Design Complexity and Fracture Control in the Equine Hoof Wall

Horse evolution has led to the development of numerous morphological characteristics which enhance locomotor speed. One striking example has been the weight reduction of the limb terminus by eliminating all but a single digit. Surrounding the digit is a lightweight, truncated-cone-shaped structure called the hoof wall, which suspends the bony skeleton. The wall is in direct contact with the substratum and serves to transfer ground-reaction forces to the bony skeletal elements. Strain measurements of the outer surface suggest that the wall does not routinely approach strains which may result in failure (Thomason et al. 1992); however, the animal’s relatively large size, coupled with its particular high-speed gait patterns and the likelihood of contact with uneven substrata, lead to the potential for the development of high, localized stress concentrations. The hoof wall must therefore be capable of withstanding these high stresses which threaten failure by crack propagation.

Although the bulk of the wall is continually being regenerated proximally at the coronary border, unlike bone, hoof wall cannot be repaired or remodeled once it is formed. Any damage (such as cracks) will remain as a threat in the tissue until it is worn off distally. Failure of the wall is, however, unusual and has therefore made the wall the focus of a number of mechanical and morphometric studies (Leach and Zoerb, 1983; Bertram and Gosline, 1986, 1987; Thomason et al. 1992; Kasapi and Gosline, 1996; Douglas et al. 1996). Most of these studies focused primarily on the stratum medium (SM) region of the hoof wall, since the outermost stratum externum forms a relatively thin covering (Banks, 1993) that apparently functions to inhibit dehydration and is not load-bearing, and the innermost, lamellar stratum internum appears to be primarily responsible for transferring loads to the bony skeleton. Previous mechanical studies from our laboratory have shown that the SM is highly resistant to crack propagation (Bertram and Gosline, 1986, 1987; Kasapi and Gosline, 1996) and is one of the most fracture-resistant biomaterials known.

In a recent study (Kasapi and Gosline, 1996), we found that hoof wall test pieces fractured in a manner that was inconsistent with our current knowledge of hoof wall morphology. Specimens which incorporated most of the thickness of the wall and were notched along the hoof longitudinal–radial plane, fractured like a trilaminar ply.

Unlike man-made plies, however, distinct lamellar boundaries were not evident in fractured hoof wall specimens. As noted previously (Bertram and Gosline, 1986), the fracture path in the middle region was redirected across hollow tubules (the dominant components of the SM microstructure) along the intertubular intermediate filament plane. Thus, in the middle region, the intertubular material forms a barrier to the propagation of cracks up the hoof wall. Crack propagation in the inner and outer regions of the wall did appear to be controlled by tubules. In these regions, a notch initiated in the radial plane formed cracks which generally continued running in the radial plane along the tubule axis, suggesting that the function of the tubules may vary depending on their position through the wall thickness.

Nickel (1938a) identified three main types of tubules in equine hoof wall and noted that the fibrous molecules which fill the cells of the tubule cortex formed concentric lamellae of alternating helices. On the basis of a microscopic study on bovine hoof wall, however, Wilkens (1964) suggested that the arrangement of tubule cortex cells more closely resembles the pattern of microsporophyll arrangement of pine cones (although ‘pine cone’ was apparently inadvertently translated into ‘pin-cushion’ in the publication) whereby the tubule cell plane lies at an angle to the tubule axis. Morphological over-simplification by both investigators has led to our present knowledge of wall microstructure, which is insufficient to provide a satisfactory explanation of hoof wall mechanics.

Using microscopy, it is possible to develop a complete model of the hoof wall which may aid in the understanding of the mechanisms that confer the wall’s favorable mechanical properties. Composite theory predicts that the stiffness of a fiber-reinforced composite will depend upon the orientation of the fibers relative to an applied stress and upon the mechanical properties and volume fractions of the fiber and matrix phases (Wainwright et al. 1982). Therefore, knowledge of keratin fiber orientation is of paramount importance in understanding the mechanics of the wall, and should predict tensile and fracture behavior. Polarized light microscopy, which may be used to determine the orientation and degree of order of molecules, is useful in developing a model since the entire wall is composed of cells containing the highly ordered protein-based molecular composite α-keratin.

The key component of α-keratin is a member of a family of closely related proteins, collectively called intermediate filaments (IFs; see Grosenbaugh and Hood, 1992), and it is generally accepted that all IFs are composed of two-stranded rope-like protofibrils approximately 45 nm long and 1 nm in diameter (Parry et al. 1987) which associate helically to form supercoils approximately 7 nm in diameter (Crewther et al. 1983). These IFs are embedded in a globular, viscoelastic protein matrix. At maturation, keratin-containing cells are hard, flattened and generally elliptical structures.

The following study is the first three-dimensional documentation of the IF organization in cells of both the tubule cortex and the surrounding intertubular material of the equine hoof wall toe region. It also documents all aspects of tubule morphology and combines the findings with mechanical data from tests conducted on isolated regions of the hoof wall in order to link form and function. Our results suggest that the complex design of the structure is a reflection of an equally complex loading pattern. The intertubular material appears to play a major role in crack redirection and skeletal load transfer, while the role of the tubular components is dependent on their position through the thickness of the wall.

Equine hooves were obtained from six freshly killed horses (Equus caballus L.) of unknown age and mass (destroyed for reasons other than this study) and disinfected in 0.02 % benzalkonium chloride for approximately 1 h. Wall tissue was roughly sectioned using a bandsaw and refrigerated at 4 °C in distilled water with 0.02 % sodium azide to prevent bacterial growth. Keratinous tissues are highly resistant to microbial attack, so histological specimens could be stored in this manner for months without degradation. Samples used in one series of mechanical tests were used within 9 days of the death of the animal, and in a second series were used within 90 days after the death of the animal.

Procedures for mechanical tests followed those described previously by Kasapi and Gosline (1996), with the following modifications. Blocks of hoof tissue running the full length of the hoof wall and spanning the entire SM (see Fig. 1) were either cut circumferentially into three strips of approximately equal thickness or sectioned radially (Fig. 1A,B) using a water-cooled thin-sectioning machine (Gillings-Hamco). Material loss was minimized by using very thin (0.35 mm) circular saw blades. Compact tension (CT) and tensile tests utilized hooves from three animals, one for each of two series of CT tests, and one for the tensile tests.

Strips of hoof wall were cut to appropriate dimensions for CT tests (Fig. 1D) using the thin-sectioning machine and were notched using a razor blade affixed to an attachment on a drill press. In the first series of CT tests, 65 specimens were produced (22 inner, 31 middle and 12 outer wall) from the four hooves of one animal and were notched upwards along the radial–longitudinal plane (Fig. 1B). Notch length (a) to specimen width (W) ratios ranged from 0.23 to 0.51, from 0.10 to 0.53 and from 0.13 to 0.55 for inner, middle and outer wall specimens, respectively. This series was produced to test for differences in mechanical properties through the wall thickness. In the second CT test series, 42 test pieces were produced from the right front hoof of another animal. To test for inter-animal variability, seven pieces were obtained from the middle hoof wall region and were notched radially as described above (a/W=0.23–0.60). Thirty-five pieces were produced that spanned the wall thickness from an adjacent block to test for possible crack diversion mechanisms. Twelve specimens were notched upwards along the circumferential–longitudinal plane (series 2A; a/W=0.20–0.65), eleven were notched inwards along the circumferential–radial plane (series 2B; a/W=0.26–0.61) and twelve were notched inwards along the radial–longitudinal plane (series 2C; a/W=0.23–0.59). All tests were performed at room temperature (approximately 20 °C) with samples immersed in distilled water using the test apparatus described by Bertram and Gosline (1987). Samples were tested at a cross-head speed of 8.3×10⁻⁵ m s⁻¹ using an Instron mechanical testing machine (model 1122) with a 50 kg load cell. Data were collected at 10 Hz using PC software (Labtech Notebook) and were processed with spreadsheet software. For a full explanation of CT specimen preparation, refer to Kasapi and Gosline (1996).

