We used an open tibial mid-diaphyseal murine fracture model, which is an established model for endochondral bone healing (3, 23, 26). We and others have published extensive analysis of the healing time course in this model (3, 23, 26), and select immunofluorescence (IF) staining and histological analyses of the healing callus at different post-fracture timepoints are shown in Figures 1A-D ; S1A-D . On d7 post-fracture in this model, the areas within and immediately surrounding the fracture gap were filled with chondrocytes that formed a dense fibrocartilaginous area (soft callus) (3, 26) ( Figures 1A , S1A ). As healing proceeded to days 10 and 14, soft-callus chondrocytes underwent hypertrophic differentiation as indicated by expression of the hypertrophy markers (3, 26) (Collagen 10 [Col X] staining is shown in Figures 1B, C ). The callus areas distal to the fracture gap were composed of newly formed woven bone ( Figures 1A-C ; S1A-C ). As healing further progressed to d21, there was complete replacement of the soft callus with woven bone, resulting in bony bridging of the fracture gap and re-establishment of the intramedullary cavity and bone marrow (BM) elements ( Figures 1D , S1D ).

In a recently published work (26), we comprehensively analysed days 14 and 21 post-fracture of the same fracture model using RNA-seq, reverse transcriptase- quantitative polymerase chain reaction (RT-qPCR), and IF staining, and we showed a strong influx of immune cells to the healing callus between days 14 and 21. In this current study, we expanded our RNA-seq experiments to include days 3, 5, 7, 10, 14, and 21 to provide an overarching insight into the progression of the immune response as well as the formation of osseous elements during the healing process. Principal Component Analysis (PCA) indicated clustering of the timepoints into 3 main groups: days 3, 5, and 7 in the first group, days 10 and 14 in the second group, and d21 in the third group ( Figure S2 ). Analysis of individual days indicated that, as expected, the expression of genes associated with the immune/inflammatory response were readily detectable and highly expressed in the d3 callus ( Tables S1 , S2 ). Progression of the healing response to d5 was accompanied by an increase in the expression of genes associated with cartilage and connective tissue development ( Figures S3A, B ; Tables S1 , S2 ), but the overall pattern of expression of immune-response-associated genes was comparable to that of d3 ( Figures S3A, B ; Tables S1 , S2 ). As healing proceeded beyond this initial inflammatory phase to the cartilaginous callus repair phase (days 7 to 14), there was a marked increase in the expression of genes involved in chondrocyte differentiation, cartilage and bone development, and ossification ( Figures S4-S6 ; Tables S3 - S5 ). Consistently, k-Means clustering indicated clustering of genes involved in cartilage and bone development into a group that shows the peak expression between days 7 and 14 ( Figures 1E “cluster B”, G ; Table S6 ). These data are consistent with our histological analysis and IF staining ( Figures 1A-D , S1A-D ) as well as published data (3, 26). Progression into this cartilaginous callus phase was also accompanied by a marked decrease in the expression of genes associated with immune and inflammatory response, leukocyte migration and activation, and cytokine production ( Figures 1E “cluster A”, F ; S4-S6 ; Tables S3 - S6 ), indicating that the early immune response was partially resolved. Importantly, as we recently reported (26), the expression of genes associated with immune cells and immune response peaked again on d21 ( Figures 1E “cluster A”, F ; S7A, B ; Tables S6 , S7 ), which was further confirmed by analyzing the expression of the general immune-cell marker Cd45 ( Figure S7C ). This biphasic pattern of immune cell infiltration in the callus is interesting; while the early infiltration wave during the initial inflammatory phase is well established, the observed late infiltration phase is uninvestigated. From these data, it remains unclear whether the observed late increase in immune activity was due to re-establishment of BM elements in the d21 callus ( Figures 1D ; S1D ), or due to active recruitment of immune cells to the healing callus.

Our RNA-seq data indicated that immune infiltration peaked twice in the healing callus: during the initial inflammatory phase and on d21 post-fracture ( Figures 1E, F ; S3-7 ; Tables S1 - S7 ). However, the principal pathways identified by the RNA-seq data revealed only a high-level view of changes in the immune and inflammatory responses during the healing process. To investigate the detailed cellular composition of the immune cells within the callus at each peak of immune activation, we analyzed the callus during the initial inflammatory phase and on d21 post-fracture using multiparameter flow cytometry (FC). We chose to analyze day 5 during the initial inflammatory phase because we could isolate 5-10 times more total cells, and thus many more CD45⁺ immune cells, from the d5 callus than the d3 callus. Furthermore, the expression of genes associated with general immune and inflammatory response was comparable on days 3 and 5 ( Figures S3A, B ; Tables S1 , S2 ). We also compared the cells present in the callus at these time points to those in the contralateral unfractured bone (thoroughly washed out to remove BM) or in the BM isolated from the contralateral unfractured bone. In this way we were able to identify changes that were specific to the callus, and to assess whether the immune-cell infiltration we observed on d21 ( Figures 1E, F ; S7 , Tables S1 , S6 , S7 ) was due to re-establishment of the normal BM microenvironment, or to active recruitment of specific immune cells to the callus. Lastly, although the d5 callus consists of an inflammatory hematoma and does not contain bone marrow, we surmised that a comparison of the cellular composition of the d5 callus to that of bone and BM would provide a strong indication of cell types that may play a requisite role in efficient initiation of fracture healing during the initial inflammatory phase.

