Loss of tenomodulin expression is a risk factor for age‐related intervertebral disc degeneration

Abstract The intervertebral disc (IVD) degeneration is thought to be closely related to ingrowth of new blood vessels. However, the impact of anti‐angiogenic factors in the maintenance of IVD avascularity remains unknown. Tenomodulin (Tnmd) is a tendon/ligament‐specific marker and anti‐angiogenic factor with abundant expression in the IVD. It is still unclear whether Tnmd contributes to the maintenance of IVD homeostasis, acting to inhibit vascular ingrowth into this normally avascular tissue. Herein, we investigated whether IVD degeneration could be induced spontaneously by the absence of Tnmd. Our results showed that Tnmd was expressed in an age‐dependent manner primarily in the outer annulus fibrous (OAF) and it was downregulated at 6 months of age corresponding to the early IVD degeneration stage in mice. Tnmd knockout (Tnmd − / −) mice exhibited more rapid progression of age‐related IVD degeneration. These signs include smaller collagen fibril diameter, markedly lower compressive stiffness, reduced multiple IVD‐ and tendon/ligament‐related gene expression, induced angiogenesis, and macrophage infiltration in OAF, as well as more hypertrophic‐like chondrocytes in the nucleus pulposus. In addition, Tnmd and chondromodulin I (Chm1, the only homologous gene to Tnmd) double knockout (Tnmd − / − Chm1 − / −) mice displayed not only accelerated IVD degeneration, but also ectopic bone formation of IVD. Lastly, the absence of Tnmd in OAF‐derived cells promoted p65 and matrix metalloproteinases upregulation, and increased migratory capacity of human umbilical vein endothelial cells. In sum, our data provide clear evidences that Tnmd acts as an angiogenic inhibitor in the IVD homeostasis and protects against age‐related IVD degeneration. Targeting Tnmd may represent a novel therapeutic strategy for attenuating age‐related IVD degeneration.

Tnmd transcript is predominantly expressed in hypovascular tissues, such as tendons, ligaments, as well as eyes (Brandau et al., 2001;Shukunami et al., 2001). Interestingly, Minogue, Richardson, Zeef, Freemont, and Hoyland (2010) have demonstrated that Tnmd mRNA has abundant expression in the annulus fibrous (AF) and the nucleus pulposus (NP). Concomitantly, Nakamichi et al. (2016) showed that Mohawk (Mkx), an upstream gene of Tnmd (Ito et al., 2010), promotes the maintenance and regeneration of the outer annulus fibrous (OAF) of IVD suggesting that Tnmd may be involved in IVD homeostasis. To date, however, this hypothesis has not been investigated in detail.
In our previous studies, we compared Tnmd knockout (Tnmd −/− ) mice with their wild-type (WT) littermates and showed that the absence of Tnmd causes reduced tendon cell proliferation and density in vivo (Docheva et al., 2005), coupled with significantly lower self-renewal and augmented senescence of tendon-derived stem/ progenitor cells (TSPCs) in vitro (Alberton et al., 2015). Furthermore, we observed the pathological thickening and stiffening of collagen I (Col I) fibers in Tnmd −/− Achilles tendons resulting in running distance and time failure in Tnmd −/− mice challenged by running tests (Dex et al., 2017). Interestingly, the local absence of Tnmd in the cardiac chordae tendineae cordis (CTC) promoted angiogenesis and matrix metalloproteinases (MMPs) activation (Kimura et al., 2008), a phenomenon also observed when Tnmd −/− mice were subjected to surgically induced Achilles tendon rupture. In this model, we detected that genetic ablation of Tnmd leads to blood vessel accumulation accompanied by abnormal extracellular matrix (ECM) composition and macrophage profile during the early repair phase of injured tendons (Lin et al., 2017).
Cumulatively, the aforementioned data reveal that Tnmd plays an important regulatory role in the avascular tendogenic/ligamentogenic tissue homeostasis. Therefore, we hypothesized that Tnmd in the IVD may act to inhibit vascular ingrowth into this normally avascular tissue and maintain homeostasis. Here, we investigated the exact functional role of Tnmd in IVD in vivo and in vitro by phenotypization of Tnmd-deficient IVD tissues and IVD-derived cells.

