Patient-Specific iPSC-Derived Models Link Aberrant Endoplasmic Reticulum Stress Sensing and Response to Juvenile Osteochondritis Dissecans Etiology

Abstract

Juvenile osteochondritis dissecans (JOCD) is a pediatric disease, which begins with an osteonecrotic lesion in the secondary ossification center which, over time, results in the separation of the necrotic fragment from the parent bone. JOCD predisposes to early-onset osteoarthritis. However, the knowledge gap in JOCD pathomechanisms severely limits current therapeutic strategies. To elucidate its etiology, we conducted a study with induced pluripotent stem cells (iPSCs) from JOCD and control patients. iPSCs from skin biopsies were differentiated to iMSCs (iPSC-derived mesenchymal stromal cells) and subjected to chondrogenic and endochondral ossification, and endoplasmic reticulum (ER)-stress induction assays. Our study, using 3 JOCD donors, showed that JOCD cells have lower chondrogenic capability and their endochondral ossification process differs from control cells; yet, JOCD- and control-cells accomplish osteogenesis of similar quality. Our findings show that endoplasmic reticulum stress sensing and response mechanisms in JOCD cells, which partially regulate chondrocyte and osteoblast differentiation, are related to these differences. We suggest that JOCD cells are more sensitive to ER stress than control cells, and in pathological microenvironments, such as microtrauma and micro-ischemia, JOCD pathogenesis pathways may be initiated. This study is the first, to the best of our knowledge, to realize the important role that resident cells and their differentiating counterparts play in JOCD and to put forth a novel etiological hypothesis that seeks to consolidate and explain previously postulated hypotheses. Furthermore, our results establish well-characterized JOCD-specific iPSC-derived in vitro models and identified potential targets which could be used to improve diagnostic tools and therapeutic strategies in JOCD.

Graphical Abstract

Open in new tabDownload slide

Introduction

Juvenile osteochondritis dissecans (JOCD) is an increasingly reported cause of pain and dysfunction among skeletally immature patients (typically ages 6-19¹), and it progresses to early onset osteoarthritis (OA).^2-4 The most common location of JOCD is in the knee. The prevalence of Knee JOCD is estimated between 2.3 and 31.6 cases per 100,000 people.⁵ Increased incidence in recent years is in part due to earlier and increasingly competitive athletic endeavors and improved advanced imaging technology and diagnostic guidelines.⁶ JOCD is thought to initially involve the formation of an avascular lesion in the subchondral bone with secondary effects in the overlaying articular cartilage.⁷ As the disease progresses, JOCD is associated with disruption of joint motion and mechanics, loose body formation, mechanical wear, attrition of associated surfaces on the tibial plateau or the meniscus, and poor bone integrity.^2,8 These manifestations of JOCD significantly increase the risk of early onset OA in the third or fourth decade of patients, with an overall incidence of 0.39.⁹ Currently, the burden of OA on the U.S. healthcare system is expected to increase as the population ages. When OA affects the working or pediatric populations, the burden on lifestyle and the healthcare system is further amplified.

Microtrauma, ischemia, genetic predisposition, and abnormal ossification have all been hypothesized as possible causes of JOCD^7,10; however, the etiology of JOCD is not fully understood and believed to be multifactorial. Previous research has been primarily limited to retrospective clinical studies,¹¹ and no study to date has thoroughly characterized the mechanisms associated with the onset and progression of JOCD.

Reportedly, chondrocytes isolated from detached JOCD fragments (loose bodies) maintain similar cell viability and proliferative activity to those from normal healthy cartilage.^12-14 Nevertheless, a study by Skagen et al. observed that loose body- and JOCD cartilage biopsy-derived primary chondrocytes demonstrated abundant intracellular accumulation of matrix proteins in the endoplasmic reticulum (ER).¹⁴ Intracellular protein accumulation is a pathological feature described in other types of osteochondritis dissecans; namely, familial osteochondritis dissecans (FOCD).¹⁵ Unlike JOCD, however, FOCD is associated with a mutation in the aggrecan gene, which leads to the accumulation of its protein in the ER of FOCD chondrocytes.¹⁵ FOCD is thus considered an endoplasmic reticulum storage disease (ERSD).

