Synovial joint cavitation during limb skeletogenesis entails Na/K-ATPase ion pump expression and osmoregulatory activity

ABSTRACT

Limb synovial joints rely on a water/lubricant-filled cavity to sustain skeletal movement through life, but mechanisms generating the cavity within the primordial joint interzone during embryogenesis remain unclear. Because water accrual would be needed to create and inflate the cavity, its formation may require energy-consuming osmoregulatory mechanisms able to draw water and exert mechanical force. Our in situ hybridization and scRNAseq data reveal that interzone cells in developing mouse embryo joints strongly expressed the Na/K-ATPase ion pump genes Atp1a1, Atp1b1 and Atp1b3. There was also local and specific expression of water channel aquaporin 1 (Aqp1) and mechano-sensing genes. When pregnant mice were administered ouabain, which is a physiological glycoside that limits pump activity and osmoregulatory processes, joint cavitation in embryos was inhibited, as was lubricant gene expression. Joint development depends on signals from Indian hedgehog-expressing growth plate chondrocytes. Interference with hedgehog signaling coordinately inhibited pump, mechano-sensing and lubricant expression and cavitation. Our data provide a new understanding of joint cavitation as an energy-requiring osmoregulatory process that accrues a water-based fluid from interstitial and transcellular sources, and establishes a fluid-filled cavity in coordination with long bone development.

Summary: This study provides evidence that joint cavitation is an energy-requiring osmoregulatory process leading to the formation of a fluid-filled synovial cavity that is needed for long-term joint function and maintenance.

INTRODUCTION

The synovial joints provide the skeleton with mobility, flexibility and physical functionality, and also exert shock absorption roles (Archer et al., 1999; Longobardi et al., 2015). They exhibit diverse organizations, shapes and sizes, each fitted to anatomical requirements and type of motion, but share basic components, including articular cartilage, a capsule and a fluid-filled synovial cavity. Articular cartilage is an extracellular matrix-rich tissue that provides resilience during physical activities (Buckwalter and Mankin, 1997; Hunziker et al., 2007). The capsule surrounds the entirety of the joint, displays an intimal synovial lining composed of resident fibroblasts and macrophages, and has roles in overall joint homeostasis (Culemann et al., 2019; Kemble and Croft, 2021). The ligaments assist joint movement and also provide proprioceptive input (Johansson et al., 1991; Moffat et al., 2008). The water/fluid-filled cavity provides lubrication and sustains frictionless functioning of the joints (Rhee et al., 2005). These and other basic aspects of joint biology are well established, as is the susceptibility of joints to problematic diseases such as osteoarthritis (OA) (Loeser et al., 2012; Sandell, 2012). In contrast, much remains unclear about how the joints form during development, and acquire their distinct traits and, in particular, their water/lubricant-filled synovial cavity (Chijimatsu and Saito, 2019; Decker, 2017; Rux et al., 2019; Salva and Merrill, 2017). If such developmental information were available and clear, it would have fundamental value and could also help to design therapeutic strategies to overcome the notoriously poor repair capacity of joints (Johnstone et al., 2013).

In early mammalian and avian embryonic limbs, joint formation starts with the emergence of a mesenchymal tissue termed the interzone, flanked by cartilaginous primordia of long bones (Holder, 1977; Mitrovic, 1978). The interzone cells are bound to each other by tight junctions, have a scanty extracellular matrix and have a fibroblastic appearance (Archer et al., 1994). In collaborative studies, we showed for the first time that the interzone cells are instrumental participants in joint development (Koyama et al., 2008). Relying on the fact that the interzone cells express growth and differentiation factor 5 (Gdf5) (Storm and Kingsley, 1999), we used Gdf5⁺ cell lineage tracing to uncover the fact that the cells give rise to most, if not all, joint tissues over developmental time, including articular cartilage, intrajoint ligaments and capsule (Koyama et al., 2008). We also discovered that the initial Gdf5 interzone cell cohort was supplemented over time by additional Gdf5-expressing cells emerging from the immediate joint surroundings, contributing to the lateral expansion of the joints (Decker et al., 2017; Pacifici et al., 2006). Similar observations have been reported in other studies (Shwartz et al., 2016). It has also been shown that Gdf5⁺ cells are present in adult joints and can be recruited into action in response to joint injury (Collins et al., 2023; Decker et al., 2017; Roelofs et al., 2017). These and other studies have provided crucial insights into basic aspects of synovial joint development.

What has remained largely unsolved, however, is how a cavity emerges within the compacted interzone and becomes filled with its essential water-based lubricating fluid (Rux et al., 2019). Studies going back decades reported that joint cavitation is due to muscle-driven limb motion. Cavitation defects occurred after administration in ovo of neuromuscular paralyzing agents (Murray and Drachman, 1969; Osborne et al., 2002) or in muscle-less mouse embryo mutants (Kahn et al., 2009; Nowlan et al., 2010). Other studies proposed that limb joint cavitation is attributable to local production of hyaluronic acid (HA) (Craig et al., 1990; Edwards et al., 1994), also reflecting its water imbibition and swelling capacity (Cowman, 2017; Fraser et al., 1997). Conditional ablation of the major HA synthase Has2 throughout the limb mesenchyme with Prx1-Cre led to a delay in cavitation along with skeletal defects (Matsumoto et al., 2009). Though tantalizing, the above studies left many questions unanswered. For example, cavitation of the knee was not affected in the muscle-less mouse embryo mutants (Kahn et al., 2009; Nowlan et al., 2010), indicating that motion does not account for cavitation. In addition, HA is produced by many solid tissues, including the cartilaginous growth plates, indicating that HA per se cannot trigger cavitation. We therefore wanted to investigate what mechanisms could bring about joint cavitation.

Cavity and lumen formation has been studied in several other developmental processes and systems, and causative mechanisms have been described (Le Verge-Serandour and Turlier, 2022; Navis and Bagnat, 2015; Sigurbjornsdottir et al., 2014). For example, the gut tube in zebrafish derives from an initial solid rod of tightly packed endodermal cells (Wallace and Pack, 2003). Lumen formation has been shown to start with the emergence of multiple microlumen among the endodermal cells that coalesce into a single entity over time. The driving force is provided by micro-luminal and luminal water accrual, which is caused by electrochemical Na⁺ gradients set up by cell surface Na/K-ATPase ion pumps (Bagnat et al., 2007). Lumen formation is inhibited by treatment with the glycoside ouabain, which is an endogenous physiologic molecule that inhibits Na/K-ATPase pump function and osmoregulatory activity (Hamlyn and Blaustein, 2016). Studies on pre-implantation mouse embryos showed that formation of the fluid-filled blastocoel also starts with emergence of microlumens throughout the embryo, coalescing to produce the single blastocoel over time (Dumortier et al., 2019). Steep osmoregulatory gradients generated by Na/K-ATPase pump activity have been found to draw water, causing hydraulic fracturing of cell-cell contacts and formation of microlumens coalescing to a blastocoel thereafter (Dumortier et al., 2019; Le Verge-Serandour and Turlier, 2022; Schliffka et al., 2024). Notably, our own recent studies have revealed that joint cavitation in mouse embryo limbs also initiates with microcavities, coalescing into a single synovial cavity over time (Kim et al., 2022). As in the processes above, joint cavitation occurs within a highly compacted structure – the interzone – and would thus require active water accrual to create and inflate the developing microcavities and cavity, and, in turn, forcibly distance the flanking skeletal elements. Here, we ask whether energy-consuming osmoregulatory mechanisms driven by Na/K-ATPase ion pumps are involved in limb joint cavity development. Our data provide strong support for this previously unreported notion.

