Functional anisotropy of the elephant trunk skin: A biological blueprint for grasping, protection, and tactile sensing Authors: Matteo Lo Preti, Paule Dagenais, Behnam Kamare, Irene Bernardeschi, Simone Lantean, Michel Milinkovitch, Lucia Beccai Published: 2026 ABSTRACT AbstractMuscular hydrostats, like the elephant trunk, face a fundamental biomechanical conflict between robust protection and delicate manipulation. How the skin, a critical structural and sensory component, resolves this paradox is largely unknown. Here, we reveal that the Asian elephant's trunk skin is a functionally graded, anisotropic biocomposite. Through integrated morphological and mechanical analysis, we demonstrate a stark dorsal–ventral dichotomy: the dorsal skin is 3.14 times stiffer and significantly rougher, forming a protective shield, while the more flexible ventral skin is specialized for conformal grasping. This ventral side possesses unique dermal papillae, which our finite-element models predict act as subsurface stress amplifiers, enhancing tactile sensitivity without compromising durability. These findings establish a biological blueprint for artificial tactile skins involved in robotic grasping and manipulation tasks, demonstrating that a functionally zoned architecture with embedded microstructures is a key strategy for resolving the grasp-protect paradox in soft robotics. Elephants can use their trunks to be both powerful and remarkably gentle, but how their skin supports this versatility is a mystery. We reveal that the elephant's trunk skin is a “smart,” functionally zoned material. The top is a tough, armor-like shield, while the bottom is a flexible, sensitive surface for grasping. We discovered that the skin's unique subsurface structure, with finger-like papillae, mechanically amplifies the sense of touch, explaining its incredible sensitivity. These biological design principles provide a direct blueprint for creating the next generation of advanced robotic skins that can be both highly durable and exquisitely sensitive, resolving a key challenge in soft robotics. Introduction The elephant trunk represents a paradigm of biological engineering, a muscular hydrostat that has masterfully solved a fundamental biomechanical conflict, being capable of gently peeling a banana, yet powerful enough to transport heavy logs (1, 2). As the trunk lacks a rigid skeleton, its skin is not merely a passive covering but a critical component of its structure, providing protection, enabling motion, and mediating sensation. This unique combination of robustness, deformability, and sensitivity makes the elephant trunk skin a compelling model for advanced material design (3). Previous work has established that the trunk exhibits distinct dorsal–ventral asymmetries. The dorsal surface is characterized by deep folds, while the ventral surface features a network of finer wrinkles, and this structural difference is reflected in the mechanical properties of the skin (4, 5). The skin's layered composition is also known to be exceptionally thick compared with other mammals, and it is densely innervated with mechanoreceptors, particularly at the tip (3, 6, 7). However, a comprehensive, multiscale understanding of how the skin's architecture (from surface topography to its fibrous microstructure) governs its graded mechanical behavior remains lacking. It is not yet known how these disparate features work in concert to resolve the grasp-protect paradox. Here, we test the hypothesis that the Asian elephant trunk skin is a functionally graded, anisotropic biocomposite, structurally and mechanically zoned to satisfy the conflicting demands of manipulation and protection. We combine histology, plastination, second harmonic generation (SHG) microscopy and mechanical testing to build a comprehensive architectural and biomechanical profile of the trunk along its dorsal–ventral and proximal–distal axes. We then leverage this experimental data to develop a finite-element analysis (FEA) that reveals the mechanistic role of key microstructures. This integrated approach allows us to establish a biological blueprint for the design of the next generation of robust, sensitive soft robotic systems. Results Our integrated investigation, which combines high-resolution imaging, mechanical testing, and computational modeling, reveals that the Asian elephant trunk skin is a highly sophisticated, functionally zoned material. The following sections detail the evidence supporting this conclusion. We first establish the skin's multiscale, anisotropic architecture. We then demonstrate how this architecture creates a stark mechanical dichotomy tuned for protection and grasping. Finally, we use modeling to propose a plausible mechanism for the trunk's renowned tactile sensitivity. The trunk skin is a graded, anisotropic biocomposite across multiple scales The architecture of the Asian elephant trunk skin is not uniform; rather, it is a complex biocomposite with distinct properties that grade along both its proximal–distal and dorsal–ventral axes. Looking at the trunk in its relaxed position, at the macroscale the skin's morphology is different between the dorsal and ventral regions: the dorsal surface is characterized by macroscopic folding (outer curvature), whereas the ventral surface features microscopic wrinkling (inner curvature) suitable for conformal contact (Fig. 1A). This asymmetry extends to the histological level, which identifies the skin's constituent layers: the epidermis, comprising the stratum corneum