The diverse origin of bone-forming osteoblasts

Abstract

Osteoblasts are the only cells that can give rise to bones in vertebrates. Thus, one of the most important functions of these metabolically active cells is mineralized matrix production. As osteoblasts have a limited lifespan, they must be constantly replenished by pre-osteoblasts, their immediate precursors. As disruption of the regulation of bone-forming osteoblasts results in a variety of bone diseases, a better understanding of the origin of these cells by defining the mechanisms of bone development, remodeling, regeneration is central to the development of novel therapeutic approaches. In recent years, substantial new insights into the origin of osteoblasts – largely owing to rapid technological advances in murine lineage-tracing approaches and other single-cell technologies – have been obtained. Collectively, these findings indicate that osteoblasts involved in bone formation under various physiological, pathological, and therapeutic conditions can be obtained from numerous sources. The origins of osteoblasts include – but are not limited to – chondrocytes in the growth plate, stromal cells in the bone marrow, quiescent bone-lining cells on the bone surface, and specialized fibroblasts in the craniofacial structures, such as sutures and periodontal ligaments. As osteoblasts can be generated from local cellular sources, bones can flexibly respond to regenerative and anabolic cues. However, whether osteoblasts derived from different cellular sources have distinct functions remains to be investigated. Currently, we are at the initial stage to aptly unravel the incredible diversity of the origins of bone-forming osteoblasts.

1. Introduction

Osteoblasts –– “bone forming cells” in Greek –– are the only cells that can give rise to bones in vertebrates. These cells are responsible for the formation of over 200 pieces of bone that account for approximately 15% of the overall body mass in human; no other cell in any other organ is responsible for contributing to as much weight as osteoblasts. Osteoblasts produce a substantial volume of macromolecules in the bone matrix, including collagenous and non-collagenous proteins that provide a scaffold for matrix mineralization through the deposition of calcium phosphate in the form of hydroxyapatite^(1–3). The specialized structure of osteoblast enables the execution of their core function, including their large cuboidal-shape polarized cytoplasm, nuclei rich in euchromatin, large Golgi apparatus, and an enrichment of rough endoplasmic reticulum and mitochondria⁽⁴⁾. Osteoblasts are connected to adjacent cells via gap junctions to generate an osteogenic front on the bone surface^(5–11). Therefore, osteoblasts specialize in matrix production and mineralization for the formation of strong bones (Figure 1).

Figure 1. Active cuboidal osteoblasts on the bone surface.

(Left) Fluorescent pseudo-confocal microscopy images of the endosteal surface of the femur, osteocalcin (Bglap)-GFP mice at postnatal day 21 (3 weeks of age). A layer of large cuboidal osteocalcin-GFP⁺ osteoblasts (green) overlay the mineralized bone matrix (visualized by differential interference contrast, DIC, gray). Note that osteocalcin-GFP is also expressed in osteocytes embedded in the bone matrix. In the lower panel, osteoblasts are often covered by spindle-shaped pre-osteoblasts (yellow dotted line). Green: GFP; gray: DIC. Scale bar: 20 μm. The images were captured by Dr. Yuki Matsushita, University of Michigan.

(Right) Diagram showing the cellular component of the endosteal space. The illustration was created by Dr. Naoko Sakagami, University of Michigan.

Osteoblasts have a limited life span. In humans, these cells last for only 3 months during a normal remodeling cycle⁽¹²⁾. After bone formation, osteoblasts face the following three fates: 1) become embedded in the bone matrix, undergo structural changes, and become osteocytes⁽¹³⁾; 2) undergo programmed cell death (also known as apoptosis)⁽¹⁴⁾ or 3) become inactive on the bone surface as flat and elongated bone lining cells⁽¹⁵⁾. Therefore, osteoblasts must be constantly replenished by their precursors to maintain building bones. Mitotic activities are exclusively found in their immediate precursors, pre-osteoblasts, that overlay mature osteoblasts^(16–18). One of the most important purposes of all the stem and progenitor cells of the osteoblast lineage, generally termed as “skeletal stem and progenitor cells”^(19–22), is to continue forming new osteoblasts by constantly providing pre-osteoblasts. Skeletal stem cells are generally defined as self-renewing cells with the “trilineage” potential to differentiate into chondrocytes, osteoblasts, and marrow stromal cells or adipocytes in vitro⁽²³⁾.

One of the most intriguing features of osteoblasts is that they can be derived from many different cellular sources in many different bone compartments. Multiple cellular sources of osteoblasts might exist through a specific osteogenic program requiring Runx2, which enables cells to differentiate in response to various cues, such as development, homeostasis, and regeneration at various locations. To accommodate for this unique feature, many cell types can acquire an osteogenic fate⁽²⁴⁾, although non-skeletal cells do not form bone unless treated with bone morphogenetic protein (BMP). Importantly, dysregulation of osteoblast precursor populations underlies many bone diseases. Changes in osteoblast number can contribute to osteoporosis, underscoring the concept that the birth and death of osteoblasts are essential for maintaining proper bone mass throughout life⁽²⁵⁾.

In recent years, numerous papers have been published regarding the origin of osteoblasts, largely owing to recent technological advances in a mouse model of in vivo lineage-tracing and other single-cell technologies. This review highlights the diverse origins of osteoblasts, focusing on development, remodeling, and regeneration, under physiological, pathological, and therapeutic conditions.

2. Intramembranous and endochondral bone formation: Two pathways for osteoblast formation

Bone formation plays an important role in the survival of vertebrates by enabling and assisting essential functions, such as locomotion, respiration, and mastication. The first osteoblasts appear relatively late in development during the fetal stage, i.e., only after mesenchymal cells form condensations and establish bone anlage^(26,27). Subsequently, bone formation continues to occur asynchronously throughout. Therefore, osteoblasts with multiple developmental origins, such as in the paraxial and lateral plate mesoderm, as well as in the neural crest are formed via the action of multiple pathways.

Two distinct pathways of bone formation during development, intramembranous and endochondral bone formation, provide insights into the origin of osteoblasts. Intramembranous bone formation is a relatively simple and straightforward mechanism in which undifferentiated mesenchymal cells directly differentiate into osteoblasts. This mode of ossification evolved early, starting with dermal bones in early fish. This intramembranous pathway is still active in flat bones in some parts of the skull and clavicle in humans. In contrast, endochondral bone formation is a relatively new and has evolved as a pathway to generate bone cells on a large scale in an organized manner. This endochondral pathway is extensively adopted in higher vertebrates, including birds and mammals, and may serve as a convenient approach to exponentially grow their bones before and after birth. Except for a fraction of cartilages persisting into adulthood on the articular surface and elsewhere (permanent cartilage), most fetal cartilages are eventually replaced by bones and are termed transient cartilage.

