TNFα and IFNγ cooperate for efficient pro- to anti-inflammatory transition of macrophages during muscle regeneration

Significance

The release of cytokines from various immune cells is known to be essential for successful muscle regeneration. However, the precise cellular origin of some of these cytokines and how they influence each other are still not clear. One such cytokine is IFNγ, which is most commonly thought to promote inflammation. Here, we show that IFNγ is primarily produced by a type of white blood cells called natural killer cells following muscle damage. In addition, we demonstrated that IFNγ cooperates with another cytokine, TNFα, in inducing the transition of macrophages from an inflammation-promoting phenotype to a regeneration-promoting phenotype. Therefore, the ways in which IFNγ regulates inflammation appear to be more diverse than previously believed.

Abstract

IFNγ is traditionally known as a proinflammatory cytokine with diverse roles in antimicrobial and antitumor immunity. Yet, findings regarding its sources and functions during the regeneration process following a sterile injury are conflicting. Here, we show that natural killer (NK) cells are the main source of IFNγ in regenerating muscle. Beyond this cell population, IFNγ production is limited to a small population of T cells. We further show that NK cells do not play a major role in muscle regeneration following an acute injury or in dystrophic mice. Surprisingly, the absence of IFNγ per se also has no effect on muscle regeneration following an acute injury. However, the role of IFNγ is partially unmasked when TNFα is also neutralized, suggesting a compensatory mechanism. Using transgenic mice, we showed that conditional inhibition of IFNGR1 signaling in muscle stem cells or fibro-adipogenic progenitors does not play a major role in muscle regeneration. In contrast to common belief, we found that IFNγ is not present in the early inflammatory phase of the regeneration process but rather peaks when macrophages are acquiring an anti-inflammatory phenotype. Further transcriptomic analysis suggests that IFNγ cooperates with TNFα to regulate the transition of macrophages from pro- to anti-inflammatory states. The absence of the cooperative effect of these cytokines on macrophages, however, does not result in significant regeneration impairment likely due to the presence of other compensatory mechanisms. Our findings support the arising view of IFNγ as a pleiotropic inflammatory regulator rather than an inducer of the inflammatory response.

Tissue regeneration is an extremely complex and dynamic process that requires the interaction of multiple cell types. Skeletal muscle has a remarkable capability to fully regenerate following acute injury. On the other hand, it undergoes fibrofatty degeneration when subjected to chronic damage. This makes skeletal muscle a well-studied model of tissue regeneration and fibrosis (1). The regenerative capacity of skeletal muscle is largely attributed to an adult stem cell population called satellite cells or muscle stem cells (MuSCs). These cells are located between the sarcolemma and basal lamina of myofibers and are naturally quiescent. Following muscle damage, they become activated and divide to eventually fuse with ailing myofibers and facilitate muscle repair, whereas a subset will return to their niche to maintain steady-state numbers (2). The regenerative function of MuSCs is heavily supported by other cells including fibro/adipogenic progenitor cells (FAPs) and immune cells. FAPs are a population of stromal cell progenitors that are present in healthy muscle and are quiescent under homeostatic conditions. Following muscle damage, these cells become activated, rapidly proliferate, and release trophic factors that support MuSCs (3). In successful muscle regeneration, excess FAPs undergo apoptosis as the population returns to steady-state levels. However, when FAPs fail to undergo apoptosis, they differentiate to form adipocytes and fibroblasts, causing fibrofatty infiltration and fibrosis (4).

Innate immunity plays a crucial role in orchestrating muscle regeneration. Inflammatory cells invade damaged muscle within minutes of injury and undergo dynamic changes during the regeneration process (5, 6). Muscle regeneration can be divided into the following two major phases: a proinflammatory phase and an anti-inflammatory phase. The early proinflammatory milieu supports the engulfment of necrotic fibers by invading neutrophils and macrophages as well as the uptake of damage-induced apoptotic cells by resident macrophages (7), and it activates resident MuSCs to proliferate. A switch to the anti-inflammatory phase is required for the differentiation and fusion of MuSCs with regenerating myotubes and triggers the production of provisional extracellular matrix by FAPs (5, 6). Several proinflammatory cytokines including TNFα, IL1β, and IL6 are believed to play a role in the early phase of muscle regeneration (8). Muscle regeneration defects seen in TNFα and TNFα receptor null mice (9–11) suggest important roles of TNFα signaling in MuSC differentiation (9) or in regulating FAP apoptosis (4). IFNγ, another cytokine traditionally known for its proinflammatory properties, has been widely studied in antimicrobial and antiviral immunity. It has been shown to have a synergistic effect with TNFα in different contexts (12–14). In contrast to TNFα, however, our knowledge of its role in the context of sterile tissue injuries is limited. Here, we thoroughly investigate the source, kinetics, and function of IFNγ in muscle regeneration following a sterile injury.

Results

Circulating NK Cells Infiltrate Damaged Muscle and Are the Main Source of IFNγ.

Different cell types including natural killer (NK) cells, T cells, macrophages, and type 1 innate lymphoid cells (ILC1s) are capable of producing IFNγ. In contrast to macrophages and T cells, the presence of NK cells or ILC1s in muscle regeneration has not been well studied. Therefore, we used flow cytometry to detect cells consistent with the features of these two cell types during damage-induced muscle regeneration. We detected a population of CD45⁺NK1.1⁺CD11B^Low cells invading the muscle at day 3 after notexin (NTX) injury. Further analysis showed that this population of cells expressed CD49b confirming their identity as NK cells and not ILC1s (Fig. 1A) (15). To determine the source of NK cells in damaged muscle, we surgically joined CD45.2 and CD45.1 mice in parabiosis. Two weeks after blood chimerism was established, we injured the tibialis anterior (TA) muscles of both partners. Three days after injury, we harvested the TA muscles of the CD45.2 partner and assessed the contribution of cells from the CD45.1 partner, which can only reach the tissue through the bloodstream, to the damaged muscle. The chimerism of NK cells in damaged muscle was normalized to the proportion of donor-derived NK cells in the circulation (i.e., blood from the CD45.2 partner). We found that 96.34 ± 11.12% (mean ± SD) of NK cells in damaged muscle are derived from the blood and not from the proliferation of any resident cells (SI Appendix, Fig. S1A).

