IKKβ-NF-κB signaling in adult chondrocytes promotes the onset of age-related osteoarthritis in mice

Abstract

Canonical nuclear factor κB (NF-κB) signaling mediated by homo- and heterodimers of the NF-κB subunits p65 (RELA) and p50 (NFKB1) is associated with age-related pathologies and with disease progression in post-traumatic models of osteoarthritis (OA). Here, we established that NF-κB signaling in articular chondrocytes increased with age, concomitant with the onset of spontaneous OA in wild-type mice. Chondrocyte-specific expression of a constitutively active form of Inhibitor of κB kinase β (IKKβ) in young adult mice accelerated the onset of the OA-like phenotype observed in aging wild-type mice, including degenerative changes in the articular cartilage, synovium, and menisci. Both in vitro and in vivo, chondrocytes expressing activated IKKβ had a proinflammatory secretory phenotype characterized by markers typically associated with the senescence-associated secretory phenotype (SASP). Expression of these factors was differentially regulated by p65, which contains a transactivation domain, and p50, which does not. Whereas the loss of p65 blocked the induction of genes encoding SASP factors in chondrogenic cells treated with Interleukin 1β (IL-1β) in vitro, the loss of p50 enhanced the IL-1β–induced expression of some SASP factors. The loss of p50 further exacerbated cartilage degeneration in mice with chondrocyte-specific IKKβ activation. Overall, our data reveal that IKKβ-mediated activation of p65 can promote OA onset and that p50 may limit cartilage degeneration in settings of joint inflammation including advanced age.

Introduction

Osteoarthritis (OA) is a debilitating degenerative joint disease and leading cause of chronic disability worldwide (1). Knee OA, in particular, accounts for much of the global OA burden and is characterized by articular cartilage loss as well as changes to the synovium, meniscus, and subchondral bone (2), resulting in pain and loss of mobility that is managed mostly through palliative care. Although advanced age is recognized as one of the most important risk factors for OA development (3), the mechanisms responsible for OA onset and progression during the aging process are unclear and understudied.

Nuclear factor κB (NF-κB) was previously identified as the transcription factor for which activation is most associated with aging across multiple tissue types in both mice and humans (4). Additionally, aberrant NF-κB activation is documented in several age-associated pathologies (5). One mechanism by which NF-κB promotes age-related tissue degeneration and chronic inflammation is through its role as a master regulator of the senescence-associated secretory phenotype (SASP), a phenotype of senescent cells characterized by the secretion of proinflammatory cytokines and chemokines, growth factors, and matrix metalloproteinases (MMPs) (6, 7). Through the SASP, senescent cells can alter the proliferative status and viability of neighboring cells, remodel the ECM, and initiate immune responses (8). The SASP is suggested to be involved in OA pathogenesis because clearance of senescent cells can inhibit the development of OA in mice following joint injury, and intra-articular injection of senescent fibroblasts can induce an OA-like phenotype in uninjured murine joints (9, 10).

Inhibitor of κB kinase β (IKKβ, also known as IKK2) is a serine-threonine kinase that positively regulates canonical NF-κB signaling. IKKβ is part of the IκB kinase (IKK) complex binding to IKKα and IKKγ (also known as NF-κB essential modulator, NEMO). In unstimulated cells, NF-κB transcription factors are sequestered in the cytoplasm by inhibitory IκB proteins, but in cells stimulated by proinflammatory cytokines and chemokines or stress-regulated signals, the IKK complex phosphorylates IκB proteins, inducing their ubiquitin-mediated proteasomal degradation. This, in turn, allows the translocation of NF-κB transcription factors from the cytoplasm to the nucleus where they can bind DNA to positively or negatively regulate the transcription of target genes (reviewed in (11)). The two NF-κB family transcription factors involved in canonical signaling are p65 (also known as RELA) and p50 (also known as NFKB1), which function as homo- or heterodimers (12, 13). Both p65 and p50 are able to bind DNA through their Rel homology domains, but only p65 possesses a transcriptional activation domain. Therefore, whereas all dimer subtypes containing p65 activate transcription of target genes, p50 homodimers act as repressors of transcription to limit inflammatory responses (14). Consistently, global deletion of Nfkb1, the gene encoding p50, leads to heightened inflammatory responses following hepatic, renal, or tendon injury (15–17) and results in several accelerated aging phenotypes in uninjured mice due to increased chronic systemic inflammation (18–20).

Canonical NF-κB signaling is known to positively regulate the expression of transcripts encoding factors catabolic to the articular cartilage matrix, including Mmp13 (21, 22). It is also suggested to be responsible for cytokine-mediated suppression of Sox9, a gene whose protein product is essential for maintenance of the chondrogenic phenotype because it directs the transcription of genes encoding cartilage ECM components, such as Col2a1 (Collagen type II α 1) and Acan (aggrecan) (23, 24). Although inhibition of p65 in chondrocytes is reported to suppress cartilage degeneration in murine models of post-traumatic OA (PTOA) (25, 26), a direct relationship between activation of canonical NF-κB signaling in joint tissues and age-related joint degeneration has not been established. Given its relationship with the SASP, it seems likely NF-κB activation could be responsible for onset of joint degeneration during the aging process. Here, we rigorously characterized the aging knee joint phenotype of wild-type C57BL/6J mice and showed an increased number of chondrocytes with active canonical NF-κB signaling in aged articular cartilage. We also showed that activation of IKKβ-NF-κB signaling in chondrocytes accelerated the onset of age-related joint tissue degeneration and that p50 deficiency further exacerbated cartilage proteoglycan loss. Transcripts and proteins of several SASP-related factors were increased upon IKKβ activation in chondrocytes and, in vitro, we found that although loss of p65 suppressed the expression of these factors, p50 either increased or did not affect their expression. Collectively, our data point to a causal role for IKKβ-NF-κB activation in age-related degeneration of the joint likely due to chondrocyte secretion of proinflammatory, SASP-like factors that affect not only the cartilage, but all tissues within the synovial joint.

Results

Aged C57BL/6J mice have a knee joint phenotype with features of early-stage osteoarthritis

To better understand the changes associated with spontaneous age-related joint degeneration, we examined the knee joint phenotypes of male and female wild-type C57BL/6J mice at ages 3, 6, 15, and ≥24 months. Histology revealed that knee joints from male and female mice ≥24 months of age had decreased staining for Safranin O, which labels proteoglycans and glycosaminoglycans in cartilage, within regions of the unmineralized articular cartilage relative to younger mice (Fig. 1A; fig. S1A). Histology also revealed that with increasing age both male and female C57BL/6J mice developed synovial hyperplasia, a phenotype commonly observed in OA (27), as well as meniscal hypertrophy (Fig. 1A; fig. S1A). Although not typically included as a hallmark of OA, meniscal hypertrophy in humans is reported to be associated with OA (28–30).

Fig. 1. The knee joints of aged wild-type C57BL/6J mice have features of early-stage OA.

(A) Representative Safranin O (SAF O) and Fast Green (FG) staining of knee joint sections from male and female C57BL/6J mice at 3 and 27 months of age, as indicated. The images in the left column show low-magnification views of the entire joint; the higher magnification views show specific areas within the joint. Arrows point to loss of articular cartilage proteoglycan staining, meniscal chondrocytic metaplasia, and synovial hyperplasia. The images in the left column show low-magnification views of the entire joint; the higher magnification views show specific areas within the joint. Scale bars, 50μm. (B to D) Total (B), unmineralized (C), and mineralized (D) tibial articular cartilage areas of male and female C57BL/6J mice at 3 (N = 10 males; N = 9 females), 6 (N = 10 males; N = 9 females), 15 (N = 8 mice per group), and ≥24 (N = 11 males; N = 10 females) months of age, as indicated. (E and F) Safranin O–positive (SafO⁺) total (E) and unmineralized (F) tibial articular cartilage areas of male and female C57BL/6J mice at 3 (N = 5 mice per group), 6 (N = 5 mice per group), 15 (N = 3 mice per group), and ≥24 (N = 6 males; N = 5 females) months of age, as indicated. (G) Modified OARSI scores of male and female C57BL/6J mice 3 (N = 5 mice per group), 6 (N = 5 mice per group), 15 (N = 3 mice per group), and ≥24 (N = 6 males; N = 5 females) months of age, as indicated. All data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 for comparisons within the same sex; two-way ANOVA followed by Tukey’s multiple comparisons test. ^#p < 0.05, ^###p < 0.001 for comparisons between sexes of the same age; two-way ANOVA followed by Bonferroni’s multiple comparisons test.