and viable epidermis, and the dermis (or corium), which includes the papillary dermis (PD) and reticular dermis (RD). The dorsal PD forms prominent interlocking dermal ridges (DRs), a feature nearly absent on the ventral side. In contrast, the ventral skin exhibits large, distinct dermal papillae (DPs) that protrude into the epidermis, creating a unique subsurface topography (Fig. 1B). The multiscale anisotropic architecture of elephant trunk skin. A) Plastinated trunk cross-section (50 cm from tip) revealing the convex, folded dorsal morphology and the concave, wrinkled ventral morphology. B) Comparative histology of the dorsal skin (left), dominated by large DRs (arrows), and the ventral skin (right), characterized by prominent DPs (arrows). C) Quantitative analysis of structural gradients, showing changes in corium thickness and DP/DR dimensions along the proximal–distal axis. D) SHG microscopy revealing the underlying collagen fiber organization, with a circumferential alignment perpendicular to the surface (top) and a cross-hatched arrangement parallel to it (bottom). TS, tendinous sheath. These architectural features vary systematically along the length of the trunk. The thickness of corium increases significantly from the tip (mean: 2 mm) to the proximal base (mean: 7 mm, P < 0.001), while the ventral DPs decrease in height by ∼67% toward the proximal region (Fig. 1C). Further underlying this structural complexity is the organization of collagen I fibers within the RD. Using SHG microscopy, we found that parallel to the skin surface, the fibers form a cross-hatched lattice with organizational peaks separated by ∼90°, while perpendicular to the surface, they align circumferentially (Fig. 1D). This complex, multiscale architecture provides a clear structural basis for the functionally critical mechanical anisotropy detailed below. Structural anisotropy creates a stark mechanical dichotomy for grasping and protection The architectural anisotropy of the trunk skin directly translates into a functional mechanical dichotomy. Uniaxial tensile tests reveal that the dorsal skin is significantly stiffer than the ventral skin, supporting the primary role of protection; this difference is most pronounced in the transverse direction (Fig. 2A). Quantitatively, the dorsal skin's initial Young's modulus (E1) is 0.22 MPa, which is 3.14 times greater than that of the more flexible ventral skin (E1 = 0.07 MPa), as detailed in the comparative analysis shown in Fig. 2C and D. Although the secondary modulus (E2) remains comparable across regions, likely reflecting their similar collagen composition, the stark difference in E1 ensures the dorsal skin resists initial deformation more strongly, providing a “shield” effect before large deformations occur. Mechanical testing confirms a functional dichotomy. (A) Average tensile stress-strain curves show that dorsal skin is significantly stiffer than ventral skin in the transverse direction. (B) Example of a representative stress-strain (σ-ɛ) curve showing the procedure for data analysis to define the initial Young's modulus (E1), secondary modulus (E2), and first changepoint (CP). (C) Box plots and individual data (small circles) quantifying the mechanical parameters identified in the previous plot across the sample groups in (D): Longitudinal Dorsal (LD), Longitudinal Ventral (LV), Transverse Dorsal (TD), and Transverse Ventral (TV). Small circles represent individual data points indicating distribution. € Indentation tests on isolated layers reveal that the epidermis is substantially stiffer than the underlying corium. While longitudinal tensile tests showed greater sample-dependent variability (standard deviation of 0.41 MPa at 30% strain), the ventral skin consistently exhibited lower initial stiffness, a property that facilitates bending and conforming around objects. Functional zoning is further corroborated at the layer level: indentation tests on isolated tissues show that the epidermis is ∼8.6 times stiffer than the underlying corium (Fig. 2E). Together, these results provide quantitative evidence that the elephant trunk skin is a functionally specialized material, divided into a stiff dorsal “protective shield” and a flexible ventral surface optimized for conformal grasping. Modeling suggests subsurface papillae as a plausible mechanism for sensory amplification To investigate a potential mechanism linking the unique architecture of the ventral skin to its sensory function, we developed an FEA model as a hypothesis-generating tool to test the role of subsurface microstructures. As outlined in our methods (Computational Modeling for Mechanistic Hypothesis Testing), the model does not serve to perfectly replicate the skin but to isolate geometric effects (Fig. 3A). Simulations predict that the DPs, which are prominent only in the ventral skin, function as effective stress amplifiers. The FEA model indicates that increasing the aspect ratio (height/width) of DPs leads to a proportional increase in Von Mises stress concentration at the dermal–epidermal junction, a region known to house mechanoreceptors (Fig. 3B and C). This suggests a plausible mechanism whereby the skin's geometry is tuned to amplify tactile signals before they reach the sensory nerves. Furthermore, the model predicts that surface features also play a key role, with wrinkles influencing traction forces and DRs contributing to strain distribution across the surface (Fig. 3D). Together, these computational results support the hypothesis that the