The respective antitumoral and protumoral roles of M1 and M2 tumor-associated macrophages (TAM) typify the complexity of macrophage function in cancer. In lung cancer, density and topology of distinct TAM phenotypes at the tumor center (TC) versus the invasive margin (IM) are largely unknown. Here, we investigated TAM subtype density and distribution between TC and IM in human lung cancer and TAM associations with overall survival. Macrophages isolated from adjacent nontumor tissue (NM), the TC (TC-TAM), and the IM (IM-TAM) were analyzed with RNA-sequencing (RNA-seq). Lung tumor tissue microarrays from 104 patient samples were constructed. M1 and M2 TAMs were identified using multiplex immunofluorescence staining and a tumor cell-TAM proximity analysis was performed. RNA-seq identified marked differences among NM, TC-TAM, and IM-TAM. On the basis of a panel of five selected markers (CD68, IL12, CCR7, CD163, and ALOX15), M2 predominance over M1 and M2 proximity to tumor cells was observed, especially at IM. Tumor cell proximity to TAM was linked with tumor cell survival and hypoxia was associated with accumulation of M2 TAM. Notably, lower density of M1 TC-TAM and higher proximity of tumor cells to M2 IM-TAM or lower proximity to M1 IM-TAM were linked with poor survival. In addition, three novel molecules (UBXN4, MFSD12, and ACTR6) from RNA-seq served as potential prognostic markers for lung cancer, and M2 predominance and juxtaposition of M2 TAM near tumor cells were associated with poor survival. Together, our results reveal the marked heterogeneity of TAM populations in different tumor regions, with M2 TAM predominance, particularly at IM.

Significance:

This study underlines the significance of the density, spatial distribution, and gene expression of TAM phenotypes as prognostic factors for overall survival in lung cancer.

Despite advances in understanding the molecular mechanisms, and improvements in diagnostics and treatment, lung cancer remains the leading cause of cancer-related morbidity and mortality worldwide (1). Lung cancer is classified into small-cell and non–small cell lung carcinoma (NSCLC) in which adenocarcinoma, squamous cell carcinoma, and large-cell carcinoma are the major histologic subtypes. Recent studies suggest the tumor microenvironment (TME) plays pivotal roles in lung cancer progression and is considered as a prognostic biomarker (2).

Tumor-associated macrophages (TAM) are the most abundant stromal cell populations in TME and two discrete activation states of macrophages based on their immune responses were identified. TAMs that inhibit angiogenesis and activate antitumoral immunity are defined as M1 TAMs, and those that facilitate tumor growth, invasion, and metastasis are defined as protumoral M2 TAMs (2). We reported TAM infiltration correlated with lung tumor stage and metastasis (3). Importantly, macrophage depletion via clodronate liposomes or employing transgenic macrophage Fas-induced apoptosis in mice inhibited lung tumor growth and metastasis (4). These studies demonstrate the central role of TAMs in lung cancer growth and metastasis. However, a deeper understanding of the heterogeneity and topography of TAM phenotypes in human lung cancer tissues and its correlation with patient survival is missing.

Recent literatures suggest topologically distinct distribution of immune cells within the TME. For example, the prognostic impact of CD8+ T-cell density with survival was highly significant at invasive margin (IM) in contrast with tumor center (TC) of lung cancer (5). However, distribution patterns of TAMs and TAM phenotypes between TC and IM in lung cancer remain unexplored (6).

In addition to genomic heterogeneity of immune cells, their spatial distribution may be particularly relevant to tumor progression (6, 7). Close interactions among immune and tumor cells generate complex ecological dynamics that can ultimately influence tumor progression and response to treatment (8, 9). Hence, the proximity of immune cells to tumor cells may have profound influence on both cell types as this allows them to interact via soluble factors or cell–cell contact.

As of now, a comprehensive analysis of lung cancer, characterizing the composition, location, and transcriptome signature of TAMs (i) in distinct regions of the tumors and (ii) with respect to TAM–tumor cell proximity, and (iii) linking these findings with survival, has not been performed. Undertaking such an approach in human lung adenocarcinoma, squamous cell carcinoma, and large-cell carcinoma samples, we noted marked heterogeneity of TAMs between IM and TC, with high M2 TAM density in particular at IM. Lower abundance of M1 TAMs or closer distance of M2 TAMs to tumor cells were correlated with poor survival.

Specimen collection

Tumor tissue specimens, from which TC and IM as well as the adjacent nontumor based on the histopathologic review, were collected from patients with lung squamous cell carcinoma at the time of surgery before chemotherapy after obtaining written informed consent at the University Hospital Giessen (Giessen, Germany). The three specimens used for RNA-sequencing (RNA-seq) analysis were obtained from male patients, aged 76, 75, and 60 years, with tumor stages II, IV, and IV, respectively. The four specimens used for qRT-PCR detection were obtained from two males and two females, aged 60, 67, 73, and 74 years, with tumor stages IV, II, II, and II, respectively. The IM was defined as the region centered on the border separating host tissue from malignant tissue, extending 1 mm in all directions (5, 10, 11).

FACS

Single-cell suspensions were prepared from tissue specimens using the Human Tumor Dissociation Kit (Miltenyi Biotec). After blocking Fcγ receptors for 15 minutes, the cells were stained for 15 minutes at 4°C with the following antibodies: CD1c-PE/Dazzle594 (BioLegend, 331531), CD15-FITC (BD Biosciences, 560997), CD33-BV510 (BD Biosciences, 563257), CD45-AF700 (BioLegend, 368514), CD326-FITC (BioLegend, 324203), HLA-DR-APC/Fire750 (BioLegend, 307658), MerTK-BV421 (BioLegend, 367603), CD14 PerCP-Cy5.5 (BD Biosciences, 561116), and CD64 BV605 (BD Biosciences, 740406). Flow cytometry–based cell sorting was performed using a BD FACSAria III fluorescence-activated cell sorter. The sorting strategy involved the exclusion of debris and cell doublets based on light scattering, and cell viability was assessed using 7-aminoactinomycin D (BD Biosciences). The macrophages were sorted as CD45+CD15+CD33+HLA-DR+MerTK+CD14+CD64+CD326CD1c cells.

