A FoxA2+ long-term stem cell population is necessary for growth plate cartilage regeneration after injury

ARTICLE A FoxA2+long-term stem cell population is necessary for growth plate cartilage regeneration after injury Shanmugam Muruganandan 1, Rachel Pierce1, Dian Astari Teguh 2, Rocio Fuente Perez3, Nicole Bell4, Brandon Nguyen5, Katherine Hohl 2,6, Brian D. Snyder7, Mark W. Grinstaff 6, Hannah Alberico1, Dori Woods1, Yiwei Kong1, Corneliu Sima 8, Sanket Bhagat 9, Kailing Ho10, Vicki Rosen 10, Laura Gamer10 & Andreia M. Ionescu 1✉ Longitudinal bone growth, achieved through endochondral ossification, is accomplished by a cartilaginous structure, the physis or growth plate, comprised of morphologically distinct zones related to chondrocyte function: resting, proliferating and hypertrophic zones. The resting zone is a stem cell-rich region that gives rise to the growth plate, and exhibits regenerative capabilities in response to injury. We discovered a FoxA2+group of long-term skeletal stem cells, situated at the top of resting zone, adjacent the secondary ossification center, distinct from the previously characterized PTHrP+stem cells. Compared to PTHrP+ cells, FoxA2+cells exhibit higher clonogenicity and longevity. FoxA2+cells exhibit dual osteo-chondro-progenitor activity during early postnatal development (P0-P28) and chondrogenic potential beyond P28. When the growth plate is injured, FoxA2+cells expand in response to trauma, and produce physeal cartilage for growth plate tissue regeneration. https://doi.org/10.1038/s41467-022-30247-1 OPEN 1Department of Biology, 134 Mugar Life Sciences Building, Northeastern University, 360 Huntington Ave, Boston, MA 02115, USA. 2Centre for Advanced Orthopedic Studies, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, MA 02215, USA. 3Division of Pediatrics, University of Oviedo, Oviedo 33206, Spain. 4New York University College of Dentistry, 345 E.24th St, New York, NY 10010, USA. 5Moderna Therapeutics, One Upland Rd, Norwood, Ohio, MA 02062, USA. 6Departments of Biomedical Engineering, Chemistry, and Medicine, Boston University, 590 Commonwealth Ave, SCI 518, Boston, MA 02215, USA. 7Department of Orthopedic Surgery, Boston Children’s Hospital, 300 Longwood Ave, Boston, MA 02115, USA. 8Department of Oral Medicine, Infection, and Immunity, Harvard School of Dental Medicine, 188 Longwood Avenue, Boston, MA 02115, USA. 9Ultragenyx Pharmaceutical, 840 Memorial Drive, Cambridge, MA 02139, USA. 10 Department of Developmental Biology, Harvard School of Dental Medicine, 188 Longwood Avenue, Boston, MA 02115, USA. ✉email: [email protected] NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications 1 1234567890():,; Longitudinal bone growth is achieved by endochondral ossification, implemented by the physis or growth plate (GP), a cartilaginous structure located between the epiphysis and metaphysis at the ends of long bones forming the appendicular skeleton. Human GP is comprised of five morphologically distinct zones: resting (RZ), proliferation (PZ), hypertrophy (HZ), calcification (CZ), and ossification (OZ)1. Physeal injuries are a major cause of skeletal morbidity in growing children, manifesting as angular or shortened limb deformities, due to partial or complete arrest of GP function2. Clinically, GP injuries are classified according to the Salter–Harris (SH) system, predicated on fracture morphology involving the physis ± epiphysis and/or metaphysis: SH1) physeal shear separation; SH2) physis +metaphysis; SH3) physis +epiphysis; SH 4) physis +epiphysis +metaphysis; SH 5) crush injury to physis. SH1 fractures have the best prognosis, healing without disrupted growth. The cleavage plane is primarily through the HZ, structurally the weakest zone comprising the GP, owing to lack of calcification and limited collagenous matrix. Stiffened by a collagen matrix, the adjacent RZ and PZ are unaffected3,4. In contrast, SH 3–5 injuries disturb all zones, resulting in disturbed bone growth3,4. These clinical scenarios suggest that involvement of the RZ affects the ability of the GP to regenerate in response to injury. In rabbits, even when the PZ and HZ from ulnar GP were excised, the restoration of the GP was still possible, as long as the RZ remained intact5,6. As the name implies, the “resting zone”houses a quiescent progenitor cell population1. Several studies have identified in mouse7, rat8, and rabbit9slow-cycling cells present in the RZ immediately adjacent to the secondary ossification center (SOC). In pulsechase experiments, animals were administered [3H] thymidine or 5-bromo-2-deoxyuridine (BrdU) for a prolonged period sufficient to allow slow-cycling stem cells to complete one cycle. At the end of the chase, the label diminishes in proliferating cells, but remains undiluted in slow-cycling cells. While these reports demonstrate the presence of slow-cycling, potential progenitor cells at the RZ, their genetic signature remains undefined, as there is no biomarker to isolate and characterize these cells. Recently Newton et al., using clonal genetic tracing with a col2CRE;ERT;R26-Confetti mouse line and functional perturbations, established that formation of the SOC triggered GP chondroprogenitors to undergo a radical shift in clonality and to acquire self-renewal capabilities10. However, since “Confetti” labeling is based on collagen type II expression, an uber-marker for all GP cells, there was no unique biomarker to isolate and/or characterize individual stem cell subgroups. Next, Mizuhashi et al. employed a