Differential regulation of iron homeostasis during human macrophage polarized activation

Authors


Abstract

Iron metabolism in inflammation has been mostly characterized in macrophages exposed to pathogens or inflammatory conditions, mimicked by the combined action of LPS and IFN-γ (M1 polarization). However, macrophages can undergo an alternative type of activation stimulated by Th2 cytokines, and acquire a role in cell growth and tissue repair control (M2 polarization). We characterized the expression of genes related to iron homeostasis in fully differentiated unpolarized (M0), M1 and M2 human macrophages. The molecular signature of the M1 macrophages showed changes in gene expression (ferroportin repression and H ferritin induction) that favour iron sequestration in the reticuloendothelial system, a hallmark of inflammatory disorders, whereas the M2 macrophages had an expression profile (ferroportin upregulation and the downregulation of H ferritin and heme oxygenase) that enhanced iron release. The conditioned media from M2 macrophages promoted cell proliferation more efficiently than those of M1 cells and the effect was blunted by iron chelation. The role of ferroportin-mediated iron release was demonstrated by the absence of differences from the media of macrophages of a patient with loss of function ferroportin mutation. The distinct regulation of iron homeostasis in M2 macrophages provides insights into their role under pathophysiological conditions.

Introduction

Macrophages play a critical role in body iron homeostasis by recovering iron from old red blood cells and returning it to the circulation for binding to transferrin, which delivers the metal to the cells that need it for various functions, thus contributing more than 80% to daily iron turnover 1–3.

Iron retention in the reticuloendothelial system is the main response of body iron homeostasis to inflammation and is regarded as a host's attempt to withhold iron from the invading pathogens 4. This restricts iron availability for erythroid progenitor cells and may contribute toward causing the common condition of inflammation-related anaemia 1, 5, 6. Increased iron retention within inflammatory macrophages is due to increased iron uptake and decreased iron export 7, and is favoured by the induction of the iron storage protein ferritin (Ft) 8, 9. The blockade of macrophage iron release is mainly due to the interaction between the acute phase protein hepcidin and the iron exporter ferroportin (Fpn) 1–3, as the increase in circulating hepcidin triggered by inflammatory cytokines causes the internalization and degradation of Fpn 10, the exporter of non-heme iron 11, thus blocking iron release from macrophages. Macrophage Fpn is also negatively regulated at transcriptional and post-transcriptional levels by inflammatory mediators 12–14. It is still unknown whether the inflammatory response affects the feline leukemia virus, subgroup C, receptor that exports heme from macrophages 15 and whether the heme transporter HRG1 proteins play a role in macrophage iron metabolism 16. On the other hand, the downregulation of HCP-1, a folate transporter and heme carrier protein that transports iron out of the endosome, may impair iron recycling and contribute to greater iron sequestration within inflammatory macrophages 14.

The effect of inflammatory mediators on macrophage iron uptake has not been clearly defined. In particular, the role of transferrin receptor 1 (TfR1), which mediates transferrin-bound iron uptake 1–3, is not fully understood. Studies of murine and human reticuloendothelial cells 8, 17–19 have shown that, despite an early NF-κB and HIF-1-dependent transcriptional upregulation 20, TfR1 expression is post-transcriptionally downmodulated by exposure to inflammatory stimuli for 10–24 h, an effect that would impair a major iron uptake mechanism. However, inflammation may increase the uptake of other forms of iron. Enhanced divalent metal transporter1-mediated acquisition of non-transferrin-bound iron has been described in cytokine-stimulated human monocytic cell lines 19 and the increased erythrophagocytosis reported in TNF-α-stimulated macrophages could increase the acquisition of heme iron 21.

Most of the information available concerning the molecular regulation of macrophage iron metabolism during inflammation has come from studies in which macrophage cell lines or freshly isolated monocytes/macrophages have been challenged with LPS/IFN-γ, a model system that mimicks exposure to micro-organisms and/or Th1 pro-inflammatory cytokines (see 6 for a recent review). However, cells of the monocyte-macrophage lineage are characterized by marked phenotypical and functional heterogeneity 22–24. Classical activation by microbial agents and/or Th1 cytokines is associated with the production of oxygen radicals and the pro-inflammatory cytokines involved in cytotoxicity and microbial killing (M1 polarization), but macrophages can also follow a different activation pathway after stimulation with the Th2 cytokines IL-4 or IL-13 (M2 polarization) 22–24. A role in regulating adaptive immunity and in the control of cell growth and tissue repair has been suggested for these latter cells. Moreover, tumour- associated macrophages (TAM) that infiltrate tumours, and are involved in tumour growth, progression and invasion, also acquire a polarized M2-like phenotype 22. All of these findings suggest that investigating whether iron homeostasis is regulated differently in M1 and M2 macrophages may provide a better understanding of the changes in iron metabolism that take place under inflammatory conditions, particularly those characterized by the presence of M2 macrophages, such as tissue repair and tumour growth.

