Tumor progression despite massive influx of activated CD8⁺ T cells in a patient with malignant melanoma ascites

Abstract

Although melanoma tumors usually express antigens that can be recognized by T cells, immune-mediated tumor rejection is rare. In many cases this is despite the presence of high frequencies of circulating tumor antigen-specific T cells, suggesting that tumor resistance downstream from T cell priming represents a critical barrier. Analyzing T cells directly from the melanoma tumor microenvironment, as well as the nature of the microenvironment itself, is central for understanding the key downstream mechanisms of tumor escape. In the current report we have studied tumor-associated lymphocytes from a patient with metastatic melanoma and large volume malignant ascites. The ascites fluid showed abundant tumor cells that expressed common melanoma antigens and retained expression of class I MHC and antigen processing machinery. The ascites fluid contained the chemokines CCL10, CCL15, and CCL18 which was associated with a large influx of activated T cells, including CD8⁺ T cells recognizing HLA-A2 tetramer complexes with peptides from Melan-A and NA17-A. However, several functional defects of these tumor antigen-specific T cells were seen, including poor production of IFN-γ in response to peptide-pulsed APC or autologous tumor cells, and lack of expression of perforin. Although these defects were T cell intrinsic, we also observed abundant CD4⁺CD25⁺FoxP3⁺ T cells, as well as transcripts for FoxP3, IL-10, PD-L1/B7-H1, and indoleamine-2,3-dioxygenase (IDO). Our observations suggest that, despite recruitment of large numbers of activated CD8⁺ T cells into the tumor microenvironment, T cell hyporesponsiveness and additional negative regulatory mechanisms can limit the effector phase of the anti-tumor immune response.

Introduction

Tumor cells can be recognized and eliminated by the immune system, and immunotherapy has been used successfully in some patients to eradicate established tumors (for reviews, see for example [1, 2]). Many tumor antigens have been identified and a large number of clinical trials designed using various strategies to attempt to generate and boost anti-tumor immune responses. However, the majority of patients treated with immunotherapy do not respond clinically, often despite efficient generation of anti-tumor specific T cells as measured by commonly used immune monitoring tools such as tetramer staining and ELISpot. Several events are necessary for generating an efficient anti-tumor immune response and the reasons for a lack of clinical responses have not always been clear. In some cases, there might be defects in T cell trafficking, so that even if tumor antigen-specific T cells are found in peripheral blood, they are unable to enter tumor sites. In other cases, the tumor cells may escape the immune system through lack of expression of components of the antigen processing machinery and hence poor antigen presentation. Finally, even if the T cells are able to enter the tumor, lack of co-stimulation and presence of inhibitory factors or regulatory T cells present in the tumor microenvironment might inhibit T cell effector function.

Careful analysis of both the tumor microenvironment and of tumor antigen-specific T cells in patients who do not respond clinically should provide important insights into mechanisms of failed immune-mediated tumor regression. However, such an analysis is often difficult due to limited numbers of tumor cells and fresh tumor-infiltrating lymphocytes (TIL) from the same patient. In the current report, we have taken the opportunity to investigate the T cell response in a patient with metastatic melanoma who developed massive malignant ascites containing abundant tumor cells and activated T cells. This patient allowed for a unique opportunity to explore several hypotheses regarding tumor resistance to T cell effector function that could explain tumor escape.

Material and methods

Patient information

The patient was a 31-year-old Caucasian male with metastatic melanoma. He had initially been diagnosed with stage III (lymph node-positive) disease and was treated with surgery, radiation therapy, and adjuvant interferon-α. He then recurred with metastatic disease to the lung and received a vaccine composed of a gp100 peptide emulsified in Montanide that was given along with low-dose IL-2. After again developing progressive disease he was treated with anti-CTLA-4 mAb plus IL-2. Approximately 6 weeks after this treatment the patient developed progressive disease that included large volume malignant ascites requiring frequent paracentesis. He was consented to provide ascites fluid through a correlative science protocol approved by the University of Chicago Institutional Review Board.

Real-time RT-PCR

Total RNA was extracted from ascites cells using GenElute™ Mammalian Total RNA Miniprep Kit (Sigma-Aldrich Corp., St. Louis, MO, USA). The RNA was DNase I treated and cDNA was synthesized from 1 μg total RNA using MLV-Reverse Transcriptase (Invitrogen Corp., Carlsbad, CA, USA) according to the manufacturer’s directions. As a control, a mock batch of cDNA was made without adding reverse transcriptase (RT). 1/20 of the total amount of cDNA, RT positive or RT negative, was used in each real-time PCR reaction. The reactions were run on an ABI PRISM® 7700 Sequence Detection System machine and analyzed using Sequence Detector v1.7a software (Applied Biosystems, Foster City, CA, USA). Primers and probes were synthesized by Applied Biosystems and are listed below:

