Human tissues and animal experiments

Human tissues were obtained with Institutional Review Board (IRB) approval and informed consent from all subjects. TMJ tissues were obtained from Boston University (IRB H33300). TMJ tissues were selected on the basis of clinical features such as TMJ pain and mobility and confirmed by pathological diagnostic findings as healthy or TMJ-OA, at the Department of Pathology, Boston University. The exclusion criteria included a history of any rheumatic disease or cancer. Histological slides from 16 patients (8/condition) including TMJ normal (Osteoarthritis Research Society International (OARSI) grade I) and TMJ-OA (OARSI grade V) from right and left joints were obtained. Freshly isolated cells from TMJ-OA cartilage from five patients were used for animal experiments. Histological slides of tissues from hip joints were obtained from the Department of Orthopaedic Surgery, School of Medicine, Boston University, Boston, MA, USA (IRB H-32517). Tissue sections included both uninvolved and involved regions of the femoral head based on micro computed tomography (microCT) mapping of the regions of cartilage erosion relative to bone cysts. Histological sample slides were obtained from six individuals. The OA cartilage histological tissue slides from the knee joints were obtained from Georgia Medical College, Augusta, GA, USA under IRB approval (657441-5). Tibial epiphyses from OA-affected regions (OARSI grade V) and the adjacent healthy region (OARSI grade I) were collected from five patients. All mouse experiments were performed with guidance and regulations through Boston University Institutional Animal Care and Use Committee (IACUC), approval number AN-15387.

Immunofluorescence of human tissue sections

Data are reported as the fold-change expression compared to respective controls.

RNA isolation and analysis

Total RNA was extracted from paraffin-embedded sections of the femur, tibia, and TMJ condyle using Trizol (Qiagen) and FFPE RNA isolation kit (Qiagen), according to the manufacturer’s instructions. Briefly, tissue sections were deparaffinized with Xylenes for 10 min, washed with Xylenes then ethanol, and air-dried. Tissues were treated with proteinase K and RNAase inhibitor overnight and extracted with Trizol. Finally, RNA was precipitated using 10 μg glycogen and isopropanol, washed with ethanol, and reconstituted in nuclease-free water. Quantitative real-time PCR (RT-qPCR) analysis was performed using TaqMan gene expression assays from Life Technologies, according to a standard protocol [16]. TaqMan® Gene Expression Assay IDs for human and mouse primers were: Sex determining region Y-box containing gene 9 (SOX9) (Hs01001343_g1 and Mm00448840_m1), matrix metalloproteinase (MMP)13 (Hs00233992_m1 and Mm00439491_m1), aggrecan (ACAN) (Hs00153936_m1 and Mm00545794_m1), collagen type II α1 (COL2A1) (Hs00264051_m1 and Mm01309565_m1), and LOXL2 (Hs00158757_m1).

Human cell cultures

Human articular chondrocytes (HAC) from three OA patients (HAC-OA) (Cell Application Inc. from lot no. 2356, 1777, and 2881) were used. Normal HAC (HAC-N) cells were from Lonza (CC-2550). Cell lines were grown in chondrocyte growth medium (CGM™ BulletKit™, CC-3216, Lonza Inc.) or in chondrocyte differentiation medium (CDM BulletKit, CC-3225, Lonza), both of which contain 5% fetal bovine serum (FBS), R3-insulin-like growth factor-1, human recombinant fibroblast growth factor-beta; insulin-transferrin-selenium (ITS), and gentamycin/amphotericin-B for in vitro studies. The chondrocyte differentiation medium also contains TGF-β3. These supplements have been optimized for HAC [17, 18]. The medium, with a similar composition, from Lonza and Cell Applications optimized for HAC (HAC-O and HAC) has been used in several in vitro and in vivo studies [19, 20] and in our experiments. Primary TMJ cells were obtained as outgrowths from explants of TMJ condylar cartilage within 24 h after collection as 1-mm to 3-mm punch biopsies, and cultured in chondrocyte growth medium with 5% FBS (Lonza Inc.).

