Expression, signaling proficiency, and stimulatory function of the NKG2D lymphocyte receptor in human cancer cells

Abstract

The stimulatory natural killer group 2 member D (NKG2D) lymphocyte receptor and its tumor-associated ligands are important mediators in the immune surveillance of cancer. With advanced human tumors, however, persistent NKG2D ligand expression may favor tumor progression. We have found that cancer cells themselves express NKG2D in complex with the DNAX-activating protein 10 (DAP10) signaling adaptor. Triggering of NKG2D on ex vivo cancer cells or on tumor lines which express only few receptor complexes activates the oncogenic PI3K–protein kinase B (PKB/AKT)–mammalian target of rapamycin (mTOR) signaling axis and downstream effectors, the ribosomal protein S6 kinase 1 (S6K1) and the translation initiation factor 4E-binding protein 1 (4E-BP1). In addition, as in lymphocytes, NKG2D ligand engagement stimulates phosphorylation of JNK and ERK in MAP kinase cascades. Consistent with these signaling activities, above-threshold expression of NKG2D–DAP10 in a ligand-bearing tumor line increases its bioenergetic metabolism and proliferation, thus suggesting functional similarity between this immunoreceptor and tumor growth factor receptors. This relationship is supported by significant correlations between percentages of cancer cells that are positive for surface NKG2D and criteria of tumor progression. Hence, in a conceptual twist, these results suggest that tumor co-option of NKG2D immunoreceptor expression may complement the presence of its ligands for stimulation of tumor growth.

The stimulatory natural killer group 2 member D (NKG2D) lymphocyte receptor and its ligands are important mediators of tumor immunity but also are instrumental in promoting tumor immune evasion and immune suppression (1, 2). Human NKG2D in complex with the DNAX-activating protein 10 (DAP10) signaling adaptor is expressed on most natural killer (NK) cells and CD8 T cells (3, 4). DAP10 signals upon Tyr phosphorylation of YXXM motifs in its cytoplasmic tail by binding either the p85 subunit of PI3K or the growth factor receptor-bound protein 2 (Grb2), thus activating protein kinase B (PKB/AKT) and MAP kinase cascades (4–6). NKG2D ligands include the MHC class I-related chains A and B (MIC, A and B), which are absent from the surface of most normal cells but are induced by generic responses to cellular stress in diseased cells such as epithelial tumor cells (1, 7). Additional NKG2D ligands comprise six members of the UL16-binding protein (ULBP) family, which can be tumor associated as well (8).

MIC frequently are expressed in cancers and stimulate NK-cell and CD8 T-cell responses against tumor cells in vitro (3, 7). Evidence for NK-cell– and T-cell–mediated protective functions of NKG2D in tumor immunity has been derived from mouse model experiments and a study of NKG2D-deficient mice (9–12). At least in humans, however, tumor progression coupled with persistent surface expression and shedding of soluble MICA point to changes in dynamics that may signify host tumor susceptibility instead of resistance (13). NKG2D is subject to ligand-induced down modulation. Because of the ubiquitous presence of soluble MICA (sMICA) in many cancer patients, NKG2D function becomes systemically impaired (13). In addition, tumor-associated MICA and sMICA drive proliferative expansions of normally rare NKG2D⁺ CD4 T cells that have immunosuppressive functions (14). Taken together, these negative imprints on the immune system may account, at least in part, for poor clinical outcomes that have been associated with the presence of NKG2D ligands in cancer patients (1, 15–17).

As tumor immune surveillance fails in advanced cancer patients, we suspected that persistent NKG2D ligand expression, in addition to promoting immune evasion, might constitute another, perhaps more fundamental, tumor survival asset. We report here that breast, ovarian, prostate, and colon cancer cells express NKG2D–DAP10 receptor complexes. In vitro experimental and correlative clinical data support the idea that this immunoreceptor complements the presence of its ligands on cancer cells for stimulation of tumor growth, thus suggesting functional similarity to oncogenic growth factor receptors.

Results

Expression of NKG2D–DAP10 on Cancer Cells.