In contrast, JOCD is not considered a genetic disease; there is no known associated gene mutation. A recent study found that only 14% of patients have family history,¹⁶ and a preliminary genome-wide associated study failed to identify any SNPs at a genome-wide significance.¹⁷ Moreover, unlike FOCD, which affects every joint and gives rise to short stature,¹⁵ JOCD is not systemic. JOCD patients may have multiple lesions within the same joint or in more than one joint, but skeletal development abnormalities have not been identified. Therefore, it remains unclear whether JOCD is also a disease of the ER, and if so, whether this is a primary or secondary feature of the disease. Protein accumulation leads to ER imbalance and subsequent stress, but the reverse is also true. Cells can accumulate protein intracellularly in response to ER stress, induced due to changes in the internal and external microenvironment of the cells caused by both physiological and pathological conditions.¹⁸ The ER-stress response, referred to as the unfolded protein response (UPR), is activated physiologically by chondrocytes and osteoblasts in the course of their differentiation.^19-21 Additionally, the UPR has been shown to activate transcription of several bone specific genes.²²

Research thus far into JOCD pathology has focused on isolated primary chondrocytes and cartilage tissue, yet this pediatric disorder is hypothesized to initiate in—or, at the very least, involve—the secondary ossification center of the subchondral bone. As JOCD patients are skeletally immature, endochondral epiphyseal ossification is still in progress. Therefore, the objective of this study was to establish JOCD-specific-induced pluripotent stem cell (iPSC)-derived in vitro models, in order to investigate potential pathological processes in this disorder. Given that these are pediatric patients, stem cells were not a viable source. Instead, we generated patient-specific iPSCs from skin biopsies, and these provided us with disease-specific stem cells and the ability to differentiate them along a number of lineages to study different aspects of the same disease. Our study focus was endochondral ossification (EO) processes, as these are highly relevant to the site where JOCD lesions initially develop (epiphyseal ossification), and chondrogenesis, since the formation of an adequate cartilaginous template is a key step in EO. We also developed JOCD ER stress-induction models in order to mechanistically dissect the response to ER stress and protein accumulation in JOCD.

Materials and Methods

iPSCs and iMSCs Generation, Characterization and Validation

Skin biopsies were obtained from 3 patients diagnosed with JOCD of the knee and no known family history (donor demographic information in Supplementary Table S1), in collaboration with the Children’s Healthcare of Atlanta (IRB study #13-098). Nine JOCD-iPSC lines were generated, with 3 clones per patient (Supplementary Table S2). Three previously published^23-25 control (non-disease) iPSC lines were kindly gifted by the Shen lab (National University of Ireland Galway, Ireland). Two additional control iPSCs lines were purchased (ACS-1025 and ACS-1026; American Type Culture Collection (ATCC), VA, USA). Details of iPSC generation, validation and maintenance are presented in Supplementary Data File. Briefly, all 9 JOCD-iPSCs and 5 control-iPSCs were karyotyped and assayed for their gene and protein expressions (qRT-PCR and immunohistochemistry) as well as their ability to differentiate into all three germ layers via in vitro embryonic body formation and in vivo teratoma formation.

iMSCs (iPSC-derived MSCs) were differentiated from iPSCs (Supplementary Table S3) and validated following the International Society for Cell and Gene Therapy’s minimal criteria²⁶ (Supplementary Data File). iMSCs were validated at passages (P)4-6 and used at P6 for all experiments. Every single iMSC line—thus, every iPSC line—was included in every experiment and tested with the indicated number of technical replicates (Supplementary Data File).

iMSC Differentiation Models

Chondrogenic Model

3D chondrogenic modeling of iMSCs (hereon referred to as chondrogenic pellets (CHPs)) was performed using the high cell-density micromass culture model.¹⁵ Briefly, 2 × 10⁶ iMSCs in 10 μl of MSC Maintenance Media (Supplemental Data File) were incubated at 37°C for 2 h in low-adherence round-bottom 96-well plates. Then, 200 μl of MSC Maintenance Medium was added to each well, followed by overnight incubation. Medium was replaced by Chondrogenic Medium (Supplementary Data File) with 10 ng/ml TGF-β3 the following day (day (D)0) and was changed every 2-3 days for 49 days. Sample collection and storage are presented in Supplementary Data File.

Endochondral Ossification (EO) Model

The 3D EO differentiation model of iMSCs (hereon referred to as EO pellets (EOPs)) was carried out as described by Freeman et al²⁷ (Supplementary Fig. S1A). The micromass cultures for the EO model were initiated as per the chondrogenic differentiation model and cultured with Chondrogenic Medium. At D21, medium was replaced by Osteogenic Medium (Supplemental Data File) and changed every second day for an additional 28 days. Samples were collected and stored with the same methods as with CHPs pellets (Supplementary Data File).

Microarray Gene Expression (GE)

For both CHPs and EOPs, total RNA was extracted using RNeasy MinElute Clean-Up Kit with QIAzol Lysis Reagent (Qiagen). Then, normalized RNA was synthesized to cDNA using the RT² First Strand Kit (Qiagen). RNA expression was quantified using the Taqman real-time PCR Fluidigm Dynamic Array Integrated Fluidic Circuits (BioMark, Fluidigm), against the arrays indicated in Supplementary Tables S4 and S5 (Invitrogen). All reactions were carried out in duplicate, and AccuRef human universal cDNA and ultrapure water were used as positive and negative controls, respectively. Samples were distributed randomly in the microarray plates at the time of the runs.

