A multivariate study of differentiating characters between three European species of the genus Lasiochernes Beier, 1932 (Pseudoscorpiones, Chernetidae)

Abstract Morphological variation in three rarely collected European species of the genus Lasiochernes Beier, 1932 is thoroughly examined in the present study. Detailed descriptions of previously ignored morphological characters of Lasiochernes cretonatus Henderickx, 1998, Lasiochernes jonicus (Beier, 1929) and Lasiochernes pilosus (Ellingsen, 1910) are presented. The female of Lasiochernes cretonatus and the nymphs of Lasiochernes pilosus are described for the first time. Multivariate morphometric techniques (principal coordinate analysis and discriminant analyses) were employed to confirm morphological differentiation of the three Lasiochernes species and to identify the most reliable characters for their separation. The usefulness of particular body parts for species identification was evaluated. An identification key for the females of the Lasiochernes species studied is provided. Geographic distribution and habitat preferences of the three species are summarized.


Introduction
The genus Lasiochernes Beier, 1932 belongs to the subfamily Lamprochernetinae, as defined by Harvey (1994). Until now, ten species of the genus have been discovered (Harvey 2013). They are rarely collected, usually being found in the nests of small mammals or in caves. The genus is characterized by the presence of a long tactile seta on pedal tarsus IV, a pair of long tactile setae on tergite XI, five setae on the hand of the chelicera, secondary sexual dimorphism of the setation of the palps, with male palps bearing a long, dense setation, and a T-shaped spermatheca in females. Most of the known species are recorded from only one or two countries: L. anatolicus Beier, 1963 andL. villosus Beier, 1957 from Turkey; L. turcicus Beier, 1949 from Turkey and Israel; L. congicus Beier, 1959 andL. punctiger Beier, 1959 from the Democratic Republic of Congo; L. jonicus (Beier, 1929) and L. cretonatus Henderickx, 1998 from Greece; L. graecus Beier, 1963 from Albania and Greece and L. siculus from Italy (Harvey 2013). Only L. pilosus (Ellingsen, 1910) occurs in several European countries (Harvey 2013).
Detailed morphological descriptions of European pseudoscorpion species are rare. This holds true for both the adults and nymphal stages. These descriptions of adults and all nymphal stages are available mainly for the families Chthoniidae, Neobisiidae and Cheliferidae (e.g. Gabbutt and Vachon 1963, 1965, 1967, 1968, Gabbutt 1970, rarely for the family Chernetidae (Sezek andÖzkan 2007, Christophoryová et al. 2012).
Material of three Lasiochernes species was obtained during our study: L. cretonatus, L. jonicus and L. pilosus. L. cretonatus was described from a single male collected in a cave in Crete (Greece) (Henderickx 1998). L. jonicus was briefly described by Beier (1929), based on several adult specimens from Corfu, Greece. L. pilosus is distributed in several European countries (Harvey 2013) and it shows a degree of host-specificity, since it is almost exclusively found in subterranean mole-nests with a particular content of dead leaves. Many adults and nymphal stages of the latter species had been collected, but there had been no detailed description of nymphs and some characters of the adults remained unknown.
Morphological differences between species of pseudoscorpions, as reported in taxonomic descriptions, are often based on quantitative traits. Multivariate morphometric methods are an effective tool to compare the role of numerous quantitative and qualitative characters and allow in-depth examination of morphological variation of phenetically similar taxa. In recent years, many papers have successfully employed multivariate morphometrics in the taxonomy of invertebrates, such as mites (Klimov et al. 2004, Stekolnikov et al. 2010, Jagersbacher-Baumann 2014, flies (Castañeda et al. 2015, Van Cann et al. 2015, beetles (Sha et al. 2016) and spiders (Hamilton et al. 2016). The applicability of these methods for differentiation of pseudoscorpion species has been studied on the family of Chthoniidae. Muster et al. (2004) used multivariate analyses to separate two European species of the genus Chthonius.
The aims of this study are to (1) assemble detailed morphological descriptions of the adults of the three investigated Lasiochernes species, (2) describe all the nymphal stages of L. pilosus, (3) assess the extent of morphological differentiation between adults of the three species, (4) identify the morphological characters that are most relevant for the differentiation of the three species and (5) provide an identification key for the females of the three species.

