Carcinoma of the exocrine pancreas has an especially poor prognosis. The overall 5-year survival rate is <1% with a median survival of 4–6 months. Even after surgical intervention with a curative intention, the 2-year survival rate in specialized centers is at best 20%.1 Therefore, alternative treatment strategies are needed, and current investigations focus, among others, on the development of immunomodulatory approaches.2, 3 However, further understanding of the mechanisms how tumour cells interact with the immune system is required to optimize immunotherapy approaches of pancreatic carcinoma.
Natural cytotoxicity mediated by the innate immune system contributes substantially to anti-tumour immune responses.4 There is increasing evidence that natural cytotoxicity is exerted not only by NK cells but also by γδ T cells. This is at least partly due to interactions of the NKG2D receptor on cytotoxic lymphocytes with its tumour-expressed ligands thereby even overcoming inhibitory signals by MHC class I molecules.5, 6 Although most tumour associated ligands engaging activating NK cell receptors are unknown, the MHC class I-related ligands of NKG2D have been extensively studied in the recent years. Among the human NKG2D ligands (NKG2DL) are the stress-inducible, MHC-encoded surface glycoproteins MICA and MICB, which are broadly expressed by epithelial and hematopoietic tumors,7, 8, 9 and the more recently described UL16 binding proteins (ULBPs), a multigene family with at least 6 functional members.6, 10 NKG2D associates in humans with the DAP10 adaptor protein allowing transduction of activating signals.11 The discovery of NKG2D-mediated immunity following binding to its stress-inducible ligands has caused considerable scientific interest prompting a “renaissance of the immune surveillance &#128;hypothesis.”12
Recently, it has been described that tumor cells shed MICA in a soluble form (sMICA) and by that might escape from NKG2D-mediated surveillance.13 Two mechanisms have been proposed, by which, release of soluble MIC (sMIC) could reduce immunogenicity of tumor cells: shedding of MICA leads to decreased NKG2DL expression levels on the tumor cell surface, which directly affects tumor cell lysis, since it has been shown that the strength of an anti-tumour immune response in mice was critically dependent on NKG2DL surface levels.14 Secondly, high levels of sMIC detectable in sera of patients with certain gastrointestinal malignancies13, 15, 16 were reported to cause systemic down regulation of NKG2D surface expression on CD8 αβ T cells and NK cells thereby impairing lysis of tumor cells.15, 16 In this study, we propose a third mechanism by which release of sMIC enables immune escape of tumor cells, namely direct blocking of NKG2D by sMIC regardless of receptor down regulation.
Up to now, nothing is known regarding the expression, release and function of NKG2DL in pancreatic carcinoma. In this study, we investigated the expression and release of sMIC in patients with pancreatic carcinoma. In addition, we investigated the susceptibility of pancreatic carcinoma cells to NKG2D-mediated lysis by cytotoxic lymphocytes with a focus on γδ T cells and determined the effect of pancreatic carcinoma-derived sMIC on natural cytotoxicity. Furthermore, we analyzed the effect of immunotherapy with IFN-α on NKG2D-mediated lysis of pancreatic tumor cells.
Material and methods
Reagents and monoclonal antibodies
The mAb, specific for human HLA-ABC (clone G46-2.6), γδ TCR (clone B1) and NKG2D (clone 14918), were obtained from Becton Dickinson, Heidelberg, Germany. The mAb AMO1 (anti-MICA), BMO1 (anti-MICB), AUMO1 (anti-ULBP 1), BUMO2 (anti-ULBP 2), the mAb BMO2 (anti-MICB) and BAMO3 (anti-MICA/B), the mAb CUMO2 (anti-ULBP 3) and recombinant sMICA*04 and sMICB*02 were a kind gift from H. Salih (Tübingen, Germany) and produced as previously described.8 Blocking anti-NKG2D monoclonal antibody (mAb) was obtained from R&D Systems (Wiesbaden, Germany) and mouse IgG1 was obtained from SIGMA (Munich, Germany).
Blood samples from patients with adenocarcinoma of the pancreas were collected at the Department of Surgery. Diagnosis was confirmed by histology performed by an independent pathologist. The study was approved by the local ethics advisory board and patients were enrolled in the study after written informed consent was obtained. Samples were obtained prior to surgical intervention or chemotherapy and sera were frozen at −80°C until further analysis.
