A reassessment of the phylogenetic utility of genus-level morphological characters in the family Bogidiellidae (Crustacea, Amphipoda), with description of a new species of Eobogidiella Karaman, 1981

Abstract Bogidiellidae is the most diverse and cosmopolitan family of stygobiotic amphipods, and inhabits a variety of subterranean biotopes, especially interstitial habitats. While the family is characterized by considerable sexual dimorphism, this dimorphism has adversely affected our understanding of the systematics of the group. Most species have restricted geographic ranges and occur in difficult to sample habitats, so it is common for individual species descriptions to be based on a single sex. In this work we revisit an analysis of morphological characters in an attempt to clarify their phylogenetic utility in resolving taxonomic relationships among genera by introducing a new species, two additional characters, and phylogenetic statistical support values. Eobogidiella venkataramani sp. n., from a spring fed brook in the Shirawati River basin along the escarpment of the Western Ghats (Karnataka, India) differs from the only known congener, Eobogidiella purmamarcensis, from Argentina, in the structure of mouthparts, the shape and ornamentation on gnathopods and characters of the telson. Our phylogenetic analyses indicate that the available morphological characters are not sufficient to resolve phylogenetic relationships within Bogidiellidae, thus these characters alone cannot be used to determine the phylogenetic placement of Eobogidiella venkataramani sp. n. within the family. Nevertheless, Eobogidiella venkataramani sp. n. shares diagnostic characters with Eobogidiella, supporting placement of the new species in this genus. Our findings point towards a critical need to resolve relationships within the family using molecular approaches, along with the development of a suite of additional morphological characters for Bogidiellidae. This is the third species of Bogidiellidae from southern India.


Introduction
The family Bogidiellidae Hertzog, 1936 has an intriguing history of study that shaped the systematics of the group (e.g., Hertzog 1933;Holsinger and Longley 1980;Karaman 1981;Stock 1981;Ruffo 1973;Koenemann et al. 1998;Koenemann and Holsinger 1999;Iannilli et al. 2006;Jaume et al. 2007;Vonk and Jaume 2010;Leijs et al. 2011;Senna et al. 2014), but this work has not led to a coherent understanding of relationships within the family (Lowry and Myers 2013). The Bogidiellidae includes 37 genera and 113 described species, with the phylogenetic relationships among the genera discussed by Stock (1981) and a phylogenetic tree produced by Koenemann and Holsinger (1999).
Only two Bogidiellidae species are known from India: Bogidiella indica , recorded from bore wells in Andra Pradesh, and the minute species Bogidiella totakura Senna et al. 2013, from a nearby locality Andhra Pradesh, southern India. The only other stygobiotic amphipod species of India are the gammaroid Indoniphargus indicus (Chilton 1923) (Mesogammaridae), reported from various groundwater habitats (e.g., springs, well water, and a mine pit) in the north-eastern states of Bihar, West Bengal and Odisha (formerly Orissa) (Stephensen 1931;Straškraba 1967), and the crangonyctoid Kotumsaria bastarensis Messouli et al. 2007 (Kotumsaridae), from Kotumsar Cave, in the east-central state of Chhattisgarh (Messouli et al. 2007;Senna et al. 2013).
Below we describe Eobogidiella venkataramani sp. n. from a spring-fed freshwater habitat in southwest India and evaluate the phylogenetic utility of the available morphological characters (Koenemann and Holsinger 1999, and two characters added in the present study) in hopes of gaining insights into the placement of our new species within the family.

Specimen sampling
A sample containing the stygobiont (one specimen) was collected in December 2008 from a spring-fed brook in the state of Karnataka in southwest India (Figs 1, 2) using a hand-made hemispherical scraper and preserved in a 4% solution of formaldehyde.

