The ischemic tissue response consists of distinct vascular, metabolic and inflammatory components organized in a spatio-temporal fashion, which can be analyzed in the HLI model (9). To further characterize the regional ischemic response in the upper vs. the lower limbs we performed comparative analysis of the proximal semimembranosus (SM) and distal tibialis anterior (TA) muscles. Following HLI, hypoxia-inducible factor Hif1a expression increased rapidly and significantly only in the distal TA muscle, but not in the proximal SM muscle, which never increased to levels above contralateral limb ( Figure 1A ), suggesting a hypoxic niche in the distal limb muscle, but not in the proximal limb muscle of the same leg. The rapid induction of hypoxia in TA muscle was accompanied by robust but transient recruitment of monocytes, which differentiate into a sustained population of macrophages (8) ( Table 1 ), while both populations showed only moderate and transient increases in SM muscle ( Figure 1B ). By histology, few CX₃CR1⁺ monocytes and CX₃CR1⁺F4/80⁺ macrophages were detected in between muscle fibers or surrounding collateral arteries (α-SMA⁺) in ischemic SM (iSM). In comparison, while the pattern of localization was maintained in ischemic TA muscle (iTA), cell abundance was remarkably higher ( Figure 1C ). Interestingly, monocytes and macrophages from iTA also expressed higher levels of cyclin-dependent kinase Cdk2 ( Figure 1D ), suggesting higher proliferation activity in the ischemic niche.

To analyze the associated growth factor milieu, we measured key myeloid growth factors. Expression of macrophage colony-stimulating factor Csf1 increased rapidly after induction of HLI and maintained high levels until d3, paralleling the peak in monocyte recruitment and the development of ischemic macrophages, while expression of Csf1 in iSM did not increase over time ( Figure 1E ). Protein levels of CSF-1 in iTA showed similar dynamics, with an expression peak on d3, but serum levels of CSF-1 did not change significantly over time ( Figure 1F ), indicating local ischemic production. In contrast, levels of myeloid growth factors CSF-2 and IL-34 remained low in iTA and serum throughout the time course after HLI ( Figure 1F ). Thus, the ischemic muscle niche is characterized by specific expression of Csf1, suggesting an important role in regulating the ischemic inflammatory response. Interestingly, dynamics of CSF-1 on gene expression and protein levels was correlating with the dynamics of Ccl2 expression ( Figure 1G ), and both were coinciding with the peak of Ly6C^hi monocytes and the development of macrophages (d3). These data indicate an early and specific ischemic niche response, promoting accumulation of Ly6C^hi monocytes and their conversion into macrophages (8).

To determine the muscle-resident cell types producing CSF-1 during ischemia we performed cell sorting and gene expression analysis, employing cell type specific markers CD45 and CD31 in combination with podoplanin (PDPN), a marker for stromal cell populations (39). We thus defined vascular EC (Ter119^-, CD45^-, CD31⁺), lymphatic EC [Ter119^-, CD45^-, CD31⁺PDPN⁺, i.e., double positive (DP)], PDPN⁺ stromal cells (Ter119^-, CD45^-) and double negative (DN) CD31^-PDPN^- cells ( Figure 2A ). In gene expression analysis, PDPN⁺ stromal cells showed the highest expression of Csf1, followed by lower expression in DP cells, while vascular EC and DN cells did not show relevant levels of expression. Furthermore, PDPN⁺ stromal cells expressed high levels of Vim, Col1 and Pdgfra, but low levels of Pdgfrb, consistent with fibroblasts (40) ( Figure 2B , Supplementary Figure 1A ). Interestingly, PDPN surface expression on CD45^- cells isolated from iTA muscle increased during ischemia ( Figure 2C ). To characterize PDPN⁺ cell types we performed tissue staining and fluorescence microscopy. Consistent with a stromal fibroblast population, PDPN⁺ cells surrounded blood vessels and were scattered in between muscle fibers (interfiber space, IFS) ( Figure 2D ), were found around and within nerve fibers ( Figure 2E ) or were co-expressing LYVE-1/VEGFR3 in tubular structures ( Figure 2F ), consistent with lymph vessels. In line with flow cytometry data, microscopy also demonstrated upregulation of PDPN expression in ischemic tissue ( Figure 2D versus Figures 2E, F ). Along with CD45^- cells, PDPN was also found on CD45⁺ cells ( Figure 2D ). Furthermore, spatial analysis revealed that CSF-1 expression co-localized within PDPN⁺ cells located in the proximity of blood vessels in iTA muscle, while no expression was detected in contralateral TA muscle ( Figure 2G ). Expression of PDFGRα and PDFGRβ also were found in perivascular (PV) PDPN⁺ cells ( Figure 2H ). However, after separation of PDPN⁺PDFGRα⁺ and PDPN⁺PDFGRα^- subpopulations ( Supplementary Figure 1B ), both demonstrated Csf1 expression, which excluded PDFGRα as a general marker for definition of CSF-1 producing PDPN⁺ cells ( Figure 2I ).

