Fish, fans and hydroids: host species of pygmy seahorses

Abstract An overview of the octocoral and hydrozoan host species of pygmy seahorses is provided based on literature records and recently collected field data for Hippocampus bargibanti, Hippocampus denise and Hippocampus pontohi. Seven new associations are recognized and an overview of the so far documented host species is given. A detailed re-examination of octocoral type material and a review of the taxonomic history of the alcyonacean genera Annella (Subergorgiidae) and Muricella (Acanthogorgiidae) are included as baseline for future revisions. The host specificity and colour morphs of pygmy seahorses are discussed, as well as the reliability of (previous) identifications and conservation issues.


Introduction
Pygmy seahorses (Hippocampus spp.) (Pisces: Syngnathidae) are diminutive tropical fish that live in close association with octocorals, colonial hydrozoans, bryozoans, sea grass and algae (Lourie and Kuiter 2008), but little information is available about their host specificity. Most host organisms are notoriously hard to identify because of a lack of clear morphological characters, which leads to the risk of obtaining erroneous identifications. Therefore there is an urgent need for taxonomic revisions of these host species.
This study deals with the octocoral (Cnidaria: Anthozoa: Octocorallia) and hydrozoan (Cnidaria: Hydrozoa) hosts of the pygmy seahorses H. bargibanti, H. denise, and H. pontohi. The taxonomic problems in the octocoral host genera Muricella (Acanthogorgiidae) and Annella (Subergorgiidae) are addressed, and type material is re-examined and depicted. In addition, a literature review of all documented host species is provided, as well as accounts on newly recorded associations. The distribution records of pygmy seahorses are updated with four localities in Indonesia and Malaysia.

Material and methods
The majority of the pygmy seahorse records in the present study was obtained during fieldwork in Raja Ampat, West Papua, Indonesia (2007) (Fig. 1).
Soft corals, gorgonians and hydrozoans were thoroughly searched for pygmy seahorses to a maximum depth of 40 m (using SCUBA), with the help of local dive guides where available (Raja Ampat, Bunaken). In situ photographs were taken of both the hosts and the associated seahorses (Fig. 2). The total number of seahorses per host colony was counted, the height of each host colony was estimated and a sample was taken for identification and as voucher material. All material is stored on 70% ethanol in the collections of NCB Naturalis, Leiden (catalogue numbers RMNH Coel.). Subsamples of the Ternate material are deposited in the collections of Museum Zoologicum Bogoriense (Java, Indonesia). For the identification of the octocoral hosts, microscope slides and SEM photographs of the sclerites were made. These were obtained by dissolving the octocoral tissue in 10% sodium hypochlorite, after which they were rinsed five times with tap water and five times with double-distilled water. The sclerites were subsequently dried on glass microscope slides on a hot plate. After drying, the sclerites were brushed on a SEM stub and coated with platinum. A JEOL JSM6480LV electron microscope operated at 10 kV was used for the SEM photography. The hydrozoans were identified using a dissecting microscope.

Results
In the literature eight pygmy seahorse species have been recorded as associates of hydroids and octocorals (Table 1)

