Macrophage origin limits functional plasticity in helminth-bacterial co-infection

Rapid reprogramming of the macrophage activation phenotype is considered important in the defense against consecutive infection with diverse infectious agents. However, in the setting of persistent, chronic infection the functional importance of macrophage-intrinsic adaptation to changing environments vs. recruitment of new macrophages remains unclear. Here we show that resident peritoneal macrophages expanded by infection with the nematode Heligmosomoides polygyrus bakeri altered their activation phenotype in response to infection with Salmonella enterica ser. Typhimurium in vitro and in vivo. The nematode-expanded resident F4/80high macrophages efficiently upregulated bacterial induced effector molecules (e.g. MHC-II, NOS2) similarly to newly recruited monocyte-derived macrophages. Nonetheless, recruitment of blood monocyte-derived macrophages to Salmonella infection occurred with equal magnitude in co-infected animals and caused displacement of the nematode-expanded, tissue resident-derived macrophages from the peritoneal cavity. Global gene expression analysis revealed that although nematode-expanded resident F4/80high macrophages made an anti-bacterial response, this was muted as compared to newly recruited F4/80low macrophages. However, the F4/80high macrophages adopted unique functional characteristics that included enhanced neutrophil-stimulating chemokine production. Thus, our data provide important evidence that plastic adaptation of MΦ activation does occur in vivo, but that cellular plasticity is outweighed by functional capabilities specific to the tissue origin of the cell.

Introduction Macrophages (MF) are central to many immune and homeostatic processes and adopt a variety of activation phenotypes. During bacterial infections 'classically activated' MF produce anti-microbial effector molecules, exhibit enhanced antigen presentation capacity and produce proinflammatory cytokines. In contrast, during helminth infection MF are activated by IL-4Rα-signaling and called M(IL-4), or 'alternatively activated' [1,2]. M(IL-4) express low levels of co-stimulatory molecules, produce molecules associated with wound healing, and are considered anti-inflammatory [1]. Of note, the MF activation phenotype is not fixed and the prevailing view is that MF can adopt a variety of activation states in response to their environment [3][4][5][6].
Although MF plasticity is well established, most of the data supporting this concept are based on in vitro findings or ex vivo derived cells. Moreover, full activation, especially in infectious settings in vivo, rarely occurs in 100% of the MF present [7][8][9]. Thus, the observed plasticity might be due to hitherto quiescent subsets of MF responding rather than true plasticity of previously activated cells. Furthermore, recent data have highlighted distinct mechanisms for MF accumulation in different infection settings [10]. During bacterial infection bone marrow-derived blood monocytes infiltrate from the vasculature and differentiate into anti-bacterial effector MF or dendritic cells (DC) [11,12] while the tissue resident MF population is usually lost from the site of infection in a process called the MF disappearance reaction [11,13,14]. In contrast the type 2 immune response associated with helminth infection can result in proliferative expansion of tissue resident MF with minimal recruitment of monocytederived cells [8,15]. Of note tissue resident peritoneal MF can originate from either prenatal sources or bone marrow derived precursors depending on the age of the animal [16,17] but the use of the term 'origin' in the context of this study refers to the tissue of origin (blood vs. peritoneal cavity) within the time frame of the infection. Critically, the functional significance of resident cell expansion vs monocyte recruitment is not yet clear. The finding that some helminth infections also lead to recruitment of blood monocytes [9] and that monocyte-recruited MF show marked disparity in the transcriptional response to recombinant interleukin 4 (IL-4) as compared to tissue resident MF [18], suggests important functional differences. Thus, the distinct tissue origin of MF during infection combined with a current lack of in vivo evidence for cell-intrinsic changes in activation state raised two questions. 1) Does MF activation plasticity occur in vivo at the level of the individual cell and 2) is plasticity and/or MF origin relevant to infection outcome?
To address these questions, we turned to co-infection models, which provide physiologically relevant insight into MF polarization and plasticity [19]. We utilized murine co-infection with Heligmosomoides polygyrus bakeri (H. polygyrus) and attenuated Salmonella enterica subsp. enterica serovar Typhimurium (S. enterica ser. Typhimurium; SL3261, ΔaroA). H. polygyrus is a natural gastrointestinal nematode parasite of mice inducing a strong type 2 immune response; infection leads to pronounced proliferative expansion and M(IL-4) activation of tissue resident MF in the peritoneal cavity [10,15,20]. Infection with Salmonella species induces a potent pro-inflammatory response required for bacterial clearance [21]. In co-infection experiments, mice were infected with H. polygyrus and at peak expansion of resident peritoneal MF SL3261 was inoculated i.p. Cell-intrinsic plasticity of the M(IL-4) MF was observed in response to bacterial infection but plasticity did not appear relevant for protection within the first 5 days of SL3261 infection. Indeed, inflammatory recruitment of monocyte-derived MF and neutrophils was unhampered in co-infected mice. Moreover, the inflammatory influx was accompanied by the marked disappearance of the helminth-expanded resident MF. Microarray and functional analyses demonstrated that nematode-expanded MF do adapt their activation phenotype in co-infection settings, but the tissue origin places limitations on the functional response of these cells.

