FOXO are required for intervertebral disk homeostasis during aging and their deficiency promotes disk degeneration

Abstract Intervertebral disk (IVD) degeneration is a prevalent age‐associated musculoskeletal disorder and a major cause of chronic low back pain. Aging is the main risk factor for the disease, but the molecular mechanisms regulating IVD homeostasis during aging are unknown. The aim of this study was to investigate the function of FOXO, a family of transcription factors linked to aging and longevity, in IVD aging and age‐related degeneration. Conditional deletion of all FOXO isoforms (FOXO1, 3, and 4) in IVD using the Col2a1Cre and AcanCreER mouse resulted in spontaneous development of IVD degeneration that was driven by severe cell loss in the nucleus pulposus (NP) and cartilaginous endplates (EP). Conditional deletion of individual FOXO in mature mice showed that FOXO1 and FOXO3 are the dominant isoforms and have redundant functions in promoting IVD homeostasis. Gene expression analyses indicated impaired autophagy and reduced antioxidant defenses in the NP of FOXO‐deficient IVD. In primary human NP cells, FOXO directly regulated autophagy and adaptation to hypoxia and promoted resistance to oxidative and inflammatory stress. Our findings demonstrate that FOXO are critical regulators of IVD homeostasis during aging and suggest that maintaining or restoring FOXO expression can be a therapeutic strategy to promote healthy IVD aging and delay the onset of IVD degeneration.

We recently reported that expression of FOXO1 and FOXO3 in IVD is reduced during aging and degeneration in humans and mice (Alvarez-Garcia, Matsuzaki, Olmer, Masuda, & Lotz, 2017). However, the function of FOXO in IVD is unknown. In this study, we used in vivo and in vitro approaches to investigate the role of this important family of transcription factors in IVD biology and pathogenesis of IDD.

| FOXO deficiency impairs postnatal IVD maturation and homeostasis
To model and study the consequences of the reduced expression of FOXO in NP, AF, and EP of degenerated IVD (Alvarez-Garcia, Matsuzaki, Olmer, Masuda, et al., 2017), we generated mice with deletion of FOXO1, 3, and 4 in IVD tissues under the Col2a1 promoter Col2a1Cre-FOXO KO mice were viable at birth and had similar body size as Col2a1Cre −/− littermates with no overt skeletal abnormalities. IVD from Col2a1Cre-FOXO KO mice were indistinguishable from those in control mice at postnatal day 1 (P1) and 7 (P7) (Supporting information Figure S1b in Appendix S1). Starting at 1 month of age, lumbar IVD from Col2a1Cre-FOXO KO mice exhibited a mild enlargement of the NP and a modest increase in disk height . The increased disk height of mutant mice became more marked at 2, 4, and 6 months of age and was concomitant with significantly higher cellularity in the NP (Figure 1c-d). At 4 and 6 months of age, Col2a1Cre-FOXO KO mice showed histological features of degeneration that included disruption of the NP/AF interface, disorganized AF lamellae with abundant hypertrophic cells in the inner AF, and cell loss and calcification of the EP (Figure 1b d). In addition to the cell loss in the EP, there was a significant reduction in cellularity in the NP of Col2a1Cre-FOXO KO mice at 6 months when compared to 4-month-old mice (Figure 1d). FOXO deficiency led to severe spine deformities with abnormal curvature of the spine and kyphosis in 6-month-old mice (Figure 1e). In addition, deletion of all FOXO isoforms resulted in abnormal cell organization in vertebral growth plate, increased vertebral diameter, increased trabeculae number, and trabecular thickness in subchondral bone at 4 and 6 months of age (Supporting information Figure S2 in Appendix S1).
To examine the function of each individual FOXO isoform in IVD growth and maintenance, we generated Col2a1Cre-FOXO1 KO, Col2a1Cre-FOXO3 KO, and Col2a1Cre-FOXO4 KO mice.
Col2a1Cre-FOXO1 KO mice showed an increase in NP cellularity and cell hypertrophy in the inner AF (Supporting information Figure S3 in Appendix S1), suggesting that FOXO1 is the predominant isoform in the regulation of postnatal cell proliferation and differentiation. Reduced cellularity and mild degenerative changes were also observed in the EP of mice with FOXO1 deletion (Supporting information Figure S3 in Appendix S1). However, these changes were less severe than those in Col2a1Cre-FOXO KO mice, indicating compensation by FOXO3 or FOXO4. On the other hand, Col2a1Cre-FOXO3 KO and Col2a1Cre-FOXO4 KO mice did not show any structural abnormalities in IVD up to 6 months of age. However, degenerative changes were observed in lumbar IVD from Col2a1Cre-FOXO3 KO mice, but not Col2a1Cre-FOXO4 KO, at 12 months (Supporting information Figure S4a in Appendix S1). These changes more severe at 18 months recapitulated histological hallmarks of age-related disk degeneration such as decreased NP cellularity, decreased disk height, abnormal AF organization, and ossification of the EP and resulted in increased histopathological scores (Supporting information Figure S4b-d in Appendix S1). These results suggest that FOXO3 is dispensable for normal IVD postnatal development, likely due to compensation by FOXO1, but is essential for IVD homeostasis, and its deletion results in accelerated IDD during aging.