Material and methods
Lasiochernes cretonatus: Greece, Crete, Azogires (Fig. 1) Populations of Lasiochernes collected from mole nests in Belgium and Slovakia were identified as L. pilosus (Beier 1963, Christophoryová et al. 2011) based on the setation on male palps and the habitat preference of this species. The taxonomic assignment of these two populations to L. pilosus is also in agreement with the known geographic distribution of this species (Harvey 2013). The studied population of Lasiochernes from Crete is from the type locality of L. cretonatus, a single cave at Azogires. The identification of this population as L. cretonatus is supported by morphological characters mentioned in the original description of this species, namely the setation of the male palp and the position of the tactile seta on the tarsus of leg IV (Henderickx 1998). The fourth Lasiochernes population was found in Pelion in Greece. It was identified as L. jonicus (Beier 1929, Mahnert 1978, due to the pedipalpal setation of the male specimens, which provides the main character distinguishing L. jonicus from L. cretonatus.
The chelicera, palp, leg I and leg IV were removed from the left side of the body of all specimens examined. In the case of L. pilosus, these appendages were mounted as permanent slide mounts using Swann's fluid as the medium. The rest of the body was studied as a temporary slide mount using lactic acid, after which it was returned to 70% ethanol. The body and the dissected appendages of L. cretonatus and L. jonicus were studied as temporary slide mounts using lactic acid, after which they were returned to 70% ethanol.
Measurements were taken from photographs using the Zeiss AxioVision 40LE application (v. 4.6). These photographs were made using the Canon EOS Utility software and a digital camera (Canon EOS 1100D) connected to a Zeiss Stemi 2000-C stereomicroscope or a Leica ICC50 camera connected to a Leica DM1000 stereomicroscope using Leica LAS EZ 1.8.0 software. Figures 4, 5 and 6 were drawn using a Leica drawing tube. Figure 2A was made with an FEI Quanta 200 scanning electron microscope at the Royal Belgian Institute of Natural Sciences, Brussels; ESEM scanning was performed in low pressure/low temperature water vapor (100% saturation, 4°C). Figures 2B, C and 2D are photographs of living specimens, taken on a glass plate with flash illumination, using a Canon Eos 5D mark III with a Canon MP-E 65 mm f2.8 lens. Nomenclature for all taxa follows Harvey (2013). The material is deposited in the zoological collections of Comenius University, Bratislava.
Methods of multivariate morphometrics (Marhold 2011) were used to examine the differentiation of 19 adult specimens assigned to three Lasiochernes species (five specimens of L. cretonatus, two specimens of L. jonicus and 12 specimens of L. pilosus) and to evaluate the importance of particular morphological characters. The morphological characters measured or scored included those reported as taxonomically relevant within the genus in identification keys and other treatments. The distribution of long and dense setation on the palps of males, the main character used for taxonomic identification of the studied samples, was omitted from the statistical analyses to avoid circular reasoning. Altogether, 92 quantitative characters were measured or scored (Table 1), of which 51 were continuous (see Table 1 in Results) and 34 were discrete (see Morphological descriptions in Results). Out of these, seven characters were invariable between measured specimens (number of blades in cheliceral rallum, number of setae on hand and movable finger of chelicera, number of trichobothria on both chelal fingers, presence of a pair of long tactile setae on tergite XI and sternite XI) and only the remaining 85 characters were used for further statistical analyses.
The statistical analyses were performed as follows: (1) As the first step, the Shapiro-Wilk statistic for the test of normality of distribution was computed for each character.
