This study helps to lay the groundwork for developing a database of DNA sequence data from POE diatoms. The use of DNA sequence data enables application of these epizoic diatom assemblages as proxies for host and ecology without the need for an extensive knowledge of benthic diatom taxonomy. We generated DNA sequence data from 16 of the known POE diatom species for sea turtles and maintained 12 of these taxa in culture. This opens the door for future studies on the molecular, genomic and physiological nature of the epizoic diatoms. Moreover, our ability to culture POE diatoms without their hosts invites several questions about diatom-host interactions. Specifically, POE diatoms evidently do not require the host to survive and so we propose four non-exclusive hypotheses that could explain the specificity of POE to particular hosts in the wild:
H1: The epizoic environment creates a unique nutrient supply regime affecting bottom-up competition, with POE taxa being better competitors in the epizoic environment and/or worse competitors in a non-epizoic environment.
There are several ways that the epizoic host environment may increase nutrient availability relative to a more static environment. Firstly, there may be leakage of nutrient compounds through the skin of the host, directly raising N or P concentrations near the host. Although mammalian skin is generally highly impermeable, the barrier requirements of glabrous marine animals living in the slightly hypertonic seawater may be less rigorous than in terrestrial species . There is some evidence to suggest mammals and reptiles have similar skin permeability . Secondly, the movement of the host through the water will constantly reduce nutrient gradients along the host (and near the diatoms) much as stream flow or wave action does. As the animal-derived nutrients are unlikely to include silicon, the Si:P and/or Si:N supply ratios may be very low close to the animal surface, a condition that the POE diatoms may be better adapted to than opportunistic taxa. In the marine environment, rivers are one of the major silica sources  and higher Si supply rates are typical of shallow-water coastal habitats and non-mobile substrates.
Regardless of the particular mechanism(s) that might provide a different competitive environment, given that we have been able to culture so many of the epizoic diatoms we have encountered, the hypothesis that epizoic diatoms have different nutrient requirements and uptake capability than closely related benthic relatives should be testable.
H2: The POE flora requires prokaryotic microbes derived from the epizoic environment.
There is ample evidence that benthic microalgae, including diatoms, exist in a complex network of interactions (the “phycosphere”) with other microorganisms either adhering to their membranes and theca or attached via extracellular matrices [34, 35, 36]. Therefore, the biochemical signals and links to the epizoic habit may be derived from bacteria specific to the host rather than the host itself. There is some evidence already that large aquatic vertebrates can harbor unique bacterial assemblages [37, 38, 39, 40], but the interaction between epizoic bacteria and algae has barely been investigated. The epizoic diatoms isolated for this study were not grown axenically, so phycosphere-associated bacteria were transferred into culture with the diatoms. This association might have allowed the POE diatoms to grow in the absence of the epizoic habit.
H3: The POE flora has special adaptations to epizoic lifestyle.
Most of the POE species we have studied do not require the epizoic host to survive and divide. Thus, it is possible that these taxa are found on vertebrate hosts not because they are favored by such an environment, but because they are simply tolerant of it. While there is no evidence the epizoic diatoms are directly harming the host, unchecked growth of diatoms (and other algae) could increase drag and negatively affect rheological properties of the actively moving animals. Raphid diatoms are often biofilm stabilizers, depositing secreted polysaccharides which allow for further attachment by bacteria, fungi, macrophytes and invertebrates, including potential pathogens and organisms that can damage animal tissue. In addition, settlement of macroepibionts may significantly increase the host’s weight, thus increasing the energetic cost of swimming and diving. Therefore, marine vertebrates have developed various mechanisms to prevent the settlement of pioneering surface-conditioning organisms and thereby limiting epizoic diatom growth [41, 42].
While biochemical deterrents to biofouling have evolved in many organisms, there is little evidence for such specialized secretions in sirenians and sea turtles. Production of bioactive antifouling compounds is especially unlikely in the case of the sea turtle carapace scutes, which are built from physiologically inactive anucleate cells . Cetacean skin, which also harbors POE diatoms, is believed to lack any glands and its biochemical adaptation depends largely on keratinocytes present in both the basal layer of the skin and the most external stratum corneum. The nucleated cell layers are responsible for production of a gel-like matrix containing hydrolytic enzymes, such as glycosidases and peptidases . Although it has been suggested that the enzymes might degrade adhesive polymers produced by some of the biofoulers , their main role seems to be the initiation of the desquamation process . The latter could be one of the most efficient defense strategies against biofouling developed by many long-lived marine macroorganisms [42, 46]. In both glabrous marine mammals and reptiles, skin scales show highly homogenous topology with few microniches. Moreover, micro- and nanoornamentations (dermatoglyphics ) of the skin scales may decrease the number of contact points and thus prevent the attachment of microscopic biofilm precursors .
