Mechanical and biological characterization of a composite annulus fibrosus repair strategy in an endplate delamination model

Abstract This study compares the mechanical response of the commonly used annulus fibrosus (AF) puncture injury model of the intervertebral disc (IVD) and a newly proposed AF failure at the endplate junction (delamination) on ex vivo bovine IVDs. Biocompatibility and mechanics of a newly developed repair strategy comprising of electrospun polycaprolactone (PCL) scaffold and fibrin‐genipin (FibGen) adhesive was tested on the delamination model. The study found no significant difference in the mechanical response to compressive loading between the two models. Primary goals of the repair strategy to create a tight seal on the damage area and restore mechanical properties, while showing minimal cytotoxicity, were broadly achieved. Postrepair, the IVDs showed a significant restoration of mechanical properties compared to the injured samples for the delamination model. The FibGen glue showed a limited toxicity in the AF and produced a resilient and mechanically stable seal on the damaged area.


| INTRODUCTION
Intervertebral disc (IVD) herniation is a very common and painful condition, where the outer layers of the IVD known as the annulus fibrosus (AF) are ruptured, which results in leakage of the inner nucleus pulposus (NP) material and consequently deterioration of the mechanical function of the organ. This damage is often sustained during intensive physical activity or due to tissue degeneration as a result of a pathology or aging. Almost 80% of the population experience lower back pain, of which approximately 40% is a result of IVD degeneration. 1,2 Herniation of the IVD can occur through the endplate in the cranial-caudal direction, or laterally through AF failure.
The injury of the AF and NP herniation are often followed by further degeneration due to immune responses, provoked inflammation and an array of circulating cytokines. [3][4][5] Severe back pain symptoms arise as a result of these processes as well as NP material directly impinging on the nerve root, and vascular and nerve invasion of the injury site. [6][7][8] To evaluate effective treatment methods, the injury model and the repair strategy need to be first considered individually and subsequently tested in conjunction.
To establish an ex vivo injury model which is most representative of the naturally occurring case, several types of AF ruptures that can lead to circumferential NP extrusion were considered. Two general rupture types were identified in the literature, including failure at the endplate junction and failure through rupture of the AF in central region. It has been shown that 65% of discs rupture at the endplate, where the most common mode is through rim fracture. 9 These findings were supported by other studies documenting a frequent presence of cartilage tissue in herniated material, 10,11 as well as a weak interface at the endplate in extension loads. 12 Failure through AF rupture in the central region is commonly investigated in the literature.
Multiple studies have characterized the pathomechanism of the injury 4,13-15 as well as proposed repair strategies. The repair strategies can be divided into the strategies focused primarily on mechanical repair and sealing, largely reliant on synthetic materials, and hybrid strategies that aim not only to provide mechanical support but to also promote remodeling of the damage site through loading synthetic or tissue engineered matrices with live cells. Approaches adapted from existing strategies such as sutures 16 have shown promising results in preventing re-herniation in the short term. The use of hydrogels made up of collagen [17][18][19][20] and fibrin-genipin (FibGen) 21 as fillers for AF defects have shown to limit biomechanical deterioration. The fillers have also been enhanced with a retention scaffold on the IVD surface [22][23][24] to further reinforce the injury site. The second class of strategies applies tissue-engineering principles to replace the injured AF with cell-laden scaffolds [25][26][27][28] or cell sheets. 29  To repair the ruptured disc, a system that is biologically and biomechanically compatible must be chosen. Considering the inherent properties of the AF, mechanical design criteria that will allow for restoration of the original mechanical properties can be defined. The material of choice should be capable of withstanding large strains up to 40%, based on the surface strains observed in IVDs. 30,31 Additionally, a tight seal on the injured area is required to prevent nerve and soft tissue ingrowth into the injury, as well as prevent further degeneration leading to NP extrusion which can both cause severe pain. 6,7,32 Furthermore, the repair construct should exhibit a modulus which is lower than that observed in the outer layers of the AF. Based on the review by Long et al, 33 a range of mechanical properties suitable for the repair were identified.
Electrospun (ES) poly(ε-caprolactone) (PCL) was chosen to fulfill the role of principal mechanical support. PCL is commonly used as a substrate providing mechanical support and guiding cellular activity in tissue engineering (TE) and regenerative medicine. 34 The ES networks can be produced with wide variety of fiber diameters and morphologies depending on the production process, and thus they can closely mimic the scale of structures and fibrous morphology of the extra cellular matrix. [35][36][37] The mechanical properties of these scaffolds can be tuned through alteration of spinning parameters to achieve various morphologies, fiber diameters, and alignments. Previous research has shown that PCL can achieve mechanical properties in the required range of 1 to 40 MPa 38 ; furthermore, the anisotropy requirements have been shown to be fulfilled through fiber alignment. 39 The required strain range of up to ±40% strain is also within the capabilities of the scaffolds. To attach the ES scaffolds on the surface of the IVD, FibGen hydrogel glue was chosen as it allows good fixation through chemical bonding with the IVD surface and mechanical interlock with the ES scaffolds. FibGen has been previously shown to have strong and tunable fixation for collagen rich soft tissues, 40,41 and mechanical properties in the range of native AF tissue. 42 We hypothesized that the new endplate delamination model would provide a medically relevant damage scenario, which would be significantly different from those previously used. 4,13-15 Furthermore, we hypothesized that the repair strategy devised would allow for the restoration of the mechanical response of the damaged IVD, create a tight seal on the damage area and would be cytocompatible. To investigate these hypotheses, a bovine tail IVD model was chosen, as it is well established and provides organs of similar size to human IVDs, with comparable mechanical properties. 43,44 2 | METHODS

