Unexpected high abyssal ophiuroid diversity in polymetallic nodule fields of the Northeast Pacific Ocean, and implications for conservation

Abstract. The largest and commercially appealing mineral deposits can be found in the abyssal seafloor of the Clarion-Clipperton Zone (CCZ), a polymetallic nodule province, in the NE Pacific Ocean, where experimental mining is due to take place. In anticipation of deep-sea mining impacts, it has become essential to rapidly and accurately assess biodiversity. For this reason, ophiuroid material collected during seven scientific cruises from five exploration license areas within CCZ, one area protected from mining (APEI3, Area of Particular Environmental Interest) in the periphery of CCZ and the DIS-turbance and re-COLonisation (DISCOL) Experimental Area (DEA), in the SE Pacific Ocean, was examined. Specimens were genetically analysed using a fragment of the mitochondrial cytochrome c oxidase subunit I (COI). Maximum Likelihood and Neighbour Joining trees were constructed, while four tree-based and distance-based methods of species delineation (ABGD, BINs, GMYC, mPTP) were employed to propose Secondary Species Hypotheses (SSHs) within the ophiuroids collected. The species delimitations analyses concordant results revealed the presence of 43 deep-sea brittle stars SSHs, revealing an unexpectedly high diversity and showing that the most conspicuous invertebrates in abyssal plains have been so far considerably under-estimated. The number of SSHs found in each area varied from 5 (IFREMER area) to 24 (BGR area), while 13 SSHs were represented by singletons. None of the SSHs was found to be present in all 7 areas, while the majority of species (44.2 %) had a single-area presence (19 SSHs). The most common species were Ophioleucidae sp. (Species 29), Amphioplus daleus (Species 2) and Ophiosphalma glabrum (Species 3), present in all areas except APEI3. The biodiversity patterns could be mainly attributed to POC fluxes that could explain the highest species numbers found in BGR (German contractor area) and UKSRL (UK contractor area) areas. The five exploration contract areas belong to a mesotrophic province, while in contrary the APEI3 is located in an oligotrophic province which could explain the lowest diversity as well as very low similarity with the other six study areas. Based on these results the representativeness and the appropriateness of APEI3 to meet its purpose of preserving the biodiversity of the CCZ fauna are questioned. Finally, this study provides the foundation for biogeographic and functional analyses that will provide insight into the drivers of species diversity and its role in ecosystem function.


sixteen license areas for the exploration of polymetallic nodules have been approved by the ISA within the CCZ, each up to 75,000 km 2 in size (Wedding et al., 2013). In its environmental management plan for the CCZ (Lodge et al., 2014), the ISA 65 adopted nine large protection areas defined as Areas of Particular Environmental Interest (APEI), where mining will not be permitted (Lodge et al., 2014). These APEIs are large enough (each of them 400 x 400 km) and far enough away apart from potential mining areas that they will not be affected by deep-sea mining (Wedding et al., 2013). In order to be effective as source populations for the recolonization of impacted areas, however, APEIs should harbour a representative subset of the fauna found in the potential fields. In addition to these protection measures, the ISA has stipulated that prior to exploitation, 70 a benthic biological baseline study must be undertaken for each exploration contract area, and the possible environmental impacts arising from exploration should be assessed. Nodule mining carries significant environmental concerns, including negative direct and indirect impacts on the biodiversity (Ramirez-Llodra et al., 2011;Vanreusel et al., 2016;Van Dover et al., 2017;Niner et al., 2018). The removal of the nodules and associated organisms could result in habitat loss, fragmentation, or modification, while the generation of sediment plumes may bury the organisms or clog their feeding 75 apparatuses and thus disrupting the food-webs (Ramirez-Llodra et al., 2011;Vanreusel et al., 2016;Van Dover et al., 2017;Niner et al., 2018;Stratmann et al., 2018). Unfortunately, accurate documentation of species diversity, which comprises the first step in understanding patterns and structures in different levels of biodiversity, biogeographical and ecological processes and is essential for marine ecosystems' management, remains poor across the CCZ . To date, Taboada et al. (2018), although dealing with a single hexactinellid sponge species, is the only study that has accessed the effectiveness 80 of an APEI (#6) or investigated connectivity with the adjacent potential mining areas. Thus, prior to exploitation, there is an urgent need to obtain baseline data on faunal biodiversity at local and regional scales in order to assess and predict the effects of mining on deep-sea organisms.
The Ophiuroidea (brittle stars and basket stars) are amongst the most emblematic mobile megafaunal inhabitants of the deep sea regarding species diversity and individuals' numbers (Gage and Tyler, 1991;Rex and Etter, 2010;Vanreusel et al., 85 2016). They constitute the most diverse echinoderm class, numbering more than 2064 species found in all oceans, from intertidal to hadal depths (Stöhr et al., 2012;Jamieson, 2015). In fact the first deep-sea species to be described was the basket star Gorgonocephalus caputmedusae (Linnaeus, 1758) (as Astrophyton linckii, Müller & Troschel, 1842) in 1600 m water depth (Menzies et al., 1973). Since then, at least 1412 species have been recorded from the deep sea of which only 109 are from abyssal depths, despite abyssal plains being the most extensive ecosystem in the world (Stöhr et al., 2012). Studies 90 describing the tropical abyssal Northeast Pacific ophiuroid fauna are scarce, including few historical studies resulting from the great expeditions of the late nineteenth and early twentieth centuries such as the HMS Challenger (Lyman, 1878(Lyman, , 1879(Lyman, , 1882Ludwig, 1898Ludwig, , 1899 and the Albatross (Clark, 1911;Clark, 1949). Limited recent studies exist (Amon et al., , 2017Glover et al., 2016), reporting only a small number of species. Consequently, the diversity of the deep-sea ophiuroid fauna in the CCZ is only poorly known. Thus the main objectives of this study were to: 1) ensure the future molecular 95 species identification for all different life-cycle stages by matching morphology-based species identifications of adult ophiuroids with molecular species assignments using COI sequences and consequently compiling a comprehensive reference https://doi.org/10.5194/bg-2019-360 Preprint. Discussion started: 26 November 2019 c Author(s) 2019. CC BY 4.0 License. library; 2) determine species ranges; 3) describe the ophiuroid biodiversity patterns of the CCZ and the DEA; 4) explore the usefulness of APEI3 for the preservation of the ophiuroid nodule fauna in the CCZ in the case of deep-sea mining.

