Gorgoniapolynoe caeciliae revisited: The discovery of new species and molecular connectivity in deep-sea commensal polynoids from the Central Atlantic

Gorgoniapolynoe caeciliae (Fauvel, 1913) is a deep-sea commensal polynoid that lives in association with several genera of octocorals from the order Alcyonacea. The species has been recorded in the Caribbean and in both Atlantic and Indian Ocean basins. The wide geographic range of G. caeciliae, coupled with it having multiple host coral species and the evolution of its taxonomic description, hints that it could potentially be a species complex. This study investigated the morphological and genetic differentiation in 82 specimens of G. cf. caeciliae, sampled from four seamounts in the Central Atlantic separated by thousands of kilometres. Our combined morphological and molecular analyses, including species delimitation models (ABGD and bPTP) using COI and a phylogenetic approach using four molecular markers ( COI , 16S , 28S , and 18S ), agreed in identifying three distinct species; two supported by morphological and molecular data and a third species, using molecular data only, from the Indian Ocean which had been previously identified as G. caeciliae. We formally describe a new species in the genus, Gorgoniapolynoe pseudocaecliae sp. nov., the most common taxa found in our study. Our morphological analyses of some members of the genus Gorgoniapolynoe revealed the presence of elytra with possible photocytes (bioluminescent cells) and conspicuous macropapillae with long cilia emerging from them, whose function is discussed here. Our demographic analysis using COI for two Gorgoniapolynoe sp. detected a high potential for dispersal for G. pseudocaecliae sp. nov., with sites approximately 3000 km apart being well connected. Unusually there was also no genetic differentiation across their bathymetric range of over 1500 m. All in all, our study highlights the importance of applying integrative taxonomy to poorly studied deep-sea species.


Introduction
In the age of molecular analysis, many new deep-sea species are being discovered in previously, seemingly, widespread single species (Vrijenhoek, 2009). These cryptic species were 'hidden' due to a lack of morphological differentiation, the traditional delimiter used by taxonomists since the inception of taxonomy. Even when minor morphological differences between individuals were observed, they were often ascribed to intraspecific variation and plasticity (Castelin et al., 2017;Dueñas and Sánchez, 2009;Oug et al., 2017;Vrijenhoek, 2009). Poor descriptions, due to limited and often damaged holotypes and assumed morphological variation/plasticity, are known to have led to an underestimation of biodiversity in the deep sea (Vrijenhoek, 2009). This underestimation, along with the lack of basic ecological knowledge, is a major barrier towards being able to predict how deep-sea habitats will react under increased anthropogenic pressure from climate change and deep-sea mining (Howell et al., 2020). Marine annelids are a group which have historically been viewed by taxonomists to have high intraspecific morphological variation and wide geographic ranges, particularly deep-sea species (Hutchings and Kupriyanova, 2018). One such species which falls under this description is the deep-sea annelid Gorgoniapolynoe caeciliae (Fauvel, 1913).
Gorgoniapolynoe caeciliae, within the family Polynoidae and member of the Suborder Aphroditiformia (Annelida), is a scale worm, named as such due to the presence of scale-like elytra on their dorsal surface. Polynoids are a highly diverse group within scale worms (Gonzalez et al., 2018;Zhang et al., 2018) with over 871 species (Martin et al., 2021). Polynoids are found in all marine habitats but are prominent members of the deep sea, with 16% of all described species found below 1000m (Martin et al., 2021). They are often found in association with other marine taxa in commensal or parasitic symbiotic relationships Britayev, 1998, 2018). Gorgoniapolynoe caeciliae is a deep-sea commensal species found exclusively in association with deep-sea octocorals from the order Alcyonacea (Fauvel, 1913;Pettibone 1991). They are one of the 490 marine annelid species known to be involved in 1229 commensal relationships with other marine invertebrates Britayev, 1998, 2018). All species in the genus Gorgoniapolynoe are colonial, meaning many individuals can be found living on a single alcyonacean host (Pettibone 1991). The worms live inside tunnels on coral branches, which are not excavated into the coral, rather the worm appears to induce the coral to modify their sclerites to form these tunnel abodes (Barnich et al., 2013;Pettibone, 1991). Of all the species of Gorgoniapolynoe, G. caeciliae has accumulated the most records in the literature (Fauvel, 1913;Hartmann-Schroder, 1985;Stock, 1986;Pettibone, 1991;Eckelbarger et al., 2005;Simpson and Watling, 2011;Britayev et al., 2014;Tu et al., 2015;Macpherson et al., 2016;Serpetti et al., 2017). From these studies, G. caeciliae has been recorded in association with ten host species of Alcyonacea: Acanthogorgia armata Verrill, 1878, Acanthogorgia aspera Pourtalès, 1867, Candidella imbricata (Johnson, 1862), Hemicorallium bayeri (Simpson and Watling, 2011), Hemicorallium niobe (Bayer, 1964), Hemicorallium tricolor (Johnson, 1899), Pleurocorallium johnsoni (Gray, 1860), Pleurocorallium secundum (Dana, 1846), Narella sp. and Isididae sp.. Among marine annelids, of the 618 symbiotic relationships recorded, 57% are monoxenous i.e., having a single host species, 19% have two host species (polyxenous, having two or more relationships), and in total 33% have between two and five host species Britayev, 1998, 2018). Thus, the number of host species recorded for G. caeciliae makes it unusual. Gorgoniapolynoe caeciliae is also unusual as 72% of known marine annelid associates with deep-sea alcyonacean corals are monoxenous (Molodtsova et al., 2016). Fauvel (1913) first described G. caeciliae (under the synonym Polynoe caeciliae) over 100 years ago. There is a significant correlation between the time a deep-sea species is first described and the size of its range (Higgs and Attrill, 2015). Higgs and Attrill (2015) suggested this correlation may be explained due to wide-ranging species being more likely to be encountered earlier and the longer a species has been described, the more records it will accumulate. However, this hypothesis does not explain G. caeciliae's > 18,000 km range as the majority of recordings came after Pettibone's 1991 re-description. Before Pettibone's re-description, G. caeciliae was known from the Eastern Atlantic (Fauvel 1913;Hartmann-Schroder 1985) and the Indian Ocean (Stock, 1986). Pettibone (1991) expanded the range to the Western Atlantic and the Caribbean. The remaining records (60%) have all been in the last 15 years and did not expand the range any further (Eckelbarger et al., 2005;Simpson and Watling 2011;Britayev et al., 2014;Tu et al., 2015;Macpherson et al., 2016;Serpetti et al., 2017). Pettibone's more detailed description would have theoretically allowed for a more precise identification and this, along with increased exploration, would explain the subsequent increase in recordings.
The taxonomic description of G. caeciliae has fallen into some of the known problems related to marine annelid descriptions: poor early descriptions; composite generic re-descriptions based on two or more geographically distant specimens; and an assumption of wide variation of characters within species, discussed in Hutchings and Kupriyanova (2018). Fauvel's 1913 description was based on two specimens from the Eastern Atlantic. When the species was re-described by Pettibone (1991) the non-type specimens examined and illustrated showed two different morphologies from either side of the Atlantic (Pettibone 1991, Figs. 12, 13, 14); this expanded the species range. However, the written description was a composite, making no mention of the morphological differences observed.
Given the potential that the descriptive taxonomic history based on morphology for G. caeciliae may not represent the true taxonomy, the aims of this study were: (i) to utilize modern molecular techniques to analyse the genetic variation/divergence in G. caeciliae collected from both the eastern and western basins of the Atlantic; and (ii) to carry out detailed morphological analysis, in tandem with molecular analysis, to determine if morphological characters could define any new species boundaries found. This study will add to the few wide-scale morphological and genetic investigations into deep-sea commensal marine annelids (e.g., Hatch et al., 2020;Lindgren et al., 2019). The results will increase the knowledge of deep-sea marine annelids and contribute to a better understanding of biodiversity patterns and biogeography of deep-sea fauna.

