The proinflammatory cytokine macrophage migration inhibitory factor (MIF) stimulates tumor cell proliferation, migration, and metastasis; promotes tumor angiogenesis; suppresses p53-mediated apoptosis; and inhibits antitumor immunity by largely unknown mechanisms. We here describe an overexpression of MIF in ovarian cancer that correlates with malignancy and the presence of ascites. Functionally, we find that MIF may contribute to the immune escape of ovarian carcinoma by transcriptionally down-regulating NKG2D in vitro and in vivo which impairs NK cell cytotoxicity toward tumor cells. Together with the additional tumorigenic properties of MIF, this finding provides a rationale for novel small-molecule inhibitors of MIF to be used for the treatment of MIF-secreting cancers.
Epithelial ovarian carcinoma (OvCA)3 is the most common cause of death from gynecological malignancy. Even with extended surgery and chemotherapy, 5-year survival rates do not exceed 20–40% (1). Only noninvasive OvCAs of low malignant potential, here designated as borderline tumors, bear a more favorable prognosis (2). Because long-term survival strongly correlates with favorable immunological parameters (3), immunomodulatory therapies would be of major clinical interest. On the one hand, they should seek to relieve tumor-induced immunosuppression (4, 5); on the other hand, they need to consider inflammation, which not only is central to tumorigenesis (6) but also contributes to the formation of ascites (7, 8), thus correlating with a poor prognosis (9).
Macrophage migration inhibitory factor (MIF), a cytokine that promotes both processes (3), has initially been described as a mediator responsible for lethal shock syndrome after exposure to bacterial endotoxins (10). MIF has also been found to be critical for wound healing (11). A polymorphism in the MIF promoter could lead to high levels of MIF that antagonize glucocorticoid functions and contribute to autoimmune diseases like rheumatoid arthritis, inflammatory bowel disease, systemic lupus erythematosus and atopic dermatitis (12, 13). In addition, MIF can recruit inflammatory cells and promote atherosclerosis via noncognate binding to chemokine receptors CXCR2 and CXCR4 (14). In normal ovarian tissue, MIF is secreted by granulosa cells and is found in follicular fluids (15). In tumor biology, MIF has been proposed as a biomarker for prostate cancer (16). Further, MIF plays a critical role for angiogenesis, stimulates tumor cell migration, suppresses p53 activity (17, 18), and activates cyclin D1 and E2F transcription factors (Ref. 19 ; summarized in Ref. 20). The MIF-CD74 ligand-receptor complex (21) activates downstream nonreceptor tyrosine kinases via CD44 (22), a coreceptor that has been proposed as a tumor stem cell marker for breast cancer (23), leukemia (24), and other malignancies. The importance of MIF for tumorigenesis is strikingly demonstrated by the resistance of MIF-deficient fibroblasts toward malignant transformation induced by c-myc, H-ras, or dominant negative p53 (25).
MIF also impairs antitumor immunity by inhibiting CTL and NK cell responses (26, 27), an effect proposed to be caused by MIF-induced T cell activation followed by activation-induced cell death (28), a mechanism that might reconcile immune inhibitory with proinflammatory potential. However, activation of NK cells by MIF has not been observed, which suggests that the MIF-mediated inhibition of NK cells is effected via a different mechanism (26).
The capacity of NK cells to potentially clear malignant cells depends on the recognition of stress- or transformation-induced molecules, most notably ligands for the activating NK cell receptor NK group 2D (NKG2D;4). These ligands (MICA/B, ULBP1-4) are induced by DNA damage (29) and are found on virus-infected and tumor cells (30, 31, 32). In human cells, ligation of NKG2D leads to the activation of the adaptor protein DAP10. DAP10 signaling initiates a perforin-mediated cytolytic response (30, 32) that can lead to NK cell-mediated tumor clearance without prior activation or Ag-specific tumor cell recognition. CD8+ αβ T cells (CTL), in contrast, must be activated by their cognate Ag presented on MHC class I in the presence of costimulation which can also be provided via NKG2D (33). In mice, NK cell activation and T cell costimulation are mediated via two different isoforms of NKG2D that recruit specifically either DAP10 or DAP12. In humans, however, the NKG2Dshort-DAP12 pathway does not appear to exist (34). The relevance of NKG2D-mediated reactions against OvCAs has recently been demonstrated when adoptive transfer of transgenic T cells expressing a chimeric NKG2D-CD3 ζ receptor induced the rejection even of late-stage tumors. What makes this strategy particularly appealing is that it does not depend on the presence of known tumor Ags because NKG2D ligands provide both TCR activation and costimulation (35). However, both murine and human tumors can counteract NKG2D-dependent antitumor immune reactions by the shedding of NKG2D ligands (36) or by the secretion of cytokines that induce a down-regulation of NKG2D (37, 38).
We here provide evidence that MIF is significantly overexpressed in OvCA and show that MIF may contribute to the inhibition of antitumoral CD8+ T and NK cells by down-regulation of NKG2D in vitro. These findings may be of particular relevance because small-molecule inhibitors of MIF are in preclinical development (39).
