The biological basis of degenerative disc disease: proteomic and biomechanical analysis of the canine intervertebral disc

Sample collection of NCD and CD intervertebral disc nucleus pulposus

All procedures involving the sacrifice of the NCD and CD dogs and tissue handling were carried out in accordance with established animal use protocols at the University Health Network, University of Toronto, ON, Canada, and in compliance with the regulations of the Canadian Council on Animal Care. We removed the T5-L6 spinal motion segments immediately after the animals were sacrificed, and then we placed all the specimens on ice. All canines used for the study were between 20 and 24 months of age and equally divided by sex. For biomechanical studies, the L4-L5 motion segment was recovered in its entirety, cleared of all soft tissues including ligaments, labeled, and immediately stored at −80 °C. For proteomic and biochemical analyses, we performed a wide laminectomy on the remainder of the spine using a Stryker saw (Stryker, Kalamazoo, MI, USA) followed by clearing of all soft tissues. After the posterior elements were removed, the spine was washed with Clidox (Pharmacal Research Laboratories, Waterbury, CT, USA) and Betadine (Purdue Products, Stamford, CT, USA) and cooled on ice. Thereafter the discs were incised immediately adjacent to the inferior end plate with a number 10 -blade, and the NP was removed in its entirety as per our established methods [16, 17]. The NP was placed into warmed physiological phosphate-buffered saline (PBS; 0.1 M, pH 7.4), rinsed, and then divided for isobaric tags for relative and absolute quantitation (iTRAQ) or immunohistochemical analysis. We were meticulous in the selection of the IVD NP used for these experiments, such that all NCD discs used were of the jellylike, mucoid, translucent appearance and all CD discs appeared fibrocartilaginous and pulpy, consistent with the degenerative phenotype widely used to describe this animal. For histological preparation, the removed IVD NP was fixed in 10 % formalin, placed in appropriately labeled cassettes, and processed using 10 % neutral buffered formalin, ascending grades of ethanol, two changes of the clearing agent xylol, and infiltrated with three changes of molten paraffin (Paraplast) for 12 h. After processing, the samples were placed into an automated embedding unit, and one sample at a time was moved from its cassette into a metal mold, appropriately oriented, filled with molten paraffin, and cooled on a cold plate and then sent for analysis.

Sample processing and iTRAQ labeling

The reference sample consisted of equal amounts of total protein from each of the eight samples used in this study, thereby providing a global reference that was common to all three sets. We randomized the labeling of the reference sample in each set to prevent any inadvertent introduction of labeling bias. We then pooled the labeled samples to form three 4-plex iTRAQ sets.

Mass spectrometry and data analysis

Each iTRAQ set was processed by two-dimensional liquid chromatography–tandem mass spectrometry analysis as described earlier [18]. Briefly, the samples were separated offline using strong cation exchange (SCX) chromatography and fractionated into 30 SCX fractions. Each of these SCX fractions was then dried and redissolved in a minimal volume (20–30 μl) of 0.1 % formic acid. The redissolved fractions were injected into a reversed-phase (RP) trap column, where they were desalted before being eluted into an analytical nano liquid chromatography (LC) RP column (75-μm internal diameter (i.d.) × 150 mm). Elution and separation on the nano LC column were effected using a binary gradient of water and acetonitrile with 0.1 % formic acid at a 200 nl/min flow rate. Eluting peptides were analyzed on a 5600 Triple time-of-flight instrument (SCIEX) in information-dependent acquisition mode. Each set was analyzed a minimum of two times, and the resulting MS data were then analyzed using the ProteinPilot software (v4.0) package from SCIEX using a mammalian database. A false discovery rate (FDR) was calculated to provide a measure of confidence in the proteins reported. To minimize redundancy between proteins reported in the three individual iTRAQ sets and to ensure consistency of reported isoforms from one set to the next, the results of the three sets were aligned using a Microsoft Excel-based (Microsoft, Redmond, WA, USA) protein alignment template as described elsewhere [19]. We first generated a master list of all the proteins identified in this study by performing a search of the combined data from all three iTRAQ sets and duplicate runs using ProteinPilot.

