Geometric morphometrics throws light on evolution of the subterranean catfish Rhamdiopsis krugi (Teleostei: Siluriformes: Heptapteridae) in eastern Brazil

Study area

The distribution of Rhamdiopsis krugi is in an area geographically known as Chapada Diamantina, in the central portion of Bahia state, eastern Brazil. With an area of ca. 38 000 km², the Chapada Diamantina landscape is dominated by karst landforms formed by the dissolution of carbonatic and, to a lesser extent, siliciclastic rocks. Typical karst features include exokarst (karst in surface) landforms such as fluted rock outcrops and large springs that form waterfalls at vertical walls, enclosed depressions and sinking streams (Pereira, 1998), and endokarst, such as caves and networks of smaller spaces that define subterranean habitats (Juberthie, 2000).

Rhamdiopsis krugi specimens inhabit lentic waters in the upper phreatic zone of a large karstic area (over 300 km²), connected to the surface through caves (Bockmann & Castro, 2010). These catfish are found either in deep and large lakes with clean water or in isolated, soft-bottomed pools, showing very distinct physical conditions and behaviour between them. Rhamdiopsis krugi catfishes from Poço Encantado Cave are distributed from 0.3 until 15 m of depth, calmly swimming near the substrate and frequently, at the mid-water (average speed approximately 0.03 m.s^–1; Mendes, 1995), rarely are shown close the surface; populations from Gruta Moreno, Lapa do Bode, Gruta Natal and Canoa Quebrada caves are distributed from 0.3 to 1.8 m depth and show preference to swim close the substrate formed mainly by silt, limestone riffles and guano piles and, less frequently, at the mid-water and surface (M. E. Bichuette unpubl. data).

The five study caves are developed in exposed limestones of Neoproterozoic age, belonging to the Una Geological Group, Salitre Formation. They are situated in the Irecê (Canoa Quebrada Cave – 12°25′29.7″S 41°33′28.2″W; altitude 690 m) and Una–Utinga metasedimentary basins: Gruta do Moreno (12°48′32.7″S 41°09′53.1″W; altitude 310 m), Lapa do Bode (12°56′06.5″S 41°03′53.9″W; altitude 281 m), Poço Encantado (12°56′41.8″S 41°06′17.3″W; altitude 385 m) and Gruta Natal (12°59′32.4″S 41°05′32.8″W; altitude 290 m) (Fig. 1). These two basins are separated by much older, Mesoproterozoic exposed rocks of the Chapada Diamantina Group, disposed in three layers, two of which are formed by sandstones (respectively, Morro do Chapéu and Tombador formations) intercalated by limestones of the Caboclo Formation (Schobbenhaus et al., 1984) (Fig. 2A).

The presence of caves in the Chapada Diamantina sandstones, such as the 2000 m long Torras and Parede Vermelha caves (Gallão & Bichuette, 2012), indicates karstification of these rocks. According to Ford & Williams (2007), siliceous sandstones and quartzites can develop dissolutional karst landforms, but in a more limited range of forms and hydrogeological characteristics than calcareous ones. These authors also mention that the quartzose rocks of most caves are vadose features (karst features developed in the unsaturated zone, above the phreatic). This would be the case of caves in the Chapada Diamantina Formation, although the presence of scallops in some of them (R. Fraga pers. comm.) may indicate some karstic activity under phreatic conditions.

The distributional area of Rhamdiopsis krugi is within the range of the Rio Paraguaçu basin, that belongs to the eastern Brazilian region which comprises several parallel river basins running directly to the Atlantic Ocean (Ribeiro, 2006). Bockmann & Castro (2010) hypothesized that this species is derived from a clade distributed throughout a region comprising the Upper São Francisco, Upper Paraná basins and other eastern basins, which arrived in the Paraguaçu basin after an event of hydrological capture caused by neotectonism, when the Santo Antônio basin (located in the Irecê basin) became part of the Paraguaçu basin.

