Different Tumor Microenvironments Contain Functionally Distinct Subsets of Macrophages Derived from Ly6C(high) Monocytes

Myeloid cells are frequently found to infiltrate tumors and have been linked to diverse tumor-promoting activities (1). In particular, tumor-associated macrophages (TAM) are an important component of the tumor stroma, both in murine models and human patients (2). TAMs can promote tumor growth by affecting angiogenesis, immune suppression, and invasion and metastasis (2, 3). However, it seems unlikely that these diverse functions are performed by a single cell type, and the existence of distinct TAM subsets, linked to different intratumoral microenvironments, has been predicted (4). Nevertheless, studies identifying spatially and functionally distinct TAM subpopulations are currently lacking.

Tissue-resident macrophages can be maintained through local proliferation or differentiation in situ from circulating monocytic precursors (5). Importantly, discrete subsets of blood monocytes have been described. Mouse monocytes can be classified as Ly6C^lowCX₃CR1^hi (CCR2⁻CD62L⁻) or Ly6C^hiCX₃CR1^low (CCR2⁺CD62L⁺) and are shown to have distinct functions and migration patterns (6). However, information on the nature and dynamics of the monocytic TAM precursors is lacking thus far.

Macrophages are plastic cells that can adopt different phenotypes depending on the immune context. Microenvironmental stimuli can drive a macrophage either toward a “classic” (M1) or an “alternative” (M2) activation state, two extremes in a spectrum (7). M1 macrophages are typically characterized by the expression of proinflammatory cytokines, inducible nitric oxide synthase 2 (Nos2), and MHC class II molecules. M2 macrophages have a decreased level of the aforementioned molecules and are identified by their signature expression of a variety of markers, including arginase-1 and mannose and scavenger receptors. It has been suggested that TAMs display an M2-like phenotype (8), although it is not clear whether these findings can be generalized and are applicable to TAMs in different tumor regions. In addition, the processes and signaling pathways that are driving the M2 phenotype of TAMs are not yet fully understood. A factor that is believed to be crucial in shaping the TAM phenotype is tumor hypoxia (9). Although hypoxia is known to have dramatic effects on the activation and function of macrophages, it remains to be determined how this relates to the M2-like orientation of TAMs.

In this study, we show the existence of molecularly and functionally distinct TAM subsets, located in different intratumoral regions, and uncover Ly6C^hi monocytes as their precursors. These results might prove important for therapeutic interventions targeted at specific TAM subsets or their precursors.

Female BALB/c and C57BL/6 mice were from Harlan. BALB/c CX₃CR1^GFP/GFP mice were provided by Dr. Grégoire Lauvau (Université de Nice-Sophia Antipolis, Nice, France). The BALB/c mammary adenocarcinoma TS/A (10) was provided by Dr. Vincenzo Bronte (Istituto Oncologico Veneto, Padova, Italy); BALB/c 4T1 mammary carcinoma (11) was provided by Dr. Massimiliano Mazzone (VIB-KULeuven, Leuven, Belgium); and 3LL-R clone of the C57BL/6 Lewis Lung carcinoma was derived as described previously (12). Cells were injected subcutaneously (s.c.) in the flank (3 × 10⁶) or in the mammary fat pad (10⁶).

Tumors were treated with 10 U/mL collagenase I, 400 U/mL collagenase IV, and 30 U/mL DNaseI (Worthington). Density gradients (Axis-Shield) were used to remove debris and dead cells. To purify TAMs, CD11b⁺ cells were MACS-enriched (anti–CD11b microbeads, Miltenyi Biotec) and sorted using a BD FACSAria II (BD Biosciences). To purify dendritic cells, spleens were flushed with 200 U/mL collagenase III (Worthington). CD11c⁺ cells were MACS-enriched (anti-CD11c microbeads, Miltenyi Biotec), and CD11c⁺MHC II^hiB220⁻Ly6C⁻ dendritic cells were sorted.

Antibodies used are listed in Supplementary Table S1.

For tumor necrosis factor α (TNFα) stainings, TAMs were cultured 5 hours with Brefeldin A (BD Biosciences). For inducible nitric oxide synthase (iNOS), TAMs were cultured with 10 U/mL IFNγ and/or 10 ng/mL lipopolysaccharide (LPS) for 12 hours.