The overall cellularity of callus vs. bone vs. bone marrow differs widely with the tissue being examined, with the BM being the tissue with the highest cellularity among the three. Therefore, a direct comparison of the cell numbers within individual populations among different tissues is not informative. Rather, we portrayed our results as % of single cells to allow comparison between tissues and to ascertain the proportion and relative enrichment of any individual cell type and, thus, the contribution of individual cell populations to each tissue.

Initially we examined the proportion of myeloid cells within d5 or d21 callus vs. bone and BM using FC. The staining panel used in the analysis is shown in Table S8 . Myeloid cells, such as neutrophils, monocytes and macrophages, are a large constituent of the BM and play well-described roles in fracture healing (2, 5, 27–33). FC analysis distinguishing polymorphonuclear granulocytes (PMNs) from monocytes within an F4/80^- population is shown in Figure 2A , but PMNs could not be distinguished from myeloid cell precursors using this staining panel. PMNs, potentially including neutrophils, eosinophils and basophils, were the largest single population within bone, BM and the calli on d5 and d21. Unfractured bone contained a slightly higher proportion of PMNs than BM, and this proportion was further increased in the d5 callus ( Figure 2B ). However, by d21 there was a marked reduction in the proportion of PMNs to levels significantly below those found in BM ( Figure 2B ). There was no significant increase in the proportion of bulk monocytes in the callus (either d5 or d21) vs. BM, and an unexpected reduction in both the calli of d5 and d21 compared to the unfractured bone ( Figure S8A ).

We also analyzed the activation of monocytes by measuring the cell surface expression of costimulatory molecules, including CD86, GITRL, and cell surface major histocompatibility complex (MHC) Class II. CD86 levels in bulk monocytes were higher in bone than in BM, with levels in the callus of d5 sitting in between the two, and levels in the callus of d21 comparable to those in BM ( Figure S8B ). The number of MHC Class II-expressing monocytes were similar in all tissues ( Figure S8C ), but the cell surface levels of MHC II were higher in the monocytes of d5 callus than in either bone or BM ( Figure 2C ), indicating activation of some monocytes in the d5 callus. This was further confirmed by a marked upregulation in the expression of GITRL in a small proportion of monocytes in the d5 callus ( Figure 2D ). Monocytes can be phenotypically and functionally separated into classical (Ly6C^Hi) and non-classical (Ly6C^Lo) ( (34) and Figure 2A ). The proportion of Ly6C^Lo non-classical monocytes was comparable in BM, unfractured bone, and d5 callus, and significantly reduced in d21 callus ( Figure 2F ). On the other hand, there was marked enrichment of the classical monocytes in both unfractured bone and d5 callus relative to BM or d21 callus ( Figure 2G ).

Macrophages are typically produced from two sources: yolk sac or liver progenitors during embryonic development, or following differentiation from monocytes. The F4/80⁺ macrophages in the calli of d5 or d21 were present at proportions indistinguishable from those found in unfractured bone or BM ( Figures 2E, H ). When we examined the proportion of activated callus macrophages, using MHC Class II and costimulatory molecules as a measure of activation, we found that the proportion of MHC Class II⁺ macrophages in the d5 and d21 callus was substantially below that found in BM, and similar to that found in bone ( Figure S9A ). However, the proportion of macrophages expressing the costimulatory molecules CD80 ( Figure 2I ), GITRL ( Figure 2J ), CD40 ( Figure S9B ) and CD86 ( Figure S9C ), as well as the expression levels of MHC Class II ( Figure 2K ), were all higher in macrophages that populated the d5 callus as compared to those that populated unfractured bone and BM, indicating that macrophages become activated during the inflammatory phase of fracture healing. Expression of MHC Class II and all the costimulatory molecules were significantly reduced on d21 vs d5, but the proportion of CD80⁺ ( Figure 2I ) and CD40⁺ ( Figure S9B ) macrophages, as well as levels of MHC Class II ( Figure 2K ), remained higher on d21 relative to BM.