(2) Principal coordinate analysis, PCoA (Podani 2000(Podani , 2001, based on 85 characters, was used to obtain possible groupings of the 19 studied specimens. The data were standardized by a standard deviation of variables, and Euclidean distance was used to compute the secondary matrix. PCoA, unlike the better known PCA method (principal component analysis), can be also used for qualitative and mixed characters, as well as in cases when p>n (p = number of characters, n = number of objects). (3) Correlation between the principal coordinate axes of PCoA and original quantitative characters was computed using Pearson correlation coefficient (Zar 1999) in order to identify the characters that are the most responsible for the groupings of specimens along the first three principal coordinate axes. (4) Discriminant analyses (Klecka 1980, Marhold 2011 were employed to assess the morphological differentiation between the three Lasiochernes species. The discriminant analyses applied included canonical discriminant analysis (CDA) and classificatory discriminant analysis (classificatory DA). In CDA, the discriminant functions were derived to express the extent of morphological differentiation between the predefined groups (the three Lasiochernes species) and to identify the most important differentiating characters. Nonparametric k-nearest neighbors classificatory discriminant analyses were performed to estimate the percentage of specimens correctly assigned to the predefined groups. A cross-validation procedure was used, in which the classification criterion was based on n−1 individuals and then applied to the individual left out. Discriminant analyses generally require a multivariate normal distribution of the characters; nevertheless, they have been shown to be quite robust against deviations in this respect (Thorpe 1976, Klecka 1980. Due to the limited number of available specimens (19) and the chosen number of predefined groups (three), we had to lower the number of characters in primary matrices to 15 (or less) in order to satisfy the requirements for number of objects (n), number of predefined groups (g) and number of variables (p) in discriminant analyses [p < (n−g)]. Therefore, the original dataset of all measured characters was divided into eight partial matrices corresponding to eight parts of the body. Each partial dataset contained no more than 15 characters and each was analyzed in a separate CDA and classificatory DA. The following eight body parts were selected: carapace (six characters), chelicera (six characters), palp (nine characters), chela (11 characters), leg I (15 characters), leg IV (12 characters), tergites (ten characters) and sternites (12 characters). As a result, eight CDAs (CDA 1-CDA 8) and eight classificatory DAs (DA 1-DA 8) were performed to identify both the body parts and the characters that are most important for the differentiation of the three species. Altogether, 81 characters (out of the original 85) were included in these analy-ses. Four characters were omitted. The character "length of the whole body" was inapplicable for the parts of the body and three other characters (posterior width of carapace, length of palpal hand with pedicel, length of patella of leg I), were excluded because they were invariable within one or more predefined groups (species) and might have distorted the discriminant analyses. Based on the results of the eight CDAs (CDA 1-8), the 15 most important characters were selected and a final matrix, combining all body parts, was assembled. This total-body matrix was analyzed in CDA 9 and classificatory DA 9. Prior to the discriminant analyses of all the datasets mentioned above, the Pearson and nonparametric Spearman correlation coefficients (Zar 1999) were computed to reveal correlation structure among the selected characters and to ensure that no very high correlations (> 0.95) were present (potentially distorting the analyses). The discriminant analyses were performed using SAS 9.1.3 software SAS/STAT v.9.2 (SAS Institute, 2009). (5) Finally, descriptive statistics were computed for adults of the three Lasiochernes species, and for nymphs of L. pilosus. Variations in the morphological characters that differentiate between them are shown as box-and-whisker plots. The minimum and maximum values for the measured characters are reported in identification key and morphological descriptions. The analyses were performed using SAS 9.1.3 software SAS/STAT v.9.2 (SAS Institute, 2009).

Multivariate morphometrics
Most of the measured characters showed departures from a normal distribution. Therefore, the nonparametric correlation coefficient (Spearman) (apart from the Pearson parametric coefficient) and nonparametric classificatory discriminant analyses were used.
The ordination diagram of PCoA of the three Lasiochernes species, based on 85 morphological characters for 19 adult specimens, showed two large groupings of specimens separated along the first principal coordinate axis (Fig. 7). The first grouping consisted of L. pilosus specimens and the second comprised both L. cretonatus and L. jonicus. However, the specimens of the latter two species were not intermingled, being divided in accordance with their taxonomic assignment along the second and partly the third principal coordinate axis. The calculations of the correlation between the principal coordinate axes of PCoA and the original quantitative characters revealed the characters most responsible for the grouping of specimens along the first three axes. The characters most correlated with the first axes are: carapace length, length and width of femur of leg I, length of femoropatella of leg IV, length of palpal hand with and without pedicel, chelicera width, width of trochanter of leg I, posterior width of carapace and length of trochanter of leg I. The characters most correlated with the second axis are: numbers of setae on sternite X, tergite VIII, tergite VII, tergite VI and sternite IX; and those most correlated with the third axis are: body length, number of setae on anterior and posterior genital opercula, length/width ratio of tibia of leg IV and number of setae on sternite IV.
