New classification of the family Sejidae (Acari: Mesostigmata) based on morphological analyses

Relationships in the family Sejidae are examined based on analysis of 74 morphological characters. Synonymy of Willmannia Balogh and Liroaspis Banks with Sejus Koch is supported, while synonymy of Epicroseius Berlese is not. Nine species of Sejus are transferred to Epicroseius, and Sejus congoensis and S. solaris are transferred to Africasejus n. gen. A total of five genera is recognized in Sejidae: Sejus, Epicroseius, Adenosejus Lekveishvili and Krantz, Africasejus, and Zuluacarus Trägårdh. New diagnoses are proposed for the family and its genera.


Introduction
The family Sejidae is a moderately diverse group of mesostigmatid mites. The family is widely distributed in the warm temperate and tropical regions of the world, with its constituent species often associated with wood. The second nymphal instar, the deutonymph, may be dimorphic, with one ''normal'' (homeomorphic) morph and a heteromorphic morph specialized for phoresy, often on cerambycid beetles.
Recent studies (Lekveishvili and Klompen 2004b) have clarified the family level groupings in the infraorder Sejina (including the surprising affinities with the Heterozerconidae and Discozerconidae), but relationships within Sejidae remain to be resolved. The primary aim of the current study is to resolve relationships within the Sejidae and to review the existing generic classification for the family.
The current study is based on morphological data only. To do this the taxon set is drastically increased relative to the infraorder revision study (Lekveishvili and Klompen 2004b). In addition, the number of morphological characters is also expanded. A secondary goal of this study is to discuss explicitly these characters in terms of homology decisions and character state designations.

Historical review
The taxonomic history of the family Sejidae is one characterized by wild swings in generic concepts, and a distinct lack of integration. The family was defined by Berlese (1913) based on the genus Sejus Koch, 1836 with type species Sejus togatus Koch, 1836. He also included a second genus Epicroseius Berlese, 1905, but did not mention Liroaspis Banks, 1902, with type species L. americana Banks, 1902. Confusion regarding the type species and genus ensued when Oudemans (1938) noted that Koch (1843) had designated Sejus viduus Koch, 1843 as type species for Sejus and proposed the name Dwigubskia Oudemans, 1938 as a replacement name for Sejus Koch, 1836. The issue was resolved by Lindquist and Evans (1965) who noted that the assignment by Koch (1843) of Sejus viduus as type species for Sejus was contrary to the rules of the International Code of Zoological Nomenclature, as the generic name was already associated with Sejus togatus in 1836. Thus Dwigubskia is a junior synonym. They also synonymized Sejus Koch, 1843with Cheiroseius Berlese, 1916. In subsequent years two more genera were proposed: Zuluacarus Trägårdh, 1906 andWillmannia Balogh, 1938. In a revision of the entire family Hirschmann and colleagues (Hirschmann 1991;Hirschmann et al. 1991) described many new species, redescribed several others, and synonymized Epicroseius, Liroaspis, Zuluacarus, and Willmannia with Sejus. Instead, Hirschmann proposed 10 species groups, without designating any formal taxonomic rank for any of these. Finally, after examining specimens of Sejus krantzi Hirschmann, 1991 andS. manualkrantzi Hirschmann, 1991, both described based solely on drawings in a reference book, Lekveishvili and Krantz (2004) synonymized the two species, and erected the genus Adenosejus Lekveishvili and Krantz, 2004, to accommodate the species. This has brought the total number of nominal genera to six.

Taxon selection
Twenty-seven species of Sejidae, two species of Ichthyostomatogasteridae, one species each of Uropodellidae and Heterozerconidae, and Archaeopodella scopulifera were included in the analysis. Inclusion of Heterozerconidae was based on the results of our previous study where heterozerconine families are included in Sejina (Lekveishvili and Klompen 2004b). Efforts were made to include the type species of all named sejid genera and representatives of all of Hirschmann's (Hirschmann 1991) ''species groups''. Exceptions include the type species of Epicroseius and Zuluacarus (E. angelioides Berlese, 1905 andZ. termitophilus Trägårdh, 1906) which were excluded because of insufficient data for analysis. Species of Sejidae for which fewer than 50% of characters could be coded were also excluded from the analysis. In addition to described species, five new species were included. Microgynium incisum Krantz, 1961 (Microgyniina) and Zercon sp. (Zerconina) were selected as outgroups. Including more distant outgroups was problematical in terms of establishing defensible homologies.
Characters were coded by direct observation of specimens, or based on original descriptions and drawings (when specimens were not available). Seventy-four characters were coded, including gnathosomal, idiosomal and leg characters of all instars except the heteromorphic deutonymphs. Description of characters and character states and the final character matrix are presented in Appendices 1 and 2, respectively.

