Goat BMSCs (gBMSCs) were harvested under sterile conditions from the sternum of skeletally mature, female, Saanen goats. Briefly, excised bone marrow was cut into small pieces using a scalpel. The marrow pieces were then gently rotated for 5 min in high glucose Dulbecco’s modified eagle’s medium (hgDMEM) GlutaMAX supplemented with 10% v/v FBS, 100 U mL^–1 penicillin, 100 μg mL^–1 streptomycin (all Gibco, Biosciences, Dublin, Ireland) and 5 ng mL^–1 bFGF2 (Prospect Bio) (XPAN) to help liberate the cellular components. The culture medium was then aspirated and passed through a 40 μm cell sieve prior to counting and plating at a density of 57 × 10³/cm² and expanded under physioxic conditions (37°C in a humidified atmosphere with 5% CO₂ and 5 % pO₂) for improved chondrogenic differentiation. Following colony formation, gBMSCs were trypsinised using 0.25% (w/v) Trypsin Ethylenediaminetetraacetic acid (EDTA). gBMSCs for microtissues were expanded from an initial density of 5,000 cells/cm² in XPAN medium under physioxic conditions until P3. gBMSCs were initially used to validate the capacity of the microwells to support initial cell-aggregation and subsequent growth under chondrogenic culture conditions, while human BMSCs (hBMSCs; see below) were used for all other studies.

hBMSCs were isolated from unprocessed human bone marrow aspirate (Lonza) on the basis of plastic adherence. Briefly, unprocessed bone marrow was plated at 2.5 × 10⁵ cells/cm² (estimated ~4,000–5,000 MSCs/cm²) in XPAN medium and expanded under physiological oxygen conditions (37°C in a humidified atmosphere with 5% CO₂ and 5% pO₂). Following colony formation, hBMSCs were trypsinised using 0.25% (w/v) Trypsin Ethylenediaminetetraacetic acid (EDTA) and tripotentiality was confirmed as previously described (Vinardell et al., 2009). hBMSCs used within the fibrin-based hydrogel to help support the development of a vascular network in vitro (see below) were seeded at 5,000 cells/cm² XPAN and expanded under normoxic conditions (37°C in a humidified atmosphere with 5% CO₂ and 20% pO₂) and used at P4. hBMSCs used for hypertrophic cartilage spheroids were expanded in the same media but under lower oxygen conditions (5% pO₂) and aggregated into microtissues at P3.

Human Umbilical Vein Endothelial cells (HUVECs) (Lonza, Walkersville, MD) were cultured at 2,500 cells/cm² in Endothelial Growth Medium (EGM-2) which had been supplemented with EGM-2 BulletKit (Lonza) at 37°C in a humidified atmosphere with 5 % CO₂ and 20 % pO₂. HUVECs were used at P4.

To enable the use of fibrin as a printable bioink, a gelatin carrier method was adapted from Kang et al. (2016). Briefly, hyaluronic acid (Sigma) was added to high glucose Dulbecco’s Modified Eagle Medium (hgDMEM) (Gibco, Biosciences) at a concentration of 3 mg mL^–1 and stirred overnight at 37°C. 10% (v/v) glycerol (Sigma) was then added and the solution was stirred for 1 h at room temperature. Gelatin type A (175 g bloom) (Sigma) was added at a concentration of 40 mg mL^–1 and stirred for 2 h at 37°C until fully dissolved. Before use, Fibrinogen (Sigma) was added to this thawed carrier gel at a concentration of 30 mg mL^–1 and stirred for 2 h at 37°C. To enhance the angiogenicity of the bioink, D-erythro-sphingosine-1-phosphate (S1P; Avanti Polar lipids, AL, United States) was added at a concentration of 125 μM mL^–1. For both cast constructs and printed constructs this fibrin-gelatin-hyaluronic acid formulation (from here on in termed a “fibrin-based” hydrogel or bioink) was used. To fabricate the prevascularising (PV) bioink, HUVECs and hBMSCs were loaded into the bioink at 3 × 10⁶ cells mL^–1 and 1.5 × 10⁶ cells mL^–1, respectively, as previously described (Nulty et al., 2021).

A micro-computed tomography (μCT) scan of a rat femur was first segmented and converted to STL file format using 3D Slicer free software¹. This STL was then converted to G code using BioCAM^TM and BioCAD^TM software packages (RegenHU, Switzerland). This scan was carried out as part of a protocol and study that were reviewed and approved by the ethics committee of Trinity College Dublin and the Health Products Regulatory Authority (HPRA) in Ireland (Approval—AE19136/P087).