skin's microarchitecture is finely tuned to create a sensitive, responsive interface for interaction with the outer world. Modeling a mechanism for tunable tactile sensation. A) Schematic of the inverse FEA workflow used to derive material parameters for the epidermis from experimental indentation data and key assumptions used in the simulation. B) Model predictions showing that larger DPs, characteristic of the ventral skin, lead to higher Von Mises stress concentrations (highlighted at the DP base in the schematic). Boxplots summarize the peak stress distribution across varying geometries. C) Effect of DRs on normal strain. A model with varying widths and heights of DRs was compared with a model without them. D) Simulations demonstrating the influence of surface features, where wrinkles affect traction (shear) forces (φ, orientation relative to pulling direction; d, depth; w, width) and DRs reduce surface strain. Discussion The skin of the Asian elephant’s trunk is not a simple protective covering but a functionally graded biocomposite. We show that its architecture and mechanics are systematically zoned to solve the fundamental “grasp-protect” paradox inherent to muscular hydrostats. The evidence points to a clear design strategy: a stiff, durable dorsal shield is seamlessly integrated with a flexible, sensitive, ventral, conformable side. This section synthesizes our findings, proposes a blueprint for bio-inspired robotics, and outlines a roadmap for future research. An integrated solution to the grasp-protect paradox The elephant's ability to perform both powerful and delicate tasks is enabled by a profound structural and mechanical dichotomy in its skin. Our findings reveal that this is achieved through a synthesis of features at all scales, starting with the dorsal shield. The trunk’s dorsal outer curvature naturally exposes it to environmental impacts, and the skin in this region is perfectly adapted for this role. It is 3.14 times stiffer than its ventral counterpart, exhibiting a characteristic J-shaped mechanical response common to collagenous tissues (8). This behavior is first governed by the flexibility of an elastin-like network, followed by rapid stiffening as cross-hatched collagen fibers (Fig. 1D) align and engage under load. Combined with a thicker dermal layer and prominent interlocking DRs, this creates a tough, abrasion-resistant shield (Figs. 1 and 2). The rougher surface texture further enhances this protective function, potentially aiding in thermoregulation through mud retention, as suggested by earlier studies (9). In contrast, the concave ventral surface is well-suited for manipulation. Its significantly lower stiffness allows it to readily conform to objects, maximizing the contact area essential for a stable grip (10). This flexibility is complemented by its unique subsurface architecture. The large DPs found exclusively on the ventral side represent a key adaptation. Although early anatomists like Smith hypothesized the tactile function of the skin (11), our computational modeling provides a clear mechanistic basis for this claim. Our FEA model predicts that these DPs act as stress amplifiers, concentrating contact forces at the dermal–epidermal junction where mechanoreceptors are located (Fig. 3). This geometry-mediated effect functions like a subsurface mechanical lens: the dome-like shape of the papilla funnels the widely distributed force from the skin's surface, concentrating it onto a small, high-pressure focal point at its base. This presumably allows a weak, gentle touch on the surface to be transformed into a strong, distinct signal precisely where the nerve endings are located. This is a clever biological strategy: instead of evolving more sensitive neurons, the skin's structure mechanically “preprocesses” and amplifies the signal, a principle also seen in human fingertip skin (12). This integration of macroscale flexibility with microscale sensory amplification of the ventral skin allows the trunk to be both a strong gripper and a sensitive “finger.” This elegant biological solution, which balances passive material properties to achieve complex functions, not only provides insight into the animal's evolution but also offers a powerful lesson in energetic efficiency and mechanically mediated signal processing for engineered systems. A biological blueprint for multifunctional robotic skins The design principles observed in the trunk skin offer a direct and sophisticated blueprint for the next generation of soft robotic systems, which face similar paradoxes. Our findings can be distilled into two key design principles, visually summarized in Fig. 4. Zoned stiffness for functional efficiency: Current robotic skins are often made of uniform materials, representing a compromise between durability and flexibility. Our results advocate for a zoned approach. By incorporating a stiff, abrasion-resistant dorsal surface with a compliant, high-friction ventral surface (e.g. via multimaterial printing), a robotic gripper can achieve robust protection without sacrificing the delicate conformability needed for manipulation. This strategy also improves energetic efficiency by relying on passive material properties rather than complex active control to manage different tasks. Zoned stiffness for functional efficiency: Current robotic skins are often made of uniform materials, representing a compromise between durability and flexibility. Our results advocate for a zoned approach. By incorporating a stiff, abrasion-resistant dorsal