Analysis and results
The analysis yielded 73 equally most parsimonous trees (Figure 1) (length 259; CI50.37; RI50.66). Sejina s.s., that is without Heterozerconina, is supported, in contrast to molecular-based analyses which support inclusion of Heterozerconina within Sejina (Lekveishvili and Klompen 2004b). Within the limits of this analysis, Heterozerconina is the closest relative of Sejina s.s., and this relationship is relatively well supported (BS53).
The families Uropodellidae, Ichthyostomatogasteridae, and Sejidae are monophyletic, with Archaeopodella included in Ichthyostomatogasteridae. Ichthyostomatogasteridae and Sejidae are sister groups although support for this relationship is relatively weak (BS51). The family Ichthyostomatogasteridae itself is also weakly supported (BS51). New sp. 1 from the Philippines groups with Archaeopodella, and may belong in that genus. Relationships of Archaeopodella, Asternolaelaps, and Japanasternolaelaps are not resolved.
The family Sejidae is relatively well supported (BS53). There are four distinct groups defined in the family, three of which correspond to existing genus-level groupings. The (nova-meso) lineage ( Figure 1) is moderately well supported (BS52). It includes the type species of the genera Sejus, S. togatus, Willmannia, W. sejiformis, and Liroaspis, L. americana. It will be referred to as Sejus. Similarly, the (zimm-marq) lineage is well supported (BS54) and includes all species previously assigned to the genus Epicroseius. This clade also includes some species (e.g. S. klakahensis, S. marquesanus, and S. tanganicus) previously classified within Sejus. These two lineages, Sejus and Epicroseius, form a well-supported clade (BS53). The two remaining lineages, one including S. solaris and S. congoensis, the other Adenosejus krantzi and n. sp. 2, are weakly supported (BS51). Both occupy more basal positions relative to Epicroseius and Sejus.
Relationships among the species of the (nova-meso) lineage are not well resolved, although a few species groups are defined. Sejus posnaniensis and Willmannia sejiformis are clustered together (BS51). Sejus stebaevi, L. americana, S. polonicus, S. togatus, S. hinangensis, and S. rafalskii form a clade which is also weakly supported (BS51). The lineage of (aust-meso) has relatively better Bremer support (BS52), although relationships within the group are not resolved. Relationships of S. novaezealandiae, n. sp. 3 and n. sp. 4 relative to other species of Sejus are not resolved.
Overall resolution and support levels are relatively weak, a situation resulting from inclusion of several taxa with large amounts of missing data. Exclusion of such taxa (as done in the more limited morphology based analysis by Lekveishvili and Klompen 2004b), achieves some improvements in both resolution and support level (results not shown), but is far less general, and results in some odd relationships, including inclusion of Archaeopodella in Sejidae. Overall, we prefer the total evidence analysis as explaining more of the overall character state pool and being more general, even if resolution and support levels are lower.

Relationships
Our recent analysis of relationships among the families of Sejina and Heterozerconina (Lekveishvili and Klompen 2004b) suggested the possibility that the family Sejidae was paraphyletic if it excluded Archaeopodella. This result was somewhat surprising given that Archaeopodella is generally considered to be intermediate between Ichthyostomatogasteridae and Sejidae (Athias-Henriot 1977). Notably, this result was based entirely on morphological characters, as molecular data were not available for the only described species in this genus, Archaeopodella scopulifera. The expansion of the current morphological character and taxon set suggest a different arrangement of Archaeopodella (within Ichthyostomatogasteridae), and monophyly of the family Sejidae. Notably, the arrangement of Archaeopodella scopulifera and n. sp. 1 (with all instars found on a Philippine rat), appears to make ecological sense, as it groups all taxa associated in some form with vertebrates into a single family, the Ichthyostomatogasteridae. Asternolaelaps collections include several records of vertebrate nest associations (Sellnick 1953;Womersley and Domrow 1959), and the only known collection of Japanasternolaelaps was from a stable (Hirschmann and Hiramatsu 1984).
Within Sejidae, the established pattern of relationships is somewhat consistent with the species groups of Hirschmann (1991), with one exception: the odd placement of his Sejus krantzi (posnaniensis group) and S. manualkrantzi (solaris group). These two taxa have been synonymized (Lekveishvili and Krantz 2004), and re-classified within the genus Adenosejus, a lineage that is quite distinct from most other Sejidae (e.g. Lekveishvili and Klompen 2004b). The current analysis also provides some higher-order arrangements for the multitude of species groups proposed by Hirschmann. For example, the (sola-cong) lineage ( Figure 1) unites the solaris and congoensis species groups, the (nova-meso) lineage unites the posnaniensis, camerunis, boliviensis, togatus, rafalskii, and stebaevi groups, and the (zimm-marq) lineage (our Epicroseius) is equivalent to Hirschmann's tanganicus group.