We expanded our analysis to dendritic cells (DC), which are the most closely related populations to the monocyte/macrophage axis. DC act as a crucial linkage between the innate and adaptive immune response (35). DC sense tissue damage, become activated, and either present antigen to T lymphocytes (which subsequently aid in B lymphocyte responses) or mediate DC-specific effector functions (35). As DC are a heterogenous cell population, we analyzed bulk DC as well as subpopulations (defined in Table S8 ). We found that bulk DC constitute ~1% of the BM single cell population ( Figures 3A, B ) a proportion consistent with DC in secondary lymphoid organs such as spleen or lymph node (36). Interestingly, bulk (i.e. CD11c⁺) DC were enriched in bone vs. BM, and further enriched in the d5 callus (~2.5-fold higher in the d5 callus than in the BM) ( Figures 3A, B ). However, by d21 the proportion of DC in the callus fell significantly relative to d5 ( Figure 3B ).

We then analyzed conventional and effector DC subpopulations. Within the conventional DC (cDC) population, cDC1 are primarily resident within secondary lymphoid organs and are important in initiating CD8⁺ T cell responses (37). As expected, we found few cDC1 in BM, bone or the callus ( Figure 3C ). Plasmacytoid DC (pDC) are effector DC that are primed to rapidly produce Type I interferons (38), which can play an important role in bone biology (39). We observed no enrichment of pDC in the fracture callus vs BM or bone ( Figure 3D ). An analysis of the proportion of pDC that express MHC Class II revealed no discernable increase in the d5 or d21 callus over unfractured bone, but showed a mild, yet statistically significant, increase in d21 callus over BM and d5 callus ( Figure S10A ). We also observed a mild yet significant increase in the levels of MHC Class II on the surface of pDC in the calli of d5 and d21 relative to either BM or unfractured bone ( Figure S10B ). In contrast, the proportion of PDCs expressing the costimulatory molecules ( Figure S10C for CD40; data not shown for other costimulatory molecules) was comparable in unfractured bone, d5 callus, and d21 callus, and was substantially higher in the 3 tissues than in BM. Therefore, given the inconsistent expression patterns of the markers of DC activation, there is no evidence of enrichment or activation of pDC during fracture healing vs normal homeostatic conditions.

To explore the DC subpopulations that are responsible for the increased bulk DC infiltration we observed on d5, we examined the cDC2 ( Figures 3E, F ) and the monocyte-derived DC (Mo-DC) ( Figures 3E, J ) subpopulations. cDC2 migrate between peripheral tissues and secondary lymphoid organs and are important players in the initiation of CD4⁺ T cell responses (40, 41). There was a marked enrichment of cDC2 in the d5 callus ( Figure 3F ). The proportion of cDC2 was reduced in the d21 callus but remained substantially higher than that in the BM or bone ( Figure 3F ). On d5 and d21, a higher proportion of cDC2 in the callus than in the BM or bone expressed MHC Class II ( Figure 3G ) at elevated levels ( Figure S11A ). The proportion of CD80⁺ ( Figure 3H ), CD86⁺ ( Figure 3I ), CD40⁺ ( Figure S11B ), and GITRL⁺ ( Figure S11C ) cDC2 was substantially higher in the callus of d5 than in BM or unfractured bone. Meanwhile, the proportion of cDC2 that expressed most of these costimulatory molecules decreased on d21 to the background levels detected in BM and bone ( Figures 3H, I , S11B, C ), indicating recruitment and activation of cDC2 during the initial inflammatory phase of healing, which then reduced to background levels during the later phases.

