Introduction

Extracellular matrices (ECMs) are complex networks of proteins providing tissue structural and mechanical support, but also acting as growth factor storing and presenting entities¹. As such, ECMs are receiving increasing attention in tissue engineering as templates capable of guiding homeostasis, remodeling, and regenerative processes^2,3,4.

ECMs can be derived from native tissue or organs (native ECMs, nECMs) by applying a decellularization step which effectively removes the cellular fraction. The resulting nECMs are commonly used as biomaterials in regenerative medicine^5,6, offering a natural biocompatible structure. However, they suffer from substantial batch-to-batch variation with their properties being largely affected by the tissue source and the decellularization process^5,7. Importantly, while displaying a high level of biological complexity and fidelity, the composition of nECMs cannot be tailored to specific needs.

A valuable option emerged from recent advances in bioengineering which led to the development of synthetic ECMs (sECMs). These tunable materials are predominantly composed of biopolymers such as poly (lactic acid) (PLA), poly (glycolic acid) (PGA), poly (lactic co-glycolic acid) (PLGA) and Poly (ethylene glycol) (PEG), which can be functionalized with bioactive substances such as peptides, growth factors or enzymes^8,9,10,. The possible modulation of sECM’s composition and chemistry offers precise control over their mechanical properties, degradability, and temporal release of instructive molecules towards improved tissue repair. Despite holding great promises, yet their structure and function remain simplified compared to their native counterpart ^11,12.

With the aim of combining high biological fidelity and design flexibility, engineered ECMs (eECMs) have become a credible alternative. Those result from the exploitation of stem/progenitor populations capable of depositing ECMs under specific in vitro culture conditions^13,14,15,16, which can be subsequently isolated from the cellular fraction and exploited as a cell-laid product. Early studies were conducted using eECM derived from primary cells, exhibiting variable performance associated with the inter-donor variability^17,18.

Most recently, we demonstrated the possibility to standardize the production of eECMs through the engineering of dedicated human mesenchymal stromal/stem cell (hMSC) lines¹⁹. Precisely, the Mesenchymal Sword of Damocles – BMP2 (MSOD-B) line offered the generation of eECMs in the form of human cartilage templates²⁰. Those cartilage templates can be further devitalized, lyophilized, and stored as an off-the-shelf tissue, retaining remarkable bone formation capacity by instructing endochondral ossification^21,22.

The exploitation of cell lines as an unlimited cell source for eECM generation offers unprecedented standardization in graft production and performance. Meanwhile, it also consists of a robust cellular platform facilitating their further genetic engineering, to achieve the production of eECMs tailored in composition and function. Towards this objective, CRISPR/Cas9 appears as a versatile and efficient gene-editing tool. While this technology has been extensively used for genetic screening or disease modeling^23,24,25, the possibility to edit hMSCs and study the compositional impact on deposited eECMs has not been investigated so far.

In this study, we aim at demonstrating that CRISPR/Cas9 can be applied as an editing tool for the engineering of custom eECMs. As proof-of-principle, our strategy relies on the CRISPR/Cas9-guided editing of the MSOD-B line for knocking out Vascular endothelial growth factor (VEGF) and Runt related transcription factor 2 (RUNX2), respectively a key pro-angiogenic protein and transcription factor critically involved in the endochondral ossification pathway. The resulting hMSC lines will be evaluated for their capacity to generate hypertrophic cartilage in vitro and instruct tailored skeletal tissue formation in vivo as cell-free and lyophilized templates. If successful, our study is expected to validate the use of CRISPR/Cas9 for eECMs edited in composition, while providing a novel platform towards decoding the necessary molecular signals driving effective tissue repair.

Results

CRISPR/Cas9 editing of hMSC lines lead to efficient VEGF knockout in cartilage tissues

VEGF is a known master regulator of angiogenesis²⁶ and a key mediator of endochondral ossification. When reaching hypertrophy, chondrocytes highly express VEGF, prompting vasculature invasion and subsequent osteoprogenitor recruitment. This was proven to be essential for cartilage template remodeling into bone and bone marrow²⁷. Here, we thus first aimed at engineering a hMSC line CRISPR/Cas9-edited for VEGF knockout and evaluate the corresponding impact on cartilage and endochondral bone formation (Figure 1A). To this end, specific guide RNAs (gRNAs) targeting different regions of the VEGF gene but conserved across all isoforms of VEGF were designed based on a previously established protocol²⁸. Three gRNAs targeting exon 1 (VEGF_1.1, VEGF_1,2, VEGF1.3), one targeting exon 2 (VEGF_2.1), and one targeting exon 8 (VEGF_8.1) were designed (Figure 1B, Supplementary Figure 1A). These gRNAs were cloned into the pU6-(BbsI)_CBh-Cas9-T2A-mCherry vector encoding the Streptococcus pyogenes Cas9 (SpCas9) machinery. Following transfection in MSOD-B cells, single cell clones were sorted based on the transient mCherry expression and expanded for characterization (Figure 1A). Out of 163 single clones, 14 could be successfully expanded (8.5 %). From these clones, five were retrieved from the gRNA 1.1, six from gRNA 1.3 and three from gRNA 8.1. In order to assess a successful editing and directly correlate it to a knockout of VEGF, an ELISA assay was performed to measure the concentration of VEGF in the supernatant of expanded MSOD-B clones (Figure 1C). In non-edited cells (MSOD-B, control), 5000 pg/mL were secreted and detected by ELISA. From the successfully expanded clones, most of them showed reduced but non-abolished secretion of VEGF. However, two clones exhibited undetectable levels of VEGF, both resulting from the gRNA targeting the Exon1 of the VEGF gene. These clones were defined as MSOD-BΔV1 and MSOD-BΔV2 and further characterized for their capacity to form cartilage.

