Bone marrow and periosteal skeletal stem/ progenitor cells make distinct contributions to bone maintenance and repair

SUMMARY A fundamental question in bone biology concerns the contributions of skeletal stem/progenitor cells (SSCs) in the bone marrow versus the periosteum to bone repair. We found that SSCs in adult bone marrow can be identiﬁed based on Lepr cre and Adiponectin-cre/creER expression while SSCs in adult periosteum can be identiﬁed based on Gli1 creERT2 expression. Under steady-state conditions, new bone arose primarily from bone marrow SSCs. After bone injuries, both SSC populations began proliferating but made very different contributions to bone repair. Drill injuries were primarily repaired by LepR + /Adiponectin + bone marrow SSCs. Conversely, bicortical fractures were primarily repaired by Gli1 + periosteal SSCs, though LepR + / Adiponectin + bone marrow cells transiently formed trabecular bone at the fracture site. Gli1 + periosteal cells also regenerated LepR + bone marrow stromal cells that expressed hematopoietic niche factors at fracture sites. Different bone injuries are thus repaired by different SSCs, with periosteal cells regenerating bone and marrow stroma after non-stabilized fractures. + cells to fracture repair reﬂect the activities of both LepR + bone marrow SSCs and a subpopulation of LepR + periosteal SSCs. Our results are consistent with recent reports that Lepr cre recombines in a subset of SSCs in the periosteum (Ortinau et al., 2019; Mo et al., 2021). In the current study, we conﬁrmed that LepR + cells within the bone marrow give rise to transient trabecular bone within the bone marrow at the fracture site. LepR + bone marrow SSCs and Gli1 + periosteal SSCs thus both contribute to the stabilization of bicortical fractures in the ﬁrst few weeks after injury by forming trabecular bone and cartilage, respectively, at the fracture site.


In brief
In adult mice, different skeletal stem/ progenitor cells (SSCs) are responsible for the repair of distinct bone injuries. LepR + Adiponectin + bone marrow SSCs are responsible for steady-state osteogenesis and drill injury repair while Gli1 + periosteal SSCs are primarily responsible for bicortical fracture repair and bone marrow stromal cell regeneration at fracture sites.

