Phylogenetic relationships of psammosteid heterostracans (Pteraspidiformes), Devonian jawless vertebrates

Psammosteid heterostracans are a group (suborder Psammosteoidei) of Devonian-age jawless vertebrates, which is included in the order Pteraspidiformes. The whole group of psammosteids is represented by numerous species (more than 40); their phylogenetic relationships are still poorly known and deserve further study. Classical researchers of the psammosteids, such as D. Obruchev, E. Mark-Kurik and L. Halstead Tarlo, had different views on the phylogeny of the group (e.g. origins and evolution of Psammosteus). To check the modern hypotheses of psammosteid origins from various Pteraspidiformes and to clarify psammosteid interrelationships, the most complete phylogeny of this group (38 ingroup taxa + juvenile Drepanapsis) is presented here. Different methods of data analysis were used to explore the psammosteid data set, including equally weighted characters versus implied weighting. According to the results of the phylogenetic analysis, the monophyletic status of the group and their early development from the Pteraspidiformes are supported. The diagnoses and interrelationships of many taxa are clarified. Two new genera are proposed (Vladimirolepis gen. nov. and Elgaia gen. nov.), and two subfamilies are erected (Placosteinae subfam nov. and Psammosteinae Traquair).

In the last version of his psammosteid classification, Halstead Tarlo (Halstead, 1993) excluded weigeltaspidids from the order. Thus, the order Psammosteiformes became similar in its composition to the suborder Psammosteida sensu Halstead Tarlo, 1964a or the order Psammosteida sensu Obruchev, 1964. The suborders Psammosteida and Obrucheviida were not recognized (Halstead, 1993): Order Psammosteiformes Berg, 1940 Family Drepanaspididae Traquair, 1899 Family Guerichosteidae Halstead Tarlo, 1964 Family Pycnosteidae Tarlo, 1962 Family Psammolepididae Tarlo, 1962 Family Psammosteidae Traquair, 1896 Family Obrucheviidae Halstead Tarlo, 1964 With the discovery of anchipteraspids (Elliott, 1984), it became clear that pteraspids evolved from cyathaspids. The hypothesis of independent evolution of cyathaspids and pteraspids from tesselated heterostracans was untenable. Psammosteids are grouped with pteraspids according to Obruchev's hypothesis. Alain Blieck in cooperation with David K. Elliott (Blieck, 1984, Fig. 61) made the first (hand-constructed) cladogram of the Pteraspidiformes, in which the psammosteid clade (coded Drepanaspis gemuendenensis) is a sister group (order Psammosteiformes) of Pteraspidiformes (consistently positioned at the base of the tree). This author compared Drepanaspis with Cosmaspis. He noted in detail the similar body plan of pteraspids (Protopteraspis) and psammosteids (Blieck, 1984, p. 99). In the subsequent work (Blieck, Elliott, and Gagnier, 1991), data on anchipteraspids were included, and the protopteraspidids were considered as the sister group of the higher pteraspids and psammosteids on the basis of the primitive location of pineal canal. In the proposed phylogenetic scheme by P. Janvier (1996, Fig. 4.9), psammosteids were placed as the sister group to protaspidids (as a development of Obruchev's traditional hypothesis) and combined with them in the new superfamily Protaspidoidea.
Vincent N. Pernègre (2002) compiled a new character-taxon matrix and produced a cladistic analysis that included Drepanaspis gemuendenensis. In the most parsimonious tree, recovered by him, this psammosteid was placed as a sister taxon to Doryaspis nathorsi in the order Pteraspidiformes. The close relationship between Doryaspis and psammosteids was criticized (Elliott and Mark-Kurik, 2005, p. 107) due to differences in squamation and position, pattern and number of plates in their cephalothoraxes. Pernègre and Elliott (2008) published a strict consensus tree including Drepanaspis gemuendenensis, based on a substantially improved and expanded matrix. According to their cladogram, psammosteids are also included in the order Pteraspidiformes and placed between the basal family Anchipteraspididae and the family 'Protopteraspididae' (Pernègre and Elliott, 2008, Fig. 5). Emma Robert Sansom (2016, 2017), using discrete, continuous and discretized continuous characters, encoded 47 taxa of pteraspidiforms including two well known species of psammosteids -Drepanaspis gemuendenensis and Psammosteus megalopteryx. In all resulting trees using different characters / coding methods and methods of data analysis (equal character weights or implied character weighting), psammosteids fall into the clade Pteraspidiformes (Randle and Sansom, 2016, Figs. 4-6). In most cases (5 of 6 presented trees), the position of psammosteids in a clade with Doryaspis and Woodfjordaspis is maintained (development of Pernègre's hypothesis). It was proposed that psammosteids should be included in the new superfamily Doryaspidae (most correct -Doryaspidoidea).
