Male C57BL/6 mice (6–8 weeks) were provided by Janvier (Le Genest, France). GPR40 knockout mice with C57BL/6N background (6–8 weeks) were previously described (7, 18). Frozen paw tissue from previously published experimentation with G2A knockout mice (The Jackson Laboratory, Bar Harbor, ME) were used for MELC analyses (7). Mice were treated according to the International Association for the Study of Pain guidelines. All the ethics guidelines for investigations in conscious animals were observed and the procedures were approved by the local ethics committee (Regierungspräsidium Darmstadt). All animals had free access to water and food. The room temperature (23 ± 0,5 °C) and light (7:00 am and 7:00 pm) were controlled. The animals were randomized for the experiments. Inflammation was induced by injection of 10 µl Zymosan A (3mg/ml in PBS, Merck, Darmstadt, Germany) subcutaneously into the plantar side of one hind paw. For neutrophil depletion, 500 µg Ly6G antibody clone 1A8 InVivoPlus™ from Bio X Cell (Dartmouth, NH) was injected intraperitoneal (i.p.) 24 hours before the zymosan injection. The same amount of 500 µg rat IgG2α isotype clone 2A3 (Bio X Cell, Dartmouth, NH) was used as control and was injected 24 hours before zymosan injection as well. Baricitinib was administered orally (10 mg/kg body weight in carboxymethyl cellulose (CMC).

In all behavioral tests the experimenter was unware of the treatment of the animals. The edema measurements were done at the indicated times after zymosan injection in one hind paw. Edema volumes were measured with a 37140 plethysmometer from Ugo Basile (IITC Life Science, Woodland Hills, CA) by immersion of the mouse hind paw (19). The zymosan-induced thermal hypersensitivity was determined with the Hargreaves test using an IITC Plantar Analgesia Meter (Hargreaves test; IITC Life Science, Woodland Hills, CA, USA) and the maximal temperature was at 32°C with a cut off time at 20 seconds.

To score the effect of interventions on the resolution of zymosan-induced hypersensitivity published data (3–7) were reanalyzed for the time points 24, 48 and 72 hour after zymosan injection. A first scoring (Score 1) was applied to significant (Two Way ANOVA, Bonferroni post hoc test) changes in the paw withdrawal latency (PWL) compared to the respective control mice. It was scored with 1 when the intervention increased PWL toward baseline, 0 when there was no change and -1 when the PWL decreased. A second score (Score 2) was calculated in case the PWL returned to baseline earlier in the intervention group than the control group an additional score of +2 was applied per day (One Way ANOVA, Bonferroni post hoc test). In case the PWL returned to baseline earlier in the control group score of -2 was applied per day. Both scores were added to generate the final resolution score.

Multi-epitope-ligand cartography (MELC) is an automated immunohistological imaging method that can be used to visualize high numbers of antibodies on the same sample (3, 10). Briefly, 10 µm tissue sections on silanized coverslips were fixed with 4% paraformaldehyde in PBS for 10 minutes, permeabilized with 0.1% Triton X-100 in PBS for 10 min, and blocked with 3% bovine serum albumin (all from Merck, Darmstadt, Germany) in PBS for 1 hour. Tissue samples were imaged using a DMi8 microscope (Leica Microsystems, Wetzlar, Germany) and a HC PL FLUOTAR L 20x/0,040 CORR PH1 objective with a cooled sCMOS camera (2048 × 2048 pixels). The sample was then incubated with up to 3 antibodies, each labeled with different bleachable fluorescence-tags, and washed with PBS. Phase-contrast and fluorescence images were collected, the fluorescence signals were bleached and post-bleaching images were recorded. The process was repeated until all antibodies were imaged. For data analysis, the post-bleaching images were subtracted from the following fluorescence image. The antibodies used on the MELC system are listed in the Supporting Data S1 .