Histomorphometry showed an increase in the total tibial cartilage area between 3- and 6-month-old male mice followed by a decrease in cartilage area in 15- and ≥24-month-old male mice (Fig. 1B). The unmineralized cartilage area (area above the tidemark, a basophilic line separating the unmineralized and mineralized cartilage) was also decreased in 15- and ≥24-month-old male mice relative to 6-month-old male mice (Fig. 1C). At 6 months of age, females had less total and unmineralized tibial articular cartilage compared to males (Fig. 1, B and C). No significant differences in the mineralized cartilage areas were observed at any age in males or females (Fig. 1D). Histomorphometry for Safranin O–positive (SafO⁺) cartilage area was performed on two separate sets of tissue sections from different animals. The results from the two sets were not directly compared to each other, because the slides were cut at different thicknesses (3μm versus 5μm) and stained at different times. For tissue sections cut at 5μm (the thickness used for all other studies reported here), total tibial articular cartilage SafO⁺ area was decreased in ≥24-month-old male mice relative to 6-month-old male mice (Fig. 1E). Unmineralized tibial articular cartilage SafO⁺ area was decreased in ≥24-month-old male mice relative to 3-, 6-, and 15-month-old male mice, and 6-month-old female mice had significantly decreased SafO⁺ unmineralized tibial articular cartilage area relative to 6-month-old male mice (Fig. 1F). The thinner (3μm) tissue sections revealed decreased total tibial articular cartilage SafO⁺ area by 15 months in the male mice and 24 months in female mice relative to male and female mice at 6 months, respectively (fig. S1B). Male mice also had decreased unmineralized tibial articular cartilage SafO⁺ area by 15 months relative to male mice 6 months of age (fig. S1C). These articular cartilage phenotypes culminated in a modest but significant increase in modified OARSI (Osteoarthritis Research Society International) scores in ≥24-month-old mice relative to younger mice (Fig. 1G; fig. S1D). Because OARSI scoring is, in part, dependent of Safranin O staining intensity, the two sets of slides were analyzed separately for this parameter.

Because empty chondrocyte lacunae and regions of acellularity were present in the articular cartilage of the aged mice (Fig. 1A), we performed TUNEL staining to examine chondrocyte death throughout the aging process. The percent of TUNEL⁺ tibial articular chondrocytes was increased in 15- and ≥24-month-old male and female mice relative to younger mice (Fig. 2, A and B; fig. S1E). By normalizing the data to total cartilage area, we found significant decreases in DAPI⁺ cell numbers throughout aging in both male and female mice, but little change in TUNEL⁺ cell numbers, suggestive of a low amount of continual articular chondrocyte apoptosis throughout the aging process (Fig. 2C; fig. S1F).

Fig. 2. Articular chondrocyte death occurs throughout aging in the knee joints of wild-type C57BL/6J mice.

(A) Representative TUNEL staining of knee joint sections from male C57BL/6J mice at 3 and 27 months of age, as indicated. White dashed lines outline articular cartilage and meniscal surfaces. Scale bar, 50μm. (B) Quantification of TUNEL⁺ cells as a percentage of total cells and (C) the number of DAPI⁺ cells/mm² in the tibial articular cartilage of male and female C57BL/6J mice at 3, 6, 15, and ≥24 months of age, as indicated (N = 6 mice per group). All data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; two-way ANOVA followed by Tukey’s multiple comparisons test.

Further histologic analysis of hindlimbs from C57BL/6J mice aged 21 months or older revealed that ~22% of male mice (5/23) developed a much more severe joint phenotype than that described above, with a near complete loss of tibial and femoral articular cartilage (fig. S2). These mice also had pronounced synovial hyperplasia and meniscal hypertrophy associated with clusters of SafO⁺ chondrocytes, consistent with chondroid metaplasia (fig. S2). This phenotype was not observed in any of the female mice. Collectively, these data indicate that male and female C57BL/6J mice develop early-stage OA as part of the normal aging process, with a subset of male mice developing more aggressive OA.

Age-related catabolic changes in articular cartilage ECM are accompanied by increased canonical NF-κB signaling

To further assess compositional changes in the articular cartilage ECM during aging, we performed COL2A1, COL10A1, and MMP13 immunohistochemistry (IHC). Whereas COL2A1 is a long-chain fibrillar collagen that is a normal component of healthy cartilage, COL10A1 is a short-chain hexagonal network–forming collagen associated with chondrocyte hypertrophy and cartilage degeneration. MMP13 degrades type II collagens and is associated with cartilage degeneration. COL2A1 increased in the ECM of the mineralized layer of articular cartilage and in the pericellular matrix of the unmineralized layer between 3 and 6 months of age, and decreased in both layers of the articular cartilage by 27 months of age (Fig. 3A). Consistent with this, SOX9 decreased with age (Fig. 3A). Conversely, COL10A1 increased in the articular cartilage ECM at 6 months of age and was evident in the pericellular matrix of chondrocytes in the unmineralized cartilage of 27-month-old mice (Fig. 3A). Similarly, MMP13 increased with age and was evident in chondrocytes in the unmineralized articular cartilage of 27-month-old mice (Fig. 3A).

Fig. 3. Knee joints from aged mice have enhanced cartilage catabolism and increased numbers of chondrocytes with active NF-κB signaling.

(A) Representative IHC for COL2A1, SOX9, COL10A1, and MMP13 in the articular cartilage of knee joint sections from 3-, 6-, 15-, and 27-month-old male C57BL/6J mice (N = 3–7 mice per group). Black arrows point to COL10A1 and MMP13 pericellular staining. Scale bars, 50μm. (B) Representative IHC for IκBα in the articular cartilage of knee joint sections from 3-, 6-, 15-, and 27-month-old male C57BL/6J mice (N = 4–5 mice per group). Scale bars, 50μm. (C) GFP fluorescence (shown with DAPI counterstain) on knee joint sections from 3- and 18-month-old NF-κB–GFP-luciferase (NGL) reporter mice, as indicated. White dashed lines outline the articular cartilage and meniscal surfaces; white arrows point to GFP⁺ cells. Scale bars, 50μm. (D and E) Quantification of GFP⁺ cells in the articular cartilage (D) and meniscus (E) of knee joints from NGL mice (N = 6 joints, 1 male and 5 female for 3-month-old group; N = 12 joints; 2 male and 10 female for 18-month-old group). Data are shown as mean ± SEM. ***p < 0.001, ****p < 0.0001; two-tailed unpaired Student’s t-test.

Based on the established role of canonical NF-κB signaling in various age-related pathologies and its involvement in the regulation of Sox9 and Mmp13 expression (21–24), we examined NF-κB signaling in the articular chondrocytes of aged mice. IHC revealed that Inhibitor of κB α (IκBα), an endogenous inhibitor of canonical NF-κB signaling, decreased in the articular chondrocytes throughout the aging process (Fig. 3B). Because NF-κB–mediated transcription is regulated at multiple levels, however, loss of IκBα does not necessarily indicate increased NF-κB transcriptional activity. To further investigate this, we utilized the previously described NF-κB–GFP-luciferase (NGL) transgenic reporter mouse (31) in which cells with active canonical NF-κB–dependent transcription can be identified through GFP or luciferase expression. Whereas histology revealed few GFP-positive (GFP⁺) cells in knee joint tissues of 3-month-old NGL mice, a modest number of GFP⁺ cells were found throughout the knee joint in 18-month-old NGL mice (Fig. 3C). Quantification confirmed a significant increase in GFP⁺ cells present in the tibial and femoral articular cartilage as well as in the menisci of aged NGL mice relative to young NGL mice (Fig. 3, D and E). These cells were predominantly found near the tidemark of the articular cartilage.

