Examined material

A total of 127 females and 114 males of wild Synodontis spp. (S. irsacae, S. melanostictus, S. multipunctatus, S. petricola and S. polli) were included in the present study for histological examination (Table 1). The original dataset included three S. irsacae males, one S. melansotictus female and one S. multipunctatus female, which were unavailable for histological examination due to technical issues, reducing the total sample of wild fish to 241. Wild fish were caught using a hand net during scuba diving or collected in minnow traps in the southern part of Lake Tanganyika, Zambia (9 km perimeter, 8°43′47.2″ S 31°09′24.6″ E; collection permit: K-4335/18 KA/K.48/18). The sampling was conducted in two different seasons: the dry season (October 2021) and the rainy season (March 2022), and the samples were distributed equally among both seasons. Fish examined in the present study were the subsample of the largest caught individuals of each sex-species-season combination from the total sample of wild-caught fish. Wild fish material was supplemented with three male S. multipunctatus (90-100 mm TL) maintained from hatching up to the age of four years in the breeding facility of the Institute of Vertebrate Biology, CAS Brno (accreditation 52699/2018-MZE-17214). The number of individuals examined per species and year of collection are summarized in Table 1.

The sex of each fish was first visually determined from the size and shape of the urogenital papilla (Cavaco et al. 1998; Fig. 1) upon collection by an expert (R. Blažek) and then confirmed by histological examination. The concordance between the visual inspection and histological assessment was tested by a generalised linear model (GLM) with binomial error distribution with a logit-link function. The agreement between visual inspection and histological assessment (1-concordance, 0-discordance) was fitted as the response variable and species (4-level factor), sex (male or female as determined by histology, 2-level factor) and body length (TL: from snout to the tip of the caudal fin, continuous) as predictors. Tukey post-hoc comparisons were performed with emmeans v 1.5.4 (Lenth 2021). The full model contained all possible pairwise interactions (sex:species; sex:TL; species:TL), but all were subsequently dropped as they were not significant (P > 0.31). The total sample size was 242 because all four S. melanostictus were excluded from this analysis due to the low sample size, and it was possible to determine sex (but no other histological information) from slides with technical issues.

Table 2.

The results of a Generalized Linear Model with binomial error distribution for the precision of visual sex determination based on the shape of urogenital papillae. The visually determined sex was compared to gonad histology. Synodontis polli is set as a reference level to which other coefficients are compared. SE is the standard error of the mean. Values are on a logit scale.

Histology

All wild-caught fish were euthanised in clove oil and fixed in 10% neutral buffered formalin, with the belly opened ventrally to facilitate fixative penetration to internal organs. In the laboratory, fish were dissected, gonads of males and females were separated from other viscera and weighed to the nearest 0.01 g. The left and right ovaries were weighed separately to assess potential differences in mass. Their mass was compared by fitting a Linear Mixed Effect Model (LMME) from the lme4 package v 1.1.26 (Bates et al. 2015). Individual ovary mass was the response variable, and ovary side (left or right, factor) and its interaction with species (4-level factor) were included as predictors. Individual fish ID was included as a random factor to correct for the dependency of measurements within each individual. All statistical analyses were performed in the R environment v 4.2.2 (R Core Team 2022).

Whole testes and right ovary were then submerged into 70% ethanol and processed by the standard paraffin technique using Histoplast (Sigma) as an embedding material. The standard fixative procedure for fresh organs removed from euthanized fish was appliedtothreefishfromthebreedingfacility.Multiple (60-80) semi-serial sections were examined from each block. The oogenesis and spermiogenesis staging followed widely accepted criteria of Grier (1981) and Grier et al. (2009, 2018) related to the terminology of reproductive phases (as standardized by Brown-Peterson et al. 2011). Tissue samples of Synodontis gonads were subjected to a two-step examination, a non-targeted screening of the whole assemblage of Synodontis species aimed at identification of common and contrasting structural features and a second step focused on species-specific features.