All grayscale antibody channel images were edited with ImageJ v1.53q (National Institutes of Health [NIH], Bethesda, MD, USA) to remove background fluorescence, noise and artifacts for the analyses. CellProfiler (v4.2.1) is an open-source software for analyzing cell images. It was used to create a cell mask and lightning corrections (20). CellProfiler is necessary for additional illumination correction and the generation of a cell mask for single-cell segmentation using the propidium iodide (cell nuclei) and CD45 (cluster of differentiation). The cell mask was imported in histoCAT (v1.7.3) (11) with the corresponding antibody channel images. All images, except images used for single-cell mask generation, were z-score normalized and used for Barnes-Hut t-SNE (BH t-SNE) (21) and PhenoGraph analysis (22) in histoCAT. PhenoGraph defines cell clusters based on single-cell mask and marker colocalization (k set between 15 and 30). BH t-SNE scatter plot was overlaid with a colored PhenoGraph cluster map and clusters were classified as cell types based on their marker expression. To calculate the relative number of cells per cell type, the number of objects per cluster was normalized to the total number of objects in the cell mask. The z-score normalized images were exported to FlowJo software v10.8.1. Pairwise interactions between cell phenotypes were calculated for each cell and their neighbors in a distance of four pixel and were imported into Cytoscape (v3.8.2) to produce dual-centered neighborhood networks showing relative distances to the defined centers (23).

Cell isolation and preparation from blood and paws was done as described previously (1, 9, 19). Briefly, inflamed paws were cut into <1 mm³ pieces and incubated for 45 minutes at 37°C in 500 µl lysis buffer (3 mg/ml Collagenase (from Clostridium histolyticum Typ IA, Merck, Darmstadt, Germany) in RPMI 1640 medium). Lysis was stopped by addition of 5 ml 10% FBS in DMEM. The cells were passed through a cell strainer (70 µm) and incubated in erythrocyte lysis (ACK) buffer for 5 minutes at room temperature. The cells were centrifuged, fixed and then permeabilized at 4°C for 10 minutes using the BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences, Heidelberg, Germany). Unspecific binding was blocked by incubation with 60 µl of 2% Fc-blocking reagent Mouse BD Fc Bloc (BD Pharmingen, NJ, USA) in PBS for 10 minutes at 4°C. Antibodies ( Supplementary Table S1 ) were incubated for 20 minutes at 4°C. Samples were acquired with a flow cytometry system (FACSymphony A5 Cell Analyzer; BD Biosciences, Heidelberg, Germany) and analyzed by using FlowJo software v10 (BD Biosciences, Heidelberg, Germany). Unstained controls and fluorescence minus one (FMO) controls were used to establish the gating strategy. Cell fragments and cell clusterings were excluded based with FCS/SSC gating based on the particle sizes. Additional Live dead staining was not included. According to this gating strategy the signals considered as live, single cells comprised around 60% of all signals after isolation of cells from paws. To detect phagocyting cells pH-sensitive pHrodoTM Red Zymosan A BioParticles conjugate (Invitrogen, Eugene, Oregon, USA) was used for injection and cells isolated from an inflamed paw injected with unlabeled zymosan were used as FMO control.

Bone marrow cells were isolated from femur and tibia from the hind legs of wild type and GPR40-knockout mice. The cells were passed through a cell strainer (70 µm), incubated in erythrocyte lysis (ACK) buffer for 5 minutes at room temperature, centrifuged, washed and resuspended in PBS-0,5% BSA. Unspecific binding was blocked by incubation with 60 µl of 2% Fc-blocking reagent Mouse BD Fc Bloc on ice. FACS sorting was performed using a FACS Diva (BD Biosciences) based on granularity and Ly6G^high expression (anti-Ly6G APC-cy7 (1A8)). RNA isolation, sequencing, quantification of mapped reads and differential transcript expression analysis was done by GenXPro GmbH (Frankfurt, Germany). Pathway analysis of the results was done using EnrichR (24, 25) and String (26) mRNA Sequencing data are available in the GUDe database (https://gude.uni-frankfurt.de/handle/gude/591).

Statistically significance was calculated using one-way or two-way analysis of variance (ANOVA) by GraphPad Prism v9.0.1. For post hoc analysis Bonferroni correction for multiple comparisons was used. Comparisons between two groups were performed by unpaired two-tailed Student’s t-test with Welsch´s correction.