The endochondral pathway involves two steps for the systematic building of bones. The first step establishes cartilage templates in a vasculature-free environment while the second step gradually replaces the cartilage templates with bone. This pathway uses a unique approach to build a framework for future bones as mesenchymal condensations, which serve as a systematic way to mass-produce undifferentiated mesenchymal cells. Condensing mesenchymal cells soon differentiate into chondrocytes upon the action of SOX9^(28,29), which continue to proliferate and grow cartilage templates. At this stage, the perichondrium is formed adjacent to the cartilage template as a highly vascularized fibrous tissue. The perichondrium provides the first osteoblasts in endochondral bones in an area immediately adjacent to the prehypertrophic zone, and later develops into the bone collar and periosteum⁽³⁰⁾. The unique partnership between the cartilage template and the perichondrium provides the anatomical foundation of endochondral bones by forming the marrow space inside and the cortical bone outside, ultimately establishing two distinct compartments of trabecular and cortical bones.

3. Fetal cartilage as a robust source of bone-making osteoblasts

Most cartilage templates in the fetal stage are eventually replaced by bones. However, how do chondrocytes within the template play a role in this process? The fate of chondrocytes within the fetal cartilage has been a matter of debate for decades; however, studies in the last few years have provided definitive answers by identifying fetal chondrocytes as an important origin of osteoblasts.

Undifferentiated mesenchymal cells within condensations later become chondrocytes within the cartilage template and perichondrial cells surrounding the cartilage. This was demonstrated by the Akiyama group who mapped the fate of Sox9⁺ cells in early limb development, initially by a constitutively active Sox9-cre and later by an inducible Sox9-creER^(31,32). Interestingly, Sox9⁺ cells at a later fetal stage lose their potential to become perichondrial cells, as Sox9⁺ cells at E10.5, but not at E12.5, can differentiate into perichondrial cells at E16.5, as demonstrated by Sox9-creER-based lineage-tracing experiments⁽³²⁾. Therefore, Sox9⁺ cells in condensations serve as precursors for chondrocytes and perichondrial cells, while maintaining their potential to become chondrocytes in later stages.

There has been a century-long debate on whether chondrocytes in the fetal cartilage template uniformly die due to programed cell death, or whether some of them can survive and transform into osteoblasts during ossification⁽³³⁾. Early evidence that hypertrophic chondrocytes could form osteoblasts was obtained from studies by the Roach^(34,35), Bianco⁽³⁶⁾ and Holmbeck⁽³⁷⁾ groups. In 2014, Cheah and colleagues provided evidence that hypertrophic chondrocytes in the fetal cartilage can become osteoblasts in the primary ossification center⁽³⁸⁾. This paradigm shift discovery was made possible by two important mouse genetic tools: Type X collagen alpha 1 chain (Col10a1)-cre and Col10a1-creER, developed for in vivo lineage-tracing experiments. Transformation of fetal chondrocytes into osteoblasts was independently confirmed by other groups, using other “chondrocyte-specific” genetic tools available in the field, including Aggrecan (Acan)-creER^(39,40) and Type II collagen alpha 1 chain (Col2a1)-creER⁽⁴⁰⁾. These series of in vivo lineage-tracing studies established the now commonly accepted concept that many fetal chondrocytes can transform into osteoblasts of the nascent marrow space of the primary ossification center, without uniformly undergoing programmed cell death^(41,42). The Cheah group reported that approximately 15% of hypertrophic chondrocytes become osteoblasts in the fetal stage when Col10a1-creER is employed⁽³⁸⁾. It is possible that only a subset of fetal chondrocytes located at a special location has a better capability to transform into osteoblasts, although the detailed mechanisms underlying this process remain largely undefined.

4. Fetal perichondrium: An important source of periosteal and trabecular osteoblasts

The perichondrium surrounding the fetal cartilage template is an equally important source of osteoblasts, contributing both to the bone collar that later develops into the periosteum and the cortical bone, and to the marrow space of the primary ossification center that later develops into the trabecular bone⁽³⁰⁾. Studies over the last decade have provided insights into the important roles that perichondrial cells play as an origin of osteoblasts during the fetal stage.

The perichondrium is composed of multiple layers of elongated fibroblasts, a part of which later develops as the osteogenic perichondrium in an area adjacent to where chondrocytes undergo hypertrophy. This is where the first osteoblasts and their precursor cells are formed in endochondral bones; the perichondrium forms the periosteal collar, but it is formed via intramembranous bone formation. The formation of osteoblast precursors in the osteogenic perichondrium is executed at least partly due to the actions of Indian hedgehog (Ihh) released from prehypertrophic chondrocytes within the cartilage template, as shown by the Long group^(43–47). Osterix (Osx)⁺ cells located in the innermost layer of the perichondrium represent the most committed precursor cells of the osteoblast lineage. As blood vessels invade into the cartilage template attracted by factors released from the hypertrophic layer such as vascular endothelial growth factor (VEGF), these Osx⁺ perichondrial precursor cells concomitantly translocate into the nascent marrow space, and their descendants eventually populate the primary ossification center as osteoblasts⁽⁴⁸⁾. Importantly, these Osx⁺ perichondrial cells are transient and do not stay in the perichondrium in subsequent stages. As a result, their contribution to the trabecular and cortical compartments might not persist in the long term^(49,50). The identity of earlier progenitor cells that constantly replace Osx⁺ cells in the perichondrium remains unidentified; as-yet identified populations of stem and progenitor cells, which may partially overlap with those identified in the periosteum^(51,52), are likely to provide a perpetual origin of osteoblasts in the cortical bone and cells in the periosteum, orchestrating the construction of strong cortical bones.

5. Establishment of marrow stroma: Primitive and definitive stroma

Osteoblasts are not the only cell fate that chondrocytes and perichondrial cells take when the primary ossification center is formed, associated with the nascent marrow space. Their progeny contributes to undifferentiated cells, generally termed as bone marrow stromal cells (BMSCs), coinciding with the initiation of marrow hematopoiesis. In developing primary ossification centers, Osx⁺ fetal perichondrial cells contribute not only to osteoblasts, but also to BMSCs^(40,48–50). Many BMSCs in the perivascular region express Leptin receptor (LepR)⁽⁵³⁾, the cognate receptor for leptin, a hormone that is often found in fat cells, and C-X-C motif chemokine ligand 12 (CXCL12)⁽⁵⁴⁾, an essential factor for bone marrow hematopoiesis. Hence, Osx⁺ fetal perichondrial cell-derived cells contribute to the developing marrow environment^(40,49,50). Although CXCL12⁺LepR⁺ BMSCs are suggested to differentiate into osteoblasts and contribute to the remodeling of adult bone tissues, fetal Osx⁺ perichondrial cells may bypass CXCL12⁺LepR⁺ BMSCs and directly differentiate into mature osteoblasts in response to active bone formation in the developing bone⁽⁴⁰⁾.