The aim of this study was to characterize the differential expression of the genes involved in iron homeostasis and changes in iron trafficking in M1 and M2 polarized human macrophages.

Results

Differentially expressed genes of iron metabolism in M1 and M2

We recently reported the transcriptional profiles associated with macrophage M1 or M2 cell polarization induced by exposure to LPS plus IFN-γ or IL-4 for 18 h 25. From this data set we extracted a total of 54 genes related to iron metabolism (see Table 1). Analysis of this data set revealed that a number of these genes were differentially expressed during polarized macrophage activation, as shown in Fig. 1 where red indicates upregulation and green downregulation. The mRNA levels of 21 genes were higher in M2 macrophages and thus were assigned to the M2 phenotype, whereas 12 genes were preferentially expressed in M1 cells and hence labelled with the M1 phenotype.

Figure 1.

Cluster analysis of iron metabolism-related genes. The expression of genes involved in iron metabolism extracted from a previously published transcriptional profile performed on human macrophages from three donors (D1, D2, D3) and polarized to M1 and M2 macrophages for 18 h 25. The genes differentially expressed in M1 and M2 macrophages are highlighted in the phenotype column. The procedure details, including the statistical analysis, are reported in the Materials and methods section. See Table 3 for details on GO categories.

Table 1. Genes involved in iron metabolisma)
 GO categoryGenes within categoryGenes regulated in M1/M2
  • a)

    a) A total of 54 genes related to iron metabolism have been identified according to GO categories with relevance for iron homeostasis included in Molecular Functions (A–H) or Biological Processes (I–P).

  • b)

    b) A group of genes included in the analysis according to literature evidence. The last column reports the number of genes in each category differentially expressed in M1/M2 macrophages (FDR<0.05; fold change>1.5).

AGO:0004322-ferroxidase activity11
BGO:0005381-iron ion transmembrane transporter activity11
CGO:0005506-iron ion binding63
DGO:0008198-ferrous iron binding32
EGO:0008199-ferric iron binding42
FGO:0034986-iron chaperone activity11
GGO:0051537-2 iron, 2 sulfur cluster binding11
HGO:0015232-heme transporter activity20
IGO:0006826-iron ion transport97
JGO:0006879-cellular iron ion homeostasis159
KGO:0006880-intracellular sequestering of iron ion22
LGO:0010039-response to iron ion32
MGO:0016226-iron-sulfur cluster assembly11
NGO:0018283-iron incorporation into metallo-sulphur cluster22
OGO:0055072-iron ion homeostasis11
PGO:0006783-heme biosynthetic process77
b)Genes from literature92
 Total (without redundancy)5433

Among these differentially expressed genes, we focused our interest on heme oxygenase 1 (HO1), ceruloplasmin (Cp) and Fpn, because they had the most significant differences in terms of microarray analysis, are directly involved in macrophage iron traffic and are known to be regulated by inflammatory stimuli. These genes were analysed by means of quantitative RT-PCR in M1 and M2 macrophages polarized for 18 h (in line with our previous study 25), and in differentiated but non-polarized cells (M0); their differential expression was also tested at protein level by immunoblotting at later times (72 h after addition of the polarising stimuli, in line with our previous findings showing a lag between transcription and protein accumulation 25).

In comparison with M0 cells, the transcript levels of HO1, which catalyses the breakdown of heme into biliverdin, carbon monoxide and iron 26, were decreased in M1 and increased in M2 cells, thus leading to a considerable difference between the two populations (Fig. 2A). Immunoblotting analysis showed that the differences in HO1 expression were maintained at protein level (Fig. 2B).

Figure 2.

Differential expression of heme oxygenase (HO1) and ceruloplasmin (Cp) in polarized macrophages. (A and C) HO1 mRNA levels in M0, M1 and M2 macrophages 18 h after polarization. The expression was normalized to GAPDH and expressed in relation to the M0 cells, arbitrarily defined as 1. (Mean+SD. *p<0.001, p<0.05, n=4). (B and D) Immunoblot analysis of proteins from M0, M1 and M2 macrophages 72 h after polarization reacted with antibodies against HO1 and Cp, respectively. Blot B was re-probed with an antibody against β-actin as a loading control. Blot D was stained with amido black to assess equal protein loading. Each panel shows one representative blot and the densitometric quantification (mean+SD of M0 values. *p<0.001, n=5).

The mRNA levels of Cp, a serum copper-containing protein endowed with ferroxidase activity that is involved in iron mobilization 27, were induced by LPS/IFN-γ, but not significantly modulated by IL-4 (Fig. 2C) and immunoblotting analysis of the Cp released in the culture medium revealed a similar pattern, but the differences were less marked (Fig. 2D).

