Inter-lamellar behaviour has usually been imaged in the circumferential or axial direction, thus visualizing the lamellar cross sections. In the present work, samples were imaged in the radial direction. Visualization of inter-lamellar interface was possible because lamellae are discontinuous, i.e., they do not run in a single layer all around the disc (Marchand and Ahmed 1990) (Fig. 8). If, ad absurdum, layers were continuous, (1) no inter–lamellar interface would exist in the middle of the annulus and (2) it would not be possible to image two lamellae simultaneously in the radial direction (as the more external lamella would hide the adjacent inner lamella).
Two types of interfaces were found: a distinct separation between lamellae (Fig. 3b–d) or an interdigitation of sub-bundles (Fig. 3f). These interfaces must be very thin and correspond to a thinning of the lamella at its end, because the depth of field of the SHG set-up was 7.2 \(\upmu \hbox {m}\) (as calculated from the equation given by Squier and Müller 2001). This is quite thin compared to the lamellar thickness, which is 280 ± 90 \(\upmu \hbox {m}\) in adult humans and 157 ± 83 \(\upmu \hbox {m}\) in young cow tail (Adam et al. 2015). The ends of discontinuous lamellae, where these inter-lamellar boundaries occur, indeed tend to thin out (Adam et al. 2015), which explains why imaging these boundaries was possible. Older human lamellae are thicker than cow tail lamellae, which might be the reason why inter-lamellar interfaces were rarer in human samples: if the interface is too thick, it cannot be imaged in a single two-dimensional image.
Fibres bending and directly linking lamellae (Fig. 3d, Online Resource 1) have not been previously reported, but the existence of such stiff structures can explain why no sliding occurs at the inter-lamellar interface. Similar structure and mechanics can be seen in the Online Resource 1 of a previous work, which employed a different experimental protocol (Vergari et al. 2016). In that work, large sliding occurred between bundles of fibres in cow tail samples. In the present work as well, inter-bundle shearing (5.2% [3.6, 9.6]% median [1st, 3rd quartiles]) was significantly higher than intra-bundle shearing (3.3% [2.1, 5.6]%, \(p = 0.02\)) confirming that inter-bundle sliding is the main straining mechanisms. Bundles in human samples did not appear to slide, as confirmed quantitatively by the values of intra- and inter-bundle shearing (3.4 and 3.8%, respectively), which showed that the same shearing occurred within the bundle as between bundles.
Healthy young cow annulus and adult human annulus showed different microstructure. The most striking difference is the complete lack of fibre crimping in human annulus. Fibre crimping in the annulus has been described for cow tail, canine and human annulus (Bruehlmann et al. 2002; Cassidy et al. 1989, 1990); the study by Cassidy et al. on human annulus, however, does not report whether the included discs were healthy. Therefore, it is possible that the lack of crimps is an effect of disc degeneration. Tendons show crimped structure of collagen fibres, aligned with its main axis. Crimp angle at rest can decrease with age and exercise in equine superficial digital flexor tendon (Patterson-Kane et al. 1997a, b). However, only a small reduction in angle was reported, not a complete loss of crimping, and in human Achilles tendon no effect of rupture was observed on crimp angle, apart from a straightening at the site of the rupture. (Magnusson et al. 2002; Järvinen et al. 2004). On the other hand, loss of crimp and a small reduction in crimp angle was described in ruptured canine ligament (Hayashi et al. 2003). Still, the comparison of annulus fibrosus and tendon crimps is not appropriate since crimps in tendon have a clear mechanical role: they increase the structure’s initial compliance by straightening.
Crimp in healthy cow tail can straighten if a single lamella is loaded in the direction of the fibres (Pezowicz et al. 2005). This straightening within bundles of fibres, however, did not occur at 6% applied strain of larger multi-lamellar samples tested in the present work, nor did it occur in a previous study with strains up to 28% (Vergari et al. 2016). Therefore, it is unlikely that fibres can straighten in vivo under physiological loading, although it must be noted that both these works used tail discs; cow lumbar discs might have a different behaviour. This might mean that the lack of crimp in degenerated human disc does not necessarily imply a lack of initial compliance for the lamella. In fact, straight parallel fibres can slide more easily against each other, thus increasing the lamellar compliance. While this hypothesis is still speculative, it is corroborated by the same shearing observed within the bundle as between bundles in degenerate human samples.