Fig. 1.

NK cells and T cells are the lone producers of IFNγ following muscle damage. (A) Gating strategy to identify NK cells. (B) The kinetics of NK cells during muscle regeneration. The number of NK cells obtained per TA muscle is shown (n = 2, 5, 11, 13, 11, 11, and 3 mice for day 0 to day 6). (C) The concentration of IFNγ in regenerating TA muscles detected by ELISA (n = 6, 7, 9, 9, 8, 9, and 5 mice for day 0 to day 6). (D) The expression of Ifng by different cell types sorted from TA muscles on day 3 after damage (n = 2 mice). (E) The concentration of IFNγ in regenerating TA muscles in the absence of different cell types detected by ELISA (n = 5 mice per group, ordinary one-way ANOVA followed by Tukey’s multiple comparisons tests). Blank (dotted line) shows the absorbance value of a well that did not receive muscle lysate, ns = not significant, ***P < 0.001, ****P < 0.0001.

To determine the kinetics of NK cells in muscle, we quantified their numbers at several time points after acute injury. Analysis with flow cytometry showed that these cells are not present in muscle preceding damage (day 0) and peak at 3 d postinjury (p.i.) before decreasing in abundance (Fig. 1B). NK cells are known to be a prominent source of IFNγ in infection (16). To evaluate their contribution of IFNγ in sterile damage, we measured IFNγ levels throughout muscle regeneration using enzyme-linked immunosorbent assay (ELISA). IFNγ remained low in abundance on day 1 and day 2, before drastically increasing to peak on day 3 following damage (Fig. 1C). This peak of IFNγ coincided temporally with the peak of NK cell numbers, supporting the hypothesis that NK cells may play a significant role in producing IFNγ. Indeed, quantification by droplet digital PCR showed that, along with T cells, NK cells are a main source of Ifng following muscle damage (Fig. 1D and SI Appendix, Fig. S1B). In order to confirm that NK cells and T cells were the sole producers of IFNγ, we used recombination-activating gene 1 (Rag1) knockout (KO) mice, which lack mature T cells. Rag1 KO mice or C57BL/6J mice were treated with either NK1.1 neutralizing antibodies (to deplete NK cells) or an isotype control. Administration of the NK1.1 neutralizing antibody every 4 d starting at 8 d before muscle injury efficiently decreased the number of NK cells found in damaged muscle (SI Appendix, Fig. S1C). A reduction in IFNγ production was observed in Rag1 KO isotype–treated mice, confirming that T cells are a source of this cytokine in regenerating muscle (Fig. 1E). In treated Rag1 KO animals with the NK1.1 neutralizing antibody, IFNγ was drastically reduced to undetectable levels (Fig. 1E). Strikingly, IFNγ was also drastically reduced in C57BL/6J mice upon NK cell depletion. These findings confirm that NK cells are the major source of IFNγ in regenerating muscle and suggest that they may also be required for T cells to efficiently produce IFNγ. The complete absence of IFNγ in nonobese diabetic (NOD) severe combined immunodeficiency (SCID) gamma mice (NSG mice) (SI Appendix, Fig. S1D), which lack both T cells and NK cells, compared to NOD SCID mice, which only lack T cells, further confirmed that NK cells and T cells are responsible for the production of all IFNγ found in damaged muscle.

NK Cells Are Not Required for Efficient Muscle Regeneration.

Our understanding of NK cells and their role in muscle regeneration and fibrosis is limited. A few studies assessed the importance of IFNγ in muscle regeneration with conflicting results. Two independent studies by Cheng et al. (17) and Zhang et al. (18) suggested that IFNγ is required for efficient muscle regeneration. On the other hand, Villalta et al. (19) found that IFNγ promotes muscle damage during the regenerative stage of muscular dystrophy and a null mutation of IFNγ improves motor function of Dmd^mdx mice, a mouse model for human Duchenne muscular dystrophy. To assess the importance of IFNγ-producing NK cells in muscle regeneration, we treated wild-type (WT) C57BL/6J animals with the neutralizing antibody against NK1.1 to deplete these cells. We did not detect any impairment in muscle regeneration in the absence of NK cells (Fig. 2 A and B) despite published reports pointing to the importance of endogenous IFNγ in muscle regeneration (17, 18). Although the amount of IFNγ was drastically lower in NK cell–depleted mice than WT mice (Fig. 1E), we asked whether the small amount of IFNγ produced by T cells may be sufficient to support muscle regeneration. Therefore, we compared muscle regeneration in NSG mice, which completely lack IFNγ (SI Appendix, Fig. S1D), and NOD SCID mice in which IFNγ-producing NK cells are still present. We did not see any impairment in the regenerative process comparing the two mouse strains (Fig. 2C), suggesting that IFNγ is dispensable for muscle regeneration.

Fig. 2.

The absence of NK cells does not have a major effect on muscle regeneration or fibrosis development. (A) Picrosirius red and nuclei staining of TA muscle sections from damaged control (isotype treated) and NK-depleted WT mice. (B) Distribution of centrally nucleated myofiber cross sectional area (CSA) of damaged muscles from control and NK-depleted WT mice (n = 5 mice per group, Brown–Forsythe and Welch ANOVA tests followed by Dunnett’s T3 multiple comparisons test on selected bins). (C) Distribution of centrally nucleated myofiber CSA of damaged muscles from NOD SCID and NSG mice (n = 4 mice per group, Brown–Forsythe and Welch ANOVA tests followed by Dunnett’s T3 multiple comparisons test on selected bins). (D) Picrosirius red and nuclei staining of diaphragm sections from control (isotype treated) and NK-depleted Dmd^mdx mice. (E) Distribution of myofiber CSA from control and NK-depleted Dmd^mdx mice (n = 5 mice for control and n = 6 mice for NK-depleted Dmd^mdx, Brown–Forsythe and Welch ANOVA tests followed by Dunnet’s T3 multiple comparisons test on selected bins). (F) Ex vivo functional analysis of dystrophic diaphragm muscle from control and NK-depleted Dmd^mdx mice (each dot represents one mouse, unpaired t test).