RESULTS

Single cell RNAseq identifies diverse cell populations in developing joints

To better define the cell populations involved in joint formation and cavitation, we first carried out single cell RNAseq (scRNAseq) analyses of cells isolated from E13.5, E14.5 and E16.5 mouse embryo prospective knee joints (Fig. 1A,C,E). The joints were dissected out along with surrounding tissues, including epiphyseal cartilage, facilitating dissection (Fig. 1B,D,F). The E13.5 joints displayed a still solid interzone (Fig. 1A,B), the E14.5 joints displayed a less compacted interzone (Fig. 1C,D, arrowheads) and the E16.5 joints were nearly fully cavitated (Fig. 1E,F). E15.5 joints were not included here, owing to their phenotypic closeness to E16.5 (Kim et al., 2022). Cells (a minimum of 10,000 per sample) were used to create libraries with Chromium Controller (10x Genomics) that were then barcoded and sequenced on an Illumina NovaSeq platform (Klein et al., 2015). Data were processed with Cell Ranger to obtain unique molecular identifiers (UMIs). We used the Seurat analysis package V4 (Stuart et al., 2019) for filtering, variable gene selection and dimensionality reduction, and for eliciting uniform manifold approximation and projection (UMAP) and graph-based clustering. We first superimposed the datasets from the three developmental stages after removing non-mesenchymal lineage cells and identified eight major cell population clusters (Fig. 1G) based on differential gene expression (Fig. S1). The bulk of synovial joint progenitor cells demarcated by Gdf5 was present in cluster 1, whereas chondrocytes at various stages of proliferation and maturation were in clusters 2, 3 and 4 (Fig. 2 and Fig. S1), referred to as chondrogenic 1, 2 and 3. Cells with phenotypic traits shared with osteogenic, tenogenic, fibrogenic and periosteal progenitors were in the remaining clusters (Fig. 1G and Fig. S1).

Fig. 1.

scRNAseq identifies cell populations during knee joint development. (A-F) Histological sections of developing knee joints in E13.5, E14.5 and E16.5 mouse embryos stained with Safranin O and Fast Green (Saf-O/FG) to reveal cartilaginous and non-cartilaginous tissues, respectively. (B,D,F) Higher magnifications of the images in A,C,E, showing a largely intact interzone at E13.5 (B), a less compacted interzone at E14.5 (D, arrowheads) and a cavitated interzone at E16.5 (F). (G) Combined UMAP plot of knee cells at the three stages above. Eight major cell populations are identified based on lineage-characteristic genes (see Fig. S1), including joint progenitors demarcated by Gdf5 expression in cluster 1 and chondrogenic populations at proliferation, maturation and hypertrophy stages in clusters 2, 3 and 4, respectively. fe, femur; ti, tibia. (A-F) n=6 independent replicates. (G) n=1 per embryonic age. Scale bars: 100 µm (bars in C and D apply to C,E and D,F, respectively).

Fig. 2.

UMAP plots depict genes defining cell identity in distinct clusters. (A) Individual UMAP plots of E13.5, E14.5 and E16.5 knee cell populations. (B) Gdf5-expressing joint progenitors in cluster 1 at each stage. (C) Prg4 characterizing joint progenitors and chondrocytes in upper cluster 2 region. (D) Cartilage matrix marker aggrecan (Acan) characterizing chondrocytes in clusters 2, 3 and 4. (E) Transcription factor Creb5 labeling joint progenitors in cluster 1 and other clusters, including fibrogenic cells. (F) Ubiquitin ligase Wwp2 marking chondrocytes within clusters 2, 3 and 4, including those in the upper cluster 2 region. (G) Growth plate marker gene Ucma characterizing clusters 2 and 3 chondrocytes, particularly by E16.5. n=1 per embryonic age.

Examination of E13.5, E14.5 and E16.5 datasets separately (Fig. 2A) showed that the Gdf5-expressing cells in cluster 1 were present at each stage, as expected (Fig. 2B). Their articular descendants expressing the characteristic anti-adhesive glycoprotein lubricin (Prg4) became evident by E16.5 (Decker et al., 2017; Koyama et al., 2008) and occupied part of cluster 2 (Fig. 2C), an otherwise large cluster containing chondrocytes expressing the major cartilage proteoglycan aggrecan (Acan), and likely including articular and non-articular cells (Fig. 2D). It became important to consider the above data in the context of recent studies on other regulators of joint development and articular chondrocyte origin and function, specifically CREB5 and WWP2 (Mokuda et al., 2019; Zhang et al., 2022). In good agreement, Creb5-expressing cells were present at every stage and largely overlapped with Gdf5-expressing joint progenitors in cluster 1 (Fig. 2E). Wwp2-expressing cells were also present at every stage but mainly occupied clusters 2, 3 and 4, where proliferating and maturing chondrocyte populations resided (Fig. 2F). Wwp2 did express the upper zone of growth plate- and matrix-associated marker gene Ucma by E16.5 (Fig. 2G) (Seuffert et al., 2018).

Next, we carried out developmental trajectory analyses to more firmly establish Gdf5⁺ progenitor/cell-progeny relationships (Fig. S2A). The Gdf5⁺ progenitors occupied the trajectory start and bifurcated into two main branches (Fig. S2B): one directed toward chondrogenic cells, and a second one directed toward tenogenic and fibrogenic cells, which are components of joint tissues (Fig. S2B). Such trajectories are in line with our original genetic cell lineage-tracing studies (Koyama et al., 2008). To further investigate this issue, we selected out and interrogated in greater detail the Gdf5 progenitor population in cluster 1 and the chondrogenic population in clusters 2, 3 and 4, using Monocle. Pseudo-time analysis revealed that the proportion of Gdf5-expressing progenitors decreased over time while the proportion of cells expressing cartilage marker genes such as Acan, Col2 and Sox9 increased (Fig. S2C-E). This indicates a loss of progenitor characteristics and acquisition of differentiated traits over time, and again suggests a cell progenitor/progeny relationship among these clusters. Further analysis using heatmap visualization of gene expression along pseudo-time highlighted several transcription factors that displayed dynamic changes of expression and might regulate the progenitor-to-chondrocyte transition (Fig. S2F). In the Gdf5-expressing progenitor population (cluster 1), such genes included Sox4 and Sox11, which are known for roles in maintaining stemness (Fig. S2F, bottom panel) (Foronda et al., 2014; Tsang et al., 2020). In the Sox9-expressing chondrogenic population (cluster 2) (Lefebvre and Smits, 2005), the genes included Runx2, Runx3 and Rarg, which are known for their roles in chondrogenesis and cartilage phenotypic expression (Williams et al., 2009; Yoshida et al., 2004).