Specimen collection
A total of 82 specimens of Gorgoniapolynoe were examined. Sample collection was undertaken using ROV Isis, deployed from the RRS James Cook during the JC094 cruise in October/November 2013. Sampling locations were all between 05 • N and 15 • N in the central Equatorial Atlantic ( Fig. 1; Table 1), on either side of the Mid-Atlantic Ridge, with the seamounts Vayda and Vema being in the western basin and Carter and Knipovich in the eastern basin. The sampling depths ranged from 600 to 2340 m. All worms were found in association with deep-sea octocorals from the families Acanthogorgiidae, Primnoidae and Coralliidae (Order Alcyonacea). The specimens were removed from tunnels within the coral tissue, with multiple individuals recovered from each coral. Once removed, worms were preserved in 70% ethanol and stored at − 20 • C.
In addition, all the specimens examined by Pettibone for her 1991 redescription, which are part of the Smithsonian Institute's National Museum of Natural History collection, were also examined here. These specimens included Fauvel's original type material (USNM 80098) and a specimen from the type locality (USNM 133356).
Additionally, a specimen identified as Harmothoe cf. bathydomus found in association with an unidentified holothurian was collected at 1,012 m at the Cantabrian Sea (43 • 54.449 ′ N, 6 • 15.494 ′ W) in June 2017. The specimen was collected on board the B/O Á ngeles Alvariño using a Beam trawl as part of the SponGES project. The polynoid was preserved in 96% ethanol and stored at − 20 • C.

DNA extraction, amplification and sequencing
DNA was extracted from 82 specimens, using approximately 20 segments from the posterior of the body. DNA extraction was done using a DNeasy® Blood and Tissue kit (QIAGEN, Germany), following the manufacturer's protocol except for the final elution stage, where a final elution in 75 μL of Buffer AE was used.
The Folmer region (Folmer et al., 1994) of the mitochondrial Cytochrome c oxidase subunit I (COI) gene was amplified for these 82 specimens. These specimens represented individuals from each location and from multiple hosts (Table 1). Additionally, a fragment of the ribosomal mitochondrial gene 16S rDNA (16S) and the nuclear ribosomal genes 18S rDNA (18S) and 28S rDNA (28S) were amplified for five specimens (Table 1). These five specimens were selected based on genetic divergence inferred from analysis of the COI data and from features identified during morphological analyses. PCR protocols and primers are presented in Table 2  GelRed® (Biotium, USA) was used to stain the PCR products, which were visualised using 1.5% agarose gel electrophoresis. The PCR  Table 1.

Table 1
Summary of specimens Gorgoniapolynoe whose morphology was examined and for which COI marker was amplified, as well as specimens, including G. caeciliae type specimen, from National Museum of Natural History (USNM) and the additional COI markers sourced from GenBank. Parent ID of host corals, coral host identification, sampling depth, coordinates, specimen numbers (prefix) and number of individuals, are also indicated. *Indicates that one individual was also sequenced for 16S, 18S and 28S; + indicates that one individual was used for SEM, # indicates that one individual was used for histological analysis.  Kearse et al., 2012). All consensus sequences were run through BLAST (Altschul et al., 1990) to check for contamination.

Species delimitation analysis
COI samples included in this analysis are found in Table 1 and in Supplementary Material Table S1. Antipathypolyeunoa sp. (GenBank accession number KU738202, supplementary material, Table S2) was used as an outgroup. Sequences were aligned in Geneious using the inbuilt MAFFT v.7.309 program (Katoh and Standley, 2013) and Q-INS-I option. The best partition schemes (including codon positions for protein-coding genes) and associated substitution models under AICc criterion were evaluated with PartitionFinder (Lanfear et al., 2016), using the greedy algorithm (Lanfear et al., 2012).
Once aligned and trimmed, COI sequences were 666 base pairs (bp) long. The sequences were checked for stop codons manually using the translate function (DNA to protein) in MEGA-X 10.0.5 (Kumar et al., 2018). Phylogenetic analyses were carried out using the model-based approach of maximum likelihood (ML), implemented using RAxML v8.2.12 (Stamatakis, 2014). RAxML was run using the GTR + GAMMA + I for all partition schemes as determined by PartitionFinder. The multiple tree search consisted of 100 alternative runs and the multiparametric bootstrap analysis had 1000 iterations.
Automatic Barcode Gap Discovery (ABGD, Puillandre et al., 2012) and a Bayesian implementation of the Poisson tree processes model (bPTP, Zhang et al., 2013) were used on the COI haplotype data to delimit species. ABGD was run online at https://bioinfo.mnhn.fr/abi /public/abgd/abgdweb.html, using the aligned COI sequences as the input. The program relies on the inclusion of the priors, P max and P min , where P is the divergence of intraspecific diversity. P min was set at 0.001 and P max at 0.37. These values were based on the range of intraspecific distances calculated for Polychaeta as a whole (Kvist, 2016). Twenty steps were run with a relative gap width (X) of 1.5. The number of bins for distance distribution was set to 50. Kimura (K80) model with TS/TV = 2 was used to calculate distance.
The bPTP model was run online at https://species.h-its.org/. The input file was the resulting tree from the COI RAxML analysis, converted into Newick format in Figtree v1.4.3 (Rambaut and Drummond, 2012). A haplotype data file was used to remove identical sequences, and only specimens collected for this study and the sequences identified as G. caeciliae from GenBank, were used for the bPTP model. The number of MCMC generations was set at 500,000, with a thinning of 100 and a burn-in of 0.25. The outgroup was removed during this analysis. The mean distance between the resulting species was calculated using MEGA-X 10.0.5 (Kumar et al., 2018)