Chondrocyte-specific IKKβ activation accelerates the onset of the age-related, OA-like joint phenotype

Given that knee joints from aged wild-type mice exhibited changes consistent with early-stage OA and that articular cartilage in aged mice had an increased number of chondrocytes with active NF-κB signaling, we next sought to determine if activation of canonical NF-κB signaling in chondrocytes was sufficient to promote age-related changes and/or OA onset in the knee joints of young mice. To do so, we used the cartilage-specific Acan^CreERT2 knock-in allele in combination with the ROSA26^Ikk2ca (R26^Ikk2ca) allele to drive expression of constitutively active IKKβ (32, 33). We induced Cre-mediated recombination with tamoxifen at 2 months of age, a time point largely following the rapid skeletal development that occurs within the first two months of age, allowing for the investigation of the effects of IKKβ gain-of-function (GOF) on tissue homeostasis rather than development. R26^{Ikk2ca/Ikk2ca} mice (control) and Acan^CreERT2/+; R26^{Ikk2ca/Ikk2ca} mice (IKKβ GOF) were harvested 2 weeks after tamoxifen induction to confirm efficient expression of the Ikk2ca knock-in allele. Cells expressing the knock-in allele were found in the superficial and middle zones of the articular cartilage, as well as in the superficial region of the meniscus, but not in cells in the lateral regions of the menisci or within the synovium (fig. S3A). The articular chondrocytes from the IKKβ GOF mice also had increased phosphorylated (active) p65 relative to chondrocytes from control mice, confirming canonical NF-κB activation (fig. S3B).

To assess the long-term effects of sustained IKKβ activation in chondrocytes, 2-month-old control and IKKβ GOF mice were administered tamoxifen and sacrificed 6 months later. Histology and histomorphometry of the knee joints revealed decreased Safranin O staining within the unmineralized articular cartilage of both male and female IKKβ GOF mice relative to control mice (Fig. 4, A and B), resulting in significantly increased modified OARSI scores for the IKKβ GOF groups (Fig. 4C). Although IKKβ GOF resulted in decreased unmineralized articular cartilage area in male mice, it did not result in loss of overall articular cartilage area in either male or female mice (Fig. 4, D and E). Female control mice had significantly decreased total articular cartilage area relative to male control mice, however, consistent with results from the 6-month-old wild-type C57BL/6J mice (Figs. 4E and 1B). Finally, histology revealed that the IKKβ GOF mice developed hypertrophic menisci and hyperplastic synovium relative to control mice (Fig. 4A). Overall, the knee joint phenotype of the IKKβ GOF 8-month-old mice resembled the spontaneous early-stage OA phenotype seen in aged wild-type C57BL/6J mice (Fig. 1A).

Fig. 4. Chondrocyte-specific IKKβ activation accelerates the onset of age-related early-stage OA phenotypes.

(A) Representative Safranin O (SAF O) and Fast Green (FG) staining of knee joint sections from R26^{Ikk2ca/Ikk2ca} (Control) and Acan^CreERT2/+; R26^{Ikk2ca/Ikk2ca} (IKKβ GOF) at 8 months of age (6 months following tamoxifen administration). The images in the left column show low-magnification views of the entire joint; the higher magnification views show specific areas within the joint. Arrows point to loss of articular cartilage proteoglycan staining, meniscal chondrocytic metaplasia, and synovial hyperplasia. Scale bars, 50μm. (B to E) Safranin O–positive (SafO⁺) unmineralized tibial articular cartilage area (B), modified OARSI scores (C), unmineralized tibial articular cartilage area (D), and total tibial articular cartilage area (E) of male and female control and IKKβ GOF mice, as indicated (N = 4 mice for male control group, N = 3 mice for male IKKβ GOF group, N = 5 mice for female control group, N = 6 mice for female IKKβ GOF group). (F) Representative TUNEL staining of knee joint sections from control and IKKβ GOF mice at 3 and 8 months of age, as indicated. White dashed lines outline articular cartilage and meniscal surfaces. Scale bars, 50μm. (G and H) Quantification of TUNEL⁺ cells as a percentage of total cells (G) and DAPI⁺ cell numbers/mm² (H) in the tibial articular cartilage of control and IKKβ GOF mice at 3 months of age (N = 3 males and 3 females for control group; N = 3 males and 4 females for IKKβ GOF group) and 8 months of age (N = 3 males and 3 females for control group; N = 3 males and 3 females for IKKβ GOF group), as indicated. All data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 for comparisons within the same sex (B, C, D) or age (G, H); ^#p < 0.05, ^###p < 0.001, ^####p < 0.0001 for comparisons between sexes (B, E) or ages (G, H) of the same genotype; two-way ANOVA followed by Tukey’s multiple comparisons test.

Empty chondrocyte lacunae were also observed in the articular cartilage of the IKKβ GOF mice (Fig. 4A). TUNEL staining showed an increase in the percentage TUNEL⁺ cells in the articular cartilage of 8-month-old (6 months post-induction) IKKβ GOF mice compared to that of control mice and a significant decrease in DAPI⁺ cells (Fig. 4, F to H). TUNEL staining on knee joint sections from 3-month-old (1 month post-induction) mice showed an increase in the total number and proportion of TUNEL⁺ chondrocytes in the IKKβ GOF cartilage relative to control cartilage, but no change in total DAPI⁺ cells (Fig. 4, F to H; fig. S4A). Of note, the total number and proportion of TUNEL⁺ chondrocytes in the 3-month-old IKKβ GOF cartilage were similar to those of the 8-month-old control cartilage, suggesting that IKKβ GOF may accelerate the chondrocyte apoptosis that gradually occurs throughout the aging process (Fig. 4G; fig. S4A). With regard to the collagen composition of the cartilage ECM, we did not find an obvious difference in COL2A1, however, we did find a significant increase in COL10A1 in the IKKβ GOF cartilage (fig. S4, B and C).

To further examine the hypertrophy and apparent enhanced mineralization of the menisci in both aged and IKKβ GOF joints, we performed micro-computed tomography (micro-CT) analyses. Representative micro-CT images showed that mineralization of the menisci appeared to increase throughout the aging process in both male and female wild-type C57BL/6J mice (Fig. 5A; fig. S5). Subsequent quantitative analyses confirmed significantly increased mineralized meniscus volumes in the anterior and posterior regions of both the medial and lateral menisci of 27-month-old male and female mice relative to 3-month-old male and female mice, respectively (Fig. 5B). No differences in mineralized meniscal volumes were found between the 3- and 6-month-old mice. However, there are significantly increased mineralized volumes in the 15-month-old mice compared to the 3- and 6-month-old mice in most meniscal compartments. The only significant differences found between males and females were in the lateral anterior and medial posterior meniscal compartments where females had increased and decreased mineralization, respectively, compared to males at 27 months of age. The mineralized structure of the IKKβ GOF knee joint appears similar to that of the 27-month-old wild-type C57BL/6J joint (Fig. 5C). Accordingly, significant increases in mineralized volumes of all meniscal compartments were found in the IKKβ GOF mice relative to control mice (Fig. 5D).

Fig. 5. Aging or chondrocyte-specific IKKβ GOF promotes mineralization of the menisci and synovial hyperplasia.

