Based on recent discoveries, Bujurquina appears to be the most widely distributed and species-rich cichlid genus in the western Amazon of South America. In this study, using a large representative sample of Bujurquina covering its whole distribution area, we use morphological and molecular data to test the hypothesis that each major western Amazon basin includes multiple endemic Bujurquina species arranged along an elevational river gradient and that these species are upland- and lowland-adapted in their ecomorphology. The hypothesis derives from two lines of evidence, i.e. observations of distribution patterns in Bujurquina and paleogeographic reorganisation of western Amazon drainage patterns. Body shape morphometrics and a biogeographic reconstruction of molecular phylogeny supported our hypothesis, confirming that upland and lowland Bujurquina show consistent differences in body shape and proportions that can be explained as repeated adaptations to local aquatic conditions within each main river basin. Ecomorphological divergence in relation to lentic and lotic waters (lowland and upland habitats) was repeated in all five basins studied, i.e. the Madre de Dios, upper Ucayali, central western Amazon-Huallaga, Marañón and Napo-Putumayo basins.
Introduction
With the most diverse and extreme terrestrial and aquatic ecosystems on Earth, the Amazon basin, and especially the western Amazon and the adjacent Tropical Andes, is the most biodiverse region on the planet (Reis et al. 2003, Hubbell et al. 2008, Hoorn & Wesselingh 2010a, b, Albert & Reis 2011, Wesselingh & Hoorn 2011). Paramount amongst its megadiverse groups are the freshwater fishes, easily the largest community of freshwater fishes in the world (Reis et al. 2003, Albert & Reis 2011, Van der Sleen & Albert 2018). However, data on Amazon species megadiversity is based on a relatively limited area of this vast and complex region (Hoorn & Wesselingh 2010a, b, Albert & Reis 2011, Van der Sleen & Albert 2018), with sampling biased toward navigable rivers and other accessible areas with areas of lowland terra firme remaining poorly explored, especially regarding freshwater fishes (Albert & Reis 2011, Crampton 2011, Van der Sleen & Albert 2018, Oberdorff et al. 2019). Consequently, our knowledge of the Amazon's present biodiversity remains limited, as does our understanding of its evolution. The general outlines of Amazon evolution have been derived predominantly from geological and paleontological data, while input from extant faunal data and phylogenies remains limited (Hoorn & Wesselingh 2010a, b).
The Andean orogeny, which has shaped the Andes and the western Amazon river network, appears to have been of paramount importance for Amazon evolution, being the primary determinant of both aquatic and terrestrial faunal diversity (Hoorn & Wesselingh 2010a, b, Albert et al. 2011, Albert & Reis 2011, Wesselingh & Hoorn 2011, Albert et al. 2018a, b). During this process, the original south-north orientation of the paleo-Amazon drainage changed with progressive Andean elevation and eastward propagation and sedimentation, forcing the river network into its present west-east orientation (Hoorn & Wesselingh 2010a, b, Wesselingh & Hoorn 2011). This large-scale reorientation of the River Amazon network must have reshaped biodiversity patterns and promoted further diversification in both the lowlands and foothills. The present mosaic of muddy, clear and black-water rivers was established from a previous, and still little-known, river network preceded in the western Amazon by a mega-wetland, the Pebas formation (Hoorn & Wesselingh 2010a, b, Wesselingh & Hoorn 2011).
Freshwater fishes are one of the best model groups for studying continental evolution as they are the least dispersal-capable of the larger organisms and hence most strictly associated with landscape evolution (Darlington 1957, Wallace 1876). To have the greatest explanatory power in terms of biogeography and landscape evolution, fish groups need to be highly philopatric; ideally, long-term endemic in the western Amazon, display small-scale endemism and high species diversity, should be found throughout all aquatic habitats (including lowlands and foothills, main rivers and their inundation zones) as well as terra firme lowland areas, and dated molecular phylogenies should be available. Among Amazon freshwater fishes, the cichlid fishes fulfil all the above requirements. Moreover, cichlids are one of the most important Neotropical fish groups, being the dominant group of larger-sized fishes in Middle America (Myers 1966, Bussing 1976, 1985, Říčan et al. 2013, 2016) and the third richest family of fishes in South America (Van der Sleen & Albert 2018).
Within the cichlid fishes, the genus Bujurquina Kullander, 1986 represents the best model group for studying Andean and western Amazon landscape evolution and its direct influence on the formation of biodiversity (Říčan et al. 2022). Bujurquina species are highly philopatric, show small-scale endemism, strong adaptation to local environmental conditions, and are highly diverse in the lowlands and Andean foothills (unlike Apistogramma, the other species-rich genus of the western Amazon). Furthermore, it is the only cichlid genus whose geographic distribution and highest diversity is centred in the western Amazon and whose distribution and endemism follows the putative course of the paleo-Amazon (Kullander 1986, Říčan 2017, Říčan et al. 2022). Consequently, Bujurquina species have a distribution unlike any other large South American cichlid genera, which generally (not Apistogramma) show the highest diversity and endemism in the older eastern Amazon within the Brazilian and Guiana shield regions (Stawikowski & Werner 1998, 2004, Kullander et al. 2018, Van der Sleen & Albert 2018, Říčan et al. 2021). These factors make Bujurquina the best candidate for studying group-specific and generalised patterns and processes in the evolution of western Amazonia. Studying the diversity patterns and evolution of Bujurquina will strengthen our understanding of cichlid and fish diversification processes in this area and provide a baseline for faunal and biogeographic evolution in general in the western Amazon.
