Introduction

The prevalence of lumbar disc herniation is estimated at 3–5%.26 Herniated lumbar discs and resultant radiculopathy have resulted in almost 15 million office-based physicians visits per year, creating a financial burden on society exceeding US$50 billion in the US annually.13 Approximately 300,000 lumbar discectomies are performed each year in the USA, making it the most common procedure performed by spine and neurosurgeons.17

Disruption of the annulus fibrosus (AF) is manifested as circumferential and radial tears, and rim lesions.21 These tears are present after cumulative 34 or sudden pre-herniation 2,33 events that lead to herniation. Circumferential tears are common and were observed in cadaver discs from the teenage years.36 It is thought that delamination of adjacent lamellae are the first steps towards the development of circumferential tears, which can lead to early disc degeneration.7 The failure mechanism of circumferential tears is believed to arise from high inter-lamellar shear stresses, which lead to propagation of these tears.16 Delamination is a known failure mechanism of composite, laminate structures, suggesting that the region at highest risk of failure initiation is at the boundary between lamellae, which is referred to as the inter-lamellar matrix (ILM).29 Circumferential disruption of the ILM has been observed as a likely pathway for the nucleus material to follow during herniation.33,34,37,38,39 Microstructural observations have revealed new insights into AF tissue disruption during herniation,32,35 as well as the delamination strength of isolated AF samples of the ILM.9 While these studies have improved our understanding of where herniation may initiate, the possible pathways of its progression, and ILM delamination strength, no studies have measured the mechanical failure properties of the ILM, nor of the adjacent lamellae during progression to herniation. A new approach is required to understand the mechanical roles of the ILM and lamellae during herniation, where the disc should be pre-herniated without herniation to create microstructural damage, and compared to herniated and unloaded controls.

Therefore, the aim of this study was to develop a new multiscale, biomechanical and microstructural understanding of the effects of progressive herniation on the ILM and the lamellae in control, pre-herniated and herniated discs. The following hypotheses were proposed:

  1. 1.

    The regional micro-mechanical properties of the ILM, compared to the lamella, will be significantly reduced during progression to herniation.

  2. 2.

    Pre-herniation of the disc will cause micro tissue damage and structural disorganization.

Materials and Methods

An ovine model was used for this study based on its structural, biochemical and biomechanical similarities to the human disc.7,32,34,35,40 Fresh lumbar spines from 2 year old animals were obtained from a local abattoir, stored at − 20°C and then thawed at room temperature. Soft tissue was removed and spines were cut into L3–4 and L4–5 vertebra-disc-vertebra functional spinal units (FSUs) with intact posterior elements and longitudinal ligaments. Within each spine, L4–5 FSUs were allocated to macro- and micro-mechanical testing (N = 7 pre-herniation, and N = 7 herniated), with the adjacent L3–4 FSUs serving as controls that only underwent micro-mechanical testing (Fig. 1). Since it has been shown that not all discs will herniate (herniation rate of 60%),38 we tested additional spines to failure until we achieved herniated discs (N = 7). In total 19 ovine spines were used.

Figure 1
figure 1

Schematic drawing of the study design (a). The pre-herniation (N = 7) and herniation (N = 7) FSUs were subjected to macroscopic loading. Micro-tensile testing on individual lamella and the ILM from anterior (AN) and posterolateral (PL) regions for control discs (N = 14), pre-herniation and herniation FSUs was performed in radial and oblique circumferential directions in 224 samples (14 spines × 2 discs per spines × 2 regions per disc (AN, PL) × 2 locations per region (lamella, ILM) × 2 test loading directions per location (radial, oblique circumferential) = 224 samples). The same slicing technique was used for PL region (not shown). Tissue segment (b) and prepared ILM (c) and lamella (d) samples for micro-mechanical testing was shown and the ILM and lamella denoted by ** and *, respectively.

Full size image

Prior to testing, FSUs were potted in cups, hydrated overnight, and installed in a hexapod robot.4,18,41 Twelve FSUs in the herniation group underwent 13° of flexion about the geometric centre of the disc, followed by a 3 mm compressive displacement at 400 mm/min to create herniation.38 In the pre-herniation group, the same protocol was used, however 50% of the compressive displacement in the herniation group was applied based on pilot studies finding no evidence of herniation at 1.5 mm of displacement. After loading, the FSU then underwent micro-mechanical testing.