In the course of examining primary breast and epithelial ovarian cancer specimens for infiltrating lymphocytes by immunohistochemistry, we observed unambiguous cancer tissue staining for the NKG2D receptor using the specific 1D11 mAb and HRP-conjugated secondary reagent (Fig. 1A) (3). The staining patterns were similar to those recorded for the MIC ligands of NKG2D that were detected with the bispecific 6D4 mAb (7). In comparison, tumor-infiltrating CD3⁺ lymphocytes, among which most CD8 T cells are positive for NKG2D, were infrequent and scattered. Staining for NKG2D of normal breast, ovary, and prostate tissue sections including well-recognizable epithelial areas gave negative results (Fig. 1B).

Fig. 1.

Cancer-cell expression of NKG2D–DAP10 and stimulation of PI3K-dependent AKT phosphorylation. (A) Micrographs of Ab staining for NKG2D and MIC of breast and ovarian cancer tissue cryosections. Staining for CD3 identifies tumor-infiltrating T cells. (B) Staining of normal breast, ovary, and prostate tissue sections for NKG2D. (C) Flow cytometry of freshly isolated breast (BT), ovarian (OT), prostate (PT), and colon (CT) cancer cells (gated for EpCAM⁺CD45⁻; see upper left dot plot for a representative example) for surface NKG2D and MIC. Numbers in dot plots indicate percentages of cells in quadrants. (D) Detection of mRNA for NKG2D, DAP10, and control GAPDH by RT-PCR in NKL NK cells and in breast (BT), colon (CT), ovarian (OT), and prostate (PT) cancer cells. Note that sample OT91 is essentially negative for NKG2D–DAP10. RNAs were prepared from sorted EpCAM⁺CD45⁻ cancer-cell suspensions. Panels at right show minimal NKG2D–DAP10 expression in control breast, ovary, and skin tissue specimens nondepleted for CD45⁺ cells. Numbers at left indicate cDNA amplicon sizes (in bp). (E) NKG2D immunoprecipitations (IP) with bead-coupled mAb 5C6 and immunoblotting (IB) for NKG2D and DAP10 using cell lysates of control NKL cells (1 × 10⁶ cells) and EpCAM⁺CD45⁻ cancer cells (3–5 × 10⁶ cells) corresponding to those shown in D. Panels at right show protein data from an additional breast cancer and matched nonaffected tissue (NAT) control. Numbers at left indicate molecular masses (in kDa) of NKG2D and DAP10. (F) Stimulation of purified EpCAM⁺CD45⁻ breast, ovarian, and prostate cancer cells by cross-linked anti-NKG2D mAb 1D11 results in detection of P-AKT (S473) on total cell lysate immunoblots. P-AKT was not detected in ovarian cancer cells (OT25⁻) sorted for absence of NKG2D. Exposure to insulin provides positive control activation except with the nonresponsive prostate cancer cells. The PI3K inhibitor wortmannin (Wort) inhibits AKT phosphorylation. DMSO is added as solvent control. (G) Detection of P-AKT (S473) after stimulation of purified breast and ovarian cancer cells with cross-linked anti-NKG2D mAb 5C6 F(ab′)₂ fragments. Ig lanes in F and G represent cells exposed to primary mouse control IgG and secondary goat anti-mouse F(ab′)₂.

Cell-surface expression and composition of NKG2D receptor complexes were examined with cancer-cell suspensions sorted for an epithelial cell adhesion molecule (EpCAM)⁺CD45⁻ phenotype to ensure analysis of epithelial tumor cells and exclusion of hematopoietic cells. The epithelial nature of EpCAM⁺CD45⁻ cells was confirmed separately by staining for pan-cytokeratin. Flow cytometry analysis showed that 12/12 breast, 14/14 colon, 29/30 epithelial ovarian, and four of four prostate cancer specimens included cancer-cell populations that were positive for surface NKG2D, with proportional ranges of 4–83% (mean 23%, SD ± 22.3), 4–39% (mean 16%, SD ± 10.6), 1–65% (mean 18%, SD ± 17.3), and 3–33% (mean 11%, SD ± 14.7), respectively (Fig. 1C). Sizeable proportions of all cancer-cell suspensions expressed the MIC ligands of NKG2D.