β-actin

Sense: 5′-GGATGCAGAAGGAGATCACTG-3′

Anti-sense: 5′-CGATCCACACGGAGTACTTG-3′

Probe: 6FAM-CCCTGGCACCCAGCACAATG-TAMRA

gp100

Sense: 5′-CCCCGCATCTTCTGCTCTT-3′

Anti-sense: 5′-ACTCAGACCTGCTGCCCACT-3′

Probe: 6FAM-CCCATTGGTGAGAATAGCCCCCTCC-TAMRA

MAGE-3

Sense: 5′-TCCTGTGATCTTCAGCAAAGCTT-3′

Anti-sense: 5′-GGGTCCACTTCCATCAGCTC-3′

Probe: 6FAM-CAGTTCCTTGCAGCTGGTCTTTGGCAT-TAMRA

Melan-A

Sense: 5′-GCCACTCTTACACCACGGCT-3′

Anti-sense: 5′-CAGTAAGACTCCCAGGATCACTGTC-3′

Probe: 6FAM-AGGCCGCTGGGATCGGCATC-TAMRA

NA17-A

Sense: 5′-TGTGGTTTTCTGTCTTACAGTTGTTTG-3′

Anti-sense: 5′-AGCACATCCATTTGCGATAGC-3′

Probe: 6FAM-CCTTACGAGGGCCCAGCTCCCC-TAMRA

Arginase I

Sense: 5′-CAGAGCATGAGCGCCAAGT-3′

Anti-sense: 5′-ATCACACTCTTGTTCTTTAAGTTTCTCAA-3′

Probe: 6FAM-CCTTTCTCAAAGGGACAGCCACGAGG-TAMRA

FoxP3

Sense: 5′-GGCACTCCTCCAGGACAG-3′

Anti-sense: 5′-GCTGATCATGGCTGGGCTCT-3′

Probe: 5′-FAM-ATTTCATGCACCAGCTCTGAACGG-TAMRA-3′

IDO

Sense: 5′-GCATTTTTCAGTGTTCTTCGCATA-3′

Anti-sense: 5′-TCATACACCAGACCGTCTGATAGC-3′

Probe: 5′-FAM-TCTGGCTGGAAAGGCAACCCCC-TAMRA-3′

PD-L1

Sense: 5′-TCTGGCACATCCTCCAAATG-3′

Anti-sense: 5′-CAGTGCTACACCAAGGCATAATAAG-3′

Probe: 5′-FAM-AAGGACTCACTTGGTAATTCTGGGAGCCA-TAMRA-3′

Cell surface marker analysis

Lymphocyte subsets were analyzed by flow cytometry. The lymphocyte population was gated based on size and granularity. The percentages of CD19⁺, CD56⁺, and CD56⁺ CD3⁺ lymphocytes were determined using FITC anti-CD19, PE anti-CD56, and PerCP anti-CD3. The fractions of HLA-DR⁺ CD8⁺ or CD4⁺ cells were determined using a combination of PE-conjugated anti-HLA-DR, PerCP anti-CD3, and either FITC anti-CD8 or FITC anti-CD4. CD25⁺ CD4⁺ cells were stained using a combination of PE-conjugated anti-CD25 or isotype matched control antibody, PerCP anti-CD3 and FITC anti-CD4. All these antibodies were purchased from BD Biosciences (San Jose, CA, USA) and used according to the manufacturer’s directions. Intracellular FoxP3 expression was assessed with a FITC-coupled antibody from BioLegend (San Diego, CA, USA). For MHC class I expression analysis, cells were stained using FITC anti-HLA-A2 (One Lambda Inc., Canoga Park, CA, USA) and PE anti-HLA-A,-B,-C (BD Biosciences) antibodies. Data were acquired using a FACScan flow cytometer (Becton Dickenson Immunocytometry Systems, San Jose, CA, USA) and analyzed with CellQuest software (© Becton Dickenson Immunocytometry Systems 1994–1999).

Intracellular cytokine staining

Whole blood or ascites fluid was centrifuged, and pelleted cells were resuspended in RPMI with l-glutamine and monensin (2 μM). Half the sample was activated to induce T cell cytokine production by addition of PMA (40 ng/ml) and ionomycin (1.6 μg/ml); the other half was set aside as a resting control. Cells were incubated for 4 h at 37°C, were then stained with FITC anti-CD8 and PerCP anti-CD3, and the RBCs were lysed with FACS Lysing Solution (BD Biosciences). The remaining cells were washed, permeabilized, and stained for intracellular cytokines using PE-conjugated anti-IL-2, anti-IL-4, or anti-IFN-γ. All antibodies were purchased from BD Biosciences.