(A) Representative micro-CT images from 3-, 6-, 15-, and 27-month-old male C57BL/6J mice. Yellow arrows point to mineralized meniscal compartments. Scale bars, 500μm. (B) Mineralized volumes of the lateral anterior, lateral posterior, medial anterior, and medial posterior meniscal compartments quantified from the micro-CT scans of 3-, 6-, 15-, and 27-month-old male and female C57BL/6J mice as indicated (N = 5 mice per group). Data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 for comparisons within the same sex; two-way ANOVA followed by Tukey’s multiple comparisons test. ^#p < 0.05, ^##p < 0.01 for comparisons between sexes at the same age; two-way ANOVA followed by Bonferroni’s multiple comparisons test. (C) Representative micro-CT images from control and IKKβ GOF mice at 8 months of age (6 months following tamoxifen administration). Yellow arrows point to mineralized meniscal compartments. Scale bars, 500μm. (D) Mineralized volumes of the lateral anterior, lateral posterior, medial anterior, and medial posterior meniscal compartments quantified from the micro-CT scans of 8-month-old control and IKKβ GOF mice, as indicated (N = 3 males and 3 females for control group; N = 2 males and 5 females for IKKβ GOF group). Data are shown as mean ± SEM. **p < 0.01, ***p < 0.001, ****p < 0.0001; two-tailed unpaired Student’s t-test. (E) Representative IHC and quantification of Ki-67⁺ cells in the entire synovium of representative tissue sections from 8-month-old (6 months following tamoxifen administration) control and IKKβ GOF mice (N = 3 males and 5 females for control group; N = 3 males and 7 females for IKKβ GOF group). Arrows point to Ki-67⁺ cells. Scale bar, 50μm. (F) Representative IHC and quantification of CD45⁺ area as a percentage of total area in the synovium of 8-month-old control and IKKβ GOF mice (N = 3 males and 5 females for control group; N = 2 males and 7 females for IKKβ GOF group). Scale bar, 50μm Data are shown as mean ± SEM. *p < 0.05; two-tailed unpaired Student’s t-test. p = 0.0532 for CD45⁺ area (G).

To investigate the synovial hyperplasia observed in the IKKβ GOF knee joints, we assessed proliferation and immune cell infiltration by performing IHC for Ki-67 and the pan-leukocyte marker CD45, respectively. Relative to control synovium, IKKβ GOF synovium had a significant increase in the number of cells positive for Ki-67 (Fig. 5E). Several, but not all, of the IKKβ GOF mice also had increased CD45-positive (CD45⁺) staining in the synovium (Fig. 5F).

IKKβ activation in chondrocytes leads to a proinflammatory secretory phenotype

Because the chondrocyte-specific IKKβ GOF mice exhibited age-related phenotypes in joint tissues not directly targeted in our genetic model (for example, hyperplastic synovium with CD45⁺ cells) and because canonical NF-κB signaling is associated with the generation of proinflammatory secretory phenotypes, such as the SASP, we examined the factors secreted from IKKβ GOF chondrocytes. Primary sternal chondrocytes from R26^{Ikk2ca/Ikk2ca} neonatal mice were infected with control adenovirus encoding GFP or adenovirus encoding Cre recombinase. Western blotting confirmed efficient Cre-mediated recombination and expression of constitutively active IKKβ as well as the endogenous chondrogenic transcription factor SOX9 (Fig. 6A, fig. S6A). The Cre-infected cells also had decreased IκBα, confirming the activation of canonical NF-κB signaling, as well as increased COX-2, relative to GFP-infected cells (Fig. 6A and fig. S6A). COX-2 is encoded by the direct NF-κB transcriptional target Ptgs2 and promotes the synthesis of prostaglandin E₂ (PGE₂), a factor associated with the SASP.

Fig. 6. IKKβ GOF in chondrocytes results in increased expression and secretion of proinflammatory cytokines, chemokines, and MMPs.

(A) Representative Western blotting for the indicated proteins in lysates from sternal chondrocytes isolated from R26^{Ikk2ca/Ikk2ca} mice, infected with adenovirus encoding GFP or Cre (MOI 100), and harvested two days later. β-actin is a loading control. (B) Imaging and quantification of staining intensity in spots for the indicated proteins from an array of antibodies to mouse cytokines and chemokines. The antibodies were spotted in duplicate and incubated with conditioned media from R26^{Ikk2ca/Ikk2ca} sternal chondrocytes infected with adenovirus encoding GFP or Cre (MOI 100) two days before harvesting. Quantification data represent relative pixel intensities and are presented as mean ± SD. (C) Relative expression of the indicated genes as determined by RT-qPCR with mRNA isolated from R26^{Ikk2ca/Ikk2ca} sternal chondrocytes infected with adenovirus encoding GFP or Cre (MOI 100) and harvested three days later (N = 3 biological replicates per group). Data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ****p < 0.0001; two-tailed unpaired Student’s t-test. (D) Representative IHC for CCL20 and MMP13 and quantification of CCL20⁺ and MMP13⁺ area as a percentage of total area within the tibial articular cartilage of 3-month-old (1 month following tamoxifen administration) control and IKKβ GOF mice (N = 1 male and 2 females for control group; N = 1 male and 2 females for IKKβ GOF group). Yellow and red boxes are high magnification images of the indicated regions within the adjacent images taken at lower magnification. Scale bars, 50μm. Quantification Data are shown as mean ± SEM. *p < 0.05; two-tailed unpaired Student’s t-test.

Conditioned media from GFP- or Cre-expressing cells were incubated with a commercial membrane array spotted with antibodies against various mouse cytokines and chemokines. Among those tested, CCL20, MMP3, and IL-6 were the three most abundant molecules present in the Cre-conditioned media relative to media from GFP-expressing cells (Fig. 6B and fig. S6B). The collective secretion of these molecules has previously been associated with the development of a SASP (34–36). Gene expression analyses confirmed increased expression of Ccl20, Mmp3, and Il6 and also showed an increase in other genes commonly associated with the SASP, such as Cxcl1, Il1a, Mmp13, and Ptgs2 (Fig. 6C). Activation of canonical NF-κB signaling and an increase in SASP gene expression was also seen in Ikk2ca-expressing chondrogenic ATDC5 cells (fig. S6, C and D).

To determine whether IKKβ GOF articular chondrocytes also expressed these SASP-associated factors in vivo, we examined the articular cartilage of knee joints from control and IKKβ GOF mice one month after tamoxifen induction, a timepoint that precedes significant chondrocyte loss (Fig. 4, F and G). At this timepoint, the articular chondrocytes from IKKβ GOF mice had decreased IκBα as well as increased phosphorylated p65 relative to chondrocytes from control mice, confirming canonical NF-κB activation in vivo (fig. S7A). CCL20 appeared increased in the femoral and tibial articular cartilages of the IKKβ GOF knee joints and was also present in the superficial regions of the menisci (Fig. 6D). MMP13 was increased in the IKKβ GOF articular cartilage, though it was still largely localized to the mineralized cartilage regions (Fig. 6D). Histomorphometric quantification of both factors in the tibial articular cartilage supported that they were increased due to IKKβ GOF (Fig. 6D). To examine the expression of additional SASP factors, we performed RT-qPCR on mRNA isolated from articular cartilage pooled from control or IKKβ GOF knee joints. Increases in Ccl20, Mmp13, Il6, Il1a, Mmp3, and Cxcl1 expression were detected in the cartilage from IKKβ GOF animals (fig. S7B). Given the similarities in knee joint phenotypes between aged wild-type C57BL/6J and young IKKβ GOF mice, we also examined the gene expression of SASP factors in young and aged wild-type C57BL/6J articular cartilage, finding increases in Il6 and Il1a (fig. S7C). Finally, mRNA isolated from the femoral heads of wild-type C57BL/6J mice revealed a reduction in Col2a1 and increases in Ccl20, Mmp3, Il6, Cxcl1, and Mmp13 gene expression in tissues from aged animals compared to tissues from young animals (fig. S7D). Given that the femoral head tissues contain subchondral bone and aged bone has been reported to contain senescent cells (37), it is not possible to determine whether the increases in SASP factor expression originate in the aged chondrocytes or senescent osteoblasts and osteocytes. These data further support the increased expression of these factors in the aged joint environment, however.

Loss of Rela or Nfkb1 differentially affects IL-1β–induced expression of genes encoding cytokines, chemokines, and MMPs

IKKβ directly regulates the canonical arm of the NF-κB pathway, leading to transcription of NF-κB target genes through the RELA (p65) and NFKB1 (p50) transcription factors. To explore the roles of p65 and p50 in regulation of proinflammatory genes in chondrocytes, we transfected ATDC5 chondrogenic cells with siRNA targeting Rela or Nfkb1 and subsequently treated the cells with or without IL-1β, a direct effector of IKKβ activation. Pathway activation and p65 and p50 knockdown were confirmed by Western blotting (Fig. 7, A and B). Knockdown of p65 significantly reduced the magnitude by which IL-1β induced multiple genes encoding SASP factors, bringing their expression down to near baseline amounts and suggesting that p65 is essential for IL-1β–induced expression of these NF-κB target genes (Fig. 7C). In contrast, knockdown of p50 further increased IL-1β–induced expression of Ccl20 and Mmp13 (Fig. 7D), suggesting that p50 may act as a transcriptional repressor of these genes in chondrocytes. Significant increases or decreases were not observed for other SASP factor genes examined. Whether this was because p50 does not function as a repressor of these genes in chondrocytes or because the experimental conditions allowed p65 activation to overcome p50 repression is unclear.

