Human Renal Cell Carcinoma Induces a Dendritic Cell Subset That Uses T-Cell Crosstalk for Tumor-Permissive Milieu Alterations

Donors and Tissues

RCC tissues were clear cell RCCs from untreated patients (Table 1). Nontumor kidney cortices (NKCs) were from tumor-free areas of tumor-bearing kidneys. Tumor peripheral areas were selected macroscopically and covered the tumor “pseudocapsule” and the adjacent regions of nontumor and malignant tissue (see Supplemental Figure S1A at http://ajp.amjpathol.org). Tissues were snap frozen after nephrectomy and stored at −86°C. Peripheral blood mononuclear cells were from healthy individuals. Sample collection was performed after informed consent and approved by the ethics committee.

Antibodies

Primary and secondary antibodies and their application are listed in Table 2.

IHC

Cryosections (5-μm thick) were stained using the alkaline phosphatase antialkaline phosphatase (APAAP) method²¹ or dual-labeling immunohistochemistry (IHC) (see Supplemental Figure S1, B and C, at http://ajp.amjpathol.org). Cryosections were incubated with primary antibodies [rabbit-anti-human CD3 and mouse-anti-human CD208/DC-Lamp, diluted in Tris-buffered saline with human serum (HS)] followed by secondary antibodies alkaline phosphatase-conjugated anti-rabbit immunoglobulin (Immunoresearch, West Grove, PA) and peroxidase-conjugated anti-mouse immunoglobulin (Immunoresearch). Detection was done with substrate Fast Blue followed by substrate aminoethylcarbazol (both from Sigma-Aldrich, Taufkirchen, Germany). Stained tissue sections were mounted using Immunomount (Vector Laboratories, Burlingame, CA). NK cell quantification in tissue was not possible by single-marker histologic analysis for the following reasons. The commonly used NK cell marker CD56 did not work reproducibly in cryosections. Moreover, CD56 is also expressed by a subset of cytotoxic T lymphocytes (CTLs)²² and thus gives cell counts not necessarily related to the NK cell number in a tissue.²³ NKp46 is a marker that is exclusively expressed by NK cells; however, it is subjected to regulation by the tumor milieu.²⁴ In particular, we observed down-regulation on a large proportion of NK cells of tumor-infiltrating lymphocytes (TILs) from RCC tissues using flow cytometry (P.P., unpublished observation). Therefore, the NK cell content in a tissue was determined as the percentage of NK cells (defined as CD3⁻CD56⁺) within TILs using flow cytometry.

Multiparameter Immunofluorescence Histology, Confocal Microscopy, Subset Evaluation, and Contact Quantification

Cryosections 5-μm thick were fixed in ice-cold 100% acetone and blocked with 2% bovine serum albumin (BSA) in PBS before incubation with primary antibody combinations, followed by corresponding combinations of secondary isotype- or species-specific fluorescent-labeled antibodies. Secondary reagents showed no cross-reactivity. All antibodies were diluted in 12.5% HS (Cambrex, East Rutherford, NJ) in PBS and incubated at room temperature. The following antibody combinations were used: primary mouse-anti-human antibodies CD209/DC-SIGN, CD14, and CD163 followed by secondary antibodies anti-IgG2b-A488, anti-IgG2a-A568, and anti-IgG1-A647; primary mouse-anti-human antibodies CD209/DC-SIGN, CD14, and polyclonal rabbit-anti-CD3ε followed by secondary antibodies anti-IgG2b-A488, anti-IgG2a-A568, and anti-rabbit-Cy5; primary mouse-anti-human antibodies CD209/DC-SIGN, CD163, and polyclonal rabbit-anti-CD3ε followed by secondary antibodies anti-IgG2b-A488, anti-IgG1-A568, and anti-rabbit-Cy5; primary mouse-anti-human antibodies CD209/DC-SIGN, CD14, and FOXP3 followed by secondary antibodies anti-IgG2b-A488, anti-IgG2a-A647, and anti-IgG1-A568; primary mouse-anti-human antibodies CD209/DC-SIGN and CD3ζ, and rabbit-anti-CD3ε followed by secondary antibodies anti-IgG2b-A488, anti-IgG1-A568, and anti-rabbit-Cy5. Detailed antibody information is in Table 2. After fixation [4% paraformaldehyde (PFA)] and nuclear staining with DAPI (Sigma-Aldrich), slides were mounted with Vectashield (Vector Laboratories). Fluorescence images were captured with a laser scanning microscope TCS SP2 (Leica Microsystems, Wetzlar, Germany) using HCX PL APO 63 × 1.40 oil immersion objective lens, pinhole 1.0 Airy units, 512 × 512 pixel image format, and four-frame averaging at a magnification of ×630. Sequential recording was applied to avoid fluorescence spillover, and Z-stacks were scanned to detect cells and intercellular contacts across different planes of the visual field. Image editing of contrast and brightness was applied to the whole image using Leica LCS Lite software.