Along with Csf1, ischemia also induced expression of Ccl2, a mediator of monocyte recruitment. Although PDPN⁺ cells showed the highest level of Ccl2 expression, EC, LEC and DN cells also showed detectable levels ( Supplementary Figure 1C ). Of note, GFP⁺CD11b⁺ cells isolated from Cx3cr1^gfp/+ reporter mice ( Figure 2A ), comprising monocytes and macrophages (8) expressed inflammatory cytokines, Tnfa, Il1b and Il6, but also expressed Ccl2 ( Supplementary Figure 1D ), potentially contributing to recruitment of circulating monocytes. In contrast, these cells did not express relevant levels of Csf1 ( Supplementary Figure 1D ). Separating PDPN⁺ cells into PDGFRα cell subsets revealed only quantitative, but not principle, differences in Ccl2 expression ( Supplementary Figure 1E ).

To determine the functional role of CSF-1 in the ischemic niche we next inhibited CSF-1 actions. CSF-1 signals through cFMS receptor tyrosine kinase (41), which is inhibited by the specific cFMS receptor tyrosine kinase inhibitors GW2580 (42). Per os treatment with GW2580 ( Figure 3A ) resulted in delayed normalization of TA muscle edema, measured as muscle mass ( Figure 3B ), and delayed distal limb perfusion recovery by Laser Doppler imaging (LDI, Figure 3C ), while Csf1 expression in GW2580 treated mice was slightly increased ( Supplementary Figure 2A ). To quantify muscle damage and regeneration we analyzed three distinct features of muscle fibers: shape, structure and nucleus, enabling to distinguish: 1) intact fibers, with oval shape, homogenous structure and peripheral nucleus; 2) regenerative fibers with oval shape and centered single, or polarized multiple nuclei and 3) disintegrated fibers with irregular shape and disintegrated structure ( Supplementary Figure 2A ). Compared to control mice, GW2580 treated mice showed lower numbers of regenerative fibers and significantly higher frequencies of disintegrated fibers, suggesting reduced debris clearance and regeneration with CSF-1 inhibition ( Figure 3D ). This was associated with strong changes in myeloid cell dynamics. The post-ischemic rise in Ly6C^hi monocytes in peripheral blood (PB) and muscle, and differentiation of muscle monocyte into ischemic macrophages observed in control mice (7–9) were severely impaired in GW2580 treated mice. Also, the population of Ly6C^lo monocytes practically disappeared with GW2580 treatment from all tissue compartments ( Figure 3E , Supplementary Figures 3B, C ). In contrast, while initial recruitment of neutrophils was unchanged in GW2580 treated mice, their persistence in ischemic tissue, but not in PB and spleen, was prolonged ( Figure 3E , Supplementary Figures 3B, C ), reflecting prolonged inflammation. Notably, F4/80⁺ splenic macrophages were also reduced after GW2580 treatment, underscoring the role of CSF-1 signaling in sustaining resident macrophage populations ( Supplementary Figure 3C ).

In line with reduced frequencies of monocytes and macrophages in iTA and their ability to produce CCL2 ( Supplementary Figure 1D ), expression of Ccl2 in muscle of GW2580 treated mice 3 days after HLI was significantly reduced ( Figure 3F ).