INTRODUCTION
Skeletal stem/progenitor cells (SSCs) that are capable of contributing to bone repair are present in the periosteum on the outside surface of bones (Duchamp de Lageneste et al., 2018;Debnath et al., 2018;Ortinau et al., 2019;Matthews et al., 2021;Zhang et al., 2005), as well as inside the bone marrow (Chan et al., 2015;Zhou et al., 2014;Matsushita et al., 2020;Bianco and Robey, 2015;Friedenstein et al., 1970). However, the relative contributions of these SSC populations to bone repair have been debated, in part because few markers have been available to distinguish these cell populations. Much of the fate mapping that has been performed was done on early postnatal or juvenile mice when different cells are responsible for the formation and repair of bone as compared to adulthood (Ono et al., 2014a(Ono et al., , 2014bMizoguchi et al., 2014;Leucht et al., 2008). Moreover, few studies have performed quantitative, side-by-side comparisons of the contributions of different SSC populations to bone repair, making it difficult to compare results among studies.
In fetal and postnatal mice prior to one month of age, Gli1 creERT2 marks SSCs in the perichondrium that promote bone growth and repair as well as giving rise to LepR + stromal cells in the bone marrow (Shi et al., 2017). After one month of age, Gli1 creERT2 expression is restricted to the metaphysis and declines quickly in young adult mice (Shi et al., 2017). Gli1 creERT2 also labels SSCs in calvarial sutures that contribute to bone growth and repair . Gli1 is expressed by a subset of cells in the periosteum of adult bones, but it has not yet been tested whether these cells have SSC activity (Shi et al., 2017;Xu et al., 2022). These studies raise the question of whether Gli1 creERT2 labels cells that contribute to the repair of adult bones beyond the calvarium, which is unusual as it forms largely via intramembranous ossification.
Bone injuries can heal through intramembranous ossification, in which SSCs differentiate directly to bone, or endochondral ossification, in which SSCs form a cartilage intermediate and then bone (Serowoky et al., 2020). Drill-hole injuries are small, stabilized fractures (Matsushita et al., 2020;Leucht et al., 2008) that heal through intramembranous ossification (Colnot, 2009;Serowoky et al., 2020;Thompson et al., 2002). Non-stabilized or partially stabilized bicortical fractures heal primarily through endochondral ossification with the formation of a cartilage callus followed by osteogenesis (Colnot, 2009;Serowoky et al., 2020;Kuwahara et al., 2019). The observation that different bone injuries induce different modes of repair raises the question of whether they are healed by different SSCs.
We performed lineage tracing with 11 Cre alleles to identify markers that distinguish periosteal from bone marrow SSCs. Most Cre alleles labeled both types of SSCs. However, we found that Gli1 creERT2 preferentially labeled periosteal SSCs in the diaphysis of adult long bones while Adiponectin-cre and Adiponectin-creERT specifically labeled LepR + bone marrow SSCs. Lepr cre labeled bone marrow SSCs but also recombined in a minority of periosteal SSCs. These markers allowed us to compare the contributions of periosteal versus bone marrow SSCs to the repair of drill-hole injuries and bicortical fractures.
Adult Osterix-creERT2; tdTomato mice exhibited labeling in less than 4% of bone marrow stromal cells, 5%-10% of LepR + stromal cells in the bone marrow metaphysis, 5%-15% of CFU-F in the metaphysis and diaphysis, and 60%-100% of osteoblasts, though the percentage of labeled osteoblasts declined over time ( Figure S1C). These data are consistent with Osterix-creERT2 recombination in osteoblasts (Park et al., 2012) and suggest Osterix-creERT2 recombines in a minority of SSCs in adult bone marrow. Gli1 creERT2 ; tdTomato mice exhibited recombination in less than 1% of bone marrow stromal cells, LepR + cells, or CFU-F in the adult bone marrow diaphysis ( Figure S1D). A higher level of recombination was observed transiently in the bone marrow metaphysis, where in the first few weeks after tamoxifen treatment up to 10% of bone marrow stromal cells, 13% of LepR + cells, 26% of osteoblasts, and 12% of CFU-F were Tomato + (Figure S1D). However, the amount of labeling in the metaphysis declined over time. These results are consistent with the observation that Gli1 creERT2 labels a subset of SSCs in the metaphysis but not in the diaphysis, and that Gli1 creERT2 expressing cells in the metaphysis are rapidly depleted in young-adult mice (Shi et al., 2017).
In contrast to these markers that labeled few adult bone marrow SSCs, Lepr cre and Adiponectin-cre labeled most of the CFU-F in the metaphysis and diaphysis at 8 to 24 weeks of age ( Figures 1A and 1B). Adiponectin-creERT also labeled 55%-65% of CFU-F at 4 to 12 weeks after tamoxifen treatment in the metaphysis and diaphysis ( Figure 1C). Lepr cre and Adiponectin-cre each labeled less than 13% of all bone marrow stromal cells but 65%-90% of LepR + stromal cells in the metaphysis and diaphysis (Figures 1A and 1B). Consistent with prior studies (Zhou et al., 2014;Mukohira et al., 2019), they also labeled an increasing percentage of endosteal osteoblasts over time in adult mice, including 15%-40% of osteoblasts in the metaphysis and diaphysis of 24-week-old mice (Figures 1A and 1B). Adiponectin-creERT labeled less than 5% of all bone marrow stromal cells but 30%-40% of LepR + cells and 15%-30% of osteoblasts in the metaphysis and diaphysis at 4 and 12 weeks after tamoxifen treatment.   Figure 1E), and 41% of LepR + cells in Adiponectin-creERT; tdTomato mice ( Figure 1F) stained positively for Tomato. Analysis of bone marrow sections from Adiponectincre; tdTomato and Adiponectin-creERT; tdTomato mice showed that the Tomato + cells were perivascular and distributed throughout the bone marrow ( Figure S2A) with a morphology and location similar to LepR + stromal cells we described in prior studies (Ding et al., 2012;Ding and Morrison, 2013;Acar et al., 2015). Thus, LepR and Adiponectin expression strongly overlap in the bone marrow.
LepR is expressed by a subset of periosteal SSCs, but Adiponectin is not To begin to characterize SSCs in the periosteum, we analyzed by flow cytometry the expression of Sca1 and CD51, markers associated with periosteal SSCs (Matthews et al., 2021;Mo et al., 2021). Within the periosteum of 8-week-old mice, Sca1 + CD51 + cells were abundant, accounting for 16 ± 6.2% of all cells that were negative for hematopoietic (CD45 and Ter119) and endothelial (vascular endothelial [VE]-cadherin) markers. We tested the frequency of CFU-F among periosteal stromal cells and found that Sca1 + CD51 + cells were enriched for CFU-F (12 ± 4.6% of Sca1 + CD51 + cells formed fibroblast colonies), while Sca1 À CD51 + cells had limited CFU-F activity (2.5 ± 1.8% formed colonies) and Sca1 À CD51 À cells had no CFU-F activity (Figures 2A and 2B). The Sca1 + CD51 + cells were present in both the cambial and fibrous layers of the periosteum ( Figure S2B).
(I-K) The percentages of bone marrow or periosteal stromal cells (I), bone marrow or periosteal CFU-F (J), and periosteal Sca1 + CD51 + cells (K) that were Tomato + in 10-to 14-week-old Gli1 creERT2 ; tdTomato mice. Each dot reflects a different mouse that was analyzed in 1-3 independent experiments per genetic background.
(L) Representative images of Tomato and anti-Sclerostin antibody staining in cortical bone from femur diaphysis of 16-week-old Adiponectin-cre; tdTomato and 16-week-old Gli1 creERT2 ; tdTomato mice that were treated with tamoxifen at 8 weeks of age. Data are representative of 3 images per mouse with 3 mice per genotype.
(M) The percentage of Sclerostin + osteocytes that were Tomato + in the experiment from (L), and in 20-week-old Adiponectin-creERT; tdTomato mice treated with tamoxifen at 8-12 weeks of age.
(N and O) The percentages of Sclerostin + osteocytes that were Tomato + in 20-week-old Gli1 creERT2 ; tdTomato mice that were treated with tamoxifen at 8 weeks of age (N) or in 24-week-old Adiponectin-cre; tdTomato mice (O). All data represent mean ± SD. Statistical significance was assessed using a Kruskal-Wallis test followed by a Dunn's multiple comparisons adjustment (B and F-H) or an ordinary one-way ANOVA followed by Tukey's multiple comparisons correction (M). *p < 0.05, **p < 0.01, ***p < 0.001.
Periostin is expressed in the periosteum but not in the bone marrow, and its expression in periosteal cells increases after fracture (Duchamp de Lageneste et al., 2018;Horiuchi et al., 1999). In 8-to 10-week-old Periostin-cre; tdTomato mice, less than 1% of stromal cells, LepR + bone marrow cells, and CFU-F were Tomato + in the periosteum and the bone marrow ( Figure S2E). Even after drill injury in the tibia diaphysis, we still only observed rare Tomato + stromal cells or osteoblasts (data not shown). Periostin-cre, therefore, was not an effective marker of adult periosteal SSCs despite Periostin expression by these cells.
aSMA-creERT2 is expressed by a subset of adult periosteal cells that are enriched for CFU-F activity, and the number of positive cells increases after fracture (Ortinau et al., 2019;Matthews et al., 2021). We treated 8-to 10-week-old aSMA-creERT2; tdTomato mice (Wendling et al., 2009) with tamoxifen and observed that approximately 10% of stromal cells and CFU-F in the periosteum were Tomato + 3-6 days after tamoxifen treatment ( Figure S2F). In the bone marrow, less than 3% of stromal cells, LepR + cells, and CFU-F were Tomato + in the diaphysis and metaphysis ( Figure S2F). Under steady-state conditions, most of the Tomato + cells in the bone marrow of aSMA-creERT2; tdTomato + mice were periarteriolar stromal cells, consistent with smooth muscle cells. Fewer than 5% of the Tomato + cells in the bone marrow stained positively for LepR by flow cytometry.
Although few cells in the periosteum or bone marrow recombined with aSMA-creERT2 under steady-state conditions (Figure S2F), we tested whether administration of tamoxifen after injury would result in increased recombination, as shown previously with a different aSMA-creERT2 allele (Matthews et al., 2016(Matthews et al., , 2021. We performed drill injuries or bicortical fractures in aSMA-creERT2; tdTomato mice then treated with tamoxifen for up to 24 h after the injury ( Figures S3A and S3B). We observed a substantial increase in the number of Tomato + cells near the injury sites as compared to uninjured bones at 7 to 10 days after the injury (compare Figures S3D and S3G to Figure S2F). How-ever, these Tomato + cells were present in both the periosteum and the bone marrow near the injury sites. The Tomato + cells in the bone marrow could not have derived from Gli1 + periosteal cells as these cells do not give rise to cells within the bone marrow 10 days after drill injuries ( Figure S3C). This suggests that Acta2 (which encodes aSMA) was induced in bone marrow stromal cells and periosteal cells after the bone injuries. To assess this, we analyzed published single-cell RNA sequencing data that compared the gene expression profiles of Tomato + stromal cells from fractured and unfractured bones/bone marrow of Lepr cre ; tdTomato mice (Mo et al., 2021). The data included one cell cluster (#4) of periosteal cells and two clusters of LepR + bone marrow stromal cells (#7 and #10) ( Figures S4A-D) that exhibited a strong induction of Acta2 after fractures as compared to unfractured or irradiated bones ( Figure S4E). These data are consistent with our lineage tracing in suggesting that aSMA (Acta2) is induced in both LepR + bone marrow stromal cells and periosteal cells after bone injuries. aSMA-creERT2; tdTomato + cells gave rise to osteoblasts after drill injuries (Figure S3E) and osteoblasts and chondrocytes after bicortical fractures ( Figures S3H and S3I), but this CreER allele did not distinguish between periosteal and bone marrow SSCs after bone injuries.
When we treated 8-week-old Gli1 creERT2 ; tdTomato mice with tamoxifen, we observed Tomato expression in the periosteum but not in the bone marrow diaphysis ( Figures 2E and 2I). In the periosteum, 19 ± 7.5% of all cells that were not hematopoietic or endothelial cells ( Figure 2I), 44 ± 13% of CFU-F ( Figure 2J), and 36 ± 12% of Sca1 + CD51 + cells ( Figure 2K) were Tomato + 1-2 weeks after tamoxifen treatment. Gli1 creERT2 ; tdTomato + periosteal cells were present in both the cambial and fibrous layers of the periosteum ( Figure S4F). In contrast, within the bone marrow diaphysis, less than 2% of stromal cells and CFU-F were Tomato + (Figures 2I and 2J). Consistent with Figure S1D, we observed modestly increased labeling in the bone marrow metaphysis, where 3.6 ± 3.1% of stromal cells ( Figure 2I) and 21 ± 11% of CFU-F ( Figure 2J) were Tomato + . These data suggest that in the diaphysis of adult bones, Gli1 creERT2 preferentially labeled periosteal SSCs.
To determine whether there is any overlap between Gli1 + periosteal cells and LepR + periosteal cells, we stained sections of bones from Lepr cre ; tdTomato mice with an antibody against Gli1. Some Tomato + periosteal cells stained positively for Gli1 ( Figures S4I and S4J), suggesting substantial overlap between the Gli1 + and LepR + cells in the periosteum.
To address the identity of the 50%-60% of Sca1 + CD51 + cells in the periosteum that were Tomato negative in Gli1 creERT2 ; tdTomato mice ( Figures 2J and 2K), we treated them with tamoxifen and, 3 days later, isolated periosteal Tomato + Sca1 + CD51 + and Tomato negative Sca1 + CD51 + cells. We then measured the expression of Gli1 in these two cell populations by quantitative real-time-PCR. Tomato + Sca1 + CD51 + cells expressed the highest levels of Gli1 ( Figure S4H). Tomato negative Sca1 + CD51 + periosteal cells expressed lower levels, but nonetheless had levels of Gli1 significantly higher than LepR + bone marrow stromal cells ( Figure S4H). This suggests incomplete recombination by Gli1 creERT2 in Gli1 + periosteal cells, with recombination occurring preferentially in the cells with the highest levels of Gli1.