In line with previous results obtained in mouse spleen macrophages 12, human monocytic cell lines 19 and human monocyte-derived macrophages 14, LPS/IFN-γ stimulation of human macrophages led to much lower Fpn expression levels than in unpolarized M0 cells; however, M2 polarization repressed Fpn less, and thus there was a tenfold difference in Fpn mRNA expression between M1 and M2 macrophages (Fig. 3A). Immunoblotting with a previously validated antibody 28 confirmed the downregulation in M1 cells, albeit to a lesser extent than at the mRNA level, whereas Fpn protein levels were increased in M2 macrophages (Fig. 3B). Therefore, we found a considerable difference in Fpn levels between M1 and M2 macrophages both at the mRNA and protein levels. The stronger difference in Fpn at the mRNA level (analysed at 18 h) than at the protein level (analysed at 72 h), prompted us to evaluate Fpn expression at an intermediate time point. Analysis of both Fpn mRNA and protein at 48 h (Fig 3C and D) showed a difference between the various macrophage populations similar to that observed at the other time points, thus indicating that the different extent of Fpn modulation at the mRNA and protein levels is maintained at times ranging from 18 to 72 h after polarization.

Figure 3.

Differential expression and subcellular localization of ferroportin in polarized macrophages HO1 and Cp. (A and C) HO1 and Cp mRNA levels, in M0, M1 and M2 macrophages 18 and 48 h, respectively, after polarization. The expression was normalized to GAPDH and expressed in relation to the M0 cells, arbitrarily defined as 1. (Mean+SD. *p<0.001, p<0.05, n=4 for (A) and 3 for (C). (B and D) Immunoblot analysis of proteins from M0, M1 and M2 macrophages 72 and 48 h, respectively, after polarization reacted with antibodies against Fpn and β-actin as a loading control. Each panel shows one representative blot and the densitometric quantification (mean+SD of M0 values. *p<0.001, n=5 for (B) and 3 for (D). (E) Immunoblot analysis of membrane Fpn from M0, M1 and M2 macrophages 72 h after polarization. Proteins from membrane preparations were reacted with antibodies against Fpn. The blots were stained with amido black to assess equal protein loading. The panel shows one representative blot and the densitometric quantification (mean+SD of M0 values. *p<0.001, n=3).

As Fpn functions on the plasma membrane where it translocates after iron loading 29, we examined expression of Fpn in the membrane fraction. Figure 3E shows that Fpn protein levels were 30% lower in M1 cells and 2.5-fold higher in M2 macrophages, as compared with unpolarized macrophages.

We also analysed ferritin, TfR1 and iron regulatory proteins (IRP), which are proteins of iron metabolism important in intracellular iron homeostasis and known to be primarily regulated at the post-transcriptional level 30. Microarray analysis did show slightly higher L-Ft expression in M2 cells (see Fig. 1), but this was not confirmed by RT-PCR analysis (results not shown), and ELISA (Table 2) confirmed that the expression of L-Ft was not significantly different between the three macrophage populations. On the other hand, as suggested by the slight increase in H-Ft expression at microarray analysis and in line with previous findings in LPS/IFN-γ-treated human and mouse macrophages 8, 17, H-Ft levels were significantly higher in the M1 macrophages than in the M0 cells, but lower than in both M1 and M0 in the M2 macrophages (Table 2). This opposite regulation led to a remarkable difference in H-Ft levels between M1 and M2.

Table 2. Ft content of polarized macrophagesa)
 L-Ft (ng/mg)H-Ft (ng/mg)
  • a)

    a) The amounts of L-Ft and H-Ft in macrophage cell extracts were determined by ELISA as described in the Materials and methods. Mean values (±SD) for ten participants. *p<0.001 versus M0; **p<0.05 versus M0.

M01360±220623±100
M11426±205822±141*
M21345±211548±99**

Immunoblotting analysis showed that TfR1, the major iron uptake protein, was downregulated in M1 cells and slightly upregulated in M2 cells as compared with M0 macrophages (Fig. 4A). These results are in line with the difference in TfR1 expression previously found between IFN-γ-activated and IL-4-treated mouse macrophages 31.

Figure 4.

Differential expression of transferrin receptor (TfR1) and IRP in polarized macrophages. (A) Immunoblot analysis of proteins from M0, M1 and M2 macrophages 72 h after polarization reacted with antibodies against TfR1 and β-actin. One representative blot and the densitometric quantification (mean +SD of M0 values. *p<0.001, **p<0.005, p<0.05, n=5) are shown. (B) RNA-bandshift analysis of IRP activity. Extracts from M0, M1 and M2 macrophages 48 h after polarization were incubated with a 32P-labelled IRE probe and RNA−protein complexes separated on non-denaturing gels. One representative autoradiogram and the densitometric quantification (mean+SD of M0 values. *p<0.001, p<0.05, n=4) are shown. (C) Immunoblot analysis of proteins from M0, M1 and M2 macrophage 72 h after polarization reacted with antibodies against IRP2 and β-actin. One representative blot and the densitometric quantification (mean+SD of M0 values. *p<0.001, n=5) are shown.

