Research ArticleOsteoarthritis

ANP32A regulates ATM expression and prevents oxidative stress in cartilage, brain, and bone

See allHide authors and affiliations

Science Translational Medicine  12 Sep 2018:
Vol. 10, Issue 458, eaar8426
DOI: 10.1126/scitranslmed.aar8426

Oxidative stress and osteoarthritis

Osteoarthritis is a common degenerative joint disorder that affects cartilage and bone. Cornelis et al. investigated the role of ANP32A, a protein involved in multiple cellular processes, in osteoarthritis. ANP32A was decreased in osteoarthritic human and mouse tissue samples and also decreased with aging. The authors found that ANP32A promoted transcription of ATM and regulated reactive oxygen species in cartilage. Antioxidant therapy protected Anp32a-deficient mice from developing osteoarthritis and osteopenia and also rescued neurological defects caused by lack of ATM and increased oxidative stress. These results suggest that ANP32A could be a therapeutic target for correcting imbalanced reactive oxygen species and antioxidants.


Osteoarthritis is the most common joint disorder with increasing global prevalence due to aging of the population. Current therapy is limited to symptom relief, yet there is no cure. Its multifactorial etiology includes oxidative stress and overproduction of reactive oxygen species, but the regulation of these processes in the joint is insufficiently understood. We report that ANP32A protects the cartilage against oxidative stress, preventing osteoarthritis development and disease progression. ANP32A is down-regulated in human and mouse osteoarthritic cartilage. Microarray profiling revealed that ANP32A protects the joint by promoting the expression of ATM, a key regulator of the cellular oxidative defense. Antioxidant treatment reduced the severity of osteoarthritis, osteopenia, and cerebellar ataxia features in Anp32a-deficient mice, revealing that the ANP32A/ATM axis discovered in cartilage is also present in brain and bone. Our findings indicate that modulating ANP32A signaling could help manage oxidative stress in cartilage, brain, and bone with therapeutic implications for osteoarthritis, neurological disease, and osteoporosis.


Osteoarthritis is the most common chronic joint disease worldwide. Patients with this musculoskeletal disorder suffer from pain and progressive loss of joint function and mobility. It therefore features among the leading causes of chronic disability, is strongly linked with aging and the obesity pandemic, and affects more than 25% of the adult population (1). Morphological, biochemical, molecular, and biomechanical changes of the cells and extracellular matrix in the tissues of the joint lead to loss of articular cartilage, sclerosis of the subchondral bone, formation of bone spurs called osteophytes, and, in some cases, inflammation of the synovial membrane that lines the joint cavity (2). The molecular mechanisms involved in the initiation and progression of osteoarthritis are insufficiently understood. Thus, there are no effective interventions to restore degraded cartilage and to delay or halt disease progression (3). Current treatment strategies are limited to changes in lifestyle, exercise, and use of painkillers and anti-inflammatory drugs. Eventually, joint replacement surgery is often required, especially for the large joints (4, 5).

The etiology of osteoarthritis is multifactorial and includes a strong genetic component adding to the deleterious effects of joint injury, obesity, and aging (6). Understanding the biological mechanisms underlying the genetic susceptibility could not only provide novel insights into disease pathogenesis but also lead to the development of new therapies that slow or stop the progression of the disease. We previously described an association between polymorphisms in the acidic leucine-rich nuclear phosphoprotein-32A (ANP32A) gene and osteoarthritis (7, 8). ANP32A is a small multifunctional protein that plays a role in a variety of cellular processes such as caspase modulation, chromatin modifications, protein phosphatase inhibition, and intracellular transport of molecules (9), regulating cell differentiation, transcription, apoptosis, and cell cycle progression (1013). However, the potential role of ANP32A in osteoarthritis and cartilage biology remains unknown.

Here, we report that ANP32A protects against the development and progression of osteoarthritis by preventing oxidative stress in the articular cartilage. We also unravel the underlying molecular mechanism, which involves transcriptional control of the ataxia-telangiectasia mutated (ATM) gene, an important regulator of the cellular defense against oxidative stress (14). In addition, the regulatory role of ANP32A and its underlying mechanism discovered in cartilage are also present in brain and bone. Thus, our results may have therapeutic implications not only in chronic joint disorders but also in bone and neurological diseases.


ANP32A is decreased in osteoarthritic and aging cartilage

To test whether the expression of ANP32A is different between osteoarthritic and non-osteoarthritic joints, we performed gene expression analysis on human articular cartilage. ANP32A expression was reduced in hip cartilage from patients with osteoarthritis compared to trauma patients with macroscopically intact cartilage, both undergoing hip replacement surgery (Fig. 1A; for patient characteristics, see table S1). These observations were confirmed at the protein level by Western blot (Fig. 1B). Immunohistochemical analysis similarly showed that the strong immunoreactive signal in control non-osteoarthritic cartilage was decreased in cartilage from patients with osteoarthritis (Fig. 1C). Furthermore, in patients with knee and hip osteoarthritis, ANP32A expression was reduced in damaged as compared to preserved areas of the articular cartilage (Fig. 1D). Immunoreactivity for ANP32A was decreased in the articular cartilage of wild-type mice with osteoarthritis triggered by surgical destabilization of the medial meniscus (DMM) compared to sham-operated animals (Fig. 1E) and in the articular cartilage of 12-month-old mice compared to 8-week-old mice (Fig. 1F). Together, these findings suggest that high ANP32A expression is positively associated with cartilage health.

