Objectives Osteophytes are highly prevalent in osteoarthritis (OA) and are associated with pain and functional disability. These pathological outgrowths of cartilage and bone typically form at the junction of articular cartilage, periosteum and synovium. The aim of this study was to identify the cells forming osteophytes in OA.
Methods Fluorescent genetic cell-labelling and tracing mouse models were induced with tamoxifen to switch on reporter expression, as appropriate, followed by surgery to induce destabilisation of the medial meniscus. Contributions of fluorescently labelled cells to osteophytes after 2 or 8 weeks, and their molecular identity, were analysed by histology, immunofluorescence staining and RNA in situ hybridisation. Pdgfrα-H2BGFP mice and Pdgfrα-CreER mice crossed with multicolour Confetti reporter mice were used for identification and clonal tracing of mesenchymal progenitors. Mice carrying Col2-CreER, Nes-CreER, LepR-Cre, Grem1-CreER, Gdf5-Cre, Sox9-CreER or Prg4-CreER were crossed with tdTomato reporter mice to lineage-trace chondrocytes and stem/progenitor cell subpopulations.
Results Articular chondrocytes, or skeletal stem cells identified by Nes, LepR or Grem1 expression, did not give rise to osteophytes. Instead, osteophytes derived from Pdgfrα-expressing stem/progenitor cells in periosteum and synovium that are descendants from the Gdf5-expressing embryonic joint interzone. Further, we show that Sox9-expressing progenitors in periosteum supplied hybrid skeletal cells to the early osteophyte, while Prg4-expressing progenitors from synovial lining contributed to cartilage capping the osteophyte, but not to bone.
Conclusion Our findings reveal distinct periosteal and synovial skeletal progenitors that cooperate to form osteophytes in OA. These cell populations could be targeted in disease modification for treatment of OA.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
What is already known about this subject?
Osteophytes are a key feature of osteoarthritis. Several skeletal stem and progenitor cell populations have been identified; however, their contribution to osteophyte formation remains unexplored.
What does this study add?
This study shows that descendants of the Gdf5-expressing cells of the embryonic joint interzone form osteophytes in osteoarthritis, and reveals contributions of two distinct progenitor cell populations residing in periosteum and synovium.
How might this impact on clinical practice or future developments?
This study defines the progenitor cell subsets forming the osteophytes, which could be targeted as part of disease modification strategies for treatment of osteoarthritis.
A characteristic feature of osteoarthritis (OA) is the formation of osteophytes, which are osteocartilaginous outgrowths that typically form at the joint margins, in the region where the synovium attaches to the edge of the articular cartilage and merges with the periosteum. Osteophytes are established through growth of an initial cartilage template that is at least partially replaced with bone containing marrow cavities.1 2 At later stages, the bone is typically covered by a cartilage cap that can merge with the articular cartilage.1 2 Despite their high prevalence, the cell populations giving rise to osteophytes in OA remain to be defined.
Several skeletal stem/progenitor cell (SSC) populations have been identified which vary in their ability to form cartilage and bone during skeletal development, maintenance and repair. These include perivascular cells marked by expression of Pdgfrα and Sca1, Nestin (Nes) or leptin receptor (LepR), and Gremlin-1 (Grem1)-expressing cells.3–9 In addition, during bone regeneration, Sox9-expressing progenitors in the periosteum initiate cartilage callus formation by giving rise to cells that co-express chondrocyte and osteoblast markers.10 11 Recently, we used a Gdf5-Cre model that is active in the embryonic knee joint interzone, but not in the adult normal or OA knee to show that Gdf5-lineage descendants include SSCs in the adult with ability to repair a focal cartilage defect.12–16 The adult Gdf5 lineage includes Prg4-expressing progenitors in the superficial zone of articular cartilage and synovial lining.12 17–19
Here, we show that osteophytes derive from Pdgfrα-expressing Gdf5-lineage cells. These include Sox9-expressing progenitors in periosteum that give rise to hybrid skeletal cells that molecularly resemble those observed during bone repair, and Prg4-expressing cells from the synovial lining that supply chondrocytes to the outer cartilage layer but do not contribute to bone. Thus, our data define the progenitor cell subsets contributing to osteophyte formation in experimental OA.
