Paraffin slides were provided by the Lung Biobank Heidelberg, a member of the Biomaterial bank Heidelberg (BMBH) and the Biobank platform of the German Center for Lung Research (DZL). Tissue microarrays (TMAs) were provided by the tissue bank of the National Center for Tumor Diseases (NCT, Heidelberg, Germany), in accordance with the regulations of the tissue bank and the approval of the ethics committee of Heidelberg University. TMA classification was performed according to the sixth edition of the Tumor Node Metastasis staging system for NSCLC. Paraffin slides from 19 patients with NSCLC (Table 1) were analyzed under the microscope and divided into iron positive and iron negative according to the visible detection/absence of intracellular iron. Each paraffin slide included tumor center, invasive front and tumor periphery.

Human adenocarcinoma tumors were obtained from NSCLC patients (n = 4) who underwent resection for primary lung cancer at the Thoraxklinik of the University Hospital, Heidelberg, Germany. Fresh tumors (human and mouse) were mechanically dissociated and digested with DNAse (SIGMA) and Hyaluronidase (SIGMA). Tumor suspensions were strained using a 70 µm cell strainer (Becton Dickinson) and washed with PBS. For the isolation of human leukocytes, cell suspensions were layered over a density gradient solution (Biocoll Separating Solution, 1.077 g/ml, Biochrom AG, Germany) in a 1:1 volume ratio (450 g, 30 min, RT). Leukocytes were collected, washed twice (PBS) and resuspended (PBS, 4°C). Cells were further processed for magnetic isolation. For the preparation of mouse tumors, a gradient purification using a Lympholyte solution (Cederlane) was performed to remove RBCs and dead cells. Briefly, 7 ml of tumor suspension were added on top of 7 ml of Lympholyte solution and centrifuged (1,500 × g, 25 min, 20°C). The layer of live cells was removed and washed again in cold PBS. The cell pellet was resuspended in PBS (4°C) and kept on ice for the respective procedures.

Iron-loaded macrophages were isolated from cell suspensions as described in Ref. (30). Briefly, cell suspensions were resuspended in 5 ml of PBS and passed through an LS column attached to a magnetic board (Miltenyi Biotech). Columns were washed three times with 5 ml of PBS. Cells that were adherent to the column (magnetic fraction) were flushed with 5 ml PBS. Cells were either resuspended in PBS and centrifuged for cytospin preparations or lysed for RNA extraction.

Female and male [C57BL/6N and Slc40a1^C326S/C326S (31)] mice were used at 8- to 10-weeks of age. Experiments were approved by “Regierungspräsidium Karlsruhe” under the project number G267/12. Lewis lung carcinoma (LLC) cells were injected (1 × 10⁶ in 100 µl PBS) subcutaneously into the flanks of mice. When indicated, LLC cells were co-injected with cross-linked iron oxide (CLIO)-FITC nanoparticles (8 mg of iron/kg of mouse), or only in PBS. Tumor size was assessed by caliper measurements at the indicated time points and volumes were calculated using the following formula: V = 1/2(length [mm] × width [mm]²) as previously described (32). Mice were sacrificed at the indicated time points. Mice that developed ulcers or necrotic tumors were sacrificed and not considered for the experiments. Blood was removed directly from the heart by cardiac puncture. Subcutaneous tumors were resected and dissected carefully to avoid tissue damage and bleedings induced during animal preparation. Tumors were transferred to PBS on ice and tumor weight (g) was measured on a scale. Tumors were processed for cytospin, FACS analysis and FACS sorting, snap frozen until further analysis or fixed in formalin for immunohistochemistry and histological analysis.

After careful resection, LLC tumors were washed in PBS and placed in a petry dish under a stereo microscope (Olympus SZ51). Tumors were evaluated for the presence of hemorrhagic areas. Hemorrhagic (H) areas were distinguished from non-hemorrhagic (NH) areas by a strong red coloration. With the help of tweezers and scalpel, H and NH areas from the same tumor were dissected, separated, and stored accordingly for further processing. As validation and quality control for the dissection procedure, heme was quantified.