Nematode-elicited MΦ can be classically activated independent of prior activation state
In order to examine the capacity of nematode-elicited MF (NeMF) to change their M(IL-4) phenotype, peritoneal exudate cells from mice infected 14 days previously with H. polygyrus were stimulated in vitro with lipopolysaccharide (LPS), recombinant IL-4 (rIL-4) or without additional stimulus. As a reference for the magnitude of the response we used non-polarised thioglycollate-elicited peritoneal MF, which respond well to polarising stimuli [22]. LPS stimulation induced upregulation of NOS2 expression in NeMF (Fig 1A), which varied in magnitude relative to control thioglycollate-elicited MF between experiments (S1 Fig). Importantly, induction of NOS2 expression in NeMF was consistently observed, indicating that NeMF retain their capacity to respond to bacterial stimuli. NeMF also remained responsive to rIL-4 as demonstrated by enhanced Relm-α expression ( Fig 1A). Of note, re-polarization of classically activated MF was less evident; MF isolated from animals harbouring a bacterial infection (S. enterica ser. Typhimurium, SL3261) showed strong upregulation of NOS2 expression following stimulation with LPS, notably in excess of the response observed in thioglycollate elicited MF (Fig 1B). In contrast only very limited non-significant upregulation of Relm-α was observed in response to rIL-4 ( Fig 1B). Thus, bacterial stimuli seemed to provide a more restrictive activation signal, placing limitations on the plasticity of the response and favouring anti-bacterial outcomes of activation.
To assess whether the upregulation of NOS2 in NeMF was restricted to a subpopulation that had not previously responded to IL-4Rα signaling, we analysed co-expression of NOS2 and Relm-α ( Fig 1C). Irrespective of whether the isolated MF expressed Relm-α or not both populations showed equal upregulation of NOS2 in response to LPS (Fig 1D) indicating that classical activation was not restricted to previously non-responding cells.
To directly confirm that M(IL-4) can switch their phenotype in vivo, resident MF that had been expanded and activated by in vivo delivery of IL-4 complex, were transferred into the peritoneal cavity of SL3261 infected animals. The transferred cells showed equivalent induction of NOS2 expression as host MF (S1 Fig). Thus, activation plasticity of M(IL-4) also occurred in vivo despite potential competition with host MF for activating stimuli Bacterial co-infection leads to the loss of F4/80 high nematode-expanded MΦ As described previously H. polygyrus infection leads to the proliferative expansion of peritoneal, tissue resident MF [15] whereas S. enterica ser. Typhimurium induces influx of blood monocyte-derived MF [23]. Thus, to address whether the presence of large numbers of nema- Effects on MF were analysed 5 days later, a timepoint when T cell-independent, partially NOS2-dependent control of bacterial growth can be observed [24].
All infections significantly increased the total number of cells in the peritoneal cavity as compared to naïve animals (Fig 2A). As expected there was preferential influx of eosinophils in nematode infected animals as compared to a more neutrophilic inflammation in SL3261 infected mice. In co-infected animals neutrophil influx was not impeded by prior H. polygyrus infection indicating normal recruitment of these cells from the circulation. In contrast the number of eosinophils was reduced by the presence of bacteria confirming crossregulation between the anti-helminth and anti-bacterial immune response in our model. The number of MF in the peritoneal cavity increased both in bacterial and helminth infection, but co-infection had no additive effect (Fig 2A). Consistent with the differences in tissue origin, MF accumulating in singly-infected nematode or bacterial infections differed phenotypically [8], expressing high or low surface levels of F4/80 respectively ( Fig 2B). Consecutive co-infection led to the simultaneous appearance of both MF populations (F4/80 high and F4/80 low ) ( Fig 2C). Furthermore, similar to neutrophils, F4/80 low MF were expanded to an equivalent level between single SL3261 and co-infected animals (Fig 2C). At the same time a significant reduction in the number of F4/80 high MF was observed in co-infected relative to single H. polygyrus infected animals. Moreover, the degree of F4/80 low cell recruitment depended on the inoculating dose of SL3261 and correlated with the loss of F4/80 high MF ( Fig 2D).

