Antibody combinations for optimized staining of macrophages in human lung tumours

The analysis of tumour‐associated macrophages (TAMs) has a high potential to predict cancer recurrence and response to immunotherapy. However, the heterogeneity of TAMs poses a challenge for quantitative and qualitative measurements. Here, we critically evaluated by immunohistochemistry and flow cytometry two commonly used pan‐macrophage markers (CD14 and CD68) as well as some suggested markers for tumour‐promoting M2 macrophages (CD163, CD204, CD206 and CD209) in human non–small cell lung cancer (NSCLC). Tumour, non‐cancerous lung tissue and blood were investigated. For immunohistochemistry, CD68 was confirmed to be a useful pan‐macrophage marker although careful selection of antibody was found to be critical. The widely used anti‐CD68 antibody clone KP‐1 stains both macrophages and neutrophils, which is problematic for TAM quantification because lung tumours contain many neutrophils. For TAM counting in tumour sections, we recommend combined labelling of CD68 with a cell membrane marker such as CD14, CD163 or CD206. In flow cytometry, the commonly used combination of CD14 and HLA‐DR was found to not be optimal because some TAMs do not express CD14. Instead, combined staining of CD68 and HLA‐DR is preferable to gate all TAMs. Concerning macrophage phenotypic markers, the scavenger receptor CD163 was found to be expressed by a substantial fraction (50%‐86%) of TAMs with a large patient‐to‐patient variation. Approximately 50% of TAMs were positive for CD206. Surprisingly, there was no clear overlap between CD163 and CD206 positivity, and three distinct TAM sub‐populations were identified in NSCLC tumours: CD163+CD206+, CD163+CD206− and CD163−CD206−. This work should help develop macrophage‐based prognostic tools for cancer.


| INTRODUCTION
Tumour-associated macrophages (TAMs) are heterogeneous in phenotypes and functions and may exert either tumour-promoting or tumour-suppressive activity. [1][2][3][4] Therefore, immunoprofiling of TAMs has a great potential as a prognostic tool and as a predictor of immunotherapy efficacy for individuals with cancer. Lung cancer is the most frequent cause of cancer-related death worldwide, and non-small cell lung cancer (NSCLC) accounts for about 85% of all lung cancer cases. 5 Several immunohistochemistry (IHC)-based analyses of NSCLC tumour sections suggested that the density of TAMs may potentially be used to predict patient survival after the operation. Six studies reported that the survival of NSCLC patients was prolonged when more macrophages had infiltrated the tumour tissue. [6][7][8][9][10][11] In contrast, seven studies showed that high numbers of TAMs were correlated with shorter survival. 7,9,[12][13][14][15][16] Five other studies reported no association between TAM density and prognosis. [17][18][19][20][21] Thus, although these reports collectively support a key role of TAMs for tumour development, the available data are contradictory and it remains unclear why high numbers of TAMs were associated with better or worse prognosis in the different studies.
The heterogeneity of macrophages poses a challenge for quantitative and qualitative analyses. In fact, the use of different molecular targets and antibody clones to stain TAMs may explain some of the discrepancies in the literature concerning the role of TAMs in human cancer. Most published IHC studies in NSCLC have used CD68 as a pan-macrophage marker. CD68 is a transmembrane glycoprotein that is mainly located in the endosomal/lysosomal compartment of macrophages. 22,23 However, CD68 has been reported to be expressed by several other cell types such as neutrophils, dendritic cells, fibroblasts and endothelial cells. [24][25][26][27] Therefore, the use of CD68 can potentially lead to an overestimation of the number of macrophages, in particular in lung cancer tumours, which may be highly infiltrated by neutrophils. 28 Accordingly, the validity of CD68 as a pan-macrophage marker in lung tumours needs to be clarified.
According to a popular but controversial model proposed separately by Mills and Mantovani, macrophages with a tumour-suppressive phenotype are called M1, whereas tumour-promoting macrophages are called M2. [29][30][31] Most TAMs are considered to have an M2 or M2-like phenotype. [30][31][32] Several molecular markers have been suggested to be specific for M2 TAMs, including the haemoglobin scavenger receptor CD163, [33][34][35] the scavenger receptor-A CD204, 15,[36][37][38] macrophage mannose receptor CD206, 16 CD209 (also known as dendritic cell-specific intercellular adhesion molecule-1-3 grabbing non-integrin, DC-SIGN) 39 and TREM-1. 40 In accordance with the Mills & Mantovani model, specific quantification of M1 or M2 TAMs should provide more precise prognostic information compared to total TAM numbers. The validity of this strategy has received support from several reports indicating that high densities of M1 TAMs were associated with prolonged survival, 34,35,41 whereas high numbers of M2 TAMs were associated with shorter survival. [36][37][38][40][41][42] Unfortunately, a major limitation for clinical implementation of the approach is that there is currently no consensus concerning which markers should be used to identify tumour-suppressive M1 TAMs and tumour-promoting M2 TAMs, respectively. For example and to illustrate the problem, two studies came to the surprising conclusion that high numbers of TAMs with an M2 phenotype were associated with better prognosis. 11,34 Macrophages reside in different histological locations in lung tumours. Inside the tumour itself, some TAMs infiltrate the islets of cancer cells, but most TAMs are present in the tumour stroma which consists of normal, non-cancerous cells. 7 When the tumour contains areas with preserved lung tissue, tissue-resident alveolar macrophages may be found. Macrophages are also present in the tertiary lymphoid structures (TLS) that are frequently observed at the periphery of lung tumours. 43,44 Macrophages in the germinal centres of TLS are called tingible body macrophages and their function is to engulf apoptotic cells. 45,46 Several reports indicated that macrophages in different histological areas of a tumour, such as tumour islets versus stroma, may play opposite roles in tumour development. 7,9 Therefore, the different compartments of a tumour should be taken into consideration when analysing TAMs.
In the present study, we critically evaluated by IHC and flow cytometry two commonly used pan-macrophage markers (CD14 and CD68) as well as several suggested markers for M2 macrophages (CD163, CD204, CD206 and CD209). The data presented here should represent a useful resource for scientists aiming at staining macrophages in human tumours and normal lung tissue. positivity, and three distinct TAM sub-populations were identified in NSCLC tumours: CD163 + CD206 + , CD163 + CD206 − and CD163 − CD206 − . This work should help develop macrophage-based prognostic tools for cancer.

