Cellular plasticity in musculoskeletal development, regeneration, and disease

Abstract

In this review, we highlight themes from a recent workshop focused on “Plasticity of Cell Fate in Musculoskeletal Tissues” held at the Orthopaedic Research Society’s 2019 annual meeting. Experts in the field provided examples of mesenchymal cell plasticity during normal musculoskeletal development, regeneration, and disease. A thorough understanding of the biology underpinning mesenchymal cell plasticity may offer a roadmap for promoting regeneration while attenuating pathologic differentiation.

Introduction

The Orthopaedics Research Society 2019 annual meeting featured a workshop on “Plasticity of Cell Fate in Musculoskeletal Tissues.” This workshop covered multiple model systems and discussed the circumstances under which differentiated musculoskeletal lineages exhibit plasticity, or the ability to undergo a fate change from one differentiated fate into another. In this review, we synthesize these topics and focus on mesenchymal plasticity (dedifferentiation and transdifferentiation) in the context of in-vitro reprogramming, as well as during in vivo biological processes related to the musculoskeletal system. Using examples from normal development (skeletal), regeneration (digit tip), and disease (heterotopic ossification), we highlight common themes throughout these processes as they relate to cell plasticity, including mechanisms triggering cell fate changes, relevant signaling pathways, and the origin of cells that undergo fate changes. We find that cellular or environmental stressors provide cues leading to abnormal cell differentiation; for example cellular stress results in impaired hypertrophic chondrocyte differentiation and delayed endochondral ossification while environmental stress (such as local or systemic inflammation) is an important trigger for heterotopic ossification. Another common theme is the intriguing notion that cell transitions from one fate to another may require an initial dedifferentiation stage. In the context of in-vitro reprogramming, the induction of pluripotency greatly enhances subsequent differentiation compared to direct attempts at transdifferentiation. De-differentiation also occurs in the context of in-vivo digit tip regeneration, which is mediated by lineage-restricted cells. In addition, there is emerging evidence that the transdifferentiation of hypertrophic chondrocytes to osteoblasts may also require a transition to a mesenchymal progenitor-like state.

For many of the studies cited throughout the review, lineage tracing in mouse was carried out to define the origin of cells and track their ultimate fate. The most commonly used genetic tool for these purposes is the Cre-loxP system ^1;2. For lineage tracing, a tissue or cell-specific Cre recombinase allele is combined with a Cre reporter (commonly engineered into the Rosa26 locus, which is ubiquitously expressed by all cells) such that upon expression of Cre, the reporter molecule is activated. Recombination by Cre is a permanent event; thus all daughter cells will express the reporter regardless of later Cre expression. Reporters for detection of labeled cells include β-galactosidase or fluorescent molecules ². By combining Cre reporter labeling with cell-specific reporters (such as Col2-GFP) or immunostaining for cell differentiation markers, the origin of a cell can be identified as well as its current state.

With the widespread use of genetic tools to study musculoskeletal development, regeneration, and disease, exciting new evidence for cell plasticity (and cell restriction) have emerged that challenge previously held assumptions. A thorough understanding of the molecular biology underpinning mesenchymal plasticity may offer a roadmap to promote regeneration while attenuating pathologic differentiation.

Plasticity in differentiation, somatic reprogramming, and transdifferentiation

Deepak A. Kaji and Alice H. Huang

Although cell plasticity is a normal part of in-vivo processes, cell plasticity (in terms of induced pluripotency, fate re-specification, or transdifferentiation) is also often a feature of in-vitro manipulations (Fig. 1). By modulating signaling pathways, transcription factors, or microenvironmental cues, committed cells can be induced to adopt alternative fates in the dish. The most famous and dramatic example is the reprogramming of terminally differentiated somatic cells toward a pluripotent fate (iPSCs) by forced expression of key stem cell transcription factors ³. During reprogramming, promoters of pluripotent genes are demethylated and reprogrammed cells re-establish the bivalent chromatin domains which poise them to rapidly activate new genetic programs ⁴. Complete reprogramming therefore requires large-scale chromatin remodeling ^3;5. In fact, iPSC reprogramming efficiency can be improved by using epigenetic chromatin modulators such as histone deacetylase and methyltransferase inhibitors ^6;7. Moreover, the relatively open chromatin state which is required to maintain stemness is a root cause for a stem cell’s susceptibility to signals which allow for differentiation ⁸. This suggests that chromatin modulation may be an important part of improving a tissue’s resistance to pathologic transdifferentiation and may explain why different somatic cell types exhibit varying levels of stability. While late passage human iPSCs (hiPSCs) retain some epigenetic programming from the cell of origin, mouse iPSCs (miPSCs) are nearly identical to their embryonic counterparts and both hiPSCs and miPSCs are fully capable of differentiating along all three germ layers ⁹.