Panc 1, DAN-G, PatScl (pancreatic carcinoma), NK 92 and K562 were purchased from DSMZ (Braunschweig, Germany). The cells were maintained in RPMI 1640 supplemented with 10% foetal calf serum (FCS, PAA, Coelbe, Germany). NK 92 cells were stimulated every third day with 100 U/ml IL-2.
For the preparation and propagation of γδ T cells, peripheral blood mononuclear cells were obtained from buffy coats of healthy donors by Ficoll density gradient centrifugation. Nonadherent cells were taken for positive selection of γδ T cells, which was performed using magnetic activated cell sorting from Miltenyi Biotec (Bergisch Gladbach, Germany), according to the manufacturer's instructions, yielding a purity of >90% γδ T cells. Cells were cultured in RPMI 1640 supplemented with 10% foetal calf serum, 25 mM Hepes. After 24 h of incubation, 50 ng/ml of anti-CD3 (Orthoclone OKT 3, Cilag GmbH, Sulzbach, Germany) and 100 U/ml IL-1β and 300 U/ml IL-2 (Roche, Mannheim, Germany) were added. Every third day, fresh complete medium including cytokines with additional 300 U/ml IL-2 was added.
Detection of sMICA and sMICB was performed using a previously described sandwich-ELISA.8 In brief, for detection of sMICA, the monoclonal antibodies AMO1 and BAMO3 were used at 5 and 1 μg/ml, respectively, with recombinant sMICA*04 as a standard. For determination of the levels of sMICB, mAb BAMO1 and BMO2 were used at 2 and 1 μg/ml, respectively, with recombinant MICB*02 as a standard. Both assays were processed using anti-mouse IgG2a-HRP (1:8.000) and developed using the TMB peroxidase substrate system (KPL, Gaithersburg, MD). The absorbance was measured at 450 nm.
Flow cytometric analysis
Human pancreatic cancer cell lines and γδ T cells were stained using mAb directed against human γδ TCR, δ1 and δ2 receptor, NKG2D, MICA, MICB and ULBP 1–3 or the respective isotype control followed by incubation with the goat anti-mouse-PE conjugate and finally analyzed on an Epics XL (Beckman Coulter, Krefeld, Germany).
A standard chromium release assay was used to determine the lytic activity of γδ T cells. In brief, tumor cells were labeled with 100 μCi 51Cr for 2 h. Ten thousand target cells per well were incubated in triplicates with effector cells. After 4 h, the supernatants were collected and counts per minute were determined (Packard, Dreieich, Germany). The ratio between maximal and spontaneous release was generally >5.
Inhibition of sMIC-NKG2D interaction by antibodies
To prevent binding of sMIC to NKG2D, 15 μl/ml of the anti MICA/B antibody BAMO1 that blocks MIC/NKG2D-interaction or isotype control was added to MIC containing sera, 5 min prior to addition of sera in the cytotoxicity assay. Blockage of NKG2D and ULBP 3 was performed by adding 15 μl/ml anti-NKG2D or anti-ULBP 3, 5 min prior to the coculture to the effector cells or to the target cells, respectively.
Nonparametrical analysis (Wilcoxon-test), paired t test and Pearson correlation on SPSS 11.5 were used to analyze statistical significance where appropriate.
All 49 patients studied received their ultimate diagnosis based upon clinical data as well as morphological investigations of pancreas biopsies. Twenty-six were male, 23 were female; the median age was 66 years with a range from 38 to 83 years. Five patients had unknown tumor stage, one patient had T2, 30 patients T3 and thirteen T4. The carcinoma of 3 patients was well differentiated, whereas 19 had moderately and 10 poorly differentiated cancer (unknown in 17 patients).
Presence of soluble MIC in sera of patients withpancreatic carcinoma
Since elevated serum levels of sMIC have been reported in various malignancies, we screened sera from 49 patients for the presence of sMICA and from 35 patients for the presence of sMICB by ELISA. Significant levels of sMICA and sMICB were detected in the patients' sera as opposed to the sera of healthy individuals (Table I). Also, the sMICA and B levels correlated significantly with the tumor stage. Furthermore, sMICB levels were significantly more elevated in patients with poorly differentiated tumors, which could not be found for sMICA levels.