| FOXO control IVD cell proliferation and maturation
To test whether the increased cellularity in the NP of mice lacking FOXO was due to changes in cell proliferation, 3-week-old Col2a1Cre-FOXO KO and Col2a1Cre −/− control mice were injected with 5-bromo-2'-deoxyuridine (BrdU) intraperitoneally for five con- FOXO deletion in the IVD also resulted in histological changes suggestive of abnormal cell differentiation. In mice, the NP compartment is composed of a large cluster of cells in the center of the NP that is surrounded by a layer of proteoglycan-rich matrix (Tam et al., 2017). During postnatal growth and maturation, cells within the NP undergo a morphologic shift from a round and vacuolated to a more flattened cell when skeletal maturity is reached at 4 months of age ( Figure 2b). However, this shift did not occur in Col2a1Cre-FOXO KO mice where large round cells could still be seen at 4 months of age ( Figure 2b). In addition, whereas the cell area/NP area ratio was progressively reduced in control mice, no significant changes were  Figure 2g). Moreover, mRNA expression of collagen type II (Col2a1) and X (Col10a1) was increased in the AF of Col2a1Cre-FOXO KO mice when compared with controls at 4 months of age ( Figure 2h). Collectively, these data indicate that FOXO deletion in IVD tissues severely impairs normal postnatal IVD growth and maturation.