Figure 1. Epigenetic ‘subway map’ of in vitro differentiation from pluripotency, reprogramming, and transdifferentiation.

The arrows and their directionality between differentiated cell types are based on the literature showing successful fate transitions in-vitro. For example, fibroblasts can be induced toward chondrogenesis but have not been differentiated toward erythrocytes. In most cases, cells do not undergo direct transdifferentiation, but pass through a de-differentiated state prior to re-differentiation.

Differentiation strategies for pluripotent cell types (including iPSCs and embryonic stem cells) generally recapitulate developmental signaling in-vitro to progressively induce a series of progenitor states that ultimately lead to the desired fate. The timing and order of signaling molecules is critical as cells must be in the right state to interpret these signals appropriately. For example, BMP-4 is a positive modulator of stemness in mouse embryonic stem cells (mESCs), but different levels of BMP-4 also specify many different mesenchymal lineages from the primitive streak. Although each successive fate specification brings cells closer to their terminal fate, cells often also lose the ability to take on other fates and become progressively restricted. However, this is not always the case. In differentiating mESCs, specification is often flexible and cells may remain receptive to respecification. For example, the musculoskeletal lineages are mesechymal in origin and are derived from the primitive streak during mouse development ¹⁰. An activin/nodal gradient along the anterior-posterior axis of the primitive streak helps to specify distinct mesenchymal lineages ¹¹. At the anterior-most portion of the primitive streak, a completely different germ layer, endoderm, is specified with the highest levels of activin/nodal signaling. Interestingly, even after the undifferentiated in-vitro mesenchyme has been specified to posterior primitive streak (destined to become blood), this fate can be switched by exogenously applied activin signaling and redirected towards endoderm, a completely distinct germ layer ¹².

Similar examples of mesenchymal plasticity can be observed later in musculoskeletal differentiation from mESCs. As undifferentiated mesenchymal cells of the primitive streak ingress, they are exposed to varying levels of BMP, and are further specified to one of a few mesodermal intermediates such as paraxial, intermediate, or lateral plate mesoderm ^13–16. Each of these progenitors can produce different musculoskeletal lineages. While paraxial mesoderm is specified by blocking BMPs, intermediate mesoderm and lateral plate mesoderm require increasing levels of BMP-4 for specification ^17–19. Surprisingly, when paraxial mesoderm is cultured ex-vivo with BMP-4, it will turn off its paraxial mesodermal markers and will instead express the lateral plate marker GATA-4, thus reversing its specification and adopting an alternative fate ²⁰. In this experiment, the undifferentiated mesenchyme had already been specified in-vivo as paraxial mesoderm before the introduction of BMP. However, a genuine lineage restriction did not occur since lateral plate mesoderm (an alternate mesodermal fate) was still programmable. Since mesenchymal plasticity can be observed in embryological derivatives cultured ex-vivo, this phenomenon is unlikely to be an artifact of in-vitro stem cell differentiations, but rather a genuine capacity of mesenchymal cells.

Interestingly, fate change in differentiation is often unidirectional. For example, we could not identify reported instances of lateralized mesoderm being respecified as paraxial mesoderm. The reasons for asymmetric plasticity remain unknown. Moreover, it remains unclear why certain fate specifications entail lineage restriction and others do not. However, several examples of plasticity described in this review required invoking a stem or progenitor program. Even in skeletally mature mammals, mesenchymally derived tissue stem cells are more susceptible to pathologic transdifferentiation in vivo than their more differentiated counterparts ^21–25.

With the discovery of iPSCs, the interest in transdifferentiation methods has waned considerably. Historically, transdifferentiation in vitro required the forced expression of transcription factors to reprogram differentiated cells (commonly fibroblasts) directly towards other cell fates ²⁶. Much of transdifferentiation research involved identifying the minimum number of transcription factors to efficiently reprogram functional, differentiated cells. The relative ease in reprogramming fibroblasts may shed light on why fibroblastic lineages such as tenocytes or muscle connective tissue fibroblasts are frequently sources of pathologic transdifferentiation in vivo after injury ^21;22;25;27.