Mo-DC ( Figures 3E, J ) are effector DC that differentiate from inflammatory monocytes. They produce TNFα and reactive nitrogen species (42) and can also alter functional differentiation of T lymphocytes locally (43, 44). The proportion of Mo-DC in the d5 callus was markedly higher than that in bone, which in turn was substantially higher than that in the BM ( Figure 3J ). Indeed, the increased proportion of cDC2 and Mo-DC cumulatively accounted for almost all the increase in the bulk DC population that we detected in d5 callus ( Figure 3B ). There was also a marked increase in the proportion of MHC Class II⁺ ( Figure 3K ), CD80⁺ ( Figure 3L ), CD86⁺ ( Figure 3M ), CD40⁺ ( Figure S12A ), and GITRL⁺ ( Figure S12B ) Mo-DC, as well as the levels of MHC Class II ( Figure S12C ), in the d5, but not in the d21, callus relative to bone or BM. Interestingly, the proportion of Mo-DC expressing MHC Class II and all the aforementioned co-stimulatory molecules was substantially higher in unfractured bone than in the BM ( Figures 3K-M , S12A, B ). Therefore, similar to cDC2, Mo-DC are recruited and activated during the early phases of healing. It is noteworthy that the proportion of MHC Class II⁺ Mo-DC in the d5 callus was ~40% of that in the cDC2 population ( Figures 3G, K ), but the number of Mo-DC was larger than that of cDC2 population ( Figures 3F, J ). Taken together, our results demonstrate that innate immune cells, primarily monocytes, macrophages, and DC, are activated during the inflammatory phase of fracture healing and return to close to near-homeostatic numbers and activation status during the later phases. To corroborate these results, we used RT-qPCR to measure the expression level of (C-X-C motif) ligand 16 (Cxcl16), a chemokine secreted at high levels by activated macrophages and DC. Cxcl16 expression on days 10 and 21 was 3-4-fold less than its expression on days 3 and 5 ( Figure S12D ), confirming that recruitment and activation of innate immune cells occur mainly during the early repair phases.

Because myeloid and DC responses shape the subsequent lymphoid response, in which Natural Killer (NK) cells (45–48), T cells (14–16, 49) and B cells (16, 17) are known to impact fracture healing, we investigated the proportion of lymphoid cells in the fracture callus (phenotypes are outlined in Table S8 and shown in Figures 4A, D, G ). We started with the analysis of d5 callus and found no significance difference in the proportion of NK cells there vs. in either bone or BM ( Figures 4A, B ), which indicates re-establishment of the normal proportion of NK cells found in bone/BM in the healing hematoma. The proportion of NK T cells, a heterogenous population of cells that share a number of properties of both T cells and NK cells, in the callus was comparable to that in bone and significantly less than the proportion in BM ( Figure S13A ). On the other hand, the proportion of bulk T cells was slightly, but significantly, higher in the d5 callus than in the BM ( Figures 4A, C ). The T cell population is generally composed of CD4⁺ and CD8⁺ T cells, but bone marrow is enriched in a CD4^- CD8^- T cell subpopulation (50, 51). Analysis of these T-cell subpopulations indicated comparable proportions of CD4⁺ and CD8⁺ T cells in the d5 callus vs bone or BM ( Figures 4D-F ). The only T-cell subpopulation whose proportion was higher in the callus than in the BM or bone was the CD4^- CD8^- T cells ( Figure S13B ), which might account for the small increase in the bulk T cells observed in the d5 callus relative to the BM ( Figure 4C ). Analysis of B cells indicated comparable proportions of bulk B cells in d5 callus, bone, and BM ( Figures 4G, H ). We also analyzed the B-cell subpopulation B1b B cells, a subset of innate lymphoid cells that displays a wider specificity and responds earlier after insult than conventional B cells. We observed a small, yet significant, increase in the proportion of this subpopulation in the d5 callus vs. bone and bone marrow ( Figure 4I ). Collectively, these results indicate that the normal proportions of lymphocytes that exist in unfractured bone/BM are re-established in the healing hematoma of d5.

In contrast to the myeloid and DC response, which was increased on d5 and reduced on d21, the proportions of NK ( Figures 4A, B ) and bulk T cells ( Figures 4A, C ) increased on d21 relative to d5, and the proportions of both populations were substantially higher in d21 callus than in BM or bone ( Figures 4A-C ). Unlike on d5, the increase observed in the proportion of bulk T cells on d21 could be accounted for by an increase in CD4⁺ ( Figure 4E ) and CD8⁺ ( Figure 4F ) T cells, whereas the proportion of CD4^- CD8^- T cells was reduced on d21 vs d5 ( Figure S13B ). Interestingly, the proportion of B cells was dramatically (~450%) increased in the d21 callus relative to the d5 callus, but also increased in the BM (~ 260%) and bone (~200%) harvested from d21 mice as compared to those harvested from d5 mice ( Figures 4G, H ). Notably, the proportion of B cells in d21 callus remains significantly higher than their proportions in bone or BM ( Figures 4G, H ). In addition, B1b B cells expanded with similar kinetics in all tissues analyzed ( Figure 4I ). Therefore, these data suggest that there is a systemic conventional B cell response triggered by bone fracture. We also examined activation of B cells, measured by upregulation of the costimulatory molecules CD40 and CD86. In doing so, we detected upregulation of CD40 expression in (~13%) of B cells in the d21 callus, which was ~150-fold higher than the proportion of CD40⁺ B cells in BM (~0.08%) or bone (~0.07%) ( Figure 4J ). A much smaller proportion of B cells upregulated CD86 ( Figure S13C ) and other costimulatory molecules (not shown). In summary, our data support a strong lymphoid response in the NK and T cell compartment in the d21 callus, with a much smaller response in any lymphoid compartment during the early inflammatory phase. Consistently, RT-qPCR analysis showed 4- to 9-fold increase in the expression of C-C motif chemokine ligand 5 (Ccl5) and interferon gamma (Ifng), both of which are highy produced by activated T cells, on d21 relative to days 3, 5, and 10 ( Figures S13D, E ). Notably, the expression of tumor necrosis factor (Tnf) also increased 2 to 6 fold on d21 relative to days 3, 5, and 10 ( Figure S13F ). Importantly, the lymphoid response observed on d21 cannot be attributed to re-establishment of BM elements at this timepoint, as the callus contained cells that were highly enriched compared to the BM. The B cell response on d21 becomes systemic but is greatly enhanced in the callus at this timepoint.

