Classically, symbiosis refers to the close association of two different species living together, with organisms being involved as hosts or symbionts (de Bary 1879). A symbiotic relationship can have different forms (parasitism, mutualism and commensalism), but they are part of a broad continuum and these associations cannot always be arranged in adjacent drawers (Parmentier and Michel 2013). The black lip pearl oyster Pinctada margaritifera (Linnaeus, 1758) (Mollusca: Bivalvia: Pteriidae) is widely distributed in tropical Indo-West Pacific regions, living in coral reef areas (Gervis and Sims 1992, Southgate and Lucas 2008). This species occurs as large populations in many atolls of French Polynesia, where it is one of the most characteristic benthic bivalve molluscs due to its economic importance for the pearl farming industry (Salvat 2009). Some cases of symbiotic organisms living in association with P. margaritifera have been reported in the past, and these include both vertebrate and invertebrate symbionts.

Pearlfishes (Ophidiiformes: Carapidae) are eel-like fishes that mainly occur in shallow to moderately deep waters of tropical seas (Markle and Olney 1990). Within this family, several genera (Onuxodon spp., Carapus spp. and Encheliophis spp.) share a remarkable peculiarity: they are able to penetrate and live inside different invertebrate hosts such as echinoderms (holothurians, starfish) and bivalves (Fowler 1927, Tyler 1970, Trott and Trott 1972). Based on stomach content analysis (Trott 1970, Vanden Spiegel and Jangoux 1989), morphological descriptions of the buccal and pharyngeal jaw apparatus (Parmentier et al. 1999, 2000) and stable isotope analysis (Parmentier and Das 2004), some Carapus spp. and Encheliophis spp. living inside echinoderms have been considered commensal or parasite, depending on the species. Basically, commensal species use their host as a shelter and leave it for foraging whereas parasitic species are known to feed on the internal tissues of their host (Smith 1964, Trott 1970, Parmentier et al. 2000). Among this fish family, members of a third genus (Onuxodon) are also known to live inside bivalves, being located between the mantle and the shell (Fowler 1927, Trott 1970, Tyler 1970). Fowler’s pearlfish, Onuxodon fowleri (Smith, 1955) (Carapinae: Echiodontini), lives inside representatives of the pearl oyster P. margaritifera (Fowler 1927, Parmentier et al. 2000, Kéver et al. 2014). Onuxodon fowleri is considered a commensal species that uses its host as a shelter and leaves it to feed on small benthic preys such as annelids and small crustaceans (Trott 1981, Parmentier et al. 2000). According to scientific evidence, no apparent harm caused by the fish to its host has ever been reported. However, the exact nature of this host/symbiont association has not yet been experimentally demonstrated and additional data are needed to gain further insight into the type of symbiosis taking place.

Stable isotope analysis has become a powerful tool for tracing dietary sources by providing an integrated measure of the dietary components over a long period of time. This method clearly shows that the isotope ratios of a consumer are related to those of its food (DeNiro and Epstein 1978, 1981, Peterson and Fry 1987). Stable isotope analysis gives an average estimate of the dietary preferences of an organism that is less subject to temporal bias (Pinnegar and Polunin 1999), but it does not provide a detailed picture of the food ingested by this organism. For this purpose, stomach content analysis can be used as a complementary tool (Frédérich et al. 2009). The combination of the two methods has the advantage of compensating for the inaccuracy of each one (Frédérich et al. 2012). Interestingly, an approach that combines stomach content analysis and the use of stable isotope ratios of carbon (¹³C/¹²C) and nitrogen (¹⁵N/¹⁴N) has proved to be a valuable tool to get more information on the symbiotic relationship (i.e. commensal or parasite) between carapids and their hosts (Parmentier and Das 2004).

The present study aimed to establish the symbiotic relationship between the carapid fish O. fowleri and the pearl oyster P. margaritifera through an approach using stable isotope ratios of carbon (¹³C/¹²C) and nitrogen (¹⁵N/¹⁴N). This stable isotope approach was also complemented by the analysis of stomach contents in the carapid fish.

Over the two sampling campaigns, we found a total of 57 O. fowleri individuals sheltered inside the 209 P. margaritifera that were collected on the 13 reef pinnacles. Therefore, the overall ratio between the number of occupied hosts and the number of collected hosts was about 1:4, with an overall occupation rate of 27.3%. A more detailed listing of the occupation rates observed on the different reef pinnacles is presented in Table 1. All fish were encountered in pearl oysters collected in front of Arikitamori Pass. The highest numbers of fish (occupation rate ≥50 %) were observed inside bivalves collected close to the Pass, whereas very few fish (occupation rate <15 %) were identified in pearl oysters collected at some of the reef pinnacles located further away from the Pass (Fig. 1). Moreover, no fish (occupation rate 0 %) were observed inside bivalves collected on two reef pinnacles: one of these pinnacles was located far from the Pass and the other one was not in its alignment (Fig. 1).

