Excessive Activation of TGFβ by Spinal Instability Causes Vertebral Endplate Sclerosis

Narrowed intervertebral disc (IVD) space is a characteristic of IVD degeneration. EP sclerosis is associated with IVD, however the pathogenesis of EP hypertrophy is poorly understood. Here, we employed two spine instability mouse models to investigate temporal and spatial EP changes associated with IVD volume, considering them as a functional unit. We found that aberrant mechanical loading leads to accelerated ossification and hypertrophy of EP, decreased IVD volume and increased activation of TGFβ. Overexpression of active TGFβ in CED mice showed a similar phenotype of spine instability model. Administration of TGFβ Receptor I inhibitor attenuates pathologic changes of EP and prevents IVD narrowing. The aberrant activation of TGFβ resulting in EPs hypertrophy-induced IVD space narrowing provides a pharmacologic target that could have therapeutic potential to delay DDD.

μCT. The lower thoracic and whole lumbar spine from mice were dissected, fixed in 10% buffered formalin for 48 h and then transferred into PBS, examined by high-resolution μ CT (Skyscan1172). The ribs on the lower thoracic were included for identification of L 4 -L 5 IVD localization. Images were reconstructed and analyzed using NRecon v1.6 and CTAn v1.9, respectively. Three-dimensional model visualization software, CTVol v2.0, was used to analyze parameters of the L 4 -L 5 IVD with half height of L 4 and L 5 vertebrae. The scanner was set at a voltage of 49 kVp, a current of 200 μ A and a resolution of 6.8 μ m per pixel to measure the IVD and EP. A resolution 16.8 μ m of per pixel was set for the whole L 5 vertebral body measurement. Coronal images of the L 4 -L 5 IVD were used to perform three-dimensional histomorphometric analyses of IVD and cartilage EP while sagittal images of L 5 vertebra were used for those of vertebral body. IVD volume was defined by the region of interest (ROI) to cover the whole invisible space between L 4 -L 5 vertebrae. Cartilage EP volume was defined to cover visible bony plate close to the vertebrae. L 5 vertebral body TV was described to figure out the medial compartment excluding cortical bone, transverse and spinous processes. A total of five consecutive images of ROI were used for showing three-dimensional reconstruction of IVD space and EP. Three-dimensional structural parameters analyzed included: TV (total tissue volume), and Trabecular separation distribution.
CT-based microangiography. Blood vessels in EP sites were imaged by angiography of microphil-perfused bones. In detail, the thoracic cavity of mice was opened after anesthesia, and the inferior vena cava was severed. The vascular system was flushed with PBS containing heparin sodium (100 U/ml) through a needle inserted into the left ventricle. The specimens were then pressure fixed with 10% neutral buffered formalin which was washed off from the vessels by heparinized saline solution. A radio paque silicone rubber compound containing lead chromate (Microfil MV-122, Flow Tech) was injected to label the vasculature. Samples were stored at 4 °C overnight for contrast agent polymerization. Mouse C 8-9 along with EPs were dissected and decalcified in 10% EDTA after fixation in 10% neutral buffered formalin for 24 h. Images were obtained using a μ CT imaging system (Skyscan 1172).