Cytokine and chemokine protein array

RayBio™ Human Cytokine Array V (RayBiotech, Inc., Norcross, GA, USA) was used to simultaneously detect the presence of 79 different cytokines and chemokines according to the manufacturer’s directions. Briefly, a membrane arrayed with antibodies against the various proteins was blocked for 30 min, blocking buffer removed, and ascites fluid (1 ml) added to the membrane and incubated for 1 h. The membrane was then washed, incubated with a cocktail of biotinylated secondary antibodies for 1 h, washed again and incubated with HRP-conjugated streptavidin for 30 min. Blots were then washed a final time and incubated with detection solution (ECL, ECL™ is a trademark of Amersham Pharmacia Biotech). Presence of bound cytokines was visualized by exposing the membrane to film. The film was scanned in at 1,200 dpi and spots quantitated by densitometry analysis using UN-SCAN-IT gel automated digitizing system software (Silk Scientific, Orem, UT, USA). The average pixels in the positive control spots (row 1, columns A–D and row 8, columns J–K) were compared to the cytokine and chemokine spots, after first subtracting the background average pixels from the negative control spots (row 1, columns E–F and row 8, column I).

Chemokine receptor expression

Expression of chemokine receptors CXCR1-4 and CCR1-5 and of CXCR4 ligand SDF-1α was detected by RT-PCR of cDNA prepared as described for real-time RT-PCR above, using primers and probes from Maxim Biotech, Inc. (South San Francisco, CA, USA) according to the manufacturer’s directions.

Tetramer staining

Mononuclear cells were prepared from either peripheral blood or ascites fluid by centrifugation over a Lymphoprep™ (Axis-Shield, Oslo, Norway) gradient. The cells were washed with PBS and tetramer staining was performed using iTAg™ PE-conjugated tetramers (Beckman Coulter Immunomics Operations, San Diego, CA, USA) according to the manufacturer’s directions. Each sample was also stained with FITC anti-CD8 and PerCP anti-CD4, anti-CD14, and anti-CD56 (BD Biosciences). The tetramers were manufactured as a complex of the HLA-A*0201 allele of human MHC class I with HLA-A*0201 restricted peptides derived from EBV BMLF1 (GLCTVLAML), gp100 (KTWGQYWQV), MAGE-3 (FLWGPRALV), Melan-A (AAGIGILTV), and NA17-A (VLPDVFIRCV). As a negative control, a tetramer containing a HLA-A*0201 restricted peptide not known to correspond to any epitope recognized by CD8⁺ T cells was used (cat. no. T20224, Beckman Coulter Immunomics Operations). PerCP-positive cells were excluded from the analysis, and lymphocytes were gated based on size and granularity. The percent of tetramer-positive CD8⁺ cells out of the total CD8⁺ cells was determined for each sample. For multicolor analysis, there was no PerCP staining, and the following antibodies were used: APC anti-CD28, APC-Cy7 anti-CD27, PE-Cy5 anti-CD45RA (eBioscience, San Diego, CA, USA) and APC anti-CD62L, PE-Cy7 anti-CCR7 (BD Immunocytometry Systems). For perforin staining, PE-Cy7 anti-CD8 was used together with the tetramers and then FITC anti-perforin (BD Biosciences Pharmingen, San Diego, CA, USA) used to stain the cells after first permeabilizing the cells. For multicolor and perforin staining, data were acquired using an LSR II cytometer (BD Immunocytometry Systems) and analyzed with FloJo software (© Tree Star, Inc. 1997–2002).

RT-PCR of antigen presentation pathway components

Primers and probes used have been described previously [3]. cDNA was prepared as described for real-time RT-PCR above. PCR was performed using 35–40 cycles of 1 min at 94°C, 1 min at primer pair annealing temperature, 1 min at 68°C, followed by final extension at 72°C for 10 min.

ELISpot assays

ELISpot assays and analysis of the ELISpot plates were performed as previously described [4]. Purified CD8⁺ T cells were plated in triplicate at 1×10⁵ cells/well together with T2 cells pulsed with peptide (described above) as APCs. In one experiment described, tumor cells from the tumor cell line were used as stimulators. The cells were also stimulated with PMA (50 ng/ml) + ionomycin (0.5 μg/ml) as a positive control for T cell activation. T cell lines specific for HLA-A*0201 restricted peptides were generated as previously described [5] from normal donor T cells and were plated at 1×10⁴ cells/well together with A2.1-transfected K562 cells or patient-derived tumor cell line cells either untreated or pulsed with the relevant peptide (50 μM).

Results

Melanoma antigen transcripts and tumor cells are present among the ascites cells

The ascites fluid obtained from the patient was quite cellular. Clinical cytology examination confirmed the presence of malignant cells consistent with the known history of metastatic melanoma. To confirm whether the tumor cells expressed defined melanoma antigens, we first performed real-time RT-PCR analysis for various melanoma tumor antigens on RNA derived from the total population of ascites cells. There was expression of gp100, MAGE-3, Melan-A, and NA17-A, suggesting that some of the cells present in the ascites fluid were antigen-expressing melanoma tumor cells (Fig. 1a). To verify the presence of tumor cells, total ascites cells were cultured continuously for several weeks to establish a stable tumor cell line. When the resulting cell line was analyzed for expression of tumor antigens by real-time RT-PCR, the same pattern of expression was seen as in the total cells, but at higher relative levels (Fig. 1b). The fact that these tumor cells were recognized by antigen-specific T cell lines (see below) argues that proteins encoded by these transcripts were also expressed. Expression of Melan-A and gp100 proteins was confirmed by immunohistochemistry (data not shown). These findings confirmed that viable antigen-expressing melanoma tumor cells were present in the ascites fluid.

