Taxonomy of Ixinandria Isbrücker & Nijssen ( Loricariidae : Loricariinae ) based on morphological and molecular data

The genus Ixinandria was described in 1979 to include I. steinbachi from the upper río Juramento in northern Argentina and I. montebelloi from the upper río Bermejo in southern Bolivia. We used a multivariate morphometric analysis (Principal Components Analysis) and a molecular comparison of the Cytochrome Oxidase I gene to investigate variation among distinct populations from the known distribution range of both species. We conclude that the populations are not significantly distinct and do not deserve separate species recognition, and we place I. montebelloi in the synonymy of I. steinbachi.


Introduction
Ixinandria Isbrücker & Nijssen, 1979 was described to accommodate two species, Loricaria steinbachi Regan, 1906, the type species, and Canthopomus montebelloi Fowler, 1940.Ixinandria was considered by the authors to be very close to Rineloricaria Bleeker, 1862, and diagnosed on the basis of the absence of abdominal plates, the degree of development of odontodes, and the shape of the head in sexually mature males (Isbrücker, 1979).Isbrücker (1979) placed Ixinandria in the subtribe Rineloricariina, which also included Spatuloricaria Schultz, 1944, Dasyloricaria Isbrücker &Nijssen, 1979, and Rineloricaria.The validity of Ixinandria or its relationships with other loricariines were never further investigated.
Loricaria steinbachi was placed in the subgenus Rhineloricaria by Eigenman (1911), a few years after its original description, and remained as Rineloricaria until it was moved to Ixinandria by Isbrücker (1979).On the other hand, Canthopomus montebelloi remained mostly unnoticed until Isbrücker's (1979) new combination by its inclusion in the genus Ixinandria.
Loricaria steinbachi was described from the río Juramento basin, a tributary to the río Salado drainage, near Cachi, Salta, Argentina, at an elevation of 2,500 meters above sea level, based on three specimens.Isbrücker (1979) later designated the larg-est of those three syntypes as the lectotype.Canthopomus montebelloi was described from tributaries of the upper río Bermejo at the Monte Bello Farm in Tarija, southern Bolivia, based on the holotype only, also from high elevation.
The validity of those species was later questioned by Isbrücker (1981a), who suggested that they might represent a single species.In this paper, we revised the two species and placed Ixinandria montebelloi in the synonymy of I. steinbachi, based on both morphological and molecular evidence.