Figure 1.

To this end, clones were expanded and seeded on collagen scaffold and induced for three weeks towards chondrogenic differentiation. As anticipated with the MSOD-B tissues, histological assessment revealed the formation of a collagenous matrix (Masson’s Trichrome) with cartilage features (Safranin-O, Figure 1D). Similarly, a successful chondrogenic differentiation and deposition of glycosaminoglycans could be observed in the MSOD-BΔV1 (Figure 1D) constructs. A quantitative assessment (blyscan assay) confirmed the comparable content in glycosaminoglycans in MSOD-B and VEGF-edited clones (Figure 1E). Importantly, VEGF quantification in corresponding pellets validated the successful knockout of the protein, barely detectable in the MSOD-BΔV1 and ΔV2 tissues (47.64 pg/pellet in edited clones versus 552.7 pg/pellet in MSOD-B, Figure 1F).

To further examine the potential impact of VEGF-editing on tissue formation, we performed immunostaining analysis of the engineered cartilages. Confocal microscopy revealed strong deposition of Collagen Type I (COL1) and Collagen Type X (COLX) in all samples, characteristic of mature hypertrophic cartilage tissues (Figure 1G). However, while the immunostaining revealed the VEGF deposition in the MSOD-B ECM, the protein could not be detected in the MSOD-BΔV1 samples (Figure 1G).

Taken together, this data indicates the successful generation of MSOD-B lines knocked-out in VEGF. The MSOD-BΔV1 clones exhibited frank cartilage formation capacity with minimal level of VEGF. This validates the use of CRISPR/Cas9 as a precision tool to edit the composition of cartilage tissue.

VEGF knockout cartilage tissues retain bone remodeling capacity despite reduced early-stage vascularization

To evaluate the functional impact of VEGF knockout in engineered constructs, we assessed both the angiogenic and bone formation assays. Since the two clones exhibited similar in vitro tissue formation capacity, only the MSOD-BΔV1 was selected for further performance assessment.

We first performed the Chorioallantoic Membrane (CAM) assay as ex vivo assessment of angiogenic potential (Figure 2A). This assay provides insights into the biological activity and neovascularization potential of each construct, through quantitative evaluation of vascular density around engineered grafts After 4 days, we observed an extensive vessel formation in the periphery of both MSOD-B and MSOD-BΔV1 eECMs (Figure 2B). Quantification of vascular densities ²⁹ suggested a reduced vessel formation in MSOD-BΔV1 samples but without reaching significance (Figure 2C and Supplementary Figure 2A). Thus, the CAM assay indicated that both MSOD-B and MSOD-BΔV1 cartilages retain angiogenic potentials.

Figure 2.

We next investigate in vivo the impact of the VEGF knockout on the graft capacity to undergo endochondral ossification, by implanting cartilage grafts subcutaneously in an immunodeficient mouse model. Importantly, prior to implantation, tissues were lyophilized using a pre-established protocol²¹. This allowed us to assess the performance of the generated tissue itself, in line with the idea of developing off-the-shelf substitutes. Samples were extracted after two and six weeks as early and late development timepoints (Figure 2D). After two weeks in vivo, we first evaluated the angiogenic potential of tissues by quantitative confocal microscopy imaging of 100 µm thick sections stained for CD31, a well-defined vascular marker. This allowed us to evidence a dense vascular network in MSOD-B constructs, covering the entirety of the grafts (Figure 2E). Instead, MSOD-BΔV1 displayed a more limited vascularization, pre-dominantly at the tissue periphery (Figure 2E). Quantification confirmed these observations with MSOD-B exhibiting a significantly higher volume of vessels per total section volume (0.523 µm³ versus 0.231 µm³ for MSOD-B and MSOD-BΔV1 respectively, Figure 2F).

Using microtomography, we further assessed the amount of mineralized tissue formed in a temporal fashion. At two weeks, the formation of a cortical ring could already be observed in both MSOD-B and MSOD-BΔV1 tissues (Figure 2G). Quantifications revealed a similar bone/mineralized volume normalized to the total volume of the graft (BV/TV, Figure 2H, supplementary figure 2C and 2D) with 17 % and 16 % in MSOD-B and MSOD-BΔV1 samples respectively. At the six-week timepoint, trabecular structures could be observed in reconstructed 3D scans from both groups (Figure 2G). Remarkably, histological analysis confirmed the similar development of the tissues with frank bone formation (Massońs trichrome) already at two weeks post-implantation and minimal remnants of cartilage (Safranin O) (Figure 2I). At six weeks both samples remodeled into fully mature bone organs, characterized by presence of cortical and trabecular structures as well as a bone marrow tissue filling the cavity (Figure 2I).

These results indicate that the absence of VEGF in cartilage tissue can delay the early vascularization of MSOD-BΔV samples. However, this did not prevent nor impact the remodeling of the lyophilized grafts into bone and bone marrow tissues indicating that VEGF is non-essential in order to efficiently instruct endochondral ossification.

RUNX2 knockdown does not prevent chondrogenic differentiation but impairs hypertrophy

We next investigated whether the composition and thus function of a graft could be modified by editing transcriptional factors involved in mesenchymal cell differentiation. Using CRISPR/Cas9, we thus targeted the RUNX2, a known master regulator particularly important for chondrocyte differentiation and hypertrophy^30,31. We designed gRNAs targeting the coding regions involved in cell signaling integration, activation, and inhibition domain which are conserved in both isoforms of RUNX2 (Figure 3A, Supplementary Figure 3A).

Figure 3.