RNA extraction and RNA-seq

After FACS, total RNA was isolated from macrophages using the Quick-RNA-MicroPrep Kit (Zymo Research). The RNA quality and quantity were assessed using Labchip GX Touch (PerkinElmer). RNA-seq libraries were constructed using approximately 2 ng total RNA as input for the SMARTer Stranded Total RNA-Seq Kit (Pico Input Mammalian; Takara Clontech). Sequencing was performed using an Illumina NextSeq500 Instrument using v2 chemistry, resulting in a minimum of 36 × 106 reads per library with a 75-bp single-end setup.

High-quality reads were aligned to Ensembl human genome version hg38 (GRCh38; ref. 12). The number of reads aligning to genes were counted with featureCounts 1.4.5-p1 using the Subread package (13). DESeq2 was used to estimate fold changes in expression (14) and P values were adjusted for multiple testing with the Benjamini–Hochberg. The RNA-seq data were deposited in the Gene Expression Omnibus archive (accession number GSE137343). For preparation of heatmaps, a count matrix representing all transcripts identified by RNA-seq was prepared for the macrophage samples from paired adjacent nontumor tissues and TC and IM samples. The mean log2 fold change was calculated for each transcript, and the FDR (probability of incorrectly accepting a difference among the macrophages isolated from the nontumor, TC, and IM tissues) for each transcript was calculated according to Storey method. Genes with FDR values less than 0.05 were considered to be differentially expressed.

Tumor specimens and tissue microarrays

A cohort of 104 patients with stage I–IV lung cancer was included in this study. The Ethical Committee of the University Hospital Munich (Munich, Germany) approved the collection and analysis of all samples, in accordance with the national laws and the Good Clinical Practice/International Conference on Harmonisation guidelines. Informed consent was obtained from all patients. Tissue microarrays (TMA) were constructed for the TC and IM of lung cancer samples. In standard paraffin sections, the TC and IM regions were histomorphologically analyzed. TMAs were prepared using 1 mm tissue cores with 5 μm thickness using standard procedures. To evaluate tumor heterogeneity, three representative cores from IM and TC were used to construct TMAs for each patient.

Multiplex staining and multispectral imaging

Seven-color multiplex fluorescence staining was performed using the PerkinElmer-Opal-Kit according to the manufacturer's instructions. Briefly, the TMA slides were dewaxed with xylene, rehydrated through a graded ethanol series, and fixed with 10% neutral-buffered formalin prior to antigen retrieval that was performed with Opal-AR6 Buffer (PerkinElmer) using microwave incubation. After blocking with 1% BSA in PBS for 20 minutes at room temperature, the staining process was performed six times in a serial fashion, including incubation with a primary antibody, a horseradish peroxidase–conjugated secondary antibody, and sequentially an Opal fluorophore (Supplementary Table S1). Finally, the nuclei were counterstained with DAPI. The seven-color Opal slides were visualized using the Vectra Quantitative Pathology Imaging System (PerkinElmer). Spectral unmixing was applied to distinguish the seven different fluorescence signals. The unmixed images were processed using the InForm (PerkinElmer) image analysis with tissue segmentation, cell segmentation, and phenotyping. Tissue segmentation based on the epithelial cell marker cytokeratin was used to differentiate the parenchyma from the stroma, and the DAPI-based cell segmentation was used to improve phenotyping. The cells were phenotyped into the following subsets: M1 TAMs, CD68+IL12hiCCR7hiCD163lowALOX15low; M2 TAMs, CD68+CD163hiALOX15hiIL12lowCCR7low; and tumor cells, Cytokeratin+CD68. Median intensities were used to set cut-off values for the stained markers. Sequentially, the counts of M1 and M2 TAMs were normalized to the total cell counts for the total TC and IM areas to generate the density of TAMs per 1,000 cells. The proximity distance between the tumor cells and TAMs was measured using HALO software. The proximity distance was defined as the average number of tumor cells distributed within a 30-μm radius from the nuclear center of any given M1 or M2 TAMs (15).

qRT-PCR

The qRT-PCR reaction mixture was prepared using PowerUp SYBR Green Master Mix (Applied Biosystems). Human intron-spannin primers were used, as listed in Supplementary Table S2. The master mix and the cDNA template were combined in nonskirted 96-well plates, and the reaction was run in an Applied Biosystems StepOne Real-time PCR machine. Data were analyzed using StepOne software v2.3 and normalized against the expression of a housekeeping gene, hypoxanthine phosphoribosyltransferase1 (HPRT). mRNA expression levels are presented as 2−ΔΔCt values.

Statistical analysis

The Kolmogorov–Smirnov test determined that the TAM density and proximity variables were not normally distributed (P < 0.05); therefore, the Wilcoxon signed-rank test for two-paired samples was used to analyze TAM density and proximity to tumor cells. The Mann–Whitney U and the Kruskal–Wallis tests were used to compare two and multiple independent samples, respectively. Spearman rank correlation coefficient was calculated to assess the correlations between the TAM-related variables and tumor size. The Kaplan–Meier method was used to estimate overall survival, and differences were assessed using the log-rank test. The independent prognostic values of the density and proximity of M1/M2 TAMs were estimated using univariate and multivariate Cox proportional hazard regression models. Statistical analyses were performed with SPSS ver. 17.0 (SPSS), Prism 6.0 (GraphPad Software), and R software. All statistical tests were two-sided, and P < 0.05 was considered significant unless otherwise specified. For two-group comparisons among multiple independent samples, adjusted significance levels (P = 0.01 for six-group comparisons and P = 0.017 for four-group comparisons) were used to avoid increasing the type I error.

Additional methodologic details on cell lines and immunocytochemistry are provided in an online Supplementary Data.