CD209⁺ Subset Quantification

CD209, CD14, and/or CD163 multiparameter-stained tissue sections were used. CD209⁺ cells were identified in three different areas of RCC-inflicted kidneys: NKCs (n = 8) and two different tumor regions, the tumor center (n = 11) and the tumor periphery (n = 6). Tissues representing the three tissue areas were selected macroscopically. NKC tissues were derived from regions with the longest possible distance away from any tumor region. Histologic sections were microscopically free of malignant cells. Histologic sections of the tumor periphery were cross-sectional cuts that microscopically encompassed nontumor kidney, the “pseudocapsule” that surrounds the tumor, separating it from the nontumor kidney area, and the tumor region (see Supplemental Figure S1A at http://ajp.amjpathol.org). Myeloid cells located in the tumor parenchyma next to the capsule (ie, the tumor periphery) were evaluated. Tumor center tissue regions were macroscopically selected to be clearly away from the “pseudocapsule.” The histologic sections showed no areas of nontumor kidney or the “pseudocapsule.”

Of each tissue area, at least 10 nonoverlapping fields (×630 magnification) containing CD209⁺ cells were evaluated. The frequency of CD209⁺ cells with co-expression of CD14 and/or CD163 was quantified by assigning CD14 and CD163 expression to each CD209⁺ cell.

Quantification of Contacts between T Cells and CD209⁺ Cell Subsets

Intercellular contacts between CD209⁺ cells and CD3⁺ or FOXP3⁺ cells were determined using multiparameter immunofluorescence-stained RCC or NKC sections. At least 10 nonoverlapping fields (×630) containing CD209⁺ cells were evaluated. The frequency of contacts between CD209⁺CD14⁺ or CD209⁺CD163⁺ cells and CD3⁺ lymphocytes was assessed by selecting CD209⁺ cells and assigning co-expression of CD14 or CD163 and contacts with CD3⁺ lymphocytes. Intercellular contacts between FOXP3⁺ and CD209⁺CD14⁺ cells were quantified by selecting FOXP3⁺ fields.

CD3ε and CD3ζ Expression on T Lymphocytes in Relation to CD209⁺ Cells

CD3ζ, CD3ε, and CD209 multiparameter immunofluorescence-stained RCC sections were evaluated. Images containing CD209⁺ and CD3⁺ cells were captured at ×630 magnification. Fluorescence intensity of CD3ζ and CD3ε of T cells in close contact to CD209⁺ cells and those away from CD209⁺ cells was compared.

Cells and Cell-Conditioned Media

Cell-conditioned media were generated from cell lines^25,26 (Table 3) by culturing 2 × 10⁶ cells in 10 mL of serum-free AIM-V (Gibco/Invitrogen, Carlsbad, CA) for 10 days. Supernatants were centrifuged to remove cells. Cytotoxic T-effector cell clones CTL-JB4 (HLA-A2 alloreactive) and CTL-A42 (HLA-A2 restricted, MELAN/MART-1 specific) (Table 3) were cultured as described²⁷ and used at a resting state, generally between days 8 and 10 after the last restimulation, when they no longer secreted cytokines spontaneously. Both are T_H1-polarized effector cells as evidenced by high expression of the chemokine receptor CXCR3 and CCR5.

Isolation of TILs

The TILs were isolated from fresh tissues immediately after surgical resection. Briefly, tissues were mechanically minced into small pieces and washed with HBSS buffer to remove contaminating blood lymphocytes. Intratumoral leukocytes were recovered from the tissue after two enzymatic digestions using collagenase IA (0.5 mg/mL) and DNase I type IV (0.19 mg/mL) (all from Sigma-Aldrich) with an intermittent step of 5 mmol/L EDTA in HBSS (without Ca²⁺ and Mg²⁺). All incubations were performed for 30 minutes at room temperature.

In Vitro Generation of Myeloid Cell Subtypes

Monocytes were isolated from peripheral blood mononuclear cells using CD14⁺ microbeads (Miltenyi) and cultivated serum free (5 × 10⁶/4 mL of AIM-V) with IL-4 (400 U/mL; CellGenix, Freiburg, Germany) and granulocyte-macrophage colony-stimulating factor (GM-CSF/Leukine; 800 U/mL; Genzyme, Cambridge, MA) to generate CD209 single-positive conventional DCs (cDCs). For tissue-conditioned cells, monocytes were cultivated with 20% cell-conditioned media or with CXCL8/IL-8 (7 ng/mL; PeproTech, Rocky Hill, NJ), IL-6 (1.9 ng/mL), and vascular endothelial growth factor (VEGF) (23.4 ng/mL) (both R&D Systems, Minneapolis, MN) or in combinations. The concentrations reflected those of RCC-26–conditioned medium. Functional analyses were performed with monocytes generated with RCC-26–conditioned medium. Myeloid cells within one experiment were derived from the same donor.