Single cell RNAseq reveals Na/K-ATPase ion pump expression patterns

As indicated above, the Na/K-ATPase ion pumps are essential cell surface components (Blanco, 2005; Kaplan, 2002). They hydrolyze ATP to force three Na⁺ ions out of the cell in exchange for two K⁺ ions entering the cell, establishing basic vital cellular processes and functions, including osmoregulatory ion concentration gradients, Na⁺-coupled solute transport, and exchange and flow of water. Each pump is composed of a 10-pass transmembrane α-subunit and a single-pass β-subunit (Clausen et al., 2017), with an additional component prominent in certain organs (Geering, 2005). Four α-subunit isoforms (α1 to α4, encoded by the Atp1a1 to Atp1a4 genes) and three β-subunit isoforms (β1 to β3, encoded by the Atp1b1 to Atp1b3 genes) occur in mammals and are expressed in varying spatiotemporal manners, providing a measure of specificity (Li and Langhans, 2015; Orlowski and Lingrel, 1988). Greater specificity of action is provided by a level of expression that can vary from cell type to cell type and is fine-tuned to meet local osmoregulatory needs (Clausen et al., 2017; El Mernissi and Doucet, 1984).

ATP1α1 is produced by many cell types and at different levels (Clausen et al., 2017; Li and Langhans, 2015). We found that Atp1a1-expressing cells were present to varying degree in all clusters at each stage, including in Gdf5-expressing cluster 1 (Fig. 3A). ATP1β1 and ATP1β3 serve as physiological partners of ATP1α1 in several tissues (Clausen et al., 2017; Li and Langhans, 2015). The overall number of Atp1b1- and Atp1b3-expressing cells was in general lower than that of Atp1a1-expressing cells, but increased from E13.5 to E16.5 and was characteristic of cluster 1 joint progenitors (Fig. 3B,C). We interrogated the datasets for expression of other Atp1 family members and found them to be minimally expressed (Fig. S3).

Fig. 3.

UMAP analysis uncovers ion pump, aquaporin and osmoregulatory gene expression patterns. Plots at E13.5, E14.5 and E16.5 were surveyed separately to identify cell populations expressing (A) Atp1a1, (B) Atp1b1 and (C) Atp1b3 in several clusters, including joint progenitors in cluster 1. (D) Aqp1-expressing cells are present in a few clusters but prominently in joint progenitor cluster 1. (E) Set of over 100 osmoregulatory genes (referred to as ‘osmotic score’) characterizing joint progenitors, particularly those in cluster 1 by E16.5. n=1 per embryonic age.

The Na/K-ATPase pumps establish electrochemical gradients that regulate key processes, including osmotic-driven transcellular flow of water via the cell-surface channel aquaporins (Agre et al., 1993; Day et al., 2014). Indeed, aquaporin 1 (encoded by Aqp1) was strongly expressed in the developing joints and specifically characterized cluster 1 joint progenitors by E14.5 and E16.5 (Fig. 3D), overlapping Atp1a1, Atp1b1 and Atp1b3 in those cells. The other members of the aquaporin family were expressed broadly but much less prominently (Fig. S4). The overlap of Aqp1 and Atp1 gene expression among cluster 1 joint progenitors likely reflected an increase in local fluid tonicity. This was strongly supported by the observation that a set of over 110 genes known to be upregulated in relation to high tissue osmolarity, collectively termed the osmotic score (GO:0006970), became very prominent among the cluster 1 Gdf5 progenitors by E16.5 (Fig. 3E).

Spatiotemporal expression of pump and aquaporin members is linked to cavitation

The data above predicted that expression of osmoregulatory pump and aquaporin genes should be spatiotemporally linked to the cavitation process. To test this thesis, we processed serial histological sections of E13.5, E14.5, E15.5 and E16.5 mouse embryo knees for RNAscope in situ hybridization. Safranin-O/Fast Green staining verified that the plane and depth of tissue sectioning captured the stepwise and stage-dependent process of joint formation and cavitation, including a compacted interzone at E13.5 (Fig. 4A,B), intermediate microcavity phase at E14.5 (Fig. 4H,I) and full cavitation occurring between E15.5 and E16.5 (Fig. 4O,P,V,W) (Kim et al., 2022). This developmental cascade was reiterated by the fact that the interzone marker Gdf5 was expressed from early stages onwards (Fig. 4C,J,Q,X), whereas the joint lubricant Prg4 [undetectable at E13.5 and E14.5 (Fig. 4D,K)] increased steadily between E15.5 and E16.5, coinciding with full cavitation (Fig. 4R,Y). In line with scRNAseq data above, we found that Atp1a1 was expressed broadly and at basic levels at E13.5 and E14.5 joints (Fig. 4E,L) but became appreciably upregulated within interzone and joint tissues with increasing developmental time (Fig. 4S,Z). In comparison, Atp1b1 transcripts were scanty in limb tissues and joints at E13.5 and E14.5 (Fig. 4F,M), but became prominent and largely restricted to cavitating joint surface tissues by E16.5 (Fig. 4T,Z1). Also in agreement with the scRNAseq data, Atp1b3 transcripts, as well as those for Aqp1, were continuously appreciable but, quite notably, were largely restricted to the interzone and immediately surrounding tissues at E13.5 (Fig. S5A,B and Fig. 4G, respectively). These patterns became more obvious with increasing time (Fig. S5C-F and Fig. 4N,U,Z2, respectively).

Fig. 4.

Developing knee joints exhibit restricted ion pump and Aqp1 spatiotemporal gene expression patterns. (A-Z2) Serial parasagittal sections of E13.5 (A-G), E14.5 (H-N), E15.5 (O-U) and E16.5 (V-Z2) knee joints stained with Safranin O/Fast Green (Saf-O/FG) to reveal cartilage and non-cartilaginous tissues or processed for RNAScope. (B,I,P,W) Higher magnification images of knee joints within the corresponding panoramic images A,H,O,V. (C,J,Q,X) Gdf5 transcripts are restricted to joint interzone and tissues at each stage. (D,K,R,Y) Prg4 expression is negligible at E13.5 and E14.5 (D,K) but specifically increases in joint cells thereafter (R,Y). Atp1a1 and Atp1b1 transcripts are present broadly and basally at E13.5 and E14.5 (E,F,L,M) but are upregulated in joint cells at E15.5 and E16.5 (S,T,Z,Z1). (G,N,U,Z2) Aqp1 expression strongly and specifically characterizes joint cells at each stage. n=6 biological replicates from independent experiments, 6-9 sections per sample, a total of 36-54 sections per embryonic age. Scale bars: 100 µm (bar in A applies to A,H,O,V; bar in B applies to B-G; bar in I applies to I-N,P-U,W-Z2).

To extend the data above, we carried out RNAscope analyses on E17.5 mouse embryo digits that exhibit interphalangeal joints at different developmental stages along their proximal-distal axis (Fig. 5A,G,M,S) (Rux et al., 2021). While all joints expressed Gdf5 (Fig. 5B,H,N,T), Prg4 expression was essentially undetectable in the youngest distal joints (Fig. 5C) and became evident in older proximal joints, including tarsal joints (Fig. 5I,O,U). Atp1a1 transcripts were broadly present in distal joints and surrounding tissues (Fig. 5D), and there was a more restricted and upregulated expression of this gene within more proximal and advanced joints (Fig. 5J,P,V). Atp1b1 expression was evident and largely restricted to the joints (Fig. 5E,K,Q,W), and this restriction was even more so for Aqp1 expression (Fig. 5F,L,R,X). Clearly, osmoregulatory pump and aquaporin genes become preferentially expressed as cavitation advances, which physically distances the articulating surfaces in both knee- and digit-developing joints.

Fig. 5.