Nematode-expanded MΦ do not alter resistance to SL3261 infection and efficiently upregulate anti-bacterial effector markers in response to bacterial infection
The loss of F4/80 high MF and concomitant recruitment of F4/80 low MF made us question whether the plastic response of NeMF observed in vitro (Fig 1) did occur in vivo. Utilising the consecutive co-infection model described above we found no statistical difference in the induction of intracellular NOS2 expression by peritoneal MF in single SL3261 or co-infected animals ( Fig 3A). Furthermore, induction of NOS2 was not restricted to newly recruited F4/80 low MF, as F4/80 high MF showed at least equal if not enhanced capacity to induce NOS2 expression (Fig 3B & 3C). Moreover, SL3261 co-infection led to loss of the M(IL-4) activation phenotype in NeMF, as measured by intracellular Relm-α expression, but induction of NOS2 was independent of previous M(IL-4) activation as indicated by equivalent expression in Relm-α positive and negative cells (Fig 3D & 3E). Thus, nematode expanded F4/80 high , resident-derived MF in co-infected animals showed clear and efficient induction of anti-bacterial effector mechanisms, which was further evidenced by enhanced expression of MHC-II ( Fig  3F). In line with the upregulation of anti-bacterial effector molecules by F4/80 high MF and unaltered recruitment of F4/80 low MF, no significant difference was found in the number of bacteria present in the spleen of co-infected animals as compared to single SL3261 infected animals ( Fig 3G). Thus, the more than 7-fold greater number of MF present in the peritoneal cavity of H. polygyrus infected mice at the time of bacterial inoculation (S3B Fig) did not provide any protection, despite their apparent activation plasticity. We considered that the inability to provide protection may be due to injection of relatively large quantities of bacteria. We therefore titrated the dose of SL3261 but no effect of helminth infection on splenic bacterial burden was detected, even when only 50 bacteria were injected into the peritoneal cavity ( Fig  3H). Similarly, when mice were co-infected with high doses of SL3261 (~3x10^6 CFU), no difference in resistance was observed (S4 Fig).
F4/80 low cells do not arise from plastic conversion of F4/80 high MΦ F4/80 low , blood monocyte derived MF exposed to tissue specific factors can give rise to tissue resident F4/80 high MF [16,17]. Thus, whilst the loss of F4/80 high MF may be due to their disappearance or death, a conversion of F4/80 high nematode-expanded MF to F4/80 low MF could have occurred under co-infection settings. To discriminate between these possibilities we utilized partially protected chimeras in which the peritoneal cavity is shielded from lethal irradiation [8,25]. This protects the tissue resident peritoneal population and prevents their replacement by bone marrow-derived cells but results in a chimeric blood monocyte population (S5 Fig). By comparing the proportion of donor cells within a given MF population with the proportion in blood monocytes it can be determined whether the population is derived from the bone marrow (ratio equal to monocytes) or from tissue resident MF (low donor ratio). Peritoneal MF from single H. polygyrus or SL3261 infected animals showed low or high proportions of donor cells, respectively, confirming their tissue resident MF or bone marrow derived origins (Fig 4A). Importantly the two MF populations present in co-infected mice displayed identical chimerism to their respective counterparts in single infected mice. The F4/80 high MF included a very low percentage of donor bone marrow cells while F4/80 low MF displayed similar chimerism to blood monocytes (Fig 4A & 4B). The data demonstrate that no conversion between these two MF populations occurred and recruitment of anti-bacterial F4/80 low MF was unperturbed by the presence of large numbers of helminth-expanded F4/80 high MF.
We further confirmed the blood monocyte origin of F4/80 low cells using CCR-2 deficient mice, which fail to recruit MF to inflammatory sites in part due to defective egress of monocytes from the bone marrow [26]. Recruitment of F4/80 low MF in response to SL3261 was ablated in Ccr2-/-animals both in single and co-infection whereas expansion of resident F4/80 high MF was unaltered during H. polygyrus infection (Fig 4C & 4D). Of note the disappearance of F4/80 high cells observed in wildtype mice was much less pronounced in co-infected Ccr2-/-animals indicating a link between recruitment of F4/80 low MF and the disappearance reaction. Enhanced cell death is likely part of the explanation for the loss of F4/80 high MF as indicated by increased Annexin V staining specifically in this population following SL3261 single or co-infection in wild-type mice (S6 Fig). Thus, F4/80 high nematode-expanded MF did not convert to an F4/80 low phenotype but were displaced from the peritoneal cavity during bacterial co-infection by new, blood monocyte derived MF. Notably, the magnitude of blood cell recruitment was the same in single bacterial or co-infected animals.