Generation of Microtumors and Monocyte Infiltration

Multicellular spheroids were generated as previously described.²⁸ In brief, 10⁵ suspended cells from exponentially growing RCC-53 monolayers were cultured on 1% solid seaplaque agarose (Biozym, Wien, Austria) in 24-well plates. After 4 days, the tight aggregates were transferred to 20 μL of AIM-V containing 10⁵ monocytes and cultured as hanging drops on the lid of a petri dish. After 24 hours, noninfiltrated monocytes were removed and the spheroids cultured for 3 more days. Thereafter, spheroids were dispersed in 5 mmol/L EDTA (mechanic disruption) and the single-cell suspension was analyzed by flow cytometry using LSRII (gated on CD45⁺ cells) (BD Pharmingen, San Diego, CA) and FlowJo (TreeStar, Ashlan, OR).

Macropinocytosis, Endocytosis, and Phagocytosis

For macropinocytosis, cells (3 × 10⁵ cells/600 μL) were incubated with fluorescein isothiocyanate (FITC)–labeled BSA (1 mg/mL; Sigma-Aldrich) for 1 hour at 37°C or 4°C (control) and analyzed by flow cytometry. Endocytosis involved FITC-labeled dextran (500 kDa; Sigma-Aldrich). For phagocytosis, the Vybrant phagocytosis assay (Molecular Probes/Invitrogen) was used.

Antigen Cross-Presentation

Antigen cross-presentation was performed as previously described.²⁹ The system involves the HLA-A2–restricted Melan-A/MART-1–specific CTL-A42 and the pep70-MART peptide, which is an extended 15mer peptide containing the HLA-A2–restricted T-cell epitope of the Melan-A/MART-1 antigen. The N-terminal extension prevents direct loading onto surface HLA-A2 molecules; thus, epitope presentation requires antigen uptake and processing by antigen-presenting cells (APCs) to achieve T-cell stimulation. T-cell stimulation results in interferon-γ (IFN-γ) secretion, which correlates with the amount of antigen cross-presented by the APCs.²⁹ Myeloid cells (2 × 10⁴/100 μL of AIM-V) were incubated with indicated concentrations of pep70-MART peptide for 1 hour at 37°C to allow uptake before addition of resting CTL-A42 (4 × 10³/100 μL of AIM-V, 24 hours, 37°C). IFN-γ in supernatants was measured by enzyme-linked immunosorbent assay (ELISA). Control samples, containing all components except the peptide, were used to determine IFN-γ background. Maximal IFN-γ secretion capacity of CTL-A42 was determined by co-culturing CTL-A42 with MEL93.04A12 (15 × 10³/100 μL), a melanoma cell line with endogenous expression and HLA-A2 presentation of the Melan-A/MART-1 antigen (Table 3).

Cell-Mediated Cytolysis

Cell-mediated cytolysis by CTL-JB4 was determined by a 4-hour chromium release assay as previously described.⁴

Flow Cytometry

Antibodies are listed in Table 2. For surface staining, 10⁵ cells were incubated with antibodies in FACS buffer (PBS, 2% HS, 2 mmol/L EDTA, 0.1% NaN₃) (20 minutes, 4°C) and propidium iodide (Sigma-Aldrich). For intracellular staining (CD3ζ), cells were incubated with 7-amino-actinomycin-D (7-AAD) (Sigma-Aldrich), fixed (1% PFA/PBS, 20 minutes, 4°C) and permeabilized (two consecutive washes with 0.1% and 0.35% saponin in 2% HS/PBS), then stained with antibodies. Apoptosis was detected by annexin V–FITC following the manufacturer's instruction (BioSource/Invitrogen).