Results
Aberrant mechanical loading in the spine leads to narrowing of IVD space. Narrowed IVD space is often seen with human aging and regarded as a gold standard of DDD. Here, we employed a lumbar spine instability mouse model (LSI) (Fig. 1A) to examine whether altered mechanical loading in the spine leads to narrowing of IVD space. In detail, the L 3 -L 5 spinous processes along with the supraspinous and interspinous ligaments were removed to induce instability of mouse lumbar spine (Fig. 1A). Mid-sagittal images of μ CT suggested that the space between L 4 and L 5 vertebral bodies decreased in LSI mice, particularly in the posterior region (Fig. 1B). Analysis by three-dimensional reconstruction images of L 4-5 EPs demonstrated that 8 weeks old mice continue to have growth of the spine, demonstrated by an increase in IVD volume 2 weeks post-sham surgery. The IVD volume then began to decrease up to 8 weeks post-sham surgery. In the LSI mice, there was no increase in IVD volume 2 weeks post-surgery, leading to statistically significant decrease in IVD volume relative to sham controls of similar age. The LSI mice continued to have a decreased IVD volume up to 8 weeks post-surgery relative to age matched sham controls (Fig. 1C,D).
Narrowed IVD space is due to EP hypertrophy. To explore if the narrowed IVD space in the unbalanced mechanical loading environments was caused by surrounding hard tissue enlargement, we examined the sizes of both cranial and caudal EPs as well as L 5 vertebral body. The results showed that both cranial and caudal EP volumes were significantly increased post-surgery ( Fig. 2B,C). An increased size of cavities within the EP were also noted in μ CT 3-D images ( Fig. 2A). The volume of L 5 vertebral body was not affected (Fig. 2D). Further characterization of EP morphology by μ CT analysis revealed an increase in trabecular separation distribution (Fig. 2E). Specifically, the percentage of values between 0.048 mm to 0.089 mm and above 0.089 mm reached to . This value of EPs in 2 w-LSI mice was similar to that in 4 w-Sham mice, indicating LSI surgery accelerated the development of cavities in EPs. The percent of values greater than or equal to 0.089-were increased, but subsequently declined, suggesting it might be increased activity of EP remodeling in LSI mice.

EP sclerosis is caused by accelerated EP bone remodeling.
To test if the change in EP cavity size was due to accelerated bone remodeling, we performed histological studies to examine how these cavities form. EP score, a histologic assessment of EP degeneration was applied for evaluation to account for pathological changes such as degrees of bony sclerosis, structure disorganization, and neovascularization 1 . We found EP score increased in the LSI mice compared to the Sham mice at similar time points, confirming EP degeneration (Fig. 3A). Safranin O and fast green staining suggested that EPs began to undergo endochondral ossification by 2 weeks post-surgery, as indicated by green-stained bone matrix surrounding the cavities in LSI mice relative to sham group (Fig. 3B). Interestingly, Trap and cathepsin K positive cells were noted beginning at 2 weeks post-surgery in LSI mice, whereas Trap and cathepsin K positive cells were rarely detected in sham controls at any time points (Fig. 3C-F), suggesting osteoclast resorptive activity increases after spinal instability.
The organization of type II collagen, which is normally detected in cartilage, was altered, with loss of type II collagen in the observed cavities and a change in spatial orientation in the remainder of the EP in LSI mice (Fig. 3G). Chondrocyte hypertrophy as assessed by type X collagen positive staining in LSI EPs decreased relative to the Sham EPs (Fig. 3I). Moreover, the distribution of the type X collagen positive cells was altered. In the Sham EPs, type X collagen positive cells were uniformly distributed (Fig. 3H Left), whereas in the LSI EPs, the positively staining cells were noted at sites where the chondrocytes merged together (Fig. 3H Right upper), but no positive cells were noted in larger EP cavities (Fig. 3H Right lower). Furthermore, expression of type I collagen, which is normally detected in ossified bone, was increased around the cavities in LSI EPs (Fig. 3J,K), which were regions where type II collagen were no longer expressed (Fig. 3E).
We performed further studies to determine if osteoblasts lineage cells were also present in the endochondral ossfication. Immunostaining for nestin, osterix and osteocalcin, markers of mesenchymal stem cells, pre-osteoblasts and mature osteoblasts, respectively, demonstrated a similar pattern as Trap staining. Specifically, all nestin, osterix and osteocalcin positive cells were abundant in the LSI mice compared to rarely detected in the sham controls (Fig. 3L-Q). Moreover, Nestin-Cre TM Er::Rosa26-lacZ flox/flox mice were used to monitor the change of numbers and location of MSC lineage cells in EP after LSI operation. β -galactosidase (β -gal)  staining of Nestin-Cre TM Er::Rosa26-lacZ flox/flox mice revealed that β -gal + MSC lineage cells were increased in EP of LSI-operated mice, while no β -gal + staining could be seen in EP of sham-operated mice (Fig. 3R).