One gRNA was designed for targeting exon 2 (RUNX2_2.1), one targeting exon 5 (RUNX2_5.1), two targeting exon 6 (RUNX2_6.1 and RUNX2_6.2) and one targeting exon 8 (RUNX2_8.1). Similar to the VEGF set-up, gRNAs were transfected in MSOD-B cells together with a pU6-(BbsI)_CBh-Cas9-T2A-mCherry plasmid to ensure transfection efficiency and single cell sorting of positive clones. Out of 385 single clones, 62 could be successfully expanded, corresponding to a 16.1% efficiency. From these clones, five were derived from the RUNX2_5.1 gRNA, 18 from RUNX2_6.1, 15 from RUNX26.2 and 24 from RUNX2_8.1.

To identify successfully edited clones, we first screened for those exhibiting a decrease in their RUNX2 protein expression using intra cellular flow cytometry. Among the 62 clones, 17 displayed a reduced RUNX2 pattern as compared to the MSOD-B control (Figure 3B and Supplementary Figure 3B) and were further sent for sequencing analysis. This led to the identification of two successfully edited clones, demonstrating a point mutation in the exon 6 of RUNX2 gene (Supplementary Figure 3C). To confirm the knockout impact on the transcription factor structure, a Western Blot analysis was conducted on cellular extracts of in vitro cultured cells. MSOD-B cells exhibited intact RUNX2 proteins of 52 – 62 kDa^32,33(Figure 3C). Instead, the RUNX2-edited cells displayed truncated versions of 32 – 42 kDa, consistent with the expected mRNA shortening.

These two clones were defined as MSOD-BΔR1 and MSOD-BΔR2 and further assessed for their ability to form cartilage in vitro. Following 3D chondrogenic differentiation, samples were processed for histological analysis. The presence of cartilage structures embedded in a rich collagenous matrix could be observed in all MSOD-B, MSOD-BΔR2 (Figure 3D) and MSOD-BΔR1 tissues (Supplementary Figure 3D). Quantification of glycosaminoglycans revealed a similar content in all the groups (Figure 3E, Supplementary Figure 3E). Using immunostaining combined with quantitative confocal microscopy, the possible impact of RUNX2-editing on tissue hypertrophy was further investigated. First, Collagen Type II (COL2) staining confirmed the similar distribution of cartilage matrix across MSOD-B, MSOD-BΔR1 and MSOD-BΔR2 samples (Figure 3F and Supplementary Figure 3F). However, we observed a clear distinction in the Collagen Type X (COLX) expression pattern, a specific marker of hypertrophy which was hardly detectable in MSOD-BΔR tissues (Figure 3F and Supplementary Figure 3F). Subsequent quantification confirmed a significant reduction of COLX volume in MSOD-BΔR1 (0.236 µm³) and MSOD-BΔR2 (0.182 µm³) compared to MSOD-B (0.81 µm³) (Figure 3G, Figure 3H, Supplementary figure 3G-H-I).

To further characterize the functional impact of the RUNX2 knock-down, the MSOD-B and MSOD-BΔR1 osteogenic differentiation capacity was assessed in vitro. After 3 weeks of culture in osteogenic medium (or expansion medium, control), Alizarin Red staining revealed a s a marked absence of mineralization in MSOD-BΔR1 culture, while in stark contrast with MSOD-B cells (Figure 3I). Quantitative PCR analysis indicated a lower expression of COLX and ALPL in MSOD-BΔR1 (Figure 3J), key osteogenic-associated genes, of interest, the expression level of RUNX2 was only partially decreased in MSOD-BΔR1, in line with the truncated mRNA reducing but not abrogating qPCR primers binding probability.

In summary, we here report the successful RUNX2 knockout in MSOD-B lines using CRISPR/Cas9. RUNX2 did not impair chondrogenesis but prevented the tissue hypertrophy, as well as the osteogenic potential of the cells.

RUNX2 knockout in cartilage tissues disrupts effective ectopic bone formation

To assess the corresponding impact of RUNX2-edited cartilages on bone formation, MSOD-B tissues (as control) and MSOD-BΔR1 were lyophilized and implanted subcutaneously in immunodeficient mice for two to six weeks. From microtomography image reconstructions, we observed an evident reduction of mineralization in MSOD-BΔR1 compared to MSOD-B (Figure 4A) already after two weeks. This qualitative and visual difference in mineralization persisted after six weeks in vivo (Figure 4A). Subsequent quantifications confirmed these observations with a clear reduction in BV/TV as early as week 2 (16.46 % in MSOD-B and 7.01 % in MSOD-BΔR1) and persisting at week 6 (20.03% in MSOD-B and 4.25 % in MSOD-BΔR1) (Figure 4B). Histological analysis revealed an early bone formation at two weeks in both samples, but to a lower extent in the MSOD-BΔR1 group (Figure 4C). As anticipated, MSOD-B tissues underwent full remodeling after six weeks in vivo. In sharp contrast, the MSOD-BΔR1 only displayed a partial maturation with limited presence of bone and bone marrow.

Figure 4.

Altogether, this indicates an incomplete remodeling of RUNX2 edited samples with significantly delayed cortical and trabecular structure formation. This correlates with the impaired hypertrophic phenotype in the corresponding in vitro samples.

RUNX2 knockout in cartilage tissues leads to better cartilage regeneration in a rat osteochondral defect

In order to assess the performance of CRISPR-Cas9 edited eECMs in a relevant skeletal regenerative context, a proof-of-concept study in the rat osteochondral defect was performed. MSOD-B and MSOD-BΔR1 tissues were lyophilized and placed in sub-chondral defect in the distal femur of a rat. The defect consisted in a drill hole of 1mm diameter and 2mm depth (Figure 5A and supplementary figure 5A). Positive controls consisted of untreated healthy rats.

Figure 5.

