Osteochondroprogenitor cells and neutrophils expressing p21 and senescence markers modulate fracture repair

Abstract

Cells expressing features of senescence, including upregulation of p21 and p16, appear transiently following tissue injury, yet the properties of these cells or how they contrast with age-induced senescent cells remains unclear. Here, we used skeletal injury as a model and identified the rapid appearance following fracture of p21+ cells expressing senescence markers, mainly as osteochondroprogenitors (OCHs) and neutrophils. Targeted genetic clearance of p21+ cells suppressed senescence-associated signatures within the fracture callus and accelerated fracture healing. By contrast, p21+ cell clearance did not alter bone loss due to aging; conversely, p16+ cell clearance, known to alleviate skeletal aging, did not affect fracture healing. Following fracture, p21+ neutrophils were enriched in signaling pathways known to induce paracrine stromal senescence, while p21+ OCHs were highly enriched in senescence-associated secretory phenotype factors known to impair bone formation. Further analysis revealed an injury-specific stem cell-like OCH subset that was p21+ and highly inflammatory, with a similar inflammatory mesenchymal population (fibro-adipogenic progenitors) evident following muscle injury. Thus, intercommunicating senescent-like neutrophils and mesenchymal progenitor cells were key regulators of tissue repair in bone and potentially across tissues. Moreover, our findings established contextual roles of p21+ versus p16+ senescent/senescent-like cells that may be leveraged for therapeutic opportunities.

Keywords: Aging; Bone biology; Bone disease; Cellular senescence.

Figure 1. p21⁺ and p16⁺ cells appear in a divergent manner during fracture healing.

(A) t-SNE visualization of clustered cell populations across murine fracture healing by CyTOF. (B) Heatmap representation of identification and (C) senescence and SASP marker median expression across all clusters. (D) t-SNE visualization of nonimmune (CD45^–CD11b^–) callus cells across fracture healing, overlaid with p16⁺ (black) and p21⁺ cells (red). (E) p16⁺ and p21⁺ cell abundances across fracture healing in nonimmune cells. (F) SASP marker median expression throughout fracture healing in all nonimmune cells and (G) in p16⁺ and p21⁺ nonimmune cells. Day 3: n = 12 mice (6 female, 6 male); day 7: n = 12 mice (6 female, 6 male); day 14: n = 12 mice (6 female, 6 male); day 28: n = 11 mice (6 female, 5 male). *P < 0.05; ***P < 0.001 by 1-way ANOVA with Tukey’s correction for multiple comparisons (F) or Mann-Whitney U test (G).

Figure 2. Clearance of p21⁺ cells accelerates fracture healing by increasing bone formation rates and reducing osteoclast numbers.

(A) Schematic of the p21-ATTAC transgene and overall study design. p21-ATTAC mice (4–6 months old) were used to selectively clear p21⁺ cells through AP administration twice weekly over a 5-week fracture healing time course. (B) qRT-PCR measurement of p21^Cip1 and GFP (p21-ATTAC transgene) mRNA expression after AP treatment on day 14. (C) Fracture healing score (described by Wehrle et al.; ref. 21), and (D) callus area as measured by weekly x-rays. (E) μCT of callus bone volume (BV). (F) Tibial stiffness and maximal torque measured by biomechanical testing. (G–I) Histomorphometric analysis of bone formation rate per bone surface (BFR/BS) and mineral apposition rate (MAR) through weekly injections of bone-labeling dyes (see Methods). Scale bars: 50 μm. (J) Histological quantification of osteoblasts through Masson’s trichrome staining. (K) Histological quantification of osteoclasts through tartrate-resistant acid phosphatase (TRAP) staining. Scale bars: 100 μm (J and K). Arrows in J and K indicate osteoblasts and osteoclasts, respectively. (L) Telomere-associated foci (TAF) staining (day 14) for DNA damage. Scale bar: 2 μm. (M) Quantification of cells exhibiting 3 or more TAF per cell. n = 8–11 (B and F–M) or n = 22–25 (C and D) per treatment, equally split by sex. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by Mann-Whitney U test (B, D, E, and J–M), 2-way ANOVA with Šidák’s correction (C, D, and F), or multiple t test with FDR correction (H and I).

Figure 3. Clearance of p21⁺ cells suppresses factors driving osteoclast recruitment and inhibition of bone formation through targeting OCH cells and neutrophils.

(A) Schematic outlining CyTOF analysis of callus cells after p21⁺ cell clearance in p21-ATTAC mice. (B) CITRUS analysis reveals reduced expression (blue dots; FDR < 0.05) of proteins in OCH and Neutrophil cell clusters. (C) CITRUS expression plots for key identification markers. (D) CITRUS results of differential expression between vehicle- and AP-treated groups in OCH (D) and Neutrophil (E) clusters; all FDR < 0.05. (F) t-SNE visualization and FlowSOM clustering of callus cells from p21-ATTAC mice treated with either vehicle (Veh) or AP. (G) Quantification of changes in cell cluster percentages after AP treatment (log₂[fold change]; 2-way ANOVA with Šidák’s multiple-comparison test). (H) Quantification of absolute changes in the Neutrophil-4 cluster after AP treatment (Mann-Whitney U test). n = 16 vehicle-treated mice (8 female, 8 male), n = 13 AP-treated mice (7 female, 6 male). *P < 0.05.

Figure 4. p21⁺ callus cells are largely highly secretory OCH cells and mature neutrophils.

