Correlations Between Quantitative T2 and T1ρ MRI, Mechanical Properties and Biochemical Composition in a Rabbit Lumbar Intervertebral Disc Degeneration Model

Abstract

Improved diagnostic measures for intervertebral disc degeneration are necessary to facilitate early detection and treatment. The aim of this study was to correlate changes in mechanical and biochemical properties with the quantitative MRI parameters T2 and T1ρ in rabbit lumbar discs using an ex vivo chymopapain digestion model. Rabbit lumbar spinal motion segments from animals less than 6 months of age were injected with 100 μL of saline (control) or chymopapain at 3 U/mL, 15 U/mL or 100 U/mL (n=5 per group). T2 and T1ρ MRI series were obtained at 4.7T. Specimens were mechanically tested in tension-compression and creep. Normalized nucleus pulposus (NP) water and GAG contents were quantified. Stepwise multiple linear regression was performed to determine which parameters contributed significantly to changes in NP T2 and T1ρ. When all groups were included, multiple regression yielded a model with GAG, compressive modulus, and the creep time constants as variables significantly impacting T2 (multiple r² =0.64, p=0.006). GAG and neutral zone (NZ) modulus were identified as variables contributing to T1ρ (multiple r²=0.28, p=0.08). When specimens with advanced degeneration were excluded from the multiple regression analysis, T2 was significantly predicted by compressive modulus, τ_1, and water content (multiple r²=0.71, p=0.009), while no variables were significant predictors in the model for T1ρ. These results indicate that quantitative MRI can detect changes in the mechanical and biochemical properties of the degenerated disc. T2 may be more sensitive to early stage degenerative changes than T1ρ, while both quantitative MRI parameters are sensitive to advanced degeneration.

Introduction

Back pain is the second most common cause of adult disability in the United States, with a prevalence greater than that due to heart conditions, stroke and cancer combined.^1,2 The economic burden of back pain is substantial, accounting for nearly $200 billion in yearly medical costs and lost wages.³ Back pain is commonly associated with degeneration of the intervertebral disc, a progressive age-related cascade which leads to degradation of disc structure and a loss of mechanical function.^2,4 Degeneration of the intervertebral disc is a ubiquitous process associated with aging, and is also prevalent in asymptomatic individuals.⁵ Consequently, the role of disc degeneration as a pain generator is a source of clinical controversy, and there is a need for quantitative assessment of both age- and trauma- associated changes to further characterize this process.

Clinical diagnosis of disc degeneration is largely limited to qualitative imaging modalities, such as plain radiographs or T2-weighted magnetic resonance imaging (MRI). On T2-weighted MRI images of the spine, the healthy intervertebral disc exhibits high signal intensity in the inner water- and proteoglycan-rich nucleus pulposus (NP), with low signal intensity in the fibrocartilaginous annulus fibrosus (AF). With degeneration, a reduction in signal intensity is observed in the NP, with eventual loss of distinction between the AF and NP.⁶ Such imaging provides primarily structural insights into disc health, which typically occur with advanced degeneration, and as such is therefore relatively insensitive to early-stage degeneration.

Compositional changes, primarily reductions in NP water and proteoglycan content, are some of the first known changes that occur with degeneration, and precede gross structural changes to the disc.⁷ Given that NP proteoglycan content directly impacts the mechanical properties of the disc, these compositional changes are also accompanied by insufficiencies in disc mechanical function.⁸ Experimental treatments for disc degeneration currently garnering significant research interest, including stem cell, biomaterial, or growth factor injection, aim to restore both disc composition and mechanical function.^9–11 However, such regenerative strategies necessitate intervention at the early stages of degeneration. Consequently, there is a need to both identify individuals that may benefit from therapeutic interventions at an early disease stage, as well as a method to noninvasively assess the treatment effect longitudinally. As such, developing improved, sensitive, non-invasive imaging modalities to quantify mechanical and biochemical changes to the disc is of critical importance to advance the diagnosis and treatment of intervertebral disc degeneration.