Single cell suspensions (200 µl) were centrifuged (500 r.p.m, 5 min) in a Cytospin Cytocentrifuge (Thermo Scientific). Tissues were fixed for 24 h in 10% neutral buffered formalin (Sigma-Aldrich), dehydrated, and embedded in paraffin. Tissue sections (3–5 µm) were stained for iron using Accustain Iron Stain No. HT20 (Sigma-Aldrich) following manufacturer’s instructions. When indicated, Perls’ blue staining was further enhanced using the DAB peroxidase substrate kit SK-4100 (Vector Labs). Quantification of iron staining was performed using the Image Pro-Premier 3D software. The software calculated the area of pixels corresponding to blue staining (iron staining). For immunostaining, cytospin samples were fixed and permeabilized in ice-cold acetone for 5 min, washed in PBS and treated with H₂O₂ to block endogenous peroxidase. For immunohistochemistry, sections were treated for 10 min with 3% H₂O₂ (Sigma-Aldrich) and subjected to microwave-mediated antigen retrieval using the Citraplus reagent (Biogenex). Immunostaining was performed according to the instructions of the Vectastain ABC mouse, rat, and rabbit kits (Vector Labs). Anti-mouse ferroportin staining was performed using MTP11-A rabbit polyclonal antibody (Anti-Mouse Metal Transporter Protein1/Ferroportin (MTP1/IREG1/Fpn) from Alpha Diagnostics); rabbit IgG was used as isotype control for the ferroportin staining. Anti-human CD68 staining was performed using the monoclonal mouse anti-human PG-M1 clone (DAKO). Tissue slides were developed using the Vector AEC substrate (Vector Labs), rinsed with distilled water, counterstained with hematoxylin, washed in PBS, and mounted using the VectaMount AQ mounting medium (Vector Labs). Images were acquired with a Ni-E Nikon microscope.

Bone marrow cells were flushed from tibia and femur using ice-cold HBSS and filtered through a 70 µm cell strainer. Cells were seeded at a density of 700,000 cells/ml in equal volumes of conditioned media (CM) from LLC cells (or DMEM used as control) and complete RPMI1640-Glutamax medium (Life Technologies) [supplemented with 10% of heat-inactivated FBS (Thermo Scientific), 1% penicillin/streptomycin (Sigma-Aldrich) and 10 ng/ml M-CSF (Sigma-Aldrich)]. For 12 well plates, cells were incubated with 0.5 ml of CM plus 0.5 ml of complete RPMI. For 6 well plates, cells were incubated with 1 ml of CM plus 1 ml of complete RPMI. After 4 days, the medium was removed and centrifuged (1,500 r.p.m for 5 min) to remove cells in suspension. Macrophages were incubated in the same media with additional CM for 24 h (0.5 ml for 12 well plates and 1 ml for 6 well plates). For each independent experiment, BMDM were prepared from three different mice. At least three independent experiments were performed for each figure panel.

Red blood cell aging was performed as described (33). Mouse blood was collected on EDTA tubes, washed with PBS (3×), and resuspended in Hepes buffer (10 mM Hepes, 140 mM NaCl, BSA 0.1%, pH 7.4). Cells kept at 4°C overnight were considered non-aged RBCs. For in vitro RBCs aging (aRBC), cells (1 × 10⁸ cells/ml in Hepes buffer) were incubated overnight at 30°C with 2.5 mM calcium and 0.5 mM of the ionophore A23187 (Calbiochem). Before incubation with TAMs, both RBCs fractions were washed twice (PBS, 1,500 r.p.m, 5 min) and resuspended in RPMI.

Data are shown as mean ± SEM, and the number of mice (n) is indicated. Statistical analyses were performed using Prism v.6 (GraphPad). Comparisons between two groups were performed with two-sided Welch t-tests, and among three or more than three groups with one-way ANOVA, followed by Bonferroni post-test. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001 are indicated.