| Patients and lung tissue samples
Tissue samples from lung tumour and from a non-cancerous part of the lung were collected from eleven NSCLC patients following surgical resection of lung lobes at the Department of Cardiothoracic Surgery at Oslo University Hospital, Norway. The diagnosis of lung cancer was based on histopathology criteria, and TNM stage of the cancers varied from I to IIIB. Five patients were diagnosed with squamous cell carcinoma and six patients with adenocarcinoma. Non-cancerous lung tissue was sampled from a part of the lung specimen located as far away from the tumour tissue as possible. The patients were smokers or former smokers, and they had not received neoadjuvant chemotherapy, preoperative radiotherapy or immunosuppressive therapy. Written informed consent was obtained from all patients before inclusion. The study has been approved by The Norwegian Regional Ethical Committee (ref: S-05307).

| Immunohistochemistry (IHC)
Tissue samples from primary lung tumour and non-cancerous lung tissue were fixed in 10% formalin for 24 hours and embedded in paraffin before 4-µm-thick serial sections were made. Staining of the sections was performed using a Ventana Discovery Ultra automated slide stainer (Ventana Medical System, Roche, Cat. No. 750-601). After deparaffinization of the sections, heat-induced antigen retrieval was performed by using cell conditioning 1 buffer (CC1, Ventana Medical System, Cat. No. 950-500) for 32 minutes, before staining with the primary monoclonal antibodies (mAbs). Next, the tissue sections were incubated with secondary antibody conjugated with peroxidase. The primary and secondary antibodies used are listed in Tables S1 and S2. Bound antibodies were visualized using the kit ChromoMap DAB Kit (Ventana Medical System, Cat. No. 760-159).
For paired immunostaining, we established the procedure shown in Table S3. The primary and secondary antibodies used are listed in Tables S4 and S5. Briefly, the primary antibodies against CD14, CD163 or CD206 were applied on the tissue sections, before incubation with the secondary antibodies. Bound antibodies were visualized using Discovery Teal-HRP detection kit (Ventana Medical System, Cat. No. 760-247). Next, to block for potential binding of subsequently added antibodies to the already applied primary and secondary antibodies, the tissue sections were treated with cell conditioning 2 buffer (

| Image analyses of tissue sections
Images of the tissue sections stained by IHC were captured by an Olympus BX51 microscope, model BX51TF using a Colorview digital camera (Olympus); or by a Nikon Eclipse microscope, model N i-U using an Infinity 2 digital camera (Lumenera Corporation). Images of the tissue sections labelled with immunofluorescence were captured by a Nikon Eclipse model 80i microscope (Nikon) using an Axiocam 506 mono digital camera (Zeiss). For quantitative analysis, we selected five representative areas in total with both tumour cell areas and tumour stroma. Images of the representative areas were captured with an Olympus BX51 microscope at a 400× magnification. If a tissue section had areas with both low abundance and high abundance of stained cells, both areas were selected as representative areas and included in the analysis. Individual cells were identified by strong brown stain and manually counted by using AnalySIS Pro software (Olympus). The cell counting was repeated three times for each area by two scientists. All images were analysed in a blinded fashion.

| Statistical analysis
For quantitative analysis, the number of counted cells is presented as mean ± standard deviation for each marker. Groups were compared and potential differences were identified using the t test with Bonferroni-Dunn to correct for multiple comparisons. All statistical analysis was carried out using GraphPad Prism, version 6. P < .05 was considered statistically significant.

CD68, CD14 and CD163
We first performed a literature search to determine which mAb have previously been used for IHC of TAMs in NSCLC prognostic studies. We found in total 25 studies and most investigators (20 reports) had used an anti-CD68 mAb to identify macrophages ( Table 1). The anti-CD68 mAb clone KP-1 was utilized in 11 studies whereas the anti-CD68 clone PG-M1 was used in 5 studies (Table 1).
In the four remaining reports, the mAb clone name was not specified. Most studies used CD68 alone, but, in a few reports, CD68 was used in combination with another marker in order to identify M1 or M2 macrophages (Table 1). Taken together, the anti-CD68 mAb clone KP-1 has so far been the most commonly used mAb to label TAMs in prognostic NSCLC studies. Three potential pan-macrophage markers (CD68, CD163 and CD14) were evaluated for IHC labelling of lung tumour sections. We tested the anti-CD14 mAb clone EPR3653, 47 the anti-CD163 mAb clone 10D6 48 and two anti-CD68 mAbs (clones PG-M1 and KP-1). 49,50 Tumour samples were obtained from 6 patients with NSCLC, and a series of adjacent tumour sections were made in order to compare the different mAbs. In all 6 patients examined, TAMs were mainly observed in the tumour stroma and were positive for CD14, CD163 and CD68 ( Figure 1). Staining with the anti-CD14 mAb clone EPR3653, anti-CD163 clone 10D6 or anti-CD68 clone PG-M1 resulted in similar numbers of positive cells ( Figure 1A,C,E). In contrast, the anti-CD68 mAb clone KP-1 gave a high background in tumour cell areas and throughout the whole section, as well as a strong stain in neutrophil-rich areas, suggesting that KP-1 may stain both TAMs and neutrophils as previously reported 24,25 (Figure 1G,H). Notably, since CD14 and CD163 are membrane-associated molecules, the anti-CD14 and anti-CD163 mAbs labelled cell membranes, whereas the anti-CD68 mAbs resulted in granular stains in the cytoplasm (Figure 1). In summary, a high signal-to-noise ratio for immunostaining of TAMs was obtained with the anti-CD14 mAb EPR3653, the anti-CD163 mAb 10D6 and the anti-CD68 mAb PG-M1, whereas the widely used anti-CD68 mAb KP-1 gave high background and seemed to stain both TAMs and neutrophils.