Polychromatic flow cytometry was used to analyze TILs. The TILs from RCC tissues were stained with antibody combinations anti-CD45-phycoerythrin (PE)-Cy7, anti-CD11c-FITC, anti-CD209-APC, anti-CD14-Pacific blue (PB), anti-CD163-PE, anti-CD3-AmCyan, and human leukocyte antigen (HLA)-DR-PE (for myeloid cell analysis) or anti-CD45-PE-Cy7, anti-CD3-PB, and anti-CD56-APC (for NK-cell quantification). Dead cells were stained with 7-AAD. Myeloid cells were selected by gating on CD45⁺ 7-AAD⁻ and CD11c⁺ cells. Within the myeloid gate, the intensity of CD209, CD14, HLA-DR, and costimulatory molecules (CD40, CD80, CD86) was determined for the gated CD14⁺CD209⁻, CD14⁻CD209⁺, or CD14⁺CD209⁺ subpopulations. The frequency of NK cells among TILs was determined as the percentage of CD3⁻CD56⁺ cells among the gated lymphocyte population (CD45⁺, 7-AAD⁻, CD14⁻).

To assess maintenance of CTL function when exposed to myeloid APCs, CTL-JB4 was co-cultured with cDCs, ercDCs, or no APCs at a ratio of 1:0.3 in 300 μL of AIM-V with 10% HS (6 × 10⁵ CTLs) in polystyrene round-bottom tubes. After 24 hours, co-cultures were harvested and used to determine phenotypic markers (CD8, CD3ζ, CD3ε, perforin, and granzyme B), as well as cognate functional capacity (degranulation, cytokine production). For phenotype, 24-hour CTL co-cultures were surface stained with antibodies to CD45, CD8, and CD3ε (Table 2) and 7-AAD in FACS buffer. Cells were fixed with 1% PFA followed by permeabilization with saponin and stained with antibodies to perforin and granzyme B. Data acquisition and analysis were performed with LSRII (BD Bioscience) and FlowJo (TreeStar). CTLs were selected by gating on CD45⁺ 7-AAD⁻ cells. The expression levels of markers were determined as the median fluorescence intensity of gated CTLs.

To assess CTL function, CTL-JB4 (3 × 10⁵) (precultured with APCs) were mixed with cognate tumor cells (HLA-A2⁺ RCC-26) at a ratio of 1:1 in the presence of GolgiStop and Brefeldin-A (BD Biosciences, Franklin Lakes, NJ) in AIM-V with 10% HS. To detect degranulation, anti-CD107a-FITC and anti-CD107b-FITC antibodies were added during the stimulation phase. After 5 hours of stimulation, cells were stained for membrane markers (anti-CD45-AmCyan, anti-CD8-PB, anti-CD11c-PE) and 7-AAD, followed by staining for intracellular proteins [anti-IFN-γ-PE-Cy7, anti-IL-2-APC, and anti-tumor necrosis factor (TNF)-α-A700].

T-Cell, Target Cell, and Myeloid Cell Co-Culture

CTL-JB4 (3 × 10³) and cognate HLA-A2–positive RCC-26 (15 × 10³) were cultured for 24 hours in 96-well plates with or without HLA-A2–negative myeloid cells generated in parallel from monocytes of the same donor (0.5 × 10³ or 3 × 10³). Peripheral blood lymphocytes (PBLs) were cultured with allogeneic myeloid cells (ratio 48:1) in 96-well plates for 4 days. Supernatants were harvested for cytokine and chemokine measurements.

Cytokine and Chemokine Measurements

Cytokine and chemokine measurements were performed using the 27-Plex bead array (Bio-Plex; Bio-Rad, Hercules, CA) or ELISA (BD OptEIA, BD Pharmingen) according to the manufacturers' instruction.

MMP-9 Measurement

Gelatinase zymography was performed as previously described.²⁶ Culture supernatant of the human fibrosarcoma cell line HT1080 was used as a reference to detect the position of the matrix metalloproteinase 9 (MMP-9) digestion area. Digestion areas were quantified using Image J software. Supernatants were from APCs (5 × 10³/200 μL, 24 hours) and RCC lines cultured in medium with recombinant TNF-α (rTNF-α).

CTL Migration Using the Modified Boyden Chamber

CTL-JB4 (7.5 × 10³), RCC-26 (37 × 10³) expressing the cognate CTL-pMHC ligand, and either ercDCs or cDCs (7.5 × 10³) (not expressing the cognate ligand) were co-cultured at a final volume of 500 μL of AIM-V in 24-well tissue culture plates (Corning Costar, Cambridge, MA). Parallel cultures involved every combination with two cell types and each cell type alone. All cultures were prepared in duplicates. After 24 hours, Transwell inserts (6.5-mm diameter; 3.0-μm pore size) (Corning Costar) were placed in each well, and carboxyfluorescein diacetate succinimidyl ester (CFDA-SE)–labeled CTL-JB4 (4 × 10⁵ cells/100 μL) was added to each insert and incubated at 37°C/6.5% CO₂. After 30 minutes and a brief shaking of the incubation plate, the inserts were removed and Counting Beads (CALTAG, Invitrogen, Camarillo, CA) were added to the lower chambers. Then cell suspensions were harvested from the lower chamber, and the amount of CFDA-SE⁺-CTL-JB4 was determined by flow cytometry (LSRII). Absolute CFDA-SE⁺-CTL counts were calculated with the help of the counting beads following the manufacturer's protocol. The migration index was determined as the ratio of each absolute cell number to the cell number in the control (medium). CTL-JB4 was labeled with 2 μmol/L CFDA-SE (10⁷ cells/mL) (Molecular Probes/Invitrogen) at 37°C for 10 minutes. The reaction was stopped by adding an equivalent volume of fetal calf serum (FCS) and subsequent washing with RPMI 1640 medium.