Ct values greater than 37 were treated as 37, namely as null. In order to correct experimental differences observed between biological replicates and runs, the means of each replicate experiment were median-centered with JMP Genomics software (SAS Institute, NC).^28,29 The ΔΔCt method was used to analyze GE. To calculate the relative GE (ΔCt), Ct values were normalized by the housekeeping gene Rps18. Among the housekeeping genes tested, RSP18 was the only one whose raw Ct values were not statistically different (P < 0.05) across different timepoints, differentiation conditions, or between groups. ΔCt value of each target was normalized by the ΔCt of the calibrator to get the ΔΔCt for each gene. For both models of chondrogenic differentiation and EO, the calibrator of each iMSC line was the respective micromass culture at 2 h postseeding (Supplementary Fig. S1A).

Protein Analyses

Immunofluorescence (IF) staining of CHPs and EOPs was carried out for COL1A1 (1:200, Abcam #ab34710) and COL2A1 (1:200, Santa Cruz #52658), as detailed in Supplementary Data File. Proteomic analysis of EOPs was conducted by the Georgia Institute of Technology’s Systems Mass Spectrometry Core Facility (Atlanta, GA). CHPs were further analyzed by transmission electron microscopy (TEM) by the University of Emory’s Robert P. Apkarian Integrated Electron Microscopy Core (Atlanta, GA).

DNA Quantification

CHP and EOP samples were digested overnight in papain at 60°C. Quant-IT Picogreen dsDNA assay kit (Thermo) was then used, according to the manufacturer’s protocol.

Sulfated Glycosaminoglycan (sGAG) Quantification

Dimethylmethylene blue was performed as previously described,¹⁵ with the same papain-digested CHPs samples that were used for DNA quantification. To assess the chondrogenic capability of JOCD- and control-iMSCs during chondrogenic differentiation, sGAG content was normalized to DNA per technical replicate.

Calcium Quantification

Calcium deposition of EOPs was determined by the colorimetric Arsenazo III reagent (Diagnostic Chemicals). To assess the osteogenic capability of JOCD- and control-iMSCs during EO, calcium content was normalized to DNA. As these two assays were not performed from the same pellets, the mean and SD of each assay were calculated separately. Then, the ratio was computed by dividing the mean of the calcium content by the mean of the DNA concentration, for each group. To calculate the SD of the ratio, Taylor’s expansion formula was used with the assumption of independent variables.^30,31

ER Stress-Induction Models

iMSCs were seeded at a 5000/cm² cell density overnight in 6-well plates and with MSC Maintenance Medium. The next day, the medium was replaced by MSC Maintenance Medium plus ER-stressor drug (1× Protein Transport Inhibitor Cocktail (PTIC; eBioscience), 10 µM Thapsigargin (TAG; Sigma-Aldrich), or 5 ng/ml Tunicamycin (TUN; Sigma-Aldrich)), as indicated in Supplementary Data File. Sample collection and storage are presented in Supplementary Data File.

Microarray Gene Expression (GE)

Total RNA was extracted using RNeasy Mini Kit. Then, normalized RNA was synthesized to cDNA using the RT² First Strand Kit. The experiment setup and quantification were performed as described in section “iMSC differentiating models,” with the difference of array indicated in Supplementary Table S6. GE data were also treated and analyzed with the ΔΔCt method. Housekeeping genes were evaluated in each treatment condition. In PTIC treatment, GAPDH and RSP18 were chosen for normalization as their expressions were not statistically different (P-value < .05) among timepoints or in between groups; in TAG treatment, ACTB and RSP18; and in TUN treatment, only RSP18. For ER stress-induction models, the calibrator was the respective iMSC line at 0 h, before stress induction.

Immunofluorescence (IF)

Fixed samples were stained for GRP78 BiP (1:200, Abcam #ab21685) as detailed in Supplementary Data File.

Cytotoxicity Assay

Quantification of LDH in the supernatants of samples treated for 24 h with PTIC, TAG, or TUN was performed using the LDH Cytotoxicity Detection Kit (Sigma-Aldrich #11644793001). The background control was the MSC Maintenance Medium; the low control, which accounts for spontaneous LDH release, was the supernatant from iMSCs cultured in MSC Maintenance Medium (no drug); and the high control, which represented the maximum amount of LDH release, was the supernatant from iMSC cultured in DMSO. Samples were evaluated in technical triplicates, and the experiment was carried out in duplicate. Data were analyzed per the manufacturer’s protocol.