Eight canonical (CDA 1-CDA 8) and classificatory discriminant analyses (DA 1-DA 8) were performed to identify the characters and body parts that are most important for the differentiation of the three species, and to evaluate the degree of differentiation in each case. The three character pairs (length and posterior width of carapace, length of palpal hand with and without pedicel, length of patella and tibia of leg I) exceeded the correlation threshold of 0.95 in datasets with the body parts and, therefore, three characters (posterior width of carapace, length of palpal hand with pedicel and length of leg I patella) were excluded from further analyses. In CDAs (CDA 1-8), three species mostly formed their own clouds in the ordination space without overlaps (Fig. 8A-H), showing that all the body parts are useful for the differentiation of the three species. The best differentiation of the three species was reached in CDA 6, based on characters measured for leg IV (Fig. 8F), and the weakest differentiation was obtained in CDA 7, based on characters of the tergites (Fig. 8G). For the characters most correlated with the canonical axes and thus contributing to the differentiation of the three species, see Table 2. For details of the correlations of all characters with the axes, see Suppl. material 1. In almost all the classificatory DAs based on the body parts, the percentage of correctly classified specimens reached 100% for all three species. The only exception was the classificatory DA based on characters measured for tergites, for which 80% of specimens were correctly classified into L. cretonatus, 100% into L. jonicus and 58.3% into L. pilosus.
Finally, the classificatory DA 9 and CDA 9 were computed to assess the differentiation of the three species based on the selection of the most important characters from all the parts of the body, as revealed in CDA 1-8. In the classificatory DA 9, the classification success rate reached 100% for all the specimens. The three species were clearly separated in the ordination space of CDA 9 (Fig. 9). The characters most highly correlated with the first and second canonical axis are those in bold type in Table 3.
The variations in morphological characters that are most useful for differentiation of the three Lasiochernes species are shown in Fig. 10.

Identification key to females of L. cretonatus, L. jonicus, and L. pilosus
Based on all the results obtained, nine morphological characters that differentiate females of the three species were selected (Table 4). The values of two of them, namely the length of cheliceral movable finger and the length of the palpal hand with pedicel, do not overlap and therefore allow the unambiguous identification of three Lasiochernes females.

Distribution and habitat preference
Lasiochernes cretonatus was described from Souré Cave (Cave of 99 Holy Fathers) in Crete, based on one male collected under a small piece of stone near the cave wall (Henderickx 1998). Šťáhlavský et al. (2005) studied karyotypes of one female and one male tritonymph of L. cretonatus from the same cave. New specimens were found between organic material, pigeon feathers, dry leaves and pieces of branches in another corner of the same upper cave room, less than six meters from where the holotype was found. Specimens were sifted from leaf litter and collected by vacuuming cracks with a modified portable electric vacuum cleaner. Lasiochernes jonicus was described as Chelifer (Trachychernes) jonicus by Beier (1929) from Agios Mattheos, Corfu, Greece. The types were collected by sifting maquis litter. Later, Beier (1963) specified that, besides the maquis litter, a rotten mouse nest was sifted as well. Altogether 25 males, 19 females and 12 nymphal stages were collected (Beier 1929). Mahnert (1978) recorded three males, one female and 33 nymphs from soil samples in a nameless cave near Profitis Elias church, on Mount Ossa, Thessaly, Greece. The find of our specimens in the Tsouka cave in Pelion, Greece, represents the third known locality of L. jonicus. The specimens in the Tsouka cave were sifted from material (leaves, small branches and rock fragments between ingrown tree roots) in an upper dry room of the cave. Ellingsen (1910) described one male of Chelifer (Trachychernes) pilosus from the vicinity of the town of Görz in Austria (now Goricia in Italy) and did not mention the habitat type or the collecting method. Heselhaus (1914) found females and nymphs in mole nests in Netherlands, and described them as Chelifer falcomontanus. Later, Berland (1925) recorded