Two Human ULBP/RAET1 Molecules with Transmembrane Regions Are Ligands for NKG2D

Molecular cloning of the ULBP family

The 5′ end of RAET1G was predicted by alignment of the established sequence tag (EST) sequences AW510737, BE711112, and BF513861 and the genomic DNA (contig NT_023451.10). The predicted sequence matched with two IMAGE clones, 3070730 and 2911855. IMAGE clone 3070730 had a truncated 3′ end and missed the stop codon. The correct 3′ end was predicted from EST AA583860 and confirmed by PCR. The signal peptide was predicted with Signal PV1.1 (http://cbs.dtv.dk), and the predicted TM region was detected with TMpred (http://ch.embnet.org). The exon structures were analyzed with the GCG program (Wisconsin package, version 10.3; Accelerys, Cambridge, U.K.).

The alignment was based on the global amino acid sequences or on local domains of RAET1L (NM_130900.1), RAET1E (AY176317), RAET1G (AY172579), ULBP1 (NM_0225218.1), ULBP2 (NM_025217.2), ULBP3 (NM_024518.1), MICA (BC016929), MICB (NM_005922.1), and MULT1 (AK020784) and conducted using ClustalW (http://ebi.ac.uk/clustalw/) and PileUp. The UPGMA (unweighted pair group method with arithmetic mean) tree was constructed by the program MEGA version 2.1 (31). The consistency of the branches was assessed by bootstrap based on 1000 samples/runs.

The following clones were obtained from the I.M.A.G.E clone collection (HGMP, Hinxton, U.K.): RAET1G, IMAGE 3070730 and 2911855; RAET1E, 3464637; and ULBP2, 4747126 (GenBank accession no. BF513861, AW 510737, BE545401, and BG675590).

DNA sequencing was performed using BigDye and ABI 377 sequencer analysis (Applied Biosystems, Warrington, U.K.) using Sequence Navigator software. Full-length receptor constructs were cloned as FLAG epitope fusions in vector p3XFLAG-CMV-9 (Sigma-Aldrich, Peck, U.K.). Topo cloning of PCR fragments was performed using the Topo cloning kit (Invitrogen, Paisley, U.K.) and the manufacturer’s instructions.

RT-PCR

The following PCR primers were used for determining tissue distribution: 1G forward, 5′-AGCCCCGCGTTCCTTCTA; reverse, 5′-TGTATACAAGGCAAGAGGGGC; 1E forward, 5′-TATCCCTGACTTCTAGCCCT; reverse, 5′-GCCACTCACCATTTTGCCAC; and GAPDH forward, 5′-ACCACAGTCCATGCCATCAC; reverse, 5′-TCCACCACCCTGTTGCTGTA.

Cell line RNAs were made as previously described (32). Tissue cDNA was a gift from Dr. D. Simmons (Celltech Group, Berkshire, U.K.). The expected sizes of the products were 935 bp for RAET1G, 835 bp for RAET1G2, and 810 bp for RAET1E. GAPDH was used as a control reaction for each cDNA.

Transfections and flow cytometry

Transient transfections were performed into CV-1 cells using Lipofectamine 2000 (Invitrogen) and the manufacturer’s standard protocol. Flow cytometry was performed on a FACSCalibur machine (BD Biosciences, Oxford, U.K.). Detection of the full-length receptors was performed via a FITC-conjugated mAb to the FLAG epitope (Sigma-Aldrich). Detection of NKG2D binding was via the human Fc fusion, using an anti-human IgG FITC-conjugated polyclonal Ab (DakoCytomation, Ely, U.K.). Detection of UL16 binding was via anti-6-histine mAb (Qiagen, Crowley, U.K.) and R-PE-Cy5 anti-mouse secondary Ab (DakoCytomation).