Following explantation, histological stainings and micro-tomography were performed on all sample groups. Hematoxylin and Eosin (H&E) and Masson’s Trichrome indicated de novo bone formation in the defect area for both the MSOD-B and MSOD-BΔR1 groups (Figure 5B and supplementary figure 5B). The presence of fibrotic tissue as well as a reduced marrow compartment suggested an incomplete remodeling at that timepoint (Figure 5B). Upon Tartrate-resistant acid phosphatase (TRAP) staining, we observed elevated level osteoclastic activities in MSOD-B samples compared to MSOD-BΔR1 indicating reduced rate of active bone formation in MSOD-BΔR1 (Figure 5C). The micro-CT quantification revealed a 30% repair of the damaged bone in the defect for both MSOD-B and MSOD-BΔR1 groups (Figure 5D, Supplementary figure 5C). Of interest, a significantly higher trabecular separation was observed in MSOD-B and MSOD-BΔR1, confirming an ongoing bone remodeling process (Figure 5E).

While no statistical differences in bone formation could be identified between the two groups, the MSOD-BΔR1 tissue led to a superior regeneration of the cartilage area, as revealed by Safranin-O staining (Figure 5F) in the chondral zone (Figure 5F, a,b and c magnification). Interestingly, the MSOD-BΔR1 group exhibited a higher remnants of GAGs in the subchondral area (Figure 5F, d,e and f magnification).

Quantification of cartilage tissue within condyle surface area confirmed a poor-to-no repair in the MSOD-B group (1.05%, Figure 5G). In sharp contrast, the MSOD-BΔR1 implanted eECMs initiated a chondral regeneration reaching approximately 20.67% of the total healthy cartilage area. The integration and regeneration potential of engineered constructs was further evaluated by performing a semi-quantitative histological assessment following a pre-existing grading system³⁴ (Table 1 and Supplementary table 1). This approach allows for the systematic evaluation of critical repair tissue parameters, offering a comparative measure of the regenerative efficacy of the engineered constructs against healthy tissue benchmarks. In the assessment of stained femur condyle sections, cellular morphology and matrix staining of the MSOD-BΔR1 group (66% and 50%, respectively) were significantly better compared to the MSOD-B one (33% and 16%, respectively). Although both constructs exhibited reduced cartilage thickness and subchondral bone regeneration, MSOD-BΔR1 consistently outperformed the MSOD-B grafts (33.33% versus 8.33% and 25% versus 16.66%). Surface regularity and integration with adjacent cartilage also revealed better outcomes in MSOD-BΔR1 (41% and 50%) as opposed to MSOD-B (50% and 33%), further indicating an enhanced regenerative potential of the MSOD-BΔR1 constructs.

Table 1.

In conclusion, while both grafts yielded comparable bone regeneration within the osteochondral defects, the RUNX2 deletion in cartilage grafts led to a superior cartilage regeneration.

Discussion

We here report the possibility to edit the composition and function of eECM by CRISPR/Cas9 engineering of human mesenchymal lines. VEGF knockout led to successful in vitro formation of cartilage with targeted protein depletion. Upon implantation, this impacted the graft vascularization onset but not the endogenous instruction of the endochondral program. In turn, RUNX2 knockout prevented cartilage hypertrophy in vitro, significantly delaying ectopic bone and bone marrow formation in vivo. This strategy was validated in a functional osteochondral defect model, whereby prevention of cartilage hypertrophy by RUNX2 deletion led to superior cartilage regeneration.

A clear interdependency of angiogenesis and osteogenesis occurs upon bone formation^35,36. For this reason, VEGF-enrichment has been naturally proposed as a strategy to accelerate or increase bone graft vascularization^37,38. Our study demonstrates that VEGF knocked-out cartilage templates retained full osteoinductive potential in a challenging ectopic environment. Those results are in sharp contrast with intramembranous strategies^39,40, including our previous work¹⁹, whereby grafts remain insufficient in promoting complete bone remodeling even upon VEGF-enrichment. This suggests that the first stage of host progenitor recruitment by our eECM is not VEGF dependent, although subsequent tissue vascularization and remodeling may be orchestrated in a second step by endogenous cell secretion. Other pro-angiogenic proteins embedded in the matrix may also have compensated for the lack of VEGF, such as BMP-2²⁰ known to stimulate endothelial cell proliferation, migration, and differentiation^{41 ,42}.

Conversely, RUNX2 deletion was shown to fully prevent cartilage hypertrophy, in line with published mouse models also reporting a lack of endochondral ossification and lethality at birth upon RUNX2 knockout^43,44. Here, MSOD-BΔR cartilage templates still exhibited frank bone formation capacity, but with cortical and trabecular structures largely reduced as compared to controls. Future work may clarify if this strictly results from the absence of COLX, or if an additional specific cartilage matrix component (e.g. matrix metalloproteinases) impaired the instruction of endochondral ossification.

The performance in a regenerative context was further evaluated in a rat osteochondral defect model⁴⁵. We defined an early timepoint of 3-week as providing the opportunity to assess the early contribution of the grafts to both the cartilage and bone tissue regeneration. Strikingly, only the MSOD-BΔR could contribute to a superior chondrogenesis. We hypothesize that the absence of hypertrophic features prevented the template degradation and favored its integration to native cartilage. Despite the repair performance being limited to an approximate 20% of native cartilage restoration, this is remarkable in light of the BMP-2 content in our ECM prompting endochondral ossification ^46,47,48. In fact, preventing the in vivo remodeling of engineered cartilage templates remains challenging in the skeletal regeneration field, with bone marrow MSCs (BM-MSCs) cartilage systematically reaching hypertrophy and subsequent ossification. Our results warrant further studies comprising additional timepoints to decipher the repair dynamic, as well as the long-term stability of newly formed tissue. It also prompts a comparison with other recently proposed off-the-shelf strategies, based on ECM⁴⁹ or synthetic scaffold materials⁵⁰, in order to comprehend the feasibility to exploit our grafts for stable cartilage / joint repair^51,52. Taken together, while the relevance of CRISPR/Cas9 eECMs in cartilage repair remains to be further validated, our study demonstrates the relevance of genetically-edited ECMs in regenerative contexts.