| FOXO deletion in skeletally mature mice leads to spontaneous IVD degeneration
As mice with Col2a1Cre-mediated FOXO deletion already developed spontaneous spine abnormalities before 4 months of age that were at least in part due to FOXO functions in postnatal growth and maturation, we used the AcanCreER model to analyze the role of FOXO F I G U R E 1 Impaired intervertebral disk maturation and spontaneous degeneration in mice with conditional deletion of FOXO. (a) Safranin O staining in lumbar intervertebral disk (IVD) samples isolated from Col2a1Cre −/− and Col2a1Cre-FOXO KO mice at 1, 2, 4, and 6 months of age (n = 6-8 mice per group). NP: nucleus pulposus; AF: annulus fibrosus; EP: endplate. Magnification bar = 100 µm. (b) Histopathological scores in the nucleus NP/AF and EP of lumbar IVD samples isolated from Col2a1Cre −/− and Col2a1Cre-FOXO KO mice at 1, 2, 4, and 6 months of age (n = 6-8 mice per group). (c) Measurement of disk height of lumbar IVD samples from Col2a1Cre −/− and Col2a1Cre-FOXO KO mice at 1, 2, 4, and 6 months of age (n = 6-8 mice per group). (d) Quantification of cellularity in the NP, AF, and EP of lumbar IVD samples isolated from Col2a1Cre −/− and Col2a1Cre-FOXO KO at 1, 2, 4, and 6 months of age (n = 6-8 mice per group). (e) Photograph (left) and µ-CT image (right) of 6-month-old Col2a1Cre −/− and Col2a1Cre-FOXO TKO mice. Values shown are mean ± SD. Statistical comparisons were assessed by oneway analysis of variance (ANOVA) followed by a post hoc Tukey's test. *p < 0.05 in the maintenance of mature IVD. Four-month-old skeletally mature Histomorphometric quantification of the ratio of the NP cell area to total NP area in IVD from Col2a1Cre −/− and Col2a1Cre-FOXO KO mice at 1, 2, 4, and 6 months of age (n = 6-8 mice per group). (d) Gene expression analysis of anabolic and catabolic genes in the NP of lumbar IVD samples from Col2a1Cre −/− and Col2a1Cre-FOXO KO at 4 months of age (n = 4 mice per group). (e) Safranin O staining in annulus fibrosus (AF) of lumbar IVD isolated from Col2a1Cre −/− and Col2a1Cre-FOXO KO mice at 4 and 6 months of age. Magnification bar = 100 µm. (f) Picrosirius red staining in lumbar IVD isolated from Col2a1Cre −/− and Col2a1Cre-FOXO KO mice at 4 and 6 months of age (n = 6 mice per group). Images were obtained under polarized light to show collagen birefringence. Magnification bar = 100 µm. (g) Collagen type X (left panel) and RUNX2 (right panel) immunohistochemistry in lumbar IVD from Col2a1Cre −/− and Col2a1Cre-FOXO KO mice at 4 and 6 months of age (n = 5 mice per group). Magnification bar = 100 µm. (h) Gene expression analysis in the AF of lumbar IVD samples from Col2a1Cre −/− and Col2a1Cre-FOXO KO at 4 months of age (n = 4 mice per group). Values shown are mean ± SD. Statistical comparisons were assessed by an unpaired, two-tailed t-test after testing for equal variance using an F-test. Values are mean ± SD. *p < 0.05 (Supporting information Figure S6 in Appendix S1). Histopathological scores of the AcanCreER-FOXO KO mice were significantly increased at 12 months for all IVD structures ( Figure 3d). These findings suggest that FOXO are required for the maintenance of IVD integrity and that loss of FOXO in IVD in mature mice results in early onset IDD. Moreover, the number of LC3 puncta, a marker of autophagosomes and autophagy activation (Mizushima & Yoshimori, 2007), was significantly decreased in the NP of Col2a1Cre-FOXO KO mice at 4 months of age ( Figure 4d).