Precision of the visual sex determination

The visual identification of sex by inspection of their urogenital papilla was less precise in females (GLM, back-transformed model estimated mean of error rate 6.5%, (SE = 2.6%), z = 2.83, P = 0.004, n = 242) than in males (0.8%, SE = 0.6%), and in smaller individuals (Table 2). The lower precision of sex determination in females may be ascribed to the shape similarity of papillae between subadults and females. At the species level, S. polli was the most difficult species for identification of its sex upon visual examination of the urogenital papilla (back-transformed model estimated mean of error rate: 22.8% (SE = 11.8%)) and tended to be significant from other species (Tukey post hoc pairwise comparisons from binomial GLM, all z > 2.36, P = 0.005-0.08). All other species (S. irsacae, S. multipunctatus, S. petricola) had an error rate < 2.5%, which did not differ significantly among each other (all P = 0.207-0.882, Table 1). The lower success of visual sex identification in S. polli comes from less developed inter-sexual differences in the shape of urogenital papillae in small adults of this species. The results are in accordance with our personal experience, where sex determination from the shape of the urogenital papilla is relatively easy in S. multipunctatus and S. petricola. In contrast, in S. irsacae, it is difficult (but still feasible), and it appears not to be a useful character for sex determination in S. polli, except in large (> 150 mm) individuals.

Maturation and reproductive state

Of 241 histologically examined individuals, 199 contained mature gametes in their gonads. The other 38 were likely adults but outside their reproductive state, as their gonads contained multiple developmental stages of gametes, or they were also considerably (> 5 mm) larger than the smallest individuals with the presence of mature gametes in gonads. For example, there was one S. polli male with the exclusive presence of spermatogonia in the testes, but it was likely an adult outside the reproductive phase due to its relatively large size (TL = 115 mm). The remaining four individuals of Synodontis spp. were likely subadults; they were small (at the lower range of total size of examined fish) and had small gonads with exclusive presence of the earliest developmental stage of gametes (oogonia: 1 × S. multipunctatus, TL = 67 mm; 1 × S. irsacae, TL = 82 mm; spermatogonia: 1 × S. irsacae, TL = 78 mm; 1 × S. polli, TL = 98 mm).

Gross morphology of ovaries

Data on the shape, size, and colour of organs from dissected fresh material (commonly used for a simple gross evaluation of gonad developmental stages) were not available for an adequate number of individuals. Therefore, our description is based on detailed observations of organs preserved in formaldehyde solution.

The ovaries, which are not always of equal shape, are a ventrally positioned paired organ with a slightly compressed left side to fit within the body cavity with the gut (Fig. 2a, b). The difference between the left and right ovaries is only in their shapes since their mass did not differ (LME, χ² ₁ = 1.01, P = 0.316) and was consistent among species; the interaction term between ovary side and species was not significant (χ² ₃ = 1.16, P = 0.763). In all studied species, the ovaries were sacculiform, narrowed caudally and joined at their posterior end to form a single common oviduct terminating at the urogenital pore. Most ovaries included oocytes at variable maturity stages in a similar ratio, except S. irsacae, where late vitellogenic stages of oocytes (when present) were the dominant maturity stage. In such cases, middle oocyte maturity stages were in the minority, and oogonia and primary oocytes were primarily the other oocyte stages present. The dominance of late vitellogenic stages in S. irsacae was recorded in the rainy season of 2022 (13/14) but not in the dry season of 2021 (0/13). This observation indicated that, in contrast to the other species examined, S. irsacae has seasonal reproduction.

Gross morphology of testes

The testes were observed as a flat, bilaterally symmetrical organ with uneven lateral outlines. Separated from surrounding connective tissue, the lobate shape of the testes was seen in a large flat anterior part (Fig. 2c, d), whereas a caudal part resembled catfish seminal vesicles documented macroscopically in catfishes from the family Clariidae (Van den Hurk et al. 1987) and Auchenipteridae (Meisner et al. 2000).

