A truncated human NKG2D splice isoform negatively regulates NKG2D-mediated function

Abstract

NKG2D is a stimulatory receptor expressed by NK cells and a subset of T cells. NKG2D is crucial in diverse aspects of innate and adaptive immune functions. Here, we characterize a novel splice variant of human NKG2D that encodes a truncated receptor lacking the ligand-binding ectodomain. This truncated NKG2D (NKG2D^TR) isoform was detected in primary human NK and CD8⁺ T cells. Overexpression of NKG2D^TR severely attenuated cell killing and IFNγ release mediated by full-length NKG2D (NKG2D^FL). In contrast, specific knockdown of endogenously expressed NKG2D^TR enhanced NKG2D-mediated cytotoxicity, suggesting that NKG2D^TR is a negative regulator of NKG2D^FL. Biochemical studies demonstrated that NKG2D^TR was bound to DAP10 and interfered with the interaction of DAP10 with NKG2D^FL. In addition, NKG2D^TR associated with NKG2D^FL, which led to forced intracellular retention, resulting in decreased surface NKG2D expression. Together, these data suggest that competitive interference of NKG2D/DAP10 complexes by NKG2D^TR constitutes a novel mechanism for regulation of NKG2D-mediated function in human CD8⁺ T cells and NK cells.

Introduction

NKG2D is a homodimeric, stimulatory type II transmembrane receptor expressed by most NK cells, CD8⁺ T cells, γδ T cells, and NKT cells in humans (1, 2). NKG2D transmits activating signals through its interaction with the signal-transducing adaptor molecule, DNAX-activated protein of 10 kD (DAP10), which can trigger cell-mediated cytotoxicity and cytokine production. NKG2D recognizes stress-inducible ligands structurally related to MHC-I molecules including MICA, MICB, and ULBP1,2,3,4 in humans and retinoic acid early-inducible (Rae)-1, H60, and MULT-1 in mice. These ligands are minimally expressed by normal tissues but can be upregulated by cellular stress and DNA damage response pathways, such as those initiated during viral infection, tumorigenesis, and exposure to DNA alkylating chemotherapeutic agents (3-5).

NKG2D plays a crucial role in many aspects of immune function and immune surveillance. For example, NKG2D blockade or deficiency has been shown to enhance susceptibility towards inducible and spontaneously growing tumors (6-9). NKG2D also contributes to viral immunity. In cytomegalovirus infection studies, deletion of virally encoded proteins that prevent expression of NKG2D ligands moderates virulence during early infection, an effect that can be reversed by NKG2D blockade (10, 11). In addition to NK cells, NKG2D also plays an important role in CD8⁺ T cell cytotoxicity. Human CD8⁺ T cells activated ex vivo by TCR crosslinking in the presence of IL-2 and IFN-γ up-regulate cytotoxic effector molecules and potently lyse a broad range of malignant cell lines and primary tumor samples (12-14). The cytotoxic effect of these human CD8⁺ T-cells is NKG2D- dependent, since NKG2D knockdown attenuates their killing activity (15).

NKG2D gene expression is subject to post-transcriptional regulation by splicing variation (16). In mice, NKG2D isoforms generated by alternative mRNA splicing include long (NKG2D-L) and short (NKG2D-S) variants, which allows NKG2D-S to pair with DAP12 in addition to DAP10 (17, 18). The human NKG2D gene is expressed from at least 3 distinct alleles, and several isoforms at the mRNA level have been described, including an alternatively spliced variant that introduces a nonsense mutation resulting in a protein that lacks the entire extracellular ligand binding domain (19). To date, the significance of this splicing variation is not well understood and the functional aspects of the human splice variants have not been analyzed.

In this study, we characterized a novel splice variant of human NKG2D in CD8⁺ T cells and NK cells. The splice variant encodes a truncated form of NKG2D (NKG2D^TR), which acts as a dominant negative regulator of NKG2D function by down-regulating full-length NKG2D (NKG2D^FL) surface expression through intracellular retention. This is the first example of a dominant negative splice variant. Thus, NKG2D^TR acts as a negative regulator of NKG2D-mediated function. Our data provide novel insights into the regulatory mechanisms of cytolysis as a potential opportunity to strengthen the broad anti-tumor potential of these cells.

