Macrophage necroptosis was found in tendon lesions in burn/tenotomy mice

A mouse burn/tenotomy model was established in accordance with the relevant literatures (Fig. S1A), and sham surgery was served as the control group. To investigate the underlying mechanisms of HO, high-throughput whole transcriptome sequencing was conducted to compare the control group with tendon lesions at 7 days post-injury. Gene set enrichment analysis (GSEA) revealed that the "response to tumor necrosis factor" and the "necroptotic signaling pathway" were significantly enriched in burn/tenotomy mice, compared to the control group (Fig. 1a). This finding was further supported by an elevated expression and concentration of Tnf-α, as confirmed by IHC staining (Fig. 1b) and ELISA (Fig. 1c). Both the heatmap and volcano plot (Fig. 1d), along with WB analysis (Fig. S1B), further demonstrated elevated levels of p-Ripk1, p-Ripk3, and p-Mlkl at 7 days post-injury. Moreover, as one of the most crucial immune cells involved in HO formation, macrophages were observed to undergo necroptosis during the early stages of inflammation. This was confirmed through IF staining for p-Ripk1, p-Ripk3, and p-Mlkl, which co-localized with F4/80, a specific marker used for detecting macrophages (Fig. 1e). Similarly, necroptotic macrophages were also detected and confirmed at 3 weeks post-surgery (Fig. S1C-G). In light of the higher level of necroptosis observed in macrophages at 7 days compared to 3 weeks post tendon injury, we selected 7 days as the time point for subsequent analysis of necroptotic macrophages in the HO model.

Now that we have confirmed the presence of necroptotic macrophages in the early osteogenic microenvironment, we aimed to determine the effects of cellular necroptosis on trauma-induced HO formation in vivo. To accomplish this, we evaluated mature HO and osteogenic indicators between wild-type (WT) mice and Mlkl mice (general knockout models deficient in the key necroptosis mediator protein) using a consistent burn/tenotomy model. IF staining for Runx2 (Fig. 2a, c), SOFG staining (Fig. S2A), micro-CT (Fig. 2b, d), and H&E staining (Fig. S2B) all demonstrated that depletion of Mlkl resulted in a significant reduction of osteogenic behavior at 3 weeks post-injury, as well as diminished maturation of HO at 10 weeks following tendon injury.

In light of the fact that the effects of macrophages are primarily mediated through their paracrine functions, we investigated whether the secretome from necroptotic macrophages contributed to the osteogenic potential of TSPCs. TSPCs (Fig. 2e and Fig. S2C, E) and BMDMs (Fig. 2f and Fig. S2D) were isolated and characterized. The coculture of BMDMs and TSPCs was established utilizing the Transwell system (Fig. 2g). The induction of macrophage necroptosis was achieved through stimulation with TNFα in conjunction with zVAD-fmk, which was confirmed by IF staining and WB for p-Mlkl (Fig. S2F, G). TSPCs were also cultured in an osteo-inductive medium. The results demonstrated that following the induction of necroptosis in BMDMs, the osteogenic activity of TSPCs co-cultured with necroptotic BMDMs was significantly greater than that co-cultured with phosphate-buffered saline (PBS)&dimethyl sulfoxide (DMSO)-stimulated BMDMs. This finding was substantiated by a marked increase in WB analysis (Runx2, Opn, and Ocn; Fig. 2h and Fig. S2H), IF staining (Runx2, Opn, and Ocn; Fig. 2i and Fig. S2I), as well as ALP and ARS staining (Fig. 2j, k). The aforementioned results suggested that macrophage necroptosis enhanced trauma-induced HO by augmenting the osteogenic potential of TSPCs. Given that the secretome of macrophages could be primarily categorized into soluble fractions (SFs, conditioned medium devoid of EVs) and EVs, subsequent experiments aimed to investigate their respective roles in the osteogenic behavior of TSPCs and to determine which component exerted a more significant influence.