| Tnmd is expressed in the IVD OAF and NP
Tnmd expression was observed by us, along with other researchers, in tendons, ligaments, eyes, and CTC (Brandau et al., 2001;Kimura et al., 2008;Shukunami et al., 2001). In the vertebral column, previous studies have localized Tnmd gene expression to areas of the IVD (Minogue et al., 2010;Nakamichi et al., 2016), as well as Tnmd immunostainings carried out in neonate mice detected robust protein expression in the OAF (Yoshimoto et al., 2017). To further determine the precise distribution of Tnmd in the postnatal and adult IVD, we first performed immunolocalization studies on IVD tissues from WT mice at distinct stages of skeletal development ranging F I G U R E 1 Tnmd is mainly expressed in OAF, and its loss leads to age-related IVD degeneration. (a and b) Immunofluorescence of Tnmd protein expression in developing adult and aged IVDs of WT mice shows that Tnmd is mainly expressed in OAF and to a lower extent in the NP, and as expected, Tnmd was not detected in Tnmd −/− IVDs. Fluorescence intensity analysis revealed an expression peak at 1 month and expression downregulation from 6 months onwards (n = 3-5 animals). (c) Western blot confirmed that Tnmd protein expression is higher at 1 month than at 6 months (n = 3 animals). (d and e) IHC of the human IVDs confirmed that TNMD protein is mainly found in the OAF and to a lesser extended in the NP (two-tailed nonparametric Mann-Whitney test; n = 5 samples). (f) Representative H&E stainings of human IVD (n = 5 samples). (g) H&E staining demonstrates greater degenerative changes in Tnmd −/− IVDs when compared to WT at 6 and 18 months. (h) Safranin O staining reveals small roundish chondrocyte-like cells in IAF and NP of Tnmd −/− IVDs. (i and j) Histological grading and disc height index calculation show in Tnmd −/− mice significantly widespread degeneration compared to WT at both examined stages (two-tailed nonparametric Mann-Whitney test; 6-month-old mice, n = 5 animals; 18-month-old mice, n = 3 animals). ** p < .01; *** p < .001. d, day; H&E, hematoxylin-eosin; IAF, inner annulus fibrous; IHC, immunohistochemistry; IVD, intervertebral disc; mo, month; NB, newborn; NP, nucleus pulposus; OAF, outer annulus fibrous; WT, wild-type. Scale bar, 200 μm from newborn to 18 months of age. We observed that Tnmd is predominantly produced and deposited in the ECM of the OAF, as well as to a lesser extent in the NP regions. Scarce positive signals in the inner annulus fibrous (IAF) and the cartilaginous endplate (EP) were also detected; however, those were primarily cellular and not in the ECM (Figure 1a). Notably, Tnmd signals in the OAF and the NP gradually peaked at 1 month of age, while it dropped at 6 months of age corresponding to the early IVD degeneration stage in mice, and then further decreased at 12 and 18 months ( Figure 1b). As expected, Tnmd was not detected in Tnmd −/− IVDs (Figure 1a). Western blotting of IVD tissue protein extracts confirmed that the protein levels of Tnmd are higher at 1 month, when compared to 6 months ( Figure 1c) as well as that in the IVD; similar to Achilles tendon and CTC, Tnmd is fully processed to its 16 kDa C-terminal portion. We also assessed the expression of TNMD in human lumbar discs (Table   S1). Consistent with the observation in mice, TNMD protein was predominantly found in the ECM of the OAF and to a lesser extent in the NP regions (Figure 1d-f).