We then performed RNA-bandshift assays to assess the activity of IRP, which regulate intracellular iron homeostasis by binding to iron-responsive elements (IRE) in target mRNA 30. In comparison with M0 macrophages, the combined binding activity of IRP1 and IRP2, which co-migrate in bandshift assays, was lower in cells exposed to LPS/IFN-γ (M1) (in line with previous results 17) and slightly increased in M2 (Fig. 4B). Accordingly, immunoblotting of cell extracts showed that IRP2 protein levels were higher in M2 macrophages than in M0 and M1 cells (Fig. 4C).

Although we cannot rule out that the decreased IRP2-binding activity in M1 macrophages is, at least in part, due to factors other than iron 32, taken together the above results indicate that M2 macrophages have lower iron levels than M1 macrophages, possibly because of their higher Fpn expression.

Iron release is higher in M2 macrophages

To assess whether the difference in the expression of Fpn, which is the main and possibly exclusive iron exporter 11, is functionally involved in modulating iron release, we incubated the macrophages with the iron donor transferrin labelled with 55Fe, and then monitored the kinetics of cell 55Fe release into the media in the presence of the non-permeable iron chelator high molecular weight desferrioxamine (HM-DFO), which better solubilizes iron and prevents re-uptake 33. Measurements of iron release showed that after 4 h M2 cells released approximately four times as much iron as M1, and M0 cells showed an intermediate activity (Fig. 5). Trypan blue exclusion testing showed that the integrity of the cell membrane was not affected by the various treatments (results not shown). These findings suggest that the lower expression of Fpn in M1 macrophages leads to iron retention, whereas M2 polarization leads to more efficient Fpn-dependent iron export.

Figure 5.

Iron release activity of polarized macrophages. Release of 55Fe in medium from M0, M1 and M2 macrophages 72 h after polarization was measured as described in the Materials and methods and is expressed as percentage of the total amount of 55Fe (cells+supernatant). Mean±SD of initial values, n=6 *p<0.001.

Ferroportin-mediated iron release affects the capacity of conditioned media to sustain cell growth

In order to gain insight into the biological implications of the different iron release and storage capacity of the M1 and M2 macrophage populations, we hypothesized that (in vivo) the iron released by polarized macrophages could affect the proliferation of cells present in the tissue microenvironment, because iron is an essential cofactor for DNA synthesis 4. In line with this, the addition of small (micromolar range) amounts of iron to the culture medium dose-dependently increased cell proliferation in the RCC10 renal carcinoma cell line (Fig. 6A) and NIH3T3 fibroblasts (results not shown). Conversely, the addition of the cell-impermeant iron chelator HM-DFO, which did not significantly affect cell viability when present alone (results not shown), abolished the enhancing effect of the iron supplementation. As these results suggested that small increases in the availability of iron accelerate cell growth rate, we used this in vitro assay to test the cell growth promoting capacity of polarized macrophages. Twenty-four hours' incubation with the conditioned media of the different macrophage populations had significant effects on the growth rate of RCC10 cells: in comparison with M0-conditioned medium, the number of viable cells decreased 30% in M1-conditioned medium and increased 40% in M2-conditioned medium (Fig. 6B), thus leading to a twofold difference in cell number between M1 and M2. Cell counting as an alternative method to measure cell growth confirmed these observations (results not shown). The same differences were observed after incubation for 48 h (results not shown). Importantly, the incubation of macrophages in the presence of HM-DFO reduced the proliferation induced by the M2-conditioned medium by 40%, and that induced by the M0- and M1-conditioned media by 10–15%. Therefore, iron chelation blunted the differences in RCC10 cell growth obtained with the various conditioned media. Conversely, the addition of 1 μM ferric ammonium citrate (FAC) restored the capacity of the M1-conditioned medium to sustain cell proliferation, but did not show significant effects on the M2-conditioned medium (Fig. 6B). The results were similar in the case of NIH3T3 cells, thus indicating that the effect of the different macrophage-conditioned media on cell growth was not restricted to the RCC10 tumour cell line (Fig. 6C). These findings suggested that differences in Fpn-mediated iron release (see Fig. 5) may contribute to the distinct effect of the various macrophage-conditioned media on cell growth. In order to substantiate this hypothesis, monocytes from a patient with Fpn disease, which is characterized by a loss of function mutation in the Fpn gene (L233P) that prevents cell surface localization and iron release 34, were purified and polarized towards M1 and M2. The significant differences in growth observed with the media of polarized M1 and M2 macrophages prepared from normal subjects (see Figs. 6B and C) were absent when RCC10 (Fig. 6D) and NIH-3T3 cells (Fig. 6E) were grown in the presence of the media of the two macrophage populations purified from the patient.

Figure 6.