After the induction of necroptosis in BMDMs, EVs and SFs derived from necroptotic macrophages (NecroMφ-EVs, NecroMφ-SFs) were isolated (Fig. 3a). Characterization of the isolated EVs was achieved through TEM, NTA, and WB analysis. TEM successfully captured membrane-bound, cup-shaped nanoparticles that conformed to the typical morphology associated with EVs (Fig. 3b). NTA measured the particle size of the isolated EVs, revealing a distribution within the range of 30-150 nm (Fig. 3b). WB analysis confirmed the abundant expression of EV markers, including Cd9, Cd81, Alix, and Tsg101, alongside their parental cell marker F4/80. In contrast, there was bare expression of Calnexin, an endoplasmic reticulum marker (Fig. 3c). This collective data demonstrated the successful isolation of macrophage-derived EVs. Furthermore, we validated the uptake of these EVs by TSPCs. The intracellular space within TSPCs exhibited localization of Dil-labeled EVs, indicating effective internalization regardless of their origin (Fig. 3d).

To investigate how the secretome derived from necroptotic macrophages influenced the osteogenic behavior of TSPCs, we separately incorporated NecroMφ-EVs and NecroMφ-SFs into the osteo-inductive medium in vitro, as well as performed local tendon injections using a burn/tenotomy model in vivo (Fig. 3e). In vitro, the addition of NecroMφ-EVs or NecroMφ-SFs resulted in enhanced osteogenic activity compared to the addition of PBS, as demonstrated by WB analysis (Runx2, Opn, and Ocn; Fig. S3A), IF staining (Runx2, Opn, and Ocn; Fig. S3B), as well as ALP and ARS staining (Fig. 3f, h). In vivo studies revealed that, compared to PBS injection, the levels of mature HO and osteogenic indicators were significantly elevated following the injection of additional NecroMφ-EVs or NecroMφ-SFs. Notably, the enhancement effects were markedly greater with NecroMφ-EVs. This was corroborated by IF staining (Runx2; Fig. S3C), SOFG staining (Fig. S3D), micro-CT (Fig. 3g, i), and H&E staining (Fig. 3). All these findings suggested that NecroMφ-EVs played a significant role in the paracrine effects of macrophage necroptosis.

Moreover, to additionally further explore the route by which NecroMφ-EVs are internalized into TSPCs, it was monitored after treatment with inhibitors. The cellular internalization of EVs primarily occurs through three major endocytic pathways: clathrin-dependent endocytosis, caveolae-dependent endocytosis, and macropinocytosis. After incorporating NecroMφ-EVs into the osteo-inductive medium of TSPCs in vitro, we employed pharmacological inhibitors targeting each pathway: pitstop 2 (clathrin inhibitor), nystatin (caveolae inhibitor), and EIPA (macropinocytosis inhibitor). Results showed that both pitstop 2 and EIPA significantly reduced the osteogenic activity compared to the addition of DMSO, while nystatin treatment showed only marginal inhibition without statistical significance, as demonstrated by ALP and ARS staining (Fig. S3F). Similar trends for mature HO were also found after incorporating NecroMφ-EVs and each pharmacological inhibitor in vivo (DMSO as control, pitstop 2, nystatin, and EIPA), as demonstrated by micro-CT (Fig. S3G). These findings revealed that the cellular internalization of NecroMφ-EVs in TSPCs predominantly occurred through clathrin-dependent endocytosis and macropinocytosis routines, and they were also the major uptake mechanisms in burn/tenotomy mice. Therefore, subsequent experiments focused on investigating the internal components of NecroMφ-EVs.

Then, an important question that remains to be addressed was the specific content that mediated the biological effects of EVs. In vitro, proteomic sequencing was conducted to compare the EVs derived from PBS&DMSO-stimulated BMDMs (ConMφ-EVs) with those from necroptotic BMDMs (NecroMφ-EVs). Meanwhile, in vivo, proteomic sequencing was also conducted to compare control group with tendon lesions at 7 days post-injury. By integrating the results of the aforementioned in vitro and in vivo proteomics sequencing, bioinformatics analysis revealed a total of 173 differentially expressed proteins that were commonly identified in both models. The selection criteria included a P value of less than 0.05 and a fold change >2, together with an emphasis on upregulation. Furthermore, we intersected the aforementioned 173 proteins with the results of in vivo transcriptomics sequencing (control group vs. tendon lesions at 7 days post-injury, the same data as in Results "Macrophage necroptosis was found in tendon lesions in burn/tenotomy mice"), applying the criterion of upregulation in expression. At last, we identified a total of 113 proteins. Next, we ranked these 113 proteins based on their fold-change values in both in vivo and in vitro proteomics sequencing, ultimately identifying Pak4 and Sfrp1 as the top ten proteins in both environments. Given that Pak4 ranked higher and consistently placed within the top five across both proteomics sequencing, we finally chose to focus our subsequent investigations on Pak4 (Fig. 4a and Fig. S4A). Pak4 has been identified as a key player in fundamental cellular processes, particularly in modification and transcription regulation. This includes crucial roles in osteogenesis and stem cell differentiation. WB analysis (Pak4, Fig. S4B) further confirmed a higher content of Pak4 in NecroMφ-EVs compared to ConMφ-EVs, while there was a significant reduction observed in InhibNecroMφ-EVs (EVs derived from Mlkl BMDMs stimulated with TNFα and zVAD-fmk). Similarly, IHC staining for Pak4 (Fig. S4C) further confirmed that the expression of Pak4 was significantly elevated in burn/tenotomy mice compared to sham surgery. Additionally, a notable reduction in Pak4 expression was observed both in Mlkl mice and macrophage depletion mice compared to WT mice following burn/tenotomy.