| FOXO regulate expression of antioxidant and autophagic genes in NP
NP cells reside in a unique microenvironment defined by low oxygen and low nutrients (Risbud, Schipani, & Shapiro, 2010). Adaptive mechanisms to this environment include activation of HIF1A (hypoxia-inducible factor 1α subunit) signaling and increase in autophagic activity (Choi et al., 2016;Risbud et al., 2010). To determine whether FOXO expression and activity are modulated by hypoxia, human NP cells were cultured in 20% or 1% oxygen for 24 hr. Gene expression analysis showed increased mRNA levels of ACAN as well as of HIF1A targets VEGF, SLC2A1, and ENO1 under hypoxic F I G U R E 3 Spontaneous intervertebral disk degeneration in skeletally mature mice with conditional deletion of FOXO. (a) Safranin O staining in lumbar intervertebral disk (IVD) from AcanCreER −/− and AcanCreER-FOXO KO mice at 6 and 12 months of age (n = 6-8 mice per group). NP: nucleus pulposus; AF: annulus fibrosus; EP: endplate. Magnification bar = 100 µm. (b) Quantification of cellularity in NP, AF, and EP of lumbar IVD from AcanCreER −/− and AcanCreER-FOXO KO mice at 6 and 12 months of age (n = 6-8 mice per group). (c) Measurement of disk height of lumbar IVD samples AcanCreER −/− and AcanCreER-FOXO KO mice at 6 and 12 months of age (n = 6-8 mice per group). (d) Histopathological scores in the NP/AF and EP of lumbar IVD from AcanCreER −/− and AcanCreER-FOXO KO mice at 6 and 12 months of age (n = 6-8 mice per group). Values shown are mean ± SD. Statistical comparisons were assessed by an unpaired, two-tailed t-test after testing for equal variance using an F-test. Values are mean ± SD. *p < 0.05 ALVAREZ-GARCIA ET AL.
| 5 of 13 conditions ( Figure 5a). Notably, FOXO3 but not FOXO1 expression was upregulated by hypoxia ( Figure 5a). This upregulation was not dependent of HIF1A as NP cells transfected with specific siRNA for HIF1A (siHIF1A) showed no differences in FOXO3 levels (Supporting information Figure S7a in Appendix S1). In addition, hypoxia increased the expression of autophagic genes (MAP1LC3, BNIP3, and PRKAA2) and autophagic activity as evidenced by decreased protein levels of p62 and increased levels of lipidated LC3 (LC3-II) . In agreement with previous reports (Choi et al., 2016), this induction of autophagy by hypoxia was not regulated by HIF1A (Supporting information Figure S7b in Appendix S1).
Next, we sought to investigate whether FOXO regulate the expression of homeostatic genes in NP cells during hypoxic conditions. Human NP cells were transfected with siFOXO1, siFOXO3, or a combination of siFOXO1 + 3 and incubated in 1% oxygen for 24 hr. FOXO3 and FOXO1 + 3 knockdown significantly decreased expression of SESN3, SOD2, and MAP1LC3, whereas siFOXO1 + 3transfected cells also showed lower levels of BECN1 and PRKAA2 ( Figure 5d). In addition, a significant reduction in LC3-II levels was observed upon knockdown of FOXO3 and FOXO1 + 3 (Figure 5e).
Conversely, ectopic overexpression of tamoxifen-inducible forms of FOXO1 (FOXO1-ER) or FOXO3 (FOXO3-ER) in human immortalized NP cells (Sakai et al., 2004) increased mRNA levels of SESN3, MAP1LC3, GABARAPL1, and PRKAA2 (Figure 5f). This transcriptional activation of homeostatic genes was likely a direct function of FOXO as overexpression of a FOXO3 mutant that lacks the DNA binding region (FOXO3-ΔDBD-ER) (Tran et al., 2002) did not change gene expression (Supporting information Figure S8 in Appendix S1). In Statistical comparisons were assessed by an unpaired, two-tailed t-test after testing for equal variance using an F-test. Values are mean ± SD. *p < 0.05 addition, LC3-II levels were significantly elevated in cells overexpressing FOXO1 or FOXO3 (Figure 5g), indicating that FOXO is sufficient to stimulate autophagy in NP cells.

| FOXO promote resistance to oxidative and inflammatory stress in NP cells
To investigate whether a decrease in FOXO would compromise NP cell viability under stress conditions, human NP cells were transfected with siControl, siFOXO1, siFOXO3, or siFOXO1 + 3 and treated with increasing concentrations of H 2 O 2 or with a cytokine cocktail (10 ng/ml of TNF-α and 10 ng/ml of IL-1β) in serum-free media for 24 hr. Simultaneous knockdown of FOXO1 and FOXO3 resulted in a significant decrease in cell viability upon H 2 O 2 or TNF-α or IL-1β stimulation (Figure 6a, b). In addition, at higher H 2 O 2 concentrations, cell viability was significantly lower in siFOXO3-treated cells, but not in cells with FOXO1 knockdown, suggesting that FOXO3 is the main isoform responsible for oxidative stress resistance in NP cells.
To examine the potential therapeutic benefit of targeting FOXO to preserve IVD homeostasis, human NP cells were transduced with adenovirus encoding for GFP (Ad-GFP) or constitutively active