Preparation of lentivirus and adenovirus

Adenoviruses for transient expression of LOXL2 (Ad-CMV-RFP-CMV-hLOXL2-His; referred to as Ad-RFP-LOXL2), and its empty vector (EV) control (Ad-RFP-EV) were custom-synthesized (ADV-214438) by Vector Biolabs. These adenovirus particles were amplified in 293 T cells and quantified using an adenovirus quantification kit (Cell Biolabs). The lentivirus packaging was performed using the 3-plasmid protocol [16]. Lentiviral vector for LOXL2 (CMV-LOXL2) and CMV-EV)) were purchased from Genecopoeia Inc. (EX-Y2020-Lv128 and LV128;). The expression of LOXL2 protein was evaluated by western blotting of the medium and cell layer, and enzymatic activity of purified LOXL2 was evaluated by the Amplex Red assay [21].

Cell proliferation

The CyQuant cell proliferation reagent assay (Molecular Probes), was used to measure the cell proliferation. HAC or HAC-OA cells were seeded at a density of 20,000 cells per well in 24-well plates and transduced with Adv-RFP-EV or Adv-RFP-LOXL2 in the respective group. After 24 h, they were washed with PBS, and the plates with cell layer were stored at −80 °C. For measurements, plates were thawed at room temperature for 30 min and 200 μl of CyQuant GR dye/cell-lysis buffer added to each well. Fluroscence measured as per manufacturer’s instructions with excitation at 420 nm and emission detection at 535 nm.

Apoptosis assay

HAC-N or HAC-OA were infected with the Ad-RFP-EV or ad-RFP-LOXL2 at a multiplicity of infection (MOI) of 50 in triplicate wells. The next day, cells were treated with 10 ng/mL of IL-1β, 10 ng/mL of TNF-α, or vehicle for 24 h in serum-free medium. Cells were lysed using lysis buffer provided in the kit and were analyzed with a bioluminescence-based caspase-3/7 glow apoptosis assay (Promega) according to manufacturer’s instructions.

Molecular signaling analysis

HAC-OA, transduced with either Ad-RFP-LOXL2 or Ad-RFP-EV at MOI 25, were incubated in serum-free growth medium for at least 16 h and then left untreated (vehicle) or treated for 10 min with 10 ng/mL of IL-1β or TGF-β1. Proteins were extracted for western blot analysis.

Western blotting

Tissue or cell layers were extracted with SDS-PAGE sample buffer (0.1 mM Tris-HCl, 4% SDS, 10% glycerol, 5% β-mercaptoethanol) and boiled for 3–5 minutes. Protein concentrations were determined using Nano Orange assay kits (Molecular Probes, Eugene, OR, USA). From tumors, approximately 20 μg of protein per experimental group were then subjected to 10% SDS-PAGE and western blotting with primary antibodies (Cell Signaling Technology). The antibodies used were for evaluation of phospho-NF-κB-p65 (S536) and NF-κB-p65 (total), phospho-ERK1/2, total extracellular signal-related kinase (ERK) and β-actin. Anti-rabbit and anti-mouse secondary antibodies were purchased from Cell Signaling Technology. Quantification was performed using a digital densitometry system (Versadoc; BioRad, Hercules, CA, USA) and Image J software.

Gene array experiments

To evaluate the genes that were differentially regulated by LOXL2, HAC-OA from three different OA patients were cultured in chondrocyte growth medium (Lonza) until they reached confluence and were infected with Ad-RFP-EV or Ad-RFP-LOXL2 at MOI of 25 in medium supplemented with 5% FBS. After overnight incubation, cells were washed once with PBS and cultured overnight in the fresh serum-free medium. Cells were processed for total RNA extraction using the RNeasy kit (Qiagen). The sample integrity was verified using RNA 6000 Nano Assay RNA chips run in an Agilent 2100 Bioanalyzer (Agilent Technologies). Analysis of mRNA was performed on microarrays and by TaqMan RT-qPCR. All microarray procedures were performed at the Boston University Microarray Resource Facility, according to the protocol for the GeneChip® WT Plus Kit Reagent Manual (Affymetrix,). The total RNA (100 ng) was reverse-transcribed using GeneChip® WT Plus Reagent Kit (Affymetrix,). The cDNA was used as a template for in vitro transcription to antisense cRNA, which was purified using Purification Beads (Affymetrix,) and then used as a template for reverse transcription to produce single-stranded DNA in the sense orientation. Microarrays were immediately scanned using Affymetrix Gene Array Scanner 3000 7G Plus (Affymetrix). Log2 (expression) was computed across all samples and laid over a colored representation (heatmap), which was scaled so that red and blue indicated expression values ≥2 standard deviations above and below, respectively, the row-wise mean (white).