Previously it was shown that neutrophil depletion using the anti-Ly6G/Ly6C antibody clone GR1 strongly decreased zymosan-induced edema formation without affecting extent or resolution of hypersensitivity (15). Here we used the more specific anti-Ly6G antibody clone 1A8, which reduced the number of neutrophils by >90% in blood and inflamed paws 24 hours after zymosan injection ( Figures 1A, B and Supplementary Figure S2A, B ). Notably, we used the anti-Ly6G/Ly6C antibody GR1 for FACS and immunohistochemical detection of neutrophils. This antibody recognizes neutrophils without interference from the 1A8 antibody, which was used for neutrophil depletion ( Supplementary Figure S3A ). In agreement with previous results (15), the zymosan-induced edema formation was strongly decreased in the depleted mice ( Figure 1C ) whereas onset and resolution of zymosan-induced thermal hypersensitivity were not affected by neutrophil depletion ( Figure 1D ). To determine the effect of neutrophil depletion on the inflammatory architecture, the high-content imaging technology MELC was used with an established panel of 29 antibodies recognizing immune cells and non-immune cells ( Supplementary Table S1 ) (3–5, 27). For single cell phenotyping, cluster analysis, quantification and neighborhood analysis the previously established workflow was used ( Figure 1E ) (3–5). Visual fields for MELC imaging were selected to cover the zymosan-containing region (approximately 30% of the total image) and the neighboring areas. MELC analysis of paws 24 hours after zymosan injection confirmed a reduction of the neutrophil number in the tissue by >90% ( Figures 1F, G and Supplementary Figure S3B ). Neutrophil depletion did not change the number of cells in clusters representing eosinophils and dendritic cells (DCs) in the observed areas of the inflamed paws ( Figures 1H, I ) whereas the total number of macrophages significantly increased ( Figure 1J ). Further bioinformatics analysis of the MELC data showed that cell clusters representing M1-like macrophages (Siglec F^–/F4-80⁺/CD86⁺/CD206^–) and M0 macrophages (Siglec F^–/F4-80⁺/CD86⁺/CD206⁺) were responsible for this increase, while the number of M2-like macrophages (Siglec F^–/F4-80⁺/CD86^-/CD206⁺) did not change ( Figures 1K–M ). As reported previously, no innate lymphoid cells, B cells, T cells and NK cells were detected in this area of the tissue during this phase of the zymosan-induced inflammation (3, 4).

To determine the effect of neutrophil depletion on the inflammatory structure, the likelihood of cells neighboring each other was compared to randomized cell distributions (15). For visualization of the neighborhood of zymosan, linear distance maps of the relative distance between zymosan and the identified cell clusters were generated based on scores ranging from 0 (direct neighbors) to 2 (no neighbors). Based on these scores, the relative distance between zymosan and M2-like macrophages can be used to define the size of the PI region (3–5). Importantly, neutrophil depletion did not change the size of the PI region ( Figures 2A–C ). Neutrophil depletion also led to surprisingly small effects on the relative distance to zymosan for most of the other immune cell types ( Figures 2A, B ). Only CD86/CD206 double positive M0 macrophages showed a significant decrease in their relative distance to zymosan ( Figures 2A, B ), suggesting a reduced pressure of the microenvironment in the PI region on macrophages to polarize toward M1-like macrophages. Also the dual-centered network visualization, which combines the cellular neighborhoods of zymosan and M2-like macrophages, and allows a clearer presentation of the distribution of the immune cells in the tissue, showed, with the exception of the M0 macrophages, no relevant changes in the network localization of the different cell types ( Figure 2D ). One of the major functions of neutrophils in inflammation is the removal of pathogens by phagocytosis. Therefore, we determined which cell types are involved in zymosan phagocytosis in our model using pHrodo-zymosan, whose pH-sensitive fluorescent label is activated upon internalization into lysosomes. FACS analysis showed that the number of pHrodo-zymosan⁺ immune cells did not change after neutrophil depletion ( Figure 2E ), with an increased proportion of pHrodo-zymosan⁺ macrophages ( Figure 2F ; Supplementary Figures S4A, B ). This increase in phagocytic macrophages was not detected in MELC analysis ( Supplementary Figures S5A–D ), which might be due to the incomplete coverage of the core region in the field of visions. Since the timely pathogen clearance was unaffected in neutrophil-depleted mice as reflected by the normal resolution of zymosan-induced thermal hypersensitivity, it can be speculated that the phagocytic capacity of macrophages and other phagocytic immune cells, which is otherwise used to clear apoptotic neutrophils, can be redirected toward phagocytosis of zymosan.