NK cell cytotoxicity assay

Human NK lymphocytes were isolated from peripheral blood using standard Ficoll isolation of mononuclear cells, followed by staining with anti-CD3-FITC and anti-CD56-CyChrome (BD Biosciences) mAbs. Stained cells were analyzed on a MoFlo cell sorter (DakoCytomation) and the CD3⁻ CD56⁺ lymphocyte population was isolated. These cells were incubated in RPMI 1640 medium containing penicillin and streptomycin, 10% human AB serum, and 100 U/ml rIL-2 for 3 days at 37°C in 5% CO₂. Flow cytometric analysis demonstrated that this polyclonal population of NK cells were uniformly CD3⁻CD56⁺NKG2D⁺ (results not shown).

The ability of novel human RAET1 molecules to induce NK cell-mediated lysis was assessed by an in vitro nonradioactive cytotoxicity assay (33). Four wells of labeled target cells alone were set up to determine spontaneous release, and each labeled target cell was assessed in duplicate at a range of E:T cell ratios. For mAb blocking, NK cell effectors were incubated at room temperature for 30 min in the presence of an NKG2D-specific mAb (a gift from Dr. T. Spies, Fred Hutchinson Cancer Research Center, Seattle, WA) before addition of labeled target cells. The percent specific lysis for each E:T cell ratio was then calculated as: % specific lysis = 100 − % survival.

Production of recombinant proteins

Soluble recombinant extracellular domain of UL16 was produced as a six-histidine, N-terminal fusion. This was produced as an inclusion body in E. coli BLR(DE3) using vector pT7His derived from pGMT7 (34) and the insoluble protein extracted with 6 M guanidine hydrochloride. Purification and refolding in situ was performed using Ni-NTA resin (Qiagen) by sequential dilution to PBS before elution using PBS plus 250 mM imidazole. The NKG2D-Fc fusion protein was produced from 293T cells using calcium phosphate transfection. The fusion was C terminal to human IgG1 hinge-CH2-CH3 domains in pCDNA3.0. The recombinant protein was purified using protein A-Sepharose (Amersham Pharmacia Biotech, Little Chalfont, U.K.), and eluted protein was transferred to PBS by passing through a coarse gel filtration matrix (PD10 column; Amersham Pharmacia Biotech). SDS-PAGE and Western blot analysis verified the integrity of the expressed Fc fusion. The His-tagged protein molecular weight and purity were confirmed by SDS-PAGE on 12% acrylamide gels based on the protocol of Laemmli. Western blotting was conducted as wet blotting transfer to Immobilon-P membrane (Millipore, Watford, U.K.). Detection was using anti-human Fc HRP-conjugated Ab (DakoCytomation).

GenBank accessions

The sequences of RAET1G, RAET1G2, and RAET1E have been filed with GenBank as accession numbers AY172579, AY172580, and AY176317, respectively.

Two RAET1 proteins with TM regions

Initial analysis of the ULBP/RAET cluster called for six expressed genes encoding GPI-linked molecules (Fig. 1⇓) (29). More detailed analysis of the sequences identified potential TM regions in RAET1E and RAET1G. An analysis of the genes encoding RAET1E, RAET1G, and ULBP2 (RAET1E, RAET1G, and RAET1H, respectively) revealed a conserved exon structure, where exon 1 encoded the signal peptide and the start of the protein, exons 2 and 3 encompassed the α1 and α2 domains, and exon 4 encoded a hydrophobic sequence. In the ULBPs, this exon encoded the GPI anchor region and 3′ UTR, but in both RAET1G and RAET1E, the sequence was compatible with a TM as well as a short CYT. Exon 5 in RAET1G encoded the remainder of the putative cytoplasmic domain. The equivalent exon in RAET1H was silent. Although the RAET1L locus also potentially encoded an intact protein, no RAET1L EST sequences have been identified, and it is not considered further in this study.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

Upper panel, Arrangement of expressed genes in the RAET1 cluster on chromosome 6q24.2-q25.3. All genes have four coding exons, except RAET1G, which has five. Exon 5 of ULBP2 is not coding. RAET1 pseudogenes are not shown in this study (29 ). Lower panel, Phylogenetic tree of murine and human NKG2D ligands. The alignment was based on the global amino acid sequences using ClustalW (http://ebi.ac.uk/clustalw/). The tree was computed from the matrix of distances values using the UPGMA (unweighted pair group method with arithmetic mean) method of the program MEGA2 version 2.1 (31 ). Bootstrap values were used to determine the consistency of the tree and are based on 1000 sample runs.