Gene expression profiling indicates heterogeneity among macrophage populations at the TC, IM, and nontumor regions

RNA-seq analysis identified differentially expressed genes (DEG) among TC (TC-TAMs), IM (IM-TAMs), and adjacent nontumor tissue–derived macrophages (NM). The tumor epithelial cell marker CD326 (16), the neutrophil marker CD15 (17), and the common leukocyte marker CD45 (18) were used to exclude tumor cells; and CD33 (19) and HLA-DR (20) were used to gate myeloid cells. In addition, dendritic cells were excluded on the basis of CD1c expression (21), and the macrophage markers MerTK (22), CD14, and CD64 (23) were used to purify macrophages using FACS (Fig. 1A). The four-way plot determined differences in gene expression for IM-TAMs and TC-TAMs relative to NMs revealed 835 and 651 genes that were exclusively highly expressed in IM-TAMs and TC-TAMs, respectively. A total of 357 genes were expressed at comparable levels between TC-TAMs and IM-TAMs (Fig. 1B). The top 50 differentially expressed protein-coding genes are shown in Fig. 1C. Furthermore, distinct cellular signaling pathways were differentially activated among NMs, TC-TAMs, and IM-TAMs. For instance, the expression levels of genes that govern the cadherin and Wnt signaling pathways were significantly different between TC-TAMs and IM-TAMs (Supplementary Fig. S1A). These findings revealed a substantial heterogeneity between the TAMs residing at TC versus those at IM, even in the same lung cancer sample. Therefore, in addition to comparing gene expression patterns in macrophages between tumor and adjacent nontumor tissues, differences in the spatial distribution of TAMs between TC and IM may also require assessment. In addition, we compared our dataset with recent data reported by Lavin and colleagues (7) who described differentially regulated transcripts in lung adenocarcinoma–derived TAMs compared with NMs, including the downregulation of LILRB2, LMNA, FCGR3A, VIM, LST1, HLA-DRA, and FCN1, and the upregulation of TNFRSF1A, GPX1, CTSD, IER3, SPP1, CEBPB, CD163, and TREME2 (Supplementary Fig. S1B and S1C; ref. 7). These datasets were highly comparable and confirmed the gene expression tendencies between NMs and lung squamous cell carcinoma–derived TC/IM-TAMs. Hence, the transcriptional signature of lung adenocarcinoma–derived TAMs is similar to the transcriptional signature of lung squamous cell carcinoma–derived TAMs.

Five specific markers are sufficient for distinguishing M1 and M2 TAM subtypes in lung cancer

To evaluate cell markers that might be appropriate for distinguishing TAMs, the mRNA expression levels of 12 M1 and M2 macrophage marker genes were evaluated in NMs, IM-, and TC-TAMs (23, 24). Compared with NMs, the expression levels of IL12B, CCR7, ALOX15, and CD163 were significantly altered in TAMs (Supplementary Fig. S1C). Therefore, the expression patterns of IL12B, CCR7, ALOX15, and CD163 were selected for examination by multiplex staining. In addition, immunocytochemistry was performed to visualize cytokeratin, CD68, IL12, CCR7, CD163, and ALOX15 in A549 lung cancer cells and M1 versus M2 peripheral blood monocyte–derived macrophages. Cytokeratin was exclusively expressed in A549 cells, whereas CD68 was solely expressed in macrophages (Supplementary Fig. S1D). Moreover, IL12 and CCR7 were highly expressed in M1 macrophages, whereas CD163 and ALOX15 were preferentially expressed in M2 macrophages (Supplementary Fig. S1D). Therefore, cytokeratin and CD68 were sufficient for distinguishing cancer cells from macrophages, and the remaining examined marker set was capable of distinguishing M1 and M2 macrophage subtypes.

In-line with the wide use of CD68 to identify monocyte lineage, the combination of CD68 and a single M1- or M2-related marker has been applied to distinguish TAM subtypes (Fig. 2A and B; refs. 25–27). To examine whether multiple staining for five macrophage markers is more effective than using three markers, scatter plots and Wilcoxon signed-rank test were used to compare M1 and M2 TAM densities (TAM counts per 1,000 total cells). The axes in the plots represented the cell densities for the respective categories, and the data points located along the diagonal indicated similar cell densities between these two groups, suggesting a higher specificity if those data points were more distant from the diagonal. Compared with the TAMs defined by three markers (CD68, with a single marker for M1 and M2), the densities of TAMs defined by five markers were substantially lower (Fig. 2C and D). More precisely, considering five markers as 100% specific, the specificity was decreased between 52% and 62% by three markers and 13%–46% by four markers (Fig. 2C and D; Supplementary Fig. S1E). In addition to the CD68+IL12hiCCR7hiCD163lowALOX15low-defined M1 TAMs and CD68+CD163hiALOX15hiIL12lowCCR7low-defined M2 TAMs, intermediate macrophage subpopulations were identified (Fig. 2E). A significant benefit (P = 0.010) on overall survival was observed for patients with higher densities of CD68+IL12hiCCR7hiCD163lowALOX15low-defined TAMs, which corresponded to a 40% reduction in the risk of death (Fig. 2F). Together, these findings indicated that M1/M2 TAMs could be sufficiently distinguished from the non-M1/M2 populations using five markers.

Higher spatial density of M1 TAMs is associated with significantly longer overall survival of patients with lung cancer

Table 1 lists the baseline clinicopathologic characteristics of 104 patients with NSCLC that were enrolled in this study. The median age was 65 years (range, 38–83), and the median overall survival time was 37 months (range, 0–162). None of the patients underwent preoperative chemotherapy or radiotherapy. The histologic grades assessed using the World Health Organization classification were adenocarcinoma, squamous cell carcinoma, and large-cell carcinoma in 50% (n = 53), 37% (n = 36), and 10% (n = 10) of patients, respectively. The heterogeneity of TAM density and distribution between TC and IM regions of lung adenocarcinoma, squamous cell carcinoma, and large-cell carcinoma were assessed using seven-color multiplex fluorescence staining (Fig. 3A and B; Supplementary Fig. S2). To visualize the distribution of TAMs and cancer cells, phenotype maps were generated on the basis of previously described cell markers (Figs. 2B and 3B).

For both TC and IM regions of all involved cancer subtypes, M2 TAMs showed a dominant density compared with M1 TAMs. The density of M2 IM-TAMs was significantly increased compared with that of M2 TC-TAMs, whereas the M1 TAM density did not differ significantly between TC and IM regions (Fig. 3C). Moreover, M1 and M2 TAMs infiltrated significantly more in the stroma than in the parenchyma (Fig. 3D). The M1 TC-TAM density of adenocarcinoma tended to be greater than that in squamous cell carcinoma, whereas the M2 TAM density was comparable among all examined lung cancer subtypes (Fig. 3E and F).