Fig. 1.

Real-time RT-PCR of total ascites cells and of the cell line established from ascites. cDNA was analyzed for expression of melanoma tumor antigens gp100, MAGE-3, Melan-A, and NA17-A, and for β-actin by real-time RT-PCR. a Real-time RT-PCR from total ascites cells. b Real-time RT-PCR from the melanoma cell line

A large proportion of activated CD8⁺ T cells are present in the ascites fluid as compared to peripheral blood

Clinical hematology laboratory cell counts revealed that the ascites fluid contained a large percentage of lymphocytes. To determine the constitution of these cells, flow cytometric analysis was performed. There was not a significant number of B cells present among the lymphocytes in the ascites fluid (0.5% in the ascites vs. 17.9% in the blood, Fig. 2a), and there was an increased number of CD3⁺ lymphocytes as compared to the peripheral blood (86.7 vs. 65.0%, Fig. 2a). The lack of FITC-labeled B cells among the ascites cells caused an apparent shift in PE staining of the PerCP-labeled CD3⁺ cells due to the compensation settings used. This phenomenon made it hard to quantitate the number of CD56⁺ CD3⁺ NK T cells, but the fraction of NK cells remained similar in ascites and blood (11.4 vs. 11.0%). Among the T cells, there was also a relatively large increase in the proportion of CD8⁺ cells, with 34% in the ascites fluid compared to 22.1% in the blood (Fig. 3 and data not shown).

Fig. 2.

Lymphocyte subset analysis and T cell activation status. a The percent of lymphocytes among total ascites and peripheral blood cells was determined by flow cytometry (top panel). Among these lymphocytes, the percent of cells staining positive for CD19 (B cells) and CD3 (T cells and NK T cells) were analyzed (middle panel), as well as the percent of cells that were CD56⁺ CD3⁻ (NK cells), CD56⁺ CD3⁺ (NK T cells), and CD56⁻ CD3⁺ (T cells) (bottom panel). b The percent of activated T cells was determined by staining for MHC class II (HLA-DR) expression among either the CD8⁺ (top panel) or CD4⁺ (bottom panel) CD3⁺ cells

Fig. 3.

Flow cytometric analysis of intracellular cytokine production. a Ascites cells were either left unstimulated (resting) or stimulated with PMA + ionomycin (activated), surface stained, and permeabilized for intracellular staining. They were then incubated with antibodies against either IL-2 (upper panels), IL-4 (middle panels), or IFN-γ (lower panels) prior to analysis by flow cytometry. The percent CD8⁺ and CD8⁻ cells producing each cytokine out of the total pool of CD8⁺ or CD8⁻ cells were determined and are indicated in the dot plots. b Peripheral blood cells were treated and analyzed in the same manner, and in parallel with, the ascites cells in panel a

To determine the activation status of these T cells, the level of expression of MHC class II (HLA-DR) was assessed (Fig. 2b). Whereas the peripheral blood lymphocytes had modest levels of HLA-DR expression, with 10.5% of all CD4⁺ T cells and 15.3% of all CD8⁺ T cells expressing some level of HLA-DR, the lymphocytes in the ascites fluid displayed a greatly elevated HLA-DR expression, with 42.6% of all CD4⁺ T cells and 63.0% of all CD8⁺ T cells expressing HLA-DR, suggesting that a large proportion of T cells present in the malignant ascites were activated.

Primed T cells are thought to have an increased ability to produce cytokines compared to naïve or resting T cells. To determine whether the T cells in the ascites fluid had an increased cytokine producing potential, intracellular cytokine levels were analyzed in response to PMA and ionomycin. IL-2 and IFN-γ were produced by a much greater fraction of ascites T cells compared to peripheral blood T cells, and some IL-4-producing cells also were detected (Fig. 3). These results are consistent with the activated surface phenotype of the ascites T cells. However, stimulation with PMA and ionomycin could bypass a potential anergic state, and this general functional assay does not directly address the status of tumor antigen-specific T cells.

T cell-attracting chemokines and some cytokines are present in the ascites fluid

The presence of large numbers of activated T cells and other inflammatory cells in the peritoneal cavity suggested that specific chemokines might have been present and mediated preferential accumulation in this compartment. To address this question, a sensitive cytokine protein array was used to simultaneously assay for the presence of 79 cytokines and chemokines (Fig. 4a). The resulting spots that corresponded to each factor were analyzed by densitometry and compared to assay-specific positive controls to determine relative expression levels (Fig. 4b).

