Abstract
The functions of activating members of the killer cell Ig-like receptor (KIR) family are not fully understood, as the ligands for these receptors are largely unidentified. In this study, we report that KIR2DS2 reporter cells recognize a ligand expressed by cancer cell lines. All cancer targets recognized by KIR2DS2 were also recognized by KIR2DL2 and KIR2DL3 reporters. Trogocytosis of membrane proteins from the cancer targets was observed with responding reporter cells, indicating the formation of KIR2DS2 ligand–specific immunological synapses. HLA-C typing of target cells showed that KIR2DS2 recognition was independent of the HLA C1 or C2 group, whereas targets cells that were only recognized by KIR2DL3 expressed C1 group alleles. Anti–HLA class I Abs blocked KIR2DL3 responses toward C1-expressing targets, but they did not block KIR2DS2 recognition of cancer cells. Small interfering RNA knockdown of β2-microglobulin reduced the expression of class I H chain on the cancer targets by >97%, but it did not reduce the KIR2DS2 reporter responses, indicating a β2-microglobulin–independent ligand for KIR2DS2. Importantly, KIR2DL3 responses toward some KIR2DS2 ligand–expressing cells were also undiminished after β2-microglobulin knockdown, and they were not blocked by anti–HLA class I Abs, suggesting that KIR2DL3, in addition to the traditional HLA-C ligands, can bind to the same β2-microglobulin–independent ligand as KIR2DS2. These observations indicate the existence of a novel, presently uncharacterized ligand for the activating NK cell receptor KIR2DS2. Molecular identification of this ligand may lead to improved KIR-HLA mismatching in hematopoietic stem cell transplantation therapy for leukemia and new, more specific NK cell–based cancer therapies.
Introduction
Natural killer cells have the capacity to recognize and kill tumor cells and infected cells (1–4). They are also important modulators of adaptive immune responses by release of cytokines (1, 2). Moreover, NK cells can recognize and kill MHC-disparate bone marrow cells, and thus play an important role in nonautologous hematopoietic stem cell transplantation (5–7). The propensity to kill targets with a different MHC background allows an NK cell–mediated graft-versus-leukemia effect in allogeneic transplantation therapies for leukemia (8, 9). Recognition of target cells relies on an array of NK cell receptors, where the response toward a target cell is determined by the balance between activating and inhibitory signals. A striking feature of many NK receptor families is that they contain both activating and inhibitory members, some of which are found as pairs with highly similar ligand-binding domains but with opposite signaling functions (10). Examples include CD94/NKG2A versus CD94/NKG2C. The same is true for the rodent receptors KLRE/I1 versus KLRE/I2 (11), and among members of the NKR-P1 and Ly49 receptor families (10). Although inhibitory members of the Ly49 (12), CD94/NKG2 (13), and killer cell Ig-like receptor (KIR) (14) families were rapidly found to bind to MHC class I, ligands for their activating siblings mostly remain unidentified and only a few have shown binding to MHC (10).
The human KIR family consists of 14 members, encoded by 12 separate loci (15–17). KIR2DS1, -2, -3, -4, and -5 associate with the activating adaptor protein DAP12 and activate NK cytotoxicity and cytokine release (18), whereas the inhibitory KIR with two Ig domains (KIR2DL1, -2, and -3) bind to normal allelic variants of HLA-C. KIR2DL1 binds HLA-C alleles belonging to the C2 group, originally defined by asparagine and lysine residues in positions 77 and 80, respectively. KIR2DL2 and KIR2DL3, conversely, bind alleles of the C1 group, carrying the 77S, 80N motif but may also bind C2 group alleles with varying affinities (15, 19, 20). In contrast to the 2DL members, the activating 2DS KIRs do not seem to bind HLA-C with high affinity. Similar to its inhibitory sibling KIR2DL1, KIR2DS1 can bind to C2 group alleles, albeit possibly with a lower affinity (21–24). Despite some observations that suggest interaction with C1 group alleles (25), KIR2DS2 has not been directly demonstrated to bind HLA-C. The inhibitory KIR2DL3 ectodomain differs from the activating KIR2DS2 by only 3 aa residues, and swapping the tyrosine in position 45 of KIR2DS2 with phenylalanine (as in 2DL2 and 2DL3) enabled binding to HLA-C (26). A recent report has demonstrated cocrystallization of KIR2DS2 with HLA-A11:01 (27). The functional consequence of this interaction remains to be established.