Osx⁺ fetal perichondrial cells might contribute to the marrow environment only transiently, as these cells appear to have limited self-renewing capacity. As a result, BMSCs derived from Osx⁺ perichondrial cells eventually disappear from the marrow space^(40,50). BMSCs derived from Osx⁺ fetal perichondrial cells are thus classified as the “primitive stroma” which acts only during the formation of the nascent marrow space. As a corollary, there is another cell population responsible for the continuous supply of BMSCs in the adult bone marrow referred to as the “definitive stroma”. Cells in the cartilage template and the perichondrium are marked by Col2a1-creER when pulsed during the early fetal stage of mouse embryonic day (E) 11.5 and E12.5; these cells contribute to osteoblasts and perivascular BMSCs in the primary ossification center, in a manner similar to Osx⁺ perichondrial cells⁽⁴⁰⁾. Interestingly, the Col2a1-creER-derived stromal cell lineage persists in the marrow stroma for a long time. These observations raise two possibilities: 1) Col2a1⁺ cells function as the upstream of Osx⁺ cells in the fetal perichondrium, resulting in the permanent bone marrow stromal lineage and the definitive stroma; or 2) cells that give rise to the definitive stromal population, independent of those in the fetal perichondrium, are labeled by Col2a1-creER. Borderline chondrocytes in the postnatal growth plate may also contribute to the primitive stroma (in addition to Osx⁺ fetal perichondrium), as these cells contribute transiently to osteoblasts and BMSCs⁽⁵⁵⁾. Recent studies have revealed that osteoclasts, which couple with osteoblasts to execute bone remodeling, have distinct origins during the developmental and adult stages^(56,57). These findings suggest that, similar to osteoclasts, there are two distinct pathways of osteoblast precursor differentiation: one corresponding to the developmental stage, when the precursors are provided constantly, and the other at the remodeling stage of adult bones, when these precursors are required at a low level. Therefore, distinct waves of marrow stromal progenitor cells might exist during bone marrow development.

6. How efficient is chondrocyte-to-osteoblast transformation in the postnatal stage?

By the end of the fetal stage, the central portion of the cartilage template is mainly replaced by bones. As a result, two separate units of cartilage are left at the edges of long bones, as a structure called the growth plate^(26,58). The growth plate undergoes a radical transformation during the postnatal stage and changes the way in which they produce their progeny.

The Chagin group elegantly demonstrated that, during the fetal stage, growth plate chondrocytes undergo exhaustive cell divisions and gradually deplete themselves through a consumption program⁽⁵⁹⁾; these cells can either transform into osteoblasts or disappear by undergoing apoptosis, contributing to rapid truncation of the cartilage template. In contrast, shortly after birth, these cells initiate self-renewal divisions upon the formation of a stem cell niche within the epiphysis, and undergo renewal⁽⁵⁹⁾. As a result, the postnatal growth plate was maintained as a thin disk of cartilage composed of characteristic clones of columnar chondrocytes. The resting zone at the top of the postnatal growth plate serves as a reservoir of stem cells, contributing to other chondrocytes in the growth plate⁽⁶⁰⁾. From the perspective of the origin of osteoblasts, how efficient chondrocyte-to-osteoblast transformation occurs in the postnatal stage, and what the contribution of postnatal chondrocytes is to osteoblasts are important open questions.

In 2014, the possibility that postnatal growth plate chondrocytes can translocate to the metaphysis (the area immediately beneath the growth plate) was raised using a lineage-tracing approach with Col10a1-creER mice⁽³⁸⁾ as well as Col2a1-creER mice with a multicolor Confetti reporter⁽⁶¹⁾. Further evidence for the transformation of postnatal growth plate chondrocytes was provided in 2018, based on a lineage-tracing study using a PTHrP-creER line, which can specifically mark PTHrP⁺ chondrocytes in the resting zone of the postnatal growth plate⁽⁶⁰⁾. These PTHrP⁺ resting chondrocytes clonally generate columnar chondrocytes in the proliferating and hypertrophic zones, and some of their descending hypertrophic chondrocytes further become osteoblasts and marrow stromal cells in the central metaphyseal marrow space (Figure 2). Current evidence also supports the concept that there are multiple routes of transition from the growth plate to the marrow space. One particularly important routes is borderline chondrocytes at the periphery of the growth plate⁽³⁶⁾; this was demonstrated using the same PTHrP-creER line⁽⁵⁵⁾. PTHrP⁺ borderline chondrocytes at birth behave as transient skeletal precursor cells, as these cells can translocate into the marrow space and become osteoblasts and marrow stromal cells, while their descendants are significantly reduced in adulthood. Therefore, chondrocytes of the postnatal growth plate can also transform into osteoblasts and marrow stromal cells through multiple pathways, perhaps via multiple mechanisms.

Figure 2. Chondrocyte-to-osteoblast transition at the end of the growth plate.

(Upper left) Fluorescent confocal microscopy images of the proximal growth plate of the femur, parathyroid hormone-related protein (Pthrp)-creER; R26R^tdTomato mice at postnatal day 36, after a tamoxifen pulse at postnatal day 6 (one month of chase). Lineage-marked PTHrP⁺ chondrocytes (red) form the entire layer of columns of chondrocytes in a clonal fashion, originating from the resting zone. In right panel, a large hypertrophic chondrocyte encased in “cocoons” is observed to escape from its sheath at the bottom of the growth plate, and then dramatically change its shape and produces its progeny in the underlying metaphyseal marrow space. Red: tdTomato; gray: DIC. Scale bars: 50 μm (left) and 10 μm (right). The images were captured by the author N.O.

(Lower left) Fluorescent pseudo-confocal microscopic images of the proximal metaphyseal marrow space of the femur, Cxcl12^GFP/+; Pthrp-creER; R26R^tdTomato (left) and Col1a1(2.3kb)-GFP; Pthrp-creER; R26R^tdTomato (right) mice at postnatal day 67, after a tamoxifen pulse at postnatal day 6 (two months of chase). Green: Cxcl12-GFP (left) or Col1a1(2.3kb)-GFP (right), red: tdTomato, gray: DIC. Scale bars: 20 μm. The images were captured by Dr. Koji Mizuhashi, University of Michigan.

(Right) Diagram demonstrating chondrocyte-to-osteoblast and stromal cell transformation at the sub-hypertrophic zone of the growth plate. The illustration was created by Dr. Naoko Sakagami, University of Michigan.