Adult cortical osteocytes arise primarily from bone marrow SSCs under steady-state conditions
To assess the contribution of Gli1 creERT2 -expressing periosteal SSCs to the production of osteocytes under steady-state conditions, we treated 8-week-old Gli1 creERT2 ; tdTomato mice with tamoxifen and then analyzed cortical bone in the femur diaphysis at 16 weeks of age. Only 2.3 ± 2.2% of osteocytes in the distal femur and 2.2 ± 1.8% of osteocytes in the proximal femur were Tomato + , and these were consistently observed near the endosteum, not near the periosteum ( Figures 2L and 2M). This suggests that these rare Tomato + osteocytes arose from rare Gli1 creERT2 -expressing cells in the bone marrow, where Gli1 creERT2 labeled up to 6% of osteoblasts and 1% of bone marrow CFU-F in the femur diaphysis ( Figure 1D). A similar analysis performed at 20 weeks of age found that less than 2% of osteocytes in the distal and proximal femurs were Tomato + (Figure 2N). We thus detected little contribution of periosteal cells to the production of osteocytes under steady state conditions in young adult femurs.
We also assessed the contributions of Adiponectin-cre-and Adiponectin-creERT-expressing bone marrow SSCs to steadystate osteogenesis in the cortical femur diaphysis. We found that in young-adult mice, new osteocytes overwhelmingly arose at the distal end of the femur, near the knee, as compared to the proximal end of the femur, near the hip ( Figure 2M). While few osteocytes arose from Gli1 + periosteal cells, 19 ± 8.2% of osteocytes were Tomato + in the distal femur of 16-week-old Adiponectin-cre; tdTomato mice, and 16 ± 3.0% of osteocytes were Tomato + in the distal femur of Adiponectin-creERT; tdTomato mice 16 weeks of age (8 weeks after tamoxifen treatment) ( Figures 2L and 2M). At 24 weeks of age, 27 ± 7.6% of osteocytes were Tomato + in the distal femur of Adiponectin-cre; tdTomato mice ( Figure 2O). These Tomato + osteocytes were consistently observed near the endosteum, not the periosteum ( Figure 2L). These data suggest that most new cortical bone osteocytes in the femur diaphysis of young-adult mice arise from bone marrow SSCs.