Study areas
The study areas are located within the Clarion and Clipperton Fracture Zone (CCZ) in the northeast equatorial Pacific Ocean and at the DISCOL Experimental Area (DEA) in the Peru Basin (Fig. 1), at depths varying from 4050 m to 4933 m (Hein et al., 2013). The CCZ is characterized by gradual changes of environmental conditions (e.g. differences in surface-water productivity, depth and sediment characteristics) across an east-west and a north-south axis, that corresponds to a variation in 105 nodule size and coverage, as well as variations in faunal composition along these gradients (Wedding et al., 2013).
Ophiuroid samples were collected during seven scientific cruises from five different exploration licence areas and one area protected from mining (APEI3). Specifically ophiuroid samples were collected during the following cruises: BioNod on R/V Federation, and Slovakia), IFREMER and APEI3. Furthermore, the area Discol Experimental Area (DEA) in the Peru Basin, in which the German project DISCOL (DISturbance and reCOLonisation experiment) was performed in the late 1980ies (Thiel and Schriever, 1990;Thiel et al., 2001), was recently revisited in the framework of the JPIO Pilot Action "Ecological Aspects of Deep-Sea Mining". Ophiuroid samples were collected from the DEA during two cruises, SO241/1 and SO241/2 120 on the R/V Sonne from 28th July to 25th August 2015 and 28th August to 01st October 2015 respectively.

Specimen sampling and processing
Small-sized ophiuroid samples were collected using a Brenke-type epibenthic sledge (EBS; Brenke, 2005) from the UKSRL (5 deployments), BGR (15 deployments), IFREMER (4 deployments), GSR (4 deployments), and IOM (1 deployment) license areas, the APEI3 (3 deployments) and the DEA (9 deployments) ( Fig. 1), following standard deployment procedures 125 (Brenke, 2005). The cod ends of the supra-and epi-net were sieved through a 500 μm-and 300 μm-mesh with cold (+10°C) sea water and immediately transferred to pre-cooled (-20°C) 96% EtOH. Large-sized ophiuroid samples were collected with a remotely operated vehicle (ROV Kiel 6000, GEOMAR) using either the ROV's suction sampler or the ROV's manipulator https://doi.org/10.5194/bg-2019-360 Preprint. Discussion started: 26 November 2019 c Author(s) 2019. CC BY 4.0 License. arm by direct picking, manipulating scoops, shovels and nets. Large specimens were also preserved in pre-cooled 96% EtOH. For all specimens the ethanol was decanted after 24 hours and replaced with new 96% EtOH to guarantee high 130 ethanol concentration for preservation of high-quality DNA, and subsequently stored at -20°C. In the laboratory at Senckenberg am Meer, Germany, an integrative molecular-morphological approach was implemented for the identification of the ophiuroid specimens. In total, DNA was extracted from 525 specimens. For species delimitation analyses 300 sequences were selected (see below). All the ROV-collected specimens were photographed on-board, while the EBScollected specimens were photographed in the lab using a Leica binocular stereo-microscope or a Keyence digital 135 microscope, VHX-5000. The voucher specimens are stored in Senckenberg am Meer, DZMB, Wilhelmshaven, Germany.

Morphological species identification
All ophiuroid individuals collected were morphologically identified to the lowest possible taxonomic level (primary species hypotheses, PSH, Puillandre et al., 2012;Boissin et al., 2017). Where possible, individuals were assigned to named species, however, in many cases because of their very small size, their early developmental stage (post-larval individuals) or their 140 unique morphology, assignment in a morphological operational taxonomic unit was possible only at a higher taxonomic level, i.e. genus or family level. For a small number of damaged specimens, morphological identification beyond class was not possible. Following the DNA analyses (see below), all individuals within the same morphospecies that appeared to be genetically distinct from one another were re-examined and if necessary reassigned to different morphospecies, while some were considered to be true cryptic species in which clear morphological differences were not identified. Finally, the 145 integrated approached allowed the assignment of damaged specimens into different operational taxonomic units. Taxonomic and systematic remarks for each SSH are given in the supplementary material.