Materials and Methods

Reagents, cell lines, and flow cytometry

Monoclonal antibodies were purchased from e-biosciences (San Diego, CA): anti-human OKT3, anti-CD28 clone CD28-6; BD Biosciences (San Jose, CA): anti-CD16 FITC clone 3G8 BD, anti CD56 PE clone B159, anti-CD314/NKG2D PE or allophycocyanin, BD clone 1D11, isotype controls (IgG) as either FITC or PE conjugates; R&D systems (Minneapolis, MN): anti-human NKG2D clone 1 49810; Sigma Aldrich (St Louis, MO): anti-gamma Tubulin clone 4D11; Fisher Scientific (Pittsburg, PA) anti-HA HRP clone HA7, anti-Flag clone M2 HRP, anti-Myc HRP A5598; Santa Cruz Biotechnology (Santa Cruz, CA): anti-human NKG2D clone N-20, NKG2D peptide N-20, DAP10 peptide; Abcam (Cambridge, MA): anti-human NKG2D 5C6. All cell culture reagents and chemicals were purchased from Invitrogen (Grand Island, NY) and Sigma Aldrich, respectively, unless otherwise specified. COS-7 cells and HEK293T cells were cultured in complete DMEM media (cDMEM; 10% FCS; Invitrogen), 2 mM L-glutamine, 100 U/mL penicillin, 100 μg/mL streptomycin). The P815 murine mastocytoma cell line was purchased from the American Type Culture Collection (Manassas, VA) and cultured as recommended. Flow cytometry was performed on antibody-labeled or transduced cells by a BD LSR-II instrument (BD Biosciences). Transduced cells were cell-sorted with a FACSAria (BD Biosciences).

Generation of activated CD8⁺ T cells and NK cells

Human peripheral blood lymphocytes were obtained from buffy coats of healthy donors by Ficoll-Hypaque density centrifugation. CD8⁺ T cells were enriched with CD8-coated MicroBeads followed by cell sorting by FACS. CD8⁺ T cells were activated either with plate- bound anti-CD3 plus anti-CD28 (10ng/ml) or with 1000 U/mL human interferon-γ (IFN-γ) (Genentech, South San Francisco) followed on day 1 by stimulation with anti-CD3 (25 ng/mL; OrthoBioTech, (Horsham, PA) and recombinant human IL-2 (300 U/mL; Chiron, (Emeryville, CA) as previously described (20). Every 3 to 5 days thereafter, new cRPMI media was added and supplemented with IL-2 (300 U/mL) to maintain a cell density of 1.5 to 2 × 10⁶/mL for a total of 14 days. NK cells were positively selected by incubating PBMCs with CD16-coated MicroBeads using MACS (Miltenyi Biotech, Auburn CA). NK cells were further purified by cell sorting for the CD56⁺ population as described previously (15). Consent was obtained from all donors of peripheral blood lymphocyte buffy coats under protocols approved by the Institutional Review Board at the University of Massachusetts (IRB2011-1051).

PCR and sequence analysis of NKG2D cDNA

RNA was isolated from purified CIK-cells using the RNeasy Mini Kit (Qiagen) and reverse transcribed with the iScript cDNA Synthesis Kit (BIO-RAD). Primers used for PCR amplification of NKG2D cDNAs are Forward: CTGGGAGATGAGTGAATTTCATA Reverse: GACTTCACCAGTTTAAGTAAATC PCR amplification was performed as previously described(16). PCR products were separated by agarose gel electrophoresis and subsequently cloned into TA vectors (Invitrogen). Cloned inserts were sequenced with an automated DNA sequence analyzer. DNA sequences of human NKG2D splice variants were analyzed using Vector NTI software (Invitrogen). Primers used for NKG2DTR Forward: ACGATGGCAAAAGCAAAGATGTCC. Reverse ACGGATTCCCATGGCTACAGCG

Specific knockdown of the NKG2D^TR alternative splice variant by shRNA

The oligonucleotide sequence TCCAGAAATTTGTAAACTA (136-155 bp of intron 4) was designed to target a unique position within intron 4 specifically retained in NKG2D^TR mRNA, but has no sequence homology to NKG2D^FL mRNA as determined by BLAST sequence analysis. This sequence was incorporated within a short hairpin structure, using the stem loop sequence 5'-TCAAGAGA-3', and cloned in the plasmid pSICO-GFP, as previously described (21). HEK293T cells were transfected using the calcium phosphate method with 20μg pSICO lentivirus, 10 μg of Vesicular Stomatitis Virus G plasmid, and 15 μg of CMV ΔR8.74 according to standard protocols. After 16 hours, the medium was changed, and recombinant lentivirus vectors were harvested 24-48 hours later. Lentiviral infection of activated CD8⁺ T cell or NK cell was performed three times at 24-hour intervals. For each infection, cells were plated in 48-well plates at 1 × 10⁵ cells/well and infected in the presence of protamine and hexadimethrin. Spin infection was performed at 1200 rpm for 90 min at 37°C. Four days after the first infection, transduced cells were isolated by FACS sorting of GFP⁺ cells to >97% purity.

Generation of sMICA

The cDNA encoding the human NKG2D ligand MICA was obtained by RT-PCR on RNA from normal human B cells and subcloned into the pFUSE-hFC1 plasmid (Invivogen) to create a MICA-Fc-fusion protein between MICA and the Fc region (CH2 and CH3 domains) of the human IgG1 heavy chain. HEK293T cells were transduced with the pFUSE-hFC1-MICA plasmid using Lipofectamine 2000. Supernatants containing sMICA-Fc fusion protein from transfected HEK293T cells were collected and sMICA-Fc was concentrated using Millipore ultra centrifuge filter (10k) and was further purified by an Fc-binding column (Hiprop; GE Healthcare life Science). The purified sMICA was used to stimulate CD8⁺ T cells.

