Animals, human cartilage, and recombinant Atsttrin

We performed all animal studies under institutional guidelines. All of the protocols were approval by the Institutional Animal Care and Use Committee, New York University, NY, USA. We generated, maintained, and genotyped the mice with the genetic background of C57BL/6 wildtype (WT), PGRN-deficient (PGRN^–/–), TNFR1-deficient (TNFR1^–/–), and TNFR2-deficient (TNFR2^–/–) mice as reported previously [6]. Sprague–Dawley rats were obtained from Charles River (Wilmington, MA, USA). Eight-week-old male mice and 14-week-old male rats were used for this experiment [19, 20]. For human primary chondrocyte culture, human cartilage samples were harvested from patients receiving total knee joint replacement surgery from New York University, Hospital for Joint Diseases (NY, USA). Acquisition and use of human tissue was conducted in accordance with an Institutional Review Board (IRB#12758) approved protocol. Recombinant Atsttrin was manufactured and provided by Atreaon, Inc.

Noninvasive anterior crucial ligament rupture rat model

The noninvasive OA rat model was established as described previously [20]. Animals were anesthetized and maintained using isoflurane, and the noninvasive anterior crucial ligament rupture model was established using the indicated machine: Electroforce 3200 (Bose Corp., MN, USA), Solidworks (Dassault Systemes, MA, USA), or Mojo 3D printer (Stratasys, MN, USA). After the model was established, we intraarticularly injected PBS or recombinant Atsttrin once a week for 4 weeks in total. After 4 weeks of treatment, the rats were sacrificed for histological evaluation.

Surgically induced OA mouse models

For the surgically induced destabilization of medial meniscus (DMM) mouse model, we took advantage of 8-week-old male PGRN^–/– mice and their age-matched WT control littermates. The medial meniscotibial ligament of the right knee joint was cut to generate a destabilized medial meniscus. Six mice were used in each group. After surgery, WT mice received local delivery of 6 μl PBS via intraarticular injection, while PGRN^–/– mice received local delivery of 6 μl PBS or recombinant Atsttrin (1 μg/μl). Four weeks after model induction, mice were sacrificed and knee joints were collected. The tissues were then processed for histological analysis.

To investigate the preventative as well as the therapeutic effect of Atsttrin, we also established the anterior cruciate ligament transection (ACLT) mouse model. To determine which TNFR was predominantly responsible for Atsttrin’s effect, we established the ACLT mouse model in age-matched WT, TNFR1^–/–, or TNFR2^–/– male mice (n = 6, respectively). To address the preventative potential of Atsttrin in OA, we intraarticularly injected 6 μl Atsttrin or PBS once a week for 4 weeks beginning on the day of surgery, as based on our previous study [16]. For examination of Atsttrin’s therapeutic effect, 6 μl Atsttrin or PBS were intraarticularly injected once a week for 4 weeks beginning 4 weeks postoperatively, as based on our previous study [16]. Ambulatory behavior of mice was monitored and recorded throughout the study. After 4 weeks of treatment, mice were sacrificed for dorsal root ganglia (DRG) harvest and histological evaluation.

Sandwich ELISA for cartilage oligomeric matrix protein

The serum concentration of cartilage oligomeric matrix protein (COMP) was analyzed by our established sandwich ELISA [21]. Protein A agarose (Invitrogen) purified rabbit anti-COMP pAb and anti-COMP type III mAb 2127F5B6 were used as capture and detection antibodies, respectively. Anti-COMP type III mAb 2127F5B6 was labeled with horseradish peroxidase (HRP) using the Lightning-Link™ Horseradish Peroxidase Labeling Kit (Innova) as per the manufacturer’s protocol. Results were interpreted based on the linear range of the standard curve. All of the samples were assayed in triplicate.