| The absence of Tnmd leads to age-related IVD degeneration
To analyze the potential involvement of Tnmd during naturally occurring IVD degeneration in mouse, 6-and 18-month-old lumbar IVDs were first examined by hematoxylin-eosin (H&E) and safranin O staining for pathological changes. Importantly, Tnmd −/− mice showed higher levels of degenerative changes compared to WT mice. Lamellae appeared thinner, looser, and fibrillated in Tnmd −/− AF ( Figure 1g). In the IAF and NP areas of Tnmd −/− IVDs, small round cells, morphologically resembling chondrocytes, were visible.
Cumulatively, such abnormalities are often described as degenerative changes (Nakamichi et al., 2016), and they were even more evident in the Tnmd −/− discs at 18 months ( Figure 1h). By implementing the histological scoring system for mouse IVD (Tam et al., 2018), we detected higher degenerative scores at 6 and 18 months in Tnmd −/− mice, reflective of a significantly widespread IVD degeneration compared to WT (Figure 1i). In addition, calculation of disc height index (Masuda et al., 2005) showed lower values for Tnmd −/− IVDs at both examined stages ( Figure 1j). Thus, our results demonstrate that Tnmd is associated with IVD homeostasis and its loss leads to profound tissue degeneration that advances during the aging process.

| Tnmd deficiency results in abnormal diameters and biomechanical properties of IVD collagen fibrils, accompanied by reduced expression of multiple IVD-and tendon/ ligament-related genes in the OAF
Since Tnmd is highly expressed in the OAF and a disorganized OAF morphology characterized by thinner collagen fibers ( Figure S1a,b) was observed in Tnmd −/− mice, we next examined the nanotopographical and biomechanical properties of 6-month-old Tnmd −/− and WT IVDs. We applied indentation-type atomic force microscopy (IT-AFM) to quantitatively assess collagen fibril diameters and compressive stiffness of AF. Height images revealed that the collagen fibrils of Tnmd −/− OAF were more frayed and interrupted by gaps, and vertical deflection images indicated that the collagenous network of Tnmd −/− OAF was less dense compared to WT (Figure 2a). The fibril diameters in OAF and IAF were significantly smaller in Tnmd −/− than in WT mice ( Figure 2b). Indentation measurements on native IVD tissues indicated a bimodal stiffness distribution in both genotypes ( Figure 2c). In Tnmd −/− OAF and IAF, the proteoglycan stiffness peaks were detected at 1.240 ± 0.098 and 0.272 ± 0.004 MPa and those of collagen network at 2.972 ± 0.033 and 0.532 ± 0.049 MPa, respectively. In WT OAF and IAF, the average proteoglycan stiffness peaks were 8.134 ± 0.307 and 0.285 ± 0.013 MPa, while the collagen stiffness peaks were 14.019 ± 0.493 and 0.581 ± 0.0063 MPa, respectively. In sum, the overall compressive stiffness of OAF was markedly lower in Tnmd −/− than WT mice, but the IAF biomechanical properties were not significantly different between genotypes.
Taken together, these findings demonstrate that Tnmd is a critical factor required to maintain the structural and biomechanical properties of the OAF collagen fibrils likely through the modulation of ECM gene expression.

| Manifestation of hypertrophic chondrocytes and Col X-rich matrix in the NP of Tnmd −/− mice
It has been shown that ectopic calcifications in IVDs are a known characteristic of IVD degeneration (Hristova et al., 2011;Illien-Jünger et al., 2016). The observed small round cells, resembling chondrocyte morphology, in the Tnmd −/− NP prompted us to test whether the loss of Tnmd is accelerating hypertrophic chondrocyte-like occurrence. The major proteoglycan of the NP is aggrecan (Acan), which due to its highly anionic glycosaminoglycan content provides osmotic properties, enabling the NP to maintain height and turgor against compressive loads (Bedore et al., 2013).