IFNγ production by activated CD8⁺ T cells

CD8⁺ T cells were retrovirally transduced with pQCN-NKG2D^FL-Flag-IRES-GFP or pQCNNKG2D^TR-myc-IRES-GFP. Transduced cells were sorted for GFP and 5 × 10⁶ sorted cells were cultured as described earlier. Cells were washed and stimulated with plate-bound anti-NKG2D (5 μg/ml), sMICA (10 μg/ml), or anti-CD3 (10 μg/ml). 12 hours later, cell-free supernatants were collected and IFNγ content was measured by ELISA (eBioscience) according to the manufacturer's instructions.

Expression vector, retrovirus, and lentivirus construction and transduction

Flag-tagged NKG2D^FL and Myc-tagged NKG2D^TR were cloned into pcDNA TOPO (Invitrogen, Carlsbad, CA) and subcloned into the retroviral vector pQCNXIX (Clontech, Mountain View, CA) generating pQCN-NKG2D^FL-Flag-IRES-GFP and pQCN-NKG2D^TR-myc-IRES-GFP. In addition, Myc-tagged NKG2D^TR and HA-tagged DAP10 were cloned into pC-GFP plasmid (Clontech) to create in-frame C-terminal GFP fusion proteins and subcloned into the retroviral vectors pQCNXIX, pQCN-NKG2D^TR-myc-IRES-GFP, or pQCN-NKG2D^FL-Flag-IRES-GFP generating the constructs pQCN-DAP10^HA=GFP, pQCN-NKG2D^FL-Flag-IRES-NKG2D^TR-Myc=GFP, NKG2D^FL-Flag-IRES-DAP10^HA=GFP, or NKG2D^TR-Myc-IRES-DAP10^HA.GFP. DAP10^HA was also cloned into the pLVX-Dsred-Monomer-N-1 (Clontech) to create an in-frame N-terminal RFP fusion protein. DAP10^HA=RFP was subcloned into the retroviral vector pQCNXIX generating pQCN-DAP10^HA=RFP. To create retroviral vector particles, Phoenix-A cells were transfected with these retroviral vector plasmids using Lipofectamine 2000 (Invitrogen). 48-72 hours post-transfection, supernatants were collected and, along with 5μg/ml protamine sulfate, added to CD8⁺ T cells. Cells were centrifuged for 90 min at 200 g, and resuspended in RPMI media. The transduction efficiency ranged from 20-40%. Transduced cells were sorted by flow cytometry for expression of GFP, RFP, Flag-tag, myc-tag, or NKG2D C-terminus using a specific antibody. In experiments where all 3 proteins were transduced (NKG2D^FL, NKG2D^TR, and DAP10), the cells were transduced with two separate retroviruses. In experiments involving transient transfections, HEK293T and COS-7 cells were transfected with the constructs using Lipofectamine 2000. pcDNA3.1 vectors encoding NKG2D^FL-Flag, NKG2D^TR-Myc, and DAP10^HA were used in transient transfection experiments using COS7 and HEK293T cells.

⁵¹Cr release cytotoxicity assay

P815 cells were labeled with ⁵¹Cr (DuPont-NEN, Boston, MA), washed, re-suspended in cRPMI, and plated in 96-well plates at a concentration of 1 × 10⁴ cells/ml in triplicates. Anti-NKG2D antibody (5 μg/ml) was added and incubated for 30 minutes before washing and plating. Effector cells were added at specified ratios and incubated at 37°C for 4 hours. At the end of each assay, supernatants were collected and counted using a gamma counter (Cobra/AII). Cytotoxicity was calculated by the following equation: % specific lysis = 100 × ((test release) - (spontaneous release)) / ((maximal release) - (spontaneous release)).

Western blotting and immunoprecipitation

Cells were lysed in freshly prepared lysis buffer (10 mM Tris pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% Nonidet-P40, 0.5% deoxycholate, 0.1% SDS, complete Protease Inhibitor Cocktail [Roche, Palo Alto, CA], and 500μM phenylmethylsulfonylfluoride) and centrifuged for 10 minutes at 4°C. Aliquots containing 70 μg of protein were separated on an 18% denaturing polyacrylamide gel and transferred to polyvinylfluoride membranes for immunoblot analysis using antibodies specific to Flag, myc, HA, or different regions of NKG2D. In experiments involving peptide blockade, anti-NKG2D antibody (1 μg/ml) was mixed with either specific (Anti NKG2D peptide N-20:P; Santa Cruz Biotechnology) or non-specific peptide (Abcam aggrecan peptide; Abcam) at 2 μg/ml in 2 ml of blocking buffer. After agitation overnight at 4°C, Western blot analysis was performed on the cell lysates. For immunoprecipitation, Flag-tagged, myc-tagged, and HA-tagged proteins in lysates of transduced cells were recovered by incubation with 10μg of anti-Flag, anti-myc, or anti-HA antibody for 2 hours at 4°C followed by binding to protein G agarose beads (Pierce). After washing, immunoprecipitated lysates were separated and resolved by Western blot analysis as described above.

