

Tests of mimicry
To test the hypothesis that P. yucatanicus is a cleaner mimic, we make specific assumptions about mimicry in cleaning symbioses. We expect that aggressive (e.g. the fangtooth blenny) and visual protective mimicry (i.e. potential predators mistakenly recognize P. yucatanicus as a cleaner and do not prey upon them) are the two most plausible types of mimicry in a cleaner-mimic system. Other types of protective mimicry, such as Batesian, and Müllerian mimicry or reproductive mimicry seem less plausible because true cleaner shrimps are unlikely to be venomous or unpalatable models, or to directly use the cleanerclient relationship to enhance reproduction. Given these assumptions, we established the following expectations for how behavioural interactions between mimetic and client species should proceed under both aggressive and protective mimicry scenarios, recognizing that these may be overly simplified and not mutually exclusive alternatives. (i) Aggressive mimicry, defined here as a species with increased foraging success due to mimetic coloration. Aggressive mimics cheat posing clients and remove live tissue instead of ectoparasites [17][18][19][20]. Interactions never result in parasite removal. Under this scenario, we expect client reef fishes to pose at aggressive mimic stations with similar regularity as true cleaners, followed by an immediate cheating event and a rapid termination of the interaction. A large retaliatory response from the client typically follows. (ii) Visually protective mimicry, defined here as a species with reduced predation rates and increased fitness due to convergent coloration with a co-occurring true cleaner species [17]. Under this scenario, we expect client reef fishes to pose in front of mimetic species with similar frequency to true cleaners, but mimetic species should be indifferent towards, and not engage, posing fish. One caveat to our behavioural expectations of an aggressive mimic is that if client fishes do not rapidly terminate the cleaning interaction following an immediate cheating event, and instead continue to pose and allow repeated cheating events, it may not be possible to distinguish between periodic cheating from a true cleaner species versus cheating from a cleaner mimic through behavioural observations alone. Under this scenario, additional types of data (e.g. gut content analyses) would be needed.

Results
Periclimenes yucatanicus was slightly more abundant at our study site (1 shrimp/53 m 2 ) than A. pedersoni (1 shrimp/62 m 2 ), and their host anemone C. gigantea was significantly larger (141.3 cm 2 TCSA, IQR: 73-241 cm 2 ) than the primary anemone host of A. pedersoni, B. annulata (46.7 cm 2 TCSA, IQR: 27-64 cm 2 ; U = 34.5, p < 0.001). In total, we recorded 83 h of video at P. yucatanicus (52.48 h) and A. pedersoni (30.98 h) cleaning stations. We observed more than 300 visits and 80 cleans at both types of cleaner stations (table 1). Visitation rate (h −1 ) and cleaning rate (h −1 ) were significantly greater at stations hosting A. pedersoni (A. pedersoni: visitation rate = 5.3 h −1 , IQR: 1.6-11.6 h −1 ; cleaning rate = 1. figure 2). There was no significant correlation between anemone size and visitation rate to P. yucatanicus cleaning stations (r 2 = 0.06; p = 0.24; figure 3a), but there was a significant positive correlation between anemone host size and visitation rate to A. pedersoni cleaning stations (r 2 = 0.51, p < 0.01; figure 3b). Additionally, we recovered a significant positive correlation between cleaner group size and visitation rate in A. pedersoni (r 2 = 0.34, p < 0.05; figure 4). While these trends for A. pedersoni were statistically significant, we suggest caution in over-interpreting these trends as sample sizes were small (n = 12 stations). Cleaning interactions were recorded at fewer than half (7/23) of the stations containing P. yucatanicus and at all but one (11/12) station hosting A. pedersoni (χ 2 = 3.42, p = 0.06). Interaction (clean) length and cumulative time (s) spent cleaning were also significantly greater at stations hosting A. pedersoni (table 1; interaction length: U = 285, p < 0.0001; cumulative cleaning time: U = 30.5, p < 0.0001). Of the 19 total cleans observed at stations hosting P. yucatanicus, 50% were shorter than 10 s in total duration. Both P. yucatanicus and A. pedersoni were observed cheating client reef fishes. We observed cheating in 16         interaction. The proportion of flinches that appeared to terminate cleans at P. yucatanicus stations was not significantly greater than at A. pedersoni stations, but only marginally so (Fisher's exact test, p = 0.052). The numbers of reef fish families and species that visited P. yucatanicus and A. pedersoni stations were similar (table 1 and tables 2