Fig. 1 Loss of ANP32A increases osteoarthritis severity and susceptibility.

(A and B) Real-time polymerase chain reaction (PCR) (A) and immunoblot (B) analysis of ANP32A mRNA and protein (with actin as loading control) in human articular chondrocytes from hip osteoarthritis (OA) and fracture patients (non-OA) (n = 4 per group; *t6 = 3.79, P = 0.009, t test). (C) Immunohistochemical staining for ANP32A in OA and non-OA human cartilage (n = 4 per group; scale bar, 400 μm). (D) ANP32A expression by RNA sequencing in paired preserved and damaged cartilage from hips (circles) and knees (triangles) from osteoarthritis patients [log2-fold change (Log2FC): damaged versus preserved] (n = 21, P < 0.001, Benjamini-Hochberg–adjusted paired t test). (E and F) Immunohistochemical staining for ANP32A in male wild-type (WT) mice 12 weeks after induction of osteoarthritis by DMM (Surgery) compared to sham-operated mice (Sham) (E) and in 8-week-old male or 12-month-old female mice (F) [n = 5 (E) and 3 (F) per group]. (G to J) Hematoxylin–safranin-O–stained sections (G and I) and quantification by Osteoarthritis Research Society International (OARSI) severity grade (H and J) of cartilage damage in knee joints after DMM and after 12-month aging of mice [n = 8 and 10 male mice, *t34 = 5.16, P < 0.001, Bonferroni-corrected for six tests in one-way analysis of variance (ANOVA) (H); n = 10 female mice per group, *t18 = 3.05, P = 0.007, t test (J)]. (E to G and I) Scale bars, 200 μm. Error bars indicate mean ± SD. IgG, immunoglobulin G.

Loss of Anp32a increases severity of and susceptibility to osteoarthritis in mice

To study the impact of ANP32A deficiency on articular cartilage, we challenged Anp32a−/− mice (mice with a global gene deletion of ANP32A) in different models of osteoarthritis. Anp32a−/− mice showed increased cartilage damage compared to wild-type controls in the gradually progressive DMM disease model driven by joint instability, in the rapidly progressive model triggered by intra-articular collagenase injection driven by severe instability and inflammation, and in the direct cartilage matrix damage model induced by intra-articular papain injection (Fig. 1, G and H, and fig. S1, A to D). ANP32A deficiency also predisposes to an accelerated development of spontaneously occurring osteoarthritic changes upon aging, as Anp32a−/− mice developed more severe signs of osteoarthritis by 12 months of age compared to wild-type mice (Fig. 1, I and J). Collectively, these data indicate that loss of ANP32A increases the severity of and the susceptibility to osteoarthritis.

ANP32A regulates Atm expression and oxidative stress in cartilage

To explore the mechanisms underlying the impact of ANP32A on the onset and progression of osteoarthritis, we compared genome-wide transcriptome profiles from the articular cartilage of Anp32a−/− mice and wild-type mice at 8 weeks of age (geonr: GSE108036). At this age, no differences in the expression of healthy chondrocyte markers collagen 2 or aggrecan were detected between the groups (fig. S2A). In Anp32a−/− mice, expression levels of 118 genes were different compared to control mice when applying a threshold of twofold change and a nominal P value of ≤ 0.01 (Fig. 2A). The gene set included 106 down-regulated genes and 12 up-regulated genes (top ranked genes are shown in table S2). Comparative pathway analyses of differentially regulated transcripts using the PANTHER database revealed an enrichment in genes associated with the oxidative stress response and the ubiquitin proteasome pathway (Fig. 2B), which is known to play a pivotal role in the recognition and degradation of oxidized proteins (15). Biological process categories found to be enriched in PANTHER analysis included DNA metabolic process, phosphate-containing compound metabolic process, catabolic process, and cellular protein modification process (fig. S3). In addition, pathways from the Reactome database of curated biological processes in humans that are found to be enriched included mitochondrial translation, regulation of p53, and oxidative stress–induced senescence (fig. S3). Thus, the transcriptome analysis suggests that oxidative stress and dysregulation of phosphorylation events play a role in the deleterious effects associated with ANP32A deficiency in cartilage.

Fig. 2 ANP32A deficiency reduces ATM and triggers oxidative stress in cartilage.

(A and B) Volcano plot (A) and PANTHER pathway analysis (B) of microarray data comparing articular cartilage of 8-week-old male Anp32a-deficient to wild-type mice (n = 4 per group). (C and D) Real-time PCR (C) and immunoblot analysis (D) of ATM mRNA and protein (with actin as loading control) in mouse articular chondrocytes. (E to G) Immunohistochemical staining for ATM (E), 8-hydroxydeoxyguanosine (8-OHdG) (F), and DHE staining (G) in knees from 8-week-old male mice (n = 3). DAPI, 4′,6-diamidino-2-phenylindole. (H and I) Real-time PCR (H) and immunoblot (I) analysis of ATM mRNA and protein in articular chondrocytes from hip osteoarthritis (OA) and fracture patients (non-OA) [n = 4 per group; *t6 = 9.04, P < 0.001, t test]. (J and K) ATM expression (J) and correlation with ANP32A expression (K) by RNA sequencing in paired preserved and damaged cartilage from hips (circles) and knees (triangles) from osteoarthritis patients [log2-fold change (Log2FC) of damaged versus preserved] [n = 21, P = 0.008, Benjamini-Hochberg–adjusted paired t test (J), Pearson’s correlation = 0.536, P < 0.001 (K)]. (L to N) Real-time PCR analysis of ANP32A (L) and ATM expression (M) and 2′,7′- dichlorofluorescin diacetate (DCFDA) ROS activity staining (N) in human articular chondrocytes transfected with ANP32A (siANP) or scrambled siRNA (siSCR), treated with H2O2 and recombinant ATM protein (rATM) {0.5 μg/ml [low dose (LD)] or 1 μg/ml [high dose (HD)]} (n = 2, error bars indicate mean ± SD of three technical replicates). Scale bars, 200 μm (F and G) and 50 μm (E).