Immunofluorescence-based cell imaging

COS-7 cells (CRL-1651, ATCC) were transfected using Extreme Gene (Roche) following the manufacturer's instructions. Transfected cells were plated on glass coverslips coated with 10 μg/ml of bovine fibronectin (Sigma Aldrich). Cells were fixed in PBS with 3.7% formaldehyde, permeabilized in PBS containing 0.5% Triton X-100, blocked with PBS containing 1% BSA and 10% goat serum, followed by incubation with anti-Flag antibody. Cells were washed and stained with goat anti-mouse Dilight-488 secondary antibody. Nuclei were stained using DAPI. Cells were mounted in Vectashield (Vector) and imaged using a Zeiss 200M inverted microscope with the Axiovision software.

Statistical analysis

Statistical significant differences in the observed mean values of experimental groups were determined by ANOVA using Excel (Microsoft, Seattle, WA) or Prism computer software (GraphPad, San Diego, CA).

Results

Identification of an NKG2D splice isoform in human CD8⁺ T cells and NK cells

Human NKG2D is encoded by a single gene consisting of 10 exons located in the NK gene complex (22). Exons 2-4 encode the intracellular/transmembrane domain, followed by exons 5-8 that encode the ligand-binding ectodomain (16) (Fig. 1A). Analysis of NKG2D gene transcripts expressed from cells in human peripheral blood identified an additional unexpected NKG2D mRNA transcript by RT-PCR using primers specific to regions in exons 4 and 5 (Fig. 1B). DNA sequence analysis demonstrated that the unexpected PCR product resulted from an alternatively spliced NKG2D transcript that included intron 4 retained between exons 4 and 5. This intron encoded a translational stop codon, potentially creating a truncated form of NKG2D (NKG2D^TR) without the extracellular ligand-binding domain (exons 5-8; Fig. 1A). PCR using primers specific to regions in intron 4 showed a band in mRNA from CD8⁺ T cells only in the presence of reverse transcriptase, confirming the presence of NKG2D mRNA with retained intron 4 (Fig. 1C).

Figure 1. A truncated NKG2D splice isoform is detected in lysates from PBMCs, NK cells, and CD8⁺ T cells.

(A) A schematic representation of the human NKG2D gene is depicted. Shown are the exons encoding NKG2D^FL mRNA and the protein product. The alternatively spliced NKG2D^TR mRNA includes intron 4 that introduces a premature stop codon (denoted by a black hexagon). The truncated NKG2D^TR protein includes the cytosolic (CY), transmembrane (TM) but lacks the stalk (ST), C-terminal extracellular C-type lectin-like (EC), and ligand-binding domain (LBD). (B) RT-PCR with primers against NKG2D exons 4 and 5 was performed on RNA isolated from activated CD8⁺ T cells from 4 donors. (C) PCR with primers against NKG2D intron 4 was performed on RNA isolated from activated CD8⁺ T cells from 2 donors with (left 2 lanes) or without (right 2 lanes) reverse transcriptase (RT). (D) Western blot analysis was performed using human PBMCs and purified peripheral blood NK cells or (E) activated CD8⁺ T cells from 4 donors by using an antibody to the N-terminus of NKG2D. The molecular weight (MW) markers are shown on the left lane. % NKG2D^TR was calculated based on densitometric quantification: (NKG2D^TR density) / (NKG2D^TR density + NKG2D^FL density) × 100%. γ-tubulin served as a loading control for the Western blot analyses.

To determine whether a truncated protein corresponding to this transcript was detectable in these cells, we probed for NKG2D protein in PBMCs, NK cells, and activated CD8⁺ T cells. Using an NKG2D N-terminus-specific antibody, Western blot analysis revealed 2 bands with molecular weights of ~40 kD and ~10 kD, corresponding to the expected sizes of full-length NKG2D (NKG2D^FL) and NKG2D^TR, respectively in PBMCs, CD8⁺ T cells, and NK cells (Fig. 1D, E). Both bands corresponded to specific recognition by the antibody, since the addition of an NKG2D-derived but not a non-specific peptide specifically blocked detection of both bands (Supplemental Fig. 1). Next we investigated whether the relative abundance of the truncated NKG2D isoform is influenced by cell activation. CD8⁺ T cells from 3 different donors were stimulated using anti-CD3 in the presence of IFN-γ and IL-2. Cell lysates from CD8⁺ T cells activated for 0, 24, 48 or 72 hours were analyzed for NKG2D^FL and NKG2D^TR protein expression. Western blot analysis showed that the relative percentage of NKG2D^TR to NKG2D^FL did not change after activation (Supplemental Fig 2).