PTHrPCre.ERT;Tomatof/+mouse strain to reveal that PTHrP marked a subset of stem cells located at the bottom of the RZ11. Labeling, tracing and isolating RZ cells based on their PTHrP expression both in vivo and in vitro, demonstrated that a subgroup PTHrP+(positive) stem cell evolved predominantly into columnar chondrocytes11. Together with another yet unidentified PTHrP−(negative) subgroup located at the top of the RZ, adjacent to the SOC, PTHrP+cells, located primarily at the bottom of the RZ, contributed to long-term GP growth11. In the present work, we prove that a population of FoxA2+col10−cells, located at the top of the RZ, are the PTHrP- (negative) stem cell population, and represent a subgroup of longterm skeletal stem cells (LTSSC) capable of dual osteo-chondroprogenitor activity during early postnatal development (P0-P28) and chondrogenic potential beyond P28, fundamental to GP turnover and regeneration following injury. Results FoxA2 expression prefigures SOC formation. We previously discovered that FoxA transcription factors are key regulators of chondrocyte hypertrophy12. FoxA1-3 are highly expressed in the hypertrophic zone (HZ) of newborn mice GP (Fig. 1a, f, k). However, unlike FoxA3, which is expressed in a broader domain throughout the GP (Fig. 1f), or FoxA1, which is highly expressed in the HZ but very little elsewhere (Fig. 1k), FoxA2 is expressed in two separate domains: the HZ and a discrete periarticular domain located at the ends of the long bones (Fig. 1a). Throughout postnatal development (P0-P14), this distinct FoxA2+population is continuously enlarging, possibly prefiguring the formation of the SOC (Fig. 1a–e). FoxA1 and FoxA3 also correlate with SOC development, but emerge in the epiphyseal cartilage around P5 (Fig. 1f–o), much later than FoxA2, which is expressed from birth (P0) (Fig. 1a–e). In the presumptive SOC, the first col10+cells appear by P7, as shown in Tg.col10mCherry mice, but FoxA2+cells appear as early as P0 (Fig. 1p–t). Comparison between FoxA2+ cells in the periarticular region with the columnar GP cells, did not reveal any histological or morphological differences and no noticeable change in Ki67 expression and proliferation, prior to P5 (Fig. 1u–y). However, after P5 FoxA2+cells downregulate the expression of Ki67, in preparation for hypertrophic differentiation (Fig. 1w, x). Altogether, these experiments suggest that FoxA2 expression marks the epiphyseal cartilage, most of which will form the secondary ossification center, far sooner than col.10 expression. FoxA2+cells give rise to the GP RZ top compartment.To determine whether FoxA2 marks the fate of the cells destined to become SOC, we labeled and traced FoxA2+cells, using a FoxA2CreERT/+mouse line consisting of a tamoxifen-inducible CreERT2 driver knocked-in the FoxA2 3’UTR13. Two tamoxifen pulses, administered P3-4 or P7-8, to FoxA2CreERT/+; ZsGreenfl/+;Tg.col10mcherry mice gave rise, by P18, to ZsGreen+ cells present primarily in the SOC, with a few ZsGreen+cells present at the top of the GP RZ (Fig. 2A(a, b, e)). It was previously published that hypertrophic chondrocytes become bone cells and contribute to the osteogenic lineage14–16. FoxA2 is not expressed in the epiphyseal bone of the SOC, but instead it is Ki67 P0 P3 P5 P7 FoxA2 Tg.Col10 mcherry abcd prst uvwx y s P14 e FoxA3 FoxA1 fgij Immunohistochemistry FoxA2 / Hoescht h klmn o q Fig. 1 FoxA1-3 expression domain predicts the formation of the SOC. Immunohistochemistry for FoxA2 (a–e), FoxA3 (f–j), FoxA1 (k–o), Ki67 (u–y), and fluorescence microscopy on Tg.col10a1mcherry mice (p–t)on postnatal day P0, P3, P5, P7, P14 tibia sections. FoxA1-3 (yellow), Ki67 (white), Hoechst (blue), mcherry fluorescence (red). Scale bars, 100 µm. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 2NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications restricted to the GP hypertrophic cartilage and a small discrete domain at the end of the long bones, as shown by immunohistochemistry for FoxA2 (Fig. 1a–e). Thus, the presence of FoxA2+progeny in the epiphyseal SOC it is likely the result of FoxA2+hypertrophic chondrocytes contributing to the osteogenic lineage. To confirm this, we administered 2x tamoxifen injections to FoxA2Cre.ERT/+;Tomatof/f;Tg.col1GFP mice at P3-4, P7-8, or P13-14 and we followed them 3 months later, demonstrating that Tomato+progeny (derived from FoxA2+cells) are col.1+expressing bone cells (Fig. 2B(a–c)). By P13, during SOC enlargement, we see a significant build-up of ZsGreen+cells in the GP cartilage. In FoxA2CreERT2/+;ZsGreenfl/+; Tg.col10mcherry mice injected with tamoxifen P13-14 and harvested at P18, we counted 126 ± 40 cells/hindlimb. Five tamoxifen injections, administered P13-17 to FoxA2CreERT2/+;ZsGreenfl/+;Tg.col10 mcherry mice, led to a further increase in ZsGreen+cells (387 ± 109 cells/ hindllimb) (Fig. 2A(c, d, e)). As FoxA2+cells are located on the border between GP and SOC, we sought to delineate the cartilage/bone interface by counterstaining with 5-DTAF(4,6-Dichlorotriazinyl Aminofluorescein) dye, which labels efficiently the bone matrix but not the cartilage matrix in AggrecanCreERT/+;Tomatofl/+mice (Supplementary Fig. 1A(a)). A P13-17 tamoxifen pulse administered to FoxA2CreERT2/+; Tomatofl/+mice, revealed that 75% of FoxA2+cells are in the GP cartilage, outside the 5-DTAF labeled bone domain, and 25% are in the green labeled SOC (Supplementary Fig. 1A(b, c)). This FoxA2+population located at the top of the GP RZ, at the cartilage/bone interface, it is not only