However, an important finding from these in vivo lineage-tracing studies is that the contribution of growth plate chondrocytes to osteoblasts during the postnatal stage appears to be modest; for example, descendants of PTHrP⁺ chondrocytes in the resting zone contribute to only several hundred osteoblasts and marrow stromal cells in the marrow space⁽⁶⁰⁾, while those of PTHrP⁺ borderline chondrocytes essentially disappear in adulthood⁽⁵⁵⁾. These results are somewhat dichotomous from existing fate-mapping studies using “chondrocyte-specific” cre lines, such as Col2a1-cre⁽⁴⁰⁾ and Col10a1-cre^(38,39,62), which mark a large number of osteoblasts, osteocytes, and marrow stromal cells in the postnatal marrow space. The remaining questions are, how efficient chondrocyte-to-osteoblast transformation is in the postnatal stage, and how is this transformation is molecularly regulated.

7. Chondrocyte-related progenitor cells as a major source of osteoblasts in growing bones

Even after chondrocyte-to-osteoblast transition wanes in the postnatal stage, bones continue to grow for a long time, indicating that there is another mechanism sustaining the continued production of osteoblasts in actively growing bones. What is the identity of undifferentiated cells that provide a robust origin for osteoblasts in growing bones? Current findings point to the theory that a new group of skeletal progenitor cells, which are particularly dedicated to bone growth, is established within the growing bone marrow⁽⁶³⁾.

This starts when fetal chondrocytes and perichondrial cells translocate to the nascent marrow space. The concept that hypertrophic chondrocytes can transform into a progenitor cell-like state in the fetal cartilage was proposed in 2015, by the von der Mark group. using Col10a1-cre, which they termed “chondrocyte-derived osteoprogenitor cells (CDOPs)”⁽⁶²⁾. Consistent with this concept, most CFU-Fs in the perinatal bone marrow are marked by Col2a1-cre; therefore, the derivative of fetal chondrocytes and perichondrial cells robustly contributes to this fraction⁽⁴⁰⁾. Interestingly, Col2a1 mRNA and COL2A1 protein are found in bona fide osteoblasts without chondrocyte-like morphology, making it possible that not all Col2a1-expressing cells are derived from fetal chondrocytes⁽⁶⁴⁾. However, the identities of these “early” bone marrow skeletal progenitor cells within the marrow space remain largely undefined.

Interestingly, most of the markers for skeletal progenitor cell populations identified in postnatal growing bones are expressed both by growth plate chondrocytes and undifferentiated marrow stromal cells, particularly those located immediately below the growth plate. Early postnatal cells marked by representative “chondrocyte-specific” transgenes, such as Col2a1-creER, Sox9-creER and Acan-creER, contribute to multiple skeletal cell types and provide a long-term source of osteoblasts^(39,40). Similarly, Gli1 is expressed by chondrocytes, cells located immediately beneath the growth plate, and progenitor cells termed metaphyseal mesenchymal progenitors (MMPs)⁽⁶⁵⁾. Grem1 is also expressed by both articular chondrocytes and “osteo-chondroreticular (OCR) stem cells in the marrow space⁽⁶⁶⁾. These in vivo lineage-tracing studies cannot formally distinguish the contribution of true immature marrow skeletal progenitor cells from growth plate chondrocytes; however, based on the reason discussed above, there is a possibility that chondrocyte-to-osteoblast transition does not account for all the contribution to osteoblasts that these studies observed⁽⁶⁷⁾. The identities of growth-associated skeletal progenitor cells, which reside in the metaphyseal marrow space and serve as the predominant source of osteoblasts, particularly for the trabecular bone, should be carefully defined in future studies.

8. Adult marrow skeletal progenitor cells support the life-long process of bone remodeling

All bones eventually stop growing in adulthood and enter the life-long maintenance phase. Although the demand for new osteoblasts is not as high as that in actively growing bones, osteoblasts need to be continuously generated in a highly coordinated manner throughout life, to execute everlasting cycles of bone remodeling⁽⁶⁸⁾. In adulthood, cells of the skeletal cell lineage are compartmentalized as a result of bone development in the prior stage. For example, BMSCs are an important source of osteoblasts during bone remodeling.

The basic multicellular unit (BMU) is a fundamental temporary structure driving bone remodeling, which is composed of osteoclasts at the leading front followed by a convoy of osteoblasts and their precursors in the rear⁽¹²⁾. Approximately 1 million BMUs operate at any moment in healthy humans, both in the cortical and trabecular bones. In each BMU, progenitor cells are recruited from the bone marrow to the resorption site, which differentiate into pre-osteoblasts and subsequently to osteoblasts. In adult bones, cells derived from growth plates are no longer abundant in the milieu.

What type of cells do adult bones utilize as a primary cellular source during normal remodeling cycles? Current evidence indicates that there is a transition from growth-associated skeletal progenitor cells to adult skeletal progenitor cells in terms of the origin of osteoblasts in adulthood. These adult skeletal progenitor cells represent a subset of BMSCs that are located near blood vessels, which coincide with “skeletal stem cells” that have been mainly characterized in cultured conditions and though transplantation experiments^(19,69). BMSCs in adult bone marrow are highly heterogeneous cell populations; these cells include, for example, reticular cells abundantly expressing CXCL12, which are located in a space surrounding bone marrow sinusoidal vessels^(54,70). Evidence that a subset of BMSCs is the major source of osteoblasts in adult bone marrow was initially demonstrated by the Morrison⁽⁷¹⁾ and Frenette groups⁽⁵⁰⁾. A large number of BMSCs express LepR, which negatively regulates osteoblast differentiation of BMSCs⁽⁷²⁾. Studies using LepR-cre identified that, interestingly, LepR⁺ cells contribute little to osteoblasts at the beginning of adulthood (i.e. two months of age in mice); however, their contribution to osteoblasts progressively increases with age⁽⁷¹⁾. These LepR⁺ cells overlap extensively with CXCL12-abundant reticular “CAR” cells^(54,70) in the adult bone marrow. Using an inducible Ebf3-creER that marks all CXCL12⁺ cells, the Nagasawa group demonstrated that Ebf3⁺ cells contribute to osteoblasts in adult bone marrow⁽⁷³⁾. Therefore, these studies established the concept that CXCL12⁺LepR⁺ adult marrow stromal cells behave as adult skeletal progenitor cells, and provide as an important origin of osteoblasts during bone remodeling in adulthood.