Developing interphalangeal joints express osmoregulatory genes. (A-X) Parasagittal sections through E17.5 tarsal regions display developmentally older joints proximally and younger joints distally. While Gdf5 is expressed in all joints (B,H,N,T), Prg4 transcripts are evident proximally (O,U) but less so in distal joints (C,I). Atp1a1 and Atp1b1 transcripts are broadly expressed but are clearly more abundant in the joints (D,J,P,V and E,K,Q,W, respectively). As seen in the knee, Aqp1 expression is very strong and restricted within the digit joints (F,L,R,X). n=6 biological replicates from independent experiments, 6-9 sections per sample, a total of 36-54 sections examined. Scale bars: 100 µm (bar in A applies to A-R; bar in S applies to S-X).

Mechano-sensing genes are expressed in developing joints

Piezo 1 and 2 are cell-surface mechano-sensing channels expressed and activated in cells experiencing biophysical forces, including high osmotic pressure and mechanical stretch (Nourse and Pathak, 2017; Syeda et al., 2016). Indeed, Piezo1 and Piezo2 were specifically expressed in interzone cells and joint tissues at E13.5 (Fig. 6A-D), E14.5 (Fig. 6F,G,I,K) and E15.5 (Fig. 6N-Q), becoming more restricted to cavitating cells and tissues along the joint line by E16.5 (Fig. 6S-V, arrowheads). Selectivity of expression was reaffirmed by analysis of adjacent growth plates in which Piezo1, but not Piezo2, was expressed in the hypertrophic zone, as reported previously (Fig. 6H,J, respectively, arrowhead) (Zhou et al., 2020). Piezo1 and Piezo2 operate through their downstream effector Yap1 (Nourse and Pathak, 2017), which was preferentially expressed in the interzone (Fig. 6E,M,R), becoming prominent in cavitating tissues by E16.5 (Fig. 6W, arrowhead). Yap1 expression also characterized the hypertrophic zone (Fig. 6L, arrowhead), in line with previous studies (Zhou et al., 2020). Increases in cluster 1 cells expressing Piezo1, Piezo2, Yap1 and Wwtr1 (a Yap1 interacting partner) (Todaro et al., 2018) over joint developmental time were also evident in our scRNAseq data (Fig. S6A-D). Together, the data provide further support to the notion that interzone cells experience physical forces and do so from early stages on, likely caused first by neighboring muscle contractions (Kahn et al., 2009; Nowlan et al., 2010), and then by increasing fluid tonicity and osmolarity as cavitation proceeds.

Fig. 6.

Mechanosensing genes characterize developing joints. Serial sections of E13.5 (A-E), E14.5 (F-M), E15.5 (N-R) and E16.5 (S-W) knee joints were processed for histochemical Safranin O/Fast Green staining or RNAScope. (B,G,O,T) Higher magnification images of A,F,N,S, focusing on the joint area. Piezo1, Piezo2 and Yap1 expression characterizes interzone and joint tissues at E13.5 (C-E), E14.5 (I,K,M) and E15.5 (P-R), becoming restricted to joint cells along the cavity by E16.5 (U-W, arrowheads). Piezo 1 and Yap1 transcripts characterize the hypertrophic zone of the growth plate (H,L, arrowheads) but Piezo 2 transcripts are undetectable (J, arrowhead), data are in line with previous studies. n=6 biological replicates from independent experiments, 6-9 sections per sample, a total of 36-54 sections per embryonic age. Scale bars: 100 µm (bar in B applies to B-E; bar in G applies to G,I,K,M; bar in H applies to H,J,L; bar in O applies to O-R; bar in T applies to T-W).

To establish the general validity of the above observations, we re-examined Visium spatial transcriptomic data from human embryonic limbs described in a recent atlas (Zhang et al., 2023). In the study, developing knees were specifically examined at post-gestation weeks 6.2-8, when they are actively undergoing cavitation (Gasser, 1975). As expected, strong expression of the stereotypic joint marker genes GDF5 and PRG4 was apparent (Fig. S7A,B). Strikingly, there was also preferential expression of PIEZO1, PIEZO2 and YAP1, as well as of ATP1A1, ATP1B1, ATP1B3 and AQP1 in the joint compared to adjacent tissues, including most epiphyseal cartilage (Fig. S7C-I). We also re-examined a follow-up multi-omic study from the same group that focused on developing human embryonic skeletogenic cells (To et al., 2024). In it, the authors examined knee cell populations isolated from pre-cavitation interzone stages (about 7.3-7.6 post-conception weeks, PCW) to after-cavitation stages (up to about 11.2 PCW). Analysis of those datasets verified that GDF5-expressing interzone cells were most numerous at interzone stages, up to about 7.3 PCW, and then decreased over time (Fig. S8). In contrast, there were increases in cells expressing AQP1, ATP1B1, ATP1B3, PIEZO1 and WWTR1 as the joints underwent cavitation over time, whereas ATP1A1⁺ and YAP1⁺ cell number did not change markedly (Fig. S8), as also seen in mouse. Finally, we re-examined scRNAseq datasets from a previous study on developing mouse embryo knee joints at E12.5, E13.5, E14.5 and E15.5 that made use of sorted Gdf5-YFP-expressing joint cells (Bian et al., 2020). Whereas the number of Atp1a1-expressing cells did not change much over time, there was a clear increase in Atp1b1- and Aqp1-expressing cells (Fig. S9). Taken together, the above studies concur with the notion that osmoregulatory genes are an integral part of mechanisms regulating joint development and cavitation in both human and murine limbs.

Ouabain administration delays cavitation in knee and digit joints

To directly test the roles of osmoregulatory mechanisms, we used ouabain, an endogenous glycoside that normally limits the activity of the Na/K-ATPase pumps and, in turn, osmoregulatory gradients (Hamlyn and Blaustein, 2016). Ouabain has long been used clinically (Smith, 1988) and is used experimentally to interrogate ion pump function in diverse contexts, including cavity formation in developmental processes, as noted above (Bagnat et al., 2007; Dumortier et al., 2019; Le Verge-Serandour and Turlier, 2022). Accordingly, we administered ouabain (1 or 5 mg/kg) (Guo et al., 2020) or vehicle to pregnant mice by daily subcutaneous injection from E12.5 to E15.5 of gestation and harvested the embryos at E16.5 (Fig. 7A). Histochemical staining showed that control E16.5 knee joints had a nearly complete cavity (Fig. 7C, arrowheads) and displayed the expected expression patterns for Gdf5, Prg4, Atp1a1, Atp1b1, Aqp1, Piezo1, Piezo2 and Yap1 (Fig. 7E to L). Knee joints in embryos following administration of 1 mg/kg ouabain were essentially normal (see Fig. 7B). In contrast, joint cavitation was visibly delayed in embryos from 5 mg/kg ouabain-receiving mice, with the opposing cartilaginous epiphyseal ends still connected by a largely continuous interzone tissue (Fig. 7D, arrowheads). This delay was accompanied by decreases in expression of Prg4, Atp1a1, Atp1b1, Piezo1, Piezo2 and Yap1 (Fig. 7N-P,7R-T), but expression of Gdf5 and Aqp1 was unaffected and remained at control levels (Fig. 7M,Q). Phenotypic changes of such impact were observed in over 60% of embryos from 5 mg/kg ouabain-treated mice in three independent experiments, as determined quantitatively (Fig. 7B).

Fig. 7.