Host behavior may also influence settlement. West Indian manatees, for example, move from marine to brackish to freshwater habitats in response to seasonality and water temperature, potentially at a rate and duration to which opportunistic diatoms may not be able to adapt . Basking behavior of sea turtles as well as passive drifting close to the ocean surface typical of sea snakes , which can also carry epizoic diatoms , can expose the diatoms to damaging levels of heat, irradiance and desiccation. Deep diving, long-distance migrations and nesting behavior shown by many sea turtle species may induce intolerable rates of change in temperature, salinity, nutrient availability (particularly silica, which can be limiting in the open ocean) and hydrostatic pressure on non-adapted, opportunistic diatom species. The POE diatoms may also be better adapted to resist drag, friction, and high shear regime induced by the animal movement as well as frequent dermal abrasion caused by grooming practices employed by either the host animal itself  or specialized grazers feeding on the epizoic flora of the host [27, 50].
Resistance to these antifouling measures as well as high tolerance to rapidly and continuously changing environmental conditions would allow the POE taxa to successfully colonize and dominate the animal surfaces. However, they would provide no competitive advantage against opportunistic diatom species on other substrates. This could explain why the POE taxa thrive in both their native epizoic habitat and artificial monoculture.
H4: There is a non-epizoic reservoir of these taxa that we have yet to discover
Large areas of the world’s marine shallow benthic environment are poorly studied for diatoms, and therefore we cannot exclude the possibility that the POE taxa do exist outside of epizoic habitats. Even in localities that are relatively well-studied for benthic diatoms, variation in the composition and relative abundance in an assemblage due to substrate specificity and seasonality make the assembly of an exhaustive diatom flora extremely difficult. While a published flora exists for one of our collection sites, Florida Bay , and many small-celled diatom species have been reported, the relatively small size of the Tripterion clade POE taxa (< 20 µm) and their gross morphological similarity to other non-epizoic gomphonemoid diatoms (e.g. Gomphoseptatum Medlin, Cuneolus Giffen) makes it possible that valves in low concentrations, such as those that might have sloughed off host animals and entered the tychoplankton, might have been overlooked.
Environmental DNA surveys, such as metabarcoding, have an advantage over microscope-based surveys with regards to relatively small-sized taxa. The results of this study, however, indicate that POE diatoms can also be undetected in eDNA surveys. Based on the molecular phylogeny of the Tripterion clade, it is easy to see how these taxa might have remained undetected in a BLAST-based bioinformatic summary of OTUs, as there is significant genetic difference between the Tripterion clade and the only other sequenced representatives of the Rhoicospheniaceae—the freshwater taxon Rhoicosphenia abbreviata (C.Agardh) Lange-Bertalot. In fact, there are no morphological characters exclusive to the taxa in the molecular clade containing Tursiocola and the Tripterion clade that would cause a diatomist to expect a close match in sequence identity to the POE taxa. With curated sequence data now available for the most common POE taxa, we may find evidence for their occurrence in non-epizoic habitats through eDNA studies.
However, it should be remembered that the presence of a POE specimen (or POE DNA) in benthic samples does not discount that these taxa may be epizoic specialists. In fact, it is highly unlikely that, given enough time, one would not encounter a Tripterion or Tursiocola specimen in the benthos or plankton of Florida Bay. If a cell of a POE diatom species had just been shed—living or dead—this would have little bearing on whether the POE taxa are obligate epizoics. Similarly, the low relative abundances of POE taxa valves that might have been overlooked in the surface sediment, seaweed, and seagrass samples would suggest that non-epizoic substrates do not provide optimal habitats for POE species. While we clearly value floristic studies, the presence-absence data cannot unequivocally determine if a species is a POE.
The hypotheses listed above are not necessarily exclusive. Different species may be affected by different conditions. Moreover, potential specific adaptations to epizoic lifestyle developed by POE diatoms suggested by (H3), do not directly explain why they have not been found in other benthic habitats. It is possible that some trade-off in obtaining those adaptations makes the POE taxa less competitive in non-epizoic benthic environments (H1). We know little about how the phycosphere might affect the competitive ability of diatoms, and/or whether the phycosphere may itself manufacture some critical compound only in an epizoic community (H2). Nonetheless, we argue that these hypotheses can form the beginning of a conceptual framework to understand POE taxa functioning as well as their ecological role.
What do POE diatoms tell us about raphid diatom evolution?
Based on our molecular phylogeny, it appears that the epizoic habit has evolved several times and in several different raphid diatom morphotypes: elongate biraphid (Proschkinia and Tursiocola, Fig. 3f & 3g, respectively) and monoraphid frustules (Achnanthes, Fig. 1e), asymmetric, clavate biraphid frustules (Tripterion complex, Fig. 3a) and thin oval monoraphid frustules (Bennettella, Epipellis ). These independent gains of the epizoic habit could be driven by the host biology and evolution. The various epizoic diatom lineages, if eventually resolved to be closely linked to a specific type of host animal, might have diverged from non-epizoic taxa under different ecological and evolutionary constraints and at different times corresponding to the emergence of various groups of marine megafauna.