9. Diverse origins of osteoblasts in adult bones

How many functionally distinct cell populations are there within CXCL12⁺LepR⁺ and other BMSCs? Single-cell sequencing approaches provide an important platform to address this issue and help advance this research field. CXCL12⁺ stromal cells labeled by Cxcl12-GFP are composed of two major groups of pre-adipocyte-like and pre-osteoblast-like cells, as identified by single-cell RNA-seq analysis⁽⁷⁴⁾; these cells may correspond to “Adipo-CAR” and “Osteo-CAR” cells identified by a spatial transcriptomics study⁽⁷⁵⁾. Intriguingly, the Cxcl12-creER bacterial artificial chromosome (BAC) transgenic line marks a highly dormant pre-adipocyte-like subset of CXCL12⁺ cells (Cxcl12-creER⁺ cells⁽⁷⁴⁾). During normal bone development and remodeling, Cxcl12-creER⁺ pre-adipocyte-like stromal cells contribute to osteoblasts in the trabecular bone, but not to osteoblasts in the cortical bone⁽⁷⁴⁾. This perhaps indicates that osteoblasts in the trabecular and endocortical compartments have separate origins, with distinct mechanisms involved. Moreover, Qin et al. identified that adiponectin (Adipoq)⁺ cells, termed “marrow adipocyte lineage precursors (MALPs)”, maintain the marrow vasculature and suppress osteoblastic differentiation of marrow stromal progenitor cells through a cell-non-autonomous mechanism⁽⁷⁶⁾. Therefore, a pre-adipocyte-like subset of CXCL12⁺LepR⁺ BMSCs is likely to have unique functionality, regulating neighboring cells non-cell-autonomously under physiological conditions, while serving as a robust source of osteoblasts under regenerative conditions. Importantly, other subsets of BMSCs also provide an important origin for osteoblasts in adulthood. Recent single-cell RNA-sequencing studies identified cellular heterogeneity within BMSCs, revealing not only CXCL12⁺LepR⁺ cells, but also other cell types^(77–79).

What is the role of these “other” BMSCs as a source of osteoblasts? The first putative origin is pericytes, which surround the small arteries of the marrow space, known as periarteriolar stromal cells. Pericytes expressing alpha smooth muscle actin (αSMA) are a source of osteoblasts, as demonstrated by Kalajzic et al. using an Acta2-creER line⁽⁸⁰⁾; however, these cells appear to be transient and gradually disappear from the marrow space in adulthood. Neural-glial antigen 2 (NG2), encoded by Chondroitin sulfate proteoglycan 4 (Cspg4), is expressed by bone marrow pericytes; however, whether these cells provide an origin of osteoblasts has not been rigorously tested with the Cspg4-creER allele⁽⁸¹⁾. In addition, pericytes expressing platelet-derived growth factor receptor beta (PDGFRβ) provide a wide variety of bone cells in a manner responsive to endothelial cell-derived PDGF, as demonstrated by Pdgfrb-cre⁽⁸²⁾. More recently, a periarteriolar subset of LepR⁺ cells expressing the osteogenic cytokine Clec11a/Osteolectin (Oln) was demonstrated to represent rapidly-diving and short-lived osteogenic progenitor cells, as shown by Oln-creER-based lineage-tracing experiments⁽⁸³⁾.

The second putative origin is the bone-lining cells. Bone-lining cells represent a poorly defined quiescent population on the bone surface, which is converted from active osteoblasts upon completion of bone formation. Using a Dmp1-creER line and a long-chase protocol, Kalajzic et al. demonstrated that quiescent bone-lining cells represent an important source of osteoblasts in adulthood⁽⁸⁴⁾. Spindle-shaped N-cadherin⁺ osteoblastic (SNO) cells on the bone surface were originally identified as regulators of hematopoietic stem cells⁽⁸⁵⁾, although subsequent studies found no hematopoiesis-supportive role of these SNO cells^(86–89). Using a Cdh2-creER line (Cdh2 encodes N-cadherin), Li et al. showed that N-cad⁺ stromal cells generate both osteoblasts and marrow adipocytes during homeostasis in adult bones⁽⁹⁰⁾; whether this population truly represents bone-lining cells as defined above remains to be studied. The other putative source is the stromal cells in the transcortical channel of the cortical bone. Dmp1⁺ perivascular cells can migrate out of the transcortical channel and expand and differentiate into osteoblasts in bone organ cultures⁽⁹¹⁾.

In summary, BMSCs in adult bone marrow are highly heterogeneous, and a number of functionally distinct subsets located at different locations can collectively participate in osteoblastogenesis during normal cycles of bone remodeling in adult bones. Many of the identified cell populations described above might overlap at least partially, making it even more complicated to interpret the overall significance of each cell population. It will be important to map the fate of BMSCs at single-cell resolution in an unbiased manner in future studies, to better understand the origin(s) of osteoblasts during the adult bone remodeling cycles.

10. The origin of osteoblasts in bone anabolism

Understanding the origin of osteoblasts in adulthood is important for defining the mechanisms of and facilitating the anabolic actions of clinically approved therapeutic agents such as parathyroid hormone (PTH) and anti-sclerostin (SOST) antibodies. Physiologically, PTH maintains calcium metabolism homeostasis. However, when PTH is administered intermittently as a therapeutic agent, it exerts a bone anabolic effect^(92,93). For example, a biologically active amino acid 1–34 fragment of human PTH [hPTH (1–34): teriparatide] is widely used in patients with osteoporosis. PTH binds to the G-protein-coupled type 1 PTH/PTH-related protein receptor (PTH1R)⁽⁹⁴⁾ and stimulates osteoblastogenesis, thereby increasing bone volume. However, the origin of osteoblasts under PTH-induced bone anabolism is not well understood, and the mechanism for the expansion of mature osteoblasts is largely controversial⁽⁹³⁾.

Genetic lineage tracing approaches have opened an avenue for dissecting the hierarchical relationship among BMSCs in PTH-treated conditions. Many mature osteoblasts under PTH-induced bone anabolism originate from postnatal marrow stromal cells labeled by Sox9-creER at mouse postnatal day (P) 42⁽⁹⁵⁾. PTH directly regulates Sox9⁺ osteoblast precursor cells, as conditional deletion of PTH1R in Sox9⁺ cells abolishes PTH-induced bone anabolic actions. Furthermore, PTH administration suppresses the apoptosis of Sox9⁺ osteoblast precursors and their descendants, but does not promote the proliferation of these cells. This is consistent with previous reports demonstrating that PTH administration increases the number of mature osteoblasts via its anti-apoptotic effects^(96,97). PTH administration also accelerates the differentiation of LepR-cre-labeled bone marrow stromal cells into mature osteoblasts^(98,99). Unlike Sox9⁺ cells, LepR-cre-labeled BMSCs proliferate in response to PTH treatment, consistent with previous findings that PTH administration accelerates the proliferation of osteoblast precursors^(100–102). The differential effects of PTH treatment on Sox9-creER-labeled versus LepR-cre-labeled stromal populations may be attributed to the highly heterogeneous nature of the LepR-cre-labeled population. Further studies are required to validate the proliferative effect of PTH on osteoblast precursors, which leads to the expansion of mature osteoblasts.

