Evaluation of human cartilage endplate composition using MRI: Spatial variation, association with adjacent disc degeneration, and in vivo repeatability

Abstract Cartilage endplate (CEP) biochemical composition may influence disc degeneration and regeneration. However, evaluating CEP composition in patients remains a challenge. We used T2* mapping from ultrashort echo‐time (UTE) magnetic resonance imaging (MRI), which is sensitive to CEP hydration, to investigate spatial variations in CEP T2* values and to determine how CEP T2* values correlate with adjacent disc degeneration. Thirteen human cadavers (56.4 ± 12.7 years) and seven volunteers (36.9 ± 10.9 years) underwent 3T MRI, including UTE and T1ρ mapping sequences. Spatial mappings of T2* values in L4‐S1 CEPs were generated from UTE images and compared between subregions. In the abutting discs, mean T1ρ values in the nucleus pulposus were compared between CEPs with high vs low T2* values. To assess in vivo repeatability, precision errors in mean T2* values, and intraclass correlation coefficients (ICC) were measured from repeat scans. Results showed that CEP T2* values were highest centrally and lowest posteriorly. In the youngest individuals (<50 years), who had mild‐to‐moderately degenerated Pfirrmann grade II‐III discs, low CEP T2* values associated with severer disc degeneration: T1ρ values were 26.7% lower in subjects with low CEP T2* values (P = .025). In older individuals, CEP T2* values did not associate with disc degeneration (P = .39‐.62). Precision errors in T2* ranged from 1.7 to 2.6 ms, and reliability was good‐to‐excellent (ICC = 0.89‐0.94). These findings suggest that deficits in CEP composition, as indicated by low T2* values, associate with severer disc degeneration during the mild‐to‐moderate stages. Measuring CEP T2* values with UTE MRI may clarify the role of CEP composition in patients with mild‐to‐moderate disc degeneration.


| INTRODUCTION
The biochemical composition of the cartilage endplate (CEP) is believed to be a key factor that influences intervertebral disc health.
For example, nutrients entering the avascular disc and exiting metabolites must pass through the CEP, 1 and deficits in CEP composition that result in lower water content or porosity, such as fibrosis or calcification, can block solute passage. This could hinder disc cell survival and function 2 as well as limit the success of disc regenerative therapies that increase nutrient demands. 3,4 However, despite the theoretical importance of CEP biochemical composition, little is known about how CEP composition varies spatially within or between individuals. Likewise, the precise relationship between CEP biochemical composition and adjacent disc health also remains unclear. Characterizing the spatial variation in CEP composition and determining the relationship between CEP composition and adjacent disc health could provide insight into disc degeneration etiology and uncover diagnostic targets that may be useful for selecting ideal discs for treatment.
The biochemical composition of the CEP consists mainly of type II collagen, glycosaminoglycans (GAGs), and water. 5 The relative quantities of these constituents vary spatially, the GAG and water contents being higher in the CEP tissue adjacent to the nucleus pulposus (NP) than in the CEP tissue adjacent to the annulus fibrosus. 6,7 The composition of the CEP also depends on age and the stage of disc degeneration. For example, during early stages of degeneration, the GAG and water contents in the central CEP decrease, 6,8 which coincides with reductions in CEP permeability 9 that are presumed to imperil disc nutrition. However, detailed understanding of intra-and inter-CEP variations in biochemical composition is lacking, owing in part to the difficulty of sampling the CEP tissues in a robust manner. And although the links between CEP composition and adjacent disc health are being uncovered, the clinical relevance is unclear. This is because the CEP is not visible with conventional magnetic resonance imaging (MRI) sequences currently used in the clinic, and thus, it is difficult to identify patients with disc degeneration in whom deficits in CEP composition may be a contributing factor.
Recent advances in MRI enable visualization of the CEP 10-12 and noninvasive estimation of its biochemical composition. 11 For example, T2* relaxation times derived from ultrashort echo-time (UTE) MRI are positively correlated with CEP hydration and negatively correlated with the ratio of collagen-to-GAGs. 11 Notably, both characteristics-the relative amount of collagen in the CEP and the degree of hydration or porosity-influence solute uptake 13 and diffusion. 2,7 In the present study, we used T2* mapping to study the biochemical composition of CEP ex vivo and in vivo. We first quantified spatial variations in CEP T2* relaxation times within and between CEPs. Next, to determine how CEP composition impacts the extent of adjacent disc degeneration, we compared T1ρ relaxation times of the NP to the adjacent CEP T2* values. T1ρ values are positively correlated with the GAG and water contents of the NP, 14 and thus, T1ρ assessment is sensitive to NP biochemical changes that typify early disc degeneration. 15,16 Last, we quantified the in vivo repeatability of CEP T2* measurements to gauge the clinical feasibility of using T2* values to noninvasively assess CEP composition.