| Alveolar macrophages are strongly positive for CD68 and CD163 but show variable expression of CD14
Alveolar macrophages constitute a substantial portion of macrophages in normal lungs and can also be present in lung tumours. Therefore, we examined the expression of CD14, CD68 and CD163 by alveolar macrophages using the same mAbs as against TAMs. For this purpose, we used non-cancerous lung tissue from patients with lung cancer, because there are more alveoli and thereby alveolar macrophages in non-cancerous lung tissue than in lung tumours. Noncancerous lung from 7 patients was examined. Alveolar macrophages were identified by their distinct morphology and intra-alveolar localization. Immunostaining of CD14 revealed heterogeneous signal intensity between patients (Figure 2A-C). In patient #1, nearly all alveolar macrophages were labelled by the anti-CD14 mAb (Figure 2A), whereas the signal for CD14 was minimal or absent in the 6 other patients investigated ( Figure 2C and data not shown). In all 7 patients, alveolar macrophages were labelled by the anti-CD163 mAb (C) In patient #2, the alveolar macrophages (black arrowheads) were not detected by the anti-CD14 mAb clone EPR3653, whereas (D) the anti-CD163 mAb clone 10D6 and (E) the anti-CD68 mAb clone PG-M1 as well as (F) the anti-CD68 mAb clone KP-1 labelled the alveolar macrophages (black arrowheads). All mAbs detected tumourassociated macrophages (black arrows) in the stroma (C-F). The results obtained in patient #2 are representative for 6 NSCLC patients: 4 with squamous cell carcinoma (patients #2, #3, #4 and #5) and 2 with adenocarcinoma (patients #6 and #7). Original magnification, 400×. Scale bars, 100 µm ( Figure 2D), the anti-CD68 mAb clone PG-M1 ( Figure 2E) and the anti-CD68 mAb clone KP-1 ( Figure 2F). As previously observed in tumour tissue ( Figure 1G), the anti-CD68 KP-1 mAb gave a high background in normal lung tissue ( Figure 2F). Thus, alveolar macrophages are strongly positive for CD68 and CD163, whereas CD14 expression varies from patient to patient.

| Macrophages in the TLS around lung tumours express CD68 but not CD163
Macrophage-containing TLS are commonly formed at the periphery of lung tumours. 43,44 We tested whether the different mAbs could detect macrophages in TLS from the same 6 patients examined in Figure 1. Inside TLS, a weak CD14 stain was found in macrophages present in areas resembling germinal centres (white arrowheads in Figure 3A). These cells were presumably tingible body macrophages because such macrophages are typically found in germinal centres. 45,46,51 The other macrophages in the TLS, hereafter termed TLS-associated macrophages, identified by morphology and haematoxylin and eosin (H&E) stain were not labelled by the anti-CD14 mAb ( Figure 3A and data not shown). In contrast, TAMs in the areas surrounding the TLS were labelled by the anti-CD14 mAb (black arrowheads in Figure 3A). The anti-CD163 mAb did not stain any cell in TLS, including in germinal centres, whereas the TAMs in the areas surrounding the TLS were labelled (black arrowheads in Figure 3C). When using anti-CD68 mAbs (clones PG-M1 or KP-1), the signal was strong for tingible body macrophages and for TAMs surrounding the TLS, but weaker for TLS-associated macrophages ( Figure 3E,G). Thus, TLS-associated macrophages are weakly positive for CD68 and negative for CD14 and CD163. Tingible body macrophages in the germinal centres of TLS are positive for CD68, weakly positive for CD14 and negative for CD163. All data presented so far are summarized in Table 2.

| The anti-CD68 mAb clone KP-1 labels neutrophils
Because CD68 has been reported to be expressed by neutrophils, 24,25 we investigated whether the anti-CD68 mAbs PG-M1 and KP-1 labelled neutrophils in lung cancer tissues, and performed IHC on tissue sections from the same 6 patients as above. Representative data from one patient are shown in Figure 4. Using H&E stain, we first identified tumour areas containing numerous neutrophils characterized by their multilobular nuclei ( Figure 4A,B). Adjacent lung cancer tissue sections showed that neither the anti-CD14 mAb ( Figure 4C), the anti-CD163 mAb ( Figure 4D) nor the anti-CD68 mAb clone PG-M1 ( Figure 4E) stained neutrophils. In contrast, use of the anti-CD68 mAb clone KP-1 resulted in a relatively strong stain of neutrophils ( Figure 4F). Thus, the anti-CD68 mAb clone KP-1 does not only label macrophages (as shown above) but also neutrophils. In contrast, the anti-CD68 mAb clone PG-M1 appears to be more macrophage-specific since it does not stain neutrophils. Therefore, we made use of clone PG-M1 to stain CD68 in subsequent analyses of tumour tissue sections.