We next tested whether cancer cells also express the DAP10 signaling adaptor (4–6). By standard 30-cycle RT-PCR, NKG2D and DAP10 cDNA amplicons were readily detected with six of seven breast, ovarian, colon, and prostate EpCAM⁺CD45⁻ cancer-cell suspensions. Upon sequencing, we found no changes in the canonical amino acid coding regions. Only faint signals were recorded with mRNAs from non-lymphocyte-depleted normal breast, ovarian, and skin tissues (Fig. 1D). Cancer-cell expression of NKG2D–DAP10 complexes was confirmed by immunoprecipitation using bead-coupled anti-NKG2D 5C6 mAb (3) followed by SDS/PAGE and sequential immunoblot probing for NKG2D and DAP10. Direct comparison of a breast cancer sample with matched nonaffected tissue control further illustrated the malignancy-associated expression of NKG2D–DAP10 (Fig. 1E).

Triggering of NKG2D Stimulates PI3K-Dependent Phosphorylation of AKT.

Our findings gave rise to the idea that cancer cells might co-opt expression of NKG2D to exploit the presence of its ligands for self-stimulation of tumor growth. In NK cells and T cells, phosphorylation of DAP10 activates branched signaling cascades that include the PI3K–AKT axis (4, 5, 18). We tested the signaling capacity of NKG2D–DAP10 in sorted EpCAM⁺CD45⁻ breast, ovarian, and prostate cancer cells after serum deprivation by 1D11 mAb-mediated receptor cross-linking and subsequent Ab probing of total cell lysate immunoblots for phosphorylation of AKT (19). Phospho-AKT (P-AKT) was induced in all four cancer samples tested but not in matched ovarian cancer cells sorted for absence of surface NKG2D. Its appearance was sensitive to wortmannin and thus was dependent on PI3K (Fig. 1F) (4, 19). Exposure to insulin provided positive control activation except for unresponsive prostate cancer cells. Crosslinked F(ab′)₂ fragments of the anti-NKG2D 5C6 mAb also were effective in P-AKT induction, thus precluding Ab Fc region/Fc receptor-mediated or unspecific stimulatory events (Fig. 1G).

Minimal NKG2D–DAP10 Expression Is Sufficient for AKT Activation in Tumor Lines.

A thorough investigation of NKG2D–DAP10 signaling and its physiological effects required serial analyses and approaches for which ex vivo cancer cells were unsuitable. We thus switched to studying tumor lines. However, unlike cancer cells, 15 breast, colon, gastric, ovarian, and prostate tumor lines tested by flow cytometry were either negative for surface NKG2D or displayed only minor shifts in the fluorescence intensity profile. No increased expression was apparent after treatment of cells with inhibitors of proteasomal or lysosomal degradation or after exposure to IL-15, which induces T-cell NKG2D (13, 20, 21). By real-time quantitative PCR (qPCR), 12 randomly selected tumor lines contained an average of no more than one copy of NKG2D and DAP10 mRNA per cell. In comparison, five freshly prepared NKG2D⁺EpCAM⁺CD45⁻ cancer cell suspensions contained averages of 15 and 100 copies of NKG2D and DAP10 mRNAs, respectively. Nonetheless, with a subset of tumor lines (breast MCF-7, BT-20, and MDA-MB-453; ovarian HTB-78; colon DLD-1; and gastric AGS) chosen because of low constitutive AKT phosphorylation, PI3K-dependent induction of P-AKT was readily observed after desensitization of cells and 1D11 mAb- or 5C6 F(ab′)₂-mediated NKG2D cross-linking (Fig. 2A).

Fig. 2.