Fig. 7. Loss of Rela or Nfkb1 differentially affects proinflammatory cytokine, chemokine, and MMP gene expression in response to IL-1β.

(A and B) Representative Western blot images and quantification of the indicated proteins from ATDC5 cells transfected with non-targeting control (NTC) siRNA or siRNA targeting Rela (p65) (A) or Nfkb1 (p50) (B) before treatment with vehicle or IL-1β. Quantification is represented as a ratio of the indicated protein relative to β-actin (N = 3 independent samples). (C and D) Relative expression of the indicated genes as determined by RT-qPCR with mRNA isolated from ATDC5 cells transfected with non-targeting control (NTC) siRNA or siRNA targeting Rela (p65) (C) or Nfkb1 (p50) (D) then treated with vehicle or IL-1β (n = 3 independent samples). Data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; one-way ANOVA followed by Tukey’s multiple comparisons test.

Global Nfkb1 deletion alone does not affect joint homeostasis but exacerbates age-related OA in IKKβ gain-of-function mice

Previous studies report that Nfkb1 deletion in mice accelerates the onset of aging phenotypes in several tissue systems due to an increase in proinflammatory factors in the tissue microenvironment (18–20). Further, it was shown that p50 exhibits decreased DNA binding activity in tissues from aged wild-type mice versus young wild-type mice (18). To determine if p50 loss-of-function is sufficient to accelerate the onset of age-related OA, we assessed the knee joint phenotypes of male and female Nfkb1^−/− (p50^−/−) mice at 12 months of age. Histology and histomorphometry revealed that the Nfkb1^−/− mice had no discernible changes in joint phenotype relative to wild-type or Nfkb1^+/− littermates (Fig. 8, A to E).

Fig. 8. Nfkb1 deletion alone does not affect the knee joint phenotypes of middle-aged mice.

(A) Representative Safranin O (SAF O) and Fast Green (FG) staining of knee joint sections from wild-type (WT), Nfkb^+/− (p50^+/−), and Nfkb^−/− (p50^−/−) mice at 12 months of age. The images in the left column show low-magnification views of the entire joint; the higher magnification views show specific areas within the joint. Scale bars, 50μm. (B to E) Total (B), Safranin O–positive (SafO⁺) total (C), unmineralized (D), and SafO⁺ unmineralized (E) tibial articular cartilage areas in knee joints from male and female WT (N = 4 males; N = 4 females), p50^+/− (N = 5 males; N = 5 females), and p50^−/− (N = 3 males; N = 4 females) mice at 12 months of age, as indicated. All data are shown as mean ± SEM. ^#p < 0.05 for comparisons between sexes of the same genotype; two-way ANOVA followed by Bonferroni’s multiple comparisons test. Two-way ANOVA followed by Tukey’s multiple comparisons test was performed for comparisons within the same sex.

Given that knockdown of p50 in ATDC5 cells did not induce the expression of proinflammatory genes in the absence of IL-1β treatment (Fig. 7D), we hypothesized that only in the context of an NF-κB stimulus would an effect due to loss of p50 be observed. Thus, we generated IKKβ GOF;Nfkb1^+/− and IKKβ GOF;Nfkb1^−/− double-mutant mice. Tamoxifen was administered at 2 months of age and knee joints harvested 6 months later. The IKKβ GOF knee joints had decreased articular cartilage proteoglycan content as well as hyperplasia of the synovium and meniscal hypertrophy as described above (Fig. 9A and Fig. 4A). The IKKβ GOF;Nfkb1^+/− and IKKβ GOF;Nfkb1^−/− double-mutant knee joints had these phenotypes as well, but with a more severe articular cartilage phenotype than that seen in the IKKβ GOF mice (Fig. 9A). This included enhanced loss of proteoglycan staining that resulted in a significant decrease in both overall and unmineralized tibial articular cartilage SafO⁺ area in the IKKβ GOF;Nfkb1^−/− mice (Fig. 9, B and C). In the IKKβ GOF;Nfkb1^+/− mice, the SafO⁺ area was only decreased in the unmineralized tibial articular cartilage (Fig. 9, B and C). Both IKKβ GOF;Nfkb1^+/− and IKKβ GOF;Nfkb1^−/− mice had significantly decreased unmineralized tibial articular cartilage SafO⁺ area relative to the IKKβ GOF mice (Fig. 9C). Although histomorphometry did not reveal enhanced loss of unmineralized or overall articular cartilage area in the double-mutant mice relative to IKKβ GOF or wild-type mice (Fig. 9, D and E), respectively, two of the IKKβ GOF;Nfkb1^−/− female mice did have thinner articular cartilage as shown by histology (Fig. 9A, arrows). Increased sample numbers are needed to determine whether combined IKKβ GOF and p50 loss leads to overall articular cartilage thinning and whether this may be sex-specific. Collectively, this data set demonstrates that loss of p50 in vivo exacerbates the spontaneous age-related OA phenotype observed in IKKβ GOF mice.

Fig. 9. Nfkb1 deletion exacerbates the age-related OA knee joint phenotype in chondrocyte-specific IKKβ GOF mice.

(A) Representative Safranin O (SAF O) and Fast Green (FG) staining of knee joint sections from control (Cre-negative; Nfkb1^+/+, WT), Acan^CreERT2/+; R26^{Ikk2ca/Ikk2ca} (IKKβ GOF), Acan^CreERT2/+; R26^{Ikk2ca/Ikk2ca}; Nfkb1^+/− (IKKβ GOF/p50^+/−) and Acan^CreERT2/+; R26^{Ikk2ca/Ikk2ca}; Nfkb1^−/− (IKKβ GOF/p50^−/−) mice at 8 months of age (6 months following tamoxifen administration). The images in the left column show low-magnification views of the entire joint; the higher magnification views show specific areas within the joint. Arrows span from the articular cartilage surface to the chondro-osseous junction of the subchondral bone. Scale bars, 50μm. (B to E), Safranin O–positive (SafO⁺) total (B), SafO⁺ unmineralized (C), unmineralized (D), and total (E) tibial articular cartilage areas in WT, IKKβ GOF, IKKβ GOF/p50^+/−, and IKKβ GOF/p50^−/− mice, as indicated (N = 1 male and 2 females for WT group; N = 4 females for IKKβ GOF group; N = 2 males and 1 female for IKKβ GOF/p50^+/− group; N = 1 male and 2 females for IKKβ GOF/p50^−/− group). Data are shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; one-way ANOVA followed by Tukey’s multiple comparisons test.

Discussion

Although aging is a significant risk factor for OA onset, the effects of aging on knee joint tissues as well as the mechanisms driving knee OA pathogenesis during the aging process are relatively understudied. This is likely due in part to the time and cost involved in aging animals to the point at which they develop OA without the intervention of an injury. Previous studies have examined spontaneous age-related OA development in male C57BL/6 inbred lines or NIA (National Institute on Aging) colonies and have reported varied severity in the aged joint phenotype, ranging from full thickness articular cartilage defects to proteoglycan loss without significant loss of thickness (38–42). Here, we characterized the progressive effects of aging on the knee joint phenotypes of male and female C57BL/6J mice, finding an early-stage OA phenotype in mice 24 months of age and older. These age-related changes consisted of articular cartilage proteoglycan loss, hypertrophy and mineralization of the menisci, and synovial hyperplasia in addition to enhanced articular chondrocyte apoptosis and increased expression of cartilage catabolic factors. Histology revealed a subset (22%) of aged male C57BL/6J mice, not included in the histomorphometry or micro-CT analyses, developed a more severe OA phenotype with full-thickness articular cartilage loss (fig. S2). The occurrence of the two contrasting age-related phenotypes could explain why previous studies have reported differences in which phenotypes constitute murine age-related OA. Although the mice from this study are of the same genetic background and were housed in the same environment, they were not housed individually. It is established that male mice can demonstrate aggressive behaviors when group-housed (43, 44). Thus, mice involved in aggressive behaviors may have experienced injury to their knee joints resulting in the onset of PTOA.