We reported previously that DIO mice exhibit defective fracture healing that is typified by accelerated/pre-mature resorption of the soft callus (3). Accelerated soft callus resorption can be observed by comparing representative images of lean callus ( Figures 1A-C ; S1A-C ) and DIO callus ( Figures 5A-C ; S14A-C ). DIO mice also show reduced bone formation (compare Figures 1D ; S1D to Figures 5D ; S14D ) and aberrant structure of collagen fibers (3). To gain a more comprehensive understanding of fracture healing in DIO mice, we performed RNA-seq using the healing calli of DIO mice at the same timepoints we used with the lean mice. We analyzed the overall pattern of gene expression over the time-course of healing in both lean and DIO groups by applying a generalized linear model fitted with a negative binomial distribution, followed by linear regression to model gene expression over the healing time-course ( Table S9 ). We identified 3909 genes that exhibited significantly different expression patterns over the course of healing in lean vs. DIO mice ( Table S9 ). Gene ontology (GO) analysis demonstrated that these genes are primarily enriched in pathways involved in immune response and leukocyte activation ( Figure 5E ; Table S10 ). These data indicate that the pattern of immune response in DIO mice during the healing course deviates from the normal pattern in lean mice. Comparison of individual days indicated that DIO mice exhibited defective immune responses and reduced activation of several immune system components between days 3 and 7 post-fracture ( Figures S15-17 ; Tables S11 - 16 ). Importantly, the most significant changes in the expression of genes involved in immune response and leucocyte activation between lean and DIO mice were observed on d5 ( Figure S16 , Tables S13 , 14 ). On the other hand, DIO mice showed accelerated chondrocyte differentiation, cartilage development, and upregulated proteoglycan biosynthesis on days 5 and 7 ( Figures S16, 17 ; Tables S13 - 16 ).

Consistent with our published data (3) and histological analysis ( Figures 1A-C , 5A-C , S1A-C, S14A-C ), DIO mice exhibited accelerated/premature resorption of the soft callus as evidenced by significantly reduced expression of genes involved in chondrocyte differentiation, cartilage development, and production of cartilage extracellular matrix relative to lean mice on day 10 ( Figure S18 ; Tables S17 , 18 ). Analysis of the chondrocyte markers Aggrecan (Acan) and collagen, Type II, alpha 1 (Col2a1) demonstrated that the expression of both genes peaked earlier in the callus of DIO mice ( Figure S18C ), confirming accelerated formation of the soft callus in DIO mice. On d10, when the expression of both Acan and Col2a1 peaked in the lean callus, the mRNAs of both genes declined sharply in the DIO callus, further confirming our published data (3) and histological analysis ( Figures 1A-C , 5A-C ; S1A-C, S14A-C ) showing comparatively earlier resorption of the soft callus in DIO mice.

Importantly, the early resorption of the soft callus in DIO mice was accompanied by reduced bone morphogenesis, which occurred as early as d10 ( Figure S18B ; Table S18 ). The calli of DIO mice exhibited abnormally upregulated differentiation of fat cells on d10 ( Figure S18B ; Table S19 ), consistent with published data showing increased adipogenic commitment of mesenchymal stromal cells in DIO mice (23). Intriguingly, the calli of DIO mice showed clear and extensive inhibition of translation and ribosome biosynthesis on d14 ( Figure S19 ; Tables S20 , 21 ). While lean mice showed substantial upregulation in the expression of genes involved in immune response between days 14 and 21 ( Figures 1E, F , S7A-C ; Tables S6 , S7 ), DIO mice showed significant reduction in the immune response on d21 relative to lean mice ( Figure S20 . Tables S22 , 23 ). Furthermore, while genes involved in ossification and mineralization peaked on d14 in the callus of the lean mice ( Figures S7A, B ; Table S7 ), these genes showed delayed expression in the callus of the DIO mice ( Figure S20 ; Tables S22 , 23 ). Taken together, the RNA-seq data corroborated the results of the IF staining and histological analyses and revealed additional defects in the healing callus of DIO mice, including defective immune responses during the early inflammatory phase as well as during the second infiltration wave of immune cells between days 14 and 21.