Classification
Based on results from the current and previous (Lekveishvili and Klompen 2004b) analyses, we support the synonymy of Willmannia Balogh and Liroaspis Banks, as proposed by, respectively, Hirschmann  and Lindquist and Evans (1965). The type species of these genera are included in the (nova-meso) lineage. On the other hand, we reject the proposed synonymy  of Epicroseius with Sejus. The (zimm-marq) and the (nova-meso) lineages are quite distinct. Moreover, the (zimmmarq) lineage includes all species previously referred to as Epicroseius. Although the type species of Epicroseius, E. angelioides, is not explicidly included in the analysis, the limited amount of available data (e.g. absence of claws on legs I, two-pronged gnathotectum, divided pygidial shield) are fully consistent with placement in the (zimm-marq) lineage. This lineage can thus be classified as Epicroseius Berlese. It also includes n. sp. 5 from Australia. Its sister group, the (nova-meso) lineage, which includes the type species of Sejus, S. togatus, will retain the generic designation of Sejus. It includes n. sp. 3 and n. sp. 4, both from Australia. Sister group relationships between Epicroseius and Sejus are supported by the shared presence of posterior projections in the protonymph. n. sp. 2, recovered for all instars from litter in the Great Smoky Mountains National Park, groups with A. krantzi, and will be classified within Adenosejus. Among the species included in this analysis, this leaves only S. solaris and S. congoensis, which do not fit into any of the generic groups proposed above. For this group we propose a new genus, Africasejus, n. gen.
The final named genus group in Sejidae is the monotypic genus Zuluacarus (Trägårdh, 1906) described from South Africa. As noted above, the species and genus descriptions of Zuluacarus are very incomplete, preventing us from including this species in the analysis. Only a few characters can be scored. The presence of a two-pronged gnathotectum is a synapomorphy shared with Epicroseius, but the presence of an ambulacrum on legs I (a plesiomorphic character) is inconsistent with such a placement. None of the other characters that can be scored (dorsal and posteromarginal shields in adults, two pairs of projections, position of genital orifice) are informative, and so we tentatively place Zuluacarus near Epicroseius. Notably, this arrangement is similar to that proposed by Hirschmann (1991), who placed S. termitophilus in the tanganicus group (our Epicroseius). The genus and species are classified as incertae sedis.

Diagnoses
Based on the above, we propose the following updated diagnoses for the family Sejidae and the genera included in the Sejidae.
Sejidae Berlese, 1913 Diagnosis Posterior edge of the female genital shield at the level of the posterior edge of coxae IV; female st1 platelets (if present) not fused with each other or with the st2 platelets (except in S. stebaevi, in which the st1 platelets are fused to each other); posterior projections (tails) present in at least the larva (except in S. congoensis which never carries such projections).

Remarks
Most species have posterior idiosomal projections in all instars, but the larval projections may be lost in the protonymph (Sejus solaris, Adenosejus) or deutonymph (S. posnaniensis and possibly S. sejiformis). The number of genital setae in the female (at least three pairs) has been used as a key character for the family Sejidae (Krantz 1978;Evans and Till 1979). However, Adenosejus krantzi, S. hinangensis, S. rafalskii, Epicroseius porosus, E. tanganicus, E. abinashi, and n. sp. 5 have only two pairs of genital setae and S. congoensis has only one pair. Clearly this character is not unambiguous.

Remarks
Setiform hypostomal setae 1 is a plesiomorphic character shared by most groups of Mesostigmata including Uropodellidae and Ichthyostomatogasteridae. Presence of large, spiniform leg setae and enlarged idiosomal glands are characters not found in any of those groups or in other Sejidae. Adenosejus appears to be most basal in the family Sejidae.