| METHODS
This study was approved by our Institutional Review Board. Informed consent was obtained from each subject.
Major exclusion criteria for the subjects and cadavers included pregnancy, spondylolisthesis, scoliosis, prior lumbar surgery, disc herniation, and compression fracture. Additional exclusion criteria that were applied to the subjects included diabetes, positive smoking status, cancer, and active use of osteoporosis medication(s).

| Image acquisition
Clinical MRI was performed on a Discovery MR 750 3T scanner using an 8-channel Cervical Thoracic Lumbar Coil (GE Healthcare). Cadaver MRI was performed on a Discovery MR 750W 3T scanner using a Geometry Embracing Method coil contained within the table. Imaging consisted of 3D multi-echo UTE Cones 17 and 3D T1ρ mapping sequences. 18 The multi-echo UTE cones sequence 17 had 21 423 conical trajectories per echo, the repetition time was 6.3 ms per trajectory, and fat suppression was applied every five trajectories. The T1rho sequence is a 3D magnetization-prepared angle-modulated partitioned k-space spoiled gradient recalled acquisition that has been described previously. 18 Parameters for the mapping sequences are summarized in Table 1. To assess in vivo repeatability of T2* mapping, subjects were scanned twice on the same day after exiting and re-entering the scanner between scans. Subjects were scanned between 8 and 11 AM, with the duration between repeat scans ranging from 15 to 30 minutes. Clinical fast spin-echo images with T 2 weighting (echo-time 61.6 ms, repetition time 2500 ms, echo train length 8, acquisition matrix 256 × 256, slice thickness 3 mm) were also acquired in the sagittal orientation and used for Pfirrmann grading.
Specifically, the orientation of the selected CEPs with respect to the magnetic field (range: 48°-60°) was close to 54.7°, the magic angle, thereby ensuring the resultant T2* values mainly reflect unbound water. 11 UTE images were interpolated to a voxel size of 0.5 × 0.5 × 0.5 mm 3 . Custom software was written in IDL (Harris Geospatial Solutions) to generate CEP T2* maps. The algorithm included the following steps: (a) CEP segmentation: we analyzed all sagittal slices that contained CEP voxels, which ranged from 51 to 88 slices depending on the CEP. On each sagittal MRI slice, the anterior and posterior ends of the CEP were located, and a contour line (thickness = 1 voxel) was created by tracing the voxels with high signal intensity (SI) between the two identified points. The resultant contour was masked over the original image, and the masked area was subsequently dilated using a voxel-based region-dependent inclusion criteria. For example, for a voxel C on the contour line, suppose voxel N is one of its cranially or caudally adjacent neighbors. N was included in the masked area if the following condition was met:

SI
SI T if N is adjacent to the vertebra, and where SI N and SI C are the SI values for N and C. Location-specific T 1 and T 2 refer to the SI difference between CEP and bony endplate and CEP and NP, respectively, and account for the varying image intensity and contrast across the endplate.
(b) T2* mapping: T2* relaxation times of the segmented CEP voxels from all sagittal slices were calculated on a voxel-by-voxel basis by fitting the measured SI to an exponential decay function. 11 (c) Establish coordinate system: The principal axes of the segmented CEP were identified, and the CEP was rotated so that the principal axes aligned with the x-, y-and z-axes, which coincided with the anterior-posterior, inferior-superior, and medial-lateral axes, respectively.
(d) Define reference template: to map the CEP voxels into standard reference template and exclude non-CEP voxels, Cartesian coordinates for each voxel were converted to polar coordinates.

Voxels with polar coordinates outside of the area bounded by
Mizrahi's equation 19 were excluded.  To illustrate T2* maps in 2D, the CEP voxels were projected onto the z-x plane. The mean T2* value was displayed when two or more CEP voxels projected onto one z-x location.

| NP T1ρ mapping
T1ρ maps of the NP adjacent to the index CEPs were created from three midsagittal slices using voxel-wise fitting of the signal intensities to decay curves. 18 T1ρ values were averaged for the NP voxels in the half-disc bordering the central CEP subregion (Figure 2).