| TAMs and alveolar macrophages in lung tumours co-express CD68 and CD163
Single-stained tumour tissue sections shown in Figure 1 suggested that TAMs may be positive for both CD68 and CD163. To investigate this further, we used fluorescence microscopy and performed double staining of CD68 (using clone PG-M1) and CD163 of lung tumour tissue from the same 6 patients. Representative data from one patient are shown in Figure 5. Numerous CD68 + CD163 + double-positive TAMs were observed ( Figure 5A). The signal intensity for CD68 and for CD163 varied within each tissue section. Several TAMs showed a strong signal for CD68 and a weak signal for CD163 (white arrowhead in Figure 5B), while other TAMs exhibited the opposite pattern (white arrow in Figure 5B). Moreover, several CD68 + TAMs without CD163 expression were detected in tumour stroma (white arrows in Figure 5C). The proportion of CD163 positive cells varied among patients. In the tumour tissue from the 6 patients, we found, by manual counting of 8 microscopy fields (magnification x400) per section, that 5%-30% of the CD68 + TAMs were negative for CD163, whereas only 1%-5% of the CD163 + TAMs were negative for CD68 (data not shown). All alveolar macrophages were strongly labelled by both anti-CD68 and anti-CD163 mAbs ( Figure S1A). In accordance with the data obtained by single stain (Figure 3), TLS-associated macrophages were positive for CD68 but negative for CD163 (white arrows in Figure S1B). CD68 + CD163 + double-positive TAMs were found surrounding the TLS (white arrowheads in Figure S1B). Isotype-matched control mAbs showed no signal in the examined tissue areas ( Figure S2A-C). Taken together, these results indicate that the anti-CD68 mAb clone PG-M1 detects most TAMs and label a higher number of TAMs than the anti-CD163 mAb. Alveolar macrophages co-express CD68 and CD163, while TLSassociated macrophages only express CD68.

| Combined staining of CD68 with a plasma membrane marker improves the identification of macrophages
The results presented above provide support for CD68 as the most reliable pan-macrophage marker in lung tumours, because CD68 was shown to be expressed by most TAMs, alveolar macrophages, TLS-associated macrophages and tingible body macrophages (Table 2). In contrast, these four macrophage subpopulations were not all consistently detected by anti-CD14 or anti-CD163 mAbs (Table 2). Unfortunately, CD68 labelling poses a challenge for cell counting because CD68 molecules are present intracellularly and appear as cytoplasmic granules or dots in the cells, a pattern that may be difficult to distinguish from non-specific background stain ( Figure 1E). In contrast, cells stained with mAb specific for a molecule present on the plasma membrane, such as CD14 and CD163, are easier to count because of the distinct labelling of the cell contours ( Figure 1A,C). Therefore, we hypothesized that combining anti-CD68 mAb with a mAb specific for a macrophage-associated plasma membrane marker (such as CD14, CD163 or CD206) may improve the detection and quantification of macrophages in tissue sections. Indeed, the identification of macrophages was greatly enhanced by such a double stain, both for alveolar macrophages and TAMs ( Figure 6). To evaluate this double stain approach for cell counting, we quantified the number of CD68 + macrophages ( Figure 7A,D), CD163 + macrophages ( Figure 7B,E) and double-stained CD68 + CD163 + macrophages ( Figure 7C,F) in adjacent sections from lung tumour from 6 patients. Five representative areas were selected and counted manually for each section. The quantification revealed that the mixture of the two mAbs resulted in a higher macrophage number than the single stain approach with anti-CD68 or anti-CD163 mAb in many of the selected areas in the tissue sections from all 6 patients ( Figure 7G and data not shown). As CD68 is the most commonly used marker for detection of macrophages, the mean of the number of CD68 + cells was set to 100% for all patients ( Figure 7H). The average number of macrophages detected by the mAb mixture was significantly increased to 130%-190% for patients #3 and #4 ( Figure 7H). For patients #2, #5 and #7, staining with the mAb mixture resulted in a tendency towards more macrophages. Notably, the data also showed that higher numbers of macrophages typically were counted in sections stained with anti-CD163 mAb alone compared to anti-CD68 alone ( Figure 7G,H). This may seem surprising since the immunofluorescence data presented in Figure 5 indicated that most macrophages in lung tumours expressed CD68 whereas only a sub-population (70%-95%) expressed CD163. A likely explanation for this apparent paradox is that more macrophages stained with anti-CD68 mAb are missed during manual quantification, as compared to anti-CD163 stain, illustrating the challenge of using an intracellular granule stain such as CD68 to quantify cells. In summary, the data indicate that macrophages are easier to identify and count by combining the cytoplasmic CD68 stain with a membranous surface stain.