Microtissue constructs were fabricated by both casting and by 3D bioprinting. Hypertrophic microtissues were loaded into either a fibrin-based hydrogel/bioink (microtissue only group) or a PV hydrogel/bioink (PV microtissue group) at 17 × 10³ spheroids mL^–1 (680/Construct). For the PV only group, PV bioink was used without any additional spheroids. Both cast and bioprinted constructs were supported by a 3D printed polycaprolactone (PCL) scaffold. Briefly, scaffolds were designed using BioCAD^TM software package and produced using the 3D Discovery multi-head bioprinting system (Regen Hu, Switzerland). Porous PCL (CAPA Ingevity, SC, United States) disc scaffolds were printed at 80°C at 0.5 MPa using a 27G needle (1 mm height, 4 mm diameter). Scaffolds were sterilised using ethylene oxide (EtO) gas prior to the addition of cell-laden hydrogel/bioinks. For cast constructs, a volume of 12.6 μL of cell/spheroid-laden hydrogel was manually deposited to fill the cylindrical PCL scaffolds. For bioprinted constructs, the fibrin-based bioinks were printed using a pneumatic driven syringe with a 25G needle at temperatures between (10–30)°C under pressures of between 0.05 and 0.2 MPa. To assist the accuracy of the printing process, 4 % (w/v) agarose (Sigma) well inserts were fabricated using custom-designed moulds printed with Form2 SLA printer (FormLabs, MA, United States). These agarose inserts were soaked in thrombin (20 U mL^–1) prior to printing, as a means of improving print fidelity of the fibrin-based bioink. After printing/casting, all constructs were immersed in a thrombin bath (20 U mL^–1) for 30 min at RT to allow the thrombin-catalysed polymerisation of fibrinogen to fibrin. The 3D Discovery was encased in a laminar flow hood to ensure sterility throughout the biofabrication process.

To investigate the effect of prevascularisation on a cartilage template in vivo, three groups were compared: A prevascularised fibrin-based hydrogel disc (PV only); a fibrin-based hydrogel disc containing hypertrophic microtissues (microtissues only); and a prevascularised fibrin-based hydrogel disc containing hypertrophic microtissues (PV microtissues). All disc constructs were fabricated by casting the microtissue laden hydrogels into moulds as described in section “Biofabrication of Prevascularised Constructs”. In addition, 3D bioprinting (see section “Biofabrication of Prevascularised Constructs”) was used to generate constructs mimicking the geometry of a segment of a rat femur. All constructs were then cultured in EGM-2 media supplemented with 50 ng mL^–1 of VEGF for 7 days under normoxic conditions to allow for the formation of microvessels. Constructs were then implanted subcutaneously into nude mice (see section In vivo Subcutaneous Implantation) and explanted at 4 and 8 weeks for analyses.

Constructs were implanted subcutaneously into the back of Balb/c female nude mice (Harlan, United Kingdom). Briefly, two subcutaneous pockets were made along the central line of the spine, one at the shoulders and the other at the hips. Two constructs were inserted into each pocket. Six constructs were implanted per group and constructs were harvested 2, 4, and 8 weeks post-implantation. Mice were anaesthetised using an intraperitoneal injection of xylazine hydrochloride and ketamine hydrochloride, Carprofen was added to water for 24 h post-surgery, and mice were sacrificed by CO₂ inhalation. This protocol and study were reviewed and approved by the ethics committee of Trinity College Dublin and the Health Products Regulatory Agency (HPRA, approval number AE19136/P069). Post-implantation samples were fixed in 4% PFA for 24 h.

Micro-computed tomography (μCT) scans were performed using a Scanco Medical 40 μCT system (Scanco Medical, Bassersdorf, Switzerland) in order to visualise and quantify mineral content and to assess mineral distribution within all constructs.

Constructs were scanned in 50% EtOH, at a voxel resolution of 12 μm, a voltage of 70 kVp, a current of 114 μA. Reconstructed 3D images were generated to visualise the mineral content and distribution throughout the constructs. A Gaussian filter (sigma = 1.2, support = 2) was used to suppress noise and a global threshold of 482.