Specimens of Rhamdiopsis krugi have been found in 12 out of 19 caves with water bodies prospected by M. E. Bichuette in the Chapada Diamantina region (Fig. 1; Table 1), but with no records in sandstone caves of the Chapada Diamantina Formation, where other troglobitic catfishes occur, the Copionodontinae Glaphyropoma spinosum Bichuette, de Pinna & Trajano, 2008 and the undescribed Copionodon sp. (Trichomycteridae) (Bichuette, de Pinna & Trajano, 2008b; M. E. Bichuette in prep.). The main physical characteristics of the localities of the R. krugi localities are in Table 2. Specimens are deposited at Laboratório de Estudos Subterrâneos of Universidade Federal de São Carlos – Ichthyological collection (LESCI) and Laboratório de Ictiologia de Ribeirão Preto da Universidade de São Paulo (LIRP), both in Brazil.

Sample collection

In total, 79 specimens were collected through hand-netting during several expeditions from 2004 to 2008: Gruta do Moreno Cave – GM (17 specimens), Poço Encantado Cave – PE (13 specimens), Lapa do Bode Cave – LB (18 specimens), Gruta Natal Cave – GN (13 specimens) and Canoa Quebrada Cave – CQ (18 specimens). The specimens were killed by over-anesthesia in benzocaine solution, fixed with 10% formaldehyde and subsequently preserved in 70% ethanol. Because troglobitic populations and subterranean habitats are generally fragile, we chose to collect the minimum number of specimens possible.

In order to test if there are differences in body shape, we used landmark-based geometric morphometric techniques to represent the shape of the catfishes, using the coordinates to test for significant differences in body size and shape (Monteiro & Reis, 1999).

For comparison, we included in the morphometric analyses seven specimens from an undescribed epigean species of Rhamdiopsis (F. A. Bockmann pers. comm.), collected in Rio Caldeirão (12°39′33′′S 041°22′12′′W; altitude 320 m) (six specimens), tributary of Rio São José, and in Rio Capivara near the dirt road Lençóis-Andaraí (12°37′26′′S 041°22′35′′W; altitude 341 m) (one specimen), both tributaries of the Rio Paraguaçu in Lençóis County (Fig. 1), which is basically a sandstone area (Fig. 2A).

Localities of troglobitic Rhamdiopsis in the Irecê sedimentary basin were not included in the original description of R. krugi (see Bockmann & Castro, 2010). Therefore, in order to confirm the identification of catfish from Canoa Quebrada Cave, we checked the presence of the anatomical autapormophies in two specimens from this cave, including a cleared and stained one (using the method of Taylor & van Dyke, 1985).

Data analyses

To capture images, a digital camera with a resolution of 8.1 megapixels was used. Each specimen was photographed in dorsal view, horizontally aligned, and measured with a digital caliper with centesimal precision to obtain a size reference for the analyses.

Twelve landmarks were chosen in dorsal view to determine the general outline of the body and main features, such as relative size of head and anterior and posterior segments of the body, position of the nostrils, opercle openings and insertion of fins (Fig. 3): 1. Snout tip; 2. anterior margin of anterior nostril (right); 3. Same as 2 (left side); 4. anterior margin of posterior nostril (right side); 5. Same as 4 (left side); 6. posterior tip of supraoccipital bone process; 7. posterior tip of right opercle; 8. Same as 7 (left side); 9. origin of first ray of right pectoral fin; 10. origin of first ray of left pectoral fin; 11. origin of first dorsal fin ray; 12. Anterior edge of caudal fin.

Morphometric data were acquired as homologous landmark coordinates of two dimensions, using the software TPSDig 2, version 2.10 (Rohlf, 2007).

Size was computed as centroid size (Bookstein, 1991), and compared among epigean and cave populations. To determine the degree of morphometric differentiation among species, the partial warp matrix, including the uniform component, was used in a multivariate analysis of variance (MANOVA). Canonical variates analysis was employed to verify the differences among groups and shape among different samples. It was explored by calculating least-squares Mahalanobis distances and employing those in a cluster analysis by the unweighted pair-group method, using arithmetic averages (UPGMA) to represent the similarities between configurations. All analyses were performed in the R programming environment (R Core Team, 2014) using the geomoph package for geometric morphometric analyses (Adams & Otarola-Castillo, 2013).