Discs from the control and pre-herniation/herniation groups were dissected from the vertebral bodies. Tissue segments (approximately 10 mm length with the depth to the nucleus pulposus region) from the anterior (AN) and posterolateral (PL) were separated from each disc, placed in plastic containers (20 mm × 20 mm × 10 mm length × width × thickness) and optimal cutting temperature compound (OCT, Tissue-Tek®, Sakura, Japan) was poured over them. The moulded samples were stored at − 20°C for an hour and adjacent frozen samples (N = 4,) were cut from each region (AN, PL) of all discs using a hand microtome at an oblique angle. The mean (95% CI) dimensions of the samples were approximately a thickness of 1 mm, a width of 10.2 (0.8) mm. The oblique cutting angle was chosen to facilitate the identification of both the in-plane single lamellae and the ILM24 during sample preparation for mechanical testing under light microscopy observation. For the herniated group, samples were taken adjacent to the PL region of nucleus extrusion. Sample preparations for the micro-mechanical tests were conducted using a stereomicroscope (Motic, SMZ-168, China). From each adjacent sample, a single lamella or two adjacent lamellae including the ILM (ILM-lamellae complex) region were identified. Under microscopic observation, a small needle was used to spread cyanoacrylate adhesive, which was mixed with black ink, beyond the identified region (single lamella or the ILM-lamellae complex). Sand paper (250 grit) was placed over the adhesive and bonded to the sample carefully. The same process was employed for another edge of the sample. Sample preparation was performed under microscopic observation to avoid adhesive leakage into the identified region.28 The micro-mechanical properties of both lamella and the ILM samples were measured in radial (i.e., tension) and oblique circumferential (circumferential) (i.e., shear) loading directions. There was no indication of sample slippage during mechanical testing as identified from observation of samples during tests and analysing testing curves. The total number of samples was 224 (Fig. 1). Samples were hydrated and then attached to a micro-mechanical testing machine (BioTester, CellScale, Canada), preloaded to 100 mN, followed by failure at a strain rate of 10% s−1 at a sampling rate of 100 Hz. Preload was applied to remove the ‘slack’ in the sample prior to testing and the magnitude of 100 mN was chosen based on pilot studies.28 The displacement required to remove this slack region was not included in the strain/displacement measurement.

Four discs from each of control, pre-herniated and herniated groups (N = 12) were used for microstructural analysis after both macro- and micro-mechanical tests. All discs were fixed using 10% formalin in phosphate buffered saline. Tissues from the PL region were dissected from adjacent sections to those harvested for micromechanical testing after being fixed. All dissected tissues were moulded with OCT at approximate angles of 0° and 30° to the transverse plane. Samples (thickness 50 μm) were cut using a cryostat microtome (Leica Biosystems, CM3050, Germany) from each disc and mounted on microscope slides. Microstructural analysis was performed using light microscopy analysis (SZX 10, Olympus, Japan). Microscopy images of control and pre-herniated samples were analysed to measure lamellae width and orientation using a quantitative method (ImageJ).23,27,30 The width was calculated by selecting approximately 30 plot profiles from the control and pre-herniated images of each sample (e.g., Fig. 6a). The plot profile (Fig. 6b) displays a two dimensional graph of the intensities of the pixels along a line selection. The X axis represents the horizontal distance along the line and the Y axis shows the vertically averaged pixel intensity. Skeletonized images were used to compare the overall orientation of lamellae and its histogram distribution, relative to X axis (lateral), in control and pre-herniated samples.5 Failure stress, modulus, and toughness were calculated from engineering stress–strain curves. Failure stress was defined as the peak stress recorded during the test. The modulus measures was calculated as the slope of the best-fit line using linear regression. The toughness, which is a measure of energy absorption, was calculated as the area under stress–strain curve at failure stress.

Failure stress, modulus, and toughness were calculated from engineering stress–strain curves. A univariate ANOVA (IBM SPSS Statistics, USA) was conducted for each mechanical parameter having fixed factors of group (control vs. pre-herniation vs. herniation), direction of load application (radial vs. circumferential), type of tissue (lamella vs. ILM) and region (anterior vs. posterolateral) using an alpha of 0.05, with post hoc multiple comparisons conducted using a Bonferroni correction on alpha. An unpaired t test was performed between control and pre-herniated samples for lamellae width and orientation using an alpha of 0.05.