Presented here is a new most complete cladistic analysis, which has been carried out in order to determine the phylogenetic relationships of psammosteid heterostracans. All reasonably well known species of psammosteids have been included. These phylogenetic relationships of psammosteids were presented for the first time at the Early Lower Vertebrates Symposium in Poland (Glinskiy, 2017;abstract + poster). The results of a preliminary analysis (Anchipteraspididae were not included ) confirm the data of A. Blieck and D. K. Elliott (1984), that psammosteids are a monophyletic group (order Psammosteiformes), and that pteraspids (Pteraspidiformes) are their sister group.

DEVELOPMENT OF VIEWS ON PHYLOGENETIC INTER-RELATIONSHIPS WITHIN PSAMMOSTEIDS
Walter Gross (1930Gross ( , p. 127, 1933a was the first to point out the homology of the branchial plates of the Middle Devonian psammosteids with those of Drepanaspis. He also made suggestions about the evolution of these exoskeletal elements, which involved their contraction and widening. Gross (1933a) divided psammosteids according to the length of their branchial plates into Psammolepis and Psammosteus. He noted similarities of the branchial plates of Schizosteus striatus (syn. Psammolepis striata sensu Gross) with those of Drepanaspis. Obruchev (1940) established the genus Schizosteus, which he treated as a transitional form between Drepanaspis and other psammosteids. He pointed out (Obruchev, 1940, Fig. 4, 1943b variability in the depth of the ventral plates' posterior notch, showing that this notch is similar in Drepanaspis and Schizosteus. Obruchev (1943a) described a 'pteraspid' Yoglinia; later it was assigned to psammosteids (Mark, 1955), attributed to the family Psammosteidae (Obruchev, 1964) and synonymized with Psammosteus (Obruchev and Mark-Kurik, 1965). Obruchev (1943c) produced in his thesis the first phylogeny of the psammosteid genera Pycnosteus, Ganosteus and subsequently Psammolepis and Aspidosteus (syn. Obruchevia) and showed that they consistently branched out from the stem line from Drepanaspis to Psammosteus (Fig. 1A). According to this scheme, psammosteids are descendants of pteraspids. This phylogeny quite accurately anticipates the phylogenetic relationships of the main psammosteid genera (with the exception of Obruchevia) obtained in the present work (Fig. 3). Mark-Kurik (Mark, 1955) described many new psammosteids from the Baltic region in her dissertation and presented a detailed discussion of their evolutionary relationships, including a phylogenetic scheme (Fig. 1C). She followed the view of Obruchev (1943c) about independent evolution of two groups (Pycnosteus + Ganosteus and Psammolepis) from Schizosteus (the data on the concentric lines of growth on branchial and ventral plates at juvelile stages of the development was taken into account). A suggestion was proposed that Schizosteus striatus is the ancestor of Psammolepis proia (syn. Vladimirolepis proia comb. nov.), and 'Psammolepis' heteraster, 'Psammolepis' alata and 'Psammolepis' undulata (syn. Placosteus) represent a separate phylogenetic line within Psammolepis. Mark-Kurik reckoned that Tartuosteus luhai (syn. Elgaia luhai, comb. nov.) is the ancestor of the Psammosteus line, and Yoglinia (syn. Psammosteus) is close to Psammosteus (ibid., p. 17), but she somehow placed Yoglinia between Tartuosteus and Psammolepis in her scheme. Halstead Tarlo (Tarlo, 1957, 1962 described psammosteids (Guerichosteidae) from the Emsian of Poland and negated Obruchev's view about the evolution of psammosteids from Drepanaspis (Tarlo, 1957, p. 228). He further suggested that psammosteids evolved into two independent lineages -marine (represented by Drepanaspis), which didn't produce descendants, and freshwater, represented by guerichosteids and their descendants that produced other psammosteids (Tarlo, 1957, 1962Halstead Tarlo, 1964a, Figs. 30-32) (Fig. 1D). The views of Halstead Tarlo on psammosteid evolution are also based on the variation of the shape of median plates, development of their tesserae