Skeletal lineage plasticity in development and disease

Zhijia Tan and Kathryn Cheah

Plasticity of the chondrocyte lineage: transformation to become osteogenic cells

This phenomenon of mesenchymal cell plasticity is not limited to in-vitro manipulations, but is also an essential element of normal development. In this section, we summarize recent findings from skeletal development. The bones form in two distinct ways: intramembranous ossification and endochondral ossification, both of which begin with condensation of mesenchymal precursors. For intramembranous ossification, mesenchymal progenitors directly develop into osteoblasts to form flat bones, such as calvarial skull, facial bones, the jaw, the clavicle and parts of the pectoral girdle ²⁸. Endochondral bone development occurs by the initial establishment of a cartilaginous template that is then replaced with bone. For endochondral bones, growth is fueled by a specialized structure, the growth plate, which is composed of highly organized chondrocytes that undergo coordinated and sequential differentiation phases of proliferation, maturation, cell cycle exit and hypertrophy to form an avascular cartilaginous growth plate that mediates longitudinal bone growth (Fig.2) ^29–31.

Figure 2. Schematic diagram showing the plasticity of hypertrophic chondrocytes in development and in chondrodysplasia.

During normal endochondral bone formation, chondrocytes undergo a cascade of proliferation, maturation and hypertrophy. HCs then are transformed by an unknown mechanism to become osteoblasts and osteocytes. In a MCDS mouse model, HCs revert to a pHC-like state, thereby alleviating ER stress. The ISR, through its preferential translation of ATF4 directly causes ectopic activation of Sox9, which mediates reprogramming of HCs. The cells survive but endochondral ossification is delayed. PC: proliferating chondrocyte; pHC: prehypertrophic chondrocyte; HC: hypertrophic chondrocyte. Genes specifically expressed in each cell type are listed.

The origin of osteoblasts in bone formation has been addressed by lineage tracing approaches employing genetically modified mice expressing Cre recombinase under the control of different cell-type specific regulatory elements. Such studies revealed that during bone formation, blood vessels in the bone collar invade into the cartilage, bringing in osteoblast progenitors from the perichondrium ³², which lay down bone matrix to form the primary ossification centre (POC). Concurrently, the cartilage matrix is degraded and the growth plate comprising layers of differentiating chondrocytes and spongy/trabecular bone (the primary spongiosa) forms. The mature osteoblasts in the perichondrium form cortical bone on the outer circumference. Thereafter, linear bone growth continues by endochondral ossification mediated by the growth plate (Fig. 2).

However, the additional direct contribution of growth plate chondrocytes to bone and the fate of hypertrophic chondrocytes (HCs) have been long debated – does the hypertrophic stage represent terminal differentiation followed by cell death or is it simply a transitional stage toward the osteogenic fate? The established dogma held that the transition to bone is driven by replacement of chondrocytes by osteoblasts. However, evidence in the last few years has challenged this assumption. Robust lineage tracing studies have demonstrated that HCs can survive and undergo a cell fate switch toward the osteogenic lineage ^33–36. Further, these cells persist into adulthood and contribute to both osteoblasts and osteocytes ³⁴.

The Col10a1 gene is specifically expressed in hypertrophic chondrocytes. This specificity was used to determine precisely whether HCs contribute to the osteoblast lineage in vivo. We made use of the Col10a1-Cre to tag the HCs specifically and chase their fate. We showed that during normal endochondral bone formation, HCs can survive and the descendants of HCs can become osteoblasts and osteocytes in prenatal and postnatal stages ³⁴. This novel concept has since been supported by other studies using different lineage tracing mouse reagents (Col2a1-CreERT, Osx-CreERT, Sox9-CreERT, Acan-CreERT, Col10a1-Cre, Col10a1-CreERT and Pthrp-CreERT), establishing the existence of a chondrocyte lineage continuum, where HCs survive and differentiate to form functional osteogenic cells in vivo under physiological conditions ^{31;33–35;37–39}. The contribution of HC derived osteoblasts to the trabecular and cortical bones peaks at embryonic and neonatal phases but declines in postnatal stage. The HC derived osteoblasts also play essential roles during bone fracture healing, allowing more efficient osteogenesis and vascularization ^34;35. In addition to osteoblasts, HCs also display the capacity to differentiate into Perilipin⁺ adipocytes, PDGFR-β⁺ pericytes, Endomucin⁺ endothelial cells and stromal cells ^33;38. Lineage tracing using a PTHrP-creERt line identified PTHrP⁺ resting chondrocytes as mouse skeletal stem cells. These PTHrP⁺ cells form columnar structure, undergo hypertrophy and became osteoblasts and stromal cells ³⁷, suggesting the stemness and plasticity of chondrocytes in the growth plate.