DNA extraction, amplification and sequencing
For the mtDNA COI analyses genomic DNA was extracted from arm tissue in individuals larger than 1-2 mm or from whole 150 individuals when smaller than 1-2 mm. DNA extractions were carried out using 30 μl Chelex (InstaGene Matrix, Bio-Rad) according to the protocol of Estoup et al. (1996) and directly used as DNA template for PCR. All DNA samples were stored at −20°C. In the cases where the whole individual was used, 20-25 μL of the supernatant was first separated from ophiuroid's voucher specimen, while the individual, which was generally intact, was transferred to 96% ethanol and stored as a voucher for morphological identifications. A fragment of 657bp of the mitochondrial cytochrome c oxidase subunit 155 (COI) was amplified by polymerase chain reaction (PCR). Amplifications were performed using Illustra PuReTaq Ready-To-Go PCR Beads (GE Healthcare) in a 25-μL volume containing 22 μL ddH 2 O, 0.5 μL of each primer (10 pmolμL −1 ) and 2 μL of DNA template or AccuStart PCR SuperMix (ThermoFisher Scientific) in a 25-μL volume containing PCR SuperMix amplification the forward primer LCOech1aF1 and the reverse primer HCO2198, tailed with M13F and M13R-pUC, 160 respectively (Folmer et al., 1994;Layton et al., 2016) were used. The amplification conditions consisted of an initial denaturation step of 3 min at 94°C, 35 cycles of 30 s at 94°C, 60 s at 42-47°C and 1 min at 72°C, followed by a final extension step of 5 min at 72°C. All PCR products were purified using ExoSap-IT (ThermoFisher Scientific). The amplified fragments were sequenced in both directions at Macrogen Europe Laboratory (Amsterdam, The Netherlands).

Alignment, genetic divergence 165
The obtained COI sequences were searched against the GenBank nucleotide database using BLASTN (Altschul et al., 1990).
Forward and reverse sequences for each individual were assembled and edited using Geneious v.9.1.7 (www.geneious.com; Kearse et al., 2012). The edited COI sequences were aligned using MAFFT v7.308 under E-INS-i and G-INS-I algorithms (Katoh et al., 2002), while alignments were further manually edited. Our dataset was supplemented with 18 COI ophiuroid sequences from the study of Glover et al. (2016). Sequence data are available in GenBank (Accession numbers 170 XXXXXXX-XXXXXXX). The sequences, trace files, collection data and photos for each specimen are listed in the datasets CCZ_Ophiuroidea and DEA_Ophiuroidea in BOLD.

Putative species delimitation
Congruent support across a range of species delimitation approaches is assumedly provides more reliable results than a single method (Carstens et al., 2013;Fontaneto et al., 2015). Therefore, four different species delimitation analyses, 175 including both distance-and tree-based approaches, were performed on the COI dataset, to allocate sequences into genetic species (secondary species hypotheses, SSH Puillandre et al., 2012;Boissin et al., 2017). Distance-based approaches detect the distance at which the 'barcode gap' occurs and sort the sequences into putative species based on this distance, whereas tree-based approaches use a phylogenetic tree from which the fit of speciation and coalescent processes are modelled to delineate species based on the branching rate of the tree (Carstens et al., 2013;Tang et al., 2014). 180

Distance-based approaches
A neighbour-joining tree was constructed in MEGA7 using a p-distance substitution model, treating gaps/missing data with "pairwise deletion", and by running 1000 bootstrap replicates. Automatic Barcode Gap Discovery (ABGD) analysis was implemented on the web interface: http://wwwabi.snv.jussieu.fr/public/abgd/ with default parameters, under the p-distance model (Puillandre et al., 2012). Barcode Index Numbers (BINs) were assigned on the registered DNA dataset automatically 185 using the BOLD v.4 workbench (ww.boldsystems.org; Ratnasingham and Hebert, 2013).

Maximum Likelihood tree
The COI barcode data available for all 300 samples was adequate to show genetic diversity patterns within and among closely related species but was not sufficient to accurately reconstruct relationships and genetic distances among the many https://doi.org/10.5194/bg-2019-360 Preprint. Discussion started: 26 November 2019 c Author(s) 2019. CC BY 4.0 License. divergent lineages in this biota. Hence to provide an all barcode sample maximum likelihood tree better reflecting these 190 divergences the barcode samples were appended to a powerful phylogenetic framework: the 48,475-site exon-28SrDNA-COI supermatrix dataset used by Christodoulou et al. (2019). This dataset comprised 200 species that outlined the ophiuroid family-level phylogeny (O'Hara et al., 2017) and 49 CCZ-DEA barcode samples with both COI and 28S sequences included (Christodoulou et al., 2019). The COI and 28S allowed the barcode only samples to be linked to the phylogenomic exon data. A maximum likelihood tree was constructed by IQ-TREE 1.6.9 (Nguyen et al., 2015;Hoang et al., 2018)