Phylogenetic analysis
Phylogenetic analyses were conducted using five specimens of Gorgoniapolynoe from our dataset and a selection of sequences from individuals of Gorgoniapolynoe found online at the NCBI within the family Polynoidae (Supplementary material, Table S1). The five specimens from our dataset were selected to represent the genetic variation seen in the COI tree constructed for species delimitation. Additional material identified as Harmothoe cf. bathydomus from the Cantabrian Sea was also sequenced and included in the analysis. All alignments were manually trimmed and concatenated in Geneious. Gblocks v.0.91b (Castresana, 2000) was run separately for the alignments of the non-coding genes (16S, 18S and 28S). Gblocks was used to clean poorly aligned positions and to eliminate divergent regions, following previous papers for comparative purposes (Gonzalez et al. 2018;Serpetti et al. 2017;Taboada et al. 2020a). Given that the data set has individuals belonging to different families within the order Aphroditiformia, without the use of Gblocks both methods did not return meaningful trees. The "minimum length of a block" was set at 5, "allowed gap positions" was set at "with half", "maximum number of contiguous non-conserved positions" was set at 10 and finally the "minimum number of sequences for a flanked position" was set at 2 n + 1, where n = total number of sequences. Once each marker was aligned, trimmed and run through Gblocks, they were concatenated, resulting in a 3,553 bp length alignment (16S = 438 bp, 18S = 1,648 bp, 28S = 900 bp and COI = 567 bp).
RAxML only allows for a single model of rate heterogeneity in partitioned analysis, so GTR + I + G was chosen as it was suggested for four out of the six partitions and had a low AICc score for the remaining two partitions. The other parameters were set as previously described for COI. In MrBayes, the Monte Carlo Markov Chains (MCMC) were run for ten million generations, with trees sampled every thousand generations using the following evolutionary models for each partition, 16S, 18S, 28S and COI (3rd position) = GTR + I + G, COI (1st position) = HKY + G and COI (2nd position) = GTR. Across the partitions, the parameters of substitution rates, nucleotide frequencies, invariant-sites proportion and gamma shape were all unlinked. Burn-in was set at 25%. Tracer v1.7.1 (Rambaut et al., 2018) was used to check for convergence. Analysis was stopped once the standard deviation of split frequencies was <0.01, the potential scale reduction factor PSRF was around 1.00, the effective sample sizes were >200 and the trace plot had the appearance of a "hairy caterpillar".

Demographic analyses
The COI alignment of Gorgoniapolynoe from the central Equatorial Atlantic was used to construct unrooted haplotype networks with the program PopART (Leigh and Bryant, 2015) by implementing the Templeton, Crandall and Sing (TCS) method (Templeton et al., 1992;Bandelt et al., 1994).
The levels of DNA polymorphism, polymorphic and parsimonyinformative sites were calculated for each sampling region and lineage using DNASP vs. 5.10.1 (Rozas et al., 2017) and included the number of haplotypes (H), private haplotypes (Hp), haplotype diversity (Hd) and nucleotide diversity (π).
An analysis of molecular variance (AMOVA) was used to determine differentiation between areas and hosts for each of the species identified in the morphological and phylogenetic analyses. AMOVA was run using Arlequin v3.5.2.2 (Excoffier and Lischer, 2010) under the hierarchy of location (seamount) > population (host). F-statistics were utilised to  Struck et al. (2006) estimate the proportion of variability found between locations (F ct ), among populations within locations (F sc ) and within populations (F st ). Data was grouped into depth bins (500-1000 m, 1000-1500 m, 1500-2000 m and >2000 m) and an AMOVA was also run to determine if depth was a factor in any genetic variation, under the hierarchy of depth > host.

Morphological analysis
The macroscopic morphological characters of all collected specimens were examined using a Leica MZ6 stereomicroscope (Leica Microsystems, Germany). Parapodia and elytra were removed from selected specimens for further examination under an Olympus BX43 compound microscope (Olympus Corporation, Japan). Specimens were photographed using an Olympus UC50 camera and the cellSens Standard interface (Olympus Corporation, Japan) for both microscopes.
The syntype of Gorgoniapolynoe caeciliae (USNM 80098), designated by Fauvel in 1913, was reexamined, as well as all the material examined by Pettibone (1991 ; Table 1) which included specimen (USNM 133356) from the type locality. All literature on G. caeciliae, from the original description by Fauvel (1913) to the most recent work by Britayev et al. (2014) was used as references throughout the morphological analysis.
Four specimens were selected for analysis using a scanning electron microscope (SEM) ( Table 1). In order to assess if any genetic variability corresponded with differences in morphology, specimens were chosen based on their positioning within the COI tree. Specimens were dehydrated in an ascending ethanol series from 70 to 100%, critical-pointdried in a Balzers CPD-030 dryer (Bal-Tec AG, Liechtenstein), mounted on stubs and coated with gold (20 nm). Imaging was performed using Zeiss Ultra Plus Field Emission SEM (Zeiss Group, German) in the NHMUK imaging facilities.

Histological preparations
Two specimens were selected for histological analysis. One specimen was selected from each of the two species (G. cf caeciliae and G. pseudocaeciliae sp. nov.) observed as a result of the morphological and phylogenetic analyses, with the criteria for selection being the presence of a pair of modified first elytra and the subsequent smaller unmodified second pair. The anterior portion of each two specimens (prostomium and following five to six segments) was dehydrated in an increasing ethanol series (50%, 70%, 96% and 100%), cleared in xylene and embedded in melted paraffin, then sliced into 5 μm sections using an Autocut Reichert-Jung microtome 2040 (R. Jung GmbH, Nubloch, Germany), stained with haematoxylin-eosin and mounted with DPX. Prepared slides were investigated using Olympus BX43 compound microscope (Olympus Corporation, Japan) and imaged using an Olympus UC50 camera and the cellSens Standard interface (Olympus Corporation, Japan).