| DISCUSSION
Chronic pain and disability due to IDD is one of the most common age-associated conditions in humans (Dagenais et al., 2008;Manchikanti et al., 2014). Aging is a major risk factor for the initiation and progression of IDD (Miller et al., 1988), and given the increase in life expectancy worldwide (2015), it has become a priority to identify the causes and mechanisms of IDD to reveal new therapeutic targets and approaches. Previous studies have proven useful in demonstrating the importance of changes in the expression of genes encoding structural components of ECM and growth factor signaling mediators in the pathophysiology of IDD (reviewed in Vo et al., 2013;Daly, Ghosh, Jenkin, Oehme, & Goldschlager, 2016). However, the precise molecular events that lead to cell dysfunction and potentially initiate the degenerative cascade during aging remain unknown.
To our knowledge, this study provides the first evidence of specific transcription factors that regulate IVD aging and homeostasis. The focus on FOXO was based on their critical role in cellular homeostasis and aging (Eijkelenboom & Burgering, 2013;Webb & Brunet, 2014) and on our previous findings that the expression of FOXO is reduced in degenerated human IVD and during mouse spine aging (Alvarez-Garcia, Matsuzaki, Olmer, Masuda, et al., 2017). Here, we F I G U R E 6 FOXO promote resistance to oxidative and inflammatory stress in human nucleus pulposus cells. (a, b) Cell viability analysis in human NP cells transfected with siRNA specific for FOXO1 (siFOXO1), FOXO3 (siFOXO3), or a combination of both (siFOXO1 + 3) and treated with different H 2 O 2 concentrations (a) or TNF (10 ng/ml) and IL-1β (10 ng/ml) (b) for 24 hr. (c) Cells from normal human NP were transduced with adenovirus encoding green fluorescent protein (GFP) or a constitutively active FOXO1 mutant (FOXO1-AAA) and stimulated with IL-1β (1 ng/ml). RNA was isolated after 6 hr for qPCR analysis. Values shown are mean ± SD of three different experiments performed in duplicate. Statistical comparisons were assessed by an unpaired, two-tailed t-test after testing for equal variance using an F-test. *p < 0.05. (d) Schematic representation of FOXO mechanisms in regulating intervertebral disk (IVD) maturation and homeostasis. NP: nucleus pulposus; AF: annulus fibrosus; EP: endplate report novel functions of FOXO in maintaining IVD integrity during aging by promoting essential homeostatic mechanisms in IVD cells, including autophagy, adaptation to the hypoxic environment, and protection against oxidative and inflammatory stresses.
Our data using the Col2a1Cre mouse model showed that loss of FOXO resulted in NP and EP cell death and tissue degeneration, suggesting a central role of FOXO in IVD homeostasis. However, these degenerative changes were also associated with profound abnormalities in growth and maturation of IVD and vertebrae with FOXO1 being the predominant isoform responsible for these changes as suggested by our isoform-specific KO mouse models. As these developmental abnormalities could negatively impact the health of the IVD, we used the AcanCreER model to delete FOXO in skeletally mature mice. AcanCreER-FOXO KO mice exhibited degenerative features that closely resembled the histological phenotype found in age-related IDD characterized by a reduction in NP and EP cellularity, loss of NP/AF demarcation, and ossification of the EP (Alvarez-Garcia, Matsuzaki, Olmer, Masuda, et al., 2017;Boos et al., 2002;Tam et al., 2017), but had no changes in the vertebral bodies.
In addition, mice with conditional deletion of FOXO3 did not exhibit any IVD defects in postnatal growth or maturation but showed early onset of IDD with aging also characterized by a marked reduction in NP and EP cellularity, further supporting a critical role of FOXO in promoting mature IVD homeostasis by directly regulating cell survival.