present in mice, but in rabbits as well (Supplementary Fig. 1B). To investigate whether the FoxA2+cells present in the RZ in early postnatal development (P0-P14) still persist in the GP cartilage 3 months later, we performed a long-term pulse chase of the FoxA2CreERT/+;Tomatofl/+mice. Tamoxifen pulses at P3-4 (during SOC prefiguration), at P7-8 (SOC hypertrophy), and P13-14 (SOC enlargement) to FoxA2CreERT/+;Tomatofl/+mice gave rise to Tomato+labeled cells which, 3 months later, were still present in the RZ either as single cells, clusters or columns of progeny (Fig. 2B(d–f)), or became col.1+bone cells in the SOC (Fig. 2B(a–c)). Altogether, these experiments suggest that a FoxA2+population, left behind from epiphyseal cartilage A a*** *** Tam P3-4 P7-8 P13-14 P13-17 Quantification of GP RZ FoxA2+ cells ( ) v. SOC FoxA2+ cells ( ) using FoxA2Cre.ERT/+; ZsGreen f/+Tg.col10mcherry mice a1 a1 b1 b1 c1 c1 d1 d1 FoxA2+ cells from GP FoxA2+ cells from SOC bcd e * Tam Harvest P3-4 P18 Tam Harvest P7-8 P18 Tam Harvest P13-14 P18 Tam Harvest P13-17 P18 # FoxA2+ cells Tam Harvest P3-4 P90 Tam P7-8 P90 Harvest P13-14 P90 BDual contribution of FoxA2+ cells to osteogenic and chondrogenic lineages FoxA2Cre.ERT/+Tomatof/f Tg.Col1GFP mice FoxA2Cre.ERT/+Tomatof/f 5-DTAF mice Harvest P3-4 P90 Harvest P7-8 P90 Harvest P13-14 P90 abcdef 1 a1 2 a2 b1 b2 211 2 c1 c2 d1 d2 2 1 e1 e2 1 2 f1 f2 2 1 0 200 400 600 800 1000 *** *** Tam Tam Tam Tam Harvest Fig. 2 A population of FoxA2+col10−cells prefigures SOC formation and gives rise to the top compartment of the GP RZ. A Tibia sections harvested from FoxA2CreERT2/+;ZsGreenfl/+; Tg.Col10mcherry mice treated with 2x tamoxifen (Tam) injections at postnatal days P3-4 (a), P7-8 (b), P13-14 (c), or 5x tamoxifen injections P13-P17(d), and harvested at P18. Hoechst (blue), ZsGreen fluorescence microscopy (green), mcherry fluorescence microscopy (red). Scale bars, 100 µm. Representative details, from the interface between GP and SOC, are shown in numbered insets (a1–d1). Quantification of the number of FoxA2+(ZsGreen+) cells in GP is achieved by counting ZsGreen+cells located in GP (extending 100 µm away from the GP/SOC interface, towards the GP). The number of ZsGreen+cells represents a sum of n=8 sections per mouse hindlimb. Similar quantification is performed for FoxA2+cells in SOC. Data presented as mean ± SD, n=3 mice. The asterisks indicate significant difference: One-way ANOVA, Tukey test; ***p=0.000003 (P3-4 GP vs P3-4 SOC), 0.000005 (P7-8 GP vs P7-8 SOC), 0.000569 (P3-4 GP vs P13-17 GP), 0.000828 (P7-8 GP vs P13-17 GP), and *p=0.011031(P13-17 GP vs P13-17 SOC). Complete statistical information is provided in Supplementary Table 1 (e). BTibia sections from FoxA2CreERT2/+;Tomatofl/+;Tg.Col1GFP mice treated with 2x tamoxifen injections P3-P4 (a), P7-P8 (b), P13-P14 (c), and harvested at P90. Tibia sections from FoxA2CreERT2/+;Tomatofl/+mice treated with 2x tamoxifen injections P3-P4 (d), P7-P8 (e), P13-P14 (f), harvested at P90. Tomato (red), Hoechst (blue), GFP (green) (a–c), 5-DTAF (green) (d–f). Scale bars, 100 µm. Representative details are shown in numbered insets (a1–f1). NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 ARTICLE NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications 3 postnatal development (during SOC prefiguration, hypertrophy and enlargement), contributes to the GP RZ top compartment. FoxA2+long-term stem cells v. PTHrP+short-term stem cells. Given that FoxA2+cells reside in the RZ, a known location for stem cells5,11, and that FoxA2+cells are located in close proximity to the SOC, which is viewed as a signaling center of RZ stemness10, we asked whether FoxA2+cells are different than the previously characterized PTHrP+stem cell subset11. Using a PTHrPmcherry mouse line, Mizuhashi et al. have recently shown that PTHrP+cells are restricted to the bottom of the RZ, and they are dedicated, at least to some degree, to giving rise to columns of progeny11. Here we show that FoxA2+cells are located at the top of the RZ and they are geographically separated from the previously characterized PTHrP+cells11, in both tibia and ulna of FoxA2CreERT/+;ZsGreenfl/+;PTHrPmcherry mice (Fig. 3A(a, b)). To confirm our histological findings, we performed FACS analysis on cells isolated from GP tissue of FoxA2CreERT/+;ZsGreenfl/+;PTHrPmcherry mice injected with tamoxifen P13 to P17, and sacrificed at P18. We sorted the FoxA2+(ZsGreen+) cells, the PTHrP+(mCherry+) cells and the double positive (DP) cells. Consistent with the histological data, we observed no overlap between FoxA2+cells and PTHrP+cells (Fig. 3A(d)). The percentage of double positive (DP) cells (0.017 ± 0.004%) is the same as intrinsic (background) fluorescence of double negative (DN) control cells (0.014%). To investigate whether FoxA2+cells can become PTHrP+cells over time, we pulse-chased FoxA2CreERT/+;ZsGreenfl/+;PTHrPmcherry mice, by injecting them with tamoxifen P13 to P17, and sacrificing them at P40 (Fig. 3A(c, e)). Histological data reveals the presence of yellow FoxA2+PTHrP+cells, while FACS analysis confirms an 0.1% of double positive (DP) cells (Fig. 3A(c, e)), confirming that FoxA2+cells can become PTHrP+cells over time. To inquire about FoxA2+cells self-renewability and multipotency, we used triple transgenic mice FoxA2CreERT2/+;ZsGreenfl/+; Tg.col10mCherry, treated with tamoxifen P13-P17 and harvested at P18, that allowed either green labeling of RZ FoxA2+col10−cells, or yellow labeling of HZ FoxA2+col10+cells (Supplementary Fig. 2A). To