Drill injuries are healed primarily by bone marrow SSCs
To characterize the contributions of bone marrow and periosteal SSCs to the repair of bone injuries, we used Gli1 creERT2 as a marker of periosteal SSCs and Adiponectin-cre as a marker of bone marrow SSCs in adult femur and tibia diaphysis. While Adiponectin-creERT also marked bone marrow SSCs, the recombination efficiency of Adiponectin-cre was higher ( Figure 1C). We thus compared the contributions of bone marrow and periosteal SSCs to the repair of drill injuries in the tibia diaphysis of 8-to 12-week-old Adiponectin-cre; tdTomato and Gli1 creERT2 ; tdTomato mice ( Figure 3A). In this injury model, the periosteum was left intact, and a small dental drill with 0.8 mm diameter was used to create a hole all the way through the bone cortex on one side of the tibia diaphysis, exposing the bone marrow. In non-injured contralateral control tibias from these mice, Adiponectin-cre-labeled bone marrow stromal cells and Gli1 creERT2labeled periosteal stromal cells were relatively quiescent, with fewer than 4% of cells incorporating a 2-day pulse of 5 0 -ethynyl-2 0 -deoxyuridine (EdU) ( Figures 3B and 3C). However, in the tibia with the drill injury, 7.7 ± 3.6% of Adiponectin-cre-labeled bone marrow stromal cells and 20 ± 9.0% of Gli1 creERT2 -labeled periosteal cells in the diaphysis incorporated a 2-day pulse of EdU, initiated 24 h after the drill injury ( Figures 3B and 3C). Periosteal SSCs and bone marrow SSCs near the injury site thus became more proliferative after the injury.
When we performed a larger drill injury (1.6 mm diameter), we observed similar results: most osteoblasts derived from cells that expressed Adiponectin-cre, Adiponectin-creERT, or Lepr cre , and very few osteoblasts derived from cells that expressed Gli1 creERT2 (Figures S6G-S6L). Osteoblasts do not express Lepr or Adiponectin (Zhou et al., 2014;Tikhonova et al., 2019;Zhou et al., 2017), and only around 20% of osteoblasts are Tomato + in the diaphysis of young-adult Adiponectin-cre, Lepr cre , or Adiponectin-creERT mice ( Figures 1A-1C). While pre-existing osteoblasts may make a small contribution to the repair of bone injuries (Matthews et al., 2021), the approximately 20% of osteoblasts that would have been Tomato + prior to the injury could not account for the 50%-80% of osteoblasts that were Tomato + after the repair of the injury. The data, therefore, suggest that undifferentiated LepR + /Adiponectin + bone marrow SSCs are primarily responsible for the repair of drill injuries.