covering, and changes in propor-tions of branchial plates (Halstead Tarlo, 1964a, Figs. 30, 31;Halstead, 1973). The resultant phylogenetic scheme (Halstead Tarlo, 1964a, Fig. 32) differs from the previous two only slightly in details -in terms of the evolution of Yoglinia (syn. Psammosteus) and Tartuosteus luhai (syn. Elgaia, gen. nov.) (Fig. 1D). The genera Guerichosteus, Schizosteus, and Psammolepis should be considered paraphyletic, according to his scheme. The species of the genus Schizosteus (member of Guerichosteus) gave the different evolutionary lineages -pycnosteids (some Schizosteus, Pycnosteus, Ganosteus, Tartuosteus, Yoglinia) and psammolepidids-psammosteids (some Schizosteus, Psammolepis, Rohonosteus, Traquairosteus, Obruchevia, Psammosteus, Karelosteus). The genus Psammosteus, according to Halstead Tarlo, evolved from Psammolepis (Halstead Tarlo, 1964a, p. 103), and Karelosteus is considered a sidebranch of Psammosteus. Traquairosteus with Obruchevia, Rohonosteus and Crenosteus are considered prospective descendants of the basal and more derived psammolepidids, respectively. His views on the evolution within Psammosteus in this work are systemless. Halstead Tarlo (1964a considered the similar characters of the branchials  Obruchev's thesis (1943c). B, Phylogeny of psammosteid families (Drepanaspidae, Psammosteidae, Obrucheviidae) by Obruchev, 1967 (in both cases the new names of taxa are used). C, Phylogeny of psammosteid species from E. Mark thesis (1955) excluding species -junior synonyms (Obruchev and Mark-Kurik, 1965). D, Tentative phylogeny of psammosteid species by Halstead Tarlo (1964a) with corrections of Tarlo (1967) and Halstead (1993) and excluding some species, junior synonyms and the Famennian non-psammosteid vertebrate 'Psammolepis' granulata (Blieck, 1991).
Robert H. Denison (1968, pp. 279, 281) attributed Psephaspis to pteraspids (this genus had previously been attributed by T. Ørvig to psammosteids) based on the absence of tesserae (they were mistakenly identified due to the fractured state of the remains). Denison also proposed that 'Psephaspis' bystrowi (Drepanaspis sp. according to Bystrow, 1959), described based on a tessera, should be named ? Drepanaspis sp. Obruchev in his works (Obruchev, 1967(Obruchev, , 1968) considered phylogenetic relationships within evolutionary lineages of psammosteids (Fig. 1B). According to him, drepanaspids gave rise to psammosteids, and obrucheviids evolved from psammosteids as a 'late aberrant offshoot' (Obruchev, 1967, p. 42). In that work Obruchev also showed recapitulation of ancestral characters in the Middle Devonian psammosteids. In his next article Obruchev specified that obrucheviids were classified as 'aberrant descendants of psammosteids, evolved from their earliest representatives' (Obruchev, 1968, p. 26). Halstead Tarlo (Halstead, 1973, p. 291) dis-covered pedomorphism in branchial plates of psammosteids Ganosteus and Psammosteus (Yoglinia). Scientists traced (Halstead Tarlo, 1964, p. 105-107;Halstead, 1973, p. 291) two evolutionary lineages in Psammosteus based on the shapes of branchial plates. Larisa I. Novitskaya (2004) used the classification by Obruchev (1964) in her revision of psammosteids. The genera Yoglinia and Crenosteus were synonymized with Psammosteus, Pycnolepis with Schizosteus and Rohonosteus with Tartuosteus. It was proposed that the most correct name for the suborder Psammosteida is Psammosteoidei (Novitskaya, 2004, p. 171). New data on Obruchevia and Perscheia (Elliott, Mark-Kurik, and Daeschler, 2004) showed that obrucheviids are more likely to be connected to a group including Pycnolepis, Pycnosteus and Tartuosteus. Sergey V. Moloshnikov (2009) introduced the genus Oredezhosteus which was later rejected (Glinskiy and Mark-Kurik, 2016). New morpho-histological characters of the plates of some representatives of Psammosteus made it possible to compare the histology of these taxa with Traquairosteus pustulatus and divide the genus Psammosteus on the two evolutionary lines (genera in the future) (Glinskiy and Nilov, 2017).