Monophyly of Rhamdiopsis krugi: colonization of the subterranean habitat and speciation

Rhamdiopsis krugi was described on the basis of samples from Poço Encantado (the type-locality), Lapa do Bode and Natal Caves (Bockmann & Castro, 2010). The osteological and external autapomorphies described by these authors are also present in specimens from Canoa Quebrada Cave (M. E. Bichuette pers. obs.), as well as the behaviours observed in habitat – life in lentic habitat, non-cryptobiotic and non-photophobic behaviours.

Among the 24 autapomorphies listed by these authors to define the species, eight clearly represent troglomorphisms: eyes completely absent and correlated atrophied optic foramen; body coloration absent (regressive, neutral character states; Wilkens, 2010); fatty tissue broadly spread throughout the body, probably as an adaptation to starvation (Weber, 1996; Hüppop, 2012); adults of small to medium body size, reaching 53.5 mm of standard length (SL) (most heptapterids are larger than 5 cm SL – Bockmann & Guazelli, 2003), probably as an adaptation to life in small spaces; non-photophobic behaviour and poorly developed circadian rhythmicity, probably regressive characters (Trajano et al., 2009); non-cryptobiotic behaviour (idem), associated to marked mid-water activity, possibly correlated to another autapomorphy, the deep and rounded anal fin; life in lentic habitats (most catfishes, including heptapterids, inhabit water bodies with medium to fast velocity).

Troglomorphic organisms in general are highly homoplastic, widely known for reduced eyes and melanic pigmentation, a phenomenon also observed in the behavioural traits mentioned, which are common traits of troglobitic heptapterids (Trajano & Bockmann, 1999) and of several other siluriforms (Parzefall & Trajano, 2010). Such is not the case with the several osteological and external morphological character states unrelated to the subterranean way of life, which, when combined, encompass a strong phylogenetic signal (e.g. subpreopercle absent; posterior limb of the transverse process 4 undivided, with spatulated shape; posterolateral corner of the posterior portion of the posterior branch of the transverse process of vertebra 4 extending approximately to mid-length of the transverse process of vertebra 5; posterior lobe of the adipose fin straight; hypural 5 usually co-ossified to hypural 4 at its distal portion; dorsal hypural plate typically with seven rays – respectively, autapomorphies 5, 7, 8, 10, 12 and 13 in Bockmann & Castro, 2010).

Emphasis must be given to the peculiar, widely exposed pseudotympanum. This is a most remarkable autapomorphy of R. krugi, probably involved in the enhancement of sound sensitivity (Bockmann & Castro, 2010) with low probability of homoplasy. These observations also support the presumed monophyly for the presently studied taxon. Therefore, there is little doubt that all studied populations diversified as a result of a single colonization and subsequent speciation event.

In addition to the smaller size, a slender and elongate body (due to the elongation of the anterior portion of the trunk), a broad head and snout, and dorsal and pectoral fins displaced backwards with a large space between the opercular opening and the origin of the pectoral fin, as shown by the morphometrics results, may be added as putative autapomorphies for R. krugi, pending on further confirmation, based on examination of specimens of other epigean species, such as R. microcephala and R. moreirai. A broadened head in comparison with epigean relatives is relatively frequent among specialized troglobitic vertebrates, including representatives of diverse fish families such as amblyopsids, bythitids, synbranchids, characids, ictalurids and, among heptapterids, Rhamdia spp., and salamanders (Durand, 2005; Christiansen, 2012). Its adaptive value may be related to the expansion of the cephalic surface allowing for the increase in number of sensorial structures, such as taste buds, free neuromasts and electroreceptors (described for the intensively studied Mexican tetra characin, Astyanax jordani (Hubbs & Innes, 1936), cave amblyopsids, and the iconic European salamander, Proteus anguinus Laurenti, 1768, among others (e.g. Boudriot & Reutter, 2001; Niemiller & Poulson, 2010); the enlargement of the mouth allowing for capture of larger prey, an important adaptation in the typical food-poor subterranean environment; and/or an enhanced hydrodynamic effects in front of the head while swimming (Wilkens, 2005). All of these explanations may apply to the broadened snout of R. krugi.