RCC Cell Proliferation

Proliferation of RCC cells was measured using the cell screen system (Roche Innovatis AG, Basel, Switzerland). It consists of an inverted phase microscope (Ix 50, Olympus, Hamburg, Germany) equipped with a motorized X-Y stage and a CCD camera. The cell density is determined by measuring the area in a well that is occupied by adherent cells using the pattern recognition software (PA V1.7; Roche Innovatis AG). CCA23 RCC cells³⁰ (4 × 10³ cells in AIM-V/1% FCS) were seeded in 96-well plates. After 15 hours, medium was replaced by AIM-V/1% FCS supplemented with rTNF-α (0.5, 1, 5, 10, and 20 ng/mL) or CTL/APC co-culture supernatants (supplemented with 1% FCS). Each group was set up in six replicates. rTNF-α and co-culture supernatants were replenished after 48 hours. Cell proliferation was monitored over 4 days, each time scanning the full area of each well (×4 objective). Measurements started directly after the addition of rTNF-α or supernatants (0 hours) and were repeated after 24, 48, 72, and 96 hours. For blocking experiments, anti-human TNF-α rabbit immune serum (1:1000, gift of Dr. H. Engelmann, Munich, Germany) or rabbit preimmune serum was added to 2 ng/mL of rTNF-α medium or ercDC/CTL co-culture supernatant and the proliferation was measured over 72 hours.

Statistical Analysis

Nonparametric statistical methods were used (Prism Windows 5.01; GraphPad, La Jolla, CA). For comparison of two unmatched groups, the two-sided Mann-Whitney U-test was applied. Contingency tables, composed of two pairs of categories, were analyzed by Fisher's exact test. Comparison of one parameter between more than two groups was performed with the Kruskal-Wallis test followed by Dunn's posttest. For analysis of multiple comparisons, the Mann-Whitney U-test was applied followed by corresponding P value corrections. Error bars are the mean deviation (MD) or as indicated. RCC tissues were divided into tumors with either low (CD209-cell^low) or high (CD209-cell^high) CD209⁺ cell content. The cutoff for CD209-cell^high tumors was set at 13 cells per high-power field (HPF) as determined by receiver operating characteristic analysis.

RCCs with High CD209⁺ Cell Content Have Advanced Tumor Stages and Lower Fractions of CD8⁺ and NK Effector Cells

RCCs typically present with a prominent immune cell infiltrate, which includes CD8⁺ T cells, NK cells, macrophages, and DCs. Unlike what has been reported for other tumor types,^31,32 the absolute number of intratumoral CD8⁺ T cells in RCC does not correlate with better prognosis. Rather, more intratumoral CD8⁺ cells are generally found in RCC tumors of higher stage and poor prognosis.³ We speculated that the composition of the infiltrate could be different in high- versus low-stage RCC tumors. Using IHC, the DC subsets present in the tumors were characterized by applying antibodies specific for interstitial immature DCs (CD209) and mature DCs (DC-Lamp, CD83). As previously reported,²⁰ DC-Lamp⁺ and CD83⁺ cells were found to be rare and confined to lymphocyte clusters within the tumor parenchyma (see Supplemental Figure S1B at http://ajp.amjpathol.org). In contrast, CD209⁺ cells were numerous and regularly dispersed throughout the tumor parenchyma alongside T lymphocytes but were absent from clustered CD3 accumulations (see Supplemental Figure S1C at http://ajp.amjpathol.org). We then determined the presence of specific lymphocyte populations in tumors with high or low CD209⁺ cell content. CD209-cell^high tumors had more T cells than did the CD209-cell^low tumors but a similar number of NK cells (Table 4), indicating that an increase in one lymphocyte subset was not similarly paralleled by that of other subsets, thereby changing the composition of the infiltrate. To describe the infiltrate composition, the ratios of CD3⁺, CD8⁺, or NK cells to CD209⁺ cells were calculated for each tumor (Table 4 and Figure 1A). The ratios varied among individual tumors, indicating individual infiltrate compositions that were independent of the absolute numbers of each immune cell subset. Importantly, comparing tumors with high numbers of CD209⁺ cells to those with low numbers of CD209⁺ cells, it was observed that CD209-cell^high tumors had lower ratios of CD3⁺, CD8⁺, or NK cells relative to CD209⁺ cells than did CD209-cell^low tumors. Specifically, CD209-cell^low tumors harbored on average three times more CD8⁺ or NK cells for each CD209⁺ cell when compared with the CD209-cell^high tumors (Table 4). The difference between both groups was statistically significant (Figure 1 and Table 4). The change in infiltrate composition suggested a reduced level of cytotoxic effector cell recruitment in the CD209-cell^high tumors.

