Monsters in the dark: systematics and biogeography of the stygobitic genus Godzillius (Crustacea: Remipedia) from the Lucayan Archipelago

Citation for published version (APA): Ballou, L., Iliffe, T. M., Kakuk, B., Gonzalez, B. C., Osborn, K. J., Worsaae, K., Meland, K., Broad, K., BrackenGrissom, H., & Olesen, J. (2021). Monsters in the dark: systematics and biogeography of the stygobitic genus Godzillius (Crustacea: Remipedia) from the Lucayan Archipelago. European Journal of Taxonomy, 751, 115139. https://doi.org/10.5852/ejt.2021.751.1383


Introduction
The crustacean class Remipedia is an enigmatic stygobitic group consisting of 29 species, 12 genera and eight families. Remipedes predominantly dwell within anchialine cave habitats (i.e., subterranean estuaries) (Bishop et al. 2015;Brankovits et al. 2017;van Hengstum et al. 2019). Like most anchialine fauna, remipedes exhibit a globally disjunct distribution, inhabiting submerged cave systems in the Caribbean, West Atlantic Ocean, Canary Islands and Western Australia (Koenemann & Iliffe 2014). A majority of remipedes (20 of 29 species) are reported from the Lucayan Archipelago (Bahamas and Turks and Caicos), suggesting a potential biodiversity hotspot for the group (Reid 1998). The karst dominated landscapes of these islands, as well as the presence of freshwater/saltwater mixing layers, provide optimal conditions for rapid dissolution and cave formation (Mylroie & Carew 1990;Mylroie & Mylroie 2011).
The clade Godzilliidae is one of four families endemic to the Lucayan Archipelago. Godzilliidae currently consists of two genera, Godzilliognomus Yager, 1989 andGodzillius Schram, Yager &Emerson, 1986. The family's name is attributed to the great size (43.2 mm) of the type species, Godzillius robustus Schram, Yager & Emerson, 1986, which is the largest observed remipede species to date (Schram et al. 1986). There are two previously described species within Godzillius: G. robustus and G. fuchsi Gonzalez, Singpiel & Schlagner, 2013. All members of Godzillius are found within the Lucayan Archipelago and are known to inhabit anchialine cave systems. Godzillius robustus occurs exclusively in Cottage Pond, North Caicos Island, Turks and Caicos Islands, while G. fuchsi inhabits the Dan's Cave and Ralph's Sink sections of the Dan's Cave System, Abaco Island, Bahamas (Fig. 1). Recent exploration of a subseafloor marine cave off Andros Island, Bahamas, revealed an unknown member of the genus Godzillius, described here.
Cryptic speciation can create taxonomic concerns for stygobitic fauna; thus, integration of morphological and molecular approaches (DNA barcoding) are useful in distinguishing species (Juan et al. 2010;Cánovas et al. 2016). Within Remipedia alone, Xibalbanus fuchscockburni (Neiber et al., 2012), and X. cozumelensis Olesen et al., 2017, were recognized as cryptic/pseudocryptic when compared to other members of their genus using mitochondrial genes (Neiber et al. 2012;Olesen et al. 2017). Since the use of highly specialized technical cave diving technology is essential to access underwater cave systems, comprehensive comparisons across taxa are challenging and often absent from studies of Remipedia. Of the 29 previous remipede species descriptions, only four have included genetic data for species level identifications. Herein, we describe Godzillius louriei sp. nov. based on morphological (LM, SEM) and molecular techniques (16SrRNA and histone 3). Furthermore, we provide a morphological overview and molecular phylogenetic analysis of Godzillius with remarks on the biogeographic distribution of the genus.