The role of IFNγ has been widely studied in the context of tissue fibrosis (20–22), with several clinical trials assessing its potential as a therapeutic (23, 24). Indeed, numerous studies have suggested an antifibrotic role for NK cells in the liver through the killing of hepatic stellate cells that generate myofibroblasts, which are responsible for collagen deposition (25–28). Fibrosis is a common outcome of chronic muscle damage and is invariably observed in muscular dystrophies (29, 30). Therefore, we further studied whether NK cells in Dmd^mdx mice may also have an antifibrotic effect. As the long-term twice-a-week injection of a neutralizing antibody is not practical, we initially tested whether NK cells could be depleted with less frequent antibody administration. Dmd^mdx Utrophin^+/− mice were used as a sensitized system to evaluate the efficiency of NK depletion with less frequent doses as they develop more extensive muscle inflammation and injury. We found that the administration of a single dose of the NK1.1 neutralizing antibody every 2 wk is sufficient to efficiently deplete NK cells in dystrophic muscle (SI Appendix, Fig. S2A). Therefore, we started the treatment of Dmd^mdx mice at 4 to 6 wk of age with either the anti-NK1.1 neutralizing antibody or an isotype control every 2 wk and continued it for 6 to 7 mo. At the end of treatment, the diaphragm muscle, which is particularly affected by the disease and develops extensive fibrosis, was harvested for histological and functional analysis. We did not detect any difference in the size of myofibers (Fig. 2 D and E) as well as in the maximum tetanus force production or muscle active stiffness (Fig. 2F and SI Appendix, Fig. S2B), suggesting that the depletion of NK cells does not appear to have a significant effect on muscle regeneration and fibrosis development in muscular dystrophies.

TNFα Compensates for the Absence of IFNγ in Muscle Regeneration.

Synergistic effects of IFNγ and TNFα have been documented in different contexts such as autoimmune diabetes (12), inflammatory atherogenesis (13), and inflammatory bowel disease (14). The highest levels of TNFα in regenerating muscle are detected on day 2 to 3 p.i. (SI Appendix, Fig. S3A), close to the peak of IFNγ abundance. To examine the role of IFNγ in the absence of TNFα, we evaluated muscle regeneration in NSG mice, which completely lack IFNγ, and NOD SCID mice, which have IFNγ-producing NK cells. Both mice were treated with TNFα neutralizing antibodies daily for 6 d following NTX injection. Partial versus complete neutralization of TNFα has been proposed to have opposite effects (31, 32). Therefore, we initially tested two doses of TNFα neutralizing antibodies (a low dose of 50 μg daily or a high dose of 150 μg daily). In NSG mice treated daily with 50 μg of TNFα neutralizing antibodies, we observed a significantly higher frequency of smaller muscle fibers compared to anti-TNFα–treated NOD SCID mice (Fig. 3A), indicating inefficient muscle regeneration. Our initial analysis showed that muscle regeneration impairment was not as evident in NSG mice treated daily with 150 μg of TNFα neutralizing antibodies compared to anti-TNFα treated NOD SCID mice; therefore, we continued using a daily administration of a low dose of the antibody (50 to 70 μg) for TNFα neutralization. Neutralizing TNFα alone in WT mice had no effect on the size of regenerating fibers (Fig. 3B). However, a delay in tissue regeneration was noted as early as 8 d p.i. in WT mice treated with neutralizing antibodies against both IFNγ and TNFα compared to the mice that were injected with only the TNFα neutralizing antibody, although the delay was small and did not reach statistical significance (Fig. 3C). These findings, together with our data showing the dispensability of IFNγ (Fig. 2C), led us to conclude that TNFα is capable of partially compensating for the absence of IFNγ during muscle regeneration. Furthermore, following TNFα neutralization, the absence of NK cells (Fig. 3A) resulted in a larger regeneration impairment compared to the absence of IFNγ per se (Fig. 3C), suggesting that besides the production of IFNγ, NK cells have other roles during muscle regeneration.

Fig. 3.

TNFα compensates for the absence of IFNγ. (A) Picrosirius red and nuclei staining and quantification of the distribution of centrally nucleated myofiber CSA of damaged muscles from NOD SCID or NSG mice subjected to TNFα neutralization (n = 8 mice per group, Brown–Forsythe and Welch ANOVA tests followed by Dunnett’s T3 multiple comparisons test on selected bins). (B) Quantification of distribution of centrally nucleated myofiber CSA in damaged muscles from WT CD45.1 mice subjected to TNFα neutralization or isotype control on frozen sections stained for laminin (n = 5 mice per group, ordinary one-way ANOVA followed by Sidak’s multiple comparisons tests on selected bins). (C) Picrosirius red and nuclei staining and quantification of distribution of centrally nucleated myofiber CSA of damaged muscles from isotype-treated control or anti-IFNγ–treated WT CD45.2 (C57BL/6J) mice subjected to TNFα neutralization (n = 5 mice per group, Brown–Forsythe and Welch ANOVA tests followed by Dunnett’s T3 multiple comparisons test on selected bins), *P < 0.05, **P < 0.01.

Conditional Inhibition of Interferon Gamma Receptor 1 Signaling in MuSCs or FAPs Does Not Impair Muscle Regeneration.