Tree-based approaches
The General Mixed Yule Coalescent (GMYC; Pons et al., 2006) method was implemented using the R package SPLITS 200 (Fujisawa and Barraclough, 2013), under the single-threshold model (stGMYC), and with the required ultrametric tree being produced in BEAST v.2.5. Settings were as follows: strict clock, Yule speciation model, GTR+G substitution site model, two independent MCMC chain runs for 50,000,000 generations, sampled every 1,000 steps (10% was discarded as burn-in period). The multi-rate Poisson Tree Processes (mPTP; Kapli et al., 2017) analysis used the rooted "supermatrix backbone" IQ-TREE phylogeny (see above). The mPTP was implemented on the web server: https://mptp.h-its.org using the multi-rate 205 Poisson tree process model and following default settings.

Genetic distances
Sequence divergences (Table 1, Tables S1-S2) were estimated using uncorrected p-distances and under the K2P model using MEGA7 according to the secondary species hypotheses.

Community structure and diversity analyses 210
Comparison of the ophiuroid communities between areas was performed in R using the package 'vegan' (Oksanen et al., 2008). As the sampling effort was very different between areas, the community table (Table 2), including the specimens of each species found in each area, was subjected to 'Chord' transformation to explore differences in relative abundance and to 'presence-absence' transformation related to faunistic differences. After transformation nMDS ordination was achieved with Euclidean distance (Legendre and Gallagher, 2001). As the number of specimens found differs greatly between areas, 215 diversity comparison was achieved using rarefaction curves, together with standard diversity indices Shannon (H´), Simpson (D) and Jaccard's Evenness (J). The expected number of species per area was inferred using the extrapolation methods Chao1 (Chao, 1994;Colwell and Coddington, 1994) and ACE (Chazdon et al., 1998). Chao1 uses the proportions of singletons and doubletons in the sample to estimate expected species richness, while ACE is an abundance-base coverage estimator. For the analysis of beta (regional) diversity, the total multiple-site beta diversity βSOR was calculated using the modified Sørensen Index (Sørensen, 1948;Balseaga and Orme, 2012), and βSOR was decomposed into its additive components "multiple-site species turnover" βSIM (Simpson Index: Simpson, 1943) and "multiple-site nestedness" βSNE using the R package "betapart" (Balseaga, 2010;Balseaga and Orme, 2012). In order to explore the relative contribution of every area to species turnover and nestedness, these values were calculated taking one area out each time in a jackkniffe approach.
Changes in turnover and nestedness were then attributable to the area that was excluded from the analysis. 225

Species delineation
The species delineation data set was comprised of 300 barcode sequences ( Fig. 2), out of which 287 were novel sequences (BOLD datasets: CCZ_Ophiuroidea, DEA_Ophiuroidea), ranging from 547 to 657bp in length (92% has 657bp length).
Both trees produced by neighbour-joining, NJ ( To be conservative, we focus primarily on the results of initial partitioning (35 SSH) as they were consistent across the parameter settings and congruent with other species delimitation methods (Puillandre et al., 2012;Kekkonen and Hebert, 2014). Nevertheless, for comparative reasons, the results of the recursive 240 partition with prior divergence 0.0359-0.0599 and which suggested 40 SSH are also presented here (Fig. 2, Fig. S1).
In BOLD, the 300 barcodes were assigned to 49 BINs (Fig. 2 Depending on the applied method, the numbers of different putative species ranged from 35 to 49. Arranging the implemented methods by increasing conservativeness gives the following: BINs (49) < stGMYC (47) < mPTP (42) < ABGD (35). In the present study a consensus dataset of species that were delineated by at least three of the four above-mentioned 250 approaches was selected, as species delineation methods tend to overestimate the number of species present in a dataset. In the few cases that the methods were inconsistent, the most conservative approach was adopted after taking into account the genetic distance between the potential species. The results were cross-referenced with the topology produced by both the NJ and ML trees. It is worth mentioning that 27 SSH were congruent throughout all methods and 34 SSH were consistent when excluding ABGDi (initial) which was the most conservative method. In total 43 SSH were recovered from the CCZ and the 255 DEA, of which some were PSHs splitted from two up to five SSHs each. Noticeably, the PSHs Amphioplus daleus, Ophiuroglypha cf. polyacantha and Ophiosphalma glabrum, Ophiocymbium sp. revealed cryptic lineages between their populations in the CCZ and the DEA. The 43 SSHs (Figs 5-17) are grouped in 11 families, Amphilepididae, Amphiuridae, Euryalidae, Ophiernidae, Ophiohelidae, Ophiolepididae, Ophioleucidae, Ophiopyrgidae, Ophioscolecidae, Ophiosphalmidae, Ophiotomidae, attributed to all the ophuroid orders (Fig. 3), Amphilepidida, Euryalida, Ophiacanthida, Ophioscolecida, 260 Ophiurida (see also Taxonomic and systematic remarks, Supplementary Material).