F I G U R E 2
Tnmd deficiency causes altered ECM nanostructure and mechanical properties of the OAF in 6-month-old mice. (a) AFM height images (upper panels for both genotypes) show that the collagen fibrils in Tnmd −/− OAF were more frayed and interrupted by gaps, and vertical deflection images (lower panels) demonstrate that the collagen network in this region was less dense in Tnmd −/− compared to WT. (b) Comparison of the collagen fibril diameters reveals significantly smaller average size in Tnmd −/− than in WT AF (two-tailed unpaired Student's t test; n = 3 animals, and 200 fibrils were analyzed per genotype). (c) Plots of compressive stiffness data obtained by indentation AFM demonstrated that ECM compressive stiffness was markedly lower in the OAF regions of Tnmd −/− IVDs but not noticeably different in the IAF regions (two-tailed unpaired Student's t test; n = 3 animals). (d) Significant downregulation of numerous IVD-and tendon/ligamentrelated genes was detected in Tnmd −/− OAF by qRT-PCR analysis. For calculation of fold changes, WT was set to 1 (two-tailed unpaired Student's t test; n = 3 animals). * p < .05; ** p < .01; *** p < .001. AF, annulus fibrous; Col, collagen; ECM, extracellular matrix; IAF, inner annulus fibrous; IVD, intervertebral disc; OAF, outer annulus fibrous; PG, proteoglycan; qRT-PCR, quantitative real-time PCR; WT, wild-type. Scale bar, 1 μm Therefore, we explored Acan expression by immunohistochemistry (IHC) in IVD sections from both genotypes. At 6 months, no noticeable change in Acan localization was found in Tnmd −/− NP compared to WT NP. However, the deposition of Acan was significantly decreased in Tnmd-deficient NP at 18 months ( Figure 4a,b).
More specifically, it has been proposed that Acan production ratio within the NP to hyaline cartilage is approximately 27:2 (Mwale, Roughley, & Antoniou, 2004). Hence, Acan downregulation can lead to appearance of hypertrophic-like chondrocytes, which subsequently contribute to calcification, due to the available free calcium ions (Hristova et al., 2011). In order to track hypertrophic chondrocyte-like differentiation in the NP, we next implemented immunofluorescence analysis for Sox9, the key pro-chondrogenic transcription factor (Takimoto, Oro, Hiraki, & Shukunami, 2012), indicating clearly substantial IVD cell dysfunction.

| Tnmd and Chm1 double mutant mice display not only accelerated IVD degeneration but also ectopic bone formation
Our findings suggest that Tnmd contributes to protect the IVD from vascularization and inflammation. However, our present data do not exclude the possibility of the existence of other anti-angiogenic factors. Chm1, the only Tnmd homologous protein (Brandau et al., 2001;Shukunami et al., 2001), is a cartilage-specific angiogenesis inhibitor (Hiraki et al., 1997;Yoshioka et al., 2006) that has been previously shown to be also highly expressed in the IVD during the gestational period and gradually downregulated after maturation (Takao, Iwaki, Kondo, & Hiraki, 2000). Immunofluorescence Based on our novel data, we concluded that simultaneous loss of Tnmd and Chm1 causes a more progressive IVD degenerative phenotype than Tnmd single knockout. It remains to be investigated to what extent these mutant variants compare to Chm1 single knockout.