Table I. Levels of sMIC in Healthy Volunteers and Patients with Pancreatic Cancer
Flow cytometric analysis of pancreatic carcinoma and γδ T cells
To evaluate the expression of NKG2DL on pancreatic carcinoma cell lines, we used a set of NKG2DL-specific mAb for flow cytometry. All cell lines tested (DAN-G, Panc 1, PatScl) constitutively expressed MICA and MICB as well as ULBP 1–3. The expression ranged from 20 to 30%, for mean fluorescence see Figure 1. HLA-ABC was expressed on Panc-1 and DAN-G but not on PatScl cells. Mean fluorescence for PatScl was 1.0 with and without IFN-α stimulation, for Panc 1 MFI was 3.5 ± 0.3 in untreated cells and 7.7 ± 2.6 after IFN-α stimulation. DAN-G showed the highest expression with 12.7 ± 3.6 in untreated cells and 21.7 ± 1.8 after IFN-α stimulation. All tumor cells were negative for HLA-DR (data not shown).
γδ T cells separated from peripheral blood from healthy donors and expanded in the presence of IL-2 over 14 days express γδ TCR [(92 ± 6%)], NKG2D [(79 ± 13)%], the γ9/δ2 receptor [(60 ± 16)%], and δ1 receptor [(18 ± 4)%; Fig. 2]. After incubation with MIC positive serum for 4 h, the expression of NKG2D on γδ T cells decreased insignificantly to (72 ± 12)% (data not shown). NK92 cells in our hands did not express CD3, CD4 or CD8. However, the majority expressed NKG2D as well as NKG2A (99%) and CD11a (88%). Inhibitory receptors such as CD158a/b were expressed by only 15 and 13% of cells, respectively.
Influence of soluble MIC in patient serum on natural cellular cytotoxicity
To analyze whether natural cytotoxicity is influenced by sMIC molecules in patient's serum, we performed cytotoxicity assays in the presence of sMIC-containing patient serum or sMIC negative serum from healthy volunteers. The lytic activity of NK cells against K562 and against pancreatic carcinoma cells was nearly completely abrogated in the presence of sMIC positive serum [from (40 ± 0)% to <4%, p < 0.01, and from (18 ± 1)% to <2%, p < 0.01, at effector to target ratio (E:T) of 40:1, respectively, Fig. 3a]. This was also true for the lytic activity of γδ T cells: In agreement with the data obtained with the NK cells, γδ T cells showed cytotoxic activity against pancreatic carcinoma cells in the presence of the sMIC negative serum, which could be abrogated completely by the addition of sMIC-containing serum [from (21 ± 6)% to maximal (6.5 ± 3.3)% for PatScl cells at E:T 30:1, p < 0.01 Fig. 3b].
As a possible reason for the diminished cytolytic activity might be the down-regulation of NKG2D in the presence of sMIC, we determined the expression of NKG2D on γδ T cells after 4 h of incubation with sMIC positive sera. No significant difference on NKG2D expression could be shown.
Effect of blocking soluble MIC in patient serum by antibodies on cellular cytotoxicity
To further confirm and extent these results, we performed experiments in which we prevented binding of sMICA to NKG2D on γδ T cells by addition of blocking anti-MICA antibody using patient serum that was sMICA positive and sMICB negative. Like in previous experiments, presence of high levels of sMICA in serum nearly completely abrogated the lysis of pancreatic carcinoma cells by γδ T cells [from (79 ± 3)% to (5 ± 3)% at an E:T ratio of 40:1; p < 0.001]. Addition of the anti-MICA mAb AMO1 to the sMICA-positive patient serum prevented NKG2D-sMICA interactions and thus restored the lytic capacity of the γδ T cells [(74 ± 10)% lysis; p < 0.001; Fig. 4]. To exclude an unspecific interaction with the antibody, anti-ULBP 3 was added as an irrelevant antibody to show no significant difference in lysis [(65 ± 7)% at E:T 40:1]. A similar reduction of γ/δ T cell-mediated lysis could be achieved by direct blockage of the NKG2D receptor [(11 ± 1)%; Fig. 4].
Influence of IFN-α on NKG2DL surface expression and sMIC release in pancreatic carcinoma
As IFN-α is presently clinically evaluated in combination therapy of pancreatic carcinoma,2 we investigated the effect of IFN-α on expression and release of NKG2DL. Three different pancreatic cancer cell lines were incubated for 24 h with IFN-α and the culture supernatants were analyzed by ELISA. Interestingly, we found no difference in the detectable amount of released sMIC in treated compared to untreated cell lines. Student's t-test did not reveal statistically significant differences between the detectable sMICA and sMICB levels in culture supernatants of treated versus untreated pancreatic carcinoma cells (123 ± 98 pg sMICA/ml/106 cells versus 126 ± 122 pg/ml/106 cells and 173 ± 72 pg sMICB/ml/106 cells versus 162 ± 70 pg/ml/106 cells; Fig. 5a–5c).