Fig. 4.

Cytokine/chemokine array of ascites fluid. Ascites fluid was analyzed for the presence of 79 different cytokines and chemokines using a membrane spotted with antibodies directed against the cytokines and chemokines. After blocking the membrane, it was incubated with undiluted ascites fluid, then washed and incubated with a cocktail of biotin-conjugated secondary antibodies. After further washing and incubation with HRP-conjugated streptavidin, the presence of bound cytokine was detected using ECL and the membrane exposed to film. a Image of the membrane, showing the cytokine and chemokine spots. b Densitometry analysis, showing the relative pixel intensity compared to the positive control spots (1A–D, 8J–K), adjusted for the negative control spots (1E–F, 8H)

A number of T cell attracting chemokines were detected, including IP-10 (CCL10), PARC (CCL18), and MIP-1δ (CCL15). There were also high levels of certain inflammatory factors, including IL-6, IL-8, and NAP-2. However, despite the large numbers of activated T cells present in the malignant ascites, there was no detectable production of IL-2, IL-4, IL-5, IL-13, or IFN-γ. The only potential T cell-derived cytokine present was IL-10; however, IL-10 could have been produced by non-T cells of the monocytes/macrophage lineage. The paucity of T cell-derived cytokines argues that in situ activation of T cells in response to tumor cells within the peritoneal cavity was unlikely to have been occurring.

Chemokines binding to CXCR1, CXCR2, CXCR3, and CCR1 were identified in the ascites fluid, which led us to investigate the expression of these chemokine receptors (Fig. 5). In the total ascites cells, CXCR3 and CXCR4 were expressed at high levels, whereas CXCR1 and CXCR2 expression was minimally detected. Neither of the CXCR family members was expressed in the tumor cells, suggesting that CXCR3 and CXCR4 were expressed by the infiltrating inflammatory cells present in the ascites. CCR1–5 were all present in the total ascites cells, whereas the tumor cells only showed detectable expression of CCR1. Thus, CCR2–4 were also likely expressed by host inflammatory cells.

Fig. 5.

Analysis of chemokine receptor expression of the total ascites cells and on the cell line. RNA extracted from the cells was analyzed for expression of chemokine receptors by RT-PCR, as compared to control cDNA expressing all of the genes analyzed. GAPDH was used as a control for cDNA quality, and RNA prepared using no RT (RT−) used to control for genomic DNA contamination. a RNA from total ascites or cell line analyzed for expression of CXCR1-4 and the chemokine SDF-1α. b RNA from total ascites cells or the melanoma cell line analyzed for expression of CCR1-5

Large numbers of tumor antigen-specific T cells are found in the malignant ascites that have an activated phenotype

Class I MHC/peptide tetramer staining was performed to determine whether CD8⁺ T cells specific for defined melanoma antigen peptides were present among the activated T cells in the ascites. HLA-A2 tetramers were utilized that were bound to peptides derived from EBV, gp100, MAGE-3, Melan-A, and NA17-A. A commercially available negative control tetramer was also used. Tetramer staining of peripheral blood mononuclear cells revealed a small population (0.52%) of EBV-specific CD8⁺ T cells, as well as even smaller numbers of Melan-A specific and NA17-A specific CD8⁺ T cells (Fig. 6b). Tetramer staining of mononuclear cells from the ascites fluid resulted in higher background staining, as seen using the negative control peptide. Despite the background staining, it was clear that a large proportion of Melan-A-specific and NA17-A-specific CD8⁺ T cells were present among the ascites cells (Fig. 6a). The fraction of antigen-specific cells was thus increased in the ascites fluid, although there did not appear to be preferential recruitment of tumor antigen-specific T cells since EBV-reactive T cells were also present (Fig. 6a).

Fig. 6.

Tetramer analysis of antigen-specific CD8⁺ T cells. a Mononuclear cells from ascites fluid were stained with the indicated tetramers and anti-CD8 antibody. The percent tetramer⁺ CD8⁺ cells among all CD8⁺ cells were determined and is shown. b Peripheral blood cells were stained and analyzed in the same manner as, and in parallel with, the ascites cells in panel a

To further dissect the phenotype of the antigen-specific T cells present in the ascites, multicolor flow cytometric analysis was performed to examine expression of CD45RA, CD62L, CD27, CD28, and CCR7. Among the total CD8⁺ T cells, a majority of cells were CD45RA⁻ CCR7⁻, and these cells were CD27⁺, CD28⁺, and CD62L^lo (Fig. 7 and data not shown). This distribution was also seen among the CD8⁺ T cells that are labeled with tetramers complexed with EBV, Melan-A, and NA17-A peptides (Fig. 7). This pattern of expression is most consistent with that of early effector T cells [6].

Fig. 7.