In vivo samples were decalcified using “Decalcifying Solution-Lite” (Sigma) for approximately 1 week. Samples were frequently x-rayed to determine if any mineral content remained. When no mineral was visible the sample was considered decalcified. All samples (in vitro and decalcified in vivo) were then dehydrated in graded series of ethanol solutions (70–100%), cleared in xylene, and embedded in paraffin wax (all Sigma-Aldrich). Sections (5 μm) were rehydrated in graded series and stained with haematoxylin and eosin, 1 % (w/v) alcian blue 8GX in 0.1 M HCL to assess sulphated glycosaminoglycan (sGAG) content with a counter stain of 0.1% (w/v) nuclear fast red to assess cellular distribution, 0.1% (w/v) picrosirius red to assess collagen distribution, 0.2% (w/v) Safranin O to assess sGAG content post-implantation and Goldner’s trichrome (Groat’s iron haematoxylin, Fuchsine, Orange G, Fast Green; all from Sigma) for visualising bone (Mineralised bone = dark green, Fibrous Tissue = Light green, Osteoid = orange/red, erythrocytes = dark red). Immunohistochemistry was performed for collagen type I (col I) (Abcam ab90395 1:400), collagen type II (col II) (Santa Cruz- sc52658 1:400), and collagen type X (col X) (Abcam 49945, 1:200). Slides were then imaged using an Aperio ScanScope slide scanner. Samples stained for picrosirius red were imaged using polarised light microscopy (PLM) to determine collagen fibre orientation. In vivo sections were evaluated for vessel infiltration by counting vessels visible across an entire section using Aperia ImageScope and ImageJ software. Bone fraction of sections taken of the entire implant was determined using Goldner’s trichrome staining and MIPAR Image Analysis software (MIPAR Software LLC, United States) using native bone as a positive control for thresholding. It should be noted that PCL is cleared during the tissue processing and leaves empty spaces in constructs as a result.

Statistical analyses were performed using GraphPad Prism software (GraphPad Software, CA, United States). To analyse significant differences between two groups at one timepoint a standard two-tailed t-test was performed. To analyse variance between > 2 groups at one timepoint a one-way analyses of variance (ANOVA) was performed with Tukey post hoc test. To analyse variance between > 2 groups at multiple timepoints, two-way ANOVA was used with Tukey post hoc test. Numerical and graphical results are displayed as mean ± standard deviation unless stated otherwise. Significance was accepted at a level of p < 0.05.

Custom-made agarose microwell inserts were fabricated using a 3D printed positive mould (Figure 2). These moulds were designed to fit into a 6 well plate, creating 401 microwells per well (Figure 2A). The designed positive moulds were successfully fabricated using a consumer SLA printer, with excellent resolution and surface finish (Figure 2B). The conceived moulding process is easily implemented and highly reproducible, allowing the production of reliable microwells within each well from a non-adherent material. In addition, the reusable nature of the positive moulds and other assembly components (holders and screw) make this a cost-effective method for generating microwells. The addition of the chamfer to the top of each microwell created an upper surface without flat regions. This in-turn increased the seeding efficiency of the system by collecting all cells into the microwells and preventing their aggregation outside of the well. An early optimisation of the design comparing a chamfer and non-chamfer (flat top) design can be found in the supporting information, where the chamfer plus, centrifugation was found to enhance the development of cellular spheroids (Supplementary Figure 1).

We next sought to validate that the microwells supported the formation of BMSC spheroids and the growth of these cellular aggregates in vitro. Over 21 days of culture in chondrogenic media, the cellular spheroids were found to increase in size for all microwell cell seeding densities. Their final size was found to depend on the initial cell number used to generate the spheroid, with higher seeding densities generating larger spheroids (Figures 2D,E).

We next sought to assess the capacity of the microwell system to support the development of hypertrophic cartilage microtissues using hBMSCs at different cell densities (Figure 3A). The diameters of microtissues containing 1,000, 2,000, and 4,000 cells (per microtissue) were highly uniform after 24 h (SD < 0.025 mm) (Figures 3B,C). All microtissues increased in diameter with time in culture (Figures 3B,C), and following the 5-week culture period, the 1,000, 2,000, and 4,000 cells per microtissue had a mean diameter of 220, 320, and 360 μm, respectively (Figure 3C). To assess tissue development histologically, multiple microtissues were combined within a cell-free fibrin-based hydrogel and processed immediately. The microtissues stained positive for the canonical cartilaginous extracellular matrix markers, sulphated glycosaminoglycan (sGAG) and type II collagen, whilst calcium deposition (Figure 3D) indicated that the in vitro culture protocol successfully facilitated the production of hypertrophic cartilage microtissues. The presence of collagen types I and II, and lower levels of collagen type X, was detected using IHC after the 5 weeks of in vitro culture (Figure 3D).