SL3261 infection induces a unique gene expression profile in F4/80 high MΦ
The recruitment of monocyte derived MF in conjunction with the displacement of F4/80 high MF following bacterial inoculation of H. polygyrus infected animals strongly suggested that these two cell populations have distinct capacities and functions during co-infection. To elucidate these differences we subjected F4/80 high and F4/80 low MF populations from co-infected animals to microarray gene expression analysis. Both populations were isolated from the same animal by fluorescence activated cell sorting (Fig 5A). F4/80 high MF from naïve or H. polygyrus single infected as well as F4/80 low MF from SL3261 single infected animals were used as controls. The quality and reproducibility of the data was confirmed by hierarchical clustering analysis of global expression profiles, which grouped according to biological conditions (S7 Fig). Principal component analysis showed F4/80 high MF in naïve and H. polygyrus infected animals clustering together, while F4/80 low MF in single SL3261 or co-infected animals clustered together, reaffirming the different tissue origins of these cell populations. Notably, F4/80 high MF in co-infected animals clustered separately from either of these populations revealing a unique response profile in response to bacterial infection (Fig 5B).
To specifically address whether F4/80 high MF could effectively contribute to resistance against SL3261 infection we compared expression of known resistance-associated genes [27] across all experimental groups (Fig 5C). F4/80 low MF from both single SL3261 and co-infected animals showed similar, largely enhanced expression of these genes. F4/80 high MF in coinfected animals also showed enhanced expression of most resistance genes, including Nos2, as compared to MF from naïve and H. polygyrus infected animals, confirming plastic adaptation of the activation phenotype. However, the overall response in F4/80 high MF was muted and most resistance-associated genes had less differential expression than in F4/80 low MF. Specifically Slc11a1 (i.e. Nramp), which was significantly upregulated in F4/80 low MF by bacterial infection, was virtually unchanged in tissue resident derived MF. Co-infection also induced a divergence with regard to the production of IL-1 in which Il1b expression was highly increased in F4/80 low MF while Il1a expression was dramatically increased in F4/80 high MF.
To further highlight functional differences, we performed pathway analyses on differentially expressed (DE) genes (q-value (FDR)<0.01, log2FC ±0.5) in a pairwise comparison of F4/80 high and F4/80 low MF from co-infected animals utilising Ingenuity Pathway Analysis (IPA). To avoid bias caused by different cellular origins, we compared the list of DE genes between F4/80 low and F4/80 high MF in our dataset with DE genes in F4/80 low and F4/80 high MF in naïve animals obtained from publicly available datasets (Immunological Genome Project, http://www.immgen.org [28]) and restricted the analysis to genes unique to the coinfection setting (S8 Fig & S1 Table). This analysis revealed enrichment of DE genes in several pathways associated with anti-bacterial or pro-inflammatory responses (Fig 5D). The pattern of gene expression between F4/80 low and F4/80 high MF within these pathways provided further evidence that the anti-bacterial response was less potent in F4/80 high MF during co-infection. Of note, pairwise comparison of F4/80 high MF from co-infected vs F4/80 high MF from single H. polygyrus infected animals revealed anti-microbial pathways as differentially up-regulated (Fig 5E). Once again, this confirmed that F4/80 high MF during co-infection adopt anti-microbial characteristics. Interestingly, other affected pathways in  Overall the gene expression analysis revealed that F4/80 high , parasite expanded MF altered their activation phenotype to adapt to the bacterial infection. However, the anti-bacterial response was limited in comparison to monocyte-derived cells.