We used Col2-CreER, 20 Pdgfrα-CreER, 21 Nes-CreER, 22 LepR-Cre, 23 Grem1-CreER, 9 Gdf5-Cre, 13 Prg4-GFP-CreER, 18 and Sox9-CreER mice,24 Cre-inducible TdTomato, 25 Confetti, 26 and mTmG reporter mice,27 and Pdgfrα-H2BGFP 28 and Nes-GFP 29 mice (see online supplementary table 1). Wild-type SWR/J mice were used for in situ hybridisation experiments. Mice were maintained on a 12:12 light-dark cycle, in a temperature-controlled room, with water and food ad libitum.
Tamoxifen administration and surgery
Administration of tamoxifen, dissolved in corn oil, was optimised for each mouse strain based on published studies9 11 18 30 31 and pilot experiments to achieve efficient labelling of intended target cells while minimising impact on animal welfare (see online supplementary table 2 for details). Adult male mice underwent surgical unilateral destabilisation of the medial meniscus (DMM) through resection of the medial menisco-tibial ligament,32 with the contralateral knee serving as unoperated or sham-operated control (see online supplementary table 2 or figure legends for age at surgery). Mice were anaesthetised with ketamine (50 mg/kg) and medetomidine (0.67 mg/kg) with atipamezole (1 mg/kg) postoperatively, ketamine (90 mg/kg) and xylazine (10 mg/kg), or isoflurane with 0.1 mg/kg buprenorphine subcutaneously for analgesia. Proliferating cells were labelled by subcutaneous injection of 2 mg bromodeoxyuridine (BrdU) immediately after surgery, followed by 1 mg/mL BrdU in drinking water for 2 weeks. Mice were euthanised for analysis 1, 2 or 8 weeks after surgery. Experimenters were not blinded.
Histology and immunohistochemistry
Tissues were fixed in 4% paraformaldehyde and decalcified in 4%–10% ethylenediaminetetraacetic acid (EDTA) in phosphate-buffered saline (PBS) or 33% (v/v) formic acid with 13.5% (w/v) trisodium citrate dihydrate. Samples were embedded in paraffin or frozen in OCT and sectioned. Sections were stained with safranin-O (Sigma), with or without fast green (Sigma). Fluorescent proteins were either detected by their native fluorescence in cryosections, or via immunofluorescence staining on paraffin sections,33 34 using antibodies listed in online supplementary table 3. Sections were counterstained with 4',6-diamidino-2-phenylindole (DAPI), Hoechst or TO-PRO-3.
Images were acquired using a Zeiss Axioscan Z1 slide scanner, Axioskop 40 with Progress XT Core 5 camera, Axio Imager 2.0 and Axio Observer Z1 with AxioCam MRm and AxioCam Mrc, 710 META Laser-Scanning Confocal Microscope, or Nikon AZ100 Microscope with a Nikon Digital sight DS-Fi1 camera. Image analysis was performed using ZEN (Zeiss), ImageJ and QuPath softwares.35 All analyses were performed on a minimum of three sections per sample. Percentages of labelled cells were calculated from the summed cell counts of all sections analysed for each sample, with marrow spaces excluded.
RNA in situ hybridisation
Fluorescence RNA in situ hybridisation was performed on 7 mm paraffin sections as described.36 RNA probes11 were generated following kit instructions (Sigma-Aldrich: 11277073910 and 11685619910) and were detected with Anti-Digoxigenin-POD (Sigma-Aldrich: 11207733910) and Anti-Fluorescein-POD (Sigma-Aldrich: 11426346910). For double-fluorescence in situ hybridisation, the TSA Cyanine 3 and Fluorescein system from Perkin Elmer was used as directed (NEL753001KT).
Cell isolation and flow cytometry
Cells were isolated from mouse knees as described,12 and stained with fixable viability dye eFluor780 (eBiosciences) and antibodies listed in online supplementary table 4. Data were acquired on a BD Fortessa flow cytometer and analysed using FlowJo v10 software. Unstained and single-labelled cells or antibody-labelled CompBeads (BD Biosciences) were used to set compensation. Debris and doublets were excluded based on Forward and Side Scatter parameters, dead cells were excluded based on viability dye staining, and gates were applied based on Fluorescence-Minus-One controls (online supplementary figure 1).
Statistical analysis was performed using GraphPad Prism V.5. Graphs show data points of individual mice, with mean±95% CI. N numbers indicate number of mice.