Following 5 weeks of in vitro priming, microtissues were combined with either a control fibrin-based hydrogel (microtissues only) or a prevascularised fibrin-based hydrogel (PV microtissues; previously described in Nulty et al. (2021) and then cast into 3D printed polycaprolactone (PCL) discs (4 mm (ø) × 1 mm; Figure 4A). The PV fibrin-based hydrogel contained HUVECs and hBMSCs at a concentration of 3 × 10⁶ and 1.5 × 10⁶ cells mL^–1, respectively. These two groups, along with a PV only control without any microtissues, were cultured for 7 days in EGM-2 media containing VEGF (50 ng mL^–1) to allow for the formation of a microvascular network. We have previously demonstrated that such a fibrin-based hydrogel can support the development of vascular network when seeded with such a co-culture of HUVECS and BMSCs (Nulty et al., 2021). Micro-computed tomography (μCT) analysis performed on the constructs following 6 weeks of in vitro culture showed no detectable mineral deposition (data not shown). All groups were subsequently implanted subcutaneously into nude mice.

Following 4 and 8 weeks in vivo, negligible mineral was observed in the PV only controls. In contrast, it could be observed macroscopically that the cartilage microtissues had mineralised in vivo, taking on a bony appearance (Figure 4B). μCT analysis confirmed this observation (Figure 4C), with prevascularisation of the microtissues prior to implantation leading to significantly higher levels of mineralisation after 4 weeks in vivo (Figures 4C,D). No significant differences in construct mineralisation between the microtissue only group and the prevascularised microtissues group was observed at the 8-week time point. However, there was an increase in bone volume from 4 to 8 weeks in both groups (1.51 ± 0.73 mm³ and 1.82 ± 0.54 mm³ for microtissues only and PV microtissues, respectively), and a trend toward more bone in the PV microtissue group at the later timepoint (Figures 4C,E).

Histological analysis was carried out on the in vivo microtissue constructs to investigate vascularisation and bony-tissue formation after subcutaneous implantation. Safranin O staining suggested little or no cartilage remained in any implants (Figures 5A, 6A). Instead, Goldner’s trichome (GT) staining indicated that the hypertrophic microtissues had mineralised by week 4 (Figure 5A), and that this calcified tissue persisted at week 8 (Figure 6A). Blood vessels (yellow arrows) were present in all groups at both time points (Figures 5A, 6A). There was a trend toward increased vascularisation (quantified by a higher number of vessels and larger vessels in the prevascularised groups) at the 4 week time point (Figures 5C–E), although these differences were not statistically significant, and no such trends were observed at the 8 week time point (Figures 6C–E). Fusion between adjacent hypertrophic cartilage spheroids was noted in both microtissue groups at both time points (Black arrows, Figures 5A, 6A). At the earlier in vivo time point (4 weeks), histological quantification of bone fraction corroborated the μCT findings, indicating that prevascularisation accelerated mineralisation (Figure 5B). Furthermore, histological analysis at 8 weeks post-implantation demonstrated enhanced mineralisation by the prevascularised cartilage microtissues (Figure 6B).

As well as undertaking this proof of concept study to assess whether prevascularisation enhances the capacity of hypertrophic cartilage microtissues to generate bone in vivo, we also assessed whether 3D bioprinting could be leveraged to fabricate larger scale, anatomically defined constructs. To this end, hypertrophic microtissues were bioprinted into a 3D printed polycaprolactone (PCL) scaffold. The geometry of the implant was defined from a scan taken of a healthy rat femoral bone, a region used commonly as a defect model for studying long bone regeneration. Briefly, a μCT scan was taken of a rat femur, this scan was then segmented and converted to STL file format using 3D Slicer software (see text footnote 1). This STL was then converted to G code using BioCAM^TM and BioCAD^TM software packages and 3D printed (Supplementary Figure 3A). The polymer framework was designed to recapitulate the long bone cortex, as such, the thermopolymer printhead toolpath followed the outer perimeter of the model bone. This created an anatomically accurate PCL shell into which hypertrophic microtissues, combined with a PV bioink containing HUVECs and MSCs, were bioprinted (Supplementary Figure 3B). This bioprinted scaffold was then implanted subcutaneously into a nude mouse. After 4 weeks in vivo, there was evidence of mineralisation within construct showing, as proof of principle, that this approach was compatible with 3D printing technology without compromising biological efficacy (Supplementary Figures 3C,D).