Arginase activity was measured as described earlier (13).

Latex labeling of monocytes was described earlier (14, 15). For Ly6C^low monocyte labeling, mice were injected intravenously (i.v.) with 250 μL 0.5 μm yellow-green microspheres (Polysciences; 1:25). Twenty-four hours later, TS/A was injected s.c. For Ly6C^hi monocyte labeling, mice were injected i.v. with 250 μL clodronate liposomes (16) and 18 hours later with latex microspheres (i.v.) and TS/A (s.c.).

Mice were injected intraperitoneally with 1 mg bromodeoxyuridine (BrdUrd) (Sigma), and 0.8 mg/mL BrdUrd was administered to drinking water. To stain for BrdUrd, Ki67 (BD Biosciences), or propidium iodide (Invitrogen), cells were fixed/permeabilized using the BD Biosciences BrdUrd labeling kit.

These tests were performed as described earlier (17). Gene-specific primers are listed in Supplementary Table S2.

For hypoxia stainings, mice were injected with 80 mg/kg body weight pimonidazole [hypoxyprobe-1 (HP-1), HPI, Inc.]. Two hours later, tumors were snap frozen, sections were acetone fixed, and stained. Pictures were acquired with a Plan-Neofluar 10×/0.30 or Plan-Neofluar 20×/0.50 (Carl Zeiss) objective on a Zeiss Axioplan 2 microscope equipped with an Orca-R2 camera (Hamamatsu) and Smartcapture 3 software (Digital Scientific UK). For HP-1 fluorescence-activated cell sorting (FACS) measurements, cells were fixed/permeabilized using the BD Biosciences Fix/Perm kit and rat anti–HP-1/FITC (HPI) was added (30–37°C).

Chorioallantoic membrane (CAM) assays were performed as described earlier (18). Gelatin sponges (1–2 mm³; Hospithera) with 5 × 10⁴ sorted TAM subsets were placed on the CAM. PBS/0.1% bovine serum albumin (BSA; 50 μg/CAM) and recombinant human vascular endothelial growth factor (VEGF)-A₁₆₅ (5 μg/CAM) served as controls. At day 13, membranes were fixed and analyzed using a Zeiss Lumar V.12 stereomicroscope with NeoLumar S 1.5× objective (15× magnification).

For allo–mixed leukocyte reaction (MLR) assays, 2 × 10⁵ MACS-purified CD4⁺/CD8⁺ C57BL/6 T cells were added to 5 × 10⁴ sorted TAMs or conventional dendritic cells and 3 days later were [³H]thymidine pulsed.

For T-cell suppression assays, 1 × 10⁵ to 1.25 × 10⁴ (1:2–1:16) sorted TAMs/conventional dendritic cells were added to 2 × 10⁵ naive BALB/c splenocytes with 1 μg/mL anti-CD3 and were [³H]thymidine pulsed 24 hours later. l-NMMA (0.5 mmol/L, Sigma) and/or NorNoha (0.5 mmol/L, Calbiochem) were added in a 1:4 ratio. Relative percent suppression of proliferation was calculated as described earlier (19).

Significance was determined by Student's t test.