Species delimitation
Both the ABGD and bPTP (Supplementary Material Fig. S1) analyses supported Gorgoniapolynoe pseudocaeciliae sp. nov. as a single species distinct from all other groups in the dataset. ABGD delimited the group as a single species for all values of intraspecific divergence (P) up to 0.0307 (3.07%) (Supplementary Material Fig. S2). Gorgoniapolynoe pseudocaeciliae sp. nov could also be distinguished from G. cf. caeciliae using morphological characters. Gorgoniapolynoe pseudocaeciliae sp. nov has a smaller modified area on the first elytra, a more pronounced and digiform neuropodial postsetal lobe and none of the notochaeta have spinous pockets (see section 3.2.) An intraspecific divergence range of P = 0.001-0.0225 (0.1-2.25%) in ABGD returned two apparent molecular operational taxonomic units (MOTUs) within G. cf. caeciliae, despite no apparent morphological difference (supplementary material, Fig. S2). The bPTP model also inferred the presence of two species within G. cf. caeciliae (supplementary material, Fig. S1) At P = 3.07% ABGD no longer delimited G. cf. caeciliae into two MOTUs, while at P = 4.19% ABGD returned the complete data set as a single species (supplementary material, Fig. S2). Both ABGD and bPTP infer that Gorgoniapolynoe caeciliae from the Indian Ocean obtained from GenBank (Serpetti et al., 2017) are a distinct species. Henceforth within this text, they will be referred to as Gorgoniapolynoe 'Indian Ocean'. The ABGD model delimited all individuals labelled as G. corralophila as a single distinct species.
The mean intraspecific distance between the different delimited species ranged from 7 to 15%, with the smallest distance being between the two MOTUs in G. cf. caeciliae (supplementary material, Table S2).

Haplotype network analyses
A total of 26 haplotypes were recovered out of the 33 individuals identified as Gorgoniapolynoe cf. caeciliae (Fig. 2). The haplotype network of the larger MOTU in G. cf. caeciliae (Unit 1) recovered a diffuse topology, with 22 haplotypes from 27 individuals. The haplotype diversity (Hd) in G. cf. caeciliae Unit 1 was 0.98 ± 0.0003 and all the haplotypes, except one, were private ( Table 3). Out of the 22 Unit 1 haplotypes, only one was recovered from both seamounts. The six specimens of the smaller G. cf. caeciliae MOTU (Unit 2) had four haplotypes ( Fig. 2), with a Hd of 0.867 ± 0.129, all of which were found on a single host from Vema. The nucleotide diversity (π) for G. cf. caeciliae Unit 1 and Unit 2, was 0.01512 ± 0.00189 and 0.00494 ± 0.00119, respectively (Table 3) and the number of mutations separating them was 17.
For Gorgoniapolynoe pseudocaeciliae sp. nov. the haplotype network topology was also diffuse, returning 22 haplotypes out of the 49 individuals (Fig. 2), with a Hd = 0.929 ± 0.02. There were 16 private haplotypes. The nucleotide diversity was 0.00977 ± 0.000001 (Table 3). Three haplotypes were shared across all three seamounts from which G. pseudocaeciliae were collected, Carter and Knipovich in the east and Vayda in the west. These three shared haplotypes were found across the entire sampled depth range, ≈700-2,340 m (Fig. 2). There were three other haplotypes shared between seamounts, but these were only observed for the East Atlantic seamounts, separated by approximately 600 km and found between 1,360 and 2,340 m.

AMOVA
AMOVA results for G. pseudocaeciliae sp. nov. showed no significant variation between seamounts or between coral hosts. There was significant variation on individual hosts (Table 4). Depth was found not to be a significant factor to genetic variation in G. pseudocaeciliae sp. nov. (Table 4). The results for G. cf. caeciliae Unit 1 also showed no significant variation between seamounts, between hosts, or between depth bins, with the only significant variation found to be on individual host corals. No analysis was run on Gorgoniapolynoe cf. caeciliae Unit 2, as it was only found at Vema on a single host.

Description
Body ranging in length from 11 to 22 mm, 2-2.7 mm wide (inclusive of chaetae), with up to 52 segments. Body almost cylindrical anteriorly, becoming progressively more dorso-ventrally flattened from segments 15-20 (Fig. 4A). Prostomium bilobed, with rounded lobes without cephalic peaks, wider than long. Two pairs of large eyes, with anterior pair generally larger. Median antenna long and tapering, attached to ceratophore in the notch between prostomium lobes. Lateral antennae subulate and approximately a third the length of the median antenna. Ceratophores of the lateral antennae robust and cylindrical, slightly wider in diameter than antennae, inserted lateroventrally, slightly separated from the median antenna. Two stout palps dorsoventrally to the proboscis, as long as the median antenna ( Fig. 4A and B). Robust tentaculophores lateral to the prostomium on the achaetous first segment. A pair of dorsal and ventral tentacular cirri. Dorsal cirri as long/longer than the median antenna. Ventral pair slightly shorter than the dorsal pair. Segment 2 with the first pair of elytrophores and the first biramous parapodia. Subulate buccal cirri emerge from short cylindrical cirrophores on the anterior ventral surface of the neuropodia, of similar length to the dorsal tentacular cirri.
A total of 15 pairs of elytra on segments 2, 4, 5, 7, then every second segment until 23, after which on segments 26, 29 and 32. First pair of elytra, larger than the subsequent ones, covering the prostomium. A large bean-shaped transparent chitinous area covers over half the elytra area. Chitinous area contains scattered rounded microtubercles and more regularly spaced elongated, barrel-shaped macropapillae (Fig. 4B, C, D). Following smaller elytra oval with down-curved borders and with barrel-shaped macropapillae randomly and sparsely scattered across the dorsal surface. Area of the elytra above elytrophore darker in colour.