Ferroportin-mediated iron release affects the capacity of conditioned medium of polarized macrophages to sustain cell proliferation. (A) RCC10 cells proliferation (evaluated by MTT assay) after addition of FAC in the presence and absence of HM-DFO. The values (mean±SD, *p<0.001, p<0.05, n=4) are expressed as a percentage of the growth of cells incubated in medium alone. (B) RCC10 cells proliferation after incubation with conditioned media from M0, M1 and M2 macrophages 48 h after polarization. When appropriate, 1 μM FAC or HM-DFO were added. The values (mean+SD, *p<0.001, n=7) are expressed as percentage of the growth of cells incubated with conditioned media from M0 macrophages. (C) NIH3T3 cells proliferation evaluated as described for (B) (mean+SD, *p<0.001, n=7). (D) RCC10 cells proliferation after incubation with conditioned media from M0, M1 and M2 macrophages, purified from a patient with Fpn disease, 48 h after polarization. The values (mean+SD) of three experiments using different macrophage preparations from the same subject are expressed as percentage of the growth of cells incubated with media from M0 macrophages. (E) NIH3T3 cells proliferation evaluated as described for (D).

Discussion

In the context of a specific immune response, the cytokine milieu compels mononuclear phagocytes to express specialized and polarized functional properties 22–24. Activated polarized M1 macrophages, which are induced by IFN-γ alone or together with microbial stimuli such as LPS or cytokines such as TNF-α and GM-CSF, are proficient producers of effector molecules and inflammatory cytokines, and mediate resistance against intracellular parasites and tumours 22–24. On the contrary, the alternative M2 form of macrophage activation arising from exposure to IL-4 or IL-13 and various other stimuli generally shows a high degree of scavenging capacity, facilitates tissue repair and angiogenesis, and favours tumour progression 22–24.

The expression profile of differentially polarized human macrophages has been characterized in relation to cytokines, chemokines and growth factors 25. We investigated the genes associated with iron homeostasis and found considerable differences in the expression of genes involved in iron storage (Ft), traffic (Fpn, Cp) and regulation (IRP2, HO1), which also led to large differences in both intra- and extracellular iron availability. The differential expression of several genes involved in heme biosynthesis was noted, but the characterization of this pathway was beyond the aim of this study.

The molecular signature of M1 macrophages provided by our study extends previous findings relating to the expression of individual proteins in human cell lines and primary monocytes/macrophages exposed to pro-inflammatory stimuli 6, and is substantially in line with previous data showing changes in gene expression (such as Fpn repression and Ft induction) that may favour iron sequestration in reticuloendothelial cells, a hallmark of inflammatory disorders 5, 6. The molecular control of iron accumulation in inflamed macrophages has been reviewed in detail elsewhere 6 and our data show that it can be also applied to M1 macrophages.

Conversely, the expression profile of iron genes in M2 macrophages is the opposite of that observed in M1 macrophages, and greatly enhances iron release. Although the effect of anti-inflammatory cytokines on the expression of some genes involved in iron metabolism has been previously investigated in mouse macrophages 35, to the best of our knowledge, this is the first extensive analysis of human M2 macrophages, a leukocyte population whose biological role is completely different from that of inflammatory macrophages 22–24.

The iron release-prone phenotype of M2 cells is mainly due to the upregulation of Fpn and HO1 and the downregulation of H-Ft. The molecular basis of the downregulation of H-Ft in M2 cells is possibly represented by the increased levels and binding activity of IRP2, the main regulator of intracellular iron homeostasis at the post-transcriptional level 30, although a small difference in H-Ft mRNA levels was revealed by the microarray analysis. Similarly, increased IRP2 expression possibly represents the molecular basis of the upregulation of TfR1 in M2 cells. Like the role of TfR1 downregulation in iron accumulation within reticuloendothelial cells under proinflammatory conditions (see Introduction), the significance of TfR1 upregulation in M2 macrophages remains to be determined.

HO1 expression is not controlled by IRP2, but its differential expression in polarized macrophages seems to be co-ordinately and functionally involved in determining their particular phenotype. The induction of HO1 expression in M2 macrophages is in line with findings showing that HO1 favours iron release 36 and massive iron overload in HO1 knock-out mice 37. On the other hand, its inhibition in M1 cells is consistent with findings indicating that it mediates the anti-inflammatory effect of IL-10 (see 26 for a review on the anti-inflammatory role of HO1).

Conversely, the role of Cp in the altered iron metabolism associated with polarized macrophages is unclear. As expected on the basis of previous evidence showing Cp induction in a monocytic cell line treated with IFN-γ 38, Cp was upregulated in human M1 macrophages, but substantially unaltered in M2 macrophages. These results are in line with clinical data indicating high Cp levels in patients with inflammatory disorders (reviewed in 27), but are not consistent with the protein's role in facilitating iron release that has recently been substantiated by the demonstration that Cp ferroxidase activity is necessary to facilitate Fpn-mediated iron export 2. Our finding that Cp expression in human polarized macrophages is the opposite of that of Fpn and iron release confirms the fact that further studies of its role in inflammation are necessary.