Bicortical fractures are healed primarily by periosteal SSCs
We next performed bicortical fractures in the tibia diaphysis of 10-to 14-week-old Gli1 creERT2 ; tdTomato; Col1(2.3)-GFP mice, Lepr cre ; tdTomato; Col1(2.3)-GFP mice, and Adiponectin-cre; tdTomato; Col1(2.3)-GFP mice. These fractures healed over a period of 5 weeks ( Figure 4A) through endochondral ossification in which a cartilage callus formed at the fracture site followed by a boney hard callus and, finally, cortical bone and bone marrow stromal cell regeneration (Figures S5A and S5F-S5H) (Colnot, 2009). Both periosteal SSCs and bone marrow SSCs in the diaphysis began to proliferate after the fracture, with approximately 40% incorporating a 48-h pulse of EdU starting 24 h after ll OPEN ACCESS Article the fracture (Figures 4B, 4C, and S7A-S4D). In each genetic background, we analyzed the percentage of Aggrecan + chondrocytes in the cartilage callus that were Tomato + 7 days after the fracture. We detected no Tomato + chondrocytes in Adiponectin-cre; tdTomato mice ( Figures 4D and 4G). In Lepr cre ; tdTomato mice, 9.8 ± 8.9% of chondrocytes were Tomato + ( Figures 4E and 4G), and in Gli1 creERT2 ; tdTomato mice, 31 ± 14% of chondrocytes were Tomato + (Figures 4F and 4G). Chondrocytes in the callus thus arose primarily from periosteal SSCs that expressed Gli1 or Gli1 and Lepr.  (K) The percentage of osteocytes that were Tomato + in the regenerated bone cortex 35 days after the injury. Each dot reflects a different mouse. All data represent mean ± SD. Statistical significance was assessed using a twotailed Student's t test (C), an ordinary one-way ANOVA followed by Tukey's multiple comparisons correction (H-J), or a Brown-Forsythe and Welch ANOVA followed by a Dunnett T3 multiple comparisons correction (K). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.
We also analyzed the contributions of these cell populations to Col1(2.3)-GFP + osteoblasts in the hard callus 14 days after the fracture. In the woven bone of the hard callus, less than 2% of osteoblasts were Tomato + in Adiponectin-cre; tdTomato; Col1(2.3)-GFP mice ( Figures 5A and 5D At 35 days after the fracture, when it was largely healed, only 1.9 ± 1.5% of the osteocytes in the new cortical bone at the fracture site were Tomato + in Adiponectin-cre; tdTomato mice ( Figures 5F and S6D), and 21 ± 15% were Tomato + in Lepr cre ; tdTomato mice ( Figures 5F and S6E). In contrast, 54 ± 13% of new cortical osteocytes at the fracture site were Tomato + in Gli1 creERT2 ; tdTomato mice (Figures 5F and S6F). By day 35, the trabecular bone that had formed within the bone marrow at the fracture site had largely been resorbed. Thus, the new cortical bone that persisted at the healed fracture site was derived mainly from periosteal SSCs.
To test the functional importance of Gli1 creERT2+ periosteal SSCs for bicortical fracture repair, we generated Gli1 creERT2 ; iDTA mice, in which Gli1 creERT2 -expressing cells were ablated by expression of the intracellular A subunit of diphtheria toxin following tamoxifen-mediated induction (Voehringer et al., 2008). Three days after the last tamoxifen treatment, periosteal CFU-F and Sca1 + CD51 + cells were depleted as compared to littermate controls ( Figures 6A and 6B). Following bicortical fracture, these mice had delayed fracture healing, with significantly reduced total callus and callus bone volumes at 14 days ( Figures 6C, 6D, and 6F). These data indicate that Gli1 + periosteal SSCs contribute functionally to bicortical fracture repair.
We next tested if osteogenic differentiation by periosteal Gli1 creERT2+ cells contributes to bicortical fracture healing. Wnt/ b-catenin signaling promotes osteogenesis during fracture healing (Matsushita et al., 2020;Chen et al., 2007), though to our knowledge it has not yet been tested whether b-catenin is necessary in periosteal SSCs. To test this, we analyzed fracture healing in Gli1 creERT2 ; Ctnnb1 flox/flox mice. Thirty-five days after bicortical fracture, the regenerating bone within the fracture callus of Gli1 creERT2+ ; Ctnnb1 flox/flox mice appeared discontinuous, suggesting incomplete healing ( Figure 6G). We quantified bone volume as a fraction of total volume (BV/TV) in the fracture callus and found it was significantly reduced in Gli1 creERT2+ ; Ctnnb1 flox/flox mice as compared to Ctnnb1 flox/flox controls ( Figures 6E and 6G). These data further support the functional importance of osteogenesis by Gli1 + periosteal cells for fracture repair.
Periosteal SSCs regenerate LepR + bone marrow stromal cells during fracture repair During the healing of bicortical fractures, bone marrow also regenerated at the fracture site at later time points after the resorption of the callus and the trabecular bone that transiently stabilized the fracture at the early stages of repair (Figures 7A and 7B). We thus analyzed the contributions of Adiponectin-creERT-expressing bone marrow cells and Gli1 creERT2 -expressing periosteal cells to the regeneration of bone marrow stromal cells at the fracture site. Tamoxifen was administered 2-4 weeks before the fracture. In normal adult tibia diaphysis bone marrow, nearly half of LepR + bone marrow stromal cells were Tomato + in Adiponectin-creERT; tdTomato mice ( Figures 7C and 1C), but only rare bone marrow stromal cells were Tomato + in Gli1 creERT2 ; tdTomato mice ( Figures 7D and 2I). Conversely, in bone marrow that regenerated at the fracture site 35 days after the fracture, we observed widespread labeling of bone marrow stromal cells by Gli1 creERT2 ( Figure 7D) but not by Adiponectin-creERT (Figure 7C). This suggests that bone marrow stromal cells regenerate largely from periosteal SSCs.   The stromal cells that arose from Gli1 creERT2 -expressing periosteal cells in regenerated bone marrow were perivascular, with an appearance and localization similar to LepR + stromal cells in normal bone marrow (Figures 7E and 7F). To test whether these cells expressed LepR, we stained sections through regenerated bone marrow in Gli1 creERT2 ; tdTomato mice with anti-LepR antibody. Many Tomato + perivascular stromal cells stained positively for LepR at 5 weeks ( Figure 7E) and 12 weeks ( Figure S7E) after fracture. We also stained sections with anti-endomucin antibody; endomucin is an endothelial marker that is highly expressed by sinusoidal endothelial cells (Kusumbe et al., 2014). The Tomato + cells were often adjacent to sinusoidal blood vessels in regenerated bone marrow ( Figure 7F). By flow cytometry, approximately half of Tomato + bone marrow stromal cells in the (F) The percentage of osteocytes that were Tomato + in the regenerated bone cortex 35 days after the fracture. Each dot reflects a different mouse analyzed in 1-3 experiments per genetic background. All data represent mean ± SD. Statistical significance was assessed using a Kruskal-Wallis test followed by Dunn's multiple comparisons adjustment (D) or a one-way ANOVA followed by Tukey's multiple comparisons correction (E-F). **p < 0.01, ***p < 0.001, ****p < 0.0001. diaphysis of Gli1 creERT2 ; tdTomato tibias stained positively with anti-LepR antibody at 7 to 12 weeks after fracture ( Figures S7F  and S7G). These Tomato + cells represented 8%-20% of all bone marrow LepR + cells in the diaphysis of these tibias at 5 to 12 weeks after fracture ( Figure 7G). Gli1 + periosteal cells thus regenerate LepR + bone marrow stromal cells after fracture repair. We next tested whether the LepR + bone marrow stromal cells that regenerated from Gli1 + periosteal cells acquire the properties of hematopoietic stem/progenitor cell (HSPC) niche cells. We isolated Tomato + stromal cells from the tibial diaphysis bone marrow of Gli1 creERT2 ; tdTomato mice 7 weeks after bicortical fracture. SCF and Cxcl12 are niche factors that are strongly expressed by LepR + bone marrow stromal cells and required for the maintenance of HSPCs (Ding et al., 2012;Ding and Morrison, 2013). We found that Tomato + stromal cells from the bone marrow of Gli1 creERT2 ; tdTomato tibias lost the expression of Gli1 as compared to periosteal cells ( Figure 7J), but began expressing Scf and Cxcl12, genes that exhibit little or no expression in periosteal cells ( Figures 7H and 7I). Indeed, 95 ± 0.7% of Tomato + and LepR + stromal cells from the bone marrow of Gli1 creERT2 ; tdTomato mice were Scf-GFP + at 5 weeks after the fracture ( Figures S7H, 7K, and 7L). Gli1 + periosteal cells thus gave rise to bone marrow stromal cells during bicortical fracture repair that had properties similar to HSPC niche cells, losing the

DISCUSSION
In this study, we confirmed that LepR marks SSCs in adult bone marrow ( Figure 1A) and that these cells give rise to most of the osteoblasts and osteocytes that contribute to the maintenance of the adult skeleton under steady-state condi- tions ( Figures 1A and 2M). We also found that LepR + bone marrow SSCs are the cells primarily responsible for the repair of drill injuries ( Figure 3). However, Gli1 + periosteal SSCs are primarily responsible for the repair of bicortical fractures (Figures 4 and 5). We previously reported that LepR + bone marrow SSCs produce most of the osteoblasts responsible for fracture repair in adult mice (Zhou et al., 2014). However, in the current study we found that in addition to the SSCs within the bone marrow ( Figure 1A), Lepr cre also labels a minority of SSCs in the periosteum ( Figures 2F-2H). Thus, the contributions of LepR + cells to fracture repair reflect the activities of both LepR + bone marrow SSCs and a subpopulation of LepR + periosteal SSCs. Our results are consistent with recent reports that Lepr cre recombines in a subset of SSCs in the periosteum (Ortinau et al., 2019;Mo et al., 2021). In the current study, we confirmed that LepR + cells within the bone marrow give rise to transient trabecular bone within the bone marrow at the fracture site. LepR + bone marrow SSCs and Gli1 + periosteal SSCs thus both contribute to the stabilization of bicortical fractures in the first few weeks after injury by forming trabecular bone and cartilage, respectively, at the fracture site.

