Step-Wise Chondrogenesis of Human Induced Pluripotent Stem Cells and Purification Via a Reporter Allele Generated by CRISPR-Cas9 Genome Editing

Abstract

The differentiation of human induced pluripotent stem cells (hiPSCs) to prescribed cell fates enables the engineering of patient-specific tissue types, such as hyaline cartilage, for applications in regenerative medicine, disease modeling, and drug screening. In many cases, however, these differentiation approaches are poorly controlled and generate heterogeneous cell populations. Here we demonstrate cartilaginous matrix production in three unique hiPSC lines using a robust and reproducible differentiation protocol. To purify chondroprogenitors produced by this protocol, we engineered a COL2A1-GFP knock-in reporter hiPSC line by CRISPR-Cas9 genome editing. Purified chondroprogenitors demonstrated an improved chondrogenic capacity compared to unselected populations. The ability to enrich for chrondroprogenitors and generate homogenous matrix without contaminating cell types will be essential for regenerative and disease modeling applications.

Grphical Abstract

Introduction

Chondrocytes produce and maintain the hyaline articular cartilage that lines the surface of diarthrodial joints. Conditions that lead to articular cartilage degeneration, such as arthritis, arise from dysregulated chondrocyte homeostasis, resulting in chondrocyte apoptosis, hypertrophic differentiation of articular chondrocytes, and aberrant activation of inflammatory cascades [1–3]. Ultimately, these changes lead to pain and loss of joint function that necessitates total joint replacement with an artificial prosthesis [4]. Many cell-based strategies for cartilage repair and regeneration are designed to preserve the joint [5–9]. Human induced pluripotent stem cells (hiPSCs) are promising cell sources for cartilage regenerative therapies and in vitro disease-modeling systems due to their pluripotency and unlimited proliferation capacity [10,11]. Furthermore iPSCs provide a means of developing patient-specific or genetically-engineered cartilage for applications in drug screening for disease-modifying osteoarthritis drugs (DMOADs) [12]. Therefore, methods to rapidly and efficiently differentiate hiPSCs into chondrocytes in a reproducible and robust manner will be critical for their application in joint repair.

An important goal of such protocols is to minimize variability in hiPSC differentiation potential, which may arise from characteristics of the donor and/or reprogramming method [13–15]. Although several approaches have been employed to generate articular chondrocytes from PSCs, most differentiation protocols have been based on trial-and-error delivery of growth factors without immediate consideration of the signaling pathways that direct and inhibit each stage of differentiation. Accordingly, chondrogenic differentiation is often dependent on the specific cell lines used, and broad application of iPSC chondrogenesis protocols has not been independently demonstrated with multiple cell lines [16].

Recently, critical insights from developmental biology have elucidated the sequence of inductive and repressive signaling pathways needed for PSC lineage specification to a number of cell fates [17–19]. By reproducing these reported signaling pathways in vitro, in combination with existing chondrogenic differentiation approaches, we sought to establish a rapid and highly reproducible protocol for hiPSC chondrogenesis that is broadly applicable across various hiPSC lines. Furthermore, as hiPSC differentiation processes are inherently “stochastic” and can often produce heterogeneous cell populations over the course of differentiation, we hypothesized that purification of committed chondroprogenitors would improve hiPSC chondrogenesis. Previously, mouse iPSC-derived chondroprogenitors (CPs) were enriched by using cells isolated from a transgenic mouse carrying a reporter gene at the Col2a1 locus, which encodes for type II collagen – an important structural constituent of articular cartilage [20–22]. However, this transgenic approach is not feasible for human iPSCs, which motivated our pursuit of gene editing methods to create a knock-in reporter of collagen II production at the human COL2A1 locus.

In this study, we demonstrate the development and application of a step-wise differentiation protocol validated in three unique and well-characterized hiPSC lines. We examined gene expression profiles and cartilaginous matrix production during the course of differentiation. To further purify committed CPs, we then used CRISPR-Cas9 genome engineering technology to knock-in a GFP reporter at the collagen type II alpha 1 chain (COL2A1) locus to test the hypothesis that purification of CPs could enhance articular cartilage-like matrix production. This type of approach for purification of hiPSC-derived CPs in conjunction with a reproducible differentiation protocol may facilitate the development of in vitro chondrogenesis platforms for disease modeling and drug screening.