The stress intensity factor K was found using the equation provided by ASTM standard E399-90 (ASTM, 1994b); the procedure for determining the J-integral and the equations for calculating the initial circumferential stiffness or modulus E_i,C and the initial radial modulus E_i,R are outlined in Kasapi and Gosline (1996). Poisson’s ratio was estimated as 0.40, 0.45 and 0.47 for inner, middle and outer hoof wall, respectively (M. A. Kasapi and J. M. Gosline, unpublished data). A Poisson’s ratio of 0.45 was used for CT specimens which spanned the wall thickness, since the notch was introduced along the middle region. By filming the notch front and corresponding force records for some of the CT tests, we determined that the critical displacement (the point at which crack extension is initiated) was the point on the load–displacement curve where the first decrease in load was observed. Used CT test specimens were prepared for the scanning electron microscope (SEM) by first dehydrating the tissue with a standard ethanol series. After two final washes in 100 % ethanol for 1 h each, specimens were critical-point-dried and sputter-coated with gold. Samples were mounted onto aluminum stubs with SPI silver paint and viewed on a Cambridge Stereoscan 250T scanning electron microscope.

Samples for tensile tests were milled to shape (Fig. 1C,E) using a brass template that served as a guide for a milling machine. Tests were conducted with the Instron using a 50 kg load cell, and specimens were held with pneumatic grips. Strain was measured directly with a strain gauge displacement transducer mounted to the front of each specimen. Very fine hypodermic needles (23G1) were attached to the ends of the transducer to prevent slippage between the transducer arms and the specimen. Slight needle penetration acted to concentrate stress; therefore, ultimate data were not used from specimens that failed at the needle marks or near the grips. Tests were conducted on 22 fully hydrated specimens (eight inner, seven middle and seven outer; obtained from all four hooves of an animal) at room temperature with the Instron cross-head rate of 8.3×10⁻⁵ m s⁻¹ (corresponding to a tensile strain rate of 2.0×10⁻³±0.3×10⁻³ s⁻¹; mean ±1 S.D.). Yield strengths were determined using the offset method suggested by ASTM standard E8M-94a (ASTM, 1994a); in our tests, an offset strain of 0.5 % was arbitrarily chosen. After mechanical testing, water contents were determined. Although specimens were tested at 100 % relative humidity (RH), bulk water may accumulate in the medullary cavity of tubules and distort water content measurements. Therefore, specimens were dehydrated slightly in a 97 % RH environmental chamber before weighing to ensure that no bulk water was present (i.e. water content measurements are for tissue at 97 % RH). Samples were then dehydrated at 100 °C for 5 days, and water content was calculated as (wet mass minus dry mass)/dry mass.

Specimens used for microscopy were taken from the lateral toe region of the right front foot of three horses. The wall was arbitrarily sectioned distally into three levels of equal length: a, b and c (Fig. 2). It was also divided radially (through the wall thickness) into six arbitrary regions of equal thickness: Ia, Ib, IIa, IIb, IIIa and IIIb. Strips of tissue were sectioned perpendicular to the tubule longitudinal axis (i.e perpendicular to the outer wall surface) approximately 5–8 μm thick using a microtome. All sections for light microscopy were observed without staining.

The axes as defined in this study are illustrated in Fig. 2: the stationary, longitudinal tubule axis L, and two non-stationary orthogonal axes, the radial axis R, and the tangential or circumferential axis, C. Tubules run along the length of the wall, parallel to the outer surface. Since the wall tilts back by approximately 40 ° from vertical in the toe region (of the front foot), the L axis here lies approximately 40 ° from vertical (Fig. 2D). The R axis represents the radial axis of the foot, running perpendicular to the outer surface of the wall, and is non-stationary since there is more than one radial axis; the C axis lies tangential to the outer surface of the wall (and is therefore also non-stationary). The azimuthal angle, θ, is measured from the R–L plane and is in the plane of section (C–R plane). If observed along the R axis from the inside to the outside of the wall, fibers oriented to the left of the R–L plane are considered positive and fibers angled to the right are considered negative. In this manner, fiber orientations from −90 ° to +90 ° encompass all possible values of θ. An angle ϕ defines the deviation from the L axis. When observed radially from the inside to the outside of the wall, fibers tilting upwards are defined as negative whereas those tilting downwards are positive; the C–R plane is at ϕ=90 ° (a 1 ° slope upwards from this plane is −89 °, 1 ° downwards is 89 °; refer to Fig. 2B).

A universal rotating stage is a high-precision accessory for light microscopes equipped with plane-polarized light optics and is used to determine the three-dimensional orientation of ordered molecules without requiring sectioning in orthogonal planes. Positioning was determined by rotating a specimen within and out of the plane of the microscope stage to find the angle of extinction. This occurred when the primary axis of ordered molecules was oriented parallel to the microscope axis. In our study, a Leitz UT 5 universal stage (angular resolution ±1 °) was mounted to the rotating stage of a Leitz Orthoplan-Pol polarizing microscope (see Canham et al. 1991). Extinction angles were determined on specimens magnified 400 times, allowing resolution of IF orientation in areas of approximately 1 μm².

Plane-polarized light is useful in determining the axial alignment of ordered molecules; however, there are drawbacks to using this illumination technique. Regions where molecules are aligned parallel or perpendicular to the plane of polarization appear dark. Therefore, to produce images that show only the orientation of molecules relative to the plane of section, photographs of hoof wall samples were taken using circularly polarized light. This technique produced images in which the lightest areas indicate molecules close to the plane of section and darker areas show molecules oriented more towards the axis of the microscope; areas which were not birefringent (such as the background) appeared black using both polarized and non-polarized light optics.

Pixilated images were produced by scanning photographic negatives of histological samples with a Kodak RFS 2035 Plus film scanner at 1000 dots per inch (d.p.i.). Digital images were processed using a Macintosh PC and arranged using Adobe Photoshop 3.0 software. One image of the entire SM and stratum internum (SI) was obtained by producing a 20 μm thick specimen that was stained with Toluidine Blue (in distilled water with 1 % sodium borate) for 3 min, placed in 100 % ethanol overnight and then mounted in a projector slide. The slide was then scanned at 2000 d.p.i. with the Kodak scanner and the image was processed by the PC (see Figs 2C, 4A).

Tubule dimensions were measured from hoof wall specimens sectioned along the C–R plane, by digitizing pixilated images with a PC video frame-capturing program (V for Windows) and then processing point coordinates using a spreadsheet. Images were generated using a video camera (Panasonic model WV-BL200) mounted on the polarizing microscope and interfaced with a Matrox PIP-1024 frame grabber. Tubules were approximated as elliptical in cross section, so that only major and minor axes dimensions were necessary to estimate cross-sectional area. The dimensions of tubules from different regions were tested for significant differences using an analysis of variance (ANOVA).