(A) Schematic of p21⁺ cell isolation and scRNA-seq using the p21 reporter mice. (B) GFP⁺ cells were significantly higher in the fractured compared with the unfractured contralateral side and an unfractured mouse tibia. n = 10 fractured (n = 4 female, n = 6 male), n = 8 contralateral sides (n = 4 female, n = 4 male), n = 6 unfractured (n = 3 female, n = 3 male). (C) p21^Cip1, Cxcl2, Vegfa, and Tnfa mRNA expression was significantly enriched in GFP⁺ cells. n = 8 mice (n = 4 male, n = 4 female). (D) scRNA-seq analysis was performed on 5,994 total callus cells from n = 4 mice (n = 2 male, n = 2 female). (E) Differentially upregulated mRNA transcripts in p21⁺ cells. (F) Proportion of p21⁺ cells and (G) SenMayo gene enrichment analysis among clustered cell populations. (H and I) Predicted secretory strength relationships in callus cells among all signaling pathways and (J) TGF-β signaling by CellChat. *P < 0.05; **P < 0.01; ***P < 0.001 by 1-way ANOVA with Tukey’s correction (B) or Mann-Whitney U test (C, GFP⁺ vs. GFP^–).

Figure 5. OCH cells are a proliferative population with a senescent-like phenotype.

(A) Expression of cell proliferation gene set (GO_0008283) among callus cell populations identified by scRNA-seq (see Figure 4). n = 4 mice. (B) Percentage Ki67⁺ and Ki67 mean expression between OCH cell and Neutrophil-4 callus cell clusters identified by CyTOF (see Figure 3). (C) Senescence-related and SASP protein expression between OCH cells and Neutrophil-4 clusters. (D) Gating strategy for p21⁺/Ki67⁺, p21⁺Ki67^–BCL2⁺, and p21⁺Ki67^–BCL-XL⁺ cell populations in fractured mice. (E) Heatmap demonstrating mean expression of senescence-associated proteins in p21⁺ subsets by CyTOF. (F) t-SNE visualization of FlowSOM-clustered callus cells overlaid with p21⁺Ki67⁺, p21⁺Ki67^–BCL-XL⁺, and p21⁺Ki67^–BCL2⁺ cells (red). n = 16 mice (B–F). *P < 0.05, ***P < 0.001, ****P < 0.0001 by unpaired, 2-tailed t test (B and C: p53, p-ATM, PAI-1, TGF-β1, p-STAT1) or Mann-Whitney U test (C: p21, IL-1β).

Figure 6. OCHs expressing SSC markers define an injury-specific senescent-like population.

(A) t-SNE visualization of FlowSOM-clustered callus cell populations collected from fractured WT C57BL/6 mice. (B and C) Heatmap representation of mean protein expression of (B) identity and (C) senescence-associated markers. (D and E) Quantification of senescence-associated proteins among all clustered cell populations. (F) Gating strategy for Lin^–Sox9⁺ cells in CyTOF samples isolated from either unfractured or fractured bones from young (4- to 6-month-old) mice. (G) Quantification of manually gated OCH-Stem clusters in both unfractured and fractured samples. (H) Quantification of senescence-associated proteins in manually gated OCH-Stem (Lin^–Sox9⁺CD51⁺) cell populations compared to all cells in both unfractured and fractured samples. n = 5 fractured, n = 4 unfractured, all female. *P < 0.05; **P < 0.01; ****P < 0.0001 by 2-way ANOVA with Tukey’s multiple-comparison test (D, E, and H) or Mann-Whitney U test (G).

Figure 7. Detrimental effects of p21⁺ cells on bone metabolism are aging independent.

(A) p21⁺ cell clearance was performed in old p21-ATTAC mice treated at 20 months of age for 4 months with either vehicle (Veh) (n = 32 mice: 16 male, 16 female) or AP (n = 32 mice: 16 male, 15 female) twice weekly until sacrifice at 24 months. (B) qRT-PCR measurement of p21^Cip1 mRNA expression. (C–G) Skeletal parameters measured at the femur by μCT: (C) diaphyseal (Dia) cortical thickness (Ct.Th), (D) Dia cortical area (Ct.Ar), (E) metaphyseal (Met) trabecular bone volume per total volume (BV/TV), (F) Met Ct.Th, and (G) Met Ct.Ar. (H) Trabecular BV/TV measured at the L5 lumbar vertebra. (I) Schematic for clearance of p21⁺ cells in old mice undergoing fracture repair; 24-month-old p21-ATTAC mice were used to selectively clear p21⁺ cells through treatment with either vehicle (n = 16 mice; 8 male, 8 female) or AP (n = 14 mice; 7 male, 7 female) twice weekly over a 5-week fracture healing time course. (J) Fracture healing score measured by weekly x-ray. (K) μCT of callus bone volume. (L) Tibial stiffness measured by biomechanical testing (n = 13 Veh: 6 male, 7 female. n = 13 AP: 7 male, 6 female). (M) CyTOF analysis of OCH-Stem population abundances among CD45^–Lin^– nonimmune cells isolated from the digested hind limbs of young (6-month-old) and old (24-month-old) WT C57BL/6 mice (n = 4 mice per group, all female). (N) Schematic of results from injured young bone versus intact aging bone. OCHs, osteochondroprogenitors; OBs, osteoblasts; OCs, osteoclasts; OCYs, osteocytes. Note that this figure is a schematic and only provides a depiction of the cell populations rather than quantitative data. *P < 0.05; **P < 0.01; ***P < 0.001 by Mann-Whitney U test (B and K–M), unpaired, 2-tailed t test (C–H), or 2-way ANOVA with Šidák’s correction (J).