PI3K-dependent AKT phosphorylation correlates with minimal NKG2D–DAP10 expression in tumor lines. (A) Immunoblot detection of P-AKT (S473) in lysates of desensitized MCF-7, BT-20, MDA-MB-453, HTB-78, DLD-1, and AGS cells but not of MDA-MB-231, PC3, and A375 cells after mAb 1D11 or mAb 5C6 F(ab′)₂ cross-linking. Transfection of NKG2D–DAP10 restores AKT phosphorylation in A375-TF cells. Insulin provides positive control activation. Ig control lanes represent cells exposed to mouse IgG and secondary goat anti-mouse F(ab′)₂. DMSO is added as solvent control. As with the cancer-cell suspensions, AKT phosphorylation is sensitive to wortmannin (Wort). (B) AKT phosphorylation correlates with the detection of small amounts of NKG2D–DAP10 protein complexes by immunoprecipitation with bead-coupled anti-NKG2D mAb 5C6 from lysates of typically 5 × 10⁷ cells (2 × 10⁷ BT-20 and 5 × 10⁶ A375-TF cells) and sequential immunoblot probing for NKG2D and DAP10.

These results implying functional NKG2D–DAP10 expression were difficult to reconcile with the minimal expression of the corresponding mRNAs. However, the mRNAs may be distributed unevenly among tumor cells and template multiple translation cycles. In fact, protein expression of NKG2D–DAP10 was detected when lysates of large numbers of cells (∼5 × 10⁷ per SDS/PAGE lane; with cancers, we used ∼3–5 × 10⁶ cells) were used in immunoprecipitation and immunoblot experiments using a highly sensitive chemiluminescent reagent (Fig. 2B). Altogether, these results indicate that few receptor complexes are sufficient for signal transduction in tumor lines, possibly because of their sensitized activation status. As exemplified by the CTLA-4 regulator of T-cell activation, flow cytometry can be insufficiently sensitive to detect minimal expression of functionally active cell-surface receptors (22).

Complementary evidence was obtained with the breast MDA-MB-231, prostate PC3, and melanoma A375 tumor lines that lacked detectable NKG2D–DAP10 complexes and showed no inducible AKT phosphorylation (Fig. 2 A and B). Taken together, these experiments with tumor lines replicated the results obtained with freshly isolated cancer cells, except for the much lower expression of NKG2D–DAP10.

We also tested for expression of NKG2D receptors in mouse cancer specimens including 7,12-dimethylbenz(a)anthracene/12-O-tetradecanoyl-phorbol-13-acetate (DMBA/TPA) carcinogen-induced squamous cell carcinoma, transgenic adenocarcinoma of the mouse prostate (TRAMP) model aggressive and late-arising autochthonous prostate cancers, and human epidermal growth factor receptor 2 (HER-2)/neu–transgenic mammary carcinoma (23–25). By RT-PCR, all T-cell– and NK-cell–depleted cancer-cell samples were devoid of both the long and short variants of murine NKG2D and its DAP10 and DAP12 signaling adaptors (26) (Fig. S1A). No NKG2D receptor protein was detected in any lysates (5 × 10⁶ cancer cells) by immunoprecipitation with bead-coupled Ab and immunoblot (Fig. S1B). It thus appears that NKG2D receptor expression does not occur in mouse models of cancer.

Genetic Confirmation of NKG2D–DAP10 Signaling in Tumor Cells.

Because tumor cell expression and signaling proficiency of NKG2D–DAP10 may have profound implications, we sought definitive experimental proof. Ectopic expression of NKG2D–DAP10 in stable transfectants of the A375 melanoma line (A375-TF cells) restored PI3K-dependent AKT phosphorylation after Ab-mediated receptor cross-linking (Fig. 1 A and B). In a complementary approach, lentiviral transduction of siRNAs in breast tumor MCF-7 cells linked NKG2D and DAP10 depletion to loss of inducible AKT phosphorylation (Fig. S2A). Demonstration of protein depletion, as done for NKG2D (Fig. S2B), was not feasible for DAP10 because the DAP10 siRNA-transduced MCF-7 cells proliferated poorly and could not be expanded to the ∼5 × 10⁷ cells required. Hence, we used NKG2D–DAP10 MCF-7 cell transfectants (MCF-7–TF cells) and a CD8 T-cell line to demonstrate fully the efficacy of RNAi targeting of NKG2D and DAP10 (Fig. S2 B–D). Taken together, these results confirmed the signaling capacity of NKG2D–DAP10 in tumor cells.