PTH-induced expansion of osteoblasts is also caused by the conversion of quiescent bone-lining cells to active osteoblasts^(103,104). These observations were confirmed by a lineage-tracing study using Dmp1-creER, which can label the descendants of mature osteoblasts after a long chase, including bone-lining cells and osteocytes⁽¹⁰⁵⁾. Kalajzic et al. identified that Dmp1-creER-labeled perivascular osteoprogenitors localized in the trans-cortical channel are mobilized by PTH treatment and increase mature osteoblasts⁽⁹¹⁾. These results indicate that the diverse origins of osteoblasts contribute to the PTH-induced bone anabolic actions.

Previous in vitro studies have demonstrated that PTH suppresses adipogenesis and enhances osteoblastogenesis^(106–108). In addition, PTHrP heterozygosity increases marrow adiposity and reduces bone volume⁽¹⁰⁹⁾. Similar phenotypes were demonstrated using conditional deletion of PTH1R in the entire bone marrow stromal lineage with Prrx1-cre, suggesting that PTH regulates the osteogenic fate of the immature bone marrow stromal population⁽¹¹⁰⁾. However, the subpopulation of BMSCs modulated by PTH1R signaling remains unclear. The aforementioned Sox9-creER⁽⁹⁵⁾ or LepR-cre⁽⁹⁹⁾-labeled populations give rise to adipocytes in addition to osteoblasts, and PTH significantly decreases their contribution to adipocytes. Marrow adipogenesis is negatively regulated by Wnt/β-catenin signaling in Osx-cre-labeled BMSCs⁽¹¹¹⁾. PTH receptor signaling facilitates Wnt/β-catenin signaling^{(101,112–114)}, suggesting that PTH simultaneously suppresses adipogenesis via activation of canonical Wnt signaling. Interestingly, upon the cessation of PTH administration, the number of Sox9⁺ cell-derived adipocytes significantly increased in the marrow space⁽⁹⁵⁾, associated with a remarkable reduction of active β-catenin levels in these cells. Collectively, these results suggest that the PTH-induced cell fate decision of BMSCs is at least partially regulated through the canonical Wnt signaling pathway.

Canonical Wnt signaling is triggered upon the binding of Wnt ligands to the co-receptor, low-density lipoprotein-related proteins 5 and 6 (LRP5/6), and frizzled (FZD), leading to β-catenin stabilization⁽¹¹⁵⁾. Sclerostin (SOST) secreted by osteocytes⁽¹¹⁶⁾ acts as a Wnt inhibitor by directly binding to LRP5/6 on osteoblasts and negatively regulating bone formation^(117–120). Neutralizing antibodies for sclerostin (Romosozumab⁽¹²¹⁾ and Blozosumab⁽¹²²⁾) block the binding of sclerostin to LRP5/6 and promote osteogenesis by activating the canonical Wnt signaling pathway. In fact, rapid increases in the number of mature osteoblasts and subsequent bone formation are observed after treatment with sclerostin antibodies^(123,124). This is due to the increased availability of Wnt ligands in the milieu. Osteoblastogenesis is enhanced in transgenic mice overexpressing Wnt10b in mature osteoblasts⁽¹²⁵⁾ and Osx-cre-labeled cells⁽¹²⁶⁾. In addition, similar to the PTH-induced bone anabolic process, Dmp1-creER-labeled quiescent bone-lining cells are transformed into activated osteoblasts upon exposure to sclerostin antibodies ⁽¹²⁷⁾. However, despite continued dosing of sclerostin antibodies, the levels of bone-forming activities gradually decrease, and bone volumes eventually reach a plateau^(128–130). Although underlying regulatory mechanisms remain unclear, osteoblast precursors may become depleted due to the anti-proliferative effects of sclerostin antibodies^(124,131). However, recent genetic lineage-tracing studies focusing on osteoblast precursors revealed that sclerostin antibody increases the number of osteoblast precursors and their maturation⁽¹³²⁾. Further studies would provide more mechanistic insights into the mechanisms of anabolic actions associated with sclerostin antibodies.

11. The origin of osteoblasts in bone regeneration

Bones are particularly susceptible to various degrees of tissue damage due to their primary functions, ranging in severity from microfractures to macrofractures that completely disrupt tissue continuity. Bone regeneration capitalizes on two pathways of bone formation depending on the demand for osteoblasts to complete the repair. Most small and mechanically stable fractures heal by intramembranous bone formation, whereas large and unstable fractures also involve endochondral bone formation in which fibrocartilages and soft callus are newly generated near the fracture site to bridge bone fragments^(133,134). In both processes, a diversity of cells across many stages of the osteoblast lineage are recruited to the injury site, and collectively participate in osteoblastogenesis to achieve regeneration of the damaged skeletal structure^(135,136). The two major sources of osteoblasts for bone repair are periosteal cells overlaying the cortical bone surface, and BMSCs located in the marrow space and the endosteal space, which have distinct osteogenic and chondrogenic potentials depending on their microenvironment, as demonstrated by the Colnot group⁽¹³⁷⁾.

The periosteum is composed of an outer fibrous layer and an inner cambium layer, which contains osteoblasts, pre-osteoblasts and putative periosteal stem cell populations that provide a robust origin of osteoblasts under regenerative conditions. Further, the periosteum contains a heterogeneous population of progenitor cells, as demonstrated by Matthews et al⁽¹³⁸⁾. For example, cells expressing αSMA in the cambium layer provide a source of osteoblasts during fracture healing⁽¹³⁹⁾. In particular, the Mx1⁺αSMA⁺ subset in the perivascular region migrate to and repopulate the injury site for the long term, as demonstrated by the Park group⁽¹⁴⁰⁾. Periosteal stem cells marked by Cathepsin K (Ctsk)-cre normally form bones through the intramembranous pathway, but acquire the capacity to form bones through the endochondral pathway in response to injury, as shown by the Greenblatt group⁽⁵²⁾. In addition, periosteal stem cells marked by Prrx1-cre possess a higher regenerative potential than BMSCs that are regulated by periostin⁽⁵¹⁾. In the murine ribs, Sox9-expressing periosteal cells function as “messenger cells” that orchestrate large-scale bone regeneration by stimulating differentiation of neighboring cells in a cell-non-autonomous manner, as shown by the Mariani group⁽¹⁴¹⁾. These osteoblast precursor cells in the periosteum are originally derived from the perichondrium, and it is generally believed that the perichondrial groove of Ranvier provides the ultimate source of these cells^(142–144), although direct evidence is still missing.

The bone marrow stroma is also the origin of new osteoblasts during bone regeneration, as LepR^+(50,71), Mx1⁺⁽¹⁴⁵⁾, PDGFRβ⁺⁽⁸²⁾ and Clec11a/Osteolectin⁺⁽⁸³⁾ BMSCs robustly contribute to bone fracture healing. Interestingly, the quiescent pre-adipocyte-like subset of CXCL12⁺ BMSCs in the central marrow space does not normally become cortical bone osteoblasts; however, these cells are enlisted for cortical bone regeneration by transforming their identities into a skeletal stem cell-like state in response to injury. This suggests a role for cellular plasticity in this process⁽⁷⁴⁾. Other osteoblast precursor cells, such as Dlx5⁺ and Osx⁺ cells on the bone surface, are concomitantly recruited to the injury site and collectively participate in regeneration⁽¹³⁶⁾.