Materials and Methods
All tumor tissues and ascites were obtained in the Department of Obstetrics and Gynecology of the University Hospital (Würzburg, Germany). Patients were between 25 and 82 years of age (median age, 59.5 years). Solid tumor tissues and ascites-derived cells were assessed by at least two experienced pathologists. Of the 38 epithelial ovarian malignancies included, 21 were classified as tumors of the serous-papillary subtype, 11 as mucinous, and 6 as endometrioid. Noninvasive OvCAs of low malignant potential, so-called ovarian borderline tumors, bear a more favorable prognosis and were therefore regarded separately (n = 19). The experiments were approved by the local ethics committee, and all patients gave informed written consent.
Primary tumor cells were derived from ascitic fluid from ovarian cancer patients. Cells were centrifuged, washed with PBS, and transferred to MCDB105-M199 medium (PAA) supplemented with 10% heat-inactivated FCS (Biochrom), penicillin (100 IU/ml), streptomycin (100 μg/ml), and 0.02% sodium pyruvate (all from PAA). Nonadherent cells were removed by washing after 72 h. The immortalized cell lines SK-OV-3 (derived from a 64-year-old Caucasian woman diagnosed with ovarian adenocarcinoma, American Type Culture Collection (ATCC) HTB-77), PA-1 (derived from ascites of a 12-year-old Caucasian diagnosed with ovarian teratocarcinoma, ATCC CRL-1572) and OAW-42 (derived from ascites of a patient diagnosed with ovarian cystadenocarcinoma, European Cell Culture Collection 85073102) were cultured in RPMI 1640 with 10% FCS.
Generation of cell culture and tissue supernatants (SN)
For the generation of SNs, 105 cells/well were seeded in a 6-well plate and adhered overnight before the cell culture medium was removed and replaced by 2 ml of fresh serum-free RPMI 1640. SNs were collected at 72 h after the medium exchange. Tissue SNs were obtained by placing 2 mg of freshly resected tumor for 72 h into 2 ml of fully supplemented RPMI 1640.
Generation of MIF small interfering RNA and pcDNA3-MIF transfectants
The pSUPER plasmid was obtained from R. Agami (The Netherlands Cancer Institute, Amsterdam, The Netherlands) (40). A puromycin cassette was inserted into the Nae1 site. The MIF-specific oligonucleotide sequences GATC CCCTCAACTATTACGACATGAAttcaagagaTTCATGTCGTAATAGT TGATTTTTGGAAA and TCGATTTCCAAAAATCAACTATTACGACATGAAtctcttgaaTTCATGTCGTAATAGTTGAGGG (nucleotides 385- 403) were obtained from Sigma Genosys and cloned into the BglII and SalI sites of pSUPER. The MIF-specific parts of the sequences are bold and underlined. For the generation of stable small interfering MIF transfectants, pSUPERpuro control or small interfering MIF plasmids were introduced using Transfectin transfection reagent (Bio-Rad). The cells were selected in medium containing 2 μg/ml puromycin (Carl Roth).
Total RNA was prepared using the RNeasy system (Qiagen) and transcribed with the RevertAid First-Strand cDNA Synthesis kit (Fermentas). For real-time PCR, cDNA amplification was monitored using SybrGreen chemistry on the ABI PRISM 7500 Sequence Detection System (Applied Biosystems). The conditions for these PCR reactions were: 40 cycles at 95°C for 15 s, 60°C for 1 min, using the following specific primers: 18S up, 5′-CGGCTACCACATCCAAGGAA-3′ (nucleotides 450–469); 18S down, 5′-GCTGGAAT TACCGCGGCT-3′ (nucleotides 636–619); MIF up, 5′-GCCCGGACAGGGTCTACA-3′ (nucleotides 272–289); MIF down, 5′-CTTAGGCGAAGGTGGAGTTGTT-3′ (nucleotides 349–328); NKG2D up, 5′-TCTCGACACAGCTGGGAGATG-3′; NKG2D down, 5′-GACATCTTTGCTTTTGCCATCGTG-3′. Data were analyzed using the ΔΔCT method for relative quantification. Briefly, threshold cycles (CT) for 18S rRNA (reference and MIF) were determined in duplicates, and the relative change (rI) in copy numbers was calculated according to the formulae rI = 2 exp[(CTMIFhealthy tissue − CT18Shealthy tissue) − (CTMIFOvaCA − CT18SOvCA)] or rI = 2 exp[(CTNKG2Duntreated PBMC − CT18Suntreated PBMC) − (CTNKG2DMIF-treated PBMC − CT18SMIF-treated PBMC)].
Dissociation curves and PCR products were analyzed to confirm the presence of a single specific PCR product at the expected size.