CHARACTERS, CODING METHODS AND DATA ANALYSIS
The new matrix is composed of 49 taxa and 120 discrete characters (49 characters are new) (see Appendix 1). The list of encoded taxa, specimen information and museum numbers are organized in a table (on-line only Supplements 1 and 2). 38 species from 16 genera (constituting the ingroup taxa) are coded as belonging to the morphologically best-known psammosteid heterostracans. To understand the relationship between pteraspids and psammosteids, juvenile Drepanapsis gemuendenensis was encoded as a separate taxon, because juvenile psammosteids show recapitulations (Gross, 1963;Obruchev, 1967). Thus, all the well known psammosteid species (except Traquairosteus pustulatus and species of Rohonosteus) are included here. Most included psammosteid taxa are known from numerous various elements of the exoskeleton; the exception is a group of taxa from the family Psammosteidae, known primarily from the branchial plates and tesserae: Psammosteus levis, P. tenuis, Karelosteus weberi, and the group of species 'Psammosteus' ramosus -'P. ' falcatus. For testing various hypotheses of derivation of the psammosteids (Janvier, 1996;Pernègre, 2002;Pernègre and Elliott, 2008;Randle and Sansom 2016), eight taxa of the order Pteraspidiformes, which may form the sister group to psammosteids, were encoded: Anchipteraspis crenulata, Doryaspis nathorsti, Errivaspis waynensis, Gigantaspis laticephala, Protaspis bucheri, Protopteraspis vogti, Woodfjordaspis felixi, and Xylaspis prima. Two cyathaspidiform taxa, Anglaspis PALAEONTOLOGY maccoulloughi (outgroup taxon in TNT) and Nahanniaspis mackenziei were chosen as outgroups, because they are very useful in polarizing character states within the analysis (Pernègre and Elliott, 2008;Randle and Sansom, 2016).
The character-taxon matrix was coded in NDE 0.5.0. by Roderic D. M. Page, 2001, and exported in TNT format implemented through the program Mesquite 3.2. Parsimony tree searches were conducted in TNT 1.5 (Goloboff, Farris, and Nixon, 2008) with space for holding 10000 trees. Two methods of data analysis were carried out-with equally weighted characters and with implied weighting (Goloboff, 2014). In both analyses the heuristic search mode (Traditional search) was used with the 'branch swapping' option (multiple Tree Bisection Reconnection, TBR) with 1000 replications and trees from RAM successively. In addition, the Nelsen strict consensus tree was calculated. Bremer tree support and standard Bootstrap (with traditional search, 5000 rep.) were implemented.
To increase resolution, avoiding polytomies in some clades (e.g. in Psammosteus), and to recover any phylogenetic signal, a second implied weighting analysis was conducted (Fig. 3B). Application of implied weighting resulted in 3 MPTs with a score of weighting (k=4) (MPTs length 23.55, CI = 0.610, RI = 0.850 and RC = 0.519. Relationships in the groups of taxa are somewhat better resolved in the current strict concensus tree than in the previous analysis. Protopteraspis vogti is shown here as the sister taxon to the clade, consisting of psammosteids and derived pteraspids. These results of the implied weighting data analysis support the cladistic Reconstructions of the cephalothorax of the major psammosteid genera with measurements of main plates. Reconstructions were made on the basis of the original material given in the list of taxa (On-line only Supplement 1). Numbers signify the selected characters, the following numbers in brackets represent character states. The slashes indicate variation of the character states within the given genus; character state, which is shown in the picture, is highlighted in bold. The variability of the branchial plates' shape in dorsal and ventral view is associated with their dorsally convex shape in the transverse-longitudinal direction. For median plates of some taxa, the juvenile states of the plates are shown. The red dotted lines indicate measurements of ratio as well as proposed canals of the lateral line system; the black dotted lines indicate concentric lines of growth, reflecting the morphology of the plates in ontogenesis.  Fig. 6B). The varied concavity value (k from 4 to 20) was implemented in the implied weighting analysis (length, topology, CI and RI indexes of MPT's and resulted strict consensus trees did not change). Shown here (Fig. 3B) is the result with concavity value k = 4. The positions of the psammosteid taxa are consistent with their positions in the equally weighted analysis. The results of the implied weighting data analysis with k from 1 to 3 represent some topological incongruence between trees from all previous analyses. According to these results, Doryaspis nathorsti is the most basal representative of Pteras- pidoidei clade (family Doryaspididae is not supported). This result appears only with three concavity values and thus requires additional testing. Tree support (standard Bootstrap with traditional search, 5000 rep.) shows high values of absolute frequencies for most clades of psammosteids (Fig. 3). Following the results of the phylogenetic analyses, previous classifications of psammosteids (Halstead, 1993;Novitskaya, 2004) are revised below.