Genetic distances
Summaries of uncorrected pairwise distances for the ophiuroid species (SSHs) are shown in Table 1 and Fig. 4, with the full data available in the Supplementary Material Section (Table S1-S2). Mean interspecific genetic distances ranged from 0.050 to 0.370 (p-distance) and 0.052 to 0.512 (K2P distance) with the lowest divergence value observed between Ophiosphalma 265 glabrum and Ophiosphalma cf. glabrum (Species 3 vs 36) and the highest between Ophiacantha cosmica (Species 27) and Ophiohelidae sp. (Species 32). Mean intraspecific variability ranged from 0.00 to 0.055 (p-distance) and 0.00 to 0.057 (K2P distance), with the highest values observed in the ophiuroid Amphioplus cf. daleus. It should be mentioned that there were 14 SSHs represented by only one sample (singletons).

Ophiuroid communities and diversity 270
The community table (Table 2) shows the counts of each species by area. The diversity values are summarised in Table 3. A total of 55 sites were sampled in seven areas. Sampling effort was uneven, with most samples deriving from the BGR area (18) and the DEA (14) in the Peru Basin. For all other areas, 3-6 sites were sampled. A total of 543 specimens were assigned to the 43 species. None of the species was recorded in all seven areas, while the most common species were Species 29 (Ophioleucidae), Species 2 (Amphioplus daleus) and Species 3 (Ophiosphalma glabrum) which were found in 6 areas, all of 275 them absent in APEI3. Is worth mentioning that the majority of species (44.2 %) was present only in one of the area (19 SSHs). Highest species numbers were found in the BGR and UKSRL areas (24 and 21 respectively), where also the highest number of specimens were recorded (219 and 158 respectively). Lowest values were found in the IFREMER area, with 13 specimens being attributable to 5 species. While the number of species was a function of the number of specimens, less species were recorded in the IFREMER and IOM areas than would be expected if they were to follow the same pattern as at 280 other sites (Fig. 18). This was corroborated by the rarefaction analysis (Fig. 19), which shows that for the same number of specimens, the IFREMER and IOM areas have fewer species. The rarefaction curves of all other areas were very similar.
Low diversity in the IFREMER and IOM areas was also indicated by the lowest Shannon Diversity, Simpson Diversity and Evenness values, while highest diversity values were recorded in the areas UKSRL, BGR and DISCOL (Table 4) extrapolation analyses predicted a total of 57 species (Chao1 index) and 53.5 species (ACE index) for all areas together. 285 Lowest extrapolated numbers of species were again obtained for the IFREMER and IOM areas (6.5-12 and 8.5-11.5, respectively), whereas highest numbers were obtained for the BGR and UKSRL areas (57-51.1 and 27.2-30.5, respectively).
The highest number of unique species (species found only in one area) was found in the areas BGR (6 species), UKSRL (5 species) and APEI3 (5 species), while no unique species were observed in the IFREMER and GSR areas.
The faunistic similarity is summarised in Table 4, showing the number of shared and unshared species between areas. APEI3 290 showed the lowest numbers of shared species (0-2) and the highest number of unshared species (13-32) compared with other areas. The most distant area, DISCOL in the Peru basin, shared 3-11 species with CCZ exploration areas, but none with APEI3.
Beta diversity decomposition is shown in Fig. 20. The total multiple-site beta diversity was high (β SOR =0.782), with a higher component of turnover (β SIM =0.640) versus nestedness (β SNE =0.142). To explore the relative contribution of each area to 295 total beta diversity, each area was taken out once and beta diversity was re-calculated. The relative change in turnover and nestedness was then attributable to the omitted area. Results of this exercise are shown graphically in Fig. 20 and numerically in Table 4. Removing most of the areas one by one (excluding APEI3) did not result in a drastic change in turnover and nestedness (β SIM =0.604-0.663; β SNE =0.121-0.177). Only the exclusion of APEI3 resulted in a substantial reduction of turnover and increase of nestedness (β SIM =0.488; β SNE =0.229). 300 The nMDS plot in Fig. 21 shows the quantitative community analysis using Chord distance (relative abundance). The BGR and UKSRL areas were close together, but also close to the areas of DISCOL, IFREMER and IOM, while greater dissimilarity occurs with the GSR and APEI3 areas. The boxplot in Fig. 22 shows the variation in Chord distance of each area to other areas, evidencing that APEI3 was most different to any other area (see median and extent of whiskers) than other areas between each other. 305 The ordination using presence/absence transformed data placed the areas with less unique species (IFREMER, GSR and IOM) in the middle of the plot and spreaded the areas with highest number of unique species at the outer margins and apart from each other (Fig. 23). The boxplot in Fig. 24 shows that APEI3 was the most dissimilar in terms of presence/absence of species, but the median value (black horizontal bar inside the boxes) was as high as UKSRL and BGR areas, which, however, displayed less variation. 310

