ConclusionsThe invention of harpoons is a fundamental milestone not only in the evolution of natural life but also for humans[48]. In Nature, evolution produced a large variety of harpoons and puncturing devices of increasing complexity, ranging from the passive and fairly simple cacti spines[49], to the more sophisticated offensive tools such as spider [41] and viper fangs[50], scorpion stinger[51], and even including drilling machines as in mosquito or parasitic wasps [52,53]. Particularly fascinating are the chitinous penetrating tools of arthropods, which often allow piercing through several layers of a similar material, and possessing multiple modifications to fulfill such task [54]. The spikes emerging from the spearing appendage of the mantis shrimps are biological harpoons deployed very quickly to impale and grab moving fishes, which have the bad luck to swim nearby. This is a challenging task requiring the ability not only to resist the initial impact with the scaly skin of the prey, but also to penetrate it for several millimeters to avoid that the fish swims away. Here, we have investigated the spike using imaging and material characterization techniques, to elucidate the many design strategies of this biological tool. From the material point of view, the spike cuticle of stomatopod is a chitin-based multilayered biocomposite, reinforced by different minerals together with compositional gradients and specific fiber arrangement[36]. A hard heavily mineralized outer shell of crystalline FAP, essentially lacking fibers, is designed to resist to impact and wear. This exterior cover is combined with inner fiber-rich regions characterized by lower mineral content and by amorphous minerals (ACP and ACC). Two distinct fiber architectures are observed: a unidirectional arrangement bordered by helicoidal twisted plywood of dissimilar size. The latter is ubiquitous in arthropod [12] and presents two main biomechanical advantages [45]: starting from a strongly anisotropic elementary building block (the chitin-protein fiber), it provides the cuticle with in-plane isotropic mechanical behavior and high shear stiffness[41,55]. Moreover, the helicoidal twisted plywood arrangement has remarkable fracture resistance obtained by interacting with cracking at different levels[56], essentially enhancing crack driving force[57] and reducing strain energy release rate[58]. Conversely, the parallel-fiber organization is less usual [45] and is used here, well oriented along the long axis of the spike from tip to base, to enhance stiffness and bending resistance. The combination of a parallel fiber with a twisted plywood region is therefore well-suited to support a complex mechanical environment with axial, bending and torsional loading. A central finding of our work is that, in the spike cuticle, the parallel fibers region is “joined” to the highly mineralized cover by a thin fiber-rich twisted plywood region, showing compositional modifications which make it the most compliant zone of the entire spike. In the cuticle, plywood (or Bouligand structures) are generally thick regions occupying a major part of the cuticle width and previous works have demonstrated the superior fracture toughness of Bouligand structure [34,59,60]. The unique feature observed here is the presence of an extremely thin plywood structure (i.e., less than 10 μm in width) sandwiched between two very dissimilar regions and allowing the integration of a highly mineralized brittle outer layer with a less mineralized but highly anisotropic straight fibers region. Not only the plywood is very effective to stop crack propagation (both from the striated region to the hard shell and vice-versa), as demonstrated by nanoindentation fracture experiment and by 3-point-bending tests on spike-inspired synthetic systems, but it also increases the force required to penetrate both the stiff layer and the underlying complaint fibrous matrix (as assessed on 3D printed replicas). Considering the attachment of different materials, introducing a more compliant region at the interface joining two dissimilar components is a construction principle common to other biological systems[61]. Tendon, for example, attaches to bone through a transition zone which is not only more compliant than bone but also than tendon [62]. This region, which co-localizes with the unraveling and splaying out of tendon fibers into smaller fibrils [61] and which is made up of fibrocartilage [63], is believed to protect the attachment region by reducing stress-concentration, effectively strengthening the interface [64]. In analogy with the tendon-bone attachment, the thin and more compliant twisted plywood region may offer protection against stress localization at the transition between the hard and the parallel fibers region, hence increasing the damage tolerance of the spike. Helix reinforced composites are common in engineering applications and the construction principles of biological materials can improve the performance of the synthetic counterparts [65]. Focusing on the mantis shrimp dactyl club, Bouligand and herringbone arrangements as well as nanoscale features of the impact surface, have inspired the design of impact resistance man-made composites[66]. In light of the endless advancement in nano- and micro-scale manufacturing methods, the biological tool investigated here could inspire the design of new synthetic harpoons for example based on environmentally friendly and recyclable building units as sees in the spike cuticle, with improved wear resistance and puncture abilities for repeated piercing on different surfaces.