Many questions are still outstanding about the molecular controls underlying the transition from HC to osteoblasts. There are some indications that HCs re-enter the cell cycle ³⁹. Do they enter a progenitor/stem-like state? It is unclear how the mitotically active progenitors of small size (4-6 um in diameter) can be derived from the substantially larger HCs with diameters of 15-20 um. It has been speculated that autophagy would be the underlying mechanism as indicated by expression of Beclin1 and LC3B in the HCs ³⁹. Morphological and molecular analyses have suggested that the transdifferentiation of HCs to osteoblasts involves a transition to a mesenchymal-like state and cell-cycle re-entry ^39–41. Lineage tracing analyses in mice employing inducible Cre that are expressed in the chondrocytes and perichondrial precursors, showed descendent contribution to osteoblasts, stromal cells, adipocytes and bone marrow mesenchymal progenitors ^33;38, suggesting that chondrocytes and perichondrial precursors can contribute to multiple mesenchymal lineages.

During skeletal development, Wnt signaling plays an important role in controlling osteogenesis. Removal of β-catenin from early osteoblastic precursors results in an arrest of osteoblast differentiation ^42;43. Interestingly, activation of Wnt signaling by stabilization of β-catenin results in inhibition of preadipocyte differentiation into mature adipocytes via reducing PPARγ expression ⁴⁴, raising the possibility that Wnt signaling determines the cell fate of osteoblastic and adipogenetic progenitors, priming them towards osteoblasts. Recent studies have provided in vivo evidence that removal of β-catenin in the osteoblastic (Osx-Cre) or chondrogenic (Col10a1-Cre) lineages cause low bone mass and increased bone marrow adiposity ^45;46. Although the HC origin of adipocytes in the Col10a1-Cre; β-catenin conditional null mutants was not established ⁴⁵, it is possible that a cell fate shift from the HCs derived osteoblasts into adipocytes may partly contribute to the observed phenotypes.

Hypertrophic chondrocyte differentiation plasticity: an adaptive response to ER stress

The cascade of chondrocyte differentiation steps is tightly controlled by gene regulatory networks, involving stage specific expression of powerful transcription factors such as SOX9, RUNX2, FOXA2, GLI and MEF2C (reviewed by Liu et al and Tan et al) ^47;48. Mechanistic studies on the impact of activating ER stress in hypertrophic chondrocytes have revealed differentiation plasticity of HCs. The integrated stress response (ISR) is at the core of a cell’s response to various stresses such as endoplasmic reticulum (ER) stress. ER stress can be induced in cells when extracellular matrix (ECM) synthesis is high or when there are mutations in secreted proteins that impair their assembly and folding. Misfolded proteins then accumulate in the ER, activating the unfolded protein response (UPR), slowing protein translation, and activating transcription factors that upregulate the production of chaperones. Three sensors in the ER execute the UPR: protein kinase RNA-like endoplasmic reticulum kinase (PERK), inositol-regulated enzyme 1α (IRE1α) and activating transcription factor 6 (ATF6). The key control is phosphorylation of the α-subunit of eukaryotic translation initiation factor 2 (eIF2α). Cell death is triggered under severe cellular stress where the adaptive response is overwhelmed. We and others, studying the metaphyseal chondrodysplasia type Schmid (MCDS) have revealed that mutations in the COL10A1 gene that causes the accumulation of misfolded collagen X protein in the ER causes stress and the activation of UPR ^49;50. By studying the MCDS mouse model, we revealed the mechanism(s) by which the unfolded protein response and in particular, the ISR arm of the process, causes aberrant chondrocyte differentiation (Fig.2). The activation of the ISR triggers a cascade of regulatory changes that reverses the hypertrophic differentiation program causing chondrodysplasia. ISR mediated preferential translation of ATF4 directly transactivates ectopic expression of SOX9 in HCs. The inappropriate SOX9 expression reverts the HCs to a prehypertrophic state, disrupting the chondrocyte differentiation cascade resulting in chondrodysplasia. These changes reveal that hypertrophy is not an irreversible state in vivo and HCs are plastic ^50;51. The potential of a cell fate switch of HCs under pathological circumstances is also supported by studies showing inactivation of Sox9 in prehypertrophic chondrocytes resulted in a cell fate switch from chondrocyte to an osteoblastic phenotype ⁵².