The Fpn gene showed the greatest differential regulation in the studied cell populations. The post-translational inhibition of Fpn mediated by liver-derived hepcidin is probably the key regulatory mechanism of Fpn expression under inflammatory conditions in vivo10. Furthermore, studies showing TLR4-dependent hepcidin expression in mice myeloid cells 39 and that Fpn expression at the cell membrane in inflammatory macrophages is negatively affected by the autocrine up-regulation of hepcidin 40, 41 indicate that “local” mechanisms may contribute to iron sequestration within monocytes. As our microarray analysis showed a slight difference in hepcidin mRNA levels, the divergent Fpn expression in polarized macrophages could be hepcidin-mediated, at least in part. However, we also found a considerable difference in Fpn mRNA level between M1 and M2 macrophages. Fpn expression is also subject to transcriptional control 4, 42, and decreased Fpn mRNA levels have previously been found under conditions of LPS-induced inflammation 12, 39. As Fpn is also post-transcriptionally regulated by the IRE/IRP system (see 30 for recent review), our finding of a less marked difference in Fpn protein levels between M1 and M2 macrophages than the difference in the corresponding mRNA at all the time points examined indicates that the Fpn gene is actively transcribed in M2 macrophages and that the impaired translation of Fpn mRNA due to increased IRP2-binding activity is not sufficient to abolish the increase in protein levels. The difference in Fpn regulation between M1 and M2 cells is in line with the recent demonstration that different mouse macrophage subpopulations positively or negatively regulate Fpn expression in response to IFN-γ and intracellular pathogens 43. It is worth noting that the differences in Fpn membrane levels among the various macrophage populations are more marked than those found in total cell extracts (see Fig. 3E). Moreover, iron release experiments (see Fig. 5) demonstrated that this differential expression is accompanied by functional effects on iron efflux.

The higher Fpn-dependent iron release appears to be the major determinant of the lower iron retaining capacity of M2 macrophages (see Fig. 5 and also Fig. 6 reporting experiments with cells purified from a Fpn-deficient patient). This is apparently not in line with the fact that M2 macrophages express 30–40% more TfR1 than M1 cells and hence should internalize more iron. However, greater iron uptake would presumably lead in turn to a higher iron export activity, because iron induces Fpn in macrophages 13, 29, 44.

Overall, our results suggest that M2 macrophages lack the typical iron-withholding mechanisms present in classically activated M1 macrophages. Moreover, increased iron availability in the extracellular milieu of M2 macrophages may have relevant physiopathological consequences (see below). Our cell growth experiments in the presence of conditioned media with or without iron chelator showed that the high level of iron release was important in sustaining cell proliferation of both tumoural and non-tumoural cell lines. The use of conditioned media of polarized cells obtained from a patient with an inactivating mutation resulting from a defect of Fpn trafficking to the cell surface that prevents iron export 34 further indicates the role of Fpn-mediated iron release in this setting.

Given the high iron requirement of many species of bacteria, it is thought that the retention of iron in reticuloendothelial cells under inflammatory conditions represents an antibacterial response 4, as also recently indicated by the finding that bacterial stimulation of macrophages triggered a TLR4-dependent reduction in Fpn 39. This defence mechanism directed against extracellular microrganisms could however facilitate the growth of intracellular bacteria 45, so that macrophages infected by pathogens (with the help of IFN-γ 46) activate mechanisms aimed at restricting iron availability to bacteria 47, 48.

In an attempt to answer the question as to what might be the pathophysiological implications of the differential regulation of iron metabolism in polarized macrophages and, in particular, the lower iron retaining capacity of M2 macrophages, we envision two possibilities. The first is that iron influences the distinct phenotypical heterogeneity of the two macrophage populations in terms of cytokine production. A study of inflammatory responses in a murine model of hemochromatosis has shown that low intracellular iron levels in macrophages of mutant mice impair the TLR4-activated signalling pathway 49, the translation of IL-6, TNF-α 50, and consequently attenuate the inflammatory response. In line with this finding, it has also been found that the synthesis of these cytokines in M2 macrophages is less than in M1 macrophages 25, and the capacity of some bacteria to reprogram macrophages toward an M2 profile is associated with the persistence of pathogens in tissues and the chronic evolution of infectious diseases 51.

The second possibility is that the greater iron releasing capacity of M2 macrophages may affect other cells in the microenvironment, such as neoplastic cells. Iron is an essential component of many of the proteins involved in cell growth and replication, and neoplastic cells require more iron than normal cells because they generally proliferate more rapidly. This is why iron chelators have shown promising anti-neoplastic activity in cell cultures and clinical trials (reviewed in 52). Interestingly, the recruitment of pathways that reduce the availability of intracellular iron is one of the mechanisms utilized by p53 to induce cell cycle arrest 53.

