Mutualism promotes site selection in a large marine planktivore

Abstract Mutualism is a form of symbiosis whereby both parties benefit from the relationship. An example is cleaning symbiosis, which has been observed in terrestrial and marine environments. The most recognized form of marine cleaning symbiosis is that of cleaner fishes and their clients. Cleaner species set up cleaning stations on the reef, and other species seek out their services. However, it is not well understood how the presence of cleaning stations influence movements of large highly mobile species. We examined the role of cleaning stations as a driver of movement and habitat use in a mobile client species. Here, we used a combination of passive acoustic telemetry and in‐water surveys to investigate cleaning station attendance by the reef manta ray Mobula alfredi. We employed a novel approach in the form of a fine‐scale acoustic receiver array set up around a known cleaning area and tagged 42 rays. Within the array, we mapped structural features, surveyed the distribution of cleaner wrasse, and observed the habitat use of the rays. We found manta ray space use was significantly associated with blue‐streak cleaner wrasse Labroides dimidiatus distribution and hard coral substrate. Cleaning interactions dominated their habitat use at this site, taking precedence over other life history traits such as feeding and courtship. This study has demonstrated that cleaning symbiosis is a driver for highly mobile, and otherwise pelagic, species to visit inshore reef environments. We suggest that targeted and long‐term use of specific cleaning stations reflects manta rays having a long‐term memory and cognitive map of some shallow reef environments where quality cleaning is provided. We hypothesize that animals prefer cleaning sites in proximity to productive foraging regions.