Upregulation of active TGFβ leads to EP degeneration and narrowed IVD space. Excess activa-
tion of TGFβ has been found to contribute to the pathogenesis of sclerotic subchondral bone leading to OA 7,8 . In our LSI mouse model, the total TGFβ 1 protein level was elevated 2 weeks post-LSI surgery, but was not significantly changed at later time points (Fig. 4A). The percentage of TGFβ that was active was increased at 2, 4 and 8 weeks post-LSI surgery (Fig. 4B). Downstream TGFβ signaling was confirmed by IHC examination, showing pSmad2/3 positive staining in both hypertrophic chondrocytes and bone marrow cells in EP cavities (Fig. 4C). To determine if the elevated TGFβ was involved in EP degeneration, we utilized a transgenic mouse model (CED) that results in overexpression of active TGFβ driven by a 2.3-kb type I collagen promoter 13 , as type I collagen is mainly expressed in the outer layer of the annulus fibrosis and is in direct contact with the EPs 15 . The EP phenotype in the CED mice was similar to the LSI mice. Specifically, mid-sagittal μ CT scan revealed that the IVD space between L 4 and L 5 vertebral bodies significantly narrowed in CED mice relative to their WT littermates (Fig. 4D,E). Three-dimensional reconstruction and analysis of L 4-5 EPs revealed a significant increase in both cranial and caudal EP volumes in CED versus WT littermates (Fig. 4F-H). Increased EP score was observed in CED mice compared to WT littermates (Fig. 4I). Histologic analysis confirmed that the decreased IVD height was associated with endochondral ossification of EPs in CED mice relative to their WT littermates (Fig. 4J). Taken together, the data indicate that upregulation of TGFβ is involved in sclerosis and hypertrophy of EP, which occurred in association with narrowing of IVD space.  Inhibition of TGFβ signaling attenuates EP degeneration in LSI mice. We then examined if EP degeneration could be prevented by inhibiting TGFβ signaling. TGFβ type I receptor (Tβ RI) inhibitor (SB, 1 mg/kg) was systemically injected in the LSI mice. Lumbar spine samples were collected at 2, 4 and 8 w post-surgery. We found IVD volume increased after 2 and 4 weeks treatment of inhibitor in LSI mice relative to LSI treated with vehicle (Fig. 5A). Cranial EP volume was preserved with inhibitor treatment as indicated by no statistically significant difference relative to sham controls (Fig. 5B). However, no effect was were observed in caudal EP volume, as EP volumes in LSI SB groups were similar to LSI sham controls (Fig. 5C). The ossification and degeneration of EPs were improved, although not completely abrogated by Tβ RI inhibitor as indicated by EP scores intermediate between sham and LSI treated with vehicle controls (Fig. 5D,E). Osteoclast numbers were decreased to sham controls levels in the LSI Tβ RI inhibitor treated mice, but this effect was not sustained at later time points (Fig. 5F,G). On another hand, the number of osteocalcin + cells was decreased in LSI mice by inhibitor treatment for 4 weeks (Fig. 5H,I). Osteogenesis is coupled with angiogenesis 16 . Therefore, we also evaluated if the blood vessel was affected by TGFβ inhibition. We found CD31 + staining was increased after LSI operation, but was preserved by inhibitor treatment, indicating the inhibitory effect of Tβ RI inhibitor in angiogenesis (Fig. 5J,K).

Inhibition of excess activation of TGFβ attenuates EP degeneration in CSI mice.
To further evaluate the effect of TGFβ inhibitor in EP degeneration, we employed second spinal instability model, CSI and injected with either vehicle or Tβ RI inhibitor (Fig. 6A). In this model, caudal 7 th -8 th (C 7-8 ) IVD was injured by annular stab and NP removal to induce caudal spine instability. IVD and EP volumes were preserved with similar volumes to sham controls when LSI mice were treated with Tβ RI inhibitor (Fig. 6B-E). We also evaluated the effect of TGFβ inhibitor on the blood vessel volume. After injury, more blood vessels could be seen in both cranial and caudal EPs of CSI mice whereas they were decreased after Tβ RI inhibitor treatment (Fig. 6F).