To evaluate whether the spatial TAM density differences could predict prognosis, all the patients in the entire cohort were dichotomized on the basis of the median TAM density. The Kaplan–Meier estimates revealed significant differences in overall survival according to the M1 TAM density. Patients with a high M1 TC- and IM-TAM densities had a significant overall survival benefit compared with those with a low M1 TC- and IM-TAM densities. However, there were no significant associations between the M2 TAM densities and overall survival (Fig. 3G). We also investigated the relationship between the spatial TAM density and overall survival among the different cancer subtypes and found that patients with adenocarcinoma with increased M2 IM-TAM densities had a poorer prognosis, whereas higher M1 TC-TAM densities were associated with longer survival in patients with squamous cell carcinoma (Supplementary Fig. S3A). However, the TAM density was neither significantly correlated with overall survival in patients with large-cell carcinoma nor with the aforementioned clinicopathologic characteristics (Supplementary Fig. S3A; Supplementary Table S3). Furthermore, when measured at an infiltration depth of 100 μm, more M2 TAMs than M1 TAMs were identified infiltrating the parenchyma, indicating that the tumor cells in the analyzed samples were closer to the M2 TAMs than to the M1 TAMs (Fig. 3B; Supplementary Fig. S3B).

Spatial distributions of M1/M2 IM-TAMs are independent survival predictors

The phenotype maps allowed us to determine the spatial proximity distance between TAMs and tumor cells. Proximity was defined as the average number of tumor cells distributed within a 30-μm radius from the nuclear center of any given M1 or M2 (15). Overall, the tumor cells were located more proximally to M2 IM-TAMs than to M1 IM-TAMs (Fig. 4A and B). More precisely, the tumor cells were closer to M2 IM-TAMs than to M1 IM-TAMs in squamous cell carcinoma (Fig. 4C and D). In addition, spine plots revealed that the incidence of metastasis increased significantly with the increasing proximity of tumor cells to either M2 TC-TAMs or M2 IM-TAMs (Fig. 4E; Supplementary Table S3). Besides, larger tumor size differences were significantly correlated with the increased proximity of tumor cells to M2 IM-TAMs (Fig. 4F; Supplementary Table S3). Furthermore, among overall lung cancer cohort, the survival was significantly longer among patients with tumor cells that were closer to M1 TC-TAMs or more distant from M2 TC/IM-TAMs (Fig. 4G). This profile was true for all histologic cancer subtypes; although significance levels were only partially reached because of the low sample numbers (Supplementary Fig. S3C).

Multivariate Cox proportional hazard analysis was applied to determine whether the spatial density and distribution of TAM subtypes were independently associated with overall survival time (Table 2). Along with the density of TAMs and the proximity of tumor cells to TAMs, age, gender, tumor stage, tumor size, metastatic and recurrent status, and histologic subtypes were included in the multivariate analysis. Univariate Cox regression analysis revealed that tumor stage, tumor size, metastasis, and the proximity of tumor cells to M1 TC-TAMs or M2 TC/IM-TAMs had significant impacts on overall survival. Multivariate analysis indicated that tumor stage (HR, 1.728; P = 0.001), metastasis status (HR, 2.304; P = 0.040), histologic subtype (HR, 0.652; P = 0.014), M1 TC-TAM density (HR, 0.986; P = 0.030), proximity of tumor cells to M1 IM-TAMs (HR, 0.503; P < 0.001), and proximity of tumor cells to M2 IM-TAMs (HR, 2.049; P < 0.001) were independent predictors of overall survival.

Proximity of tumor cells to TAMs directly affects tumor cell survival

The proliferation marker, Ki67, and the apoptosis marker, cleaved Caspase-3, were used to evaluate tumor cell turnover. However, technical limitations associated with the seven-color staining protocol confined our ability to stain additional markers. Therefore, we were forced to omit one macrophage marker to stain for either Ki67 or cleaved Caspase-3. As determined previously, the omission of IL12, CCR7, ALOX15, or CD163 would lead to 46%, 13%, 32%, and 23% reduction in specificity, respectively, indicating that CCR7 was the least important macrophage marker (Supplementary Fig. S1E). Therefore, staining for CD68, IL12, ALOX15, CD163, cytokeratin, DAPI, and either Ki67 or cleaved Caspase-3 was performed using the tissue microarrays. Cleaved Caspase-3+ tumor cells were more proximal to M1 TAMs than to M2 TAMs (Fig. 5A). In contrast, Ki67+ tumor cells were more distal to M1 TAMs than to M2 TAMs (Fig. 5B). These results indicated that M1 TAMs promote apoptosis in proximal tumor cells, whereas M2 TAMs establish a favorable environment that allows tumor cells to survive and proliferate. Therefore, the proximity of TAMs to tumor cells directly affected tumor cell survival.

Hypoxia contributes to the accumulation of M2 TAMs

To investigate the relationship between hypoxia and the accumulation of M2 TAMs, especially at IM, the oxygenation of tumors was assessed by analysis of hypoxic marker expression patterns, including carbonic anhydrase 9 (CA9) and hypoxia-inducible factor 1-alpha (HIF1α; refs. 28, 29). The expression levels of CA9 and HIF1α at IM were significantly elevated compared with those at TC (Fig. 5C and D). In addition, positive correlations between M2 TAM density at IM and the expression of either CA9 or HIF1α was observed (Fig. 5E). These findings suggested that hypoxia contributed to the accumulation of M2 TAMs.