To characterize monocyte/macrophage phenotype changes resulting from CSF-1 inhibition, we sorted GFP ⁺ CD11b⁺ cells from iTA muscle at d3 as described ( Figure 2A ) and performed gene expression analysis. CSF-1 inhibition resulted in strongly reduced expression of Ccl2 ( Figure 3G ), which was in line with downregulation of Ccl2 expression in iTA tissues at the same time point ( Figure 3F ). CSF-1 is known to stimulate glucose uptake by macrophages (43), and glucose uptake is increased in the ischemic muscle, to which Ly6C^hi monocytes and macrophages are significant contributors (9). We therefore analyzed expression of genes involved in glucose uptake and metabolism. Treatment with GW2580 significantly reduced expression of metabolic genes Hk1, Pkm1, Ldh1 and Slc2a1, encoding glucose transporter GLUT1, in GFP ⁺ CD11b⁺ cells ( Figure 3G ), suggesting reduced metabolic activity. Although Pkm2 was highly expressed, it remained unchanged upon inhibition of CSF-1 signaling. Overall, these data demonstrate impaired metabolic adaptation and reduced migratory activity of monocytes/macrophages with CSF-1 inhibition.

Since GW2580 treatment may have off-target effects on other growth factor receptors we next employed CSF-1 blockade with specific neutralizing antibody injected subcutaneously into the ischemic limb ( Figure 3H ). Compared to control treatment, mice treated with CSF-1-neutralizing antibodies showed slightly increased Csf1 expression in ischemic muscle ( Supplementary Figure 3D ). Overall, myeloid cell dynamics recapitulated the findings in GW2580 treated mice; namely, strongly reduced numbers of post-ischemic Ly6C^hi monocytes and macrophages, systemic reduction of Ly6C^lo monocytes, increased Csf1 expression but reduced expression of Ccl2 in ischemic muscle of aCSF-1-treated mice ( Figure 3I, J , Supplementary Figures 3D–F ). This confirms specificity of the observed effects for CSF-1 inhibition. To conclude, CSF-1 inhibition impairs the differentiation of ischemic macrophages, which is associated with impaired tissue perfusion and muscle regeneration. This indicates that instructive signals for macrophage differentiation play an essential role in perfusion recovery and tissue regeneration after ischemia.

Notch signaling regulates macrophage differentiation and maturation from Ly6C^hi monocytes during ischemia, which is regulated by Notch ligand Delta-like 1 (Dll1) expressed by arterial EC. In fact, Dll1-deficient mice show impaired arteriogenesis, increased numbers of macrophages with immature differentiation profile and decreased ischemic muscle regeneration (8, 38). Muscle Dll1 expression is upregulated upon induction of ischemia, reaching its peak at day 3, which coincides with the temporal pattern of ischemic macrophages differentiation from infiltrating Ly6C^hi monocytes (8) and the induction of CSF-1 production ( Figures 1E, F ). To test whether Dll1/Notch and CSF-1 have non-redundant functions, we performed experiments in Dll1^LacZ/+ haploinsufficient mice (38, 44). Compared to wild-type (WT) controls, Dll1^LacZ/+ mice showed prolonged perfusion defects after HLI ( Figure 4A ), as described previously (38). This was associated with increased expression of Hif1a and its target gene Hmox1 at d1 after HLI, consistent with a pronounced hypoxic response ( Figure 4B , Supplementary Figure 4A ) (45). Furthermore, Dll1-haploinsufficient mice also showed higher expression of Csf1 in iTA muscle ( Figure 4B ). This was paralleled by increased cytokine expression ( Supplementary Figure 4B ) and a pronounced cellular inflammatory response ( Figure 4C ). At the same time, numbers of live ECs were significantly and persistently reduced in Dll1^LacZ/+ mice, suggesting impaired neoangiogenesis ( Figure 4C ).