Bujurquina distribution follows an arc from Argentina, Paraguay and Bolivia in the south, through its highest currently known centre of diversity and endemism in Peru (Kullander 1986) and Ecuador (Říčan et al. 2022), and continues to the north through Colombia and Venezuela (Kullander 1986, Stawikowski & Werner 1998, Van der Sleen & Albert 2018). Molecular-based dating and biogeographic analyses (Musilová et al. 2015, Říčan et al. 2022) support the hypothesis that most of the genus' evolution occurred throughout the period of major Amazonian river network reorganisation. Both molecular and colouration-pattern data demonstrate a division of the genus into two main phylogenetic groups, the Southern and Northern groups, that divide within the present Ucayali basin of Peru (Říčan et al. 2022). These two Bujurquina lineages are characterised by different colouration patterns, chiefly on the dorsal fin, which is unpatterned in the Southern group but ornamented in the Northern group.
In addition to the main biogeographic division of the genus into Southern and Northern groups, the biogeography and endemism of Bujurquina plays out in two main biogeographic and ecological zones. There is a separate band of Bujurquina endemism in the Andean foothills, which has a distinct fauna from the supposedly more widely distributed species in the Amazonian lowlands (Kullander 1986, Říčan et al. 2022). The two zones host different Bujurquina faunas, with the lowland species usually not found in the adjacent foothills and vice versa (Kullander 1986, Říčan et al. 2022). Thus, foothill areas have a different Bujurquina fauna than lowland areas within the same river basins, at least in the centre of known Bujurquina diversity in Peru, where Kullander (1986) first proposed the hypothesis (Říčan 2017), and in Ecuador (Říčan et al. 2022).
Fig. 1.
Distribution of valid species, the two main groups of Bujurquina, and four of the putative species identified. The distribution of the newly identified Ecuadorian species is shown in Fig. 2. All species were included in morphological analyses, and those with an asterisk were also included in molecular phylogeny. (H) and (L) identify highland and lowland species, respectively, with highland and lowland areas approximately demarcated by the grey line. See Fig. S1 for a map identifying i) the division between highland (H) and lowland (L) species within areas of the western Amazon, and ii) the main river types in the Amazon. Examples of the main highland and lowland Amazon stream types where Bujurquina species occur are shown in Fig. S2.
Bujurquina currently includes 18 described species; however, this species diversity is significantly underestimated, as demonstrated by Říčan et al. (2022). If we generalise biogeographic patterns to all valid Bujurquina species, the highland (foothill) band of species includes B. eurhinus Kullander, 1986; B. labiosa Kullander, 1986; B. huallagae Kullander, 1986; B. ortegai Kullander, 1986; B. zamorensis Regan, 1905; and the northernmost species B. mariae Eigenmann, 1923 (Orinoco basin in Colombia and Venezuela). The lowland and supposedly more widely distributed endemic band includes B. vittata Heckel, 1840; B. oenolaemus Kullander, 1987; B. cordemadi Kullander, 1986; B. tambopatae Kullander, 1986; B. robusta Kullander, 1986; B. megalospilus Kullander, 1986; B. apoparuana Kullander, 1986; B. hophrys Kullander, 1986; B. moriorum Kullander, 1986; B. peregrinabunda Kullander, 1986; B. syspilus Cope, 1872; and B. pardus Arbour, Salazar & López-Fernández, 2014 (Fig. 1, Figs. S1-S2).
Typical highland species of Bujurquina occur in the Andean foothills, generally between 400 m and up ca. 1,200 m elevation, in streams and rivers with seasonally clear and permanently running water, with river beds composed of stones and pebbles (Kullander 1986, Crampton 2011, Říčan 2017; Figs. 1-2, Figs. S1-S2, Table S1). Typical lowland species occur below 200-300 m elevation, in rivers with low to zero visibility or in small rainforest streams with clear water. The lowland rivers and streams are generally slow-flowing or stagnant with beds composed of soft material (mud, sand, leaves, etc.), with high habitat complexity due to accumulated tree trunks, branches, etc. (Kullander 1986, Crampton 2011, Říčan 2017; Figs. 1-2, Figs. S1-S2, Table S1). At elevations between 200-400 m, there are zones of intergradation between typical highland and lowland habitats (Fig. S2).
Fig. 2.