UBXN4, ACTR6, and MSFD12 are potential prognostic indicators for lung cancer survival

Having established the relevance of spatial TAM subset distribution in tumors, we wondered whether new prognostic markers would emerge from RNA-seq analysis of IM- and TC-TAMs and NMs. To this end, we identified the top DEGs (50 from each group, in total 150 DEGs) from the RNA-seq analysis. Furthermore, we selected 48 of the 150 DEGs using in silico analysis for relevance to cancer microenvironment and cancer biology. In silico resources, such as The Human Protein Atlas and The Cancer Genome Atlas (TCGA), and single-immune cell transcriptomes in lung cancer (30) were analyzed for this selection. As a next step, we evaluated the mRNA expression of these top 48 DEGs in M1 and M2 macrophages generated from human peripheral blood mononuclear cells (Supplementary Fig. S4A). Compared with M1 macrophages, CPD, SERPINB9, WARS, HIVEP1, PAG1, ERAP2, ACTR6, SPATA2, UBXN4, TMEM189, IQGAP2, SDF4, AP153, LRRFLP2, TM2D3, STK38, UBR2, IST1, MED16, METAP2, DBF4, PIHID1, ZNF37A, PUS7L, and SEL61A were significantly downregulated in M2 macrophages, whereas KMT2C, PLOD1, PACS2, BCL2L1, PAQR4, HAMP, MFSD12, and UBTD1 were significantly upregulated in M2 macrophages (Supplementary Fig. S4B). After confirming their expression in M1 and M2 macrophages, we determined the strength of correlations between each validated gene and TAM-related markers (CSF1R and CD163), using cBioPortal for Cancer Genomics (http://cbioportal.org) and Illumina HiSeq_RNASeqV2 of Lung Adenocarcinoma (TCGA, PanCancer Atlas; ref. 31, 32).

We identified that among the 48 DEGs, three displayed a strong correlation with TAM-related markers, namely ACTR6, UBXN4, and MFSD12. ACTR6 (actin-related protein 6) and UBXN4 (UBX domain protein 4) were negatively correlated, whereas MFSD12 (major facilitator superfamily domain containing 12) was positively correlated with either CSF1R or CD163 (Supplementary Fig. S4C). However, the roles of ACTR6, UBXN4, and MFSD12 in cancer cell development and progression, and in macrophage biology, as well as their influences on the TME are unknown. Moreover, immunocytochemistry confirmed that ACTR6 and UBXN4 were preferentially expressed in M1 macrophages, and MFSD12 was highly expressed in M2 macrophages at the protein level (Supplementary Fig. S5A–S5C). Therefore, ACTR6, UBXN4, and MFSD12 can serve as potential macrophage subtype markers and novel lung cancer prognostic biomarkers.

Furthermore, to investigate UBXN4, ACTR6, and MFSD12 as potential lung cancer prognostic biomarkers, fluorescence staining was performed on the TMAs (Supplementary Fig. S5D–S5F). In a Kaplan–Meier analysis, high expression levels of UBXN4 at IM, high expression levels of ACTR6 at TC, and reduced expression levels of MSFD12 at TC were significantly associated with increased overall survival time among patients with lung cancer (Supplementary Fig. S5D–S5F). In addition, patients with lung cancer with high expression levels of UBXN4 in CD68+ cells at IM, and the increased production of ACTR6 in CD68+ cells at either TC or IM presented with significant overall survival benefits (Supplementary Fig. S5D and S5F). However, no significant association between the expression of MSFD12 in CD68+ cells and overall survival was observed (Supplementary Fig. S5E). These findings indicated that the general expression of UBXN4 at IM, the expression of ACTR6/MSFD12 at TC, and the CD68+ cell expression of UBXN4 at IM and ACTR6 at either TC or IM are potential prognostic indicators for lung cancer patient survival.

This study identified four key findings. First, gene expression profiling showed marked differences among NMs, TC-TAMs, and IM-TAMs. Second, M2 phenotypic predominance over M1 was observed, particularly at the IM, and hypoxia was associated with the accumulation of M2 TAMs. Third, M2 IM-TAMs were more proximal to tumor cells than M1 IM-TAMs, and the proximity of tumor cells to the different TAM phenotypes was correlated to tumor cell survival. Fourth, the reduced density of M1 TC-TAMs, increased proximity of tumor cells to M2 IM-TAMs, and reduced proximity of tumor cells to M1 IM-TAMs were independent lung cancer survival predictors. In addition, gene expression profiling and subsequent experiments identified three molecules, including UBXN4, MFSD12, and ACTR6, as potential novel prognostic markers for lung cancer.

A four-way plot deciphered the gene expression differences among NMs, IM-TAMs, and TC-TAMs based on the RNA-seq results. The functions of some of the top differentially regulated genes are known, with respect to macrophage biology and TAM–tumor cell interaction. For instance, STK38 facilitates Smurf1-mediated MEKK2 ubiquitination and degradation, negatively regulating TLR9-mediated immune responses in macrophages (33). In addition, HAMP is associated with the central role played by TAMs in the regulation of tumor iron homeostasis during breast cancer, supporting increased iron-targeting therapeutic approaches with regard to TAM modulation (34). Furthermore, CYP1B1 was expressed at a lower level in IM-TAMs than in TC-TAMs, and its deficiency is known to impair the phagocytic activity of macrophages (35). We identified UBXN4, MFSD12, and ACTR6 as novel biomarkers that may serve as potential prognostic indicators for lung cancer patient survival. ACTR6 possesses an evolutionarily conserved role in heterochromatin formation, and high MFSD12 expression has been positively associated with shorter survival and lung metastasis in patients with melanoma (36, 37). The findings in this study indicate that the general expression of UBXN4/ACTR6/MSFD12 and the expression of UBXN4 and ACTR6 in macrophages may serve as potential prognostic biomarker for lung cancer. However, the mechanisms through which UBXN4/ACTR6/MSFD12 expression influence on lung cancer progression and TAM biology require further investigation. Therefore, future studies will focus on the determination of TAM heterogeneity at the genomic level, and the roles played by the major differentially expressed transcripts in TAM biology, with regards to their impact on the TME and cancer progression.

TAMs exhibit functional plasticity, with both antitumoral and protumoral effects, depending on a variety of external factors (2, 38). In addition to these opposing effects, the distribution of TAM is another important factor to consider when evaluating TAMs for the prediction of clinical outcomes (26). We constructed TMAs of lung cancer tissues and identified an association between TAM density and spatial distribution and overall survival. TMAs are relatively cost-effective and efficient, and are commonly used as high-throughput assays in histochemical studies, incorporating different tissues or tissue regions from different patients. However, it remains unclear whether TMA data are as reliable as whole-tissue sections for clinicopathologic correlations and survival. Therefore, to provide a satisfactory representation of the specimens, three punches obtained from each TC and IM region for individual patient samples were utilized to construct TMA in this study.