NKG2D^TR suppresses NKG2D-dependent function

Given that NKG2D^TR lacks an extracellular domain, we predicted that expression of NKG2D^TR would impact the function of NKG2D^FL. To test the role of NKG2D^TR in NKG2D-dependent cytotoxicity, we employed a gain and loss of function approach in CD8⁺ T cells. Overexpression of NKG2D^TR significantly decreased NKG2D-dependent cytotoxicity (redirected lysis by anti-NKG2D antibody) by activated human CD8⁺ T cells. Decreased cytotoxicity by NKG2D^TR was observed even when NKG2D^FL and DAP10 were overexpressed in these cells (Fig. 2A, A Figure 3). Similar results were obtained with cytotoxicity against K562 cells, which endogenously express the NKG2D ligands MICA and MICB (Supplemental Figure 3B). Importantly, the transduction of activated CD8⁺ T cells with NKG2D^TR was not associated with global adverse effects on cytotoxic function, since TCR-dependent lysis of HLA-A2-restricted, influenza M1 peptide-specific activated CD8⁺ T cells against antigen-pulsed target cells was not affected by transduction with NKG2D^TR (Fig. 2B).

Figure 2. NKG2D^TR suppresses NKG2D-dependent cytotoxicity.

(A) Human CD8⁺ T cells were untransduced or transduced with GFP-expressing vector or different combinations of lentiviruses encoding NKG2D^TR-Myc, NKG2D^FL-Flag, DAP10^HA=GFP, or DAP10^HA=RFP. The cytotoxicity of the transduced human CD8⁺ T cells against P815 cells labeled with or without anti-NKG2D antibody was determined. One representative of n = 6 independent experiments is shown. (B) Activated HLA-A2-restricted, influenza M1-peptide-specific CD8⁺ T cells were transduced with vector alone or NKG2D^TR-Myc and cytotoxicity against peptide-pulsed or unpulsed target cells was determined. One representative of n = 3 experiments is shown. (C) Activated CD8⁺ T cells were untransduced or transduced with retroviruses encoding NKG2D^TR-Myc, or NKG2D^FL-Flag. Cells were stimulated with either plate-bound anti-NKG2D, sMICA, or anti-CD3. Cell-free supernatants from were tested for IFNγ by ELISA and are shown as mean concentration (pg) ± SD of triplicate determinations. One representative of 2 different donors is shown. (D) Human CD8⁺ T cells were transduced with scrambled shRNA, or shRNA against NKG2D^TR. NKG2D protein expression was detected by Western blot analysis with the ~40 kD and ~10 kD bands representing NKG2D^FL and NKG2D^TR, respectively. One representative of n = 3 donors is shown. (E) The cytotoxicity of shRNA-transduced human CD8⁺ T cells at a 40:1 E:T ratio against P815 cells labeled with anti-NKG2D antibody was determined. One representative of n = 3 donors is shown. All cytotoxicity assays were performed at a 40:1 E:T ratio and are represented as mean % specific lysis ± SD of triplicate determinations. *, **, and *** indicate significance of p<0.05, p<0.01, and p<0.001 by ANOVA, respectively. NS = not significant.

Figure 3. NKG2D^TR downregulates surface expression of NKG2D^FL.

(A) Human CD8⁺ T cells were untransduced, transduced with vector (GFP), or with different combinations of retroviruses encoding NKG2D^FL-Flag, NKG2D^TR-myc=GFP, NKG2D^TR-myc, and DAP10^HA=GFP, and analyzed for NKG2D^FL surface expression. NKG2D^TR-myc=GFP was used in double transduction of NKG2D^TR and NKG2D^FL. Filled histograms and solid lines represent cells stained with IgG control antibody and anti-NKG2D, respectively. One representative of n = 6 independent experiments is shown. (B) Human CD8⁺ T cell cells were transduced with vectors expressing either scrambled shRNA (solid line) or NKG2D^TR shRNA (dotted line) and analyzed for NKG2D^FL surface expression. The filled histograms represent cells stained with IgG control antibody. One representative of n = 3 donors is shown. (C) Human CD8⁺ T cells were transduced with GFP alone or with NKG2D^TR-Myc. Surface and intracellular NKG2D was detected by flow cytometry. Filled histograms and solid lines represent cells stained with IgG control antibody and anti-NKG2D, respectively. One representative of n = 3 experiments is shown.

Since NKG2D induces cytokine production from activated CD8⁺ T cells, we next examined whether NKG2D^TR impacts NKG2D-mediated IFNγ release by CD8⁺ T cells. CD8⁺ T cells were retrovirally transduced with either NKG2D^TR or NKG2D^FL and stimulated with plate-immobilized anti-NKG2D, sMICA, or anti-CD3 for 12hours. Compared to CD8⁺ T cells untransduced or transduced with NKG2D^FL, CD8⁺ T cells overexpressing NKG2D^TR produced little IFNγ in response to NKG2D stimulation (Fig. 2C). In contrast, anti-CD3 stimulation induced similar IFNγ levels in all CD8⁺ T cells, suggesting that the inhibitory effect of NKG2D^TR is specific to NKG2D stimulation.

