Evolutionary radiations are one plausible explanation for the rich biodiversity on Earth. Adaptive radiations are the most studied form of evolutionary radiations and ecological opportunity has been identified as one factor permitting them. Competition among individuals is supposedly highest in populations of conspecifics. Divergent modes of resource use might minimize trophic overlap, and thus intersexual competition, resulting in ecological character displacement between sexes. However, the role of intersexual differentiation in speciation processes is insufficiently studied. The few studies available suggest that sexual niche differentiation exists in adaptive radiations, but their role within the radiation, and the extent of differentiation within the organism itself, remains largely unexplored. Here, we test the hypothesis that multiple morphological structures are affected by sexual niche differentiation in “roundfin” Telmatherina, the first case where sexual niche differentiation was demonstrated in an adaptive fish radiation. We show that sexes of two of the three morphospecies differ in several structural components of the head, all of these are likely adaptive. Sexual differentiation is linked to the respective morphospecies-specific ecology and affects several axes of variation. Trait variation translates into different feeding modes, processing types and habitat usages that add to interspecific variation in all three morphospecies. Intrasexual selection, i.e. male-male competition, may contribute to variation in some of the traits, but appears unlikely in internal structures which are invisible for other individuals. We conclude that intersexual variation adds to the adaptive diversity of roundfins, and might play a key role in minimizing intersexual competition in emerging radiations.

µ-CT Imaging

The skulls of 13 specimens of each morphospecies T. antoniae “small”, T. antoniae “large” and T. prognatha were used for 3D µ-CT analyses. µ-CT scanning was performed with Skyscan 1272 and Skyscan 1173 scanners (Bruker). All specimens were scanned in 70 % Ethanol. Five male and five female specimens per species were stained with 0.3 % Phosphotungstic acid (PTA) in advance. Five male and five female specimens per species were scanned without any prior staining. The resolution ranged between 11 µm and 23 µm depending on the size of the specimen. Selected rotation steps varied between 0.2, 0.3 and 0.4 degrees over 180°. The chosen voltage ranged between 60 kV and 100 kV and the current between 80 µA and 166 µA. Detailed scanner settings for each individual can be viewed in Dryad. The projections were reconstructed with NRecon ver. 1.7.1.0 (Bruker). Data size was then reduced with the software Dataviewer ver. 1.5.2.4 by Bruker and ImageJ ver. 1.51f by NIH (Schindelin et al., 2015). Segmentation and volume rendering of the resulting 3D models was accomplished with the software packages Amira ver. 6.5.0 by Thermo Fisher Scientific (Stalling et al., 2005) and VG Studio 3.2 by Volume Graphics. Surface rendering was performed with the software packages Checkpoint ver. 17.04.21 (Stratovan Corporation) and Amira ver. 6.5.0 by Thermo Fisher Scientific (Stalling et al., 2005).

Classical and geometric morphometrics

To identify variation in the opercle bone and the pharyngeal jaw, linear morphometric measurements and geometric morphometric analyses were conducted based on surface rendered 3D models created by the software Checkpoint ver. 17.04.21 (Stratovan Corporation) out of µ-CT tiff image stacks. The following traits of the cranial skeleton were quantified by linear measurements: skull length, left opercle height, left opercle length, left opercle circumference, left opercle surface area, lower right pharyngeal jaw length, lower right pharyngeal jaw width, lower right pharyngeal jaw height and lower right pharyngeal jaw circumference. The number of teeth on the right lower pharyngeal jaw was counted. All measurements were carried out with the software Checkpoint.

In order to test on interspecific and intersexual shape differences, 14 landmarks were placed at homologous points on the pharyngeal jaws of the 30 unstained µ-CT scanned specimens. The outline shape and circumference of the left opercle and the pharyngeal jaw were analysed with 80 semi landmarks respectively. Patches were used to measure the surface area of the opercle in order to quantify its overall size between species and sexes.

Buccal cavity measurements

All classical morphometric measurements and geometric morphometric analyses of the buccal cavity were based on surface rendered 3D models created by Amira ver. 6.5.0 by Thermo Fisher Scientific (Stalling et al., 2005) out of µ-CT tiff image stacks. In order to quantify interspecific and intersexual variation in buccal cavity size and shape, the 30 stained specimens were used for creating volume rendered models of the cranial region with the software Amira. Surface rendered models of the buccal cavity were created with the help of the semi-automatic segmentation tool of Amira. Background artefacts were removed by applying the “remove islands” and “fill holes” options of Amira on the segmented 3D model. The length, width, height and volume of the buccal cavity were measured for every priory stained specimen. Two female specimens of T. prognatha were removed from the analysis because they showed deformations of the buccal cavity due to a slightly opened mouth.

NA = not applicable:

In files 1, 2, 4 and 5 the column "Morphospecies" for T. prognatha is empty since this species only contains one morphospecies.

In file 2 "Linear morphometric measurements" some data rows are missing because some specimens were scanned in an unstained and stained condition, while other specimens were only scanned in one of the two conditions. Specimens which were only scanned in an unstained condition were not used for the buccal cavity measurements. Specimens which were only scanned in a stained condition were not used for the opercle and pharyngeal jaw measurements.