several specimens of C. falcomontanus from mole nests in Luxem-   Beier (1929) recorded several adults and nymphs in mole nests from Austria and synonymized C. falcomontanus with C. pilosus. Beier (1929) indicated that the species occurs in mole and ground-squirrel nests. Ressl (1965) and Ressl and Beier (1958) later found many specimens in mole nests with leaf content in Austria. Caporiacco (1949) recorded two L. pilosus males in the rotten trunk of an oak at Lipizza, Italy (now Lipica, Slovenia). Later Ćurčić (1974) listed L. pilosus in Slovenia (without providing collecting details) in his catalogue of the former Yugoslavian fauna. There is no mention of this species occurring in Slovenia in the current version of the world pseudoscorpion catalogue (Harvey 2013). The occurrence of L. pilosus in mole nests in Italy was recorded by Beier (1963) and Inzaghi (1981). In the Netherlands, L. pilosus was typically collected in mole nests (Van der Hammen 1969). Ventalló (1934) recorded the species for the first time from Spain, based on 14 specimens found in a cave. L. pilosus also occurs in mole nests in Germany (Hesse 1941, Weidner 1954, Weygoldt 1966, Rehage and Renner 1981. Weygoldt (1966Weygoldt ( , 1969 reported that the species can be found in water vole nests. L. pilosus was also collected in Belgium, in mole nests in forests (Cooreman 1946, Leleup 1948, Henderickx 1998, 1999, Šťáhlavský et al. 2005). The locality from which the material studied here was collected is a new record for the geographic distribution of L. pilosus in Belgium. The locality is located on a hilltop, all specimens were sifted from a mole nest between the roots of a tree on the hilltop, next to a road. Krumpálová and Krumpál (1993) extracted this species from a mole nest for the first time from Slovakia (at Borinka, the same locality as in the current paper). Christophoryová and Krumpál (2010) sifted two females from leaf litter in the Nature Reserve Šúr, Slovakia. New specimens from Borinka were extracted from a mole nest situated in the ecotone between forest and grassland.

Morphological variation
The original description of L. cretonatus was based on one male (Henderickx 1998). Comparison of our newly found male with the holotype showed similarity in a majority of the characters (palpal teeth numbers, morphometrics of palps and leg IV, length of body and chelicera and position of tactile seta on tarsus IV). Exceptions were the higher setae number on posterior carapace margin of the newly found male (13 versus 12) and higher number of paraxial accessory teeth on movable chelal finger of the newly found male (4 versus 3). The length of the palpal trochanter was incorrectly given by Henderickx (1998) as 0.84 mm (0.53 mm in the new male). In the present paper, several characters of this species are described for the first time: morphometrics of leg I and carapace, width of chelicera, length of cheliceral finger; number of setae on chelicera, form of galea, rallum, serrula exterior; complete trichobothrial pattern, complete chaetotaxy of carapace, tergites and sternites; numbers of setae and lyrifissures on genital opercula. The female is described here for the first time. Beier's (1929Beier's ( , 1963 descriptions of L. jonicus provided information concerning the cheliceral rallum, serrula exterior, galea, setation and shapes of palpal parts and the position of the tactile setae on tarsus IV and tergite XI. The mean values (for an unspecified number of specimens) of palpal measurements, length of body and carapace of males and females were given by Beier (1929). Mahnert (1978) described one male, giving measurements of the palps and leg IV, and the numbers of marginal and accessory teeth on the chelal fingers. Most of the characters of our male and female correspond with previous descriptions (Beier 1929, 1963, Mahnert 1978; some differences in measurements are probably related to the number of specimens studied. Mahnert (1978) counted more marginal teeth (47 on the fixed finger, versus 42 here, and 50 on the movable finger, versus 47 here), more paraxial accessory teeth (7 on fixed finger, versus 6, and 5 on movable finger, versus 4) and more antiaxial teeth on movable finger (11 versus 10) of the male. In contrast, there are more antiaxial teeth on the fixed finger of our male (12 versus 11). Our results provide information on several new characters: measurements of chelicera, carapace width, measurements of leg I, setae number on carapace, chelicera, tergites, sternites, and genital operculum anterior and posterior.