Primary cultures of chondrocytes

Human articular chondrocytes were harvested by enzymatic digestion in accordance with established methodology [16]. Briefly, human cartilage slices were cut into small pieces and washed several times with PBS, pH 7.4. Minced tissues were incubated with agitation in digestion medium comprised of 0.25% collagenase II in DMEM medium with 5% FBS in a spinner flask for 16 hours at 37 °C with 5% CO₂. After digestion, the suspended cells were collected and seeded into six-well plates for subsequent study. For mouse primary chondrocyte culture, knee joints were collected from 6-day-old WT, TNFR1^–/–, or TNFR2^–/– mice following sacrifice. Under magnification, the cartilage samples were isolated with special attention to avoid damaging the subchondral bone and tissues were rinsed completely three times in PBS. Primary cartilage samples were placed in a 10-cm dish containing the aforementioned digestion buffer and incubated for 16 hours at 37 °C with 5% CO₂. After full digestion, suspended cells were collected and seeded in a six-well plate for use. All chondrocytes used for experiments are first-generation cells.

von Frey test

von Frey filaments (Stoelting, Wood Dale, IL, USA) were applied with ascending force intensities on the plantar surface of the hind paw to determine the tactile pain threshold as based on a previous publication [22]. Rapid withdrawal of the hind paw was recorded as a positive response. Hind paws were subjected to 10 trials at a given intensity with a 30-second interval maintained between trials. The number of positive responses for each von Frey filament’s stimulus was recorded. Animals were considered to have reached tactile threshold when five in 10 trials generated a positive response. The examiner was blinded to the groups.

Dorsal root ganglia isolation

Eight weeks after ACLT surgery, mice were sacrificed and L3–L5 DRG were isolated based on a previous publication [23]. Briefly, mice were anesthetized using isoflurane and fur was cleared from the dorsal surface. A longitudinal incision was made, the spinal column was exposed, and L3–L5 DRG were extracted and tissues flash-frozen using liquid nitrogen. These tissues were processed using the Qiagen RNeasy kit (Qiagen, Valencia, CA, USA) for RNA extraction.

Luciferase assay

Luciferase assay was performed as reported previously [24]. Lipofectamine2000 DNA transfection reagent was used to cotransfect NF-κB and renilla plasmids in C28I2 cells according to the manufacturer’s protocol (Life Technologies). Eighteen hours after transfection, C28I2 cells were treated without or with 10 ng/ml TNFα in the absence or presence of 200 ng/ml recombinant Atsttrin. After 24-hour incubation, we measured luciferase activity using the Reporter Assay System of Dual-Luciferase® in accordance with the manufacturer's instructions (Promega).

Histological analysis and immunostaining

Histological analysis was conducted as described previously [16]. Briefly, knee joints were fixed immediately after sacrifice in 4% PFA at room temperature for 48 hours. After washing three times in PBS, the tissues were decalcified at 4 °C in 10% w/v EDTA for 2 weeks. Tissues were measured using a vernier caliper before paraffin processing. Knee joints were dehydrated and embedded; the blocks were trimmed to the midpoint as calculated previously from caliper measurements and serial 5-μm sections were placed on slides for staining. H&E or Safranin O/fast green staining was performed following the established protocol. The extent of synovitis was determined using a graded scale based on H&E staining: grade 0, no signs of inflammation; grade 1, mild inflammation with hyperplasia of the synovial lining without cartilage destruction; and grades 2–4, increasing degrees of inflammatory cell infiltrate and cartilage/bone destruction. For immunohistochemistry staining, sections were pretreated with 0.1% trypsin for 30 minutes at 37 °C. Sections were washed with PBS three times, followed by treatment with 0.25 U/ml chondroitinase ABC (Sigma-Aldrich) for 1 hour and then 1 U/ml hyaluronidase (Sigma-Aldrich) for 1 hour at 37 °C. To reduce nonspecific staining, sections were blocked at room temperature with 20% normal horse serum diluted in 3% BSA for 1 hour. Without washing after blocking, Col X antibody (1:200 dilution; DSHB), MMP-13 antibody (ab3208, 1:200 dilution; Abcam), and affinity-purified monoclonal anti-COMP fragments (1:200 dilution) were diluted in 20% normal horse serum with 3% BSA at 4 °C overnight. Sections were prepared for detection using the Vectastain Elite ABC kit following the manufacturer’s guidelines at 25 °C for 1 hour. Immunorecativity was visualized using 0.5 mg/ml 3,3′-diaminobenzidine (DAB) in 50 mM Tris–HCl substrate, pH 7.8. Methyl green (1%) was used for counterstaining.