| Study design
To investigate the effect of the injury on the IVD, two types of injuries were induced (puncture and delamination). The IVDs were extracted from bovine tails and separated into three groups (intact, delaminated, and punctured) with four samples in each group. The intact group was kept undamaged as a control; the punctured group used the traditional central AF injury model, while the delaminated group applied the new damage model near the endplate. The samples were tested on a schedule described in Table 1. For the investigation of the repair strategy, samples were also separated into three groups (intact, injured, repaired) with eight samples in each group. Only the delamination model was used as the injury and is therefore referred to as just the injured group. The samples were tested on the schedule described in Table 2.

| IVD preparation
IVDs comprising cartilaginous endplates were harvested from bovine tails (10-12 months old), obtained from a local abattoir within 2 hours of death, and washed in phosphate buffered saline (PBS) containing 10% penicillin-streptomycin (Gibco, Zug, Switzerland) for 10 minutes, followed by a second wash in PBS with 1% penicillin-streptomycin.
Three bovine tails were dissected for the injury model comparison, while five bovine tails were dissected for the repair strategy efficacy investigation. Six discs were collected from each tail. Average diameter was 18.36 mm ± 2.28 SD, average height was 10.90 mm ± 1.34 SD. The IVDs were assessed visually for signs of degeneration or abnormal growth. Discs were cultured overnight in Dulbecco's modified Eagle's medium (4.5 g/L glucose) with 2% fetal bovine serum (Gibco), 1% insulin transferrin selenium, and 0.2% Primocin (Invivogen, Nunnigen, Switzerland).

| Puncture model
The IVDs were injured using a biopsy punch (diameter 4 mm, length 7 mm, Kai Medical, Gifu, Japan) in the central region of the AF, penetrating to the NP region ( Figure 1D). The methodology was adapted from Li et al. 13

| Delamination model
The IVDs were injured using a No. 10

| IVD repair procedure
For the repair procedure, FibGen glue in two parts and the scaffold were prepared separately. Prior to application, the scaffold was soaked in thrombin-genipin solution, followed by application of fibrinogen-PBS on the IVD surface. The scaffold was then applied and the repaired IVDs were securely wrapped in sterilized paraffin film to create a tight seal for 20 minutes after which the repaired samples were kept in a six well plates with gauze soaked in PBS + 1% penicillin-streptomycin (Gibco) for at least 1 hour at 37 C.
The samples were then placed back in culture medium in the incubator. The resulting repair can be seen in the hystological section in