H. colemani
Australia ( RMNH Coel. 39868,39871,39874) is characterized by wide, plump, capstans from the adaxial layer, up to 0.12 mm long (Fig. 5). Secondly, Muricella sp. 2 (RMNH Coel. 39869, 39873) is characterized by small, slender adaxial capstans, up to 0.05 mm long (Fig. 6). Thirdly, Muricella sp. 3 39870,39872), is characterized by adaxial capstans intermediate in shape between the first two, up to 0.10 mm long, and big spindles with rounded ends (Fig. 7). The latter are lacking in the first two species. Muricella plectana has similar plump spindles in the coenenchyme but differs from the present material by lacking the bent spindles from the polyp (Fig. 3). M. paraplectana differs from all other material by having spindles with pointed ends (Fig. 4).
The taxonomic history of the genus Annella is puzzling. Ellis and Solander (1786) described Gorgonia reticulata and added a drawing of the habitus without further description or indication of its type locality. The type specimen of Gorgonia reticulata is presumably lost. Subsequently, Gray (1857[1858]) described the genus Annella, with A. reticulata as type species, but it is unknown whether he associated this species with Gorgonia reticulata. Later, Nutting (1910) described Euplexaura reticulata (Fig. 8), probably without considering a possible homonymy involving A. reticulata and G. reticulata. Stiasny (1937) synonymised Gorgonia reticulata and Euplexaura reticulata, based on the external morphology, and transferred the species to Suberogorgia reticu-lata. Grasshoff (1999) eventually placed G. reticulata, A. reticulata, E. reticulata and S. reticulata in the genus Annella. The species is therefore currently known as Annella reticulata (Ellis and Solander, 1786). Here the holotype of E. reticulata is re-examined and considered different from A. reticulata, based on the morphology of the double head sclerites (Figs 8, 10). Due to the netlike structure of these gorgonians, it is not surprising that the different authors independently chose 'reticulata' as epithet, so adding to the confusion. Nutting (1910) described a different species as Euplexaura mollis (type locality Moluccas). Stiasny (1937) transferred this species to Suberogorgia [= Subergorgia] (Bayer, 1981), and subsequently Grasshoff (1999) placed it in the genus Annella. The species is therefore currently known as A. mollis (Nutting, 1910).
A taxonomic revision of Annella has not yet been made, but Grasshoff (2001) provided an overview of the sclerite diversity observed within this genus. He suggests that the morphological diversity of the sclerites within these two species is correlated with their geographical distribution in the Indo-Pacific. To the best of our knowledge this would be the first and only case in octocoral taxonomy, in which sclerite morphology varies geographically. Following Grasshoff's (2001) overview of the sclerites, the Annella specimens were compared with the nearest region used by Grasshoff (2001), viz. the Moluccas. Based on those characters five specimens are identified as A. reticulata (RMNH Coel. 39878-80, 39882, 39952; Fig. 10). Likewise, two specimens are identified as A. mollis (RMNH Coel. 39875-76; Fig. 11), although the double heads of the examined specimens are less developed compared to Grasshoff's A. mollis from the Moluccas. Two of the specimens with an A. mollis colony form had sclerites like the ones depicted for specimens from the Maldives (Grasshoff 2001). These two specimens are provisionally identified as A. cf. mollis (RMNH Coel. 39877, 39881; Fig. 12) and share similarities with the holotype of E. reticulata (Fig. 8).
The sclerites of the holotype of Euplexaura mollis from the Moluccas (= Annella mollis sensu Grasshoff 1999Grasshoff , 2000 (Fig. 9) were also examined and compared with those pictured by Grasshoff (2001). These sclerites resemble the sclerites in drawings of A. reticulata from the Moluccas instead of those of A. mollis, whereas the habitus resembles A. mollis. Based on our presented material and additional material from the NCB Naturalis collection it seems unlikely that the sclerites of the two Annella species differ according to locality. Most varieties, as described by Grasshoff concerning the geographic areas, are also found in Indonesian and Malaysia's seas (unpublished data). Additional material from other locations is needed to test Grasshoff's hypothesis on geographically determined sclerite morphotypes.

Hydrozoa
On four occasions specimens of Hippocampus pontohi were observed and three of their hosts were collected. Two records of H. pontohi individuals are from a colony of Thyroscyphus fruticosus (Esper, 1793) (RMNH Coel. 39883-4), a common littoral species on coral reefs with a distribution range throughout Indonesia (Prof. W. Vervoort, pers. comm.). A single individual from Kri Island (Raja Ampat) was found on a specimen of T. fruticosus intertwined with a specimen of the hydroid Lytocarpia phyteuma (Kirchenpauer, 1876) (RMNH Coel. 39886), therefore both co-host species are listed in Table  2. Lytocarpia phyteuma is an uncommon hydrozoan, which can be found at 0-50 m depth, especially in eastern Indonesia (Prof. W. Vervoort, pers. comm.). The H. pontohi individual from Mioskon Island was found on specimens of Clytia cf. gravieri (Billard, 1904) (RMNH Coel. 39885), a common hydrozoan on coral reefs with a wide (sub-) tropical distribution range. Due to the small amount of collected material, a positive identification is not possible. This hydroid species was also recorded during previous expeditions in Indonesia, such as the Snellius II expedition (1983-84) (unpublished data Prof. W. Vervoort). The host of H. severnsi was unfortunately not sampled and therefore its identity remains unknown.