Osteophytes derive from Pdgfrα-expressing progenitors in periosteum and synovium
In the DMM model of OA in mice, osteophytes form at the joint margins, often merging with the articular cartilage, similar to human OA.1 2 We combined the DMM model with genetic cell-labelling and tracing models, to unravel the cell populations giving rise to osteophytes. Pdgfrα is broadly expressed by mesenchymal stem and progenitor cells. Pdgfrα-H2BGFP mice were, therefore, used to identify and trace Pdgfrα-expressing cells by long-lived GFP expression. We observed GFP+ cells throughout synovium and periosteum of the adult knee (figure 1a). In parallel, we tested whether pre-existing articular chondrocytes contribute to osteophytes using Col2-CreER;ROSA26:loxP-STOP-loxP-TdTomato (Tom) mice treated with tamoxifen at 2 weeks of age, when articular chondrocytes still express high levels of Col2a1 and are efficiently labelled (figure 1b).30 Two weeks after DMM in tamoxifen-induced Pdgfrα-H2BGFP;Col2-CreER;Tom mice, GFP+ chondrocytes were present throughout the osteophyte and were negative for Tom (figure 1c). Furthermore, there was negligible contribution from Tom+ pre-existing chondrocytes to mature osteophytes at 8 weeks post-DMM (figure 1d). These findings indicate that osteophytes do not develop from articular cartilage but derive from Pdgfrα-expressing stem/progenitor cells.
Next, we induced adult Pdgfrα-CreER;Confetti mice with tamoxifen to trace individual Pdgfrα-expressing cells through stochastic expression of membrane-bound CFP, nuclear GFP, cytoplasmic YFP or cytoplasmic RFP. Negligible fluorescence was detected in the absence of tamoxifen. In unoperated knees of tamoxifen-induced mice, fluorescently labelled cells were observed in periosteum and synovium (figure 1e). Consistent with previous studies utilising the R26-Confetti reporter,37 38 GFP was rarely detected and omitted from analysis. Two weeks after DMM, 23.9% (95% CI 17.3% to 30.6%, n=4) of cells in the osteophyte were labelled with either CFP, YFP or RFP (figure 1f,g). This may reflect the low sensitivity of the R26-Confetti reporter to Cre-mediated recombination,17 and aided the identification of distinct clonal cell populations. Intriguingly, clusters of identically coloured chondrocytes, likely derived from individual stem/progenitor cells, were observed in the deep periosteum at the bone surface, while distinct clusters of chondrocyte-like and fibroblast-like cells were found in the overlying synovial tissue (figure 1f). These data indicate that Pdgfrα-expressing stem/progenitor cells clonally expand and give rise to chondrocytes that form osteophytes in experimental OA, and suggest a dual contribution from cells in periosteum and synovium.
SSCs marked by expression of Nes, LepR or Grem1 do not give rise to osteophytes
Since SSC populations marked by expression of Nes, LepR or Grem1 express Pdgfrα,3 7–9 we investigated whether they contribute to osteophyte formation. At 2 or 8 weeks post-DMM, cells marked by Nes-GFP or Nes-CreER;Tom expression remained confined to vascular niches in synovium, periosteum and bone marrow, likely including pericytes and endothelial cells,8 9 39 with no detectable contribution to either cartilage or bone of the osteophyte (figure 2a–c). We next analysed LepR-Cre;Tom mice, since LepR-traced cells have been reported to make significant contributions to adult bone turnover and repair following fracture.8 LepR-traced cells were present in synovium and periosteum but negligible contribution to osteophytes was observed (figure 2d,e). We also used Grem1-CreER;Tom mice to trace Grem1-expressing SSCs, as they are distinct from Nes-GFP+ cells and contribute to fracture repair.9 Following tamoxifen induction at 7 weeks of age, we found no contribution of Grem1-traced cells to osteophytes at 2 weeks post-DMM (figure 2f). These data show that Nes-, LepR-, and Grem1-expressing SSCs do not form osteophytes.