Relative quantification of proteins, which was simultaneously performed using ProteinPilot, was based on the areas of the iTRAQ signature ion peaks. Only peptides not shared with other reported proteins (or group) contribute to the overall ratios reported for the protein. The overall protein ratios reported represent a weighted average of the ratios of the contributing unique peptides, where the weighting factor is determined by the percentage error of the individual peptide ratios. Further, the reported ratios were normalized using a factor, termed the applied bias, that is calculated based on the assumption that the majority of the proteins being compared between the samples in a set are expressed at similar levels.

Western blot analysis

As further validation of the iTRAQ data, we chose to probe the same NP homogenate supernatants from which we generated samples for iTRAQ for the level of proteins secreted, as determined in our proteomic experiments. Equal amounts of tissue homogenates from one representative sample each of NCD and CD NP that was used for iTRAQ analysis and immunohistochemistry were subjected to Western blot analysis and probed for the levels of secreted fibromodulin, biglycan, and hyaluronan and proteoglycan link protein 1 (HAPLN1) as additional verification of our iTRAQ results. Briefly, 50 μg of total protein obtained from IVDs were resolved on SDS-PAGE gels. Proteins were then electrotransferred onto polyvinylidene difluoride membranes (Bio-Rad Laboratories, Hercules, CA, USA). After blocking with 5 % non-fat powdered milk in Tris-buffered saline (TBS; 0.1 M, pH 7.4), blots were incubated with either rabbit polyclonal antibody against fibromodulin (1:1000, catalog number sc-33772; Santa Cruz Biotechnology, Santa Cruz, CA, USA), goat polyclonal antibody against biglycan (1:1000, catalog number sc-27936; Santa Cruz Biotechnology), and goat polyclonal antibody against HAPLN1 (1:1000, catalog number sc-46826; Santa Cruz Biotechnology) at 4 °C overnight. We used protein lysates obtained from canine articular cartilage as positive controls. Membranes were incubated with the corresponding secondary antibody, horseradish peroxidase (HRP)-conjugated rabbit/goat anti-immunoglobulin G (anti-IgG; Bio-Rad Laboratories) diluted 1:10,000 in 1 % bovine serum albumin (BSA) for 2 h at room temperature. After each step, blots were washed three times with TBS with Tween 20 (0.1 %). Protein bands were detected by the enhanced chemiluminescence method (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) on Kodak Hyperfilm (Amersham; GE Healthcare, Little Chalfont, UK).

Histology and immunohistochemistry

To validate the cellular morphology and characteristics of the discs obtained from both NCD and CD animals, we stained representative sections of the discs using hematoxylin and eosin and Safranin O. We performed immunohistochemical analysis for fibromodulin, HAPLN1, biglycan, decorin, and aggrecan using independent sets of paraffin-embedded sections of IVD NP from the same animals used for iTRAQ analysis. In brief, the sections were deparaffinized in xylene, hydrated in gradient alcohol, and pretreated in a microwave oven for 10 minutes at maximum power in Tris–ethylenediaminetetraacetic acid (EDTA) buffer (0.01 M, pH 9.0, 0.05 % Tween-20) for antigen retrieval. The sections were incubated with hydrogen peroxide (0.3 % vol/vol) in PBS (0.1 M, pH 7.2) for 15 minutes to quench the endogenous peroxidase activity, followed by blocking with 5 % BSA to preclude non-specific binding. Thereafter the slides were incubated with fibromodulin, HAPLN1, biglycan, decorin, and aggrecan antibodies for 16 h at 4 °C. The primary antibody was detected using the streptavidin–biotin complex with the Dako LSAB+ kit (Dako Cytomation, Glostrup, Denmark) and diaminobenzidine as the chromogen. All procedures were carried out at room temperature unless otherwise specified. The slides were washed three times after every step using PBS containing 0.025 % Triton X-100, and finally the sections were counterstained with Mayer’s hematoxylin and mounted with DPX mountant (distrene, plasticizer, and xylene). We replaced the primary antibody with isotype-specific non-immune goat/rabbit IgG for negative controls, and all sections were evaluated by light microscopic examination.