Circulating monocytes recruited from the circulation into the tumour differentiate into TAM, whose phenotype closely resembles that of M2 macrophages 22. The molecular basis of the TAM functions favouring tumour growth has been related to an immunosuppressive and tumour-promoting phenotype characterized by a distinct repertoire of growth factors, cytokines, chemokines. 22–24. The high Fpn-dependent iron release in M2 cells shown by our results suggests that an increase in Fpn expression may also occur in the TAM, which are closely related to M2 macrophages. Therefore, in an in vivo situation of tumour growth, a high level of iron release may provide an unrestricted source of iron for the multiplication of tumour cells and may thus represent a further and previously unknown mechanism underlying their tumour-promoting activity. Moreover, TAM accumulate in the hypoxic regions of tumours, and hypoxia triggers a HIF-1-dependent pro-angiogenic program in these cells 22. As iron is required for HIF-1 degradation 54, a relative iron deficiency may contribute to a (hypoxia-induced) increase in HIF-1 activity and the expression of its target genes (e.g. vascular endothelial growth factor) in M2 macrophages, and be an additional mechanism promoting tumour progression. Further studies are needed to address the role of TAM in these settings.

Materials and methods

Subjects

We studied 39 healthy blood donors aged 20–59 years (59% male) with no history of iron metabolism disorders and normal serum iron indices. A previously characterized patient with ferroportin disease was also investigated 34. Informed consent was obtained from all the donors, and the study protocol was approved by the Ethics Committee of the University of Milan and Verona.

Cell cultures and conditioned media

The monocytes were purified from buffy coats by means of sequential centrifugation (Ficoll-Paque followed by a solution consisting of RPMI 1640 and Percoll), resuspended in RPMI 1640 containing 20% human serum 17, and kept for 6 days in 5% CO2 at 37°C in the presence of 100 ng/mL M-CSF for differentiation to macrophages (M0). Macrophages were then stimulated for different times (18–72 h) in RPMI medium containing 5% FCS with 1 μg/mL LPS (E. coli strain 055:B5, Sigma) plus 100 U/mL IFN-γ (for M1 polarization) or 20 ng/mL IL-4 (for M2 polarization) 25. The conditioned media of macrophages polarized for 48 h were collected, centrifuged at 1000×g for 10 min, and aliquots were stored at −80°C. The cell culture media (Biochrom, Milano, Italy Cambridge, UK) and FCS (HyClone Logan, UT, USA) contained less than 0.125 endotoxin units/mL, (Microbiological Associates Farmington Hills, MI, USA). Recombinant human cytokines were obtained from PeproTech Neuilly-Sur-Seine France.

Clustering of iron homeostasis-related genes

The experimental details of transcriptional profile analysis have been reported previously 25. The genes of relevance for iron homeostasis were identified using the GeneOntology (GO) Biological Processes and Molecular Functions categories related to iron homeostasis (Table 1), which identified 54 genes. The GO annotations were retrieved from the consortium website (www.geneontology.org) using Biomart/ENSEMBL (http://www.ensembl.org/biomart/martview). Nine genes not included in the GO search but involved in iron homeostasis on the basis of published evidence were also included in the study (listed in Fig. 1 and Table 1). The expression data of selected genes were extracted from a transcriptional profile analysis data set performed in three independent human polarized macrophage preparations using the Human Genome U133 A and B arrays (HG-U133; Affymetrix Santa Clara, CA, USA) 25 (available at www.ncbi.nlm.nih.gov/geo under the accession number GSE5099). The expression data were analysed using Student't T test and false discovery rate (FDR) correction for multiple testing, with the genes with a FDR of <0.05 and a fold change of >1.5 being considered differentially expressed. A hierarchical complete linkage clustering analysis of this set of genes was made using Pearson's correlation as a similarity measure and MultiExperiment Viewer 3.0 software (Institute for Genomic Research Rockville, MD, USA) 55.

Cell growth analysis

Human renal clear cell carcinoma (RCC10) or mouse NIH-3T3 cells (2×104 cells/well) were seeded in quadruplicate in RPMI 1640 medium containing 5% FCS. The medium was changed after 24 h and the cells were grown in the conditioned media of M0, M1 or M2 macrophages in the presence or absence of 100 μM HM-DFO (Biomedical Frontiers, Minneapolis, MN, USA), a non-permeable hydroxyethyl starch conjugate of desferrioxamine. When appropriate, various amounts of FAC, in the presence or absence of HM-DFO, were added 24 h after seeding. After 24 or 48 h of growth, cell proliferation was evaluated using thiazolyl blue (MTT, Sigma) as an indicator of mitochondrial function as previously described 56. Alternatively, the cells were harvested with trypsin/EDTA and counted in triplicate using a hemocytometer.

