Biomechanical, Structural, and Composition Characterization of the Vertebra‐Disc Interface

ABSTRACT

Background

The spinal endplate region consists of the cartilage endplate (CEP), bony endplate (BEP), and subchondral trabecular bone (STB). Although this region is often involved in vertebral fractures and disrupted with disc degeneration, the biomechanics of these tissues are incompletely understood. The goal of this study was to characterize the mechanical, structural, and compositional properties of the spinal endplate region across aging and disc degeneration.

Methods

Cadaveric human lumbar spine segments (n = 45) were imaged with magnetic resonance imaging to assess disc degeneration. BEP and STB microstructure were analyzed with micro‐computed tomography. BEP specimens were subjected to four‐point‐bend testing, and CEP specimens to confined compression testing. BEP and STB composition was measured with ash analysis.

Results

Several BEP mechanical properties, including flexural modulus, were associated with structural measures such as bone volume fraction, but showed few correlations to composition. Measures of structure and composition were generally different but correlated between the BEP and STB. Superior and inferior endplates differed from each other in structure but not composition. Apart from the STB microstructure, which became sparser with age in superior endplates, none of the endplate region properties correlated with age or disc degeneration, despite large ranges in values observed for many of these properties, particularly, BEP fracture strain (1%–52% strain).

Conclusions

Structural measures are more reliable indicators of endplate region properties than are compositional measures, age, or disc degeneration. The BEP may play an important, independent role in spine function, due to its range in ductility and limited correspondence in properties to those of neighboring tissues.

1. Introduction

The spinal endplate region, consisting of the cartilage endplate (CEP), bony endplate (BEP), and subchondral trabecular bone (STB), forms the junction between the intervertebral disc and vertebral body (Figure 1). This region is biomechanically critical: it prevents the disc from bulging into the vertebral body and distributes applied loads to the adjacent bone and disc tissues [1, 2], while also serving as the primary pathway for nutrient delivery and waste removal for the disc [3]. The endplate region is often the site of initiation of vertebral fracture [4, 5, 6]. Vertebral fracture is the most common type of osteoporotic fracture [7] and an injury whose prevalence increases sharply with age [8, 9]. In addition, changes in the endplate region, such as structural disruptions in the BEP, are frequently noted in motion segments with degenerated discs [2, 10, 11]; in fact, recent evidence suggests that abnormalities in the CEP and BEP may precede degenerative changes in the disc [12]. Taken together, these observations on the function of the endplate region and the damage associated with aging and disc degeneration motivate study of the properties of the endplate region's tissues.

FIGURE 1.

Anatomy of the vertebral endplate region.

The mechanical properties of the components of the endplate region affect their behavior during loading and injury. Based on evidence from many types of osseous and cartilaginous tissues [2, 13, 14, 15, 16], these properties are likely to depend on the tissues' structure and composition. Indeed, there is some evidence of structure–function relationships in the endplate region in human spines. For example, CEP aggregate modulus decreases with increasing porosity [17] and BEP flexural modulus and yield strength increase with apparent density [18]. It is well established that age and disc degeneration are associated with changes in bone density [19, 20] and microstructure [20, 21] in the entire vertebra, but less is known about whether these findings apply to the endplate region. In fact, evidence suggests that the properties of bone in the endplate region are different from those in the rest of the vertebra [18, 22, 23].

Initial data suggest that the porosity of the BEP [24] and STB [23] increases with age, and that the BEP flattens as disc degeneration advances; the latter is thought to be a response to the altered loading that occurs with degeneration [25, 26, 27, 28]. Age and disc degeneration have also been correlated to changes in CEP composition and structure, such as decreases in water content and permeability [3, 29, 30, 31], but few data of this kind are available on BEP and STB composition. Additionally, studies relating structural properties of the endplate region to age and/or disc degeneration have often produced mixed findings, perhaps due to cohort differences, sample size, and varying definitions of the components of the endplate region. For example, studies comparing BEP thickness to age have found no correlation [28, 32, 33], a negative correlation [34], or a positive correlation in superior endplates but none in inferior [35]. Whether aging and disc degeneration affect the superior endplate region differently from the inferior one is also not clear, despite evidence of lower BEP thickness and STB density [35, 36] and higher prevalence of vertebral fracture [37, 38, 39] in the superior region.

The goal of this study was to characterize the mechanical, structural, and compositional properties of the endplate region. Specimens from the superior and inferior BEP, CEP, and STB in human lumbar spine segments spanning wide ranges in donor age and disc health were studied. Our specific objectives were (1) to evaluate the influence of BEP structure and composition on mechanical properties; (2) to evaluate correlations between and differences among the properties of the different tissues of the region; and (3) to assess the correspondence between tissue properties and aging and disc degeneration.