CRISPR/Cas9 editing of human mesenchymal cells has previously been explored for gene and cell therapy applications^53,54, for tailoring the immunomodulatory and/or differentiation capacity of engineered cells. However, no studies have described the exploitation of editing strategies for the generation of eECMs, whereby modified cells are absent from the final grafting product. Our work illustrate that this can be achieved by direct targeting of secreted factors typically embedded upon ECM deposition, as demonstrated with VEGF. Alternatively, we also propose the knockout of key transcription factors as a strategy for custom eECM generation by impacting their tissue developmental/maturation stages.

A clear implication of our work lies in the possibility to decipher the necessary factors capable of instructing de novo tissue formation. Those findings will be of high relevance for the design of eECMs tailored in composition. Beyond bone repair, our study also bears high relevance in other regenerative contexts. In fact, an exciting opportunity also lies in harnessing CRISPR/Cas9 for editing eECMs and tuned their immunogenicity. Key inflammatory components could be turned down, leading to improved efficacy of repair in line with the immuno-engineering principles^55,56.

Importantly, our study describes the exploitation of eECMs in a lyophilized form, thus conferring an off-the-shelf storage solution. Together with the standardization of production conferred by stable cell sources, our concept offers exciting translational opportunities. In fact, instrumental to this work is the use of dedicated hMSC lines. The genetic modification of primary cells is laborious, and their limited lifespan ex vivo challenges their selection, characterization, and exploitation for tissue engineering applications^57,58. Here, the MSOD-B was harnessed as an unlimited cell source with robust differentiation potential. Nonetheless, the editing of eECMs remains tedious in part due to the limited CRISPR/Cas9 efficiency and potential off-target effects, calling for a systematic clonal selection. In addition, while large CRISPR/Cas9 screening can be performed in other stem cell systems^59,60, the critical cell mass required for effective cartilage formation affects parallelization and leads to limited throughput. Nonetheless, after the identification and characterization process, the resulting cell lines can be banked and used for unlimited tissue manufacturing.

The MSOD-B line remains the only human cell source capable of priming endochondral ossification by engineering living or cell-free grafts. This was demonstrated to be driven by the combined low dose of BMP-2 embedded in the tissue (∼40 ng/tissue) together with glycosaminoglycans and other thousands of identified ECM proteins²⁰. The BMP-2 amount is thus far below the typical amount used in sECMs approaches for effective osteoinduction, falling in the microgram to milligram range²¹. This is one key advantage of eECM graft, exhibiting a biological complexity that so far has not been matched by sECMs. Ideally, the two approaches are complementary as the editing of eECMs could inform on the necessary but sufficient factors driving effective regeneration. Those would in turn be embedded in a sECMs strategy and offer a fast and possibly cost-effective solution.

To conclude, the present work offers a platform for decoding factors involved in tissue regeneration and generate tailored eECM. Here, illustrated in the context of skeletal repair, our study may offer similar opportunities in other regenerative situations.

Materials and Methods

Cell expansion

MSOD-B cells and their modified progeny were cultured in a humidified incubator at 37 °C and 5% CO₂ using complete medium consisting of α-minimum essential medium (αMEM) supplemented with 10% fetal bovine serum, 1% HEPES, 1% sodium pyruvate, 1% penicillin-streptomycin-glutamine solution, and 5 ng/ml of fibroblast growth factor-2 (FGF-2) (all from Gibco). Cells were seeded at a density of 3200 cells/cm² until they reached 90% confluency. The medium was replaced twice a week.

Chondrogenic and osteogenic differentiation

MSOD-B cells and their modified progeny were harvested from culture flasks by adding Trypsin-EDTA (0.25%) (Gibco) and subsequently seeded on cylindrical Collagen Type I scaffold (Avitene^TM Ultrafoam^TM Collagen Sponge, Davol) of 6 mm in diameter and 3 mm in thickness at a density of 2 × 10⁶ cells per scaffold in 12 well plates coated with 1 % agarose (Sigma) for chondrogenic differentiation and in a 12-well plate for osteogenic differentiation. Tissue constructs were cultured for three weeks in chondrogenic medium (DMEM supplemented with 1% penicillin-streptomycin-glutamine, 1% HEPES (1M), 1% sodium pyruvate (100 mM), 1% ITS (100x) (Insulin, Transferrin, Selenium) (all from Gibco), 0,47 mg/ml linoleic acid (Sigma), 0.12% bovine serum albumin (25 mg/mL) (BSA) (Sigma), 0.1 mM ascorbic acid (Sigma), 10^-7 M dexamethasone (Sigma) and 10 ng/ml TGF-β3 (Novartis)). The cells in the 12 well plates were supplemented for three weeks with osteogenic (or expansion medium, control) differentiation medium (ɑ-minimum essential medium (ɑMEM) with 10% fetal bovine serum, 1% HEPES (1M), 1% sodium pyruvate (100 × 10^-3M) and 1% penicillin-streptomycin-glutamine solution (100X), supplemented with 0.01M β-dexamethasone and 0.1M ascorbic acid). Media were replaced twice a week.

Lyophilization

After three weeks of chondrogenic differentiation, the tissue constructs were rinsed twice with PBS 7.2 (without calcium/magnesium Gibco), snap frozen in liquid nitrogen for five minutes and then lyophilized using a freeze dryer (Labconco) (−80 °C and 0.05 mbar) overnight. Thereafter, the lyophilized tissue constructs were stored at 4 °C.