Species delimitation method performance
The results obtained here were consistent with many other studies showing that different species delimitation methods can produce different delimitation scenarios when employing single-locus data (Hofmann et al., 2019). The single-locus species delimitation methods tested here, although they are extensively used throughout the literature, including for the Ophiuroidea 315 (Khodami et al., 2014;Laakman et al., 2016;Boissin et al., 2017), are each subject to potential biases and differing conditions inherent in the empirical datasets (Hofmann et al., 2019). The four species delimitation methods used here generally recovered the same number of SSH and despite some degree of incongruence observed in the numbers of SSHs, they were consistent in recovering more SSHs than the number of species originally recognised. Given the lack of information regarding the biodiversity of and the relationships between deep-sea ophiuroids, it was not surprising that more 320 lineages were inferred than are currently recognised. It is likely that many of these SSHs correspond to undescribed cryptic species, but simultaneously some may be the result of genetic drift or isolated populations currently undergoing speciation.
Noticeably, the BIN method in BOLD recovered a higher number of species than all other approaches. BOLD and specifically BINs can greatly improve the Linnaean taxonomic assignment in many animal groups, including echinoderms (Layton et al., 2016;Laakman et al., 2016). The low intra-cluster divergence (2.2%) at the initial cluster step of RESL 325 methodology (Ratnasingham and Hebert, 2013;Song et al., 2018) could be the reason why, in some cases the BIN method overestimated species number, especially since there appears a small overlap between intraspecific and interspecific distance in our data (Fig. 4). This could be the case in the delimited Ophiocymbium spp. (species 24, 25, 40; Fig. 2, Table S2), which were separated into numerous lineages despite the relative low divergence between them. Generally, barcodes are well defined when the lowest interspecific distance exceeds the highest intraspecific distance, and in such cases a species 330 delineation 'threshold' will be clear. But, as the threshold can be lineage-specific, a universal threshold that fits all the branches may not exist, as coalescent depths among species will vary greatly due to differences in population size, mutation rate and speciation times (Colins and Cruickshank, 2012). Similarly, GMYC recovered a relatively high number of species (47 vs 49 BINs). Arguably, GMYC and especially the single-threshold version of the method is a robust species delimitation method (Fujisawa and Barraclough, 2013). GMYC performance depends on a single-locus ultrametric tree which tends to 335 compress the coalescent events towards the tips of the tree, making it especially difficult to distinguish closely related species (Boissin et al., 2017). It has been argued that the PTP methods generate diversity estimates that are more robust to different phylogenetic methods, while GMYC is more sensitive, but provide consistent estimates for BEAST trees (Tang et al., 2014). Nevertheless, unresolved nodes can affect both GMYC and PTP estimates, although seem to have a greater effect on GMYC estimates (Tang et al., 2014). In contrast it seems that ABGD (initial partition) has underestimated the species 340 number in this study, although the performance of the method improved when the recursive partition option was used.
ABGD has been reported to over-lump speciose datasets with high speciation rates (Dellicour and Flot, 2018). ABGD's conservatism and GMYC overestimates have also been shown on reef brittle stars (Boissin et al., 2017), while both studies indicated that PTP methods show a small advantage as the most stable, suggesting the presence of additional cryptic species but without over-splitting taxa. Summarising, despite the differences in the number of delimited species, overall the methods 345 recovered a broadly similar number of SSH. Congruence among different delimitation methods is a strong indication that the delimitation is correct, allowing the designation of cryptic species and rectification of taxonomic problems (Dellicour and Flot, 2018), always when possible taking into account the morphology.

Taxonomic Implications
The abyssal Eastern Pacific harbours a highly diverse ecosystem. The number of ophiuroid species reported from the 350 polymetallic nodule fields of the Pacific has now increased by 433%, from 10 Amon et al., 2017) to 43 in this paper. This is the largest collection of any megafaunal taxon in the CCZ and the only one that has been studied in such detail using a comprehensive combination of morphological and genetic evidence. Remarkably, from the species reported here, 32 (75%) are probably new to science and some represent hitherto unknown old evolutionary lineages (see also Christodoulou et al., 2019). The discovery of new species is the direct result of increased sampling effort, in which a 355 greater number of specimens deriving from a larger sampling surface were collected than during any previous studies in the DEA or in CCZ, spanning over five exploration contract areas and one APEI. Furthermore, the use of new sampling gears, i.e. Epibenthic Sledge (EBS), in relation to past historical expeditions that took place in the area, permitted the collection of fragile and minute specimens, while new DNA barcoding approaches allowed the identification of post-larvae and juveniles that lacked adult morphological characters. Overall, these data show that the brittle-star biodiversity in the deep sea is still 360 greatly underestimated, while supporting the use of DNA barcoding as an effective and time-efficient method of species delimitation to complement morphological studies. Although we do not wish to suggest that single mitochondrial locus data should be used on its own to draw taxonomic conclusions, in much the same way as using single morphological characters is  (Smith et al., 1995), proving that there is still a lot to be discovered and reevaluated within this group. Specifically in our study, DNA barcoding proved to be necessary since a significant proportion of the specimens are post-larvae juveniles, making their identification based on morphological characteristics quite challenging. DNA barcoding not only allowed species delimitation but also aided in matching post-larvae individuals with 370 their corresponding adults, as for example in the case of Ophiosphalma glabrum where individuals ranging from 0.5 mm to 20 cm were collected (Fig. 24). Furthermore, the large-sized brittle-stars collected with the remotely operated vehicle were matched with their in-situ photos allowing a more accurate estimation of morphospecies which in turn could facilitate the more accurate annotation of photos and video transects used in various biodiversity assessment studies (Tilot et al., 2018) Mean COI genetic intraspecific distances (K2P) of brittle stars (0.00-0.057) were concordant with previous ophiuroid 375 studies (0.00-0.042: Khodami et al., 2014 and 0.00-0.064: Boissin et al., 2017), while the mean COI interspecific genetic distances (0.052-0.512) were found to be noticeably higher. This could be attributed both to the great phylodiversity of ophiuroids collected from the polymetallic fields, spanning over 11 families and 5 orders, and to the discovery of previously undescribed diversity up to the family level (Christodoulou et al., 2019).