To clarify the sequences of the expressed gene products, we fully sequenced clones corresponding to RAET1E and RAET1G (29). We confirmed that RAET1G was very similar to ULBP2 over the first four exons. A comparison of the novel protein sequences with those of existing murine and human NKG2D ligands showed that RAET1G was most closely related to ULBP2 (85% overall similarity). The highest level of amino acid identity was in the α1/α2 domains. The remaining translated sequence encoded a TM and a 100-aa CYT (Fig. 2⇓). Similarly, the RAET1E sequence encoded two α domains, then a hydrophobic TM, followed by a cytoplasmic domain of 20 aa. RAET1E was the most divergent member of the cluster, sharing <43% identity with the other ligands, whereas ULBP1-3 shared ∼55–60% identity with each other. The conserved amino acid sequences were aligned to the key structural elements of the α1 and α2 domains by performing a ClustalW alignment using the European Bioinformatics server (EBI, Hinxton, U.K.) and subsequent comparison to the known crystal structure of ULBP3. The key structural features of ULBP3 are shown in the protein sequence alignment (Fig. 2⇓) where the NKG2D binding residues of ULBP3 are highlighted. Like RAET1G, the murine NKG2D ligand MULT1 also has a long CYT. No significant sequence similarity was found between the cytoplasmic regions of the two proteins. The cytoplasmic domain of RAET1G did not show homology to any protein or domain when searched by BLAST or through Prosite. We searched for known signaling motifs in the CYT regions of RAET1E and RAET1G. No classical ITIM or ITAM were identified. There was a proline-rich PxPxxP region in the cytoplasmic domain of RAET1G, which corresponded to a consensus Src homology 3 kinase binding motif (35). Downstream of this were two pairs of hydrophobic residues similar to those attributed to basolateral targeting of MICA (36).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

Upper panel, Sequence alignment of RAET1E and RAET1G with ULBP2 and ULBP3. AA sequences derived from cDNA clones were aligned to compare main features. Both splice variants of RAET1G are shown. Symbols indicate proposed α-helical (black cylinders), 3₁₀ helical (gray cylinder), and β strand (gray arrow). These features were derived from the crystal structure of ULBP3. The amino acids underlined in ULBP3 are those in contact with NKG2D. Lower panel, Exon structures of RAET1G and RAET1G2. Alternative splicing in RAET1G leads to a soluble protein product. The cDNA clones are consistent with contiguous sequences for the first three exons. The splice acceptor sites in the fourth exon differ by 100 bp, resulting in a truncated RAET1G2 protein.

Fig. 1⇑ shows a phylogenetic tree of murine and human ligands. Human and mouse NKG2D are ∼60% identical; however, their ligands are substantially different, showing 25–35% identity. Therefore, despite showing some similar features, such as GPI anchors or TM regions, substantial duplication and variation have occurred after speciation between mice and humans.

Alternative splicing of RAET1G

The sequence of IMAGE clone 2911855 was colinear with that of RAET1G except for a 100-bp deletion at the start of exon 4. This arrangement is compatible with alternative splicing at this boundary, with a second potential splice start shifted 3′ by 100 bp. Translation of this deleted form of RAET1G showed that the alternative splicing caused a frameshift and premature termination of the protein sequence. This truncated protein is predicted to be soluble, as the frameshift causes termination before the TM region. This splice form is termed RAET1G2, and its alternative sequence ending is shown below that of RAET1G in Fig. 2⇑. Exon structures for RAET1G and -1G2 are shown also in Fig. 2⇑. Truncated NKG2D ligands such as this may have a role in blocking receptor interactions, and their expression could be part of an immune evasion strategy used by tumors or infected cells (2, 6, 25).