Statistical Analysis

GraphPad Prism v.6.0 was used for all statistical analysis, except for the results from the protein mass spectrometry. Statistical parameters, including the type of tests, number of samples (n), descriptive statistics and significance are reported in the figures and figure legends. Briefly, outliers were identified with the IQR technique. As the hypothesis of the study proposed that JOCD-specific cells are different than control cells, a multiple t-test was used to evaluate differences between these two groups at each timepoint. Statistical significance was set to α = 0.05, using the Holm-Sidak method for correction of multiple comparisons.

For the protein mass spectrometry analysis, the differences in normalized protein profiles between JOCD- and control-EOPs at D49 were analyzed in JMP Genomics software v.9.0 using a volcano plot.²⁸ Significance was determined using Bonferroni correction (.05 (P-value) divided by the number of observations).

Results

Derivation of JOCD- and Control-iPSCs and iMSCs

JOCD- and control-iPSC lines did not have any abnormal chromosome changes due to reprogramming (Fig. 1C). They stained positive for pluripotent stem cell markers (Fig. 1B), and qRT-PCR confirmed the expression of these markers (Fig. 1A). Furthermore, iPSC lines were able to form embryonic bodies, which stained positive for all three-germ layers (Fig. 1D). In vivo, iPSC lines were able to form teratomas, and the three germ layers were distinguishable within them (Fig. 1E).

JOCD- and control-iMSC lines had a fibroblast-like morphology and were adherent to tissue culture plasticware. Additionally, iMSCs successfully passed the approved MSC Surface Marker Screening Panel²⁶ (Fig. 2A) and the Multipotent Differentiation Potential Test (Fig. 2B). Finally, each iMSC line had a downregulated gene expression of pluripotent stem cell markers, compared to its respective iPSC line of origin (Fig. 2C).

JOCD-iMSCs Have a Lower Chondrogenic Capability

During differentiation, the expression of chondrogenic genes was significantly lower in JOCD- than in control-CHPs (Fig. 3A). COL2A1 was expressed significantly lower in JOCD- than control-CHPs at D7, D21, D28, D35, and D49; ACAN, at D28, D35, and D49; COMP, at D7 and D14; FMOD, at D14, D21, and D28; and COL4A6, at D49. RUNX2 was expressed significantly lower in JOCD than in the control at D7, D14, D21 and D28, and SOX9, at D14 and D49.

In contrast, genes related to either hypertrophic chondrogenesis or fibrous cartilage formation were more highly expressed in JOCD- than control-CHPs (Fig. 3A). Expression of COL10A1 was significantly higher in JOCD- than control-CHPs at D49, and that of COL1A1 at D21 and D49. Inflammatory marker IL1β, which negatively regulates the expression of chondrogenic genes,³³ was expressed significantly higher in the JOCD than control group at D21, D35, and D49.

As COL2A1 is a typical marker of differentiated chondrocytes in hyaline cartilage, as opposed to COL1A1, which is associated with fibrous cartilage, the ratio of COL2A1 to COL1A1 can be utilized as a differentiation index to assess the efficiency of chondrogenesis.³⁴ JOCD-CHPs had a significantly lower ratio of gene expressions of COL2A1 to COL1A1 than control-CHPs at D7, D21, D28, D35, and D49 (Fig. 3B). IF staining of COL1A1 and COL2A1 protein confirmed this observation. Although COL1A1-positive staining increased for both groups from D21 to D49, JOCD-CHPs had qualitatively more COL1A1 staining than control-CHPs at both timepoints (Fig. 3D). JOCD-CHPs were also more likely to form a fibrous shell of COL1A1. Furthermore, COL1A1 and COL2A2 stained positively throughout the CHPs with no apparent signs of intracellular protein retention. TEM images of JOCD- and control-CHPs at D49 also showed no signs of protein accumulation in the ER (Fig. 3E). Moreover, JOCD-CHPs had significantly less sGAG content at D49 than control-CHPs (Fig. 3C).

JOCD-iMSCs Have a Different EO Differentiation Profile

JOCD- and control-EOPs had significantly different gene expression profiles during differentiation (Fig. 4A). Expression of COL1A1 was significantly higher in JOCD- than control-EOPs at D49, and that of COL10A1 at D28 and D49. However, expression of RUNX2, SP7 and FMOD was significantly lower in JOCD- than control-EOPs at D28, D35, and D49; and that of MMP13, only at D49.

EOPs did not show significant differences between JOCD and control groups in calcium deposition at D28, D35, and D49 (Fig. 4B). IF co-staining of COL1A1 and COL2A1 proteins showed no apparent differences between the JOCD and control groups in the amount of protein or location within the EOPs (Fig. 4C). At D49, both groups had substantially more COL1A1-positive staining than at D21. Moreover, from the mass spectrometry-based proteomic quantitative analysis, there were also no significant differences in the profiles of protein counts between JOCD- and control-EOPs at D49 (Fig. 4D).