Remarks. The most basal node of the ingroup (Fig. 2). The following apomorphies characterise this basal taxon of the ingroup: 6:1, 10:1, 21:3, 27:1, 30:1, 33:2, 49:2, 67:3, 76:2, 79:2, 83:1. All Drepanaspis gemuendenensis specimens (including juvenile specimens) have fields in the cephalothorax interpreted as fields of tesserae. Following the supposition of Gross (1963, p. 149), I assume that in the case of the most juvenile Drepanaspis, the smallest tesserae (more likely discrete single and complex elements) must be present in the fields of tesserae. So they probably grew more slowly than the plates, not constraining the growth of the latter (unfortunately, the juvenile exemplar of Drepanaspis was lost during WWII). The juvenile Drepanaspis differs from the adult specimens by the more consolidated cephalothorax, which has the following contacts: planar rostro-pineal (8:1), rostro-orbital (6:0), dorso-postorbital, dorso-cornual contacts; triangular pineal plate (14:0). It is possible that the juvenile specimen shows some of the characters of more distant ancestors (recapitulations) (Obruchev, 1967), which would be lost in adults of the same species. Anyway, the juvenile specimen of Drepanaspis falls well in the psammosteid clade as the basal member.
Taxa included. Guerichosteus kozlowskii Halstead Tarlo, 1964, G. heterolepis (Preobrazhensky, 1911. Remarks. These near-basal members of the ingroup are characterized by four synapomorphies (18:2, 79:3, 88:2, 109:3). 'Hariosteus' is known from fragmentary material, which makes revision difficult (presence of the dorsal lamella of the branchial plate is not confirmed), but the characters of the ornamentation indicate that this genus probably is a junior synonym of Guerichosteus. Guerichosteus heterolepis (Preobrazhensky, 1911) is included in the genus Guerichosteus on the basis of the similar morphology of the branchial plates and ornamentation pattern (Glinskiy and Pinakhina, 2018). The genus Schizosteus, previously included in this family (Halstead Tarlo, 1964a), is excluded here given its synapomorphies with higher psammosteids. Despite their similarity, the two species of Guerichosteus do not form a clade, which can be explained by the fragmentary nature of the available material.
Taxa included. Obruchevia heckeri (Obruchev, 1936); Perscheia pulla Elliott, Mark-Kurik et Daeschler, 2004. Remarks. This group is characterized by two apomorphies (33:5, 117:2). Obrucheviids have plates with a sculptured surface of pits, radial grooves or mounds, the outer layers being infilled by pleromic dentine. Obruchev (1941) believed that the plates were located in soft tissues, but later decided that the surface of the plates was not covered with epidermis (Obruchev, 1968). The presence of lateral line canals on the dorsal plate in Perscheia shows that the dorsal surface of obrucheviids was mostly covered by soft tissue in life (Elliott, Mark-Kurik, and Daeschler, 2004, p. 32). The narrow and long branchial plates of Obruchevia (Obruchev, 1968, p. 26) indicate its origin from the early psammosteids, in which the plates had not yet been shortened. I assume that the dorsal exoskeleton in the obrucheviids was strongly consolidated, and the fields of tesserae did not develop or were extremely narrow. The following characters support this conclusion: the presence of lateral canals of the lateral line on the dorsal plate in Perscheia, and the presence of the paired branchial sinus (sensu Obruchev, 1941) on the dorsal plate. Obruchevia was compared with Cardipeltis, but the similarity of dorsal plates was suggested to be caused by convergence (Obruchev, 1941(Obruchev, , 1968. Traquairosteus pustulatus (Traquair, 1897) is known only from an elongated dorsal (probably ventral) plate (Tarlo, 1961, Fig. 6) that is typical for Psammosteidae. Traquairosteus pustulatus isn't included in the current analysis. Moreover, herein Traquairosteus is the proposed generic name of the group of species 'Psammosteus' ramosus -'Psammosteus' falcatus because all these psammosteids have tubercles on aspidin mounds. The widely spaced tubercles on the main plates of Traquairosteus should be interpreted as the primordial tubercles of the rooted progressive micromeric elements (tesserae) as on the dorsal plate of 'P'. falcatus (Obruchev and Mark-Kurik, 1965, p. 66, pl. 88, Fig. 2).
Remarks. The following unique apomorphies characterize only this most basal taxon of Psammosteidae: 30:3, 57:3. The shape of the primordium (juvenile stage of plate) of Vladimirolepis proia comb. nov. is identical to the shape of the dorsal plate of Schizosteus striatus. Also, the primordium of the ventral plate has a V-shaped posterior notch (64:1) as in the more primitive forms (Schizosteus). These characters strongly distinguish Vladimirolepis gen. nov. from Placosteus. The pattern of ornamentation with tubercles fused in ridges is similar to those in Schizosteus striatus and Placosteus.