Discussion
Many sessile marine organisms contribute to the high marine biodiversity in the socalled Coral Triangle by acting as host for many associated organisms (Hoeksema 2007). Gorgonians are hosts to a variety of species, such as sponges, molluscs, hydroids, crustaceans, brittle stars and fish (Munday et al. 1997, Goh et al. 1999, McLean and Yoshioka 2007, Neves et al. 2007, Puce et al. 2008, Sih and Chouw 2009, Reijnen et al. 2010. The 'persistence' of the relationship (intermittent occurrence on host) between the associated fauna and the host organism is often largely unknown (Goh et al. 1999).
Pygmy seahorses were observed to remain on a single gorgonian for periods of at least 3-40 weeks. Information on the pygmy seahorse whereabouts after this period is lacking, and movement between different hosts was not directly observed (Baine et al. 2008). The claim that pygmy seahorses appear to parasitize their hosts (Kuiter 2000, Teske et al. 2004 has not been substantiated, just like the observations that species were seen moving over a mushroom coral (Fungia sp.) and encrusting sponges (Lourie and Kuiter 2008) do not seem to be related to real host specificity.

Host specificity
In the case of Hippocampus bargibanti two host species have been recorded in the literature, M. plectana and M. paraplectana. Three additional Muricella species from the Indo-Pacific, different from M. plectana or M. paraplectana, were found during the present study. H. bargibanti is therefore associated with at least five different Muricella spp. Unfortunately, the genus Muricella is in need of a revision (Samimi and Ofwegen 2009). The latest overview of the genus Muricella was made by Kükenthal (1924), in which species-specific characters are usually missing. This makes it impossible to identify specimens to species level.
Since Kükenthals' overview only three additional Muricella species have been described (Grasshoff 1999(Grasshoff , 2000, which are considered endemic to New Caledonia and the Red Sea. Although the current status of the taxonomy of this gorgonian genus is a large obstacle in identifying species, the results herein indicate that Muricella sp. 1, M. sp. 2 and M. sp. 3 are new host records for H. bargibanti, each based on their own unique characters (Figs 5-7).
Individuals of H. denise primarily occur on colonies of Annella spp., which they strongly resemble in colour pattern, resulting in an optimal camouflage. Based on our results Hippocampus denise lives in association with at least three different Annella species: A. reticulata, A. mollis and A. cf. mollis. No other Annella species are currently recognized, and a revision of the genus Annella is needed. This will most likely show that additional Annella species await description. One individual of H. denise was found on Muricella sp. 2 (Fig. 13). This association was already known (Table 1), but appears quite unusual. Additional host genera for H. denise are expected, based on published photographs (Kuiter 2009). A study by Sih and Chouw (2009) showed that other fish species (Bryaninops amplus Larson, 1985) associated with gorgonian hosts, select their habitat on physical properties, such as the host's size and surface area, rather than the species to which it belongs. This may explain why H. denise was encountered on a Muricella sp., instead of on its far more common host genus Annella. Fish species may generally be more associated with certain host gorgonians, but they can still be found on other hosts if the preferred host is not available.