Circularly polarized light optics proved very useful in characterizing both general and specific aspects of hoof wall morphology. This illumination technique revealed not only gross morphological changes through the hoof wall thickness, but also fine-scale design at the cellular level. Fig. 3 is an image of a non-stained cross-sectional sample from region IIIa illuminated using this technique. Medullary cavities appeared dark in the centers of the tubules and were usually devoid of cellular material, but occasionally they contained irregularly positioned cellular matter incapable of supporting load. Overlaid in Fig. 3 are tracings of cell boundaries identified from a second image of the same section illuminated using bright-field optics. Inner, middle and outer cortical lamellae types are labeled as a, b and c, respectively. There are clearly differences in shape and size between the cells of the tubule cortex and those from the intertubular material. These differences are probably due both to the orientation of the pancake-shaped cells within the wall and to slight changes in cell morphology. Also visible is the intimate relationship between the cells of the tubule and the intertubular material. Note the brightness of the intertubular cells above and below the tubules (indicated by asterisks in Fig. 3) contrasted with the mottled appearance of the intertubular material on either side (indicating a less regular pattern of IF organization). Tubule morphology, intertubular material organization and the relationship between the tubule and intertubular material were dependent on the position through the wall thickness. Complexity and variability in design, however, were best exemplified in tubule morphology.

Cells of the tubular cortex were organized generally into concentrically arranged lamellae, where each lamella was composed of one layer of cells (see Fig. 3). In cross-sectional samples viewed under circularly polarized light, these lamellae were characterized as one of three types: bright, innermost cortical lamellae which were usually one or two cell layers thick (a in Fig. 3), more numerous middle lamellae which appeared dark (b in Fig. 3), and outer, bright cortical lamellae consisting of approximately 2–3 cell layers (c in Fig. 3). In Fig. 3, the inner cortex has two lamellae, the middle cortex has approximately five lamellae and the outer cortex is composed of approximately four lamellae. To avoid confusion, the inner, middle and outer tubule lamellae will be referred to as the inner, middle and outer type, thereby distinguishing these areas of the tubule cortex from the inner, middle and outer regions of the SM. An outer cortex was not clearly distinguishable in tubules from regions Ia and Ib. Although this outer cortex appeared to be present in these regions, its extensive association with the adjacent intertubular material implied that this material was not generated by dermal papillae, but instead was probably formed by cells at the coronary border.

Cortical lamellae differed not only in their appearance under circularly polarized light, but also in cross-sectional thickness (Table 1). Inner-type mean lamellar thickness ranged from 5.1 to 7.5 μm, depending on wall region, and showed statistical differences in thickness between tubules through the thickness of the wall (P<0.001, ANOVA). The thicknesses of middle-(9.9 μm) and outer-type (13.5 μm) lamellae were not statistically different between tubules from different regions (P>0.05, ANOVA).

Although lamellar thicknesses were generally constant, tubule size was dependent on position in the wall. This resulted from a change in the number of cortical lamellae in tubules from different regions (Table 2). Tubules from region Ib were largest, averaging 0.046 mm² in whole-tubule (i.e. cortex + medullary cavity) cross-sectional area; the smallest tubules averaged 0.018 mm² and were found in region IIa near the middle of the wall. Tubule shape was also dependent on the position through the wall thickness. A weak trend towards relative widening along the C axis was evident within the tubule cortex of one animal, whereas the medullary cavity consistently increased along this axis from 37.1 μm in the innermost region to 69.0 μm in the outermost region of another animal (Table 3). Innermost tubules from region Ia were the most circular in cross section, with an average major axis to minor axis ratio of 1.15 (Table 3). Progressing outwards, tubules became more elliptical in cross section; the outermost IIIb tubules had an average ratio of 1.57. Note, however, that the axis ratio is higher than the C/R ratio in tubules from regions Ia, Ib and IIIb. This indicates that the major axis of a tubule was occasionally oriented radially, not circumferentially. In regions where the values were equal, tubules were always oriented with their major axis oriented circumferentially.

Tubule morphology is summarized in Fig. 4. Fig. 4A is a photograph of a Toluidine-Blue-stained section of the entire hoof wall, including the lamellar stratum internum. Gross changes in tubule morphology are visible, as are the differential staining patterns of each hoof wall region. Higher-magnification images produced using circularly polarized light further indicate the dependence of tubule morphology on position through the wall thickness (Fig. 4B). To the right of these images are illustrations based on average cross-sectional dimensions and the shape of tubules from each region (Fig. 4C). Molecular alignment in each lamellar type, relative tubule sizes and shapes are generalized in three-dimensional tubule renderings in Fig. 4D.

To show inter-animal variability, cross-sectional areas for two animals are plotted in Fig. 5A; these curves revealed similar trends. Tubule density (Fig. 5B) increased radially from approximately 10 to 25 tubules mm⁻². The ratio of cortex to medulla cross-sectional area shows that, although tubules in some regions may take up considerable space, the medullary cavity contributes more of the tubule space progressing outwards (Fig. 5C). Therefore, in order to determine the material contributed by the tubule cortex to the tissue, a ratio of cortex cross-sectional area to total SM area (discounting the medullary space) was calculated. A U-shaped curve resulted, with tubule cortex area as a proportion of total SM area starting near 50 % in the inner region, declining to 25–30 % in the middle regions and increasing to over 50 % in the outer region (Fig. 5D). This suggests that tubular material plays a large role in determining the mechanical properties of the outer and inner regions of the wall, whereas intertubular material probably dominates the mechanics of the mid-wall.

The entire SM of the equine hoof wall was intensely birefringent, indicating a high degree of molecular order (see Fig. 4B). IF alignment in hoof tubules did not have a significant radial component (with respect to an individual tubule), so that the ‘pine-cone’ arrangement of cells (and therefore, of IFs) noted by Wilkens (1964) in bovine hoof tubules does not exist in the equine hoof wall. Rather, the lamellar-like design offered by Nickel (1938a,b) is more appropriate. Tubule cortex IFs were usually aligned in the cell plane and wound helically around the central medullary cavity. Since a helical formation was generally found in lamellae of tubules from all regions, the term helical angle (measured from the tubule axis) will be used in place of ϕ. It should be noted, however, that in lamellae with very small ϕ angles, regular helices were not always evident. In addition, helices were observed to change direction through the thickness of a single cell (see arrow in Fig. 3), such that in one instance a single cell contained three helices with complementary angles. IF helical orientation also changed periodically around a lamella and continued across adjacent lamellae. These findings suggest that, although cellular lamellae are an obvious morphological scale of order, the lamellar structure may also exist at a smaller scale, at the level of IF organization within cells. In general, helices of adjacent cells of inner-type and outer-type lamellae were crossed (Fig. 4D).

Tubule IF helical angles changed progressively from the inner Ia region to the outer IIIb region of the wall (Table 4; Fig. 4D). Throughout most of the wall, this change occurred as a gradual morphological gradient. Tubules from regions Ia and Ib were similar in IF orientation. Often, inner-type lamellae of Ia tubules had cross-helical IF orientations, usually with helical angles between 40 ° and 60 ° (the innermost lamella usually had a right-handed helix). Adjacent lamellae from middle-type lamellae of tubules from regions Ia and Ib usually did not cross; however, crossing was observed in these lamellae in tubules from outer regions. Tubules from outer regions were characterized by crossed helices of all adjacent lamellae. Cells forming middle-type lamellae of region Ia and Ib tubules were not as flattened as those of inner- and outer-type lamellae of these tubules, or as those of middle-type lamellae of tubules from other regions. Helical angles of middle-type lamellae in regions Ia and Ib ranged from 0 ° to 12 °, and adjacent lamellae were usually wound in register with a right-handed (RH) helix (see Fig. 4D); left-handed (LH), in-register winding of adjacent lamellae was not observed.