Table 3
Nucleotide and haplotype information for each Gorgoniapolynoe species from this study. Data is presented for all locations combined and for each individual seamount. N = number of individuals, Pi = nucleotide diversity, h = number of haplotypes, Hd = Haplotype diversity, Ph = Private haplotypes, %Ph = percentage of haplotypes which are private. *private haplotypes for combined seamounts refer to haplotypes found on a single host, Ph of individual seamounts refers to haplotypes unique to that seamount. SD = Standard deviation + Gorgoniapolynoe cf. caeciliae MOTU 2 was collected from a single host coral on Vema so has no private haplotypes.  The results of analysis of molecular variance (AMOVA) on the COI data for Gorgoniapolynoe from the Equatorial North Atlantic. Species were determined using morphology and species delimitation models ABGD and bPTP. Depth bins were 500-1000 m, 1000-1500 m, 1500-2000 m and >2000 m. * marks significant variation (p < 0.05). Very faint nerves present in elytra, only visible under microscope. Paired elytra do not touch leaving the mid-dorsum exposed. Emerging from tips of all macropapillae of all elytra are 3-4 cilia, that appear to be retractable wit yond the elytra boundary. Cilia have golf club-shaped structures at the distal end that are an artefact of preservation and are called "paddle cilia", AKA "discocilia" (Fig. 4F). Dorsal cirri present on segments without elytra around three times as long as parapodia; sporadic clavate papillae present on proximal ends.
Cylindrical cirrophores along posterior side of notopodium. Starting on the third segment ventral cirri present on all segments. Base of ventral cirri slightly inflated before tapering, reach past the tip of neuropodium. From segment 10, bulbous, swollen area on base of ventral cirri causing them to project posteriorly. Transverse bands of short cilia present dorsally on random segments along the body. Posterior most segments without elytra have transverse ciliary bands with cilia over three times as long, more densely packed than is seen on any other segments (see Fig. 3. Phylogenetic tree of Aphroditiformia, recovered from analysis of the concatenated sequences of 16S, 18S, 28S and COI. The tree topology is based on Bayesian Inference analysis. Node labels are the posterior probability from BI analysis, followed by the bootstrap support from Maximum likelihood analysis. Families within Aphroditiformia and subfamilies within Polynoidae are colour coded. Individuals from the genus Gorgoniapolynoe are highlighted. All species in bold were sequenced for this study. See Supplementary Table S1 for NCBI accession numbers. (Colour should be used).  Parapodia biramous. Notopodium is a subconical acicular lobe. Neuropodium longer and wider with presetal and postsetal lobes. Neuropodial presetal lobe diagonally truncated, projecting dorsally. A shorter flattened postsetal lobe lays close to neurochaetae; rounded, projecting slightly ventrally in anterior segments so that it can just be seen when viewed from anterior (Fig. 4G). On posterior segments the postsetal lobe becomes reduced to a small bump roughly in line with the neuroacicula, no longer visible when viewed from the anterior. Reduction in the size of postsetal lobe begins approximately a third of the way down the body and is at its most reduced by the final third of the body. Notochaetae number between 3 and 7 on the first 10-15 segments (Fig. 4H), reduced to 2-4 after that. Two types of notochaetae present on anterior segments: stout acicular and slightly curved chaeta, with the groove on outside of the curve (Fig. 4H), and also stout, acicular and curved but with alternating spinous pockets from the apex of the curve to the tip ( Fig. 4H and I). Both types of notochaetae positioned anterior to the notopodium and longer than it. Neurochaetae number between 10 and 15, as stout as notochaetae. Last third becomes quill-like, with spinose rows on dorsal surface. All neurochaetae tips bidentate to varying degrees and change in form moving dorsal to ventral. Supraacicular neurochaetae have a large blunt tooth with a smaller rounded secondary tooth (Fig. 4J). Two forms of infraacicular neurochaetae present: the set closest to the supraacicular neurochaetae with a large pointed tooth slightly hooked and a secondary smaller tooth slightly more distinct than those in supraacicular neurochaetae (Fig. 4K); the most ventral infraacicular neurochaetae have a large blunt rounded tooth with a highly reduced secondary tooth barely more than a bump (Fig. 4L).
Pygidium is a blunt cone, with a pair of anal cirri at the terminus. Bases of the anal cirri closely situated almost touching, positioned slightly dorsally on the pygidium. Anal cirri as long or slightly longer than the dorsal cirri. Anus ventral to the cirri, protruding slightly.

Histological examination
Histological sections reveal darker area of smaller elytra correspond to presence of cup-like cells, possible photocytes (Fig. 4E).