Expression patterns of RAET1G/1G2 and RAET1E

Specific PCR primers were designed to establish the expression profiles of RAET1G/1G2 and RAET1E. Several tumor cell lines contained mRNA for RAET1E or RAET1G (Fig. 3⇓), and the genes were expressed independently of each other, in cells of different lineages. This is in contrast to MICA and MICB where expression appears to be restricted to cells of epithelial origin, and it is unclear whether they are expressed independently of each other (37, 38). The T cell leukemia-derived line HSB-2 expressed a truncated RAET1G transcript. This cDNA product was cloned using Topo cloning and, when sequenced, was identical with the splice form RAET1G2 in IMAGE clone 2911855. The expression of a splice variant encoding a soluble protein is potentially important given the proposed role of soluble NKG2D ligands in impairment of NK and T cell recognition of tumors (39). A limited range of normal human tissues tested showed no expression of RAET1E or the splice form RAET1G2. RAET1G was strongly expressed in colon, but not in the other tissues screened (Fig. 3⇓). An EST matching RAET1G has also been identified from a larynx cDNA library (data not shown).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Expression of RAET1E and RAET1G in cell lines and tissues. A, A panel of established cell lines was screened for expression by RT-PCR. Specific primers were used, such that RAET1G2 gave a product of 835 bp, and RAET1G gave a product of 935 bp, respectively. The RAET1E product is 810 bp. Upper panel, RAET1G; middle panel, RAET1E. GAPDH in the lower panel was used as a positive loading control. The cell lines used were: lane 1, Daudi (B cell); lane 2, 721.221 (B cell); lane 3, HSB-2 (T cell); lane 4, Jurkat (T cell); lane 5, THP-1 (monocyte); lane 6, HeLa (cervical carcinoma); lane 7, HT1080 (fibrosarcoma); lane 8, MelJuSo (melanoma); lane 9, K562 (erythroid); lane 10, NKL (NK cell like); lane 11, Raji (B cell); lane 12, RPMI-8402 (T cell); lane 13, HPB-ALL (T cell); lane 14, MOLT-4 (T cell); lane 15, MOLT-13 (T cell); lane 16, CCRF-CEM (T cell); lane 17, Fern (B cell); lane 18, Ind (B cell); lane 19, U937 (premonocyte); lane 20, YT (NK cell like); lane 21, HUVEC (umbilical vein); and lane 22, HepG2 (liver carcinoma). Track 23 (right) contains RAET1G cDNA control in the upper panel and RAET1E control in the second panel. B confirms the expression of the RAET1G2 form in the T cell leukemia line HSB-2. PCR products from left to right: IMAGE 3070730 (RAET1G), IMAGE 2911855 (RAET1G2), K562, and HSB-2 cell lines The sequences of the products matched those shown in Fig. 2⇑. C, Expression of RAET1G in tissues. Five normal tissues were screened for the presence of both novel genes. Upper panel, RAET1G expression. GAPDH (lower panel) was used as a positive loading control. No RAET1E was found in the panel tested. Tissue cDNA tested bone marrow (lane 1), colon (lane 2), PBL (lane 3), placenta (lane 4), and tonsil (lane 5). Lane 6, Positive control. RAET1G was detected in the colon.

RAET1E and RAET1G expressed on the cell surface bind NKG2D and UL16

We next asked whether RAET1E and RAET1G were capable of binding to their proposed ligand, NKG2D. Full-length cDNAs for RAET1E, RAET1G, and ULBP2 were cloned as N-terminal, FLAG-tagged fusion proteins. These reached the cell surface in transient transfections of CV-1 cells and detection with anti-FLAG Abs in flow cytometry (Fig. 4⇓). NKG2D, expressed as a recombinant soluble Fc fusion protein, bound to CV-1 cells transiently transfected with ULBP2, RAET1G, and RAET1E by flow cytometry (Fig. 4⇓). Similarly, recombinant, soluble, His-tagged UL16 bound FLAG-tagged RAET1G expressed on the surface of transfected cells, but not to untransfected cells, by flow cytometry (Fig. 4⇓). No other transfected cell lines were analyzed for UL16 binding.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