| INTRODUC TI ON
Mutualism is the exchange of goods and services between organisms that provides a net benefit to those involved. A classic example is pollination, in which an animal vector receives food in the form of pollen or nectar, in exchange for fertilizing a plant's ovules (Cushman & Beattie, 1991). Another is cleaning symbiosis, whereby a client species has ectoparasites removed by a host cleaner species (Limbaugh, 1961). On land, this symbiosis is commonly observed in bird species such as oxpeckers Buphagus spp. removing ticks and blood-sucking flies from ungulates (Sazima, 2011). Cleaning symbiosis is also found, though less commonly, in other taxa including small mammals: an example is the banded mongoose Mungos mungo removing ticks from common warthogs Phacochoerus africanus (Sazima, 2010). In the ocean, over 100 marine fishes and numerous invertebrate species act as cleaners to a wide range of taxa including cephalopod mollusks, fishes, mammals, and reptiles (Côté, 2000). Host cleaner species in marine systems are generally more site attached than their terrestrial counterparts, setting up "cleaning stations" that clients seek out and visit. However, to date there is little understanding of how this mutualism promotes site selection in large-bodied, vagile marine species.
Many species of marine megafauna have extensive home ranges, moving 100s of kilometers in search of food or undergoing reproductive migrations (e.g., leatherback turtles Dermochelys coriacea, sperm whales Physeter macrocephalus, and lemon sharks Negaprion brevirostris; (Christal & Whitehead, 1997, Houghton et al., 2006, Chapman et al., 2009). Yet large migratory species also seek out the services of site-attached cleaners (e.g., oceanic sunfish Mola mola, pelagic thresher sharks Alopias pelagicus, whale sharks Rhincodon typus, and manta rays Mobula birostris and M. alfredi; (Konow et al., 2006, O'Shea et al., 2010, Oliver et al., 2011, Araujo et al., 2020, Murie et al., 2020. Cleaning interactions typically involve the cleaner removing ectoparasites from visiting or resident clients, but cleaners may also feed on host mucus and skin, particularly at wound sites (Grutter, 1999). The ecological importance of this symbiosis in terms of body maintenance, especially for mobile client species, has been demonstrated on tropical coral reefs via experimental exclusion of the blue-streak cleaner wrasse, Labroides dimidiatus. For client species, this has resulted in decreased diversity and abundance, higher rates of fungal infection, smaller body size, and poorer health (Bshary, 2003;Bshary et al., 2007;Grutter et al., 2003;Waldie et al., 2011). Mobile client species may therefore benefit from their choice to visit reefs based on the presence of cleaner fish. In turn, reef-associated cleaners benefit from large, mobile species bringing a food source that originates from off the reef, suggesting that this symbiosis is indeed mutualistic.
The preference of mobile species for particular sites indicates they have a spatial memory of the sites and the service they have received. The species of cleaner fish at a site may influence the quality of the cleaning station, and large mobile marine species have multiple options when it comes to habitat choice. Mobile clients would likely opt for cleaning stations where they receive quality service (i.e., parasite removal with few or no adverse events, such as biting from the cleaners; (Bshary & Schäffer, 2002)) and would be less likely to return to a site where they were previously not attended to promptly (Bshary & Grutter, 2002). Perhaps then it is not surprising that clients with multiple choices of cleaning stations are given better service than clients with reduced options, suggesting that cleaners can distinguish between resident and visiting clients (Adam, 2010). A key determinant of a mobile client's attendance at a particular cleaning station site may be the memory of quality cleaning service and the animal's spatial cognition. How animals move in relation to learning and memory remains a key question in megafauna studies (Hays et al., 2016), and the role of spatial cognition in relation to fine-scale habitat use is not well understood.
Many large, mobile elasmobranch species have been observed attending cleaning stations, including spotted eagle rays Aetobatus narinari, pelagic thresher sharks, silky sharks Carcharhinus falciformis, Galapagos sharks C. galapagensis, bull sharks C. leucas, demersal lemon sharks Negaprion acutidens, whale sharks, and scalloped hammerheads Sphyrna lewini (Keyes, 1982;Oliver et al., 2011;Quimbayo et al., 2016). Ectoparasite loads may drive these visits, as individuals with high parasite loads frequent cleaning stations more regularly than individuals with low parasite loads (Grutter, 1999). Ectoparasite loads are lower after cleaning interactions (Keyes, 1982), and cleaner fish spend longer foraging on body regions with more ectoparasites (Oliver et al., 2011). Larger elasmobranchs may also receive preferential treatment from the cleaner community (Keyes, 1982;Oliver et al., 2011). Larger fish with more parasites are inspected more often and for longer than smaller fish with fewer parasites (Grutter, 1995).
Further, facultative cleaners favor interactions with planktivore clients over piscivores, which is likely to be related to a lower risk of being eaten by the client fish (Francini-Filho & Sazima, 2008).
However, drivers of site selection by large mobile clients at cleaning station habitats have not previously been investigated.
The reef manta ray is a large, mobile planktivorous species that demonstrates site affinity to reef environments, including cleaning stations ( Figure 1). This affinity has been documented at aggregation sites in the Maldives (Stevens, 2016), eastern Australia (Couturier F I G U R E 1 Reef manta ray Mobula alfredi attending a cleaning station at Lady Elliot Island, Australia et al., 2018), Indonesia (Germanov et al., 2019), the Seychelles , and Mozambique (Venables et al., 2020). These studies have documented that manta rays visit cleaning stations, but the methods used in most studies have limitations for determining behavior and fine-scale spatial use. For example, while observations by SCUBA or free divers or via remote underwater cameras can provide fine-scale information, they are limited to short observational windows. By contrast, aquatic telemetry approaches such as acoustic or satellite tagging can facilitate continuous, long-term detections, but typically lack spatial resolution. However, combining these approaches, and enhancing the fine-scale tracking of animal movements, could elucidate the site preferences and drivers of mobile client's visitation to particular sites.
Here, we investigated the role of mutualism in determining site selection in vagile marine megafauna. Our aim was to determine the role of cleaning stations as a driver of movement and habitat use in a mobile client species. Through a combination of in-water observations and a novel application of fine-scale passive acoustic tracking, we show how high-accuracy tracking can elucidate the memory of particular shallow reef habitats and space use in a largely pelagic species and provide a mechanism for nutrient exchange between reef and open ocean environments.