Phenotypic analysis of CD8⁺ T cells in the ascites. Total CD8⁺ T cells were analyzed for expression of CD45RA, CD62L, CD27, and CD28. EBV, Melan-A, and NA17-A-specific CD8⁺ T cells, as determined by tetramer staining, were also gated on and analyzed for the expression of these surface markers. The percent cells found in each quadrant or gate (shown in the dot plots) for total CD8⁺ T cells as compared to the tetramer-specific CD8⁺ T cells are shown in table format

The tumor cells have no apparent defects in antigen presentation

The paradox of simultaneous presence of large numbers of activated effector T cells specific for melanoma antigens and abundant antigen-expressing tumor cells that persisted and were not rejected led us to investigate putative mechanisms of tumor escape in this patient. First, the tumor cells were analyzed for surface expression of MHC class I (HLA-A2) by flow cytometry and were found to express high levels (Fig. 8a). They were also examined for expression of antigen processing proteins by RT-PCR analysis. TAP1, TAP2, LMP2, LMP7, and LMP10 were all found at very high levels (Fig. 8b). The ability of the tumor cells to stimulate third party T cells was examined by pulsing them with EBV peptide and measuring the ability of CD8⁺ T cells from a normal HLA-A2⁺ donor to produce IFN-γ by ELISpot analysis. As shown in Fig. 8c, vigorous stimulation was observed, arguing that the HLA-A2 expressed was functional and also that the tumor cells were not expressing sufficient levels of inhibitory factors to globally block T cell activation. As soluble inhibitory factors could have accumulated in the ascites over time (indeed IL-10 was present by protein array analysis), ascites fluid was added to this in vitro stimulation culture at a 50% concentration. In fact, no inhibition of IFN-γ production by EBV-reactive T cells was observed when ascites fluid was added along with tumor cells (Fig. 8c), suggesting that production of large quantities of soluble inhibitory factors was not the dominant mechanism of tumor escape in this patient. Finally, to determine if melanoma antigens were effectively processed and delivered to HLA-A2 molecules, T cell lines raised against Melan-A and gp100 peptides were stimulated with the ascites tumor cells and analyzed by ELISpot. As shown in Fig. 8d, these cells responded extremely vigorously, arguing that processing and presentation were not compromised. The susceptibility of these tumor cells to lysis by CTL was not examined.

Fig. 8.

Antigen presentation in ascites tumor cells. a HLA-A2 expression by tumor cells. Tumor cell line cells were stained with FITC anti-HLA-A2 or isotype control and analyzed by flow cytometry. b RT-PCR of antigen presentation pathway components. cDNA (RT+) or mock cDNA made in the absence of RT (RT−) from normal donor PBMC, the tumor cell line, or TAP deficient T2 cells was analyzed for expression of TAP1, TAP2, LMP2, LMP7, and LMP10. 1 normal donor RT+, 2 normal donor RT−; 3 tumor cell line RT+, 4 tumor cell line RT−; 5 T2 cells RT+, 6 T2 cells RT−. c Purified CD8⁺ T cells from a normal donor were cultured together with tumor cell line cells that were either not loaded with peptide or pulsed with EBV peptide. The culture was performed in normal medium or in the presence of 50% ascites fluid. The number of IFN-γ spots/10⁵ CD8⁺ T cells plated is shown. d T cell lines specific for Melan-A and gp100 peptides were stimulated with the patient’s tumor cell line and ELISpot analysis was performed. Stimulation with K562-A2 cells is shown as a control

The antigen-specific ascites T cells have functional defects and do not express perforin

Inasmuch as the melanoma tumor cell line derived from the ascites of this patient seemed capable of processing and presenting antigen to third party T cells, we examined whether they would stimulate autologous ascites-derived CD8⁺ T cells to produce IFN-γ. Based on the frequency of defined tetramer-reactive CD8⁺ T cells specific for just Melan-A and NA17-A that were present, at least 5% of the cells were expected to be tumor reactive. However, the apparent frequency of functional T cells reactive against tumor cells as assessed by ELISpot was only 0.01%, a number comparable to the background level of spots seen with nonpeptide pulsed T2 cells as stimulators (Fig. 9a). These results argue that the melanoma antigen-reactive CD8⁺ T cells present in the ascites, at least with the level of class I MHC/antigenic peptide complexes expressed by autologous tumor cells, were relatively dysfunctional.

Fig. 9.