Distribution of Bujurquina species identified in Ecuador. Dots show localities, while colours and large numbers of the same colour identify each species (species are identified in the text accompanying these numbers and by names derived from their distribution). All species were included in morphological analyses, and those with an asterisk were also included in molecular phylogeny. (H) and (L) identify highland and lowland species, respectively, with highland and lowland areas approximately demarcated by the grey line. Rivers are identified in white letters. The two main Ecuadorian western Amazon basins (Curaray-Napo-Putumayo plus extralimital Caquetá and the Santiago-Morona-Pastaza and the Tigre, all tributaries of the Marañón) are identified with the drainage divide shown.
Fig. 3.
Representative live specimens of the valid Bujurquina species (plus B. sp. Bolivia) in the Southern group (no photograph of a live specimen of B. cordemadi was available. We are not aware that the species has ever been reliably identified or photographed alive).
In this study, we use the extensive collection of Bujurquina specimens of Říčan et al. (2022) to test the hypothesis that each major western Amazon river basin (in Peru and Ecuador, where our sampling is densest) includes multiple endemic species of Bujurquina arranged along an elevational river gradient and ecomorphologically upland- and lowland-adapted, most likely through body shape. This hypothesis, originally proposed by Kullander (1986) for Peru, is based on two lines of evidence, i.e. the observed distribution patterns in Bujurquina and the paleogeographic reorganisation of western Amazon drainage patterns described above. By examining this extensive collection of specimens, we can provide, for the first time, a representative sampling throughout the genus' western Amazon distribution area (Figs. 1-2, Figs. S1-S2), thus enabling us to explore diversity within the genus and test evolutionary hypotheses by comparing morphological data obtained in this study with the molecular phylogeny of Říčan et al. (2022). Specifically, using body shape morphometrics in the context of a biogeographic reconstruction of molecular phylogeny, we test whether the upland and lowland fauna show consistent differences in body shape and proportion that could be explained as adaptations to local aquatic conditions along the elevational gradient of each main river basin (Kullander 1986, Říčan 2017).
Fig. 4.
Representative live specimens of the valid Bujurquina species in the Northern group, including the newly identified putative species B. sp. Oran from Peru and the newly identified putative species from Ecuador found in the Marañón basin (colours and numbers of the Ecuadorian species correspond to those in Fig. 2).
Fig. 5.
Representative live specimens of the newly identified putative Ecuadorian Bujurquina species in the Northern group, found in the Napo and Putumayo basins (colours and numbers of the Ecuadorian species correspond to those in Fig. 2).
Material and Methods
Species determination
Species were determined by combining a priori morphological species determinations and post-hoc species delimitation using the molecular mtDNA cytochrome b(cytb) marker (Říčan et al. 2022; see below). Specimens were identified to species using original descriptions, identification keys and comparative material of all presently recognised Bujurquina species. A large proportion of the specimens sampled for this study were identified as putatively new species in Říčan et al. (2022), where they were reported as “Bujurquina sp.” (possible new/unidentified species).
In the present study, we employed the Unified Species Concept (de Queiroz 2007), whereby species are equated with independently evolving metapopulation lineages. Thus, the operational species-delimitation criteria used were consistent morphological differences and topology and divergence analyses in molecular phylogeny.
Species and specimens
For this study, we used the collection of Bujurquina specimens of Říčan et al. (2022), which, for the first time, provides a representative sampling throughout the genus' western Amazon distribution area (Figs. 1-2; for a list of the 37 Bujurquina species studied [including 17 valid species, i.e. all but one for the morphometric data] and 446 specimens examined, see Table S1). In addition, we also provide photographs of living specimens of all previously described species in the genus (except B. pardus) and all putative new species, most of which were included in this study (Figs. 3-5). The photographs provide an overview of the range of body shapes in the genus and information on the live appearance of most valid species, which, up until now, are mostly only known from black and white photographs of long preserved specimens. All specimens were collected by the first author during a dedicated long-term project, as described in Říčan et al. (2022). The specimens are currently deposited in the fish collection of the Department of Zoology, Faculty of Science, University of South Bohemia in České Budějovice, Czech Republic. The specimens will be transferred to publicly available museum collections for permanent storage and study upon formal species description.
Molecular phylogeny
Molecular data and phylogeny based on the mitochondrial cytb marker analysed using Bayesian Inference (BI) in BEAST were previously generated and presented in Říčan et al. (2022) and are used here to compare with results based on morphological characters.
Morphological methods
Variation patterns in morphometric measurements, obtained as described by Kullander (1986), were visualised using principal component analysis (PCA) implemented in Canoco 5 (Microcomputer Power, Ithaca, NY, USA) according to Šmilauer & Lepš (2014). Measurements (see Table S2) were taken as straight-line distances, using a digital calliper to 0.1 mm, on the left side of each specimen. Body length was taken as standard length (SL), with morphometric characters expressed as a per cent of SL. Only fully-grown adult specimens > 60 mm SL with a similar size range were used for PCA analyses of proportional morphometric data. Body bar terminology follows Říčan et al. (2005), while institutional abbreviations are as those listed in Ferraris (2007) and Sabaj (2020).