Article
While Lepr cre recombined in bone marrow SSCs and a subset of periosteal SSCs, Adiponectin-cre and Adiponectin-creERT recombined exclusively in bone marrow SSCs. In a prior study, treatment of Adiponectin-creERT; tdTomato mice with low dose tamoxifen recombined in 5% of LepR + bone marrow cells that were fated to form adipocytes (Zhou et al., 2017). In the current study, we treated Adiponectin-creERT; tdTomato mice with a much higher dose of tamoxifen, recombining in approximately 40% of LepR + bone marrow cells, including in most bone marrow CFU-F ( Figure 1C). Consistent with this, nearly all Adiponectinexpressing stromal cells in the bone marrow were LepR + , and most LepR + stromal cells were Adiponectin + (Figures 1D-1F). While Adiponectin is often considered an adipocyte marker, Adiponectin + LepR + cells do not express other markers of mature adipocytes and do not have large lipid vacuoles, adipocyte size, or morphology. Adiponectin + bone marrow stromal cells have been described as marrow adipogenic lineage precursors (MALPs) based on single-cell RNA sequencing (Zhong et al., 2020), but fate mapping shows that these cells are not committed to the adipocyte lineage. Only a subset of LepR + stromal cells is fated to form adipocytes (Tikhonova et al., 2019;Baryawno et al., 2019;Baccin et al., 2020), and even this subset retains the ability to form bone in response to injury (Matsushita et al., 2020). Therefore, most LepR + Adiponectin + bone marrow stromal cells are not adipocytes or adipocyte-committed progenitors. Bone marrow SSCs are thus reminiscent of other stem cells, such as hematopoietic stem cells (Morrison et al., 1995), that sometimes express markers that are also expressed by their differentiated progeny.
Within the periosteum, Sca1 + CD51 + cells were enriched for CFU-F activity ( Figure 2B), consistent with prior reports  (G) Flow cytometric quantitation of the percentage of LepR + (antibody staining) stromal cells that were Tomato + in bone marrow from uninjured mice or mice 5-12 weeks after bicortical fracture.
(H-J) Quantitative real-time-PCR analysis of gene expression in sorted CD45 + /Ter119 + bone marrow hematopoietic cells, Sca-1 + CD51 + periosteal cells, LepR + bone marrow stromal cells from uninjured bones, and Tomato + LepR + bone marrow stromal cells from Gli1 creERT2 ; tdTomato mice 7 weeks after fracture. Each dot represents an independent sample. (K and L) Representative confocal images of a tibia from a Gli1 creERT2 ; tdTomato; SCF-GFP mouse showing an uninjured region of the bone marrow (K) or a region that had regenerated 5 weeks after bicortical fracture (L). In (G), each dot represents a different mouse, analyzed in 1-2 experiments per time point. All data represent mean ± SD. Mice were analyzed in 1-5 experiments per genetic background.

Article
Cell Stem Cell 29, 1-15, November 3, 2022 11 (Matthews et al., 2021;Mo et al., 2021). Gli1 creERT2 recombined in many of these Sca1 + CD51 + cells ( Figure 2K). Conversely, Gli1 creERT2 recombined in only rare cells within the diaphysis of long bones, indicating that it is an effective marker of periosteal SSCs in the diaphysis of adult long bones. Within the periosteal compartment, Gli1 creERT2 recombined in approximately 40% of CFU-F and Sca1 + CD51 + cells (Figures 2J and 2K). This may reflect incomplete recombination among Gli1-expressing periosteal cells ( Figure S5C). It is also possible that there is a Gli1negative subset of periosteal SSCs, potentially with properties that differ from the Gli1 + SSCs. The LepR + Adiponectin + bone marrow SSCs and the Gli1 + periosteal SSCs that we studied overlap with previously characterized SSC populations. SSCs have been isolated by flow cytometry from enzymatically dissociated fetal/neonatal and adult bones/bone marrow based on the expression of CD51 and other markers (Chan et al., 2015). The CD51 + SSCs are present in multiple regions of fetal and/or adult bones, including in the periosteum, bone marrow, and growth plate, but it is not clear if the fates of CD51 + SSCs differ among different regions of bone (Ambrosi et al., 2021a(Ambrosi et al., , 2021b. The mechanisms that ensure different SSC populations repair different kinds of bone injuries remain unclear. Gli1 + periosteal cells and LepR + Adiponectin + bone marrow stromal cells both begin dividing after drill injuries and bicortical fractures (Figures 3C and  4C). Therefore, both SSC populations sense both types of injuries even though the Gli1 + periosteal cells make little contribution to the repair of drill injuries (Figures 3G-3K). Our data show that b-catenin promotes the osteogenic response of Gli1 + periosteal cells to bicortical fractures (Figures 6E and 6G). This is consistent with the observation that b-catenin deletion in cells at bicortical fracture sites as a result of viral transduction with Cre recombinase impaired fracture healing (Chen et al., 2007). Another recent study demonstrated that deletion of b-catenin from a subset of bone marrow LepR + cells impaired the repair of drill hole injuries (Matsushita et al., 2020). Wnt pathway activation thus appears to promote the osteogenic response of both periosteal and bone marrow SSCs to bone injuries.
In addition to damaging bone, bicortical fractures damage the bone marrow at the fracture site. To our knowledge, the process by which bone marrow stroma regenerates at fracture sites has not been studied. Surprisingly, we found that Gli1 + periosteal cells gave rise to LepR + bone marrow stromal cells in regenerated bone marrow ( Figures 7D-7G and S7E-S7H). These LepR + bone marrow stromal cells that arose from Gli1 + periosteal cells had properties similar to HSPC niche cells in terms of their perivascular localization, their loss of Gli1 expression, and their acquisition of Scf and Cxcl12 expression-genes not expressed by periosteal cells (Figures 7H-7J). This suggests that Gli1 + periosteal SSCs were able to transdifferentiate into LepR + bone marrow SSCs late in the process of bicortical fracture repair.