We identified the ataxia-telangiectasia mutated serine/threonine kinase (Atm) gene in a prominent position within the top down-regulated genes (Fig. 2A and table S2). Atm encodes a protein kinase that orchestrates signaling cascades to preserve the cellular redox balance and promote the antioxidant response (14). Near absence of Atm in Anp32a−/− mice was confirmed by quantitative PCR (qPCR) (Fig. 2C). We detected a considerable reduction of ATM protein expression in articular chondrocytes and in the articular cartilage of Anp32a−/− mice compared to controls, analyzed by Western blot and immunohistochemistry, respectively (Fig. 2, D and E). This decrease in ATM expression paralleled enhanced reactive oxygen species (ROS) production in the articular cartilage of Anp32a−/− mice, assessed by immunohistochemistry and dihydroethidium (DHE) staining (Fig. 2, F and G).

We then measured mRNA expression and protein amounts of ATM in cartilage from non-osteoarthritic trauma patients and from patients with osteoarthritis. We observed that ATM expression was down-regulated in cartilage from patients with osteoarthritis compared to non-osteoarthritic cartilage (Fig. 2, H and I) and in damaged areas compared to preserved areas (Fig. 2J), paralleling the changes in ANP32A (Fig. 1, A, B, and D). Expression of ANP32A and ATM correlated (Fig. 2K). Moreover, ANP32A knockdown in primary healthy human articular chondrocytes resulted in low ATM expression (Fig. 2, L and M). H2O2 treatment in ANP32A knockdown cells strongly increased ROS production, an effect rescued by treatment with recombinant ATM (Fig. 2N). Overall, our results show a close link between ANP32A, oxidative stress, and ATM expression and function in articular cartilage.

The antioxidant system is complex and comprises multiple interacting and interdependent mechanisms (16). Thus, we investigated compensatory mechanisms in the articular cartilage of the Anp32a−/− mice using our transcriptome data. We focused primarily on molecules involved in the regulation of glutathione, the most abundant endogenous antioxidant, on the peroxiredoxin and thioredoxin families, and on catalase and the superoxide dismutases (table S3). None of the individual genes analyzed met our microarray thresholds of fold change and P value. At the group level, overall changes in gene expression suggested an increase in enzymes responsible for glutathione synthesis and reduction (fig. S4). These data suggest that oxidative stress triggered by loss of ANP32A may result in compensatory increase in the glutathione system that is insufficient to restore the cellular redox balance.

ANP32A directly promotes ATM expression in articular chondrocytes

Given the strong link found between ANP32A and ATM, we investigated whether ANP32A directly regulates ATM at the transcriptional level. In primary healthy human articular chondrocytes, chromatin immunoprecipitation–qPCR (ChIP-qPCR) demonstrated that ANP32A localized at the ATM gene promoter (Fig. 3A). To determine whether ANP32A actively participates in the regulation of ATM transcription, we evaluated the binding of the RNA polymerase II transcription machinery to the ATM promoter after small interfering RNA (siRNA)–mediated knockdown of ANP32A in human articular chondrocytes compared to controls. The recruitment of RNA polymerase II to the ATM promoter was low in ANP32A knockdown cells (Fig. 3B). These data indicate that ANP32A directly promotes the transcription of the ATM gene in human articular chondrocytes.

Fig. 3 ANP32A directly induces ATM expression to prevent oxidative stress in cartilage.

(A) ChIP-qPCR analysis of ANP32A binding to different regions of the ATM gene promoter in non-osteoarthritic human articular chondrocytes. (B) ChIP-qPCR analysis of RNA polymerase II binding to the ATM promoter in human articular chondrocytes transfected with siANP or scrambled siRNA. (C and D) Real-time PCR (C) and immunoblot (D) analysis of ATM mRNA and protein in articular chondrocytes treated over time with H2O2 or vehicle (V) for 72 hours. (E) Immunoblot analysis of cytosolic and nuclear ATM and ANP32A protein amounts in response to H2O2 treatment for the indicated times in human articular chondrocytes. The images are representative of images from three independent experiments. (F) ChIP-qPCR analysis of ANP32A binding to the ATM promoter upon H2O2 treatment in human articular chondrocytes. (A, B, and F) Data from two biologically independent experiments. (C) Data are from two experiments with three technical replicates. Error bars indicate mean ± SD.

Oxidative stress induced by H2O2 treatment increased ATM gene (Fig. 3C) and protein expression (Fig. 3D) in articular chondrocytes. The enhancement in ATM protein was observed in the cytoplasmic fraction (Fig. 3E), in agreement with the reported role of extranuclear ATM in oxidative stress defense colocalizing with peroxisomes, which are a major site of oxidative metabolism (1719). Oxidative stress induced by H2O2 treatment triggered rapid translocation of ANP32A protein to the nucleus (Fig. 3E) associated with high ANP32A binding to the ATM gene promoter (Fig. 3F). These results suggest that ANP32A may prevent osteoarthritis development and progression by increasing ATM in the articular cartilage as an endogenous protective mechanism against oxidative stress.