| Mechanical and physical disc characterization
Two types of mechanical loading scenarios were implemented. A lower stress 2-hour bioreactor loading protocol to simulate daily activity and a high stress 15-minute compression mechanical testing, which was used to evaluate the mechanical performance of the IVDs. The bioreactor loading and compression testing schedule is briefly described in Table 1 for the comparison of different injury models and in Table 2 for the investigation into the repair strategy.
Dimensional measurements were taken before and after each mechanical loading for both protocols. The discs were kept under free swelling conditions between all bioreactor loadings and mechanical tests. The discs were kept in custom-made chambers F I G U R E 2 Overview of experimental design starting with intact healthy disc, the injury was induced with scalpel blade to produce an injury 4 mm deep and 90 section as close as possible to the endplate, finally the IVD was repaired using the electrospun PCL scaffold and FibGen glue. Histology images with Safranin-O/Fast Green staining obtained after 14 days in culture and 7 days after repair was applied show the three states in which the IVDs were tested. Note the close adhesion of the repair construct to the IVD surface. * NP, ** AF, *** repair scaffold. AF, annulus fibrosus; ICD, intervertebral disc; NP, nucleus pulposus; PCL, poly(ε-caprolactone) The overview of a histological section stained with Safranin-O/Fast Green stain of the delamination injury after a total of 14 days in culture, A, here proteoglycans are stained red and collagen is stained blue, live and dead cells are stained black. A closer look at the injury site with the same stain, B, and LDH stain of the injury site with live cells stained black, C. The gradient of cell density from the highly cellularized outer AF to the NP is clearly visible in the LDH stain. An overview of the punctured disc after 7 days in culture, D. The disturbance in the NP is in present here unlike in the delamination injury. * Injury site, **AF, ***NP, EP is the endplate location. AF, annulus fibrosus; NP, nucleus pulposus; LDH, lactate dehydrogenase during dynamic loading and mechanical compression testing, and in six well plates during free swelling.

| Dimensional measurements
The cross-sectional measurements used to calculate stresses were taken directly after dissection using precision calipers. Crosssectional area was calculated assuming an oval cross-sectional shape.
Height measurements were taken daily before and after mechanical loading procedures using force limited calipers at 5 to 10 N. The height was calculated from the average of three measurements.

| Bioreactor loading
Discs were cultured in a bioreactor system with physiological loading regime. The bioreactor was maintained in an incubator at 37 C, 85% and 500 ng of total RNA according to the manufacturer's protocol.
StepOnePlus System (Applied Biosystems) was used to conduct quanti-

| Histology
After removal of the bony endplate from both sides with a drill, the IVDs were frozen in cryoembedding compound (Sysmex, Horgen, CH). Transverse and sagittal sections (10 μm) were cut with a microtome (Microm, Germany).

| Safranin-O/Fast Green staining
The extracellular matrix (ECM) in the injury and repair sites, as well as the native tissue was qualitatively evaluated by Safranin-O/Fast Green staining. Sections were fixed in 70% methanol before staining.
Sections were stained with 0.1% Safranin-O and 0.02% Fast Green to reveal proteoglycan and collagen deposition, respectively, and counterstained with Weigert's hematoxylin to reveal cell distribution.

| Lactate dehydrogenase and DAPI staining
The cell viability at the injury and repair sites, as well as the native tissue was quantitatively evaluated by staining. The cryosections were stained using lactate dehydrogenase (LDH) solution as described previously. 13 Staining was performed with LDH in 40% polypep solution (Sigma-

| Glycosaminoglycan and nitric oxide medium assays
The total glycosaminoglycan (GAG) content of the culture media, collected at each media change, was determined by the dimethylmethylene blue dye method, using bovine chondroitin sulfate as the standard, 48 to assess the release of matrix molecules from the sample into the media. Absorbance was measured with a Victor3 PerkinElmer (Waltham, Massachusetts) 1420 multilabel counter.
Levels of nitric oxide (NO) production in the conditioned medium of IVDs were determined as the concentration of its stable oxidation product, nitrite (NO 2− ), using the Griess Reagent Kit (Promega, Madison, Wisconsin). 49 The results were calculated based on the medium volume in each well or bioreactor chamber and then divided by the volume of each IVD measured after extraction on day 0.

| Data postprocessing and statistics
Statistical analysis was performed using GraphPad Prism 8.2 software (GraphPad Software, Inc., La Jolla, California), unless stated otherwise.
The viability of the cells in LDH measurements was analyzed using independent two-sample t test at each location, with Holm-Sidak multiple comparison correction, due to the large dataset (three transverse sections from one sample and three sagittal sections from another sample, n = 6). PCR data were first preprocessed to eliminate outliers using ROUT method at Q = 1%, the normality of the dataset was then assessed and was found to be non-parametric in most samples due to low sample number: n = 4. Kruskal-Wallis test was then used to assess the significance of the results with multiple comparisons. The overall GAG and NO content was assessed through a Wilcoxon matched pair test, while individual time points were analyzed using an independent two-tailed two-sample t test, with intact samples as control (n = 8).
The mechanical data were analyzed with the SPM1D package for Python 50 using a two-tailed independent two-sample t test based on random field theory, with intact samples as control (n = 8). P < .05 was considered statistically significant (*P < .05, **P < .005, ***P < .0005).