To determine whether the endogenous NKG2D^TR splice variant physiologically regulates endogenous NKG2D^FL function, we designed shRNA sequences to specifically knockdown NKG2D^TR expression, while leaving NKG2D^FL expression intact. Since the NKG2D^TR splice variant sequence retains intron 4, we targeted a unique position within this region that shares no homology with the NKG2D^FL sequence (intron 4 136-155 bp). Control and experimental shRNA oligonucleotides were delivered by lentiviral vectors into activated human CD8⁺ T cells. Immunoblotting analyses confirmed specific knockdown of the endogenous NKG2D^TR protein, as there was a 78-93% reduction of the NKG2D^TR to NKG2D^FL ratio in cells transduced with experimental shRNA compared to scrambled shRNA (Fig. 2D). Specific knockdown of endogenous NKG2D^TR in activated CD8⁺ T cells resulted in significantly enhanced cytotoxicity mediated by NKG2D^FL (Fig. 2E). Thus, NKG2D^TR acts as a physiologic dominant-negative regulator of NKG2D^FL-dependent cytotoxicity.

NKG2D^TR downregulates surface expression of NKG2D^FL

To determine how NKG2D^TR negatively regulates NKG2D^FL, we examined the cell surface expression of NKG2D on activated CD8⁺ T cells after transduction with different combinations of NKG2D^TR, NKG2D^FL, and/ or DAP10. Activated CD8⁺ T cells overexpressing NKG2D^TR, with or without DAP10 overexpression, showed a significant reduction in cell surface NKG2D expression (Supplemental Figure 4A). Importantly, the overexpression of NKG2D^FL and/or DAP10 was unable to overcome the negative effect of NKG2D^TR on NKG2D cell surface expression (Fig. 3A, Supplemental Figure 4A). Next, NKG2D^TR was specifically knocked down in CD8⁺ T cells and cell surface NKG2D was assessed. Silencing of NKG2D^TR significantly increased surface expression of NKG2D^FL (Fig. 3B, Supplemental Figure 4B). These data suggest that NKG2D^TR negatively regulates cell surface expression of NKG2D. In contrast to surface NKG2D expression, intracellular NKG2D was still detected in CD8⁺ T cells overexpressing NKG2D^TR (Fig. 3C, Supplemental Fig. 4C). This was consistent with the observation that overexpression of NKG2D^TR did not change total amounts of NKG2D^FL by Western blot analysis (Fig. 4A). These data suggest that NKG2D^TR downregulates surface NKG2D expression by retaining NKG2D^FL intracellularly.

Figure 4. NKG2D^TR does not change total NKG2D^FL expression but blocks the interaction of NKG2D^FL with DAP10.

(A). Human CD8⁺ T cells were transduced with either vector (GFP alone) or NKG2D^TR. Total cell lysates were blotted for NKG2D, γ-tubulin (loading control), or DAP10. Cell lysates were also immunoprecipitated (IP) with anti-NKG2D^FL antibody and immunoblotted (IB) with antibody against NKG2D^FL or anti-DAP10. One representative of n = 3 independent experiments is shown. (B) Total cell lysates from CD8⁺ T cells untransduced (left lane), transduced with scrambled shRNA (middle lane), NKG2D^TR shRNA vector (right lane) were blotted for NKG2D, γ-tubulin (loading control), or DAP10. Cell lysates were also immunoprecipitated (IP) with anti-NKG2D^FL antibody and immunoblotted (IB) with antibody against NKG2D^FL or anti-DAP10. One representative of n = 3 donors is shown.

NKG2D^TR pairs with DAP10 and NKG2D^FL, interferes with the association of DAP10 with NKG2D^FL, and retains NKG2D^FL in the intracellular compartment

We reasoned that NKG2D^TR could attenuate cell surface expression of NKG2D by competing for binding to DAP10 and/or by pairing with NKG2D^FL. To test these possibilities, NKG2D^TR was overexpressed and the ability of DAP10 to co-immunoprecipitate with NKG2D^FL was assessed. Compared to vector-transduced CD8⁺ T cells, NKG2D^TR overexpression markedly decreased binding of NKG2D^FL with DAP10 (Fig. 4A). Conversely, knockdown of NKG2D^TR resulted in enhanced association of NKG2D^FL with DAP10 (Fig. 4B).