To explore whether iron is detectable in the TME, Perls’ staining was performed on human NSCLC pathological specimen (n = 19 samples). Iron positive staining was detected in 11 cases while 8 samples were iron negative. Table 1 summarizes the available clinical information from these patients grouped by the iron staining results. While cancer cells were negative for iron staining in all samples, some infiltrating cells were clearly iron positive. Interestingly, iron positive cells accumulated in the vicinity of RBCs (10/11 patients) (Figure 1A; Table 1). In addition, iron staining strongly overlapped with positive immune staining for CD68, a macrophage marker (Figure 1B). To further analyze if iron accumulates in TAMs, we isolated leukocytes from fresh tumor tissue (human lung adenocarcinoma) (Table 2) and separated them by exposure to a magnetic field according to their iron content. Cells retained in the magnetic fraction (due to their high iron content) stained positive for iron while cells in the flow through were negative for iron (Figure 1C). Cells within the magnetic fraction were strongly CD68 positive demonstrating that TAMs accumulate iron in lung adenocarcinoma. In addition the macrophage population in the flow through was iron negative (Figure 1C). To explore whether iron-loaded TAMs (iTAMs) are associated with specific areas in the tumor, we performed iron staining in TMA from 116 patients with NSCLC. From each patient, three areas of the original histology block were represented: normal lung, tumor center, and the invasion front (Figure 1D). 38 of 116 patient samples stained positive for iron in the tumor center and/or the invasion front (Table 3). Quantification of iron staining revealed that signals are significantly higher in the invasion front and tumor center when compared to normal lung (Figure 1E). We next analyzed if iron content correlates with tumor size (length in centimeters). Even if it represents a relatively crude clinical parameter, the tumor size of patients scored as iron positive was significantly smaller compared to iron negative tumors (Figure 1F) independently of the histological subtype (Figure S1A in Supplementary Material) and tumor grade (Figure S1B in Supplementary Material). Taken together these data demonstrate that iron accumulates in a subset of TAMs that localize in the vicinity of RBCs in invasive areas of the tumor. Furthermore, the presence of iTAMs correlates with smaller tumor size.

To understand the molecular mechanism(s) of how iron accumulates in TAMs, we used the LLC mouse model, a widely used syngeneic model for NSCLC (34). Consistent with our findings in human NSCLC, infiltrating cells in the proximity of RBCs stained positive for iron (Figure 2A), whereas cancer cells were negative for iron staining. We next isolated TAMs from LLC tumors and sorted for the surface markers CD11b⁺/Gr-1⁻/F4/80⁺ (35) by flow cytometry (see gating strategy in Figure S2A in Supplementary Material). We further separated them into iron-spared (i(-)TAMs) and iron-loaded TAMs (iTAMs) by magnetic isolation (Figure 2B). iTAMs express elevated mRNA levels of Cd163 [the scavenger receptor for hemoglobin and haptoglobin-hemoglobin complexes (36), and Hmox1 (the inducible isoform of heme oxygenase responsible for heme degradation)], while mRNA expression of ferroportin (Fpn), was low and similar to iron-spared TAMs (Figure 2C). At the protein level, ferroportin was not detectable in TAMS, but in splenic macrophages that were analyzed as a control (Figure S3A in Supplementary Material). Ferroportin is internalized and degraded by the binding of the hepatic iron-hormone hepcidin. Inflammation increases hepcidin levels and decreases ferroportin cell surface expression causing iron retention in macrophages (37–42). Tumor-bearing mice neither showed elevated serum levels of the inflammatory cytokines IL-6 and IL-1β, known activators of hepcidin expression (43, 44) (Figure S3B in Supplementary Material) nor increased hepatic hepcidin mRNA levels (Hamp1) (Figure S3C in Supplementary Material). Consistently, the expression of ferroportin was detectable in cell types contributing to systemic iron supplies, such as splenic macrophages, Kupffer cells and enterocytes (Figure S3D in Supplementary Material). Taken together these data suggest that iron retention in TAMs does not depend on a hepcidin-dependent decrease of ferroportin expression. Consistently, TAMs (CD11b⁺/Gr-1⁻/F4/80⁺) isolated from LLC tumors of Slc40a1^C326S mice, which express a ferroportin allele with a point mutation (C326S) that causes resistance to hepcidin-binding (31), did not express ferroportin protein in most TAMs (Figure 2D) pointing toward a hepcidin-independent downregulation of ferroportin. We conclude that iTAMs are hallmarked by a phenotype of hemoglobin recycling and iron retention.