| Study site
Mobula alfredi is found in tropical and subtropical waters of the Indo-Pacific and Indian Ocean. In Australia, it is found in coastal waters north of ~30°S (Figure 2a; Armstrong et al., 2020), with the largest known aggregation on the east coast around Lady Elliot Island (LEI, Figure 2b; Couturier et al., 2014). The peak of the M. alfredi aggregation at LEI is in winter, and during summer many individuals migrate south, with North Stradbroke Island (Figure 2a) a seasonal aggregation site (Couturier et al., 2011). At LEI, previous research identified a high-use area on the western side of the island where M. alfredi individuals cruise, feed, court, and are cleaned (Couturier et al., 2018;Jaine et al., 2012). This area has a series of coral reef features 8-15 m deep ( Figure 2c). Receivers were ~76 m apart in a grid formation placed at GPSverified locations. Each receiver was paired with an acoustic transmitter or sync tag, attached 1 m above the VR2W, which emitted a unique coded pulse-train every 500-700 s. These transmissions allowed for the relative position of each receiver within the array to be determined throughout the study, to provide an estimate of the location error associated with the calculated positions of tagged animals. To assess the magnitude of the positioning error, two sentinel transmitters and temperature loggers were placed at GPS-verified locations within the array for the duration of the study. The VPS facilitates precise locational tracking of tagged animals in and around the acoustic array (Espinoza et al., 2011;Roy et al., 2014), and receivers in the LEI array were set up conservatively (close together) to ensure stable performance over a 24-hr cycle.  (Table A1). Each transmitter was attached to a Domeier umbrella-dart tag head with a 10 cm shrink-wrapped braided wire tether and was inserted into the dorsal musculature of a pectoral fin of a free-swimming ray using a modified Hawaiian sling spear. Antifoul was not applied due to the relatively short retention times of external Domeier umbrella-dart head tags on manta rays (median detection time of 121 days; Couturier et al. 2018). Prior to tagging, individual M. alfredi were photographed for subsequent identification by comparing the image with those in an existing photo-ID database (Armstrong et al., 2019). Sex was assessed based on the presence (male) or absence (female) of claspers (Marshall et al., 2011). Maturity status was determined by the clasper size (in males), and pregnancy and/or presence of mating scars (in females; Marshall & Bennett, 2010). Animal size was estimated visually to the nearest 0.5 m using stationary objects for scale, and all females were assigned as adult if their disk width was ≥3.5 m . Tagging commenced in February 2017 at North Stradbroke Island (n = 10), and the remaining tags were deployed at LEI between February 2017 and June 2018 (n = 32).

| Processing acoustic data
The position of a tagged M. alfredi within the array was calculated were used as a control to reduce the potential for making erroneous inferences regarding animal behavior (Payne et al., 2010) and to rule out the possibility that animal position estimates were heavily influenced by background reef noise. acoustic array, we conducted three surveys on SCUBA using the aforementioned habitat map to provide accurate positions for spatial analysis. Counts of L. dimidiatus were obtained and cross-checked by two divers on SCUBA. Surveys were conducted in June to coincide with the observed peak of M. alfredi visits to the study site (Couturier et al., 2011). The focus of the distribution surveys was the obligate cleaner fish, L. dimidiatus, as this species is found across all habitat zones (Green, 1996). Labroides dimidiatus establishes cleaning stations at fixed locations (Potts, 1973), whereas the other cleaner species observed in the current study are facultative cleaners, and less site attached.

| Statistical analyses
To L. dimidiatus, and substrate polygons were transformed into spatial data using a 5 × 5 m grid in the R packages "raster" (Hijmans, 2020) and "sf" (Pebesma, 2018).  This ensures that the cyclical nature of these predictors is captured, while guaranteeing that the response values predicted at the extremes of the predictor range are the same (e.g., the same prediction for Count at times of 0 and 24 hr). Wind Speed was smoothed using a natural spline in the R package "splines" (R Core Team, 2019).
Explained deviance was calculated in R using delta values in the package "MuMIn" (Barton, 2009). Final models were selected based on AIC values, and significance of variables was taken at p < .05. To visualize the generalized linear mixed effects models, we present contrast plots using the response scale in the R package "visreg" (Breheny & Burchett, 2017). Model output with confidence limits on the response scale is provided in the Results, and output on the log-link scale with residuals is available in the appendix ( Figure A3).

| Acoustic tracking data
Between

F I G U R E 4
Eight examples of fine-scale tracking of Mobula alfredi in the acoustic array off Lady Elliot Island. Each color represents an individual manta ray, and each plot a separate visitation event. The colored triangle denotes the start of the track, and the square is where the track ends. Black circles represent the locations of the acoustic receivers, and the black outlines are the mapped reef structures

| Drivers of Mobula alfredi visitation
The  (Table 1). Manta rays visited cleaning stations for longer periods during daylight hours, when winds were from a SSW direction and the wind speed was low (Total explained deviance = 14.2%, Fixed effects = 5.1%, and Random effects = 8.9%).
Analysis of sentinel transmitter positions revealed that the minimal performance of the array decreased slightly at night. The correction to the total hourly counts of animal positions using the standardized positioning frequency of the sentinel transmitters did not alter the overall pattern of attendance to the site from tagged manta rays ( Figure A6). We can thus be confident that the patterns observed are driven by the physics and biology, rather than an artifact of under-performance in the acoustic array due to reef noise.