Activation of Mammalian Target of Rapamycin and Downstream Effectors and Evidence for Self-Stimulation.

The growth factor-responsive PI3K–AKT signal transduction pathway regulates intersecting cellular processes including cell-cycle progression, metabolic activity, and survival and commonly is hyperactive in cancer (27). To evaluate further the proficiency of NKG2D–DAP10 signaling in tumor cells, we examined activation of the mammalian target of rapamycin (mTOR) kinase downstream of AKT and its catalytic activity on effectors controlling protein synthesis and cell growth, the ribosomal protein S6 kinase 1 (S6K1) and the translation initiation factor 4E-binding protein 1 (4E-BP1) (28, 29). Along this axis, mTOR is part of the rapamycin-sensitive mTOR complex 1 (mTORC1), which is activated separately by nutrient supply. The detection of target phosphorylation events following NKG2D cross-linking thus necessitated extensive prior cell starvation in serum-free medium and HBSS for 24 h and a minimum of 4 h, respectively, to reduce constitutive activation sufficiently (30). Under these conditions, freshly isolated cancer cells and most tumor lines became unresponsive to NKG2D triggering because of impaired viability and probable loss of the scarcely expressed receptor proteins, respectively. However, with MCF-7–TF cells, which were comparable to ex vivo cancer cells in surface NKG2D expression (Fig. 1C and Fig. S2C), and the HTB-78 ovarian tumor line we recorded robust induction of phosphorylation of mTOR, S6K1, and 4E-BP1 (Fig. 3A). The appearance of the phosphoproteins was both PI3K dependent and rapamycin sensitive. Thus, these results provided evidence for the capacity of NKG2D–DAP10 to stimulate the oncogenic PI3K–AKT–mTOR–S6K1/4E-BP1 signaling axis in tumor cells. By inference from this example, other effectors coupled to AKT signal transduction that promote cell-cycle progression, differentiation, and survival are likely to be affected by NKG2D–DAP10 as well.

Fig. 3.

Activation of mTOR–S6K1/4E-BP1 and MAP kinase cascades. (A) Immunoblot detection of P-mTOR (S2448), phosphorylated S6K1 (P-S6K1) (T389), and phosphorylated 4E-BP1 (P-4E-BP1) (T37/46) in lysates of HTB-78 and MCF-7–TF cells after starvation and mAb 1D11 cross-linking. Phosphorylation events are sensitive to both wortmannin (Wort) and rapamycin (Rapa). Insulin provides positive control activation. Ig control lanes are as in Fig. 2. DMSO is added as solvent control. (B) Induction of P-AKT after 10 min in pellets of A375-TF but not in mock-transfected A375 cells and inhibition by anti-MIC/ULBP mAb mixture. (C) Immunoblot detection of P-ERK1/2 (T202/Y204) and P-JNK1/2 (T183/Y185) in lysates of desensitized EpCAM⁺CD45⁻ ovarian cancer (OT63), and HTB-78 and MCF-7–TF cells after exposure to recombinant sMICA and cross-linking anti–His-tag Ab. EGF provides positive control activation. UO126 and SP600125 are inhibitors of MEK/ERK and JNK, respectively. Phosphorylation of ERK but not JNK is sensitive to wortmannin. (D) Immunoprecipitation of DAP10 and immunoblot detection of transiently associated Grb2 in lysates of MCF-7–TF cells (5 × 10⁷ cells per SDS/PAGE lane) after 2 or 5 min of Ab-mediated NKG2D cross-linking. (E) Induction of P-ERK1/2 in pellets of A375-TF cells and inhibition by anti-NKG2D ligand Ab mixture as in B. (F) Induction of P-JNK1/2 in pellets of A375-TF cells and inhibition by anti-NKG2D ligand Ab mixture.