Gene array analysis

Affymetrix GeneChip Human Gene 2.0 ST CEL files were normalized to produce gene-level expression values using the implementation of the Robust Multiarray Average (RMA) [22] in the Affymetrix package (version 1.36.1) [23] included in the Bioconductor software suite (version 2.12) [24] and an Entrez gene-specific probeset mapping (version 17.0.0) from the Molecular and Behavioral Neuroscience Institute (Brainarray) at the University of Michigan [25]. Array quality was assessed by computing relative log expression (RLE) and normalized unscaled standard error (NUSE) using the affyPLM Bioconductor package (version 1.34.0). Principal component analysis (PCA) was performed using the “prcomp R” function with expression values that had been normalized across all samples to a mean of zero and a standard deviation of one. Pairwise differential expression was assessed using the moderated (empirical Bayesian) t test implemented in the limma package (version 3.14.4) (i.e., creating simple linear models with lmFit, followed by empirical Bayesian adjustment with eBayes). Fold changes are indicated as 2 = twofold higher or −2 = twofold lower in Ad-RFP-LOXL2- than in Ad-RFP-EV-transduced cells. Correction for multiple hypothesis testing was accomplished using the Benjamini-Hochberg false discovery rate (FDR) and represented as FDR q values. All statistical analyses were performed using the R environment for statistical computing (version 2.15.1).

In vivo implantation of human articular and TMJ chondrocytes in nude mice

As an alternative to various models based on human chondrocytes [26–28], we established a model using implants of primary TMJ-OA or HAC-OA chondrocytes embedded in Matrigel. For in vivo applications, freshly isolated TMJ-OA cells or frozen cells obtained from Cell Application Inc., were expanded only once to maintain their phenotype prior to preparation of implants. Freshly prepared 10⁶ TMJ-OA or HAC-OA chondrocytes in 50 μL medium, from TMJ or knee joints of three different patients with OA, were mixed with Matrigel at a 1:1 ratio. The total 100 μL of Matrigel:chondrocyte suspension were implanted subcutaneously in the backs of nude mice (three implants/mouse) and allowed to grow for a week. These implants were treated locally with weekly injections of 30-μL suspensions of Ad-RFP-LOXL2 or Ad-RFP-EV (n = 5/condition), for 6 weeks. Transduction was confirmed by visualization of RFP by in vitro imaging systems (IVIS) each week. One implant from each mouse was then processed for RNA isolation, and the other two were prepared for histologic analysis and stained with Safranin O/Fast Green (American Mastertek Inc.).

Immunofluorescence analysis of human cartilage implants extracted from mice

Paraffin-embedded tissue sections were labeled with mouse SOX9 antibody (Abcam) or its isotype control (Vector Biolabs) detected with anti-mouse IgG conjugated to Alexa 488. To detect LOXL2 expression, the tissues were subjected to rabbit anti-LOXL2 (GeneTex), anti-RFP (Abcam), anti-phospho-SMAD2/3 (Cell Signaling Technology), anti-COL2A1 (Abcam) or its isotype control antibody and detected with anti-rabbit Alexa 488-conjugated antibody or biotin antibody followed by streptavidin-conjugated Texas red, in respective samples. Anti-fade reagent with DAPI was added to all samples. A Zeiss 710 dual scanner confocal microscope with a Plan-Apochromat objective, oil immersion lens, and a CCD detector was used to obtain confocal images. Image acquisition was performed with Zeiss Zen image analysis software (Carl Zeiss Micro Imaging). The image analysis was performed using Zeiss LSM viewer and Image J software (NIH). Z-stack images analysis and 3-dimensional reconstruction was performed by using LOCI and the 3D viewer plug-in of Image J software. Quantification was performed using Image J software as described [15].