Results

To obtain the required seven FSUs for the herniation group, 12 FSUs underwent failure testing with five specimens not herniating, and showing signs of endplate failure in the inferior vertebra following dissection. Herniation in the seven FSUs occurred in the posterolateral region of the disc. The mean (95% CI) load to failure for the herniation group was 10.5 (1.5) kN and mean (95% CI) maximum load for the pre-herniation group was 6.4 (1.2) kN (Fig. 2). Upon visual inspection of the pre-herniated discs, no evidence of herniation was observed.

Figure 2
figure 2

Example force–displacement curves from the pre-herniation and herniation groups. No evidence of disc failure was observed during the pre-herniation test, compared to the herniation failure tests.

Full size image

Within each tissue type (ILM or lamella), there were no significant differences between the two matched control groups for all mechanical parameters and both regions (p = 1.00). Therefore, the data for the two control groups were pooled. For failure stress, the overall effects of group, tissue type, and disc region were significant (p < 0.001, Table 1). Significant interactions were found between group and tissue (p = 0.004, Fig. 3), as well as region and tissue (p = 0.029). For modulus, no significant overall main effects were found (p > 0.092), with only one overall significant interaction between loading direction and tissue (p = 0.033). For toughness, significant overall effects were found for tissue (p = 0.002) and region (p = 0.020). In addition, the only significant interaction found for toughness was between group and region (p = 0.032).

Table 1 Summary of ANOVA results for the overall effects of Group (control vs. pre-herniation vs. herniation), loading direction (radial vs. circumferential), tissue type (lamella vs. ILM), and disc region (anterior vs. posterolateral) and their interactions on each of the dependent micro-mechanical test parameters (failure stress, modulus, and toughness).
Full size table
Figure 3
figure 3

Example stress–strain curves from micro-mechanical testing during circumferential loading for all three groups in the posterolateral (PL) and anterior (AN) regions. Samples taken from adjacent samples.

Full size image

Post hoc tests revealed that the failure stress for the ILM was significantly lower in the pre-herniation (40%) and herniation (49%) compared to control group during radial loading (Fig. 4) in the posterolateral region (p < 0.032). In circumferential loading (Fig. 5) the ILM failure stress was significantly decreased by approximately 28 and 48% for pre-herniation and herniation groups compared to the control (p < 0.032). However, no significant difference in lamella or ILM failure stress was found between the pre-herniation and herniation groups (p = 1.00). Significant post hoc differences were found in the ILM where control group toughness was greater (76%) than the herniation group during circumferential loading in the posterolateral region (p = 0.049).

Figure 4
figure 4

Mean (95% CI) measured micro-mechanical parameters [modulus (a, b), toughness (c, d) and failure stress (e, f)] for ILM and lamella in the posterolateral (PL) and anterior (AN) regions during loading in the radial direction for all groups (control, pre-herniated, and herniation). * denotes significant pairwise differences between- and within-groups and tissue types (Lamella and ILM) (p < 0.05).

Full size image
Figure 5
figure 5

Mean (95% CI) measured micro-mechanical parameters [modulus (a, b), toughness (c, d) and failure stress (e, f)] for ILM and lamella in the posterolateral (PL) and anterior (AN) regions during loading in the circumferential direction for all groups (control, pre-herniation, and herniation). * denotes significant pairwise differences between- and within-groups and tissue types (Lamella and ILM) (p < 0.05).

Full size image

Comparing type of tissue, post hoc tests revealed that the failure stress of the ILM, was significantly lower than that of the lamella under radial loading in the anterior region for the control (28%) (p < 0.001) and pre-herniation (33%) groups (p = 0.003). In the posterolateral region during radial and circumferential loading, ILM failure stress was significantly lower than lamella for all three groups of control (28 and 22%), pre-herniation (60 and 36%) and herniation (60 and 59%) in radial and circumferential directions, respectively (p < 0.029, Figs. 4e, 4f and 5e, 5f). Under circumferential loading, ILM failure stress was significantly lower than lamella for both pre-herniation (28%) and herniation (36%) groups in the anterior region (p = 0.016). Post hoc tests also revealed ILM toughness to be significantly lower than lamella during radial loading for the control (25%, p = 0.030) and herniation (30%, p = 0.015) groups in the anterior region (Fig. 4d).