Surface expression of RAET1 proteins, and binding to NKG2D and UL16. A, Cell surface expression in CV-1 cells. The histogram denotes transfected cells stained with anti-flag Ab: cells only (solid black line); ULBP2 (solid gray line); RAET1G (dashed gray line); and RAET1E (dashed black line). The percentages of cells in the M1 populations marked are as follows: untransfected cells, 15%; ULBP2, 85%; RAET1G, 53%; and RAET1E, 71%. B, Histogram denoting transfected cells stained with NKG2D: cells only (solid black line); ULBP2 (solid gray line); RAET1G (dashed gray line); and RAET1E (dashed black line). The percentages of cells in the M1 populations marked are as follows: untransfected cells, 12%; ULBP2, 81%; RAET1G, 50%; and RAET1E, 78%. C, Histogram denoting transfected cells stained with UL16 cells only (solid black line) and RAET1G (dashed gray line).

RAET1E and RAET1G are capable of inducing NK cell cytotoxicity via NKG2D

We next wanted to determine whether RAET1E and RAET1G were biologically functional in triggering cytotoxicity. To show this, stable transfectants of the two molecules in CV-1 cells introduced to cytotoxicity assays using NK effectors (Fig. 5⇓). CV-1 cells expressing either molecule were targets for NK killing under conditions in which untransfected CV-1 cells did not trigger a response. NK cell killing of the transfected cells was 100% blocked by the addition of saturating amounts of an anti-NKG2D antibody, showing that the killing was specifically due to the NKG2D-ligand interaction. Relative killing data for the two ligands and untransfected cells are shown in Fig. 5⇓. These data demonstrate that both TM-linked RAET molecules are potentially efficient targets for NK cells after engagement by NKG2D.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Novel RAET1 molecules confer susceptibility to NK cell-mediated lysis. CV-1 cells transfected with RAET1E (A) RAET1G (B), or vector only (C) were used as targets in a cytotoxicity assay with CD3⁻CD56⁺NKG2D⁺ human NK lymphocytes. The levels of specific lysis observed at a range of E:T cell ratios (▪) were consistent with the derived binding affinity data and could be completely reversed by incubating effector NK cells with a NKG2D-specific mAb (○). The SD of mean specific lysis observed in replicate wells was <10%.

The great diversity of RAETs: continuous evolution for disease resistance?

The diverse features of NKG2D ligands are consistent with functions in alerting the immune system to a wide variety of challenges, whether infection, stress, or malignant transformation. The marked differences in their sequences in different species indicate that some duplication of different sets of genes, namely ULBP/RAET1 and MIC in humans and Rae1, H60, and MULT1 in mice, have occurred, making it difficult to identify individual murine and human ligand orthologues. Although a number of features are shared, including a long CYT in both RAET1G and MULT1, it is likely that some gene duplication occurred after speciation of mouse and man. This evolutionary diversity, association to the membrane via a TM region rather than a GPI linkage, and varied expression patterns are all suggestive of selection, presumably in response to resistance to different pathogens. Consistent with this premise, human and mouse CMV have developed proteins that can prevent some, but possibly not all, ligands for NKG2D from reaching the cell surface (25, 41).

Role for soluble versions of NKG2D ligands

The presence of soluble MIC in the sera of patients with MIC⁺ tumors has been linked to a reduction in surface NKG2D on lymphocytes and may be a route for immune evasion by impairing the responsiveness of NKG2D-bearing NK and T cells (29). MICA is proposed to be lost from the cell surface of tumors through cleavage by metalloproteases (42), and this is likely to be the case for the TM-containing ligands, RAET1G and RAET1E. The soluble splice form of RAET1G detected in the T cell leukemia line HSB-2 could play a similar role. Genetic variation in the ULBP/RAET cluster has not been fully explored, but our preliminary data suggest that the genes are polymorphic, and this variation could have a profound effect on the interaction with NKG2D and viral products, as is proposed for variants of MICA/B (43).