To elucidate the incorporation of Pak4 into NecroMφ-EVs, IF staining was conducted to detect Pak4 co-localization with Cd63 or Eea1, markers for the endosomal system, in BMDMs subjected to necroptosis induction or not. The results demonstrated that upon necroptosis induction, an increased amount of Pak4 in the cytoplasm of BMDMs co-localized with the speckle-like distribution of Cd63 and Eea1 (Fig. 4b and Fig. S4E). This finding indicates that Pak4 is loaded into the endosomal system for EVs formation and extracellular release in response to necroptosis induction. To validate these findings in vivo, IF staining was performed on tendon tissues collected 7 days post-injury to assess co-localization of Pak4 with F4/80 and Cd63 (Fig. 4c). Compared to the sham group, injured tendons exhibited a significant elevation in Pak4 expression along with increased extracellular distribution; this change was accompanied by marked infiltration of F4/80 macrophages. In terms of histological localization within the BTT group, Pak4 colocalized with Cd63, and such colocalization was found to be situated around the concentrated areas of F4/80 macrophages. This suggested that these macrophages were responsible for releasing vesicles containing Pak4. Collectively, our findings indicated that necroptotic macrophages had the capacity to secrete EVs carrying Pak4. The processes involved in generating and packaging these Pak4-containing EVs were summarized in Fig. S4D.

Subsequently, to more accurately ascertain the role of Pak4 in EVs, we prepared Flox mice and Lyz2-cre::Pak4 mice (conditional knockout) alongside isolated corresponding BMDMs. After the induction of necroptosis in both Flox and Pak4 cKO BMDMs, NecroMφ-EVs and PAK4-EVs were respectively isolated. Subsequently, these two types of EVs were individually added into osteo-inductive medium in vitro or administered via local tendon injection in BTT model in vivo. In vitro analysis revealed that, compared to NecroMφ-EVs, PAK4-EVs significantly disabled the osteogenic behavior of TSPCs. This was evidenced through WB analysis for osteogenic markers Runx2, Ocn, and Opn (Fig. S5A), IF staining for these same markers (Fig. S5B), as well as ALP and ARS staining (Fig. 4d and Fig. S5C). Furthermore, PCR results showed that there were no significant differences among groups for Pak4 transcription in TSPCs after PBS, NecroMφ-EVs, or PAK4-EVs were added, respectively (Fig. S5D). What's more, WB analysis showed that addition of NecroMφ-EVs led to higher Pak4 protein concentration in TSPCs than either addition of PBS or PAK4-EVs, while there was no notable difference between the latter groups (Fig. S5E). The combined evidence from above demonstrated that the Pak4 protein within NecroMφ-EVs exerted an exogenous significant influence on the osteogenic behavior of TSPCs after its entry, without affecting the transcription levels of Pak4 in TSPCs themselves. In vivo studies demonstrated that PAK4-EV injection led to a reduction in mature HO and associated osteogenic indicators compared to NecroMφ-EVs injection. These findings were confirmed through IF staining for Runx2 (Fig. S5G), SOFG staining (Fig. S5H), micro-CT (Fig. 4e and Fig. S5F), and H&E staining (Fig. S5I). Moreover, comparisons between Flox mice and PAK4 cKO groups indicated a similar reduction in mature HO and osteogenic markers following the conditional knockout of Pak4 in macrophages (Fig. 4e and Fig. S5F-I). Based on these results, it could be concluded that necroptotic macrophages secreted Pak4-enriched EVs, which were directly uptaken by TSPCs and played a crucial role in promoting the osteogenic process during the formation of HO.