Results
Bujurquina phylogeny, time frame of evolution and biogeography based on mtDNA phylogeny
Phylogenetic analysis of the cytb data matrix using BI analysis provided robust results and supported the majority of morphologically determined valid and putative species (Říčan et al. 2022), while biogeographical analysis of phylogeny (Říčan et al. 2022) produced a predominantly vicariant reconstructed history of the genus (Fig. 6). The main biogeographic dichotomy within the genus was between the Southern and Northern groups, separated within the present upper Ucayali basin (Figs. 1, 6, Fig. S1). Thus, both the Southern and Northern groups have a wide ancestral area that became fragmented by vicariance. The Southern group split into Paraguay-Paraná vs. Madre de Dios-upper Ucayali at 10.2 Ma, and then into the upper Ucayali vs. Madre de Dios at 8.4 Ma, while the Northern group, whose ancestral area includes all of the northern and western Amazon plus the Orinoco, has a more complicated pattern, the oldest identifiable vicariant event being the separation of the Orinoco basin from north-western Amazon at 7.9 Ma. Further vicariant events occurred between the Marañón and the central western Amazon plus Napo, Putumayo and Caquetá at 3.6 Ma, between the central western Amazon and Napo, Putumayo and Caquetá at 2.6 Ma, and between the Marañón and the central western Amazon at 2.4 and 2.1 Ma (Fig. 6). From the seven main biogeographic areas corresponding to river basins identified through biogeographic analysis, we concentrated on five, i.e. the Madre de Dios, the upper Ucayali, the central and western Amazon up to Huallaga, the Marañón and the Napo-Putumayo, each of which included multiple endemic parapatric species as geographic units for testing the main ecomorphological hypothesis of this study.
Fig. 6.
Phylogenetic relationships, biogeography, ecology and species delimitation of Bujurquina based on the mtDNA cytb marker from a dated BEAST analysis (modified from Říčan et al. 2022). Valid species are shown in bold. (H) and (L) identify highland and lowland species, respectively. Species are colour coded based on their geographic distribution within the five major river basins of the western Amazon plus the Orinoco and Paraguay basins (see Fig. 1 & Figs. 7-11). Note the two main groups of Bujurquina (see Fig. 1), their reconstructed ancestral areas (which correspond to the sum of the major river basin areas, and thus predominantly reconstruct vicariant evolution; shown with circles at nodes and colour codes and names of the river basins) and their correspondence to colouration patterns (Fig. 1 & Figs. 3-5). Numbers at nodes show age in Ma, while grey bars at the nodes show confidence intervals of age estimates (95% HPD). Bayesian posterior probabilities above 0.95 are shown as black points on the nodes. Note that species delimitation using the molecular marker delimits 30 species, while morphology delimits 32 species. The scale bar shows 5 My.
Fig. 7.
Morphometric variation and discrimination of Southern group Bujurquina species endemic to the River Madre de Dios basin, based on morphometric proportional values of standard length (SL) in PCAs based on specimens ≥ 60 mm SL. (H) and (L) identify highland and lowland species, respectively. Colours represent individual species and correspond to those in Fig. 1. Note the separation of species into two main lotic and lentic ecomorphs. Characters in bold and underlined are associated with separation into lotic and lentic ecomorphs in all PCA analyses (cf. Figs. 8-11; ORB in four out of five). See Table S2 for data and characters; abbreviations of characters are as follows: BD = body depth/SL, CPD = caudal-peduncle depth/SL, CPL = caudal-peduncle length/SL, DSP = last dorsal-fin spine length/SL, HL = head length/SL, HW = head width/SL, INO = interorbital width/SL, ORB = orbital diameter/SL, PL = pectoral-fin length/SL, PRO = preorbital depth/SL, SNT = snout length/SL, VRAY = pelvic-fin longest ray length/SL.
Fig. 8.
Morphometric variation and discrimination of Southern group Bujurquina species endemic to the upper River Ucayali basin, based on morphometric proportional values of standard length (SL) in PCAs based on specimens ≥ 60 mm SL. (H) and (L) identify highland and lowland species, respectively. Colours represent individual species and correspond to those in Fig. 1. Note the separation of species into two main lotic and lentic ecomorphs. Characters in bold and underlined are those associated with separation into lotic and lentic ecomorphs in all PCA analyses (cf. Figs. 7-11; ORB in four out of five). Character abbreviations are given in Fig. 7.
PCA analysis of body shape morphometrics and correspondence with biogeography and ecology
PCA analyses demonstrated that the main ecomorphological divergence, i.e. lentic and lotic, was recurrent in each of the five main river basins (Figs. 7-11), suggesting repeated local adaptations to local aquatic conditions in each river basin. In all PCA analyses (Figs. 7-11), the lotic/lentic ecomorphological divergence was best correlated with caudal peduncle length (long vs. short), caudal peduncle depth (narrow vs. deep), body depth (narrow vs. deep), pectoral fin length (short vs. long) and last dorsal fin spine length (short vs. long). In four of the five basin faunas analysed (not Napo-Putumayo; Fig. 11), lotic/lentic ecomorphological divergence was also explained by orbital diameter (small vs. large).