Next, we sought to identify the cellular target of IFNγ during muscle regeneration. Our data clearly showed up-regulation of IFNγ during muscle regeneration, suggesting a role for this cytokine. Moreover, impaired muscle regeneration in the absence of IFNγ has been reported by previous studies (17, 18). However, systemic inhibition of the effect of IFNγ with or without TNFα neutralization in our study did not result in significant muscle impairment. Two possible explanations exist for our observation, as follows: IFNγ marginally affects one cell population without a significant ultimate effect on the regeneration process or alternatively IFNγ targets different cell populations with opposing effects on regeneration that cancel out each other. To discriminate between these two scenarios, we investigated the role of IFNγ on different cell types in vivo by systemic as well as cell-specific conditional inhibition of IFNγ signaling. In-vitro treatment of MuSCs with a combination of proinflammatory cytokines IL1α, IL13, TNFα, and IFNγ has been shown to improve their proliferation and regenerative capacity (33). To assess the role of IFNγ and the compensatory effect of TNFα on the proliferation of MuSCs in vivo, we measured 5-ethynyl-2′-deoxyuridine (EdU) incorporation in MuSCs after injury. Our preliminary data showed no significant difference in the percentage of EdU⁺ myogenic cells in the absence of IFNγ and TNFα compared to the absence of TNFα alone (Fig. 4A). In addition, IFNγ has been shown to inhibit MuSC differentiation in vitro (34, 35). To further confirm that the regenerating effect of IFNγ is not directly through MuSCs, we crossed Ifngr^flox/null mice with Pax7^CreERT2 (homozygous Cre) to conditionally inhibit IFNG receptor 1 (IFNGR1) signaling in MuSCs. Five doses of tamoxifen were administered intraperitoneally for CRE activation, followed by 10 d of a washout period prior to inducing muscle damage. To evaluate the efficiency of the CRE-inducible system, we assessed gene copy numbers using droplet digital PCR. We found that IFNGR1 is efficiently recombined in 98.88% ± 1.2 (mean ± SD, n = 5 mice) of MuSCs at the peak of IFNγ expression (i.e., on day 3 p.i.; SI Appendix, Fig. S4A). Then, a comparison of muscle regeneration between Pax7^CreERT2 × Ifngr1^wt/null and Pax7^CreERT2 × Ifngr1^flox/null mice was performed in the presence of TNFα neutralization. No significant difference in the distribution of the size of regenerating fibers was observed (Fig. 4B), suggesting that IFNγ does not have a major direct effect on MuSC function in vivo during regeneration.

Fig. 4.

A direct effect of IFNγ on MuSCs or FAPs does not play a major role in muscle regeneration. (A) Left: Experimental design and gating strategy to assess the effect of IFNγ neutralization on MuSC proliferation; Right: Quantification of the percentage of EdU incorporated MuSCs from damaged TA muscles of mice subjected to anti-TNFα and isotype compared to mice subjected to anti-TNFα and anti-IFNγ (each dot represents one mouse, unpaired t test). (B) Laminin staining and quantification of the distribution of centrally nucleated myofiber CSA of damaged muscles in TNFα-neutralized transgenic mice with specific inhibition of IFNGR1 signaling in MuSCs compared to proper controls (n = 13 mice per group, Brown–Forsythe and Welch ANOVA tests followed by Dunnett’s T3 multiple comparisons test on selected bins). (C) Top: Experimental design and gating strategy to assess the effect of IFNγ neutralization on the number of FAPs; Bottom: Quantification of the total number of FAPs obtained per TA muscle from mice subjected to anti-TNFα and isotype compared to mice subjected to anti-TNFα and anti-IFNγ (each dot represents one mouse, unpaired t test). (D) Quantification of the distribution of centrally nucleated myofiber CSA in damaged muscles of TNFα-neutralized transgenic mice with specific inhibition of IFNGR1 signaling in FAPs compared to controls (n = 14 to 16 mice per group, Brown–Forsythe and Welch ANOVA tests followed by Dunnett’s T3 multiple comparisons test on selected bins).

Previously, we showed that TNFα derived from infiltrating macrophages is required to efficiently induce apoptosis and clearance of FAPs after their damage-induced expansion (4). Therefore, we hypothesized that IFNγ could also play a role in the clearance of FAPs. In support of this hypothesis, we observed that IFNγ has the potential to sensitize FAPs to the apoptotic effect of TNFα in vitro (SI Appendix, Fig. S4B). This in-vitro effect was counteracted by adding TGFβ1, which is in line with what we showed previously (4). To evaluate the potential of IFNγ being involved in triggering the onset of FAP apoptosis in vivo, we quantified the absolute number of FAPs obtained per TA muscle on day 8 p.i. in mice subjected to anti-IFNγ and anti-TNFα treatments compared to a control receiving anti-TNFα and an isotype control. We did not detect any difference in the number of FAPs in the absence of IFNγ (Fig. 4C), suggesting that the role of this cytokine in muscle regeneration is not facilitating the clearance of FAPs. In the skeletal muscle, PDGFRa is exclusively expressed by FAPs (3). Therefore, we used a Pdgfra^CreERT2 system (heterozygous Cre) to exclude any potential effect of IFNγ on FAPs in vivo. The efficiency of the recombination of the IFNGR1 allele observed in Pax7^CreERT2 × Ifngr1^flox/null mice encouraged us to simply use the Pdgfra^CreERT2 × tdTomato lineage tracing system to assess the recombination efficiency. As expected, 95.6% ± 2.16 (mean ± SD, n = 3 mice) of FAPs were efficiently labeled on day 3 p.i. (SI Appendix, Fig. S4C). We then investigated muscle regeneration in Pdgfra^CreERT2 × Ifngr1^flox/null mice compared to Pdgfra^CreERT2 × Ifngr1^wt/null mice following tamoxifen induction and TNFα neutralization in both groups. Conditional inhibition of IFNGR1 signaling in FAPs did not have a significant effect on muscle regeneration (Fig. 4D), suggesting that any direct effect of endogenous IFNγ on FAPs does not play a major role in muscle regeneration.

IFNγ and TNFα Act through Macrophages to Support Muscle Regeneration.