In addition to changes within the articular cartilage, the joints from aged and IKKβ GOF mice had hypertrophic menisci with features of chondrocytic metaplasia and enhanced mineralization. Others have shown a similar phenotype in STR/ort mice, an inbred strain with early onset of spontaneous OA, and in mice subjected to joint injury (45, 46). The causes of these changes to the menisci are unclear, but are likely related to abnormal mechanical loading or inflammation. Although it is clear from the micro-CT of young mice that there is a baseline amount of mineralization within the murine menisci that is not seen in healthy human menisci, the hyperplasia and mineralized tissue volume increase that occurred in the menisci of aged or osteoarthritic murine joints may still be relevant to human OA pathogenesis. For example, micro-CT and MR (magnetic resonance) images of human menisci show that meniscal volume and height (or thickness) are increased in the knees of OA patients (28–30). Histologic examination shows that the human hypertrophied menisci have a disorganized collagen matrix with regions of cartilage tissue and chondrocytes present (28). COL10A1 and MMP13 as well as calcium deposition are also reported in human OA menisci, further suggesting that cartilaginous metaplasia may occur in the diseased tissue (47).

A limitation of our study includes the low sample numbers for many of our experiments. As mentioned in Results, for example, we were limited in the number of wild-type samples we could include per group for quantification of articular cartilage SafO⁺ staining and, thus, OARSI scoring, due to the tissues being sectioned at two different thicknesses (3μm versus 5μm) and stained at different times. Significant loss of SafO⁺ area was detected by 15 months in male joints sectioned at 3μm, but not until 24 months in male joints sectioned at 5μm, suggesting that differences in tissue thickness, not surprisingly, affect staining intensity. These results highlight the importance of normalizing all steps in tissue processing prior to performing quantitative histomorphometry. A second limitation of our study is that we did not analyze males and females separately for all experiments. It is now appreciated that OA pathogenesis in males and females may differ. To better understand why this is the case, all assays for molecular hallmarks of disease should be designed to analyze males and females separately. A third limitation of our study is that our experimental design did not allow us to conclude anything regarding the role of IKKβ-NF-κB signaling in OA progression to end-stage disease, only the initiation of OA onset. To study progression, we would need to wait longer than 6 months following tamoxifen induction to determine if joint degeneration ever progresses beyond that of an early-stage OA phenotype.

Canonical NF-κB signaling is suggested by a number of studies to play a role in OA pathogenesis. Most compelling are the loss-of-function studies showing that siRNA-mediated knockdown or heterozygous deletion of Rela can suppress the development of PTOA following surgical induction of joint injury in mice or rats (25, 26, 48). Also compelling is the use of a dual inhibitor of IKKα and IKKβ to suppress OA following knee joint injury in mice (49). Effects of this inhibitor cannot be attributed solely to its actions on IKKβ and canonical NF-κB signaling, however, given that IKKα is also required for non-canonical NF-κB signaling. Of note, IKKα has been shown to promote chondrocyte hypertrophy in vitro, and chondrocyte-specific deletion of IKKα using the Acan^CreERT2 allele in vivo protects mice from cartilage degradation following joint injury (50, 51). Deletion of IKKβ in the context of PTOA has not been reported. Several genetic mouse models resulting in modification of signaling pathways that induce canonical NF-κB activation in chondrocytes provide correlative evidence of NF-κB transcriptional activity and cartilage degeneration in the absence of injury but do not directly link NF-κB activation to OA onset (reviewed in (52)). Although one published study shows loss of proteoglycan staining upon IKKβ activation in chondrocytes (53), our study shows that this activation is sufficient to initiate the onset of an OA-like whole joint phenotype reminiscent of the joint degeneration that occurs with aging. Further studies involving IKKβ inhibition in an aging model will be required to determine if chondrocyte-specific IKKβ signaling is necessary for age-related joint degeneration.

In addition to loss of proteoglycan content within the cartilage ECM, our data showed decreased articular chondrocyte viability in aged and IKKβ GOF joints. Chondrocyte apoptosis is associated with OA, and NF-κB signaling is known to both promote and inhibit apoptosis. In mice, homozygous deletion of Rela in chondrocytes increases chondrocyte apoptosis and results in more severe spontaneous OA by 16 months likely due to decreased transcription of anti-apoptotic target genes (25). However, Fasl, the gene encoding the Fas ligand (FasL), is a direct target of NF-κB, and Fas, the gene encoding the Fas receptor (FasR), is expressed in articular chondrocytes (54). Engagement of FasL on the cell surface by FasR on neighboring cells induces apoptosis in the FasL⁺ cell. It is possible, therefore, that IKKβ GOF induces apoptosis by stimulating the production of FasL and, therefore, the engagement of FasL with FasR on chondrocytes. T cells expressing the R26^Ikk2ca allele show enhanced apoptosis through this mechanism (55). Independent of NF-κB, IKKβ can induce p53-mediated apoptosis through the phosphorylation of the p85 isoform of the kinase S6K1 under conditions of oxidative stress (56). Although the role of p85 S6K1 has not been investigated in chondrocytes, this mechanism is an attractive possibility for IKKβ-mediated articular chondrocyte apoptosis, given that aged and osteoarthritic chondrocytes experience oxidative stress (38, 57). Further, it has been demonstrated that pharmacological inhibition of IKKβ following joint injury promotes chondrocyte viability (58).

Cellular senescence and the SASP have received much attention with regard to their relationship to OA. Evidence of chondrocyte senescence in OA articular cartilage comes mostly from human tissues following chondrocyte isolation and in vitro culture(59). In mice, the elimination of cells in the joint that produce the cyclin-dependent kinase inhibitor p16^INK4a can slow OA development following injury, whereas chondrocyte-specific deletion of Cdkn2a, the gene encoding p16^INK4a, does not affect injury-induced OA development (9, 60). Collectively, these studies suggest that it is not replicative arrest of the chondrocyte that affects OA development, perhaps because they are a mostly quiescent cell population, but other aspects of senescence or other senescent cell types present in the injured joint. The SASP is a hallmark of senescent cells that can vary depending on the cell type and the senescence-inducing agent, but typically includes a core group of proinflammatory and matrix-degrading factors whose transcription is regulated by canonical NF-κB signaling (6). Our data showed that IKKβ GOF chondrocytes expressed some of these core SASP factors and, thus, through paracrine actions may be able to affect some of the changes observed in the chondrocyte-specific IKKβ GOF knee joints. In particular, synoviocyte proliferation and immune cell recruitment to the synovium are likely to be due to these factors because the cell types involved should not be targeted by the Acan^CreERT2 allele. Thus, although our study does not address chondrocyte senescence directly, it suggests that a SASP-like proinflammatory secretory phenotype in chondrocytes may affect changes to joint tissues beyond the cartilage even in the absence of injury.