Sampling and localities
A single remipede specimen (holotype) of Godzillius louriei sp. nov. was collected on 4 September 2017 in a 50 ml plastic Falcon tube from Conch Sound Blue Hole (25°07′ N, 78°00′ W), a subseafloor marine cave located 20-30 m offshore from North Andros Island, Bahamas. Conch Sound Blue Hole is the longest known subseafloor marine cave, consisting of a predominantly linear, southward trending conduit found just offshore from the northeastern coast of North Andros (Fig. 2) (Palmer 1997;Daenekas et al. 2009). The holotype was collected in the 'Collapse Room' at a water depth of 30-32 m and approximately 1600 m from the cave's only entrance. The remipede was collected in the saltwater  Schram, Yager & Emerson, 1986 within the Lucayan Archipelago. Type localities of Godzillius fuchsi Gonzalez, Singpiel &Schlagner, 2013, G. louriei sp. nov. andG. robustus Schram, Yager &Emerson, 1986 are indicated. Map constructed using the open source QGIS ver. 3.12 software (QGIS Development Team 2020) and metadata from Natural Earth (2020). zone just above a hydrogen sulfide layer. The holotype was preserved in 80% ethanol and stored in the refrigerator prior to morphological and molecular work. Additional specimens used for comparative investigations were collected from Dan's Cave, Abaco Island, Bahamas in March 2017 (LB, TMI, BK,  KM, JO) and in Cottage Pond, North Caicos Island, Turks and Caicos Islands (LB, TMI, BCG, KW, JO) in January 2019. Specimen details are provided below in 'Comparative material'.

Photography, specimens and morphology
The single specimen of G. louriei sp. nov. was used for both morphological and molecular studies. Ten limbs were removed for molecular work (see below) prior to photographing the habitus of the specimen. All specimens were photographed using a Canon EOS 5D Mark IV fitted with a Canon Macro Twin Lite MT-24EX flash and a Canon MP-E 65mm f2.8 macro lens tethered to a PC and operated using standard EOS software. Depth of field in the final images of G. louriei sp. nov. was enhanced by shooting and combining z-stacks later blended using Zerene Stacker ver. 1.04. Left side mouthparts (maxilla 1, maxilla 2, maxilliped, both mandibles) and one trunk limb were removed and prepared for SEM. Additionally, the mouthparts (maxilla 1, maxilla 2, maxilliped) of two individuals of G. fuchsi and one individual of G. robustus were prepared for comparison. All dissected appendages for SEM were dehydrated in a graded ethanol series (80%, 90%, 95%, 100%), critical point dried, mounted on aluminum stubs and sputter coated with platinum/palladium. Morphological observations and micrographs were made using a JEOL JSM-6335-F (FE) field emission SEM at the Natural History Museum of Denmark (University of Copenhagen). Selected appendages (left antenna 1, left antenna 2, trunk limbs 1, 2, 7, 28 and 29) were additionally prepared on permanent slides. Slides were photographed using an inverted Olympus microscope (IX83) with automatized stacking and stitching capabilities. Terminology follows Gonzalez et al. (2013), Koenemann & Iliffe (2014) and Schram et al. (1986). Material of the new species is deposited at the Natural History Museum of Denmark (NHMD), University of Copenhagen.

Comparative material
The following material of Godzillius fuchsi and G. robustus from NHMD and the National Museum of Natural History, Smithsonian Institution, Washington DC (USNM) were included for comparison:

Taxon selection for molecular phylogeny
In order to systematically evaluate our new material, we compared it with all other species within Godzillius and Godzilliognomus (Table 1). A total of six individuals across three species were newly sequenced: three Godzillius fuchsi, two G. robustus and one G. louriei sp. nov. Additionally, eight individuals across four species were obtained from GenBank (Benson et al. 1998) Yager, 1989 and one Cryptocorynetes haptodiscus Yager, 1987. Cryptocorynetes haptodiscus (Cryptocorynetidae) was selected as the outgroup as it was shown to be one of the closest relatives to Godzilliidae that has data available in GenBank (Hoenemann et al. 2013).
Sequences (16S rRNA and H3) of G. louriei sp. nov. (n = 2), G. fuchsi (n = 6) and G. robustus (n = 4) were visually inspected, trimmed and cleaned using Geneious Prime ver. 2019.2.3 (Kearse et al. 2012). All sequences were checked for potential contamination by running a nucleotide BLAST similarity search (Altschul et al. 1990). Protein-coding H3 gene sequences were inspected for stop codons and point mutations using Geneious Prime to reduce the risk of including pseudogenes (Song et al. 2008). All sequence data were submitted to GenBank under accession numbers MW760694-MW760699 and MW768707-MW768712. The GenBank H3 gene sequence of G. robustus (KC989960) was excluded due to probable contamination, as the sequence genetically resembled that of Godzilliognomus. Sequences from multiple individuals previously identified as Godzilliognomus schrami and Godzilliognomus frondosus were concatenated from available GenBank data for H3 and 16S rRNA sequences to avoid excessive gaps in the phylogeny. These included KC989961 + KC989998, KC989983 + KC989999 and KC989962 + KC990013. As these sequences were not from the same individuals, individual gene trees for H3 and 16S rRNA were constructed using maximum likelihood to identify potential issues from concatenation, and are provided in the supplementary material (Supplementary File 1 and Supplementary File 2). Sequences were aligned using the MAFFT ver. 7 auto-iterative alignment program (Katoh et al. 2019). MAFFT was selected due to its greater accuracy relative to other alignment programs (Pais et al. 2014). Gene alignments were subsequently concatenated within Geneious Prime (H3: 327bp, 16S: 543bp). Both Maximum Likelihood (ML) and Bayesian Inference (BI) were utilized. ML substitution models for each gene were selected based on the Akaike Information Criterion (AIC) in ModelFinder within IQ-Tree ver. 1.6.11 (Nguyen et al. 2014;Kalyaanamoorthy et al. 2017). The most optimal DNA substitution models for ML analyses of 16S rRNA and H3 alignments were GTR + F + R2 and TN + F + G4, respectively. The optimal AIC model for BI analyses of both 16S rRNA and H3 alignments was GTR + G. Individual gene trees and the concatenated gene tree were constructed using the program IQ-TREE for Maximum Likelihood (ML) analyses (Nguyen et al. 2014). IQ-TREE was selected for this analysis as it was shown to outperform other ML programs in increased likelihood values when analyzing concatenated species trees (Zhou et al. 2017). Nodal support was quantified using ultrafast bootstrapping methods (UFBoot) with 1000 replicates (Hoang et al. 2018). jModelTest ver. 2.1.10 (Guindon & Gascuel 2003;Darriba et al. 2012) was used to find the optimal BI substitution models based on AIC and the alignment was subsequently run in MrBayes ver. 3.2.6 (Ronquist et al. 2012) on XSEDE within the Cipres Science Gateway (Miller et al. 2010). Four Markov Chain Monte Carlo (MCMC) chains were run twice for 30 000 000 generations with a burn-in of 10 000 000. Convergence was evaluated using trace plots and effective sample size (ESS > 200) within the program Tracer ver. 1.7.1 (Rambaut et al. 2018).

Molecular variation in
Godzillius relative to that in other genera within Remipedia was compared using 16S rRNA sequence pairwise distances calculated using p-distance and pairwise deletion of gaps in MEGA ver. 7 (Kumar et al. 2016). All GenBank 16S rRNA material was used, with the exception of a potentially contaminated sequence of Pleomothra apletocheles Yager, 1989 KC990006 (Table 1).

Etymology
Named for Robert Lourie whose financial support of the Bahamas Caves Research Foundation contributes to furthering cave and blue hole related research in the Bahamas. The taxonomic description and underlying molecular justification for Godzillius louriei sp. nov. was prepared by LB, HBG, and JO, who are thus responsible for making the specific name louriei available. Description Cephalon (Fig. 3). Cephalic shield subtrapezoidal, posterior margins wider than anterior. Posteriolateral margins rounded; sutures absent. Anterior margin folds ventrally, covering a1 aesthetascs and bifurcated frontal filaments. Body (Fig. 3). Body length 25 mm; 29 trunk segments. Pleurotergite lateral margins pointed posteriorly. Sternal bars isomorphic. Trunk limbs bifurcated with endopods and exopods consisting of three and four segments respectively. Trunk limbs 1 and 18-29 reduced in size (Figs 3-4). Trunk limb 14 protopod with large lobate protrusion and ventrally with slender genital flap (Fig. 3F).
Antenna 1 (Fig. 4A). Biramous, located posterior to frontal filaments. Peduncle with two articles; proximal article bearing numerous aesthetascs. Distal peduncle article bifurcated, acts as base of dorsal and ventral rami. Dorsal ramus (i.e., dorsal branch) with 11 articles; girth decreasing distally through articles. Article 1 with single anteriodistal setal cluster; article 2 with one medial seta, one distal cluster; article 3 with two medial setae; article 4 with two medial setae, one distal seta, fine marginal setae; article 5 with one medial seta, one distal cluster; article 6 with three medial setae, one distal seta; article 7 possessing two to three medial setae, one distal seta; article 8 bearing one fine medial seta; articles 9 and 10 lacking setae; article 11 with terminal tuft of setae. Ventral ramus (i.e., ventral branch) with ambiguous articulation, treated as three articles (Fig. 4A). Proximal article shorter than article 2, no setae, partly fused with peduncle. Article 2 length ~2× that of proximal article. Article 3 length ~3× that of article 2, with one filiform medial seta and a distal setal tuft.