Having excluded direct effects of IFNγ on MuSCs and FAPs, the two main players of muscle regeneration, we asked whether IFNγ plays its role through the other key player—macrophages. The number of inducible CRE systems to specifically target macrophages is limited. A widely used animal model for a temporal control of the transgene KO/activation in macrophages is Cx3cr1^CreERT2 (36). CX3C chemokine receptor 1 (CX3CR1) is the receptor for the chemokine CX3CL1, and it is exclusively expressed in the mononuclear phagocyte system with minimal expression in other cell types. We used a Cx3cr1^CreERT2 × tdTomato lineage tracing system (heterozygous Cre) to further characterize Cx3cr1-expressing cells in muscle regeneration. We first assessed the proportion of macrophages that are targeted using the system on day 3 p.i. Our standard regimen of five doses of tamoxifen administration followed by 10 d of a washout period before muscle injury labeled only 1.87% ± 0.90 (mean ± SD) of macrophages on day 3 p.i. (Fig. 5A). This is consistent with the rapid replacement of recombined Cx3cr1⁺ blood monocytes by the progeny of unrecombined Cx3cr1⁻ bone marrow progenitors. Therefore, we further tested a shorter washout period of 3 d, which resulted in the labeling of 56.80% ± 5.98 (mean ± SD) of macrophages. A similar recombination efficiency was achieved when muscle injury was induced while tamoxifen was administered continually on a daily basis (Fig. 5A). Therefore, we switched to a 3-d washout period regimen to target genes of interest in macrophages using Cx3cr1^CreERT2 mice. Further characterization of inflammatory cells in regenerating muscle showed that eosinophils and neutrophils do not express Cx3cr1 (SI Appendix, Fig. S5A). As expected, CD11B⁺LY6G⁻SIGLECF⁻ macrophages constitute 93.7% ± 1.02 of tdTomato⁺ cells with NK cells and T cells accounting for only 0.16% ± 0.05 and 0.17% ± 0.06, respectively (SI Appendix, Fig. S5B), confirming that Cx3cr1 expression in regenerating muscle is mainly limited to macrophages. Then, we evaluated the potential role of IFNγ on macrophages in muscle regeneration using Cx3cr1^CreERT2 × Ifngr1^flox/null mice compared to Cx3cr1^CreERT2 × Ifngr1^wt/null mice in the presence of TNFα neutralization. Specific inhibition of IFNGR1 signaling in macrophages did not result in muscle regeneration impairment (Fig. 5B). A caveat is that this could be due to the low efficiency of the Cx3cr1^CreERT2 system in inhibition of IFNGR1 signaling in damage-induced macrophages and to the potential replacement of recombined cells by unaffected progenitors during the later phase of regeneration.

Fig. 5.

Evaluation of the direct effect of IFNγ on macrophages during muscle regeneration using the Cx3cr1^CreERT2 system. (A) Evaluation of different tamoxifen strategies to target damage-induced macrophages using the Cx3cr1^CreERT2 system. A representative flow cytometry plot is shown on the Left. Quantification of the percentage of TdTomato (TdTom)⁺ macrophages on day 3 p.i. is shown on the Right (each dot represents one mouse). (B) Quantification of the distribution of centrally nucleated myofiber CSA of damaged muscles in TNFα-neutralized transgenic mice with specific inhibition of IFNGR1 signaling in macrophages compared to proper controls (n = 14 mice per group; Brown–Forsythe and Welch ANOVA tests followed by Dunnett’s T3 multiple comparisons test on selected bins).

Considering the well-known effects of IFNγ on bone marrow–derived macrophages in vitro, where it strongly polarizes them toward a proinflammatory fate, the lack of effects of IFNGR1 inhibition in vivo was highly surprising. We reasoned that this may be due to the relatively low efficiency of the CRE system and decided to further investigate the potential effect of IFNγ in individual macrophages. To this end, we sorted PI⁻Hoechst⁺CD45⁺CD11B⁺LY6G⁻SIGLECF⁻ macrophages on day 5 and 6 p.i. from anti-TNFα–treated Cx3cr1^CreERT2 × Ifngr1^flox/null and Cx3cr1^CreERT2 × Ifngr1^wt/null mice for single-cell RNA sequencing. Considering that the level of IFNγ is highest on days 3 and 4 p.i., the two close time points (day 5 and day 6 p.i.) were selected to increase the chance of detecting the maximum differences between the two experimental groups in at least one of the time points. Datasets were pooled for Louvain clustering and dimensionality reduction, leading to the identification of 12 clusters (Fig. 6A). There was no noticeable difference in the size of clusters between Ifngr1^flox/null and Ifngr1^wt/null samples (Fig. 6A). Gene ontology analysis of differentially expressed genes in each cluster showed that clusters 7 and 8 mainly included metabolically active and proliferating cells (SI Appendix, Fig. S6A). Cells in cluster 9 highly expressed Ly6c2 (SI Appendix, Fig. S6B) and were enriched in pathways related to cell migration (Arp2/3 complex-mediated actin nucleation) and extravasation, while cells in cluster 1 expressed high levels of genes related to respiratory burst, glycolysis, and phagocytosis (SI Appendix, Fig. S6A), suggesting that these two clusters were newly recruited macrophages. Cells in cluster 4 expressed genes related to IFNα and IFNβ signaling and antiviral immunity. They also uniquely expressed Cxcl10 (SI Appendix, Fig. S6B) and therefore correspond to the IFN-responsive macrophages recently reported by Zhang et al. (18). Clusters 2 and 3 were enriched in pathways related to antigen presentation as well as the proliferation/activation of lymphocytes. Cells in cluster 3 were the only cells expressing Cd209a (SI Appendix, Fig. S6B), a gene primarily expressed by antigen-presenting cells (37). A large number of differentially up-regulated genes in cluster 0 were genes encoding transcription regulators, including Fos, Fosb, Jun, Junb, Mafb, Atf3, Cebp, Egr1, Btg2, and Nr4a1, suggesting that these cells are likely undergoing phenotypic reprograming. Cells in cluster 5 were enriched for pathways related to clathrin-dependent endocytosis. Up-regulated pathways in cluster 6 were limited to cytoplasmic translation. Cluster 10 also expressed genes related to antigen presentation and genes specific to T cell activation/proliferation. A small population of T cells was found in cluster 11, which likely arose from contamination during cell sorting.

Fig. 6.

Single-cell analysis of damage-induced macrophages in the absence of IFNγ signaling. (A) Uniform manifold approximation and projection plot of muscle macrophages on day 5 and 6 after injury. (B) Violin plots depicting the expression of key differentially expressed genes of specific clusters in Ifngr1^flox/null compared to control. (C) Violin plots depicting the expression of key differentially expressed genes considering all clusters together in Ifngr1^flox/null compared to control. (D) Violin plots depicting expression of Tsc22d3 of specific clusters in Ifngr1^flox/null compared to control.