All tissue specimens were from the tumor bank of the University of Würzburg School of Medicine (Würzburg, Germany) where they had been evaluated by at least two pathologists in routine diagnostics. Paraffin-embedded tissue samples of ovarian epithelial carcinomas, borderline tumors, and normal ovaries were cut at 2 μm, placed on slides (Superfrost; Langenbrinck), deparaffinized with xylene, and rehydrated in a descending alcohol sequence. Ags were unmasked in 10 mM sodium citrate buffer in a microwave oven. Endogenous peroxidases were left to react with 3% H2O2 in methanol for 10 min. Slides were washed in PBS, and nonspecific binding was blocked with 1% goat serum. Subsequently, slides were incubated for 1 h with the monoclonal mouse anti-MIF Ab (mAb289; R&D Systems) diluted in commercial Ab diluent (DakoCytomation) at 1/100. Biotinylated anti-mouse immunoglobulins and HRP-conjugated streptavidin (both from DakoCytomation) were used according to the manufacturer's protocol. Stainings were developed for 5–10 min with diaminobenzidine. Nuclei were counterstained with hematoxylin. Stained sections were then dehydrated by washing in graded ethanol and embedded in Vitro Clud (Langenbrinck).
Lysates were prepared from 106 cancer cells or from 10-μm-thick slices of snap-frozen tumor or control tissue. Protein lysis buffer contained 50 mM Tris (pH 7.4), 0.25 M NaCl, 0.2% Tween, 10 μg/ml leupeptin, 10 μg/ml, PMSF, and 2 μg/ml aprotinin. Protein concentrations were determined using RotiQuant Universal (Carl Roth). 10 μg of lysate per lane were separated. Western blotting was performed using 15% SDS-polyacrylamide gels in a Mini V8.10 gel chamber (Whatman). Proteins were transferred to nitrocellulose membranes (Whatman) that were then blocked with Roti-Block (Roth), incubated with 0.5 μg/ml mAb289 mouse anti-human MIF Ab (R&D Systems), washed in PBS containing 0.05% Tween 20, and developed with HRP-conjugated goat anti-mouse Ab (Santa Cruz Biotechnology) and luminol solution (88.5 mM Tris-HCl (pH 8.6), 0.2 mg/ml luminol, 1 mg/ml p-hydroxycoumarinic acid, and 0.01% H2O2). β-Actin staining (Abcam Ab8226) was performed as loading control.
The concentration of soluble MIF was determined by ELISA using mouse anti-human MIF mAb289 as capture and biotinylated goat anti-human MIF BAF289 as detection Ab (both from R&D Systems) in a MaxiSorp 96-well plate (Nunc). Streptavidin-HRP (Immunotools) and diaminobenzidine were used for color development. Optical density was recorded at 450 nm in a Sunrise microplate reader (Tecan) and quantified relative to a standard row. Both samples of ascitic fluids and cell culture SNs were examined. IFN-γ and TNF-α levels were also determined by ELISA using matched Ab pairs from Immunotools according to the manufacturer's instructions.
Flow cytometric analysis of NKG2DL expression levels
OvCA cells (106/sample) were detached with Accutase (PAA), blocked, and stained with AMO1 IgG1 anti-MICA, BAMO1 IgG1 anti-MICA/B, BMO2 IgG1, anti-MICB, AUMO1 IgG1 anti-ULBP1, BUMO1 IgG1 anti-ULBP2, and CUMO3 IgG1 anti-ULBP3 as described previously (38).
NK cell preparation and lysis assays
PBMC were obtained from healthy volunteers by density gradient centrifugation (Biocoll; Biochrom). Monocytes were depleted by adherence. PBL were cultured on irradiated (30 Gy) RPMI 8866 feeder cells to obtain polyclonal NK cell populations (41). Where indicated, 10 μg/ml blocking anti-MIF Ab mAb289 (42) or irrelevant control IgG (Immunotools) and/or 500 μl of cell culture SN from primary ascites-derived OvCA cells or from MIF-short hairpin RNA or control transfected SK-OV-3 ovarian cancer cells were added on day 8. After 48 h, NK cells were harvested and used in different killing assays: Cytotoxicity against primary ascites-derived OvCA cells was assessed in 4-h 51Cr release assays in the absence or presence of mAb used at 10 μg/ml. NK cells were pretreated with normal human IgG to prevent Ab-dependent cellular cytotoxicity before they were coincubated for 4 h with 1 × 104 51Cr-labeled target cells per well at various E:T ratios. Spontaneous 51Cr release was determined by incubating the target cells with medium alone. Maximum release was determined by adding Nonidet P-40 (2%). The percentage of 51Cr release was calculated as 100 × [(experimental release − spontaneous release)/(maximum release − spontaneous release)].
For assays involving SK-OV-3 OvCA cells, NK cells were labeled with PKH-26/Vibrant Dil solution (Cambrex), and lytic activity against CFSE-stained (Invitrogen) target cells (100.000 target cells/well) was assessed in modified 4-h FATAL assays (43), using various E:T ratios. Cells were detached, and target cell lysis was determined by flow cytometric analysis of 50,000 target cells in a FACScan flow cytometer (BD Biosciences). PKH-26-negative target cells were selected by gating, and the percentage of CFSEdim cells within this population was determined. Spontaneous leakage of CFSE was determined by incubating the target cells with medium alone.