The MCDS studies implicate cell fate change and chondrocyte plasticity that facilitates adaptation and survival under ER stress. However, the cost of the adaptive response is delayed endochondral ossification and chondrodysplasia. Recent studies have also shown that the impact of the UPR/ISR in MCDS models can be prevented or ameliorated by the action of an inhibitor of the integrated stress response or by promoting degradation of the unfolded proteins, and identify targeting ER stress response components as a viable therapeutic option. Treatment of the MCDS mice with an ISR inhibitor, acting at the level of p-eIF2a mediated over expression of ATF4, prevents the differentiation defects of HCs, abrogates the aberrant activation of ATF4 and SOX9 in HCs and ameliorates chondrodysplasia in MCDS mice (Fig.2) ⁵¹. Carbamazepine (CBZ), an FDA-approved drug used to treat epilepsy, stimulates both autophagy and proteasomal degradation pathways. CBZ treatment has been found to ameliorate partially the dwarfism phenotype of MCDS mice ⁵³. The mechanistic insights gained by studying chondrocyte plasticity under pathological conditions such as with MCDS and other congenital skeletal dysplasias associated with aberrant chondrocyte differentiation can guide the development of therapeutic interventions (reviewed in Boot-Handford et al) ⁵⁴ and have led to clinical trials in MCDS patients (https://mcds-therapy.eu/). These studies provide the inspiration for developing or repurposing compounds that can exploit chondrocyte plasticity to treat skeletal dypslasia in humans.

Cell plasticity versus lineage restriction during digit tip regeneration

Gemma L. Johnson and Jessica A. Lehoczky

Although regenerative and developmental processes are similar in that a fully formed organ is the end result, the cellular and molecular events (as well as the local environment) can be quite distinct. In this section, we focus on cell plasticity in the context of mammalian regeneration. Vertebrates have varying abilities to regenerate amputated composite tissues such as the limb. At one end of the spectrum, axolotls and juvenile xenopus can perfectly regenerate limbs following amputation ^55;56. Other vertebrates have more limited limb regenerative abilities: zebrafish regenerate distal fin rays ⁵⁷, and mammals, including mice and human children (but not adults), can regenerate amputated digit tips (Fig. 3) ^58;59. Despite the range of regeneration among these species, they all utilize blastema-mediated regeneration to replace the amputated limb tissues ^60–62. In brief, this process begins with epithelial cells forming a specialized epithelium to close the wound, at which point the blastema, a group of proliferating progenitor cells, emerges ⁶³. The blastema is essential for regeneration, as these cells go on to differentiate into the heterogeneous tissues of the regenerated limb, fin, or digit tip ⁶².

Figure 3. Overview of major tissue types during mouse digit tip regeneration.

Both images are schematized cross sections through mouse digit tips; only major tissue types are depicted: orange = epidermis (with dorsal hair follicles and ventral sweat glands); pink = nail plate; white = connective tissues; purple = tendon; blue = bone; red = blood vessels (capillaries not shown); green = nerves (distal axon branching not shown). (TOP) Unamputated digit where dotted arrow shows level of amputation permissive for regeneration. (BOTTOM) Blastema stage regenerating digit tip where progenitor cells in the blastema remain lineage restricted to their original tissue type.

The heterogeneity of tissue types within the limb and digit tip (bone, tendon, connective tissue, nerves, etc) (Fig. 3) underscores fundamental questions of blastema-based regenerative biology: where do the blastema cells come from and how do they give rise to the diverse set of tissues types necessary for limb regeneration? Hypotheses addressing these questions involve two opposing concepts: cell plasticity and lineage restriction. For example, blastema cells could be multipotent progenitor cells that can differentiate into any digit tip tissue, or they could be heterogeneous lineage restricted cells that are committed to a specific cell/tissue-type. Extensive established genetic tools, in combination with robust blastema-based digit tip regeneration, make mouse an ideal model organism to tease apart these concepts in mammals. To distinguish between blastema models of cell plasticity or lineage restriction, two putative multipotent populations have been assessed for their ability to give rise to all tissue lineages within the regenerated mouse digit tip ^64;65.

Rinkevich et al. hypothesized that multipotent blastema cells could exist within the systemic circulatory system ⁶⁴. The contribution of circulating stem cells was evaluated using parabiosis, whereby a wild-type (non-GFP) mouse and a GFP-expressing mouse were surgically connected to create a chimeric circulatory system ⁶⁶. Following digit tip regeneration in these chimeric mice, regenerated wild-type digit tips revealed that GFP-expressing cells (originating from cells derived from GFP-donor circulating stem cells) were found in the regenerated digit tip, but only within the hematopoietic lineage, not within the connective tissue or skin ⁶⁴. The conclusion from this experiment is that progenitor cells competent to differentiate into all tissues of the digit tip are not systemically circulating. Separately, we hypothesized that fibroblasts expressing the transcription factor Msx1 may serve as a multipotent progenitor pool for the digit tip ⁶⁵. The boundary of Msx1 expression in the developing digit delineates regenerative ability, and Msx1 genetically null mutant mice fail to regenerate their digits following amputation ⁶⁷. In neonatal mice, Msx1 is expressed in connective tissue directly adjacent to the nail matrix and the nail is essential for successful digit tip regeneration ^68;69. We evaluated the contribution of Msx1-expressing cells to the regenerating digit tip using genetic lineage analyses such that upon administration of tamoxifen, Msx1-expressing cells and any cells they differentiate into permanently express tdTomato. Following digit tip regeneration, we found that Msx1-expressing cells gave rise to many cells in the blastema which went on to form connective tissue and bone in the regenerated digit tip. However, these cells did not give rise to any non-mesenchymal cells in the regenerated tissue. While this experiment does not directly rule out the possibility that Msx1-expressing cells are multipotent progenitors within the mesenchymal lineage, it does demonstrate that they do not transdifferentiate between germ layers during regeneration. Taken together, these separate studies focused on circulating stem cells or Msx1-expressing cells converge on a model of lineage restriction during digit tip regeneration ^64;65.