Histological analysis and score

The articular cartilage proteoglycan content was defined on the basis of Safranin O staining. In this study, we used the well-accepted Osteoarthritis Research Society International (OARSI) scoring system [25]: 0 = normal cartilage without any damage; 0.5 = loss of Safranin O staining with no detectable structural change; 1 = small fibrillation; 2 = vertical damage of cartilage limited to superficial layer; 3 = vertical damage, no more than 25% of the cartilage surface; 4 = vertical damage, 25–50% of the cartilage surface; 5 = vertical damage, 50–75% of the cartilage surface; and 6 = vertical damage, more than 75% of the cartilage surface.

Real-time RT-PCR

Total RNA were extracted from cultured chondrocytes using the RNeasy kit (Qiagen) and reverse transcribed into cDNA using the ImProm-II reverse transcription system (Promega). Data were normalized to the internal control, GAPDH. The primers for specific amplification of murine genes are as follows: 5′-AATGCTGGTACTCCAAACCC-3′ and 5′-CTGGATCGTTATCCAGCAAACAGC-3′ for Aggrecan; 5′-ACTAGTCATCCAGCAAACAGCCAGG-3′ and 5′-TTGGCTTTGGGAAGAGAC-3′ for Col II; 5′-AATCTCACAGCAGCACATCA-3′ and 5′-AAGGTGCTCATGTCCTCATC-3′ for IL-1β; 5′-ACAGGAGGGGTTAAAGCTGC-3′ and 5′-TTGTCTCCAAGGGACCAGG-3′ for NOS-2; 5′-GCATTGACGCATCCAAACCC-3′ and 5′-CGTGGTAGGTCCAGCAAACAGTTAC-3′ for ADAMTS-4; 5′-ACTTTGTTGCCAATTCCAGG-3′ and 5′-TTTGAGAACACGGGGAAGAC-3′ for MMP-13; 5′-CATAGCAGCCACCTTCATTCC-3′ and 5′-TCTCCTTGGCCACAATGGTC-3′ for MCP-1; 5′-AGAGAGCTGCAGCAAAAAGG-3′ and 5′-GGAAAGAGGCAGTTGCAAAG-3′ for CCR-2; and 5′-AGAACATCATCCCTGCATCC-3′ and 5′-AGTTGCTGTTGAAGTCGC-3′ for GAPDH. Melting curve analysis was used to verify the PCR product. Each experiment was repeated three times.

Western blot analysis

Proteins extracted from chondrocytes were processed on 8% SDS-polyacrylamide gel, followed by electrotransfer to nitrocellulose membrane. The membrane was blocked in 3% BSA in 10 mM Tris–HCl, pH 8.0, 150 mM NaCl, and 0.5% Tween 20. After washing three times, blots were incubated at 4 °C overnight with polyclonal anti-Erk1/2 (#4695, 1:1000 dilution; Cell Signaling Technology), anti-phosphorylated Erk1/2 (#4370S, 1:1000 dilution; Cell Signaling Technology), polyclonal anti-Akt (#9272, 1:1000 dilution; Cell Signaling Technology), anti-phosphorylated Akt (#4058S, 1:1000 dilution; Cell Signaling Technology), polyclonal anti-MMP-3 (ab52915, 1:1000 dilution; Abcam), polyclonal anti-MMP-13 (ab3208, 1:1000 dilution; Abcam), polyclonal anti-ADAMTS-4 (PA1-1750, 1:1000 dilution; Thermo Fisher Scientific), polyclonal anti-NOS-2 (SC651, 1:1000 dilution; Santa Cruz Biotechnology), polyclonal anti-GAPDH (SC25778, 1:1000 dilution; Santa Cruz Biotechnology), polyclonal anti-tubulin (#5346, 1:1000 dilution; Cell Signaling Technology), or diluted polyclonal anti-lamin B (SC-6217, 1:500 dilution; Santa Cruz Biotechnology). After washing three times, blots were incubated with an appropriate HRP-conjugated anti-rabbit/mouse immunoglobulin secondary antibody at 25 °C for 1 hour. The bound antibody was visualized using an enhanced chemiluminescence system (Amersham Life Science, Arlington Heights, IL, USA).

Cartilage explant cultures

Cartilage explants were cultured as described previously [16]. Briefly, mouse femoral head cartilage was isolated and finely minced to 1 mm diameter and 1 mm thickness. The cartilage explants were then dispensed into serum-free DMEM containing 25 mM HEPES and 2 mM glutamine, in the absence or presence of recombinant Atsttrin (200 ng/ml).