Ophiuroid diversity and community structure, implications for conservation in the light of possible nodule mining 380
Mining of abyssal polymetallic nodules in the CCZ could result in severe habitat disruption and loss of benthic communities in and directly around the mined sites (Vanreusel et al., 2016). An attempt to foresee the potential impact of deep-sea mining to abyssal communities requires a profound knowledge of natural background biodiversity and ecosystem functioning, such as how many and which species are living there now? How large are the species ranges? Are there natural changes in diversity along environmental gradients? However, our knowledge of abyssal benthic communities is still so poor, that even 385 these simple questions remain unanswered for many groups of organisms in what is nonetheless considered to be an economically important and potentially endangered deep-sea region like the CCZ. We can now provide partial answers to these questions for the ophiuroids, one of the dominant megafaunal groups in the CCZ.
Our initial assumption was that ophiuroid diversity would be low (we expected around 10 species) based on the previous studies in the region Amon et al., 2017) and on a recent review of global ophiuroid distribution, in 390 which only 28 species were recorded for the whole tropical East Pacific at abyssal depths (Stöhr et al., 2012). Coupled with expectations of low diversity we assumed that connectivity would be high and that most beta diversity between sites would be composed of nestedness (high) rather than species turnover (low). Under these circumstances, the APEI3 could be a good region to host most of the CCZ species and serve as source for most of the populations. Also we assumed that the most distant DISCOL area would display a low similarity with the CCZ. 395 The results of our study do not support any of these initial assumptions, in fact, they show exactly the opposite. We recorded a fourfold higher number of species than expected and rarefaction curves show no sign of reaching an asymptote (Fig. 19).
Chao1 and ACE estimators predicted between 53 and 57 species across the region ( Table 4). The ophiuroid communities were characterised by high beta diversity that is mainly composed of high turnover between areas, rather than nestedness. This means that there was a high proportion of rare species (19 species were present in only one area and 12 species in only 400 in 2 areas), which reduces nestedness and increases the potential for damage to natural populations caused by deep-sea mining at local scale. Food availability regulated by particulate organic carbon (POC) flux seems to strongly influence diversity and abundance in abyssal ecosystems (Smith et al., 2008). The CCZ licence areas for exploration despite being all in a mesotrophic zone are not all the same. The POC flux in the CCZ shows a southeast to northwest gradient increasing towards its eastern edge (Vanreusel et al., 2016;Volz at al., 2018, in review). Volz et al. (2018, in review) after studying four 405 of the CCZ areas studied herein and the APEI3 found that they differ in POC fluxes to the seafloor ranging from as low as 1 mg C org m -2 d -1 in APEI3 to 2mg C org m -2 d -1 to BGR area. Within this study the highest diversity values were recorded in the areas UKSRL, BGR and DISCOL. The BGR and UKSL areas are located in the far east of CCZ, in a region of higher POC flux (Vanreusel et al., 2016;Volz at al., 2018, in review), that could explain higher standing stocks and higher species diversities and abundances. Food availability seems to justify why the communities of the very distant and eutrophic 410 DISCOL area (Haeckel et al., 2001), resemble the eastern CCZ more than the geographically closer APEI3. The DISCOL area shares 11 species with the BGR and 8 species with the UKSL area, while no species are shared with APEI3. In contrary to the CCZ areas APEI3 lies within an oligotrophic zone exhibiting twofold lower POC fluxes and subsequently twofold lower aerobic respiration rates in comparison to the BGR area (Volz et al. 2018, in review). APEI3 differs significantly from the rest areas in additional aspects such as lower Chloroplastic Pigment Equivalents (CPE) and total organic carbon (TOC) 415 values, lower sedimentation rates resulting in finer sediments, with higher clay content (Hauquier et al., 2019;Volz et al. 2018, in review). In conclusion the APEI3 biogeochemical features differ considerably from the other areas (Volz et al., 2018, in review) and could explain the fact that APEI3 has a very different community sharing only up to two species with the CCZ areas and DISCOL.Furthermore, APEI3 is mainly located outside the CCZ, north of the Clarion Fracture, a submarine mountain range characterised by a peak and trough surpassing 1800 m difference in elevation (Hall and Gurnis, 420 2005), which may act as a dispersal barrier for abyssal fauna.
This raised the critical question whether the APEI3 fauna is representative for the exploration licence areas in the CCZ especially as ISA created the APEIs based on environmental conditions but in the absence of any biological data (Wedding et al., 2013). Our results suggest that APEI3 may not be a good surrogate area for the CCZ nodule fauna. Only a small fraction of the total registered ophiuroid fauna was recorded in APEI3. This area is the most different in terms of species 425 composition and community structure (Figs 18, 20). Removing the APEI3 from beta analysis, results in a great reduction of turnover and increase of nestedness. This means that the remaining areas become more similar to each other, the total beta diversity decreases, and differences between sites due to missing species out of a common species pool (nestedness) increases when APEI3 is excluded. Lower species richness and abundance in APEI3 as well as very low similarity between the APEI3 and the exploration areas, independently of distance were also found in the studies of Vanreusel et al. (2016), 430 Hauquier et al. (2019) and Bonifácio et al. (in review) after studying megafaunal, nematode and polychaete assemblages respectively. Volz et al. (2018, in review) after studying the (bio)geochemical characteristics of APEI3 as well as four eastern CCZ concluded that the preservation area APEI3 does not represent the depositional conditions and bio-geochemical processes that are dominating in the investigated CCZ license areas. The results of these studies converge with ours in finding that APEI3 is ill-suited as representative area of the recovery of the potentially mined areas. Furthermore, Taboada et 435 al. (2018) found that APEI6 is inadequate to act as population source for a hexactinellid sponge species and suggest the designation of a new APEI closer to the exploration areas studied. The large geographic distance between the APEIs and the explorations areas, may hinder the exchange of individuals and the genetic flow, among remaining populations after mining. Therefore, we strongly advocate in favour of incorporating no-mining sites within the core CCZ area, having a similar nodule composition as the potential mining areas, rather only in the periphery of CCZ, as was already suggested by 440 Vanreusel et al. (2016), in order to prevent the loss of biodiversity.
Biodiversity studies that focus only on known, nominal species are problematic, as they likely overlook cryptic or undiscovered lineages involved in diversification. As shown herein, most of the brittle stars recorded in the CCZ and Peru Basin lack formal Linnaean scientific names, thus widening the gap between described species and actual biodiversity, which appears to be far greater than previously estimated. Not recognising these cryptic or undescribed taxa ensures that 445 they remain in the shadows of research and conservation policies. These taxa could be locally endemic or rare, and thus more vulnerable to human impacts such as deep-sea mining. Biodiversity studies, such as presented here, aiming to develop reference libraries while using an integrative taxonomic approach, such as presented here, will provide much-needed comprehensive and time-efficient assessments of "missed" diversity. These, in turn, may fill the gaps for adequate baseline assessment at the onset of commercial-scale mining and thus, through adequate management schemes, prevent serious 450 species declines before they have been adequately described or even discovered.
In conclusion, it is important to note that present study explored only a part of the polymetallic fields of the CCZ and DEA.
Our dataset on ophiuroid communities in the CCZ is the largest available to date, but still too small to allow for comprehensive conclusions. The conclusions for APEI3 cannot be extrapolated to other APEIs in the region. Also, our dataset is biased toward the eastern areas of the CCZ where the sampling effort was higher. Thus, although a large number of 455 specimens were examined, is highly likely that the true biodiversity is even much higher. Broader efforts, especially those that will include samples from the western parts of the CCZ, or from other APEIs, are likely to result in the discovery of additional diversity and will allow us to obtain a better understanding of connectivity and patterns of distribution across the CCZ. This will in turn refine our perception of the marine biodiversity of the abyssal plains and specifically of polymetallic nodule fields. 460