Remarks. The two genera Vladimirolepis gen. nov. and Pla co steus gen. nov. are the two early-branching members of the family Psammosteidae; they still have the eurybasal branchial plates (83:2) but strongly differ in general morphology of the juvenile median plates (primordia).

Diagnosis. As for 'Psammolepis' alata
Remarks. Placosteus alatus is very similar to Pc. undulatus but differs in branchial plate shape and some ornamentation characters. The dorsal plate of Pc. alatus was possibly fully covered by tesserae.
Genera included. Elgaia gen. nov., Psammosteus Agassiz, 1844, Karelosteus Obruchev, 1933, Traquairosteus Halstead Tarlo, 1964 Remarks. This group is characterized by the following basal apomorphies (57:5, 60:2). Currently this group (family Psammosteidae sensu Halstead Tarlo, 1965;Novitskaya, 2004) includes Psammosteus and provisionally retained Karelosteus (Glinskiy and Mark-Kurik, 2016;Novitskaya, 2004). Herein two more genera, Elgaia gen. nov. and Traquairosteus, are included. With the availability of new data on psammosteids, the list of their unique apomorphies must be refined. Presumably all representatives of the family Psammosteidae appear to lack cornual plates, as in the associated cephalothorax of Psammosteus megalopteryx (Woodward, 1911, pl IX, Fig. 1). Halstead Tarlo (1965, p. 121) redescribed this specimen and identified the area with isolated tubercles as the remains of the cornual plate. However, such isolated tubercles also can be attributed to numerous discrete micromeric elements or to the ornamentation, detached from the main plates (tubercles are weakly connected with the plates of Psammosteus, see Mark-Kurik, 1999). The branchial openings presumably are located behind the reduced branchial plates through the branchial sinus of the dorsal plates (Fig. 2). The characters of the Psammosteinae appear in Elgaia luhai (Mark-Kurik, 1965) comb. nov. In the species of Psammosteus, a full reduction of the anterior part of the median margin of the branchial plates is seen, and cyclomorial, mostly polygonal concentric tesserae of the basic morpho-histological type are formed (Glinskiy and Nilov, 2017). Traquairosteus should be nested closer to Psammosteus, as has been noted earlier (Glinskiy and Pinakhina, 2018;Glinskiy, 2018). In Traquairosteus the tesserae and other elements of progressive type may be rooted on the main plates, forming the strange unique ornamentation . Traquairosteus is grouped with a question with the species 'Psammosteus' ramosus Glinskiy, 2017, 'P'. pectinatus Obruchev, 1965, 'P'. kiaeri Halstead Tarlo, 1964, and 'P'. falcatus Obruchev in Gross, 1942, from the Upper Devonian (Frasnian) of the Main Devonian Field, Scotland and Ellesmere Island. These psammosteids have extremely stenobasal branchial plates (l:w≤0.3) and discrete tesserae of the progressive morpho-histological type (odontodelike shape), which represent the circular stage of the tesserae development (Glinskiy and Nilov, 2017). Histology of the plates of these psammosteids is very similar to that of Traquairosteus pustulatus (aspidin mounds, surmounted by tubercles with massive marginal crenulations (Traquair, 1897, pl. XI, Fig. 4) and differences of tubercle shapes (Traquair, 1897, Additional notes…, pl. XI , Fig 4)).
In addition, it should be noted that in the reconstruction of Psammosteus by Halstead Tarlo (Tarlo, 1961, Fig. 5), the proximal parts of the branchial plates are freely located in the cephalothorax. According to this author's opinion, Psammosteus has movable branchial plates (Halstead, 1973, p. 291). However, numerous specimens of Psammosteus branchial plates with the strong marginal lifetime abrasion on the substrate indicate the fixed location of these plates in the cephalothorax. The medial parts of Psammosteus branchial plates definitely grew under anterolateral lobes of the dorsal plate.

Derivation of name.
In memory of Elga Mark-Kurik, Estonian palaeontologist and stratigrapher, gender feminine.
Type species. Tartuosteus luhai Mark-Kurik, 1965. Remarks. Tesserae are not known in the median plates of Tartuosteus species (Elliott and Mark-Kurik, 2005, p. 101), the dorsal and branchial plates of Elgaia gen. nov. are different from those of Tartuosteus. This new genus is the most basal representative of the subfamily Psammosteinae and differs from Psammosteus by the relative length and shape of branchial plate medial margin (83:3) and mostly scale-like, dense rooted tesserae on the dorsal plate (60:1+2, 62:0).

Diagnosis. As for the genus.