Osteophytes arise from adult progeny of the Gdf5-expressing embryonic joint interzone
Cells in adult knees that are traced from Gdf5-expressing joint interzone cells in the embryo are present in synovium and adjacent periosteum (figure 3a), and include SSCs.12 Analysis of cells isolated from knees of Pdgfrα-H2BGFP;Gdf5-Cre;Tom mice (figure 3b) revealed that the vast majority of Tom+ Gdf5-lineage cells express Pdgfrα-H2BGFP (93.6%, 95% CI 87.6% to 99.6%, n=9), with Gdf5-lineage cells constituting approximately one-third of all Pdgfrα-H2BGFP+ cells (figure 3c,d). Both Pdgfrα-expressing Gdf5-lineage cells (GFP+Tom+) and other Pdgfrα-expressing cells (GFP+Tom-) expressed, to varying degrees, the mesenchymal stromal cell and fibroblast markers podoplanin (Pdpn/Gp38), CD90, CD73, CD51 and CD105, while neither population included haematopoietic cells (CD45+), endothelial cells (CD31+), or erythrocytes (Ter-119+) (figure 3e–h, online supplementary figure 1). These findings indicate that adult Gdf5-lineage cells are a subset of Pdgfrα-expressing cells that may form osteophytes.
We, thus, induced DMM in Gdf5-Cre;Tom mice, followed by 2 weeks of BrdU administration to label proliferating cells. At 2 weeks post-DMM, Gdf5-lineage cells had extensively proliferated and expanded (figure 4a,b), and they were major contributors to Col2-expressing chondrocytes in the osteophytes (figure 4b). Tom+ Gdf5-lineage cells constituted 82.5% (95% CI 65.1% to 99.8%, n=4), and BrdU+ cells constituted 82.4% (95% CI 69.9% to 94.8%, n=4), of cells within osteophytes at 2 weeks post-DMM (figure 4c). Tom+ Gdf5-lineage cells remained abundant in mature osteophytes at 8 weeks post-DMM (figure 4d). They included 87.7% (95% CI 80.0% to 95.4%, n=7) of cells in the cartilage cap and 70.8% (95% CI 63.6% to 78.1%, n=7) of osteocytes in the bone (figure 4e), as well as bone lining cells at endosteal surfaces (figure 4d). These data indicate that the Pdgfrα-expressing progenitors that form osteophytes are largely contained within the joint-resident Gdf5-lineage population, which undergo extensive proliferation to supply cells that form both cartilage and bone.
Sox9-expressing progenitors give rise to hybrid skeletal cells to initiate osteophytes
Next, we sought to refine which progenitor populations contribute to osteophytes. Clonal tracing of Pdgfrα-expressing cells had indicated a possible dual origin from periosteum and synovium (figure 1f). Sox9-expressing progenitors in adult periosteum supply skeletal cells to the callus during femoral fracture repair and large-scale rib regeneration.10 11 We therefore performed lineage tracing of adult Sox9-expressing cells in experimental OA by treating Sox9-CreER;Tom mice with tamoxifen prior to DMM surgery. Mice not treated with tamoxifen showed absence of Tom expression. In knees of tamoxifen-treated uninjured mice, Tom+ cells were detected in periosteum (figure 5a). At 2 weeks post-DMM, we observed Tom+ Sox9-traced chondrocytes in the early osteophyte (figure 5b), thus identifying Sox9-expressing progenitors as important contributors of osteophytes.
During regeneration of the adult mouse rib bone, Sox9-expressing periosteal cells form a callus composed of hybrid skeletal cells; these ‘hybrid’ cells are characterised by strong coexpression of chondrocyte and osteoblast genes.11 Similarly, Col2a1-expressing cells in the forming osteophyte coexpressed the osteoblast and hypertrophic chondrocyte marker Ocn as early as 1-week post-DMM (figure 5c). In contrast to growth plate chondrocytes, they also co-expressed Col1a1 (figure 5c), which was particularly apparent in the large osteophytes that typically form on the Col2-CreER background (online supplementary figure 2). At 2 weeks post-DMM, Col2a1-expressing chondrocytes were also positive for the osteoblast and hypertrophic chondrocyte marker Spp1, and the hypertrophic chondrocyte marker Col10a1 (figure 5d). We confirmed in the Sox9-CreER;Tom model that Sox9-traced cells are the source of at least some hybrid cells, based on co-localisation of Tom with Col2a1 and Col1a1 mRNA expression (figure 5e). These findings indicate that the early osteophyte contains Sox9-derived hybrid skeletal cells similar to those described for rib bone regeneration.