To gain deeper insights into the influence of Pak4 derived from NecroMφ-EVs on the osteogenic behavior of TSPCs, we conducted the in vivo transcriptome sequencing (control group vs. tendon lesions at 7 days post-injury, the same data as in Results "Macrophage necroptosis was found in tendon lesions in burn/tenotomy mice"), and searched for the meaningfully enriched item (Fig. 5a) in the disease model. In the top 10 Wikipathways Enrichments, there were 3 FAO-related items including "fatty acid oxidation" at 3st, "mitochondrial long-chain fatty acid beta oxidation" at 7st, and "fatty acid beta oxidation" at 10st. In addition, the FAO also presented in the top 10 chord diagram (Fig. S6A). The transcription of acyl-CoA dehydrogenases in FAO (Acadvl, Acadl, Acadm, and Acads; illustration diagram in Fig. 6S6B) was significantly reduced, as demonstrated in the volcano plot and heatmap (Fig. 5a and Fig. S6A). What's more, the results indicated that oxidative phosphorylation and electron transport chain were also accompanied by a reduction. All these demonstrates a significant reduction in FAO after burn/tenotomy model in the early stage. Further, high-throughput whole transcriptome sequencing was conducted again in vitro on TSPCs stimulated with osteo-inductive medium, which was supplemented separately with NecroMφ-EVs or PAK4-EVs. GSEA revealed that both the "fatty acid beta-oxidation" and the "fatty acid beta-oxidation using acyl-CoA dehydrogenases" were significantly reduced after NecroMφ-EVs addition, compared to PAK4-EVs addition (Fig. 5b), as well as the "oxidative phosphorylation" and the "electron transport chain" (Fig. S6C). These findings were accordance with the above in vivo results, and were further supported in vitro by metabolites analyses for palmitoyl-CoA and acetyl-CoA (Fig. 5c), seahorse OCR (Fig. S6D), and IF staining (Fig. S6E) for Lcad and Mcad, the important rate-limiting enzymes during FAO. Similarly, in vivo burn/tenotomy models separately received local tendon injection of NecroMφ-EVs and PAK4-EVs. The results demonstrated that IF staining (Lcad and Mcad co-localized with Pdgfr-α, a marker for identifying TSPCs; Fig. 5d and Fig. S6G) and WB analysis (Lcad and Mcad, Fig. S6F) corroborated an enhanced FAO in TSPCs following the injection of PAK4-EVs compared to NecroMφ-EVs. Furthermore, the in vivo results also indicated increased FAO in TSPCs associated with Pak4 conditional knockout in macrophages when compared to the wild-type group.

Then, the effects of FAO on the osteogenic behavior of TSPCs and the traumatic HO were evaluated. AVs in vitro and AAVs in vivo designed to downregulate Lcad were employed to inhibit FAO (Fig. S7A). In vitro, TSPCs were stimulated with an osteo-inductive medium supplemented with PAK4-EVs. Compared to the sh-NC control group, the osteogenic behavior of TSPCs was significantly enhanced upon the addition of sh-Lcad, as demonstrated by WB analysis (Runx2, Ocn, and Opn; Fig. S7B), IF staining (Runx2, Ocn, and Opn; Fig. S7C), as well as ALP and ARS staining (Fig. 5e). In vivo, in a burn/tenotomy model involving local tendon injection of PAK4-EVs, compared to the sh-NC control group, there was a significant increase in mature HO and osteogenic indicators when sh-Lcad was injected. This finding was confirmed through IF staining (Runx2; Fig. S7D), SOFG staining (Fig. S7E), micro-CT (Fig. 5f), and H&E staining (Fig. S7F). These results suggested that Pak4 from NecroMφ-EVs enhanced the osteogenic behavior of TSPCs during traumatic HO formation by reducing FAO.