surface with a compliant, high-friction ventral surface (e.g. via multimaterial printing), a robotic gripper can achieve robust protection without sacrificing the delicate conformability needed for manipulation. This strategy also improves energetic efficiency by relying on passive material properties rather than complex active control to manage different tasks. Geometric structures for sensory amplification: Achieving high tactile sensitivity without exposing fragile sensors is a major challenge in robotics. The elephant's use of subsurface DPs provides an elegant solution rooted in mechanical principles. Our model predicts that the “geometry” of these papillae-like features mechanically amplifies faint force signals at the sensor-matrix interface. This is not a matter of embedding a stiffer material but of shaping the interface itself to concentrate stress. Engineers can tune the size, aspect ratio, and spacing of these geometric features to selectively amplify certain types of stimuli, creating a robust sensor that is protected from mechanical wear. This moves beyond simple sensor placement to intelligent, mechanically mediated signal processing through structural design. Geometric structures for sensory amplification: Achieving high tactile sensitivity without exposing fragile sensors is a major challenge in robotics. The elephant's use of subsurface DPs provides an elegant solution rooted in mechanical principles. Our model predicts that the “geometry” of these papillae-like features mechanically amplifies faint force signals at the sensor-matrix interface. This is not a matter of embedding a stiffer material but of shaping the interface itself to concentrate stress. Engineers can tune the size, aspect ratio, and spacing of these geometric features to selectively amplify certain types of stimuli, creating a robust sensor that is protected from mechanical wear. This moves beyond simple sensor placement to intelligent, mechanically mediated signal processing through structural design. A biological blueprint for multifunctional robotic skins. This integrated view summarizes the paper's core findings. The elephant’s trunk solves the grasp-protect paradox through a functionally zoned architecture. The dorsal skin acts as a “protective shield,” characterized by high stiffness and durability (left pathway). The ventral skin serves as a “sensory gripper,” defined by high flexibility and subsurface microstructures that amplify tactile stimuli (right pathway), offering a design paradigm for the next generation of soft robotics. Limitations and future directions Our study not only provides a foundational framework but also illuminates critical avenues for future research. We identify three main areas for advancement. First, this investigation is a detailed case study of a trunk of a single adult Asian elephant. To establish species-wide principles, future work must conduct comparative analyses across multiple individuals, ages, and species, particularly comparing the Asian and African (Loxodonta africana) elephants, which inhabit different environments and exhibit distinct wrinkling patterns (5). Additionally, our quantitative SHG analysis was prioritized for the tip region due to its functional importance for manipulation; future mapping of collagen architecture along the full proximal–distal axis would provide a more complete biomechanical picture. Second, our mechanical characterization relied on cryopreserved tissue. Although this is a necessary method for this study, freezing can alter the properties of hydrated soft tissues (13). The next generation of studies should leverage techniques like Digital Image Correlation in fresh-tissue samples to create validated strain maps, providing more precise data for both comparative analysis and computational modeling. Furthermore, due to the logistical and regulatory constraints of working with this rare species, testing to failure was not feasible. While we characterized the physiological low-strain regime, the ultimate failure strain and damage mechanics remain an open question. Finally, our FEA, while mechanistically insightful, employed necessary simplifications. Future models should incorporate the full nonlinear, hyperelastic, and viscoelastic properties of the skin layers captured via inverse FEA on fresh tissue to fully resolve the anisotropic tensor. Crucially, these advanced models must be validated against the fresh-tissue experimental data proposed above. A fully validated model would allow for predictive design and a deeper understanding of the interplay between the complex structure of the skin and its remarkable functions. Materials and methods Specimen provenance and preparation The trunk analyzed in this study was obtained from a single adult Asian elephant (Elephas maximus) that died of natural causes at Zurich Zoo, Switzerland, in 2020. The University of Geneva obtained the trunk for anatomical studies and performed the initial sectioning. Therefore, this investigation constitutes a detailed case study. The trunk was prepared in different steps, including initial freezing at −28 °C, thawing for imaging scans, and refreezing for sectioning. Transversal serial slicing of a 127-cm-long portion yielded ∼130 slices at 1 cm intervals. The remaining proximal 45 cm was divided sagittally, with the left half cut into 20 sagittal slices and the right half into 25 coronal slices. All procedures adhered to international regulations, with re-export and import permits obtained from the Swiss and Italian