Ischemic macrophages express CD11c during maturation in vivo, which is associated with ischemic muscle repair (8). Despite higher levels of CSF-1, macrophages in Dll1^LacZ/+ mice showed impaired expression of CD11c, but normal expression of the general macrophage marker I-A (MHC class II). These changes were also seen in muscle infiltrating Ly6C^hi monocytes, but to a lesser extent ( Figure 4D ). However, Ly6C^hi monocytes showed a reduction of I-A⁺ and CD11c⁺ cell frequencies ( Figure 4E ). This indicates intact initial macrophage lineage differentiation, but impaired maturation in the absence of Dll1. Furthermore, changes in monocytes and macrophage phenotypes were reflected in a pro-inflammatory expression profile in the early post-ischemic phase, characterized by significantly increased expression of Il1b, Tnfa, Il6 and Tgfb in ischemic muscle of Dll1^LacZ/+ mice after HLI ( Supplementary Figure 4A ), while no significant difference was found in upregulation of Ccl2 ( Figure 4B ). Frequencies of DC and Ly6C^lo monocytes were identical between the two groups ( Figure 4C , Supplementary Figure 4B ) and no differences in cell frequencies in peripheral blood and spleen were found between WT and Dll1 haploinsufficient mice ( Supplementary Figure 4C ). Notably, the number of PDPN⁺ cells and subpopulations was also unchanged in Dll1-deficient mice ( Supplementary Figure 4D ).

These data suggest non-redundant function of CSF-1 and Dll1 in instructing or sustaining a mature macrophage phenotype in ischemia. We therefore employed an in vitro culture system to study the individual effects of CSF-1 and Dll1 in a defined setting. Ly6C^hi monocytes were sorted from the bone marrow of Cx3cr1^gfp/+ mice and cultured in control plates or plates pre-coated with different concentrations of recombinant DLL1, in the presence of different concentrations of CSF-1. After 48 or 72hrs of culture, GFP⁺CD11b⁺F4/80⁺ macrophages were collected and analyzed by gene expression analysis and flow cytometry ( Figure 4F ). Gene expression analysis revealed that expression of Hes1, a transcriptional target of Notch signaling indicating Notch activation, was dependent on DLL1 concentration, but not CSF-1 concentration ( Figure 4G ). Furthermore, macrophage differentiation markers CD11c and I-A were very low in the absence of DLL1, but both markers became upregulated in a dose dependent manner, while a 10-fold difference in CSF-1 dose did not significantly alter expression of either CD11c or I-A ( Figures 4H, I ), suggesting non-redundant differentiation cues provided by Notch signaling. In contrast, macrophage proliferation was stimulated by CSF-1 in a dose-dependent manner, which was counteracted by increasing amounts of DLL1 ( Figure 4J ). This suggest non-redundant functions for CSF-1 and Notch, with CSF-1 acting as proliferation or survival stimulus for the macrophage lineage, while Notch provides instructive cues for differentiation. Since no cells are obtained in the absence of CSF-1 (data not shown), this also indicates that CSF-1 is required, but not sufficient, for macrophage differentiation.

Cx3cr1^gfp/+(B6.129P2(Cg)-Cx3cr1^tm1Litt/J) (60) (C57BL/6 background) and Dll1^LacZ/+ (129-Dll1^tm1Gos/J) (SV129 background) (44) mice had been described previously. 10-12 weeks old male mice were used for the experiments. Animals were housed under specific pathogen-free conditions at 14/10hr light/dark cycle and free access to standard lab animal diet (Altromin) and autoclaved tap water. All experiments were approved by local animal welfare authorities of Hannover Medical School and Lower Saxony (LAVES).

Experiments with hind limb ischemia were carried out as described (8, 9, 61). Briefly, mice were anesthesized through intraperitoneal injection of Ketamin (80mg/kg, CPPharma), Xylavet (Xylazin 2.5mg/kg, CPPharma) and Dormicum (Midazolam 2.5mg/kg, Ratiopharm), diluted in 0.9% NaCl (Braun), 5μl per 1 gram of body weight. After a right inguinal incision, the neurovascular bundle was exposed under microscopic control and the superficial branch of the femoral artery was surgically ligated distal to the origin of the deep femoral branch. To avoid post-operative hypothermia, mice were maintained on heating pads until they had been fully awake. Perfusion was measured by Laser Doppler Imaging (LDI) of plantar regions of interests with Perimed LDPI PIM II Laser Scanner (Perimed, Sweden). Animals with less than 90% relative perfusion reduction post-surgery were excluded from the study. Experimenters were unaware of treatment allocation or genotype.