| CD163 expression is essentially restricted to macrophages
Molecular markers for macrophages were further investigated using flow cytometry. First, we examined CD163 expression in tumour tissue from 4 patients (Figure 8). Tumour tissue from the same patients was previously evaluated by immunofluorescence ( Figure 5). In this initial experiment, we used the markers CD14 and HLA-DR, which represent a commonly used gating strategy in flow cytometry for identification of macrophages as double-positive CD14 + HLA-DR + cells. 28,52,53 The nucleated cells were first gated to remove debris from all recorded events ( Figure 8A). Next, the cells were gated using a single-cell gate to exclude doublets and cell clumps ( Figure 8B). Live leucocytes were then gated by use of propidium iodide (PI) and CD45 ( Figure 8C), whereas CD19 + B cells and CD3 + T cells were excluded ( Figure 8D). From the live, non-T cell, non-B cell leucocyte gate ( Figure 8D), cells were gated into CD163positive and CD163-negative cells ( Figure 8E,H,K,N).
Among the CD163 − cells, there was a small population of CD14 + HLA-DR + macrophages, confirming that some macrophages do not express CD163 ( Figure 8F,I,L,O). In contrast, essentially all CD163 + cells were shown to express both CD14 and HLA-DR, revealing that CD163 expression is essentially restricted to macrophages ( Figure 8G,J,M,P). Thus, CD163 is a macrophage-specific marker, but it is not expressed by all macrophages.

| Flow cytometry confirms that CD163 is expressed by most but not all TAMs
The data presented in Figure 8 were further analysed to evaluate in more detail the association between CD163, CD14 and HLA-DR expression. After exclusion of CD3 + T cells and CD19 + B cells ( Figure 9A-D) the live leucocytes were separated into three populations based on HLA-DR and CD14 expression ( Figure 9E,I,M,Q). The expression of CD163 was investigated in all three populations, that is the HLA-DR + CD14 − population ( Figure 9F,J,N,R), the HLA-DR + CD14 + bona fide macrophage population ( Figure 9G,K,O,S) and the HLA-DR -CD14 − population ( Figure 9H,L,P,T). A clear difference in CD163 expression was observed, as a substantial proportion of HLA-DR + CD14 + cells were positive for CD163 whereas the two CD14 − populations were negative or contained very few CD163 + cells, confirming that CD163 expression is essentially restricted to macrophages. The analysis also verified our observation made by immunofluorescence that not all macrophages express CD163 since the proportion of CD163 + cells among CD14 + HLA-DR + macrophages varied from 48.2% to 84.9% (mean: 64%) for the 4 patients investigated ( Figure 9G,K,O,S and Figure S4). Notably, the CD163 stain did not separate the macrophages into two distinct populations ( Figure 9G,K,O,S). Instead, the intensity for CD163 varied from low to relatively high, which is consistent with the results obtained by immunofluorescence ( Figure 5). We also considered the possibility that CD163 may be expressed intracellularly by all TAMs. To test this hypothesis, we performed F I G U R E 7 Improved quantification of TAMs using a mixture of mAb against CD68 and CD163. Serial sections of tumour tissue from 6 patients with NSCLC were stained with (A) anti-CD68 mAb clone PG-M1, (B) anti-CD163 mAb clone 10D6 or (C) a mixture of anti-CD68 mAb clone PG-M1 and anti-CD163 mAb clone 10D6. (D-F) For each patient and each staining, images from 5 representative areas were selected for manual quantification. Cells with strong brown labelling were counted (yellow contours). (G) Manual counting of stained cells in each representative area was repeated 6 times (3 times by each investigator). Quantitative data obtained in tissue from 3 patients are presented as mean of the number of labelled cells ± SD for CD68 stain (white columns), CD163 stain (grey) and combined CD68 and CD163 stain (black) for each representative area. (H) The columns show average numbers in relative percentage based on counting of 5 representative areas in each section ± SD of infiltrating CD68 + cells (white columns), CD163 + cells (grey) and cells double-positive for CD68 and CD163 (black). The number of CD68 + cells was set to 100%. Sections from 6 NSCLC patients were examined: 4 patients with lung squamous cell carcinoma (#2, #3, #4 and #5) and 2 patients with adenocarcinoma (#6 and #7). * P < .05. Original magnification, 400×. Scale bars, 50 µm intracellular flow cytometry staining in fixed and permeabilized single cells from a lung tumour sample from 1 patient and found that 68% of the HLA-DR + CD14 + TAMs were positive for CD163 ( Figure S5). Thus, there is a significant fraction of TAMs, approximately one-third, that do not express CD163, neither on the cell surface nor intracellularly.

| Flow cytometry shows that CD68 expression is not restricted to macrophages
Immunostaining of tissue sections ( Figure 5) and previous reports suggested that CD68 may be expressed by other cells than macrophages. 24,54 To clarify this using flow cytometry, we examined tumour tissue from 3 NSCLC patients with the anti-CD68 mAb clone Y1/82A (Figure 10). Because CD68 is predominantly located in intracellular vesicles, fixation and permeabilization were included in the staining protocol. CD45 + leucocytes were first gated ( Figure 10A-C) and CD68 + cells were identified in the CD45 + CD3 − CD19 − population ( Figure 10D-E). Next, the expression of CD14 and HLA-DR among the CD68 + cells was assessed ( Figure 10G). This analysis revealed that the majority (58.8%) of CD68 + cells expressed both HLA-DR and CD14 (indicated in yellow in Figure 10G). However, a substantial fraction (25.8%) of the CD68 + cells was HLA-DR + and CD14 − (indicated in green in Figure 10G). In addition to the two HLA-DR + populations, a population of HLA-DR -CD14 − cells (indicated in red in Figure 10G) was also observed among  In addition to macrophages, the CD68 + HLA-DR + CD14 − population may potentially contain some CD11c + myeloid dendritic cells ( Figure 10F). Staining of CD11c revealed that 14.9% of CD11c + HLA-DR + CD14 − cells, presumably representing dendritic cells, expressed CD68 ( Figure S6). However, it should be noted that CD11c is not sufficient by itself to identify dendritic cells because most lung macrophages are also positive for CD11c. 28,54,55 Notably, the CD68 + HLA-DR + CD14 − CD11c − population contained smaller and less granulated cells ( Figure 10I) compared to the CD68 + HLA-DR + CD14 + population, which showed size and granularity typical to that of macrophages ( Figure 10H). We also analysed the small population of cells (14.8%) that lacked both CD14 and HLA-DR (indicated in red in Figure 10G). Because immunostaining revealed that one of the evaluated mAbs towards CD68 detected neutrophils (Figure 4), we included a mAb against CD11b to be able to identify CD11b + granulocytes and natural killer cells in the CD68 + CD14 − HLA-DRpopulation. However, only 0.4% of the CD68 + CD14 − HLA-DR − cells could be gated by this strategy ( Figure 10J). The remaining fraction of CD11bcells contained both large cells with a high granularity (SSC-H) and smaller and less granulated cells ( Figure 10K). Further analysis showed that no CD68 expression could be found on CD56 + natural killer cells ( Figure S7), and virtually all neutrophils (99.3%) and most eosinophils (91.7%) were found to be negative for CD68 ( Figure S8). In contrast, all monocytes in blood were shown to express CD68 ( Figure S9). Taken together, the data strongly suggest that CD68 expression in lung tumour is not strictly restricted to macrophages, in particular because CD14 − HLA-DR − cells and probably some dendritic cells also express CD68.