| Loss of Tnmd in OAF cells suppresses their proliferation and promotes cell apoptosis
To further understand the mechanisms underlying the roles of Tnmd in IVD homeostasis, we performed in vitro studies with OAF cells isolated from 12-month-old Tnmd −/− or WT lumbar IVDs ( Figure   S2a). First, we characterized the obtained cell populations by immunofluorescence staining and RT-PCR. Fmod, an accepted AFspecific marker in rodent (Leung, Tam, Chan, Chan, & Cheung, 2011;Smits & Lefebvre, 2003), was expressed in both genotypes, but its immunostaining signal was weaker in Tnmd −/− OAF cells compared with WTs ( Figure S2b,g).
WT OAF cells were strongly expressing Tnmd, while as expected Tnmd was not produced by Tnmd −/− cells ( Figure S2c). To analyze whether the synthesis of ECM proteins by the OAF cells is altered due to Tnmd deficiency, we next performed immunofluorescence analysis for Col I, which is the main protein component of OAF; Fn, which plays a pivotal role in facilitating AF cell attachment and fiber alignment (Attia, Santerre, & Kandel, 2011); and Lum, which interacts with collagen fibrils and contributes to AF mechanical properties (Sztrolovics, Alini, Mort, & Roughley, 1999). In vitro, only a slight decrease in the expression of the three proteins was observed in Tnmd −/− OAF cells ( Figure S2d-g). Semiquantitative RT-PCR revealed that Tnmd mRNA could not be detected, while Chm1 mRNA levels were reduced in Tnmd −/− OAF cells ( Figure S2h).
To further evaluate the role of Tnmd in OAF cell behavior, we carried out time-lapse imaging of random migration, followed by plotting of forward migration index (FMI). Our results showed that Tnmd −/− OAF cells were less migratory than WT cells ( Figure S3a,b).
Quantification of velocity, accumulated and Euclidean distance clearly indicated that the observed effect was significant ( Figure   S3c,d). Furthermore, during 0, 3, 5, and 7 days of culture, DNAbased CyQUANT assays showed that the proliferation of Tnmd −/− OAF cells was significantly reduced compared to that of WT cells ( Figure S3e). Lastly, TUNEL assays demonstrated that Tnmd −/− OAF F I G U R E 4 Accumulation of hypertrophic chondrocyte-like alterations in the NP of Tnmd −/− mice. (a and b) Significant downregulation of Acan protein expression was detected at 18 months by immunofluorescent imaging and fluorescence intensity analysis in Tnmd −/− NP compared to WT. (c-h) Immunofluorescence stainings with anti-Sox9, anti-Runx2 and anti-Col X antibodies reveal increased expression levels of the three hypertrophy markers in Tnmd −/− NPs compared with WT at 18 months. All quantitative histomorphometry data were assessed by two-tailed nonparametric Mann-Whitney test; 6-month-old mice, n = 5 animals; 18-month-old mice, n = 3 animals. * p < .05; ** p < .01. AF, annulus fibrous; EP, endplate; mo, month; NP, nucleus pulposus; WT, wild-type. Scale bar, 200 μm cell population contained increased number of apoptotic cells compared to WT cell population ( Figure S3f,g). Taken together, we validated the OAF phenotype of our isolated cells and concluded that the loss of Tnmd causes reduction in their proliferative and migratory potential but an increase in apoptotic risk.

| D ISCUSS I ON
Low back pain is an enormous medical and socioeconomic burden in modern society. Although there are many etiologies for the development of low back pain, a main component appears to be IVD degeneration, including neovascularization and inflammatory element