Colour morphs
Different colour morphs are recorded for the gorgonian-associated species H. bargibanti, and H. denise (Lourie et al. 2004, Kuiter 2009), resulting in the most optimal camouflage considering the colour and the polyp structure of the gorgonians, which are perfectly mimicked by the pygmy seahorses. According to Lourie et al. (2004) the pale grey, purple with pink, and red tubercles colour morphs of H. bargibanti are found on Muricella plectana, whereas species showing yellow with orange tubercles are found on M. paraplectana. Unfortunately, it remains uncertain whether this is a valid assumption, without examining the host's sclerites. Neither M. plectana nor M. paraplectana were encountered during our field studies and all our specimens of H. bargibanti belonged to the pale grey / purple colour morph. The strict association between colour morph and specific host species can therefore not be confirmed. Based on the present data such strict associations seem unlikely, since identically coloured host species are in fact often different species.
For other pygmy seahorse species new colour morphs may be encountered, since pygmy seahorses are enigmatic species that are popular objects for divers and underwater photographers. As a result they often appear in dive magazines and field guides. Occasionally these pictures show new colour morphs or maybe even new species. For future research such observations and sightings can contribute to the general knowledge and ecology of pygmy seahorses, especially when the host organisms are collected for taxonomic studies.

Reliability of identifications
Previous identifications of the hosts of H. bargibanti may well be in error since they were made by non-gorgonian specialists, except for the identifications in Lourie and Randall (2003) which were done by Dr F.M. Bayer. Based on the herein presented data and re-examination of the holotypes, it seems plausible that the published coral host records contain several errors. For the genus Annella the literature records (see Table  1) show the same host species as found in the present study (Table 2,, but previous identifications were based on the growth form (mesh shape) and not on the sclerites. These identifications should be re-assessed based on sclerite morphology. Lourie and Kuiter (2008) mention Acanthogorgia spp. as hosts for H. denise, based on a photograph (Lourie and Randall 2003: image 10;pers. comm. Sara Lourie). This identification seems erroneous, since the polyps shown in the photograph are not characteristic for the genus Acanthogorgia. Although no certain identification can be made based on a photograph, the image most likely depicts a zoanthid (Dr James Reimer, pers. comm.), which would be the first indication that pygmy seahorses might be associated with zoanthids as well. Associations with Echinogorgia sp. and Subergorgia sp. cannot be confirmed based on material in the present study.
For many organisms molecular methods can be of help to identify species, but so far barcoding of Octocorallia for the COI gene has not been successful. Even when sequences are obtained, information on species level is very limited. Most research is currently limited to three genes, which are still unsatisfactory to identify species (McFadden et al. 2011). For Octocorallia the 'traditional' taxonomy based on morphological characters remains of primary importance. When new species of pygmy seahorses are described a photograph of the whole host colony, and a close-up of its polyps and branches should also be provided, which is normally enough to identify the host to family or possibly genus level. Preferably also tissue samples of the host should be collected for taxonomic studies.

Conservation
The distribution ranges of the pygmy seahorses are largely situated within the Coral Triangle (Lourie 2001, Hoeksema 2007, which receives much attention with regard to coral reef conservation. The entire genus Hippocampus is listed in Appendix II of CITES and H. bargibanti and H. denise are listed as data deficient in the IUCN Red List (Lourie et al. 2004), whereas the other five pygmy seahorse species have not yet been assessed. One of the main threats to seahorse populations concerns habitat loss and degradation, especially for the species depending on specific host coral species. Seahorses have been increasingly used as flagship species in local and regional conservation programs to promote the protection of both the seahorses and their habitats (Scales 2010). Knowledge on the distribution of the host species can be beneficial for conservation efforts of their associated organisms.

Conclusion
This paper shows that pygmy seahorses are associated with more gorgonian and hydrozoan hosts than previously assumed, resulting in new associations; H. bargibanti is associated with five species of the genus Muricella, H. denise is associated with three Annella species, and H. pontohi with four hydrozoan and one algae species. No new records are available for H. severnsi. The presumed association of colour morphs of H. bargibanti with certain Muricella species cannot be confirmed based on our present results. Future work on pygmy seahorses should preferably include more attention for their hosts, including taking tissue samples for identification by an octocoral taxonomist.