LIMITATIONS OF THE STUDY
The data reported in this study were collected in the femurs and tibias of young-adult mice. Thus, the markers we describe for bone marrow and periosteal SSCs have been verified only in this context. It is possible that these markers are less effective in some other bones or during aging. It is also possible that the functions of bone marrow versus periosteal SSCs differ in some other bones or during aging.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Sean Morrison (sean.morrison@utsouthwestern.edu).

Materials availability
This study did not generate new unique reagents.

Data and code availability
Microscopy and flow cytometry data reported in this paper will be shared by the lead contact upon request. This paper analyzes existing, publicly available data. The accession numbers for the datasets are listed in the key resources table. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Drill injuries and fractures
For drill injuries, mice were anesthetized with isolfluorane (Covetrus), a small incision was made in the skin over the central region of the tibia diaphysis and a micro dental drill with 0.8 mm or 1.6 mm diameter was used to make a hole through one side of the tibia cortex, exposing the bone marrow. The skin was closed with 4/0 sutures (Covetrus) and 1.2 mg/kg of Buprenorphine SR (ZooPharm) was injected for analgesia. For bicortical fractures, a 27 gauge needle was inserted into the intramedullary canal of the tibia through the knee joint after anesthesia, and the tibia was fractured mid-diaphysis by 3-point bending using a Zondervan apparatus (Zondervan et al., 2018). The location of the pin was verified by X-ray radiography on an AMI HTX Optical Imaging System (Spectral Instruments Imaging). Buprenorphine SR was injected immediately after the surgery, while the mice were still anesthetized.

Tissue dissociation
To isolate bone marrow cells, bones were thoroughly cleaned to remove periosteal cells. The metaphyses of the femur and/or tibia were separated from the diaphysis with dissection scissors, and the diaphysis was flushed with dissociation buffer using a 25-gauge

Oligonucleotides
Genotyping primers IDT Table S1 qRT-PCR primers IDT needle. The diaphysis bone was then crushed and combined with the flushed marrow for dissociation. The metaphyses were separately crushed in dissociation buffer using a mortar and pestle, and the contents were transferred to a tube for digestion. Bone marrow dissociation buffer contained type I collagenase (3 mg/mL, Worthington Biochemical), dispase (4 mg/mL, Sigma-Aldrich) and DNase I (1 U/ml, Roche) in Hank's balanced salt solution (HBSS) with Ca 2+ and Mg 2+ (Corning). Samples were dissociated at 37 C for 30 min as described previously (Yue et al., 2016). Dissociated cells were then transferred into HBSS (without Ca 2+ and Mg 2+ ) plus 2% serum, centrifuged, then resuspended in HBSS (without Ca 2+ and Mg 2+ ) plus 2% serum and filtered through 40mm mesh to generate a single cell suspension.
To isolate periosteal cells, muscle was carefully trimmed from femurs and tibias. Bones were placed in HBSS (without Ca 2+ and Mg 2+ ) plus 2% serum on ice for 30 min to loosen the periosteal layer from the bone surface. Periosteum was then peeled from the long bones, placed on a nylon mesh filter, and washed with cold HBSS (with Ca 2+ and Mg 2+ ) to wash away contaminating bone marrow cells. Chunks of periosteum were then placed in periosteum dissociation buffer containing collagenase D (1 mg/mL, Roche), dispase (2 mg/mL, Sigma-Aldrich) and DNase I (1 U/ml, Roche) in HBSS with Ca 2+ and Mg 2+ (Corning). Samples were dissociated at 37 C for 45 min. The dissociated cells were then transferred into HBSS (without Ca 2+ and Mg 2+ ) plus 2% serum, washed by centrifugation and resuspension, and filtered through a 40mm mesh to generate a single cell suspension.
To isolate cells from fractured bones, the fractured region was separated from the rest of the bone with dissection scissors, then crushed in dissociation buffer using a mortar and pestle and transferred to a tube for dissociation. The fractured bone dissociation buffer contained collagenase D (1 mg/mL, Roche), type I collagenase (3 mg/mL, Worthington Biochemical), dispase (4 mg/mL, Sigma-Aldrich) and DNase I (1 U/ml, Roche) in HBSS with Ca 2+ and Mg 2+ (Corning). Samples were dissociated at 37 C for 45 min. Dissociated cells were then transferred into HBSS (without Ca 2+ and Mg 2+ ) plus 2% serum, washed by centrifugation and resuspension, and filtered through a 40mm mesh to generate a single cell suspension.
Safranin O with fast green staining Frozen sections were brough to room temperature. Sections were then stained with Weigert's iron hematoxylin (#1.15973, Sigma-Aldrich) for 10 min. Sections were washed in distilled water 3 times for 5 min each. Sections were then stained with fast green FCF (#C.I. 42053, Sigma-Aldrich) for 5 min, then rinsed quickly with 1% acetic acid (#A38-212, Fisher) diluted in distilled water. Finally, sections were stained with 0.1% Safranin O (#TMS-009, Sigma-Aldrich) solution for 5 min and washed briefly in distilled water 3 times.
Hematoxylin and eosin (H&E) staining Frozen sections were treated with Hematoxylin QS Counterstain (#H-3404-100, Vector Laboratories), followed by three washes in distilled water. Sections were then treated briefly with bluing buffer (#CD702, DAKO), followed by three washes in distilled water. Finally, sections were treated briefly with Eosin Y solution (#HT110116, Sigma) followed by three washes in distilled water.