Prg4+ progenitors contribute to cartilage but not to bone in osteophytes
The Gdf5 lineage includes Prg4-expressing cells in the synovial lining, which expand in response to acute focal cartilage injury.17 We, therefore, investigated whether Prg4-expressing cells contribute to osteophyte formation. We first used Prg4-CreER;ROSA26:loxP-membrane-Tomato-loxP-membrane-GFP (mTmG) mice induced with tamoxifen at 7 weeks of age. No GFP was detected in mice not treated with tamoxifen. At 2 weeks post-DMM, we observed expansion of GFP + Prg4-traced cells in synovium, and Prg4-traced cells were found in the outer region of the early osteophyte, while minimal contributions of these cells to the deeper hybrid skeletal cells, expressing Col2a1 and Col1a1, was observed (figure 6a,b). To confirm these findings and determine the role of Prg4-traced cells at later stages of osteophyte formation, we performed similar Prg4-tracing experiments using the Cre-inducible Tom reporter model, and extended analysis to 8 weeks post-DMM (figure 6c–h). Osteophytes in Prg4-CreER;Tom mice at 2 weeks post-DMM were typically more advanced than those observed in the Prg4-CreER;mTmG model, with a layer of cartilage surrounding a hypertrophic centre undergoing remodelling to bone (figure 6d). As well as expanding in synovium, Tom+ Prg4-traced cells constituted 41.9% (95% CI 27.7% to 56.1%, n=8) of chondrocytes embedded in a cartilage matrix immunostaining for Col2 (figure 6d,e). Consistent with the data in the Prg4-CreER;mTmG model (figure 6b), they were predominantly found in the outer region, with some contribution to Col10-expressing hypertrophic chondrocytes in deeper regions of the osteophyte (figure 6f). At 8 weeks post-DMM, Tom+ Prg4-traced cells persisted in the cartilage cap covering the mature osteophyte but were barely detected in bone (figure 6g,h). Thus, Prg4-lineage cells from the overlying synovial tissue supply chondrocytes to the forming osteophyte, and while they persist in the cartilage cap of the mature osteophyte, they make negligible contributions to bone.
Osteophytes are a key feature of OA, and their occurrence is a criterion for imaging-based diagnosis of OA.40 In peripheral joint OA, osteophytes are associated with pain, knee structural progression and incidence of joint replacement.41–43 Nonetheless, research in OA pathogenesis has largely focused on mechanisms of articular cartilage breakdown, while the extensive joint remodelling events have been considered secondary. Osteophytes, however, are not always linked to severity of articular cartilage loss.44 Understanding the biology of osteophyte formation will provide critical insights in the mechanisms underlying the structural derangements that occur in OA joints.
We show that osteophytes primarily arise from descendants of Gdf5-expressing embryonic joint interzone cells that reside in the adult knee. Together with our previous data showing that Gdf5-lineage cells underpin synovial hyperplasia and cartilage repair after injury,12 these findings point to a central role of Gdf5-lineage cells in the maintenance, repair, and remodelling of adult joint tissues. Although osteophytes typically develop close to articular cartilage, Col2a1-expressing chondrocytes from articular cartilage did not give rise to osteophytes. Instead, we show that osteophytes originate at least in part from a population of Sox9-expressing progenitors in periosteum, with progeny of Prg4-expressing synovial-lining cells supplying chondrocytes to the cartilage but not osteoblast-lineage cells that form the bone of the osteophyte.
Several SSC populations have been implicated in bone fracture repair, including SSCs identified by expression of Nes, LepR or Grem1.4 8 9 45 We observed negligible contribution of these SSC populations to osteophytes in OA. Yet intriguingly, our data indicate that the initial stages of osteophyte formation are similar to endochondral bone repair in the mouse femur10 and rib11, as well as the zebrafish lower jaw.36 During bone repair, Sox9-expressing cells in the periosteum supply chondrocytes and osteoblasts, and help to orchestrate callus formation.10 11 Furthermore, the early callus of the regenerating rib includes Sox9-lineage cells with a hybrid chondrocyte/osteoblast identity,11 similar to what we observed in the early stages of osteophyte formation.