2. Methods

2.1. Sample Preparation and Imaging

Fresh‐frozen, cadaveric whole lumbar spines (n = 45, age range 20–88, 20 male, 25 female) were acquired (NDRI; Science Care; Anatomy Gifts Registry). Spine specimens underwent magnetic resonance imaging (MRI) for assessment of disc degeneration by T2 relaxation time [40, 41] and Pfirrmann grade (Figure S1) [42]. Pfirrmann grade was evaluated from T2‐weighted Turbo Spin Echo (T2w TSE) scans with 0.5 × 0.5 × 5.0 mm resolution [43, 44]. The grade is based on disc signal intensity, homogeneity, distinction between the disc nucleus pulposus and annulus fibrosus, and intervertebral disc height. Grade 1 describes a homogeneous, hyperintense disc with normal height and a clear distinction between nucleus and annulus; Grade 2, an inhomogeneous but hyperintense disc with preserved height and clear distinction; Grade 3, an inhomogeneous less intense disc with an indistinct boundary and normal to slightly reduced height; Grade 4, an inhomogeneous hypointense disc with loss of distinction and moderately reduced height; and Grade 5, a collapsed, hypointense disc with no distinction between nucleus and annulus. Grading was performed by two trained operators. Discs were put into three groups based on their Pfirrmann grade: a score of 1–2 was considered healthy, a score of 3 mildly degenerated, and a score of 4–5 severely degenerated.

CEP, BEP, and STB samples were then dissected from each segment (Figure 2). First, the vertebral bodies were separated into superior and inferior halves with an autopsy saw by cutting through the mid vertebral body and adjacent discs. The CEP and adjacent portion of the intervertebral disc were removed from the BEP. The cartilaginous tissues were typically visually distinct from the BEP and underlying bone and could be pried apart with a metal spatula without cutting in the central endplate. Around the periphery, the cartilaginous tissues often anchored into the bone and had to be cut with a scalpel. Once removed, the CEP samples were frozen to preserve the tissue prior to mechanical testing. We note that “superior” and “inferior” locations were defined relative to the vertebra (Figure 1). Each half‐vertebra was submerged in 1X phosphate‐buffered saline (PBS) solution with a dissolved protease inhibitor tablet (Thermo Scientific; Waltham, MA) and imaged in a microcomputed tomography (μCT) scanner (Xradia 520 Versa; Carl Zeiss Inc., Pleasanton, CA; 20 μm/voxel, 160 kV, 10 W) to enable analysis of the STB microstructure. The half‐vertebrae were then further trimmed using a diamond wafering blade (IsoMet 4000; Buehler, Lake Bluff, IL) to produce a rectangular test specimen of the BEP. The cut separating the BEP from the STB was made at a depth determined visually, and it produced a BEP sample with approximate dimensions 30 mm × 13 mm × 1.5 mm. The BEP specimen was then inspected; in some cases, small cartilaginous portions of the CEP visible to the eye remained and were then removed with a scalpel. Given the pore size distribution in the BEP, a higher‐resolution μCT scan was needed to characterize the BEP microstructure; moreover, the smaller size of the BEP specimens made higher‐resolution scans possible. Each BEP test specimen was submerged in PBS solution and imaged in a μCT scanner (μCT 40; Scanco Medical, Brüttisellen, Switzerland, 16 μm/voxel, 70 kV, 114 μA). At the time of BEP dissection, a rectangular specimen of the STB was taken directly beneath the BEP specimen, with the same 30 mm × 13 mm transverse dimensions, and a height up to 6 mm. In some cases, the half‐vertebra was less than 7.5 mm in height, in which case the remaining height beneath the BEP specimen limited the height of the STB specimen. In total, 13 of the spine segments provided CEP, BEP, and STB samples from L1 superior and L3 inferior regions, and the remaining 32 produced CEP, BEP, and STB samples from L2 superior and L2 inferior regions.

FIGURE 2.

Experimental workflow.

2.2. Bone Microstructure

STB microstructure was analyzed using the μCT scans taken of the half‐vertebrae prior to the separation of BEP and STB samples. Scans were binarized using Otsu's method in MATLAB; on average, the resulting threshold was 57% of the maximum brightness value in the scan. A volume of interest (VOI) was then defined that corresponded to the location of the physical STB sample dissected experimentally: that is, a central location in the transverse plane, and beginning at the depth corresponding to the BEP‐STB boundary. The VOI matched the physical sample's 13 mm width but was shorter, only 17 mm long, due to limits of the scan field of view. To determine the depth corresponding to the start of the STB, the top and bottom of the BEP were identified on sagittal image slices using image dilation/erosion to determine the lowest depth of the BEP within the VOI bounds; the STB VOI was then taken beneath this depth, with a height up to 6 mm to match the physical sample. Bone volume fraction (BV/TV) of the resulting rectangular VOI was calculated via simple counting of voxels in Matlab, and trabecular thickness (Tb.Th), trabecular separation (Tb.Sp), connectivity density (Conn.D), and degree of anisotropy (DA) were calculated using BoneJ [45]. Trabecular number (Tb.N) was calculated as 1/(Tb.Th + Tb.Sp).