Loss of NP cellularity is one of the earliest and most important events in IDD that is hypothesized to drive the progression of the disease (Anderson & Tannoury, 2005;Ding, Shao, & Xiong, 2013;Kadow et al., 2015;Vergroesen et al., 2015). NP cells are unique in their ability to adapt to their physiological environment, defined by low-oxygen and nutrient levels (Bartels, Fairbank, Winlove, & Urban, 1998). In a recent study, Choi et al. (2016) showed that NP cells activate autophagy under hypoxic conditions through a noncanonical, HIF1A, and mTOR-independent mechanism and that this activation of autophagy was required for long-term NP cell survival. Here, we identified FOXO3 as a novel mediator of NP cell adaptation to hypoxia and demonstrated that FOXO3 and, to a lesser extent, FOXO1 regulate the expression of autophagy genes and activation in NP cells. Although a causal relationship between autophagy defects and IDD has not yet been established, our findings showed that FOXO function promotes NP cell survival under oxidative and inflammatory stresses, major drivers of NP cell apoptosis and agingrelated IDD Wang et al., 2013;Yang et al., 2014), and thus suggest that activation of FOXO in aged IVD could increase autophagy and protect NP cells from stress-induced apoptosis. Further supporting this hypothesis, various studies have shown that autophagy activation can protect NP cells against different apoptotic insults (Jiang, Jin, Wang, Jiang, & Dong, 2014;Jiang, Zhang, et al., 2014).
In the present study, Col2a1Cre and AcanCreER drivers that are active in NP, AF, and EP (Henry et al., 2009;Jin et al., 2011) were chosen to delete FOXO in IVD to model the reduction of FOXO expression seen in all IVD compartments during mouse aging (Alvarez-Garcia, Matsuzaki, Olmer, Masuda, et al., 2017). In addition to the severe degenerative changes in the NP, our data using both mouse models show that FOXO deletion led to cell loss and ossification in the cartilaginous EP. The EP has a pivotal role in regulating the nutritional status of the IVD and the age-related decrease in EP permeability due to ossification has been suggested as a contributing factor in IDD (Bartels et al., 1998;Bernick & Cailliet, 1982;Wang, Battie, Boyd, & Videman, 2011). It is thus likely that reduced EP permeability secondary to EP calcification significantly contributes to the severe degeneration observed in FOXO-deficient IVD. Importantly, the present findings imply that FOXO-targeted therapies could preserve cell viability during aging and have beneficial effects on all IVD tissues.
The Col2a1Cre and AcanCreER drivers also cause FOXO deletion in other cartilaginous structures such as in the articular cartilage of the synovial joints and the growth plates (Henry et al., 2009;Zhu, Chen, Lichtler, O'Keefe, & Chen, 2008). We have recently reported the knee articular cartilage phenotype in these mouse models and demonstrated that FOXO are critical mediators of chondrocyte homeostasis by regulating PRG4 expression and maintaining the integrity of the cartilage superficial zone (Matsuzaki et al., 2018). We thus identified common and distinct molecular mechanisms that are regulated by FOXO in spine and synovial joints. The most important common feature of joints and spine in this regard is the aging and disease-associated suppression of FOXO which could serve as a foundation to develop therapeutic interventions to prevent the onset of different aging-related skeletal diseases such as IDD or osteoarthritis.
In conclusion, this study identified FOXO as essential regulators of IVD homeostasis during aging (Figure 6c). The protective effects of FOXO against oxidative stress, stimulation of autophagy and adaptation to hypoxia, and suppression of key catabolic mediators that contribute to IDD support the concept that the aging-associated reduction in FOXO expression (Alvarez-Garcia, Matsuzaki, Olmer, Masuda, et al., 2017) has important functional implications and may represent a central mechanism of IVD aging and risk factor for IDD. Therefore, maintaining or restoring FOXO expression during aging can be a therapeutic strategy to prevent or delay the onset the disease.