For the signaling experiments throughout this study, cells typically were grown at low density to minimize NKG2D ligand-mediated self-stimulation. However, this functional activity was unproven, although it was central to our model of the biological significance of tumor expression of NKG2D. To obtain evidence for self-stimulation, we examined P-AKT induction in a time-course experiment comparing desensitized A375 mock-transfected control cells (NKG2D⁻DAP10⁻) and A375-TF cells (NKG2D⁺DAP10⁺) that were spun into pellets to enforce cell contacts mimicking solid cancer-cell compaction. Like the other tumor lines, A375 melanoma cells are positive for several NKG2D ligands (Fig. S3). Although no signal increase occurred in the negative control cells, P-AKT was induced in A375-TF cells after 10 min of incubation. This activity was blocked in the presence of a mixture of anti-MIC and anti-ULBP mAb (14), thus reflecting productive NKG2D receptor–ligand interactions (Fig. 3B).

Activation of ERK and JNK MAP Kinases.

As determined thus far, signaling pathway activation by NKG2D–DAP10 in tumor cells was similar to that in lymphocytes. In activated human NK cells and T cells, signal transduction initiated by PI3K also leads to phosphorylation of ERK, whereas alternative coupling of DAP10 to Grb2 results in phosphorylation of JNK (5, 6, 31). In cancer cells, these MAP kinases are activation targets of the EGF receptor (EGFR), among other receptor Tyr kinases, which, because of mutation or aberrant expression, frequently cause excessive tumor cell proliferation and increased motility and survival (32). To test the relevance of these pathways for NKG2D–DAP10 signaling in tumor cells, we used freshly isolated EpCAM⁺CD45⁻ ovarian cancer cells and HTB-78 and MCF-7–TF cells that were desensitized and stimulated with recombinant sMICA cross-linked by anti–His-tag Ab. As with the EGF control activation, stimulation of NKG2D–DAP10 resulted in phosphorylation of the ERK1/2 and JNK1/2 isoforms as determined by immunoblot using phosphoprotein-specific Abs (Fig. 3C). Tumor cell exposure to sMICA or anti–His-tag Ab alone had no effect. The appearance of phosphorylated ERK1/2 (P-ERK1/2) and phosphorylated JNK1/2 (P-JNK1/2) was diminished by inhibitors of the MAP kinase kinase (MEK) upstream of ERK (UO126) and JNK itself (SP600125), respectively, but not vice versa (Fig. 3C). P-ERK1/2 but not P-JNK1/2 was sensitive to wortmannin and thus, as in lymphocytes, dependent on PI3K (6). Consequently, P-JNK1/2 was in all likelihood downstream of Grb2. This arrangement was supported by immunoblot detection of Grb2 in transient association with DAP10, which was immunoprecipitated from lysates of MCF-7–TF cells after brief Ab-mediated NKG2D cross-linking (Fig. 3D). As with tumor cell ligand-induced AKT phosphorylation (Fig. 3B), P-ERK1/2 and P-JNK1/2 were detected in time-course experiments with compacted A375-TF cells in the absence but not in the presence of a ligand-masking Ab mixture (Fig. 3 E and F).

Stimulation of Cellular Proliferation and Bioenergetic Metabolism.

According to our analysis of representative components of signaling pathways associated with tumorigenesis, NKG2D–DAP10 displayed activities similar to growth factor receptors such as EGFR and insulin-like growth factor-1 receptor (IGF-1R) (32, 33). We thus explored effects of NKG2D stimulation on cellular functions using the MCF-7 mock-transfected control and MCF-7–TF model cells. Cells were plated at near confluence in the absence of growth factors with or without the relevant anti-MIC and anti-ULBP1/3/4 Abs or control Ig 24 h before assay time (Fig. S3). Cell-cycle analysis by propidium iodide (PI) staining identified significantly enlarged proportions of MCF-7–TF cells with DNA content corresponding to S- and G₂-phase transitions. This effect was reversed in the presence of the anti-NKG2D ligand Ab mixture or by RNAi targeting of NKG2D (Fig. 4A). These results confirmed ligand-mediated NKG2D stimulation and were corroborated by determinations of cellular ATP as an independent parameter of cellular proliferation (Fig. 4B).

Fig. 4.