To study the tumor-infiltrating myeloid compartment, we, at first instance, used the BALB/c mammary adenocarcinoma model TS/A. Subcutaneous tumors contained a large CD11b⁺ fraction, indicating a high infiltration of myeloid cells (Fig. 1A). Interestingly, this CD11b⁺ population was heterogeneous and encompassed at least seven subsets, which could be readily distinguished based on their differential expression of MHC class II and Ly6C (Fig. 1A). Ly6C^hiMHC II⁻ cells (gate 1: Fig. 1A) were F4/80⁺CX₃CR1^lowCCR2^hiCD62L⁺, did not express the granulocyte markers Ly6G or CCR3, and had a small size and granularity (FSC^lowSSC^low), indicating that they were Ly6C^hi monocytes (Fig. 1A and D; Supplementary Fig. S1). The CD11b⁺MHC II⁺ cells in gates 2 to 4 were reminiscent of macrophages, having an enlarged macrophage-like scatter and expressing high levels of F4/80 (Fig. 1A and D). Remarkably, distinct subsets of TAMs were clearly distinguishable: Ly6C^intMHC II^hi (Ly6C^int TAMs, gate 2), Ly6C^lowMHC II^hi (MHC II^hi TAMs, gate 3), and Ly6C^lowMHC II^low (MHC II^low TAMs, gate 4). The majority of Ly6C^lowMHC II⁻ cells were CCR3⁺CX₃CR1⁻ eosinophils (Fig. 1A, gate 5; Supplementary Fig. S1, gate E). However, Ly6C^lowMHC II⁻ cells also consisted of CCR3⁻CX₃CR1^low (Supplementary Fig. S1, gate 2) and CCR3⁻CX₃CR1^hi (Supplementary Fig. S1, gate 3) cells, the latter possibly resembling Ly6C^lowCX₃CR1^hi monocytes. However, the majority of these CX₃CR1^hi cells did not have a monocyte scatter, suggesting they were TAMs (Supplementary Fig. S1). This suggests that Ly6C^low monocytes were not present in significant amounts in these tumors. Finally, TS/A tumors were also infiltrated with CCR3⁺Ly6C^int eosinophils (Fig. 1A, gate 6) and Ly6G^hi neutrophils (Fig. 1A, gate 7).

Interestingly, the relative percentages of these distinct myeloid subpopulations dramatically changed as tumors progressed (Fig. 1B). Within the TAM compartment, the percentage of Ly6C^int TAMs decreased, whereas the Ly6C^lowMHC II^low TAM subset became gradually more prominent, reaching up to 60% of the myeloid tumor infiltrate in large tumors (>10 mm). Because the amount of tumor-infiltrating CD11b⁺ cells increased as tumors progressed (Fig. 1C), MHC II^low TAMs also strongly accumulated in absolute numbers, to a much greater extent than MHC II^hi TAMs.

Macrophages typically derive from circulating blood-borne precursors such as monocytes. The presence of Ly6C^hi, but not Ly6C^low, monocytes in TS/A tumors suggested that the former could be more efficiently recruited to tumors and function as the TAM precursor. To investigate this, we selectively labeled Ly6C^hi or Ly6C^low monocyte subsets in vivo with fluorescent latex beads, using a previously described procedure (14, 15). This method has been validated to stably label the respective monocyte subsets for 5 to 6 days in naive mice. Hence, TS/A was injected after Ly6C^low or Ly6C^hi monocyte labeling, and tumors were collected 6 days post injection. No appreciable numbers of tumor-infiltrating latex⁺ monocytes were observed when applying the Ly6C^low-labeling strategy (Fig. 2A). In contrast, Ly6C^hi labeling resulted in the detection of a significant fraction of CD11b⁺latex⁺ monocytes, illustrating that Ly6C^hi monocytes comprise the main tumor-infiltrating monocyte subset. With this approach, latex⁺ cells could be detected up to 19 days after tumor injection (Fig. 2B), allowing a follow-up of the monocyte progeny in the course of tumor growth. At day 6, latex⁺Ly6C^hi monocytes had differentiated into latex⁺Ly6C^int TAMs, and to some extent also into latex⁺MHC II^hi and latex⁺MHC II^low TAMs (Fig. 2B). From day 12 onward, the majority of latex⁺Ly6C^hi monocytes had converted into latex⁺MHC II^hi and latex⁺MHC II^low TAMs. Together, these data show that all TAM subsets can be derived from Ly6C^hi monocytes.

Remarkably, the total number of peripheral blood monocytes had significantly increased at later stages of tumor growth (≥21 days post injection; Supplementary Fig. S2A). Furthermore, around 4 weeks of tumor growth, there was a significant increase in the percentage of the Ly6C^hi monocyte subset (Supplementary Fig. S2B–C).