For lysis experiments with purified NK cells, SK-OV-3 cells were stably transfected with a firefly luciferase plasmid, kindly provided by Dr. Michael Jensen (City of Hope National Medical Center and Beckman Research Institute, Duarte, CA) (44). NK cells were isolated from peripheral blood using a magnetic NK cell isolation kit (Miltenyi Biotech) and stimulated for 48 h with IL-2 (100 IU/ml) in the absence or presence of 5 ng/ml recombinant human (rh) MIF (Lot 38220000; R&D Systems) or SN of primary ovarian cancer cells as described above for recovery experiments. Target cells were seeded into 96-well plates (104/well), and NK cells were added in triplicates at the indicated E:T ratios. After addition of cell-permeable D-luciferin (PJK) at 0.14 mg/ml, luminescence was recorded at different time points using an Orion II luminometer (Berthold). The ATP-dependent conversion of luciferin occurs only in viable luciferase-transfected cells and is thus directly proportional to cell viability (44).
Regulation of NKG2D expression by patient or control serum, ascites, OvCA SNs, or rhMIF and TGF-β
PBL were obtained from healthy volunteers or OvCA patients and incubated for 48 h with patient or control serum or ascites or with SN generated from primary ascites-derived OvCA cells or from SK-OV-3 control and MIF-knockdown cells. Likewise, rhMIF or TGF-β1 (Peprotech) were applied for 24 or 48 h. rhMIF was prepared at Yale University by a standardized protocol (45) and found to contain 32 EU/mg recombinant protein. Several lots from a commercial supplier were also tested (R&D Systems; Lots FC115011, FC115061, FC115091, and FC1107011). The TGF-β receptor I kinase inhibitor SD-208 (46) was kindly provided by Scios. Where indicated, MIF was blocked with 10 μg/ml azide-free mAb289 (R&D Systems). Anti-CXCR2 (MAB331; R&D Systems) and anti-CD74 (clone BU45; Immunotools) were also used at 10 μg/ml. For analysis of MIF mRNA levels, PBL were lysed at 24 h, and RNA was extracted and reverse transcribed as described previously. Surface expression levels of different NK cell receptors were analyzed at 48 h. The cells were stained with the following Abs: MAB139 anti-NKG2D Ab (R&D Systems), BAT221 anti-NKG2D-PE (Miltenyi Biotech), mouse anti-human NKp30-PE (Immunotech) or a mouse IgG1 isotype control (Immunotools) and washed. Indirect stainings were visualized with goat anti-mouse Cy3 (Dianova) or goat anti-mouse FITC (Caltag). After thorough washing, further Abs (all from Immunotools) were added to enable the discrimination of the various immune cell subpopulations. The following combinations were used: CD3-PeCy5/CD56-PE/NKG2D-FITC; CD3-FITC/CD8-PeCy5/NKG2D-Cy3; and CD3-PeCy5/CD4-FITC/NKG2D-Cy3. Cells were analyzed in a FACScan and gated for CD4+ T (CD3+CD4+), CD8+ T (CD3+CD8+), and NK (CD3−CD56+) cells; NKG2D expression was determined for each subset. Expression levels are indicated as specific fluorescence intensity (SFI) values (5). These values are obtained by dividing the fluorescence intensity detected with the specific Ab by the fluorescence signal measured with the isotype control Ab. Expression of NKG2DL was assessed as previously described (38).
For recovery experiments, treated immune cells were washed and cultured for another 48 h before being analyzed as above.
Experiments were performed at least three times with similar results. Analysis of significance was performed using the two-tailed Student t test or, where indicated, ANOVA (Excel, Microsoft). For the assessment of in vivo expression levels, the scores for the relative staining intensities were compared between the various tumor entities using the Kruskal-Wallis test (∗, p < 0.05; ∗∗, p < 0.01). For flow cytometry data, SDs are indicated as calculated from the raw data by Summit software (DakoCytomation).
MIF is strongly overexpressed in OvCA
To investigate the in vivo expression of MIF in OvCA, paraffin sections from OvCAs (n = 11, 7 serous, 3 mucinous, 1 endometrioid), borderline tumors (n = 10), or healthy control tissue (n = 11) were stained with the MIF-specific mAb 289 (Fig. 1 A). The percentage of MIF-positive tumor cells and the staining intensities were quantified on a scale ranging from 0 to 5 with 0 indicating the absence of staining and 5 being the maximum as outlined in Materials and Methods. Average scores were 0.3 ± 0.3 (range, 0–1) for normal ovaries, 1.2 ± 0.8 (range, 0–3) for borderline tumors (p < 0.01), and 4.6 ± 0.4 (range, 3–5) for OvCA (p < 0.001). No significant differences were found between the various OvCA subtypes.
In addition, we performed double stainings using anti-MIF mAb289 in combination with the tumor markers EpCAM (CD326) or cytokeratin 8 and with the macrophage marker CD68. Stainings of the tumor stroma showed that moderate MIF expression colocalizes with CD68 or with endothelial vessel structures. However, MIF levels were much higher in EpCAM (CD326) or cytokeratin 8-positive cells in the tumor center (Fig. 1 B).