Antioxidant treatment prevents osteoarthritis in Anp32a-deficient mice

To explore the therapeutic implications of our findings, we evaluated whether pharmacological antioxidant treatment is protective against the development of osteoarthritis in Anp32a-deficient mice. Anp32a−/− mice and controls were given the ROS inhibitor N-acetyl-cysteine (NAC) in drinking water, which reduced the severity of osteoarthritis in Anp32a-deficient mice in the DMM model (Fig. 4, A and B) and prevented ROS production in the articular cartilage of these mice (Fig. 4C).

Fig. 4 Antioxidant treatment prevents osteoarthritis in Anp32a-deficient mice.

(A and B) Hematoxylin–safranin-O–stained sections (A) and quantification by OARSI severity grade (B) of articular cartilage damage in knee joints of 20-week-old male mice 12 weeks after induction of osteoarthritis by the DMM model, treated with vehicle or NAC [n = 2 (WT-Sham NAC), 5 (WT-Surgery), 5 (WT-Surgery NAC), 2 (Anp32a−/−-Surgery), and 5 (Anp32a−/−-Surgery NAC), *t14 = 3.58, P = 0.03, Bonferroni-corrected for 10 tests in one-way ANOVA]. (C) Immunohistochemical detection of 8-OHdG to measure ROS in knees from WT and Anp32a−/− mice after induction of DMM osteoarthritis with or without NAC treatment. Scale bar, 200 μm. (D and E) Immunohistochemistry (D) and quantification by digital image analysis (E) of COLX in the articular cartilage of 16-week-old Anp32a−/− compared to WT mice with or without NAC treatment [n = 3, NAC reduced relative intensity more in knockout (KO) than in WT, interaction F1,8 = 5.78, P = 0.0429 by two-way ANOVA]. Images are representative of images from three to five different mice. Scale bars, 200 μm (A and C) and 50 μm (D). Error bars indicate mean ± SD.

Atm-deficient mice have high amounts of ROS in growth plate chondrocytes, which leads to a proliferation defect and the stimulation of chondrocyte hypertrophy (20). We observed increased immunohistochemical staining of hypertrophy marker type X collagen (COLX) in the growth plates of Anp32a-deficient mice compared to controls, and this effect was normalized by NAC treatment (fig. S5, A and B). Osteoarthritis is associated with ectopic hypertrophic differentiation of chondrocytes in the articular cartilage. Altered matrix composition and factors secreted by these hypertrophic-like articular chondrocytes, such as matrix metalloproteinase-13 (MMP-13), likely contribute to cartilage degeneration (21). Therefore, we evaluated whether Anp32a-deficient mice showed increased hypertrophic differentiation in the articular cartilage as a consequence of the high ROS production. We detected increased COLX staining in the articular cartilage of Anp32a-deficient mice compared to controls (Fig. 4, D and E). Antioxidant treatment prevented chondrocyte hypertrophy in the articular cartilage of Anp32a-deficient mice (Fig. 4, D and E). These data further demonstrate that prevention of oxidative stress is the pivotal mechanism by which ANP32A protects cartilage against osteoarthritis.

Anp32a deficiency leads to ataxia-like neurological defects that are attenuated by antioxidant treatment

We then explored whether our insights into the role of ANP32A in cartilage and its discovered link with ATM have clinical implications beyond this tissue, in particular in the context of ataxia-telangiectasia (A-T). Patients with A-T lack functional ATM protein and exhibit a pleiotropic phenotype that includes cerebellar ataxia, immunodeficiency, and premature aging (22). Most of the clinical manifestations of A-T seem to result from an inability to limit the production of ROS (23). Atm expression was strongly down-regulated in Anp32a−/− mouse brain compared to wild-type mice (Fig. 5A). ATM protein expression was strongly down-regulated in the cerebellum of Anp32a−/− mice (Fig. 5B). We also observed increased ROS production in the cerebellum of Anp32a−/− mice compared to controls (Fig. 5C).

Fig. 5 Anp32a deficiency leads to ataxia-like neurological defects, prevented by antioxidant treatment.

(A) Real-time PCR analysis of ATM expression in adult brain from 8-week-old male WT and Anp32a-deficient mice (n = 9). (B) Immunohistochemical staining for ATM in cerebellum from 8-week-old male Anp32a-deficient mice compared to WT mice. (C) Immunohistochemical detection of 8-OHdG to measure ROS in 16-week-old male cerebellum. (D) Time course of oral treatment with NAC and CatWalk automated gait analysis in WT and Anp32a-deficient (KO) mice [n = 5 (WT no NAC), 7 (KO no NAC), 5 (WT NAC), 5 (KO NAC)]. (E) Gait analysis of 8-week-old male mice. Footprint colors were assigned manually (green, right; red, left; light print, forelimbs; dark print, hindlimbs). (F) Average stride length analyzed with ANOVA accounting for genotype (WT, KO), NAC (yes, no), paw [front paw (FP), hind paw (HP)], and all interactions (genotype/NAC interaction F1,18 = 6.587, P = 0.019; *for indicated pairwise comparisons, all P < 0.001, Bonferroni-corrected for 12 tests). (G) Regularity index of walking pattern analyzed with ANOVA accounting for genotype, NAC, and their interaction (genotype/NAC interaction F1,18 = 9.227, P = 0.007; *for indicated pairwise comparisons, all P ≤ 0.006, Bonferroni-corrected for six tests). Error bars indicate mean ± SD. Scale bars, 500 μm. Details about ANOVA in data file S1.