| D ISCUSS I ON
We found that mutualism-in the form of cleaning symbiosis-can promote site selection in a large-bodied, mobile marine species. While the amount of time spent in the vicinity of cleaning stations may be small in absolute terms (tens of minutes over a 24-hr period), it appears to be a key component of the daily time-budget for manta rays. Similarly, many terrestrial species spend time each day grooming to remove parasites (e.g., chimpanzees (Foster et al., 2009;Lehmann & Boesch, 2008), with an important distinction being that in these cases animals do not need to travel to specific locations to undertake such behavior.
Quality cleaning habitat is an important driver of pelagic species visitation to inshore reefs. Here we established that the number of individual L. dimidiatus present at a particular cleaning station may influence manta ray site preference, demonstrating the disproportionate effect that a small and sparsely distributed species can have on coral reef communities (Waldie et al., 2011). Labroides dimidiatus is ubiquitous in tropical reef systems globally (Green, 1996)

D4
Sex + TimeofDay(k = 2)+WindDirection(k = 2)+WindSpeed(df = 2) 11 -14.4 (5.9) 5,638.5 studies on the spatiotemporal patterns of abundance and behaviors (e.g., foraging and cruising) of animals is key to understanding their ecological roles. For the reef environment, it may be that visits to cleaning stations from animals that spend considerable time in the open ocean (e.g., pelagic thresher sharks (Oliver et al., 2011) and manta rays (Murie et al., 2020)), provides a mechanism for nutrient exchange between these environments. This is similar to the broader-scale example of large whales (baleen and sperm whales) migrating from high latitudes translocating nutrients to oligotrophic tropical systems (Roman et al., 2014). Thus, understanding the movement ecology and site selection of threatened species is not only crucial for informing effective management strategies, but also for gaining insights into their role in ecosystem function.
Many individuals repeatedly visited the same localized sites across many weeks, and we propose that manta rays likely locate these cleaning stations using conspicuous landmarks that they remember. Relatively little is known about effects of learning and memory on the movement patterns of marine megafauna (Hays et al., 2016). An unresolved question in marine megafauna movement ecology is how learning and memory or innate behaviors drive animal movements. It is likely that manta rays have a cognitive map of particular reef areas, akin to how animals with distinct home ranges know their environment intimately (Harten et al., 2020). In common with other taxa such as sea turtles, that alternate between oceanic and coastal areas, manta rays likely use coarse-scale navigational cues in the open ocean, and precisely orientated movement in coastal areas . Manta rays visit the same reef systems across many years, as demonstrated by photo-ID records Harris et al., 2020), and such repeated site use is similar to that seen in sea turtles and many bird species that maintain strong site fidelity to particular areas interspersed with longdistance migration (Alerstam et al., 2006;Armstrong et al., 2019;Shimada et al., 2020). These observations imply that many migratory taxa, including manta rays, have a long-term memory of particular focal sites.
We found that M. alfredi prefer hard coral structure, rather than soft coral cover or continuous coral ridge substrate, and that cleaner wrasse density was also more associated with this type of substrate.
Cleaner species often use prominent coral heads or outcrops to set up their cleaning stations (Côté et al., 1998), and choosing a conspicuous location is likely to be beneficial to them for attracting clients. But for large, mobile species such as M. alfredi, these structures also provide suitable habitat to allow maneuverability and facilitate cleaning interactions. As for many other pelagic elasmobranch species, M. alfredi is a ram ventilator and has to swim continuously to irrigate its gills for uptake of oxygen. They are unable to rest on the bottom to facilitate cleaning interactions, as do some demersal elasmobranch species (Keyes, 1982;Sazima & Moura, 2000). Hard corals are particularly susceptible to the impacts of climate change, and the potential for coral reefs to recover from multiple stressors is declining (Hughes et al., 2018). Given the preference for hard coral structures, climate change could present a threat to the habitat of numerous cleaner species. Loss of habitat for cleaner species could have downstream consequences on the movements and site selection of large mobile clients like manta rays.
Wind conditions and tidal cycles are known to influence the movements of large animals, and this study also confirmed the importance of these environmental variables. There were contrasting patterns in how Wind Direction influenced reef manta ray Counts and the Duration of their visits, and this may be explained by the location of the study site on the western side of the island. Previous work has suggested manta rays favor this side of the island due to the shelter provided from prevailing winds (Couturier et al., 2018), and it may be that south to south-westerly winds are more favorable for longer cleaning station attendance, but that the protection afforded on this side of the island means manta rays may be detected during other wind regimes as well. We found that Wind Speed influenced the Duration of visits here, but did not impact their detection rate, supporting that manta rays are still present during less favorable conditions but for shorter periods of time. Foraging opportunities may present a reason for the Tidal influence of