NO and GAG medium assay
The GAG and NO content of the medium was measured daily from samples taken before and directly after the loading cycles. The total amounts were calculated by combining the results from directly after the loading and the amount released into the medium during the overnight incubation. The NO and GAG amounts per disc as a function of time can be found in Figure 10. The overall GAG in the medium measured in μg per mm 3 of an IVD was significantly higher (P < .05) for injured samples  In the puncture model, the NP and AF are severely disturbed by the puncture as observed in Figure 1D and were able to immediately protrude from the injury site, while in case of the delamination injury, the NP was largely undisturbed, as seen in Figure 1A  The two different damage models investigated in this study did not show a significant difference in mechanical response between the models or from the control that was kept intact. This is likely due to The increase in minimum strain reflects the inability of the IVD to recover height quickly after load is released, likely due to increased hydraulic permeability as noted by Grunert et al. 19 The delamination model was therefore chosen for the second phase of the study where the repair strategy would be investigated, as it showed potential for significantly different mechanical properties.

| Mechanical and biological evaluation of the endplate delamination injury and repair strategy
The Similarly, to the comparison of the injury models, the mechanical evaluation was limited to compressive loading due to biological considerations, preventing fixation at the vertebra. However, unlike the injury model investigation, significant differences were found for the minimum strain between the intact and injured samples. This measurement corresponds to the ability of the disc to recover height under load. This ability was significantly reduced for the injured samples at all time points. On the other hand, it was restored to the intact values for the repaired samples on day 14, which corresponds to 7 days after repair. This suggests that the full crosslinking process proceeds for more than 24 hours after the application. In particular, this finding could be a result of genipin crosslinking collagen in the AF matrix, which would lead to a stiffer and more elasticity in AF. The diffusion process required for the genipin to penetrate into the AF would slow down the reaction, unlike in the FibGen hydrogel where the components are liquid and the layer is thin compared to AF, as seen in Figure 3. The stiffening of the AF on genipin exposure is further supported by Fessel et al, 54 where increased stiffness was observed in collagen networks on application of genipin.
One limitation of comparing the biomechanics results in the current study to many other repair strategies that have been previously attempted in literature is the type of injury that is repaired. All studies that are cited here use various types of puncture injuries in the central AF. For small animal models, a needle puncture is used or AF tissue removal for larger defects, for large animal models a biopsy punch is used or once again AF tissue is removed directly. As discussed in the comparison of the injury models, the commonly studied puncture injury likely has a different mechanical effect than the delamination injury. With that in mind, the minimum strain in this study can be considered analogous to the height loss of the IVDs postloading, which is commonly reported in literature. Many studies have recorded disc height loss after AF injury, and disc height is considered a benchmark for both degeneration and success of the repair strategy clinically. 55 The results in this study are in agreement with multiple studies that have shown successful repair of the injuries to the central AF in vivo. 18,23,[27][28][29] Both in terms of significant height loss in injured IVDs and disc height restoration and improvement due to repair. These The overall IVD modulus measurements in this study, however, did not show significant differences between groups. This finding is in agreement with study by Likhitpanichkul et al,21 where no significant change in stiffness was also observed on a whole organ model of an IVD, although the injury and repair had different geometry than the present study.
The impact the injury and repair have on the cell viability was measured through LDH-based live/ dead staining. The injury was found to have no effect on cell viability after 14 days. However, there was a significant decrease in cell viability on all exposed surfaces of shown to promote cell infiltration and effectively seal AF defects in small animal models in vivo and restore some aspects of biomechanics. 19,27,28 Whether this approach can translate to large animal models is questionable as the biomechanical environment is far more sever.
Further doubt is cast in work by Pennicooke et al, where the efficacy of such hydrogels in ovine in vivo model was investigated and found that while histologically the quality of the NP was improve, the biomechanics and disc height loss was not significantly different from the injured IVDs. 20 This is likely due to insufficient stiffness of the collagen hydrogels required for large animal models. FibGen shows compressive elastic modulus on the scale of 100 to 200 kPa, 58 while highdensity collagen gels are an order of magnitude softer at approximately 3 kPa. 59 In addition, the need for UV light for crosslinking would not allow for the use of opaque ES membranes for resilience.
Cyanoacrylate glues are widely used clinically, however the brittle nature of these glues coupled with potential toxicity concerns 60 would not give them an advantage over the use of FibGen. As such, the compromise of using potentially toxic genipin as crosslinker has to be considered. An improved procedure minimizing tissue exposure to genipin and DMSO can provide the best possible outcome.
No cell infiltration into the scaffold was observed after 7 days in culture. The location of the repair as well as its geometry are likely the reasons for no observed infiltration in this study. In Figure 8, it can be seen that there is proportionally more live cells in inner regions of AF as opposed to the surface. Therefore, the application of the scaffold on the very outer surface of the AF in this study limits the cells access to the scaffold. Furthermore, FibGen is embedded in the ES fiber network with the porosity in the range of 1 μm, which is at the lower limit of porosity large enough to allow for cell infiltration. were not significant. 22 The downregulation of Col2 in injured samples is also consistent Pirvu et al, 26 it was also documented as part of degenerative process without inducing an injury by Lang et al. 47 Very large and significant upregulation of MMP1, ADAMTS4 and a trend of upregulation of MMP13 in the AF of the repaired and injured samples at the injury site was not consistent with previous studies, where downregulation of catabolic genes was observed with diurnal loading. 22,26 As most markers (anabolic and catabolic) were upregulated in the AF, this may be related to a response to the injury of repair and remodeling of ECM in the AF.
The GAG content measured in the medium showed a higher value in injured samples compared to intact ones at most time points. This is likely due to increased surface area and ECM breakdown/ remodeling in the injury group. The repaired samples showed a significantly lower amount of GAG content in the medium than the injured samples 3 days after repair. This further supports the idea that the full reaction and sealing of the defect happens slowly after the initial setting and adhesion of the hydrogel. The levels of NO found in the media, which were higher for injured samples than other groups are in agreement with previous work done by Likhitpanichkul et al. 21