A particularly abrupt transition from one tubule type to another was evident in the zone between Ib and IIa (see Fig. 4A). This transition region has been previously recognized and named the intermediate zone (Leach, 1980). Here, wall dominance changed from tubules that primarily reinforced the L axis to intertubular material. Vast tubule size and overall design changes occurred over such a short distance that the contrast could be observed in stained sections with the unaided eye. Samples from all other animals showed this same pattern, suggesting a unique function for this region of the wall. Helical angles of the inner-type tubule lamella of region IIa were similar to those of Ia and Ib; however, both LH and RH helices were observed. IIa tubules had reduced numbers of middle-type lamellae. In a section from one animal, helical angles in these lamellae ranged from 0 ° to 33 °. In region IIa, distinct outer-type lamellae became apparent, although middle-type lamellae were still predominant. Similar helical angles were observed in region IIb (Table 4).

A second major transition was apparent in region IIIb. Here, tubules were characterized by increased numbers of inner- and outer-type cortical lamellae (Table 2). These lamellae retained the same helical orientations as tubules from more inner regions, but helical angles in middle-type lamellae in tubules from region IIIb appeared greater (from 0 ° to approximately 50 °). Here again, although helices of adjacent lamellae tended to cross, no consistent pattern was observed. As with tubules from more inner regions, helical angles were quite variable between samples and amongst animals.

Intertubular IFs from specimens sectioned in the C–R plane formed a static pattern which resembled a ‘flow’ around the wall circumference, with tubules acting in a manner analogous to pillars, obstructing the ‘flow’ and causing ‘turbulence’. In contrast to a normal flow, this ‘turbulence’ is formed immediately behind and in front (with respect to the flow) of the tubules (refer to Figs 3, 4B). In these areas of ‘turbulence’, IF organization was unpredictable and appeared as areas of mixed light and dark patterns under circularly polarized light, whereas IF organization in areas outside these turbulent zones was consistent. As with cells of the tubule cortex, IFs lay parallel to the cell plane of the intertubular material. However, most of the SM cells of the intertubular material lay in a plane at a large angle relative to the tubule axis, and this angle varied through the wall thickness.

The intertubular IF pattern of organization is summarized in Fig. 6. The top photograph (Fig. 6A) is a cross section (C–R plane) of the entire hoof wall SM illuminated with circularly polarized light; a first-order red filter provided coloration. Areas with fibers generally aligned radially appear yellow and those running circumferentially are illuminated blue (non-birefringent areas, such as the background, and areas with fibers oriented longitudinally appear purple). Fig. 6B,C shows three-dimensional renderings of the intertubular material, based on a tracing of a portion of Fig. 4A. Fig. 6D shows a longitudinal section (R–L plane) of the SM, illuminated as in Fig. 6A. Here, areas with fibers oriented approximately radially are yellow and those running approximately longitudinally appear blue (non-birefringent areas and areas with fibers oriented circumferentially appear purple). Intertubular organization is clearly discernible in the cross section (Fig. 6A). In the innermost Ia region, intertubular IFs tended to follow the longitudinal axis of cells and generally formed a concentric path around the tubules. Note that this region appeared mostly blue, indicating that more fibers were running circumferentially than radially. In the adjacent Ib region, radial and circumferential orientations became apparent, with the presence of blue and yellow areas. Intertubular cortex of the middle (IIa and IIb) regions accounted for approximately 70 % of the total hoof wall material in that region. Intertubular cells immediately adjacent to tubule cortex still conformed to the concentric orientation, but in the large intertubular space, a predominantly circumferential IF alignment was apparent (note the predominance of blue in this region). This dominance of circumferentially aligned fibers continued towards the outermost regions (IIIa, IIIb), although overall birefringence appeared to decrease as intertubular IFs became aligned further and further out of the C–R plane.

The innermost Ia region was the only part of the intertubular SM that lay predictably in or close to the C–R plane (ϕ=−83 °; N=38); most of the intertubular material of the SM lay at an angle to the C–R plane, as illustrated in Fig. 6C. Middle regions showed planar intertubular IF angles decreasing (approaching the tubule axis) outwardly, ranging from ϕ=−81 ° (IIa; N=48) to ϕ=−62 ° (IIb; N=58; see Fig. 6D). These regularly aligned intertubular IFs of the middle region comprise approximately 70 % of the mid-wall SM (Fig. 5D) and lie in planes like unidirectional ‘mats’ which are stacked and tilted relative to the tubule axis such that, when a foot is placed on the ground, the ‘mat’ planes would orient nearly vertically. Within each mat, IFs were generally aligned circumferentially (around the hoof) within these planes and appeared to flow around tubules. This arrangement is analogous to a forest situated on a surface that is predominantly tilted at approximately 60 °; the lower side represents the inside of the wall and the higher side the outside of the wall. The lower side tapers off to a flat surface perpendicular to the tree axis, and near the end of the upper side it slopes quickly higher to almost 90 °. Now imagine a current of a fibrous liquid defying gravity and flowing along the slope (not down the slope); this flow direction would be circumferential with respect to the hoof wall. The flow is generally laminar, but turbulence develops in front of and behind the trees (with respect to the flow direction). If one of these layers represented a cell plane in the wall, then numerous layers of ‘currents’ would be stacked on top of one another. These intertubular ‘mats’ form the crack propagation barrier whose function was first recognized by Bertram and Gosline (1986). Towards the periphery of the wall, intertubular material was aligned close to the tubule L axis; although this is not easily seen in Fig. 6D, it is visible as the progressive change to purple towards the outer wall in Fig. 6A. This progression is also seen as the darkening of the intertubular material towards the outermost region in Fig. 4B. In one section, ϕ changed from −55 ° (N=64) in region IIIa to −38 ° (N=45) in IIIb and rapidly approached the tubule axis in the outer part of region IIIb. The pattern of intertubular IF orientation close to the tubule axis and the prevalence of tubules (Fig. 5D) in the outer wall suggest that this region inhibits inward crack propagation across the tubule axis.

In contrast to all other areas of the SM, intertubular IF planar alignment in region Ib was highly variable and unpredictable. Here, although uniformity in planar IF angles was observed occasionally in localized areas (see top of region Ib in Fig. 6D), angles were measured through a 180 ° arc (N=46). Therefore, no mean intertubular IF plane angle was assigned to this region. However, to show relative angular variability, standard deviations are shown below each illustration in Fig. 6C. The standard deviation in region Ib was approximately seven times that of region IIIa. To ensure that this variability was not unique to this particular specimen, the observation was verified by examining sections from three different animals. In all cases, although a general concentric order around tubules was discernible (see Fig. 6A), no regular pattern of ϕ could be quantified. Random IF order, coupled with the large tubules in this region whose IFs are generally aligned longitudinally, could form an inner-wall crack diversion mechanism. This hypothesis is tested in fracture tests that follow.

Series 1 and 2 mechanical tests were performed to test functional hypotheses generated after the collection of morphological data. Series 1 mechanical tests on inner, middle and outer wall specimens (which incorporated regions Ia and Ib, IIa and IIb, and IIIa and IIIb, respectively) quantified tensile properties and tested for the existence of crack diversion mechanisms for cracks generated up the R–L plane. By incorporating more than one region (as defined in this study) in each specimen, data from CT and tensile tests represent average mechanical properties for each third of the SM thickness. Series 2 CT experiments tested the effectiveness of possible fracture control mechanisms that became apparent after morphological data and results from series 1 tests had been obtained. All CT specimens fractured stably, with the exception of several series 2 specimens which failed at a clevis hole.