Data analysis

Data analyses were performed using two-way analysis of variance (ANOVA) with Bonferroni post-hoc analysis or Student’s t test (Graph Pad Prism 5 software). All experiments were performed three times each using cells derived from a different HAC-N, HAC-OA, or TMJ patient sample. Each data point is represented in the graphs as mean ± SEM of three experiments with significance set at P < 0.05.

Immunolocalization of LOXL2 protein in human TMJ and hip and knee joints

To determine whether LOXL2 is present in OA tissues, we evaluated its localization by immunohistochemistry in healthy (OARSI, grade I) and OA tissues (OARSI grade V) from multiple joints, including the TMJ condyle, the femoral head, and the tibial plateau of the knee joint. Safranin O/Fast green staining was used to identify the cartilage region (Fig. 1a). LOXL2 was detectable in damaged parts of cartilage, compared to the negligible LOXL2 in the healthy regions of tissues from all three joints (Fig. 1). Thus, LOXL2 is upregulated in damaged joints, similar to other anabolic factors such as BMP2 [29], fibroblast growth factor (FGF)18 [30], insulin-like growth factor (IGF)1, and TGF-β1 [31], compared to cartilage not affected by OA.

Regulation of LOXL2 mRNA by OA-related factors

Whether LOXL2 expression is modulated by anabolic and catabolic factors involved in OA [7] has not been defined. Therefore, we investigated the effects of anabolic factors (TGF-β1, IGF-1, BMP-2, and BMP-7) and catabolic factors (TNF-α, and IL-1β) on LOXL2 gene expression in chondrocytes treated for 24 h. Compared to vehicle stimulation, LOXL2 mRNA levels in HAC-OA were induced significantly by TGF-β1 (twofold) stimulation, whereas they were downregulated 0.5-fold and 0.7-fold, respectively, by TNF-α and IL-1β (Fig. 2). However, treatment with IGF-1, BMP-2, or BMP-7 produced no significant effect on LOXL2 mRNA levels in HAC-OA (not shown). These results indicate that LOXL2 gene expression can be modulated by key factors affecting OA-related processes in cartilage.

LOXL2 promotes the differentiated chondrocyte phenotype and inhibits apoptosis in human chondrocytes from OA cartilage

To evaluate if LOXL2 promotes the differentiated chondrocyte phenotype, we examined proliferation, chondrocyte phenotypic markers, and TNF-α-induced and IL-1β-induced effects on COL2A1 mRNA and apoptosis using chondrocytes isolated from healthy cartilage (HAC-N) or OA cartilage (HAC-OA) and transduced with viral vectors. HAC-OA cells infected with Ad-RFP-LOXL2 and cultured for 24 h showed no significant difference in cell proliferation compared to those transduced with Ad-RFP-EV (Fig. 3a). To permit long-term expression of the transduced LOXL2 in the HAC-OA cultures for evaluation of chondrogenic differentiation, we used lentiviral vectors lenti-CMV-LOXL2 and lenti-CMV-EV (control). HAC-OA-LOXL2 cells showed increased SOX9, ACAN, and COL2A1 mRNA levels on days 14 and 21 after lenti-CMV-LOXL2 transduction and culture in chondrogenic differentiation medium compared to cells transduced with lenti-CMV-EV vector (Fig. 3b). Next, we determined that LOXL2 transduction inhibits IL-1β-induced caspase 3/7 activity as a measure of apoptosis in both HAC-N and HAC-OA cells (Fig. 3c). However, LOXL2 transduction inhibited caspase 3/7 activity induced by TNF-α in HAC-N but not in HAC-OA cells. Finally, transduction with Ad-RFP-LOXL2, compared to Ad-RFP-EV, increased COL2A1 mRNA levels in both healthy HAC-N and HAC-OA cells treated with TNF-α for 24 h (Fig. 3d). Thus, exogenous expression of LOXL2 does not affect chondrocyte proliferation, but promotes chondrogenic lineage-specific gene expression, increases COL2A1 expression in the presence of TNF-α, and inhibits chondrocyte apoptosis.