To get more cell-specific details on the dysregulated pattern of immune response in DIO mice that we identified using RNA-seq data, we directly compared the proportions of myeloid cell types (examined in Figures 2 , 3 ) in the calli of lean and DIO mice on days 5 and 21. The largest population of immune cells in the callus, PMNs and myeloid cell precursors, was reduced ~30% in the d5 callus of DIO mice relative to lean mice ( Figures 6A, B ), a reduction that ablated the small increase observed in this population in the d5 callus of lean animals relative to bone and BM ( Figure 2B ). On d21, the proportion of PMNs and myeloid cell precursors was statistically higher in DIO mice, although the proportions in both lean and DIO remained lower than those found on d5 ( Figures 6A, B ). On d5, the proportion of total monocytes was also lower in DIO mice ( Figure S21A ), and this was accounted for by a drop in both classical ( Figure 6C ) and non-classical ( Figure 6D ) monocyte populations, with only minor differences in the monocyte proportions between lean and DIO calli on d21 ( Figures 6C, D , S21A ). Conversely, there was an increase in the proportion of activated monocytes in the d5 callus of DIO mice relative to lean mice as indicated by an increase in the proportion of monocytes expressing the costimulatory molecule GITRL ( Figure 6E ) and in the level of surface expression of MHC-II ( Figure 6F ). However, the number of monocytes expressing the co-stimulatory molecule CD86 was decreased in the DIO callus relative to the lean callus at this timepoint ( Figure S21B ), while the proportion of monocytes expressing MHC-II was comparable in the lean and DIO calli ( Figure S21C ), a finding that complicates any conclusions regarding the activation status of monocytes in DIO relative to lean mice.

The F4/80⁺ macrophage population was not altered in DIO mice (on either d5 or d21) relative to lean mice ( Figure S21D ). The proportion of macrophages that became activated on d5 was greater in DIO mice than in lean mice, as measured by an increase in the cell surface levels of MHC-II ( Figure S21E ), and an increase in the proportion of MHC-II⁺ ( Figure S21F ), CD40⁺ ( Figure S21G ), CD80⁺ ( Figure S21H ), CD86⁺ ( Figure S21I ) and GITRL⁺ ( Figure S21J ) macrophages. Any activation of these cells had returned to background levels by d21 in both lean and DIO mice ( Figures 2I-K , S21E-J ).

DC were among the most highly modulated cell populations in the callus of lean mice on d5 ( Figure 3B ), and the proportion of bulk DC in the DIO callus at this timepoint was increased ~58% over that in lean mice, returning to baseline by d21 ( Figures 6G, H ). Of the DC subpopulations, the proportions of PDC ( Figure S22A ) and cDC1 ( Figure S22B ) in the DIO callus were comparable to those in the lean callus on d5 and d21. Interestingly, on d5, there was a ~60% increase in the proportion of cDC2 ( Figure 6I ) and a ~120% increase in the proportion of Mo-DC ( Figure 6J ) in DIO callus relative to lean callus. This created a stoichiometry of DC subpopulations (i.e., the % of each subpopulation relative to total DC not to total single cells) that was substantially different between lean and DIO calli on d5, with the most noticeable difference being the expansion of the Mo-DC subpopulation in the DIO callus to constitute ~64% of the total DC ( Figure 6K ). The increase in the proportions of cDC2 and Mo-DC in the DIO callus as compared to lean callus on d5 ( Figures 6I, J ) was accompanied by a greater activated proportion of each of the two subpopulations, as measured by an increase in the proportion of cells expressing CD40 ( Figure 6L ), CD86 ( Figures 6M, N ) and MHC-II ( Figure 6O ) (data not shown for other markers). The % of MHC-II⁺ and CD40⁺ PDC and the expression level of MHC-II on PDC were comparable in the lean and DIO calli ( Figures S22C-E ). By d21, the increase in DC response observed on d5 in DIO callus had normalized, and the proportions of almost all DC and activated DC subsets in the DIO calli were either similar to or lower than their proportions in the lean calli ( Figures 6H-O ; S22A-E ). Notably, the stoichiometry of the DC subpopulations was substantially different between d5 and d21 in both lean and DIO mice: on d5, Mo-DC was the major DC subpopulation in both lean and DIO calli, while on d21, Mo-DC shrank in both lean and DIO calli to become a minor subpopulation, resulting in an increase in the % of all other subpopulations, especially PDC which, together with cDC2, were the major DC subpopulations on d21 ( Figure 6K ). In summary, during the inflammatory phase, while there is a reduction in PMNs and monocytes in the DIO callus relative to lean callus, there is an increase in the proportions of activated macrophages as well as an increase in the DC response, primarily cDC2 and Mo-DC. Consistent with this, RT-qPCR analysis detected a significant increase in the expression of Cxcl16 in the DIO callus relative to the lean callus on d5, but not at any later timepoint ( Figure S22F ).