Results

Step-wise differentiation of hiPSCs into chondroprogenitor cells

To establish a standardized protocol for hiPSC chondrogenesis, we optimized growth factor and small molecule concentrations using established principles of PSC differentiation along mesodermal lineages as described in Figure 1A [17]. To validate our differentiation approach, we measured expression of transcription factors representative of various stages of development with qRT-PCR and monitored cell morphology at multiple time points in three hiPSC lines (BJFF, ATCC, and RVR) (Figure 1B-G and S1). Over the course of differentiation, we observed a gradual decrease in expression of the pluripotency markers octamer-binding protein 4 (OCT4) and nanog homeobox (NANOG) (Figure 1B). Stage-specific expression of transcription factors was observed between days 1–3. At day 1, expression of an anterior primitive streak (PS) marker, mix paired-like homeobox (MIXL1), spiked in each cell line (Figure 1C). Subsequently, a spike in mesogenin 1 (MSGN1) expression at day 2 marked paraxial mesoderm specification, and early somitogenesis initiated at day 3, as evinced by upregulation of transcription factor 15 (PARAXIS) (Figure 1D-E). Interestingly, SRY-box 9 (SOX9) expression was highly elevated at both paraxial mesoderm and sclerotome stages as compared to hiPSC stage for ATCC and RVR lines (Figure 1F). Furthermore, platelet derived growth factor receptors (PDGFR)α and PDGFRβ gradually increased their expression levels over the course of mesodermal differentiation but PDGFRα was down-regulated, whereas PDGFRβ maintained up-regulated, in chondroprogenitor (CP) cells (Figure 1F). The CP cells also expressed the highest levels of collagen type II alpha 1 chain (COL2A1) but only a moderate increase of aggrecan (ACAN) as compared to hiPSCs for all three cell lines examined (Figure 1G).

Figure 1. Chondrogenesis of hiPSCs through the Paraxial Mesoderm Lineage.

(A) Stepwise of differentiation of hiPSCs toward of paraxial mesoderm and chondrocyte-like cells. Gene expression levels of major transcriptional factors for (B) Pluripotency, (C) Anterior PS, (D) Paraxial mesoderm, (E) Early somite, (F) Sclerotome, and (G) Chondroprogenitor cells. (H) Heat map of representative surface marker expression of hiPSC and Chondroprogenitor (CP) cells of RVR cell line. (B-G) n = 3–4 experiments per cell stages per cell line (N=3 lines). Mean ± SEM. One-way ANOVA for each cell line using differentiation stage as factor. *p < 0.05 and †p < 0.01 compared to hiPSC stage. For MSGN1 expression level at paraxial mesoderm stage, only the RVR-iPSC line is significantly different from its hiPSC stage. For ACAN expression level at chondroprogenitor stage, only RVR is not significantly different from its hiPSC stage. PS: primitive streak. Data points represent means and error bars signify SEM.

Characterization of surface markers of hiPSCs and CP cells

Upregulation of chondrogenic markers in CPs suggested that this stage may be appropriate for further chondrogenic differentiation. The RVR-iPSC line was assessed for surface marker expression levels at iPSC and CP cell stages (Figure 1H). The hiPSCs and CP cells exhibited distinct expression patterns of surface proteins (Table S1). hiPSCs exhibited a surface marker profile characteristic of primed hiPSCs (CD90⁺/CD24⁺/SSEA-4⁺/CD57⁺/CD45⁻). Pluripotency-specific markers SSEA-4/CD57/CD24 decreased in CP cells, and the CP cells displayed a moderate increase in the surface markers CD105, CD146, CD166, and CD271. Interestingly, we also observed that CP cells expressed surface proteins that are often absent on mesenchymal stem cells (MSCs) such as CD56, CD111, CD112, and CD117.

Chondrogenic gene expression during chondrogenesis

After mesodermal specification and pre-chondrogenesis of hiPSCs, we further differentiated cells in chondrogenic pellet cultures with TGF-β3 supplementation. At days 7, 14, and 28, we assessed chondrogenic gene expression. Chondrogenic markers, particularly SOX9 and COL2A1, were significantly up-regulated at early stages of chondrogenesis, while the expression of ACAN was not increased until 14 days post-induction of chondrogenic differentiation (Figure 2A-C). These findings suggested that COL2A1 could serve as critical marker and purification criterion for chondrogenic cells. These three chondrogenic markers remained highly elevated at day 28 of pellet culture as compared to the hiPSC stage. We also observed that both hypertrophic chondrocyte marker type X collagen and fibroblast matrix gene type I collagen were increased at day 28 of chondrogenesis, consistent with chondrogenic maturation (Figure 2D-E).

Figure 2. Pellet Culture Time Course of Unpurified CPs.

Gene Expression Profiles of (A)SOX9, (B) COL2A1, (C) ACAN, (D) COL1A1, and (E) COL10A1 and (F) GAG/DNA ratio of the pellets at various time points during 28 day chondrogenic pellet culture at various time points during 28 day chondrogenic pellet culture. n = 3–4 each time point per cell line. Mean ± SEM. One-way ANOVA for each cell line using differentiation stage as factor. *p < 0.05 and †p < 0.01 compared to hiPSC stage.

Histological and biochemical analyses of chondrogenesis

Safranin-O and hematoxylin staining of day 28 pellets of each cell line showed high amounts of proteoglycan production and the presence of round chondrocyte-like cells within the matrix (Figure 3). Additional biochemical analysis of pellets during chondrogenesis demonstrated increasing GAG/DNA ratio over time for each cell line. The ATCC-iPSC line produced less matrix compared to RVR- and BJFF-iPSC lines (Figure 2F).

Figure 3. Type II Collagen and Glycosaminoglycan Content of Unpurified CP Pellet Culture.

Representative safranin-O (Saf-O) and Type II collagen staining images of day 28 pellets for (A) BJFF (B) ATCC, and (C) RVR cell line. (D) Human osteochondral tissue was used as positive control for both staining. The area of the white dash-square was magnified to reveal chondrocyte-like cell morphology within the pellet for each cell line. Scale bar = 0.4 mm and 0.2 mm for left and right panel, respectively.