Towards delineating a mouse digit tip blastema model of lineage restriction, the trajectories of epidermal, mesenchymal, tendon, bone, and endothelial cell populations have each been assessed during digit tip regeneration using similar genetic lineage tracing strategies as described above ^64;65. Epidermal cells marked by Keratin14 contributed exclusively to epidermal structures, not bone or connective tissue ^64;65. Conversely, mesenchymal cells broadly expressing Prx1 or En1⁶⁴ were found to only contribute to connective tissue, and cells expressing skeletal markers Sox9 or Sp7 were found to only contribute to bone ^64;65. Further supporting a model of blastemal lineage restriction, Scleraxis expressing cells were found to specifically give rise to tendons ⁶⁴, and cells expressing Tie2 or VE-Cadherin remained restricted to the endothelial lineage ⁶⁴. Collectively, these experiments provide strong evidence that blastema cells are not plastic and do not transdifferentiate among major tissue lineages during regeneration. However, genes used as lineage markers in these experiments are expressed in broad tissue types and do not have the resolution to reveal more subtle transdifferentiation relationships, for example, whether cells are starting as one type of mesenchymal fibroblast and ending regeneration as a different type of mesenchymal fibroblast.

While these studies support a heterogeneous and lineage restricted blastema, it remains unclear where these cells come from: the de-differentiation of mature digit tissues, tissue-specific resident progenitor populations, or a mix of the two depending on lineage. These subtleties have been explored in other regenerative species. For example, zebrafish osteoblasts de-differentiate during tail fin regeneration by up-regulating pre-osteoblastic markers and becoming proliferative ⁷⁰. In axolotl, muscle is regenerated by satellite cells (resident muscle progenitor cells) ⁷¹, whereas in newts, myofibers de-differentiate to form blastemal cells and subsequent regenerated muscle ⁷². Despite these examples, this issue has been difficult to address in mouse using standard genetic lineage tracing and will likely require a more comprehensive characterization and comparison of both the blastema and unamputated digit tip cell populations. Newer techniques, such as single-cell RNA sequencing (scRNAseq), can offer deeper insight into cell plasticity and lineage relationships during digit tip regeneration ^73–77. Recently, Carr et al. used scRNAseq to define distinct cell populations in injured and uninjured nerves in the mouse digit tip ⁷⁸. They found a mesenchymal cell population specific to injured nerves, showed that nerve-derived mesenchymal cells have hallmarks of mesenchymal progenitor cells such as the ability to differentiate down the osteogenic, chondrogenic, and adipogenic lineages, and found via lineage tracing that nerve-derived mesenchymal cells can contribute to regenerated bone ⁷⁸. This data supports a more subtle cell plasticity between mesenchymal populations, and provides evidence for a nerve associated mesenchymal precursor cell population that can contribute to the blastema and non-nerve mesenchymal tissues during digit tip regeneration.

Further studies using scRNAseq or even newer technologies (for example that maintain spatial information in the single cell sequencing process ⁷⁹) have the potential to reveal much more about cell plasticity in regeneration. Moreover, computational methods exist to determine the lineage relationship of individual cells within a scRNAseq dataset, as well as to discover rare cell populations ^73;80;81. These tools, together with the ability to complement computational studies with informed genetic lineage tracing and knockout studies in mice, will provide valuable insight into the question of cell plasticity and lineage restriction in mouse digit tip regeneration.