JOCD-iMSCs Express Higher Levels of ER Stress-Related Genes During Chondrogenesis and EO

Unlike other diseases under the osteochondritis dissecans umbrella, which have been linked to protein accumulation and subsequent ER stress-related pathogenesis, JOCD-specific 3D differentiation models showed no evidence of intracellular protein retention. Nevertheless, differentiation processes involve ER stress signaling pathways, and JOCD-iMSCs showed significant differences in differentiation patterns at the gene expression level, compared to control-iMSCs. Therefore, in order to elucidate whether the ER is involved in JOCD pathology, ER-stress response signaling pathways of JOCD and control cells were assessed during differentiation. We evaluated the expression of UPR genes, as well as ER stress-related genes that regulate skeletal differentiation (gene targets listed in Supplementary Table S5).

During chondrogenesis, JOCD-iMSCs expressed higher levels of UPR than control-iMSCs (Fig. 5A and Supplementary Fig. S2A). ER-stress sensors, which initiate UPR, were expressed significantly higher in JOCD- than control-CHPs: EIF2AK3 (common name PERK) at D7, D14, D28, D35, and D49, and ERN1 (common name IRE1α), at D49. EIF2AK3 is also directly involved in the inhibition of protein translation; and ERN1, in ER stress-associated protein degradation (ERAD). EDEM1, which is another marker of ERAD, was significantly higher in JOCD- than control-CHPs at D21, D35, and D49.

Downstream of EIF2AK3, ATF4 was expressed significantly higher in JOCD- than control-CHPs at D7, D14, D21, D28, D35, and D49. DDIT3 (common name CHOP) is downstream of ATF4, and it plays a pivotal role in the induction of ER stress-associated apoptosis (ERAA).³⁵ Even though the expression of DDIT3 was significantly lower in JOCD- than control-CHPs at D7, at the later timepoints of D35 and D49, its expression was significantly higher in JOCD- than control-CHPs.

Total XBP1 expression is related to an increase in protein chaperone synthesis³⁶; which, in turn, increases protein trafficking out of the ER and relieves protein accumulation and, thus, ER stress. XBP1 expression was significantly higher in JOCD- than in control-CHPs at D49 only. Conversely, there was no significant difference in expression between the groups for molecular chaperone HSPA5. Expression of chaperone HSP90B1 was significantly higher in JOCD- than control-CHPs at day 7 but significantly lower at D28 and D49.

Finally, ER stress-related genes directly involved in chondrogenesis and chondrocyte terminal differentiation were upregulated in JOCD-CHPs compared to control-CHPs. ATF3 expression was significantly higher in JOCD- than control-CHPs at D21; GADD45B, at D7, D14, D21, D28, D35, and D49; and CREB3L1 (common name OASIS), at D7 and D49.

During EO, UPR genes were also upregulated in JOCD-EOPs compared to control-EOPs (Fig. 5B and Supplementary Fig. S2B). Stress sensor EIF2AK3 was expressed significantly higher in JOCD- than in control-EOPs at D28, D35, and D49; and stress sensor ERN1, at D28 and D49. EDEM1, ATF4, and DDIT3 were significantly higher expressed in JOCD- than control-EOPs at D28, D35, and D49. Moreover, the expression of total XBP1 was significantly higher in JOCD- than control-EOPs at D49 only. In contrast, HSPA5 was expressed significantly higher in JOCD- than control-EOPs only at D28.

ATF3, CREB3L1, GADD45B, and SERPINH1 are involved both in chondrocyte maturation and osteoblast differentiation. The gene expression of ATF3 and SERPINH1 was significantly higher in JOCD- than control-EOPs at D28 and D49, and that of CREB3L1 and GADD45B, at D28, D35, and D49.

Since DDIT3 is associated with ERAA, we further evaluated whether JOCD-iMSCs, which expressed significantly higher levels of the DDIT3 gene at later timepoints during differentiation, had higher incidence of cell death than control-iMSCs. There were no significant differences in DNA content between the groups in either the chondrogenic or EO models at D49 (Fig. 5C).