DCs can help mount immune responses against tumors,⁶ suggesting that a large DC content may correlate with an early tumor stage. However, setting the CD209⁺ cell content in relation to the TNM classification, a strong predictor of disease prognosis,³³ we observed that CD209-cell^high tumors rarely showed a TNM stage associated with good prognosis (pT1N0M0) but generally had prognostically poor tumor stages (pT2-4N0-2M0-1) (Figure 1B). By contrast, CD209-cell^low tumors were equally distributed among the two TNM categories. These results raise the possibility that the CD209⁺ DCs in RCC may have an adverse effect on tumor growth control.

Most CD209⁺ DCs in RCC Co-Express Macrophage Markers

CD209⁺ DCs were then characterized in RCC and NKC tissues using multiparameter immunofluorescence staining combining CD209 with the lineage marker CD14 for macrophages and CD163 for chronically activated macrophages and activated circulating DCs.¹⁰ Unexpectedly, conventional CD209⁺ DCs, which do not co-express these macrophage markers, were rarely found in RCC and NKC tissues. Rather, most CD209⁺ DCs co-expressed CD14, CD163, or both markers (Figure 2A). In RCC, the triple marker–positive CD209⁺CD14⁺CD163⁺ cells represented the dominant population among the CD209⁺ cells (mean, 62%; range, 26% to 80%), whereas they were significantly less frequent in NKC (mean, 19%; range, 0% to 43%) (P = 0.007) (Figure 2B, upper panel). Most CD209⁺ cells in NKCs co-expressed CD163 without CD14 (mean, 70%; range, 45% to 95%), and the frequency of this subset was significantly higher than in RCC (P = 0.006). The CD209⁺CD14⁺ phenotype predominated in the tumor center (median, 71%; range, 51% to 83%) (Figure 2B, lower panel). In the tumor periphery, the tumor area directly adjacent to the tumor “pseudocapsule,” an intermediate frequency of 46% (range, 37% to 61%) was found, whereas in NKCs only an 18% frequency (range, 0% to 88%) in this subset was observed (Figure 2B). This distinct subset distribution suggests that different regional milieus exist within renal tissue that determine marker co-expression. A histologic view depicting the different regions of RCC tumors is shown in Supplemental Figure S1A (available at http://ajp.amjpathol.org). Based on their predominance in RCC tissue, the CD209⁺CD14⁺ DCs (with or without CD163 expression) are referred to as ercDCs.

ErcDCs isolated from tumor tissues expressed the markers CD209 and CD14 at intensities similar to cells with single-marker expression (Figure 2C). This supports the proposition that the CD209⁺CD14⁺ ercDCs represent an additional subtype within the DC/macrophage differentiation spectrum. ErcDCs expressed HLA-DR and costimulatory molecules (CD40, CD80, CD86) at levels comparable or higher than cDCs (Figure 2D), and they were positive for the blood DC antigen 1 but not blood DC antigen 3 (not shown), indicating a DC lineage relationship.¹⁴

Soluble Tissue Factors Induce the Phenotypic Differentiation of ErcDCs in Vitro

Monocytes differentiate into macrophages or DCs, depending on the local tissue milieu.^6,8,9 Because CD209⁺ cells with macrophage marker co-expression represented the dominant CD209⁺ population in tumor tissues, tumor tissue–related factors are probably responsible for the induction of this phenotype. Monocyte infiltration into tumors was mimicked by adding monocytes to RCC microtumors. Subsequent analysis revealed the induction of the ercDC phenotype on microtumor-infiltrating monocytes (Figure 3A, RCC-53 spheroid).