Material and Methods
The morphometric variables were measured with a digital caliper (0.1 mm precision).The measurements and counts follow Isbrücker & Nijssen (1978), except the following: Counts of lateral plates with coalesced keels and the measurement of cleithral width were made according to Reis & Pereira (2000); and eye diameter was measured according to Isbrücker (1973).Names and counts of lateral plate rows and rostral plates follow Schaefer (1997).Osteological observations were made on specimens cleared and stained (c&s) prepared according to the method of Taylor & Van Dyke (1985).
We used Principal Components Analysis (PCA) to investigate morphometric variation among four distinct populations of Ixinandria, encompassing the area of distribution of both I. steinbachi and I. montebelloi.A total of 30 morphometric variables (Table 1) was taken from 73 specimens representing comparable size ranges: a population from the río Juramento basin, upper río Salado drainage, type-locality of Ixinandria steinbachi (n=16, 58.2-91.6 mm SL), a population from the upper río Bermejo drainage in northern Argentina (n=18,, a population from the upper río Bermejo in southern Bolivia, type-locality of Ixinandria montebelloi (n=37, 68.9-100.3mm SL), and two specimens from the upper río Pilcomayo in Bolivia (n=2,72.7 and 76.6 mm SL).Principal components analysis was performed on the co-variance matrix of the 30 log-transformed measurements.Factor scores were plotted for the PC II, III and IV, interpreted to represent size-free shape (Bookstein, 1989).
In the lists of material museum abbreviations and catalog number are presented first, followed by the number of specimens in that lot, size range, number and size range of measured specimens in parentheses, and locality.Institutional abbreviations are according to Reis et al. (2003), with the addition of Museo de Ciencias Naturales, Universidad Nacional de Salta, Salta (MCNI), and Universidad Mayor de San Simón, Cochabamba (UMSS).
Scanning electronic micrographs were made at the Facultad de Ciencias Naturales y Museo (Universidad Nacional de La Plata) and Centro de Microscopia e Microanálises (Pontificia Universidade Católica do Rio Grande do Sul).
For molecular analyses we used tissue samples from six localities as follows: 12 specimens from two localities in the río Arenales, tributary to the río Juramento near Salta, Argentina; 10 specimens from the río La Caldera and arroyo Lesser, tributaries to the río Bermejo near Salta, Argentina; and 11 specimens from the río Gaudalquivir and río Camacho, tributaries to the río Bermejo near Tarija, Bolivia.In addition, we used sequences from two outgroup species taken from GenBank (Apistoloricaria ommation EU359403 and Rineloricaria sp.EU359461).
We used a non-standard protocol for this primer pair with an initial denaturation step of 1 min at 96°C followed by 40 cycles of 94°C for 30s, annealing at 50°C for 20s, 48°C for 5s, 46°C for 5s, 44°C for 5s, 42°C for 5s, 40°C for 20s and extension at 72°C for 1 min.This was followed by a final 3 min at 72°C extension step.Amplification success was evaluated on GelRed TM (BioTium, São Paulo) or ethidium bromide-stained agarose gels (Sambrook et al. 1989).PCR products were purified using PEG8000, ExoSAP-IT ® (USB) or the ilustra TM GFX PCR and Gel Band Purification Kit (GE Healthcare, Buckinghamshire, UK).Sequencing was done using the DYEnamic TM ET dye terminator kit (GE Healthcare, São Paulo) read in a MegaBace1000 sequencer.Sequences were edited and combined using BioEdit 7.0.1 (Hall 1999).
Sequences were aligned using Clustal X 1.83 (Thompson et al., 1997) with the standard settings.PAUP*4.0b10(Swofford, 2001) was used to analyze the data with respect to the parsimony and the maximum likelihood criteria.For the maximum parsimony (MP) analysis we used the tree-bisection-reconnection (TBR) search algorithm with 1000 replicates in which taxa were added randomly to the starting tree.All characters were treated as unordered and transformations were assigned equal weight.Nodal support was evaluated with 1000 nonparametric bootstrap pseudoreplicates (Felsenstein, 1985) using the TBR search algorithm on a starting tree to which taxa were added randomly.Multiple optimal topologies were summarized through consensus methods.The maximum likelihood (ML) analysis was done using the TVM+G model found with Modeltest 3.7 (Posada & Crandall, 1998).Nodal support was evaluated with 1000 nonparametric bootstrap pseudoreplicates.

Results
Multivariate analysis.The resulting first principal component included a large proportion of the total variance (78.6%) and all variable loadings were positive and varied little in magnitude, suggesting that it represents a general size factor.Plots of factor scores of principal component II vs. III grouped specimens into four broadly overlapping clusters (Fig. 1).PC II and III included 4.5 and 4.0% of the total variance, respectively.Measurements with heavier loadings on PC II are analfin spine length (0.395), body width at dorsal-fin origin (-0.416) and body depth at dorsal-fin origin (-0.434); on PC III are dorsalmost principal caudal-fin ray (-0.407), ventralmost principal caudal-fin ray (-0.386) and eye diameter (0.395).Plots of factor scores of PC II vs. IV also grouped specimens into four overlapping clusters (Fig. 1).PC IV included 2.5% of the total variance and heavier loadings on PC IV were on thoracic length (0.247) and internareal width (-0.487).