JOCD-iMSCs Have a Heightened UPR Response to ER Stress

To elucidate whether JOCD cells sense an elevated ER stress (compared to control cells) and, subsequently, activate a higher ER-stress response signaling, we tested JOCD- and control-iMSCs under the same, controlled conditions of induced ER stress. PTIC treatment yielded significantly higher expression of all UPR genes tested in JOCD-iMSCs compared to control-iMSCs at all timepoints (2, 6, 12, and 24 h post-treatment) (Fig. 6A and Supplementary Fig. S3). TAG treatment resulted in a significantly higher expression of UPR genes in JOCD- than in control-iMSCs for all timepoints, except for ASNS (Fig. 6B and Supplementary Fig. S4). The expression of ASNS was significantly higher in the JOCD than the control group at 2, 12, and 24 h post-TAG treatment. TUN treatment also resulted in a significantly higher expression of UPR genes in JOCD- than control-iMSCs for all timepoints, with the exception of DNAJC10 and ASNS (Fig. 6C and Supplementary Fig. 5). The expression of DNAJC10 was significantly higher in JOCD- than control-iMSCs at 6 and 12 h post-TUN treatment, and that of ASNS, at 2, 6, and 12 h post-TUN treatment.

IF staining of chaperone protein GRP78 BiP (gene HSPA5) confirmed the gene expression results (Fig. 6D). At 24 h post-treatment, there was qualitatively more BiP-positive staining in JOCD-iMSCs treated with either the PTIC, TAG. or TUN than in treated control-iMSCs.

ER Stress Is More Cytotoxic to JOCD-iMSCs

Treatment with PTIC did not result in any cytotoxicity differences between JOCD and control cells. In contrast, 24 h treatment with TAG resulted in significantly higher cell death in JOCD- than control-iMSCs (Fig. 7). Twenty-four hour treatment with TUN also resulted in higher cell death in JOCD- than control-iMSCs, but this difference was not statistically significant.

Discussion

Research limited to retrospective clinical studies and histological findings has translated into limited knowledge about the pathological mechanisms underlying JOCD, which has, in turn, hampered innovation in treatment strategies. In this study, we sought to elucidate possible pathological mechanisms of JOCD using patient-specific iPSC models to study chondrogenesis and EO, and to explore the role of the ER in JOCD pathology.

During chondrogenesis, we observed that JOCD-iMSCs had significantly lower expression of chondrogenic markers and a higher expression of hypertrophic markers than control-iMSCs, which suggests a lower chondrogenic capability in JOCD-iMSCs. This result was supported by the IF co-staining of COL1A1 and COL2A1, which showed more COL1A1-positive staining and fibrous shell formation in JOCD-CHPs, and by the lower sGAG content in JOCD-CHPs. As previous research has found a weak genetic correlation in JOCD patients,¹⁶ we hypothesized that the lower chondrogenic capability of JOCD-iMSCs, compared to control-iMSCs, could be due to regulatory transcriptional changes related to the cellular stress response, which may be initiated physiologically or pathologically. Physiologically, MSCs undergo ER stress in the course of chondrogenesis and EO. Pathologically, there could be protein accumulation in the ER, as it is in the case of FOCD, and, although this accumulation may not rise to the level of visibly engorging the ER (as our TEM results did not show evidence of protein retention), it may be sufficient to influence protein production and affect transcription. When proteins accumulate in the ER, cellular stress is initiated, and response mechanisms act to downregulate these proteins, until such time as the cell can reduce ER stress.

As JOCD arises in the development of the secondary ossification center, investigating the specific processes taking place in EO of JOCD cells was of interest. We observed a lower gene expression of EO orchestrators but a higher gene expression of osteogenic markers in JOCD-EOPs, compared to the control group. Thus, we propose that the EO in JOCD-iMSCs is not necessarily impaired, but that it is different; and that there may be compensatory pathways which deliver an end product that is adequate in terms of bone formation. For example, RUNX2 is a master orchestrator of EO, and it regulates the expression of osteocalcin, an important bone protein, through the activation of ATF6.^37,38 However, ATF4 is also able to regulate the production of osteocalcin.³⁹ In our results, we observed significantly lower gene expression of RUNX2 in JOCD-EOPs and no differences in the expression of ATF6; but the gene expression of ATF4 was significantly higher in JOCD- than control-EOPs, which suggests that osteocalcin production may still take place at appropriate concentrations. In support of this hypothesis, we did not observe any significant differences between JOCD- and control-EOPs at the protein level or in calcium production.

In order to determine whether the observed differences in chondrogenesis and EO were indeed linked to cellular stress and whether this stress was induced physiologically or pathologically, we studied the expression profiles of ER stress-related genes of JOCD- and control-iMSCs during differentiation. Differentiating JOCD-iMSCs had a higher expression of core UPR markers and other ER stress-related markers involved in chondrogenic and osteogenic differentiation, than differentiating control-iMSCs. However, it was unclear whether the observed aberrant UPR was due to JOCD cells experiencing different stresses during differentiation or responding differently to the same stress since cell differentiation is a highly variable process. Even under the same differentiation-promoting conditions, there is known donor-to-donor variability and also clone-to-clone variability within the same iPSC donor.^40-43 Therefore, it was necessary to investigate the UPR of JOCD-iMSCs in a controlled environment. ER stress was, thus, induced with 3 independent ER stressors: PTIC, TAG, and TUN. Our results indicated that JOCD-iMSCs had indeed an elevated (vs. control-iMSCs) UPR activation response to the same ER stressors, suggesting that JOCD-iMSCs are more susceptible, and thereby exhibit a heightened response, than control-iMSCs to the same cellular stresses. To support this observation, we found that ER stress was more cytotoxic to JOCD- than control-iMSCs.

