Catfishes of the order Siluriformes are an exceptionally diverse group occupying significant fraction of the world's freshwater ichthyofauna. Catfish species renders a model system for phylogenetic and evolutionary studies. However, there is dearth of knowledge on the phylogenetic history with reference to morphological diversity of catfishes in the Ganga River. Geometric morphometrics is an efficient approach to quantify shape variations and has lately found application in range of taxonomic, phylogenetic and evolutionary studies. This study utilizes photographs of 39 Gangetic catfishes to evaluate inter-specific phenotypic relationship through geometric morphometrics. PERMANOVA revealed significant relationship (F = 4.50, p < 0.0001) between shape of all the species. Procrustes ANOVA revealed significant relationship between shape and size (F = 17.67, p < 0.0001) and between shape (F = 34.95, p < 0.0001) of all species. The first two principal components of PCA accounted for 69.41% (PC1 44.41% and PC2 25.00%) of the total variance while the first two canonical variates of CVA accounted for 55.89% (CV1 34.06% and CV2 21.83%) of the total variance. The CVA-extracted Mahalanobis and Procrustes distances were found to be highly significant (p < 0.0001) between distinct groups and non-significant (p > 0.0001) between closely related groups. Morphospace phenogram depicted that the species under Mystus, Sperata, Glyptothorax, Ompok, Eutropiichthys, Bagarius and Gagata clustered together in their respective genus, indicating that species of respective genus share common phenotypic traits and have evolutionary relatedness. Species like Chaca chaca, Clarias magur, Sisor rabdophorus, Heteropneustes fossilis, Silonia silondia and Clupisoma garua showed clear distinctness, which may be attributed to different set of morphological traits. The study affirms the implication of geometric morphometrics in discriminating Gangetic catfishes and showcasing phenotypic relatedness between respective groups. This study has high implications in understanding the ecological and evolutionary problems in relation to morphological diversity in catfishes.

Study area

The Ganga River is among the largest rivers in Asia, originating from icy-cave of ‘Gaumukh’ (30055’N and 7007‘E) of Gangotri Glacier in the Garhwal Himalaya at an altitude of 4,100 m. The river flows through five states of India, viz. Uttarakhand, Uttar Pradesh, Bihar, Jharkhand and West Bengal covering a total length of about 2,550 km before merging into the Bay of Bengal. Owing to its high cultural and religious value, the Ganga River was declared as “National River” of India in 2008 (Sanghi 2014). Based on topography, geography and ecology, the river is divided into six sections for monitoring aqualife (Dwivedi et al. 2019). The section wise topography of the river is listed in Table 1.

Sampling

Fish samples were collected from 14 sites in the Ganga River and its head tributary, Bhagirathi by gill nets (30 × 3.0 m) with stretched mesh size of 19–152 mm in one year duration from September, 2018 to August, 2019 (Figure 1). Identifications were done through key characters following established taxonomic literature (Hamilton 1822; Day 1878; Talwar and Jhingran 1991; Das et al. 2010; Jayaram 2009, 2010). In all, 134 specimens belonging to 39 species of the order Siluriformes were identified (Table 2). Valid nomenclature for all the species was adopted as per the Catalogue of Fishes of the California Academy of Sciences (Fricke et al. 2021). As the preserved specimens affect the geometric morphometric analyses (Fruciano et al. 2020), only fresh samples were used in this study. High resolution images were captured using a digital hand-held camera (Cyber Shot DSC-W300; Sony, Tokyo, Japan) by placing specimens laterally on a flat surface with body posture and fins teased into a natural position to ensure that the body was not bent.

Landmark based Geometric Morphometric data Generation

JPEG images of fishes were converted to the required tps format using tpsUtil (ver. 1.40, F. J. Rohlf, see http://life.bio.sunysb.edu/ee/rohlf/software.html, accessed 10 December 2019). Digitization of 14 homologous landmarks for each specimen were done through tpsDig software (ver. 2.16, F. J. Rohlf, see http://life.bio.sunysb.edu/ee/rohlf/software.html, accessed 10 December 2019; Figure 2). Only the landmarks that best describe the geometry of the body form and homologous in nature were selected for the purpose (Dwivedi et al. 2020). Hence, adipose fin was not considered to digitize the landmarks as it was absent in some catfishes. ImageJ software was used to efficiently extract 28 two–dimensional x,y coordinate data from 14 landmarks of all the specimens (ver. 1.50i, see http://imagej.nih.gov/ij/, accessed 12 December 2019).

Statistical Analysis

Procrustes superimposition was employed to standardize shape coordinate data of each specimen to unit centroid size eliminating the effect of overall body size (Rohlf and Slice 1990; Bookstein 1991). Procrustes ANOVA on size was performed to determine its effect on the shape for all the species. Permutation Multivariate Analysis of Variance (PERMANOVA) was further performed to determine the significance in overall shape difference between species using permutation test by applying “Euclidean” similarity index. Partial least squares (PLS) analysis was done to analyze the pattern of covariation between shape and size (Bookstein et al. 2003; Zelditch et al. 2004). The shape variation among species was analyzed through Principal Component Analysis (PCA) and Canonical Variate Analysis (CVA). PCA helps in reduction of geometric morphometric coordinate data (Veasey et al. 2001), and in detecting outline groups of samples (Johnson and Wichern 1998). CVA extracts the set of axes that potentially differentiate between two or more groups (Mardia et al. 1979). Cluster Analysis (CA) was applied to extract morphospace phenogram from the squared Mahalanobis distance, and group centroid PCA and CVA scores for robust interpretation of similarity–dissimilarity in shape between species (Maddison et al. 1997). Statistical analyses were performed using the software packages MorphoJ version 1.06d (Klingenberg 2011) and Past version 3.23 (Hammer et al. 2001) were used for all statistical analyses.