Typical stress–strain curves for inner, middle and outer specimens of the hoof wall SM are shown in Fig. 7. The initial longitudinal tensile modulus or stiffness, E_i,L increased from 0.30 to 0.56 GPa progressing from the inner to the outer regions of the wall (Table 5). A multiple-comparison analysis showed a statistically significant difference in E_i,L between regions I and III (P<0.05; Student–Newman–Keuls test). Data from series 1 CT tests provided the initial modulus in the circumferential direction, E_i,C; in all regions, E_i,C was lower than E_i,L (P<0.01 for all; t-test). Note, however, that the contribution of inter-animal variability to these differences has not been determined. E_i,C of the inner region was significantly lower than that of the middle and outer regions; there was no significant difference between the middle and outer regions (P<0.05 for all; Student–Newman–Keuls test).

Beyond the initial linear portion, tensile test stress–strain curves of specimens from all regions were characterized by a ‘yield’, seen as a rapid drop in instantaneous modulus. The mean stress at which yield occurred in the outer wall was 1.5 times that in the inner wall (Table 5). After the yield region, the shapes of the curves were similar. Although care was taken to ensure that data from tests in which premature failure occurred were rejected, ultimate (maximum) data were less reliable than E_i,L, since failure could have been initiated by flaws resulting from specimen preparation; ultimate data from eight specimens were rejected as a result of failure at the grips or at the site of strain-gauge attachment. There were no significant differences in total energy, maximum stress or maximum strain between specimens from different regions of the wall (P>0.05; ANOVA).

To determine whether the differences in E_i,L could be accounted for by simple hydration effects, our initial modulus data were plotted as a function of water content (Fig. 8, filled circles) and compared with Bertram and Gosline’s (1987) data for middle-region SM tissue (Fig. 8, open circles). Although it was unspecified in the 1987 publication, the middle region was tested in the study (J. E. Bertram, personal communication). To determine water content at 100 % RH, we extrapolated data from absorption isotherms (M. A. Kasapi and J. M. Gosline, unpublished data) for each of the three regions. Estimated water contents at 100 % RH were 48 %, 41 % and 35 % for inner, middle and outer samples, respectively. Data from the present study fell very close to the regression line for Bertram and Gosline’s (1987) data, suggesting that water content alone could account for the differences in initial longitudinal stiffness.

Scanning electron micrographs of CT fracture surfaces from representative specimens are shown in Fig. 9. Fig. 9A shows the fracture surface of an inner (regions Ia and Ib) CT test specimen. A notch was introduced up the R–L plane and is visible in the lower left of the specimen (notch surfaces in all specimens are indicated by asterisks). Differences in micro structure between regions Ia and Ib (left and right sides, respectively, of Fig. 9A) caused the advancing crack to deviate in two distinct directions (towards the upper right) in this specimen; one along the tubule axis (region Ib) and the other along the intertubular IF plane (region Ia). This fracture path is seen more easily in the illustration in Fig. 10, in which each test specimen is shown in a sample location within the hoof wall and then enlarged to show detail. The upper specimen in Fig. 10B illustrates the fracture path observed in the region I sample. The specimens in Fig. 10B are illustrated in their correct positions through the wall thickness. In Fig. 9B, the notched surface of a specimen from the middle (IIa and IIb) region is barely visible on the lower right side of the micrograph. Here, the crack deviated towards the circumferential axis of the wall (towards the upper left-hand side of the micrograph) and also began to follow the forward slope of the intertubular IF plane (refer to Fig. 6), passing through tubules as it progressed. This crack reorientation is seen as a twist of the fracture surface (illustrated for clarity in Fig. 10B) and results from the mid-wall crack diversion mechanism described previously. Fig. 9C shows the fracture surface of an outer hoof wall specimen (regions IIIa and IIIb). Here, the crack propagated from the upper left to lower right of the specimen, along the tubule axis (see Fig. 10B). Although region IIIa was structurally similar to the middle region, the strong L axis orientation of the outer (IIIb) tubular and intertubular components caused the crack to progress along the favored path parallel to the tubule axis. In general, regions with a high proportion of tubules showed fracture paths which followed the tubule axis (see Fig. 9A,C), whereas cracks initiated in regions dominated by intertubular material primarily followed the intertubular IF plane (Fig. 9A,B).

The J-integral parameter suggested decreasing toughness progressing outwardly (Table 6). The inner region of the hoof wall showed a statistically significant higher toughness than the middle and outer regions (Student–Neuman–Keuls test, P<0.05). Using the stress intensity factor K as a measure of fracture toughness, the inner and outer regions were statistically similar, but the middle region of the wall was significantly tougher than the inner and outer regions (Student–Newman–Keuls test; P<0.05). The issue of which parameter is better for characterizing fracture toughness has been discussed in a previous paper (Kasapi and Gosline, 1996) and will not be dealt with here; however, a point worth noting is that no decline in fracture toughness is evident with the gradual softening of the tissue towards the inner regions of the wall. Furthermore, differences between the toughest and weakest regions of the wall using J and K are relatively small (29 and 31 %, respectively). These differences may not exist at in situ hydration levels since, if an optimal hydration level for fracture toughness exists for the outer and inner regions (as it does for the middle region; Bertram and Gosline, 1987), then the in situ J value for each region could be higher and more similar.

Data from series 2 CT tests also provided initial modulus values for the middle region in the circumferential (E_i,C) and radial (Ei,R) directions. Ei,C of the middle region from series 2 CT tests (0.38±0.02 GPa; mean ±1 S.E.M.) was significantly higher (P<0.01; t-test) than that obtained from series 1 CT samples (see Table 5). This implies that inter-animal variability could contribute significantly to the differences between E_i,C and E_i,L observed in series 1 tests. E_i,R (0.23±0.01 GPa; mean ±1 S.E.M.) was significantly lower than E_i,C (P<0.0001; t-test) and is probably the effect of the circumferential alignment of intertubular IFs in this region. Inter-animal variability is not a factor here, since both parameters were obtained from the same hoof.

Representative samples from series 2 CT tests are also shown in Fig. 9D–F. Cracks in seven of the twelve CT specimens notched in the middle region up the C–L plane (series 2A) were redirected, as expected by the mid-wall crack diversion mechanism, along the intertubular IF plane towards the outer wall surface (Fig. 9D; Fig. 10C, bottom). At the outermost wall, cracks continued along the intertubular IF plane, deviating closer to the tubule axis as the crack progressed further outwards. In two specimens, the crack continued generally in the original notch plane, but showed a tendency to deviate along the intertubular IF plane. Three specimens failed at the clevis hole in the outer wall, with cracks that followed the intertubular IF plane and indicated a considerable weakness of the outer wall relative to the mid wall, along which the notch was applied. Clevis-hole failure was not observed in series 1 tests or in any of the middle-region specimens notched up the R–L plane in series 2 tests.