| CD68 is a sensitive marker for detection of most macrophages
In the next analysis, we wanted to examine whether CD68 is a suitable marker for detection of all macrophages in lung tumour tissue. CD68 expression was evaluated in tumour tissue from the same patients as presented in Figure 10, and three cell populations were examined: HLA-DR + CD14 + , HLA-DR + CD14 − and HLA-DR -CD14 − (Figure 11). Before analysis, CD3 + T cells and CD19 + B cells were excluded from the CD45 + leucocyte population ( Figure 11A-D). Essentially all CD14 + HLA-DR + macrophages expressed CD68 ( Figure 11E, F), indicating that CD68 can be used to detect all bona fide macrophages in tumour. The HLA-DR + CD14 − population contained 31.8% CD68 + cells, presumably representing macrophages and dendritic cells that did not express CD14 ( Figure 11H). Finally, some CD68 + cells (11.1%) were found in the CD14 − HLA-DR − population ( Figure 11G), suggesting that CD68 expression is not strictly restricted to macrophages, as mentioned in the previous section. Thus, these data are consistent with CD68 being expressed by virtually all macrophages in lung tumour tissue.

| A better gating strategy for TAMs in flow cytometry based on CD68 and HLA-DR
The results from the flow cytometry evaluation of CD14, CD163 and CD68 indicated that all three markers had limitations as pan-macrophage markers. CD163 was found to be a rather specific marker for the monocyte/macrophage lineage but CD163 was not detected on all macrophages in lung tumours. CD68 was not strictly specific for the macrophage/ monocyte lineage since it was also expressed by HLA-DRcells. However, CD68 was detected on a larger fraction of macrophages in tumour, as compared to CD163 and CD14. Based on these observations, we hypothesized that CD68 in combination with HLA-DR would specifically detect most macrophages, including macrophages that are negative for CD14. Therefore, such a strategy should identify more macrophages than the commonly used definition of macrophages in flow cytometry being double-positive CD14 + HLA-DR + cells. To compare the two gating strategies, flow cytometry was performed on adenocarcinoma tumours from 3 patients ( Figure 12). For this purpose, the first three gates were set to exclude debris, cell clumps and CD45 − non-leucocytes ( Figure 12A-C). Next, CD3 + T cells and CD19 + B cells were excluded from the CD45 + leucocytes ( Figure 12D). From this population, macrophages were gated by the current definition (HLA-DR + CD14 + ) ( Figure 12E,G,I) or our suggested new strategy (HLA-DR + CD68 + ) ( Figure 12F,H,J). The percentage of macrophages was calculated from the total population of leucocytes (single, CD45 + leucocytes). The HLA-DR + CD68 + based strategy identified 25%-50% more macrophages than the current definition of macrophages ( Figure 12). Thus, we conclude that the combination of HLA-DR and CD68 is a better strategy to gate most macrophages in tumours as compared to HLA-DR and CD14.

| Evaluation of commonly used markers for the M2 macrophage phenotype
M2-associated markers including CD163 have been used for the purpose of identifying tumour-promoting macrophages in lung tumours in previous studies. Because we observed that CD163 was expressed by a subset of TAMs, we wanted to compare the CD163 expression to that of other suggested M2 markers (CD204, CD206 and CD209) using the classical gating strategy (CD14 + HLA-DR + ) for macrophages. Lung cancer tissue from 3 patients was included in the analysis ( Figure 13). Live leucocytes were gated and CD19 + B cells and CD3 + T cells were excluded from the population ( Figure 13A-D). The expression levels of the M2-associated markers CD163 ( Figure 13F), CD204 ( Figure 13H), CD206 ( Figure 13J) and CD209 ( Figure 13L) differed within the macrophage population of each patient. Both CD206 and CD163 were expressed by approximately 60% and 85% of CD14 + HLA-DR + macrophages, respectively, whereas no expression of CD204 was found among the same population. Labelling of CD206 separated the macrophages into a CD206 + and a CD206 − population ( Figure 13J). A small fraction of the HLA-DR + CD14 + cells (8.8%) expressed low levels of CD209, but these CD209 + macrophages did not make a distinct population ( Figure 13L). Taken together, only the two M2-associated markers CD163 and CD206 were detected on a significant fraction of macrophages in lung tumours.