Generation and validation of a COL2A1-GFP knock-in reporter line

While successfully validating a reproducible chondrogenic differentiation protocol applicable to multiple hiPSC lines, we observed some formation of non-cartilaginous tissue from pellet culture systems within each cell line, attesting to heterogeneity inherent in PSC differentiation (Figure 3A). Therefore, we sought to further improve cartilaginous matrix production by purifying CP cells prior to chondrogenic differentiation. Using genome editing with CRISPR-Cas9, we integrated an EGFP reporter sequence downstream of the COL2A1 reading frame, separated by a P2A ribosomal skipping peptide sequence, in RVR hiPSCs (Figure 4A). Co-integration of a floxed puromycin resistance cassette allowed enrichment of successfully targeted hiPSCs, and subsequent excision of this cassette left a single loxP scar upstream of the 3’ untranslated region (UTR). Clonal populations were isolated from single cells and successful targeting of a 2A-EGFP sequence at the 3’ end of the COL2A1 reading frame was confirmed via junction PCR and Sanger sequencing (Figure S3).

Figure 4. CRISPR-Cas9 Mediated Knock-in of EGFP Reporter at the COL2A1 Locus in hiPSCs.

(A) Schematic depicting generation of the reporter line. A gRNA overlapping the COL2A1 stop codon directed Cas9 to generate a double-strand break, which was repaired with a targeting vector containing the 2A-EGFP reporter and a puromycin resistance cassette (PURO^R) flanked by ~700 bp homology arms upstream and downstream of the stop codon. After selection of clones with Puromycin, a CRE recombinase expression plasmid was transfected to excise the PURO^R cassette, leaving a residual loxP scar. (B) Brightfield and fluorescence imaging of reporter line after SOX9-mediated activation of COL2A1. Scale bar = 100 μm (C) FACS sorting of GFP+/− populations after COL2A1 activation in the reporter line and unmodified controls. (D) Fold-change of COL2A1 expression in unmodified, reporter, GFP+, and GFP-populations relative to expression in hiPSC controls. (E-G) Fold-change of chondrocyte markers (E) ACAN, (F) TRPV4, and (G) COL9A1 in unmodified, reporter, GFP+, and GFP-populations relative to expression in hiPSC controls. Groups not sharing the same letter are statistically different by one-way ANOVA (p<0.05). Error bars represent mean ± SEM (n=3 independent experiments). LHA, RHA: Left Homology Arm, Right Homology Arm

To functionally validate reporter activity, we activated endogenous COL2A1 by SOX9 lentiviral overexpression. Fluorescence imaging revealed clusters of bright round GFP-positive cells after 21 days, comprising roughly 4% of the total population (Figure 4B and 4C). Reporter and unedited RVR lines exhibited comparable activation of COL2A1 expression, suggesting that the addition of 2A-EGFP and a loxP scar upstream of the 3’ UTR did not significantly alter SOX9-dependent COL2A1 activation (Figure 4D). The GFP+ population exhibited significant enrichment (40-fold) of COL2A1 transcript over unsorted cells; conversely, the GFP-population exhibited decreased levels of COL2A1 transcripts (Figure 4D). Similar trends were observed for expression of other chondrogenic markers such as ACAN, TRPV4, and COL9A1 (Figure 4E-G).

Application of optimized directed differentiation protocol to the COL2A1 reporter line

After validating the utility of the reporter in purifying cell populations enriched for COL2A1 expression, we applied our directed differentiation protocol to the reporter line and observed reporter activation over time (Figure 5A). At D0 of differentiation, the COL2A1-GFP reporter line shows no fluorescence when compared to the unmodified line, suggesting the lack of leakiness in reporter activity (Figure 5A). GFP fluorescence in the reporter line becomes evident as early as D3, during differentiation to paraxial mesoderm and the GFP florescence intensity continues to increase at D6, finally becoming a distinct GFP+ population by D9 and 12, at the CP cell stage as described previously. At D15 we begin to observe the emergence of a brighter GFP population, which clearly bifurcates into two distinct GFP populations (high and low) at D21 (Figure 5A, Figure 6A). Mean fluorescence intensity (MFI) of the overall population increased steadily over time (Figure S4). To characterize the sensitivity of the reporter in the context of directed differentiation, we sorted various GFP populations: GFP⁻, GFP⁺, GFP^bottom (bottom 10^th percentile of GFP⁺), and GFP^top (top 10^th percentile of GFP⁺), and assayed COL2A1 transcript levels compared to the unsorted population at multiple time points (Figure 5B). Consistent with the intensifying GFP signal over time, we observed a steady increase in COL2A1 in the unsorted population. GFP⁻ and GFP^bottom populations, gated at a constant MFI, exhibited steady levels of COL2A1 transcript over time (~100 and 200-fold greater than iPSCs, respectively). Cell sorting allowed for isolation of cells with significantly greater COL2A1 transcript levels in the GFP⁺ and GFP^top populations compared to unsorted cells (2.6 and 1.2 fold at day 15; 6.4 and 2.2 fold at day 21, respectively) (Figure 5B).