To determine the turnover rate and differentiation kinetics of the monocyte/TAM subsets, BrdUrd was administered continuously to tumor-bearing animals and its incorporation was measured at consecutive time points. Tumor-infiltrating Ly6C^hi monocytes quickly became BrdUrd⁺, reaching plateau values after 48 hours of BrdUrd administration (Fig. 2C). This indicates a rapid monocyte turnover rate and/or proliferation of monocytes inside tumors. Remarkably, although intratumoral Ly6C^hi monocytes were Ki67⁺ (Fig. 2D1), none were found to be in the S-G₂-M phase (Fig. 2D2), suggesting that these cells were in the G₁ phase and not proliferating (20). TAMs were Ki67⁻, and no appreciable numbers were found in S-G₂-M phase, indicating no significant levels of proliferation. Hence, TAMs were unable to directly incorporate BrdUrd so that BrdUrd⁺ TAMs must differentiate from BrdUrd⁺ monocytes, resulting in a lag phase of BrdUrd positivity. Indeed, only a minor fraction of MHC II^hi and MHC II^low TAMs were BrdUrd⁺ upon 24 hours of BrdUrd administration (Fig. 2C). However, compared with these subsets, Ly6C^int TAMs incorporated BrdUrd at a faster rate, with a higher percentage being BrdUrd⁺ already at 24 hours. These results suggest that monocytes first give rise to Ly6C^int TAMs, which then differentiate into MHC II^hi and MHC II^low TAMs. MHC II^hi and MHC II^low TAMs incorporated BrdUrd slowly and with similar kinetics, arguing for a comparable and low turnover rate.

Although efforts have previously been made to characterize TAMs at the molecular level (21, 22), a thorough study of TAM heterogeneity is lacking up to now. Hence, we further characterized the distinct TAM subsets at the gene and protein levels. Gene expression of sorted MHC II^hi and MHC II^low TAMs (Supplementary Fig. S3A) was analyzed through quantitative reverse transcriptase-PCR (RT-PCR; Table 1). Ly6C^int TAMs, constituting only a minor fraction in larger tumors, were not included in this analysis. Interestingly, when comparing MHC II^hi with MHC II^low TAMs (Table 1, hi/low), M2-associated genes such as Arg1 (arginase-1), Cd163, Stab1 (stabilin-1), and Mrc1 (MMR) were higher expressed in the MHC II^low subset. In contrast, more M1-type, proinflammatory genes, such as Nos2 (iNOS), Ptgs2 (Cox2), Il1b, Il6, and Il12b, were upregulated in MHC II^hi TAMs. This differential activation state was also reflected at the protein level. Membrane expression of the M2 markers macrophage mannose receptor (MMR), macrophage scavenger receptor 1 (SR-A), and interleukin-4Rα (IL-4Rα) were clearly higher on MHC II^low TAMs, whereas the M1-associated marker CD11c was only expressed on MHC II^hi TAMs (Fig. 1D). Moreover, although arginase activity was observed in both TAM subsets, it was significantly higher for MHC II^low TAMs (Fig. 3A). In the same vein, TNFα, which has previously been reported to associate with a M2 phenotype in tumors (23, 24), was produced by both TAM subsets; however, a significantly higher percentage of MHC II^low TAMs were found to be TNFα⁺ (Fig. 3B). Although iNOS protein was not detected in freshly isolated TAMs, it could be induced by IFN-γ and/or LPS stimulation (Fig. 3C). Interestingly, IFN-γ or LPS induced iNOS more efficiently in MHC II^hi TAMs, with a higher fraction of these cells becoming iNOS⁺. Together, these data indicate that the identified TAM subsets have a differential activation state, with MHC II^low TAMs being more M2 oriented.

TAM subsets also showed a markedly distinct chemokine expression pattern (Table 1). Notably, mRNAs for chemokines typically involved in lymphocyte attraction, such as Ccl5, Cx₃cl1, Cxcl11, Cxcl10, Cxcl9, and the CCR4 ligands Ccl17 and Ccl22 were upregulated in MHC II^hi TAMs. In contrast, mRNAs for monocyte/macrophage chemoattractants, such as Ccl6; the CCR2 ligands Ccl7, Ccl2, and Ccl12; and the CCR5/CCR1 ligands Ccl4, Ccl3, and Ccl9 were significantly higher in MHC II^low TAMs. Furthermore, at the protein level, a differential expression of the chemokine receptors CX₃CR1 and CCR2 was observed, with MHC II^hi TAMs being CX₃CR1^hiCCR2⁻, whereas MHC II^low TAMs were CX₃CR1^lowCCR2⁺ (Fig. 1D).