To further decipher how Pak4 influenced FAO in TSPCs, we investigated the downstream binding molecules associated with EVs-shuttled Pak4. Mass spectrometry analysis was conducted to compare the control group and tendon lesions at 7 days, aiming to identify potential molecules capable of binding with Pak4 in the disease model (Fig. 6a and Fig. S8A). Results indicated that, following the intersection of results from three biological samples in both the control group and the BTT (7 days post-injury) group, there were 121 molecules identified in the control group and 224 molecules in the BTT group capable of directly binding to Pak4, respectively. We further conducted an intersection analysis of the molecules identified in the control group (121) and those in the BTT group (224). Ultimately, we identified 134 molecules that were capable of directly binding to Pak4 in the BTT group but not in the control group. Based on the capacity for direct binding with Pak4, Acta2, Fabp3, and Krt5 were ranked among the top 10 molecules across all three BTT samples. We chose FABP3 for subsequent research, considering it is also a general fatty acid metabolism-related molecule. Consistently, molecular docking (Fig. S8B) validated the direct physical binding ability between Pak4 and Fabp3. In vitro study, IF staining for Pak4 and Fabp3 (Fig. 6b) further visualized this direct interaction, revealing a higher degree of colocalization in TSPCs treated with NecroMφ-EVs compared to those receiving PBS. Co-IP (Fig. 6c) further confirmed the physical association between Pak4 and Fabp3. In vivo study, IF staining demonstrating colocalization of Pak4 and Fabp3 with Pdgfr-α (Fig. S8C) revealed increased colocalization in BTT mice compared to the sham group. Co-IP experiments also provided support for the physical association between Pak4 and Fabp3 (Fig. S8D). Collectively, these results indicated that Pak4 directly interacted with Fabp3 in TSPCs exhibiting osteogenic behavior and the disease model of trauma-induced HO.

Due to Pak4 being a type of kinase for serine/threonine proteins, phosphorylation prediction and mass spectrometry analyses were conducted to identify the potential phosphorylation site of Pak4 on Fabp3. Both approaches pointed to S122 as the target (Fig. 6d). Subsequently, we assessed changes in the phosphorylation level of Fabp3 (p-Fabp3) upon knockout of Pak4. IF staining (p-Fabp3, Fig. 6e and Fig. S8E) and WB analysis (p-Fabp3, Fig. 6f and Fig. S8F) further validated that p-Fabp3 levels were reduced following Pak4 knockout, both in vitro and in vivo. To confirm that direct phosphorylation of Fabp3 at the S122 site constitutes a downstream pathway through which Pak4 from NecroMφ-EVs affects FAO in TSPCs, we engineered a phosphomimetic mutant of Fabp3 on S122 site called Fabp3. In vitro experiments demonstrated that transfection with the plasmid encoding Fabp3 led to decreased FAO, as indicated by metabolite analyses revealing lower levels of palmitoyl-CoA and acetyl-CoA (Fig. 7a), alongside seahorse OCR supporting these findings (Fig. 7b). Furthermore, enhanced osteogenic activity was observed through WB analysis measuring Runx2, Ocn, and Opn expression levels (Fig. S9A), IF staining for osteogenic markers Runx2, Ocn, and Opn (Fig. S9B), as well as ALP and ARS staining (Fig. 7c). Consistently, in vivo assessments after transfecting AAV encoding Fabp3 revealed an increase in mature HO and osteogenic indicators, as evidenced by IF staining for Runx2 (Fig. 7d), SOFG staining (Fig. 7e), micro-CT (Fig. 7f), along with H&E staining (Fig. 7g). Collectively, these results indicated that Pak4 derived from NecroMφ-EVs reduced FAO in TSPCs by directly phosphorylating downstream target Fabp3 at the S122 site, ultimately incurring osteogenic changes of TSPCs during traumatic HO formation.

Finally, the activation of PAK4, p-FABP3, and LCAD was examined in clinical samples to establish the clinical relevance of our findings. Soft tissue-derived HO samples and control soft tissue samples (normal tendon) were collected (Fig. 8a). H&E staining revealed their characteristic histological features, which allowed for the identification and definition of the ossification region in HO samples (Fig. 8b). Consequently, stronger IF staining for PAK4 and p-FABP3, as well as reduced levels of LCAD, were observed in HO tissues compared to normal tendon tissues (Fig. 8c). This expression pattern paralleled that seen in murine models. Collectively, these data suggested that cells within human HO tissues were also activated for PAK4 and FABP3-related FAO, indicating a trend similar to that observed in murine tissue sections.