LOXL2 overexpression differentially regulates genes in OA chondrocytes

To determine the consequences of LOXL2 overexpression on potential OA-related targets, we performed global gene expression analysis on total RNA extracted from HAC-OA after transduction with Ad-RFP-LOXL2 or Ad-RFP-EV. LOXL2 overexpression resulted in the upregulation of genes such as natriuretic peptide C (NPPC), CSPG4, growth differentiation factor 5 (GDF5), TGFB3, insulin-like growth factor binding protein 5 (IGFBP5), calcium binding protein 5 (CABP5), secreted frizzled-related protein 1 (SFRP1), ACAN, COL2A1, and SOX9 (Fig. 4a, arranged by decreasing FDR value). Exogenous LOXL2 expression reduced the expression of genes (Fig. 4b) such as topoisomerase (DNA) II binding protein 1 (TOPBP1), vascular cell adhesion molecule 1 (VCAM1), E2F transcription factor 8 (E2F8), IL6, Dickkopf WNT signaling inhibitor 1 (DKK1), matrix metalloproteinase (MMP) 1, cyclin E2 (CCNE2), MMP3, antigen identified by monoclonal antibody Ki-67 (Ki67), MMP13, microtubule-associated homolog (TPX2), and N(alpha)-acetyltransferase (NAA)15. These expression changes were statistically significant (Fig. 4c). Validation of the global gene expression data by RT-qPCR analysis (Fig. 4d) showed that the key genes related to anabolic responses such as COL2A1, SOX9, and ACAN were increased after LOXL2 transduction, whereas those related to catabolic responses and chondrocyte hypertrophy such as ADAMTS5, MMP13, COL10, and RUNX2 were unchanged (Fig. 4e). Thus, our gene array analysis suggests that LOXL2 may be involved in promoting chondrogenic lineage maintenance, acting as an anabolic factor that also inhibits catabolic and angiogenic factors.

LOXL2 regulates specific gene sets

In HAC-OA transduced with Ad-RFP-LOXL2 and Ad-RFP-EV, 43 (P value <0.01) and 106 (P value <0.05) gene sets were enriched significantly; only 3 gene sets were enriched significantly at FDR <25% (0.25) (Table 1). LOXL2 transduction compared to the EV control downregulated genes known to be involved in TGF-β1 and TNF-α signaling and apoptosis (Fig. 5a-i); these include reactome signaling by TGF-β1 receptor complex, Biocarta TNF-α receptor 1 pathway, reactome activation genes by ATF4, negative regulation of cell proliferation, regulation of angiogenesis regulation of apoptosis, reactome mitogen-activated protein (MAP) kinase nuclear events mediated by MAP kinases, regulation of IkB kinase NF-ĸB cascade, and signaling by miR-191.

Name	Size	NES	Nom P value	FDR q value
Reactome cell cycle checkpoints	109	−2.478	0.0001	0.00
Reactome activation of genes by ATF4	24	−2.234	0.000	0.00
Reactome SMAD2 SMAD3 SMAD4 heterotrimer regulates transcription	25	−1.880	0.002	0.01
Reactome signaling by TGF Beta receptor complex	59	−1.748	0.0001	0.03
Reactome MAPK targets nuclear events mediated by MAP kinases	30	−1.620	0.010	0.07
ttccgtt, MIR-191	29	−1.604	0.012	0.07
Cell proliferation GO 0008283	495	−1.578	0.000	0.08
Biocarta TNFR1 pathway	29	−1.533	0.033	0.11
Negative regulation of cell proliferation	153	−1.508	0.002	0.12
aagcact, MIR-520 F	238	−1.493	0.000	0.13
Regulation of angiogenesis	26	−1.476	0.061	0.14
Reactome muscle contraction	46	2.023	0.0001	0.19
Regulation of apoptosis	332	−1.373	0.006	0.21
Regulation of cell proliferation	298	−1.371	0.010	0.21
IĸB kinase NF-ĸB cascade	109	−1.36	0.06	0.23
Vasculature development	55	−1.346	0.072	0.24
Reactome platelet calcium homeostasis	16	1.768	0.001	0.24
Reactome antigen activates b cell receptor leading to the generation of second messengers	29	1.750	0.001	0.24
Negative regulation of developmental process	189	−1.335	0.022	0.25