Lasiochernes pilosus was described from one male by Ellingsen (1910), who counted more blades on the serrula exterior than were observed here (27 versus 23-24). Beier (1963) described both sexes, mainly giving their palp measurements. The number of serrula exterior blades was modified to 25-27. The number of antiaxial accessory teeth on the chelal finger was lower than in our specimens (10 versus 11-16 on fixed finger and 8 versus 11-15 on movable finger). The present study provides a number of new details, such as measurements of leg I and IV; the number of paraxial accessory teeth on chelal fingers; the numbers of setae on the carapace, genital opercula, tergites and sternites. For the first time, all nymphal stages are described in detail.
In this paper, the potential of multivariate morphometric techniques for the diagnostic of pseudoscorpion species has been explored. Our study provides a first reference library of morphometric measurements that might be used for the identification of Lasiochernes specimens. The PCoA, which depicts the variation without prior definition of the groups in the dataset, showed rather clear differences between the three species. Two large groupings of specimens were visible in the PCoA, the first consisting of L. pilosus and the second of L. cretonatus and L. jonicus. The proximity of the latter two species in PCoA was probably caused by one specimen of L. cretonatus with significantly higher numbers of setae on the carapace (total and number on anterior disk). Discriminant analyses, which, unlike the PCoA, weight the characters to stress the between-group variation component, revealed considerable differences between the three species. These analyses were also used to identify the most differentiating body parts and the most important characters. The characters traditionally used most in identification keys to pseudoscorpions are those of the palps (Beier 1963, Christophoryová et al. 2011 and their importance was confirmed again by our data. A surprising discovery was that, from among the body parts, the best differentiation of the three species was obtained with leg IV. On the other hand, the tergites were not very useful for species differentiation, due to the high variability of setae number on each tergite. The majority of the most differentiating characters was measured or scored on the carapace, chelicera, chelal fingers and legs I and IV. Until now, the number of setae on the carapace was only rarely used in the descriptions of Chernetidae, mainly the setae number on posterior carapace margin (Beier 1963, Henderickx 1998. The whole count of setae could substantially facilitate the diagnosis of chernetid species in future. The setal counts on the tergites and sternites (except the genital ones) of Lasiochernes species showed a high degree of variability.
Multivariate morphometrics have been successfully applied in many other taxonomic studies of various invertebrates. For instance, they were very helpful in interpreting morphological differences between two cryptic species of Sancassania Oudemans, 1916, Acari (Klimov et al. 2004). Stekolnikov et al. (2010) revised a species group of chiggers (Acari) using multivariate morphometrics and developed a multivariate classification model to separate three closely related species. These analyses showed complete separation of the studied species. The characters contributing strongly to the discrimination were used in formal description of these species as well as in an identification key. Jagersbacher-Baumann (2014) analyzed four mite species of the acarorumcomplex (Scutacaridae) using traditional and geometric morphometric methods. The results showed that multivariate morphometric methods are perfectly suitable for dif-ferentiating even between morphologically similar scutacarid species, with traditional morphometrics performing better than geometric morphometrics. Van Cann et al. (2015) explored the potential of wing morphometrics for the diagnosis of morphospecies of Tephritidae (Diptera). Multivariate analyses allowed the consistent identification of a significant proportion of specimens in that study. In pseudoscorpion taxonomy, multivariate analyses were used to separate two European Chthoniidae species. Although multivariate analyses suggest specific separation, there was only one unequivocal character for discrimination, the presence or absence of a single isolated tooth on the moveable finger of the chelicerae (Muster et al. 2004).
The genus Lasiochernes is noteworthy for its sexual dimorphism (Beier 1963). Males are unambiguously identified by the presence of a long setation arranged on different palpal parts, depending on the species. The setation of the palp is normal in females, without long setae. Our aim was to find characters that could be used for a more reliable identification of the females. It should be noted that our identification key is useful mainly for differentiation of females of L. cretonatus and L. pilosus. Values of some characters measured on the female of L. jonicus are influenced by low number of specimens examined and it is possible that better sampling might show stronger overlaps in future studies. The identification key is based on the characters that were rarely or never used in previously published taxonomic treatments of Lasiochernes. Therefore, the comparison of these characters with other European species of the genus is not yet possible.
Based on the results obtained, we assume that future studies will benefit from application of multivariate morphometric analyses, and could potentially help to find new characters and contribute to a more reliable identification of pseudoscorpion species.