Stimulation of proliferation and bioenergetic metabolism. (A) Cell-cycle analysis of MCF-7 mock-transfected, MCF-7–TF, and MCF-7–TF–NKG2D RNAi cells, plated for 24 h in the presence or absence of anti-MIC/ULBP1/3/4 Ab mixture, by PI staining and quantitative evaluation of flow cytometry data based on Dean–Jett–Fox curve fitting. Data shown are representative of three experiments. P values indicate statistical significance of data pair comparisons. (B) NKG2D–DAP10 signaling is associated with increased metabolic activity. MCF-7 mock-transfected (control) and MCF-7–TF cells were compared for total cellular ATP (Left), for real-time OCR (Center), and for ECAR (Right). P values indicate statistical significance. Data shown are representative of three experiments.

Changes in bioenergetic cell metabolism were measured using an extracellular flux analyzer that allows real-time determinations of oxygen consumption rates (OCR) and of extracellular acidification rates (ECAR) as a measure of glycolysis-derived lactic acid (34). Both energy-producing pathways, oxidative phosphorylation and glycolysis, were markedly stimulated in MCF-7–TF cells (Fig. 4B). We conclude that, based on key criteria tested, in vitro stimulation of cellular proliferation, and bioenergetic metabolism, NKG2D–DAP10 is functionally similar to tumor growth factor receptors.

Correlation Between NKG2D Expression and Tumor Progression.

To obtain evidence for the pathophysiological significance of NKG2D expression in cancers, we tested for clinical correlations. Pathology reports for all 60 primary cancer specimens examined were abstracted for clinical and histopathological information, and data on tumor size/spread, lymph node involvement, and, where available, tumor metastasis used to generate American Joint Committee on Cancer (AJCC) staging (Table S1). To allow a combined analysis of all four cancer types studied, we applied tumor size/spread, node involvement, and metastasis (TNM) status rather than cancer type-specific staging. Survival/outcome information was not available. Linear regression analysis revealed significant correlations between mean percentages of NKG2D⁺ cancer cells (%2D) with tumor stage [P < 0.0001; stage I (n = 14), %2D 3.7; stage II (n = 15), %2D 10.9; stage III (n = 30), %2D 28.3; stage IV (n = 1), %2D 45.7] and tumor size/spread [P < 0.0001; T1 (n = 12), %2D 4.5; T2 (n = 13), %2D 6.3; T3 (n = 34), %2D 27.7; T4 (n = 1), %2D 34.0]. By t-test assessment, a statistically significant association also was observed with lymph node status [P < 0.009; negative (n = 33), %2D 12.3; positive (n = 18), %2D 24.6]. There were no statistically significant associations with tumor grade (P = 0.025), presence or absence of lymphatic and/or vascular invasion (P = 0.10), or patient age (P = 0.59).Together, these results provide ex vivo correlative evidence associating NKG2D expression with criteria of tumor progression, thereby lending support to its tumor growth factor receptor-like stimulatory functions.

Discussion

This study has uncovered the surface expression and signaling proficiency of the human NKG2D–DAP10 immunoreceptor in cancer cells. Because NKG2D–DAP10 activates oncogenic signaling cascades, it may complement the presence of its ligands for stimulation of tumor growth. This functional relationship is supported by in vitro proliferation and bioenergetic metabolism data and by the positive correlation of NKG2D with criteria of tumor progression. Tentatively, NKG2D thus may resemble tumor growth factor receptors such as the EGFR, IGF-1R, and TGF-β receptor type 1 (TGFβ1R). Although these receptors also have roles in tissue development and differentiation, NKG2D normally is confined to regulating lymphocyte functions alone.