BEP microstructure was analyzed using the μCT scans of the BEP test specimens, following their dissection from the surrounding tissue (Figure S2). A global threshold of 528 mg HA/cm³ (220 per‐mille) determined from an adaptive, iterative technique (Scanco Medical) was used to binarize the images. Image dilation‐erosion was used to fill the sample volume, enabling automatic contouring. BV/TV, bone mineral density (BMD), and tissue mineral density (TMD) were quantified for the central 16 mm of the specimens, which corresponds to the flexural span in the bend test. BEP thickness was measured using the three‐dimensional thickness measurement in BoneJ [18].

2.3. BEP Mechanical Testing

BEP specimens were subjected to a four‐point‐bend test as previously described in [18] (Figure S3). Briefly, each specimen was placed onto the support pins (inner span = 8 mm, outer span = 16 mm) of the bend‐test fixture (Model 5565; Instron, Norwood, Massachusetts). The surface that had been adjoined to the CEP was oriented downwards such that it would be subjected to compression during the test and mimic the concavity that typically develops in clinical vertebral fractures. After 15 preconditioning cycles to a displacement of 0.75 mm, each specimen was loaded to failure at a rate of 0.21 mm/s [46]. Force and displacement were measured with a 1 kN load cell (Instron 2525‐800 Series) and the test frame's linear variable displacement transducer, respectively, with the measured displacement being that of the outer pins. The test was stopped either when specimen failure occurred or when the displacement limit of the test fixture was reached.

Following the test, the apparent flexural modulus E, apparent flexural yield stress σ _y, ultimate strain ${\epsilon }_{\mathrm{u}}$, and the fracture strain ${\epsilon }_{\mathrm{f}}$ were calculated as previously described in [18]. In some samples, the fixture's displacement limit was reached without the sample fracturing, and pinching of the sample between the outer fixture pins and the fixture frame (Figure S3B) ensued. For these samples the strain was calculated from the still frame just before pinching commenced and was used as a lower bound on the ultimate and fracture strains. These lower bounds were omitted from statistical analyses but nevertheless provide context on the range of ductility seen among BEP samples. The tissue‐level yield stress (${\sigma }_{\text{tissue}}$) was calculated at the apparent yield point using M _y/S, where S is the section modulus, calculated with the second moment of area from the binarized μCT image, analogous to methods previously reported for torsional analysis [47, 48]. S and the corresponding tissue‐level stress were calculated for all bone voxels at the superior and inferior surfaces within the 16 mm span. ${\sigma }_{\text{tissue}}$ was reported as the 99th percentile of those tissue‐level stress values, to reduce susceptibility to outliers (Figure S4).

Some samples were excluded during dissection or mechanical testing. Complete calcification of the CEP prevented separation of the CEP and BEP in two inferior samples. Three additional endplates (two superior, one inferior) produced an untestable BEP sample due to damage or low thickness. There were also two superior BEP samples and one inferior sample that fractured during the preconditioning period of the test. These samples were included in the fractured group when testing for differences in properties between fractured and non‐fractured samples. STB samples were only obtained from roughly two thirds of the overall samples in the study due to some samples being dissected following a prior study's design that did not retain an STB sample.

2.4. BEP and STB Composition

After mechanical testing of the BEP samples, half of each BEP and STB sample was processed for measurements of tissue and ash densities. Each specimen was gently cleaned with a water jet to remove the bone marrow and then weighed (Entris II Model BCE Balance and Density Determination Kit, Sartorius Lab Instruments, Göttingen, Germany) in air and again in degassed water to calculate the tissue density (${\rho }_{\text{tissue}}$) using Archimedes' principle. Specimens were then defatted in acetone for 12 h and weighed to calculate tissue volume again before being dried in a muffle furnace (Thermolyne Furnace 47 900, Thermo Fisher Scientific, Waltham, MA) at 110°C for 2 h to obtain the dry weight. The dried specimen was then put back into the furnace at 650°C for 16 h to obtain the ash weight. Ash fraction and ash density (${\rho }_{\mathrm{ash}}$) were computed as the ratio of ash weight to dry weight and ash weight to tissue volume, respectively. Ash weight, water weight, and organic weight were then each computed as a percentage of the tissue weight after defatting.

2.5. CEP Mechanical Testing

CEP samples were prepared and tested as previously described [49]. Briefly, the CEP samples obtained from the dissection described above were sectioned on a freezing‐stage microtome to a thickness of ~0.8 mm corresponding to the transverse plane. A plug 4 mm in diameter was excised using a biopsy punch. The confined compression protocol consisted of a swelling period of 3 h, where the samples were allowed to equilibrate in the confining chamber within a PBS bath, followed by three compression–relaxation ramps. The three compression ramps, each of 5% of the initial thickness were applied over a period of 300 s (i.e., strain rate of 0.016%/s) followed by relaxation periods of 4000, 6000, and 8000 s for each ramp, respectively. Peak force of each ramp and the equilibrium force over the final 1500 s of each relaxation period was determine and converted to stress using the cross‐sectional area, calculated from measured sample height and diameter. Equilibrium modulus was calculated as the slope of the equilibrium stress–strain line. Percent relaxation was calculated for each ramp and normalized by sample height. The average of the 3% relaxation values was calculated and used for data analysis.