To quantify MIF mRNA expression relative to normal ovarian tissue (n = 3), solid tissues from borderline tumors (n = 4) and OvCA (n = 17, 9 serous-papillary, 5 mucinous, 3 endometrioid) as well as ascites-derived EpCAM-positive primary OvCA cells (n = 10, 5 serous-papillary, 3 mucinous, 2 endometrioid) were analyzed by quantitative real-time PCR revealing a highly significant overexpression of MIF mRNA in tissue from borderline tumors (7,000-fold on average; range, 4,000–10,000) and further increased MIF levels in solid OvCA tissue (25,000-fold on average; range, 3,300–70,000; Fig. 2 A). MIF mRNA expression did not differ between solid tumor tissue and purified ascites-derived primary OvCA cells (18,000-fold overexpression on average; range, 70–49,000). cDNA samples prepared from splenocytes of MIF−/− mice served as negative controls which confirmed that the low MIF levels detected in normal ovarian tissue were still >8000-fold above background.
Immunoblotting of protein lysates from normal ovarian tissue (n = 6), borderline tumors (n = 3), OvCA tissue (n = 8, 5 serous-papillary, 2 mucinous, 1 endometrioid), and ascites-derived primary OvCA cells (n = 10, 5 serous-papillary, 3 mucinous, 2 endometrioid) confirmed increased MIF protein levels in OvCA compared with borderline tumors and normal ovarian tissues (Fig. 2 B).
High MIF levels correlate with the presence of ascites
To obtain data on the presence of secreted soluble MIF, we measured MIF protein levels in ascitic fluids from 13 patients with OvCA and 5 patients with nonmalignant ascites by ELISA (Fig. 2,C). Although MIF levels in malignant OvCA ascites ranged from 0.34 to 11.30 ng/ml (median, 2.70 ng/ml), ascites from patients with liver cirrhosis or cardiac ascites contained only 0.30–0.63 ng/ml (median, 0.50 ng/ml; p < 0.05). To further assess the secretion of MIF by OvCA cells, we placed 2 mg of freshly resected OvCA tissue in 2 ml of medium and measured the resulting MIF concentrations after 72 h of incubation (Fig. 2 C). The high amounts of MIF secreted by OvCA tissue (median, 4.5 ng/ml; range, 4.3–4.9 ng/ml; n = 5,) may explain the increase in MIF serum levels already reported for OvCA patients (47).
Having found that MIF levels are high in malignant ascites, we wondered whether the formation of ascites might correlate with MIF expression levels in the primary tumor (Fig. 2 D). A two-sided, unpaired t test revealed that MIF expression levels were significantly higher in patients who had developed ascites (n = 8) at the time of tumor surgery than in ascites-free patients (n = 7; p < 0.01).
MIF inhibits NK cell-mediated killing by decreasing NKG2D expression
To elucidate a functional immunological role for MIF, we investigated whether a blockade of MIF would affect the NK cell-mediated killing of OvCA cells. To this aim, polyclonal NK cell cultures were generated and used as effector cells against 51Cr-labeled primary ascites-derived OvCA cells. Inclusion of a blocking anti-MIF Ab resulted in a moderate but significant increase in target cell lysis (Fig. 3,A). In an alternative experimental setup, polyclonal NK cells were treated for 48 h with MIF-containing SN from primary ascites-derived OvCA cells before being used as effector cells against the same primary OvCA targets (Fig. 3,B). A comparison with the previous experiment showed that target cell lysis is >50% inhibited by the tumor cell SN. However, this inhibition is greatly relieved when a blocking anti-MIF Ab is present during the preincubation period. To confirm these data with an established permanent cell line, we stably down-regulated MIF in SK-OV-3 cells. Next, we preincubated polyclonal human NK cell cultures with SNs of SK-OV-3 pSUPERpuro control (1.5 ng/ml MIF) or MIF-depleted SK-OV-3 pSUPERpuroMIF (0.3 ng/ml MIF) cells. In line with the results obtained with primary cells, we also found the subsequent lytic activity of NK cells against SK-OV-3 WT cells to be significantly lower when they had been pretreated with MIF-containing SN than when they had been exposed to SN of pSUPERpuroMIF cells (Fig. 3 C).
Accordingly, we hypothesized that MIF might down-regulate a receptor involved in the recognition of tumor cells by NK cells. One obvious candidate molecule in this context was NKG2D (30, 31), an activating receptor on NK and a costimulatory receptor on CD8+ T cells. In fact, NKG2D levels on CD8+ T cells and NK cells were substantially reduced by MIF-containing SN. This effect was abrogated when SN from MIF-depleted cells were used (Fig. 4 A).
Because TGF-β exerts similar effects on the NKG2D system (38), we assessed possible interrelations between TGF-β and MIF. TGF-β levels were unaltered in MIF-depleted cells, as assessed by the TGF-β-specific CCL-64 bioassay (Ref. 46 and data not shown). Both a blocking anti-MIF Ab (mAb289) or the TGF-β receptor I kinase inhibitor, SD-208 (46), partly prevented the reduction of NKG2D levels mediated by the SK-OV-3 SN (Fig. 4 B). Although concentration escalation experiments did not enhance the effect of either reagent alone (data not shown), the combination of SD-208 and mAb289 rescued NKG2D surface expression levels completely. Thus, MIF and TGF-β down-regulate NKG2D levels in an additive and therefore presumably independent manner.