To determine whether soluble tissue–derived factors are sufficient to elicit the ercDC phenotype, monocytes were cultivated with cell-conditioned medium from cell lines, including RCC lines (RCC-26, RCC-53), proximal tubular epithelial cells (NKC-26, HK-2), pancreatic (Colo-357) and prostate (DU-145) carcinoma, Epstein-Barr virus transformed B cells (LCL-26), and T-cell leukemia (Jurkat) (Table 3). No additional cytokines were applied. Flow cytometry revealed that all cell-conditioned media, with the exception of that produced by LCL-26, induced co-expression of CD209, CD14, and CD163 (Figure 3A shows representative results achieved with indicated conditioned media). As expected, the IL-4/GM-CSF–exposed monocytes developed into a cell population only positive for CD209, reflecting monocyte-derived cDCs (Figure 3A). Monocytes cultured in LCL-26–conditioned medium remained CD14 positive and negative for CD209 and CD163. The Jurkat-conditioned medium did not yield sufficient viable cells for analysis.

To define factors responsible for the co-induction of CD209, CD14, and CD163, the cytokine and chemokine profile of the cell-conditioned media was analyzed (Table 5). Three cytokines, IL-6, CXCL8/IL-8, and VEGF, were common to the cell-conditioned media that induced the triple-marker phenotype and absent in the LCL-26–conditioned medium, which did not induce this phenotype.

The T_H2 cytokines IL-4, IL-13, or IL-15, known inducers of CD209,^34–36 or IL-10 were detected at very low concentrations in all conditioned media, including those that did not induce the ercDC phenotype. Transforming growth factor β was detected in most cell-conditioned media but was not present in the conditioned medium of Colo-357, which induced the ercDC phenotype.

A cocktail of IL-6, CXCL8/IL-8, and VEGF was subsequently shown to induce CD209 and CD163 on monocytes with maintenance of CD14 expression (Figure 3A). A combination of either CXCL8/IL-8 or IL-6 with VEGF only partially induced the ercDC phenotype (not shown), demonstrating the importance of all three factors. Additional factors in the tumor cell-conditioned medium seem to support survival because cell yields were consistently higher when using the tumor cell-conditioned medium for differentiation of the ercDCs compared with the cytokine cocktail.

ErcDCs Express DC Markers and Have DC Functions

Surface marker analysis of in vitro generated single-CD209⁺ cDCs and triple-CD209⁺CD14⁺CD163⁺ ercDCs revealed that major histocompatibility complex class I and II, CD40, CD80, and CD86 expression was similar or higher on ercDCs (Figure 3C) differentiated in vitro, consistent with the expression pattern observed for the in vivo ercDC and cDC subsets from RCC tissues (Figure 2D). B7-H1, commonly associated with immunoinhibition,⁵ was lower on in vitro generated ercDCs, whereas CX₃CR1 was expressed at the same level (Figure 3C). CD16 and CD15 were absent on both cell types (not shown).

In vitro generated ercDCs showed similar or higher activities of macropinocytosis, receptor-mediated endocytosis, and phagocytosis than cDCs, measured by uptake of BSA-FITC (Figure 3B), dextran-FITC, or FITC-labeled Escherichia coli BioParticles (not shown).

Endocytic activity is necessary for the cross-presentation of exogenous antigens to T lymphocytes. The capacity of cDCs and ercDCs in cross-presentation was compared using an established assay consisting of the CTL-A42 and an extended pep70-MART 15mer peptide, which requires uptake and processing by APCs for epitope presentation and T-cell stimulation.²⁹ ErcDCs allowed stronger epitope recognition by CTLs than cDCs, as seen by higher CTL-IFN-γ secretion (Figure 3D). This confirms a key DC function (ie, antigen cross-presentation)¹⁴ for this cell population that co-expresses CD209 and the macrophage lineage marker CD14.

ErcDCs Are in Crosstalk with T Cells in Situ

CD209⁺ cells were found dispersed throughout the RCC parenchyma alongside T lymphocytes (see Supplemental Figure S1C at http://ajp.amjpathol.org). Contacts with T cells were frequent and equally involved CD209⁺CD14⁺ and CD209⁺CD163⁺ cells (Figure 4A, Table 6; see also Supplemental Figure S1C at http://ajp.amjpathol.org). Most of these cells are presumably CD209⁺CD14⁺CD163⁺ cells because they represent the predominant CD209⁺ subtype in RCC.

DCs can be immunosuppressive, causing T-cell inactivation or the induction of FOXP3⁺ T-regulatory cells.⁶ To identify functional consequences resulting from ercDC/T-cell contacts in situ, FOXP3⁺ cells and potential down-regulation of CD3ε and CD3ζ, representing signs of T-cell inactivation and preceding apoptosis,³⁷ were evaluated. CD3ε and CD3ζ expression was similar on T cells with or without contact to CD209⁺ cells (Figure 4B). FOXP3⁺ cells were infrequent in RCC or NKC tissues (confirming published results^16,38) and were rarely observed in close proximity to ercDCs (Table 6). Together, these data provided no evidence that the RCC-resident ercDCs would have a negative impact on the local T-cell infiltrate.