GW2580 (LC Laboratories) was administered once daily by oral gavage at a dose of 80 mg/kg in 0.1% Hydroxipropylmethylcellulose/0.1% Tween-20 as described previously (62). Treatment started 4 days prior HLI induction. Body weight was measured daily to exclude weight loss. Anti-CSF-1 (aCSF-1) antibody treatment (Clone 5A1, BioXCell) or PBS control was subcutaneously injected in the ischemic limb, at a dose of 50 μg, immediately after surgery or up to 3 days (for details see Figure 3H ).

Spleens were pressed, resuspended in PBS (Sigma) and filtered through 70um mesh (Nitex). Blood samples were filtered as indicated. Red blood cells from spleen and blood samples were removed through treatment with RBC lysis buffer (Biolegend) and subsequent washing with PBS. M. Tibialis anterior and M. semimembranosus were excised, small piece was snap-frozen in the liquid nitrogen. The rest of the tissue was minced and incubated in DMEM (Sigma) containing 500U/ml type 2 collagenase (Worthington) and 1% penicillin/streptomycin (Sigma) for 22 min at 37⁰C. Then, tissue was disintegrated using GentleMACS dissociator (Miltenyi Biotec, program “spleen 01_01”) and the incubation step was repeated for another 22 min. Finally, tissue lysate was resuspended, filtered as indicated above and cells were washed with PBS. Bone marrows were flushed from the long bones through spinning at 5000g, 30sec, filtered twice and resuspended in PBS. For sorting purposes, single cell suspension was overlaid on equal amount of Histopaque 1083 (Sigma), spun at 430g 25min. at room temperature (RT), interphase was collected and washed with PBS. Cell suspensions from all organs were counted using Countess II FL automated cell counter (Invitrogen). Dead cells were excluded through 0.4% trypan blue (Sigma) staining.

Cells were resuspended in staining buffer containing PBS supplemented with 2% fetal calf serum (FCS, Biochrom), 2mM Na₂EDTA (Roth) and 0.02% NaN₃ (AppliChem). 0.5-1x10⁶ cells were plated and stained with antibodies and streptavidin compounds, listed in Supplementary Table S1. Propidium Iodide (Fluka) was used to exclude dead cells. Flow cytometric analysis was performed using LSR II (BD biosciences; acquisition software BD FACSDiva Software v8.0.1). BD FACSAria™ Fusion Flow Cytometer was used for cell sorting. All flow cytometry data were analyzed using FlowJo software v8.0.1 (TreeStar). Cell definition is summarized in Table 1 .

12-well cell culture plates (TPP) were pre-coated with different amounts of chimeric mouse DLL1-Fc protein or Fc fragment alone (both from R&D) dissolved in PBS, at RT for 3hr. Single cell suspensions were prepared from the bone marrows of Cx3cr1^gfp/+ mice, stained for CD11b and Ly6C and GFP⁺CD11b⁺Ly6C^hi monocytes were sorted as described above. 1.6x10⁵ cells were plated in pre-coated wells, in RPMI-1640 medium (BioCell), supplemented with 10%FCS (Biochrom), 1% Glutamax (Gibco) and 1% Penicilin-Streptomycin (Biochrom), in the presence of different concentrations of recombinant mouse M-CSF (Peprotech). After 48hr of culture, medium was supplemented with the new dose of M-CSF, equal to initial amounts. Finally, cells were collected after 72hr of culture and used for flow cytometry analysis or RNA isolation. To quantify proliferation, BrdU (Sigma) was added to culture medium at a 10μM final concentration, during the last 16hr of 72hr culture and incorporation was detected by flow cytometry, using BrdU flow kit (BD Pharmingen), according to manufacturer’s instructions.