Until this study, it was not known whether the osteophyte derives from a single progenitor population, or whether multiple progenitor populations cooperate to form the different tissue layers of the osteophyte. Our data indicate that within the Gdf5-lineage population that forms all parts of the osteophyte, periosteal Sox9-expressing progenitors give rise to the deeper hybrid cells that form the ossifying cartilage template, while synovial Prg4-expressing cells supply chondrocytes but make negligible contributions to osteoblast-lineage cells. Recently, it was shown that joint development occurs through a continuous influx of new cells into the Gdf5-expressing joint interzone and flanking regions, with cells being temporally specified to contribute differentially to the multiple tissues of the joint.46 Our data show that lineage fate determination persists in the adult joint between subpopulations of Gdf5-lineage cells and suggest the coexistence of distinct progenitor cell subsets with restricted differentiation potential that may have become imprinted through development.
Human osteophytes in OA hip and knee joints share similar pathological features to those seen in mice, with endochondral bone covered by a cartilage cap that merges with or overgrows the articular cartilage.2 Notably, our molecular phenotype data in mouse are consistent with observations in human, where an overlap of Col1 and Col2 expression is found in the early stages of osteophyte formation, and chondrocytes in the osteophyte express osteocalcin.2 47 Adult human synovium and periosteum are known to contain mesenchymal progenitors,48 49 and synovium-derived mesenchymal stem cells can orchestrate a joint morphogenetic process.12 Various molecular factors and pathways have been implicated in osteophyte formation, including transforming growth factor β and bone morphogenetic protein (BMP) signalling.1 Here, we propose a model whereby joint-resident progenitor cell subsets in periosteum and synovium that ontogenetically derive from the embryonic joint interzone respond to such signals and cooperate to form osteophytes in OA (figure 7). Our data define progenitor cell subsets that could be targeted as part of disease modification strategies for treatment of OA.
The authors thank all members of the Arthritis and Regenerative Medicine Laboratory at the University of Aberdeen; Susan Clark, Shina Ardani and William Alton for their contributions to data acquisition; Animal Facility staff for care of our animals; staff in the Microscopy and Histology Facility and Iain Fraser Cytometry Centre at the University of Aberdeen.
Handling editor Josef S Smolen
Twitter @littlecb5001, @Francesca Mariani @MarianiLabUSC
AJR and KK contributed equally.
Contributors AJRo: Conceptualisation, experimental design, data acquisition, analysis and interpretation, writing of the manuscript; KK: Experimental design, data acquisition, analysis and interpretation, contributed to writing of the manuscript; AJRa and FLC: Data acquisition, analysis and interpretation, contributed to writing of the manuscript. MSa and SK: Data acquisition, analysis and interpretation; JSm, MSe, LR and CBL: Data acquisition and analysis; HW, RG: Data acquisition; CK and SMF: Experimental design and data acquisition; FVM, TP and JFB: Conceptualisation, experimental design, data interpretation; JSh: Conceptualisation, experimental design, data acquisition, analysis and interpretation; JGC: Conceptualisation, experimental design, data acquisition, analysis and interpretation, contributed to writing of the manuscript; CDB: Conceptualisation, experimental design, data analysis and interpretation, writing of the manuscript.
Funding CDB, AJRo, KK, AJRa, FLC and HW were supported by funding from Versus Arthritis, formerly Arthritis Research UK (20775, 21156, 20050, 19429), and the Medical Research Council (MR/L020211/1). TP, JSh, MSa and RG were supported by funding from the Bundesministerium für Bildung und Forschung (BMBF) Overload-PrevOP consortium (01EC1408F) and the Innovative Medizinische Forschung (IMF) Programme of the University Hospital Münster (Project I-SH121608). JGC, STK, JSm, MAS and FVM were supported by funding from the National Insitute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS: R01AR069700). JFB, LR and CBL were supported by the National Health and Medical Research Council (NHMRC: APP1063133), and part funding was provided to JFB by the Victorian Government‘s Operational Infrastructure Support Programme to the Murdoch Children’s Research Institute. CK and SMF were supported by funding from the Wellcome Trust (203151/Z/16/Z), Horizon2020 (ERC-2014-CoG-648765), Cancer Research UK (C61367/A26670) and NHS Blood and Transplant.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Patient consent for publication Not required.
Ethics approval All animal experimental protocols were approved by the UK Home Office and the Animal Welfare and Ethical Review Committees of the University of Aberdeen and University of Cambridge, the University of Southern California Institutional Animal Care and Use Committee, the Murdoch Children’s Research Institute Animal Ethics Committee, or the Animal Use Committee for University Hospital Münster.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data supporting the findings of this study are available within the Article and its online supplemental information files, or are available from the corresponding authors on reasonable request.