In our ER stress-induction models, JOCD cells experienced elevated ER stresses and had a corresponding elevated ER-stress response, which was observed throughout the entire UPR signaling under study. However, in our differentiation models, JOCD cells’ strategy for coping with the heightened ER stress being experienced, was to increase different UPR signaling cascades at different timepoints. We, thus, propose that JOCD cells undergo chondrogenesis and EO in a different manner to control cells and that these differences are directly related to both the heightened ER stresses experienced by JOCD cells and their unique response to such stresses.

This study is the first, to the best of our knowledge, to identify and attribute differences in ER-stress sensing and the associated ER-stress response as part of JOCD pathology; and to provide preliminary evidence of hitherto unknown pathomechanisms involved in JOCD etiology. We propose that the currently postulated etiologies (local ischemia, microtrauma and abnormal ossification) constitute conditions in which the microenvironment is disturbed and cellular stress is induced; and that the dysregulated UPR in JOCD cells is the underlying cause for the onset of this disorder. Ischemia is a pathological stress, and upon microtrauma, which can cause interruption of the microvasculature supply, JOCD tissue may be less tolerant of the temporary ischemia than normal tissue, in part due to the heightened cellular stress to an already “higher-than-normal” ER stress baseline in JOCD cells. Failure to cope with a heightened ER stress over prolonged periods of time may lead to ERAA in JOCD cells, thus creating the initial necrotic lesion observed in JOCD patients. This hypothesis could be further tested by studying our chondro- or osteo-organoids under ischemic conditions (low glucose or a glucose-blocker, and low oxygen). In fact, any local disruption of homeostasis during chondrogenesis or EO may introduce additional ER stresses that would be able to, at the very least, alter the differentiation process locally. Therefore, reported cases of focal abnormal ossification may, in fact, be a consequence of the aberrant ER-stress sensing and response of JOCD cells. Our concluding hypothesis, therefore, provides an explanation for the previously accepted multifactorial etiology and local nature of JOCD and accounts for the observed cases of multiple JOCD lesions on the same joint or multiple affected joints.

Furthermore, this study raises two questions: first, if chondrogenic capability is indeed lower in JOCD cells, then why do JOCD patients not show any skeletal abnormalities? In vitro cell culture—and, by extension, in vitro differentiation—is known to induce additional cellular stress on cultured cells than in vivo.⁴⁴ A significant finding in this study is that JOCD cells experience and respond to cellular stresses differently than control cells. Therefore, the cellular stress induced by regular cell culture practices, in addition to the stress induced by differentiation, may accentuate the observed differences in vitro compared to in vivo. Second, if JOCD cells are more susceptible to ER stress, then why is JOCD a musculoskeletal disorder and not observed in other organ systems? To decisively answer this question, the ER-stress response of other JOCD-cell types such as hepatocytes, cardiomyocytes, and so forth would need to be evaluated. We hypothesize that aberrant ER-stress sensing and response are only features of the lateral plate mesoderm and not an issue—at least not significantly—in other cell lineages in JOCD.

The question remains whether there is a genetic component in JOCD, epigenetic or both. While genetic studies have been inconclusive, candidate loci have been identified for further investigation.¹⁷ We believe that JOCD may be either a polygenic disease involving more than one precipitating genetic mutation and an interplay of various combinations of precipitating factors, or purely epigenetic. In all cases, our results demonstrate that JOCD onset is at least partially caused by environmental cues. In the present study, we aimed to replicate the appropriate cellular microenvironment to stimulate and elucidate the native epigenetic response of JOCD cells. Modeling polygenic and sporadic diseases remains a challenge due to iPSC’s inherent immaturity problems and lack of their respective native environmental cues.⁴⁵ For these reasons, our approach included differentiating JOCD and control iPSCs across an array of cell types (iMSCs, chondrogenic and osteogenic cells), replicating biological processes as closely as possible (3D chondrogenesis and endochondral ossification models), and purposely inducing ER stress to mimic our hypothesized natural-occurring environmental cues. In this regard, we can conclude that some epigenetic response was indeed stimulated by modulating the cellular microenvironment; thus, producing the appropriate JOCD iPSC-based disease models.