As we observed substantial changes in innate myeloid cell populations in the callus of DIO mice, we hypothesized that there would be likely changes in the lymphoid responses that are primed and shaped by the early myeloid cell responses. Therefore, we examined the proportion of NK cells, NK T cells, T cells and B cells in the DIO vs. Lean callus on days 5 and 21. The proportions of almost all the analyzed lymphoid cell populations on d5 in the DIO mice were similar to those in the lean mice, except for the innate B1b B cells whose proportion increased in the DIO callus over the lean callus at this timepoint ( Figures 7A-E , S23A, B ). On the other hand, many differences were observed between lean and DIO calli in the lymphoid populations on d21. The proportion of NK cells was significantly lower in DIO mice than in lean mice, and, unlike in the lean mice, the proportion of this population did not increase on d21 relative to d5 in the DIO mice ( Figure 7A ). There was no significant difference in the NK T cell population in the DIO callus vs lean callus ( Figure S23A ). In the T-cell compartment, relative to the lean mice, DIO mice showed a lower proportion of CD4⁺ ( Figure 7B ), but similar proportions of CD8⁺ T cells ( Figure 7C ) and CD4^-CD8^- T cells ( Figure S23B ). No discernable difference was observed between lean and DIO calli in the proportion of the bulk B cells ( Figure 7D ). However, we did observe an alteration in the constituents of the B cell population. The proportion of the innate B1b B cells were reduced in the DIO callus as compared to the lean callus at this timepoint ( Figure 7E ). We also observed that, in the lean callus, the majority of B cells expressed high levels of CD45R (B220), indicative of a mature status ( Figures 7F, G ). In contrast, in the DIO callus, the proportion of mature cells was reduced and replaced by immature B cells expressing intermediate levels of CD45R, which likely represent Pro B cells that are not fully functional ( Figures 7F, H ). Therefore, in the DIO callus there is an enhanced early d5 B1b B cell response, but deficits in the NK, CD4+ T, B1b B cell, and mature B cell responses on d21, indicating that the adaptive immune response observed on d21 in lean mice ( Figure 4 ) is defective in DIO mice. RT-qPCR results were in total agreement with this conclusion and showed significantly reduced expression of Ccl5, Ifng, and TNF on d21 in the callus of DIO mice as compared to the callus of lean mice ( Figures S23C-E ).

Lean and DIO, male C57BL/6J mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA) at the age of 18 weeks (high-fat diet feeding of DIO mice started at the age of 6 weeks). Mice were allowed to acclimate to the care facility for 2 weeks, during which DIO mice were placed on high-fat diet (60 kcal % fat; Open Source Diets, Research Diets Inc.), while the lean controls received lean diet (10 kcal % fat; Open Source Diets, Research Diets Inc.). At the end of the 2-week acclimation period, obesity was confirmed, and the body weight of DIO mice was significantly higher than that of lean mice ( Figure S25 ). The animals were provided with ad libitum access to pelleted feed and water, and they were housed in a 12-hour light/dark cycle at 21.1°C to 22.8°C and 30% to 70% humidity. All animal protocols were approved by the University Committee on Animal Resources (IACUC) at the Pennsylvania State University College of Medicine. Notably, only male mice were used as female mice are more resistant to the development of obesity-associated metabolic changes (70, 71).

Open tibial fracture osteotomy was induced in the right hindlimb according to the established protocol that we described previously (3, 26). Fractures were stabilized using an intramedullary nail, and stabilization of the fracture site was confirmed by taking X-ray images post-operatively and at the harvest time using UltraFocus DXA system (Faxitron Bioptics). Harvest timepoints were decided based on studies published by us and others on the healing time course of this fracture model (3, 23, 26). Mice were kept on the corresponding lean or high-fat diet until harvest time.