Quantitative MRI techniques, such as T2 and T1ρ mapping, can in principle provide information on both disc structure and composition in a non-invasive manner. T2 relaxation is a physical chemistry time constant that relates to the transverse relaxation of magnetization with atomic and molecular interactions and correlates largely with water content and tissue anisotropy.^12,13 The T1ρ time constant is based on spin-lattice relaxation (T1) in a rotating frame achieved by a magnetic field applied at an off-resonant component (spin-lock).¹⁴ In complex, hydrated tissues such as the intervertebral disc and articular cartilage, several factors can contribute to T2 and T1ρ relaxation, including collagen organization, water and proteoglycan concentration.¹⁵ As such, there are inherent challenges to the biophysical interpretation of T2 and T1ρ values. T2 is thought to be predominantly influenced by collagen ultrastructure and water content in the intervertebral disc, while T1ρ is considered more sensitive to proteoglycan content.^12,13,16 With regards to the mechanical function of the intervertebral disc, T2 and T1ρ have been shown to correlate with NP hydraulic permeability and compressive modulus.¹⁷ T1ρ has also been correlated with the fixed charge density and osmotic pressure of the intervertebral disc.¹⁸ T2 and T1ρ have both been shown to be sensitive to pathologic changes to the disc, with values decreasing with increasing severity of degeneration.^16,19,20

The relationships between quantitative MRI parameters including T2 and T1ρ and disc biochemistry, structure and mechanical function are complex and warrant continued investigation in order to translate such imaging modalities into clinical use. Ex vivo degradation of the nucleus pulposus provides a platform by which to study the effects of disc biochemistry and function on MRI relaxation. Chymopapain is a proteolytic enzyme which functions by cleaving the non-collagenous protein connections of proteoglycans, and has been clinically used for the treatment of disc herniation.^21,22 The purpose of this study was to correlate changes in mechanical and biochemical properties with T2 and T1ρ values in rabbit lumbar discs using an ex vivo chymopapain digestion model.

Methods

Specimen Preparation

Rabbit lumbar spines from animals less than six months of age were obtained from a local abattoir or from animals used in studies unrelated to the spine. The spines were dissected into individual bone-intervertebral disc-bone motion segments, and the posterior boney elements removed. Chymopapain (Sigma-Aldrich, St. Louis, MO) was resuspended in phosphate buffered saline at a range of concentrations. The NP of each intervertebral disc was injected with either 100 μL of phosphate buffered saline (control) or 100 μL chymopapain at 3 U/mL, 15 U/mL or 100 U/mL with a 27G needle.²³ Motion segments were then wrapped in saline-soaked gauze to ensure hydration of the disc and incubated at room temperature for 15 hours. A total of 28 lumbar motion segments were utilized in this study, and were randomly assigned to each experimental group. Five motion segments from each group underwent sequential MRI, mechanical testing and biochemical assays. An additional two motion segments per group were utilized solely for histology.

Magnetic Resonance Imaging

Following incubation with saline or chymopapain, MRI scans were performed using a DDR console (Agilent, Palo Alto, CA) interfaced to a 4.7T horizontal bore magnet (Magnex Scientific Limited, Abington, UK) using a custom-made 17 mm diameter solenoid coil. A series of mid-coronal slices for T2 mapping (TE=i*8.2 ms, i=1, 2…16) were obtained with an in-plane resolution of 195 μm and a 0.5 mm slice thickness. A second series of mid-coronal slices was obtained for T1ρ mapping (Spin Lock Time=i*10 ms, i=1,2…6) with an in-plane resolution of 234 μm and a 1 mm slice thickness. The bandwidth for both experiments was 100 kHz. The T1ρ pulse sequence was validated by quantifying the T1ρ relaxation time of phantoms of varying agarose percentage (1%, 2%, 4%), as described previously.²⁴ T2 and T1ρ maps of one mid-coronal slice from each sample were then generated in ImageJ in a pixel-wise fashion using a simplex algorithm (MRI Analysis Calculator plugin, NIH, Bethesda, MD). Average T2 and T1ρ maps of the intervertebral discs in each experimental group were generated using custom Matlab software, as previously described.²⁵Briefly, for each individual T2 and T1ρ map, the disc was manually segmented and mapped to a standardized grid, from which T2 or T1ρ values could be averaged among experimental groups on a pixel-by-pixel basis. The NP region was then manually contoured, as shown in Figure 1, to calculate mean T2 and T1ρ values.