Subsequently, the expression of MIC and ULBP molecules was determined by flow cytometry. Stimulation with IFN-α led to a significant increase of MICA, MICB, ULBP 1–3 expression in all 3 cell lines (p < 0.001; Fig. 1a–1c).
Cytotoxicity of γδ T cells against IFN-α-treated tumour cells
Since the differential modulation of NKG2DL expression and release following treatment with IFN- α poses an explanation for the beneficial immunomodulatory effects of IFN-α treatment, we tested whether the increase in NKG2DL expression was mirrored by an enhanced susceptibility of pancreatic carcinoma to γδ T cell-mediated killing. The pattern of increase in NKG2DL expression differed in the investigated cell lines. ULBP 1 and ULBP 3 was especially upregulated in Panc-1 cells, ULBP 2 and MIC in PatScl cells, which did not respond regarding ULBP 1 and ULBP 3. DAN-G showed for all NKG2DLs but MICA moderate upregulation. Two pancreatic carcinoma cell lines were tested in a cytotoxicity assay with or without preincubation with IFN-α. Enhanced NKG2DL expression following stimulation with IFN-α regardless of the particular involved ligands was mirrored by increased lysis [from (27 ± 6)% without treatment to (34 ± 12)% after stimulation with IFN-α and from (29 ± 8)% to (42 ± 1)% at E:T 20:1 for DAN-G and PatScl, respectively, Fig. 6]. Blockage of NKG2D on γδ T cells by addition of anti-NKG2D antibody resulted in a significant decrease of cytotoxicity [(10 ± 2)% and (11 ± 2)% at E:T 20:1 for DAN-G and PatScl, respectively, p < 0.001].
MICA/B and other ligands of NKG2D are broadly expressed on various tumors. Recently, it was demonstrated that tumor cells release the NKG2D ligands MICA and MICB in soluble form by proteolytic shedding from the tumor cell surface by metalloproteases.13 Elevated levels of sMIC can be detected in sera of patients with certain gastrointestinal malignancies8, 13, 16 and various other tumor entities such as prostate cancer,17 neuroblastoma18 and various hematopoietic malignancies.8 In prostate carcinoma, high serum levels of sMICA correlate with tumor grade and invasiveness/presence of metastasis.17 Here, we show that the same is true for pancreatic carcinoma: poor differentiation and high tumor stage correlate significantly with the level of sMICB. Also, larger tumors are associated with higher levels of sMIC. Thus, our data show that a worse prognosis in pancreatic carcinoma is associated with a high level of soluble NKG2D ligands. Although this is a highly significant finding, sMIC seems not be specific enough to serve as a prognostic tumor marker.
However, this could well be one of the mechanisms by which tumor cells escape from the immune surveillance. Release of soluble NKG2DL will inhibit the interaction with the activating receptor and protect the tumor cell from cytolysis if it is NKG2DL positive. We confirmed that pancreatic carcinoma cells have a relevant surface expression of NKG2DL, which renders them to be possible targets for effector cells expressing NKG2D (Fig. 1a–1c).
Pancreatic carcinoma cells are therefore targets for various effector cells expressing NKG2D. These include NK cells, some CD8+ αβ T cells and γδ T cells. γδ T cells occur in the intestine (δ1 TCR) and in the peripheral blood (δ2 TCR). Recently, it has been proposed that MIC expression delivers both the T cell receptor (TCR) dependent signal 1 and the NKG2D dependent costimulatory signal 2 for a subset of Vδ1 γδ T cells.19 However, there is a strong expression of NKG2D in peripheral blood γ2/δ2 T cells as well. Here, we show that polyclonal γδ T cells exhibit a marked lysis of pancreatic carcinoma cells, similar to NK cells (Fig. 3). These data are in accordance with Kabelitz et al. who found an NKG2D-dependent lysis of pancreatic carcinoma cells by γδ T cell lines.20
To investigate the effect of the release of sMIC on the cytotoxicity of NK cells and γδ T cells, chromium release assays were performed in the presence of sMIC positive serum (Fig. 3). Here, cytotoxic activity of NK cells and γδ T cells was significantly reduced in the presence of sMIC positive patients' sera. This tumor escape may be due to several mechanisms: Firstly, shedding of MIC molecules reduces cell surface expression of NKG2DL on tumor cells lead to diminished immunostimulatory signals for cytotoxic lymphocytes.14, 21 In addition, NKG2D-mediated immunity is proposed to be impaired by sMIC via down modulation of NKG2D on effector cells.15, 16 To investigate if this occurs in our setting, we incubated γδ T cells with sMIC positive sera for 4 h and re-evaluated the expression of NKG2D. In our experiments, no significant difference could be detected after this 4 h incubation as it is used in cytotoxicity assays; thus this effect seems unlikely.