To identify motor impairments in Anp32a−/− mice, we performed CatWalk gait analysis (Fig. 5, D to G) to reveal subtle coordination defects related to cerebellar dysfunction (24, 25). This approach suggested the presence of ataxia in Anp32a−/− mice, as their stepping pattern was less consistent than in control mice. The average distance between each stride (stride length) was shorter in Anp32a−/− mice compared with wild-type controls (Fig. 5, E and F). Anp32a−/− mice also showed a lower regularity index (Fig. 5, E and G). Additional parameters of the CatWalk analysis that were altered in Anp32a−/− mice are shown in table S4. Consistent with the observations in Atm-deficient mice (26, 27), Anp32a−/− mice appeared to have a normal cerebellar architecture and did not show histological signs of neuronal degeneration (fig. S6).

Using oral administration of NAC, a drug that can cross the blood-brain barrier (28, 29) and reduce oxidative stress in the brain at the dosage used (30), we studied whether antioxidant treatment also protected Anp32a−/− mice from the development of ataxia-like features (Fig. 5, C to G). Oral NAC supplementation in Anp32a−/− mice from the age of 3 weeks until mice were 8 weeks old (Fig. 5D) decreased ROS production in the cerebellum (Fig. 5C) and ameliorated the gait disturbances associated with loss of ANP32A (Fig. 5, E to G, and table S4). An exploratory experiment, albeit with limitations linked to the number of mice included, suggested that pharmacological intervention with NAC in Anp32a-deficient mice at an older age, from 8 to 12 months old, also ameliorated gait impairments (fig. S7). Thus, ataxia-like features in Anp32a−/− mice can be linked to loss of ATM and excessive ROS production.

Anp32a deficiency leads to osteopenia in mice that is prevented by antioxidant treatment

Atm−/− mice exhibit an osteopenic bone phenotype, mainly due to impaired bone formation (31, 32). To further validate the link between ANP32A and ATM, we examined whether Anp32a deficiency also affected bone homeostasis. Twelve-week-old Anp32a−/− mice screened using dual-energy x-ray absorptiometry (DEXA) demonstrated reduced bone mineral density, bone mineral content, and lean body mass compared to wild-type littermates (Fig. 6A). Peripheral quantitative computed tomography (pQCT) of femora from 12-week-old Anp32a−/− mice confirmed decreased trabecular bone content, density, and area as well as decreased cortical bone content and area (Fig. 6B). These observations were corroborated by histology of the tibiae dissected from 16-week-old mice (Fig. 6C). These data support that the ANP32A-ATM axis also protects bone homeostasis by limiting ROS production. Of note, an exploratory analysis by in vivo microcomputed tomography (μCT) suggested that NAC intervention may improve bone health in the Anp32a-deficient mice (Fig. 6D). Further, end-stage analysis by histomorphometry, however, could not provide consistent confirmation for this observation (Fig. 6, C and E).

Fig. 6 Anp32a deficiency leads to osteopenia that is responsive to antioxidant treatment.

(A) Subcapital DEXA analysis of bone mineral density (BMD), bone mineral content (BMC), and lean body mass in 12-week-old female Anp32a−/− mice compared to WT littermates (n = 11 and 9, *all t18 ≥ 2.224, all P ≤ 0.039, t test). (B) pQCT of trabecular and cortical bone parameters in femora from 12-week-old female Anp32a−/− mice and WT littermates (n = 10 and 9, *all t17 ≥ 2.215, all P ≤ 0.041, t test). (C) Hematoxylin–safranin-O staining of tibiae from 16-week-old male Anp32a−/− and WT mice with or without NAC treatment for 13 weeks. Scale bar, 250 μm. (D) In vivo μCT of tibiae from female Anp32a−/− mice with or without NAC treatment for 12 weeks from the age of 3 until 6 months [bone volume/tissue volume (BV/TV)] (n = 2). (E) Histomorphometry analysis of tibiae from male Anp32a−/− mice treated or not with NAC for 13 weeks from the age of 3 until 16 weeks (n = 5 per group, *t8 = 2.366, P = 0.046, t test). Error bars indicate mean ± SD.


Here, we demonstrate that ANP32A maintains cartilage homeostasis and protects against the development of osteoarthritis in rodent models of posttraumatic and age-related joint disease. ANP32A’s protective role can be attributed to promoting the expression of ATM in the articular cartilage, to preserve the cellular redox balance (fig. S8)—a molecular mechanism suggested by genome-wide microarray analysis on articular cartilage from Anp32a−/− mice and further elucidated using primary human articular chondrocytes. The ANP32A-ATM axis prevents excessive accumulation of ROS that may directly damage DNA and proteins by oxidation and alter gene transcription programs. We provide evidence that the regulatory role of ANP32A on Atm expression and on oxidative stress not only is limited to cartilage but also exists in brain and bone. Anp32a deficiency leads to ataxia-like cerebellar dysfunction and osteopenia. The identified disease-causing cascade of events seems to be responsive to therapeutic intervention, because cartilage damage and ataxia are effectively prevented and ameliorated by oral antioxidant treatment.