Figure 1.

The raw image of the first echo of the T2 series is show in (A). From the series of T2 images, T2 maps (B) for each specimen were generated in a pixel-wise fashion. The NP region was manually contoured in MATLAB, as denoted by the outlined region in (B). A similar procedure was followed for T1ρ.

Mechanical Testing

Following MRI, the specimens were frozen for storage until mechanical testing. After thawing, the cranial and caudal vertebral bodies of each motion segment were potted in custom fixtures using a low-melting temperature indium casting alloy (McMaster-Carr, Robbinsville, NJ) for biomechanical testing (Instron 5948, Instron, Norwood, MA). Specimens were subjected to a loading protocol consisting of 20 cycles (19 cycles of preconditioning, 20^th cycle used for analysis) of tension/compression (+21N to −42N) at 0.5 Hz, followed by a creep test consisting of a 1 second step load to −42N (~0.48 MPa) and a 10 minute hold.²⁶ All testing was conducted in a bath of phosphate-buffered saline at room temperature. A high-resolution digital camera (A3800; Basler, Exton, PA) and a custom Matlab texture tracking program were used to optically track the axial displacements of two ink marks placed on each vertebral body immediately adjacent to the intervertebral disc.

The 20^th cycle of tension-compression was analyzed using custom Matlab software by fitting the force-displacement curve to a sigmoid function, as previously described.²⁷ Compressive range of motion (ROM) was defined as the displacement between the inflection point of the sigmoid curve and the maximum compressive displacement. Compressive stiffness was defined as the slope of the raw data between these two points. Total ROM was defined as the displacement between maximum tensile and maximum compressive loads. The compressive and tensile boundaries of the neutral zone (NZ) were defined by the maximum and minimum 2^nd derivatives of the sigmoid function, allowing for the calculation of NZ range of motion (ROM) and NZ stiffness.

Creep behavior was fit to a 5-parameter viscoelastic constitutive model in Matlab.²⁸ The model consists of an elastic response, a fast exponential decay response (characterized by time constant τ₁ and damping stiffness S₁), and a slow exponential decay response (characterized by time constant τ₂ and damping stiffness S₂). Creep displacement was measured from the raw optical displacement data, and was normalized by disc height measured by MR imaging. For both tension-compression and creep data, stiffness was normalized to disc height, and axial area measured from macroscopic digital images of the disc cross-section. Average force-displacement and creep curves for each experimental group were generated by LOWESS smoothing (local weighted scatterplot smoothing) in GraphPad Prism 6 (GraphPad Software, Inc., La Jolla, CA).

Biochemistry and Histology

Following mechanical testing, a scalpel was used to cut through the intervertebral disc immediately adjacent to the endplate, and the NP tissue was manually isolated. A digital photograph of the disc cross-section was obtained prior to NP tissue removal and used to quantify disc and NP area using a custom Matlab program. NP volume was approximated by multiplying NP area by disc height measured from MRI. NP water content was quantified by measuring wet weight of the NP tissue immediately following isolation and subtracting the dry weight measured after lyophilization. Lyophilized NP tissue was then digested in proteinase-K at 60°C. Sulfated glycosaminoglycan (GAG) content was measured using the dimethylmethylene blue (DMMB) assay.²⁹ NP water and GAG content were normalized to NP volume (NPVol) for each specimen.

Additional samples (n=2) from each experimental group were fixed in 10% neutral buffered formalin (Sigma-Aldrich, St. Louis, MO), decalcified (Formical-2000, StatLab, McKinney, TX) and processed for paraffin embedded histology. 7μm thick mid-sagittal sections were cut and stained with Safranin-O (proteoglycans) and fast green (collagen) to qualitatively assess the effects of chymopapain digestion on disc composition.