manta ray detections at the site. Prior research has shown zooplankton concentrations in the vicinity of the current study site are found to peak prior to low tide, and manta rays are more commonly observed feeding during this tidal phase (Armstrong et al., 2016). It may be that manta rays attend the cleaning station for longer periods either side of this foraging opportunity. Tide and current movements have been implicated for cleaning behavior at other locations (Murie et al., 2020;O'Shea et al., 2010;Rohner et al., 2013), as moderate currents are favorable to a manta ray's ability to hold station and facilitate cleaning. This suggests that the preference of M. alfredi for prominent hard coral structures in the current study may be related both to suitable habitat for the cleaner wrasse and the hydrodynamics of the location that facilitates cleaning interactions.
We found a clear diurnal signal for M. alfredi attendance within the study region. First arrival occurred in the morning, after sunrise, and individual visits gradually declined throughout the day. Early in the day was also when individuals were more likely to spend longer periods of time in the cleaning station area. Similar findings have been reported for manta rays in other regions (Venables et al., 2020) and for other elasmobranchs (Oliver et al., 2011). Diurnal visitation likely reflects the behavior of the cleaner fish, since L. dimidiatus individuals are inactive at night, and do not return to their cleaning station habitats until after dawn (Potts, 1973). However, it is also likely a product of the behavior of the manta rays themselves, as they move offshore to forage at night. Satellite tracking has revealed night-time diving behavior in manta rays (Braun et al., 2014;Stewart et al., 2016), and investigations using stable isotope analysis have suggested manta rays source their food from deep, benthic or epipelagic environments (Burgess et al., 2016;Couturier et al., 2013;. Planktivores from a range of taxa exhibit diurnal patterns in foraging behavior (Brierley, 2014;Hays, 2003), to take advantage of diel vertical migrating zooplankton that come into shallower waters at night.
We showed the utility of automated high-accuracy acoustic tracking, which contrasts with historic active acoustic tracking that often has low accuracy, is limited in time and is labor intensive (Nelson et al., 1997). Levels of location accuracy we achieved in a reef environment (within a few meters) are similar to that recorded by others using the VPS approach in freshwater (Espinoza et al., 2011;Roy et al., 2014). By comparison, modern satellite tracking approaches such as Fastloc-GPS, where locations are typically within a few 10s of meters of the true location (Dujon et al., 2014;Thomson et al., 2017), can provide continuous broad-scale tracking at the cost of such precision. These modern approaches are transforming our understanding of the patterns of small-scale space use for a range of marine species. For manta rays, a blended use of acoustic arrays and satellite tracking would provide a comprehensive understanding of space use over a range of spatial scales, and further clarify links between cleaning stations and adjacent feeding grounds .
Combining tagging methods could also offer redundancy for tag failure or loss. Almost 20% of the tags that we deployed were never detected after their initial deployment, likely because of animal migration, tag shedding or tag failure. Understanding why transmitters stop relaying data is important to help drive improvements to tag design and deployment (Hays et al., 2007). Three of the eight tagged manta rays that were not detected by the VPS array were identified at the site by photo-ID within the study period, showing that for these individuals at least, the issue was tag shedding or failure (tag attachment was not confirmed via photo-ID). Nevertheless, for 34 individuals (81% of those tagged), tracking revealed repeated visits to cleaning areas. This large sample size, when compared to many tracking studies from a recent review (Sequeira et al., 2019), suggests that fidelity to cleaning areas is a general feature of manta ray ecology.
We hypothesize from the findings of the current study, together with other recent work (Murie et al., 2020;Stevens, 2016), that preferred cleaning station sites are likely paired with rich feeding grounds nearby. The current study location is about seven kilometers from the shelf edge, and the mesoscale oceanographic feature of the Capricorn Eddy (Weeks et al., 2010). The productivity of the Capricorn Eddy is a result of increased frontal activity and upwelling, providing foraging opportunities to seabirds such as wedgetailed shearwaters Puffinus pacificus (McDuie et al., 2018), and has been associated with foraging of manta rays .
Cleaning station environments, where manta rays sometimes stay in close proximity for long periods (i.e., weeks to months), would need to be in the vicinity of places that fulfill the multiple biological and ecological functions of these animals. Therefore, the feeding-cleaning hypothesis-where mobile species select cleaning sites close to productive foraging opportunities-may also explain the habitat preferences of other large, mobile client species.

ACK N OWLED G M ENTS
We would like to acknowledge the support of Lady Elliot Island were not involved in the design of the study and collection, analysis, and interpretation of data, or in writing the manuscript.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

O PEN R E S E A RCH BA D G E S
This article has earned an Open Data Badge for making publicly available the digitally-shareable data necessary to reproduce the reported results. The data is available at https://doi.org/10.14264/ 2e79b25.