We discovered a role for ANP32A as a positive transcriptional regulator of the Atm gene. This stimulatory function of ANP32A on gene transcription markedly contrasts with its earlier-defined role as transcriptional repressor. ANP32A was first identified as a transcriptional repressor upon purification and characterization of the inhibitor of histone acetyltransferase (INHAT) complex, a multiprotein complex that potently inhibits specific histone acetyltransferases (33). Further mechanistic insights revealed that ANP32A blocks histone acetylation by binding to histone tails and sterically inhibiting the histone modifier enzymes (34, 35). Other reports suggested that ANP32A represses transcription upon its recruitment to gene promoters by different transcription factors (3638). In transcriptome analysis of articular cartilage from Anp32a−/− mice compared to controls presented here, most of the genes that changed were down-regulated, pointing out the unexpected role of ANP32A as a positive regulator of transcription in cartilage. To our knowledge, only one study reported ANP32A to enhance gene transcription of interferon-stimulated genes (39).

ANP32A has been portrayed as a multifunctional protein that, in addition to the above-highlighted role in regulation of gene transcription, also inhibits phosphorylases, modulates caspase activity, and affects intracellular transport (9). However, the specific effects on joint homeostasis reported here appear to be largely dependent on ANP32A’s direct positive involvement in the regulation of ATM expression, as demonstrated by ChIP analysis and the rescue experiments performed.

Our results reveal an endogenous system to control oxidative stress in cartilage. We demonstrated that ANP32A stimulates ATM protein production in articular cartilage as a protective mechanism to enhance the antioxidant capacities of chondrocytes. In addition to its known role as a sensor of DNA double-strand breaks and thereby as a regulator in the DNA damage response, ATM forms active homodimers (23) and manages stress via various pathways, including modulation of mitogen-activated protein kinase (MAPK) pathways, rerouting of carbon metabolism from glycolysis to the pentose phosphate pathway to increase the production of NADPH (reduced form of nicotinamide adenine dinucleotide phosphate) that is essential in the antioxidant defense and nucleotide synthesis, and repression of mTOR (mammalian target of rapamycin) that leads to decreased mitochondrial activity and subsequent ROS production (40). Here, loss of Anp32a resulted in strong down-regulation of Atm and increased ROS production. Earlier studies have demonstrated that excessive ROS contribute to the onset and progression of osteoarthritis (41). Furthermore, ROS production and oxidative stress are elevated in patients with osteoarthritis (4244), and conversely, antioxidant enzymes are reduced in the joints of patients with osteoarthritis, confirming the role of oxidative stress in disease pathogenesis (4347).

Articular cartilage is not vascularized, and consequently, its oxygen supply is limited. Nevertheless, oxygen from the synovial fluid diffuses into articular cartilage, and chondrocytes have mitochondria and respire in vitro, producing ROS (48). In normal conditions, ROS are produced at low concentration in articular chondrocytes and have a homeostatic function, regulating gene expression, extracellular matrix synthesis and breakdown, cytokine production, and chondrocyte apoptosis (49). However, excessive ROS trigger oxidative damage of cellular proteins and also oxidize lipids, carbohydrates, and DNA in human articular cartilage (50). Moreover, high concentration of ROS contributes to cartilage degradation by inhibiting matrix synthesis, cell migration, and growth factor bioactivity, by directly degrading matrix components, activating MMPs, and inducing cell death (51). Given the dual role of ROS in cartilage, it is translationally relevant to gain insights into how the ROS balance is regulated.

Several factors may contribute to excessive ROS production in the onset and progression of osteoarthritis. Chondrocytes can produce abnormal amounts of ROS in response to local oxygen variations (52), mechanical stress (53, 54), and proinflammatory cytokines (55). Aging, a key risk factor for the development of osteoarthritis, is associated with elevated oxidative damage of DNA, proteins, and lipids, and accumulating evidence indicates that oxidative stress is a major physiological inducer of aging (56). We observed reduced expression of ANP32A in aged mouse cartilage and in human cartilage from patients with osteoarthritis, and we showed that Anp32a-deficient mice develop spontaneous osteoarthritis upon aging. Thus, ANP32A can be considered as a key coordinator of oxidative stress and aging in joints.

Both chondrocyte death and the loss of the specific molecular identity of the articular chondrocyte contribute to osteoarthritis. Differentiation of articular cartilage cells toward cells with molecular characteristics of growth plate hypertrophic chondrocytes results in the synthesis of an extracellular matrix that will calcify and impairs the optimal biomechanical characteristics of the bone-cartilage unit in the joint (57). We demonstrated increased chondrocyte hypertrophy in the articular cartilage of Anp32a-deficient mice due to loss of Atm expression and increased oxidative stress. Morita et al. (20) reported that ROS stimulate chondrocyte hypertrophy in growth plate cartilage, which can be prevented in vitro and in vivo by antioxidant treatment. Increased ROS in growth plate cartilage of Atm-deficient mice and antioxidant rescue of subsequent increased hypertrophy were also reported (20).

Antioxidants such as NAC have been successfully used to prevent chondrocyte death in explants after excessive loading stress (54) or in a rat model of osteoarthritis (58). NAC has also inhibited hypertrophic differentiation of articular chondrocytes by oxidized low-density lipoprotein (LDL) (59). These observations are in line with our NAC rescue experiments in the Anp32a−/− mice. Repeated overloading of cartilage and severe cartilage injury are associated with oxidative stress and mitochondrial dysfunction (60, 61). NAC treatment seems to be effective in providing protection for the cartilage in these clinically and translationally relevant settings.