Next, we compared the expression profiles between the two experimental groups. We focused on genes that have been recently shown to be differentially expressed by macrophages at different time points following sterile muscle injury (38). We found that genes including Ly6e, Irf7, Ms4a4c, Chil3, Spp1, and Cxcl2 that are expressed by macrophages at earlier time points of regeneration (day 0 to day 3.5 p.i.) (38) as well as the proinflammatory gene Il1β were significantly up-regulated in Ifngr^flox/null mice (mainly in cluster 0 that includes cells undergoing phenotypic reprograming) compared to controls (Fig. 6B). A subset of these genes was also expressed at higher levels globally in Ifngr^flox/null mice (Fig. 6C). Interestingly, we also found a higher expression of Tsc22d3, a transcriptional regulator that mediates immunosuppressive effects of corticosteroids (39, 40), in several clusters in control mice compared to Ifngr^flox/null (Fig. 6D). To focus on the cells directly impacted by the transgenic system, we focused on Cx3cr1⁺ cells for further analysis. Genes related to IFNα and/or IFNβ signaling such as Ifi204, Ifit2, Ifit3, and Ifi47, as well as genes related to macrophages at early phase (day 0 to day 3.5 p.i.) of muscle regeneration including Irf7, Ly6e, Spp1, Ms4a4c, Ctsl, Chil3, and Il1rn (38), comprised a large proportion of genes that were differentially up-regulated in Cx3cr1⁺ cells from Ifngr1^flox/null compared to Ifngr1^wt/null on day 5 p.i. On the other hand, out of eight genes that were differentially up-regulated in Ifngr1^wt/null compared to Ifngr1^flox/null mice on day 5, three genes (Tmem176b, H2-Eb1, Cd81) were related to the macrophage signature at day 10 p.i (38).

Although single-cell RNA sequencing provides a higher resolution of cellular differences and a better understanding of the function of an individual cell, it has limitations including a bias of transcript coverage and low sequencing depth (41). Furthermore, the Cx3cr1^CreERT2 system targets only half of macrophages in damaged muscle. Therefore, to better understand the effect of TNFα and IFNγ on macrophages, we performed bulk RNA sequencing of macrophages sorted from damaged muscle of Ifngr^null/null or Ifngr^wt/null mice in the presence of the neutralizing anti-TNFα antibody or the isotype control. Surprisingly, neither genotype nor treatment resulted in defined transcriptomic differences among samples following the initial principal component analysis (PCA; Fig. 7A, Left). Throughout the project, we noticed differences in response to muscle damage across mouse litters of the same colony. As such, we performed batch normalization (42) to correct for differences attributed to litters, which resulted in a clearer distinction of samples (Fig. 7A, Middle). Interestingly, this revealed that the absence of TNFα and/or IFNγ receptor signaling had minimal effect on specifically one of the litters (litter D; Fig. 7A, Middle), perhaps as the result of an unidentified cage-level stressor event. Indeed, when this litter was excluded from the analysis of differentially expressed genes, the distinction between samples in the PCA plot was significantly improved, as was the difference between the principal components (Fig. 7A, Right). The absence of TNFα has larger effect (85% of the variation) on macrophages compared to the absence of IFNγ receptor signaling (9% of the variation), although the smaller effect of IFNγ receptor signaling on macrophages could result from prenatal genetic compensations, a common phenomenon in response to gene KOs (43). We initially focused on genes that were consistently differentially expressed between Ifngr^null/null and Ifngr^wt/null mice in both the absence of anti-TNFα antibody (comparing isotype-treated Ifngr^null/null with isotype-treated Ifngr^wt/null mice) and the presence of anti-TNFα antibody (comparing anti-TNFα-treated Ifngr^null/null with anti-TNFα-treated Ifngr^wt/null mice). Gene ontology analysis points to genes (Igtp, Gbp3, Gbp4, Gbp5, Gbp6, Gbp8, Gbp9, Nlrc5) related to the cellular response to IFNγ (Fig. 7B) that were significantly down-regulated in Ifngr^null/null compared to Ifngr^wt/null mice. This suggests that IFNγ signaling is active in macrophages following muscle injury. Then, we evaluated the genes that were consistently differentially expressed following TNFα inhibition in both Ifngr^null/null and Ifngr^wt/null mice. Apart from genes related to the MHC class II protein complex that have been shown to be expressed by macrophages at a later phase of regeneration (44), the absence of TNFα paradoxically resulted in the acquisition of a proinflammatory phenotype as shown by the up-regulation of genes related to IL1/TNFα-mediated signaling (Il1a, Il1b, Tnfa, Traf1, Traf6, Traf3ip3, Il1r2, Egr1, Rps6ka5, Myd88, Traf6, Vrk2), leukocyte chemotaxis (Ccl2, Ccl3, Ccl4, Ccl5, and Ccl7), Toll-like receptor signaling (Tlr3, Tlr9), and inflammasome complex (Aim2, Casp4, Gbp5, Mefv, Naip5, Nlrc4, Nlrp1b, Nlrp3, and Myd88) (Fig. 7B). Furthermore, genes related to the collagen metabolic process and extracellular matrix organization, including Prolyl 3-hydroxylases (P3h1, P3h2), Lysil oxidases (Lox, Loxl1, and Loxl2), and matrix metalloproteinases (Mmp2, Mmp9, Mmp12, and Mmp14), were down-regulated in macrophages in the absence of TNFα (Fig. 7B). Together, our single-cell and bulk RNA sequencing data suggest that TNFα and IFNγ cooperate for a proper pro- to anti-inflammatory transition of infiltrating macrophages during muscle regeneration.

Fig. 7.

TNFα and IFNγ cooperate for phenotype transition of macrophages during muscle regeneration. (A) Bulk RNA sequencing and PCA analysis of damage-induced macrophages in the absence of TNFα and/or IFNγ signaling. Letters refer to different mouse litters. (B) Heatmap showing differentially expressed genes (DEGs) categorized by their annotations to Gene Ontology pathways.