Among others, the eco-physiological constraints shaping epizoic diatom speciation through adaptive radiation would include the nature and character of the animal substrate. Variations of the dermal layer of sirenians and sea turtles including the ultrastructure, topology, physiology (e.g. shedding patterns), and biochemistry (e.g. enzymatic activity) would require different attachment and colonization (and re-colonization) strategies, thus encouraging the development of specific adaptations. Such a specific adaptation is evidenced by Melanothamnus maniticola Woodworth, Frankovich & Freshwater, an epizoic red alga on manatees that has unique skin penetrating rhizoids that anchor the thallus to the deeper epidermis and permit the alga to persist as the host surface skin cells are shed . In sea turtles, the carapace scutes are often retained or shed periodically, while the skin scales are either shed continuously (sea turtles) or the epidermis is renewed completely in a process called ecdysis (sea snakes ). These patterns differ from those observed in marine mammals in which skin shedding may be regulated by external factors such as temperature . Similarly, animals with different diving regimes may host diatoms with different physiological and metabolic adaptations as various stages of photosynthesis will be differently affected by changes in hydrostatic pressure related to the depth, duration, and frequency of dives .
Moreover, the diversification dynamics in POE diatoms may be linked to the host animal behavior and lifestyle. The high niche heterogeneity, biodiversity, productivity, and nutrient concentrations typical of shallow-water habitats occupied by sirenians and some sea turtles may increase colonization rates by new species and favor benthic diatom immigration to the epizoic community, thus spurring the observed diversity of diatom forms associated with manatees [24, 25] or sea turtles using neritic foraging habitats (e.g. loggerheads; ). The opposite phenomenon could explain low epizoic diatom diversity on leatherback sea turtles [5, 30], and pelagic sea snakes  that spend significant (though not exclusively  time feeding in the pelagic zone rather than on benthic organisms . This follows the general pattern of low macro-epibiotic diversity on leatherbacks . Epizoic diatom diversity might also be driven by intrinsic biotic factors, such as gregariousness and range of the host species as both factors may affect the new species encounter and colonization rates. However, in these systems in which epizoic diatom species richness is driven mainly by speciation rates as opposed to benthic species immigration, the total epizoic diatom diversity may remain low. The higher number of diatom taxa observed on neritic megafauna species as compared to open-water animals seem to support this hypothesis .
Currently, taxon sampling is still scattered, and while strains were isolated from multiple geographic localities (focused in the southeastern USA and South Africa at this time), much of the strain diversity in species-level clades come from a single collection. The Florida Poulinea lepidochelicola clade, for example, represents strains isolated exclusively from the Turtle Hospital rehabilitation facility in Marathon, Florida. Among the South African P. lepidochelicola strains, six strains (Majewksa 14C, Majewska 20C, HK630, HK638, HK639 and HK640) came from collections from three turtles at the uShaka Sea World facility in Durban, and likely represent one population. It is curious to note, however, that a morphological difference does exist between the sequenced Medlinella amphoroidea strains from South Africa and the type population of Florida Bay. The valve areolae of the former appear to be occluded by hymenes (Fig. 3d) as opposed to the volae of the type population . Whether this corresponds to a genetic, and perhaps species differentiation remains to be seen, once the Florida Bay population is sequenced.
What might POE diatoms tell us about their hosts?
While we do not yet have enough information to assign any sort of host specificity to certain POE diatom taxa, we have enough DNA sequence data to suggest that some genetic differentiation among POE diatoms is occurring. While we do not know if the genetic distance between the Florida, Mediterranean and South African Poulinea strains is driven by speciation or intraspecific biogeography, they are genetically distinct. Data collected from loggerheads suggests little mixing between sea turtle individuals across ocean basins , with the Mediterranean population being distinct from the northeast Atlantic one, which is then distinct from northwest Atlantic (including the Gulf of Mexico) population. Even within closer geographic boundaries, such as the western Atlantic, there is demonstrated genetic distance between POE strains (C. caribeana of Florida and the Bahamas; Achnanthes elongata of Florida and Georgia) in DNA sequence markers which are generally considered too conserved to show intraspecific variation in diatoms [60, 61].
Future studies and molecular information from a larger number of POE diatom strains may reveal whether genetic diversity in epizoic diatoms reflects biogeographic, ecological, and behavioral patterns observed in the host animal populations. For example, it was demonstrated that sea turtle phylogeography is shaped by the sea turtle species thermal regime and habitat preference . Provided the close relationship between epizoic diatoms and sea turtles holds up under the scrutiny of increased data sampling, it may be expected that POE diatoms associated with the cold-tolerant leatherbacks, which are able to use the southwestern corridors to migrate across the oceans, will be characterized by lower genetic diversity than diatom taxa growing on tropical species such as green turtles, hawksbills, and olive ridley sea turtles, whose Atlantic and Indo-Pacific populations appear to be genetically distinct . This knowledge may significantly advance our understanding about evolutionary relationships between diatoms and their animal hosts as well as shed more light on the mechanistic processes of divergence and adaptive evolution of diatoms and other marine microbes.