2.6. Statistical Analyses

To examine structure–function relationships in the BEP, linear regression was used to test for the dependence of BEP mechanical properties (E, σ _y, ${\epsilon }_{\mathrm{u}}$, and ${\epsilon }_{\mathrm{f}}$) on BEP BMD, and multiple linear regression was used to test for the dependence of BEP mechanical properties on BEP BV/TV and TMD. BMD, BV/TV, and TMD were chosen as the independent variables, because they are the main, non‐invasively obtained descriptors of the intrinsic quantity of bone mineral. To examine correspondence with composition, linear regression was used to test whether BEP mechanical properties were dependent on ash fraction and, separately, organic content. Water content was omitted in the statistical analyses, because organic and water contents were strongly inversely correlated in the tested samples. Given the variability in BEP flexibility observed even qualitatively during the bend tests, t‐tests were used to compare structural and compositional properties between the BEP samples that fractured vs. those that did not, and Pearson correlation analysis was used to test the correspondence between fracture strain and flexural modulus.

To examine correspondence among properties of adjacent pairs of tissues (BEP and CEP, and BEP and STB) in the endplate region, multiple linear regression was used to test the dependence of CEP modulus on the flexural modulus E, thickness, and BV/TV of the BEP, and the dependence of CEP relaxation on thickness and BV/TV of the BEP. These analyses explored whether the stiffness and transport properties of the CEP are correlated with those of the BEP. Linear regression was used to test for a correlation between CEP swelling force and BEP BV/TV, because we surmised that the swelling force may be related to the BEP's porosity. Linear regression was also used to test whether CEP swelling force was dependent on BEP organic content to investigate a potential relationship in organic content of the two tissues, as CEP swelling force is related to its proteoglycan content. Differences between and associations among the compositional and structural properties of the BEP vs. STB were assessed using paired t‐tests or Wilcoxon signed‐rank tests and Pearson correlation analyses, respectively.

To assess the correspondence between tissue properties and aging and disc degeneration, multiple linear regression was used to evaluate the effects of age and T2 relaxation time on CEP, BEP, and STB properties. ANOVA or Welch's test (based on the results of Levene's test of equality of variances) was used to test for differences in T2 relaxation time and CEP, BEP, and STB properties across groups of Pfirrmann grade. t‐tests or Mann–Whitney U tests, depending on the normality of the data, were used to test whether fractured samples and nonfractured samples in the BEP bend test differed in age, disc T2 time, structure, or composition.

All the above statistical analyses were performed separately for superior vs. inferior specimens. Subsequently, paired t‐tests or Wilcoxon signed‐rank tests were used to test for differences in CEP, BEP, and STB properties between superior and inferior endplate regions. All tests were carried out only after verifying that the assumptions underlying each were valid for the dataset used.

Due to the large number of tests included in correlations versus age and T2, tests across Pfirrmann grade groups, and comparisons between superior and inferior endplates, the significance level for these groups of tests was adjusted using Hommel's method [50].

3. Results

Multiple BEP mechanical properties depended on BEP structure and, to a lesser extent, composition. With increasing BMD, BEP E increased in both superior (p = 0.006) and inferior (p = 0.004) samples (Figure 3A), and ${\epsilon }_{\mathrm{u}}$ (p = 0.004) and ${\epsilon }_{\mathrm{f}}$ (p = 0.003) decreased in superior samples only (Figure 3B). σ _y was not correlated with BEP BMD (p > 0.05). In the multiple regressions with BEP BV/TV and TMD as the independent variables, no influence of TMD was found, and the influence of BEP BV/TV was similar to that of BEP BMD. In superior endplates, BEP E (p = 0.026) and σ _y (p = 0.007) increased with ash fraction (Figure 3C), and ${\epsilon }_{\mathrm{u}}$ increased with organic content (p = 0.019, Figure 3D). However, in inferior endplates, none of the BEP mechanical properties correlated with compositional measures.

FIGURE 3.

BEP mechanical properties versus BEP structure and composition: (A) flexural modulus (E) and (B) fracture strain (${\epsilon }_{\mathrm{f}}$) versus BMD; (C) apparent yield stress (σ _y) versus ash fraction; (D) ultimate strain (${\epsilon }_{\mathrm{u}}$) versus organic content. Points labeled with an arrow in (B) correspond to specimens that did not reach their fracture point before the end of the test, which have their final strain plotted. These points were not included in statistical analysis.

No consistent differences in structure or composition were found between BEP samples that fractured and those that didn't. In superior endplates, fractured BEP samples had greater BEP thickness (p = 0.029), while in inferior endplates, fractured samples had lower BEP thickness (p = 0.008), BV/TV (p = 0.015), and BMD (p = 0.022). Composition was not different in fractured versus non‐fractured BEP samples in superior or inferior endplates (p > 0.05). BEP ${\epsilon }_f$ decreased with E in inferior endplates (p = 0.002) but not superior endplates (p = 0.699).