The fractured limb was harvested from the mid-femur to the tibiotalar joint to avoid disruption of the healing callus. Most of the surrounding soft tissues were removed, and the isolated tissue was fixed in 10% (v/v) neutral buffered formalin (Thermo Fisher Scientific). The intramedullary pin was removed, and the fixed tissue was then decalcified in 14% w/v EDTA tetrasodium (Thermo Fisher Scientific) and embedded in paraffin. Sagittal sections with 5-µm thicknesses spanning the center of the callus were collected. The sections were stained with Hematoxylin/Safranin-O/Fast Green or subjected to IF staining as we described previously (3, 26). Antibodies used in IF staining are listed in Table S24 .

Soft tissues surrounding the healing callus were completely removed, and the callus was harvested, avoiding the surrounding intact bone, and immediately snap frozen in liquid nitrogen. RNA was extracted according to the protocol we detailed previously (72). RNA-seq was performed as we described previously (26, 73) using three biological replicates (i.e., three callus tissues). Briefly, the Illumina^® Stranded mRNA Prep and Ligation kit (Illumina) was used to prepare cDNA libraries as follows. Oligo (dT) beads were used to purify poly(A) RNA from 200 ng total RNA, and the purified fraction was subjected to fragmentation and reverse transcription, followed by end repair, 3′ adenylation, and adaptor ligation. The unique dual index sequences (IDT^® for Illumina^® RNA UD Indexes Set A, Ligation, Illumina) were incorporated in the adaptors for multiplexed high-throughput sequencing. PCR amplification and SPRI bead purification (Beckman Coulter) were then performed. Size distribution and concentration of the final product were assessed using BioAnalyzer High Sensitivity DNA Kit (Agilent Technologies). Pooled libraries were diluted to 3 nM using 10 mM Tris-HCl (pH 8.5), denatured, loaded onto an S1 flow cell (Illumina NovaSeq 6000), and run for 2x53 cycles.

RNA-seq data were analyzed as we described previously (26). Briefly, Illumina bcl2fastq (released version 2.20.0.422) was used to generate De-multiplexed and adapter-trimmed sequencing reads, allowing no mismatches in the index read. Low quality sequences were trimmed/filtered using BBDuk (ourceforge.net/projects/bbmap/) and the “qtrim=lr trimq=10 maq=10” option. HISAT2 (version 2.1.0) was used to align the filtered reads to the mouse reference genome (GRCm38), applying –no-mixed and –no-discordant options. HTSeq was used to calculate read counts, and DESeq2 was used to determine differentially expressed genes.

For analysis of gene expression profile over the time course, we first applied a generalized linear model (GLM) to raw read counts fitted with a negative binomial distribution and obtained TMM (trimmed mean of M-values)-normalized read counts using an R package TCC (74). Another R package maSigPro (75) was then used to identify differentially expressed genes over the time course between the lean and DIO groups. maSigPro applies linear regression to model gene expression and selects differentially expressed genes. Using Polynomial degree=2 and Q=0.05 as significance cutoffs, we identified the differentially expressed genes between the lean and DIO groups. For GO analysis, the integrated Differential Expression and Pathway analysis (iDEP) (76) and ShinyGO (77) were used, and heatmaps were prepared using iDEP.

cDNA was prepared using ~ 1 μg RNA and SuperScript IV Reverse Transcriptase (Thermo Fisher Scientific). qPCR was performed using TaqMan Fast Advanced Master Mix and TaqMan Gene-Expression Assays (Thermo Fisher Scientific).

The mice were sacrificed, and the healing hematoma of d5 was collected or the healing callus of d21 was shaved. The surrounding cortical bone was avoided. The contralateral, unfractured tibia was also collected from each mouse, the BM was completely flushed out, and each tissue was processed and analyzed separately. The soft tissues were completely removed from the calli and bones. The harvested tissues were minced and digested for 1 h at 37° C in a mixture of collagenase/dispase (1 µg/ml, catalog: 10269638001, Roche) and collagenase D (3 mg/ml, catalog: 11088882001, Roche). The red blood cells were then lysed using 1X RBC lysis buffer (Catalog # 00-4333-57, eBioscience) according to the manufacturer’s protocol. The cells were counted, blocked using 24G2 hybridoma and 20% serum, and stained using the antibodies described in Table S24 . Isotype controls and fluorescence-minus-one (FMO) controls were employed. The cells were sorted using 23-color BD FACS Symphony (BD Biosciences, San Jose, CA), and the results were analyzed using FlowJo Software 10.6.2 (Treestar, Ashland, OR). The experiment was repeated for 3 biological replicates, and 5 calli were pooled for each replicate.

ANOVA (followed by Tukey’s post hoc test) was used to determine statistical significance among three or more groups. Statistical analyses were performed using the GraphPad Prism software. The following symbols were used to indicate significance: (ns) P ≥ 0.05; (*) P < 0.05; (**) P < 0.01; (***) P < 0.001.