The precise identification of TAM phenotypes is challenging, because no single specific marker can be utilized to distinguish M1 and M2 TAM subtypes. Most studies to date have used single or double immunostaining of macrophage markers, such as CD68, CD163, and HLA-DR, to identify TAMs (25–27, 39–41). An insufficient number of markers increases the possibility of misidentifying non-TAM populations as TAMs. Therefore, we used a combination of five markers to maximize the accuracy of TAM phenotype identification.

Previous studies demonstrated that hypoxia promotes M2 TAM infiltration and upregulates the expression of HIF1α, VEGF, glucose transporter-1, von Hippel-Lindau protein, and lactate dehydrogenase-A at IM compared with TC (2, 29). These findings are consistent with our observation that M2 TAMs displayed a greater density at IM than at TC. Tumor hypoxia occurs when uncontrolled cell proliferation predominates, limiting the supply of oxygen and nutrition. In this study, we observed that more M2 TAMs accumulated at IM, which displayed an increased hypoxic status compared with TC. The density of M2 TAMs was positively correlated with the expression of the hypoxia markers CA9 and HIF1α. Similarly, a significantly increased hypoxic status was observed at IM compared with TC for colorectal cancer (29). Three plausible mechanisms could explain how hypoxia contributes to M2 TAM accumulation (28). First, the hypoxic TME is enriched in cytokines, such as CCL2 and CSF1, which attract macrophages. Second, macrophage mobility is hampered within a hypoxic niche, due to hypoxia-dependent disruptions of the macrophage expression of CCR2, CCR5, and neuropilin-1 (NRP1). Third, hypoxia promotes macrophages to express the protumoral phenotype through the increased production of growth factors, such as VEGF, which supports tumor cell proliferation, and different matrix metalloproteinases, such as MMP7, to enhance tumor cell migration and invasion. Finally, TAMs might also directly contribute to the induction of tumor hypoxia, induced by the activation of AMP-activated protein kinase, which enhances the oxygen consumption rate in TAM mitochondria (42, 43).

Patients with higher M1 TAM densities at both TC and IM showed significantly better overall survival rates. Lower densities of M2 IM-TAMs were associated with better survival, although significance was only achieved in patients with adenocarcinoma. In addition, our results are consistent with previous studies, which demonstrated that a high M1 TC-TAM density was positively associated with better survival and that the infiltration of M2 TAMs was associated with reduced overall survival in lung cancer (26, 27, 39). Besides, the lack of significance for TAM density as a predictor of survival in patients with large-cell carcinoma can be attributed to the limited number of samples of this tumor subtype in this study.

The proximity measurements performed in this study shed light on the spatial TAM distribution. Patients with tumor cells that were more proximal to M1 TC-TAMs or more distant to M2 TC/IM-TAMs had better overall survival rates. The multivariate analysis identified both M1 TC-TAM density and the proximity of tumor cells to M1/M2 IM-TAMs as independent predictors of survival. These results also indicated that the reprogramming of protumoral M2 TAMs to yield an antitumoral phenotype, or specifically depleting M2 TAMs, could represent potentially effective therapeutic strategies for lung cancer (2).

In summary, this study underlines the significance of the density, spatial distribution, and gene expression of TAM phenotypes as prognostic factors for overall survival in lung cancer. The multiplex profiling of macrophages in combination with other immune cells may facilitate the stratification of patients with lung cancer. Focusing on the roles played by TAMs in TME may offer novel treatment strategies for lung cancer.

W. Seeger reports personal fees from Actelion, Novartis, Vectura, Medspray, and United Therapeutics outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

X. Zheng: Data curation, software, formal analysis, investigation, visualization, methodology, writing-original draft, writing-review and editing. A. Weigert: Data curation, software, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. S. Reu: Resources, methodology. S. Guenther: data curation, software, formal analysis, methodology. S. Mansouri: Formal analysis, investigation, methodology. B. Bassaly: Resources. S. Gattenlöhner: Resources. F. Grimminger: Resources, supervision, funding acquisition, project administration. S. Savai Pullamsetti: Conceptualization, resources, funding acquisition, writing-original draft, project administration, writing-review and editing. W. Seeger: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, writing-original draft, project administration, writing-review and editing. H. Winter: Conceptualization, resources, formal analysis, writing-original draft. R. Savai: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing.

R. Savai, S. Savai Pullamsetti, and W. Seeger were supported by the Max Planck Society. R. Savai and S. Savai Pullamsetti received Von-Behring-Röntgen-Stiftung (projects 66-LV06 and 63-LV01). R. Savai and A. Weigert were supported by Frankfurt Cancer Institute (LOEWE). R. Savai, S. Savai Pullamsetti, W. Seeger, F. Grimminger, and S. Guenther were supported by Cardio-Pulmonary Institute. R. Savai, S. Savai Pullamsetti, W. Seeger, F. Grimminger, and H. Winter were supported by the German Center for Lung Research (DZL). R. Savai and S. Savai Pullamsetti were supported by the German Research Foundation (DFG) by the CRC1213 (Collaborative Research Center 1213), projects A01 (to S. Savai Pullamsetti), A05 (to S. Savai Pullamsetti), and A10* (to R. Savai). S. Savai Pullamsetti received European Research Council Consolidator Grant (866051). The authors thank Drs. Jianning Zhao, Duo Chen, and Jochen Wilhelm for their assistance with statistical analyses using SPSS and R, and Praveen Mathoor, Margarete Mijatovic, Dr. Annika Franziska Fink, and Dr. Yingying Han for their technical support. The authors also thank Drs. Bernard A. Fox, Shawn M. Jensen, and Carmen Ballesteros-Merino for helpful discussions on macrophage markers and multiplex staining.

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.