Dimethylmethylene blue assay of GAG

The mouse cartilage culture medium was collected and GAG release was quantified by dimethylmethylene blue assay (DMMB) (Polysciences, Warrington, PA, USA). Hyaluronidase (0.5 unit/ml; Seikagaku, Tokyo, Japan) was incubated with collected medium for 3 hours at 37 °C to remove hyaluronan in order to reduce inhibition of the DMMB assay. The DMMB signal from digests was measured at 520 nm using a SpectraMax 384 Microplate Reader (Molecular Devices, Sunnyvale, CA, USA). The GAG content was calculated based on linear regression of readings from chondroitin-6-sulfate standards from Shark cartilage (Sigma Aldrich, St. Louis, MO, USA). Each sample was read in triplicate. The average values of the triplicates were normalized to the standard curve.

Statistical analysis

Results were expressed as mean ± SEM. Statistical analysis included Student’s t test performed by SPSS software (SPSS Inc., Chicago, IL, USA). p < 0.05 was considered statistically significant.

Atsttrin rescues accelerated OA caused by PGRN deficiency

Previously, we reported that PGRN is expressed in both human and mouse articular cartilage, and its level is elevated, relative to healthy controls in both human OA cartilage and in surgically induced OA model mice. In addition, we have reported that loss of PGRN led to enhanced OA in both “aged” mice and surgically induced OA models [16]. In the present study, PGRN expression is slightly, but not significantly, elevated in 1-year-old aged mice as compared with 4-month-old mice (Additional file 1: Figure S1A, B). To determine whether Atsttrin, an engineered protein derived from PGRN [6], could prevent the accelerated OA seen in our surgically induced, PGRN-null OA models, the DMM model was generated in PGRN^–/– and age-matched WT mice. As illustrated in Fig. 1a, Safranin O-stained sections demonstrated that Atsttrin effectively prevented loss of proteoglycan content in PGRN^–/– mice. Statistical analysis of proteoglycan loss and the OARSI score obtained from Safranin O-stained sections indicates that Atsttrin significantly reduced articular cartilage destruction (Fig. 1b, c). Moreover, as shown in Fig. 1d, application of recombinant Atsttrin also reduced the serum levels of COMP-degradative fragments as assayed by ELISA.

Atsttrin prevents OA development in surgically induced OA mouse model

Recently, Xia et al. [18] reported that local delivery of Atsttrin-transduced MSCs could effectively protect against OA development in a murine OA model. This finding promoted us to determine whether local delivery of recombinant Atsttrin could prevent OA development. To address this issue, we established the ACLT OA model in C56LB/6 WT mice, followed by intraarticular injection of recombinant Atsttrin (6 μg) once per week for 4 weeks. As shown in Fig. 2a, histological analysis indicated that Atsttrin effectively protected against cartilage loss. Additionally, the OARSI score and proteoglycan loss score based on the histology revealed a significant improvement in the Atsttrin-treated group (Fig. 2b, c). Furthermore, as shown in Fig. 2d, immunohistochemical staining demonstrated that Atsttrin reduced COMP degradation, type X collagen (Col X) expression, as well as matrix-degrading enzyme matrix metalloproteinase 13 (MMP-13) expression. Notably, subchondral bone provides support for articular cartilage [26]. Studies indicate that the thickness of the subchondral bone plate significantly decreases during the first phase of ACLT-induced OA in mice [27] and loss of subchondral bone leads to reduced support and more deterioration of articular cartilage in OA pathogenesis [28, 29]. Interestingly, Atsttrin effectively inhibited subchondral bone loss (Fig. 2e). Moreover, we reported previously that Atsttrin inhibited osteoclastogenesis in vitro [6] and it is believed that osteoclastogenesis is involved in subchondral bone remodeling in ACLT mice [30]. To demonstrate whether Atsttrin-mediated subchondral bone protection occurs through inhibition of osteoclast formation, we performed tartrate-resistant acid phosphatase (TRAP) staining. As illustrated in Fig. 2f, Atsttrin significantly inhibited osteoclast activity. Synovium also plays an important role in the pathogenesis of OA; to further determine the effect of Atsttrin on synovitis, we performed H&E staining and found that Atsttrin effectively inhibited synovial inflammation (Fig. 2 g, h). Behavioral observations and pain marker levels were also recorded to determine whether Atsttrin could reduce OA-related pain. Atsttrin treatment was associated with significantly greater rearing times, as compared to the PBS-treated group when monitored for free ambulation in an open field testing box, as indicated in Fig. 2i. In addition, von Frey filament testing demonstrated that the pain threshold was decreased after ACLT surgery and that Atsttrin effectively improved the pain threshold (Fig. 2j). Further, inflammatory cytokines such as chemokine monocyte chemoattractant protein 1 (MCP-1), chemokine receptor 2 (CCR-2), and interleukin-1β (IL-1β) in DRG play an important role in OA-associated pain. It is reported that upregulated transcriptional levels of MCP-1, CCR-2, and IL-1β in DRG correlate to exhibition of pain-associated behaviors in murine OA models [31]. As indicated in Fig. 2 k, l, the transcriptional level of MCP-1 and CCR-2 in the ipsilateral L3–L5 DRG was significantly upregulated compared to the control group, whereas Atsttrin significantly reduced their expressions to levels comparable to those of the control group. Collectively, these results indicate that Atsttrin is likely to be a disease-modifying and also a symptom-modifying molecule.