Proteoglycan extraction and SLRP assay

As described above, entire NPs were removed from five spinal segments of each of three CD and NCD donor animals, all of which had the same gross morphological appearance and were within the same age range as the donors of our iTRAQ samples. Proteoglycans were then extracted using GuHCl according to the method described by Melrose et al. [6, 20]. NCD and CD NP were diced, weighed (approximately 0.4 g wet weight), and extracted with 10 volumes of 4 M GuHCl 100 mM acetate buffer (pH 5.8 containing 20 mM EDTA, 100 mM 6-amino hexanoic acid, 25 mM benzamidine) for 40 h at 4 °C with constant end-over-end mixing. The tissue residue was removed by centrifugation, and the supernatant was concentrated by ultrafiltration (Amicon Ultra-15, 15 ml, 3000 kDa, catalog number UFC900308; EMD Millipore, Billerica, MA, USA), dialyzed against three changes of deionized cold water using a D-Tube Dialyzer Maxi (MWXO 3.5 kDa, catalog number 71508-3; EMD Millipore) with a fourth overnight dialysis. The dialyzed extracts were then freeze-dried. The freeze-dried proteoglycans thus extracted were dissolved in 2.1 ml of digestion buffer (0.1 M Tris, 0.03 M acetate, pH 6.5) overnight at 4 °C with constant shaking, and the protein content of the dissolved extracts was calculated using a Bradford assay. Briefly, 500 μg of protein from NCD and CD samples were digested with 1.0 U of chondroitinase ABC (C3667; Sigma-Aldrich) and 0.05 U of keratinase (G6920; Sigma-Aldrich) overnight at 37 °C. Next, 30 μg of total protein obtained from the extracted proteoglycans (digested and undigested) were mixed with SDS loading buffer and dithiothreitol, electrophoresed, and visualized by Western blotting as described above. For the decorin and biglycan Western blot primary antibodies, we used a rabbit polyclonal antibody against decorin (1:200, catalog number NBP1-57923; Novus Biologicals, Littleton, CO, USA) and a goat polyclonal antibody against biglycan (1:200, catalog number sc-27936; Santa Cruz Biotechnology). The secondary antibodies were both HRP conjugates. We used a goat anti-rabbit IgG (heavy and light chains, catalog number (170-6515; Bio-Rad Laboratories) and a rabbit anti-goat IgG (heavy and light chains, catalog number#A24452; Life Technologies, Carlsbad, CA, USA), both used at 1:10,000 dilution.