Morphology and taxonomic terms
Body length was recorded while holding the specimen straight and measuring the distance along the dorsal side of the body from the base of the first antenna to the base of the telson using an ocular micrometer in a Lomo MBS-9 dissecting microscope. Appendages were drawn using a Carl Zeiss NU-2 compound microscope equipped with a drawing device as described in Gorodkov (1961).
Due to improper storage, the specimen was entirely dry upon initial examination. We followed the method described by Namiotko et al. (2011) to rehydrate the specimen. A permanent preparation was made using polyvinyl lactophenol (PVL) and a methylene blue staining solution was used as mounting medium.
The term "palmar angle" of the gnathopod propodi refers to the angle formed at the end of the palm and beginning of the posterior margin or the point at which the tip of the dactylus closes on the propodus (Birstein 1941). The fore-gut lateralia comprise a potentially useful morphological character in the phylogenetic analysis (Coleman 1991). We use the term "sternal humps" (Holsinger 1989;Sidorov 2010) to refer to the "pulvinate sternal epithelium" of Kikuchi et al. (1993) and Kikuchi and Matsumasa (1997), which is homologous to the "mediosternal processes" of Koenemann and Holsinger (1999) and Pérez-Schultheiss (2013). Koenemann and Holsinger (1999) took the view that the mediosternal gills of Paracrangonyx Stebbing, 1899 are autapomorphous, with a different physiological function and morphological structure, but they do not cite the works of Kikuchi. Fenwick's (2001) diagnoses of the genus Paracrangonyx includes the presences of "Single, simple, elongate sternal gills medially on peraeonites 2-7." Bousfield (1977) observed that sternal gills are present in several families of amphipods which are not closely related -including Crangonyctidae (e.g., Holsinger 1977), Hyalellidae, and Pontogeneiidae -and suggests that these structures arose independently in the different groups. Homologies of mediosternal gills and sternal humps within and across families of freshwater amphipods remains problematic, and effective use of these characters in phylogenetic analyses requires further study.
We used the Bogidiellidae sensu lato in our analysis, including Artesiidae, as its acceptance as a distinct family has been questioned (Stock 1981;Botosaneanu and Stock 1989), as well as the genus Kergueleniola Ruffo, 1974 which is sometimes placed in a separate family Kergueleniolidae (Lowry and Myers 2013). We were unable to test the validity of the inclusion of the Salentinellidae in Bogidielloidea within the Senticaudata: the uniramous uropod 3 in Parasalentinella Bou, 1971 does not fit with core bogidiellid features, and Salentinella Ruffo, 1947 species lack apical robust setae on uropods 1-2 (cf., Salentinella anae Messouli et al. 2002). Bogidiella indica , the sole member of the indica-group sensu Holsinger et al. (2006), recently has been attributed to the niphargoides-group based on the shared absence of rami on pleopods 1-3 (Senna et al. 2013). Therefore, the indica-group was not considered in our analysis, as it is instead included in our analysis within the niphargodes-group. The genera Paracrangonyx Stebbing, 1899 (Paracrangonyctidae), Pseudingolfiella Noodt, 1965 (Pseudingolfiellidae) and Dussartiella Ruffo, 1979 (Dussartiellidae) were excluded from the analysis. The recent placement of these genera in different families (see Koenemann and Holsinger 1999;Iannilli et al. 2011; Myers 2012), supports a higher-level analysis of the Senticaudata, in which the Pseudingolfiellidae is not considered even to be a member of the suborder Senticaudata (Lowry and Myers 2013), whereas the Paracrangonyctidae and Dussartiellidae fall into the Gammarida instead of the Bogidiellida in the analysis of Lowry and Myers (2013).
Following Koenemann and Holsinger (1999), we conducted two phylogenetic analyses, treating all characters as unweighted: first with unordered character states and an 'alternative' analysis with ordered character states. The parsimony analyses (unordered and ordered) of 46 taxa, including 37 genera of Bogidiellidae, 2 genera of Artesiidae, and 1 genus of Kergueleniolidae, and the hypothetical ancestor outgroup used by Koenemann and Holsinger (1999), were based on 29 morphological characters (Suppl. material 1). Both analyses were conducted in PAUP*4.0a146 (Swofford 2002) using a heuristic search, random stepwise addition with 1000 replicates and TBR branch swapping. Advances in computer power and processor speeds and have allowed us to reevaluate Koenemann and Holsinger's (1999) original cladistic analysis with modern and more rigorous methods that incorporate statistical measures of branch support. Bootstrap and Jackknife resampling methods for branch support were performed with PAUP*4.0a146 using the "Fast" stepwise-addition search (1,000,000 replicates). PAUP* command files for Decay/Bremer support indices were generated with TreeRot. v3 (Sorenson and Franzosa 2007), input with strict consensus trees, edited to run each heuristic search for 500 replicates with TBR branch swapping, and executed in PAUP*4.0a146.
Acronym used for the collection FEFU Zoological Museum of the Far East Federal University, Vladivostok.