Figure 5. Time course of Directed Differentiation Applied to Reporter Line.

(A) FACS sorting of GFP^+/− population at several time points. Beginning at D6 the top (GFP^top) and bottom (GFP^bottom) 10^th percentiles of GFP+ cells were sorted (shaded orange and blue, respectively). (B) Differential COL2A1 transcript enrichment in various GFP populations over the course of differentiation. (C) Scatter plot showing correlation of COL2A1 transcript enrichment and change in mean fluorescence intensity (MFI) in GFP^top samples compared to MFI of undifferentiated population (r² > 0.91, p<0.0001). Curves peripheral to trendline represent 95% confidence intervals. (D,E) Differential ACAN and SOX9 transcript enrichment in GFP populations over the course of differentiation. Two-way ANOVA with Tukey’s post hoc comparison for each chondrogenic marker demonstrated significance along both time and population factors (p<0.05). Asterisks (*) represent significant enrichment (p<0.05) of GFP^high compared to each other population at a given time point. Error bars represent mean ± SEM (n=3 independent experiments).

Figure 6. Evaluation of Chondrogenic Differentiation Capacity of Purified Chondroprogenitors.

(A) Schematic of Experimental Design. D21 GFP^high/low/mid populations were isolated and cultured as pellets or seeded for expansion. Pellets were generated from expanded cells at each passage up to 5 passages. (B) Safranin-O/Fast Green stain of pellets from unsorted and GFP populations after FACS (P0) and 1 or 2 passages (P1 and P2, respectively) Scale bar (200 μm) (C) Biochemical analysis of GAG content normalized to DNA content in unsorted and GFP populations. Two-way ANOVA with Tukey’s post hoc comparisons for each chondrogenic marker demonstrated significance along both time and population factor (p<0.05). Asterisks (*) represent significant enrichment (p<0.05) of GFP^high compared to each other population at a given passage. Error bars represent mean ± SEM (n=3 independent experiments). (D-G) Type II Collagen (D), Type VI Collagen (E), and Type I Collagen (F), and Type X Collagen (G) immunohistochemistry (IHC) of pellets derived from unsorted cells and GFP^high cells at P0-P5. Scale Bar (200 μm)

To assay the utility of reporter fluorescence to serve as a surrogate for COL2A1 expression, we compared the MFI and COL2A1 fold-enrichment of the GFP^top population at each measured time point. MFI of GFP^top cells was highly correlated to COL2A1 transcript enrichment (r²=0.91) (Figure 5C). Taken together, these results suggest that our knock-in reporter is sensitive to relatively small changes in COL2A1 expression and functions with high fidelity over a wide dynamic range.

GFP⁺ and GFP^top populations also exhibited significant enrichment of other chondrocyte-associated transcripts, with robust ACAN activation and steady up-regulation of SOX9 from D12 to D21 (Figures 5D and 5E). We also examined expression of Runt-related transcription factor 2 (RUNX2), a marker for hypertrophic chondrogenic or osteogenic differentiation, during the course of differentiation (Figure S4). Although we observed enrichment of RUNX2 in the GFP^top at D21, we did not observe activation of its downstream targets including Matrix Metallopeptidase (MMP)-13 and COL10A1 (Figure S4). These findings imply that increased RUNX2 expression by itself may not indicate hypertrophic or osteoblastic differentiation in the GFP^top cells. Moreover, transcripts of COL1A1, a fibroblastic and osteogenic marker, were significantly depleted in the GFP^+/top compared to unsorted controls (Figure S4).

Purification of chondroprogenitors improves cartilaginous matrix production

Given the generation of two distinct GFP populations at D21 (termed GFP^low/high) of monolayer differentiation (Figure 6A) and differential expression of chondrogenic markers within these populations (Figure 5B, D, and E), we selected this time point to isolate committed CP cells. GFP^low and GFP^high populations, along with an intermediate GFP^mid population, were sorted and either directly used for chondrogenic pellet culture (P0 pellets) or expanded up to 5 passages prior the induction of chondrogenesis (P1-P5 pellets) (Figure 6A). We evaluated pellets after 28 days of culture histologically and biochemically (Figures 6B-E). Pellets created from GFP^high cells at all passages (P0-P5) were larger than those created from unsorted and GFP^low/mid populations, showed homogeneous matrix rich in glycosaminoglycan (GAG) and were void of other tissue types (Figure 6B and S5A). Biochemical analysis confirmed the histology with significantly higher GAG/DNA ratio in the GFP^high pellets compared to unsorted ones. Interestingly, chondrogenesis of GFP^mid cells improved with expansion, suggesting that this population may have been a less differentiated population compared to GFP^high. Meanwhile GFP^low populations failed to produce cartilaginous matrix from P0-P5 (Figure 6C). In addition to GAG content, normalized collagen content was increased in GFP^high pellets at P1 and P2 (Figure S5B).