Ouabain administration delays cavitation. (A) Schematic showing that ouabain was administered to pregnant mice daily at 1 or 5 mg/kg from E12.5 to E15.5, and embryos were harvested on E16.5 along with embryos from mice receiving vehicle. (B) Quantification of full or delayed cavitation at E16.5, with significant delay in cavitation observed in 5 mg/kg ouabain group compared to other groups (P<0.05, n=12 embryos). Chi-square and Fisher's Exact tests were used. (C,D) Frontal (left images) and parasagittal (right images) knee sections from vehicle- or 5 mg/kg ouabain-treated mice processed for histochemical Safranin O/Fast Green staining demonstrate complete cavitation in controls (C, arrowheads) and markedly reduced cavitation after ouabain treatment (D, arrowheads). (E-T) RNAScope analysis showing that control E16.5 joints exhibit the expected patterns of expression of Gdf5 (E), Prg4 (F), Atp1a1 (G), Atp1b1 (H), Aqp1 (I), Piezo1 (J), Piezo2 (K) and Yap1 (L). The expression of all genes was markedly decreased in ouabain-treated joints (N-P,R-T), with the exception of Gdf5 and Aqp1 (M,Q). (U-Z6) Underlying tibia growth plates from same samples were examined by Safranin O/Fast Green staining (U-V,Z1-Z2) and expression of the mitotic marker Ki67 (W,Z3), aggrecan (Acan) (X,Z4), Ihh (Y,Z5) and collagen X (ColX) (Z,Z6). Growth plate structure and expression patterns were all invariant in vehicle versus ouabain-treated specimens. (Z7,Z8) Quantification of the height of the Ihh-expressing prehypertrophic zone and ColX-expressing hypertrophic zone, revealing lack of statistically significant differences in control versus treated samples. Unpaired Student's t-test; P<0.05, n=5 embryos/group. Data are shown with 95% confidence intervals. For histological and gene expression analyses, n=3 biological replicates from independent experiments, 6-9 sections per sample, a total of 18-27 sections examined at E16.5. Scale bars: 100 µm (bars in C apply to C,D; bar in E applies to E-T; bar in U applies to U,Z1; bar in V applies to V-Z,Z2-Z6).

To test whether a similar delay in cavitation would occur in developing digit joints, ouabain was administered to pregnant mice (5 mg/kg) daily from E12.5 to E18.5 and embryos were harvested at E19.5 (Fig. S10A). This longer treatment schedule was needed because digit joints develop later than knees. Ouabain treatment did delay digit joint cavitation (Fig. S10N) compared to vehicle-receiving controls (Fig. S10E) in over 55% of embryos analyzed (Fig. S10B). This was linked to reduced expression of Prg4, Atp1a1, Atp1b1, Piezo1, Piezo2 and Yap1 expression (Fig. S10P-R,T-V) versus control (Fig. S10G-I,K-M), but Gdf5 and Aqp1 were largely unaffected (Fig. S10O,S) versus control (Fig. S10F,J).

Ouabain is a potent biological agent and could elicit side effects or changes in other cellular mechanisms (Blaustein and Hamlyn, 2020). Aiming to exclude generalized side effects, we further assessed the overall structure and phenotype of the cartilaginous elements in the same control and treated embryos as above, using histochemical and gene expression analyses. There appeared to be no major organizational differences in overall long bone organization and growth plate phenotype in vehicle versus 5 mg/kg ouabain-treated embryos, except for a slightly shorter diaphyseal region (compare Fig. 7U,V with Fig. 7Z1,Z2, arrowheads). Likewise, there were no major changes in patterns of Ki67-expressing proliferating cells (Fig. 7W vs 7Z3) or in the gene expression of aggrecan (Acan) and collagen X (Col X), which normally characterize upper and hypertrophic growth plate zones, respectively (compare Fig. 7X,Z with Fig. 7Z4,Z6) (Kronenberg, 2003). Differences were noted in the pre-hypertrophic zone expressing Indian hedgehog (Ihh) that varied depending on peripheral versus central growth plate location (compare Fig. 7Y with Fig. 7Z5). Quantification of serial sections confirmed that there were no significant differences in control and treated growth plate zones (Fig. 7Z7,Z8). Finally, we found that the number of dying TUNEL-positive cells was very low within joints and growth plates in both control or ouabain-treated E16.5 embryos in both long bones and digits (compare Fig. 8A-H with Fig. 8I-P), confirmed by quantification (Fig. 8Q-U). The only consistent presence of TUNEL-positive cells occurred at the chondro-osseous border in the growth plate where some hypertrophic chondrocytes normally undergoes programmed cell death before replacement by bone cells (Fig. 8F,N, arrowheads) (Shapiro et al., 2006). This was also invariant in control versus ouabain-treated specimens.

Fig. 8.

Cell death is minimal in developing joints and long bones. (A-P) Knee, ankle, tarsal and phalangeal joints, and adjacent skeletal elements in embryos from vehicle- (A-H) and 5 mg/kg ouabain-administered (I-P) mice were examined by histological Safranin O/Fast Green staining and for the presence of TUNEL-positive dying cells. Virtually no dying cells are present in most joints, regardless of ouabain treatment, except for a few positive cells in phalangeal joints (H,P). Consistent presence of TUNEL-positive cells characterizes the chondro-osseus border of growth plates (F,N, arrowheads), in line with the physiological apoptosis known to occur but invariant after ouabain treatment. (Q-S) Quantification of TUNEL-positive cells shows no significant differences in samples from control versus ouabain-treated (1 or 5 mg/kg) mice. n=6 biological replicates from independent experiments, 6-9 sections per sample, a total of 36-54 sections examined. There were no statistically significant differences among samples. Scale bars: 100 µm (bar in A applies to A-C,E-G,I-K,M-O; bar in D applies to D,H,L,P).

To further exclude the possibility of side effects, we asked whether the delay in cavitation caused by ouabain administration was reversible upon drug withdrawal. Pregnant mice were given ouabain (5 mg/kg) or vehicle from E12.5 to E15 daily, and their embryos were then harvested at E19.5 (Fig. S11A,B). We found that cavitation was evident and seemingly normal in all knees (Fig. S11C-E), which was further confirmed by quantification (Fig. S11F).

Inhibition of hedgehog signaling delays cavitation

Limb joint development is known to require hedgehog signaling stemming from Ihh-expressing pre-hypertrophic chondrocytes in neighboring long bone growth plates (St-Jacques et al., 1999). To test whether hedgehog signaling is an upstream regulator and coordinator of osmoregulatory gene expression and cavitation, we administered a single dose of the clinically relevant hedgehog signaling antagonist LDE225 (Byrne et al., 2018) to pregnant mice at E13.5 of gestation and harvested the embryos at E16.5, as described above (Fig. 9A). The E13.5 stage was chosen as starting point because it is immediately before the onset of cavitation and because it tolerates LDE225 administration well compared to earlier stages (Jiang et al., 2019). In line with Jiang et al.’s study, the overall body appearances of the control versus LDE-treated E16.5 embryos were similar (Fig. 9B,D), but the limb length and joint angulation at elbow and knee were appreciably reduced in treated embryos (Fig. 9B,D, arrowheads). Long bones and knee joints in controls had their typical cavitated appearance (Fig. 9F,G), whereas the knee joint was clearly un-cavitated in LDE225-treated mice (Fig. 9H-M). Data from three companion embryos are shown to demonstrate the consistency of findings (Fig. 9H-L). There was a marked decrease in expression of Prg4, Atp1a1, Atp1b1, Piezo1, Piezo2 and Yap1 in knees from LDE225-treated mice (Fig. 9W-Y,Z1-Z3) as compared to controls (Fig. 9O-Q,S-U). As also seen in the experiments with ouabain above, expression of Gdf5 and Aqp1 was largely unaffected by LDE225 administration (compare Fig. 9N,R with Fig. 9V,Z).