The CEP properties were not correlated with any of the BEP properties (p > 0.05); however, there was correspondence in structure and composition between BEP and STB. BEP BV/TV and STB BV/TV were positively correlated in superior (p = 0.003) and inferior (p = 0.001) endplates (Figure 4A), though BEP thickness was not correlated to STB BV/TV (p > 0.05, Figure 4B). In superior endplates, ash fraction (p = 0.028) and ash density (p = 0.025) were positively correlated between BEP and STB (Figure 4C); in inferior endplates, ${\rho }_{\mathit{ash}}$ (p = 0.014), ${\rho }_{\text{tissue}}$ (p < 0.001), and water content (p = 0.022) were positively correlated between the two (Figure 4C,D). Despite the correlations, BEP and STB samples differed in composition (Figure 5): compared to STB samples from the same endplate, BEP samples had higher organic content (p < 0.001 in both superior and inferior) and lower water content (p = 0.001 in both) and lower ash fraction (p < 0.001 in both).

FIGURE 4.

(A) BEP BV/TV and (B) BEP thickness versus STB BV/TV; (C) BEP ash density versus STB ash density; (D) BEP water content versus STB water content.

FIGURE 5.

Organic, water, and mineral content in BEP versus STB samples from the same endplate: (A) superior endplates; (B) inferior endplates. Ternary plots of content fractions: (C) BEP and (D) STB data.

Superior and inferior endplates differed from each other in structure but not composition or mechanical properties (Figure 6, Tables 1, 2, 3). Superior endplates had lower BEP thickness (p < 0.001), BEP BV/TV (p < 0.001), BEP BMD (p = 0.002), and STB BV/TV (p < 0.001), and greater STB Tb.Sp (p < 0.001) and STB DA (p < 0.001) compared to inferior endplates from the same spine segment. Composition and mechanical properties were not different between superior and inferior BEP samples (p > 0.002).

FIGURE 6.

(A) BEP thickness and BV/TV and (B) STB BV/TV in superior versus inferior endplates. Samples from the same donor are connected by lines.

TABLE 1.

Disc and CEP properties.

	Superior		Inferior		Superior vs. Inferior
	n	Mean ± SD (range)	n	Mean ± SD (range)
T2 (ms)	44	84.5 ± 39.1 (39.2, 232.2)	45	89.8 ± 33.3 (50.5, 196.2)	—
F swell (N)	38	1.69 ± 1.36 (0.42, 6.84)	31	1.41 ± 1.16 (0.17, 4.86)	0.301
Pct R (%)	38	69.8 ± 10.5 (44.4, 86.6)	31	70.0 ± 12.6 (44.9, 88.4)	0.251
E eq (MPa)	38	1.56 ± 0.83 (0.41, 3.93)	31	1.51 ± 0.93 (0.20, 3.98)	0.304

Note: Rightmost column lists p‐values for paired t‐tests comparing values between superior and inferior endplates.

Abbreviations: E _eq, CEP equilibrium modulus; F _swell, CEP swelling force; Pct R, average CEP percent relaxation; T2, Disc T2 relaxation time.

TABLE 2.

BEP properties.

	Superior		Inferior		Superior vs. Inferior
	n	Mean ± SD (range)	n	Mean ± SD (range)
E (MPa)	38	138 ± 116 (19.1, 713)	36	143 ± 110 (13.7, 475)	0.199
σ y (MPa)	38	3.22 ± 1.90 (0.50, 7.77)	36	3.95 ± 3.07 (0.52, 15.18)	0.082
ε u (−)	33	0.105 ± 0.058 (0.005, 0.241)	27	0.117 ± 0.070 (0.013, 0.279)	0.281
ε f (−)	21	0.120 ± 0.091 (0.010, 0.389)	14	0.203 ± 0.133 (0.045, 0.524)	0.500
σ tissue (MPa)	37	13.72 ± 7.83 (2.66, 41.60)	36	16.29 ± 15.78 (2.56, 82.63)	0.357
BV/TV (−)	45	0.263 ± 0.075 (0.116, 0.489)	41	0.315 ± 0.136 (0.135, 0.707)	< 0.001
Thickness (mm)	44	0.196 ± 0.050 (0.109, 0.349)	41	0.248 ± 0.100 (0.122, 0.549)	< 0.001
BMD (mg HA/ccm)	45	315 ± 76 (161, 513)	41	361 ± 131 (184, 707)	0.002
TMD (mg HA/ccm)	45	992 ± 35 (928, 1076)	41	983 ± 31 (938, 1072)	0.006
ρ tissue (g/ccm)	45	1.499 ± 0.123 (1.111, 1.699)	41	1.511 ± 0.115 (1.225, 1.755)	0.294
p (−)	45	0.602 ± 0.058 (0.429, 0.693)	41	0.610 ± 0.063 (0.394, 0.756)	0.299
ρ ash (g/ccm)	45	0.696 ± 0.148 (0.367, 1.007)	41	0.689 ± 0.159 (0.261, 0.981)	0.261
Water (%)	45	29.1 ± 16.1 (2.9, 74.3)	41	27.7 ± 14.7 (4.3, 65.5)	0.412
Organic (%)	45	27.8 ± 6.5 (14.7, 49.5)	41	27.8 ± 6.4 (13.6, 52.8)	0.096
Mineral (%)	45	43.1 ± 11.6 (11.0, 64.2)	41	44.6 ± 11.5 (20.5, 67.3)	0.216
fAGE	32	517 ± 375 (191, 1619)	30	520 ± 253 (227, 1174)	0.118

Note: fAGE content is measured in ng quinine/mg collagen. Rightmost column lists p‐values for paired t‐tests comparing values between superior and inferior endplates.