p50 homodimers can repress the expression of proinflammatory factors in some tissues, resulting in accelerated aging phenotypes or unresolved inflammation following acute injury in Nfkb1^−/− mice (14–18, 20, 61). The best-defined mechanism whereby p50 homodimers are able to repress gene transcription is through interaction with Histone Deacetylase 1 (HDAC1) or B cell lymphoma 3 (Bcl-3) on κB DNA binding motifs, thereby preventing activating NF-κB dimers from binding to these sites (62). Although the p50-p65 heterodimer is the most abundant and widely-studied of the NF-κB activating dimers, p65 is able to form transcriptionally active homodimers as well as heterodimers with p52 and c-Rel (62). Thus, in the absence of p50, activation of gene transcription without repression by p50 homodimers can occur. To our knowledge, the repressive function of p50 has not been studied in chondrocytes, but our data suggest it might be important for cartilage homeostasis in response to an inflammatory stimulus. We found that Nfkb1 heterozygous or homozygous deletion alone did not accelerate age-related OA onset in mice by 12 months of age; however, in combination with IKKβ GOF, heterozygous or homozygous Nfkb1 deletion resulted in more severe cartilage degeneration than IKKβ GOF alone. Previous studies reveal that some tissues of middle-aged Nfkb1^−/− mice have phenotypes associated with advanced age likely due to enhanced replicative senescence (18–20). Most of these tissues have high cell turnover and depend on progenitor cells for tissue maintenance, such as the epidermis, intestine, and bone. As discussed above, cartilage is less reliant on chondrocyte proliferation for tissue maintenance and, thus may not be as susceptible to degeneration as these other tissues following loss of p50. Upon receipt of a signal promoting canonical NF-κB activation, however, proinflammatory cytokines and chemokines and factors catabolic to the cartilage matrix such as MMP13 will likely be more highly expressed in Nfkb1^+/− or Nfkb1^−/− mice due to the reduced presence or complete absence of the p50 repressor complex. Here, IKKβ GOF in chondrocytes served as an effector of canonical NF-κB activation. Proteoglycan content, in particular, was significantly reduced in mice with Nfkb1 deletion in addition to IKKβ GOF compared to IKKβ GOF alone. Whether cartilage degeneration progresses differentially in p50-deficient males and females in response to an effector of canonical NF-κB signaling remains unanswered as our current study was not powered sufficiently to address this question. As mentioned above, this is a major limitation of this study.

All of our current studies were performed in uninjured joints. Whether inflammatory factors found in the injured joint environment can effectively activate NF-κB signaling to accelerate the progression of PTOA in Nfkb1^+/− or Nfkb1^−/− mice remains to be determined. Of note, DNA binding by p50 homodimers is reported to be decreased in tissues from aged wild-type mice (18). Whether this could be a reason that PTOA is accelerated in the knees of aged mice compared to young mice (63) also remains to be determined and is a translationally intriguing question given that advanced age is one of the most important risk factors for OA onset in humans. Whether the effects of p50 loss on cartilage are due to a loss of chondrocyte-specific p50 activity also remains an open question, given the global deletion of Nfkb1 in our studies and the possible crosstalk between the chondrocytes and other cell types in the joint. Finally, our data justify further characterization of the mechanisms underlying p50 function in joint tissues; this should be explored with consideration of whether the anti-inflammatory activity of p50 could be harnessed as a therapeutic approach to OA.

Materials and Methods

Mice

All animal studies were approved by the University of Rochester Committee on Animal Resources. Mice were housed in groups of up to 5 animals and kept at 21°C-23°C with a 12-hour light/dark cycle. They had ad libitum access to food and water at all times. Wild-type C57BL/6J mice were obtained from The Jackson Laboratory (JAX Stock No: 000664) and bred in-house with exception of some 15- and 27-month-old mice (N = 5 males and females per group) used for the histology/histomorphometry and micro-CT analyses; these mice were obtained from the National Institute on Aging at 12 and 24 months and aged an additional 3 months in-house. The NGL (JAX Stock No: 027529), Acan^CreERT2 (JAX Stock No: 019148), Rosa26^Ikk2ca (R26^Ikk2ca, JAX Stock No: 008242), and Nfkb1^−/− (JAX Stock No: 006097) mice were previously described(31–33, 64). All mice were maintained on a C57BL/6J background with exception of the NGL mice which were maintained on an FVB/NJ background. To induce Cre-mediated recombination, tamoxifen (Sigma) was dissolved in corn oil (Sigma) and administered via intraperitoneal (IP) injection (0.1 mg/g body weight) for five consecutive days when the mice were 2-months-old. For the IKKβ GOF studies, both male and female mice were used; because IKKβ GOF and Nfkb1 deletion had similar effects on male and female knee joints, some analyses involving these genetic models compared knee joints regardless of sex.

Histology

For paraffin histology, hindlimbs were cleaned of soft tissue and fixed in 10% neutral buffered formalin (NBF) for 3 days followed by decalcification for 1 week in 14% EDTA (pH 7.4–7.6). Fixation and decalcification were both carried out at room temperature with gentle agitation. Tissue was then processed and embedded into paraffin with hindlimbs oriented sagittally and then sectioned from the medial side at 3 or 5 microns. Sections were stained with Safranin O/Fast green to visualize morphology of the bone, cartilage, and soft tissues within the knee joint. Slides were scanned and imaged using an Olympus VS120 whole slide imaging system.

For frozen histology of the NGL joints, hindlimbs were cleaned of soft tissue and fixed for 2 hours in 10% NBF followed by decalcification for 4 days in 14% EDTA (pH 7.4–7.6). Fixation and decalcification were both carried out at 4°C with gentle agitation. Samples were then placed in a 30% sucrose/phosphate buffered saline (PBS) solution overnight at 4°C prior to embedding into Shandon Cryomatrix Frozen Embedding Medium (Thermo Fisher). Samples were sectioned at 10 microns using the previously described tape transfer method and Cryofilm Type 2C(10) (65). Tape sections were adhered to glass slides, tissue side up, using a solution of 1% chitosan/0.25% acetic acid. Sections were stained with NucBlue Live Cell Stain ReadyProbes reagent (Thermo Fisher) and immediately coverslipped with Prolong Gold Antifade Mountant (Thermo Fisher). Slides were imaged using a Zeiss AxioImager 40 fluorescent microscope. Sections from littermate mice lacking the NGL transgene were also analyzed at 18 months of age to discern autofluorescence from GFP fluorescence.

Modified OARSI Scoring

Semi-quantitative modified OARSI scoring was performed using the grading system described previously where a score of 0.5 indicated proteoglycan staining loss without other changes to the articular cartilage(66). Briefly, 2 to 3 different sections from distinct levels per sample were stained with Safranin O/Fast green and randomized for blinded scoring. Scores were averaged among three independent scorers for each slide and an average score for all slides from each sample was generated to obtain each sample’s score.

Immunohistochemistry and TUNEL Staining

Details regarding vendor, concentration, and antigen retrieval for each primary antibody used are provided in Table S1. Otherwise, the following general protocol was used for IHC. Paraffin tissue sections were deparaffinized and rehydrated, then air-dried. The specified antigen retrieval for the antibody of interest was performed, and slides were rinsed in deionized water. Sections were incubated with BLOXALL for 10 minutes to quench endogenous peroxidases (Vector Labs), rinsed in deionized water, and blocked in 2.5% normal horse serum (Vector Labs) or 2.5% normal goat serum (Vector Labs; for CD45 IHC) for 30 minutes prior to incubation with primary antibody. Sections were rinsed in PBS + 0.1% Tween-20 (PBST) and incubated with ImmPress IgG peroxidase polymer secondary antibody (Vector Labs) against the species of the relevant primary antibody or with Vectastain ABC-HRP reagent (Vector Labs) for 30 minutes. After subsequent rinses in PBST and deionized water, the antigen of interest was detected by ImmPact DAB peroxidase HRP substrate (Vector Labs). All slides were counterstained in a quick dip of Mayer’s hematoxylin (Electron Microscopy Sciences), rinsed, and air-dried prior to coverslipping. For negative control slides, the relevant species of IgG (Vector Labs) was used at the same concentration as the primary antibody of interest. TUNEL staining was performed on paraffin sections per manufacturer’s instructions using the In-Situ Cell Death Detection Kit, Fluorescein (Sigma Roche). All IHC and TUNEL staining was performed on sections from at least 3 or 6 animals per group, respectively.