H. polygyrus expanded F4/80 high MΦ retain tissue-sentinel / neutrophil recruiting properties
Although F4/80 high MF showed an overall limited anti-bacterial response, certain genes did not follow this pattern (Fig 5C). Specifically Il1a showed divergent, unexpectedly enhanced expression (Fig 5C) in F4/80 high MF from co-infected animals. This made us question whether these cells adopted functionalities other than anti-bacterial effector mechanisms. In this context Schiwon et al. recently highlighted a two-step model of inflammation with tissue resident Ly6C-and newly recruited Ly6C+ MF adopting different, non-redundant roles in the recruitment and activation of neutrophils during uropathogenic E. coli infection [29]. In line with these findings F4/80 high MF from co-infected animals in our model expressed enhanced levels of key neutrophil chemotactic factors (e.g. Cxcl1, Cxcl2, Pf4) [30] or enzymes involved in the generation of neutrophil chemotactic factors (e.g. Alox5, Ptgs1) (Fig 6A & 6B). Thus, F4/80 high , helminth expanded tissue resident derived MF seemed to retain their tissue sentinel capacity and may contribute to resistance to bacterial infection through recruitment of neutrophils and other inflammatory cells. The distinct capacity of F4/80 high resident derived MF to promote neutrophil recruitment was supported by data from Ccr2-/-mice. Recruitment of neutrophils was only marginally reduced in SL3261 infected or co-infected Ccr2-/-mice despite failing to recruit F4/80 low MF (Figs 6C & 4C).