ErcDCs Do Not Suppress Cognate CTL Function in Vitro

The conclusion drawn from the in situ observation that ercDCs appear to not negatively affect the local T-cell infiltrate was supported by in vitro studies. In co-culture with T cells, in vitro generated ercDCs did not cause down-regulation of CD3ε or CD3ζ on T cells and there was no elevated T-cell apoptosis compared with cDC co-cultures (Figure 4C). In addition, ercDCs did not inhibit cognate stimulation of CTLs by tumor cells because similar amounts of IFN-γ were detected in co-cultures with ercDCs, with cDCs, or without APCs (Figure 4D). Moreover, cytotoxicity of CTLs against tumor cells was enhanced by both cDCs and ercDCs (Figure 4D). Maintenance of CTL effector function was also not negatively affected: cytokine response and degranulation were unchanged (Figure 5A) and lytic activity of CTLs that had been exposed to ercDCs or cDCs for 24 hours was higher than that of unexposed CTLs (Figure 5B). This higher lytic activity may result from the increased perforin and granzyme B observed in this context (Figure 5C).

ErcDCs Produce High Levels of MMP-9 and Regulate TNF-α and Chemokines in T-Cell Co-Culture

MMP-9 is a matrix degrading enzyme that can be produced by tumor cells or by tumor-infiltrating myeloid cells. Its expression is associated with more aggressive tumor growth.^39,40 Using gelatinase zymography, ercDCs were found to secrete higher levels of MMP-9 (mean of 5 donors: 4.6-fold; range, 1.3–11.2) than did cDCs (Figure 6A).

The presence of ercDCs in CTL/tumor-cell co-cultures led to an increase in TNF-α (Figure 6, B and D), which did not occur in co-cultures containing cDCs. TNF-α is described as a key cytokine involved in tumor promotion and angiogenesis, and it is also known to stimulate MMP secretion and proliferation in RCC tumors.⁴¹ Using rTNF-α, a concentration-dependent induction of MMP-9 secretion and proliferation of RCC was documented in vitro (see Supplemental Figure S2, A and B, at http://ajp.amjpathol.org). Furthermore, CTL/ercDC co-culture supernatants were shown to induce a stronger proliferation of an RCC line in vitro compared with co-culture supernatants from CTLs with cDCs (see Supplemental Figure S2C at http://ajp.amjpathol.org). Using a blocking anti-TNF-α antiserum, the proliferation induced by rTNF-α was completely abrogated. Using CTL/ercDC co-culture supernatants, the induced proliferation was attenuated but not completely blocked, suggesting that other factors in the ercDC co-culture supernatant also promote RCC cell proliferation (see Supplemental Figure S2D at http://ajp.amjpathol.org).

In addition to increased TNF-α, CTL/tumor cell co-cultures containing ercDCs showed a reduction in CXCL10/IP-10 and CCL5/RANTES (Figure 6, B and D). Other chemokines, such as CCL2/MCP-1, were present at similar levels in co-cultures containing ercDCs or cDCs (Figure 6D). CXCL10/IP-10 and CCL5/RANTES are important chemokines involved in the recruitment of T_H1-polarized effector cells and the antitumor immune response.⁴² Using the modified Boyden chamber assay and the T_H1-CTL clone JB4 (expressing the CXCL10/IP-10 receptor CXCR3), less CTL mobilization was observed when using CTL/tumor cell co-culture supernatants containing ercDCs (migration index, 2.6 ± 0.8) compared with supernatants of CTL/tumor cell co-cultures containing cDCs (migration index, 4.0 ± 1.4) (Figure 6C).

The changes in TNF-α and the chemokines moderated by ercDCs required co-culture with CTLs or PBLs because ercDCs or cDCs alone did not produce significant levels of these factors (see legend to Figure 6). Thus, ercDCs, although not compromising T-effector lymphocyte function, appear to use the crosstalk with T cells to cause milieu alterations that favor tumor proliferation and can attenuate T_H1 responses.

Close contacts of ercDCs with T cells were observed in tissue sections. Thus, one might speculate that similar milieu alterations could result from these interactions in vivo. The CD209-cell^high tumors were of more advanced tumor stages and showed lower infiltration with CD8⁺ cells and NK cells (Figure 1). These lymphocytes are by majority CXCR3⁺⁴³ and equipped with lytic granules (P.P. unpublished) and, thus, are Th1-polarized cytotoxic effector lymphocytes responsive to CXCL10/IP-10. These observed characteristics of CD209-cell^high tumors could be a consequence of the modulatory effects evoked by the presence of ercDCs and their interaction with T cells in situ.