Remarks
Africasejus occupies a basal position relative to Sejus+Epicroseius. Both species included were described from specimens discovered under the bark of tropical hardwood, imported into Poland. The OSAL collection includes an additional specimen of Africasejus congoensis (OSAL 14363) recovered from tropical hardwood imported from tropical Africa into a US harbour.

Epicroseius Berlese, 1905
Diagnosis Two-pointed gnathotectum; dendritic processes lateral of tritosternum; presence of seta pv3 on tarsus IV. The latter is an unique character for Mesostigmata.

Remarks
Validity of several of these species is unclear since they were described from single specimens in poor condition. These include: Epicroseius klakahensis, E. marquesanus, E. vitzthumiangelioides, E. vizthumiseurati, E. oblitus, and E. javensis.

Diagnosis
Sternal platelets st1 and st2 in female are not fused (character 55); posterior mesonotal shields in female are not fused to each other.

Remarks
Balogh separated the genus Willmannia from Sejus based on the absence of posterior projections in adults (immatures are unknown). Since these projections are present in the larva and protonymph of S. posnaniensis, a species which is very close to W. sejiformis (it may even be a synonym), it can be assumed that they are present in the same stages of W. sejiformis as well. The absence of posterior projections in the adults is apomorphic for S. sejiformis and S. posnaniensis, but continued recognition of Willmannia would leave a paraphyletic Sejus.
The species group including Sejus stebaevi, S. americana, S. polonicus, S. togatus, S. hinangensis, and S. rafalskii has characteristic anchor-shaped lateral extensions of the tritosternum in adults; in males the mesonotal and pygidial shields are fused forming an opisthonotal shield (except in S. stebaevi; male of S. rafalskii is unknown).
Sejus bakeriarmatus is based on a drawing of Liroaspis armatus in a textbook (Baker and Wharton 1952). Validity of this species is unclear.
Chelicera (Characters 1, 2) 1. Pilus dentilis: [0] present (Figure 2A  The derived state is shared by Sejus togatus, S. polonicus, S. hinangensis, S. rafalskii, and S. stebaevi. Gnathotectum (Characters 3-6, Figure 3) The shape of the anterior edge of the gnathotectum varies between families of Mesostigmata and within the Sejidae. It can be curved ( Figure 3B, G-I), triangular ( Figure 3C-F), or blunt ( Figure 3A). It may also have from one to three prominent points arising from the anterior margin. Finally, the anterior margin may be serrate or smooth.  In Sejidae seta hyp1 is modified as a membranous structure with a broad base and a curved tip (Figure 4). The shape of that tip varies a little among species and instars. Adenosejus krantzi and n. sp. 2 are the only members of the Sejidae with a setiform hyp 1 seta, the state found in all outgroups. The slightly inflated seta hyp1 of Archaeopodella seems to be somewhat intermediate between membranous and setiform, and is coded as semi-membranous. 8. Corniculi: [0] horn-shaped; [1] massive, bifurcate or trifurcate; [2] flat, lobed, membranous.
Horn-shaped corniculi are typical for most Mesostigmata and are retained in Sejidae (Figure 4). Modifications include massive trifurcate corniculi in Ichtyostomatogasteridae, and flat, lobed, and even membranous forms in Heterozerconidae.
Dorsal shields (Characters 9-26, Figure 5) Sejina can have up to eight dorsal shields. The usually somewhat triangular podonotal shield covers the anterior half of the dorsum, the posterior half may carry up to four median (mesonotal) shields plus a pygidial shield. Larvae have podonotal and pygidial shields, but lack mesonotal shields. The mesonotals may or may not be added in protonymphs, but they are never fused. In deutonymphs and adults, the mesonotals can be fused to each other and/or to the podonotal or pygidial shields. Deutonymphs and adults of Uropodellidae add separate shields lateral to the podonotal and opisthonotal shields, lateral shields. The only species of Sejidae where elements of these shields may be present is n. sp. 2, which has laterally extended mesonotal shields. Coding for these characters can be problematic when shield fusion occurs. The coding approach chosen uses area as the basic unit, for example anterior mesonotal area or pygidial area. Thus a holodorsal shield, as found in adult Asternolaelaps and Heterozerconina, is coded as present for podonotal, all mesonotals, lateral, and pygidial shields as all those regions are covered by shields. There is only one sejid species, Africasejus solaris, that has the anterior mesonotal shields fused to the podonotal shield in the male. This type of fusion has not been observed in any other species of the family. This hypothesis of homology is based on (1) the presence of slightly differentiated areas resembling the mesonotal shields on the large anterior shield, and (2) the presence of a large number of additional setae on that anterior shield which can only be explained by assuming fusion of the anterior mesonotals and the podonotal.