Discussion
EPs function as transitional tissue that absorbs hydrostatic pressure resulting from mechanical loading of the spine. In our previous study, EP hypertrophy was found in rat lumbar IVD when the axial force was increased upon the spine 9 . Cadaveric lumbar spines also show a similar trend of more severe IVD degeneration associated with greater thickness in both the cranial and caudal EPs 17 . In this study, we revealed that unstable mechanical loading in the spine induces EPs hypertrophy and is associated with IVD degeneration. We systematically analyzed the changes of cartilaginous EP, IVD space and vertebral bodies in the spine in two different spine instability animal models, considering them as a functional unit. The responses of EP resulted in narrowing of the IVD space and likely generated pathological static compression on the IVDs.
In pathological conditions, static compression stress activates excess TGFβ 18,19 . In the physiologic state, TGFβ maintains chondrocyte homeostasis to preserve EP structure and function [20][21][22] . In our current study, we found that elevated TGFβ was associated with accelerated endochondral ossification and vascularization in EP regions as indicated by hypertrophic chondrocytes activity and the presence of osteoclasts, osteoblasts, endothelial cells in LSI mouse model. These changes were observed without spinal instability surgery in the transgenic mice with overexpression of TGFβ and attenuated when TGFβ signaling was inhibited in the LSI mice. These changes are similar to those found in subchondral bone of OA patients and animal models 7 . TGFβ released from latent extracellular matrix mobilizes and recruits MSCs [23][24][25] . MSCs are thought to localize with vasculature 26 . We found that endothelial cells and osteoblast lineage cells appeared at a similar time in EPs to support osteogenesis. As TGFβ signaling exerts its primary effect on MSC migration, rather than osteoclast activity 25 , we did not, as expected, see effect on osteoclast numbers when TGFβ signaling was inhibited. The temporary early decrease in osteoclast numbers observed in the LSI + SB group may have been due inhibition of immunological reactions associated with osteoclast function 27 .
EP degeneration is thought to begin with abnormal calcification 28 , a process where calcium crystals salts are deposited into pores of EP. The calcified EP undergoes ossification and are eventually replaced with bone during aging 6,17 . The process is thought to reduce nutrients transport from vertebral marrow to IVD 6 . In our study, we found spinal instability accelerated ossification of EPs in the LSI mouse model and contributed to EP hypertrophy and sclerosis. The CSI mouse model was utilized as a second spinal instability model. The caudal spine in mice, unlike the lumbar spine, undergoes spontaneous ossification at a younger age with complete ossification at the 8 week age studied 29 . Similar to the LSI model, the CSI model also demonstrated EP hypertrophy likely from abnormal bone remodeling in the osseous EP area. Although bone remodeling is coupled with angiogenesis and reduces the distance between circulating nutrients and IVD 30,31 , the nutrient diffusion is actually impaired as a consequence of the replacement with cortical bone matrix 32 . The cortical bone matrix type I collagen in conjunction with loss of local type II collagen and change in spatial orientation contributes to EP sclerosis.
TGFβ has been found to increase proteoglycan expression and exert an anabolic effect to alter IVD development and degeneration 20,22,[33][34][35] . However our study suggests that supraphysiologic levels of TGFβ can also be detrimental to the IVD. Further supporting this idea are reports that high levels of TGFβ 1 are present in the IVDs from DDD patients [36][37][38][39] and a rabbit model of IVD benefit with TGFβ inhibition 33 . It is most likely that TGFβ has a functional versatility on the metabolism of IVD cells, where both too little and excessive signaling are detrimental.
We found that alterations in mechanical loading increase TGFβ early in the course of DDD. TGFβ is associated with ossification of the EP, leading to EP hypertrophy and likely static compression of the IVD by narrowing IVD space. Inhibition of TGFβ activation in the early phase of DDD attenuated EP and IVD degeneration.