Discussion
Inflammatory MF are commonly recruited through influx and differentiation of blood monocytes [12,31] whereas some helminth infections, and Th2 cytokines, lead to the proliferative expansion of tissue resident MF [10]. The functional importance of this divergence in recruitment is not completely clear, yet helminth infections can last years and avoidance of chronic inflammation might be the evolutionary driver of this phenomenon [10,32]. Furthermore MF are well known to flexibly adapt their activation phenotype to changes in their environment in vitro [33,34]. Nonetheless, limited data exist on the true plasticity of individual cells in vivo, and on the relative importance of plasticity vs. cellular recruitment.
Here we show that H. polygyrus-expanded peritoneal MF can effectively upregulate antibacterial defense mechanisms (i.e. NOS2, MHC-II) in response to bacterial stimulation in vitro and in vivo. Importantly, plasticity was not restricted to newly recruited or previously non-activated cells. However, the magnitude of the response was limited in the resident population and MF instead adopted specific characteristics dependent on their tissue origins. Thus, plasticity of MF activation as defined by a change in activation phenotype did exist in vivo, albeit with certain restrictions on the degree of repolarization. Limitations on MF activation make evolutionary sense to allow fine tuning of the ensuing immune response depending on the persistence and virulence of the invading pathogen [29].
Altered, often reduced responses of MF to a later challenge following primary stimulation have been described before [35]. IL-4 and IFNγ have been shown to induce distinct non-overlapping enhancers of gene transcription which are retained even when stimulation ceases [36]. Hence stimulation of MF can generate an epigenetic memory, which influences and dictates future responses [37,38]. Indeed, helminth infections are in general assumed to impart a detrimental effect on resistance to bacterial infections in part by altering MF activation [19,[39][40][41][42]. This is also evident in our findings. F4/80 low MF from co-infected animals showed reduced induction of anti-bacterial effector genes as compared to F4/80 low MF from single SL3261 infected animals. However, the limiting effect of previous cytokine exposure was overlaid by the much more profound effect of the immediate origin of the MF (tissue resident vs. blood monocyte derived). F4/80 high and F4/80 low MF in co-infected animals responded in unique ways to bacterial infection indicating differing functional roles. Notably, tissue resident colon MF, although originally blood monocyte derived, have been shown to maintain a similar restricted and preferentially anti-inflammatory phenotype in the face of inflammation [43] as do the peritoneal MF discussed here. Furthermore exposure of MF to tissue environmental factors has been described to affect and shape MF responses to infection [16]. Therefore, independent of embryonic or bone marrow derived origins, tissue MF responses seem largely determined by previous exposure to tissue factors and adoption of a resident phenotype. Thus, plasticity of MF activation as defined by the adoption of a full anti-bacterial phenotype by helminth expanded MF did not happen in vivo. Rather tissue resident derived MF, whether expanded by helminth infection or not, responded in a unique, non-redundant fashion to the bacterial infection likely necessary for an optimal induction of the immune response.
Whether these unique properties of tissue resident MF are of functional relevance to the expansion of these cells in some helminth infections [10] but not others [9] remains unclear.
Previous studies suggest that early innate recognition of bacteria has the potential to overcome the normally detrimental impact of helminth infection on resistance to bacterial infection [44,45]. Furthermore helminth expanded resident MF have been shown to exert a protective effect in models of sepsis [46,47]. Thus, expansion of these cells might serve the dual purpose of rapid initiation of anti-bacterial effector mechanisms while avoiding excessive, potentially self-harming immune activation dependent on the virulence and persistence of the invading pathogen.
In this context it is interesting to note that in our co-infection experiments the disappearance reaction of tissue resident MF seemed linked to the recruitment of monocytes and monocyte derived MF. Although the exact mechanism remains unclear, recruited MF appear to displace the resident population in a time and dose dependent manner likely through the induction of apoptosis. In the light of their activation plasticity and non-redundant role in triggering an anti-bacterial immune response their displacement raises the question of why tissue resident MF are removed in a persisting bacterial infection? In line with the muted proinflammatory response found here the MF disappearance reaction might reflect differences of tissue resident and recruited MF in the capacity to deal with intracellular pathogens [48,49] or their interaction with the adaptive immune system [50]. Thus, although relevant in early phases of the immune response, persistence of tissue resident MF may render the host more susceptible to the infection or dampen the ensuing T cell response.
Taken together our data indicate that plastic adaptation of MF responses to consecutive coinfections does occur in vivo but is outweighed by cellular recruitment due to functional restrictions imposed by the tissue origin of the MF.

Ethics statement
All animal experiments were performed in accordance with the UK Animals (Scientific Procedures) Act of 1986 under a Project License (60/4104) granted by the UK Home Office and approved by the University of Edinburgh Ethical Review Committee. Euthanasia was performed by giving an overdose of anaesthetic (Ketamine/Medetomidine; 1/1; v/v).

Mice and infection
C57BL/6 mice were bred and maintained in specific pathogen-free facilities at the University of Edinburgh. Experimental mice were age and sex matched.
Heligmosomoides polygyrus bakeri life cycle was maintained in house and infective thirdstage larvae (L3) were obtained as described elsewhere [51]. Mice were infected with 200 H. polygyrus L3 by oral gavage. Fecal egg burden was determined on day 13 of the infection using a McMaster counting chamber (Hawksley).
The attenuated, aroA deficient Salmonella enterica serovar Typhimurium strain SL3261 [52] was cultured as stationary overnight culture from frozen stock in Luria-Bertani broth. Unless indicated otherwise animals were injected i.p. with~3-5x10^4 CFU diluted in PBS. Infectious doses and splenic bacterial burdens were enumerated by plating inocula or tissue homogenates in 10-fold serial dilutions in PBS on LB-Agar plates.
For in vitro experiments mice were injected with 400 μL 4% Brewer modified thioglycollate medium (BD Biosciences) three days prior to necropsy.