Based on our initial investigation at the protein level, we could
show that Tnmd is strongly expressed in the ECM of the OAF, while Chm1 is predominantly found in the NP. We could detect to a lower extent a signal for Tnmd also in the NP; however, this should be taken with caution considering the highly conserved C-terminus between both proteins against which the primary antibody was raised.
Moreover, we could show, similar to the Achilles tendon and CTC (Dex et al., 2016), that Tnmd processing in the IVD undergoes a proteolytic cleavage at its 16kDa C-terminal portion. Thus, we report for the first time the expression pattern of Tnmd protein in the ECM of the IVD, but also revealing that Tnmd is downregulated at 6 months of age.  In general, IVD degeneration starts with proteoglycan breakdown leading to diminished water retention and disc dehydration. Our previous nanostructural and biomechanical analyses of the collagen matrix in Achilles tendons demonstrated that the absence of Tnmd causes thickening and stiffening of the collagen fibrils (Docheva et al., 2005). Therefore, we analyzed whether the lack of Tnmd affects the collagen fibrils of the AF. In contrast to our previous results, we found out that Tnmd −/− OAF contain collagen fibrils with smaller calibers and lower compressive stiffness, which cannot provide enough strength and proper load distribution over large parts of the OAF. According to these results and our previous finding that Tnmd is a mechanosensitive and mechanoregulated gene (Dex et al., 2017), we propose that the above-described phenotypic differences between tendon and IVD tissues are the consequence of the different biomechanical environments, namely IVDs are subjected mainly to compressive stress while tendons to tensile forces.
Altered OAF matrix can hinder the ability of the tissue to withstand mechanical load and in turn can increase the likelihood of tear formation (Gruber & Hanley, 1998;Guterl et al., 2013;Iatridis, Nicoll, Michalek, Walter, & Gupta, 2013;Roberts et al., 2008;Shukunami et al., 2018). Interestingly, the absence of Tnmd in tendon versus OAF tissues caused the opposite effect on the mRNA expression levels of several ECM genes such as Col14a1, Col15a1, and Prg4, as well as the transcription factors Scx, a direct transactivator of Tnmd (Shukunami et al., 2018), and Mkx, both known to regulate collagens. Furthermore, we detected elevated expression of MMP-3 and MMP-9 in Tnmd −/− OAF cells; thus, soluble matrix proteins can be extruded from the tissue by mechanical loading and enhance the degenerative process (Kamper et al., 2016).
Notably, the glycosaminoglycan to hydroxyproline ratio within the NP of young adults is approximately 27:1, whereas the ratio within the hyaline cartilage of the same individuals is about 2:1 (Mwale et al., 2004). In our study, decreased Acan content in the Tnmd −/− NP, paralleled by increased Sox9, Runx2, and Col X levels, is indicative for a transition from a hydrated gel-like NP to a more hypertrophic chondrocyte-like matrix that is involved in IVD degeneration.
Altogether, in such circumstances, sprouting of blood vessels in the abnormal OAF matrix and further toward the center of the IVD can facilitate ectopic ossification via intermediate hypertrophy state (Roberts et al., 2008).
Apoptosis is known to be a component of IVD degeneration (Gruber & Hanley, 1998). Previously, we have reported a reduced proliferative rate in Tnmd −/− tendon tissues and derived cells, as well as an increased number of p53-positive cells in Tnmd −/− Achilles tendons (Alberton et al., 2015;Docheva et al., 2005). Therefore, another important point that we examined was the effect of Tnmd on cell proliferation and apoptosis in the IVD.
Interestingly and similar to tendon tissues, our in vivo and in vitro investigations convincingly prove that Tnmd is a positive regulator of cell proliferation and its loss accelerates apoptosis, which in turn leads to propagation of IVD aging and degenerative process.
However, our study is impeded in providing an explanation of the exact molecular mode of action of Tnmd, due to lack of known binding partners. Therefore, further studies are promptly required to determine the signaling pathways in which Tnmd participates during IVD homeostasis.
Taken together, our findings provide new insights into the protective role of Tnmd in IVD degeneration. Understanding the precise Tnmd-dependent mechanisms can form the basis of developing new therapeutic strategies for prevention or treatment of IVD degeneration.

| Human samples
Samples comprising 5 IVDs (Table S1) were collected from 5 patients undergoing vertebral reconstruction (IVD tissues are removed and discarded) due to burst fractures or lumbar tumors in the Orthopaedic Center of People's Liberation Army, the Affiliated Southeast Hospital of Xiamen University. There were 3 males and 2 females (21-32 years of age). All patients received magnetic resonance imaging scans to confirm intact and healthy status of the IVDs. Samples were fixed immediately after removal in 4% paraformaldehyde (PFA; Merck) overnight at 4°C and then embedded in paraffin. From each patient, an informed consent was obtained.
Sample collection and experimental methods were authorized by the Ethics Committee of the Xiamen University.
Both cell types were cultured at 37°C and 5% CO 2 , kept up to 70% confluency and supplemented with fresh culture media every third day.