To test the effectiveness of the structural discontinuity at the intermediate zone and region Ib tubules in stopping cracks from propagating inwards across tubules, 11 specimens were produced that spanned the entire SM thickness and were notched inwards along the C–R plane (series 2B). Four samples failed along the intertubular IF plane at a clevis hole in region III. In all the others, the notch was redirected downwards. In specimens with short notches (through the outer and middle regions), the crack deviated downwards and inwards along the intertubular IF plane until it met the large inner tubules of region Ia. Upon encountering these structures, the crack then continued straight downwards along the tubule axis. In specimens with long notches that entered region Ib, cracks were immediately redirected perpendicular to the notch surface along the tubule axis (see Figs 9E, 10C). These results clearly establish the existence of an inner-wall crack diversion mechanism.

The effectiveness of the mid-wall crack diversion mechanism in redirecting cracks initiated inwards along the R–L plane (series 2C) was tested by notching 12 specimens from the lateral toe region in this plane (see Fig. 10D). In specimens in which the notch front terminated in the middle regions, the crack was redirected circumferentially (following the predominant, intertubular IF ‘grain’ of the middle region), twisting into the intertubular IF plane (see Fig. 6A). All of these specimens were cut to the right (lateral) of the mid-line, and in all cases cracks were redirected towards the back of the hoof. In specimens with notches that extended into the inner region, cracks were only weakly redirected. Fig. 9F shows a micrograph of a fracture surface, and an illustration is also provided in Fig. 10D. In four specimens, failure occurred at the clevis holes in region III by crack propagation between tubules.

J and K values obtained from the middle region of series 2 CT samples notched up the R–L plane were statistically similar to those obtained from similar specimens in the series 1 tests (see Table 6), indicating that inter-animal variability was not a factor affecting this parameter and that toughness is not compromised by storing specimens in this manner over a period of 3 months. J values from series 2 specimens notched in the C–L plane (in the middle region) were significantly higher (Table 6; P<0.001; t-test). Results for the stress intensity factor K, however, suggested that these specimens were less tough than specimens notched in the R–L plane. Fracture data for specimens notched inwards along the C–R and R–L planes (Fig. 9E,F, respectively) are not presented because in these tests fracture toughness was dependent on notch length (since cracks were initiated in different regions, depending on the original notch length). This violated a critical assumption necessary for determining a single, representative value of fracture toughness.

The primary mechanical function of the equine hoof wall is to transfer ground-reaction loads to the bony skeleton. While doing so, it must resist the formation and propagation of cracks yet also allow for the necessary process of wear. The transfer of loads is a relatively simple task; fracture control and wear management are more complex. Although load transfer and wear management are functionally important issues, fracture control appears to be the major driving force in the development of morphological complexity in this tissue. Therefore, the following discussion considers all issues, but concentrates most heavily on crack control.

To act effectively as a load transfer element, the hoof wall must be stiff enough routinely to withstand large loads without undergoing large-scale or irreversible deformation. It has been documented here and in other studies (Leach, 1980; Douglas et al. 1996) that a gradient of stiffness exists through the wall thickness. This gradient is due to the proximity of the different regions to the source of moisture (the vascularized tissues adjacent to the stratum internum and coronary border) and, as seen in this study, to the properties of the keratin in each region.

Although the hoof wall must be stiff enough to support incurred loads, differential mechanical demands on the tissue through the wall thickness may require particular mechanical properties in specific regions. The outer wall probably encounters a variety of mechanical insults from many directions. To resist crack initiation and minimize abrasion, this region must be both strong and stiff (i.e. hard). Loading of the inner wall, however, is more predictable since loads become diffuse before reaching this region. Here, loads are ultimately transferred (primarily by shear and tension) to the collagenous, dermal suspension system, which links the hoof wall to the bony skeleton. To minimize the strain differential (and potentially high stresses) which results from an abrupt transition in modulus from one load-bearing element to another, the two stiffnesses must be similar. Consequently, the inner wall tissue is highly hydrated, to approach the stiffness of the dermis. A stiffness gradient is also necessary through the hoof wall thickness, because high stresses would result at the interface between the stiff outer wall and the soft inner wall. Note that our stiffness values for fully hydrated specimens are underestimates of the true in situ stiffnesses. Bertram and Gosline (1987) noted a 36-fold increase in E_i,L from fully hydrated to completely dehydrated specimens of middle equine hoof wall. To appreciate the magnitude of this stiffness gradient fully, we may estimate the in situ wall stiffnesses.

Since specimens from regions I, II and III placed in environments with the same relative humidity differ in water content, the proportion of IFs to IF-associated proteins (non-ordered molecules) changes from one region to the next, the protein constituents may vary, or both of these may occur. Although protein constituents vary between the SI and SM (Grosenbaugh and Hood, 1992), it is unknown whether these differences exist through the SM thickness.

Fig. 8 suggests that differences in E_i,L are due to water content and not to morphological or biochemical differences. Assuming that this is true, we may use estimates of water content from Leach’s (1980) data and the curve from Bertram and Gosline’s (1987) data to predict the in situ E_i,L. From Leach (1980), water content estimates for inner, middle and outer specimens are 33.1 %, 23.5 % and 13.9 %, respectively. Using the equation from Bertram and Gosline’s (1987) data, E_i,L will increase from 0.30 to 0.66 GPa (a 2.2-fold increase), from 0.43 to 1.2 GPa (a 2.8-fold increase) and from 0.56 to 3.0 GPa (a 5.4-fold increase) for inner, middle and outer regions, respectively. This predicts an in situ stiffness differential of approximately 2.3 GPa through the wall thickness.

Locally, IFs appear to function by increasing material stiffness along their axis of orientation. In the middle wall region, which is dominated by circumferentially aligned intertubular IFs, circumferential stiffness (0.38 GPa) is significantly higher than radial stiffness (0.23 GPa). However, through the wall thickness, IF alignment does not correlate with stiffness. The inner region, which has a large proportion of SM cortex with IFs aligned nearly parallel to the L axis, should be considerably stiffer than the middle region, which is dominated by intertubular material with IFs aligned at relatively large angles to the L axis (large ϕ). This is not the case. Our results show that the innermost third of the SM is almost half as stiff as the outer third at the same RH and would probably be equally stiff at the same water content (refer to Fig. 8). In addition, although ultimate properties are also expected to correlate with IF alignment, no significant differences were found in maximum stress and maximum strain between regions of the wall (refer to Table 5). Yield stress data further show that IF orientation and mechanical properties are unrelated in the wall; the yield stress of the inner region, which had a high proportion of IFs aligned close to the axis of applied stress, was lower than that of the middle region, which was dominated by intertubular material with IFs aligned at a steep angle to the stress axis.

It may therefore be concluded that the primary function of IFs is not to increase stiffness in the axis of orientation. This has major implications for the tubular components of the wall whose primary mechanical role was believed to involve offering reinforcement along the wall L axis (Nickel, 1938a,b; Bertram and Gosline, 1986). The functions of tubules and the role of IF orientation within both the tubule cortex and intertubular material are more complex than previously imagined. Our results suggest that these elements are not expressions of demands for rigidity in the hoof wall, but rather they arise from the need to control fracture processes and increase fracture toughness.

It should first be noted that tubules are also present in other keratinous, non-homologous (developmentally non-related) structures such as rhinoceros horn (Ryder, 1962) and bovine horn (Kitchener, 1987; Kitchener and Vincent, 1987), but are absent in homologous structures such as human nail (Hashimoto, 1971a,b). Although the exact functions of tubules are still under debate, current dogma holds that they facilitate hydration of the wall (Evans et al. 1990) and/or provide mechanical reinforcement along the wall length.