Besides the multigene receptor families, NK cells express several activating receptors, including NKG2D, NKp30, NKp44, and NKp46 (10). To the extent that the ligands for these receptors have been determined, they are encoded in the cellular genome and are either induced by transcriptional regulation or posttranscriptional or posttranslational modifications in stressed cells and/or cancer cells (28–30).
Among rapidly advancing strategies for treating cancer, Ag-specific mAbs and chimeric Ag receptor–expressing T cells have proven to be especially effective for the treatment of hematological malignancies. These therapies are tailored toward cancer-specific Ags that are not expressed by normal cells, and there is an urgent demand for the identification of new cancer-specific Ags. In this study, we report that a β2-microglobulin–independent ligand for the activating KIR2DS2 receptor is widely expressed by cancer cell lines. To our knowledge, this is the first description of a cancer cell ligand for KIR2DS2. We discuss the implication of these findings for tumor recognition by NK cells.
Materials and Methods
Cell lines
The cancer cell lines PC-3 (31), DU145 (32), WM9 (33), T47D (34), OVCAR-3 (35), SK-BR-3 (36), and WM35 (33) were provided by Z. Suo and E. Emilsen (Oslo University Hospital). All cancer cell lines were authenticated at the Oslo University Hospital cell typing core facility. The HLA-ABC–deficient cell line 721.221 as well as 721.221 cells stably transfected with HLA-C*03:04, -C*04:01, and -C*15:03 (37) were provided by K. Kärre (Karolinska Institute). Chinese hamster ovary (CHO)-K1 cells were obtained from the American Type Culture Collection, and the BWN3G cell line (BW5147 mouse thymoma cells stably transfected with enhanced GFP [EGFP] under control of a 3× NFAT response element promoter) has been described previously (38). All cell lines were routinely screened for mycoplasma infection and cultured in RPMI 1640 supplemented with 1 mM sodium pyruvate, 1% antibiotic/antimycotic solution, and 10% FBS (Invitrogen) (hereafter termed complete medium) at 37°C in humidified air containing 5% CO2.
Antibodies
The following purified mAbs were used: DX17 (anti–HLA-A, -B and -C) (39); EGFR.1 (anti– epithelial growth factor [EGF] receptor); TÜ99 (anti–β2-microglobulin); 6D4 (anti–MHC class I polypeptide-related sequence A and B); anti–HLA-G (all obtained from BD Biosciences); B9.12.1 (anti–HLA-A, -B and -C; Beckman Coulter) (40); 67A4 (anti–E-cadherin; Beckman Coulter); M2 (anti-FLAG; Sigma-Aldrich); DT9 (anti–HLA-C and -E; Millipore) (41); 16B12 (IgG1 control; Covance); W6/32 (anti–HLA class I) (42); HC10 (anti–HLA-B and -C) (43); HCA2 (anti–HLA-A and -B) (43); GL183 (anti-KIR2DL2/KIR2DL3/KIR2DS2) (44); EB6 (anti-KIR2DL1/KIR2DS1) (44); FES172 (anti-KIR2DS4; Beckman Coulter); WEN30 (IgG2b control); OX34 (IgG2a control). mAbs A6-136 (anti–HLA class I) and 6A4 (anti–HLA class I) were used as undiluted supernatants (45). A polyclonal Alexa Fluor 647–conjugated goat anti-mouse Ig secondary Ab (Life Technologies) was used in flow cytometry.