| Histology, immunohistology, and histomorphometry
Mouse spines were obtained after euthanasia (newborn, 15 days, 1 month, 12 months, and 18 months of age: n = 3 animals; 6 months of age: n = 5 animals). The samples were fixed in 4% PFA overnight at 4°C. Following fixation, specimens were decalcified in 10% ethylene diamine tetraacetic acid (EDTA)/phosphate-buffered saline (PBS) pH 8.0 (Sigma-Aldrich, Munich, Germany) for 14 days, and embedded into paraffin or cryo-media, and then the mouse and human samples were sectioned with 6 μm (paraffin) or 10 μm (cryo), and stained with H&E and safranin O using standard protocols. Quantification of histological score was based on a new scoring system specialized on histomorphology of mouse IVD (Tam et al., 2018). In brief, the scoring system included the following evaluation: NP structure (0 point, single-cell mass; 1 point, cell clusters ˂50%; 2 points, cell clusters ˃50%; 3 points, matrix-rich with little cells NP; 4 points, mineralized NP), NP clefts/fissures (0 point, none; +1 point, mild; +2 points, severe), AF/NP boundary (0 point, clear cut boundary; 1 point, round chondrocyte cells at the boundary; 2 points, loss of boundary), AF structure (0 point, concentric lamellar structure; 1 point, serpentine, widened or rounded AF lamellae; 2 points, reversal of lamellae; 3 points, undefinable lamellar structure or penetrating the NP; 4 points, mineralized or lost AF), and AF clefts/fissures (0 point, none; +1 point, mild; +2 points, severe). Quantification of disc height index was performed on H&E images (Masuda et al., 2005). In general, all histology, immunohistology, and histomorphometry experiments, unless specified otherwise in the text, were reproduced in 3 sections per sample with 5 or 3 samples per group for investigation. mounted with fluoroshield (Sigma-Aldrich). To analyze apoptotic cells numbers, TUNEL kit was applied following the manufacturer's instructions (Abcam, ab66110). Photomicrographs were taken on the Observer Z1 microscope equipped with the Axiocam MRm camera (Carl Zeiss). Quantitative histomorphometry was carried out via an automated quantitative image analysis according to algorithms from literature (Hsieh et al., 2016;Lin et al., 2017). In brief, using ImageJ (National Institutes of Health), the following algorithm was applied: (a) area of interest was manually designated using the "drawing/selection" tool; (b) "set measurements" for area, integrated density and mean gray value was selected from the analyze menu; and (c) lastly, the corrected total cryosections fluorescence (CTCF) representing the Acan, Col I, Col X, Fmod, Fn, Lum, MMP3, MMP-9, p65, and Tnmd expression were exported and calculated in Excel

| IT-AFM
Indentation-type atomic force microscopy was performed on 14 μmthick frozen tissue sections from 6-month-old Tnmd −/− and WT mice (3 animals per group; each animal represented by 3 tissue sections).
Measurements were taken with NanoWizard AFM instrument (JPK Instruments) mounted on an inverted optical microscope (Axiovert 200, Zeiss) as described in detail previously (Dex et al., 2017;Gronau et al., 2017;Kamper et al., 2016). Briefly, 625 indentation curves per sample were recorded in 2 × 2 μm 2 area using pyramidal tips with nominal radius of 20 nm and silicon nitride cantilevers (MLCT microcantilever, Bruker). The spring constant of each cantilever was determined prior to the experiment using the thermal noise method, and the sample stiffness was determined via a modified Hertz model using the JPK Data Processing software (V4.2.20, JPK Instruments), as described in detail previously (Gronau et al., 2017;Kamper et al., 2016). Based on these results, histograms were plotted and the two maxima were identified by fitting a linear combination of two Gaussian functions using OriginLab software (version 6). As previously reported, the lower stiffness peak attributes to the proteoglycan moiety, while the higher stiffness peak relates to collagen fibrils (Gronau et al., 2017;Kamper et al., 2016).

| Western blot analysis
Mouse IVDs from WT (1 and 6 months of age) and Tnmd −/− (6 months of age) animals were snap-frozen in liquid nitrogen. Using a mortar and pestle, the tissue was pulverized and resuspended in 8 M urea, 50 mM Tris-HCl (pH 8.0), 1 mM dithiothreitol, and 1 mM EDTA. Protein (22 μg) aliquots were loaded on a 12% SDS-polyacrylamide gel and transferred to Immobilon-p transfer membrane (Merck Millipore).
The membrane was blocked for 90 min with 5% Milk/TBS-T buffer.