In human ovarian cancer patients, both MIF and TGF-β may affect NKG2D expression in vivo. Consequently, we found NKG2D surface expression levels to be greatly reduced in these patients (Fig. 4,C). To obtain more information on the role of MIF in this context, we incubated PBLs from ovarian cancer patients (n = 3) or healthy donors (n = 5) with OvCA ascites (n = 5) or serum (n = 3) in the absence or presence of the blocking anti-MIF or an isotype control Ab. Serum from healthy donors (n = 4) and RPMI 1640 with FCS were also included as controls. These experiments showed that ovarian cancer patients display drastically reduced NKG2D expression levels on freshly isolated PBL and on ascites-derived tumor-infiltrating lymphocytes (TIL) (Fig. 4, C–E). However, NKG2D expression recovers upon brief (48 h) in vitro culture in the absence of tumor-derived factors (Fig. 4, D and E). In contrast, NKG2D expression on immune cells from healthy donors is decreased by ascites or serum from tumor patients (Fig. 4, D and E). Moreover, when PBL from either donor are cultured in the presence of ascites or serum, addition of a neutralizing anti-MIF Ab partly rescues NKG2D expression. This effect is more pronounced with serum from tumor patients than from healthy controls (Fig. 4 E). This confirms that MIF contained in serum or ascites from OvCA patients is partly responsible for the observed down-regulation of NKG2D, at least under ex vivo conditions. Another part of the effect may be due to TGF-β which was also detected in SNs from primary OvCA cells (median concentration for TGF-β1, 0.5 ng/ml; range, 0.08–1.25 ng/ml; n = 7).
Importantly, the effect of MIF-containing SN or ascites on NKG2D expression was mimicked by rhMIF, which induced a concentration-dependent down-regulation of NKG2D on CD8+ T cells and NK cells. Surface levels of NKp30 were also reduced by treatment with TGF-β1 (37), but not by MIF (Fig. 5 A, lower right). Neither did MIF (4 ng/ml, 48 h) affect the expression of NKG2A, NKG2C, NKp44, NKp46, or NKp80 (data not shown).
To further dissociate MIF-dependent from TGF-β-mediated effects, we incubated PBLs with rhMIF or TGF-β 1 (4 ng/ml each) for 48 h in the presence or absence of 1 μM SD-208 or the blocking anti-MIF Ab mAb289 used at 10 μg/ml. For each treatment, only the specific antagonist prevented the reduction of NKG2D surface levels on CD8+ T and NK cells (Fig. 5 B). The increased NKG2D surface expression on SD-208-treated cells can be explained by SD-208 blocking auto and paracrine effects of lymphocyte-derived TGF-β.
Another difference between MIF and TGF-β became apparent when we investigated the recovery of NKG2D expression on cytokine-treated cells: The effect of MIF was reversed 48 h after the cells had been switched to RPMI 1640 with 10% FCS + 100 IU/ml IL-2. In contrast, the effect of TGF-β largely persisted even 48 h after the medium exchange (Fig. 5,C). This suggests that MIF and TGF-β may exert their effects on NKG2D via different mechanisms. In fact, PBLs that had been treated with MIF showed a greatly reduced number of NKG2D transcripts at 24 h (Fig. 5 D), whereas NKG2D mRNA was only slightly down-regulated by TGF-β 1 (p = 0.17). Accordingly, MIF seems to have a strong transcriptional effect on NKG2D, whereas TGF-β appears to modulate NKG2D expression levels largely by posttranscriptional mechanisms.
Impaired cytotoxic capacity of MIF-treated NK cells is due to an MIF-dependent reduction in NKG2D expression
To confirm the functional significance of NKG2D levels at the surface of NK cells, we first characterized the expression of NKG2D ligands in SK-OV-3 cells. SFI values were 1.4 ± 0.2 for MICA, 1.1 ± 0.04 for MICB, 1.1 ± 0.1 for ULBP1, 1.9 ± 0.3 for ULBP2 and 1.7 ± 0.1 for ULBP3. Autocrine regulations of NKG2D ligands (38) were not observed between control-transfected and pSUPERpuroMIF-transfected SK-OV-3 cells (data not shown). Having confirmed that ligands for NKG2D are expressed by OvCA cells, we preincubated magnetically sorted NK cells for 48 h with 100 IU/ml IL-2 (to activate them) or with 100 IU/ml IL-2 + 4 ng/ml rhMIF or with 500 μl of tumor cell SN in the absence or presence of 10 μg/ml anti-MIF Ab or isotype control. When one-half of the cells were now used as effectors in a 4-h biophotonic lysis assay against SK-OV-3-fLuc cells, the MIF-treated cells showed clearly reduced lytic capacity (p < 0.05; Fig. 5,E, top). However, the remaining cells were cultured for a further 48 h in medium low-dose IL-2 (10 IU/ml), but no MIF. Then, a subsequent lysis assay confirmed that the recovery of NKG2D surface expression coincided with an almost complete restoration of the killing capacity of MIF-treated NK cells (p > 0.05; Fig. 5 E, bottom). Likewise, when NK cells had been treated for 48 h with tumor cell SN (in the absence or presence of MIF Ab), MIF-dependent differences in target cell lysis almost disappeared after 48 h of recovery (data not shown). However, SN from primary ascites-derived OvCA cells also induced MIF-independent long-term effects, such as reduced viability of NK cells and diminished killing capacity (data not shown).