Quantitative RT-PCR analysis

DNAse1-treated RNA was reverse transcribed using the cDNA Achieve kit (Applied Biosystems Monza, Italy). Real-time PCR was performed in triplicate using 100 ng cDNA on the DNA Engine Opticon 2 System (MJ Research Waltham, MA, USA) using the FastStart Taqman Probe Master Mix (Roche Molecular Biochemicals Monza, Italy), 100 nM Universal Probe Library probes (Roche Molecular Biochemicals) and 300 nM primers (presented in Table 3) designed using the ProbeFinder 2.20 software (Roche Molecular Biochemicals). The samples were amplified by means of one cycle at 95°C for 10 min, 39 cycles at 95°C for 15 s and 60°C for 1 min. The results were normalized on the GAPDH housekeeping gene, and expressed as transcript fold changes in relation to the expression observed in M0. Four macrophage preparations from different donors were analysed.

Table 3. Primers and probes used in real-time PCR experiments
GenePrimersProbe
Tfr15′-TGAGAAAACAATGCAAAATGTGA-3′ctgggcaa
 5′-CCCAGTTGCTGTCCTGATATAGA-3′ 
IRP25′-GAAATATGGTTCAGGAAACTCCA-3′ctgggctg
 5′-GCCAAAACAGCTTTCACACC-3′ 
HO15′-GGGTGATAGAAGAGGCCAAGA-3′catccagc
 5′-AGCTCCTGCAACTCCTCAAA-3′ 
Cp5′-TTACAACCCCCAGAGCAGAA-3′cttcagcc
 5′-TTCATAGGTGAATGTTTCTGTGG-3′ 
SLC40A15′-GGGGTCGCGTAGTGTCAT-3′ggatgctg
 5′-CAGGTAGTCGGCCAAGGAT-3′ 
GAPDH5′-TCCACTGGCGTCTTCACC-3′ctggggct
 5′-GGCAGAGATGATGACCCTTTT-3′ 

Immunoblot analysis

The cell lysates were prepared as previously described 56. For the preparation of membrane extracts, homogenates were centrifuged at 2000 g for 10 min and the postnuclear supernatants were ultracentrifuged at 150 000 g to pellet crude membrane fractions 29.

Equal amounts of protein extracts were electrophoresed, blotted and incubated with antibodies against TfR1 (1:1000; Zymed Laboratories San Francisco, CA, USA), HO1 (1:4000; Stressgen Victoria, Canada), IRP2 (1:100) 56, Fpn (1:250; Alpha Diagnostic San Antonio, TX, USA) and β-actin (1:10 000; Sigma). To detect Cp, aliquots of culture medium were concentrated on Centricon YM-30 (Millipore Vimodrone, Italy), processed as described above and incubated with anti-human Cp (1:5000; Sigma). The filters were stained with amido black to assess equal protein loading. After incubation with the appropriate secondary antibodies, the antigens were detected using ECL Plus (GE Healthcare Milano, Italy) and quantitated by densitometry with the values being calculated after normalization to the amount of β-actin or amido black-stained proteins.

Determination of ferritin subunit content

Ferritin concentrations were determined in cell lysates by means of an ELISA assay based on monoclonal antibodies against H and L ferritin subunits 28.

RNA-protein gel retardation assay

Proteins from cytosolic extracts were incubated with a 32P-labelled probe encompassing the IRE of the human H-Ft, and treated as previously described 17. After electrophoretic separation, the RNA–protein complexes were visualized by autoradiography and quantitated by means of direct nuclear counting using an InstantImager (Packard Instruments Meriden, CT, USA).

Analysis of 55Fe release

Macrophages were polarized for 72 h and 1 μM 55Fe-labelled transferrin prepared as previously described 56 was added during the last 20 h of culture. At the end of iron loading, the macrophages were extensively washed with PBS and, after the addition of 100 μM HM-DFO, incubated for a further 4 h. At various time points, triplicate aliquots of the medium were centrifuged and the supernatant was saved for radioactivity counting. At the end of the incubation, the cells were thoroughly washed with PBS, pelletted, and homogenized with the buffer used for the immunoblot analysis. Radioactivity in the cell extracts and in the supernatants was counted. The percentage of iron release was calculated as [(cpm in supernatant)/(cpm in supernatant+cpm in cells)]×100. Trypan blue exclusion testing was performed to check the integrity of the cell membrane.

Statistical analysis

The data are expressed as means±SD and were analysed using Student's T test with p values of <0.05 being considered statistically significant. These statistical comparisons were made using Stata Statistical Software (Stata Corporation, College Station, TX, USA).

Acknowledgements

This study was supported by grants from Telethon Italy (GGP06213) and the Cariverona Foundation to D.G.; the Fondazione Cariplo (NOBEL project) to M.L., the Programma Straordinario di Ricerca Oncologica 2006-RNBIO project to M.L. and L.Z., and the MIUR (FIRB and PRIN projects) to M.L. and G.C.

Conflict of interest: The authors declare no financial or commercial conflict of interest.

Ancillary