Comparing disc region, for the lamella, one post hoc significant regional difference was found where the failure stress in the anterior region was greater (17%) than the posterolateral for the control group during radial loading (p = 0.015). For the ILM, the anterior region failure stress was significantly higher than the posterolateral in all three groups: control (18 and 23%, p < 0.021), pre-herniation (43 and 32%, p < 0.008), and herniation (58 and 40%, p < 0.015, Figs. 4e, 4f and 5e, 5f) under both radial and circumferential loading, respectively. In addition, toughness in the anterior region was significantly greater (78%) than the posterolateral during circumferential loading in only the herniation group (p = 0.029) and during radial loading in both herniation (72%, p = 0.036) and pre-herniation groups (65%, p = 0.030, Fig. 4d).

Visual inspection of lamella width from light microscopy images (Fig. 6) revealed wider lamellae (55%) for pre-herniated compared to control samples. Skeletonized images identified that pre-herniated samples demonstrated a greater level of lamellae disorganization compared to control, where the boundaries of the ILM appeared to merge with lamellae. Results obtained from surface profile plots were consistent with measurements of average lamella width (Table 2), which showed an increase in lamellae width after pre-herniation. The lamella width was significantly larger after discs were pre-herniated, compared to control samples (p = 0.001). To visualize the degree of lamellae disorganization after pre-herniation, the lamellae orientation and its histogram distribution was measured for control and pre-herniated samples (Fig. 6, Table 2). A significant difference in the orientation of lamellae between control (43.5(0.8)°) and pre-herniated (81.6(6)°) samples (p = 0.002) was found.

Figure 6
figure 6

Light microscopy images (a) and surface profile plots (b) comparing a control and pre-herniated sample in the outer posterolateral (PL) region revealed a significant effect of pre-herniation on lamellae width (black arrows). The skeletonised images (c) and orientation histogram (d) indicated that pre-herniation altered lamellae orientation. The ILM and lamella were shown by black lines and arrows, respectively.

Full size image
Table 2 Mean (95% CI) measurements of lamellae width and their orientation (relative to the horizontal axis) across four samples for control and pre-herniated FSUs.
Full size table

Discussion

The first hypothesis for this study was that the regional micro-mechanical properties of the ILM, compared to the lamellae, will be significantly reduced during progression to herniation. We found that the ILM failure stress, in both radial and circumferential loading, was significantly reduced in pre-herniation and herniation groups, compared to control. However, the failure stress of the ILM after pre-herniation was not significantly different to after herniation (Figs. 4, 5), suggesting that the ILM was susceptible to mechanical damage at loads below those required to create herniation. In contrast, the lamella exhibited no reduction in failure stress between all three groups.

The second hypothesis was that pre-herniation of the disc would cause micro tissue damage and lead to structural disorganization. Light microscopy images revealed significant widening of lamellae, and a significant increase in structural disorganisation from a highly preferential lamellae alignment in control, to a random distribution in pre-herniated discs (Fig. 6). This micro tissue damage and disorganization resulted in a loss of distinction between respective lamellae and ILM boundaries.

The maximum reported axial compressive failure load that resulted in posterolateral herniation during flexion at a loading rate of 400 mm/min was 11.6(1.0) kN,38 which was comparable with our data (10.5(1.5) kN). In addition, our micro-mechanical results (Tables 1 and 2) were consistent with other studies. Radial stretching of ox tail lamella indicated that the cohesion stress was 250 kPa, which was similar to lamella failure stress in the anterior and posterolateral regions from this study (250 and 206 kPa, respectively).22 The reported range of failure strength of porcine and human lamella was from 0.18 to 1.88 MPa,11 and 0.14 to 0.46 MPa,8 respectively, in which our results fall. In addition, the peel strength of rabbit ILM, was significantly lower in the posterolateral region than the anterior region, corresponding to the same trends found in this study.10