Both TAM subsets expressed many potentially proangiogenic genes, including Vegfa, Mmp9, Pgf, Spp1, and cathD (Table 1). However, several angiostatic factors such as angpt2, Cxcl9, Cxcl10, and Cxcl11 were upregulated in the MHC II^hi fraction. One of the most differentially expressed genes (higher in MHC II^low TAMs) was Lyve1, a marker previously associated with angiogenic/hypoxic macrophages (25).

We conclude that MHC II^hi and MHC II^low TAMs have a distinguishing profile of molecules involved in inflammation (M1/M2), chemotaxis, and angiogenesis.

To extrapolate these findings to orthotopically grown tumors, TS/A was injected in the mammary fat pad. Orthotopic tumors contained identical myeloid subsets, which accumulated with comparable kinetics (Supplementary Fig. S4A) and retained their differential expression of surface markers (Supplementary Fig. S4B).

We then investigated whether differentially activated TAM subsets were present in other tumor models. Interestingly, orthotopic 4T1 mammary and subcutaneous 3LL lung carcinoma tumors contained distinct granulocyte and monocyte/macrophage subsets (Supplementary Figs. S5A and S6A), including Ly6C^hi monocytes (gate 1), Ly6C^int TAMs (gate 2), MHC II^hi TAMs (gate 3), and MHC II^low TAMs (gate 4). 3LL tumors also contained a population of Ly6C^intMHC II^low TAMs (Supplementary Fig. S6A, gate 5), possibly representing an alternative differentiation path from Ly6C^hi monocytes to Ly6C^low TAMs. As with TS/A, the progression of 3LL tumors was linked with an accumulation of MHC II^low TAMs (Supplementary Fig. S6B). Surprisingly, progressing 4T1 tumors gradually increased their MHC II^hi TAM content (Supplementary Fig. S5B), indicating that the relative increase of TAM subsets over time is tumor dependent.

Interestingly, the 4T1 and 3LL tumor-derived MHC II^hi and MHC II^low TAM subsets remained differentially M1- versus M2-like activated, shown by enhanced M2 marker gene expression in MHC II^low TAMs (CD163, Stab1, Arg1, Mrc1, IL4Ra, Il10) and M1 marker upregulation in MHC II^hi TAMs (Il1b, Il12b, Cox2; Supplementary Tables S3 and S4). Of note, although Nos2 mRNA was higher in the 3LL MHC II^low TAMs, iNOS protein levels were higher in MHC II^hi TAMs, supporting their more M1-like activation (Supplementary Fig. S7). Finally, in both models, MHC II^low TAMs expressed higher levels of MMR and IL4Rα protein, whereas CD11c was restricted to MHC II^hi TAMs (Supplementary Figs. S5C and S6C). Hence, TAM subsets in three unrelated tumor models had a high level of similarity, with a consistent differential activation of MHC II^hi and MHC II^low TAMs.

Tumors often harbor regions of hypoxia, a factor that is known to influence macrophage function (9). To visualize hypoxia in TS/A tumors, tumor-bearing mice were injected with pimonidazole (HP-1) and tumor sections were stained for hypoxic adducts and blood vessels. Figure 4A shows that tumors indeed contained a large number of hypoxic cells, primarily in regions with a less developed vasculature. Interestingly, staining sections for HP-1, CD11b, and MHC II showed that many CD11b⁺MHC II⁻ cells (which in large tumors are mainly MHC II^low TAMs) were HP-1⁺ (Fig. 4B). Interestingly however, the majority of CD11b⁺MHC II⁺ cells were HP-1⁻. This indicates that whereas a significant fraction of MHC II^low TAMs resided in hypoxic areas, MHC II^hi TAMs were mainly normoxic. Importantly, HP-1 adducts could also be detected through intracellular flow cytometry on freshly isolated TAMs. Again, the highest signal was seen in MHC II^low TAMs, confirming they were the most hypoxic TAM subset (Fig. 4C).