We observed that iTAMs co-localize with RBCs in the TME. We next quantified tumor micro-bleedings in response to LLC tumor growth by high field magnetic resonance imaging (MRI). Micro-bleedings were detectable at day 7 after LLC inoculation and increased significantly in number with tumor progression (Figures 3A,B). In addition, dynamic contrast-enhanced (DCE) imaging showed that tumor vessels were more permeable than those within muscle tissue (Figure 3C), and thus more fragile and leaky, leading to the occurrence of micro-bleedings within tumors. Heme, a product of RBC degradation, promotes inflammation by activating macrophages, neutrophils, and endothelial cells (23, 45). We next dissected hemorrhagic areas (H) and NH areas from the same tumor (Figure 3D) and analyzed their properties. Hemorrhagic areas showed increased heme levels and Hmox1 mRNA expression (Figures 3E,F), as well as elevated Cd163 levels (Figure 3F), consistent with the accumulation of iTAMs in areas of RBC extravasation. In addition, the percentage of Gr-1⁺ cells (gated as CD11b⁺/Gr-1⁺) was increased (Figure 3G). The marker Gr-1 (Ly-6C/Ly-6G) is expressed in neutrophils and granulocytes and myeloid-derived suppressor cells (35). Interestingly, expression levels of chemokines known for their neutrophil and myeloid cell chemoattractant activity, KC (Cxcl1) and MIP-2 (Cxcl2) were increased in hemorrhagic compared to NH areas (Figure 3H). In addition, expression of M-CSF (Csf1) and GM-CSF (Csf2) that drive macrophage differentiation were also elevated (Figure 3I). In hemorrhagic areas TAMs (CD11b⁺/Gr-1-/F4/80⁺) expressed less CD206, an M2 polarization marker, suggesting a shift of macrophages toward a pro-inflammatory phenotype (Figure 3J). Consistently, mRNA expression of the M1 markers Nos2 and Il-6 were increased in H areas (Figure 3K). In summary, we show that hemorrhagic areas occur due to RBC extravasation from permeable vessels in the TME and are characterized by an infiltration of leukocytes, the accumulation of CD206^low iTAMs, as well as increased inflammation.