Fig. 9.

Interference with hedgehog signaling derails cavitation. (A) Schematic showing that LDE225/Sonidegib (100 mg/kg) was administered once to pregnant mice at E13.5, and embryos were harvested at E16.5 along with control embryos. (B-E) Gross anatomical images showing largely normal appearance of body and limbs in control (B,C) and LDE-treated (D,E) embryos, except for a slight change in joint angulation (compare B with D, arrowheads). (F-M) Histochemical images showing a normal cavitated knee processed for histochemical Safranin O/Fast Green staining in controls (F,G) and un-cavitated knee joints in LDE-treated embryos (H-M). Images from three individual treated embryos (H-M) show data consistency. (N-Z3) RNAScope data showing that control E16.5 joints exhibit expected expression patterns of Gdf5 (N), Prg4 (O), Atp1a1 (P), Atp1b1, (Q) Aqp1 (R), Piezo1 (S), Piezo2 (T) and Yap1 (U). (V-Z3) The expression of all genes was decreased in the joints of LDE-treated mice (W-Y,Z1-Z3), with the exception of Gdf5 and Aqp1 (V,Z). For histological and gene expression analyses, a total of 12 limbs from independent experiments were used (n=12), with 6 sections per slide and 8-10 slides per limb across two experimental groups. Scale bars: 100 µm (bar in F applies to F,H,J,L; bar in G applies to G,I,K,M; bar in N applies to N-Z3).

Analysis of digit joints in E19.5 embryos from pregnant mice injected once with LDE225 at E13.5 (Fig. S10C) showed that the hedgehog antagonist delayed cavitation in most joints (Fig. S10D). This was accompanied by downregulation of Prg4, Na/K-ATPase ion pumps, Piezo1, Piezo2 and Yap1 gene expression (Fig. S10Y-Z1,Z3-Z5), with a lower or no effect on Gdf5 and Aqp1 expression (Fig. S10X,Z2).

DISCUSSION

The cavitation process in developing limb joints has remained an intriguing but unsolved puzzle for years and, thus, our data represent a previously unreported and major step ahead in deciphering its mechanisms. The data now reveal that cavitation entails local upregulation of osmoregulatory Atp1a1, Atp1b1 and Atp1b3 genes, along with strong and joint-restricted expression of the water channel Aqp1. When ouabain was acutely administered with the goal of interfering with ATP1 ion pump activity, cavitation was delayed, as demonstrated by persistence of a largely solid interzone and by decreased Prg4 expression. As indicated above, the Na/K pumps establish and govern electrochemical ion gradients that regulate osmotic pressure and movement of solutes and water within cellular, pericellular and tissue milieus (Blanco, 2005; Kaplan, 2002). It is thus likely that the upregulation of Atp1a1, Atp1b1 and Atp1b3 in developing joints could locally strengthen the Na⁺ gradient and increase the concentration of Na extracellularly, providing a driving osmotic force to attract and accrue water for cavity initiation, progression and expansion (see our working model in Fig. 10). Ouabain would inhibit this basic but crucial mechanism, delaying cavitation. Our data and conclusions relate very well to the studies cited above showing that gut lumen development in zebrafish and blastocoel formation in preimplantation mouse embryos rely on similar ion pump-driven osmoregulatory and water-drawing mechanisms, all blocked by ouabain treatment (Bagnat et al., 2007; Dumortier et al., 2019; Schliffka et al., 2024). Thus, fundamentally similar mechanisms of cavitation appear to be used in a variety of biological circumstances and are conserved across species, including humans, as indicated by our re-analysis of embryonic limb spatial transcriptome atlases (To et al., 2024; Zhang et al., 2023).

Fig. 10.

Working model of joint development and cavitation. (A) At E12.5-E13.5, the Gdf5-expressing mesenchymal interzone in developing limb joints is solid and composed of tightly bound cells, flanked by most epiphyseal chondrocytes (Ch). Green indicates where further joint development would be promoted by the influence of continuous Ihh signaling from the growth plate. In addition, actions of local proteases would reduce the cell-cell links and interactions among the interzone cells. (B) At E14.5-E15.5, upregulation of Na/K ion pump expression and activity would increase ion concentration extracellularly, and elicit recruitment and accrual of water via transcellular and intercellular routes, leading to formation of microcavities. (C) From E16.0 onwards, continuous Na/K pump activity would permit further water accrual, promoting the establishment of a joint-wide single cavity. Production of lubricants, including Prg4/lubricin and hyaluronic acid, would provide the synovial fluid with physiological anti-adhesive properties, sustaining long-term friction-less joint motion.

By virtue of their ion pump capacity, the Na/K-ATPases are vital effectors and regulators of osmoregulatory mechanisms. However, these integral membrane proteins can exert other important roles in certain contexts and influence phenotypic activities such as cell signaling (Clausen et al., 2017; Li and Langhans, 2015; Pierre and Xie, 2006). Their α- and β-subunits can interact with Src, protein phosphatase 2, epithelial growth factor receptor and other regulatory proteins, in turn activating signaling pathways, including Akt, Erk and MAPK, and affecting cell behavior and gene expression (Haas et al., 2002; Kimura et al., 2011; Li and Langhans, 2015). In addition, the extracellular domain of the α-subunit acts as the receptor for ouabain and other endogenous or exogenous glycosides, and the resulting interactions can influence Na/K-ATPase roles in cell signaling or as modulators of K⁺ and Ca²⁺ channels in several biological contexts (Blaustein and Hamlyn, 2020). In skeletal biology, these and other ion channels continue to receive much attention in view of their function in adult joint and articular cartilage homeostasis as well as age-related joint disease (Zhou et al., 2024), although their possible involvement in joint development remains to be tested. While we interpret our data as evidence that the Na/K-ATPases act primarily as ion pumps and elicit osmotic and mechanically driven formation of the synovial cavity in embryonic limbs, there is a possibility that these proteins may also exert non-pump functions in joint formation, e.g. influencing Erk signaling that is known to be activated at cavitation (Bastow et al., 2005).