Light microscopy of ovaries

The ovaries of the study species shared features standard to the ovaries of most teleost fishes. They were covered with connective tissue, walls subdivided by ovigerous lamellae/folds, in which oogenesis started with the proliferation of oogonia/oogonia stem cells. Oogonia were round; their cytoplasm was pale, and a large diameter nucleus contained a prominent nucleolus (Fig. 3a, b). Formed by the mitotic division of much smaller germ cells (difficult to demonstrate in histological sections), oogonia were grouped in nests located under the wall of ovaries and in ovigerous lamellae protruding into deep layers of this organ. Oogonia differed substantially from the next stage of oogenesis, i.e. from primary oocytes with strongly basophilic cytoplasm (Fig. 3b-e). This stage transformed into early oocytes characterized by an increased volume of cytoplasm with Balbiani bodies (e.g. Schisa 2012) and round nuclei that contained nucleoli on their periphery (Fig. 3c, d). Early oocytes were covered by a delicate structure comprising a simple layer of squamous follicle cells. Oogonia and primary oocytes belonging to the previtellogenic stages of ovarian development were usually found together with advanced, i.e. vitellogenic, stages. Unlike in previtellogenic stages, the cytoplasm of early vitellogenic oocytes was pale and contained numerous tiny yolk droplets densely accumulated around the nucleus, whereas the peripheral part of the cytoplasm contained numerous vacuoles, also called vesicles or cortical alveoli (Fig. 4a-d). In advanced vitellogenesis, the oocytes are enlarged due to the progressive accumulation of yolk substances and the growth of yolk droplets/globules. The oocyte membrane differentiated to form a well-visible layer called zona radiata (a term used in original papers and fish histology textbooks due to its radial striations described at the subcellular level: Chaudhry 1956, Anderson 1967, Flügel 1967, Matsuyama et al. 1991). Note that zona radiata is termed zona pellucida in mammalian histology textbooks. The follicle cells surrounding the zona radiata, which were initially flat/squamous, became cuboid, contained prominent nuclei and formed continuous oocyte envelope (Fig. 4d, e). In the course of oocyte late vitellogenesis, follicle cells increased in size, became cylindrical and underwent conspicuous changes manifested by the presence of multiple vacuoles in their cytoplasm and the loss of staining properties (Fig. 5). The comparison of Synodontis vitellogenic oocytes revealed a direct correlation between the stage of vitellogenesis as manifested by the size of yolk globules and the morphology of follicular layer. In oocytes with the largest yolk globules, the most advanced changes in the follicular layer were seen in terms of increasing cell height, increased vacuolization of cell cytoplasm, and loss of cell stainability (Fig. 5). Similar changes of follicular layer were documented in Neotropical catfishes (Melo et al. 2011) which linked follicle cell characteristics to their reproductive strategies (pelagic spawners, demersal spawners, and internally fertilizing species). Unfortunately, no data were available about the exact synchronisation of the vitellogenic stages used for comparison with histometric data collected on follicular cells.

The post-ovulatory fate of follicular cells has rarely been addressed. Shelton (1978) summarized varieties of development and transformation of follicular cells in fishes and described in detail the contribution of follicular cells persisting after ovulation to egg adhesiveness in the clupeid fish Dorosoma petenense. A comparison of the most developed ovarian oocytes of S. multipunctatus collected in the wild with eggs of the same species produced in the laboratory during artificial reproduction by applying gentle pressure on the abdominal cavity (Blažek et al. 2018) revealed the absence of cellular follicular envelope in artificially spawned eggs. Under the light microscope, the zona radiata of artificially spawned eggs was covered with a layer of homogeneous acellular material (Fig. 6c, d) considered to be a product of follicular cells in most bony fishes (Shelton 1978). Similar advanced transformation of the cellular follicular envelope into a homogeneous acellular layer was observed in intraovarian oocytes of two S. polli individuals of this study (Fig. 6a, b).