Remarks. The dorsal plate of E. luhai (Mark-Kurik, 1965), comb. nov., is completely covered by scale-like, cyclomorial tesserae with eccentric position of the primordial tubercles (Fig. 4 G-I), they are consolidated with aspidin layer. A few cyclomorial tesserae are with concentric position of the primordial tubercles as in Psammosteus (Fig. 4 G). Branchial plates are stenobasal ( Fig. 4 A, B, E, F) with length/width ratio 0.6.
Species included: P. bergi (Obruchev, 1943); P. maeandrinus Agassiz, 1844;and P. praecursor Obruchev, 1947. Provisionally included species: Psammosteus asper Obruchev, 1965;P. cuneatus Obruchev, 1965;P. levis Obruchev, 1965;P. livonicus Obruchev, 1965;P. megalopteryx (Trautschold, 1880); and P. tenuis Obruchev, 1965. Remarks. Currently, there is no way to divide the genus Psammosteus into several genera, as there is not enough information on the median plates of many species, their natural margins and interspecific variation. The different morphologies of the juvenile and adult branchial plates and the morpho-histological types of tesserae (Glinskiy and Nilov, 2017) suggest several evolutionary lines (potentially genera). The first group (P. bergi,P. precursor,P. maeandrinus;Fig. 2) has triangular branchial plates with an oblique margin of ornamentation on their dorsal side (Psammosteus); the second group (all other species provisionally included in Psammosteus + Karelosteus weberi Obruchev 1933) has extremely stenobasal branchial plates (81:4) and small zones of ornamentation (87:2). According to the results of phylogenetic analysis, Karelosteus weberi Obruchev 1933 is included in the Psammosteus clade. However, the genus Karelosteus is provisionally retained, because it is not known how many species of the large Psammosteus clade will be divided into several genera following a future analysis of all available material. Psammosteus livonicus was described in Oredezhosteus (Moloshnikov, 2009), and later this genus was revised (Glinskiy and Mark-Kurik, 2016). Possibly, in the future P. livonicus and the related species P. asper might be included also in Karelosteus. The current analysis included corrected data on the dorsal plate of Psammosteus livonicus, which should be expanded by the narrowed part in the caudal direction (error orientation: Mark-Kurik 1968, Fig. 11A;Elliott and Mark-Kurik 2005, Fig. 4
Remarks. This group is characterized by the following apomorphies: 108:4, 114:4. This terminal group of psammosteids has extremely stenobasal branchial plates (like other derived psammosteids) with large tubercles that are located on aspidin mounds (detailed histological and topographic comparisons are needed; this work is held jointly with D. K. Elliott). Widely spaced tubercles on the median plates of Traquairosteus should be interpreted as rooted progressive micromeric elements (Glinskiy and Nilov, 2017) as on the dorsal plate of Traquairosteus ? falcatus (Obruchev and Mark-Kurik, 1965, p. 66, pl. 88, Fig. 2). The holotype of Traquairosteus pustulatus probably has imprints from the detached tesserae. Apparently Traquairosteus pustulatus and Traquairosteus ? falcatus Obruchev in Gross, 1942 are the same species named on different exoskeleton plates. Differences in the ornamentation of Traquairosteus pustulatus might be a consequence of topographic variability (branchial plates with complicated ornamentation in contrast to median plates with simple ornamentation) and different conditions of preservation between the locality of Scaat Craig (Scotland) and the Main Devonian Field. The described elongated plate (Tarlo, 1961, Fig. 6) of T. pustulatus is probably the ventral plate.
Psammosteidae family incertae sedis 'Schizosteus' perneri (Růžička, 1929) Diagnosis. As the species description given by Vaškaninová and Kraft (2016).  (Růžička) might be attributed to guerichosteids on the basis of the presence of the typical ornamentation (fairly smooth margins of tubercles without marginal crenulations ) and extremely narrow branchial plates. However, the branchial plates of this species have lost the dorsal lamella (79:4) and have more isometric eurybasal shape (81:1); the derived Middle Devonian psammosteids have such synapomorphies. Obruchev and Mark-Kurik (1965, p. 15) stated that 'S'. perneri should be attributed to the more basal genus (close to Drepanaspis) and did not include this species in the genus Schizosteus. Ornamentation of branchial plate 'S. ' perneri (holotype) is fairly abraded (Vaškaninová and Kraft, 2016, pl. 1, Figs. 2d-f). The limits of variation in the ornamentation on the ventral side of plate are unknown. More data on 'S' . perneri are needed. Furthermore it is necessary to postpone naming of a new genus for 'S. ' perneri until the new information on the problematic 'psammosteid-like' genus Sedowichthys and the dorsal side of the branchial plate of 'Hariosteus' becomes known. In any case, the guerichosteids and 'Schizosteus' perneri are the early branching lineages derived from the basal psammosteids.