1.
Ferlay
J
,
Colombet
M
,
Soerjomataram
I
,
Mathers
C
,
Parkin
DM
,
Pineros
M
, et al
Estimating the global cancer incidence and mortality in 2018: GLOBOCAN sources and methods
.
Int J Cancer
2019
;
144
:
1941
53
.
2.
Zheng
X
,
Turkowski
K
,
Mora
J
,
Brune
B
,
Seeger
W
,
Weigert
A
, et al
Redirecting tumor-associated macrophages to become tumoricidal effectors as a novel strategy for cancer therapy
.
Oncotarget
2017
;
8
:
48436
52
.
3.
Banat
GA
,
Tretyn
A
,
Pullamsetti
SS
,
Wilhelm
J
,
Weigert
A
,
Olesch
C
, et al
Immune and inflammatory cell composition of human lung cancer stroma
.
PLoS One
2015
;
10
:
e0139073
.
4.
Schmall
A
,
Al-Tamari
HM
,
Herold
S
,
Kampschulte
M
,
Weigert
A
,
Wietelmann
A
, et al
Macrophage and cancer cell cross-talk via CCR2 and CX3CR1 is a fundamental mechanism driving lung cancer
.
Am J Resp Crit Care
2015
;
191
:
437
47
.
5.
Donnem
T
,
Hald
SM
,
Paulsen
EE
,
Richardsen
E
,
Al-Saad
S
,
Kilvaer
TK
, et al
Stromal CD8(+) T-cell density- a promising supplement to TNM staging in non-small cell lung cancer
.
Clin Cancer Res
2015
;
21
:
2635
43
.
6.
Yang
M
,
McKay
D
,
Pollard
JW
,
Lewis
CE
. 
Diverse functions of macrophages in different tumor microenvironments
.
Cancer Res
2018
;
78
:
5492
503
.
7.
Lavin
Y
,
Kobayashi
S
,
Leader
A
,
Amir
ED
,
Elefant
N
,
Bigenwald
C
, et al
Innate immune landscape in early lung adenocarcinoma by paired single-cell analyses
.
Cell
2017
;
169
:
750
65
.
8.
Gentles
AJ
,
Newman
AM
,
Liu
CL
,
Bratman
SV
,
Feng
W
,
Kim
D
, et al
The prognostic landscape of genes and infiltrating immune cells across human cancers
.
Nat Med
2015
;
21
:
938
45
.
9.
Fridman
WH
,
Pages
F
,
Sautes-Fridman
C
,
Galon
J
. 
The immune contexture in human tumours: impact on clinical outcome
.
Nat Rev Cancer
2012
;
12
:
298
306
.
10.
Hendry
S
,
Salgado
R
,
Gevaert
T
,
Russell
PA
,
John
T
,
Thapa
B
, et al
Assessing tumor-infiltrating lymphocytes in solid tumors: a practical review for pathologists and proposal for a standardized method from the International Immuno-Oncology Biomarkers Working Group: Part 2: TILs in melanoma, gastrointestinal tract carcinomas, non-small cell lung carcinoma and mesothelioma, endometrial and ovarian carcinomas, squamous cell carcinoma of the head and neck, genitourinary carcinomas, and primary brain tumors
.
Adv Anat Pathol
2017
;
24
:
311
35
.
11.
Gorris
MAJ
,
Halilovic
A
,
Rabold
K
,
van Duffelen
A
,
Wickramasinghe
IN
,
Verweij
D
, et al
Eight-color multiplex immunohistochemistry for simultaneous detection of multiple immune checkpoint molecules within the tumor microenvironment
.
J Immunol
2018
;
200
:
347
54
.
12.
Dobin
A
,
Davis
CA
,
Schlesinger
F
,
Drenkow
J
,
Zaleski
C
,
Jha
S
, et al
STAR: ultrafast universal RNA-seq aligner
.
Bioinformatics
2013
;
29
:
15
21
.
13.
Liao
Y
,
Smyth
GK
,
Shi
W
. 
featureCounts: an efficient general purpose program for assigning sequence reads to genomic features
.
Bioinformatics
2014
;
30
:
923
30
.
14.
Love
MI
,
Huber
W
,
Anders
S
. 
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
.
Genome Biol
2014
;
15
:
550
.
15.
Feng
ZP
,
Bethmann
D
,
Kappler
M
,
Ballesteros-Merino
C
,
Eckert
A
,
Bell
RB
, et al
Multiparametric immune profiling in HPV-oral squamous cell cancer
.
JCI Insight
2017
;
2
:
e93652
.
16.
Patriarca
C
,
Macchi
RM
,
Marschner
AK
,
Mellstedt
H
. 
Epithelial cell adhesion molecule expression (CD326) in cancer: a short review
.
Cancer Treat Rev
2012
;
38
:
68
75
.
17.
Amsalem
H
,
Kwan
M
,
Hazan
A
,
Zhang
JH
,
Jones
RL
,
Whittle
W
, et al
Identification of a novel neutrophil population: proangiogenic granulocytes in second-trimester human decidua
.
J Immunol
2014
;
193
:
3070
9
.
18.
Nakano
A
,
Harada
T
,
Morikawa
S
,
Kato
Y
. 
Expression of leukocyte common antigen (CD45) on various human leukemia/lymphoma cell lines
.
Acta Pathol Jpn
1990
;
40
:
107
15
.
19.
Lajaunias
F
,
Dayer
JM
,
Chizzolini
C
. 
Constitutive repressor activity of CD33 on human monocytes requires sialic acid recognition and phosphoinositide 3-kinase-mediated intracellular signaling
.
Eur J Immunol
2005
;
35
:
243
51
.
20.
Gowda
M
,
Payne
KK
,
Godder
K
,
Manjili
MH
. 
HLA-DR expression on myeloid cells is a potential prognostic factor in patients with high-risk neuroblastoma
.
Oncoimmunology
2013
;
2
:
e26616
.
21.
Collin
M
,
McGovern
N
,
Haniffa
M
. 
Human dendritic cell subsets
.
Immunology
2013
;
140
:
22
30
.
22.
Crittenden
MR
,
Baird
J
,