Histological evaluation of the female reproductive development in Synodontis spp. identified previtellogenic stages of oogenesis in 14 out of 66 females collected in October 2021. This means that 21% of females were not in a reproductive phase (2/14 S. polli, 9/14 S. irsacae, 3/14 S. multipunctatus). The examination of females collected in March 2022 revealed exclusively the co-occurrence of early and late vitellogenetic oocyte stages; hence, all those females were developmentally mature and soon prepared to spawn.

The oocyte dysmorphism (Fig. 7) was found in all four Synodontis species (technical issues prevented examination of a single S. melanostictus) – in one female (S. polli) in 2021 (out of 66 examined females of all Synodontis spp.) and 13 (out of 59) females in the 2022 dataset. Overall, oocyte dysmorphism was identified in 4/27 S. irsacae, 2/34 S. multipunctatus, 3/30 S. petricola and 4/34 S. polli. This aberrant oocyte morphology linked to an atypical course of meiosis was classified as irreparable, i.e. meiotic resumption was not expected.

Ovarian regressive changes such as those affecting oviducts, the presence of so-called postovulatory follicles (reported, e.g. in siluriform fish by Sales et al. 2016) and atretic oocytes, were difficult to estimate accurately in Synodontis spp., due to low general cohesion of ovarian tissues and multiple artefacts arising from sectioning mature oocytes containing large yolk globules (Fig. 6e and Fig. 8).

Unusual ovarian cellularity, presumably representing an inflammatory response, was found in three individuals (S. polli, S. irsacae and S. multipunctatus) in association with cross-sections of unidentifiable nematodes localized in the abdominal cavity (Fig. 9).

In summary, the comparison of oogenesis structures revealed no significant differences among the study species. Based on the material examined, a single finding of an advanced degree of intraovarian follicular layer transformation cannot be considered species-specific.

Light microscopy of testes

The histology of male gonads was more difficult to investigate than that of females. Well-visible lobes of the anterior spermatogenic part of testes represented incompletely divided segments of left and right parts fused near the system of efferent ducts (Fig. 10). In all study species, testes structure corresponded to the unrestricted type sensu Grier (1981) and Uribe et al. (2014). Consequently, spermatogonia developed along the entire length of seminiferous tubules. The earliest stage of spermiogenesis, i.e. exclusive occurrence of germinal primordia with proliferating spermatogonia, was detected in only two individuals (one S. polli and one S. irsacae) (Fig. 11). In all other individuals, more advanced stages of the gonadal development with spermatogonia, spermatocytes, spermatids and spermatozoa in seminiferous tubules were found so that these fish could be considered maturing or mature (Figs. 12-14). A pronounced dominance of either early or advanced spermatogenic stages was observed in 45 of 114 examined males. Among those, spermatogonia, spermatocytes and spermatids dominated in five of 45 males, while spermatozoa dominated in 40 of 45 males. Testes of the 69 remaining individuals contained all developmental stages without an apparent dominance.

Pronounced differences were observed among particular lobes of testes (Fig. 15). In tubules of some lobes, spermatogonia plus spermatocytes and spermatids prevailed; other lobes had tubules filled with spermatozoa and rarely earlier spermatogenetic stages or with eosinophilic material considered testicular plasma sensu Mansour et al. (2004). Three individuals of S. multipunctatus reared under laboratory conditions for four years were additionally examined, and all showed signs of advanced stage of maturity. Like individuals from the wild, laboratory-reared individuals had asynchronous development of testicular lobes.