| CD163 and CD206 expression identifies three macrophage populations in lung tumours
The observation that the two M2 markers CD163 and CD206 were expressed by approximately 60%-85% of macrophages in tumour ( Figure 13) suggested that the two molecules may be co-expressed by the same population of TAMs. To test this hypothesis, CD3 + T cells and CD19 + B cells were excluded from CD45 + leucocytes in lung tumour tissue from 3 patients ( Figure 14A-D). After gating macrophages as HLA-DR + CD14 + cells ( Figure 14E,I,M), expression of CD163 and CD206 revealed three populations of TAMs: CD163 + CD206 + , CD163 + CD206 − and CD163 − CD206 − cells. The proportion of macrophages being double-positive for CD163 and CD206 varied from 16.2% to 58.8% ( Figure 14F,J,N). Single-positive CD163 + cells constituted 21.3%-34.6% of the macrophages, whereas macrophages lacking both CD206 and CD163 were observed in 9.1%-44.9% of the TAMs ( Figure 14F,J,N). Finally, we wanted to examine the expression levels of CD163 and CD206 on blood monocytes as most macrophages in tumour likely originate from those cells. The blood samples were from the same patients as the tumour tissue presented in Figure 14. More than 85% of the CD14 + monocytes were found to express CD163 ( Figure S10F,I,L), in accordance with previous reports. 56,57 In contrast, monocytes were negative for CD206 ( Figure S10G,J,M). Thus, three TAM sub-populations were identified in NSCLC tumour (CD163 + CD206 + , CD163 + CD206 − and CD163 − CD206 − ), whereas blood monocytes from the same patients were mostly CD163 + CD206 − . isotype-matched isotype control mAb for CD209. The percentage of HLA-DR + CD14 + macrophages was calculated from the total number of live leucocytes (CD45 + PI -). All other percentages (blue) were calculated from the population of HLA-DR + CD14 + macrophages. All per cent values are means of data obtained on lung squamous cell carcinoma and lung adenocarcinoma from 3 patients (patients #4, #5 and #8)