Transfection of MSOD-B cells with gRNAs

MSOD-B cells were seeded at a density of 400,000 cells per well in 12 well plates to reach a minimum of 80% confluency the following day. The medium was replaced before transfection. The transfection was performed with a / ratio of 2 µL lipofectamine to 1 µg DNA. A DNA mix was prepared for each of the five gRNAs and the two plasmid controls. DNA mixtures were composed of 1 µg of plasmid, 2 µL of P3000 reagent and 50 µL of OptiMEM (Thermo-Fischer). A lipofectamine cocktail was prepared for all DNA mixtures consisting of 24 µL of lipofectamine 3000 in 400 µL of OptiMEM. Lipofectamine and the DNA mixtures were added at a 1:1 ratio. After 48h, cells were analyzed and mCherry positive clones were FACS sorted as single cells in 96-well plates using ARIAIII (BD Biosciences). Successfully expanded clones were then further characterized.

ELISA

VEGF protein content was measured in supernatant collected from cells seeded at 570 cells/cm² in T175 flask and cultured for 3 days. Content from engineered cartilage tissues was assessed following digestion in RIPA buffer. The Quantikine ELISA kit for Human VEGF-A Immunoassay from the R&D System was used according to the manufacturer instructions to determine protein concentration.

Intracellular flow cytometry

MSOD-B and MSOD-B ΔRUNX2 cells were trypsinized, fixed and permeabilized using the Fixation/Permeabilization Kit (BD Biosciences). Following a blocking step in 10% normal donkey serum (Sigma), cells were stained with a primary antibody against RUNX2 (Rabbit anti-human, Thermo-Fischer PA5-82787) for one hour at room temperature. After primary antibody incubation, cells were washed and stained with an Allophycocyanin (APC) labeled secondary antibody (Donkey anti-rabbit IgG DyLight™ 649, Biolegend 406406) for 30-45 minutes at room temperature in 2% normal donkey serum. The Control samples consisted of unstained cells (negative control) and cells incubated only with the secondary antibody (secondary Ab control). Data were recorded on a BD LSRFortessa^TM Cell Analyzer (BD Biosciences). FCS files were analyzed using the FlowJo^TM software (FlowJo LLC, 10.5.3, BD Biosciences).

Western blotting

MSOD-B and MSOD-B ΔRUNX2 cells were trypsinized and washed with ice cold PBS twice and lysed on ice for 10 min in RIPA buffer (#10017003, Thermo Fisher Scientific) supplemented with 1× proteinase and phosphatase inhibitor cocktail (#78440, Thermo Fisher Scientific). The lysates were centrifuged at 16, 000g for 15 min +4 °C, and the supernatants were collected. Sample buffer (Laemmli buffer (#161-0737, BioRad) supplemented with 5% 2-mercaptoethanol, 1× proteinase and phosphatase inhibitor cocktail (#78440, Thermo Fisher Scientific) was added to the supernatant at 1:1 ratio. Samples were boiled at 95 °C for 5 minutes and stored at −80 °C or kept on ice until gel loading. Proteins were separated using Bolt gels according to manufacturer’s protocol (#NW04122, #B0002, Thermo Fisher Scientific). iBlot2 system was used to transfer the proteins on polyvinylidene fluoride membrane (PVDF) membrane according to manufacturer’s protocol (#IB24001, Thermo Fisher Scientific). PVDF membrane was washed once in 1× PBST buffer (#28352, Thermo Fisher Scientific) and blocked in 2% blocking solution (#10156414, Thermo Fisher Scientific) for 1 h at room temperature. Membranes were incubated overnight at +4 °C with primary antibodies at recommended concentrations in 1% blocking solution. Membranes were washed three times (5 min for each wash) with 1× PBST buffer, and secondary HRP-conjugated antibodies in 1% blocking solution were added to the membranes at 1:5000 concentration for 1 h incubation at room temperature. Membranes were washed three times with 1× PBST and proteins were detected by chemiluminescence according to manufacturer’s protocol (#RPN2232, Thermo Fisher Scientific). The following antibodies were used. primary antibodies: RunX– (D1L7F - 12556) from Cell Signaling Technology and Actin (612656) is from Becton Dickinson. The secondary antibodies were anti-Mouse (GENA931) from Sigma-Aldrich, anti-Rabbit (NA9340V) from Thermo Fisher Scientific.

Biochemical analysis (GAG analysis)

Lyophilized tissue constructs were digested by overnight incubation in 0.5 ml of Proteinase K solution (1 mg/mL Proteinase K, Sigma; 10 ug/ml pepstatin A, Sigma; 1 mM EDTA, Sigma; 100 mM Iodoacetamide; 50mM Tris) at pH 7.6 and 56 °C. The GAG content of digested samples was analyzed using Glycosaminoglycan Assay Blyscan kit (Biocolor) following the manufacturer instruction.

Mice

FoxN1 KO BALB/C (nude mice) 6–8-weeks old, were obtained from Charles River Laboratories. All mouse experiments and animal care were performed in accordance with the Lund University Animal Ethical Committee guidelines (ethical permit #M15485–18). Mice were housed at a 12-hour light cycle in individually ventilated cages at a positive air pressure and constant temperature. Mice were fed with autoclaved diet and water ad libitum. During the implantation procedure, anaesthesia was performed with 2-3 % Isoflurane (Attane). The mice were kept on a heating pad during the procedure to avoid the heat loss.