Expression patterns suggest different roles in different tissues

The expression pattern of the ULBP/RAET genes presented in this study and in previous studies (9, 28) shows that multiple ligands for NKG2D can be expressed on one target cell. The ligands are also clearly capable of independent expression. The data are consistent with different NKG2D ligands expressed on different tissues. MIC products are generally expressed on epithelial cells. ULBP/RAET1 can be expressed on epithelial cells, but are also expressed in cell lines of nonepithelial origin, providing a rationale for roles distinct from MICA/B; for example, in immune responses to lymphoid malignancies and to viruses that infect lymphocytes. It may be simplistic to consider the ligands for NKG2D merely as flags of cellular stress or transformation, because a number appear to be expressed in normal tissues. MICA is known to be constitutively expressed in the intestinal epithelium and can be recognized by intestinal intraepithelial lymphocytes, including γδ T cells (38). It is proposed that MICA is up-regulated on mucosal epithelial surfaces upon recognition of bacterial products (24), where it contributes to T cell alertness (44). RAET1G, a new ligand for NKG2D, is also strongly expressed in the colon, and it could have a parallel role to MICA in the gut. The identification of novel ligands for NKG2D with TM domains raises questions about the functional implications of different ligand topologies.

In the mouse, GPI-anchored proteins have modest to low affinities for NKG2D, whereas the ligands possessing CYT domains have high affinity, and the human NKG2D ligands may well conform to this pattern. In a situation where one cell bears multiple ligands with differing affinities for NKG2D, there would be competition for the binding of the receptor, and the ligand with more favorable kinetics would gain occupancy (26). However, this model is complicated by potential differences in cellular distribution between GPI-anchored and TM proteins. GPI-anchored proteins tend to be found in localized cholesterol-rich regions known as lipid rafts. Localized concentrations of specific surface proteins and signaling apparatus perform distinct roles in what has been referred to as the immune synapse (45). The ULBPs may be found preferentially in lipid rafts on cellular activation and synapse formation and would be found at high density in contact regions between cells. Their affinities may reflect the clustered nature of proteins within lipid rafts, where low to moderate affinity for NKG2D would be augmented by an increased avidity. If ligands such as RAET1G, MICA, and MULT1 have an increased affinity over GPI-anchored molecules, they would have a greater ability to bind NKG2D in the absence of the increased avidity afforded by such clustering.

Within polarized epithelial layers there is an additional difference in the subcellular localization of GPI-anchored and TM proteins. GPI-anchored proteins have a greater tendency to be found at the apical surface, whereas some TM proteins, if they posses one of several specific motifs, will be targeted to the basal or lateral sides of the cell. Basolateral targeting studies on MICA are consistent with the idea that TM ligands behave differently from those only on one side of the membrane (36). Normal wild-type MICA is targeted to the basolateral membrane of gut epithelium, in proximity to NK or T cells, where it can signal cell destruction. However, a naturally occurring allele that has a TM motif, but no cytoplasmic tail, does not target to the same part of the cell, but moves to the apical surface, where there is a greater concentration of lipid rafts and less contact with NKG2D-bearing cells. This targeting is shown to be effected by two pairs of leucine/valine residues. RAET1G could be similarly targeted to the basolateral surface, because it has a similar dihydrophobic motif. In effect, it could be a good front-line indicator of bacterial challenge. In a polarized cell layer, such as epithelial surfaces in the gut, the differences in anchorage of NKG2D ligands allow differential distribution in the same cell, different possible signaling pathways and hence, differential availability to lymphocytes. The distribution of ligands on a cell could change on bacterial challenge, transformation, or lymphocyte engagement. The relative distribution of NKG2D ligands in distinct tissues and cellular domains may be fundamental to understanding NKG2D-mediated immune recognition. The unique CYT tail carried by RAET1G provides the potential to transmit signals within the cell to modulate other molecules involved in responses to pathogens. We are currently investigating the signaling capability of the RAET1G cytoplasmic tail. This is the first NKG2D ligand described with evidence of its own signaling capability.

In conclusion, we have described two novel ligands of human NKG2D. Both ligands possess TM and CYT domains, making them functionally divergent from the ULBPs. The novel sequences contained within the CYT domain of RAET1G have implications for our understanding of the specificity of activation by NKG2D. It will be important to determine the precise conditions for induction of expression of the entire ULBP/RAET family of molecules and their distribution in normal and challenged tissues.