Molecular analysis.
We obtained an alignment with 35 sequences with 709 contiguous nucleotides.The character matrix contained 42 parsimony-informative characters.The MP analysis resulted in 26 equally parsimonious trees with a length of 151 steps.Their summarized strict consensus tree is shown in Fig. 2. The single best tree found in the ML analysis has a -ln L score of -1643.94356and is shown in Fig. 3.
Based on the geographic distances we considered our samples to be from three populations: Juramento (río Arenales), Bermejo 1 (río Caldera and arroyo Lesser) and Bermejo 2 (río Gaudalquivir and río Camacho).The Network showed the population Bermejo 2 separated from Juramento and Bermejo 1 which, in turn, could not be separated (Fig. 4).
The AMOVA showed that the sequences from the samples from Tarija are significantly different (p=0.0000)from the other samples.The samples from Salta did not show any significant difference (p= 0.39640 ± 0.0243).Our sequences showed an avereage K2P distance of 1.1%.Spatuloricaria, and Loricaria sometimes lack a preanal plate, but small abdominal plates are present.Some species of Rineloricaria have a naked abdomen, but a preanal plate is always present, except in R. setepovos.This species, however, has a caudal-fin filament and 3-4 premaxillary teeth.Ixinandria can also be distinguished from other loricariines by the shape and number of premaxillary teeth.Ixinandria has wide and almost symmetrical tooth cusps, and up to 18 teeth in each premaxilla (vs.premaxilla with up to 13 narrow and clearly asymmetrical teeth in other loricariines, except Rineloricaria aequalicuspis which has up to 15 narrow teeth with nearly symmetrical cusps, and some hartiini, which also have many asymmetrical premaxillary teeth).et al., 2003: 45 [listed].-Ferraris, 2003: 336 [listed].-Menni, 2004: 195, 198  Outline of head rounded in dorsal view.Head strongly flattened in mature males.Paired anterior postrostral and cheek plates scarcely developed ventrally.Odontodes curved, conical, and pointed, densely covering head, trunk, and fin rays, making fish somewhat hispid.Body odontodes scarcely distributed and grouped centrally in plates (Fig. 6).Thick skin sometimes covering plates partially, hiding plate limits and covering odontodes.Snout tip with big globular protuber-  ance of naked skin.This roundish naked area not reaching most anterior pore of infra-orbital ramus of sensory canal.Gill opening small.Mouth big and roundish.Lower lip margin with short digitiform papillae.Upper lip well developed, folded inwards with row of globular papillae.Big globular papillae arranged in regular rows on surface of lips.Numerous small globular papillae on mouth roof, bigger posteriorly.Maxillary barbel very small, mostly coalesced with lower lip.Teeth bilobed, with wide cusps, mesial more developed and slightly longer than lateral (Fig. 7).Premaxilla with six to 18 (mean=12.4,n=71) teeth in functional series.Dentary with six to 15 (mean=10.3,n=46) teeth in functional series.Abdomen completely naked, except for lateral abdominal plates.Preanal plate absent.Two to 11 (mean= 7.0, n= 72) lateral abdominal plates on each side.27 to 32 (mean=29.4,n=72) plates in median lateral series, with well developed keels formed by hypertrophied odontodes.Five lateral series, sometimes broken and appearing to have six or seven lateral series; only posterior portion of median and midventral series with keels.Keels coalesced in last seven to 12 rings (mean=9.5, n=72).Middorsal series with 5-10 plates (mean=7.8,n=9) (better visible in cleared and stained material).Entire body with thick skin between the plates, especially in the lateral series.Predorsal plates usually arranged in irregular pattern, not forming rows.Supraoccipital and predorsal plates without keels.Transversal diameter of nasal capsule equal or slightly larger than eye diameter including notch (better visible in cleared and stained material).

Ixinandria
Posterior margin of dorsal fin straight or slightly rounded, generally with first or second branched ray longest.Tip of dorsal fin, when depressed, reaching third or fourth (exceptionally fifth) plate posterior to fin insertion.Dorsal-fin spinelet absent.Posterior margin of pectoral fin convex, first or second branched ray longest, reaching to or slightly beyond pelvic-fin origin.Posterior margin of pelvic fin rounded; first or second branched ray longest, reaching to or falling slightly short of anal-fin origin.Posterior margin of anal fin rounded, with longest first or second (rarely) branched ray.Tip of anal fin, when depressed, reaching fifth or sixth plate posterior to fin insertion; three or four ventral plates along its base.Pos-Fig.6. Scanning electron micrographs of the body odontodes of Ixinandria steinbachi (MCNI 99, 65.9 mm SL).A) Two plates of the median and mid-ventral lateral series with odontodes; B) detail of odontodes.terior margin of caudal fin truncated or slightly concave.Upper and lower unbranched rays never extending as filament.Two or three supracaudal plates covering base of caudal-fin rays, median plate sometimes absent.
Color in alcohol.Background color of dorsal surface of head and body dark brown with five darker transverse bars; first one inconspicuous on supraoccipital, others distributed between origin of dorsal fin and caudal fin.Last three bars sometimes continuing across ventral surface and forming dark brown rings.Some specimens with black spots on head and snout.Dorsal fin with dark gray circular spots.Pectoral and pelvic fins yellowish with dark circular spots.Sensorial pores on head and beginning of lateral canal pigmented with black.Ventral surface of body light brown in naked regions.Anal and caudal fins with dark dots arranged in bars on branched and unbranched rays.Ventral plates darkened around anal-fin base.
Color in life.Background color of head and body light brown with five dark brown transverse bars distributed as described above.Other markings as described above (Fig. 8).
Sexual dimorphism.Males with head wide and flattened; long hypertrophied odontodes on sides of head and on upper side of branched pectoral-fin rays; short odontodes on remaining of dorsal surface of body, especially on predorsal region; unbranched pectoral-fin ray hypertrophied.Papillae in mouth cavity more numerous in males than in females.Outer margin of tooth cusps straight in females, rounded in males (Fig. 7).