The remaining characters (i.e. interocular distance, head width, head length, preorbital distance, snout length, and first ventral fin ray length) were generally not correlated with lotic/lentic ecomorphological divergence. The least correlated character was the length of the first ventral fin ray (uncorrelated 4×, 1× positively correlated with the lentic ecomorph), followed by head length (uncorrelated 3×, 2× positively correlated with the lentic ecomorph). Snout length was positively correlated 2× with the lentic ecomorph, 2× with the lotic ecomorph and 1× uncorrelated; head width and interorbital distance were positively correlated 3× with the lentic ecomorph and 2× uncorrelated; and preorbital distance positively correlated 2× with the lentic ecomorph, 1× with the lotic ecomorph and 2× uncorrelated.
Fig. 9.
Morphometric variation and discrimination of Northern group Bujurquina species endemic to the central western River Amazon basin (plus the River Huallaga), based on morphometric proportional values of standard length (SL) in PCAs based on specimens ≥ 60 mm SL. (H) and (L) identify highland and lowland species, respectively. Colours represent individual species and correspond to those in Fig. 1. Note the separation of species into two main lotic and lentic ecomorphs. Characters in bold and underlined are those associated with separation into lotic and lentic ecomorphs in all PCA analyses (cf. Figs. 7-11; ORB in four out of five). Character abbreviations are given in Fig. 7.
Fig. 10.
Morphometric variation and discrimination of Northern group Bujurquina species endemic to the River Marañón basin, based on morphometric proportional values of standard length (SL) in PCAs based on specimens ≥ 60 mm SL. (H) and (L) identify highland and lowland species, respectively. Colours represent individual species and correspond to those in Fig. 2. Note the separation of species into two main lotic and lentic ecomorphs. Characters in bold and underlined are those associated with separation into lotic and lentic ecomorphs in all PCA analyses (cf. Figs. 7-11; ORB in four out of five). Character abbreviations are given in Fig. 7.
Based on geographic distribution, phylogenetic analysis and PCA of body shape morphometrics (Figs. 1-2 and 7-11), endemic Bujurquina species in the foothills include the following lotic species: B. oenolaemus (Paraguay basin, Bolivia); B. eurhinus (Madre de Dios basin, Peru); B. robusta and B. labiosa (upper Ucayali basin, Peru); B. huallagae, B. ortegai and B. sp. Oran (central western Amazon, Peru); B. zamorensis, B. pardus and B. sp. Santiago (= B. sp. Ecuador 14) (Marañón basin, Peru and Ecuador); B. sp. Upper Napo (= B. sp. Ecuador 2); B. sp. Upper Coca (= B. sp. Ecuador 5); B. sp. Upper Curaray (= B. sp. Ecuador 8); B. sp. Villano (= B. sp. Ecuador 9); B. sp. Lower Curaray (= B. sp. Ecuador 10) (Napo-Putumayo basins, Ecuador); and the northernmost species, B. mariae (Orinoco basin in Colombia and Venezuela). Lowland endemic Bujurquina species (Figs.1-2and7-11)includethefollowinglenticspecies: B. vittata (Paraguay basin, Argentina, Paraguay and Bolivia); B. cordemadi, B. tambopatae and B. sp. Bolivia (Madeira basin, Peru and Bolivia); B. megalospilus, B. apoparuana and B. hophrys (upper Ucayali basin, Peru); B. moriorum, B. peregrinabunda and B. syspilus (central western Amazon, Peru, Ecuador, Colombia, Brazil); B. sp. Middle Pastaza-Morona (= B. sp. Ecuador 12); B. sp. Lower Pastaza-Morona (= B. sp. Ecuador 13) (Marañón basin, Peru and Ecuador); and B. sp. Aguarico-Napo (= B. sp. Ecuador 4), B. sp. Upper Aguarico (= B. sp. Ecuador 6) and B. sp. Putumayo (= B. sp. Ecuador 7) (Napo-Putumayo basins, Ecuador).
Fig. 11.
Morphometric variation and discrimination of Northern group Bujurquina species endemic to the River Napo basin (plus one species shared with the Putumayo basin), based on morphometric proportional values of standard length (SL) in PCAs based on specimens ≥ 60 mm SL. (H) and (L) identify highland and lowland species, respectively. Colours represent individual species and correspond to those in Fig. 2. Note the separation of species into two main lotic and lentic ecomorphs. Characters in bold and underlined are those associated with separation into lotic and lentic ecomorphs in all PCA analyses (cf. Figs. 7-11; ORB in four out of five). Character abbreviations are given in Fig. 7.
