This study investigates the morphology, material, and mechanical properties of radular teeth in six lineages of Chondrinidae, a family of terrestrial gastropods. We examined three species that forage on biofilms from mixed substrates (Abida secale, Granopupa granum, Granaria frumentum) and three that specialised in scraping lichen off rock faces (Chondrina arcadica, Rupestrella rhodia, Solatopupa similis). Using scanning electron microscopy (SEM), nanoindentation, and energy-dispersive X-ray spectroscopy, we analysed tooth morphology, hardness H, Young’s modulus E, and elemental composition. Our findings confirm that the central radular region reflects substrate preferences, with rock-scraping species exhibiting reduced tooth denticles and significantly higher E and H values compared to mixed substrate feeders. Correlations between mechanical properties and elemental composition (increased calcium or silicon content) suggest convergently evolved strengthening mechanisms in rock-scraping species. The results support hypotheses of parallel evolution in radular adaptations, highlighting Chondrinidae as a valuable model for studying the interplay of morphology, function, ecology, and evolution.

Energy dispersive X-ray spectroscopy

To analyse the elemental composition of the teeth using energy-dispersive X-ray spectroscopy (EDX, EDS), the same radulae were used that were previously documented in SEM. In total, 1495 small areas (1 x 1 µm) of three different radular regions (central, lateral, and marginal regions; see Supplementary Figure 2 for radular regions) were successfully analysed (for the exact quantity of measurements per region and species, see Supplementary Table 2).

Following established procedure [37,51], radulae were attached to glass slides with double-sided adhesive carbon tape, commonly used in SEM analysis. The adhesive was used because it appeared not to infiltrate the teeth. Each radula was surrounded by a small metallic ring, which was filled with epoxy resin (Reckli Epoxy WST, RECKLI GmbH, Herne, Germany) to fully encase the radula. After the resin polymerized for three days at room temperature, we removed the glass slides and adhesive tape. The samples were then polished using sandpapers of varying roughness until the cross-sections of the teeth became visible, followed by further smoothing with a 0.3 μm aluminium oxide polishing powder on a polishing machine (Minitech 233/333, PRESI GmbH, Hagen, Germany) to achieve a uniformly smooth surface. This embedding and smoothing process was essential to ensure a flat sample surface, minimizing electron scattering during subsequent EDX analysis.

The embedded samples were cleaned in an ultrasonic bath for five minutes, mounted on SEM sample holders, and sputter-coated with a 5 nm layer of platinum. Elemental composition analysis was carried out using the SEM Zeiss LEO 1525, equipped with an Octane Silicon Drift Detector (SDD) (microanalysis system TEAM, EDAX Inc., Mahwah, USA). Consistent settings were applied across all tests (e.g., 20 kV acceleration voltage, same working distance, and lens opening). The detector was calibrated using copper prior to analysis.

We performed point analyses (without mapping) on small areas on the cross-section of cusps (for region see Figure 1A) to identify the presence of various elements, including Al (aluminium), C (carbon), Ca (calcium), Cl (chlorine), Cu (copper), F (fluorine), Fe (iron), K (potassium), Mg (magnesium), N (nitrogen), Na (sodium), O (oxygen), P (phosphorus), Pt (platinum), S (sulphur), Si (silicon), and Zn (zinc) (for representative EDX spectra, see Supplementary Figures 3 and 4). Elements like C, N, Na, and O were not discussed in detail as they are integral to chitin and proteins, while Al and O could come from the polishing powder. Additionally, C, N, and O are light elements; they cannot be reliably quantified. Quantitative analyses and statistical evaluations were restricted to heavier elements, which can be robustly measured with this technique.

For quality control, we conducted 10 additional EDX tests on the epoxy to check for contamination from mechanical application, embedding, or polishing. No Si (from the sandpaper), Al (from the polishing powder), or any of the discussed elements (Ca, Cl, Cu, Fe, K, Mg, P, S, Si, Zn) associated with the resin composition were detected, confirming that the elements found in the teeth were not due to contamination.

Due to the overlap of phosphorus (P) and platinum (Pt) peaks, the software could not distinguish between these two elements reliably. Therefore, P content was discussed together with Pt (P+Pt). To estimate the proportion of P in the teeth, we measured the Pt content in 20 areas of pure epoxy, yielding a mean value of 0.14 ± 0.02 atomic %.

The samples were then used for nanoindentation of the teeth.

Nanoindentation

For estimating the mechanical properties of the cusps (for the region see Figure 1A), nanoindentation was carried out on the embedded and polished samples previously analysed by EDX. We followed previous protocols [51,57]. A nanoindenter SA2 (MTS Nano Instruments, Oak Ridge, USA), equipped with a Berkovich indenter tip and a dynamic contact module (DCM) head, was used for the measurements. Both effective hardness (H) and Young’s modulus (E) were determined from force-distance curves obtained using the continuous stiffness mode (CSM). E was calculated using the Oliver–Pharr method [68]. The Poisson’s ratio was set to 0.3, and the allowable drift rate was set to 0.1 nm/s. All tests were conducted under normal room conditions (relative humidity 28–30%, temperature 22–24 °C), with each indent and corresponding curve manually monitored. Due to the Pt sputter coating, effective E and H were determined at penetration depths ranging from 600 to 800 nm. The metallic coatings may influence nanoindentation if the coating thickness is comparable to the indentation depth. In our case, the sputter coating was ~5 nm thick, while indentation depths were 600–800 nm (>100× thicker). Given that the Oliver-Pharr analysis relies on contact depth at these scales, the thin Pt layer should have had no measurable influence on hardness or modulus. Approximately 40 values were obtained at this indentation depth for each site, and the average was used to calculate one mean H and one mean E value per indent. In total, nanoindentation was successfully performed on 1495 cross-sections of the cusps.

Statistical analyses

All statistical analyses were performed using JMP Pro, Version 14 (SAS Institute Inc., Cary, USA, 1989–2007). The elemental composition data obtained via EDX were recognized as compositional, given that the measurements represent proportions that sum to a constant total (100%). To address the inherent closure constraint and avoid spurious correlations associated with standard statistical methods, we processed the data using compositional data analysis. This was not done for the mechanical property data, Young’s modulus, and hardness. For the elemental data, we replaced components with zero values using a small imputation value (0.0001) to allow for log-ratio transformation. Then, we calculated the geometric mean g(x) of all components of each measurement and applied the centred log-ratio (clr) transformation: clr(x_i)=ln(x_i/g(x)). The conducted statistical analyses were performed with the clr-transformed data. The Shapiro-Wilk test was used to evaluate normality. As the data were non-normally distributed, a Kruskal-Wallis test was applied. Pairwise comparisons were then conducted using the Wilcoxon method. Additionally, correlation coefficients and relationships between parameters were determined using JMP.