A defining feature of the interzone is that it is initially very compact and composed of connexin 43-bound mesenchymal cells (Archer et al., 1994). To explain cavitation, it was proposed long ago that many interzone cells would undergo cell death, liberating the flanking epiphyseal cartilaginous ends of long bones from each other and creating a tissue discontinuity able to be turned into a cavity (Abu-Hijleh et al., 1997). However, the occurrence of cell death was not confirmed in follow up studies (Ito and Kida, 2000; Kavanagh et al., 2002) and we do not observe it here either (see Fig. 8). Yet the interzone cells need to separate along the future joint line, in turn favoring the emergence of microcavities first and a single, consolidated, joint-wide synovial cavity thereafter. But how could this occur? In the study on blastocoel formation cited above, the authors relied on the microscopic transparency and approachability of the murine blastulas to directly measure the hydrostatic pressure/tonicity of fluid in the microlumens between blastomeres and found it to be 10-fold higher than basal physiological value; such hypertonicity was lost upon ouabain treatment (Dumortier et al., 2019). This and other data led the authors to conclude that the hydrostatic pressure was strong enough on its own to cause hydraulic fracturing of cell-cell contacts, facilitating microlumen formation first and then microlumen coalescence into a single blastocoel thereafter. If hydrostatic pressure/tonicity of such a magnitude were to occur in the joint interzone because of local upregulated Na/K-ATPase pump expression, it could theoretically break cell-cell contacts by itself. Our data showing the strong and preferential expression of mechano-sensor components such as Piezo1, Piezo2 and Yap1 indicate that the cavitating interzone cells do very likely experience and sense strong physical forces (Roelofs et al., 2017; Young and Kobayashi, 2023). Other studies have also demonstrated that the interzone cells at cavitation express proteases such as Adamts-1 that are able to degrade pericellular components, including versican, thus reducing physical cell-cell contacts and links (Capehart, 2010; Nagchowdhuri et al., 2012). Considering all the above, it is more likely that limb joint cavitation involves multiple steps, with a first one being the loosening of cell-cell contacts by protease action without affecting interzone cell vitality and density (model in Fig. 10). This would be accompanied by or immediately followed by local increases in Na/K-ATPase ion pump expression and activity, establishing the electrochemical gradients that are able to accrue water via interstitial and transcellular routes to physically distance the cells, and create microcavities first (Kim et al., 2022) and a consolidated cavity thereafter. Muscle-driven joint motion could indirectly help cavitation by angulating the opposing epiphyseal ends (Nowlan et al., 2010). Subsequent upregulation of HA production and Prg4 expression would add lubricating macromolecules to the incipient watery synovial fluid (Kim et al., 2022), sustaining frictionless joint sliding short and long term (Blewis et al., 2007) (Fig. 10). These sequential steps would act in concert with hedgehog signaling provided by the growth plate, coordinating joint formation and cavitation with long bone development and elongation.

In contrast to energy-consuming Na/K-ATPase pumps, the aquaporins are cell surface channels that permit passive transcellular flow of water and other small molecules via facilitated diffusion, driven by electrochemical gradients (Agre et al., 1993; Day et al., 2014). The aquaporin family is large and includes 13 members that are expressed differentially in tissues and organs, have intrinsic characteristics, and can be modified post-translationally, fine-tuning their roles and action to local needs and requirements (Li et al., 2020). For example, aquaporin 2 is mainly expressed in kidney and regulates water diffusion, and aquaporin 8 is expressed in pancreas and glands, and regulates water and ammonia diffusion and exchange (Fushimi et al., 1993; Liu et al., 2006). Aquaporin 1 is expressed in several tissues where it controls diffusion of water but also CO₂ (Geyer et al., 2013). In view of these studies, the strong and restricted expression of Aqp1 in the interzone prior and during cavitation stands out for its level and topographical specificity. In addition, Aqp1 expression in the interzone was unaffected by ouabain or LDE225 treatments, unlike genes such as Prg4, which was downregulated. Given these characteristics, what roles could aquaporin 1 play in the developing joints? It should be remembered that the initial interzone is not only compacted and with a scant extracellular matrix, but is also exceedingly hypoxic, also a reflection of the lack of blood vessels (Provot et al., 2007). Thus, an abundance of aquaporin 1 from early stages on may permit flow and exchange of water, and other small solutes and metabolites, including CO₂, among the compacted interzone cells and surrounding environment, sustaining cell vitality and function. Later in development, the protein could also facilitate transcellular water flow into the incipient synovial cavity, and contribute to fluid buildup and cavity inflation, as proposed above. Ongoing studies aiming to conditionally ablate Aqp1 should help to sort through these intriguing possibilities.

Once created and established, the synovial fluid is needed throughout life to protect the joints and maintain frictionless motion (Blewis et al., 2007; Rhee et al., 2005). Not surprisingly, there continues to be much interest in studying synovial fluid composition, as well as its deterioration in pathological conditions such as OA, and there is little doubt that loss of lubricating macromolecules in disease contributes to joint pathologies (Catterall et al., 2010; Loeser et al., 2012). Supplementations of lubricants such as HA is one of the standard-of-care clinical treatments for joint disease (Teeple et al., 2011). On the other hand, much less attention has been paid to the equally crucial issue of how the overall volume of synovial fluid, and its water content in particular, is maintained through life. Water is the largest component of synovial fluid and is required for solubility and functioning of its lubricants (Tamer, 2013). Because the joints freely exchange water, solutes and nutrients with surrounding tissues and vessels (Geborek et al., 1989; Haywood and Walsh, 2001), mechanisms are needed to retain and control water within the cavity and prevent its reabsorption into, and loss to, the general circulation, in turn maintaining overall synovial fluid volume. An answer to this conundrum comes from the few but felicitous studies showing that synovial fluid in normal healthy adults is hyperosmotic (410-440 mOsm) when compared to basal isotonicity of interstitial tissue fluids or blood, which is about 300 mOsm (Deng et al., 2022; Eltawil et al., 2015; Shanreld et al., 1988). Very interestingly, the tonicity of synovial fluid in OA patients was found to drop to about 300 mOsm. It is thus the normal hypertonicity of synovial fluid that most likely permits its homeostatic retention and volume stabilization in healthy adult joints, properties that are hampered in disease and accounting for fluid volume loss in joint disease. Members of the Na/K-ATPase ion pump family have long been known to be expressed in adult articular chondrocytes where they are found to sustain cell functions including coping with a proteoglycan-driven high ionic microenvironment (Deng et al., 2022; Mobasheri et al., 1997). However, the pumps in joint cavity lining tissues could also maintain the hypertonicity of synovial fluid (see model in Fig. 10), enabling it to stabilize water content and cavity fluid volume. This is reminiscent of what the pumps would do in the genesis of the cavity itself as our data now reveal. Drugs exist that enhance Na/K-ATPase ion pump activity (Li and Langhans, 2015; Therien and Blostein, 2000). They could be tested for their potential to ameliorate joint disease on their own or in a combination therapy with lubricants.

MATERIALS AND METHODS

Study approval

The experimental animal protocols were approved by the Institutional Animal Care and Use Committees at the Children's Hospital of Philadelphia. The experiments were performed in the animal facilities of this institution, which implement strict regimens for animal care and use. In accordance with the standards for animal housing, mice were group housed at 23-25°C with a 12 h light/dark cycle and allowed free access to water and standard laboratory food pellets.

Single cell transcriptome analyses

Knee regions were dissected from E13.5, E14.5 and E16.5 C57BL/6 mouse embryos using microsurgical procedures with the aid of a Leica dissecting microscope, regions that included the developing joint itself and neighboring tissues. Samples were finely minced and dissociated into single cell suspensions by treatment with dispase (50 U/ml in HBSS) for 2 h at 37°C. Cells were then suspended in 5% FBS/DMEM at a concentration of 700-1200 cells/µl. Cell suspensions were immediately delivered to the Center for Applied Genomics (Children's Hospital of Philadelphia) for determination of cell viability and sample processing. Viability of cells was >95% in all samples. cDNA libraries of cells from each group were generated by Chromium controller (10X Genomics), barcoded and purified as described by the manufacturer, and sequenced using a 2×150 pair-end configuration on an Illumina NovaSeq platform at a sequencing depth of ∼400 million reads. Cell Ranger (version 7.1.0) was used to de-multiplex reads, followed by extraction of cell barcode and unique molecular identifiers (UMIs). The cDNA insert was aligned to a modified reference mouse genome (mm10).