Body shape morphometrics and diagnostics
All species within the Madre de Dios basin (B. eurhinus, B. tambopatae, B. cordemadi, B. sp. Bolivia) were diagnosable using solely morphometric data, with no overlap in the PCA plot (Fig. 7). Within the upper Ucayali basin, however, just three out of five species (B. megalospilus, B. robusta, B. labiosa but not B. apoparuana and B. hophrys) were diagnosable using solely morphometric data, with no overlap in the PCA plot (Fig. 8), and four of six species (B. moriorum, B. peregrinabunda, B. syspilus, and B. sp. Oran) within the central western Amazon basin, B. huallagae, B. ortegai showing partial separation (Fig. 9). Species in the Marañón basin showed least diagnosability, with none of the four species included (B. zamorensis, B. sp. Middle Pastaza-Morona = B. sp. Ecuador 12, B. sp. Lower Pastaza-Morona = B. sp. Ecuador 13, B. sp. Santiago = B. sp. Ecuador 14) possible to identify based solely on morphometric data, all of them showing substantial overlap in the PCA plot (Fig. 10). Three species (B. sp. Aguarico-Napo = B. sp. Ecuador 4, B. sp. Upper Aguarico = B. sp. Ecuador 6, B. sp. Lower Curaray = B. sp. Ecuador 10) in the Napo-Putumayo basin were diagnosable using solely morphometric data with no PCA overlap, with three to four species showing some overlap (B. sp. Upper Napo = B. sp. Ecuador 2, B. sp. Upper Coca = B. sp. Ecuador 5, B. sp. Upper Curaray = B. sp. Ecuador 8, B. sp. Villano = B. sp. Ecuador 9), though this was mainly attributable to B. sp. 2 as B. sp. 5 did not overlap with B. sp. 8 and 9 (Fig. 11). One species, B. sp. Putumayo = B. sp. Ecuador 7, was assessed with the inclusion of just one adult specimen. As such, proportional morphometric data were relatively robust in distinguishing different Bujurquina species, with the majority of species, including many of the putative new species, possible to diagnose based solely on morphometric data.
Discussion
Phylogeny and the two main Bujurquina groups
Based on colouration patterns, particularly the presence/absence of dorsal fin ornamentation, Říčan et al. (2022) hypothesised that Bujurquina are naturally divided into two groups. Our phylogenetic analysis of mtDNA supported this supposition (Říčan et al. 2022; Fig. 6), with the Northern group characterised by an ornamented dorsal fin and a mid-lateral stripe generally running to the dorsal margin of the caudal peduncle, while the Southern group was characterised by an unpatterned (hyaline, i.e. without any markings) dorsal fin and a mid-lateral stripe that always ran toward the posterior insertion of the dorsal fin (i.e. more distinctly dorsal). The two groups were almost entirely separated by lower lip colouration, with most Northern group species having a distinctly azure blue lower lip. In contrast, Southern group species always had upper and lower lips of the same colour (i.e. never blue). Consequently, colouration patterns alone were sufficient to easily delimit and diagnose both valid and putative species in the Northern group as these displayed many forms of patterning on the dorsal fin and other body areas such as the lower head, i.e. the snout and cheek (Říčan et al. 2022). For the valid species, a detailed description of colouration patterns can be found in Kullander (1986). Taxonomic descriptions of the undescribed putative species referenced in this study are being prepared.
Southern group species are harder to delimit and diagnose solely by the colouration patterns as they are mainly limited to the cheek area; nevertheless, most species in the group can be diagnosed by the colouration patterns (Kullander 1986). For example, B. apoparuana (named based on this character state), B. labiosa and B. megalospilus all lack any (or most) trace of a suborbital stripe or blotch; B. robusta, B. eurhinus, B. tambopatae, B. oenolaemus and B. vittata all have a faint but complete suborbital stripe; while B. hophrys has an incomplete stripe in the form of a suborbital blotch in the posteroventral position (B. hophrys was named based on another unique colouration pattern on its head; Kullander 1986). This variation in suborbital stripe development, together with variation in opalescent cheek colouration, is sufficient to diagnose and delimit all species in the Southern group, despite the lack of dorsal fin colouration.
Based on molecular clock dating (Fig. 6; Říčan et al. 2022), the two main clades of Bujurquina appear to have diverged around 16.6 Ma; hence, the diagnostic colouration patterns of the two main species groups are very old and thus unlikely to correlate with present environmental conditions. Separation probably occurred in the region between the Upper Ucayali river basin (the northern terminus of the Southern group) and the adjacent Huallaga river basin and the Lower Ucayali river basin (the southern terminus of the Northern group), making this the oldest area of divergence within Bujurquina. This same area is also where the divergence from its sister genus, Tahuantinsuyoa, appears to have occurred, its two species being found in the foothills of the Pachitea river basin, the major tributary basin of the Upper Ucayali.
Value of purported diagnostic characters for Bujurquina species delimitation
In the first descriptions and review of the valid Bujurquina species, Kullander (1986) mainly used colouration patterns and proportional measurements to distinguish, delimit and diagnose each species. However, this seminal publication (or any publication since) did not provide any statistical analysis demonstrating the effectiveness of these factors as diagnostic indicators. Here in this study, we have reviewed and tested these data types and demonstrated their very high delimitation and diagnostic value.