Friedman
D
,
Savage
T
,
Uhde
L
,
Alice
A
, et al
Mertk on tumor macrophages is a therapeutic target to prevent tumor recurrence following radiation therapy
.
Oncotarget
2016
;
7
:
78653
66
.
23.
Gautier
EL
,
Shay
T
,
Miller
J
,
Greter
M
,
Jakubzick
C
,
Ivanov
S
, et al
Gene-expression profiles and transcriptional regulatory pathways that underlie the identity and diversity of mouse tissue macrophages
.
Nat Immunol
2012
;
13
:
1118
28
.
24.
Giroux
M
,
Descoteaux
A
. 
Cyclooxygenase-2 expression in macrophages: modulation by protein kinase C-alpha
.
J Immunol
2000
;
165
:
3985
91
.
25.
Dai
F
,
Liu
L
,
Che
G
,
Yu
N
,
Pu
Q
,
Zhang
S
, et al
The number and microlocalization of tumor-associated immune cells are associated with patient's survival time in non-small cell lung cancer
.
BMC Cancer
2010
;
10
:
220
.
26.
Ma
J
,
Liu
L
,
Che
G
,
Yu
N
,
Dai
F
,
You
Z
. 
The M1 form of tumor-associated macrophages in non-small cell lung cancer is positively associated with survival time
.
BMC Cancer
2010
;
10
:
112
.
27.
Jackute
J
,
Zemaitis
M
,
Pranys
D
,
Sitkauskiene
B
,
Miliauskas
S
,
Vaitkiene
S
, et al
Distribution of M1 and M2 macrophages in tumor islets and stroma in relation to prognosis of non-small cell lung cancer
.
BMC Immunol
2018
;
19
:
3
.
28.
Henze
AT
,
Mazzone
M
. 
The impact of hypoxia on tumor-associated macrophages
.
J Clin Invest
2016
;
126
:
3672
9
.
29.
Rajaganeshan
R
,
Prasad
R
,
Guillou
PJ
,
Scott
N
,
Poston
G
,
Jayne
DG
. 
Expression patterns of hypoxic markers at the invasive margin of colorectal cancers and liver metastases
.
Eur J Surg Onc
2009
;
35
:
1286
94
.
30.
Zilionis
R
,
Engblom
C
,
Pfirschke
C
,
Savova
V
,
Zemmour
D
,
Saatcioglu
HD
, et al
Single-cell transcriptomics of human and mouse lung cancers reveals conserved myeloid populations across individuals and species
.
Immunity
2019
;
50
:
1317
34
.
31.
Gao
JJ
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
32.
Cancer Genome Atlas Research Network
,
Weinstein
JN
,
Collisson
EA
,
Mills
GB
,
Shaw
KR
,
Ozenberger
BA
, et al
The Cancer Genome Atlas Pan-Cancer analysis project
.
Nat Genet
2013
;
45
:
1113
20
.
33.
Wen
M
,
Ma
X
,
Cheng
H
,
Jiang
W
,
Xu
X
,
Zhang
Y
, et al
Stk38 protein kinase preferentially inhibits TLR9-activated inflammatory responses by promoting MEKK2 ubiquitination in macrophages
.
Nat Commun
2015
;
6
:
7167
.
34.
Mertens
C
,
Mora
J
,
Oren
B
,
Grein
S
,
Winslow
S
,
Scholich
K
, et al
Macrophage-derived lipocalin-2 transports iron in the tumor microenvironment
.
Oncoimmunology
2018
;
7
:
e1408751
.
35.
Ward
JM
,
Nikolov
NP
,
Tschetter
JR
,
Kopp
JB
,
Gonzalez
FJ
,
Kimura
S
, et al
Progressive glomerulonephritis and histiocytic sarcoma associated with macrophage functional defects in CYP1B1-deficient mice
.
Toxicol Pathol
2004
;
32
:
710
8
.
36.
Ohfuchi
E
,
Kato
M
,
Sasaki
M
,
Sugimoto
K
,
Oma
Y
,
Harata
M
. 
Vertebrate Arp6, a novel nuclear actin-related protein, interacts with heterochromatin protein 1
.
Eur J Cell Biol
2006
;
85
:
411
21
.
37.
Wei
CY
,
Zhu
MX
,
Lu
NH
,
Peng
R
,
Yang
X
,
Zhang
PF
, et al
Bioinformatics-based analysis reveals elevated MFSD12 as a key promoter of cell proliferation and a potential therapeutic target in melanoma
.
Oncogene
2019
;
38
:
1876
91
.
38.
Yuan
A
,
Hsiao
YJ
,
Chen
HY
,
Chen
HW
,
Ho
CC
,
Chen
YY
, et al
Opposite effects of M1 and M2 macrophage subtypes on lung cancer progression
.
Sci Rep
2015
;
5
:
14273
.
39.
Ohri
CM
,
Shikotra
A
,
Green
RH
,
Waller
DA
,
Bradding
P
. 
Macrophages within NSCLC tumour islets are predominantly of a cytotoxic M1 phenotype associated with extended survival
.
Eur Respir J
2009
;
33
:
118
26
.
40.
Li
Z
,
Maeda
D
,
Yoshida
M
,
Umakoshi
M
,
Nanjo
H
,
Shiraishi
K
, et al
The intratumoral distribution influences the prognostic impact of CD68- and CD204-positive macrophages in non-small cell lung cancer
.
Lung Cancer
2018
;
123
:
127
35
.
41.
Rakaee
M
,
Busund
LR
,
Jamaly
S
,
Paulsen
EE
,
Richardsen
E
,
Andersen
S
, et al
Prognostic value of macrophage phenotypes in resectable non-small cell lung cancer assessed by multiplex immunohistochemistry
.
Neoplasia
2019
;
21
:
282
93
.
42.
Jeong
H
,
Kim
S
,
Hong
BJ
,
Lee
CJ
,
Kim
YE
,
Bok
S
, et al
Tumor-associated macrophages enhance tumor hypoxia and aerobic glycolysis
.
Cancer Res
2019
;
79
:
795
806
.
43.
Herzig
S
,
Shaw
RJ
. 
AMPK: guardian of metabolism and mitochondrial homeostasis
.
Nat Rev Mol Cell Biol
2018
;
19
:
121
35
.