Discussion
We provide the first field-based evidence that reef fish in the tropical western Atlantic visit P. yucatanicus with some regularity, and that in combination with its unambiguous signalling (electronic supplementary material, video S1), this species engages in true symbiotic cleaning interactions with reef fishes. Our video observations demonstrate that P. yucatanicus wave their long white antennae, signalling to client reef fishes that they are available to clean (electronic supplementary material, video S1). This behavioural signal is the same signal used by A. pedersoni, is akin to other overt signalling behaviours that true cleaner shrimp use to engage client reef fish (e.g. [5,25,28,52]) and meets the newly revised definition of a cleaning symbiosis proposed by Vaughan et al. [11], which requires communicative behaviours for both cleaner and client. Combined with reciprocal poses from clients and engagement in cleaning activity, our data confirm the designation of P. yucatanicus as a true cleaner shrimp species. However, we only deployed cameras during mid-morning and only for a limited period of time, and so we lack the temporal context required to determine whether P. yucatanicus is a dedicated or facultative cleaner shrimp. Ancylomenes pedersoni, a dedicated cleaner, shows no temporal patterns of cleaning services throughout the day in Honduras, receiving regular visits and engaging in cleaning bouts from dawn to dusk [25]. Other dedicated cleaners (e.g. gobies) show increased cleaning activity at dawn, coinciding with gnathiid isopod abundance (e.g. [53,54]), one of their preferred parasitic food sources [44]. It is unclear at this point how cleaning varies throughout the day for P. yucatanicus, and it is certainly possible that its cleaning behaviour may be more temporally restricted. Nocturnal activity should also not be discounted as cleaner shrimp have been demonstrated to forage and clean at night [55,56]. To date, no nocturnal cleaning activity has been documented or observed in the Caribbean for cleaner gobies, juvenile wrasse or A. pedersoni [27].
Previous work has shown considerable overlap in clientele between A. pedersoni and cleaner gobies in the Caribbean [25]. We demonstrate similar overlap in clientele between A. pedersoni and P. yucatanicus. Nine families of fish had members who visited both cleaner species at our study site in Curacao, in contrast to more than 20 families whose members have been documented visiting A. pedersoni stations throughout its range [25,[28][29][30]. Although the relatively low client diversity here does limit our ability to discern the extent of the community of western Atlantic reef fish that visit P. yucatanicus stations, cleaning rates observed in A. pedersoni largely mirror those recovered in Honduras [25] and US Virgin Islands [28]. Thus, the rate at which client reef fish visit A. pedersoni appears to be independent of region, reef site and local reef fish diversity, giving us confidence that our video data set accurately captures the scope of the usage of cleaner shrimp stations in Curacao. A follow-up study in a higher diversity setting would illuminate whether any client families specialize on specific shrimp species, but to date no client families have been documented as cleaner-shrimp specialists. If P. yucatanicus service the same community of reef fishes as A. pedersoni and cleaner gobies, as our data suggest they may, understanding how these (apparently) ecologically redundant cleaner species co-occur, compete for clients, and partition resources will be an important avenue for future research. One possibility is that different cleaners specialize on different parasites. Gobies clean more actively during dawn and show reduced cleaning activity as gnathiid isopod abundance naturally decreases throughout the day [53,54,57,58]; by contrast, A. pedersoni cleans regularly from dawn to dusk, which may suggest this species targets parasites that are more permanently attached (e.g. flatworms) [25]. More systematic investigation is needed into the ecological role of P. yucatanicus, as well as all co-occurring cleaner species, to determine whether specialization allows for overlapping clientele, or whether there are sufficient resources (i.e. parasite loads) to allow for such redundancy in service provisioning.
While there is similarity in client diversity, the differences in visitation and cleaning rates between both species are striking. While some of these differences could be attributed to differences in host use, host size is not an appropriate explanation here. Anemone size has a positive effect on cleaning activity for A. pedersoni hosted by B. annulata, but because C. gigantea is the same size or larger than B. annulata, host size is not a general explanation for the differences in visitation or cleaning rate between shrimp species. Whether host species identity impacts cleaning activity is unclear and difficult to parse in our data because of the non-random association we see between host anemone and shrimp. However, fish may use sea anemones as visual cues to locate A. pedersoni shrimp, so anemone size may matter on some reefs [29]. This may be especially true for B. annulata which has thinner and more transparent tentacles than C. gigantea, and would explain why we observed a significant increase in fish cleaning rate with body size in B. annulata but not C. gigantea.
Because we find regular client visitations to P. yucatanicus stations that result in cleaning bouts between shrimp and reef fish, we find no compelling evidence that P. yucatanicus is a cleaner mimic. The regularity of visitation to P. yucatanicus stations is much greater than we would expect given the species' apparent apprehension to clean (electronic supplementary material, video S1). In a typical cleaner system, high-quality service provisioning by the cleaner results in repeat visits by the client (e.g. [59]). However, visitation to P. yucatanicus stations was not rare, and the large discrepancy between visitation and cleaning frequency is somewhat consistent with the behavioural expectations we established surrounding visual protective mimicry. In our view, however, the physical engagement between P. yucatanicus and posing reef fish negates the interpretation that this is a visually protective mimic species. We further reject the hypothesis of P. yucatanicus as an aggressive mimic. While the duration of cleaning bouts was generally short some exceeded 1 min in total length, and not all cheating bouts resulted in a termination of the interaction. There is also no indication from our data that P. yucatanicus showed a significantly greater propensity to cheat than A. pedersoni, as would be predicted if it were an aggressive mimic, and cheating did not occur during every interaction. Instead, we demonstrate that reef fishes in the Caribbean appear to tolerate regular cheating occurrences by both species. Our data set is the first to quantify the frequency with which cheating occurs in a cleaner shrimp in the tropical western Atlantic. While cheating occurs frequently in both P. yucatanicus and A. pedersoni, no aggressive responses were observed by the clients towards the cleaners beyond jolting, and not all cheating resulted in the termination of the interaction (electronic supplementary material, video S2). Our findings are consistent with previous studies of cleaner gobies (e.g. [60]) and previous observations of A. pedersoni [28] that the cleaning symbioses in the Caribbean may be a system without punishment.
In conclusion, while P. yucatanicus appears to be a true cleaner shrimp, the effectiveness of this species as a cleaner and its ecological role (dedicated or facultative cleaner) remains in question. Less than 15% of all client visits resulted in an actual cleaning interaction, and the brevity of these cleans (table 1), in comparison to those performed by A. pedersoni on the same reef, probably resulted in minimal parasite removal. These observations and the findings by McCammon et al. [4] that P. yucatanicus did not significantly reduce monogenean parasite loads on surgeonfish in a semi-natural setting paint a clearer picture of the ecological importance of P. yucatanicus. Additional study will be necessary to discern whether P. yucatanicus specializes on different ectoparasites than A. pedersoni or whether it varies in its cleaning activities across day or season, but we are sceptical that this species plays a role in ecosystem health that equals that of A. pedersoni.