Notably, neutrophil depletion results in their phagocytosis by macrophages, which might alter the course of inflammation. Therefore, GPR40 (FFAR1) knockout mice were tested as a second model to interfere with neutrophil functions. The G protein-coupled receptor (GPCR) GPR40 is highly expressed in neutrophils and positively regulates neutrophil functionality (28–30). FACS analysis showed that although neutrophil numbers were upregulated in the blood of GPR40 knockout mice ( Figure 3A ; Supplementary Figure S6A ), neutrophil recruitment to zymosan-injected paws was reduced in the GPR40 knockout mice ( Figure 3B ). Analogous to the findings in the antibody-mediated neutrophil depletion model, FACS analysis also showed increased numbers of macrophages in the inflamed paw ( Figure 3C ), suggesting a similar compensatory mechanism. As seen in the antibody-mediated neutrophil depletion model, the edema formation was strongly reduced in GPR40 knockout mice ( Figure 3D ) while the resolution of the zymosan-induced thermal hypersensitivity was not altered ( Figure 3E ). To investigate the mechanisms underlying the decreased neutrophil recruitment in GPR40 knockout mice, we determined the mRNA expression levels by RNAseq of neutrophils isolated from the bone marrow of wild type and GPR40 knockout mice. Pathway analyses using the EnrichR and String databases showed a prominent decrease in the expression of genes associated with the tubulin cytoskeleton, including the downregulation of tubulin-α-1c and the kinesins Kif4, Kif11, Kif15 and Kif22 ( Figure 3F ; Supplementary Figure S6B and S9 ). Since the tubulin cytoskeleton is required for endosomal transport involved in receptor transport to the plasmamembrane (31), we performed FACS analysis of neutrophils isolated from the blood to study the expression of 10 chemokine and integrin receptors, that mediate neutrophil recruitment to inflamed tissue. We found a significant downregulation of CD44, CD49d, CXCR1 and CXCR4 ( Figure 3G ; Supplementary Figure S4A ), which have been shown to mediate bone marrow egress of neutrophils, neutrophil adhesion to endothelial cells, neutrophil tissue recruitment and endothelial transmigration of neutrophils (32–36). Thus, the data show that loss of GRP40 abolishes neutrophil recruitment during zymosan-induced inflammation, which appears to be based on a multifactorial perturbation of cytoskeletal functions and receptor surface expression.

MELC analysis of the inflamed paws confirmed the nearly complete absence of neutrophils 24 hours after zymosan injection ( Figures 4A, B ). Notably, the difference in the neutrophil numbers between the MELC and the FACS analyses shown in Figure 3B can be attributed to the presence of contaminating blood in the tissue preparations used for the FACS analysis. Fittingly, this was not an issue after antibody-mediated neutrophil depletion, since the blood was also depleted of neutrophils. As seen for antibody-mediated neutrophil depletion an increase of cells in clusters representing M1-like macrophages was observed ( Figure 4C ), suggesting a compensatory upregulation similar to the results for antibody-mediated neutrophil depletion. The number of M2-like macrophages, all macrophages and DCs did not change ( Figures 4D–F ), while the number of eosinophils decreased by approximately 50% ( Figure 4G ), suggesting a previously unknown role for GPR40 in their recruitment to the inflamed tissue. Most importantly, the neighborhood analysis showed only a minor shift of M2-like macrophages toward the zymosan, which did not reach significance ( Figures 4H, I ). Taken together, the data show that neutrophil depletion, either mediated by antibodies or due to GPR40-deficiency, does not significantly affect the size of the PI region as defined by the relative distance between zymosan and M2-like macrophages. Accordingly, also the resolution of the zymosan-induced thermal hypersensitivity was not affected.