Statistical Analysis

Five samples from each group were utilized for MRI, biomechanical and biochemical analysis, while an additional 2 samples per group were utilized for histologic analysis. Statistically significant differences in quantitative outcome measures were established via one-way ANOVA with Fisher’s post-hoc test using SYSTAT 13 (Systat Software, Inc., San Jose, CA). Normality was verified using the Kolmogorov-Smirnov test. Coefficients of variation for all quantitative outcome measures are reported in supplementary Table 1. Univariate regressions were calculated for all combinations of measured variables in R (https://www.r-project.org/) to generate correlation matrices of r and p values using the corrplot function. As the contributors to T2 and T1ρ relaxation are likely multifactorial, multiple regressions were performed in order to determine the relative strengths and significances of contributors to T2 and T1ρ. Forward stepwise multiple linear regression was performed in R using the step function and the Akaike information criterion for model selection (penalty term; 2k = 4).³⁰

Results

Average T2 and T1ρ maps²⁵ for each experimental group (Figure 2A) illustrate the spatial distribution of T2 and T1ρ relaxation times across the intervertebral disc, and changes with chymopapain digestion. T2 values (Figure 2B) in the NP were progressively reduced with increasing dosage of chymopapain, and were significantly different from control T2 values in the 15 U/mL (p=0.03) and 100 U/mL (p<0.001) group. T1ρ values (Figure 2C) were also reduced, but only significantly in the 100 U/mL group compared to the 3U/mL (p=0.05) and 15U/mL (p=0.04) groups.

Figure 2.

Population average T2 and T1ρ maps (A) illustrate the progressive decrease in T2 (B) and T1ρ (C) relaxation times in the nucleus pulposus with increasing chymopapain dosage. Bars denote significance between groups (p<0.05, ANOVA with Fisher’s post hoc-test).

Changes in the quantitative MRI parameters corresponded with alterations to disc biochemistry. Alcian blue and picrosirius red stained histology sections (Figure 3) illustrated a loss of proteoglycan staining in both the annulus fibrosus and NP with increasing dosage of chymopapain. Shrinkage and loss of NP material was also observed in the 15 U/mL and 100 U/mL groups on histology. Normalized NP GAG content (Figure 4A) was significantly reduced in 15 U/mL (p=0.02) and 100 U/mL (p=0.007) groups compared to controls. Normalized NP water content (Figure 4B) was also reduced in chymopapain digested discs; this reduction was significant in the 100 U/mL (p=0.04) group compared to control.

Figure 3.

Safranin-O (proteoglycans) and fast green (collagen) stained histology sections demonstrate progressive loss of nucleus pulposus material and reductions in safranin-O staining in the nucleus pulposus and annulus fibrosus with chymopapain digestion.

Figure 4.

NP proteoglycan content (A) and water content (B), normalized to NP volume, were significantly reduced with increasing chymopapain digestion. Bars denote significance between groups, (p<0.05, ANOVA with Fisher’s post hoc-test).

Mechanical properties of the intervertebral disc, both in tension-compression and in creep, were significantly altered by chymopapain digestion (Table 1). Compressive modulus and NZ modulus were progressively reduced for the 3 U/mL and 15 U/mL groups compared to control, but were increased in the 100 U/mL group compared to the 15 U/mL group. Compressive, total and NZ range of motion generally increased with increasing dosage of chymopapain. Creep strain also increased with increasing chymopapain dosage. The early (S₁) and late (S₂) creep moduli and time constants (τ₁ and τ₂) were not statistically significantly different between groups, however, in general, τ₁, S₁ and S₂ were reduced in chymopapain digested samples compared to controls. Average force-displacement and creep curves for each group, shown in Figure 5, qualitatively illustrate the variations in the mechanical response with chymopapain digestion.

Table 1.

A summary of the mechanical properties for control and chymopapain digested samples.