compare FoxA2+cells with PTHrP+cells, we isolated PTHrP+ cells from P18 GP tissue of PTHrPmcherry mice. Both FoxA2+cells and PTHrP+cells isolated from GP cartilage have high levels of cartilage specificgenes(collagen type 2,aggrecan) and low levels of bone markers (collagen type 1, alkaline phosphatase) when compared with bone marrow mesenchymal cells (BMSC) (Supplementary Fig. 2B). In a cfu (colony forming unit) assay, RZ FoxA2+col.10-(green) cells formed distinct colonies (Supplementary Fig. 3a–c), while HZ FoxA2+col.10+(yellow) cells failed (Supplementary Fig. 3d–f). This indicates that FoxA2+cells, from RZ not HZ, have the capacity to form clones when cultured ex vivo. Next we compared the self-renewability of FoxA2+cells (isolated from FoxA2CreERT2/+;ZsGreenfl/+;Tg.col10mCherry mice), with that of PTHrP+cells (isolated from PTHrPmcherry mice) (Fig. 3B, Supplementary Fig. 4A). At plating, PTHrP+cells gave rise to 40% more colonies than FoxA2+cells (51 vs. 37), but only 11% (6/51) PTHrP+primary colonies could form secondary colonies, whereas 38% (14/37) FoxA2+primary colonies could be further passaged (Fig. 3B, Supplementary Fig. 4B). FoxA2+colonies have higher clonogenicity and longevity than colonies established from PTHrP+cells. About 9% (10/112) FoxA2+colonies reach Passage 9 and beyond, whereas only 1.4% (2/143) PTHrP+ colonies could reach Passage 5 (Fig. 3B, Supplementary Fig. 4B). To inquire whether FoxA2+clones are multipotent, individual FoxA2+clones, from Passage 1 and 9, were further expanded in vitro, and generated Alcian-blue matrix, Alizarin red mineralized matrix and Oil Red droplets during culture in chondrogenic, osteogenic, or adipogenic differentiation media (4/4 clones) (Fig. 3C). Altogether, these experiments demonstrate that, in vitro, FoxA2+cells have multipotency and higher self-renewability and longevity than PTHrP+cells. Dual osteo-chondro-potential for FoxA2+cells prior to P28. To investigate the dynamics of FoxA2+cells clonality in vivo, we aimed to label sufficient FoxA2+cells to demonstrate significant contribution to the GP tissue, but sparse enough to allow assessment of single clones (cells, clusters and columns of progeny) over time. Two tamoxifen injections, administered P14-P15 to FoxA2CreERT/+;Tomatofl/+mice, labeled 7.92 ± 0.9 FoxA2+ cells, which is 8% from the total number of FoxA2+cells detected by immunohistochemistry at P14 (97.67 ± 9 cells) (Supplementary Fig. 5A(a, b, e)). Five tamoxifen injections (administered P14-P18 to FoxA2CreERT/+;Tomatofl/+mice) labeled 22.6 ± 4 FoxA2+cells, increasing the labeling efficiency relative to the number of FoxA2+cells identified via immunohistochemistry, from 8 to 23% (Supplementary Fig. 5A(a, c, e)). Five tamoxifen injections (administered P14-P18 to FoxA2CreERT2/+;Tomatoflfl mice), labeled 36.9 ± 2.3 FoxA2+cells, increasing the efficiency of labeling from 23% (one Tomato floxed allele) to 38% (two Tomato floxed alleles) (Supplementary Fig. 5A(a, d, e)). This increase in labeling efficiency was reflected in a 2-fold increase in the number of columns in FoxA2CreERT/+;Tomatoflfl mice as compared with FoxA2CreERT/+;Tomatofl/+mice harvested 9 months after the last tamoxifen injection (Supplementary Fig. 5B). Overall, increased labeling of the FoxA2+cells demonstrated increased contribution of the FoxA2+cells to the GP cartilage, but it also increased density of the Tomato+cells hindering single clone analysis. As such, in lineage tracing experiments, we labeled a moderate amount of FoxA2+cells by giving 5x tamoxifen injections (P14 to P18) to FoxA2CreERT/+; Tomatofl/+mice (with one Tomato floxed allele) (Fig. 4A). When FoxA2CreERT2/+;Tomatofl/+mice were pulsed with tamoxifen P14-P18 and chased for 1 day or 1, 3, 9 months we observed increasing numbers of doublets, clusters, and columns of progeny (Fig. 4A). Although the initial percentage of columns is 1% at 1 month, there was substantial increase in column formation over time, and by 9 months, the columns of progeny account for 26% of the total combined structures (single cells, clusters, and columns) (Fig. 4C1, Supplementary Fig. 6C). This is in contrast with PTHrP+cells, which gave rise to a large number of columns initially, but the number decreased significantly over time11. FoxA2+cells behave as LTSSC, first giving rise to progeny, which then produce columns of chondrocytes. Consequently, the percentage of FoxA2+clusters peaks early (43% by 1 month), but drops to 31% by 9 months, when FoxA2+column formation is high (Fig. 4C1). Lastly, despite forming clusters and columns, a significant number of FoxA2+cells remains as single cells. The percentage of single cells decreases rapidly over time, from 82% (1 day), to 55% (1 month), but there is still a high fraction of single cells as late as 9 months after labeling (Fig. 4C1). Altogether, these findings suggest that FoxA2+cells constitute a more quiescent, LTSSC population located near the SOC, which gives rise to an increasing number of columns of progeny over time. Next, we compared the clonality of FoxA2+cells, labeled during SOC enlargement (P14-P18), with the clonality of FoxA2+cells, labeled during SOC osteogenic maturation (P28P37). Quantification of the percentage of FoxA2+clones (single cells, clusters or columns) to the total number of Tomato+units, shows a similar dynamic. Both FoxA2+cells labeled at a younger age (tamoxifen P14-18), or at an older age (tamoxifen P28-37), ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 4NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications exhibit the same LTSSC characteristics in vivo, as shown by an increased percentage of FoxA2+columns over time (Fig. 4A–C). However, quantification of the number of FoxA2+clones highlights higher heterogeneity in the mice labeled at a younger age. A high number (138 FoxA2+clones/hindlimb) are initially labeled in FoxA2CreERT/+;Tomatofl/+mice (treated with tamoxifen P14-P18), but this number decreases significantly (to 41 FoxA2+clones/hindlimb) by 3 months after the initial labeling (Fig. 4(C3), Supplementary Fig. 6C). In contrast, in FoxA2CreERT/+; Tomatofl/+mice (treated with tamoxifen P28-P37) the number of FoxA2+clones is not significantly changed between 1-day (55 FoxA2+clones/hindlimb) and 3 months (57 FoxA2+clones/ hindlimb) after the initial labeling (Fig. 4C4, Supplementary Fig. 6C). These results raise the possibility that FoxA2+cells labeled after P28 remain mostly in GP, and their contribution to the SOC slows down significantly over time. 0 20 40 60 Colonies from FoxA2+ (Zsgreen+) stem cells Colonies from PTHRP+ (mcherry+) stem cells cfu assay for FoxA2+ cells vs PTHrP+ cells B FoxA2 Cre.ERT/+ ZsGreenf/+ PTHrP mcherry mice PTHrP+ FoxA2+ a Proximal tibia a Distal ulna PTHrP+ FoxA2+ ab a1 b1 a1 b1 A P18 Tam Harvest P13-17 a Distal ulna P40 Tam Harvest P13-17 PTHrP+ FoxA2+ c1 c1 FoxA2+ PTHrP+ Passage 1 Passage 9 Growth media Adipogenic media Osteogenic media Chondrogenic media Trilineage differentiation assay for FoxA2+ cells abc d efgh C c d. Harvest at P18 P1 P2 P3 P4 P5 P6 P7 P8 P9 Passage number Colonies / 1000 cells 10 102 103104 105 106 102 103104 105 106 ZsGreen mCherry 102 103104 105 106 mCherry 10 10 10 102 103104 105 106 ZsGreen e. Harvest at P40 1.579% 0.015% 0.681% 2.445% 0.108% 0.396% Fig. 3 FoxA2+cells are long-term stem cells, and they are distinct from the short-term PTHrP+stem cells. A Tibia (a) and ulna (b)fromFoxA2CreERT2/+; ZsGreenfl/+;PTHrPmcherry mice treated with tamoxifen (Tam) P13-P17 and harvested at P18, and from FoxA2CreERT2/+;ZsGreenfl/+;PTHrPmcherry mice treated with tamoxifen P13-P17 and harvested at P40 (c). Florescence microscopy for ZsGreen (green), mcherry (red), Hoechst (blue). Scale bars, 50 µm. Separation of FoxA2+(green) cells from PTHrP+(red) cells and double positive FoxA2+PTHrP+(yellow) cells by Fluorescence-activated cell sorting (FACS) of GP cells isolated from FoxA2CreERT2/+;ZsGreenfl/+;PTHrP mcherry mice, injected with tamoxifen from P13 to P17, and harvested at P18 (d)orP40(e). BColony forming unit (cfu) assay for FoxA2+(ZsGreen+) cells (from FoxA2CreERT2/+;ZsGreenfl/+;col.10mcherry mice treated with tamoxifen P13-P17 and harvested at P18) and PTHrP+ (mcherry+)cells (from PTHrPmcherry mice harvested at P18). Quantification of the number of colonies (per 103cells) presented as mean ± SD from n=3 experiments. CTrilineage differentiation of FoxA2+(ZsGreen+) clones, from passage P1 (a–d) and passage P9 (e–h). Four independent clones, for each passage, were tested in either growth media (a,e), or adipogenic (b,f), osteogenic (c,g) and chondrogenic (d,h) media. Scale bars, 100 µm. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 ARTICLE NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications 5 Indeed, for FoxA2CreERT2/+;Tomatofl/+mice treated with tamoxifen P14 –P18, and harvested at P19, a 1-day chase reveals that 137 cells of the FoxA2+cells are in the GP and 43 in the SOC (Supplementary Fig. 6A1). A 1-month chase increases the SOC population, from 43 to 253, suggesting that FoxA2+cells can contribute to the epiphyseal bone, during SOC osteogenic maturation (Supplementary Fig. 6A1). This contribution is no longer endochondral, as RZ FoxA2+cells no longer express col.10 by P21 (Supplementary Fig. 6B). Instead, FoxA2+cells are osteo-progenitors that support directly the osteogenic lineage. This is consistent with our previous findings demonstrating that FoxA2+cells labeled during early postnatal development gave rise to col.1+progeny in the SOC (Fig. 2B). In addition, as previously shown, the FoxA2+cells labeled P14 to P18 give rise to clusters and increasing columns of progeny over time, suggesting chondro-progenitor quality (Fig. 4A). Altogether, these findings highlight dual osteo-chondro-progenitor fate for FoxA2+cells labeled prior to P28. In contrast, FoxA2+cells labeled after P28 remain mostly in GP, and their contribution to the SOC slows down (Supplementary Fig. 6A2). For older mice, 1-day chase reveals that 55 FoxA2+cells are in the GP and 10 FoxA2+cells in the SOC above (Supplementary Fig. 6A2). A 1-month chase shows a similar number of FoxA2+cells (23) in the SOC, suggesting that Harvest: 1day ab dgjc efhi kl Lineage tracing FoxA2Cre.ERT/+Tomatof/+ mice, Tamoxifen starting P28 Quantification Tomato+ cells, clusters, columns Harvest: 1 month Harvest: 3 months Harvest: 9 months j g d ba B1 B2 B3 B4 B k ab b a c cef d ig gi fed h A1 A2 A3 jkl lj k A4 Harvest: 1day Lineage tracing FoxA2Cre.ERT/+ Tomatof/+ mice, Tamoxifen at P14 Harvest: 1 month Harvest: 3 months Harvest: 9 months A C FoxA2Cre.ERT2; Tomatof/+ Tamoxifen at P14 FoxA2Cre.ERT2; Tomatof/+ Tamoxifen at P28 # FoxA2+clones / HL # FoxA2+clones / HL C3 Number of FoxA2+ clones single cells clusters columns 0 40 80 120 160 1d 1mo 3mo 9mo C4 Percentage (%) of FoxA2+ clones FoxA2Cre.ERT2; Tomatof/+ Tamoxifen at P14 FoxA2Cre.ERT2; Tomatof/+ Tamoxifen at P28 % FoxA2+clones / HL % FoxA2+clones / HL C1 single cells clusters columns C2 0 40 80 120 1d 1mo 3mo 9mo 0 40 80 120 1d 1mo 3mo 9mo 0 40 80 120 160 