A consequence of MHC II^low TAMs being in hypoxic regions should be a reduced access to blood-transported molecules. To test this, fluorescent latex particles were injected i.v. in tumor-bearing mice. One to 2 hours later, a fraction of tumor-associated CD11b⁺ cells were found to be latex⁺ (Supplementary Fig. S8A). However, latex uptake was not equal in all TAM subsets. Indeed, in relative terms, MHC II^low TAMs phagocytosed less latex than monocytes and other TAM subsets. This was not due to an inherently reduced phagocytic capacity of MHC II^low TAMs because the latter showed the highest phagocytic latex uptake in vitro (Supplementary Fig. S8B). These data suggest that the reduced in vivo latex uptake of MHC II^low TAMs was due to a restricted access to latex particles, which further substantiates the enrichment of MHC II^low TAMs in hypoxic regions.

Hypoxia initiates an angiogenic program (26). In addition, our gene profiling revealed the expression of angiogenesis-regulating molecules in TAMs. To directly test the proangiogenic activity of both TAM subsets in vivo, we used the CAM assay. Sorted MHC II^hi or MHC II^low TAMs were implanted on developing CAMs, whereas BSA or rhVEGF served as negative and positive controls, respectively. rhVEGF induced the outgrowth of allantoic vessels specifically directed toward the implants (Fig. 5A). Interestingly, compared with BSA controls, the presence of MHC II^hi or MHC II^low TAMs significantly increased the number of implant-directed vessels, demonstrating a proangiogenic activity for both TAM subsets. However, the vessel count for implants containing MHC II^low TAMs was on average 2-fold higher than with MHC II^hi TAMs. These data show that MHC II^low TAMs had a superior proangiogenic activity in vivo.

We wondered whether the TAM subsets were able to process internalized antigens and activate T cells. Both TAM subsets took up and processed DQ-Ovalbumin at 37°C. However, examining DQ-Ovalbumin processing at consecutive time points indicated that processing occurred more slowly in the MHC II^low fraction (Supplementary Fig. S9). To investigate whether TAMs could directly activate naive T cells, a MLR assay was used. Sorted MHC II^hi or MHC II^low TAMs were cultured with purified allogeneic C57BL/6 CD4⁺ or CD8⁺ T cells. Sorted splenic CD11c^hiMHC II^hi conventional dendritic cells (Supplementary Fig. S3D) were used as a reference T-cell–stimulating population (27). Compared with conventional dendritic cells, MHC II^hi or MHC II^low TAMs induced poor proliferation of allogeneic CD4⁺ or CD8⁺ T cells (Fig. 5B), suggesting a limited antigen-presenting capacity or, alternatively, a T-cell suppressive capacity that overrules antigen presentation.

To investigate the latter possibility, T cells were polyclonally activated in the presence of TAMs or conventional dendritic cells. Interestingly, as opposed to conventional dendritic cells, both MHC II^hi and MHC II^low TAMs equally suppressed anti–CD3-induced T-cell proliferation in a dose-dependent manner (Fig. 5C). In an attempt to identify the suppressive molecules responsible for TAM-mediated suppression, inhibitors of iNOS (l-NMMA) and arginase (NorNoha) were added to the cocultures (Fig. 5D). Blocking iNOS significantly reduced T-cell suppression by MHC II^hi TAMs, demonstrating a role for nitric oxide in its suppressive mechanism. In contrast, iNOS inhibition only had a minor effect on the suppressive potential of MHC II^low TAMs, showing that both subsets use different T-cell suppressive mechanisms.

In this article, we show that the tumor-infiltrating myeloid compartment can be highly heterogeneous, with the coexistence of distinct subsets of granulocytes and mononuclear phagocytes. Furthermore, we identified the nature and dynamics of the monocyte precursor that was seeding tumors and giving rise to distinct TAM subsets. Interestingly, these subsets differed at the molecular and functional levels and were present in different intratumoral microenvironments (for an overview, see Supplementary Fig. S10).