The monophyly of the taxon composed of the populations from both Una and Irecê metasedimentary basins poses an initial problem: how could these populations, currently living in, at least partially, isolated metasedimentary basins have originated from a single ancestral population? As a first approach to this question, it is useful to examine available data about: (1) the geology of the region and rates of karstification; (2) the state-of-art of the systematics and distribution of the genus; and (3) evolution of the major South American river basins.

The genus Rhamdiopsis is distributed in two major river basins, the Paraná (running to the south, in direction of Mar del Plata) and the São Francisco, and a set of several parallel rivers running toward the east coast of eastern Brazil (Bockmann & Guazelli, 2003; Bockmann & Castro, 2010). Ancient dates have been proposed for the formation of the South American basins and the diversification of the freshwater SA ichthyofauna. For instance, according to Lundberg et al. (1998) the Paraná and São Francisco drainages were established, in configurations similar to the present-day ones, by the late Oligocene–early Miocene (from 30-20 Ma). Ribeiro (2006), analyzing Brazilian coastal basin histories, which includes Rio Paraguaçu basin, proposed a pattern where Brazilian eastern coastal rivers are inhabited by remaining taxa of an ancient biogeographic history, such as the Copionodontinae basal catfishes (Trichomycteridae), and consider a Mesozoic age (late Cretaceous) for the initial phase of diversification of this endemic ichthyofauna.

Rhamdiopsis krugi is likely a basal species within the genus (F. A. Bockmann pers. comm.), and its high degree of troglomorphism makes it reasonable to suppose that it was isolated in the subterranean habitat very early in its evolutionary history. Low denudation rates were estimated from the São Francisco Craton, around 30 ± 10 m/Myr, within the range recorded for Padre Cave, another karst area in Bahia (Auler et al., 2009). Associating these rates with the ca. 300 m difference in altitude between the top of Chapada Diamantina and the base level of the Irecê basin (Fig. 2B) and based on the history of phreatic habitats formation, the latest underground connection between Irecê and Una-Utinga limestones across the Chapada would have occurred approximately 10 Mya. In this scenario, this would be the minimum age for the origin of the R. krugi lineage; a seemingly old age, but consistent with its high degree of morphological and behavioural specialization, as well as with its basal position within the genus. This is because Rhamdiopsis catfishes, with 17 species (most of them undescribed yet), are distributed in several rivers of the eastern and Paraná basins (Bockmann & Castro, 2010), its origin consequently predating the separation of these basins in the late Miocene. Old ages have been proposed for other South American freshwater fishes, such as the armoured catfish, genus Hypostomus (up to 11.8 Myr – Montoya-Burgos, 2003). Langecker & Longley (1993) proposed older ages, in the late Early Miocene, for the highly specialized phreatobic ictalurids, Satan eurystomus Hubbs & Bailey, 1947 and Trogloglanis pattersoni Eigenmann, 1919.

Following the progressive erosion of limestones over the Chapada Diamantina sandstones, the limestones of the Una and Irecê basins became separated, causing the isolation of two sets of R. krugi populations. Nevertheless, it must be noted that because the axis of the folding sedimentary layers declines in a south to north direction, the dissection of these limestones progresses from south to north. Therefore, more recent connections may have occurred through the north in the lower areas around the Chapada Diamantina, where limestones of the Bebedouro Formation presently surface (see Fig. 2B). Even a present-day connection is possible through subterranean aquifers in subsurface limestones. According to this hypothesis, the presence of R. krugi or closely related species, besides Itaetê, Iraquara, Palmeiras and Nova Redenção, is predicted for Souto Soares, Utinga, Wagner, Ruy Barbosa and Lajedinho regions (Fig. 1, see municipality boundaries).