In complete fracture repair involving the formation of the fracture callus, chondrocyte-to-osteoblast transformation plays an important role in bone repair. Using a cartilage graft model, the Marcucio group demonstrated that graft chondrocytes transform into osteoblasts for repairing bones⁽¹⁴⁶⁾. This transformation was also confirmed by lineage-tracing experiments in non-graft-mediated fracture repair⁽³⁹⁾, with cells in the chondro-osseous border (the transition zone) of the fracture callus expressing the pluripotency factors⁽¹⁴⁷⁾. Therefore, multiple cellular sources are mobilized through multiple mechanisms to rapidly and effectively repair bone fractures in response to the potentially life-threatening emergency situation of complete fracture.

12. The origin of osteoblasts in the craniofacial bone

The craniofacial bone is formed as a result of the deliberate coordination of two distinct mechanisms of intramembranous and endochondral bone formation⁽¹⁴⁸⁾. The facial bones are of neural crest origin, while the cranial vault is of paraxial mesoderm origin. There are multiple routes of osteoblastogenesis involved in this process, including both chondrocyte-mediated and non-chondrocyte-mediated pathways in a way similar to major long bones. For example, the Msx2⁺ cells of the embryonic mandibular process contribute to diverse types of cells such as chondrocytes, perichondrial cells, osteoblasts, and suture fibroblasts in the postnatal stage⁽¹⁴⁹⁾. In addition, some chondrocytes in the craniofacial complex, such as those in the mandibular condylar cartilage, can transform into osteoblasts^(37,150). In later stages, BMSCs and periosteal cells are generally considered to provide a primary origin of osteoblasts in the craniofacial bone, although the details are not completely understood. The cellular plasticity of periosteal cells may also play a role in large-scale regeneration of the zebrafish jaw bone⁽¹⁵¹⁾. Importantly, the craniofacial bone also utilizes other unique sources of osteoblasts in development and adulthood, such as the suture and periodontal ligament.

The suture is a fibrous tissue between two adjacent bone compartments in the skull, which is composed of two osteogenic fronts, with undifferentiated cells in between. The suture provides a niche for “skeletal stem cells” of the craniofacial bone; these cells express various markers such as Gli1⁽¹⁵²⁾, Axin2⁽¹⁵³⁾ and Prrx1⁽¹⁵⁴⁾, and serve as a robust origin of osteoblasts in the skull during homeostasis and repair. The periodontal ligament is another fibrous tissue that is present between the tooth and the alveolar bone. It is derived from the dental follicle, a sac-like membranous tissue surrounding the developing tooth bud during tooth development. PTHrP⁺ cells in the dental follicle serve as an origin of osteoblasts in the specific area of the alveolar bone, such as the cryptal bone (the bone between the tooth roots) and the interseptal bone (the bone between the teeth), as well as periodontal ligament fibroblasts and some cementoblasts⁽¹⁵⁵⁾. Gli1⁺ cells function as stem cells of the adult periodontal ligament, which are regulated by osteocytes and occlusal forces, and support their tissue turnover⁽¹⁵⁶⁾. Therefore, these specialized fibrous tissues that are present between the mineralized compartments of the craniofacial bone serve as an important origin of osteoblasts as well as other mature cells supporting the functionality of the highly complex craniofacial bone.

13. The origin of osteoblasts in heterotopic ossification: when osteoblastogenesis goes awry

Osteoblast formation is highly regulated under normal conditions to ensure the formation of bones where needed. However, in the pathological condition of heterotopic ossification, bones are formed in extraskeletal sites through endochondral bone formation, within soft and connective tissues, such as in the skeletal muscle and the tendon. Heterotopic ossification is induced as a result of hereditary or acquired conditions, such as a rare genetic disorder of fibrodysplasia ossificans progressive (FOP) caused by activating mutations of a bone morphogenetic protein (BMP) type I receptor⁽¹⁵⁷⁾ or trauma and inflammatory insults⁽¹⁵⁸⁾. The origin of osteoblasts in heterotopic ossification is diverse, including Glast⁺ pericyte-like cells⁽¹⁵⁹⁾, scleraxis (Scx)⁺ cells in the tendon and the ligament⁽¹⁶⁰⁾, cells expressing Mx1⁽¹⁶¹⁾ and Tie2 residing in the interstitium of the skeletal muscle⁽¹⁶²⁾, and Wnt1⁺ cells in the endoneurium of peripheral nerves⁽¹⁶³⁾. These cells can be converted to osteoblasts through both chondrocyte-mediated and chondrocyte-independent pathways under pathological conditions, such as abnormal activation of BMP signaling. Therefore, diverse arrays of connective tissue cells can be aberrantly converted to an osteogenic fate under pathological conditions and used as the origin of osteoblasts in heterotopic ossification (of note, this topic has been extensively reviewed by the Levi group⁽¹⁵⁸⁾).

14. Conclusion

In this review, we discussed the diverse origins of osteoblasts that support bone formation under various conditions, i.e., bone development, remodeling, regeneration, anabolism, and heterotopic ossification. Current evidence based on murine in vivo lineage-tracing studies clearly demonstrates the unique feature of the osteoblast lineage, where osteoblasts can be generated from local cellular sources when necessary (Figure 3). The fact that osteoblasts can employ many different cellular sources to generate the similar material of mineralized tissues is intriguing, although subtle variations in the components of the extracellular matrix exist among bones from different sources and locations^(164,165). This property endows bones with the substantial flexibility to respond to regenerative and anabolic cues. Our future endeavors should aim to capitalize on this profound plasticity to enhance osteoblastogenesis for therapeutic purposes, or conversely, suppress it to prevent debilitating conditions such as heterotopic ossification. Whether osteoblasts derived from different cellular sources have distinct functionalities remains to be addressed in future studies and requires rigorous functional approaches at single-cell resolution. We are still at the initial stage of understanding the intricate processes that are involved in regulating the generation of bone-forming osteoblasts.

Figure 3. The diverse origin of bone-forming osteoblasts.

Diverse origins of bone-forming osteoblasts at the fetal and postnatal stages and regeneration. Osteoblasts are matrix-secreting cells with well-developed cytoplasm, organelles, and euchromatin. Osteoblasts have a limited lifespan; these cells soon become embedded in the matrix as osteocytes or undergo apoptosis. Osteoblasts are continuously provided by their immediate pre-osteoblast precursors. These pre-osteoblasts are formed from a diverse array of cells, including those termed as “skeletal stem cells,” under a variety of conditions such as bone development, remodeling and regeneration, bone anabolism, and heterotopic ossification. Osteoblasts originate from growth plate chondrocytes, bone marrow stromal cells, quiescent bone-lining cells, and other fibroblasts, such as suture fibroblasts or dental follicle cells. The illustrations were created by Dr. Naoko Sakagami, University of Michigan.