Generation of reporter cell lines
cDNA clones of human KIR2DS1, KIR2DS2, KIR2DS4, KIR2DL1, KIR2DL2, and KIR2DL3 were obtained by PCR or by gene synthesis (Eurofins). Expression constructs encoding chimeric fusion proteins consisting of an N-terminal FLAG epitope tag (DYKDDDDK) followed by the extracellular KIR regions, the transmembrane region of human CD8, and the cytoplasmic region of mouse CD3ζ were generated in the pBSRα-EN vector (46) and verified by sequencing. To obtain stable transfectants, 3 × 106 BWN3G cells (38) were resuspended in 400 μl of complete medium containing 20 μg of linearized plasmid and electroporated in 2-mm cuvettes (120 V, 960 μF, GenePulser II; Bio-Rad Laboratories). After 24 h, transfected cells were seeded in 96-well plates at between 10,000 and 1000 cells per well in selection medium (complete medium supplemented with 1.6 mg/ml geneticin [G-418 disulfate; Duchefa Biochemie] and 1 mg/ml hygromycin [Invitrogen]). Clones with bright surface expression were identified by flow cytometry (anti-KIR2DL2/-2DL3/-2DS2 mAb GL183, anti-FLAG mAb M2; Sigma-Aldrich). The clones were further tested for EGFP expression after receptor crosslinking: 96-well culture plates were first coated with 10 μg/ml secondary Ab (goat polyclonal anti-mouse IgG; Jackson ImmunoResearch) in coating buffer (50 mM sodium carbonate buffer [pH 9.3]) at 4°C overnight, blocked for 30 min at room temperature with PBS containing 10 mg/ml BSA (Sigma-Aldrich), and then coated with anti-FLAG mAb (M2; Sigma-Aldrich) at 4°C overnight and washed. Then, 5 × 104 KIR reporter cells were added per well, incubated overnight at 37°C, and analyzed for EGFP expression by flow cytometry.
Reporter assay
Target cells (5 × 104) were mixed with 5 × 104 KIR reporter cells in 200 μl of complete medium at 37°C overnight (18–22 h) in 96-well plates. Following incubation, all samples were stained with anti–HLA class I mAb W6/32–Alexa Fluor 647 (4 μg/ml in PBS with 2% FCS and 10 mM NaN3) at 4°C. To discriminate target cells (human) from reporter cells (mouse) in flow cytometry, a forward scatter/side scatter gate was set to include live reporter cells and a sequential EGFP/Alexa Fluor 647 gate was set to exclude the EGFP−, W6/32bright cancer target cells (see Fig. 2D). Trogocytosis was investigated by flow cytometry after overnight coincubation of the KIR reporter cells and the target cells. The cells were stained with W6/32, 67A4, or EGFR.1 (2 μg/ml in PBS with 2% FCS and 10 mM NaN3) followed by Alexa Fluor 647–conjugated secondary Ab. For blocking experiments, anti–HLA class I mAbs W6/32, B9.12.1, DX17, HC10, HCA2, A6-136, 6A4, or appropriate control Ig was added to the target cells and incubated for 15–30 min before the KIR reporter cells were added. The experiments were incubated overnight. A6-136 and 6A4 were used as undiluted hybridoma supernatants, whereas all other Abs were used at 20 μg/ml.
KIR and HLA genotyping
DNA was extracted from cancer cell lines by a standard protocol as previously detailed (47). KIR genotyping was performed by PCR with a sequence-specific primer (SSP) typing kit (Miltenyi Biotec) followed by agarose gel electrophoresis and visualization by ethidium bromide staining. HLA genotyping was performed by PCR with SSP typing kits (HLA-A low resolution, HLA-B low resolution, and HLA-C low or high resolution; Olerup) followed by agarose gel electrophoresis and visualization by ethidium bromide staining. Data analysis and allele scoring were performed with the SCORE program (Olerup).
Small interfering RNA–mediated knockdown
β2-Microglobulin expression was targeted using a mix of four small interfering RNAs (siRNAs) complementary to human β2-microglobulin and compared with an irrelevant control siRNA mix (ON-TARGETplus SMARTpool; GE Dharmacon). Adherent target cells were seeded at 4 × 104 cells per well in 24-well plates in 500 μl of complete medium 24 h before transfection. At transfection, 7.2 pmol of siRNA was resuspended in 60 μl of Opti-MEM (Life Technologies) and added to 60 μl of Opti-MEM containing 1.2 μl of Lipofectamine RNAiMAX (Invitrogen), left to incubate at room temperature for 20 min, mixed with 500 μl of complete medium, and added to one well of target cells. Sixty-six hours following transfection start, transfection medium was aspirated, the wells were washed once in Opti-MEM, and 600 μl of complete medium was added. At 72 h, transfected cells were released in PBS with 2 mM EDTA, centrifuged, resuspended in complete medium, and plated in 96-well plates. Transfected target cells (5 × 104) were coincubated with 5 × 104 KIR reporter cells in 200 μl of complete medium overnight (20–24 h). Parallel wells did not receive reporter cells but were assayed by flow cytometry for surface expression of β2-microglobulin and HLA class I at 72 h and 92–96 h after transfection.