Experimental Section
Stomatopods and sample preparation : The specie of spearing mantis shrimps used in this study is Lysiosquillina maculata (Fabritius, 1793) also called the striped mantis shrimp (Figure 1A). Eight living stomatopods from Kendary (Indonesia), ranging from 20 to 40 cm in length, were delivered by Marine Life (Paris, France) and were kept in captivity in proper tanks at the Functional Morphology Lab (ULiege). Spikes were harvested from 3 different individuals after euthanasia, for a total of 9 spikes analyzed with the different analyzed with the different methods. All specimens were dissected in the intermolt stage C4, corresponding to entire and mature cuticle[67]. Distal spikes were then separated from each other (Figure 1B). Samples imaged with micro-CT (n = 2) were directly embedded in epoxy resin to minimize movement artefact. Samples used for SEM (n = 4), Raman (n = 3) and nanoindentation (n=6) were dehydrated in a graded series of ethanol bath (50%, 70%, 90%, and 100%) and embedded in epoxy resin (EpoFix resin Kit, Struers Inc., Germany) under vacuum. The resin blocks where then polished (Rotopol-2, Struers, Germany) with series of silicon carbide disks of decreasing grain size (P800, P1200, P2400, P4000, Matador, Germany). Final polishing steps were performed with a diamond spray (DP-Spray P 1 μm, Struers, Germany) and an aluminum suspension (Eposil F, 0.1 µm, ATM, Germany).
Micro-CT imaging and processing: Micro-CT scans of resin embedded spikes were done at a nominal isotropic voxel size of 2 μm. The micro-CT (Skyscan 1272, Bruker, Belgium) was operated at tube voltage of 60 kV and current of 166 μA, in combination with a 0.25 mm thick aluminum filter. The samples were rotated over 180° with a rotation step of 0.2° (corresponding to 940 projections), with an exposure time of 3600 ms and a frame averaging of 4, leading to a scan time of approximately 4 h. A bigger sample including also the dactyl was scan at a lower resolution of 15 μm. Scans were reconstructed using Nrecon (v.1.7.5.2, Skyscan), and further analyzed with ImageJ (v.1.52a), Matlab (R2018a; The Mathworks, USA) and CTAn (v.1.19.4.0, Skyscan). Each virtual spike was first aligned along its principal axes of inertia using BoneJ (v.1.4.3)[68], a module of ImageJ. Images were binarized using a global threshold based on Otsu’s method[69] and filtered to extract the biggest connected component. Geometrical properties of the cross-sections were measured using the Matlab Image Processing Toolbox. Using the general formula\(I=\int{z^{2}\text{dA}}\ \) where z is the shortest distance of the area element dA from a given axis, the planar (\(I_{x}\) and\(I_{y}\)) and polar (\(J=\ I_{x}+\ I_{y}\ \)) second moments of area passing through the centroid were computed. Major and minor axis were deduced by fitting an ellipse having the same second moments of area as the current cross-section. Eccentricity was computed as the ratio of the distance between the foci of the ellipse and the length of its major axis. Principal moments of area about the minor (\(I_{\min})\)and the major (\(I_{\max})\ \)axis of each cross-section, corresponding approximately to dorso-ventral and lateral loadings, were calculated based on Mohr’s circle theory. To compute the curvature of the spike, firstly a centroid profile line was extracted smoothed and down-sampled, to reduce the noise. At each location along the profile line, the local radius of curvature was computed by fitting circles, always considering three neighboring points. The curvature vector was obtained by dividing the unit vector pointing from the middle point of the triple to the center of the circle by the radius of curvature[70]. A similar procedure was followed to compute the radius of curvature at the tip.
Scanning electron microscopy and energy dispersive spectroscopy : Observations of polished sections were carried out with an environmental scanning microscope ESEM-FEG XL-30 (FEI, Nederland) in low vacuum mode (0.4 Torr) at 20 KeV accelerating voltage. Images were taken using a back-scattered electron detector. A silicon drift detector of X-rays (Bruker, USA) with a super-ultra-thin window was used to obtain energy dispersive spectra and to perform elemental mapping using a Quantax analyzer and the software Esprit 2.1 (Bruker, USA). To measure the different elements concentration, the following energy peaks were used: Carbon Kα peak (0.277 keV), Phosphorus Kα peak (2.013 keV), Calcium Kα peak (3.690 keV), Magnesium Kα peak ( 1.253 keV), Fluorine (Kα peak at 0.677 keV) and Sodium (Kα peak at 1.041 keV). Fractured samples were sputter-coated with silver (SCD030, Balzers Liechtenstein) and observed in high vacuum mode using the same microscope with an acceleration voltage of 15 kV and a secondary electron detector. Measurements of grooves and serrations were made using ImageJ software[71] on SEM-BSE micrographs of the spike in longitudinal and transversal sections to quantify their length and depth.