The colouration pattern data enable easy delimitation and diagnosis all the species in the Northern group (= dorsal fin ornamented), and while in the Southern group species (= dorsal fin unpatterned), colouration pattern variation is more limited, nevertheless, most species in this group can be diagnosed by the colouration patterns as initially suggested by Kullander (1986, Říčan et al. 2022).
The morphometric characters analysed through PCA have also been found of very high diagnostic value, and these data indeed enable the delimitation and diagnosis of Bujurquina species. The main message from our results is that the morphometric body-shape characters should not be used to diagnose Bujurquina species in its global context but only in a basin-by-basin context. When divided by main river basins, only two valid species pairs are not diagnosable solely by morphometric characters (B. ortegai vs. B. huallagae, and B. apoparuana vs. B. hophrys). Most of the newly identified putative new species also show strong separation in morphometric data in four out of five river basin faunas (Fig. 1, Figs. 7-11). The only exception is the Marañón basin (Fig. 10), with poor morphometric separation of three putative species plus B. zamorensis.
Foothill and lowland Bujurquina diversity and corresponding morphological divergence
As our analyses have confirmed a separate band of endemism in the Andean foothills, with a distinct fauna from the supposedly more widely distributed species in the Amazonian lowlands, we can confirm Bujurquina endemism in two major biogeographic zones. Furthermore, these two bands are discontinuous since each major river basin (Madre de Dios, upper Ucayali, central western Amazon-Huallaga, Marañón, Napo-Putumayo) has an endemic set of upland and lowland species (Figs. 7-11). That the two zones appear to host separate Bujurquina fauna was first proposed by Kullander (1986) for Peru, where it was noticed that the lowland species were usually missing from the adjacent foothills and vice versa (Kullander 1986). Thus, our results support and extend the validity of this major biogeographic pattern throughout the Bujurquina distribution area, from the Paraguay basin in the south to the Napo-Putumayo basins in the north. This same pattern most likely continues well into Colombia based on the undescribed species known from there that were included in the molecular phylogeny of the present study (though not assessed in the morphological analyses).
Our study also confirmed a strong correspondence between foothill/lowland distribution and lotic/lentic ecomorphs. The only species for which this predictor was more complex were B. robusta, which was found in a lowland area within a foothill band (Fig. 8), and B. oenolaemus, B. sp. Oran and B. sp. Ecuador 10 (Figs. 9, 11), the latter two are species from terra firme islands within a lowland area, while B. oenolaemus is a species from the elevated divisor area between the Amazon and Paraguay river basins. The results for B. pardus are difficult to interpret as the species and its habitats remain little known and require further study.
The incomplete correspondence between foothill/lowland distribution and lotic/lentic ecomorph appears to be mainly due to the complexity of the transition zone between highlands and lowlands (B. robusta, B. oenolaemus) and contrasting habitats within the lowlands, e.g. floodplain vs. terra firme areas. These terra firme areas appear to exert similar morpho-ecological pressures on Bujurquina populations as in foothill areas, i.e. a generally higher and less stable water velocity, low water turbidity, higher oxygenation levels, variable but generally low insolation and high habitat complexity (Crampton 2011). The pockets of lowland type habitat within foothill areas, on the other hand, approximate pressures found in the extensive lowland floodplains, e.g. low water velocity, high water turbidity, higher water temperatures, low oxygen levels, more constant insolation and less habitat complexity (for a review, see Crampton 2011).
The repeated evolution of character states and character state complexes in the form of ecologically adapted parallel ecomorphologies, as found in Bujurquina, is well known in cichlids (Burress 2015), both from lacustrine (where parallel evolutions were pioneered, e.g. Kocher et al. 1993, Schliewen et al. 2001, Cooper et el. 2010, Hulsey et al. 2013, Elmer et al. 2014, Kusche et al. 2014, Seehausen & Wagner 2014, Machado-Schiaffino et al. 2015) and riverine environments (Říčan et al. 2016, Burress et al. 2018, 2022). In cichlids, this parallel evolution of ecomorphologies generally develops along common environmental gradients; in lakes, this is typically along a benthic-to-pelagic axis (i.e. body shape) and the trophic axis (i.e. head shape, e.g. hard-shelled to soft-bodied prey axis or herbivory axis, e.g. Kocher et al. 1993, Schliewen et al. 2001, Machado-Schiaffino et al. 2015). In rivers, however, the adaptive landscape differs greatly from lakes (Burress 2015, Burress et al. 2018, 2022), but nevertheless both repeated and parallel adaptive radiations along the same axes have been documented in both Middle American cichlids and South American cichlids (Piálek et al. 2012, 2019a, b, López-Fernández et al. 2013, Říčan et al. 2016, Burress et al. 2018, 2022). In this study, we have documented the repeated evolution of Bujurquina ecomorphologies along the body shape axis, caused by evolutionary forces from the most common environmental gradient present in rivers, i.e. lowlands vs. uplands. Along the trophic axis (i.e. head shape), most Bujurquina species appear to be more generalist, with few adaptive departures (e.g. B. oenolaemus or B. sp. Oran); however, further studies will be needed to fully identify trophic evolution in the genus.