HLA-C allele cDNA cloning and stable transfection of CHO-K1 cells
RNA was isolated from cancer cell lines using TRIzol (Ambion) according to the manufacturer’s instructions. First-strand cDNA synthesis was as previously detailed, and full open reading frames of expressed HLA-C alleles were amplified by PCR with PfuTurbo polymerase and HLA-C–specific primers (forward, 5′-GCCGAGATGCGGGTCATGG-3′, reverse, 5′-CAGGCAGCTGTCTCAGGCTTTACAA-3′), cloned (pCR2.1-TOPO vector, TOPO-TA cloning; Invitrogen), and sequenced. Expression constructs encoding an HLA-C allele (under CMV promoter control) together with human β2-microglobulin (EF-1α promoter) were generated in the pBudCE4.1 (Invitrogen) and verified by sequencing. Stable transfectants were generated by cationic lipid transfection with polyethyleneimine: 4.2 μg of plasmid DNA in 250 μl of PBS was mixed with 1.7 μl of polyethyleneimine (6.45 mg/ml) in 250 μl of purified H2O and incubated at room temperature for 30 min, then slowly added to a T25 flask that 24 h previously was seeded with 2 × 106 CHO-K1 cells in 5 ml of complete medium. After 24 h of transfection, the cells were plated in 96-well plates at 1000 cells per well in complete medium with 400 μg/ml Zeocin (Invitrogen). Stable clones with bright surface expression of HLA-C were selected by flow cytometry. The full open reading frame of HLA-A*11:01:01 was generated by gene synthesis (Eurofins), cloned into pBudCE4.1 vector with β2-microglobulin, sequence verified, and stably expressed in CHO-K1 cells as described for HLA-C above.
Results
The activating receptor KIR2DS2 responds to a ligand expressed by cancer cells
To search for cell surface–expressed ligands for activating members of the human KIR family, we generated EGFP-producing reporter cell lines (38, 48) that stably express chimeric receptors consisting of the KIR2DS1, -2DS2, -2DS4, or -2DL3 ectodomains (Supplemental Fig. 1B, 1C) fused to the cytoplasmic region of CD3ζ. Hypothesizing that activating KIR could play a role in recognition of neoplastic cells, we used the KIR reporter cells to screen a panel of human cancer cell lines. In overnight coincubation experiments followed by flow cytometric detection of EGFP, we observed that the KIR2DS2 reporter cell lines responded strongly to a subset of the cancer cell lines tested. Notably, all cancer cell lines that activated the KIR2DS2 reporters were also recognized by the KIR2DL3 reporters. The prostate cancer cell lines DU145 and PC-3 consistently induced strong responses with both KIR2DS2 and KIR2DL3. Other strong targets for both KIR2DS2 and KIR2DL3 reporters were the breast cancer line SK-BR-3 and the ovarian carcinoma OVCAR-3 (Fig. 1). Some tumor lines (T47D, breast ductal carcinoma and WM9, melanoma) were only recognized by KIR2DL3 and not the KIR2DS2 reporters. Finally, some cancer targets did not activate either the KIR2DL3 or -2DS2 reporters. These included WM35 (Fig. 1) as well as FO-1, Caco-2, Daudi, K562, THP-1, and KYSE70 (data not shown). No EGFP response was observed with KIR2DS1 or KIR2DS4 reporters for any of the cancer cell lines tested (Fig. 1).