Ontogeny. Many structures vary ontogenetically in
Ixinandria steinbachi.The premaxilla, dentary, and the pectoral girdle are ossified in a specimen of 17.2 mm SL, the smallest individual we were able to examine.At that size the pelvic girdle was not developed, the lateral abdominal plates and plates on the head and predorsal region were not formed.The lateral series of plates were represented by small platelets with one or two odontodes each.
In a specimen of 28.2 mm SL the pelvic girdle, the lateral connecting bone, the nucal plate, and the neurocranium are totally ossified.The plates on the head are represented by scarce odontodes and the lateral series of plates are almost entirely formed from the posterior region to level of the dorsal fin.Three small lateral abdominal plates are forming laterally on the abdomen, with one or few odontodes each.
Distribution and habitat.Upper río Juramento basin, río Salado drainage in Salta (Argentina), upper río Bermejo in northwestern Argentina and southern Bolivia, and upper río Pilcomayo in southern Bolivia (Fig. 9).Its presence in the río Pilcomayo is herein reported for the first time.Ixinandria steinbachi was caught in fast flowing waters between 15 and 65 cm depth, especially in places with filamentous algae, in mountainous areas at altitudes ranging from around 200 to 2900 meters a.s.l..The minimun altitude where it was recorded is 210 m a.s.l. in the río Rosario, 48 km from the city of Salta.This species lives in fast flowing and highly oxygenated waters, with stony bottom, usually hiding under the stones during the day.Under 2200 m a.s.l.one species of Hypostomus, Heptapterus mustelinus, and some characids and other loricarids were found occurring syntopically with I. steinbachi (Menni, 2004).