1d 1mo 3mo 9mo Tam Harvest P14-18 P19 Tam Harvest P14-18 P48 Tam Harvest P14-18 P108 Tam Harvest P14-18 P288 Tam Harvest P28-37 P38 Tam Harvest P28-37 P67 Tam Harvest P28-37 P127 Tam Harvest P28-37 P307 Fig. 4 Dual osteo-chondro-potential for FoxA2+cells in the early weeks of postnatal development (P14) and stronger chondrogenic potential beyond P28. A Tibia sections harvested from FoxA2CreERT2/+;Tomatofl/+mice treated with 5x tamoxifen (Tam) injections starting at P14, and harvested at 1 day (A1), 1 month (A2), 3 months (A3), and 9 months (A4) after the last injection. Hoechst (gray), 5-DTAF (green). Bars, 100 µm. Representative details from GP sections, in numbered insets (a–l). BTibia sections harvested from FoxA2CreERT2/+;Tomatofl/+mice treated with 10x tamoxifen injections starting at P28, and harvested at 1 day (B1), 1 month (B2), 3 months (B3), and 9 months (B4) after the last injection. Hoechst (gray), 5-DTAF (green). Bars, 100 µm. Representative details from GP sections, in numbered insets (a–l). CQuantification of the number (C3,C4) and the percentage (C1,C2) of single cells, clusters, columns formed at 1 day, and 1, 3, 9 months after the last injection. Percentage (%) single cells is calculated by the number of single Tomato+ cells in the GP (outside the green 5-DTAF domain) to the total number of Tomato+units (single cells, clusters and columns). Each point represents the sum of n=8 sections per mouse hindlimb. Similar quantification of percentage (%) clusters and columns. One Tomato+unit, is defined by 1 single cell, or 1 cluster comprised of two or more cells grouped together, or by 1 column of cells. ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 6NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications FoxA2+cells contribution to the epiphyseal bone slows down significantly after P28 (Supplementary Fig. 6A2). Self-renewability of FoxA2+cells in a reconstitution assay.To demonstrate self-renewal and multipotent capabilities of FoxA2+ cells in vivo, we performed a serial transplantation assay. We isolated FoxA2+(ZsGreen+) cells from FoxA2CreERT2/+; ZsGreenfl/+;Tg.col10mCherry mice, as previously described, labeled them with DiD fluorescent dye, and transplanted them subcutaneously into ZsGreenlittermates (Fig. 5A, Supplementary Fig. 7A). After 1 month, FoxA2+(ZsGreen+) cells from the grafts exhibited self-renewability and retained colony-forming ability (Fig. 5B(a)). In addition, the grafted cells have undergone multiple rounds of cell division shown by a log-fold reduced fluorescence, observed as a shift in the intensity histogram plots (Fig. 5B(c)). Peak DiD intensity at day 0 (D0), immediately before transplantation, was 1.2 × 104, and it was reduced to 1.1 × 103on day 30 (D30) post-transplantion (Fig. 5B(c)). The percentage of ZsGreen+cells with high intensity DiD fluorescence (DiDhigh) Serial transplantation of FoxA2+ cells in vivo A Primary Transplant a1 a2 a3 1 2 3 1 23 b2 b3 b1 Isolation and assessment of self-renewability potential of FoxA2+ cells B ab Secondary Transplant FACS separation Label with DiD First Transplant Explant after 1mo. In vitro colony formation and expansion Label again with DiD dye Second Transplant In vitro ZsGr+colony formation (after first transplant) Chondrocyte isolation Take cells for DiD analysis by FACS (Day 30 first transplant) Take cells for DiD analysis by FACS (Day 0 second transplant) Explant after 1 mo. Take cells for DiD analysis by FACS (Day 30 second transplant) In vitro ZsGr+colony formation (after second transplant) Transplantation of primed progenitors 105 cells Take cells for DiD analysis by FACS (Day 0 first transplant) ZsGr–littermate Secondary recipient ZsGr –littermate Tertiary recipient FoxA2+Col10Histological analysis of explants at 1mo. Lineage priming FoxA2Cre.ERT/+ ZsGreenf/+ Tg.col10mcherry mice Primary host Chondrogenic Adipogenic Osteogenic *** 0 20 40 60 80 100 *** DiDhigh D0 D30 % ZsGreen+ cells 0 20 40 60 80 100 DiDlow D0 D30 % ZsGreen+ cells 0 20 40 60 80 100 0 20 40 60 80 100 *** *** D0 D30 % ZsGreen+ cells D0 D30 ZsGreen+ cells count DiDhigh DiDlow def g h ZsGreen+ cells count c 100 50 0 100 50 0 100 50 0 100 50 0 DiD intensity 101 103105107 DiD intensity 101 103105107 100 50 0 % ZsGreen+ cells 100 50 0 Fig. 5 Serial transplantation of FoxA2+cells in vivo. A Experimental strategy for the isolation and assessment of self-renewability and differentiation potential of FoxA2+cells in vivo. BRepresentative images of colonies derived from FoxA2+cells isolated from primary (a) and secondary transplants (b). Scale bars, 1 cm. DiD fluorescence intensity histogram profiles at Day 0 (red), Day 30 (blue), and negative control (black) from primary (c) and secondary (f) transplants. Percentage of DiDhigh cells in the FoxA2+cells from primary (d) and secondary (g) transplants. Percentage of DiDlow cells in the FoxA2+ cells from primary (e) and secondary (h) transplants. Data are presented as mean ± SD for each group, n=3 mice. The asterisks indicate significant difference: two-tailed Student’s unpaired samples ttest; ***p=7.0148E−07 (DiDhigh primary transplant D0 vs D30), 7.84512E−06 (DiDlow primary transplant D0 vs D30), 3.32176E−06 (DiDhigh secondary transplant D0 vs D30), and 3.46695E−06 (DiDlow secondary transplant D0 vs D30). Complete statistical information is provided in Supplementary Table 1. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 ARTICLE NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications 7 was 75 ± 1% prior to transplantation (D0), but it was reduced to 27.3 ± 1.1% in the cells derived from grafts at D30 posttransplantation (Fig. 5B(d)). Simultaneously, the percentage of cells exhibiting low intensity DiD fluorescence (DiDlow) was significantly increased from 20.3 ± 0.6% at D0 to 42.3 ± 1.1% at D30 (Fig. 5B(e)). This suggest that FoxA2+(ZsGreen+) cells proliferate in the primary transplants and dilute the DiD fluorescent dye. Next, FoxA2+(ZsGreen+)cells from the primary graft were stained with DiD dye and transplanted into the tertiary recipient mice (Fig. 5A, Supplementary Fig. 7A). A similar reduction in DiDhigh cell population (from 75.3 ± 1.5% on D0 to 29.7 ± 1.5% on D30) and a proportionate increase in DiDlow cell population (from 19.7 ± 0.6% on D0 to 46.7 ± 1.1% on D30) were consistent with the properties observed with cells derived from primary grafts (Fig. 5B(f, g, h)). In addition, FoxA2+(ZsGreen+) cells from the secondary grafts retain colony forming ability (Fig. 5B(b)). The cells collected at the end of serial transplantation exhibited multi-lineage differentiation capabilities, as indicated by positive Alcian blue, LipidTox or Alizarin Red-S staining in cross-sections of tertiary grafts (Supplementary Fig. 7B(a–c)). These tertiary grafts were produced by transplantation of lineage-primed (chondrogenic, adipogenic, osteogenic) second graft-derived FoxA2+(ZsGreen+) cells (Fig. 5A, Supplementary Fig. 7B(a–c)). In contrast, second graft-derived FoxA2+(ZsGreen+) cells, without lineage priming, produced tertiary grafts that were negative for cell differentiation (Fig. 5A, Supplementary Fig. 7B(d–f)). Altogether, these experiments indicate that FoxA2+cells preserve both their self-renewing ability and their multipotency during in vivo serial transplantation. GP cartilage regeneration after SH1-like injury. To investigate whether FoxA2+cells expand and repair the GP cartilage in response to trauma, we developed a murine model of GP injury. Salter–Harris type1 fractures (SH1), which involve separation across the entire physis, through the HZ, have the best prognosis for repair and are the least likely to cause growth arrest3,4. We reproduced this clinical observation and developed a SH1-like surgical model, to test if FoxA2+cells could repair the GP cartilage. A 30 g needle is passed in the transverse plane, through the HZ of the proximal tibial physis, creating a discrete defect, visualized by Contrast-Enhanced Computed Tomography (CECT)17,18 (Fig. 6A, B). Quantitative analysis of sequential sagittal planeimagesestimatesthatthedefect tracks, on average, 2216 ± 195.6 μm medial to lateral, 280 ± 60 μm anterior to posterior, 165.6 ± 36.9 μm height, and 28.6° ± 5.5 angle relative to mid-coronal plane (Supplementary Fig. 8A, B). Color maps of representative CECT images of control and operated tibias reveal that injury site heals within 7 days, irrespective of mouse gender (Fig. 6C). The injury gap volume decreases significantly from 0.29 ± 0.07 mm3at 1 day, to 0.04 ± 0.04 mm3, close to background level, at 7-day post-op (Fig.6D).Overalllongitudinalgrowthwasnot affected at 7and 21-day post-op (Supplementary Fig. 8C). At 1-day post-op, the injured GP displays significant damage in the HZ (Fig. 6E(a, b)). In addition, given that the central region is more curvilinear, with a shorter GP in the middle, the injury causes more damage in this area, destroying the HZ, the PZ, and most of the RZ. At 1-day post-op, the healthy cartilage area is 42% of the total measured area (healthy plus injured) (Fig. 6F). Within days, the defect shows healing and the cartilage is regenerating. By the 3rd day, the central gap is closing in, and the healthy cartilage area is up to 63% ((Fig. 6E(c, d), F). By the 7th day, the GP is 96% regenerated, displaying all three layers (RZ, PZ, HZ) ((Fig. 6E(e, f), F). The injured GP healed with physeal cartilage, not fibrocartilage or bone. FoxA2+cells expand and regenerate the GP after SH1-like injury. After SH1-like surgery on FoxA2CreERT/+;Tomatofl/+ mice, there was substantial expansion of Tomato+cells in the injured GP (Fig. 7A, B, Supplementary Fig. 9). We quantified the number of FoxA2+cells in GP cartilage as well as in neighboring metaphyseal bone (MB) and epiphyseal bone (SOC) (Fig. 7B). At 1-day post-op, there was no difference between GP FoxA2+cells in control versus operated mice (Fig. 7A(a, d), Fig. 7B(b1)). However, there is a significant loss (94%) of FoxA2+cells at the HZ/MB border, which continues to be sustained 3-day post-op (Fig. 7B(b3)). This indicates that there is no expansion of the MB Tomato+cells in response to injury. Similarly, there is no expansion of the Tomato+cells in the SOC at 1 day or 3 days after SH1-like surgery (Fig. 7B(b2)). However, there is a 2.7-fold expansion of Tomato+cells in the GP 3 days after the SH1-like surgery (Fig. 7B(b1)). These findings are supported by a proliferation assay, which demonstrates that Tomato+cells are BrdU+(Supplementary Fig. 10A). This expansion of FoxA2+cells at 3 days post-op is the result of FoxA2+cells labeled prior to surgery who are multiplying in response to injury, and not fresh labeling of FoxA2+cells due to de novo expression of FoxA2 in the injured cartilage remnants. Immunohistochemistry for FoxA2 in the injured GP cartilage shows the expression of FoxA2 in its established RZ domain, but not in the cartilage remnants at d1, d2 or d3 post-op (a–c) (Supplementary Fig. 10B). By 7-day post-op, FoxA2+(Tomato+) cells are significantly reduced in the GP tissue, and in parallel, we see a surge