The GPCR G2A (GPR132) is expressed on macrophages and mediates efferocytosis of apoptotic neutrophils by increasing chemotaxis toward apoptotic “find-me” signals (37–40). Accordingly, G2A deficiency causes reduced macrophage migration to the zymosan-containing core region, thereby reducing the M1-like macrophage numbers in this area, which results in an enhanced resolution of inflammation (7). To determine the effect of the G2A-deficiency on the PI region, we performed MELC analyses using inflamed paws from wild type and G2A-deficient mice 24 hours after zymosan injection. Consistent with previous publications (7, 38, 39), MELC analysis showed in G2A knockout mice a significant decrease of the number of macrophages in the observed area ( Figure 5A ). Also, the number of M1-like ( Figure 5B ) and M2-like macrophages decreased ( Figure 5C ). Interestingly, the number of CD206/CD86 double positive M0 macrophages increased, suggesting an overall decreased polarization pressure on macrophages by their microenvironment ( Figure 5D ). To investigate the effect of the locally altered recruitment and polarization of macrophages on the regional structure of the inflammation, we performed a neighborhood analysis. We found an increased likelihood of M2-like macrophages to neighbor zymosan whereby the relative distance between zymosan and M2-like macrophages decreased by 50% ( Figures 5E, F ). Thus, in G2A deficient mice, the strongly enhanced resolution of inflammation (7) is accompanied by decreased numbers and polarization of macrophages as well as shrinkage of the PI region by 50%.

Since G2A-deficiency increased the resolution of inflammation and was accompanied by decreased M1-like macrophage numbers, we investigated whether inhibition of M1-like macrophage polarization is sufficient to decrease the PI region and to promote the resolution of inflammation. Therefore, we used the immunosuppressive Janus kinase (JAK) 1/2 inhibitor baricitinib, which blocks among others the M1-like promoting cytokines IL-6 and INFγ (41). MELC analysis showed that baricitinib treatment abolished M1-like macrophage polarization and caused an increase of the number of the CD86/CD206 double-positive M0 macrophages ( Figures 6A–D ). Notably, neither the number of M2-like macrophages nor the total number of macrophages was altered ( Figures 6E, F ). Similarly, there was no significant effect on the number of DCs, eosinophils, neutrophils or mast cells seen in the observed inflammatory area ( Supplementary Figures S7A–D ). Neighborhood and network analyses showed a shift of M0 macrophages toward the zymosan, which is likely due to the reduced polarization of M0 macrophages toward M1-like macrophages, which are thereby reduced in their numbers ( Figures 6G, H ). Most importantly, the neighborhood analysis showed no change in the likelihood of zymosan and M2-like macrophages being neighbors, demonstrating that the treatment does not alter the size of the PI region ( Figures 6G, H and Supplementary Figure S7E ). Finally, baricitinib treatment had no effect on the resolution of thermal hypersensitivity ( Figure 6I ), demonstrating that preventing the polarization of M1-like macrophages alone is not sufficient to promote the resolution of inflammation.

Next, we calculated the effect of the 4 different interventions presented in this study and reevaluated the data from previously published data for eosinophil depletion (3), meloxicam treatment (4), mast cell deficiency (5) and TP knockout mice (6) for the relative size of the PI region as determined by the likelihood of zymosan and M2-like macrophages. These values were plotted against a score describing the resolution of inflammation, which was calculated by scoring the change in time required for the zymosan-induced hypersensitivity to return to baseline as well as the change in paw withdrawal latency as compared to the respective control treatments.

We found that a small reduction of the size of the PI region of less than 30%, as seen in GPR40 knockout mice, after neutrophil depletion and after baricitinib treatment, correlates with a normal resolution of inflammation ( Figure 7A ; Supplementary Table S8 ). In contrast, strong changes of the size of the PI region, as observed for eosinophil depletion and meloxicam treatment (>65% decrease of the PI region) (3, 4) or in mast cell-deficient mice (50% increase of the PI region) (5), were associated with a delayed resolution of inflammation ( Figure 7A ). The only interventions leading to improved resolution of inflammation, were the G2A- and TP-knockout, which showed a decrease of 50-60% for the PI region ( Figure 7A ). Notably, no dependence of the resolution score on the number of M1-like or M2-like macrophages in the observed area was seen ( Figures 7B, C ). In summary, the data show that the size of the PI region mirrors pro- and anti-inflammatory effects of interventions, whereby small or no effects on the size of the PI region do not affect resolution of a TLR2-mediated inflammation. A moderate decrease of the size of the PI region is associated with promotion of resolution while stronger changes are associated with a reduced resolution. These effects mirrored the net effect on the immune response and was not clearly associated with changes in macrophage subtypes.