governments as required by the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES). Due to logistical constraints, tissues for mechanical testing were kept at −28 °C. Although necessary, this approach can alter the biomechanical properties of soft tissues, primarily through matrix dehydration and collagen stiffening (13). Therefore, while the relative differences reported between anatomical regions are considered robust, the absolute values of the mechanical properties should be interpreted within this context. Morphological and microstructural analysis To build a comprehensive architectural model of the skin (Fig. 1), a suite of imaging techniques was employed. Skin samples from the dorsal and ventral regions were collected at 59 and 4 cm from the tip of the trunk for histological analysis. Mallory's trichrome staining was performed, labeling different tissue components: blue for mucus, black for nuclei, gray for cytoplasm, light blue/gray for connective tissue, and red for muscles. Transversal trunk sections (1 cm thick) from various positions were dehydrated, cleared, and impregnated with methylmethacrylate. The hardened samples were sectioned to <1 mm, ground to ∼300 µm, and surface etched. The images were acquired using a digital microscope (Keyence VHX-7000 series) to measure the dimensions of DRs and DPs. To visualize the label-free collagen fiber network, SHG microscopy was performed on trunk skin samples using a Leica SP8 DIVE confocal microscope with a 60× oil-immersion objective. This method allowed for the characterization of the orientation of the fibers (Fig. 1D). Prior to mechanical testing, muscle tissue was carefully excised. The samples were thawed and rehydrated in 10 mM phosphate-buffered saline. High-resolution images were captured using a Hirox KH-8700 digital microscope. Approximately 200 epidermal islands (EIs) from top-view images and 50 EIs from cross-sections were analyzed using ImageJ software. The island aspect ratio (Ar) was calculated as Ar = √A/t. Biomechanical characterization To understand how the observed architectures translate to mechanical function (Fig. 2), we performed the following tests. The surface roughness (Ra) of six samples from dorsal and ventral regions was evaluated at three different points each using a Hirox KH-8700 digital microscope, as described in (14). Computational modeling for mechanistic hypothesis testing FEA was performed using COMSOL Multiphysics v6.0 to explore the mechanical implications of the skin's microarchitecture (Fig. 3). We utilized the model as a hypothesis-generating tool to isolate the geometric effects of specific microstructures on the skin's sensory function. As detailed in Supplementary Note S4, we adopted a hybrid dimensionality strategy: a full 3D model was used to capture stress concentrations around the complex DP geometry (Fig. 3B), while 2D plane-strain and axisymmetric approximations were strategically employed to decouple geometric variables, allowing us to study specific morphological features (e.g. ridges and wrinkles) in isolation during parametric analysis (Fig. 3C and D; variance across individual geometric configurations is detailed in Fig. S4). The skin was modeled as a bilayer structure. The epidermis was modeled using the hyperelastic Yeoh model, with parameters derived from an iterative least-squares optimization of indentation data (Fig. 3A). The parameters used were C10 = 0.45 MPa, C20 = 0.01 MPa, and C30 = 0.001 MPa. The corium was modeled as a softer, linear elastic substrate (E = 0.61 MPa, ν = 0.49) based on the linear region of corium-specific indentation tests. Boundary conditions included a fixed bottom surface and roller constraints on lateral boundaries to simulate an infinite periodic tissue continuum. Loads were applied using a rigid cylindrical indenter (3 mm diameter) for normal indentation or tangential displacement for traction simulations. Quantification and statistical analysis Data were collected from 35 adult Asian elephant trunk skin samples in five main anatomical locations. A detailed breakdown of the number of samples used for each specific analysis (e.g. microstructural, mechanical, and morphological) is provided in Table S1 in Supplementary Note S5: Detailed Sample Counts and Statistical Analysis of the Supplementary material. All statistical analyses were performed using MATLAB (The MathWorks, Inc.) with a significance level of α = 0.05. Data normality was evaluated with the Shapiro–Wilk test. As mechanical test data for parameters E1 and CP were not normally distributed, nonparametric analysis (Kruskal–Wallis) was used to assess differences. Hedges’ g was calculated as an effective effect size measure to compare these non-Gaussian data groups (15) (Fig. S5). For all normally distributed morphological data, one-way ANOVA with post hoc pairwise comparisons was used, as these variables showed a P-value of <0.001. Ethics The specimens were obtained by the University of Geneva from an adult Asian elephant only after the natural death of the animal at the Zurich Zoo in 2020. No animals were euthanized for the purposes of this study. The work did not involve live animal experimentation and was therefore not subject to approval by an Institutional Animal Care and Use Committee (IACUC). All procedures adhered to international regulations. Re-export and import permits were obtained from the Swiss and Italian governments, respectively, as required by the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES). Supplementary Material