Phylogenetic analysis of Bogidiellidae sensu lato
To investigate the phylogenetic utility of the available morphological characters and to determine the placement of our new species among the bogidiellids, we reevaluated the relationships within the family, adding new taxa and characters to the morphology matrix of Koenemann and Holsinger (1999). Our phylogenetic analysis of 29 morphological characters supports two equally parsimonious trees (length = 170) and 3235 equally parsimonious trees (length = 243) in the unordered and ordered analyses, respectively. Although the strict consensus trees (Figs 3, 4) resolved some relationships, they lack support from bootstrap (Suppl. material 2, 3), jackknife (Suppl. material 2, 3), and Bremer/decay indices (Figs 3,4), due, at least in part, to the low character to taxa ratio (29 to 46, respectively). The strict consensus tree for the ordered analysis places the new species within the genus Eobogidiella Karaman, 1981 (Fig. 3), without significant statistical support. The unordered analysis ( Fig. 4) instead places these two taxa in association with other genera (the new species with Kergueleniola; Eobogidiella purmamarcensis (Grosso & Ringuelet, 1979) with Bogidiella and other genera), also without significant statistical support.
The two additional characters (i.e., the presence or absence of a coxal endite on the maxilliped, and the morphology of coxa 5) were added to the matrix of Koenemann and Holsinger (1999), but provide little additional phylogenetic support. Although the "maxilliped, coxal endite" is an informative character as it is present for a number of genera, the morphology of coxa 5 is normal for all genera except for Glyptogidiella (for which it is hypertrophied), therefore, this autapomorphy is phylogenetically uninformative.
The above analyses revealed that the available morphological characters provide no phylogenetic utility in resolving generic relationships within the Bogidiellidae sensu lato, thus the available morphological characters do not allow us to establish the phylogenetic placement of the new species. Therefore, the resulting phylogenies (Figs 3, 4) cannot be used to inform generic placement of the new species or direct choices for generic comparisons. Instead, generic placement of the new species must rely exclusively on shared generic-level diagnostic characters. Generic-level diagnostic characters (i.e., 3 outer ramus segments in pleopods 1-3, uniarticulate and reduced inner rami of pleopods 1-3, and 1 segmented palp of maxilla 1) of the new species are shared with the South American genus Eobogidiella, suggesting a possible close relationship with E. purmamarcensis. Other bogidiellid genera were considered based on the shared presence and absence of male sexual modifications (Indogidiella) and similar geographical distributions (Bogidiella) ( Table 1). Although both Indogidiella and the new species

Bogidiella indica
Holsinger et (Table 1), the nature of the modifications of the spines on the rami of male uropods 1 and 2 of Indogidiella (Ruffo 1994, fig. 4g, h;Stock 1983, figs 23, 24) differ from the new species, possibly indicating independent origins. Furthermore, species in Indogidiella have a 2 segmented palp on maxilla 1, whereas the new species has a 1 segmented palp. The geographically proximate species, Bogidiella indica and Bogidiella totakura, do not share the same male sexual modifications and also have a 2 segmented palp on maxilla 1. Therefore, based on the diagnostic characters shared with Eobogidiella and morphological dissimilarity from Indogidiella and the more geographically proximate genus, Bogidiella, we tentatively place the new species in the genus Eobogidiella recognizing further study is required to understand generic boundaries and relationships within the family. Diagnosis. Habitus typical of a stygomorphic bogidiellid, combining a number of features found in other genera of this family. Primary characters: maxilla 1 with vestigial, single-segmented, symmetrical palps; pleopods 1-3 with single-segmented, reduced inner rami.

Species description and taxonomy
Secondary characteristics: ventral surface of pereonites 2-7 bearing sternal humps; coxal gills on pereopods 3-6; antenna 1 with reduced, single-segmented, minute accessory flagellum; mandibles with tiny, vestigial molars with 2 short spines and 1 plumose seta; maxilliped lacking coxal endite; apparent sexual dimorphism (spines on uropods 1 and 2 modified). Etymology. The specific epithet honors the former Director of Zoological Survey of India, Dr. K. Venkataraman, whose assistance was pivotal in the early stages of this research.
Variability. Unknown. Sexual dimorphism. Unknown, but modified spines on uropods 1 and 2 probably represent a male-specific trait.
Distribution and ecology. Eobogidiella venkataramani sp. n. dwells in a springfed brook habitat located on the flat bottom of a small valley in the rainforest. The biotope is a small trickling swampy stream 1-3 m wide and 0-0.05 m deep, without flow, water temperature +22 °C, and a substrate comprised of wet litter, detritus, stones, clay. Known only from type locality.