Seurat package V4 was used for filtering, variable gene selection, dimensionality reduction analysis and clustering according to standard methods (Stuart et al., 2019). Doublets or cells of poor quality (with more than 6000 genes, with fewer than 200 genes or with more than 5% of genes mapping to the mitochondrial genome) were excluded. Expression was natural log transformed and normalized for scaling the sequencing depth to a total of 1×10⁴ molecules per cell. For the integrated dataset, batch integration was performed using Harmony (version 1.0) (Korsunsky et al., 2019). The cell-cycle heterogeneity effect was also regressed out by using the Seurat Cell-cycle scoring function. Statistically significant principal components were selected as input for uniform manifold approximation and projection (UMAP) plots. Different resolutions for clustering were used to demonstrate the robustness of clusters. In addition, differentially expressed genes within each cluster relative to the remaining clusters were identified using FindMarkers within Seurat. Sub-clustering was performed by isolating the mesenchymal and skeletogenic lineage clusters using known marker genes, followed by re-analysis as described above. Gene ontology analysis was performed using the clusterProfiler package (Guangchuang et al., 2012).

To computationally delineate the developmental progression of Gdf5-expressing interzone cells into other joint cell populations and order them in pseudotime and trajectory, we used the algorithms of Monocle package V2 (Trapnell et al., 2014). We defined genes characterizing the cell differentiation trajectory by selecting those with high dispersion across cells, using a parameter of ‘mean_expression >=0.05 and dispersion_empirical >=1 * dispersion_fit’. The gene list was further used for dimensional reduction to generate the trajectory reconstruction using the nonlinear reconstruction algorithm DDRTree. Heat maps to visualize pseudotime-dependent genes were generated by using a subset of topmost significant differentially expressed genes. We performed trajectory analysis using Slingshot (Street et al., 2018). Briefly, UMAP was used as dimensional reduction after the principal component analysis (PCA) was calculated for individual or integrated datasets. Then Seurat objects were transformed into SingleCellExperiment objects. Slingshot trajectory analysis was conducted using the Seurat clustering information and with dimensionality reduction produced by UMAP.

Histochemistry, RNAScope in situ hybridization and TUNEL assay

Knee and digit samples were harvested, fixed overnight in 4% paraformaldehyde, dehydrated, embedded in paraffin wax and processed for continuous 5 µm sectioning. To minimize the risk of generating artefactual tissue separation, sections were carefully floated on DEPC-treated water on a ProbeOn Plus glass slide for 5-10 s. The water was then decanted onto paper towels, and any remaining water was gently removed with a paper tissue. The sections were then placed on a slide warmer set to a lower temperature (approximately 30°C) and were stained with Hematoxylin or Safranin-O/Fast Green for histological and morphometric evaluation.

RNAscope in situ hybridization was carried out using RNAscope2.5 HD Detection reagent-RED (Advanced Cell Diagnostics, ACD) to visualize the spatio-temporal expression of Gdf5 (ACD, 407211), Prg4 (ACD, 437661), Atp1a1 (ACD, 569611-C2), Atp1b1 (ACD, 531721), Atp1b3 (ACD, 531601), Aqp1 (ACD, 504741), Piezo1 (ACD, 500511), Piezo2 (ACD, 400191) and Yap1 (ACD, 316601) in knee and digit joints at different developmental stages. Briefly, tissue sections were pre-treated with a custom reagent and hybridized with each probe for 2 h at 40°C in a custom oven. Signal was amplified with a pre-amplifier and a multiple amplifier as per the manufacturer's protocols. Final signal was detected and visualized by reaction with Fast Red substrate after a ∼10 min incubation at room temperature. Companion sections were hybridized with positive (Advanced Cell Diagnostics, 313911) or negative control probes (Advanced Cell Diagnostics, 310043) to assure signaling specificity. Sections were counterstained with Hematoxylin, dried and sealed with Permount. Bright-field images were taken with a Nikon Eclipse Ci equipped with a Nikon camera operated with NIS-Elements software.

To assess apoptosis, a TUNEL assay was performed using the In Situ Cell Death Detection Kit (Roche, Switzerland) according to the manufacturer's instructions. Paraffin-embedded sections of embryos from ouabain- or vehicle-treated mice (1 mg/kg or 5 mg/kg) were permeabilized with Proteinase K (10 mg/ml) in 1×PBS at 37°C for 20 min. Sections were then incubated with the TUNEL reaction mixture, which included terminal deoxynucleotidyl transferase (TdT) and fluorescein-labeled dUTP, in a humidified chamber at 37°C for 1 h.

Following incubation, sections were washed three times with 1xPBS and counterstained with Hoechst to visualize nuclei. Fluorescence images were captured using a Nikon fluorescent microscope. TUNEL-positive cells, indicating DNA fragmentation, were quantified by counting the number of fluorescent cells in randomly selected fields of view, with a minimum of 5 fields per sample. A total of 6 animals per group were analyzed. Negative controls were processed in the same manner, excluding the TdT enzyme from the reaction mixture to ensure the specificity of the staining. For the positive control, DNase I recombinant (3000 U/ml in 50 mM Tris-HCl, pH 7.5, 10 mM MgCl₂ 1 mg/ml BSA) for 10 min at room temperature to induce DNA strand breaks, prior to labeling procedures.

Drug administration

Stock solutions of ouabain and the hedgehog antagonist LDE225 (ChemieTek) were prepared in DMSO at a 100 mg/ml final concentration. Stocks were subdivided into small aliquots and stored at −30°C until use. On days of treatment, enough aliquots were thawed and diluted with a mixture of DMSO/corn oil mixture (3:7) to working concentration. Ouabain was administered daily by subcutaneous injection at 1 or 5 mg/kg body weight from E12.5 to E15.5 for knee joint analysis or from E12.5 to E18.5 for digit joint analysis. LDE225 was administered at 100 mg/kg once at E13.5 by oral gavage for knee and digit joint analysis. Control mice received the same treatments of DMSO/corn oil mixture. Three pregnant mice per treatment group were administered ouabain, LDE225 or vehicle. From each dam, three or four embryos were randomly selected. Joint development and cavitation were then assessed in the knees (12 embryos in total) and digits (10 embryos in total). Mice needed for each experiment were placed in separate cages, and randomized into control and experimental groups to minimize bias. Before gavage, mice were anesthetized by inhalation of 1.5% isoflurane in 98.5% oxygen to minimize discomfort, as prescribed by IACUC. Oral gavage probes were 22 gauge and 25 mm long (Instech Laboratories). Following drug administration by injection or gavage, animals were regularly monitored for signs of overt toxicity, including lethargy and dystonia, and Institutional LAS personnel provided additional daily animal monitoring and care.

Statistical analyses

Chi-square analysis was initially conducted to evaluate joint cavitation across the treatment groups, comparing 12 knee joints (12 embryos) in Figs 7 and 9, and 10 phalangeal joints (10 embryos) in Fig. S10 per group (control vehicle, ouabain or LDE225). Fisher's Exact Test was then applied to assess the statistical significance of the observed differences. Statistical significance was determined by comparing the P-value from Fisher's Exact Test to a threshold of P<0.05.

The effects of Ouabain (5 mg/kg) on the structure and function of the growth plate were evaluated using five embryos lacking cavitation, compared to vehicle-treated controls. Statistical differences were established using an unpaired Student's t-test, with P<0.05 considered statistically significant. Additionally, 95% confidence intervals (CIs) were calculated for the mean differences between groups.