| Limitations of the study
The chosen ex vivo bovine IVD from the tail model used in this study has limitations in terms of transfer to in vivo human clinical applications. Further in vivo experiments on large animal model are required to optimize the application procedure and confirm its viability. The mechanical properties and size are similar between the human and bovine IVDs. 44 However, biologically, the tissue compositions as well as cell types vary significantly between the two models. 61 In particular, discs extracted from the tails of calves in this study are likely most representative of IVDs from healthy young humans.
The delamination model does not address the NP herniation and potential further tissue degeneration of the AF and NP. This limitation arises due to the nature of the injury being surgical rather than induced through degeneration of loading to failure. Therefore, it is representative of early stages of AF tare at the endplate. The repair in the presented form would need to be applied soon after the injury occurs and would likely not be suitable for severely degenerated discs, as it requires relatively healthy AF of adhere to for mechanical stability. In the future, it would be possible to explore injuries induced through flexural tensile loading of the disc, which is most likely to lead to delamination of the AF at the endplate. 12,62,63 To assess the degenerative effects of the injury a much longer ex vivo study is required likely involving alterations in the IVD culture to induce degenerative changes mechanically or chemically. 47

| CONCLUSION
AF rupture is a common condition that can lead to severe pain and reduction in mobility. A new injury model similar to failure of the AF at the endplate junction was investigated and was found to be mechanically not significantly different from previously described rupture of the AF in compressive loading. A repair strategy utilizing ES PCL scaffold and FibGen adhesive was investigated through a long-term ex vivo biomechanical testing protocol. The repair strategy showed a promising restoration of mechanical properties to the levels found in intact IVDs, while FibGen adhesive showed limited cytotoxicity in the AF. In addition, the adhesion of the scaffold to the injury site had a good seal and remained intact throughout the procedure. Individual components of the repair were not tested separately; hence, their specific contributions to the restoration are not reported. Furthermore, the degradation rates of the components may be significantly different and require further tuning.
The application technique used in this study may not be suitable for in vivo procedure and may require further optimization. As such, to conclusively define the efficacy and biocompatibility of the proposed approach an in vivo study would be required in the future.