A positive NES indicates that the genes in the set are predominantly upregulated with respect to the reference group, and a negative NES indicates that the genes in the set are predominantly downregulated with respect to the reference group. GSEA gene set enrichment analysis, OA osteoarthritis, FDR false discovery rate, MAP mitogen-activated protein, TGF transforming growth factor

LOXL2 inhibits signaling pathways induced by IL-1β and TGF-β1

We next validated the inhibitory effects of LOXL2 on signaling pathways detected by gene set enrichment analysis (GSEA) (Fig. 5a-i), focusing on those involving NF-κB and ERK. Transduction of HAC-OA with Ad-RFP-LOXL2 inhibited IL-1β-induced phospho-NF-κB/p65 and TGF-β1-induced phospho-ERK1/2 levels compared to cells transduced with Ad-RFP-EV, as shown by western blotting (Fig. 5j and k).

LOXL2 promotes anabolic responses in implants of the human knee and TMJ-OA chondrocytes in vivo and co-expresses with SOX9

To evaluate if LOXL2 can maintain the human cartilage phenotype long term (6 weeks) and induce specific chondrogenic differentiation in vivo, we produced a model of OA using chondrocyte/Matrigel constructs implanted subcutaneously in the backs of nude mice. We used adenoviral delivery of LOXL2 to evaluate the expression in implants of critical genes such as SOX9, ACAN, and COL2A1, which are involved in chondrogenesis. This model is an effective alternative to the human-cell-based models used previously [26–28]: it replicates the 3-dimensional cartilage environment in vivo, uses specific growth factors that preserve the differentiated phenotype for cartilage regeneration [32], and prevents the tissues from becoming fibrotic after 6–9 weeks [33] and myogenic [34]. Weekly injections of Ad-RFP-EV or Ad-RFP-LOXL2 near the chondrocyte-Matrigel implants were performed for up to 6 weeks. IVIS imaging confirmed successful transduction by Ad-RFP-EV or Ad-RFP-LOXL2 (Fig. 6a). After 6 weeks, histological sections of the retrieved implants stained with Safranin O/Fast Green and Alcian blue showed that LOXL2 transduction increased proteoglycan deposition and cartilage formation. However, implants after transduction with Ad-RFP-EV did not show Safranin O or Alcian blue staining, consistent with loss of chondrogenic phenotype and proteoglycan deposition (Fig. 6b). Masson’s trichrome staining was reduced in LOXL2-transduced implants, showing that LOXL2 promotes specific chondrogenic changes and these implants lacked mineralization and fibrosis. In contrast, implants transduced with EV appeared fibrous (Fig. 6b). Molecular analysis of implants by RT-qPCR showed that LOXL2 treatment increased the levels of LOXL2, SOX9, COL2A1, and ACAN mRNA compared to control, whereas MMP13 and ADAMTS5 mRNA levels were not affected significantly in HAC-OA implants (Fig. 6c) and TMJ-OA implants (Fig. 6d). Finally, we evaluated LOXL2 and SOX9 by immunofluorescence staining in HAC-OA and TMJ-OA implant tissue sections. LOXL2 co-localized with nuclear SOX9 (Fig. 7). Adv-RFP-LOXL2 transduction promoted SOX9 gene expression, which was not detected in Ad-RFP-EV-transduced implants, suggesting that LOXL2 maintains the chondrogenic phenotype via SOX9. However, a co-immunoprecipitation assay did not show direct interaction of LOXL2 with SOX9 (Additional file 1: Figure S1). Finally, forced LOXL2 expression resulted in downregulation of phospho-SMAD2/3 and upregulation of COL2A1 compared to Ad-RFP-EV-transduced HAC-OA (Fig. 8a) and TMJ-OA implants (Fig. 8b), which showed lower expression of COL2A1. Thus, LOXL2 transduction may be critical for expression of cartilage-specific matrix genes such as COL2A1 (Fig. 8a, b). Taken together, our results show that LOXL2 induces an anabolic response in HAC-OA and TMJ-OA chondrocytes in vivo and maintains the chondrogenic phenotype, indicating its essential involvement in cartilage formation and maintenance.