Next, to test whether NKG2D^TR associates with DAP10 and/or form heterodimers with NKG2D^FL, the 3 constructs were transiently transfected into HEK293T cells. NKG2D^FL coimmunoprecipitated with NKG2D^TR indicating that NKG2D^FL and NKG2D^TR can form heterodimers (Fig. 5, A and B). DAP10 was also co-immunoprecipitated with both NKG2D^FL and NKG2D^TR (Fig. 5, B and C), indicating that both isoforms are able to associate with DAP10. The expression of NKG2D^TR did not affect the association of DAP10 with NKG2D^FL in these experiments, suggesting that overexpression of DAP10 overcomes the inhibition of the NKG2D^FL/DAP10 interaction by NKG2D^TR seen in CD8⁺ T cells (Fig. 4A). Alternatively, these differences may arise because the latter immunoprecipitation studies (Fig. 5) were performed with HEK293T cells instead of activated CD8⁺ T cells. However, DAP10 overexpression was insufficient to reverse NKG2D^TR-mediated inhibition of NKG2D surface expression (Fig. 3A) or restore NKG2D-mediated cytotoxic activity of CD8⁺ T cells (Fig. 2A). Thus, these data suggest that NKG2D^TR interference with the DAP10/NKG2D^FL interaction is only partially responsible for the reduced NKG2D^FL surface expression by NKG2D^TR cannot fully explain why cell surface expression of NKG2D is decreased by NKG2D^TR.

Figure 5. NKG2D^TR complexes with DAP10 and NKG2D^FL.

(A) HEK293T cells were either left untransduced or transiently transfected with DAP10^HA in combination with NKG2D^FL-Flag and/or NKG2D^TR-Myc as indicated. Anti-Flag or anti-Myc immunoprecipitations (IP) were performed on transfected cell lysates, and then subjected to immunoblotting (IB) with anti-Flag, anti-Myc or (B) anti-HA antibodies. (C) Cell lysates were immunoprecipitated with anti-HA antibody and immunoblotted with anti-Flag or (D) anti-Myc antibodies. One representative of n = 4 independent experiments is shown.

To determine the intracellular localization of the full-length and truncated NKG2D molecules, we overexpressed NKG2D^TR and/or NKG2D^FL with DAP10 and membrane-targeted mCherry in COS-7 cells. Similar to the expression pattern seen in CD8⁺ T cells, only NKG2D^FL but not NKG2D^TR was found at the cell surface of transduced COS-7 cells (Fig. 6A, Supplemental Fig. 4D). Using anti-Flag antibodies to detect NKG2D^FL/FLAG, we found that the distribution of NKG2D^FL was consistent with that of a typical transmembrane protein; the majority of NKG2D^FL was clustered at the cell membrane with some perinuclear fluorescence (Golgi) observed (Fig. 6B). In contrast, in cells expressing both NKG2D^FL and NKG2D^TR, the staining was restricted to perinuclear localization and almost no surface membrane NKG2D^FL clusters were observed (Fig. 6B). These data suggest that NKG2D^TR heterodimerization with NKG2D^FL, causes intracellular retention of NKG2D^FL.

Figure 6. NKG2D^TR retains NKG2D^FL in the intracellular compartment.

(A) COS-7 cells were transfected with vectors expressing DAP10^HA, NKG2D^TR-myc, and/or NKG2D^FL-Flag as indicated. Surface NKG2D expression was detected by flow cytometry. (B) COS-7 cells were transfected with vectors expressing DAP10^HA and NKG2D^FL-Flag with or without NKG2D^TR-myc. NKG2D was visualized by structured light fluorescence microscopy using fluorescent dye-labeled anti-Flag antibodies (green). Nuclei were stained with DAPI (blue), and cell membranes marked with membrane-targeted mCherry (red). One representative of n = 2 independent experiments is shown.

Discussion

In this report, we have demonstrated a novel mechanism of negative regulation of NKG2D mediated-cytotoxic function through expression of an ectodomain-truncated isoform of NKG2D. The NKG2D^TR isoform is generated by alternative splicing which retains both the transmembrane and cytoplasmic domains but lacks the ligand-binding C-type lectin-like domain. Our data demonstrate that NKG2D^TR maintains the capacity to associate with NKG2D^FL and the adaptor molecule, DAP10. However, NKG2D^TR could not generate stimulatory signals for the induction of cytotoxicity by effector cells, likely due to its inability to bind to NKG2D ligands. Instead, NKG2D^TR formed heterodimers with NKG2D^FL and negatively modulated NKG2D^FL-dependent function, at least in part by preventing surface expression of NKG2D^FL. As such, forced expression of NKG2D^TR blocked whereas silencing of NKG2D^TR enhanced NKG2D-mediated cytotoxicity. Thus, our data support NKG2D^TR as a physiological negative regulator of NKG2D function.

The expression of receptor isoforms lacking all or parts of the extracellular ligand-binding domain results in diverse phenotypic effects in other systems. As exemplified by the RON receptor tyrosine kinase (23, 24), some receptor variants become constitutively activated due to deletion of membrane subdomains associated with down-regulation of protein function. Others such as splice variants of TrkA, Cytotoxic T lymphocyte antigen 4 (CTLA4), and CD28 allow constitutive activation in a ligand-independent manner (25-27). Our current study shows that NKG2D^TR functions similarly to splice variants of the growth hormone releasing hormone GHRH and D3 dopamine receptors, where pairing of an inactive truncated isoform with an intact full-length protein is associated with defective ligand binding (28, 29). However, in contrast to GHRH receptor and D3 dopamine receptor splice variants, negative regulation by NKG2D^TR also involves intracellular retention of functional NKG2D. This is the first example of a truncated splice variant that affects the cell surface expression of the full-length product. Thus, our results expand the known range of biological pathways that can be affected by specific splicing alterations.