Cell-isolation
Mice were sacrificed by exsanguination via the brachial artery under terminal anesthesia. After sacrifice, peritoneal cavity exudate cells (PEC) were obtained by washing the cavity with 9 mL lavage media comprised of RPMI 1640 containing 1% normal mouse serum (AbD serotec), 100 U/mL penicillin and 100 μg/mL streptomycin. Erythrocytes were removed by incubating with red blood cell lysis buffer (Sigma Aldrich). Cellular content was assessed by cell counting using a Cellometer Auto T4 Cell Counter (Nexcelom Bioscience) in combination with multicolor flow cytometry. Erythrocytes in blood samples were lysed using FACS Lyse solution (BD Biosciences). All antibodies were purchased from Biolegend UK unless stated otherwise.

Flow cytometry
Detection of intracellular Relm-α and NOS2 was performed directly ex vivo. Cells were stained for surface markers then fixed with 2% paraformaldehyde (Sigma Aldrich) and permeabilized using Permeabilization Buffer (eBioscience). Cells were then stained with purified polyclonal rabbit anti-Relm-α (PeproTech) or directly labeled Abs to NOS2 (CXNFT; eBioscience), followed by Zenon anti-rabbit reagent (Life Technologies). Expression of Relm-α and NOS2 was determined relative to appropriate polyclonal or monoclonal isotype control.
Samples were acquired on a BD LSR II or BD FACSCanto II using BD FACSDiva software (BD Bioscience) and post-acquisition analysis performed using FlowJo v9 software (Tree Star Inc.).

Partially protected bone marrow chimeras
Bone marrow chimeric mice were constructed by exposing anaesthetized C57BL/6 Cd45.1 mice to a single dose of 11.5 cGy γ radiation while shielding the abdomen, thorax, head and fore limbs with a 2-inch lead screen followed by i.v. injection of 5 x 10^6 donor bone marrow cells from congeneic Cd45.2 mice. Chimeric mice were left for 8 weeks before further experimental manipulation.

RNA-isolation and microarray analysis
MF were purified using FACS-sorting on a FACSAria cell sorter (BD Biosciences) according to their expression of surface molecules (F4/80+, SiglecF-, CD11b+, CD11c-, B220-, CD3-; all antibodies purchased from BioLegend or eBioscience) reaching purities of above 96%. Isolated MF were frozen at -70C and total RNA isolated using RNeasy mini columns (Qiagen). Sample preparation, quality control, running the microarray and initial bioinformatics analysis were carried out by the Bioinformatics and Genomic Technologies Core Facilities at the University of Manchester. In brief 10 ng of total RNA were converted to cDNA using the GeneChip WT Pico Kit (Affymetrix) and hybridized to Affymetrix GeneChip Mouse Gene 1.0 ST Array according to the manufacturer's instructions. Mouse Transcriptome Assay 1.0 data were processed and analysed using Partek Genomics Solution (version 6.6, Copyright 2009, Partek Inc., St. Charles, MO, USA) with the following options: probesets were quantile normalised and RMA background correction applied. Probesets were summarised to genes by calculating the means (log 2). Validation and gene enrichment strategies consisted of the following steps.
Step 1, to establish relationships and compare variability between replicate arrays and experimental conditions, principal components analysis (PCA) was used. PCA was chosen for its ability to reduce the effective dimensionality of complex gene-expression space without significant loss of information [54].
Step 2, Differential expression analysis was performed on annotated genes with Limma using the functions lmFit and eBayes [55]. Gene lists of differentially expressed genes were controlled for false discovery rate (fdr) errors using the Benjamini-Hochberg procedure [56].
Step 3, functional annotation of gene lists containing significantly differentially expressed genes was done with QIAGEN's Ingenuity Pathway Analysis (IPA, QIAGEN Redwood City, www.qiagen.com/ingenuity). Microarray data have been deposited to NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE85805.

Statistical analysis
Statistical analysis was performed using Prism 6 for Mac OS X (v6.0f, GraphPad Software). Differences between groups were determined by t-test or ANOVA followed by Tukey's or Dunn's multiple comparison-test. In some cases data was log-transformed to achieve normal distribution as determined by optical examination of residuals. Where this was not possible a Mann-Whitney or Kruskal-Wallis test was used. Comparison of activation markers of F4/80 high / low or Relm-α positive / negative MF within one experimental animal were considered as paired observations. Differences were assumed statistically significant for P values of less than 0.05.