Functional analysis of antigen-specific CD8⁺ T cells. a Purified CD8⁺ T cells from ascites fluid were cultured together with T2 cells that were either not loaded with peptide or pulsed with EBV peptide, or with the melanoma tumor cell line cells. b CD8⁺ T cells from the ascites were cultured in the presence of T2 cells loaded with the indicated peptides or PMA + ionomycin for 24 h. The number of IFN-γ spots was analyzed for each well, and is shown as the number of spots/10⁵ total cells. c Perforin expression of antigen-specific T cells. Ascites cells were surface stained with antibodies against CD8 and with tetramers and then stained for intracellular perforin. Lymphocytes were gated on by size and granularity and are shown, with the percent of cells stained with EBV, Melan-A, and NA17-A tetramers, on the left. The perforin staining is shown on the right, with percent cells in each quadrant indicated in each dot plot

In some instances, tumor antigen-specific CD8⁺ T cells can respond to antigenic peptide despite an inability to recognize tumor cells directly [7–9]. To test this hypothesis, ELISpot analysis was performed with T2 cells pulsed with specific epitopes as stimulator cells. EBV peptide was used as a positive control. Interestingly, stimulation with Melan-A peptide did give detectable stimulation of ascites CD8⁺ T cells by ELISpot, although the frequency was still significantly lower than expected based on tetramer staining. However, NA17-A-specific responses still were not observed (Fig. 9b). T cells specific for the NA17-A peptide were easily derived from normal donor PBL (data not shown). Even after 3 weeks of in vitro stimulation with peptide-pulsed APCs and IL-2, no T cells producing IFN-γ in response to NA17-A peptide were detected, whereas EBV and Melan-A-specific T cells were successfully expanded by this method (data not shown). Thus, CD8⁺ T cells isolated from the ascites that were specific for Melan-A were able to proliferate in response to antigen-specific stimulation, whereas those specific for NA17-A appeared irreversibly hyporesponsive.

In addition to IFN-γ-producing capability, granule-mediated cytolytic activity is thought to be a critical effector function of anti-tumor T cells. We did not obtain sufficient numbers of CD8⁺ T cells to assess cytolysis directly. However, intracellular flow cytometry was performed to analyze expression of perforin. Although some perforin-expressing T cells were present among the ascites cells, the CD8⁺ T cells that stained with Melan-A or NA17-A peptide-bound HLA-A2 tetramers were perforin negative (Fig. 9c). These T cells thus would not be expected to display granule-mediated lytic activity.

Additional negative regulatory factors present

The fact that activated tumor antigen-specific CD8⁺ T cells present in the ascites of this patient were poorly responsive to autologous tumor cells and lacked perforin expression argued that these CD8⁺ T cells were exposed to a negative regulatory environment. Although classical anergy occurring in the absence of B7 costimulation can result in T cell hyporesponsiveness, prior exposure to CD25⁺ regulatory T cells [10–13], to tryptophan depletion via indoleamine-2,3-dioxygenase (IDO) [14, 15], to arginine depletion via Arginase I [16], or to engagement of the inhibitory receptor PD-1 by PD-L1 [17–19] also can lead to this state. We therefore examined the ascites cells for evidence of these factors. Real-time RT-PCR revealed expression of the regulatory cell-associated transcript FoxP3 (Fig. 10a, b). IDO transcripts also were present, as was expression of the PD-L1 gene. Flow cytometric analysis also demonstrated the expression of CD4⁺CD25^high cells that expressed intracellular FoxP3 (Fig. 10c), consistent with regulatory cells. These data indicate that numerous negative regulatory factors were present that could have contributed to the hyporesponsiveness of tumor-specific T cells in the peritoneal microenvironment of this patient.

Fig. 10.

Negative regulatory factors present in the ascites. a Real-time RT-PCR of total ascites cells. RNA extracted from the cells was analyzed for expression of Arginase I, FoxP3, IDO, PD-L1, and for β-actin by real-time RT-PCR. b Results for each of the transcripts shown in a were quantitated for relative expression compared to β-actin. c Ascites cells were surface stained with antibodies against CD3, CD4, and CD25 and permeabilized to stain with anti-FoxP3. Analysis was gated on CD3⁺ CD4⁺ cells that were either CD25^high or CD25^low, and analyzed for the percent FoxP3⁺ cells

Discussion

We have described a patient with metastatic melanoma in whom very large numbers of activated T cells were recruited to the peritoneal cavity where abundant malignant tumor cells also were present. When considered in terms of potential mechanisms of tumor escape from immune destruction, there clearly was not a failure of T cell trafficking to the target tumor site in this case. In fact, high concentrations of specific chemokines were present in the ascites fluid that likely mediated this inflammatory cell recruitment. It is interesting to note that the activated CD8⁺ T cells that accumulated were not exclusively tumor reactive, as EBV peptide-specific T cells also were present by tetramer analysis at a higher frequency than what was found in the blood.

The tumor cell line isolated from this patient showed no obvious global defect in antigen presentation ability, expressed normal levels of defined antigen processing genes, expressed numerous defined tumor antigens by RT-PCR, and was capable of stimulating established melanoma antigen-specific T cells. Thus, although some melanoma cell lines have well characterized defects in the expression of specific molecules involved in antigen processing or presentation [20, 21], this did not appear to be the case in the current patient. It is formally possible that these tumor cells are intrinsically resistant to cytolysis, a hypothesis that has not been examined directly. Even with the addition of ascites fluid to in vitro stimulation cultures, EBV-specific CD8⁺ T cells were activated normally by this melanoma cell line, arguing that acute exposure to these tumor cells and soluble factors from the ascites was not sufficient to inhibit T cell activation. These results do not rule out a possible role for low concentrations of soluble factors that may act in the close proximity of cell–cell contact. Nevertheless, these collective results suggest that dysfunction of the ascites-derived T cells was involved and likely explained inadequate tumor rejection.