Although these data show that the down-regulation of NKG2D expression correlates with the impaired lytic activity of MIF-treated NK cells, we also sought to prove a functional relationship between reduced NKG2D levels and killing capacity. Toward this aim, we repeated the 4-h killing assay with MIF-treated NK cells in the absence or presence of a blocking anti-NKG2D Ab (10 μg/ml) or the respective isotype control (Fig. 5,F). Again, MIF-treated polyclonal NK cells displayed reduced lytic capacity when compared with untreated NK cells (p < 0.05). The addition of anti-NKG2D inhibited target cell lysis by both untreated (p < 0.01) and MIF-treated NK cells (p < 0.05), which still displayed residual NKG2D expression. Importantly, pretreatment with MIF showed no additional effect when NKG2D was blocked (Fig. 5 F), indicating that MIF does not affect the NKG2D-independent lytic activity of the NK effector cells. These data therefore lead to the conclusion that the MIF effect on NKG2D expression is directly responsible for the MIF-induced inhibition of lytic activity.
MIF was the first cytokine to be discovered (10), but its diverse role in immunity has only recently been elucidated (12). MIF-knockout mice display no obvious phenotype (48) apart from p53-dependent growth alterations (18). Nevertheless, gene array analysis of MIF-knockdown neuroblastoma cells revealed no less than 166 different genes regulated in an autocrine manner by MIF (49). Its most important physiological functions may be the sustainment of inflammatory reactions (50) and the promotion of tissue repair (11). A role for MIF in autoimmunity has been confirmed because MIF-knockout mice are resistant toward experimentally induced arthritis and autoimmune encephalomyelitis (51, 52). A critical role for MIF in tumor biology is strongly supported by the finding that fibroblasts from MIF-knockout mice cannot be transformed by various oncogenes (25). The inactivating effect of MIF on p53 has been demonstrated in a knockout mouse model (17, 18). An effect of MIF on tumor angiogenesis, proliferation, and metastasis was shown in vivo using MIF-knockdown neuroblastoma cells (53). In addition, MIF inhibits tumor cell-specific cytolytic responses (27, 28) and induces TLR4 (54), which is found on OvCA cells and is associated with an inflammatory, chemoresistance-promoting tumor milieu (7). Given these manifold tumorigenic effects of MIF, its overexpression in various human neoplasias including prostate-, bladder-, breast-, colon-, brain-, skin-, and lung-derived tumors (20) comes as little surprise.
We have analyzed MIF in OvCA and found a significant overexpression in vitro and in vivo (Figs. 1,A and 2, A–D). MIF levels were similar in solid tissue from the primary tumor and in purified ascites-derived tumor cells, suggesting that MIF expression remains unchanged when cells dissociate from the primary tumor and migrate toward the peritoneum. Immunohistochemical double staining of primary tumors revealed that MIF overexpression may largely be ascribed to EpCAM- and cytokeratin 8-positive OvCA cells, whereas the contribution from tumor-infiltrating macrophages and endothelial cells appears to be small in comparison (Fig. 1 B).
The high levels of tumor-derived MIF were also reflected by an overexpression of MIF in malignant ascites as compared with nonmalignant hepatic and cardiac ascites (Fig. 2,C). MIF expression levels correlate with malignancy. Whereas borderline tumors that are characterized by a noninvasive growth pattern overexpress MIF to a much lesser extent (Figs. 1,A and 2, A and B), we found the highest MIF mRNA expression in ascites-forming OvCA that are associated with the poorest prognosis (Ref. 9 and Fig. 2 D). It may be speculated whether proinflammatory effects exerted by MIF contribute to the formation of malignant ascites. Alternatively, lysophosphatidic acid contained in malignant ascites may superinduce MIF levels.
To address a functional immunological role for MIF in OvCA, we generated a SK-OV-3 MIF-knockdown subline that secreted ∼80% less MIF than control cells. Viability, proliferation, and morphological appearance of the MIF-depleted cells were unaltered (data not shown). SN from MIF-knockdown cells was considerably less suppressive for NK cytotoxicity than SN derived from control transfectants (Fig. 3,C). Likewise, inclusion of a blocking anti-MIF Ab enhanced the NK cell-mediated lysis of primary ascites-derived OvCA cells (Fig. 3,A). Moreover, the inhibitory effect of SN from primary ascites-derived OvCA cells could also be relieved by the blockade of MIF (Fig. 3 B). Nevertheless, the lytic activity of polyclonal NK cells was still inhibited by OvCA SN even when a blocking anti-MIF Ab was included.