Remarks
The morphology of G. cf. caeciliae parapodia appears to match the parapodial syntypes of Gorgoniapolynoe caeciliae (USNM 80098), most notably in the shape and orientation of the postsetal lobe. The morphology of the rest of the body also matches that of Fauvel's original written description, although the description is extremely limited. It cannot be said with confidence that the specimens examined in this study are G. caeciliae.
The morphological characters of G. cf. caeciliae are the same as the specimens examined by Pettibone (1991) for her redescription of the species. The presence of 3-7 notochaetae on anterior segments, observed on the Equatorial Atlantic specimens, were not mentioned or illustrated in Pettibone (1991). On re-examination of USNM 21123 (two specimens), we observed 4-6 notochaetae on segments three, five, six, eight of one specimen and segments three, five and ten on the other. The newly observed notochaetae in G. caeciliae are similar in appearance to the notochaetae seen in G. corralophila (Day, 1960), however, in the latter species they are much larger and can be easily observed under low magnification. Additionally, the specimens USNM 80090, 80091 and 133357, all of which were examined by Pettibone for her description, agree with the morphological features observed for G. cf. caeciliae.
Gorgoniapolynoe cf. caeciliae contain morphologically identical but genetically diverged individuals, which may be a cryptic species. The analysis of the COI sequences revealed that six of the 43 individuals identified as G. cf. caeciliae were genetically distinct and were inferred to be a separate species: G. cf. caeciliae Unit 1 and 2, under ABCD and bPTP models (see section 3.1.1.) As the number of specimens is so low, no morphological differences could be found and the uncertain taxonomic affinity of the specimens in the first place, we prefer to be conservative and stop short of dividing it into two species. It may be that G. cf. caeciliae has high intraspecific genetic variability, hence why we have referred to them as MOTUs.
Body almost cylindrical anteriorly, becoming progressively more dorso-ventrally flattened after approximately segment 13 (Fig. 5A). Bilobed prostomium with rounded lobes without cephalic peaks. Prostomium wider than long. Two pairs of large eyes present, with the anterior-most pair generally larger. A median antenna present, long and tapering, attached to ceratophore in a notch between prostomium lobes. A pair of short subulate lateral antennae, approximately a third the length of the median antenna inserted lateroventrally. Ceratophores of the lateral antennae cylindrical and slightly wider than the antennae. Two stout palps lateral to the lateral antenna as long as the median antenna. Robust tentaculophores present lateral to the prostomium on the achaetous first segment. Pairs of dorsal and ventral tentacular cirri. Dorsal tentacular cirri as long as the median antenna, with the ventral tentacular cirri slightly shorter than dorsal ones. Segment 2 with the first pair of elytrophores, ventral buccal cirri and biramous parapodia. Subulate buccal cirri the same length as dorsal tentacular cirri, attached at short cylindrical cirrophores. Cirrophores on anterior ventral surface of neuropodia.
Fifteen pairs of elytra on segments 2, 4, 5, 7, then every second segment until 23, after which on segments 26, 29 and 32. First pair of elytra large, covering the prostomium, modified with a crescent-shaped, transparent, chitinous area. Within the chitinous area there are irregularly spaced rounded macrotubercles and more regularly spaced elongated, cylindrical macropapillae ( Fig. 5B and C). Remaining elytra oval, (caption on next page) smaller than the first pair, with smooth borders curved downwards. Elytra pairs do not touch, leaving the mid-dorsum exposed. Area of the elytra above the elytrophore darker, the rest of the elytra is opaque. Vein-like pattern branching out from the darker area of the elytra (visible under stereoscope). All elytra with randomly and sparsely scattered cylindrical macropapillae across their dorsal surface. Emerging from tips of all macropapillae observed are 3-4 cilia-like structures. These cilia appear to be retractable and can be as long as the width of the elytra. Tips of the cilia flattened and ovoid, these are an artefact of preservation i.e., paddle cilia.
Dorsal cirri present on segments without elytra, long and extending past tips of parapodia. Cirrophores of the dorsal cirri cylindrical, projecting from the posterior side of notopodium. Ventral cirri start from segment 3, with a cylindrical cirrophore. Base of the ventral cirri slightly inflated before tapering, reaching past the tip of the neuropodium. Transverse bands of short cilia present dorsally on random segments. Last 10 segments without elytra with ciliary bands on the dorsal side with cirri, longer and more densely packed compared to other segments ( Fig. 5G and H). Small, digitate nephridial papillae from segment six, posterior-ventrally positioned on parapodium. Present on all segments past six but highly reduced in the posterior.
Parapodia biramous, with a subconical acicular notopodium and a neuropodium with a presetal and a postsetal lobe. Presetal lobe of neuropodium diagonally truncated, projecting dorsally. Postsetal lobe shorter, rounded and digitiform, projecting ventrally and away from the neurochaetae ( Fig. 5I and J). Projection of pre-and postsetal lobes cause neuropodium to appear bilobed when viewed anteriorly. Postsetal lobe reduced roughly a third of the way down the body. As the postsetal lobe becomes reduced it also takes on the same orientation as the presetal lobe, meaning it can no longer be seen when parapodia viewed from the anterior. There are 1-3 notochaetae, stout, acicular and slightly curved with a small groove on the dorsal side of the curve. Notochaetae positioned anterior to notopodium, longer than it (Fig. 5K). There are 10-15 neurochaetae, as stout as notochaetae with the last third or so becoming quill-like and spinulose. Neurochaetae tips bidentate to varying degrees, with a slight change in form moving from dorsal to ventral. Supraacicular neurochaetae have a slightly hooked, pointed larger tooth and a smaller, rounded secondary tooth (Fig. 5L). Most dorsal of the infraacicular neurochaetae have a more prominent secondary tooth, more pointed than that of supraacicular neurochaetae (Fig. 5M). Ventral most infraacicular neurochaetae more rounded larger tooth with a reduced secondary tooth (Fig. 5N).
Pygidium is a blunt cone, with a pair of anal cirri at the terminus. Bases of the anal cirri closely situated, almost touching. Anal cirri as long or slightly longer than dorsal cirri. Anus ventral to the cirri, protruding slightly.

Histological examination
Histological sections reveal that the darker area corresponds to presence of cup-like cells, possible photocytes (Fig. 5E).

Etymology
The species is named G. pseudocaeciliae due to its morphological resemblance to G. caeciliae.

Distribution
Gorgoniapolynoe pseudocaeciliae sp. nov. has been reported from Portuguese waters (Pettibone 1991) and in the central Atlantic, on both the east and west side of the Mid-Atlantic Ridge. It has a depth range of between 710 and 2,343 m.

Remarks
The distinct digit-like postsetal lobe on the anterior neuropodium, which projects ventrally to an extent that it can clearly be seen when viewing the neuropodium from the anterior is the character that distinguishes Gorgoniapolynoe pseudocaeciliae sp. nov. from G. caeciliae (Fig. 3J). The parapodial syntype of G. caeciliae (USNM 80098), has a shorter flattened postsetal lobe that lies close to neurochaetae. It is rounded and projects slightly ventrally in anterior segments so that it can just be seen when viewed from anterior. The parapodial syntype was taken from the anterior of the body, so a direct comparison could be made.
Gorgoniapolynoe pseudocaeciliae sp. nov. was previously illustrated by Pettibone (1991) as a morphotype of Gorgoniapolynoe caeciliae (USNM 133356). This current study reexamined USNM 133356 and the morphological features concur with the Gorgoniapolynoe pseudocaeciliae sp. nov collected for this study. The distinguishing characters that separate this new species from of G. cf. caeciliae are: (1) the presence of only one type of robust notochaetae, which is curved and smooth; (2) a distinct digit-like postsetal lobe on the anterior neuropodium, which projects ventrally to an extent that it can clearly be seen when viewing the neuropodium from the anterior; and (3) a small, crescent-shaped chitinous area on the modified first elytra pair, that takes up approximately a quarter of the total area of the elytra, as opposed to the one in G. cf. caeciliae, which is significantly larger. Molecular analyses presented above also support G. pseudocaeciliae sp. nov as a distinct species from G. cf. caeciliae. Pettibone's (1991) redescription When Pettibone (1991) redescribed Gorgoniapolynoe caeciliae she examined type material USNM 80098, as well as a single individual from the type locality and several specimens from the Western Atlantic. For this current study, all this material was re-examined, along with Fauvel's (1913Fauvel's ( , 1914 original description and illustrations. The parapodial morphology of the syntype USNM80098 matches that of Pettibone's Western Atlantic specimens but not the individual from the type locality. The modified first elytra were never mentioned in any of the literature until Pettibone (1991), so we do not know what characteristics the type material had. Additionally, the original descriptions lack detail, however, what little detail within them also matches the characters of Pettibone's Western Atlantic specimens. The parapodia of specimens designated as G. cf. caeciliae in this current study match those of Fauvel's syntype and those of the Western Atlantic specimens examined by Pettibone.