In addition, a pair of palaemonid shrimp (C. meleagrinae) was observed in almost each of the collected pearl oysters (pers. obs.), implying that the co-occurrence of both symbionts within the same host was frequent.

All O. fowleri individuals were found in host bivalves collected in the axis of Arikitamori Pass. Overall, the percentages of pearl oysters being occupied by fish were very low, or zero, at some reef pinnacles located far from the Pass, while the highest numbers of fish were observed inside bivalves collected close to the Pass (Fig. 1, Table 1). Moreover, no fish were observed among bivalves collected at the only reef pinnacle that was not in the alignment of the Pass (Fig. 1). These differences in the occupation rates observed at the different locations of capture might be explained by the way of life of carapids. Like most coral reef fishes, carapids have a complex life history divided into two stages: a dispersive pelagic larval stage followed by sedentary demersal juvenile and adult stages associated with the coral reef environment (Leis 1991, Leis and McCormick 2002). At the end of the pelagic stage, larvae settle on the patch reef within the lagoon and rapidly try to enter a benthic host (Smith 1964, Smith et al. 1981, Colleye et al. 2008). This behaviour appears to be essential for the growth and survival of carapids (Parmentier et al. 2004a,b, Parmentier 2016). Therefore, it is likely that the greater amount of fish found inside pearl oysters collected close to Arikitamori Pass results from the fact that O. fowleri directly seek to enter a bivalve host shortly after settlement.

A large proportion (60%) of the stomach contents were empty and very little prey material was found in the non-empty stomachs. This high percentage of empty stomachs could have been misleading: the remains of host tissues, if any, could not have been detected during stomach content analysis since these soft tissues would have been digested very fast. Nonetheless, the way of life of symbiotic carapids would not require a great amount of energy since they are quite inactive inside their host. This proportion of empty stomachs might simply reflect the infrequency of feeding due to the low metabolism of species that spend a great part of their adult life within their host (Parmentier et al. 2002). In C. bermudensis living inside holothurians, the periodicity of feeding ranges from 15 to 24 days on average but it can last up to 60 days (Smith et al. 1981). In addition, the high percentage of empty stomachs observed in O. fowleri might be related to the time of sampling (10:00 am – 3:00 pm). Recently, Kéver et al. (2014) reported that the majority of sounds produced by O. fowleri in the field were recorded between 5:00 PM and 12:00 AM, which implies a nocturnal activity in this species. Assuming that O. fowleri forages for food mainly at night, it is thus likely that most of the stomach contents were empty because they were opened during daytime.