Atsttrin is therapeutic against OA and its therapeutic effects depend on both TNFR1 and TNFR2 pathways

To determine whether recombinant Atsttrin is therapeutic against OA, we took advantage of both nonsurgically induced rat and surgically induced mouse models. Four weeks following the establishment of the unique noninvasive rat OA model, we locally delivered 60 μg Atsttrin once per week for 4 weeks. Histological analysis indicated that Atsttrin effectively protected against articular cartilage destruction (Fig. 3a) and significantly improved the arthritis score as well as the proteoglycan loss score (Fig. 3b, c).

We also took advantage of the well-accepted surgically induced ACLT mouse model. Beginning 4 weeks postoperatively, we intraarticularly injected 6 μg recombinant Atsttrin weekly for 4 weeks. As indicated in Fig. 4a–c, Atsttrin exhibited a therapeutic effect in the WT mice, evidenced by preservation of articular cartilage. To compare the efficacy of PGRN and its derived engineered protein Atsttrin in mitigating OA progression, we also compared the OARSI score between these two groups, and found no significant difference between PGRN and Atsttrin in treating OA (Additional file 2: Figure S2).

To determine whether the Atsttrin-mediated protective effect is dependent upon TNFR1 or TNFR2, or both receptors, we established the ACLT model in TNFR1^–/– and TNFR2^–/– mice. As revealed in Fig. 4a–c, TNFR1^–/– mice exhibited partial loss of Atsttrin-mediated protection while the Atsttrin-mediated protective effect was almost abolished in TNFR2^–/– mice. However, we found that, in the presence of Atsttrin, the OARSI score and proteoglycan loss score of both TNFR1^–/– mice and TNFR2^–/– mice were each significantly increased relative to those of WT mice. Sera from WT mice and from each group of TNFR null mice were collected and COMP fragment-specific ELISA revealed that Atsttrin significantly decreased COMP degradation in the WT OA mice; however, the Atsttrin-mediated protective effect against COMP degradation was slightly decreased in TNFR1^–/– mice and relatively unaffected in TNFR2^–/– mice (Fig. 4d). Notably, after Atsttrin treatment, the level of COMP fragments in TNFR1^–/– mice and TNFR2^–/– mice were significantly increased in comparison to levels in WT mice. Collectively, both TNFR1 and TNFR2 appear to be required for mediating Atsttrin’s protective effect in OA.

Atsttrin-mediated anabolism primarily depends on TNFR2-Akt-Erk1/2 signaling

Given that Atsttrin selectively binds to TNFRs and that the Atsttrin-mediated beneficial effect in OA depends on both receptors in vivo, we next sought to examine the molecular mechanisms involved through detailing the effect of Atsttrin on cartilage and chondrocyte metabolism. To determine whether Atsttrin has an anabolic effect in cartilage, we took advantage of mouse primary cartilage explants and found that recombinant Atsttrin significantly promoted GAG synthesis in WT and TNFR1^–/– cartilage while Atsttrin lost this effect in TNFR2^–/– cartilage (Fig. 5a). Additionally, Atsttrin-induced expressions of anabolic biomarkers, including type II collagen (Col II) and aggrecan (ACN), were remarkably reduced in TNFR2^–/– chondrocytes in comparison with WT and TNFR1^–/– chondrocytes (Fig. 5b, c). These data indicate that the Atsttrin-mediated anabolic effect primarily depends on the TNFR2 pathway.