| Human IVD tissues
Normal human lumbar IVD samples were collected from five cadaveric donors (age 19-45 years). Macroscopic assessment was performed according to Thompson grading (Thompson et al., 1990) and confirmed the absence of any pathological changes. Additional details of the samples are listed in Supporting information Table S1 in Appendix S1.

| Histological analyses
Lumbar spines were collected from FOXO mutant and control mice.
Samples were fixed and decalcified as described previously (Alvarez-Garcia, Matsuzaki, Olmer, Masuda, et al., 2017). The 4-µm-thick sagittal sections were stained with safranin O-fast green or picrosirius red staining for morphological analysis. Histological grading of NP/AF in mouse L4/L5 IVD was performed following the system described by Masuda et al. (2005), and degenerative changes in EP were graded according to a modification to the scoring system described by Boos et al. (2002). Briefly, the grading system in the NP and AF evaluated four different parameters: NP cellularity, border between NP and AF, matrix of the NP, and lamellar morphology in the AF. Each parameter was scored from 1 to 3 points, with 4 points (1 point in each category) representing normal histology and 12 points indicating severe degeneration. The scoring system for EP evaluated four parameters: cellularity (0-4 points), cartilage disorganization (0-4 points), cartilage continuity (0-4 points), and calcification/bone formation (0-4 points), for a combined score ranging from 0 to 16 points. At least six mice per experimental group were scored by two different observers blinded to the experimental conditions. Additional histomorphometric analysis was performed on images from L4/L5 IVD obtained under 20× magnification using IMAGEJ software. Disk height was calculated by averaging 20 parallel measurements of the distance between cranial and distal cartilage endplates.
Cellularity of each IVD tissue and vertebral growth plate was calculated by dividing number of cells in each specific tissue by the area and was expressed in cell number/mm 2 . Vertebral diameter, number of trabeculae, and average trabecular thickness were measured in the distal subchondral bone at a distance of 50 µm of the growth plate.

| Immunohistochemistry
Immunohistochemistry of lumbar spine sections was performed as

| Immunofluorescence
Lumbar spine sections were processed as described above and incu-

| Cell proliferation and apoptosis
To evaluate the rate of cell proliferation, injections of 100 mg/kg of BrdU (Sigma-Aldrich) were administered intraperitoneally to 3-week-

| Cell culture
Human NP cells were isolated following the protocol described by Sakai et al. (2012)

| Gene expression analysis
Mouse lumbar spine samples were collected at 2 and 4 months of age. NP and AF from all lumbar IVD were resected separately for each mouse and homogenized in QIAzol Lysis Reagent (Qiagen, Valencia, CA, USA). RNA was isolated using Direct-zol RNA miniprep kit (Zymo Research, Irvine, CA, USA). RNA quality was assessed by measuring the 260/280 absorbance ratio. All RNA samples used in this study had a 260/280 ratio > 1.8. Gene expression was measured by real-time PCR using predesigned TaqMan gene expression assays (Thermo Fisher Scientific). At least four samples were analyzed in duplicate for each experimental condition.

| Statistics
All data are reported as the mean ± standard deviation (SD). All data were tested for normal distribution using the Kolmogorov-Smirnov test. Statistical comparisons between more than two groups were assessed by one-way analysis of variance (ANOVA) followed by a post hoc Tukey's test. Comparisons between two groups were assessed by an unpaired, two-tailed t-test after testing for equal variance using an F-test. All statistical analyses were performed using PRISM 6 software (GraphPad Software). p-Values less than 0.05 were considered significant.

| Study approval
All mouse studies were approved by the Scripps Institutional Animal Care and Use Committee and performed in accordance with ARRIVE guidelines and with attention to the standards reported by the NIH-NINDS designed to enhance animal study transparency and reproducibility.

ACKNOWLEDGMENTS
This study was supported by NIH grants AG053747 (O.A.) and AG007996 (M.L.). We thank Josan Chung for her technical assistance. We also thank Dr. Nissim Hay, Dr. Eugene Kandel, and Dr.

CONFLI CT OF INTEREST
The authors have declared that no conflict of interest exists.

AUTHOR CONTRI BUTION
OA and ML designed the study; OA, TM, MO, KM, and SM performed experiments; OA, DS, KM, HA, and ML analyzed the data; OA and ML wrote and edited the manuscript. All authors reviewed and approved the final version of the manuscript.

S U P P O R T I N G I N F O R M A T I O N
Additional supporting information may be found online in the Supporting Information section at the end of the article.