Our results further demonstrate that not only the ANP32A-ATM axis is present in cartilage but also ANP32A regulates Atm expression and oxidative stress in cerebellum and bone. Lack of Anp32a resulted in clinical signs of ataxia, and these abnormalities were prevented by antioxidant treatment in young mice. We also documented a positive effect of antioxidant pharmacological intervention in older mice. Different mouse models of Atm deficiency have been developed, with the neurological phenotype apparently dependent on genetic background and on the mutation strategies used (26, 27). A specific Atm-mutant strain showed motor learning deficits and histological changes in the cerebellum that recapitulate some of the abnormalities seen in ataxia-telangiectasia patients at early disease stages (62). Barlow et al. (63) demonstrated that loss of Atm in mice resulted in oxidative damage in different tissues including the cerebellum, an observation that was confirmed by Kamsler et al. (64). Antioxidant treatments were shown to prevent some of these changes in respective genetic mouse models of Atm deficiency (65, 66).

The phenotype of the Anp32a KO mice reported here may seem to be in contrast with the observations published in a previous study in which phenotypic screening of this mouse strain did not show specific abnormalities in the nervous system (67). No ataxia was detected in the Anp32a-deficient mouse strain (67), although behavioral analysis to assess cerebellar function was only done by rotarod tests, a method that is likely less sensitive to detect abnormalities than the CatWalk system that we used (24). Moreover, the Anp32a-deficient mice used in our study were fully backcrossed to the C57Bl/6 background, which might not have been the case for mice reported earlier (67).

Oxidative stress is known to be a risk factor for the development and progression of osteoporosis (68). Lack of Anp32a resulted in bone loss in line with a similar phenotype observed in Atm-deficient mice, attributed mostly to defective osteoblast differentiation and lack of bone anabolism (31, 32). Our results suggest that potentiation of the ANP32A-ATM axis in bone could enhance the antioxidant capacities of the cells. This could have implications for conditions such as perimenopause, where hormonal changes are associated with increased oxidative stress (69, 70).

Some limitations apply to our study and its results. First, in the context of a translational study requiring large amounts of primary human cells from healthy controls, access to young, undamaged human cartilage is extremely difficult. Our control samples were obtained from elderly individuals suffering an osteoporotic or pathological fracture for whom hip prosthesis surgery was required. Thus, despite macroscopically appearing healthy, the tissue and cells have aged and may be different from cartilage in young individuals. Nevertheless, gene expression and protein analyses differentiated these samples from those of patients of a similar age group but requiring prosthesis surgery for osteoarthritis of the hip. Second, animal models of osteoarthritis, in particular in rodents, mimic specific aspects of the disease dependent on the trigger and are therefore not identical to the human disorder. For further development of therapeutic approaches, large-animal models in which the cartilage thickness and architecture are more closely related to that of humans will likely be required.

Third, in vivo experiments were performed with mice with a global deletion of the Anp32a gene. Although unlikely, it may be possible that the development of osteoarthritis in the mutant mice is not primarily caused by the absence of ANP32A in the cartilage. Impairments resulting from ataxia in the mutants may lead to different joint loading that could induce osteoarthritis features. However, silencing of ANP32A in non-osteoarthritic human articular chondrocytes resulted in decreased expression of healthy cartilage markers type II collagen and aggrecan, supporting the existence of a cell-autonomous effect in the development of osteoarthritis. Further analysis of the gait experiments suggested that, although intensity of weight support seemed to decrease in the mutant mice in particular in the front paws at an early age, stance duration consistently and significantly increased in both front and hind limbs (table S4). Consequently, the impulse, accounting for loading magnitude and duration, may be increased, and therefore, overall loading would be comparable if not higher in Anp32a−/− mice compared to wild-type mice. Stance duration was decreased by NAC treatment. Thus, an additional effect of ataxia on the development of osteoarthritis in the mice cannot be entirely excluded. However, in the clinical literature, there appears no evidence of a relationship between primary ataxia and the development of osteoarthritis.

To minimize mouse handling, NAC was given by the drinking water. Drinking ad libitum may not allow a fully controlled individual application but did rescue the effects of ANP32A loss. NAC rescue was demonstrated for the osteoarthritis and ataxia phenotype but could not be convincingly shown to affect osteoporosis in Anp32a−/− mice. This may be related to high variability in bone parameter measurements in older mice, the relative lack of statistical power due to the limited number of mice, or inadequate dosing or tissue penetrance of NAC.

Although we demonstrated the important role of the ANP32A-ATM axis in protecting joint, brain, and bone from oxidative stress, our experimental approaches are not likely to address the full complexity of the regulation of the redox status in cells and tissues, in particular the enzymatic activities in the system. Our microarray data suggested the presence of some compensatory mechanisms, in particular in the regulation of glutathione, the major cellular antioxidant system. Yet, this up-regulation and putative posttranslational mechanisms that affect the activity of the different redox enzymes did not restore the redox balance altered by the loss of ANP32A.