To further understand the impact of RBC extravasation and degradation in TAMs, we next established an in vitro system by differentiating isolated bone marrow cells into TAMs. We incubated cells isolated from the bone marrow with (CM) from LLC cells for 4 days, and analyzed adherent cells that differentiated into macrophages (CD11c^neg/Gr-1^neg/CD11b^pos/F4/80^pos) (Figure 4A). We observed that macrophages differentiated with CM polarized toward an anti-inflammatory, tumor-tolerant phenotype showing increased expression of Arginase 1, Ccl2, and Vegf (Figure 4A). We next mimicked conditions in hemorrhagic areas of the TME by adding RBCs to macrophages. In order to mimic senescent hemolytic RBCs (e.g., as would be expected to occur in the prooxidant inflammatory TME) we prepared aged RBC by treating RBC overnight with calcium. In vitro TAMs accumulated iron when incubated with either aged RBCs (aRBC) (to mimic senescent/hemolytic RBCs) or non-aged RBCs (Figure 4B). Heme (heme and hemoglobin) concentration was significantly higher in the supernatant of macrophages treated with aged RBCs (Figure 4C). Additionally, heme was also detected in the supernatant on non-aged RBC, suggesting that inflammation may induce RBC breakage. Aged RBCs exposed to CM from LLC cells changed shape and size (higher FSC-A) and were more prone to hemolysis when compared to control media (Figure S4A–C in Supplementary Material). Furthermore, mRNA expression of Hmox1 and Spi-c, markers for “iron recycling macrophages” was increased in macrophages treated with both RBCs sources (Figure 4D). Similar to observations in hemorrhagic areas of the tumor, the presence of RBCs increased mRNA expression of Cxcl1 (KC), Cxcl2 (MIP-2), and Csf1 (M-CSF) (Figure 4E), suggesting that the activation of macrophages by heme/iron may trigger recruitment of myeloid cells. We further observed that in the presence of aged RBCs, pro-inflammatory markers (CD86, Il6, Nos2, and Tnfa) were increased, while the M2 markers remained unchanged (Arginase 1) or were decreased (CD206, Ym1, and Il10) (Figures 4F,G). Our results show that the responses of macrophages to RBCs in the in vitro model mirrored those in the hemorrhagic areas of the TME, with iron accumulation in macrophages (iTAMs) and a shift toward a pro-inflammatory phenotype.

Pro-inflammatory macrophages in the TME can promote tumor cell death by producing ROS (46, 47). To test the effect of iron loading on tumor cell killing in vitro, we co-cultured macrophages exposed to hemolytic RBCs with LLC cells. This co-culture strongly reduced LLC cell viability (Figure 5A, gating strategy shown in Figure S2B in Supplementary Material); macrophages showed increased ROS levels (Figure 5B) and a shift toward pro-inflammatory phenotype characterized by the decreased expression of CD206 and increased expression of CD86 (Figure 5C). We next performed TUNEL staining and DAB enhanced Perls’ staining of consecutive slides of LLC tumors. Consistent with our findings in cultured cells, apoptotic areas co-localize with hemorrhagic areas and with the presence of iTAMs in the TME of LLC tumors (Figure 5D). Our findings demonstrate that macrophage iron loading in the TME correlates with and leads to a pro-inflammatory phenotype and anti-tumor activity. Thus, increasing the population of iTAMs in the TME emerged as a promising new therapeutical option to counteract tumor growth.

We tested CLIO nanoparticles as a strategy to deliver iron to macrophages without provision of this growth factor to tumor cells. CLIO nanoparticles are used for in vivo imaging by magnetic resonance and are specifically ingested by phagocytic cells such as TAMs, rather than neighboring cell types such as tumor cells or other leukocytes (48). Similar to TAMs exposed to RBCs, macrophages accumulated CLIO nanoparticles (Figure 6A) which induced decreased expression of CD206 (Figures 6A,B). We next tested whether TAMs treated with CLIO displayed anti-tumor activity in vivo and in vitro. Incubation of LLC cells with CLIO-treated TAMs significantly reduced tumor cell viability (Figure 6C), correlating with a decrease in the expression of CD206 in TAMs (Figure 6D). We injected mice with LLC cells and analyzed tumor growth with or without co-injection of CLIO. 15 days after injection, tumors from CLIO-co-injected mice were significantly smaller than control tumors (Figure 6E). The CLIO nanoparticles accumulated in TAMs (Figure 6F), which were predominantly localized in the periphery of the tumor and near the invasive front (Figure 6G). The presence of iTAMs was also detected in NT samples, near areas of RBCs extravasation (Figure 6G). In accordance, we observed a moderate reduction of CD206 expression and increased CD86 expression (Figure 6H), together with a significant increase of the CD8/CD4 T-cell ratio (Figure 6I). These data show that the injection of iron nanoparticles limits tumor growth in the LLC cell mouse model, translating the in vitro effect observed with hemolytic RBCs. Our data suggest that iron nanoparticles injection could be developed as a therapeutic strategy to inhibit tumor growth via TAMs reprogramming.