Addressing
The error in Pettibone (1991) was the designation of the Eastern Atlantic specimen (USNM 133356) as G. caeciliae. Pettibone clearly recognised the morphological differences between USNM 133356 and the Western Atlantic specimens, as she chose to illustrate them but for some reason never addressed this in the text. The morphology of the specimens designated as G. pseudocaeciliae sp. nov. in this current study match that of USNM 133356. There is no molecular data available for the syntype. Given this evidence, USNM 133356 should be redesignated as the holotype of G. pseudocaeciliae.

Gorgoniapolynoe hidden species diversity
Our study uncovered hidden species diversity within the genus Gorgoniapolynoe and increases the number of species from eight to nine. Differences in the parapodia morphology of G. pseudocaeciliae sp. nov. and Fauvels's syntype differentiate it as a distinct and separate species from G. caeciliae. (Fig. 5; see remarks in section 3.2). There is also the potential for a further two new Gorgoniapolynoe species. Further indetail morphological investigation is needed into the individuals known as Gorgoniapolynoe 'Indian Ocean' to determine what characters delimitate them from other species in this genus. Additionally, the specimens designated here as G. cf. caeciliae may represent another new species, however given the problems with the current description for G. caeciliae, confirmation or refutation of this is impossible without a detailed re-examination of the species as a whole. Furthermore, there is a high genetic diversity within G. cf. caeciliae, to the extent that it may be a cryptic species complex comprising of two morphologically nondifferentiable species i.e., G. cf. caeciliae MOTU 1 and MOTU 2. While no morphological differences were observed it could be possible that colouration in the living species may discriminate between the two MOTUs. This has shown to be the case in other marine annelids (Nygren and Pleijel, 2011;Pleijel et al., 2009). As all specimens examined here had lost their colouration due to preservation it was impossible to tell if this was the case and highlights the need to image and record colouration at time of collection. More genetic data from additional individuals is needed to determine if what is observed in this study was intraspecific variation or if G. cf. caeciliae is in fact a cryptic species complex.
According to Kvist (2016) there is an intraspecific distance range of 0-37% of the COI gene across all marine annelids, however, the median genetic distance was 0.79%. Kvist acknowledged that the wide intraspecific distance found may be a result of wrongly identified or undiscovered cryptic species within Genbank, an opinion with which we concur. Other annelid taxa studies have found intraspecific divergence of COI to be less than 2% (Brasier et al., 2016;Carr et al., 2011;Drake et al., 2007;Nygren et al., 2018). The interspecific distances in species of the genus Gorgoniapolynoe were between 10 and 15% for all pairwise comparisons, which are similar to the distances recorded by Brasier et al. (2016) and Nygren et al. (2018). The mean distance between G. caeciliae MOTU 1 and MOTU 2 of 7% is high for intraspecific distance but also is low for interspecific distance based on the aforementioned studies, hence our hesitancy to divide them into two different species without more data.
Gorgoniapolynoe corralophila is distinct within the genus as it is the only species described which has three pairs of anterior modified elytra (Pettibone 1991). It also has distinct large notochaeta with spinose rows. These differences were enough for Britayev et al. (2014) to recommend that the species be moved into its own genus. The positioning of G. corralophila within the phylogenetic tree presented here (Fig. 3), along with the discovery of the similar (although much smaller) notochaeta in the G. cf. caeciliae complex would suggest that G. corralophila is correctly placed within this genus.

Novel morphological features in Gorgoniapolynoe
On the elytra of all the specimens analysed using SEM, we observed macropapillae bearing 3-4 cilia-like structures with a distal end shaped like a golf club (Figs. 4F and 5F) The shape of the distal end of the cilia is an artefact of preservation. They are referred to as "paddle cilia" or "discocilia" and are believed to form when the fixation method causes the axoneme to coil within a distal expansion of the ciliary membrane (Short and Tamm, 1991). Some of these cilia were observed to be many times longer than the papillae from which they emerged, longer than the width of the elytra in some cases and potentially retractable (Figs. 4F and 5F). To our knowledge, these cilia emerging from the elytra have not been observed on any other member of Gorgoniapolynoe. The only other record of similar structures in a polynoid is Pholoe minuta (Fabricius, 1780), a member of the Sigalionidae (subfamily Pholoinae) in the suborder Aphroditiformia, where cilia-like structures with basal collars were described on the tips of their elytral papillae (Heffernan, 1990). The cilia-like structures of P. minuta were hypothesized to be involved in respiration and sensory function. Similarly, we suggest that the cilia in Gorgoniapolynoe might have a function in respiration. Segrove (1938) observed and documented epidermal ciliated bands on several different marine annelids. These bands of cilia were observed to produce currents over the surface of the body and were hypothesized to aid respiration via gas exchange through the epidermis. Lwebuga-Mukasa (1970) similarly observed ciliated bands on the polynoid Halosydna brevisetosa Kinberg, 1856 and agreed with Segrove (1938) that they probably performed a function in respiration. Cilia on the parapodia create a lateral flow that draws water in under the elytra, where bands on the dorsal surface of the animal further direct this flow posteriorly. No epidermal cilia were observed on the parapodia of Gorgoniapolynoe, although some were observed in bands across the dorsal side of elytra-free segments. The cilia emerging from the elytra may provide a similar function to those observed in other species in keeping water circulating within their tunnels.
Gorgoniapolynoe tunnels are open-ended and the higher density of ciliated papillae on the first elytra could be to create an initial inflow current, while the epidermal ciliated bands observed on the posterior segments may create an outflow. Circulating water through the tunnel in this way would help cutaneous respiration and also function in keeping the tunnel clean, a secondary function suggested by Segrove (1938) and Lwebuga-Mukasa (1970).
Alternatively, the macropapillae with the cilium structures might be involved in the emittance of bioluminescence in combination with the photocytes we detected in our histological analysis (Figs. 2E and 3E). Bioluminescence produced by photocytes has already been described in other polynoids (Nicol 1953;Bassot and Nicolas 1995;Kirkegaard 2001;Taboada et al., 2020a) and macrotubercles present on the elytra could act as refractory prisms thus amplifying the luminescent effect (Nicol 1953;Bassot and Nicolas 1995;Plyuscheva and Martin, 2009). Bioluminescence in polynoids has been suggested to be used as a warning or distracting mechanism allowing the animal to escape when attacked by a potential predator (Nicol, 1953;Plyuscheva and Martín, 2009), but more recently, it has also been suggested to be used by Neopolynoe chondrocladiae (Fauvel, 1943) as a lure for prey (Taboada et al. 2020a). Neopolynoe chondrocladiae is an obligate symbiont of two deep-sea carnivorous sponges of the genus Chondrocladia that might use bioluminescence to increase chances of getting more preys, which would benefit both the symbiont and the host sponge (Taboada et al., 2020a;Taboada et al., 2020b). Similarly, we propose that bioluminescence in members of the genus Gorgoniapolynoe could be either linked to dissuading potential predators to both the host and the symbiont and/or to attract a greater amount of prey for both partners. In any case, further experiments should be conducted to confirm our hypothesis and also to investigate whether the host octocorals where Gorgoniapolynoe are found do have bioluminescent capabilities, as recently described for several deep-sea anthozoans from the coast of California (Bessho-Uehara et al., 2020).