While the presence of type I collagen predominates in fibrocartilage, hyaline articular cartilage is characterized by the presence of type II collagen and the lack of type I collagen. Immunohistochemistry analysis of collagen composition demonstrated enrichment of type II/VI collagen in P0 pellets generated from GFP^high cells compared to the unsorted cells (Figure 6D,E). P0 pellets from unsorted cells stained diffusely for type I collagen, and this was limited to the periphery of the tissue from GFP^high P0 pellets (Figure 6F). No immunostaining of type X collagen was observed from either unsorted or GFP^high pellets (Figure 6G). Together, this suggests that purification of GFP^high cells specifies the generation of matrix with qualities similar to hyaline cartilage rather than hypertrophic or fibrocartilage. Pellets derived from passaged CPs (P1-P5) had less type II/VI collagen accumulation as compared to P0 pellets, but we did observe modest enrichment of type II collagen up to P3 in GFP^high pellets (Figure S5C). In addition, late passage GFP^high pellets (P3-P5) exhibited increased type I collagen accumulation compared to unsorted controls beginning at P3 (Figure S5E). Type X collagen became apparent at the periphery in P3-P5 unsorted pellets, but pellets derived from the GFP^high population did not stain for type X collagen (Figure S5F).

Discussion

Articular chondrocytes develop in vivo from a group of mesenchymal cells derived from mesoderm and ectoderm [19,23–25]. While craniofacial cartilage is generated from the ectoderm-derived neural crest, articular cartilage of the appendicular and axial skeleton is derived from paraxial and lateral plate mesoderm, respectively. Recent advances in developmental biology and the establishment of hPSC differentiation “roadmaps” can be applied to the in vitro generation of various musculoskeletal tissues through the specification of different mesodermal lineages [6,17,19]. While it is known that chondrocytes can be derived from paraxial mesoderm during embryonic development in vivo, the reproducible differentiation of hiPSCs to chondroprogenitors (CPs) has remained a major challenge to cartilage tissue engineering and arthritis disease modeling. Here, we describe a highly robust, stepwise protocol for in vitro chondrogenesis through the paraxial mesodermal lineage that is consistent among multiple hiPSC lines and further validated using a knock-in collagen II reporter system.

Our approach successfully recapitulates mesodermal differentiation, as evidenced by the sequential stage-specific up-regulation of mesodermal master transcription factors including MIXL1, MSGN1 and PARAXIS. Interestingly, the findings of down-regulation of PDGFRα as well as SOX9, but not PDGFRβ, in chondroprogenitor (CP) cells in response to BMP-4 treatment suggests possible interactions among PDGFRα, SOX9, and BMP signaling pathways during CP cell commitment. Indeed, PDGFRα expression level has been reported to be involved in regulating BMP-Smad1/5/8 signaling and induction of osteogenesis in MSCs [26]. Furthermore, using conditional deletion animal models, Lim et al. recently demonstrated that BMP-Smad4 signaling controls pre-cartilaginous mesenchymal condensation in a SOX9 independent manner [27].

CP cells expressed high levels of CD146 and CD166 but moderate to low levels of CD271, CD73, CD44, and stage specific embryonic antigen (SSEA)-4. These results suggest that the CP cells derived via our differentiation protocol are distinct from MSCs, which generally express these markers (CD271, CD73, CD44, SSEA4) [28]. Moreover, our finding of low CD44 levels in the CP cells is consistent with the results of a recent study showing that CD44 is expressed at a low level in pre-chondrocytes as compared to mature chondrocytes [29]. However, the results of surface marker analyses should be interpreted with caution, as surface markers do not necessarily reflect cell function and as stem and/or progenitor cells can be highly plastic, and lineage-origin dependent. [30,31]

The cartilaginous matrix produced by the P0 pellets showed enriched staining of type II/VI collagen and GAG, but was depleted of type I collagen, similar to hyaline articular cartilage. Similar to tissue generated from high passage primary articular chondrocytes (ACs) and bone marrow/adipose-derived mesenchymal stem cells (MSCs), pellet chondrogenesis of expanded chondroprogenitors (i.e. P3-P5) exhibited increased production of fibrocartilage as evinced by immunodetection of Type I collagen [32–34]. However, unlike MSCs and more similar to ACs, type X collagen was not abundant in pellets derived from expanded chondroprogenitors and was depleted in purified chondroprogenitors [35]. Indeed, diffuse expression of type VI collagen further suggests the production of immature “neocartilage”. Prolongation of chondrogenesis in pellet culture may improve maturation of hiPSC derived-chondrocytes. Our protocol generates articular cartilage by extended treatment with TGF-β3, which serves as an important step in the generation of articular cartilage from adult stem cells as well as human PSCs [6,20,25,36,37]. Extended treatment with BMP-4 in pellet culture on the other hand may promote a more hypertrophic phenotype [6].

It has been reported that the donor-to-donor variability for iPSCs may result from epigenetic memory of the somatic cell of origin or caused by different reprogramming methods [13,14,38]. As expected, variance among different hiPSC lines was also observed in this study. Indeed, expression levels of important transcription factors such as MIXL1, MSGN1 and PARAXIS varies among unique hiPSC lines prior to the induction of mesodermal differentiation, leading to noticeable differences in the magnitude of increase of these stage-specific markers. However, despite the intrinsic variation of different cell lines, this protocol generated articular cartilage-like tissue from multiple hiPSC lines through specification of paraxial mesodermal. Nonetheless, the subtle differences between human chondrocytes derived from paraxial/lateral plate mesoderm or ectoderm remain unclear. Future investigation of hiPSC chondrogenesis along alternative developmental lineages, coupled with comprehensive genomic, transcriptomic, and/or proteomic analysis may facilitate identification of functional differences at multiple stages of differentiation. This information may be important for functional investigation of genetic arthritis susceptibility variants, which may cause joint-specific pathology [39,40].