Limitations of the study include its restriction to in vitro assays. In vitro models may diverge in differentiation pathways due to a number of factors, such as cell source variability, composition of fetal bovine serum, etc. In order to compensate for arbitrary differences due to cell source as well as any clonal variations, we used three independent clones, which may be considered as independent biological copies, for each JOCD patient; totaling 9 JOCD lines. Furthermore, the control iPSC lines used in this study are well-established (and published) lines and were matched as closely as possible to the characteristics of our JOCD patients. To mitigate any transcriptomic or epigenetic differences resulting from culture conditions,⁴⁶ all iPSC lines were grown, validated, and differentiated under the exact same conditions, in the same lab. Yet, a major limitation of the study still remains our pool of patients. While it is common to use four to six experimental samples in studies of polygenetic and sporadic iPSC disease models,^{23–25,45,47,48} we recommend testing the hypotheses proposed by this study on larger sample sizes and more complex models of ossification. The development of the secondary ossification center is far more complex, involving blood vessels and osteoclasts, which we have not been able to fully recapitulate in this study. Lastly, it is worth pointing out that, despite successfully fulfilling all MSC criteria, iMSCs are not MSCs, but derived from iPSCs, which may introduce additional noise into our in vitro models.⁴⁹

Conclusion

Due to the degenerative changes caused by JOCD, JOCD patients can develop OA as early as 25 years old. Treatment for OA is already a heavy burden on the healthcare system, and younger patients with OA will unfortunately not only add to the direct cost of a lifetime of surgeries, but also to the indirect cost of disability. For the first time, by utilizing patient-specific iPSC-derived in vitro models, we are able to propose causative pathological mechanisms for JOCD which could explain and account for previously postulated hypotheses, as well as provide the field with well-characterized platforms for testing therapeutic interventions. The present study is the first to highlight the important role of cells of mesoderm origin (and their differentiating counterparts) on the onset of JOCD, and these preliminary findings may pave the way to novel diagnosis and treatment of this disorder. Furthermore, this study demonstrates the versality of iPSC-based disease modeling in a range of study conditions utilizing chondrogenesis, osteogenesis, and mesenchymal stem culture.

Acknowledgements

The authors would like to acknowledge Syed (Mohib) Hasnain for his assistance with cell culture and sectioning of histological samples; Dalia Afarat, with microarray gene expression experiments; David Smalley, with the analyses of mass spectrometry data; and Dr. Giuliana D. Noratto and Dr. Greg Gibson, with the analyses of gene expression data and statistics. The authors would also like to acknowledge Dr. Greg Gibson for his scientific guidance throughout the project and the staff at the Children’s Healthcare of Atlanta (CHOA) Orthopaedics and Sports Medicine unit for their help in patient sample collection. Additionally, the authors acknowledge the Genomics Core, the Histology Core, the Cellular Analysis and Cytometry Core and the Systems Mass Spectrometry Core at the Georgia Institute of Technology (Atlanta, GA, USA), as well as the Robert P. Apkarian Integrated Electron Microscopy Core at Emory University (Atlanta, GA, USA) and the Flow Cytometry Facility and the Centre for Microscopy and Imaging at the National University of Ireland Galway (Galway, Ireland).

Funding

This study was supported by the National Science Foundation Graduate Research Fellowship Program (no. DGE-1148903), the National Science Foundation/Science Foundation Ireland Graduate Research Opportunities Worldwide Grant (SFI no. 13/RC/2073s1), and a research partnership between Children’s Healthcare of Atlanta and the Georgia Institute of Technology.

Author Contributions

G.E.S.N.: conception and design, collection and assembly of data, data analysis and interpretation, financial support, manuscript writing, final approval of manuscript; C.C.N.: collection and assembly of data, data analysis and interpretation; H.Y.S.: data analysis and interpretation, administrative support, manuscript writing; M.X. and K.McD.: other (knowledge transfer); S.G.: collection and assembly of data, other (knowledge transfer); C.D. and N.N.: collection and assembly of data; S.S. and S.C.W.: provision of study material and patients; F.B. and R.E.G.: conception and design, financial support, final approval of manuscript.

Conflict of Interest

S.C.W. declared advisory role with Smith & Nephew Endoscopy, Vericel, Arthrex. F.B. declared advisory role with eQcell Inc.; Bio-Recell Ltd. and stock ownership with Orbsen Therapeutics Ltd. All of the other authors declare no potential conflicts of interests.

Data Availability

All data associated with this study are available in the main text or the supplementary materials. Further information and requests of materials can be directed to and fulfilled by the corresponding authors, Dr. Robert E. Guldberg ([email protected]) and Dr. Frank Barry ([email protected]). Information specifically regarding the control iPSCs should be directed to Dr. Sanbing Shen ([email protected]). MTAs required for cell lines.

References