Raman spectroscopy : Polished sections of the spike were analyzed by Raman spectroscopy with a green DPSS laser (λ = 532nm, LabRAM 300, HORIBA Jobin Yvon, Japan). Spectra were acquired along straight lines at several different positions within the highly mineralized region, the outer helicoidal layer and the striated region of the cuticle. Each spectrum was acquired for 5 seconds. High resolution Raman mapping was performed on the outer region of the spike cuticle with the same microscope by focusing the laser beam through an Olympus MLPlan objective (NA 0.75), using an integration time of 1 s and resulting into a nominal pixel size of 500 nm. The mapping was performed with two different polarization angles (0 and 90 deg). The obtained spectra were analyzed with Rstudio (Rstudio Team, 2015) to extract the phosphate intensity (peak intensity in the 952 to 968 cm-1region), the carbonate intensity (peak intensity in the 1077 to 1083 cm-1 region), the phosphate position (wavelength at which the phosphate υ1 vibration mode is the highest), the FWHM (width of the phosphate υ1 vibration mode at its half height) and the polarization ratio (intensity of the 960 cm-1 band in 0 / 90 deg polarization).
Depth sensing nanoindentation : Nanoindentation tests were conducted with a TriboIndenter TI-950 (Bruker, USA) on re-polished surface (the same samples used for SEM were probed), considering both transversal and longitudinal sections. We first used a Berkovich diamond tip to perform nanoindentation grids spanning across the different regions of the spike with indents spaced 10 μm apart (grid size in Figure 5A: 32x5 indents). We employed a trapezoidal load-controlled function (10s-5s-10s for loading, holding, and unloading segments) with a peak load of 5000 µN. Higher resolution nanoindentation mapping covering \(\sim\)45x45 μm areas was carried out on two locations at the interface between the highly mineralized region and the outer helicoidal region after scanning the surfaces with the tip of the indenter to check the roughness (average RMS roughness in the range \(\sim\)16-25 nm depending on the location). We used a sharper cube corner probe and a displacement-controlled indentation (10s-5s-10s) with a maximum penetration depth of 200 nm, allowing a smaller spacing of 1.5 μm between indents [72]. Both tips were calibrated in fused quartz. Force-depth curves were analyzed with the Oliver and Pharr method [73] to extract the indentation modulus and hardness. High lateral resolution nanoindentation data were used to generate 2D maps of the indentation modulus (Figures 5C and D and S8) and raw data were interpolated with the MESHGRID function of Matlab. A qualitative high load fracture study was performed on polished samples with a high load 10 N transducer (Omniprobe, Bruker, USA) equipped with a Berkovich probe. Several indentation loads ranging from 50 mN to 1000 mN were used to generate surface damage in different locations of the spike.
3D printing and testing: Spike inspired samples were designed with IronCAD (2020, USA), fabricated using a 3D multimaterial polyjet printer (Objet 260, Stratasys, USA) and tested. Specifically, we performed penetration tests with a blunt tip (5 mm tip radius) and 3-point-bending tests (MTS Criterion C43.304, USA). For the penetration tests we considered cuboid samples having the following dimensions: 4 cm x 3 cm x 3 cm (height, width and length). For 3-point-beding, we fabricated beam-shaped samples, with dimensions: height (H) 4 cm, width (W) 3 cm and length (L) 20 cm. Each sample was a multilayer composite, featuring an outer monolithic region (mimicking the heavily mineralized shell) and inner fiber-reinforced region. For the fiber reinforced part, two different architectures were considered: i) a thick region composed of unidirectional parallel fibers, all oriented perpendicular to the longitudinal axis of the sample (mimicking the striated region) and ii) a two-layer system with a thin helicoidal twisted plywood region (mimicking the outer helicoidal region) sandwiched between the monolithic part and the unidirectional fibers array. The monolithic layer as well as the fibers were printed with a rigid glassy polymer (commercial name Vero White Plus), having Young’s modulus at room temperature in the range of 2-3 GPa [74]. For the matrix embedding the fibers we used a rubbery polymer (commercial name Tango Black Plus) with a tangent modulus at room temperature of about 0.5-1 MPa [75]. Based on the printer resolution and considering the finite size (i.e., up to 150 μm) and the properties of the interface between fiber and matrix [76], the 3D printed fibers had a diameter of 600 μm. For the region mimicking the helicoidal region structure, 10 sheets were stacked with a pitch angle of 18°. At the bottom surface of the samples, a rigid layer made up of Vero White Plus was added to prevent excessive sink-in at the contact point with the supports. A notch (0.15 cm deep) was directly printed in the samples to trigger check initiation within the parallel fiber array. Penetration tests were done at 10 mm/min whereas three-point-bending tests (span length of 16 cm) were conducted at 60 mm/min, considering preload of 5 N and with a 30 kN load cell. During tests, samples were imaged with a HD camera at 15 fps and movies were used to analyze deformation mechanisms. Mechanical tests were performed 1 day after printing to have always the same post-curing time. Fabrication, storage, and testing were done in a room with controlled humidity and temperature to minimize experimental variability.