Abbreviations: BMD, bone mineral density; BV/TV, bone volume fraction; E, flexural modulus; fAGE, fluorescent AGE content; p, ash fraction; thickness, BEP thickness; TMD, tissue mineral density; ε _f, fracture strain; ε _u, ultimate strain; ρ _ash, ash density; ρ _tissue, tissue density; σ _tissue, tissue yield stress; σ _y, apparent yield stress.

TABLE 3.

STB properties.

	Superior		Inferior		Superior vs. Inferior
	n	Mean ± SD (range)	n	Mean ± SD (range)
BV/TV (−)	29	0.104 ± 0.036 (0.038, 0.175)	27	0.118 ± 0.043 (0.046, 0.211)	< 0.001
Tb.Th (mm)	29	0.132 ± 0.020 (0.093, 0.175)	27	0.135 ± 0.020 (0.099, 0.200)	0.104
Tb.Sp (mm)	29	0.889 ± 0.234 (0.604, 1.691)	27	0.838 ± 0.166 (0.582, 1.260)	< 0.001
Tb.N (mm−1)	29	1.02 ± 0.20 (0.55, 1.40)	27	1.05 ± 0.16 (0.72, 1.39)	0.014
Conn.D (mm−3)	29	14.28 ± 9.85 (2.50, 36.24)	27	13.99 ± 8.19 (3.57, 33.66)	0.438
DA (−)	29	1.66 ± 0.22 (1.22, 2.22)	27	1.47 ± 0.18 (1.22, 1.87)	< 0.001
ρ tissue (g/ccm)	32	1.534 ± 0.131 (1.254, 1.697)	30	1.502 ± 0.144 (1.153, 1.708)	0.130
p (−)	32	0.638 ± 0.026 (0.558, 0.680)	30	0.638 ± 0.025 (0.579, 0.686)	0.393
ρ ash (g/ccm)	32	0.698 ± 0.128 (0.421, 0.912)	30	0.688 ± 0.149 (0.349, 0.865)	0.311
Water (%)	32	31.4 ± 7.0 (20.6, 50.7)	30	32.1 ± 8.3 (24.4, 53.6)	0.355
Organic (%)	32	24.7 ± 2.1 (17.1, 27.7)	30	24.5 ± 2.5 (17.5, 28.8)	0.231
Mineral (%)	32	43.8 ± 5.6 (32.2, 52.4)	30	43.4 ± 6.3 (27.8, 50.6)	0.432
Pen	12	1.181 ± 0.548 (0.530, 2.234)	11	1.354 ± 0.569 (0.540, 2.593)	0.050
fAGE	32	691 ± 942 (149, 5253)	30	616 ± 773 (109, 4467)	0.496

Note: fAGE content is measured in ng quinine/mg collagen; pentosidine content is measured in mmol pentosidine/mol collagen. Rightmost column lists p‐values for paired t‐tests comparing values between superior and inferior endplates.

Abbreviations: BV/TV, bone volume fraction; Conn.D, connectivity density; DA, degree of anisotropy; fAGE, fluorescent AGE content; p, ash fraction; Pen, pentosidine content; Tb.N, trabecular number; Tb.Sp, trabecular separation; Tb.Th, trabecular thickness; ρ _ash, ash density; ρ _tissue, tissue density.

Most mechanical, structural, and compositional properties of the endplate region did not correlate to age or disc degeneration (Figure 7, Supplementary Table 1). None of the CEP or BEP properties were dependent on age or T2 (p > 0.002), nor were there differences in age or T2 between the BEP samples that fractured and those that didn't. In STB samples from superior endplates, BV/TV (p < 0.001) and Tb.N (p = 0.001) decreased and Tb.Sp (p < 0.001) increased with age (Figure 7A), but no superior STB properties correlated to T2. In inferior endplates, STB properties did not correlate to age or T2. None of the structural or mechanical properties from the CEP, BEP, or STB were different among grouped Pfirrmann grades (p > 0.002). T2 relaxation time changed with changes in grade in inferior samples (p = 0.009) but not in superior samples.

FIGURE 7.

(A) STB BV/TV versus age; (B) CEP modulus versus disc T2 relaxation time.