With the widespread availability of iPSCs and gene editing technologies, it is now possible to assay the functional effects of disease susceptibility variants on developmental and pathological phenotypes [10,41–46]. Genome-wide association studies (GWAS) and candidate association studies have identified many genetic variants that may be involved in pathologies of articular cartilage, such as rheumatoid arthritis and osteoarthritis [47,48]. Follow-up analyses of these genetic variants have been challenging in iPSC-derived CPs due to the difficulty of isolating this cell population. Indeed, the heterogeneity of in vitro chondrogenesis may preclude our understanding of the mechanisms by which genetic variation can impact cartilage development and arthritis progression. Previous attempts to purify iPSC-derived CPs relied on many rounds of sorting multiple surface markers, using reporter constructs that are activated by minimal promoter regions of chondrocyte markers, or physical dissection of self-aggregating chondroprogenitors [20,29,37,49,50]. During chondrocyte differentiation, COL2A1 expression is regulated by concerted coordination of promoter regions, distal and intronic enhancers, and UTRs [51–53]. Therefore, the knock-in reporter approach described in this study enables precise monitoring of endogenous COL2A1 expression during differentiation and purification of a homogeneous CP population.

High-throughput disease modeling and drug screening platforms of cartilage injury and repair will benefit from the ability to produce large quantities of homogeneous cartilaginous aggregates from purified CPs [12,22,25,49,54]. Our laboratory and others have demonstrated the utility of CPs as an expandable depot of cells for subsequent chondrogenesis [20,49]. We show that the day 21 purified CPs (GFP^high population) maintain improved chondrogenic capacity over the course of 5 passages compared to controls. In this manner, small molecule and cytokine screens can be readily applied to tissue derived from expanded and purified CPs in a high throughput manner [22]. Increased homogeneity of hyaline cartilage derived from purified CPs may improve fidelity and reproducibility of disease modeling and drug discovery experiments. As demonstrated in previous studies of hiPSC chondrogenesis, we expect matrix degradation upon treatment with inflammatory cytokines such as IL-1 and TNF-α concomitant with diminished reporter activity, consistent with decreased COL2A1 expression and other chondrocyte markers upon addition of cytokine [6]. However, compared to AC-derived cartilage, hiPSC-derived cartilage is resistant to inflammatory cytokines, and higher doses may be needed to observe a response [55]. In addition, future studies are needed to assess the ability of purified CPs to improve osteochondral repair in vivo. CRISPR screening technology has recently become a powerful tool to identify coding and non-coding genetic elements that perturb reporter or surface marker expression or affect viability [56–58]. Since our reporter enables purification of CP cells, it can also facilitate such screens to identify novel targets necessary or inhibitory for chondrogenesis. Validation of identified hits may provide targets to further improve hiPSC chondrogenesis protocols.

The future of novel arthritis therapeutics lies in the application of hiPSC chondrogenesis to regenerative medicine and arthritis disease modeling [6,12]. The standardization of chondrogenesis protocols that follow developmental pathways will allow comparisons to be made between studies and improve reproducibility. By applying signaling logic aimed at activating and repressing specific developmental pathways, we can reliably guide hPSC differentiations along mesodermal lineage bifurcations [17,59]. This work shows that establishment of a standardized chondrogenesis protocol combined with purification of CP cells can produce a homogenous population of chondrocyte-like cells with improved matrix quality. Future studies investigating the effects of genetic or environmental perturbations on hiPSC-derived cartilage-like tissue will benefit from the current differentiation method.

Materials and Methods

hiPSC Lines and Cell Culture

Three human iPSC lines were used in the current study: BJFF.6, ATCC®ACS-1019 and RVR-iPSC.[54,60] Both BJFF.6 (BJFF) and ATCC®ACS-1019 (ATCC) were derived from foreskin fibroblasts of male newborn infants using sendai virus. BJFF was obtained from iPSC core at Washington University in St. Louis, and the cells were maintained on vitronectin (VTN-N; Fisher Scientific, Hampton NH) coated plates in Essential 8 Flex medium (E8; Life Technologies, Carlsbad CA). The ATCC line was purchased from ATCC (Manassas VA), and the cells were maintained on CellMatrix Basement Membrane Gel (ATCC) in Pluripotent Stem Cell SFM XF/FF medium (ATCC). RVR-iPSCs were retrovirally reprogrammed from BJ fibroblasts and characterized previously.[60] The cells were maintained on vitronectin in mTeSR1 medium (Stemcell Technologies, Vancouver CA). Media were changed daily and cells were passaged at 80–90% confluency.