Typically, the δ¹⁵N values increase by approximately 2 to 5‰ with each trophic transfer between a consumer and its diet, while the δ¹³C values of an animal are close to that of its diet or slightly enriched by 1‰ (DeNiro and Epstein 1978, 1981, Vander Zanden and Hulshof 1998). Instead of being used as a reliable indicator of trophic level, δ¹³C values are generally used to indicate the relative contribution of different primary food sources (Parmentier and Das 2004, Frédérich et al. 2009, Cabanellas-Reboredo et al. 2010). Regarding carapid fishes, Parmentier and Das (2004) observed that tissues of commensal species such as C. homei and C. boraborensis were strongly ¹³C-depleted and ¹⁵N-enriched compared with the respiratory trees and gonads of their holothurian host B. argus (e.g. the mean decrease in δ¹³C ranged from 4.5‰ to 9‰, and the mean increase in δ¹⁵N ranged from 6‰ to 9‰, depending on host tissues). A similar ¹³C depletion and ¹⁵N enrichment was observed for the muscles of the commensal species C. mourlani compared with the gonads of its starfish host C. novaeguineae (e.g. the mean δ¹³C decreased by about 8‰ and the mean δ¹⁵N increased by about 7‰; Parmentier and Das 2004). On the other hand, Parmentier and Das (2004) noticed a mean increase in δ¹³C of about 1.5‰ and a mean increase in δ¹⁵N of 2.5‰ between B. argus gonads and Encheliophis gracilis muscles. These isotopic values indicated that E. gracilis could feed on its host gonads (Parmentier and Das 2004). All these observations were also confirmed by stomach content analysis and morphological characteristics (Smith 1964, Trott 1970, Parmentier et al. 1998), which supported the commensal and parasitic relationship attributed to these carapid species. In the present study, a mean increase in δ¹⁵N ranging from 4‰ to 5‰ was observed between the tissues of P. margaritifera and O. fowleri muscles. Moreover, fish muscles were significantly ¹³C-depleted compared with their host tissues. From an ecological point of view, the ¹³C depletion of O. fowleri muscles compared with some of the bivalve tissues seems to indicate that these are not the main source of food for the fish. Given that the bivalve gonads showed the same δ¹³C values as the fish muscles (Table 2), they should also be excluded from the fish diet. We also found that the δ¹³C values of O. fowleri did not match the isotopic composition of small benthic invertebrates (zoobenthos; see Fig. 2, Table 2) collected in the vicinity of the pearl oysters. Similarly, Parmentier and Das (2004) observed that adults of commensal Carapus species are ¹³C-depleted compared with the lagoon benthic invertebrates. Stable isotope ratios of carbon are known to be typically higher in species from coastal or benthic food webs than those from offshore food webs (Guo et al. 2002, Frédérich et al. 2009, 2012). It is thus likely that the δ¹³C values displayed by O. fowleri reflect a diet including pelagic preys or settling larvae entering the lagoon, which might also explain why most of the O. fowleri specimens were observed inside pearl oysters collected close to Arikitamori Pass. Regarding the δ¹⁵N difference between the carapid fish and its bivalve host, it seems very unlikely that O. fowleri specimens might feed on their host tissues, considering their strong ¹⁵N enrichment compared with P. margaritifera. Parmentier and Das (2004) measured a mean increase in δ¹⁵N of 2.5‰ between B. argus gonads and the parasite carapid E. gracilis, which is about two times less than the ¹⁵N enrichment observed between P. margaritifera tissues and O. fowleri muscles (Fig. 2). In a similar coral reef ecosystem, it is also interesting to note that fish considered as pelagic and benthic feeders showed an enrichment ranging from 2‰ to 3‰ in ¹⁵N relative to their assimilated food (Frédérich et al. 2009, 2012, Wyatt et al. 2010). In this context, O. fowleri would occupy two trophic levels higher than its bivalve host (Fig. 2).

Our isotopic analysis showed that δ¹⁵N and δ¹³C values increased between the bivalve tissues and the palaemonid shrimp C. meleagrinae (Fig. 2, Table 2). Due to the mean increase in both δ¹³C and δ¹⁵N, it appears that the shrimp occupies a higher trophic level than its host. More interestingly, C. meleagrinae showed a mean increase in δ¹³C of 1‰ and a mean δ¹⁵N enrichment of 3.0‰ compared with its host gonads. Considering these isotopic compositions, it cannot be totally ruled out that the shrimp might feed on its host gonads. From the evolutionary point of view, this assumption could explain why the palaemonid shrimp became morphologically adapted to living inside its host and why it adopted a sedentary way of life (Bruce 1976). Feeding on its host gonads would provide the shrimp with easy access to a food source rich in lipids, especially during the sexual maturation of the bivalve (Vahirua-Lechat et al. 2008). Moreover, it is not uncommon to encounter a crustacean guest parasitizing a bivalve host in nature. Although the association had regularly been considered commensalism, it was reported that the crab Zaops ostreus was parasitic on the American oyster Ostrea virginica (Stauber 1945). Likewise, the pea crab P. pisum is known to cause stress and lesions to its host bivalve, the common mussel Mytilus edulis (Bierbaum and Ferson 1986, Haines et al. 1994). At this point, further analysis of the shrimp diet using stomach contents should provide additional data in order to confirm this parasitical behaviour.

Finally, it is interesting to note that both symbionts may co-occur within the same host bivalve, which suggests a potential trophic competition between the two guests. Onuxodon fowleri seems to occupy a higher trophic level by being significantly ¹⁵N-enriched compared with C. meleagrinae (Table 2), but the carapid also showed a mean decrease in δ¹³C of 1‰ compared with the palaemonid shrimp (Table 2). As a result, these isotopic compositions indicated that the shrimp should not be part of the fish diet, a finding which was also supported by the fact that no fragments of cuticular remains were found in the stomach contents of O. fowleri.

Our results provide new data on the symbiotic relationship between Fowler’s pearlfish and the black lip pearl oyster. On the basis of the isotopic compositions measured in the present study, the commensal relationship usually attributed to P. margaritifera and its guest O. fowleri is supported. The carapid fish seems indeed to use its bivalve host as a shelter. In addition, our δ¹³C and δ¹⁵N measurements suggest that the palaemonid shrimp C. meleagrinae might feed on the bivalve tissues, especially considering the enrichment in δ¹³C and δ¹⁵N values compared with its host gonads. Ultimately, further isotopic measurements of other food sources (sessile invertebrates, zooplankton and algae) as well as stomach content analysis of the palaemonid shrimp are needed to better characterize the diet of both symbionts.