Remarks
Species of Godzillius can be distinguished by several morphological characters, most notably relating to the md and the three pairs of prehensile/raptorial post-mandibular mouthparts (Figs 9-10, Table 2). On the left md, the lacinia mobilis of both G. louriei sp. nov. and G. fuchsi have five denticulae, whereas G. robustus has six. One of the most striking distinctions between species of Godzillius is the number of conical denticles on the mx1 endite segment 4 anteriodistal margin (Fig. 9B, F, J). While G. fuchsi and G. robustus have been observed or described as having between 6 and 8 denticles along its margin, G. louriei sp. nov. has 10. Furthermore, the mx1 endite first segment has a unique spination, with 10 large spines and 3 small (Fig. 9D), contrasting with those of G. robustus (11 large, 4 small) and G. fuchsi (10 large, 2 small).
The terminal claw of mx2 in G. robustus is reported to have 10 free spines, whereas that of G. fuchsi and G. louriei sp. nov. have 7 (Fig. 9O, R, U). The mxp terminal claw in G. fuchsi has an elongate protrusion of the setal pad that is not covered by its spines (Fig. 10F); in contrast, the spines of G. louriei sp. nov. and G. robustus cover the setal pad (Fig. 10B, D). The mxp terminal claw of G. robustus has been described as a "grappling hook" with ten spines wrapping around a setal pad (Schram et al. 1986). Godzillius louriei sp. nov. has a similar arrangement, with at least 7 spines in the grappling hook arrangement (Fig. 10B). Godzillius fuchsi differs from the aforementioned species, having shorter, denticle-like spines with narrow spaces between them and not covering a distinctly protruding setal pad (Fig. 10F). We found the mxp of all three species to be composed of 9 segments (Fig. 10), modifying the previous descriptions of G. robustus and G. fuchsi, where fewer proximal segments were identified. It should be noted that this number of mxp segments coincides with what is reported for all other remipede species (Koenemann & Iliffe 2014).
16S rRNA pairwise distances revealed Godzillius louriei sp. nov. has a genetic distance of 15% when compared to all individuals of G. fuchsi and G. robustus whereas the distance between individuals of G. fuchsi and G. robustus is 13-14% (Table 3). Within Godzilliognomus, the sister genus to Godzillius (see Fig. 11), the distance between the two known species, Godzilliognomus frondosus and G. schrami, is slightly lower at 12-13%.

Molecular distinction of Godzillius louriei sp. nov.
The present study describes a third remipede species of the genus Godzillius. Both morphological and molecular approaches provide support for the recovery of G. louriei sp. nov. within Godzillius, being distinct from the two other species of the genus. There is some indication within the phylogeny that G. louriei sp. nov. may be sister to G. robustus (Fig. 11); however, further data is needed to clarify  (2021) 132 the relationships within Godzillius. We compared 16S rRNA pairwise distances within Godzillius and found them to be equal to or greater than what is observed within Godzilliognomus. In general, the 16S rRNA disparity observed within genera of Remipedia is notably high relative to other crustacean groups (Lefébure et al. 2006), which may suggest greater divergence times between remipede species.