An alternative hypothesis, less likely but still possible, is a current connection through the Chapada Diamantina sandstones. It is widely accepted that karst processes may develop in siliciclastic rocks. However, cited cases of speleogenesis in sandstones and quartzites refer to vadose caves. Although the authors do not disclose the possibility of conduit aquifers in such rocks (e.g. Ford & Williams, 2007: 389–390 – ‘most caves in quartzose rocks are vadose features’), no actual example is presently known. Nevertheless, as already mentioned, the presence of scallops in some sandstone caves of the Chapada Diamantina may indicate some karstic activity under phreatic conditions. Thus this hypothesis deserves attention. Currently, the Irecê base level is at least 300 m higher than the Una base level (Fig. 2B). Under these conditions, and considering the small size of R. krugi and the large distances between these basins, there is a higher probability of dispersion from Irecê to Una than vice-versa. This is consistent with the greater similarity between specimens of EP and CQ than in relation to specimens from Una.

Like all of South America, Bahia was subject to an alternation of dry and wet periods, well documented over the last 350 000 years (Wang et al., 2004; Auler et al., 2006). It is reasonable to suppose that climatic fluctuations have occurred during the Miocene as well. Thus the paleoclimatic model (Barr, 1968; Trajano, 1995) may apply to the presently studied catfish. According to this model, during dry periods with the disruption of surface drainage, the fauna in subsurface habitats (hyporheic and/or subterranean) could survive and, once isolated, differentiate. Accelerated divergence rates are expected because subsurface selective regimes highly contrast with surface ones, characterized mainly by absence of photoautotrophy and food scarcity (sometimes extreme), as well as the functional impossibility of visual orientation. Thus, colonizing populations would be relatively small, prone to genetic drift by founder effects and other factors.

The hypothesis of vertical colonization of the subterranean environment through the hyporheic zone is also compatible with models of parapatric speciation. In any case, the high degree of morphological, ecological and behavioural specialization of R. krugi, characterized by an elevated number of autapomorphies, well above the usual for epigean species, indicate a long period of evolution in the phreatic habitat. Bockmann & Castro (2010) stated that these autapomorphies could be functionally related, appearing in ‘packages’ due to paedomorphosis.

Divergence within Rhamdiopsis krugi: differentiation of (semi)isolated populations based on geometric morphometric analysis

Geometric morphometric analysis shows that there are significant differences in body shape between CQ, GN caves and the group formed by PE, LB and GM caves. Such differences may be due to the geographic isolation provided by geological and hydrological barriers. These barriers are represented by insoluble rocks (Irecê versus Una) and running waters (GN versus PE + LB + GM). Hydrological barriers have also been proposed to explain divergence between two molecular clades of troglobitic amblyopsid fish, Amblyopsis spelaea Dekay, 1842, separated by the Ohio River, USA (Niemiller et al., 2013). Based on multilocus genetic data, up to 15 putative cryptic species were identified by Niemiller, Near & Fitzpatrick (2012) within the nominal species Typhlichthys subterraneus Girard, 1859 (Amblyopsidae), showing how the subterranean biodiversity may be underestimated.

The water bodies in which R. krugi lives inside PE, LB and GM caves belong to the same aquifer, since the variations in water level follow those observed in the Rio de Una (Karmann, Fraga & Mendes, 2002). The divergence between GM (broader body), GN and PE caves (shorter posterior segment of the body) may be explained by underground isolation by depositional processes, since several connecting ducts are colmated by sediment (R. Fraga and F. Laureano pers. comm.).

Biomolecular analysis based on RAPD bands (Bichuette et al., 2008a) mostly agreed with our results: a bootstrapped Maximum Likelihood tree using Contml and Seqboot (1000 iterations; Phylip 3.5) arbitrarily rooted with two heptapterid species (P. kronei and an epigean species of Rhamdiopsis) showed that individuals from the same basin (Una or Irecê) generally clustered together – Moreno Cave population clustered with Natal cave population and then Moreno/Natal cave populations clustered with PE and LB cave populations. Interestingly these populations are in the same limestones of Una Basin. These four clusters were quite different from Canoa Quebrada/Torrinha cave populations, caves located in the Irecê basin (Bichuette et al., 2008a). This topology, similar to the obtained in our study (Fig. 6), corroborates the hypothesis of separation between populations from different sedimentary basins.