Table 1.

Mouse transgenic lines/alleles for in vivo lineage-tracing experiments, for the analysis of the origin of osteoblasts.

A. Development
Driver	Time point of induction	Representative cre-marked cells	Contribution in developing long bone	References
Osx-creERT2	One time at E13.5	Perichondrium	BMSCs and osteoblasts (Short lived)	48, 50
Col1(3.2 kb)-creERT2	One time at E13.5	Perichondrium	Perichondrial cells in fetal development	48
Col2a1-creERT2	One time at E13.5	Chondrocytes and perichondrium	BMSCs, chondrocytes, osteoblasts, and adipocytes	40
Col10a1-cre	N/A	Hypertrophic chondrocytes	Osteoblasts	38, 62
Acan-creERT2	One time at P3, 2 wks	Chondrocytes, perichondrium, and primary spongiosa	Chondrocytes and osteoblasts	40 (P3), 39 (2 wks)
Col10a1-creERT2	One time at E13.5	Hypertrophic chondrocytes	Osteoblasts	38
Sox9-creERT2	One time at P3	Chondrocytes, perichondrium, endosteum, periosteum, and primary spongiosa	BMSCs, chondrocytes, osteoblasts, and adipocytes	40
Grem1-creERT2	One time at P1	Chondrocytes and primary spongiosa	BMSCs, chondrocytes, and osteoblasts	66
Gli1-creERT2	One time at E13.5	Chondrocytes and perichondrium	BMSCs, chondrocytes, osteoblasts, and adipocytes	65
Pthrp-creERT2	One time at P0	Borderline chondrocytes	BMSCs and osteoblasts (Short lived)	55
Pthrp-creERT2	One time at P6	Chondrocytes in the resting zone	Growth plate chondrocytes, BMSCs, and osteoblasts	60

B. Remodeling
Driver	Time point of induction	Representative cre-marked cells	Contribution in postnatal long bone	References
LepR-cre	N/A	BMSCs	BMSCs, osteoblasts, and adipocytes	50, 71
Cxcl12-creERT2	One time at 3 wks old	BMSCs	BMSCs, trabecular osteoblasts, adipocytes	74
Ebf3-creERT2	Four times on alternate days at 10 wks old	BMSCs	BMSCs, osteoblasts, and adipocytes	73
Sox9-creERT2	One time at P3	Chondrocytes, perichondrium, endosteum, periosteum, and primary spongiosa	BMSCs, chondrocytes, osteoblasts, and adipocytes	95
Pdgfrb-cre	N/A	BMSCs	BMSCs, chondrocytes, periosteum, and osteoblasts	82
Oln-creERT2	Five consecutive days at 2 months old	Periarteriolar LepR+ cells, hypertrophic chondrocytes, periosteum, osteoblasts, and osteocytes	Osteoblasts	83
Dmp1-creERT2	Two consecutive days at 2 months old	Osteoblasts, osteocytes, and bone lining cells	Osteoblasts	84
Cdh2-creERT2	Three consecutive days	N-cad+ stromal cells	BMSCs, osteoblasts, and adipocytes	90

C. Regeneration
Driver	Time point of induction	Representative cre-marked cells	References
Grem1-creERT2	Four times on alternate days at 6–8 wks old	Endosteum	66
Gli1-creERT2	Three consecutive days at 1 month old	Periosteum	65
LepR-cre	N/A	BMSCs and periosteum	50, 71
Cxcl12-creERT2	One time at 6–8 wks old	BMSCs	74
Acta2-creERT2	Two consecutive days at 3–5 months old	Periosteum	139
Mx1-cre	Every other day for 10 days at 4 wks old	Periosteum	140
Ctsk-cre	N/A	Periosteum	52
Prrx1-cre	N/A	Periosteum	51
Mx1-cre	Every other day for 10 days	Mx1+ stromal cells	145
Osx-creERT2	One time at P5	Primary spongiosa, endosteum and periosteum	50
Osx-cre	N/A	BMSCs and periosteum	82
Pdgfrb-cre	N/A	BMSCs and periosteum	82
Oln-creERT2	Five consecutive days at 2 months old	Periarteriolar LepR+ cells, hypertrophic chondrocytes, periosteum, osteoblasts, and osteocytes	83
Acan-creERT2	Induced in fractured at 2.5 months old	Chondrocytes in fractured callus	39

D. Craniofacial
Driver	Time point of induction	Representative cre-marked cells	Contribution in craniofacial tissues	References
Msx2-creERT2	One time at E10.5	Mesenchymal cells in cranial vault and mandible	Chondrocytes, osteoblasts, perichondrial cells, BMSCs, and suture fibroblasts	149
Acan-creERT2	One time at P14	Mandibular condylar cartilage	Chondrocytes and osteoblasts	150
Pthrp-creERT2	One time at P3	Dental follicle cells	Cementoblasts, PDL fibroblasts, and alveolar bone osteoblasts	155
Prrx1-creERT2	One time at 8 wks old	Suture fibroblasts in the craniofacial bones	Calvarial osteoblasts	154
Axin2-rtTA	Three consecutive days starting at P25	Suture fibroblasts in the craniofacial bones	Calvarial osteoblasts	153
Gli1-creERT2	Four consecutive days at 1 month old	Suture fibroblasts in the craniofacial bones	Calvarial osteoblasts, periosteum, and dura	152
Gli1-creERT2	Two consecutive days at 5–8 wks old	PDL in the apical space	PDL fibroblasts, cementoblasts, and alveolar bone osteoblasts	156

E. Heterotopic
Driver	Time point of induction	Representative cre-marked cells	Contribution in heterotopic ossification	References
Glast-creERT2	Five consecutive days at older than 1 month	Muscle interstitial pericyte-like cells	Chondrocytes and osteoblasts	159
Scx-creERT2	Weaning stage	Fibroblasts in tendon and muscle	Chondrocytes and osteoblasts	160
Mx1-cre	Every other day from P7 to P21	Interstitium of skeletal muscle	Chondrocytes and osteoblasts	161
Tie2-cre	N/A	Interstitium of skeletal muscle	Chondrocytes and osteoblasts	162
Wnt1-creERT2	Three consecutive days at 6–8 wks old	Endoneurium of peripheral nerves	Chondrocytes and osteoblasts	163

E, embryonic day; BMSCs, bone marrow stromal cells; N/A, not applicable; P, postnatal day; wks, weeks; PDL, periodontal ligament.