Discussion

Recent evidence suggested a role for NK cells in liver regeneration and fibrosis where hepatic NK cells have been shown to promote liver regeneration after partial hepatectomy (45), likely through the stimulation of hepatocyte proliferation or hepatic macrophage activation (46). On the other hand, a paradoxical role has been proposed for NK cells in liver fibrosis. While they can exert a beneficial antifibrotic effect by killing myofibroblasts (47, 48), they may also enhance liver injury and fibrosis by killing hepatocytes (49). In contrast to the liver, the finding of NK cells infiltrating damaged muscle (another tissue with regenerative capacity) has only been sporadically reported, and the role of NK cells in muscle regeneration following a sterile injury is not well understood. Here, we found that although appreciable numbers of NK cells invade damaged muscle, their absence has no significant effect on muscle regeneration following acute damage or in the setting of chronic repetitive injury in Dmd^mdx mice.

The presence of IFNγ in regenerating muscle has been reported over a decade ago. Macrophages (8, 17), myoblasts (17), NK cells (17, 44), and T cells (8, 17, 18, 44) have all been proposed as a source of IFNγ in regenerating muscle. Our data showed that NK cells found in damaged muscle are the major source of IFNγ. Besides this cell population, IFNγ production is limited to a small population of T cells. However, our data suggested that NK cells are required for efficient IFNγ production by T cells in muscle regeneration. This is in line with the proposed immunomodulatory role of NK cells on T cells (50).

Up-regulation of TNFα in damaged muscle has been documented by several studies with contradicting results regarding its potential role in muscle regeneration. Two different studies showed impaired muscle regeneration and/or significant reductions in isometric muscle strength in TNFα receptor KO mice (9, 10). On the contrary, Collins et al. (51) saw no impairment in muscle regeneration in the absence of TNFα alone or TNFα and its homolog LT-α. In agreement with the study by Collins et al. (51), we also detected no muscle impairment following the neutralization of TNFα. Similar to TNFα, a few studies had suggested that IFNγ is a required positive factor for muscle regeneration (17, 18), while other studies showed it has a deleterious effect on muscle function (19). Here, we saw no impairment in muscle regeneration in the absence of IFNγ. However, our further analysis revealed that TNFα compensates for the absence of IFNγ in muscle regeneration and the role of IFNγ becomes more evident in the absence of this compensatory effect.

TNFα has been proposed to be involved in both the proliferation and differentiation of MuSCs following muscle injury (52, 53). It has also been shown to be required for apoptosis of FAPs following damage-induced expansion (4). Similar to TNFα, MuSCs have been proposed as the main target of IFNγ (17, 19, 33, 35). However, using efficient in vivo systems to specifically inhibit IFNγ signaling in MuSCs or FAPs, we found that direct effects of endogenous IFNγ on MuSCs or FAPs do not play a major role in muscle regeneration.

Common models of muscle injury include freeze injury, barium chloride, NTX, and cardiotoxin. The last three models, in particular, are more similar regarding the extent of muscle injury and the duration of regeneration process (54) in that, within 4 d of injury, the majority of necrotic myofibers are cleared and some regenerating fibers have been formed (54). As reported by others and us (4, 55), following NTX injury, a large population of monocytes infiltrates the damaged tissue a few hours after muscle injury. Initially, this population expresses proinflammatory cytokines such as TNFα and IL1β. On day 2 to 4 after injury, these cells switch to regenerative, anti-inflammatory macrophages that express high levels of TGFβ and IL10. This switch is critical for efficient muscle regeneration. IFNγ has been considered as a potential inducer of the proinflammatory phase acting within 24 h of regeneration (56). This notion mainly stems from the historical classification of M1 (proinflammatory) and M2 (anti-inflammatory) macrophages, based on early in vitro data as well as subsequent in vivo studies in models of infection (57). Specifically, M1 macrophages, can be obtained in vitro from bone marrow–derived monocytes by treatment with Toll-like receptor ligands (such as lipopolysaccharide) and IFNγ (58). Due to its simplicity, this classification has been widely used to characterize macrophages in different tissues and pathologies including muscle regeneration. However, more recent evidence calls into question the accuracy of this classification system in vivo (6, 59–61). Here, to our surprise and against commonly held notions (56, 62), we found that IFNγ is not present in the environment during the first 2 d following sterile injury when the majority of macrophages present in the tissue are proinflammatory. Instead, it peaks on day 3 when inflammation resolution occurs. This is consistent with static expression of transcripts associated with IFNγ-STAT1 signaling in macrophages sorted from damaged muscle after 24 h of injury (60). More recently, IFNγ is viewed as a pleiotropic regulator of inflammation rather than just an inducer of inflammatory response (63). Our data further support an immunomodulatory role of IFNγ in muscle regeneration.

Recently, Zhang et al. (18) described a population of damage-induced, interferon-responsive macrophages that decline with aging. They showed that this decline resulted in poor regeneration of aged muscle mainly due to the absence of CXCL10, which is required for efficient MuSC proliferation. In young mice, however, the absence of CXL10 has no effect on regeneration, likely due to the compensatory effect of CXCL9 (18, 64). We also found a similar population in our single-cell RNA sequencing data and showed that the absence of IFNγ results in a significantly lower expression of Cxcl10 by these cells. However, we found the absence of IFNγ has a broader effect on macrophages in particular on a subset that is actively undergoing reprograming. The transition of these cells toward an anti-inflammatory phenotype was delayed, and inflammatory markers related to the early phases of regeneration were increased, including the proinflammatory cytokine Il1β. An increased expression of Il1β as well as Il1α was also noted in the absence of TNFα in our bulk RNA sequencing of macrophages, potentially representing a compensatory mechanism. While we found that TNFα can compensate for the absence of IFNγ, it is possible that IL1 family cytokines can also partially compensate for the absence of both TNFα and IFNγ. As such, the inhibition of these cytokines may result in further regeneration impairment. The compensatory effect of these cytokines could also explain the conflicting results reported by different studies evaluating the isolated role of these cytokines in muscle regeneration using a variety of transgenic mouse and injury models.