We found that the failure stress of the ILM was significantly lower than that of the lamella during both circumferential and radial loading, for pre-herniated and herniated groups compared to control. It is not possible to simultaneously measure the macroscopic properties of the disc while also measuring the micro-mechanical properties at the lamellar interface. Therefore, we used macro-mechanical loading to pre-herniate and herniate the disc, and found significantly reduced mechanical properties of the ILM, which indicated partial failure at this site, compared to the adjacent lamella. This finding suggests that the ILM may be the weakest structure by providing a pathway for herniation, which has been observed in microstructural studies.31,32 In addition, the ILM in the posterolateral region of the disc was significantly weaker than the anterior region, potentially due to a higher number of incomplete lamellae.19 In support of our findings, the mechanical properties of single and multiple lamella bundles have been found to exhibit both radial and circumferential dependency, with the anterior region being stronger and stiffer than posterolateral.1,6,8,15,25 Increased mechanical strength of the outer lamellae could be a result of its structure, which is composed of highly packed collagen and elastic fibres compared to the ILM, which is mainly comprised of elastic fibres.30 A significant increase in lamella width and orientation was seen after pre-herniation (Fig. 6), which resulted in lamellae disorganization and disruption. Delamination of the ILM due to high shear stresses16 potentially severed the elastic fibre network, providing a mechanism for the nucleus to extrude through the ILM via tensile separation of the outer lamellae. Elastic fibres, which provide a well-organized orthotropic network, are the main fibrous element in the ILM. While the presence of collagen fibres in the ILM is less probable, it is unlikely that collagen fibres contribute to the mechanical properties of the ILM, unless they were translamellar bridges.29 The finding of unaffected modulus that was calculated from the linear region of loading may indicate the contribution of both collagen and elastic fibres, where at 100% strain both lamella and the ILM are loaded, which is consistent with the extensibility of elastic fibres.12 Further studies using histological staining are required to identify which components (the elastic network of the ILM or collagen fibres at the ILM interface) had failed. Moreover, extracellular matrix within the ILM may interact with elastic fibres and play a role in imparting isotropic failure properties, and likely to contribute to the failure properties of the ILM during progression to herniation.

A microstructural analysis revealed that circumferential progression of the nucleus toward the posterolateral region in the outer AF mainly occurred by ILM delamination (Fig. 7). Our biomechanical findings support this microstructural observation, indicating that in the outer AF, the ILM is the weakest structure. A study that used a flexion-compression model revealed tears in the mid region of the AF that represented direct path of extrusion in herniated discs.37 Another study that tracked the movement of pressurized gel injected directly into the nucleus32 found that there was a lack of ILM connectivity, which is supported by the biomechanical results of this study, showing a lower ILM failure stress. While this pressurization study disrupted AF lamellae via a different loading mechanism that may not be physiological, a comparison of nucleus pressures can be performed. Based on the average applied compressive load for pre-herniated discs, and an average ovine disc area of 550 mm2, the nucleus pressure was estimated to be 17.5(3.3) MPa.20,41 This pressure was higher than the injected gel study (12.2(0.4) MPa, posterolateral herniation),32 and is a factor of seven greater than the in vivo nucleus pressure of 2.3 MPa measured during lifting a 20 kg weight.41 Considering these differences to in vivo pressures, it is not surprising that we measured a significantly reduced failure stress in the ILM after pre-herniation, compared to control, which was no different to that after herniation. Therefore, a threshold of ‘safe’ compressive loading likely exists, above which a significant reduction in the failure stress of the ILM occurs, which places the disc at greater risk of herniation. The determination of this threshold may be used to support the development of manual handling guidelines to limit lifting motions/repetitions that could cause herniation in both the short- and long-term.

Figure 7
figure 7

Light microscopy (top) and macroscopic (bottom) image of posterolateral (PL) region of two herniated discs revealed an incomplete radial penetration of nucleus into the AF in the lateral (L) region (*), nucleus lateral tracking within the AF (arrows), and herniation site (×).

Full size image

The limitations of this study included the use of normal ovine discs instead of human, and a small sample size. Ovine discs represent a suitable biochemical, biomechanical, macro- and micro-structural analogy to human discs.40 While there is no consensus whether the disc is degenerated or healthy when herniation occurs,3,14 we chose 2 year-old sheep to represent herniation in normal discs without the confounding factor of degeneration. The most notable difference to human discs is geometry and this was accounted for by presenting engineering stress vs. strain. A relatively small sample size of N = 7 per group was used in this study, however we found statistically significant effects for both the mechanical and microstructural parameters. For the first time, we have demonstrated that compared to the lamella, the ILM is the weaker structure during the progression of herniation in the posterolateral region of the disc. The finding of no differences between ILM failure stress during pre-herniation and herniation suggests that there is a loading threshold above which the ILM loses its structural integrity, which has clinical relevance for recommending safe levels of lifting loads.