Molecular connectivity and dispersal in Gorgoniapolynoe spp.
Our demographic results conducted on two species of Gorgoniapolynoe agreed in suggesting that there is a certain degree of molecular connectivity between locations separated hundreds to thousands of kilometres. This is especially true for the case of G. pseudocaeciliae sp.
nov., where we found three haplotypes shared between the East and West Atlantic seamounts, crossing the Mid-Atlantic Ridge (Fig. 2G). The Mid-Atlantic Ridge (MAR) has been shown to act as a barrier to dispersal and connectivity for some abyssal species (Bober et al., 2018;Boyle, 2011). However, for taxa that are upper bathyal or have highly mobile adults the barrier appears to be more permeable (Coscia et al., 2018;Shields et al., 2013;White et al., 2011;Zardus et al., 2006). Shields et al. (2013) found there to be a level of connectivity between populations of Eunoe bathydomus (Ditlevsen, 1917), a commensal polynoid associated with holothurian Deima validum Théel, 1879, on either side of the MAR. However, the populations sampled by Shields et al. (2013) were from along the axial trough of the ridge separated by less than 300 km, with east and west populations separated by less than 70 km. Their hosts are also highly mobile and may take long migrations between food-rich areas, thus facilitating gene flow in the polynoids. Gorgoniapolynoe are assumed to be broadcast spawners, with relatively high fecundity, with >3,000 eggs recorded in a single female (Eckelbarger et al., 2005). It is therefore assumed that they have a pelagic larval stage which could potentially facilitate the connectivity we report here.
The larval dispersal between the two basins may be aided by the presence of fracture zones that transverse the ridge, particularly the Vema Fracture Zone (VFZ), north of Vema Seamount and Fifteen-Twenty Fracture Zone FTFZ, north-east of Vayda seamount. There is an exchange of waters between the two basins through the VFZ, Antarctic Bottom Water flows eastward, above which North Atlantic Deep Water flows westward (Devey et al., 2018). These offset fractures may also aid in disrupting topographic steering, whereby strong currents are forced to run parallel to the ridge, carry larvae along the ridge and prevent them crossing over it (Young et al., 2008).
Connectivity between widely separated locations at similar bathymetric depths is common in the marine realm but individuals separated vertically by only 100s of metres can have limited or no connectivity (Etter et al., 2005;France and Kocher, 1996;Rex and Etter, 2010;Roy et al., 2012;Taylor and Roterman, 2017;Zardus et al., 2006). It is most likely that this differentiation by depth is a result of the strong vertical environmental gradients e.g. temperature and pressure (Rex and Etter, 2010).
The Gorgoniapolynoe in this study went against the common finding of depth being a strong structuring factor, with both species showing no significant structure with depth. In particular, the three largest shared haplotypes in G. pseudocaeciliae were found across a depth range of 1600 m (Fig. 3). It may be that having the ability to colonise multiple different host species has increased their odds of settlement across a wide depth range. Lattig et al. (2016) investigated the population structure of Iphitime cuenoti Fauvel (1914) (Dorvilleidae) a polyxenous symbiont associated with brachyuran crabs. The results revealed little genetic structure that was independent of their crab hosts. Having multiple hosts, across a 600 m depth range appears to facilitate gene flow and prevent structure. If along with multiple host species, Gorgoniapolynoe have a long pelagic larval duration and an ability to tolerate the environmental factors associated with depth, then this would explain the lack of structure associated with depth and distance and explain their wide geographic range in the Atlantic (Lester et al., 2007). The significant genetic variation found on individual hosts suggests that the individual worms are not closely related and that recruitment of worms to the host may be continuous.

Conclusion
The current investigation, using a combined morphological and molecular approach, has increased known diversity within Gorgoniapolynoe by describing a new species, Gorgoniapolynoe pseudocaeciliae sp. nov.. The molecular analysis has uncovered a potentially new species from the deep Indian Ocean previously identified as G. caeciliae (Serpetti et al., 2017), and has revealed a wide genetic diversity within G. cf. caeciliae from the Equatorial Atlantic. We suggest that a complete reinvestigation into the G. caeciliae as a whole is needed. We have uncovered connectivity across large distances and, unusually for the deep-sea, across a wide bathymetric range. Our study highlights again the need to utilize molecular methods in combination with detailed morphological and ecological investigations (Dayrat, 2005) to reinvestigate species with unusually wide ranges and/or having an unusually high number of known hosts, as was the case for G. caeciliae. Only with this integrative approach will taxonomy be able to uncover and describe the true diversity of deep-sea life.

Funding
This work was supported by the SponGES project (Grant agreement no. 679848) -European Union Framework Programme for Research and Innovation, H2020. Some of the data collection for this work was funded by the ERC Starting Grant projects CACH (grant no. 278705). ST received funding from the Juan de la Cierva-Incorporación program (IJCI-2017-33116), the Spanish Government and received a fellowship from the Systematics Research Fund (SRF) in 2018 and funding from the grant PID2020-117115 GA-100 funded by MCIN/AEI/10.13039/ 50110001103. JM had the support of the Erasmus+ programme of the European Union.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.