Our study identifies the ANP32A-ATM axis as a critical regulator of oxidative stress in cartilage, bone, and brain, thereby suggesting ANP32A as a therapeutic target for diseases associated with oxidative stress. Further research should focus on the factors that regulate ANP32A expression and activity in the articular cartilage and other tissues to define additional therapeutic approaches to prevent cartilage damage and osteoarthritis, bone loss and osteoporosis, and cerebellar loss of function and ataxia.


Study design

The objective of this study was to determine the role of ANP32A in the pathophysiology of osteoarthritis in humans and mice. Cartilage tissue and cells from patients with osteoarthritis and non-osteoarthritic controls and genetically engineered mice were used in ex vivo studies in different models of aging and joint disease, combined with in vitro assays. Mouse models are reported following the principles of the ARRIVE guidelines ( (see table S5). For all experiments, the largest possible sample size was used, with sample sizes and disease end points selected on the basis of previous studies (71). No exclusion of animals was carried out with the exception of one animal in the bone analysis due to a technical issue (see table S5). Mice were randomly assigned to the experimental groups where applicable. Pathology analysis was performed in a blinded fashion.

Statistical analysis

Data are presented as mean and SD or as individual data points, representing the mean of technical replicates as indicated in the figure legends. Statistical analyses were performed where appropriate with R Studio (version 1.0.15) and GraphPad Prism software. A detailed overview of the statistical analyses and assumptions is provided in data file S1. Data distribution was evaluated on the basis of parameter characteristics, quantile-quantile (QQ) plots, and Shapiro-Wilk normality tests. Variances were compared using the Levene test. T tests or ANOVA tests were applied, taking into account equal or different variances (applying Welch corrections). When different groups were compared by ANOVA tests, pair-wise t tests were subsequently performed applying a Bonferroni correction for multiple comparisons to control for type I errors in rejecting the null hypothesis. Data file S1 also reports estimates of differences of means between groups (95% confidence intervals).


Materials and Methods

Fig. S1. Loss of ANP32A increases the severity of osteoarthritis in the collagenase- and papain-induced mouse models.

Fig. S2. Expression of molecular markers associated with healthy chondrocytes in the presence or absence of ANP32A.

Fig. S3. Transcriptome network analysis of articular cartilage of Anp32a-deficient mice.

Fig. S4. Compensatory regulation of antioxidant systems in the articular cartilage of Anp32a-deficient mice.

Fig. S5. Immunohistochemistry of COLX in the growth plates of 16-week-old male Anp32a−/− compared to WT mice.

Fig. S6. Calbindin immunostaining of cerebellar Purkinje cells of 16-week-old male WT and Anp32a-deficient mice.

Fig. S7. Late-stage antioxidant intervention in Anp32a-deficient mice ameliorates ataxia-related defects.

Fig. S8. Model for the role of ANP32A on oxidative stress.

Table S1. Patient characteristics.

Table S2. Top ranked genes of transcriptome network of articular cartilage of 8-week-old Anp32a-deficient male mice compared to WT mice.

Table S3. Gene expression of main antioxidants in transcriptome network of articular cartilage of 8-week-old male Anp32a-deficient mice compared to WT mice.

Table S4. Gait parameters of early-stage antioxidant intervention with NAC in Anp32a-deficient mice.

Table S5. Animal experiments: Overview, setup, and analysis details.

Table S6. Primers used in qPCR analysis.

Data file S1. Statistical analysis.

References (7278)


Acknowledgments: We are grateful to A. Hens for animal facility management. We also thank N. Dirckx for assistance with digital image analysis. We are indebted to the traumatology and orthopedic surgeons willing to contribute samples (A. Sermon, J. P. Simon, and S. Nys) and to the nursing staff (in particular M. Penninckx). We thank all study participants of the Research Arthritis and Articular Cartilage (RAAK) study. Funding: This work was supported by the Flanders Research Foundation (FWO-Vlaanderen), the Interuniversity Attraction Poles (IUAP) network Development and Repair (Belspo. grant number: IUAP-VII/07), the FP7 project Translational Research in Europe Applied Technologies for Osteoarthritis (Treat-OA, European Commission framework 7 programme grant 200800), the Dutch Arthritis Association (DAA_10_1-402), the Dutch Scientific Research council NWO/ZonMW VICI scheme (nr. 91816631/528), the Leiden University Medical Center, and a Marie-Curie Intra-European postdoctoral fellowships to S.M. Author contributions: R.J.L., F.M.F.C., and S.M. planned the study and designed all the in vitro, ex vivo, and in vivo experiments with the exception of the data from the RAAK study that was planned and designed by I.M. and R.G.H.H.N. F.M.F.C., L.-A.K.A.G., L.S., and T.P. performed the animal experiments. S.M. performed the in vitro experiments. W.d.H. performed the analysis of the RAAK study. R.J.L. is responsible for all the other statistical analyses. S.M., F.M.F.C., R.J.L., I.M., and I.J. wrote the manuscript. Competing interests: The authors declare that they have no competing financial interests. Leuven Research and Development, the technology transfer office of KU Leuven, has received consultancy and speaker fees and research grants on behalf of R.J.L. from Abbvie, Boehringer-Ingelheim, Celgene, Eli-Lilly, Galapagos, Janssen, MSD, Novartis, Pfizer, Samumed, and Union Chimique Belge. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials. Anp32a−/− mice were obtained from P. Opal (Northwestern) under a material transfer agreement. Microarray data have been deposited to the Gene Expression Omnibus ( under accession no. GSE108036.

Stay Connected to Science Translational Medicine

Navigate This Article