Biomechanical testing

We harvested the L4-L5 lumbar motion segments from CD and NCD animals within 1–4 h after they were sacrificed. These segments were removed from the spines from which the IVDs were subjected to both iTRAQ analysis and Western blot analysis. CD and NCD canines (n=6 samples each) were age-matched (20–24 months) and had average weights of 10 kg and 14 kg, respectively. Following lumbar segment harvesting, the muscle tissues, intertransverse ligaments, and supra- and interspinous ligaments were carefully dissected, leaving the posterior joints and the IVD intact. All segments were kept frozen at −20 °C until testing. The IVD dimensions of each specimen were measured, and the cross-sectional areas were calculated. Before testing, the lumbar segments were thawed at room temperature, and pilot holes were then drilled into each VEP. Each hole accommodated two anchor screws (#3 × 25.3 mm) in the superior L4 end plate and one screw (#8 × 43 mm) in the inferior L5 end plate, perpendicular to the middisc plane. Dental stone (Modern Materials; Heraeus Kulzer, South Bend, IN, USA) was prepared according to the manufacturer’s recommendations and used to cement the segments within 3.5-inch–diameter plastic pots. The two vertebra composing the motion segments (inclusive of the intervening IVD) were seated within opposing pots using a standardized template, and cement was added to the pot until two-thirds of the vertebral body height was covered, leaving the IVD exposed (Fig. 1). Each segment was kept hydrated with gauze soaked in isotonic saline and covered in plastic wrap during tissue preparation, potting, and testing. The pot containing the caudal vertebra (L5) was fixed to a stationary crossbeam, and the pot containing the cephalad vertebra (L4) was fixed to a six-axis load cell (AMTI MC3A-100, Advanced Mechanical Technology, Watertown, MA, USA), which in turn was attached rigidly to the parallel robot (Parallel Robotic Systems, Hampton, NH, USA). This robot is composed of a rigid platform suspended by six rigid struts of fixed length. Each strut was attached to an electromechanical motor that travels about a circular track to alter the position and/or orientation of the robot platform and thus the potted motion segment. Custom computer software (National Instruments, Austin, TX, USA) was employed to control robot position and orientation. The resolution of the robot is a function of the motor performance (0.05 mm with a repeatability of 0.025 mm; Mikrolar, Hampton, NH, USA), which translates into a linear resolution of less than 1 μm and an angular resolution of approximately 0.001 degree [20].

The methodology used for biomechanical testing involved a standardized procedure according to an established testing protocol [21, 22]. Specimens were tested using an unconstrained force control system that permitted the application of a desired load in a specific axis of motion without any constraints on the segment’s motion behavior, regardless of the specimen condition (e.g., intact, injured) [23]. Specifically, we employed our software to position the specimen in a neutral pose (e.g., loads and moments in all directions were minimized), after which a bending moment about all three axes of movement (i.e., flexion and extension, lateral bending, and axial rotation) (Fig. 1c) was applied until the moment obtained from the load cell reached a maximum predetermined value. When the maximum moment was reached in the positive direction, the direction of rotation was reversed to provide a corresponding negative moment. Based on preliminary tests conducted in three additional CD specimens (not included in this study), maximum moments of newton meters were predetermined for flexion and extension, ±3 Nm for lateral bending, and ±4.5 Nm for axial rotation [24, 25]. The robot’s displacement during specimen rotation was recorded directly through a software query via the robot. This query and collection of specimen moment and load data occurred at 200 Hz. Five moment cycles around each Cartesian axis were performed at 0.1/s. The first two cycles were applied to precondition the specimens and load deformation data from the last three cycles used in the statistical analysis [26, 27]. The specimens’ moment and rotation data were plotted, and specific point displacements were obtained at the neutral position, the midpoint of the ascending moment, the peak moment, and the midpoint of the descending moment for both positive and negative rotations.

Statistical analysis

Before statistical analysis, the biomechanical data were normalized to the cross-sectional area of the IVDs to account for any differences in the sizes of the respective IVDs. We used the Shapiro–Wilk normality test to verify the normal distribution of data obtained from biomechanical testing, and numeric data were represented as the mean ± standard deviation. We compared normalized displacement mean values between the CD and NCD groups using an independent t test, with p < 0.05 considered statistically significant. The proteomic data are presented as weighted averages of the ratios of the relevant peptides (ratio calculated with respect to the reference standard) as described, and the data were normalized using the applied bias with β-actin as a normalizing factor. All proteins and peptides were statistically analyzed within the iTRAQ analysis software platform (ProteinPilot) after having achieved a 95 % confidence interval and 1 % global FDR cutoff plus up- or downregulation with respect to the reference pool of at least 0.5 of greater than 2.0-fold. The proteins and peptides identified by iTRAQ and their expression ratios are therefore not directly comparable to each other.