Discussion
The only other species in this genus, E. purmamarcensis was described by Grosso and Ringuelet (1979) who placed it in the genus Bogidiella. It occurs in sandy sediments of the Rio Grande at the entrance of Purmamarca, Jujuy Province of northwestern Argentina (Fig. 1). Karaman (1981) Karaman (1982) elevated Eobogidiella to generic status.
In spite of our decision assign the new species to Eobogidiella, weak phylogenetic support for generic concepts and relationships within the family leaves us with reservations regarding this placement. The highly disparate known geographic distributions of Eobogidiella venkataramani sp. n. and E. purmamarcensis (India and Argentina, respectively) is suspicious, suggesting that some of their shared character states may be homoplasious. Furthermore, two important morphological characters may be misleading in their support of a close relationship between E. venkataramani sp. n. and E. purmamarcensis. First, the soft suture between the head and pereonite 1 described here for E. venkataramani sp. n. was not mentioned in the description of E. purmamarcensis (Grosso and Ringuelet 1979) nor in subsequent works treating the placement of this species (Karaman 1981(Karaman , 1982Stock 1981;Koenemann and Holsinger 1999). It is likely that the soft suture in E. venkataramani sp. n. is an artifact caused by the inflation of soft tissues from rehydration of the desiccated specimen. Second, we have described sternal humps as present on pereonites 2-7 of E. venkataramani sp. n., and these are not mentioned in the description of E. purmamarcensis nor in subsequent works treating the placement of this species (Karaman 1981(Karaman , 1982Stock 1981;Koenemann and Holsinger 1999). Koenemann and Holsinger (1999) included the sternal humps (as "mediosternal processes") as a character in their phylogenetic analysis, but determined that the mediosternal gills of Paracrangonyx evolved independently, coding the mediosternal processes as absent in Paracrangonyx. However, the use of sternal humps as a character in the Bogidiellidae did not come into play until well after the treatments of E. purmamarcensis by Karaman (1981Karaman ( , 1982 and Stock (1981), so the character could have been overlooked. Additionally, we suspect that the occurrence of sternal humps (or "mediosternal processes") in E. venkataramani sp. n. is likely independent and does not reflect phylogenetic proximity to the Chilean Patagongidiella and Grossogidiella (Pérez-Schultheiss 2013).
Based on our reanalysis of Koenemann and Holsinger's (1999) dataset, relationships among and within genera of the family Bogidiellidae remain unclear. Because the available morphological characters are phylogenetically uninformative, development of additional morphological characters across the family, and, especially, implementation of modern molecular phylogenetic approaches, are desperately needed to resolve relationships within the family and to better define generic boundaries. Nevertheless, it seems that the current assignment of the genera, mostly developed by Koenemann and Holsinger (1999), should be maintained until a more robust and well supported phylogeny can be produced.

Morphological character matrix
Authors: Dmitry A. Sidorov, Aron D. Katz, Steven J. Taylor, Mikhail V. Chertoprud Data type: NEXUS file Explanation note: NEXUS file including character matrix for Bogidiellidae, Artesiidae and Kergueleniolidae used in analysis.. Copyright notice: This dataset is made available under the Open Database License (http://opendatacommons.org/licenses/odbl/1.0/). The Open Database License (ODbL) is a license agreement intended to allow users to freely share, modify, and use this Dataset while maintaining this same freedom for others, provided that the original source and author(s) are credited.

Figure S1
Authors: Dmitry A. Sidorov, Aron D. Katz, Steven J. Taylor, Mikhail V. Chertoprud Data type: TIF file Explanation note: Ordered bootstrap and jackknife consensus tree. Numbers below branches are bootstrap followed by jackknife support values. Copyright notice: This dataset is made available under the Open Database License (http://opendatacommons.org/licenses/odbl/1.0/). The Open Database License (ODbL) is a license agreement intended to allow users to freely share, modify, and use this Dataset while maintaining this same freedom for others, provided that the original source and author(s) are credited.

Figure S2
Authors: Dmitry A. Sidorov, Aron D. Katz, Steven J. Taylor, Mikhail V. Chertoprud Data type: TIF file Explanation note: Unordered bootstrap and jackknife consensus tree. Numbers below branches are bootstrap followed by jackknife support values. Copyright notice: This dataset is made available under the Open Database License (http://opendatacommons.org/licenses/odbl/1.0/). The Open Database License (ODbL) is a license agreement intended to allow users to freely share, modify, and use this Dataset while maintaining this same freedom for others, provided that the original source and author(s) are credited.