The mesonotal shields in Epicroseius species are surrounded by secondarily sclerotized cuticle, an arrangement designated as ''partially coalesced''. Coding of this character has to be approached with some care. The level of secondary sclerotization appears to be age dependent. In one (new) species of Epicroseius we have observed specimens that can be assigned, conditionally, to four groups ( Figure 6) depending on level of secondary sclerotization. In the first group the four mesonotal shields are well defined. In the second group the mesonotal shields are embedded in a poorly delimited zone of secondary sclerotization. Specimens in the third group show an almost rectangular, and much better defined, zone of sclerotization around the mesonotals and in the most sclerotized specimens (fourth group) the rectangular shield of secondary sclerotization is very well developed and the original mesonotal shields are hard to distinguish.
Because type material of most Epicroseius species was not available to us we had to rely on original descriptions and drawings. Trägårdh (1951) states that E. zimmermani has ''a median shield, with two cuneiform areas which are presumably the remnants of anterior mesonotal shields, now almost completely coalesced with the postero-median shields, the lateral and posterior margins of which are rather indistinct''. In his description of E. porosus Domrow (1956) describes the median shield as ''rectangular, entirely covered by striated cuticle, except for four small transverse exposed areas''. It seems, that in older adults of Epicroseius the mesonotal shields are always coalesced, but in teneral specimens the secondary sclerotization may not always be present. In the diagnosis of Epicroseius angelioides, the type species of genus, Berlese (1905) states that the posterior (pygidial) shield is divided in two by a longitudinal strip of soft cuticle. Later, in the description of E. scutatus, he refers to the pygidial shield as two shields, fused together. Trägårdh (1951) follows Berlese in his description of E. zimmermani. In contrast, Domrow (1956) describes it as ''posterior shield with sclerotized median longitudinal strip without setae; lateral margin well sclerotized, remainder reticulated''.
All specimens of Epicroseius have a pygidial shield with a median longitudinal strip that suggests a groove rather than a line of fusion. We call this the median groove. The sclerotization level of the groove area is about the same as that of the lateral strips which are also present in all Epicroseius ( Figure 7A). The groove is absent in most other Sejina or in the outgroups. N. sp. 2 shows a similar median groove dividing the pygidial shield, but lacks the lateral strips ( Figure 7B).
Posteromarginal shields (Characters 27-29, Figure 5) Posteromarginal shields are not developed in immatures. In adults they may be present or absent. They can be fused to each other or to the pygidial or ventrianal shields. not developed.
Only one species, Sejus posnaniensis, has projections in the larva and protonymph, but loses them in the deutonymph.
Most Sejidae have barbed dorsal seta except Africasejus congoensis which has short and leaf-shaped setae, a state shared with Uropodella and Japanasternolaelaps. The presence of these processes is characteristic for members of the genus Epicroseius. They are absent in other taxa.
Most Sejidae have a row(s) of denticles (small teeth) on both sides of, or posterior to, the base of the tritosternum. This is a character shared with most Heterozerconidae.
Sternal and genital regions (Characters 42-60, Figures 10, 11) Sejidae have a divided sternal shield in both males and females. Males can have two pairs of sternal platelets (very small shields) and a larger sternal shield ( Figure 10). The platelets may be fused with each other and/or with the main sternal shield. In most species setae st1, st2, and st3 are inserted on the sternal platelets. When platelets are absent, these setae are inserted on dentate areas or on soft cuticle. Sternal setae st4 and st5 are always inserted on the large sternal shield, which may bear additional setae. The presternal genital orifice is small, circular, and devoid of genital setae.
Females may have up to three pairs of sternal platelets, which bear st1, st2, and st4, respectively, and one transverse platelet anterior to the genital shield, which bears setae st3 (Figure 11). Platelets bearing st4 may be fused with platelet bearing setae st3. If platelets are absent, the setae are inserted on dentate areas or on soft cuticle (as in the male). Seta st5 is always inserted on the genital shield. The genital shield usually bears additional setae, from