Mesodermal Lineage Differentiation

Mesodermal differentiation medium: induction of hiPSCs into mesodermal lineages is based on a protocol reported by Loh et al with slight modifications [17]. In brief, when hiPSCs reached 70% confluency, the culture medium was switched to mesodermal differentiation medium composed of IMDM GlutaMAX™ (IMDM; Fisher Scientific, Waltam MA) and Ham’s F12 Nutrient Mix (F12; Fisher Scientific) with 1% chemically defined lipid concentrate (Gibco, Gaithersberg MD), 1% insulin / human transferrin / selenous acid (ITS+; Corning, Durham NC), 1% penicillin / streptomyocin (P/S; Gibco), 450 μM 1-thioglycerol (Sigma, St. Louis MO)and 1 mg/ml polyvinyl alcohol (Sigma).

Induction of anterior primitive streak (PS) (day 0): hiPSCs were rinsed with wash medium consisting of IMDM/F12 and 1% PS, and then fed with mesodermal differentiation medium supplemented with 30 ng/ml of Activin A (Stemgent, Cambridge MA), 4 μM of CHIR99021 (CHIR; Stemgent), and 20 ng/ml human fibroblast growth factor (FGF; R&D Systems, Minneapolis MN) for 24 hrs.

Induction of paraxial mesoderm (day 1): after a brief wash with DMEM/F12, cells were fed with mesodermal differentiation medium supplemented with 2 μM SB-505124 (SB5; Tocris, Bristol UK), 3 μM CHIR, 20 ng/ml FGF, 4 μM dorsomorphin (DM; Stemgent) for 24 hrs.

Induction of early somite (day 2): after a brief wash with DMEM/F12, cells were fed with mesodermal differentiation medium supplemented with 2 μM SB5, 4 μM DM, 1 μM Wnt-C59 (C59; Cellagen Technology, San Diego CA), and 500 nM PD173074 (PD17; Tocris) for 24hrs.

Induction of sclerotome (day 3 – day 5): after a brief wash with DMEM/F12, cells were fed with mesodermal differentiation medium supplemented with 2 μM purmorphamine (Pur; Stemgent) and 1μM C59 for 3 days. Medium was changed daily.

Induction of chondroprogenitor cells (day 6 to day 12–21): after a brief wash with DMEM/F12, cells were fed with mesodermal differentiation medium supplemented with 20 ng/ml human bone morphogenetic protein 4 (BMP-4; R&D Systems). Media was changed daily.

Chondrogenic Differentiation

Monolayer cultured CP cells were dissociated with 0.02% ethylenediaminetetraacetic acid (EDTA) at 37°C, and EDTA was diluted with an equal amount of medium composed of DMEM/F-12, GlutaMAX™ (DMEM/F12; Gibco), 10% fetal bovine serum (FBS; Atlanta Biologicals, Flowery Branch GA), 1% ITS+, 55 μM 2-Mercaptoethanol (2-ME, Gibco), 1% MEM non-essential amino acids (NEAA; Gibco), and 1% P/S. The cells were then collected and centrifuged for 5 minutes at 300 × g. After aspirating the supernatant, the cells was re-suspended in chondrogenic medium composed of DMEM/F-12, 1% FBS, 1% ITS+, 55 μM 2-mercaptoethanol, 100 nM dexamethasone (DEX; Sigma-Aldrich), 1% NEAA, 1% P/S, 10 ng/ml human transforming growth factor beta 3 (TGF-β3; R&D Systems), 50 μg/ml L-ascorbic acid 2-phosphate (ascorbate; Sigma-Aldrich), and 40 μg/ml L-Proline (proline; Sigma-Aldrich), and 300,000 – 500,000 CP cells were then pelleted for chondrogenesis for 28 days. Chondrogenic medium was changed every three days.

Chondroprogenitor Expansion

The chondroprogenitor population was expanded as previously described.[20] Briefly, unsorted or sorted populations of Day 21 chondroprogenitors were seeded on tissue culture plastic precoated with 0.1% Gelatin at a density of 10,000/cm² in a media composed of DMEM/F-12, GlutaMAX™ (DMEM/F12; Gibco), 10% fetal bovine serum (FBS; Atlanta Biologicals, Flowery Branch GA), 1% ITS+, 55 μM 2-Mercaptoethanol (2-ME, Gibco), 1% MEM non-essential amino acids (NEAA; Gibco),1% P/S, 40μg/mL human bFGF(Sigma-Aldrich), and 50 μg/mL L-ascorbic acid 2-phospate(Sigma-Aldrich). Media was replaced every two days. At roughly 70–80% confluency cells were passaged for chondrogenic differentiation or further expansion.

Histology

Pellets were fixed in 10% formalin overnight and stored in 70% ethanol at 4°C. They were then embedded in paraffin wax and sectioned at 8 μm thickness. Sections were stained with Safranin-O/hematoxylin standard protocol to reveal proteoglycan matrix.

Immunohistochemistry

To perform immunohistochemistry of type II collagen (1:1, Iowa Hybridoma Bank, Iowa), type VI collagen (1:200, Fitzgerald, Cambridge UK), or type I collagen (1:800, Abcam, Cambridge MA), sections were treated with xylenes and a decreasing ethanol series, incubated for 5 min with Digest-All 3 pepsin digestion (Thermo Fisher), treated with 30% peroxidase, blocked with goat serum, incubated with primary antibody for 1 h at room temperature, incubated with secondary antibody (ab97021 for type II and type I collagen, ab6720 for type VI collagen, Abcam, 1:500) for 30 minutes, treated with HRP streptavidin and AEC Red Single (Histostain-Plus BS, Thermo Fisher), counterstained with hematoxylin, and mounted with Clearmount (Thermo Fisher). Human osteochondral sections were used as both positive and negative controls.