Histomorphometry

Semi-automated histomorphometry was performed using Visiopharm image analysis software (Version 6.7.0.2590) as previously described(67). Total, unmineralized, and mineralized tibial articular cartilage areas, cartilage area that is Safranin O⁺ (and negative, denoted background), and cartilage area that is DAB⁺ in IHC studies (and negative, denoted background) were determined using custom applications based on color thresholding. DAB⁺ areas were determined as percentages of the overall articular cartilage areas measured. For all measurements based on Safranin O/Fast Green staining, values were averaged from 2 slides per animal (aging study, 3 μm sections) or 3–5 slides per animal (aging study, 5 μm sections; IKKβ GOF study). For quantification of TUNEL⁺ and DAPI⁺ cells, a custom Visiopharm application based on both color and size thresholding to identify individual cells was used as previously described(67). Numbers of TUNEL⁺ and DAPI⁺ cells per area measured were determined as were the percentages of TUNEL⁺ cells out of total cell numbers within the articular cartilage areas measured. GFP⁺ cell numbers in the NGL knee joints were determined by manual counting of at least three sections per knee joint.

Micro-CT and Amira Analysis

Micro-computed tomography (micro-CT) scanning of the knee joints was performed with a Scanco VivaCT 40 (Scanco Medical). The scan parameters were: 1000 projections over 180°, 300 millisecond integration time, 55 kVp, and 145 microamp current. DICOMs were imported into Amira (Thermo Fisher) and calcified meniscus volume was determined by thresholding for dense tissue at 3500 Hounsfield Units (HU), then manually segmenting the menisci. Specificity of the segmentation was ensured by inspecting the 3D reconstructions of each segmentation for anatomic location and exclusion of adjacent bone (tibia, fibula or femur).

RNA isolation from mouse articular cartilage and femur heads

For isolation of the tibial articular cartilage, knee joints were disarticulated allowing articular cartilage to be shaved from the tibial plateau and immediately frozen over dry ice. For isolation of femur heads, hip joints were disarticulated and femur heads cut from the femoral neck with a bone scissor prior to freezing over dry ice. Using a Bullet Blender Gold (Next Advance), the tissues were homogenized in TRIzol (Thermo Fisher) with a mix of 2, 1, and 0.5 mm RNase-free zirconium oxide beads (Next Advance) for 10 minutes. Phases were separated using chloroform per TRIzol instructions, and the aqueous layer and one volume of 70% ethanol transferred to a RNeasy MinElute Spin Column from the RNeasy Micro Kit (QIAGEN). RNA was isolated per Micro Kit instructions.

Sternal Chondrocyte Isolation and Infection

Chondrocytes were isolated from the ribs and sterna of 4-day-old R26^{Ikk2ca/Ikk2ca} mice following the previously described method(68). Cells were counted and plated at a density of approximately 1.5 × 10⁵ cells/cm² in DMEM (Thermo Fisher) with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. The day after plating, media was supplemented with 50 μg/ml ascorbic acid and 10 μM β-glycerophosphate. Cells were infected with Ad5-CMV-GFP or Ad5-CMV-Cre adenoviruses (Baylor Vector Development Labs) at a MOI of 100 in DMEM with 10% FBS and without antibiotics, but with 10 μg/ml Polybrene (Santa Cruz) 48 hours after plating. After 24 hours, adenovirus was removed and media replaced with standard culture media until harvest for protein or mRNA isolation. For analysis of secreted factors, conditioned media were collected from the cells and incubated with the Proteome Profiler Mouse XL Cytokine Array (R & D Systems) per manufacturer’s instructions. The arrays were imaged using a ChemiDoc XRS+ (Bio-Rad) and intensities of each duplicate set of spots measured and normalized to control spots on the membrane using Image Lab software (Bio-Rad).

ATDC5 Cell Culture and Transient Transfections

ATDC5 cells (Millipore Sigma) were cultured in DMEM/F12 with 10% FBS and 1% penicillin/streptomycin (Life Technologies). For plasmid transfections, cells were plated in 6-well plates and grown to roughly 80% confluency before transfection with 2.5μg pcDNA3.1 (empty vector) or pCMV2-IKK-2 S177E S181E (Ikk2ca, a gift from Anjana Rao, Addgene plasmid #11105) using Lipofectamine 2000 (Thermo Fisher). Cells were harvested 48 hours following transfection for protein or mRNA isolation. For siRNA transfections, cells were plated in 6-well plates grown to roughly 50% confluency before transfection with mouse ON-TARGETplus Non-targeting Pool siRNA (Dharmacon), ON-TARGETplus Nfkb1 siRNA SMARTpool (Dharmacon), or ON-TARGETplus Rela siRNA SMARTpool (Dharmacon) using Lipofectamine RNAiMAX (Thermo Fisher) per manufacturer’s recommendations; 72 hours following transfection, cells were treated with vehicle (0.5% BSA in PBS) or 1 ng/ml IL-1β (R&D Systems) for 15 minutes or 3 hours prior to harvest for protein or mRNA, respectively.

Protein Isolation and Western Blotting

Protein was harvested from cell cultures using NP-40 protein lysis buffer (0.5% NP-40, 150 mM NaCl, 50 mM Tris, pH 8.0) supplemented with a protease/phosphatase inhibitor cocktail (Cell Signaling Technology). Protein concentration was measured with Bio-Rad Protein Assay Dye (Bio-Rad), a colorimetric assay based on the Bradford assay. Western blotting was performed using standard procedures and nitrocellulose membranes. Primary antibodies were diluted in 5% milk/TBST or 5% BSA/TBST if against a phosphorylated protein of interest. Primary antibodies and the dilutions at which they were used are listed in Table S2. Goat anti-mouse or goat anti-rabbit IgG HRP conjugated secondary antibodies (Bio-Rad) were diluted to 1:2000 in 5% milk/TBST. Blots were developed using SuperSignal West Pico PLUS Chemiluminescent Substrate or SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher) and then imaged using a ChemiDoc XRS+ (Bio-Rad). Western blot images were quantified using Image J. Briefly, bands were selected using the rectangle tool. The measure tool was used on the mean gray value setting. The same rectangle size was used to measure all bands of a given protein on the same blot (based on the largest band) as well as for a background measurement from a neighboring area on the blot. Pixel density measurements were inverted by subtracting the measured value from 255. Net intensity values were then obtained by subtracting the background from the inverted density measurement. Finally, relative levels for each protein were determined as a ratio of the protein-of-interest net intensity value over the control (β-actin) net intensity value for the same sample.

RNA Isolation and Quantitative Reverse Transcription PCR (RT-qPCR)

RNA was isolated from cell cultures using the RNeasy Mini Kit per manufacturer’s instructions (QIAGEN) and quantified using a NanoDrop 2000 Spectrophotometer (Thermo Fisher). cDNA was synthesized from the mRNA using the iScript cDNA Synthesis Kit (Bio-Rad) per manufacturer’s instructions. Real-time PCR was performed using a Rotor-Gene Q real-time PCR cycler (Qiagen) and the PerfeCTa SYBR Green SuperMix (Quanta Biosciences) according to manufacturer’s instructions. A list of primers used is included in Table S3. mRNA copy numbers for all genes of interest in a given sample were normalized against Actb copy numbers prior to calculating fold change among experimental samples.

Statistics

Data are presented as the mean ± SEM or SD, as noted. Statistical significance was determined in Prism (GraphPad Software, Inc.) by two-tailed Student’s t-test for comparison of 2 groups or by one-way ANOVA or two-way ANOVA for comparison of 3 or more groups. For analysis of the wild-type C57BL/6J aging datasets and the 12-month-old Nfkb1 deletion dataset, two-way ANOVA followed by Tukey’s multiple comparisons test was performed for comparisons made within sex, whereas, Bonferroni’s multiple comparisons test was performed for comparisons made between sexes of the same age or genotype, respectively. For the IKKβ GOF datasets, two-way ANOVA followed by Tukey’s multiple comparisons test was performed for all possible pairwise comparisons. For all datasets analyzed by one-way ANOVA, Tukey’s multiple comparisons test was performed post hoc for all possible pairwise comparisons. In all cases, p values of less than 0.05 were considered significant.

Data and materials availability:

All data needed to evaluate the conclusions of the paper are present in the main text of the paper and in the Supplementary Materials. An MTA between Boston Children’s Hospital and the University of Rochester Medical Center exists for pCMV2-IKK-2 S177E S181E produced by the laboratory of Dr. Anjana Rao. Raw data and protocols are available upon reasonable request to the corresponding author.