Conclusions
Four methods of species delineation showed concordant results and revealed 43 deep-sea ophuroid species in the Clarion-Clipperton Zone and the DISCOL Experimental Area (Pacific Ocean), revealing an unexpectedly high diversity and showing that the most conspicuous invertebrates in abyssal plains have been so far considerably under-estimated. This study increases the number of ophiuroid species reported from polymetallic nodule fields of the Pacific by 433%. 465 A comprehensive reference library including 287 novel ophiuroid sequences allocated to 43 species is produced. This reference library can facilitate the assessment of potential impacts and biodiversity loss due to deep-sea mining. It is the first time such an integrated reference library is produced for the CCZ and the DISCOL area including both genetic and morphological information about the most emblematic mobile megafaunal inhabitants.
The biodiversity patterns observed within CCZ could be mainly attributed to differences in POC fluxes explaining the higher 470 species numbers found in BGR and UKSRL areas. The five exploration contract areas belong to a mesotrophic province, while in contrary the APEI3 (Area of Particular Environmental Interest) is located in an oligotrophic province which could explain the lowest diversity as well as very low similarity with the other six study areas.
Based on the results of our study the representativeness and the appropriateness of APEI3 (Area of Particular Environmental Interest) to meet its purpose of preserving the biodiversity of the CCZ fauna is questioned, while the creation of no-mining 475 sites within the core CCZ area is suggested.