However, the fact that NKG2D down-regulation was observed by Doubrovina et al. earliest after 4 h and reached its maximum after 24 h15 indicates that in different settings this might contribute to the direct inhibition of sMIC. We propose a direct inhibition. To confirm this, an antibody against MICA was added to the serum prior to the chromium release assay. Thereby the reduction in cytotoxic activity could be reversed significantly.
These data are in accordance with Kabelitz et al. who also found lysis of pancreatic carcinoma cells by γδ T cell lines.20 Their data suggest a TCR dependent lysis if γδ T cells are stimulated by phosphoantigens. However, in their hands lysis of pancreatic cancer cells by γδ T cells, which are not stimulated with phosphoantigens, appears to be independent of the TCR. Our experiments appear to confirm this finding, since our γδ T cells were not stimulated by phosphoantigens and their activity could be nearly completely inhibited by blocking NKG2D. This suggests a TCR independent lysis of pancreatic cancer cells by γδ T cells not stimulated by phosphoantigens.
New therapies for pancreatic carcinoma focus on vaccination, gene therapy and especially immunotherapy using immunomodulatory cytokines like IFN-α.2, 3 IFN-α is presently evaluated in several studies with patients with pancreatic carcinoma.2, 22 However, the mechanism by which IFN-α in pancreatic carcinoma mediates its beneficial effect remained unclear. It is well known that the activity of NK cells is regulated by various cytokines, among them IFN-α,23 and recent studies have implicated IFN-α in the induction of NKG2DL expression, especially on dendritic cells leading to subsequent activation of NK cells.24, 25 We therefore investigated the effect of IFN- α treatment on NKG2DL expression on pancreatic carcinoma cells and found that treatment with this cytokine induced a marked and statistically significant induction of both MIC molecules and ULBP 1–3 on the cell surface (Fig. 1a–1c). To exclude that the potentially beneficial effect of induction of NKG2DL surface expression was hampered by enhanced release of NKG2DL in soluble form following IFN-α treatment we determined sMIC levels in the culture supernatants of the respective pancreatic carcinoma cells by ELISA. Although we found considerable levels of both sMICA and sMICB in supernatants of all investigated cell lines, treatment with IFN-α did not significantly alter the detectable levels of sMICA and sMICB (Fig. 5a–5c). This suggests that IFN-α treatment of pancreatic carcinoma induces cell bound but not sMIC expression thereby augmenting immunogenicity of the treated tumor cells for natural cytotoxicity. To determine the functional significance of increased NKG2DL surface expression, we performed cytotoxicity assays using pancreatic tumor cell lines with or without treatment with IFN-α and γδ T cells as effectors. Increased NKG2DL surface expression was mirrored by a significant increase of cytotoxicity, and lysis of tumor cells was nearly completely abrogated by addition of a blocking NKG2D antibody (Fig. 6). These results indicate that the beneficial effects of IFN-α treatment of pancreatic carcinoma might, at least in part, be due to increased NKG2DL surface expression and subsequent enhanced lysis by γδ T cells.
Taken together, our data indicate that elevated sMIC serum levels correlate with tumor stage, grade and differentiation of pancreatic carcinoma and implicate an important role for NKG2D-mediated immunity mediated by NK cells and, importantly, also by γδ T cells in pancreatic carcinoma. Furthermore, we demonstrate that presence of sMIC directly inhibits cellular cytotoxicity of NK cells and γδ T cells, which, in addition to reduction of NKG2DL surface levels and NKG2D receptor down-modulation on cytotoxic lymphocytes, constitutes a third mechanism by which release of sMIC mediates the escape of tumor cells from NKG2D-mediated immune surveillance. Since IFN-α treatment induces NKG2DL expression without increasing sMIC levels our results indicate that the beneficial effects of IFN-α immunotherapy might be, at least in part, due to modulation of NKG2D-mediated immunity and provide a rationale for further clinical studies using IFN-α in treatment of pancreatic carcinoma.
The authors are grateful to Helmut Salih and Alexander Steinle from the University of Tübingen for helpful cooperation and detection of sMIC levels in patient's sera.