Because NK cell activation depends on the balance between activating and inhibitory signals transmitted by various NK receptors, we hypothesized that MIF might induce the down-regulation of an activating receptor. Flow cytometry and quantitative real-time PCR revealed that MIF decreases NKG2D expression in NK and CD8+ T cells (Figs. 4, A and B and 5, A–C). This effect was not restricted to rhMIF or MIF contained in the SNs from SK-OV-3 cells. Serum or ascites from OvCA patients also decreased the NKG2D surface expression on CD8+ T and NK cells from healthy donors in a MIF-dependent manner. In contrast, serum from control persons had a much weaker effect which corresponds to the lower MIF levels described in the sera of healthy individuals (Fig. 4, D and E). This correlates with reduced NKG2D levels on tumor-infiltrating lymphocytes or PBLs from OvCA patients (Fig. 4, C–E). The reduced NKG2D expression levels on immune cells from tumor patients recover to normal levels when the PBLs are cultured in normal growth medium in vitro. Accordingly, the reduced NKG2D expression levels observed in vivo do not seem to be due to T or NK cell-intrinsic deficiencies in OvCA patients but rather to the presence of OvCA-derived factors including MIF. Not surprisingly, however, immune cells that are treated with tumor ascites or serum still show a moderate down-regulation of NKG2D when MIF is antagonized which indicates that further tumor-associated immunosuppressive cytokines also contribute to the observed effect. Because TGF-β has so far been the only cytokine known to affect NKG2D expression, we confirmed the presence of low to moderate levels of TGF-β1 by ELISA (median concentration, 0.5 ng/ml) and sought to differentiate between effects caused by TGF-β and MIF.
One difference appears to be a much higher specificity of MIF for NKG2D. Whereas NKG2A, NKG2C, NKp30, NKp44, NKp46, and NKp80 expression levels on NK cells were not affected by MIF, TGF-β also induces a strong down-regulation of NKp30 (37) as well as an up-regulation of NKG2A (Fig. 5,A and data not shown). The recovery of NKG2D levels on cytokine-treated cells also seems to follow a different kinetics: Whereas NKG2D expression on MIF-treated cells is restored to control levels within 48 h after withdrawal of the cytokine, TGF-β shows a more persistent effect (Fig. 5,C). Mechanistically, we show that MIF inhibits the transcription of NKG2D mRNA, whereas TGF-β appears to exert its effect on NKG2D mainly via posttranscriptional mechanisms (Fig. 5,D). In accordance with these data, both a blocking anti-MIF Ab and the TGF-β receptor kinase inhibitor SD-208 partly prevented the down-regulation of NKG2D by tumor cell SN, with additive effects (Fig. 4,B) (46). SD-208, however, did not prevent the reduction of NKG2D levels when PBLs were treated with rhMIF (Fig. 5 B). Thus, MIF and TGF-β must use different receptor mechanisms to suppress NKG2D expression. Although we have not identified the SD-208-insensitive receptor involved in the MIF-dependent down-regulation of NKG2D, CD74 can be excluded because it is not found on NK or CD8+ T cells. This leaves as primary candidates the noncognate MIF receptors CXCR2 and CXCR4 which are expressed by resting and activated NK and CD8+ T cells. In fact, a blocking Ab against CXCR2 could partly prevent the effect of MIF on NKG2D expression (Refs. 14 , 21 , and 22 and data not shown).
To test whether NKG2D expression on NK cells really correlates with lytic activity against OvCA targets, we performed killing assays using MIF-treated IL-2-activated NK cells either immediately after 48 h of exposure to the cytokine or after an additional 48 h of recovery in medium with IL-2 (100 IU/ml). As predicted, the initially reduced killing capacity of MIF-treated NK cells was largely restored after the recovery period (Fig. 5 E). When (instead of the recombinant cytokine) MIF-containing tumor cell SN was used, addition of anti-MIF Ab also attenuated the inhibition of NK cell-lytic activity at 48 h. Again, the MIF-dependent differences in target cell lysis lost their significance after additional 48 h of culture in fully supplemented RPMI 1640 + IL-2. However, OvCA SN also induced MIF-independent long-term effects including significant cell death and reduced lytic capacity (data not shown).
To further confirm the functional importance of MIF-induced down-regulation of NKG2D, we treated purified NK cells with MIF and performed a lysis assay in the presence of either a control or a blocking anti-NKG2D Ab (Fig. 5 F). This showed that the blocking of NKG2D by Ab represses the cytolytic activity of NK cells even more efficiently than treatment with MIF, suggesting that the low levels of NKG2D remaining on the surface of MIF-treated cells are still sufficient to mediate some activation and ensuing killing activity. An important observation, however, was that there was no remaining MIF-dependent effect when NKG2D was blocked by neutralizing Abs. Accordingly, MIF does not seem to affect the residual NKG2D-independent lytic activity of NKG2D-blocked NK cells which illustrates that the effect of MIF on lytic activity is mediated by the induced loss of NKG2D.
Taken together, our study is the first to demonstrate that MIF expression in OvCA increases with the grade of malignancy. Functionally, we demonstrate that MIF inhibits antitumor immunity by down-regulating NKG2D on NK and CD8+ T cells. Autocrine effects of MIF on the progression of OvCA have not been investigated, but are likely. Thus, MIF inhibitors that are currently being developed for the treatment of rheumatoid arthritis (39) may become promising agents also for the treatment of OvCA.
The authors have no financial conflict of interest.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by grants from the Interdisciplinary Center for Clinical Research Würzburg (to J.W.) and the National Institutes of Health (to L.L. and R.B.).
Abbreviations used in this paper: OvCA, ovarian carcinoma; MIF, macrophage migration inhibitory factor; NKG2D, natural killer group 2D; SFI, specific fluorescence intensity; SN, supernatant; rh, recombinant human.