		Control	3 U/mL	15 U/mL	100 U/mL
Tension/Compression	Comp. Modulus (MPa)	8.98±1.71	6.50±2.52*	4.85±1.51*	7.48±1.43
	Comp. ROM (mm)	0.13±0.03	0.18±0.05	0.35±0.18*	0.31±0.25#
	NZ Modulus (MPa)	1.50±0.34	1.63±1.09	0.38±0.39*	1.02±0.89
	NZ ROM (mm)	0.06±0.04	0.09±0.04	0.15±0.05*	0.16±0.10*
	Total ROM (mm)	0.27±0.05	0.35±0.14	0.66±0.28*	0.56±0.38#
Creep	Creep Strain (mm/mm)	−0.12±0.05	−0.13±0.07	−0.22±0.04	−0.22±0.11#
	S1 (MPa)	4.57±2.04	4.94±2.20	2.41±0.60	3.29±2.26
	τ1 (sec)	7.36±5.25	9.32±2.97	8.79±6.51	12.01±3.18
	S2 (MPa)	3.98±2.28	3.10±1.77	2.12±0.20	2.17±1.88
	τ2 (sec)	607.24±174.70	709.76±191.31	457.25±181.82	535.91±230.27

^*= p <0.05 compared to control,

^#= p<0.1 compared to control.

Figure 5.

Average force-displacement (A) and creep (B) curves for each group (generated using LOWESS smoothing) show marked changes in mechanical properties of the disc with chymopapain digestion. Most notable are the increase in neutral zone ROM in the 15 U/mL group and the increase in creep displacement with increasing chymopapain digestion.

Univariate correlations (Figure 6) between outcome measures (T2, T1ρ, mechanical properties, GAG and water content) revealed many of the mechanical properties were significantly correlated with one another. GAG and water content were also significantly correlated (p=0.002). When all experimental groups were included in the univariate correlation matrix (Figure 6A), T2 correlated significantly with T1ρ (p=0.04), water (p=0.04) and GAG (p=0.007) content. The strongest correlation for T1ρ besides T2 was for GAG content, but this did not reach statistical significance (p=0.09). When the 100 U/mL group was removed from the correlation matrix (Figure 6B), in order to exclude the severely degenerated samples, T2 values significantly correlated with compressive modulus (p=0.01), water (p=0.03) and GAG content (p=0.03). No significant univariate correlations were found with T1ρ.

Figure 6.

Univariate correlation matrices for all variables including all experimental groups (A) and including only the control, 3 U/mL and 15 U/mL groups (B) show significant univariate correlations between T2, water content, GAG content and compressive modulus. The color of the dots corresponds to the r-value, and increasing dot size indicates a decreased p-value. Statistically significant correlations (p<0.05) are denoted with an asterisk symbol in the corresponding box.

Finally, a forward stepwise multiple linear regression was performed to yield linear equations for T2 and T1ρ, in an effort to determine which mechanical and biochemical parameters were significant predictors of T2 and T1ρ values in the NP. When all groups were input into the multiple regression analysis, GAG content (p=0.10), compressive modulus (p=0.04), and the creep time constants τ₁ (p=0.01) and τ₂ (p=0.14) were included in the model as predictors of T2 (multiple r²=0.64, p=0.006):

$$\text{T}2=66.78+0.55\left(\frac{\text{GAG}}{\text{NPVol}}\right)-3.78\left({\tau }_1\right)+6.18\left(\text{Comp.}\text{Modulus}\right)+0.04\left({\tau }_2\right)$$ (1)

For T1ρ, GAG content (p=0.04) and neutral zone modulus (p=0.13) were included in the multiple regression model (multiple r²=0.28, p=0.08):

$$\text{T}1\rho =206.45+1.64\left(\frac{\text{GAG}}{\text{NPVol}}\right)-23.61\left(\text{NZ}\text{Modulus}\right)$$ (2)

When the 100 U/mL group (advanced degeneration) was removed from the multiple regression analysis, model fit for T2 was improved (multiple r²=0.71, p=0.009), with compressive modulus (p=0.02), the creep time constant τ₁ (p=0.04), and water content (p=0.16) included as predictors of T2:

$$\text{T}2=87.868+5.29\left(\text{Comp.}\text{Modulus}\right)-1.92\left({\tau }_1\right)+70.49\left(\frac{\text{Water}}{\text{NPVol}}\right)$$ (3)

None of the parameters were significant predictors of T1ρ when samples with advanced degeneration were not included in the multiple regression analysis.