Irrespective of their tissue origin, nearly all cancer specimens heterogeneously expressed NKG2D within wide proportional ranges. The identity of cancer cells was ascertained by staining for EpCAM and confirmation of absence of CD45. Because loss of surface EpCAM is common (35), proportions of NKG2D⁺ cancer cells may be underestimated. At this point, there is no hint at mechanisms of gene regulation that may underlie cancer-cell expression of NKG2D and DAP10, which might be induced by cytokines or growth factors in permissive cancer-cell subsets. It may be relevant here that the promoter region of the DAP10 gene includes an activator protein 1 (Ap-1) transcription factor-binding site (36). Conversely, constitutive expression of NKG2D–DAP10 opposed by inducible down-regulation [e.g, by TGF-β, macrophage migration inhibitory factor (MIF), or IL-21] is a formal although unlikely possibility (37–39). In contrast to ex vivo cancer cells, most tumor lines expressed only scarce amounts of NKG2D–DAP10, which nonetheless were signaling proficient as substantiated by RNAi targeting and transfection experiments. Tumor lines may down-modulate NKG2D–DAP10 as an escape from detrimental effects of ligand-mediated self-stimulation under in vitro culture conditions. We have not explored ectopic NKG2D receptor expression in hematopoietic malignancies, which has been reported for acute myeloid leukemia. However, the significance here is less clear and presumably is unrelated to our observations in cancers, because NKG2D appears to confer cytolytic ability (40).

The question arises as to whether NKG2D imparts cellular effects in cancer environments that are distinct from or merely synergistic to those conferred by prototypic growth factor receptors. Distinct from T cells, we observed no stimulatory effect of noncross-linked recombinant sMICA in suspended cancer cells and tumor lines. However, sMICA could be stimulatory, given permissive NKG2D licensing in specific cancer environments (20).

In conclusion, our results suggest that the NKG2D immunoreceptor expressed by human cancer cells may promote tumor progression and thus represent a determining factor underlying the typically poor clinical outcomes that have been correlated with tumor-associated expression of its ligands (1, 15–17). This tentative role of NKG2D will require scrutiny by in vivo experimental models of tumor growth. If confirmed, previous mouse models of NKG2D-mediated tumor immune surveillance may not account adequately for human cancer conditions, because our data suggest the absence of NKG2D expression in mouse cancer cells. We propose that in humans NKG2D-mediated promotion of tumor growth becomes effective concurrent with the failure of immune defense at advanced tumor stages. However, dynamics resulting from the contributions of NKG2D to both tumor immune surveillance and growth cannot be considered, because there is no knowledge of its expression during earlier tumorigenesis stages.

Materials and Methods

Cancer Specimens and Cell Lines, Immunohistochemistry, and Flow Cytometry.

Primary cancer, nonaffected tissue specimens, and pathology reports were provided by the Cooperative Human Tissue Network (CHTN)–Western Division. This activity was approved by the Institutional Review Board of the Fred Hutchinson Cancer Research Center. If two tumor grades were found within one cancer specimen, the higher grade was used for classification. Tumor lines were from the American Type Culture Collection. Details of cell culture conditions, immunohistochemistry, flow cytometry, antibodies, and conjugates are given in SI Materials and Methods.

Immunoprecipitations and Immunoblots.

NKG2D was immunoprecipitated using mAb 5C6 immobilized on AminoLink Plus Coupling Resin (Pierce). DAP10 was immunoprecipitated using rabbit polyclonal antibodies (FL-93; Santa Cruz Biotechnology). Immunoblots were probed with polyclonal antibodies to NKG2D, DAP10 (N-20 and N-17; Santa Cruz Biotechnology), or Grb2 (Cell Signaling Technology) and developed using secondary reagents and Supersignal West Dura Extended Duration Substrate (Pierce). Further details are given in SI Materials and Methods.

RT-PCR, NKG2D-DAP10 Transfection, and siRNA Transduction.

Details of PCR primers and conditions, qPCR, transfections, and siRNA sequences are given in SI Materials and Methods. Annealed siRNA oligonucleotides were ligated into lentiviral pRRLsin-cPPT-PGK-GFPwpre vector modified by insertion of a U6 gene promoter cassette (41).

Cell-Cycle Analysis and Metabolic Activity Assays.

Details of cell-cycle analysis are given in SI Materials and Methods. Cellular ATP was determined using the ATPlite luminescence assay system (PerkinElmer). Real-time oxygen consumption and extracellular acidification rates were measured using the Seahorse Bioscience Extracellular Flux Analyzer XF24 (34).

Statistical Analysis.

Details of statistical analysis are given in SI Materials and Methods.