Micro-CT scanning

In vivo samples were explanted and fixed overnight with 4% formaldehyde before ex vivo micro-CT analysis using a U-CT system (MILABS, Netherland) equipped with a tungsten x-ray source at 50 kV and 0.21 mA. Volumes were reconstituted at 10 µm isotropic voxel size. For total volume (TV) analysis, each sample was assessed with Blender (v2.82a, Netherland). Briefly, a mesh was created surrounding the 3D reconstruction of each sample and the volume occupied was then quantified. For bone volume (BV) analysis, the highly mineralized tissue volume was quantified using Seg3D (v2.2.1, NIH, NCRR, Science Computing and Imaging Institute (SCI))

Sample preparation for histological analysis

In vitro samples were directly fixed prior to sectioning. In vivo samples were fixed and subsequently decalcified with a 10% EDTA solution, pH = 8, at 4 °C for two weeks prior to tissue embedding.

Paraffin embedding was performed on samples fixed in 4% formalin (Solveca AB). Tissues were dehydrated by immersion in consecutive solutions of 35%, 70%, 95%, and 99.5% graded ethanol solution (Solveco AB). Immersion in a 99.5% ethanol/xylene solution (1:1, Fisher Scientific) was then performed for ten minutes followed by two rinsing in xylene (Fisher Scientific) for 20 minutes. Subsequently, tissues were embedded in paraffin at 56 °C overnight, before sectioning with a microtome (Microm HM 355 Rotary Microto–e) in 5 - 10 μm sections. The sections were then dried overnight at 37 °C before staining. Prior to staining, sections were deparaffinized by two washes in xylene for 7 minutes and once in 99.5% ethanol/xylene solution (1:1) for 3 minutes. Afterwards, sections were hydrated twice in consecutive solutions of 99.5%, 95%, 70%, and 35% ethanol, for 7 minutes each.

Agarose embedding was performed on samples fixed overnight with 4% paraformaldehyde (Thermo Scientific), using 4% low-melting agarose (Sigma). Sections of 100 µm thickness were obtained using a 7000smz vibratome (Campden) with stainless steel or ceramic blades.

Safranin O Staining

Paraffin-embedded sections were stained using Mayer’s hematoxylin solution (Sigma-Aldrich) for 10 minutes. Samples were then placed under running distilled water to remove superfluous staining from the sections. Subsequently, sections were stained with 0.01% fast green solution (Fisher Scientific) for five minutes and rinsed with 1% acetic acid solution (glacial, Fisher Scientific) for 15 seconds. After that, the slides were stained with 0.1% safranin O (Fisher Scientific) solution for 5 minutes. Dehydration and clearing were performed by immersion in 95%, 99.5% ethanol, 99.5% ethanol/xylene solution (1:1) and xylene twice successively for 2 minutes. Finally, the stained slides were mounted with glass slides using PERTEX mounting medium (PERTEX, HistoLab).’

Masson’s trichrome staining

Masson’s trichrome staining was performed using the trichrome staining kit (Sigma-Aldrich Sweden AB) according to the manufacturer’s guidelines. Briefly, tissue sections were deparaffinized and immersed in cold running deionized water for 3 minutes. Then the sections were kept in Bouin’s solution at room temperature overnight or at 56 °C for 15 minutes. The slides were washed by running tap water and stained using working Weigert’s iron hematoxylin solution (Sigma) for five minutes for nuclei detection (in black). After washing the slides, the cytoplasm was stained in red with Biebrich Scarlet-Acid fuchsin for five minutes followed by clearing the slides by immersion in working phosphotungstic/ phosphomolybdic acid solution for five minutes. Collagen stained blue by immersion in aniline blue solution for five minutes followed by clearing in 1% acetic acid (glacial, Fisher Scientific) solution diluted in distilled water (glacial, Fisher Scientific) for two minutes and washing with running deionized water. Finally, the sections were dehydrated in graded ethanol solutions (95% once, 100% twice) for two minutes each before washing with xylene twice for two minutes and mounted with PERTEX mounting medium.

Hematoxylin and Eosin staining

Paraffin-embedded sections were stained using Mayer’s hematoxylin solution (Sigma-Aldrich) for 10 minutes. Samples were then placed under running distilled water to remove superfluous staining from the sections. The slides were immersed precisely in 1X phosphate-buffered saline (PBS) for 1 minute to intensify the blue nuclei staining while preserving tissue integrity. Following this, a thorough rinse with 3 changes of distilled water effectively removed any residual PBS, preparing the sections for the subsequent counterstaining step. Subsequently, Alcoholic-Eosin (Sigma-Aldrich) was applied for 1 minute as the counterstain without any rinsing post-application, ensuring optimal interaction of the staining solution with the tis. Dehydration and clearing were performed by immersion in 95%, 99.5% ethanol, 99.5% ethanol/xylene solution (1:1) and xylene twice successively for 2 minutes. Finally, the stained slides were mounted with glass coverslips using PERTEX mounting medium (PERTEX, HistoLab).

Quantitative PCR

Total RNA isolation was conducted from in vitro engineered tissues utilizing the Quick-RNA extraction kit (Zymo Research, R1055) following the manufacturer’s protocol. Subsequently, cDNA extraction was carried out using cDNA synthesis kit (Invitrogen 11917020). The quantitative real-time polymerase chain reaction (qRT-PCR) was performed, utilizing the assay on demand from Applied Biosystems to assess the expression of specific genes. These genes include glyceraldehyde 3-phosphate dehydrogenase (GAPDH, HS02786624_G1), runt-related transcription factor 2 (Runx2, Hs00231692_m1), alkaline phosphatase (ALP, Hs01029144_m1), and collagen type X (ColX, Hs00166657_m1).