CBA analysis was performed with serum and muscle extract samples. Blood was collected without anticoagulant. After clot formation, samples were spun at 1000g, 10min. and serum was harvested. TA muscles were excised, minced and immediately placed in the ice-cold extraction buffer: 0.05% Tween-20 (BioRad), 0.1% fraction V BSA (Roth) in PBS, supplemented with complete mini protease inhibitor cocktail (EDTA-free, 1 tablet per 10ml; Roche). Then, tissue was mechanically homogenized and incubated on ice for 30min. Finally, samples were vortexed, spun at 17000g for 20min., 4⁰C and supernatants were collected.

CBA was performed using LEGENDPlex™ Mouse HSC Myeloid panel kit (Biolegend), according to manufacturer’s instructions. Samples were acquired on LSR II flow cytometer (see above). Results were analyzed using LEGENDPlex™ online software (Biolegend), according to manufacturer’s instructions. Concentrations in muscle were normalized on tissue mass and extraction volume.

Histologic and Immunohistochemistry analysis of muscle tissue was performed as previously described (61) with modifications. Tissues were fixed in 4% PBS-buffered paraformaldehyde (PFA, Sigma) and embedded in paraffin or cryopreserved in increasing concentrations (15% and 30% in PBS) of sucrose (Roth), and embedded in Tissue-tek OCT compound (Sakura). Blocks were sectioned with a rotation microtome (Leica) into 2µm (paraffin blocks) or 8 µm (frozen blocks) slices. For morphology analysis, paraffin sections were deparaffinized and stained with H&E according to routine protocol. To detect CSF-1, antigen retrieval procedure was performed on deparaffinized sections, by boiling in 10mM Citrate (Sigma) buffer pH6.0 for 16min. and subsequent cooling on ice. Blocking buffer was prepared using appropriate sera from the hosts of secondary antibodies, at 10% final concentration in 3%BSA/PBS. Fc blocking reagent TrueStainX anti-CD16/32 (Biolegend) was used if needed. For intracytoplasmic markers, fixation-permeabilization was performed using 0.15% Triton-X100 (Roth)-containing PBS respectively. After antibody staining, nuclei were counterstained with DAPI (Roth) and sections were embedded in the fluorescence mounting medium (DAKO). Primary and secondary antibodies used for IHC are listed in Supplementary Table S2 . Light microscopy was performed using Leica DFC425 C microscope, images were acquired and processed using Leica Application Suite v. 3 software. Confocal images were acquired using Leica DMi8 (Leica Microsystems) inverted microscope, with x20 immersive objective. Data were processed using Las AF Lite Software (Leica Microsystems). Sizing of microscopy images was performed using Adobe Illustrator (Adobe).

For quantification analysis of H&E stained samples, morphology of fibers was evaluated according to their shape and architecture and location of nuclei. Total amount of fibers were counted per section and % of each group: intact, regenerative or disintegrated fibers was determined based on it (for details see Supplementary Figure 2A ).

Total RNA was purified using Nucleospin RNA plus Kit (Macherey Nagel) according to manufacturer’s instructions. Qualitative and quantitative analysis was performed using Nanodrop 2000 (Thermofisher scientific) spectrophotometer. Then, RNA was transcribed into cDNA using cDNA synthesis kit (Invitrogen) and quantitative real-time PCR was performed with Fast Start Essential DNA Green Master Mix on a LightCycler 96 system (both from Roche). All procedures were run according to the manufacturer’s instructions. The expression of each specific gene was normalized to the expression of housekeeping Rps9, using 2^-<σπ>Δ</σπ>^Ct method. Primer sequences are listed in the Supplementary Table S3 .

Results were analyzed with GraphPad Prizm 9 software. All data are expressed as mean ± SEM. Difference between groups were determined using 2-way ANOVA with Tukey’s and Bonferroni’s multiple comparison test, 1-way ANOVA with Dunnett’s multiple comparison test, Mann-Witney’s or 2-tailed unpaired t-test. P<0.05 was considered to be statistically significant difference.