While it is possible that discrepancies between the tetramer and ELISpot assays in our patient could be explained by low TCR avidity, this is unlikely to be the case, as the tetramers were mutated to lack CD8 binding and thus relied solely on TCR interactions, and the peptide concentration for APC pulsing in the ELISpot was high (50 μM) and should reveal low-avidity TCR-expressing cells. Yet, several exogenous factors that negatively regulate T cell activation were present in the malignant ascites microenvironment. Soluble IL-10 was detected in the ascites fluid, perhaps indicative of a nonlytic, tolerogenic T cell environment [22, 23]. CD4⁺ CD25⁺ regulatory T cells that express FoxP3 were also present. Regulatory T cells have been found to be abundant in the ascites tumor microenvironment of ovarian cancer patients which appears to confer negative prognostic value [13]. However, the chemokine CCL22 which has been found to preferentially recruit regulatory T cells [13, 24] was not detected in ascites fluid from this patient, suggesting that other chemokines would have been responsible for recruitment of regulatory T cells in this case. It is also of interest that high levels of IL-6 were present in the ascites fluid, which should antagonize Treg function. Nevertheless, T cell dysfunction was observed in our patient. Transcripts for IDO and PD-L1 also were observed among the ascites cells. IDO catabolizes tryptophan, which has been shown to be suppressive for T cell activation and to promote T cell apoptosis [14]. Inhibition of IDO appears to promote tumor rejection in some model systems [15]. PD-L1 appears to inhibit the effector phase of anti-tumor T cell function in preclinical models [17–19], and interfering with PD-L1/PD1-1 interactions also promotes tumor rejection in vivo [17, 18]. Which of these negative regulatory influences might have been dominant in our patient is not clear, and it is also possible that classical anergy due to TCR ligation in the absence of sufficient costimulation was involved. It is important to note that, although short-term exposure of T cells to the total ascites microenvironment might not have been inhibitory based on the brisk EBV peptide ELISpot activity observed, it is conceivable that the prolonged exposure to this microenvironment that would have existed within the peritoneum in vivo could have induced a T cell hyporesponsive state over time.

One encouraging observation in the present study was that, at least for the Melan-A-specific T cells, the apparent hyporesponsive state was reversible. These T cells were easily expanded with peptide-pulsed APCs and IL-2. This suggests that it may be possible to restore effector function of anti-melanoma T cells in vivo, perhaps by blockade of defined negative regulatory processes or by promoting T cell proliferation. The recent clinical trial success using adoptive transfer of T cell lines derived from TILs supports the potential for reversibility of this dysfunctional state [25]. However, effector function might not always be restored, as evidenced by the lack of growth of the NA17-A-specific T cells from our patient. The lack of IFN-γ production by NA17-A-specific T cells in a tumor-bearing environment has previously been described [26], suggesting that this could be a common observation, perhaps correlating with the lower level of NA17-A expression seen by quantitative RT-PCR analysis [27, 28]. The reason for more complete dysfunction of some T cells and not others is not clear, but indicates that distinct mechanisms may be operating even in the same patient.

In addition to the poor cytokine production by the ascites T cells in our patient, we found that neither the Melan-A nor the NA17-A-specific T cells present in the ascites expressed perforin. This could be due to a defect in the acquisition of lytic ability during CD8⁺ T cell differentiation [22], or rather could suggest that the effector cells have degranulated and need to receive additional proper stimulation in order to reacquire cytolytic machinery. Expression of FasL was not examined, and therefore the potential for lytic activity via Fas engagement cannot be ruled out. We recently observed that circulating antigen-experienced CD8⁺ T cells from normal donors also minimally express granule proteins, suggesting that this may be a frequent feature of primed human CD8⁺ T cells. Stimulation of these perforin-negative CD8⁺ T cells with anti-CD3 plus anti-CD28 mAbs induced granule acquisition after at least 3 days in vitro, in a cell cycle-dependent fashion (unpublished observation). Thus, introducing high levels of B7 costimulation, or other strategies to promote vigorous proliferation of effector T cells within the tumor microenvironment, may be necessary to restore cytolytic potential.

Taken together, our data support a tolerogenic melanoma tumor microenvironment in which tumor-antigen specific CD8⁺ T cells are unable to acquire and/or maintain full effector function. Our observations also suggest that patients who receive antigen-specific immunotherapy should be carefully monitored not only for presence of tumor antigen-specific T cells in the circulation, but also for the differentiation status and function of those T cells within the target tumor microenvironment.