Negative regulation of NKG2D function by NKG2D^TR likely involves the ability of NKG2D^TR to heterodimerize with NKG2D^FL. Our co-immunoprecipitation and protein binding data show that NKG2D^TR associates with both DAP10 and NKG2D^FL, implying that higher order hexameric complexes can be formed that are each composed of a NKG2D^TR/FL heterodimer attached to four DAP10 subunits (30). The ectodomains of NKG2D exhibit hallmark secondary structural elements of canonical C-type lectin-like proteins, consisting of two β-sheets followed by one (human NKG2D) or two (mouse NKG2D) α-helices (31-33). NKG2D homodimers interact with their ligands through a surface formed by their C-terminal region that is symmetrically and complementarily shaped to relevant ligand surfaces, with the α-helices of each NKG2D monomer making relatively equal contributions. Thus, two NKG2D monomers, homodimerizing through a tightly complementary interface are necessary for the binding of a single monomeric ligand (34-36). This suggests that pairing of NKG2D^FL with NKG2D^TR results in a chimeric heterodimer lacking functional ligand-binding capability, in turn providing a possible “interference” mechanism for the negative regulation of intact NKG2D^FL function by the NKG2D^TR isoform.

The finding that NKG2D^TR can pair with DAP10 is not surprising, because oligomerization occurs through interactions between acidic residues at or near the DAP10 dimer interface (on the cytosolic portion of NKG2D), as well as basic residues in the NKG2D transmembrane domain, all of which are preserved in NKG2D^TR (37). Non-covalent association of NKG2D homodimers with the DAP10 adaptor molecule is required for surface expression and function of NKG2D in human NK cells and activated CD8⁺ T cells (38). As endogenous NKG2D^TR can both pair with DAP10 and form heterodimers with NKG2D^FL, the reduction in surface receptor expression could be potentially explained by the sequestration of DAP10 by NKG2D^TR away from NKG2D^FL. This is supported by our data showing that knockdown of NKG2D^TR enhances while the overexpression of NKG2D^TR inhibits the interaction of NKG2D^FL with DAP10. However, the inability of DAP10 overexpression to restore NKG2D surface expression and function argues against this notion. While the sequestration of DAP10 by NKG2D^TR could still play a role in the inhibitory function of NKG2D^TR, features of NKG2D^TR other than DAP10 binding are likely responsible for intracellular NKG2D retention. Alternatively, the association of NKG2D^TR with NKG2D^FL may cause misfolding of this heterodimeric protein, which abrogates cell surface expression of NKG2D^FL. While we do not know the exact cellular compartment where these complexes reside, the absence of cell surface expression of NKG2D^FL when NKG2D^TR is overexpressed suggest that NKG2D^TR does not make it to the cell surface. Further studies will be required to thoroughly investigate the trafficking of these complexes.

The demonstrated negative effect of NKG2D^TR on NKG2D^FL function and overall NKG2D-dependent cytotoxicity, both at physiologically relevant endogenous levels and those enforced by overexpression, suggests potential roles for NKG2D^TR under both physiological and pathological conditions. The mechanisms underlying the regulation of NKG2D^TR expression among individuals is unknown. However, the manipulation of NKG2D^TR splice variant expression may have clinical implications. Both activated CD8⁺ T cells and NK cells have considerable therapeutic potential, particularly in immune-based approaches to tumor therapy (39). Thus, it is possible that expression of NKG2D^TR in donor lymphocytes would negatively impact graft-versus-tumor effects in bone marrow transplantation. Selection of donors with relatively less expression of NKG2D^TR might be beneficial for an augmented graft-versus-tumor effect. Alternatively, the expression of NKG2D^TR could be suppressed by RNA interference before adoptive transfer into recipients to enhance the cytotoxic potential of CD8⁺ T cells and NK cells. While it is unknown whether chimeric antigen receptor-expressing T cells require NKG2D for maximal effector function, silencing of NKG2D^TR might also be an attractive option to enhance the activity of genetically engineered tumor-targeting T cells.

In summary, we have identified a novel and important regulatory mechanism that regulates NKG2D-mediated cytotoxic function in activated CD8⁺ T cells. As NKG2D^TR affects the cell surface expression of NKG2D, the full delineation of the factors affecting the interaction between NKG2D^TR and NKG2D^FL, along with an elucidation of internal and external regulators of NKG2D splicing, may open doors to specific modulation of NKG2D function. This will enhance our understanding of the regulation of immune responses, which may lead to better treatment options for several diseases, including cancer and autoimmune disorders.

Abbreviations used

DAP10

DNAX-activated protein of 10 kD

MICA

MHC class I chain-related gene A

NKG2D^FL

full-length NKG2D

NKG2D^TR

truncated NKG2D variant

sMICA

soluble MICA