CFU-F assay
Freshly dissociated bone marrow cells that were unfractionated or double sorted for combinations of cell surface markers were plated at clonal density in 6-well plates such that individual CFU-F could form spatially distinct colonies that could be counted. Unfractionated bone marrow was plated at 100,000 cells per well, and unfractionated periosteum was plated at 1000 cells per well in 6 well plates. These cell numbers led to the formation of 30-50 spatially distinct fibroblast colonies per well. The cells were cultured in DMEM with high glucose (Gibco) plus 20% fetal bovine serum (Sigma), 10mM ROCK inhibitor Y-27632 (Tocris, 1254), and 1% penicillin/streptomycin (Invitrogen) at 37 C in gas-tight chambers (Billups-Rothenberg) flushed with 1% O 2 and 6% CO 2 (balance Nitrogen) to maintain physiological oxygen levels that promoted survival and proliferation (Morrison et al., 2000). The medium was changed the next day to eliminate contaminating hematopoietic cells and dead cells. To count the colonies, the cultures were stained with 0.1% Toluidine blue in 10% formalin eight days after plating and the percentage of colonies that was Tomato + was assessed by direct fluorescence (without antibody staining) using an Olympus IX83 inverted microscope.
Quantitative RT-PCR Cells were sorted directly into 75mL of Buffer RLT lysis buffer (Qiagen) containing 1% b-mercaptoethanol and stored at À80 C. RNA was extracted using the Qiagen RNeasy Micro Kit (74,004, Qiagen) according to the manufacturer's instructions. The eluted RNA was then used to make cDNA using iScript cDNA Synthesis Kit (#170-8891, Bio-Rad Laboratories) according to the manufacturer's instructions. cDNA was then used to perform qPCR with iTaq Universal SYBR Green Supermix (#172-5124, Bio-Rad Laboratories) and results were analyzed using BioRad CFX Maestro Software.

MicroCT analysis
Tibias from mice at 0 to 35 days after drill injury or 7 to 35 days after fracture were fixed in 4% paraformaldehyde overnight at 4 C, washed with PBS, and scanned using a Scanco Medical mCT 35. Tibias were scanned at an isotropic voxel size of 3.5mm, with peak tube voltage of 55 kV and current of 0.145 mA (mCT 35; Scanco). A three-dimensional Gaussian filter (s = 0.8) with a limited, finite filter support of one was used to suppress noise. A threshold of 342-1000 was used to segment mineralized bone from air and soft tissues. To determine callus volume, Scanco Medical software was used to draw contours around the outside of the callus in the x-y plane on every 10th level on the z axis, and the morph function was used to interpolate contours for the intervening levels. Callus total volume and bone volume were then determined using the same number of z-slices for each bone in the experiment. Analysis of bone volume as a fraction of total volume (BV/TV) was performed using the entire callus volume for each bone.

Quantification of Tomato + cells in confocal images
To quantitate the percentage of cells within sections that were Tomato + , relevant representative regions from confocal images were chosen from multiple areas of the bone or the bone callus and at least 150 cells (osteoblasts and chondrocytes) or 80 cells (osteocytes) were counted. First, we counted the total number of osteoblasts/chondrocytes/osteocytes in the field of view in the absence of Tomato fluorescence, then the Tomato channel was turned on and the percentage that were Tomato + was determined. This procedure was repeated for each mouse.
Single cell RNA-sequencing analysis Single-cell RNA sequencing data from the Gene Expression Omnibus (GEO) study GSE138689 were used for the analysis (Mo et al., 2021). Samples' barcodes, features, and matrix files were downloaded from the study's GEO entry and processed using R 4.0.2 with the Seurat 3.2 package (Stuart et al., 2019). Low-quality cells with >10% of their transcripts from mitochondrial genes, <100 Unique Molecular Identifier (UMI) counts, or <100 genes per cell were excluded from analysis. Filtered samples were then normalized and integrated using Seurat's variance stabilizing transformations for single-cell UMI data (SCTransform). Clustering was performed, and samples and genes were visualized using Seurat as well.

Statistical methods
In each type of experiment, multiple mice were tested, generally in multiple independent experiments performed on different days. Mice were allocated to experiments randomly and samples were processed in an arbitrary order, but formal randomization and blinding techniques were not used. Prior to analyzing the statistical significance of differences among treatments, we tested whether the data were normally distributed and whether variance was similar among treatments. To test for normal distribution, we performed the Shapiro-Wilk test when 3 % n < 20 or the D'Agostino Omnibus test when n R 20. To test if variability significantly differed among treatments, we performed F-tests (for experiments with two treatments) or Levene's median tests (for more than two treatments). When the data significantly deviated from normality or variability significantly differed among treatments, we log2-transformed the data and tested again for normality and variability. If the transformed data no longer significantly deviated from normality and equal variability, we performed parametric tests on the transformed data. If the transformed data still significantly deviated from normality or equal variability, we performed non-parametric tests on the non-transformed data.
All the statistical tests we used were two-sided, where applicable. To assess the statistical significance of a difference between two treatments, we used t-tests (when mice were not littermates and a parametric test was appropriate), or Mann-Whitney tests (when mice were not littermates and a non-parametric test was appropriate). To assess the statistical significance of differences between more than two treatments, we used repeated measures one-way or two-way ANOVAs (when mice were littermates and/or cells were from same mice, and a parametric test was appropriate) followed by Tukey's, Dunnet's, or Sidak's multiple comparisons adjustment. All statistical analyses were performed with Graphpad Prism 8.3 or R 3.5.1 with the fBasics package. All data represent mean ± SD. Samples sizes were not pre-determined based on statistical power calculations.     Tamoxifen was administered to aSMA-creERT2; tdTomato mice on days 0 and 1 after fracture.
(H) The percentage of Aggrecan + chondrocytes that were Tomato + in the fracture callus at 7 days after fracture. (I) The percentage of Col1(2.3)-GFP + osteoblasts in the bone marrow or fracture callus that were Tomato + at 10 days after the fracture. Tomato + cells were present in both the bone marrow and the periosteum near the sites of bone injuries in aSMA-creERT2; tdTomato mice, while Tomato + cells were restricted to the periosteum and bone callus in Gli1 creERT2 ; tdTomato mice. All images in this figure were acquired as tiled images and stitched together. Each dot represents a different mouse, analyzed in 1 to 2 experiments per genetic background. All data represent mean ± standard deviation.          Adipoq-creERT