in Tomato+cells in the MB (3-fold increase) (Fig. 7B(b3)). This suggests that the Tomato+cells, that have expanded in the GP tissue at 3-day post-op, transited below in the bone spongiosa by 7-day post-op (Fig. 7A(f2)). In the center, where the wide gap was now reduced to a thin slit, the last Tomato+cells are migrating down, aligned in short (4–8 cell) stacks of clones (Fig. 7A(f1)). FoxA2+cells are needed for cartilage regeneration. To determine whether FoxA2+cells participate in regenerating the injured GP tissue via multi-lineage differentiation, we asked if FoxA2+cells differentiate into col.10+cells (chondrogenic lineage) or col.1+cells (osteogenic lineage). SH1-like surgery on FoxA2Cre.ERT/+;ZsGreenf/+; Tg.col10mcherry mice demonstrates that FoxA2+(ZsGreen+)cells generate yellow hypertrophic cells (col.10mcherry+)(Fig.8A(b)), while SH1-like surgery on FoxA2Cre.ERT/+;Tomatof/f;Tg.col1GFP mice shows that FoxA2+(Tomato+) cells transit in the metaphyseal bone to become yellow bone cells (col.1GFP+)(Fig.8A(a)). To inquire whether FoxA2+cells are needed for cartilage regeneration, we generated FoxA2CreERT/+;DTAf/f mice, in which FoxA2+cells were ablated by production of diphtheria toxin (DTA). In FoxA2CreERT2/+;DTAf/+;Tomatofl/+mice there was a 36% decrease in FoxA2+(Tomato+) cells compared with FoxA2CreERT2/+;Tomatofl/+littermates (Fig. 8B). Loss of FoxA2+ cells lead to a delay in cartilage closure in FoxA2CreERT2/+;DTAf/f mice 7-day post-op (Fig. 8C(b1, d1)). Cartilage regeneration in FoxA2CreERT2/+;DTAf/f mice was significantly reduced (down to 72%) compared with control mice (Fig. 8C, D), suggesting that FoxA2+cells are necessary for GP repair. Summary In summary, we demonstrate that a population of FoxA2+col10cells, residual from epiphyseal cartilage postnatal development, contributes to the top sub-compartment of the RZ (Fig. 9). Based on their location, FoxA2+cells are physically separated from the PTHrP+cells, which are located at the bottom of the RZ, and exhibit higher clonogenicity and longevity than PTHrP+cells. Short-term pulse-chase of the FoxA2+cells generates dyads/clusters ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 8NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications of cells, whereas long-term pulse-chase generates increasing columns of proliferating progeny, suggesting that FoxA2+cells may first give rise to progeny that subsequently differentiate into columns of chondrocytes. Prior to P28, FoxA2+cells have a dual osteo-chondro-progenitor fate, and after P28, FoxA2+LTSSC remain mostly in GP, while their contribution to the SOC winds down significantly. In a murine model of GP injury, FoxA2+cells expand in response to trauma and participate in production of physeal cartilage allowing for successful regeneration. In conclusion, FoxA2+cells are a LTSSC population necessary for both GP turnover and cartilage regeneration following injury. Discussion While POC (primary ossification centers) of the long bones form at E15.5, during mouse embryonic development, the SOCs appear sometime later, primarily after birth, around postnatal day P719,20. Chondrocyte hypertrophy (P7-P8) is followed by RZ PZ HZ SH1 – like GP injury model CECT imaging of the injury with CA4+ AB Control + CA4+Surgery + CA4+ Control - CA4+ 1a 2a 3a 1b 2b 3b CECT imaging of GPC regeneration . C ** Injury gap volume (mm3) P=0.055 d1 d3 d7 d1 d3 d7 tsopsyad-op Surgery Control Quantification cartilage volume D 0.0 0.1 0.2 0.3 0.4 0.5 Control Surgery A B d1 d3 d7 d1 d3 d7 1 2 3 ab d c ef aa. SO/FG in control vs. operated mice tsopsyad-op % GPC regeneration % SO+ area (A) vs total area (A+B) 1 2 3 E * Image J quantification F ** *** 0 20 40 60 80 100 -1000 Houndsfield Units (HU) 5000 100 80 60 40 20 0 Fig. 6 A murine model of GP injury. A Schematics of Salter–Harris type 1 (SH1)-like injury model. BCECT (contrast-enhanced computed tomography) imaging of the SH1-like injury, 1-day after surgery. Representative sagittal (1a,2a,3a) and transverse (1b,2b,3b)μCT slices of the proximal tibia, at 1-day post-surgery, before (1a,1b) and after (2a,2b,3a,3b) exposure to CA4+. Scale bars, 500 µm. CColor maps of representative contrast-enhanced μCT images of control and operated tibias at 1, 3, and 7 days post-operation, viewed from the transverse plane. The color scale corresponds to Hounsfield Units (HU). For reference, air attenuates at −1000 H and water at 0 HU. Scale bars, 500 µm. DQuantification of the injury gap volume using CECT imaging. Data are presented as mean ± SD, n=5 mice for each time point (d1,d3,d7)). The asterisks indicate significant difference: Two-tailed Student’s unpaired samples ttest; **p=0.0079 (d1 vs d7). ESafraninO/Fast Green staining of control and operated tibias, 1 day (a,b), 3 days (c,d), and 7 days (e,f) after SH1-like surgery. Inset, central area of the defect (1–3). Scale bars, 100 µm. FImageJ quantification of the injury. A is healthy cartilage area (SafraninO+), B is injured area (SafraninO−), and A+B is the total area. Percentage (%) cartilage regeneration is A/A+B. Data are presented as mean ± SD, n=3 mice. The asterisks indicate significant difference: Two-tailed Student’s unpaired samples ttest; ***p=0.000839 (d1 vs d7), **p=0.007121 (d3 vs d7), *p=0.037861 (d1 vs d3). Complete statistical information is provided in Supplementary Table 1. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-30247-1 ARTICLE NATURE COMMUNICATIONS | (2022) 13:2515 | https://doi.org/10.1038/s41467-022-30247-1 | www.nature.com/naturecommunications 9