Cell plasticity in musculoskeletal trauma and heterotopic ossification

Wesley Huang, Kaetlin Vasquez, and Benjamin Levi

Finally, while cell plasticity is a feature of normal in-vivo processes, it is also a feature of undesirable disease states and aberrant differentiation. Heterotopic ossification (HO) is the abnormal development of bone in soft tissue, which can occur after severe burn, debilitating musculoskeletal trauma, spinal cord injury, and in patients with hyperactive BMP signaling ^82–85. This pathological process is widely thought to be due to an inflammatory response to injury, which causes an influx of inflammatory cells to the site of injury where they interact with resident progenitor cells. Pro-osteogenic genes are then up-regulated leading to ectopic bone growth ^86;87. While the inflammatory environment is a key trigger of HO, the requirement for specific immune cells and their interactions with resident progenitors are still largely unknown since loss of function/ablation studies have been carried out for only select populations. A recent review highlights key immune cells that have been implicated in HO, including macrophages and mast cells ⁸⁸. Using genetic tools, the resident progenitors are now beginning to be elucidated. To focus on cell fate plasticity and transitions, we will therefore summarize some of the work that has been done in murine models to characterize and identify the main progenitor cell types involved in HO.

The earliest models of HO included the implantation and overexpression of osteogenic factors through the use of intramuscular injections of solutions of bone, bone marrow, and periosteum ^89;90. Urist et al later discovered the osteogenic capabilities of BMPs and subsequent studies have demonstrated purified BMPs alone are sufficient to induce HO ^91–93. In 2012, Chakkalakal et al created a genetic form of HO based on a mutated constitutively active BMP type receptor (ALK2/ACVR1) that closely modeled the human disease Fibrodysplasia Ossicifans Progressiva (FOP) ⁹⁴. Since then, many different possible progenitor cells of HO have been examined in the context of BMP-induced HO using Cre-loxP lineage tracing (Table 1) ¹.

Table 1.

Summary of Cre drivers used to identify HO progenitor cells

Cre driver	Progenitor cell type	Contribution to HO
Tie2-Cre 95–97	Endothelial cells	Only a subpopulation of Tie2+ cells that are CD31-/CD45-/PDGFRα+/Sca1+ contribute to HO. These cells reside in native uninjured muscle and are not derived from endothelial cells.
VE-Cadherin-Cre 97	Endothelial cells, hematopoietic cells	Endothelial cells do not contribute to HO.
Prx1-Cre 82;99;100;106	Mesenchyme	Prx1+ cells contribute to all stages of HO formation. When the FOP mutation is targeted to this driver, mice develop spontaneous HO and display phenotypically similar great toe deformations seen in FOP patients.
Dermo1-Cre 99	Mesenchyme	Dermo1-Cre mice displayed similar HO phenotype as Prx1-Cre mice.
Glast-CreERT 101;102	Pericytes, epithelial cells, interstitial cells in connective tissue	Glast+ cells contribute to all stages of HO formation. These cells extensively co-localize with mesenchymal cell markers S100A4 and STRO1.
Gli1-CreERT 101	Perivascular cells	Gli+ cells contribute to all stages of HO formation. These cells extensively co-localize with mesenchymal cell markers S100A4 and STRO1. Mesenchymal markers were reduced at later stages of HO suggesting loss of stemness as HO progresses.
Mx1-Cre 104	Skeletal muscle interstitium, bone marrow	Muscle-resident interstitial Mx1+ cells contribute to intramuscular, injury-dependent HO formation.
Scx-Cre 21;22; Scx-CreERT 22;104	Tendon cells	Scx+ cells contribute to all stages of HO formation. When the FOP mutation is targeted to this driver, spontaneous HO forms in tendons and does not form in muscle after intramuscular injury.
Nfatc1-Cre 107	Mesoderm	Nfatc1-Cre mice displayed similar HO phenotype as Scx-Cre mice. Mice developed HO in the hindlimbs in the absence of trauma or BMP administration.

Traumatic HO

Endothelial cells

Based on lineage-tracing Tie2-Cre transgenic mice where Tie2+ cells were observed to undergo osteogenic differentiation with intramuscular BMP injection, it was hypothesized that progenitor cells were of endothelial origin that underwent transformation into multipotent mesenchymal stem cells (MSCs) with the ability to form bone ^95;96. Eventually, it was discovered that only a subpopulation of Tie2+ cells that were CD31-/CD45-/PDGFRα+/Sca1+ were responsible for bone formation suggesting they were of mainly mesenchymal origin instead ⁹⁷. Wosczyna et al also showed that CD31+ endothelial cells were unable to undergo HO and using a VE-Cadherin-Cre driver, a driver for endothelial and some hematopoietic cells ⁹⁸, demonstrated that endothelial cells contributed little to HO ⁹⁷.