The anterior part of the efferent system of male gonads with efferent ductules of individual lobes collecting and emptying spermatozoa to bilateral main testicular/efferent ducts (Fig. 16) was relatively easy to follow. The posterior part of the efferent system with sperm duct was observed in a preserved condition only exceptionally (Fig. 10). The site of the testicular duct connection is missing in the documentation due to the intricate architecture of the posterior part of the testes. The structure of efferent ductules and duct walls is presented in Fig. 16.

Interpretation of the structure in the posterior part of the testes, referred to as seminal vesicles, accessory glandular organ or sperm duct glands in Siluriformes and many other teleosts (Miller 1992, Chowdhury & Joy 2007), was even more difficult than that of the anterior spermiogenic part. Adequate histological description was possible only for some of the diverse functions attributed to the fish seminal vesicles (Chowdhury & Joy 2007, Schulz & Nóbrega 2011, Uribe et al. 2014). In our Synodontis dataset, seminal vesicles principally served to store accumulated spermatozoa prior to their release for fertilization. Histological observation of the secretory activity of epithelial cells that participate in seminal vesicle fluid production was feasible but challenging due to the topography of this part of testes, spatial arrangement of vesicles (well seen in gross morphology of large individuals), and challenging removal of this fragile organ for a detailed study.

In most Synodontis males, multi-chambered vesicles with spermatogenic stages in various densities predominated in the posterior part of the testes (Figs. 17-22). Although frequently observed in fragments, they evidently functioned as seminal vesicles, i.e. repositories/storage places of most advanced spermatogenic stages. In our histological sections, whether these vesicles contained exclusively mature spermatozoa or also advanced spermatids could not be identified with certainty. In some vesicles, spermatozoa were mixed with material of pronounced eosinophily, also reported as seminal vesicle plasma, a supposed component of seminal vesicle fluid. Co-occurrence of vesicles containing spermatozoa and those also containing eosinophilic material was registered in the caudal part of the testes in 31 of 114 Synodontis spp. males under study (5/26 S. irsacae, 2/3 S. melanostictus, 4/25 S. multipunctatus, 11/30 S. petricola, 9/30 S. polli).

Posterior vesicles with the structural characteristics of an active secretory gland, similar to those reported in the comprehensive review by Chowdhury & Joy (2007) across ten teleost families, were found only in two individuals (S. irsacae and S. melanostictus). We, therefore, cannot confirm whether this is a typical structure in the examined species. Interconnected chambers of these vesicles were lined by simple cuboidal epithelium and contained acellular homogeneous acidophilic/eosinophilic material. A special situation – all vesicles of the posterior part of the testes being filled with seminal vesicle secretion/plasma – was found in one fish, S. melanostictus (Fig. 20).

In other instances, vesicles containing eosinophilic material (5/26 in S. irsacae, 1/3 in S. melanostictus, 15/30 in S. polli) differed in having small to large clusters of spermatozoa (Fig. 18a, b) within a predominating eosinophilic content. Flocculated or dense eosinophilic content was found only in several cases (Fig. 18c). Atrophy of epithelial lining responsible for seminal vesicle secretion was found in S. polli (Fig. 18d). Histological study can hardly determine composition of eosinophilic materials and relate their appearance in sections with results of detailed mucopolysaccharide-protein-lipid analyses reported in other fish species (Chowdhury & Joy 2007). Nevertheless, the density of seminal vesicle contents could be inferred from sections stained with Hematoxylin and Eosin (H & E). The liquid content of vesicles was washed out during the processing of tissue samples, leaving them empty (Fig. 22c), whereas dense and strongly eosinophilic material remained embedded in vesicles (Figs. 17, 18).

Solitary ducts filled with dense eosinophilic material were interposed between posterior seminal vesicles near the long axis of the testes and suggested to be part of the sperm duct gland. However, the site of connection with the sperm duct was not found.

In teleost fish, seminal vesicle variability and temporal changes in their morphology were always related to cycling changes (Chowdhury & Joy 2007). However, studies focused on experimentally induced changes are rare.

Comments on male gonads of individual Synodontis species