Kitchener (1987) suggested that tubules in bovine horn have functions analogous to fibers of composites. This analogy may not be appropriate for horn or hoof wall. The fibers of composites are usually stiffer than the matrix and provide rigidity and strength to the composite along the axis of orientation. The matrix is normally made from a more flexible material and aids in transferring stresses to the fibers via shear. In horn and hoof, however, there is no indication of a difference in stiffness between tubules and the intertubular ‘matrix’.

Tubule ‘pull-out’ observed in fracture tests (see Fig. 9) indicates that tubules reinforce the hoof wall along the longitudinal axis to some extent. Even as cracks produced in the middle (IIa and IIb) region find a new route along the intertubular IF plane (Fig. 9B), tubules (which form approximately 30 % of the material in this region) retard crack propagation by acting as mechanical barriers. These reinforcements offer just enough resistance to cause the advancing crack to deviate part-way along its surface (and between tubular lamellae), thereby dissipating energy and retarding crack growth in the process (see Broek, 1986). But tubules do not just reinforce along the tubule axis. Tubules from different regions are morphologically distinct, differing in IF helical angles and in the quantity of lamellae with similar helical angles.

Nickel (1938b) recognized the morphological gradient of equine hoof wall tubules, but simplified the wall model by characterizing tubules into one of two extreme classes: those with predominantly ‘steeply-angled spirals’ (low ϕ) and those with predominantly ‘low-angled spirals’ (high ϕ). Illustrations of tubules with ‘spirals at high angles’ and those with ‘spirals at low angles’ correspond best to tubules from regions Ia and IIIb, respectively (see Fig. 4D). Our study generally agrees with Nickel’s (1938b). Unfortunately, by representing averages of morphologies from more than one of the regions investigated here, they are not directly comparable. In our study, adjacent lamellae from similar tubule cortex zones also had alternate IF helical directions. However, unlike Nickel (1938b), we did not observe alternating helices in the middle-type cortical zone of inner tubules (regions Ia and Ib); rather, fibers in lamellae of this zone were in register and aligned to the right. An outer cortical zone for these tubules was also not defined in the present study (hence the term ‘middle-type’ for the outer cortical zone of these tubules), as it was impossible to locate a distinct tubule–intertubular material interface in these inner regions. Note, however, that tubule-forming papillae are continuous with the coronary epidermis that is responsible for generating intertubular cortex. It is therefore not surprising to find regions where a distinction between the two wall components is not clearly evident, suggesting a tight coupling between the two wall components in this region.

A satisfactory explanation of tubule function must also rationalize the complexity of tubule design. In this study, we have arbitrarily divided the wall into six regions through the thickness of the tissue and have offered illustrations of representative tubules by averaging the morphological dimensions of tubules within each region. Recall that a gradient of tubule morphology exists and that no abrupt morphological transitions exist in the equine hoof wall. For convenience, we shall refer to an average or ‘typical’ tubule from each region.

Generally, tubules from all regions share two elements: a medullary cavity incapable of supporting loads, and at least one abrupt transition from lamellae with low helical angles to those with high helical angles. This transition may play an analogous role to the tubules in the intertubular material by leading the crack through a more tortuous route and absorbing fracture energy in the process. The alternation of helical angles between lamellae of similar cortical zones (recall that there are three tubule cortical zones: inner, middle and outer) may also serve the same purpose. Note that lamellae from middle cortical zones of tubules from regions Ia, Ib and IIa do not alternate like those from other tubules and other cortical zones. The reason for this is unclear. Kitchener (1987) also found extensive delamination of the layers of keratinocytes after fracture, implying that relatively weak lamellar interfaces may serve as energy-dissipation and crack-blunting mechanisms.

Progressing outwards, the intertubular IF plane changes from approximately perpendicular to nearly parallel to the tubule axis. In contrast, tubules progress from structures dominated by lamellae with low helical angles to those with primarily high helical angles, so that most tubular IFs lie approximately perpendicular to the intertubular IF axis. This could create extensive strain-transition interfaces that will promote separation of tubules from the intertubular material during fracture (tubule pull-out) and absorb energy in the process.

The elliptical cross-sectional shape of outer tubules may be a response to the bending mechanics observed in the hoof wall. It has been observed that expansion at the heels occurs during loading of the equine hoof wall (Lungwitz and Adams, 1966). In combination with loading along the tubule axis, compressive loading from this expansion translates into a particular pattern of surface strains which includes large circumferential surface compression at the hoof toe (Thomason et al. 1992). Orientation of the major ellipse axis of these outer tubules along the circumference of the wall could offer tubules increased resistance to collapse along this axis.

Equine hoof wall is clearly not designed as a simple, hollow fiber-reinforced composite. Differentiated tubules and the changing planar orientation of intertubular material signify changing mechanical demands through the wall thickness. But what are these mechanical demands? Unfortunately, the specific demands are still unknown. However, the following discussion considers the morphological features of the hoof wall that control the growth of cracks initiated at the ground-contact surface propagating up the wall and at the outer surface propagating inwards.

The most extreme mechanical demands placed on the hoof wall are loads generated at the distal ground-contact surface upwards towards the coronary border. Whereas on flat surfaces these loads pose no threat to the structure, very high stresses resulting from localized loading (such as stepping on a small rock) during high-speed locomotion pose serious threats of injury to the animal. Any crack that contacts the vascularized (dermal) tissue will create the potential for infection, subsequent lameness and possible death. Bertram and Gosline (1986) showed that the intertubular material creates a mid-wall crack diversion mechanism that prevents cracks from propagating up the wall, and they suggested that this mechanism would reorient cracks into the plane of the ground-contact surface, thus facilitating the necessary process of wear. Our morphological and fracture results for the mid-wall region confirm the existence of this mechanism. We observed that cracks initiated up the hoof wall are redirected parallel to intertubular IFs, and this redirection probably occurs because the intertubular material occupies 70–80 % of the load-bearing cross-sectional area in the middle SM (Fig. 5D). However, our results clearly indicate that the intertubular IF plane and the direction of crack propagation run upwards towards the outer surface of the hoof, not parallel to the ground-contact surface. Since the intertubular material slopes at approximately ϕ=−60 °, and the hoof wall at the toe region of forehooves is sloped backwards by approximately 40 ° from vertical (Sack and Habel, 1977), the intertubular IF plane at the toe region follows a path at an angle of approximately 20 ° from vertical (Fig. 11C). This intertubular IF plane orientation is probably a direct consequence of the orientation of the generative tissue at the coronary border (see Stump, 1967; Banks, 1993; Fig. 11D).

Although the intertubular IF plane runs upwards and outwards, IF fibers are strongly aligned circumferentially within this plane (Fig. 6A). Therefore, the mid-wall crack diversion mechanism is capable of redirecting cracks initiated at the ground-contact surface in any orientation. For example, cracks initiated in the C–L plane (series 2A, Figs 9D, 10C) are always redirected along the intertubular IF plane and will, if they continue, emerge at the outer wall surface (Fig. 11C). Alternatively, cracks initiated in the R–L plane (series 1) are initially diverted circumferentially by the IF fibers, but as the crack progresses the fracture path twists towards the plane of intertubular IFs (Figs 9B, 10B). This tendency to twist redirects the circumferentially diverted cracks, forcing them to emerge at the hoof surface. Thus, irrespective of initial crack orientation, the mid-wall crack diversion mechanism redirects cracks initiated upwards between regions Ib (outer) and IIIa outwards towards the wall surface (Fig. 11C), away from the living tissues of the foot. Such crack redirection is visible in Fig. 11E. Here, a naturally formed notch shows the circumferential and forward redirection of cracks initiated in the toe region.