Discussion

With this work, we have generated a range of degenerative states from mild to moderate to severe in the rabbit intervertebral disc in vitro via a gradient of chymopapain digestions. As evidenced by quantitative MRI and histology, chymopapain injection at concentrations of 3 U/mL and 15 U/mL induced mild to moderate degeneration of the disc, while injection at a concentration of 100 U/mL induced more severe degenerative changes. Chymopapain digestion resulted in both a significant loss of proteoglycans and water in the NP, hallmarks of early stage disc degeneration in humans.⁷

Chymopapain digestion also resulted in significant changes in mechanical response of the disc to tension-compression loading. The decrease in compressive modulus shown in the 15 U/mL group compared to control was consistent with changes previously demonstrated in a mouse puncture model of disc degeneration, where a reduction in compressive modulus was accompanied by a reduction in NP GAG content 8 weeks following needle puncture.²⁷ The increase in modulus in the 100 U/mL group compared to the 15 U/mL group is likely due to a loss of NP pressure and an increase in the share of compressive loads supported by the annulus fibrosus.³¹ Increased creep strain and increased τ₁ with increasing chymopapain digestion are similar to the trends observed in degenerative human discs.²⁸ Increases in range of motion were also observed with chymopapain digestion, including a marked elongation of the neutral zone in the 15 U/mL group, consistent with the loss of function of the NP.³¹

Via univariate and multivariate regression analyses, we identified several mechanical and biochemical parameters as primary predictors of T2 and T1ρ relaxation times in the NP. Across all levels of degeneration, proteoglycan content, compressive modulus and the two creep time constants, τ₁ and τ₂, were found to be the most significant determinants of T2 relaxation time in the NP. Proteoglycan content and NZ modulus were the primary contributors to T1ρ; however, the fit of the multiple regression model was poor (r² = 0.28), with only a trend towards statistical significance (p=0.08), likely because no statistically significant univariate correlations for T1ρ existed. When the multiple regression analysis was performed with only control, mild and moderate degeneration groups (control, 3U/mL and 15 U/mL groups), model fit for T2 was improved, with compressive modulus, τ₁, and water content identified as the primary determinants of NP T2. However, no variables were included in the regression model for T1ρ, suggesting that in the intervertebral disc, T2 mapping may be more sensitive to the biochemical and mechanical changes occurring with early stage degeneration.

The results from this study build on previous work which has shown correlations between quantitative MRI parameters and local disc tissue mechanics and biochemistry. In human degenerative intervertebral discs, univariate correlations have been reported between T2,water and proteoglycan content, which we have also observed in this ex vivo digestion model.^12,32 MRI studies in human degenerative discs have also found associations between T1ρ and proteoglycan content;¹⁶ this is in contrast with other prior work in trypsin digested bovine tail discs that found no correlation of T1ρ with proteoglycan content, and our current study, which found no univariate correlations with T1ρ.¹⁷ Previous multiple regression analyses in both human degenerative discs and trypsin digested bovine tail discs concluded that water content and NP hydraulic permeability were influenced by T1, T2, T1ρ, magnetization transfer ratio and apparent diffusion coefficient, while NP shear modulus was determined by T2, T1 and apparent diffusion coefficient.^17,33

Limitations of this study include the ex vivo digestion of the NP via chymopapain, and the estimation of NP volume from digital and MRI images. Additionally, further work needs to be conducted in animal models of degeneration and in human patients to verify these results. Overall, results from this study suggest that quantitative MRI techniques, particularly T2 mapping, are sensitive to the mechanical and biochemical changes to the NP that occur with disc degeneration. While T2 and T1ρ mapping are useful for quantifying degeneration in a more rigorous manner than clinically used qualitative grading schemes, data from the multiple regression analyses conducted in this study suggest that T2 mapping may be particularly sensitive to early and intermediate stage degenerative changes compared with T1ρ. With continued development, quantitative MRI techniques may offer a powerful, non-invasive methodology to evaluate functional and compositional changes to the intervertebral disc in both animal models of degeneration and regeneration and in human patients.