4. Discussion

Considering the importance of the endplate region in spine health, the goal of this study was to characterize mechanical, structural, and composition properties of the tissues in this region. Samples from wide ranges in donor age and extent of degeneration in the adjacent discs were examined. We found that several BEP mechanical properties, including the flexural modulus (E), were associated with the primary measures of structure, BV/TV and BMD. In contrast, relatively few correlations of BEP mechanical properties to compositional measures were observed. Measures of structure and composition were generally different but correlated between the BEP and STB, indicating that the two bony tissues of the endplate region, while distinct from each other, may be influenced similarly by genetic and environmental factors. Superior and inferior endplates differed from each other in structure (i.e., BEP BV/TV, BEP BMD, STB BV/TV, STB Tb.Sp, and STB DA) but not composition. Finally, apart from the STB microstructure, none of the endplate region properties examined here correlated with age or disc degeneration, despite the large ranges in values observed for many of these properties. Overall, these results reveal deep complexity within the endplate region of the human lumbar spine.

This study has several strengths. No prior study of the endplate region has included characterization of each of the three tissues—CEP, BEP, and STB—and here we isolated each and measured their properties directly. Use of μCT imaging for analysis of BEP and STB microstructure enabled precise and consistent definition of the VOIs as well as precision in the measurement of structural properties. Isolation of the BEP and STB also allowed for the direct measurement of tissue density, ash density, and ash fraction, all of which have been reported to affect the mechanical properties of bone, rather than relying on estimates of density from CT scans [22, 28, 36]. Mixed results have been found when comparing CT‐based TMD measures to experimental density measures [51, 52]. In fact, in the present study, tissue density, ash density, and ash fraction did not correlate to the μCT‐measured TMD in superior or inferior BEP specimens, highlighting the importance of obtaining the experimental measurements. Few prior studies have measured the composition of the BEP, and none has reported on STB composition specifically. The large number of properties measured in a cohort of spine segments that spanned a wide range of donor ages and extent of disc degeneration is a major strength. While one prior study has performed similar mechanical testing of the BEP [18], no examination of age or disc degeneration was included. Similarly, prior studies that included such examination have studied only one or two of the three tissues studied here [23, 24, 28, 34].

This study also had limitations. Samples were taken from the central endplate only, and the properties of the peripheral endplate region may differ from those of the central endplate [17]. We chose to analyze the central region due to the high prevalence of VF and other damage in this area [53, 54]. Most samples were taken from the L2 level, but samples from 13 donors were taken from the neighboring L1 and L3 levels. These levels experience similar loads [55] and have similar morphology [56, 57], but some differences do exist across levels. Discs with severe disc degeneration were not well represented in the sample cohort. Moreover, consistent with prior studies [30], endplates adjacent to degenerated discs were more likely to produce untestable CEP or BEP specimens; for example, the two endplates that were completely calcified and unable to be included in any portion of the study were both adjacent to discs with a Pfirrmann grade of five. Use of the grouped Pfirrmann grade, which combined discs with a grade of four or five into one group, reduced but did not eliminate the imbalance. BEP specimens were visually inspected to check for remaining cartilaginous tissue from the CEP, which was removed if present. However, it is possible that some specimens still contained small bits of mineralized cartilage that were not visually distinct from the bone. This problem is likely unavoidable, as the interface between the BEP and CEP is not uniform and may contain fissures and protrusions of the tissues into one another [58]. Some BEP samples reached the fixture limit without fracturing, and in some cases also without reaching the ultimate point, thus limiting the sample size for fracture and ultimate strains. We also did not assess STB mechanical behavior or CEP composition. The composition of the CEP has been studied more extensively than that of the BEP and STB [3, 17, 59, 60], and the shallow depth and sparse, heterogeneous microstructure of the STB presents challenges for meaningful measurement of STB mechanical behavior. A final limitation in this study is the occurrence of multiple freeze–thaw cycles before mechanical testing due to multiple imaging and dissection steps. These steps were necessary to make full use of the spine segments, and prior studies have demonstrated minimal effect of multiple freeze–thaw cycles [61, 62, 63].

The lack of correlations of most endplate properties to age and disc degeneration is surprising, given the primary involvement of this region in vertebral fracture and the frequent observations of endplate damage adjacent to degenerated discs. T2 relaxation time was correlated with Pfirrmann grade in inferior samples, but not in superior samples; we suspect that the lack of a correlation in the superior cohort is due to insufficient representation of all Pfirmann grades. We did find that, with age, the STB in superior endplates is less dense, consisting of fewer and more sparse trabeculae, and higher DA. Theoretically, these aging‐associated differences alone can be consequential for risk of vertebral fracture, as they indicate a more compliant, weaker STB foundation on which the BEP rests and thus greater susceptibility to downward depression of the BEP in response to compression–a clinical hallmark of fracture [6, 64, 65]. Additionally, it is possible that there are aging‐ and degeneration‐associated differences in properties that we did not measure, such as tensile mechanical properties of the BEP and CEP, crystallinity [66, 67], collagen crosslinking [66, 68], and CEP composition characteristics like collagen content, collagen type, and GAG content [13, 29]. Nonetheless, the results of this study do not support age or disc degeneration as reliable indicators of most endplate properties.