| Statistical analysis
Paired t-tests with Holm-Bonferroni adjustments for multiple comparisons were used to compare summary T2* parameters (mean, interquartile range) between CEP subregions. Independent associations between mean NP T1ρ, mean CEP T2* and age were quantified by the Pearson correlation coefficient. To investigate the nonmonotonic relationship between disc degeneration and CEP composition, the discs were grouped to achieve some degree of age distinction while ensuring a similar number of discs in each group. To determine "low" and "high" CEP T2* values in each age group, the CEPs within a particular age group were rank-ordered from lowest to highest according to their mean T2* value. Then the CEPs were split into two equal-sized groups using the median T2* value for that particular age group as the cutoff between "low" and "high" T2* values. Two-way analysis of variance with interaction was used to test the effects of age and mean CEP T2* values on mean NP T1ρ values. Within each age group, independent two-sample ttests were used to compare mean NP T1ρ values and mean CEP T2* values between discs with the "low" vs "high" mean CEP T2* values. Short-term repeatability was estimated by the precision error, defined as the root-mean-square average of the precision errors in mean T2* values for the seven subjects. 20 Reliability of T2* measurements was measured by intraclass correlation coefficient (ICC) generated from random effects models 21 to account for repeated observations. We considered ICC values between 0.75 and 0.90 to be good reliability, and values more than 0.90 to be excellent reliability. 22 All analyses were performed using R software (R Core Team, 2018).
Data are given as mean ± standard deviation (SD). 3.2 | Associations between T2*, age, and adjacent disc degeneration Age was inversely associated with both mean T1ρ values in the NP (r = −.72, P < .001, Figure 4A) and mean T2* values in the central CEP (r = −.45, P = .032, Figure 4B). The relationship between NP T1ρ values and CEP T2* values for all CEPs pooled was not significant (r = .34, P = .11, Figure 4C). After splitting the discs into three  Figure 5 and Table 2). In the older age groups, NP T1ρ values were similar between discs with high vs low CEP T2* values (50-60 years, P = .62; >60 years, P = .39). Also, NP T1ρ values were significantly correlated with CEP T2* values (r = .71, P = .047), but only in the youngest age group ( Figure S1). The discs in the youngest age group had Pfirrmann grades II to III; in the two older age groups, grades III to IV with one grade II (Table 3).

| Comparisons between cadavers and subjects
The cadavers were significantly older than the human subjects

| Repeatability of T2* assessment in vivo
Repeat in vivo imaging and analysis indicated that inter-test variability in the distribution of T2* values was qualitatively lower than inter-subject variability ( Figure 6). The precision error of mean T2* assessment was 1. suggest that sGAG and water contents are likewise highest centrally and lowest posteriorly. Importantly, we also discovered that inter-  However, this could also reflect partial volume effects in the central subregion, which is known to be the thinnest region. 6,9,23 Histologic measurements indicate that mean CEP thickness is 0.62 ± 0.29 mm. 6 While the in-plane voxel size (0.5 × 0.5 mm 2 ) was smaller than the reported mean CEP thickness, and the segmented CEP masks were between 1 and 3 voxels thick (1.8 voxels on average), there may be some partial volume averaging of the boney endplate or NP that affects the interquartile range in some locations. Nevertheless, we believe any partial volume effects are likely to have a small impact on mean T2* values, since a prior study comparing histologic measures of CEP thickness to UTE measurements of CEP thickness with the same inplane resolution reported that the overestimation was not significantly different than zero (mean difference = 0.02 ± 0.13 mm, P = .32). 23 A clinically relevant finding of this study is that low CEP T2* values coincided with greater extent of disc degeneration during the mild-to-moderate stages. Previous ex vivo studies found that the relationship between disc degeneration and CEP composition and permeability is nonmonotonic. Specifically, disc degeneration associated with deficits in CEP composition and permeability, but only in the earlier stages of disc degeneration, 8,9 when increased disc degeneration is thought to proceed with reduced transport. Our results here corroborate those prior findings, since low CEP T2* values correlate with lower water content and greater collagen content in the tissue, 11  We limited our analysis to L4-L5 and L5-S1 CEPs because those CEPs were oriented with respect to the magnetic field at angles nearest to the magic angle. At that angle, the contributions of matrixbound water to T2* are minimized, and thus the measured variations in T2* values mainly reflect variations in the amount of mobile water and the collagen-to-GAG ratio. 11 Although limiting T2* analysis to L4-L5 and L5-S1 CEPs is clearly restrictive, inferring CEP composition based on T2* values at those levels is most relevant for understanding disc degeneration and regeneration. This is because disc degeneration is most prominent at L4-L5 and L5-S1, and hence, those discs are frequent targets for biologic therapies. From a practical standpoint, the orientation of the L4-L5 and L5-S1 CEPs is near to the magic angle owing to lumbar lordosis.  16 Also, certain exclusion criteria that could independently impact disc health, including diabetes and positive smoking status, were not applied to the cadavers and may have therefore contributed to the unexplained variability in the findings. A fourth limitation was that we tested short-term in vivo repeatability in only seven subjects, which precludes accurate estimation of the confidence interval around the precision errors. All scans were conducted in the morning to reduce any confounding effects of diurnal variations in disc water content, and thus, repeatability over the course of daily loading remains unknown. Last, this study was cross-sectional, and thus, we were unable to determine how changes in CEP biochemical composition relate to changes in disc health. Nevertheless, our findings here motivate T2* assessment in future longitudinal studies.
In summary, we used T2* mapping from UTE MRI to noninvasively study the biochemical composition of the CEP ex vivo and in vivo. Our findings suggest that deficits in CEP composition, as indicated by low T2* values, associate with more severe disc degeneration during the mild-to-moderate stages (Pfirrmann grades II-III) in younger individuals. Future longitudinal studies will use CEP T2* assessment to clarify the role of CEP composition in disc degeneration and in patient's response to treatment.