KIR2DS2 reporter cells respond to ligand expressed by cancer cells. Induction of EGFP expression in KIR-CD3ζ reporter cells was assayed by flow cytometry following overnight coincubation with the indicated cancer cell targets. The reporter cells expressed chimeric receptors of the CD3ζ cytoplasmic region coupled to the ectodomain from KIR2DS1, -2DS2, -2DS4, or -2DL3 as indicated. HLA-ABC–deficient 721.221 cell lines (untransfected) as well as 721.221 cells transfected with HLA-C*03:04 were included as control targets. HLA C1/C2 genotype of the target cells is indicated. One representative experiment of more than four experiments of each cell line is shown. The percentage of EGFP+ cells in each experiment is indicated in the upper right corner. Before analysis, the cells were stained with an Alexa Fluor 647–conjugated mAb toward human HLA class I in a gating strategy to exclude target cells (shown in Fig. 2D).
KIR2DL2 and -2DL3 are considered alleles of the same locus. KIR2DL2 reporter cells showed the same reactivity as KIR2DL3 reporters toward the KIR2DS2-reactive tumor targets, whereas KIR2DL1 reporter cells did not respond to these targets (Supplemental Fig. 1A). KIR2DL2 and -2DL3 are well documented to bind to C1 group alleles of HLA-C (15). Accordingly, KIR2DL3 reporter cells produced EGFP upon incubation with 721.221 cells stably expressing HLA-C*03:04 (Fig. 1), but they did not respond to HLA-C*04:01 or -C*15:03 (data not shown and Supplemental Fig. 1A). In accordance with previous findings (20), the KIR2DL2 reporter cells, in contrast to -2DL3 reporters, consistently responded to the C2 allele C*15:03 in addition to C*03:04, whereas the KIR2DL1 reporters as expected responded strongly toward C*15:03 but not C*03:04 (Supplemental Fig. 1A). The KIR2DS2 reporter cell lines did not respond to 721.221 target cells transfected with either HLA-C*03:04 (Fig. 1), -C*04:01, or -C*15:03 (data not shown and Supplemental Fig. 1A). The same pattern of recognition of cancer cell targets was observed with five independent KIR2DS2 reporter cell lines (generated at two different time points), four independent KIR2DL2 reporters, and three independent KIR2DL3 reporters. As a positive control, Ab crosslinking induced strong responses in all reporter cells (Supplemental Fig. 1B). To exclude that the extracellular N-terminal FLAG epitope tag on the KIR reporter cells influenced binding, KIR2DS2, -DL2, and -2DL3 reporters without FLAG tag were also generated and assayed with cancer cell targets. They produced responses similar to those of the reporters with tag (data not shown). Flow cytometry analysis showed no expression of mouse NKG2D or the Fc receptors CD16 or CD32 on the mouse T cell parent reporter line BWN3G (data not shown).
KIR2DS2/KIR2DL3 ligand interaction specifically induced intercellular transfer of target cell membrane proteins
As a gating strategy to separate reporter cells (mouse) from cancer cell targets in flow cytometry, we stained the mix of reporter and target cells with an mAb toward human HLA class I after EGFP induction in overnight coincubation reporter assays (Fig. 2D). Somewhat to our surprise, we in some instances observed staining of KIR reporters with the anti–HLA class I Ab W6/32, indicating that target cell HLA class I proteins had been transferred to the reporter cell membrane, presumably by trogocytosis. Similar staining was observed with five different anti–HLA class I Abs tested, including mAbs specific for HLA-ABC (data not shown).
Transfer of surface molecules from ligand-expressing cancer cells to ligand-reactive KIR reporters. Two-color flow cytometry analysis of EGFP expression versus expression of target cell–derived surface molecules on the surface of KIR reporter cells is shown. (A) HLA class I, (B) human EGF receptor, (C) human E-cadherin. The indicated KIR-CD3ζ reporter cells (mouse) were coincubated with the indicated human target cell lines, and trogocytosis of target cell Ags was visualized by subsequent staining with the indicated human-specific Abs. (D) Cell conjugates were excluded by forward scatter/side scatter gating, including only small, viable reporter cells while excluding target cells (EGFP−, HLA class Ibright) by a sequential gate (EGFP versus Alexa Fluor 647).