Potentially destructive mechanical insults may be received by the outer surface of the wall, which will tend to drive cracks inwards. As in previous studies (Nickel, 1938a; Bolliger and Geyer, 1992), we found that the outer SM (IIIb) region is characterized by a high density of elliptical tubules which are surrounded by intertubular material with IFs aligned generally parallel to the tubule axis. This combination of tubules and parallel-oriented intertubular IFs appears to form an outer-wall fracture barrier that should be effective in inhibiting the inward propagation of cracks initiated perpendicular to the tubule axis (e.g. in the C–R plane). For cracks initiated inwards in a plane parallel to the tubule axis, a degree of reinforcement may be offered by the morphology and organization of the tubules. The staggering of tubules, their elliptical shape and the prevalence of tubular lamellae with fibers aligned at high ϕ may offer a degree of resistance to these cracks. Unfortunately, it was not possible to construct CT specimens to test for the existence of an outer-wall fracture barrier. However, results from series 2B and 2C tests show how the mid-wall crack diversion mechanism and other aspects of hoof wall morphology control the growth of cracks that pass into the hoof wall from the outside.

In tests which simulated the initiation of cracks propagating inwards along the R–L plane (series 2C), cracks were redirected circumferentially and twisted towards the intertubular IF plane by the mid-wall crack diversion mechanism (Figs 9F, 10D). The predominantly circumferential orientation of the mid-wall intertubular IFs effectively eliminates the possibility of cracks initiated from the outside in the R–L plane from reaching the inner living tissues of the foot. In addition, cracks were always redirected laterally towards the back of the hoof (recall that series 2C specimens were taken lateral to the hoof mid-line).

The results of tests which simulated the initiation of cracks propagating inwards along the C–R plane (series 2B) were dependent on the initial notch length. In specimens with notches that terminated in the mid-wall (region II), crack growth progressed downwards along the intertubular IF plane. Upon encountering the morphological discontinuity at the Ib–IIa boundary, these cracks were abruptly redirected downwards along tubules (towards the ground-contact surface). In specimens with notches that terminated in the inner wall, cracks always progressed downwards along the tubule axis (Figs 9E, 10C). This behavior indicates a new fracture control strategy that we shall refer to as the inner-wall crack diversion mechanism. This mechanism is important because, without it, cracks initiated from the outside in the C–R plane could propagate downwards and inwards along the intertubular IF plane (see Fig. 11C,D) and enter the living tissues of the foot. An inadvertent consequence of this mechanism is that cracks in this region along the L axis will tend to propagate up towards the coronary border (Fig. 9A). This predisposition, however, is countered by the propensity for circumferential redirection of cracks by the mid-wall crack diversion mechanism.

Typical uniaxially aligned synthetic fiber-reinforced composites strongly resist crack growth perpendicular to the fiber axis, but they allow relatively easy crack propagation along the fiber axis. To resist crack growth in more than one direction, sheets or laminae may be adhered such that the fiber axes of adjacent laminae are crossed. Alternatively, fibers may be incorporated in a random orientation throughout the matrix; this offers equal fracture toughness in any direction, but compromises toughness in any single direction. In the hoof wall, where the potential directions of crack initiation are numerous but predictable, evolution has apparently produced a complex composite tissue with highly organized, specialized crack diversion mechanisms to control cracks initiated in specific directions. These mechanisms involve the incorporation of planes of relative weakness, which redirect cracks from an initial dangerous route to a more benign path. The predictability of these fracture paths ensures a particular mode of failure depending on the initial crack orientation, without seriously compromising the mechanical integrity of the structure. This design strategy works, however, only if crack propagation along these planes requires a high energy cost which is reflected in high fracture toughnesses.

Recall that, in all but the outer-wall specimens of the series 1 tests, cracks were redirected away from the initial notch plane. Therefore, the J and K values presented here are not accurate representations of fracture toughness along the initial notch plane, but rather more closely reflect the toughness of the relative planes of weakness along which the cracks were redirected (Table 6). It must be stressed that the favored planes of crack growth are only relative planes of weakness, and our results suggest that the energy required to propagate cracks along these planes is much higher than the fracture toughness of bone to which the hoof wall is indirectly attached (Behiri and Bonfield, 1984). It may be assumed that the resistance to crack propagation across a crack diversion mechanism (along the initial notch plane) will be much higher.

Several mechanisms are incorporated into hoof wall design which maintain high fracture toughness along a crack diversion mechanism. The mid-wall crack diversion mechanism is reinforced periodically with tubules which run through its planes of weakness, providing a more effective bond between planes. These tubules also act as small-scale crack deviation structures and further confound crack progression by directing the crack along a more tortuous route across its complex, lamellar structure. Additionally, unlike traditional homogeneous matrix phases of synthetic composites, intertubular material is formed by the adhesion of cells which are filled with a molecular composite, α-keratin. Effective bonding of these cells makes crack propagation between cells difficult, so that cracks propagate between and within cells. Propagation along the IF axis within cells is inhibited by intimate associations between the IFs and the globular IF-associated proteins. The inner-wall crack diversion mechanism is reinforced only by intertubular material whose IFs are less ordered. This offers an advancing crack minimal resistance and implies that it acts as a final safety mechanism to ensure the redirection of cracks (which have penetrated deep into the wall) away from the very sensitive mechanical junction of the wall and the skeleton.

The production of these crack diversion mechanisms is not without limitations. In hoof wall, planes of preferred crack propagation are formed by substructures whose IFs are predominantly organized in one axis or plane, and IF orientation seems to be restricted to a direction within the cell plane. Thus, crack diversion design is coupled to the surface plane of the generative tissue. This restriction may be responsible for the difference in morphology between the inner-wall crack diversion mechanism and the (hypothesized) outer-wall fracture barrier, which apparently have the same functions. These difference are probably because it is functionally impossible to produce intertubular IFs along the tubule axis in the inner region. In the outer region, this was achieved by curving the generative tissue at the outer coronary border to a plane almost parallel with the tubule axis; this design may be impossible to repeat in the inner-wall region. Instead, the inner coronary border lies at a plane approximately perpendicular to the tubule axis. All resources for reinforcing the L axis are therefore invested in the tubule cortex, which is a dominant feature of this region. In contrast to the intertubular IFs of the outer wall, inner-wall intertubular IFs are suitably disorganized (see Fig. 6C,D), since any intertubular IF order except along the tubule axis would tend to divert a crack along a plane across these tubules.

It has been known for some time that the equine hoof wall is a multi-level composite material. This study has revealed that, in addition to the advantages in increased fracture toughness resulting from its hierarchical design, mechanisms have been incorporated to control the direction of crack growth. By recruiting the intertubular component into a fracture toughness and direction control mechanism, the hoof wall has become one of the most fracture-resistant biological structures known and, upon failure, is capable of modulating crack direction along predetermined routes which are dependent on initial crack orientation.

The authors would like to thank Robert Scharein of the UBC Imager Labs for creating the computer renditions of hoof wall tubules using KnotPlot software (site available for viewing at http://www.cs.ubc.ca/nest/imager/contributions/scharein/Knot Plot.html) and two anonymous referees for their constructive comments on the manuscript. This research was conducted while M.A.K. was on a Natural Sciences and Engineering Research Council of Canada (NSERC) Post-Graduate Scholarship. This study was also supported by a NSERC grant to J.M.G.