Discussion
The principal components analysis of morphometric traits failed to discriminate clusters among the four populations of Ixinandria being investigated, indicating that morphology is highly homogeneous and suggesting no species-level separation.However, small morphological differences were found between specimens of populations from the río Bermejo in southern Bolivia and the río Bermejo in Argentina (Baritú National Park) when compared to specimens from other tributaries to the río Bermejo and río Juramento basins in Argentina.These small differences are degree of development of odontodes along the sides of the head in mature males, width of the pectoral-fin spine, and size of the naked skin area at the tip of the snout.These differences, however, are difficult to interpret and mostly overlap when many specimens are analysed, and are not enough to warrant species level separation.
The molecular analysis was similarly not able to discriminate the population from the upper río Juramento basin, type locality of Ixinandria steinbachi, from that in the the upper río Bermejo in Salta, Argentina.A significant difference, however, was found between the population in the río Bermejo in Salta and that in the tributaries to the same river in Tarija, Bolivia, type locality of Ixinandria montebelloi.This difference, however, is smaller than those usually found among species clearly distinct and is not enough to warrant separate species recognition.The mean K2P distance between our sequences was 1.1%, which falls within the typical range of intraspecific K2P distance in COI sequences of up to 2%, much smaller than the average distance of 9.9% of distinct, congeneric species (Ward et al., 2005).
An explanation for the differences found could be that the samples represent three separate populations of the same species, what corroborates the morphologic results.The differences between the populations from the río Bermejo in Salta and in Tarija could be explained by the geographic distance (>350 km) between the two sample localities.To test this hypothesis, samples from areas between the present collection stations and from the río Pilcomayo basin would be necessary.In summary, the results of the molecular analysis do not adequately support the discrimination of the populations and therefore the separation of the two species.
Both morphometric and molecular results make sense from the geological point of view.Andean orogeny and climate have had profound influence on the biota of the Central Cor-dillera in northeastern Argentina and southern Bolivia.The geological and climatic development during the late Tertiary and early Quaternary have had a significant impact on the establishment of orographic barriers and the regulation and distribution of rain in this region and, as a consequence, on the establishment and continous changing of hydrographic drainages (Alonso et al., 2006).
The presence of a single species of Ixinandria in the three river drainage basins can be explained by the fact that the Pilcomayo, Bermejo, and Juramento rivers had been connected and separated several times during the Quaternary, as indicated by the superposition of their alluvial fans (Iriondo, 1993).One example of documented river capture is the río La Caldera, presently a tributary to the río Bermejo basin that formerly used to cross the area where today the city of Salta is located draining towards the south and emptying into the río Arenales, itself a tributary to the río Juramento (Monasterio de Gonzo, 2003).Because of early Quaternary tectonism the río La Cardera changed its direction and become a tributary to the río Mojotoro, in the upper río Bermejo basin.A considerable similarity in the catfish fauna between the río Juramento and the río Bermejo drainage basins was reported by Monasterio de Gonzo et al. (1994).The geographic distribution of Ixinandria steinbachi, encompassing the upper portions of the Pilcomayo, Bermejo, and Juramento rivers, is very similar to the distributions of Parodon carrikeri, Oligosarcus bolivianus, Corydoras micracanthus, Loricaria tucumanensis (Liotta, 2006), and Jenynsia alternimaculata (Ghedotti, 1998).
The only citation of Ixinandria steinbachi for a river outside the known range of the species is that of Gomes (1947), who listed a species of Rineloricaria from the rio Maquiné in northeastern Rio Grande do Sul State, southern Brazil, as Loricaria (Rhineloricaria) steinbachi, based on 18 young specimens deposited at the UMMZ.That species was later found to be new and described as Rineloricaria aequalicuspis by Reis & Cardoso (2001).
The diagnosis of Ixinandria is based on a completely naked abdomen, which has no abdominal plates ventral to the lateral abdominal plates.Only one specimen out of 242 examined in this study, from the río Juramento basin (MCNI 1225), has small platelets on the central portion of the abdomen.Even in this specimen, a preanal plate is lacking.The validity of the genus Ixinandria is a question beyond the scope of this study, and will remain unanswered until a broad phylogenetic analysis of the loricariines can be performed.cial thanks are due to Gladys and Fernando who sent a lot of fixed material and tissue from Argentina and Bolivia for this study.Virginia Martinez and Gastón Aguilera helped collecting fishes.Ricardo Kanitz helped with the genetic programs.Tiago Carvalho took the scanning electron micrographs at PUCRS.MSR was financed by a fellowship from Universidad Nacional de La Plata (UNLP) and by a postdoctoral fellowship from CONICET/CNPq.The All Catfish Species Inventory (NSF-DEB #0315963) subsidized the visit of MSR to the North American museums.CAC was financed by a two-years doctoral fellowship from the German Academic Exchange Service (DAAD) and a one year doctoral fellowship from CNPq.RER is partially financed by CNPq (process # 301748/2004-7).SLB is partially financed by CNPq (process # 481121/2004-8).

Fig. 2 .
Fig. 2. Strict consensus of the 26 best trees from the maximum parsimony analysis.Tree length: 151.Numbers above branches are values from 1000 bootstrap replications.

Fig. 4 .
Fig. 4. Minimum spanning network of Ixinandria steinbachi haplotypes.Distances correlate to the number of mutations in the sequence as marked by open dots.Circle size corresponds to frequency of haplotypes.White: Juramento, black: Bermejo 1, grey: Bermejo 2.
Morphometric data in Table1.Head and body strongly depressed.Trunk and caudal peduncle becoming more compressed caudally.Dorsal profile of body convex from snout to end of supraoccipital, straight or slightly convex from supraoccipital to dorsal-fin spine, straight from dorsal fin to one or two plates before caudal fin.Upper edge of orbit not raised.Scarcely developed triangular postorbital notch.

Table 1 .
Descriptive morphometrics of Ixinandria steinbachi of four distinct populations.Values are given as percents of standard length or of head length.Lec = Lectotype of Loricaria steinbachi, SD = standard deviation.