| Semiquantitative and qRT-PCR
Total RNAs from OAF tissues (n = 3 animals) and OAF cells were isolated with Qiagen RNeasy Mini kit (Qiagen) and used for standard semiquantitative and qRT-PCR. For cDNA synthesis, 1 μg total RNA and AMV First-Strand cDNA Synthesis Kit (Invitrogen) were implemented. Semiquantitative PCR was performed with Taq DNA Polymerase (Invitrogen) in MGResearch instrument (BioRad, Munich, Germany). Primer sequences and PCR conditions were as follows: Tnmd forward 5′gaaaccatggcaaagaatcctccagag3′, reverse 5′ttagactctcccaagcatgcgggc3′; Chm1 forward 5′atggtagggcctgag-gacgttg3′, reverse 5′gctgcatggcatgacgactctg3′; Gapdh forward 5′ga-gaggccctatcccaactc3′, reverse 5′gtgggtgcagcgaactttat3′; PCR was performed with incubation at 94°C for 5 min followed by 30 cycles of a three-step temperature program of 1 min at 94°C, 20 s at 57°C, and 30 s at 72°C. The PCR reaction was terminated after a 7 min extension at 70°C. The band intensity of the amplified products in the gel was visualized, photographed, and analyzed using a gel imager (Vilber Lourmat). The relative gene expression was quantified by densitometry and normalized to the amount of Gapdh with ImageJ and presented as fold change to WT. Quantitative PCR of tendon-associated genes was performed using RealTime Ready Custom Panel 96 -32+ plates (https ://confi gurat or.realt imere ady. roche.com) according to the manufacturer's instructions (Roche Applied Science). Briefly, PCR reactions were pipetted on ice and each well contained 10 µl LightCycler 480 probes master mix, 0.2 µl cDNA (diluted 1:5) and 9.8 µl PCR grade water. Plates were subsequently sealed and centrifuged down for 15 s. at 2,100 rpm. The relative gene expression was calculated as a ratio to Gapdh. All PCR results have been reproduced in three independent experiments.

| Migration analysis
Migration analysis was performed similarly to our previous study (Popov, Kohler, & Docheva, 2016). For random migration,

| CyQUANT assays
A total of 1.5 × 10 3 cells (passage 1) per well were plated in 6-well plates, and CyQUANT assay detection was performed according to the manufacturer's instructions (Invitrogen) after 0, 3, 5, and 7 days cell culture. CyQUANT assay was repeated independently in 3 experiments per time point.
Before seeding HUVECs into the upper chamber, the membrane was coated with Col I (10 μg/Ml; Millipore), kept in a humidified incubator overnight, and filled with ECGM-2 an hour before introducing the 5 × 10 3 HUVECs per well. After 36 hr co-culturing, HUVECs that have migrated through the pores and adhered to the lower side of the membrane were fixed with 4% PFA and stained with DAPI.
The cell nuclei were counted using the Observer Z1 microscope equipped with the Axiocam MRm camera. Tnmd −/− or WT OAF cells in the lower chamber were analyzed for p65, MMP-3, and MMP-9 by immunofluorescence staining. The interactions of the co-cultures assays were repeated independently in 3 experiments. All cell nuclei (DAPI) in 3 images per well from 5 wells per plate were counted.

| Statistical analysis
In this study, each animal was represented with 3 different tissue sections with comparable planes between genotypes. The results were averaged per animal were presented as mean ± SD between the 3-5 animals per group. Exact animal number and experimental reproducibility is given for each result in the figure legends. Statistical differences between two groups were determined using two-tailed unpaired Student's t test, or two-tailed nonparametric Mann-Whitney test. In multiple comparisons, one-way ANOVA was followed by Bonferroni post hoc correction. Differences were considered statistically significant at the values of * p < .05, ** p < .01, and *** p < .001. Li for experimental help, Dr. D. Luo for his guidance with statistical analyses, and Dr. Girish Pattappa for English proof-reading.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

E TH I C S A PPROVA L
Mouse husbandry, handling, and euthanasia were strictly carried out according to the guidelines of the Bavarian authorities. Animals were euthanized with CO 2 and dissected for collection of whole spine and tail tendon tissues. Human sample collection and experimental methods were authorized by the Ethics Committee of the Xiamen University.