| DISCUSSION
In this study, we critically evaluated antibodies that may be used for quantitative and qualitative analysis of macrophages in human lung tumours by IHC and flow cytometry. CD68 was confirmed to represent a pan-macrophage marker that is expressed by all four macrophage populations that are present in lung tumours, namely TAMs, alveolar macrophages, TLSassociated macrophages and tingible body macrophages. In contrast, two other potential pan-macrophage markers evaluated, CD14 and CD163, were found not to be optimal because they were not expressed by all macrophages. Although CD68 was confirmed to represent a pan-macrophage marker in human lung tumours, several important limitations of this marker were observed. First, careful selection of antibody was found to be critical. Several anti-CD68 mAbs including clones KP-1 and PG-M1 which recognize different epitopes of the CD68 antigen are available. 58 The widely used anti-CD68 clone KP-1 appears to stain both macrophages and neutrophils. In contrast, the anti-CD68 clone PG-M1 was found to stain TAMs but not neutrophils. Labelling of neutrophils and other granulocytes by KP-1 has been previously reported by several laboratories. 24,25,49,59,60 In fact, it is well established that CD68 is expressed by both macrophages and neutrophils. 25,58 It is unclear why the anti-CD68 mAb clone KP-1 stains neutrophils, whereas the anti-CD68 clone PG-M1 does not, 50 but it could be due to differences in CD68 glycosylation in the two cell types as previously suggested. 24,58 For IHC of TAMs in lung tumours, we recommend the use of anti-CD68 clone PG-M1 rather than KP-1 because lung tumours may contain many neutrophils. 28 The difference in specificity between clone KP-1 and PG-M1 may in part explain the inconsistencies of the results from studies concerning the prognostic significance of TAMs in NSCLC.
Another limitation of CD68 as a marker to identify macrophages on tissue sections is that CD68 molecules are present intracellularly and appear as cytoplasmic granules or dots, a pattern that may be difficult to distinguish from non-specific background stain. Our data showed that combining anti-CD68 stain with another mAb specific for a macrophage-associated plasma membrane marker (such as CD14, CD163 or CD206) could greatly improve the identification of macrophages on tumour sections. In particular, analysis of tumour sections stained with a mixture of anti-CD163 and anti-CD68 mAbs resulted in a significant increase in the number of quantified macrophages compared to staining with each mAb alone. However, it should be kept in mind that none of the macrophage-associated plasma membrane marker tested (CD14, CD163 and CD206) was expressed by all TAMs. Therefore, although double staining of, for example, CD68 and CD163 makes it easier to count TAMs, it is likely that some TAMs (eg CD163 − TAMs) will be missed by this method. Thus, identification of all macrophages in lung tumours may require additional mAbs or different antibody combinations.
A commonly used strategy for identification of macrophages by flow cytometry is based on double expression of HLA-DR and CD14, but CD68 has also been used. 61,62 Here, we compared several gating strategies and found that all HLA-DR + CD14 + cells were positive for CD68 whereas only 64% were positive for CD163. CD163 was therefore excluded as a potential pan-macrophage marker in flow cytometry. Importantly, our data suggest that a significant fraction of macrophages in human lung tumours do not express CD14. Among all CD68 + , non-T cell/non-B cell leucocytes, we found that approximately 30% of the HLA-DR + cells did not express CD14. Our IHC data also revealed that TLS-associated macrophages and alveolar macrophages in most patients were negative for CD14. Macrophages with low or no CD14 expression have been previously reported in human non-diseased lung tissue. 55,63 Interestingly, the cytokine IL-4 has been shown to downregulate CD14 expression on macrophages in vitro. 64,65 It should also be noted that there exists a soluble form of CD14 and that proteases mediate the shedding of CD14 from the macrophage cell surface. 66 Because many macrophages in tumour lack CD14, we suggest an improved strategy for identification of macrophages in flow cytometry by use of CD68 and HLA-DR, with the purpose of including CD14-negative macrophages.
Each method has strengths and limitations and this certainly applies to IHC and flow cytometry. In order to label most TAMs, we established two different, yet related protocols adapted for each method. Both strategies take advantage of CD68 as decent, although not perfect, pan-macrophage marker. In IHC, CD68 labelling poses a challenge for cell counting because CD68 molecules are present intracellularly and appear as cytoplasmic granules or dots in the cells. We could show that combining anti-CD68 stain with another mAb specific for a macrophage-associated plasma membrane marker such as CD163 could greatly improve the identification of macrophages on tumour sections. In flow cytometry, the intracellular localization of CD68 does not pose a problem for cell counting, but cell fixation and permeabilization are required before CD68 staining. Because CD68 is not strictly macrophage-specific, we suggest an improved strategy for identification of macrophages in flow cytometry as CD68 + HLA-DR + double-positive cells.
CD163, 34,35 CD204, 15,36 CD206 16 and CD209 67 have previously been suggested as markers for human M2 macrophages. We failed to detect significant CD204 and CD209 expression on macrophages in lung tumours from 3 patients investigated, whereas previous IHC studies found both CD209 + cells and CD204 + cells in NSCLC tumours. [36][37][38][39] This discrepancy may be explained by patient-to-patient variation, by use of different mAb clones or | 21 of 24 FRAFJORD et Al. different methods. In contrast, we detected both CD163 and CD206 on subsets of macrophages. Notably, there was no perfect correlation between CD163 and CD206 expression on TAMs. Instead, the markers divided HLA-DR + CD14 + lung tumour macrophages into three sub-populations: CD163 + CD206 + , CD163 + CD206 − and CD163 − CD206 − TAMs. If CD163 and CD206 were indeed markers for M2 macrophages, and this remains to be clarified, it would  N) The expression of CD163 and CD206 on HLA-DR + CD14 + macrophages and the association between these markers was further assessed. Isotype controls were used to set the threshold for specific and positive staining of the target molecules. (G)(K)(O) The CD206 mAb was replaced with the corresponding fluorochrome-labelled isotype-matched mAb. (H)(L)(P) The CD163 mAb was replaced with the corresponding fluorochrome-labelled isotype-matched mAb. Tissue from 3 patients was analysed, and the first four gates are representative for all 3 patients. The case number and histological subtype for each patient are indicated. The percentages of HLA-DR + CD14 + macrophages were calculated from the total number of live leucocytes (CD45 + PI -). All other percentages (blue) were calculated from the population of macrophages (HLA-DR + CD14 − ). SCC, squamous cell carcinoma imply that there are at least two distinct M2 macrophage sub-populations in human lung tumours: CD163 + CD206 + TAMs and CD163 + CD206 − TAMs. Moreover, CD163 was detected on essentially all circulating blood monocytes as previously reported, 56,57 but the monocytes were negative for CD206. Most TAMs are likely to derive from circulating monocytes. Therefore, the CD163 + CD206 − population, which constituted 21%-34% of the HLA-DR + CD14 + cells, may in fact represent monocytes that have recently infiltrated the tumour.
For the flow cytometry analysis, we included some additional macrophage markers that were not used in IHC, such as HLA-DR, CD204 and CD209, because it is much easier to combine antibodies for multistaining in flow cytometry than in IHC. However, these markers, in particular HLA-DR and CD204, are certainly of interest for staining TAMs on tissue sections. It should be noted that IHC-based labelling of HLA-DR on macrophages can be challenging because i) HLA-DR levels vary on the surface of TAMs and ii) other tumour-infiltrating immune cell types such as B cells also express HLA-DR. 68 Although it is well accepted that macrophages may either suppress or promote tumour development depending on their activation phenotype, most investigators currently consider that the M1-M2 macrophage nomenclature represents an oversimplification. For example, a recent study based on single-cell RNA sequencing defined as many as 14 distinct transcriptional states of monocytes and macrophages in human lung tumours. 69 Macrophages are highly responsive to their surroundings. Therefore, it is likely that the phenotype of TAMs is influenced by the molecular composition of the tumour microenvironment. In vitro studies have shown that IL-4 may induce CD206 expression by macrophages. 64,65 Therefore, it is possible that IL-4 in the tumour induces CD206 on CD163 + monocytes, thereby generating CD163 + CD206 + TAMs. Furthermore, the presence of CD163 on blood monocytes, but its absence on many TAMs, suggests that CD163 expression is lost by some monocyte-derived TAMs. Anti-inflammatory signals such as glucocorticoids and IL-10 have been reported to induce the surface expression of CD163 on monocytes and macrophages, 56,57,64,65,70 whereas inflammatory mediators such as IFN-γ, lipopolysaccharide and tumour necrosis factor-α were shown to suppress CD163 expression. 70 Therefore, IFN-γproducing immune cells in the tumour, such as natural killer cells and Th1 cells, may potentially suppress the expression of CD163 on monocyte-derived macrophages. Accordingly, the CD163 − CD206 − sub-population, which constituted 9%-45% of HLA-DR + CD14 + cells, may potentially contain tumour-suppressive 'M1' macrophages. 2