The differences detected among the five studied populations may be related to habitat characteristics selecting for different body shapes (Table 2, Fig. 7A, B). In fact, the main differences in shape were related to body width and relative proportion of body segments, with the repositioning of the pectoral and dorsal fins. We hypothesize that these characteristics are related to the stabilization of the body as an adaptation for the activity in the water column, more advanced in the PE cave (Fig. 7A) and less in CQ cave (Table 2, Fig. 7B), for example.

Genetic drift might also exert influence over this differentiation. Leinonen et al. (2006) analyzed body differences in marine and isolated lacustrine populations of threespine sticklebacks. These authors suggest that genetic drift played a major role in phenotypic divergence among lake populations, given that these are small-sized fish, and that natural selection may become ineffective in such cases.

It is important to stress that, as with any phenogram based on similarities like those generated by molecular data, the phenogram depicted in Figure 6 is not a cladogram (i.e., a hypothesis of phylogenetic relationships resulting from the application of the comparative method). For instance, the grouping of GMa and LB caves samples may be due to the retention of a plesiomorphic body shape in relation to specimens from PE cave. The huge lake in the latter cave is the most specialized among the habitats of R. krugi populations, requiring morphological and behavioural adaptations for swimming in spaces that are very large for these tiny fish that originated from a typical bottom-dweller living in lotic epigean habitats.

Because no other population from Irecê basin was included in our analysis, one cannot state that the CQ cave population is basal in relation to the Una populations. Nevertheless, greater similarity to the epigean sample indicates that these catfish retained a more plesiomorphic body shape, with the set of Una populations presenting a more advanced condition. Further studies are needed, including additional populations from Irecê basin and a larger number of characters (e.g. myology, brain morphology). Additionally, cave localities to the north of Chapada Diamantina should be prospected in order to test the hypothesis of connection through the limestone outcrops of the Salitre Formation in the Iraquara, Souto Soares and Barra do Mendes regions.

Morphometric methods have been for the last half-century the basic toolkit for the study of organismal shape and its covariation with size, time, environment, geography, ecology, development and other factors (MacLeod, 2008). Over the last 10 years the collection of techniques denominated geometric morphometrics, grounded in a strong theoretical and mathematical framework, has provided the ability to robustly interpret, through multivariate axes and deformation grids, the differences in shape between organisms. As noted by MacLeod (2008), it is surprising that both traditional (sensu Marcus, 1990) and geometric morphometrics never explored in an appropriate way the correlation between form and taxonomy, even though the recognition of taxonomic entities is strongly based on morphology, while the use of molecular approaches yield limited results in some cases. This, combined with the need to develop new methods of identifying taxonomic entities, and also taking into account the taxonomic impediment (de Carvalho et al., 2007, 2008), suggests that morphometric approaches may play an important role, not yet properly explored in systematics.

Conservation recommendations

Because the distributional range of R. krugi is located outside protected areas such as the Chapada Diamantina National Park, these populations are not protected by law. Therefore, they are subject to local environmental impacts, including: habitat conversion due mainly to limestone exploration and garbage and debris accumulation (due to low water circulation; Trajano & Bichuette, 2006); uncontrolled exploration of the subterranean waters in that region; pollution of the aquifers by pesticides and poorly regulated speleotourism. The maintenance of genetic diversity at all levels is essential for the preservation of biodiversity; hence conservation policies should be directed at each of the R. krugi populations because, as we have shown, they differ significantly in morphology. However, conservation regulations are extremely bureaucratic and only formally named species may be included in Brazilian Red Lists. Consequently, instead of being concerned with protecting real biological units, the environmental authorities only focus on strict taxonomic determinations. Because the taxonomic impediment is especially serious in countries with mega-diversity, many biological units in Brazil are extinct before they can be officially protected, with considerable loss of such diversity. Therefore, for conservation purposes, we propose a referral to the two sets of differentiated populations of R. krugi as R. krugi ‘Una morphotype’ and R. krugi ‘Irecê morphotype’. We also recommend special attention to each population, whose protection should be prioritized.