Crimp, however, appeared to be straightening at the interface between two lamellae (Online Resource 1), facilitating the large reorientation of the lamellae. It is possible that fibre crimp act locally as a torsional spring, making the inter-lamellar interface strong, allowing no sliding, but at the same time being very flexible.
The schematic drawings of intervertebral disc structure usually reported in the literature, depicting the annulus as a neat sequence of lamellae that run around the whole disc, with precise alignment of fibres at ±30\({^{\circ }}\), are not consistent with our newer observations. In Fig. 8, we propose an update on the drawing by Burke (1969) depicting incomplete lamellae running in several different direction, with inter-lamellar intersection and regions of possible interdigitation of bundles. These aspects should be included in biomechanical numerical model of the disc, as they likely have a significant impact on inter-lamellar mechanics (Nerurkar et al. 2011). Further research should elucidate on how these inter-lamellar structures change along the radial direction.
All human samples were harvested from degenerated discs. Acaroglu et al. (1995) observed a slight decrease in elastic modulus with degeneration, albeit not significant, and this is consistent with the present results showing that degeneration had very small effects. However, all discs were highly degenerated (Pfirrmann grade 4 or 5). It is possible that mechanical differences between these two grades of severe degradation are small and that including a wider range of Pfirrmann grades would have highlighted differences with the less severely degenerated discs. Also, compressive properties of the annulus might be more sensitive to degeneration than tensile properties (Iatridis et al. 1998).
A limitation of the present work is that testing and imaging were not performed in a saline bath. Preliminary tests showed that annulus sample swell significantly in saline, this reduces the flatness of the sample’s surface, making imaging impossible because of the thin imaging depth of field. The protocol adopted in which drops of PBS were applied to the samples once they were mounted in the testing device was felt to provide the best compromise between swelling and dehydration. A preliminary test was performed by mounting a cow tail annulus sample on the rig, loading it at 1% strain and letting it rest for two hours (much longer than the average 6 minutes of the “dry” testing phase of this work). No microscopic strain change was observed during this time. This test also verified that dehydration did not affect the microstructural features described in this work.
Another limitation is that cow tail discs were adopted as healthy control for degenerated human discs. Healthy human discs are usually not excised so no such surgical residual was available. On the other hand, degenerated discs are not usually found in young cow tail. Nevertheless, cows are often used in the literature as animal model replacement for human disc biomechanics.
Applied strains did not exceed 6%, so the mechanical properties reported are relative to the toe region of the stress strain curves. The median elastic modulus at 6% applied strain was 2.7 MPa in cow tail annulus and 2.8 MPa in human samples; mean and standard deviation values were 5.0 ± 5.0 and 2.9 ± 1.7 MPa, respectively. Previous work has reported similar mean values for elastic modulus: O’Connell et al. estimated a modulus of 7.33 ± 5.50 MPa for human lumbar disc annulus strips (2.1 ± 0.3 mm thickness) (O’Connell et al. 2012). Elliott and Setton (2001) reported 2.52 ± 2.27 MPa on thicker samples (5 mm) of human lumbar annulus. This high inter-sample variability is likely due to a combination of factors: differences between subjects (age, medical history, pathology, etc.), spinal level, but also samples size measurement error, difficulties in mechanical testing reproducibility, etc.
Single lamella testing of bovine tail annulus yielded a much lower elastic modulus of 1.88 ± 0.6 MPa (mean ± standard error, Monaco et al. 2016); these samples, however, were loaded in a direction orthogonal to the fibres, thus loading the weaker inter-bundle network. Consistent with the orthotropic nature of the tissue, single-lamella elastic modulus along the direction of the fibres is much higher (53.2 ± 27.5 MPa (Pezowicz 2010)).