Within the tumor-infiltrating monocyte pool, Ly6C^hiCX₃CR1^int monocytes were the most prominent subset, whereas Ly6C^lowCX₃CR1^hi monocytes constituted only a small minority. In addition, bead labeling and BrdUrd incorporation experiments showed that Ly6C^hi monocytes were the precursors of all the distinct TAM subsets in TS/A tumors. Ly6C^hi monocytes rely on the chemokine receptor CCR2 for their migration from the bone marrow into the circulation (28). Recent studies showing that tumors grown in CCR2^−/− mice have significantly reduced numbers of TAMs (29, 30) are therefore in line with our observation that Ly6C^hi monocytes comprise the main tumor-infiltrating monocyte subset. Furthermore, TAMs (and in particular MHC II^low TAMs) had a high gene expression of the CCR2 ligands CCL2, CCL7, and CCL8, suggesting an active role in the recruitment of Ly6C^hi monocytes. Most studies focusing on infection or immunization settings show that, at the site of insult, Ly6C^hi monocytes give rise to inflammatory dendritic cells (31–34). These inflammatory dendritic cells remain Ly6C^hi, express intermediate levels of CD11c, and can be efficient antigen presenters. In addition, a recent study has shown that shortly after Listeria monocytogenes infection, Ly6C^hi and Ly6C^low monocytes enter a dendritic cell or a macrophage differentiation program, respectively (35). However, our results show that at the tumor site, Ly6C^hi monocytes exclusively gave rise to distinct subsets of inflammatory macrophages, further highlighting monocyte plasticity and the impact of the tumor microenvironment thereon.

Strikingly, MHC II^hi and MHC II^low TS/A TAMs tended to be more M1- or M2-like, respectively. At the protein level, this included the differential expression of the M1 markers MHC II, CD11c, and iNOS, and the typical M2 markers MMR, SR-A, IL-4Rα, and arginase-1. Further proof for a differential activation state was delivered by gene expression analysis: an upregulation of proinflammatory genes in MHC II^hi TAMs, whereas M2-associated genes preferentially adhered to the MHC II^low subset. Interestingly, Hagemann and colleagues described that macrophages cocultured with ovarian cancer cells obtain an M2-like phenotype reminiscent of the MHC II^low TAMs in our present study, including upregulation of MMR, SR-A, and high expression levels of TNFα (23). In follow-up studies, it was shown that inhibiting IκB kinase (IKK) β activity in these macrophages results in a switch from M2 to M1, as evidenced by enhanced expression of MHC II, iNOS, and IL-12, and a reduction in arginase, TNFα, and IL-4Rα (24, 36). Hence, MHC II^hi TAMs more closely resemble the phenotype of the IKKβ-deficient macrophages, raising the possibility that the opposing activation states of MHC II^hi and MHC II^low TAMs might be driven by a differential NF-κB activity in these subsets.

Importantly, our findings in TS/A could be translated to 4T1 and 3LL tumors. The remarkable similarities between TAM subsets from these unrelated tumors suggest that similar environmental cues might shape their respective phenotypes.

Interestingly, in TS/A tumors, MHC II^low TAMs were found to preferentially reside in hypoxic regions, as shown by pimonidazole stainings and their reduced access to blood-transported molecules. Hypoxia is known to influence gene and protein expression of macrophages: inducing expression of arginase, TNF, and proangiogenic factors while downregulating MHC II (37–39). Hypoxia-inducible factors (HIF) are the main transcription factors involved in regulating hypoxia-driven gene expression (26). Interestingly, a recent report showed that IKKβ is required for HIF-1α accumulation under hypoxic conditions, thereby uncovering a link between the hypoxic response and NF-κB (40). Hence, it is tempting to speculate that the involvement of IKKβ in shaping the M2 activation state of macrophages and its requirement for HIF-1α activity might be involved in the M2 skewing of hypoxic TAMs. Irrespective of the molecular mechanism, these are the first data linking the M2-like orientation of TAMs with a hypoxic environment.

Another striking difference between TS/A MHC II^hi and MHC II^low TAMs was at the level of chemokine and chemokine receptor expression. MHC II^low TAMs, possibly under the influence of hypoxia, had the highest gene expression of monocyte-recruiting chemokines, whereas chemokines that can recruit Th1, Th2, or natural killer cells, such as Cx₃cl1, Ccl5, Cxcl9, Cxcl10, Cxcl11, Ccl17, and Ccl22 (41), were clearly upregulated in MHC II^hi TAMs. Hence, TAM subsets might contribute differently to shaping the inflammatory tumor infiltrate. In addition, the differential membrane expression of CX₃CR1 and CCR2 on the TAM subsets possibly reflects the use of different chemokine axes for their migration.