Previous study has shown that both Akt and Erk1/2 pathways are involved in chondrocyte anabolism and PGRN slightly activates Akt signaling and strongly actives Erk1/2 signaling in chondrocytes [13]. Contrastingly, we demonstrate herein that Atsttrin strongly activates Akt signaling and slightly activates Erk1/2 signaling (Fig. 5d). Activation of these signaling pathways was lost in TNFR2^–/– chondrocytes, but no change was observed in TNFR1^–/– chondrocytes relative to WT chondrocytes (Fig. 5d). More importantly, Atsttrin’s anabolic effect was partially lost after blocking Erk1/2 signaling alone using U0126 or blocking the Akt signaling pathway alone using Wortmannin (Fig. 5e–h). Furthermore, by applying both Erk1/2 and Akt signaling inhibitors simultaneously, we found that Atsttrin-mediated anabolism in chondrocytes was almost abolished (Fig. 5i, j). Taken together, Atsttrin exhibits an anabolic effect in chondrocytes, and this effect primarily relies on TNFR2-Akt-Erk1/2 signaling.

Atsttrin inhibits TNFα-induced inflammatory catabolism

It is known that TNFα plays an important role in OA [32]. Additionally, TNFα could activate MAPK and NF-κB signaling [6], subsequently inducing matrix-degrading enzymes, including MMP-13 and ADAMTS-4, as well as other inflammatory biomarkers, such as NOS-2, in chondrocytes. The finding that Atsttrin protected against TNFα-induced cartilage loss in multiple rheumatoid arthritis mouse models promoted us to determine the interplay between TNFα and Atsttrin in OA. Human primary chondrocytes were cultured in the absence or presence of 10 ng/ml TNFα with or without 200 ng/ml Atsttrin for various lengths of time. Western blot analysis indicated that Atsttrin inhibited TNFα-activated signaling pathways (Fig. 6a). Since TNFα exerts its inflammatory action largely through activation of classical NF-κB signaling [6], we next examined whether Atsttrin altered TNFα-mediated NF-κB phosphorylation, translocation, and activity. As elucidated in Fig. 6b, Atsttrin effectively inhibited phosphorylation of NF-κB in primary human chondrocytes in vitro. Additionally, immunochemistry staining demonstrated that Atsttrin dramatically inhibited NF-κB phosphorylation in the articular cartilage of mice induced under the ACLT model (Fig. 6c).

Separately extracting cytoplasmic and nuclear protein from human primary chondrocytes also revealed differences in NF-κB translocation in treated and untreated cells. As illustrated in Fig. 6d, without TNFα treatment p65 was mainly detected in the cytoplasm; after TNFα treatment, cytoplasmic p65 was decreased and nuclear p65 was increased. In contrast, TNFα-mediated NF-κB nuclear translocation was nearly abolished in the presence of Atsttrin. We next transfected a NF-κB reporter gene into C28I2 human chondrocyte cells (provided by Dr Mary Goldring, Hospital for Special Surgery, New York, NY, USA [33]) to test whether Atsttrin altered NF-κB activity. As shown in Fig. 6e, Atsttrin significantly inhibited TNFα-induced NF-κB activity. To further elucidate the mechanism, we cultured primary human chondrocytes in the absence or presence of 10 ng/ml TNFα with or without 200 ng/ml Atsttrin for 6 hours. After this, the total RNA was extracted for real-time PCR analysis. As indicated in Fig. 6 g–j, Atsttrin significantly decreased TNFα-upregulated transcriptional levels of MMP-13, ADAMTS-4, NOS-2, and COX-2. After 48-hour incubation, Atsttrin exhibited inhibition of TNFα-induced protein levels of inflammatory matrix-degrading enzymes and inflammatory cytokines in a dose-dependent manner (Fig. 6f).