Terra firme islands and floodplains and Bujurquina diversity in lowland Amazonia
Bujurquina sp. Oran and B. sp. Ecuador 10 (= B. sp. Lower Curaray) are the first Bujurquina species specifically recognised from terra firme islands in lowland Amazonia (Říčan et al. 2022). However, these are not the only species known from terra firme islands, with several others having been identified from lowland areas of Ecuadorian and north Peruvian Amazonia, including B. peregrinabunda and B. moriorum from the same general area as B. sp. Oran, all three being separated by the floodplain species B. syspilus.
Based on the recent data presented here and in parallel studies, it would appear that the interplay between floodplain and terra firme Bujurquina species in lowland Amazonia is the rule rather than an exception and hence has high predictive power. All western Amazonian lowland areas so far explored for Bujurquina diversity host a more widespread lowland species found in the inundable portions of main rivers and tributaries, along with localised endemics in streams of terra firme islands separated (and hence effectively isolated in the case of Bujurquina) by the inundable areas of the main rivers.
Lowland Amazonian cichlid species, including those of Bujurquina, have been hypothesised to have larger distribution areas than the narrowly endemic foothill species (Kullander 1986); however, the available data do not allow confirmation of this generalisation as lowlands of the western Amazon are still poorly explored for Bujurquina species diversity. The high species diversity in Ecuadorian foothills newly recorded here supports narrow endemism in the foothills of the western Amazon also outside of Peru, where the hypothesis was initially formulated (Kullander 1986). Our knowledge of Bujurquina species diversity in the western Amazonian lowlands is further limited as large areas remain completely unexplored and, consequently, the distribution limits of the known species remain largely unknown; hence comparisons with foothill endemism are not yet possible. We have no data on the distribution limits of any lowland Bujurquina species in the western Amazon, not even for the megadiverse countries of Peru, Ecuador or Colombia. Nevertheless, it appears that some lowland species, e.g. the floodplain species B. syspilus or B. sp. Ecuador 13 (= B. sp. Lower Pastaza-Morona), do have much larger distribution areas than any of the foothill species (Kullander 1986; Figs. 1-2), and this was probably what initiated the original formulation of the hypothesis. Moreover, while the diversity of terra firme Bujurquina species appears to be higher than that of floodplain species, we do not know the sizes and limits of individual terra firme species distribution areas in most cases. Indeed, a similar situation applies to Amazonian freshwater fishes in general (Crampton 2011, Oberdorff et al. 2019).
Acknowledgements
We thank Hernán Ortega (Universidad Nacional Mayor de San Marcos, Lima, Perú) and Javier Maldonado Ocampo (Universidad Javeriana, Bogota, Colombia) for their help during our studies on Bujurquina. We also thank Zuzana Musilová for her photographs of adult specimens of the two species. This study was partially supported by a DCG (Deutsche Cichliden-Gesellschaft) grant to O. Říčan.
This is an open access article under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits use, distribution and reproduction in any medium provided the original work is properly cited.
Author Contributions
Conception, study design and sample collection, O. Říčan, Š. Říčanová, C.E. Rodriguez Haro; statistical analysis, O. Říčan, A. Pangrácová; review and editing, O. Říčan, A. Pangrácová, Š. Říčanová, C.E. Rodriguez Haro. All authors have read and agreed to the final version of the manuscript.
Literature
Appendices
Supplementary online material
Fig. S1. Distribution of valid species and the two main groups of Bujurquina, along with four of the putative species identified, on a map identifying i) the division between highland and lowland areas of the western Amazon (shown by the grey line), and ii) the main river types in the Amazon. Compare with Fig. 1.
Fig. S2. Examples of highland and lowland Bujurquina habitats.
Table S1. Specimens included in this study.
Table S2. Table of proportional measurements (% standard length) for the described Bujurquina species. While the table includes specimens > 40 mm SL, PCA analyses (Figs. 7-11) are based on specimens > 60 mm SL. SD = standard deviation. Data for B. pardus is from Arbour et al. (2014) and B. cordemadi from Kullander (1986). Colours aid in visualising the delimitation of valid species, with B. sp. Oran taken as the reference species. Values in orange are smaller than those for B. sp. Oran and show no overlap, while values in green are larger than those for B. sp. Oran and show no overlap. Consequently, values in orange and green are diagnostic between B. sp. Oran and the described species, values in grey are not strictly diagnostic, those in dark grey indicate a tendency toward separation and those in light grey indicate no separation.
( https://www.ivb.cz/wp-content/uploads/JVB-vol.-72-2023-Rican-et-al.-Figs.-S1-S2-Tables-S1-S2-1.pdf)