A recent study compared the gene expression profile of tumor-associated CD11b⁺Tie2⁺ cells (TEM) with that of CD11b⁺Tie2⁻ cells (42). Remarkably, many of the genes that are differentially expressed between TEMs and the residual CD11b⁺ fraction were also key differential genes between MHC II^low and MHC II^hi TAMs. For example, similar to MHC II^low TAMs, TEMs have a higher mRNA expression level of Lyve1, CD163, Stab1, Mrc1, Arg1, and Il4Ra, but lower Il1b, Nos2, Ptgs2, Ccl5, Cxcl10, and Cxcl11. However, TEMs are only a minor fraction of CD11b⁺ cells in tumors and are suggested to have a lineage relationship with Ly6C^lowCX₃CR1^hi monocytes (42). In contrast, MHC II^low TAMs were the most abundant tumor-associated myeloid population and were originating from Ly6C^hi monocytes. In addition, antibody staining did not reveal any Tie-2 expression on MHC II^low TAMs and Tie2 mRNA levels were slightly lower in MHC II^low compared with MHC II^hi TAMs. Although we do not exclude that TEMs might be present in the MHC II^low fraction, our results suggest that MHC II^low TAMs and TEMs are distinct populations, but have intriguing similarities in their phenotypes. Because TEMs are potent proangiogenic cells (43), it is tempting to speculate that the similarities between TEMs and MHC II^low TAMs might reflect a comparable function. Indeed, one of the key responses to hypoxia is the induction of an angiogenic program (26).

An in vivo CAM assay showed that both TS/A TAM subsets stimulated angiogenesis. Interestingly, however, in line with their localization in hypoxic regions, MHC II^low TAMs showed a significantly higher proangiogenic activity, indicating that the balance of proangiogenic versus antiangiogenic mediators was highest for this subset. At present, the exact molecular basis for the increased angiogenic potential of MHC II^low TAMs is not clear, as several proangiogenic genes were expressed at a high level in both TAM populations. However, MHC II^hi TAMs had the highest expression of the antiangiogenic CXC chemokines (Cxcl9-11; ref. 44), potentially limiting the effects of proangiogenic factors.

The induction of myeloid-derived suppressor cells (MDSC) is shown to be an important immune-evading strategy used by tumors (45, 46). We and others have previously shown that splenic CD11b⁺Gr-1⁺ MDSCs consist of two major subsets: monocytic Ly6G⁻ MO-MDSCs and granulocytic Ly6G⁺ PMN-MDSCs (19, 47). Importantly, the phenotype of TS/A tumor-infiltrating Ly6C^hi(MHC II⁻) monocytes closely resembled that of MO-MDSCs, whereas the tumor-infiltrating Ly6G⁺ neutrophils were reminiscent of PMN-MDSCs (19). However, whether these tumor-infiltrating cells have immune-suppressive potential remains to be determined. In any case, within the TS/A tumor microenvironment, cells with a MO-MDSC–like phenotype differentiate into CD11b⁺Gr-1⁻/Ly6C⁻ macrophages, suggesting a potential lineage relationship between MDSCs and TAMs. Importantly, T-cell–suppressive activity was a prominent feature of TAMs. Indeed, whereas MHC II^hi and MHC II^low TAMs were able to process antigens (albeit with different kinetics), they inefficiently activated naive T cells. In contrast, both subsets strongly suppressed polyclonal T-cell proliferation. Interestingly, in line with their M1-like activation, MHC II^hi TAMs relied to a higher extent on iNOS for suppression.

It has been predicted that TAMs in different tumor regions might have specialized functions (4). Our results provide the first evidence for this and describe markers for their discrimination in three independent tumor models. This offers the prospect of specifically targeting the M1-/M2-like or hypoxic/perivascular TAM subsets and investigating their impact on tumor biology. Eventually, this might lead to combinatorial strategies for optimally “re-educating” the TAM compartment and reverting its tumor-promoting activities.

No potential conflicts of interest were disclosed.

Grant Support: Doctoral grants from FWO-Vlaanderen (K. Movahedi and J. Van den Bossche), a doctoral grant from IWT-Vlaanderen (D. Laoui), and grants from Stichting tegen Kanker.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.