It is currently understood that proinflammatory cytokines including TNFα and IL1 can locally increase the bioavailability of active glucocorticoids by increasing the expression of local 11β-hydroxysteroid dehydrogenase type 1 (11β-HSD1), an enzyme that regulates intracellular concentrations of glucocorticoids (65). Although we did not see any difference in the expression of Hsd11b1, we detected a significant decrease in the expression of Tsc22d3 (also known as glucocorticoid-induced leucine zipper [Gilz]) by infiltrating macrophages in the absence of IFNγ signaling in our single-cell RNA sequencing data. As a glucocorticoid-inducible molecule, Gliz plays an important role in immunosuppression (39, 40). The role of endogenous glucocorticoids on macrophages during tissue regeneration has drawn attention recently (66). It has been shown that the absence of glucocorticoids receptors (GRs) from myelomonocytic cells results in a poor prognosis in a mouse model of myocardial infarction (67). Furthermore, murine macrophages with impaired GR signaling fail to acquire the anti-inflammatory phenotype in inflammatory bowel disease (68). Although the role of endogenous glucocorticoids in modulating the phenotype of macrophages in vivo during muscle regeneration has not been studied, annexin A1, which mediates the anti-inflammatory effects of glucocorticoids, has been shown to drive the transition to anti-inflammatory macrophages during muscle regeneration (69).

In summary, our study elucidated the cellular sources, kinetics, and cellular targets of IFNγ during muscle regeneration. We showed that during muscle regeneration following sterile damage, the production of cytokines traditionally characterized as proinflammatory, TNFα and IFNγ, is timely regulated to cooperate for an efficient pro- to anti-inflammatory transition of macrophages, emphasizing the need for a review of their role during sterile inflammation.

Materials and Methods

Mice and Induction of Muscle Damage.

All experimental procedures were performed in compliance with the University of British Columbia Animal Care Committee regulations. Details regarding mice strains and suppliers can be found in SI Appendix, SI Materials and Methods. Sex-matched littermates were assigned to different experimental groups wherever possible.

Muscle Injury, Tamoxifen Induction, and Neutralizing Antibody Administration.

Daily intraperitoneal injection of 3 mg of tamoxifen (Sigma-Aldrich, # T5648-5G,) was conducted in order to activate the Cre/Lox system. A total of 0.1 or 0.15 μg of NTX (Latoxan) was injected into the TA muscle to induce damage. 250 μg of the NK1.1 neutralizing antibody (AbLab, clone: PK136, # 21-0034-50) was administered intraperitoneally to deplete NK cells. Mouse IgG2a isotype control (AbLab, clone: BR24, # 21-0013-16) or mouse IgG2a anti-human CD3 antibody (AbLab, clone: OKT3, # 21-0058-96) was used as a control. To neutralize TNFα, 50 to 70 μg of neutralizing antibodies (AbLab, clone: XT31.1, # 21-0074-16) was injected intraperitoneally for 7 d starting on the day of injury. To neutralize IFNγ, 250 μg of neutralizing antibodies (AbLab, clone: XMG1.2, # 21-0046-10) was injected intraperitoneally for 6 d, starting 24 h after injury. Rat IgG1 isotype control (AbLab, clone: 3B2, # 21-0045-50) was used as a control.

Tissue Preparation for Flow Cytometry.

Details regarding muscle preparation for flow cytometry can be found in SI Appendix, SI Materials and Methods. The antibodies used are summarized in SI Appendix, Table S1.

ELISA for TNFα and IFNγ.

TA muscles were cut into small pieces and submerged in lysis buffer (20 mM Tris [pH 7.8], 137 mM NaCl, 2.7 mM MgCl₂, 1% Triton x-100, 10% glycerol, 1 mM EDTA, 1 mM DTT, 1% protease inhibitor). The samples were homogenized using metal beads with TissueLyser II (Qiagen) for 12 min at 20 Hz followed by 15 cycles of medium-intensity sonication in a Bioruptor Plus sonication device at 4 degrees. Following homogenization, samples were centrifuged at 10,000 g for 10 min at 4 degrees. The supernatant was used for protein quantification using the Pierce bicinchoninic acid (BCA) Protein Assay Kit (Thermo Scientific #23228) and ELISA using mouse IFNγ ELISA Ready-SET-Go (eBioscience, #88-7314-22) and mouse TNFα ELISA Ready-SET-Go (eBioscience, #88-7324-22) according to the instructions.

Immunofluorescence and Picrosirius Red Staining.

Details regarding staining of histology sections can be found in SI Appendix, SI Materials and Methods.

Ex Vivo Muscle Function Evaluation.

The contractile properties of diaphragm strips were evaluated using the 1300A: 3-in-1 Whole Animal System – Mouse (Aurora Scientific). Details can be found in SI Appendix, SI Materials and Methods.

In Vitro Treatment of FAPs.

FAPs were sorted from TA muscles at 3 d p.i. A total of 5,000 cells/well were cultured in Dulbecco’s modified Eagle’s medium (10% fetal bovine serum) + bFGF (2.5 ng/mL, Gibco, # 13256-059) in a 96-well plate. On day 3 of culture, cells were treated with a combination of IFNγ (200 ng/mL, R&D Systems, # 485-MI-100), TGFβ (43.75 ng/mL, eBioscience, # 14-8348-62), and/or TNFα (Sigma-Aldrich, # T7539). Forty-eight hours after treatment, cells were fixed with 4% paraformaldehyde, stained with Hoechst, and counted by a Cellomics ArrayScan VTI HCS Reader (Thermo Scientific).

Proliferation Assay.

To assess the proliferation rate of MuSCs in vivo, 0.5 mg of EdU (Invitrogen, # E10415) was administered intraperitoneally twice (12 h apart) a day before muscle harvest. After muscle digestion and preparation, samples were stained using Click-iT Plus EdU Pacific Blue Flow Cytometry Assay Kit (Invitrogen, # C10636) according to the manufacturer’s manual. Details can be found in SI Appendix, SI Materials and Methods.

Single-Cell and Bulk RNA Sequencing and Analysis.

Details regarding RNA sequencing can be found in SI Appendix, SI Materials and Methods.

Statistical Analysis.

All statistical analysis was done using GraphPad Prism 9. Rationales regarding using different tests can be found in SI Appendix, SI Materials and Methods. Throughout the manuscript a P value of <0.05 was considered significant except for the analysis of single-cell data to detect differentially expressed genes in which an adjusted P value of <0.01 was considered significant.

Data, Materials, and Software Availability

RNA sequencing data have been deposited in GEO (GSE212373) (70). All study data are included in the article and/or SI Appendix.