Mesenchymal Stem Cells

The most current findings suggest local MSCs contribute to traumatic induced HO. In determining the role of the GNAS gene on HO, Regard et al were able to form HO in mice with mesenchymal tissue specific Prx1- and Dermo1-Cre drivers ^99;100. Using a burn/tenotomy model of HO in Prx1-Cre mice, Agarwal et al showed Prx1+ cells were present at sites of HO throughout all stages ⁸². Kan et al showed Glast+ cells contributed to HO using a Glast-CreERT driver ^101;102, a promoter that can be found in pericytes ¹⁰³. These cells also expressed the mesenchymal markers PDGFRα and S100A4 ¹⁰². In examining the role of other cells of mesenchymal origin, Kan et al used a perivascular specific Gli1-CreERT promoter to show Gli1+ cells contributed to all processes of endochondral ossification in HO ¹⁰¹. Interestingly, while Gli1-CreER labeled cells also expressed the mesenchymal markers STRO1 and S100A4, these markers were reduced at later stages of mature HO, suggesting there may be a loss of stemness in these cells as HO progresses ¹⁰¹. Dey et al used a Mx1-Cre model to show a muscle-resident interstitial Mx1+ population was implicated in intramuscular, injury-dependent HO formation ¹⁰⁴.

Tendon Cells

Because it has been observed that transection of the Achilles’ tendon in mice can lead to HO ¹⁰⁵, tendon cells have also been investigated as possible progenitor cells. Using the tendon specific marker, Scleraxis (Scx), Howell et al noted that Scx-lineage traced cells gave rise to cartilage following tenectomy while Agarwal et al showed that Scx-Cre cells contribute to all phases of HO in a burn/tenectomy model ^21;22. Furthermore, restricted overexpression of BMP signaling in Scx+ cells using Scx-CreERT driver resulted in intramuscular HO after trauma ^22;104.

Genetic HO – Fibrodysplasia Ossificans Progressiva

It is thought there is a common progenitor cell for both traumatic and genetic HO, and many of the findings in FOP mouse models corroborate the results in traumatic HO models. In testing the efficacy of the experimental RAR-ƴ agonist drug, Palovarotene, on FOP, Chakkalakal et al used a Prx1-Cre driver that limited overactive BMP expression to the limb bud mesenchyme ¹⁰⁶. These mice developed severe spontaneous HO by 1 month and had phenotypically similar great toe deformations commonly seen in FOP patients ¹⁰⁶. Dey et al targeted the FOP mutation to a Scx-Cre driver and noted spontaneous HO of tendons; however, cardiotoxin-mediated injury was insufficient to generate HO suggesting Scx+ cells are not able to migrate to the site of HO formation in muscle ¹⁰⁴. Interestingly, however, Mx1+ cells with the FOP mutation were unable to form spontaneous HO and were only able to do so with intramuscular injury ¹⁰⁴. Agarwal et al provided further evidence of the mesenchymal progenitor cell with the creation of a mouse model of FOP using the mesodermal tissue specific Nfatc1-Cre driver which had a similar phenotype to Scx-Cre mice ¹⁰⁷. Mutant mice developed HO in the hindlimbs as early as P4 in the absence of trauma or BMP administration ¹⁰⁷.

The use of the Cre-loxP system has shown many subpopulations are involved in HO; however, because these Cre lines label more than one cell population, it is hard to delineate the exact progenitor cell responsible. Nonetheless, when taken all together, these studies suggest that the local stromal / mesenchymal cells within the connective tissue of skeletal muscle, fascia and/or subcutis serve as the predominant source of HO.

HO is a complex process driven by a multitude of cells, interactions, as well as mechanisms. Given the wide variety of murine HO models available, it is necessary to identify which cell progenitors contribute most to HO and which cell types serve niche functions for HO formation. Understanding the interactions between these cells and the inflammation necessary to induce HO is key to developing specific therapies for HO and will allow us to better understand tissue repair, regeneration, and abnormal cell fate.

Concluding Remarks

The examples of mesenchymal cell plasticity highlighted in the workshop and this review span in vitro differentiation biology, developmental biology, regeneration, and disease. Interestingly, several of these examples involve transitions through a stem/progenitor stage prior to redifferentiation or transdifferentiation. While complete reprogramming is now possible, the partial reprogramming that allows for de-differentiation is still poorly understood. Moreover, why some fates are more difficult to reprogram than others remains a mystery. Understanding the variables that control fate specification may help to develop strategies for preventing pathological transdifferentiation in the context of age or injury. Pluripotent stem cell differentiation in vitro offer controlled environments for understanding the molecular underpinnings that drive plasticity. Future studies of cell plasticity in both in vivo and in vitro contexts will be required to promote regenerative responses and avert pathologic changes that arise from age, injury, or disease.