The dependence of several BEP mechanical properties on BV/TV and BMD suggests these structural measures as potential biomarkers. We found that with increasing density, BEP samples are stiffer and, in the superior endplates, less ductile. The positive association between stiffness and density is consistent with what is known for many types of bone tissue. The emergence of photon counting CT, which provides high resolution assessment of bone microstructure at comparatively low radiation doses and without the limited gantry size of high‐resolution peripheral CT [69, 70], may permit sufficient measurement of BMD and BV/TV in the BEP and STB in clinical settings, for use as in vivo indicators of these tissues' stiffness and, in the case of the BEP, ductility. BEP ductility is highly relevant clinically, as vertebral fractures that do involve BEP fracture are associated with worse patient outcomes [14, 71, 72, 73, 74]. We note, however, that the full extent of BEP ductility is not known, because the flexibility of many BEP samples exceeded the capacity of the bend‐test fixture. Thus, in addition to the substantial range in fracture strains observed, 1.0%–52% strain, the upper limit on ductility could be even higher. The incompleteness of the dataset of fracture and ultimate strains could explain why no correlation was found between these strains and density measures in inferior endplates and why BEP thickness was higher for fractured versus unfractured endplates in superior samples but lower in inferior samples. Another confounding factor could be the presence of double endplates in some samples (5/45 superior and 10/41 inferior; Figure S2D), in that the mechanics of double‐endplate samples under flexion may be different, resembling more of a sandwich panel than a thin plate.

Clinical observations of fractured vertebrae note that superior endplates are much more frequently involved than inferior ones [37, 39]. Our results show that superior and inferior endplates differ in structure but not composition or BEP or CEP mechanical properties. Superior endplates had lower BEP BV/TV, BEP thickness, BEP BMD, STB BV/TV, and greater STB Tb.Sp and STB DA. These findings are consistent with prior comparisons of BEP thickness [35, 36] and BEP BV/TV [18, 36] and STB density [35], and they expand on them to encompass BEP density, STB microstructure, tissue composition, and BEP and CEP mechanical behavior. Moreover, the decline in STB BV/TV with age, while reported previously for both superior and inferior regions combined [23], is more pronounced in the superior samples (Figure 7A). Taken together, these results suggest that the higher prevalence of vertebral fractures in superior endplates may be due primarily to deficits in the superior STB structure. The lower thickness of superior BEP samples may exacerbate the problem by permitting greater deflection of the BEP and therefore more localized compression of the STB during spinal loading.

Overall, our findings provide insight into the nature of the BEP, a relatively understudied tissue in bone mechanics, both independently and relative to its adjacent tissues. Prior studies have found that BEP BV/TV does not correlate to the BMD of the full vertebra [22] and that the STB microstructure is different from that of the mid‐vertebral body [23]. Our data extend these findings to show that the composition of the BEP is distinct from that of the STB. Its lower ash fraction and water content and higher organic content may be one reason why many BEP samples had such high fracture strains. Nevertheless, correspondence was found between BEP and STB in structure and composition: endplates with dense STB are also likely to have a dense BEP, suggesting overall higher stiffness and lower permeability of these endplates. Interestingly, such correspondence did not extend to the CEP: whereas CEP permeability has previously been reported to increase with BEP porosity [24, 75], we found no correlation of CEP percent relaxation, an indicator of permeability, to BEP BV/TV. As such, the BEP can play an important and independent role in spine function, capable in some cases of extreme deformation and exhibiting limited correspondence in structure, composition, and mechanical properties to those of the neighboring tissues.

Author Contributions

Neilesh R. Frings: investigation, writing – original draft, writing – review and editing, data curation, formal analysis, software, conceptualization, methodology, validation, visualization. Harrah R. Newman: methodology, investigation, validation, data curation. Faysal M. Haque: investigation, methodology, validation, data curation. Veronica A. Caruso: investigation, writing – review and editing, visualization. Dawn M. Elliott: funding acquisition, conceptualization, writing – review and editing, project administration, supervision, resources, methodology. Elise F. Morgan: conceptualization, funding acquisition, writing – original draft, writing – review and editing, project administration, resources, supervision, methodology.

Funding

This work was funded by the National Institutes of Health (Grant R01‐AR054620).

Supporting information

Supplementary Table 1: Results of multiple linear regression of endplate region properties to both age and T2 relaxation time. Variable names are those defined in Tables 1, 2, 3; variables with a B subscript are properties of the BEP, and variables with an S subscript are properties of the STB. Values listed in the “vs. Age” and “vs. T2” columns are the p‐values for the respective model effects in the multiple regression, and the R ² values correspond to the multiple linear regression model with age, T2, and the age*T2 interaction as independent variables.

Figure S1: Representative MRI images of discs with increasing Pfirrmann grade from 1 to 5 [1].

Figure S2: Cross‐sections from binarized μCT scans of four BEP bend test samples. Scale bars in red are 50 voxels (800 μm) long.

Figure S3: (A) MicroCT scan of BEP bend test specimen; (B) BEP specimen in four‐point‐bend test fixture; (C) Force‐displacement curve from test.

Figure S4: Representative cross‐section from a binarized μCT scan of a BEP bend test sample. The neutral axis for this cross‐section is shown in blue, and the voxels with a section modulus value in the lowest 1% (resulting in a tissue yield stress value in the highest 1%) have been colored in red.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.