Rat osteochondral defect model

Ten- to 12-week-old male Sprague Dawley rats (n=11) were purchased from JANVIER LABS (France). After 7 days of acclimation the rats were anesthetized using 3% isoflurane. Then, animals were placed on a 37°C warm heating pad in prone position. Once anesthetized, isoflurane was lowered to 2 to 2.5%, and buprenorphine (Temgesic, 30 μg/kg; Indivior Europe Ltd., Dublin, Ireland) was injected subcutaneously for analgesia. The right hind limb of the animal was shaved and carefully disinfected, and an incision was made along the skin and soft tissue to expose the right knee. After cleaning the knee laterally from soft tissue, patella was displaced laterally to expose the distal femur. An articular cartilage defect with 2mm depth and 1mm diameter was created at the trochlear groove of femur. The defect was then push-fitted with the grafts, and not secured with any sutures. The wound was closed in a layered fashion using resorbable sutures (Vicryl 45-0, Ethicon, Somerville, USA) by closing the joint capsule, followed by the muscle tissue (continuous interlaced suture), followed by closing the skin (Donati suture). Animals started load bearing immediately after surgery. After 3 weeks (n=11) animals were anesthetized by isoflurane inhalation (3%), followed by CO2 asphyxiation, and right-femur were harvested prior to subsequent micro-CT and histological characterization.

Histology grading method

Histological sections from rat femur condyles were stained with Safranin-O, Masson’s trichrome, and H&E staining. A semi-quantitative analysis of the repaired tissue was performed by utilizing a customized adaptation of the histological grading system initially outlined by Wakitani et al⁶¹. This modified scale encompasses six distinct parameters: cellular morphology, matrix staining, surface regularity, cartilage thickness, regenerated subchondral bone formation, and integration with neighbouring cartilage. Each parameter was graded on a numerical scale ranging from 0 to 4 points by independent expert observers, where a maximum total score of 16 indicates the presence of tissue displaying characteristics akin to entirely healthy and normal tissue (Table 1).

Cartilage quantification method

An evaluation of three distinct sites within the condyle defect was conducted to ascertain the proportion of cartilage in each knee. Utilizing the measured femoral condyle thickness (200μm) and defect length (1mm), a standardized Region of Interest (ROI) was defined using ImageJ. The percentage of cartilage-positive regions within each defect was calculated in relation to the total ROI size, offering a quantitative assessment of cartilage regeneration. This identical approach was applied to evaluate the percentage of cartilage in the healthy control samples.

Trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp)

Image analysis was performed by ImageJ. To evaluate the defect space, a rectangular interface region (2 × 1 mm) was defined as the Region of Interest. Trabecular thickness (Tb.Th) and trabecular separation (Tb.Sp),was calculated inside the ROI using BoneJ plugin.

Vascular density quantification

To assess vascular density, the Quantitative Vascular Analysis Tool (Q-VAT)²⁹ was utilized. The samples were placed inside the chicken egg for CAM assay, and the images were acquired using a digital microscope. These images were then segmented into tiles for detailed analysis. The binary vascular masks were generated using the Q-VAT’s automated capabilities, to distinguish vascular structures within the tissue. The tool then facilitated the separation of vessel measurements based on their diameters, allowing for a differentiated quantification of macro- and microvasculature. These binary masks were then processed to calculate the vascular density, which refers to the proportion of the tissue area occupied by vessels. This quantitative metric was evaluated for each tile, and the mean vascular density was computed across all tiles for each tissue sample. As part of the analysis, we applied Q-VAT to double or triple-stained slides to quantify the overlapping percentage of vessels, comparing various time points to observe changes over the course of the experiment.

Immunofluorescence

Agarose-embedded sections were treated with 0.5% Triton X-100 (Sigma) in PBS supplemented with 20% donkey serum (Jackson ImmunoResearch) for one hour at room temperature (RT) for blocking and permeabilization. After blocking/permeabilization, sections were stained with primary antibodies (mouse anti-Collagen II (Invitrogen, MA137493), Anti-Collagen 1 (Abcam, ab138492), VEGF polyclonal antibody (Bioss Antibodies, bs-0279R), rabbit anti-Collagen Type X (ColX) (abbexa, abx101469), Goat anti mouse/rat CD31 (R&D system, YZU0119121), VEGF at 4 °C overnight. Sections were washed three times using cold PBS with 0.1% Triton X-100, 20 minutes each, followed by staining with secondary antibodies (CF568 donkey anti-rabbit (Biotium, 20098), CF567 anti-goat (Invitrogen, A11057), CF633 donkey anti-mouse (Sigma-Aldrich, SAB4600131 in 2% donkey serum for three hours at room temperature. Sections were washed three times using cold PBS with 0.1% Triton X-100, 20 minutes each once more.

All sections were mounted using the Vectashield antifade mounting medium containing DAPI (Vector Laboratories, H1200). A LSM780 confocal microscope (Zeiss) with a 10x/20x objective was used to capture images and image series of whole tissue sections (z-step = 2 µm). The IMARIS 9.5 software package (Oxford Instruments) was used to analyze the data.

Acknowledgements

We express our gratitude to the following institutions and individuals for their invaluable support and contributions to this study: Lund University, Medical Faculty, FACS Facility (Anna Hammarberg), Cell and Gene Therapy core facility (Pia Johansson) and Imaging Facility at Lund Stem Cell Centre, as well as the Lund University Bioimaging Centre, for providing access to the X-ray CT system. The funding for this research was made possible by a Knut and Alice Wallenberg grant, the Medical Faculty at Lund University, the Swedish Research Council, the European Research Council (Starting grant hOssicle #948588 to P.E.B), and Region Skåne. Figures were generated using BioRender.com. Special thanks to Ani Grigoryan and Dimitra Zacharaki for their expertise and valuable insights regarding FACS experiments, and to Steven Dupard for his advice and guidance on the design of gRNAs and molecular techniques. Christian Hansen deserves our appreciation for providing the mcherry plasmid and offering valuable guidance and knowledge related to CRISPR-Cas9.