Biochemical Analysis

Pellets were rinsed with DPBS, digested overnight at 65°C in 200 μl of papain solution consisting of 125 μg/ml papain, 100 mM sodium phosphate, 5 mM EDTA, and 5 mM L-cysteine hydrochloride at 6.5 pH, then stored at −20°C. Double-stranded DNA content was quantified using the PicoGreen™ assay (Invitrogen, Carlsbad CA). Total sulfated glycosaminoglycan (s-GAG) content was measured using the dimethymethlyene blue (DMMB) assay using bovine chondroitin sulfate (Sigma Aldrich) as a standard. Collagen content was measured using the hydroxyproline assay as previously described.[61]

Gene Expression

Cells in monolayer and pellets were rinsed with DPBS. Monolayer cells were lysed in 600 μl of Buffer RL (Norgen Biotek, Thorold Canada) and pellets were snap-frozen in 350 μl of Buffer RL which were then stored at −80°C. Pellets were homogenized using a miniature bead beater. The RNA was isolated using the Total RNA Purification Kit according to the manufacture’s recommendations (Norgen Biotek). Reverse transcription was performed using SuperScript™ VILO™ Master Mix (Thermo Fisher) per the manufacturer’s instructions. Quantitative RT-PCR was performed on the QuantStudio 3 (Thermo Fisher) and CFX96 Real Time System (Biorad, Hercules CA) using Fast SYBR™ Green Master Mix (Thermo Fisher) according to the manufacturer’s protocol. Fold changes were calculated using the ∆∆C_T method relative to hiPSCs as the reference time point and TATA-box-binding protein (TBP) as the reference gene. [62] Gene expression was probed using the primer pairs listed in the Table S2.

Flow Cytometric Analysis for Surface Markers

Surface markers of pluripotent RVR-iPSCs and CP cells derived from this line were characterized using 361 PE-conjugated monoclonal antibodies (Biolegend, San Diego CA) according to manufacturer’s instructions. Stained cells were fixed and analyzed by LSRII (BD Biosciences, East Rutherford NJ).

Fluorescence Activated Cell Sorting

At various time points during differentiation, cells were dissociated as described above. Pelleted cells were re-suspended in sort buffer containing PBS, 10% FBS, 2× P/S/F, and 1 U/mL DNase I (Thermo Fisher)). Cells were sorted on a Sony SH800S Cell Sorter (Sony Biotechnology, San Jose CA) using a 130 μm sorting chip.

Genome Editing and Clonal Isolation

To generate the COL2A1 reporter line, RVR-iPSCs were dissociated with accutase and 3×10⁶ cells were re-suspended in 200 μL complete mTeSR. 6 μg of gRNA/Cas9–2a-GFP expression vector and 3 μg of COL2A1 targeting vector were added (see Supplemental Experimental Procedures). Cell suspension was placed into a 4 mm cuvette and transfected using the Gene Pulser XL (BioRad) electroporation system. Transfected cells were plated in a matrigel-coated 10cm dish in complete mTeSR supplemented with 10μM Rock Inhibitor (Y-27632). Selection began the following the day with 0.5 μg/mL Puromycin. After roughly 10 days puromycin-resistant colonies were transfected with a CRE recombinase expression vector to remove the floxed selection cassette. Transfected cells were expanded and plated at low density for clonal isolation(180 cells/cm²). After 10–14 weeks resulting clones were mechanically picked and expanded for PCR screening of targeting vector integration using the following primer pair: PuroF2: 5’-GCAACCTCCCCTTCTACGAG; RightOutReverse: 5’-GCAAGGGACACCCAAGAGTA-3’

Statistical Analysis

To compare matrix production and gene expression levels within each cell line (Figures 1,2, and 4), one-way ANOVA using differentiation stage as the factor with Tukey’s post hoc testing was performed. For comparisons between multiple groups at various passages or time points (Figure 5 and 6), a two-way ANOVA using time point/passage and GFP population as factors with Tukey’s post hoc testing were performed as appropriate. All statistical analyses were carried out using the Graphpad Prism software. To measure correlations, linear regression was performed; R square and p-values were reported. Significance was reported at the 95% confidence level.

Significance.

Human induced pluripotent stem cells (hiPSCs) provide a promising cell source for cartilage tissue engineering and in vitro disease-modeling in a patient-specific manner. However, the low efficiency and variability of chondrogenesis in hiPSCs has been a major hurdle for these applications. Here, we demonstrate a highly efficient chondrogenic differentiation protocol via stepwise specification through a mesodermal lineage. Furthermore, we developed a genome-engineered COL2A1-GFP knock-in hiPSC line to identify a purified progenitor population with improved chondrogenic potential. The development of processes for rapid and repeatable chondrogenic induction of iPSCs will hopefully enable the identification of novel therapies for joint diseases such as osteoarthritis.