Type species. Schizosteus asatkini Obruchev, 1940. Species included. S. asatkini Obruchev, 1940, Schizosteus striatus Gross, 1933 Remarks. Psammosteids with long and narrow branchial plates. Monophyly of the genus Schizosteus is not supported due to the lack of data on the central (median) plates of the adult specimens of S. asatkini and S. striatus. Thus, it is unknown whether the character 32:1 (straight anterior margin of the dorsal plate) describes only Schizosteus striatus, or both species of Schizosteus. To clarify the position of Schizosteus asatkini Obruchev, 1940, the specimens of Schizosteus ? sp. (Glinskiy, 2014) were included, which are probably worth attributing to this species. Pycnolepis splendens (Eichwald, 1844) was previously attributed to the genus Schizosteus (Novitskaya, 2004;Obruchev and Mark-Kurik 1965). The validity of the genus Pycnolepis Halstead Tarlo, 1964 is supported on the results of this phylogenetic analysis, and it is placed in the clade Pycnosteidae. 'Schizosteus' perneri (Růžička, 1929) is attributed here to the basal evolution line of psammosteids that is close to guerichosteids.
Remarks. The current analysis includes only two species of this genus with a generally triangular shape of the branchial plates. Tartuosteus ? (Rohonosteus) ornatus (Rohon, 1899) and T. ? zheleznogorskensis Moloshnikov, 2009 are not considered here due to the poor material. The unique specimen of the last species is possibly Psammosteus with pathology of branchial plate development and imprints of detached tesserae on ventral side. Elgaia luhai (Mark-Kurik, 1965) comb. nov. is placed in the clade Psammosteidae as the basal taxon. Tesserae are not known in the main plates of Tartuosteus species (Elliott and Mark-Kurik, 2005, p. 101). Tartuosteus and Psammolepis have a synapomorphy -their postorbital plates do not have the posterior process (24:3).

Discussion and conclusions
The results of the analyses show that psammosteids belong to the Pteraspidiformes. The derived pteraspids (Pteraspidoidei sensu Pernègre andElliott, 2008 with Doryaspididae Heintz in Tarlo, 1962) form the sister group for the psammosteids. Results of both analyses indicate that psammosteids originate from the basal pteraspidiformes as Protopteraspis (Blieck, Elliott, and Gagnier, 1991). These results contradict a hypothesis that some pteraspids, representatives of Doryaspididae Heintz in Tarlo, 1962, must be united in one clade with psammosteids (Pernègre, 2002;Sansom, 2016, 2017). The present analyses do support the classically accepted monophyly of psammosteids (suborder Psammosteoidei sensu Psammosteida Kiaer, 1932emend. Tarlo, 1962 and for the most part the classical arrangement of families: Drepanaspididae; Guerichosteidae; Obrucheviidae; Pycnosteidae; Psammolepididae; and Psammosteidae. This study provides a basis for a deeper investigation into the intra-relationships of the psammosteids. The ingroup taxa have been chosen to represent only the best and most completely known genera. During the analysis the monophyly of some psammosteid genera (e.g. Schizosteus, Tartuosteus, Pycnosteus, Psammolepis) and validity of doubtful taxa (e.g. Pycnolepis splendens) were checked. From the genus Psammolepis sensu Mark-Kurik, 1965, Halstead Tarlo, 1965, are excluded taxa Vladi mi ro le pis proia comb. nov., Placosteus undulatus and Pla co steus alatus comb. nov. From Tartuosteus sensu Mark-Kurik, 1965, Halstead Tarlo, 1965 is excluded Elgaia luhai comb. nov. Elgaia gen. nov. is the most basal taxon of the subfamily Psammosteinae. Two new subfamilies (Placosteinae subfam nov. and Psammosteinae Traquar, 1896) are erected. Also proposed is inclusion of a number of 'Psammosteus' species with progressive types of tesserae and peculiar histological structure of other plates in the genus Traquairosteus (Glinskiy and Pinakhina, 2018;Glinskiy, 2018). Guerichosteus, Schizosteus and Psammosteus are still very poorly known by median plates and represent here a paraphyletic groups (additional material is needed). New material on the median plates of these taxa and Karelosteus is needed for a more correct analysis. The following taxa, used in the analysis, are not assigned to the existing families (Psammosteoidei incertae sedis): 'Schizosteus' perneri (Růžička, 1929); Schizosteus Obruchev, 1940;Tartuosteus Obruchev, 1961.