Morphological distinction of Godzillius louriei sp. nov.
The shape of the cephalic shield, articulation of the ventral ramus of antenna 1 and the digitiform maxilla 1 endite fourth segment are key characteristics of the genus Godzillius (Schram et al. 1986;Gonzalez et al. 2013) which are shared by G. louriei sp. nov. Godzillius louriei sp. nov. can be distinguished from other species of Godzillius by several minute morphological characters on the prehensile/raptorial cephalic limbs, maxilla 1, maxilla 2 and maxilliped (Table 2, Figs 9-10). These limbs exhibit notable variation and often harbor specific diagnostic characters, as Koenemann et al. (2007) concluded in their detailed morphological phylogeny. The differences between G. louriei sp. nov. and its two congeners relate to details such as the number of spines, denticles and setae on the endites of segments 1-3 of maxilla 1, and the number of spines on the terminal claws of maxilla 2 and the maxilliped (see Remarks above and Table 2, Figs 9-10). Based on new SEM examination, we identified several discrepancies between the original descriptions of G. robustus (see Schram et al. 1986) and G. fuchsi (see Gonzalez et al. 2013) relative to our newly collected topotypic material, specifically with regards to the spination and setation of maxilla 1 endites (see Table 2). These variations may be due to the use of different microscopy techniques; SEM provides alternative viewpoints of a singular structure at high magnification, capturing spines and setae that may be difficult to view in light microscopy. For instance, neither description of G. fuchsi or G. robustus report the presence of small proximal spines on the endite of segment 1, nor the presence of spines along the spatulate endite of segment 2 within maxilla 1; yet they are both observed in our SEM analyses. Based on our 16S rRNA data, the material of G. robustus is conspecific with similarly named material in GenBank (Fig. 11). A detailed examination of type material is needed to clarify whether the morphological differences are instances of intraspecific variation, or whether the original descriptions lack details in these respects.  Schram, Yager & Emerson, 1986: G. louriei sp. nov., G. robustus Schram, Yager & Emerson, 2013and G. fuchsi Gonzalez, Singpiel & Schlagner, 1986. All characters denoted with an asterisk ( * ) are from observations in this study. All other characters are from their respective species descriptions ( 1 = Gonzalez et al. 2013;2 = Schram et al. 1986).

Distribution of the genus Godzillius within anchialine habitats
Godzillius louriei sp. nov. marks the first of its genus to be found on the Great Bahama Bank, considerably expanding the known distribution of Godzillius throughout the Lucayan Archipelago (Fig 1). The presence of a potential G. louriei sp. nov. -G. robustus clade is not readily explainable zoogeographically, as the two species occur further from each other (Andros and North Caicos: 700 km) than G. louriei sp. nov. and G. fuchsi (Andros and Abaco: 135 km) (Fig. 1). All three species are found within the Lucayan Archipelago, but each occurs on separate shallow-water platforms (banks) and are isolated by deep ocean channels. The species of Godzillius are only known from their type localities. This may either suggest that they are truly endemic, possibly representing remnants of an earlier broader distribution, or that their distribution spans unexplored or unknown crevicular systems.
While most remipede species have been collected within inland anchialine cave environments (n = 26), a few have been observed in offshore subseafloor marine caves. Godzillius louriei sp. nov. marks only the third known remipede species to inhabit subseafloor marine caves, the others being Xibalbanus cokei (Yager, 2013), from Caye Chapel Cave, Belize and Speleonectes kakuki Daenekas et al., 2009, which also inhabits Conch Sound Blue Hole (Daenekas et al. 2009;Yager 2013). Interestingly, S. kakuki was collected in the same section of the cave as G. louriei sp. nov. (Daenekas et al. 2009). Both X. cokei    and G. louriei sp. nov. are only known from their type localities, whereas S. kakuki has been observed within both inland anchialine cave systems and subseafloor marine caves (Daenekas et al. 2009;Yager 2013). The summarized occurrence of remipedes in both types of cave systems, with one species (S. kakuki) spanning both, suggests that a close relationship between these habitats exists, either currently or historically. van Hengstum et al. (2019) proposed that anchialine and marine caves may be linked through allogenic succession and should be considered parts of the "anchialine habitat continuum".
The idea of a continuous or crevicular "spelean corridor" has been shown as a means for anchialine fauna to disperse throughout subterranean systems (Hart et al. 1985;Hunter et al. 2008;Gonzalez et al. 2017). Historic sea level fluctuation may also have contributed to the present day distribution of Godzillius. Anchialine habitats are shown to shift with sea level change, resulting in different community compositions within cave environments (van Hengstum et al. 2019). The type localities of G. louriei sp. nov., G. fuchsi and G. robustus (Conch Sound, Dan's Cave, Ralph's Sink, Cottage Pond) all contain large speleothems within their passages, which only form in air by dripping water (BG, BK, TI, pers. obs.; Koenemann et al. 2004;Surić et al. 2005), indicating that the caves were dry during glacial periods of low sea level. Because of these historic complexities, caution must be applied when assessing anchialine fauna distribution patterns, as we are likely only seeing a snapshot of a dynamic transgression and regression of anchialine habitats along karstic coastlines.