Importantly, the transfer of HLA class I to KIR2DS2 or KIR2DL3 reporter cells was only observed when reporters were incubated with target cells that expressed ligand, and transfer of HLA class I was not observed with reporter cells expressing KIR2DS1 when incubated with the same cancer cells. Moreover, transfer of target cell HLA class I was more pronounced with EGFPbright individual cells (Fig. 2A). Thus, transfer of target cell HLA class I correlated with individual reporter cell activation, suggesting that the transfer depended on the formation of an immunological synapse between reporter and target cell. To investigate whether the transfer was HLA class I specific we looked at two other membrane proteins expressed on the surface of the target cells, EGF receptor and E-cadherin. Similar to transfer of HLA class I, EGF receptor and E-cadherin were transferred to the reporter cells only in cases where the target cell activated the KIR2DL3 or KIR2DS2 reporters (Fig. 2B, 2C, Supplemental Fig. 2). Thus, the observed transfer of target cell membrane proteins was not specific to HLA molecules, and instead most likely reflected a strong form of cell–cell contact. Trogocytosis has previously been observed following the formation of immunological synapses between human NK cells and target cells (49), and trogocytosis in T cells has been reported to require signaling (50).
These observations suggested that KIR2DS2/KIR2DL3 recognition of cancer cells induced the formation of an immunological synapse, reflecting a physiologically relevant avidity between KIR2DS2/KIR2DL3 and the tumor cell ligand.
The KIR2DS2 response is independent of the target cancer cell HLA-C1/C2 genotype
To investigate a correlation between HLA class I expression on the cancer targets and the KIR2DS2 responses, surface expression of class I was investigated by flow cytometry using seven different mAbs. Cancer cells have frequently downregulated or lost expression of HLA class I (51, 52). Almost all the cancer target cells recognized by both KIR2DS2 and KIR2DL3 showed high levels of class I expression, whereas SK-BR-3 expressed moderate levels. With regard to HLA-C expression, PC-3 and DU145 stained well with the HLA-C– and HLA-E–specific mAb DT9, whereas SK-BR-3 and OVCAR-3 showed only weak staining. The HC10 mAb, reactive with HLA-B and -C, did not stain SK-BR-3 at a significant level (Fig. 3).
Surface expression of HLA class I by cancer cell lines. Flow cytometry analysis of the indicated cell lines using a panel of mAbs toward HLA class I Ags is shown. Cells were stained with anti-HLA or isotype control (filled curves). Data are representative of at least two experiments.
KIR2DL1 binds to C2 group (77N, 80K) alleles of HLA-C, whereas KIR2DL2 and -2DL3 bind C1 group (77S, 80N) alleles (19). Low-affinity binding of KIR2DL2/-3 to C2 alleles has also been reported (20). Our KIR2DL2 and -2DL3 reporter cells responded to the same cancer cell lines as did the KIR2DS2 reporters. Given the high level of amino acid sequence similarity between KIR2DS2, KIR2DL2, and KIR2DL3, this would point to C1 group alleles as candidate ligands for KIR2DS2. To this end, we genotyped the cancer target cell lines at the HLA-C locus by low- and high-resolution SSP typing (Table I). KIR typing of the cancer cell targets was also performed (Table II). Among the KIR2DS2/KIR2DL3–reactive targets reported in this study, two cancer cell lines were only C1 (SK-BR-3, OVCAR-3), one was C1/C2 (PC-3), and one cell line was only C2 (DU145). The HLA-C alleles found in these cells were: HLA-C*01:02:01 (C1), -C*03:04:01 (C1), -C*06:02:01 (C2), -C*07:02:01 (C1), and -C*08:02:01 (C1) (Table I). Thus, our data indicated that the KIR2DS2 reporter reactivity was independent of the C1/C2 genotype (Fig. 1). As for the KIR2DL3 reporters, all the HLA-genotyped targets contained at least one C1 allele, except for DU145. Because KIR2DL2/DL3 can bind certain C2 group alleles, we could thus far not exclude that the 2DL2 and 2DL3 reporters responded to normal HLA-C alleles expressed on the tumor cells.
Although encoded by separate loci, HLA-A and -B alleles share sequence similarity with HLA-C. Moreover, a recent report showed peptide-dependent binding of KIR2DS2 to HLA-A*11:01. We therefore genotyped the cancer target cell lines at the HLA-A as well as the HLA-B loci. None of the cancer cells were A*11 positive. Moreover, no apparent pattern of correlation was observed between the HLA-A sequence of the cancer cells and the KIR2DS2/-2DL2/-2DL3 reactivity (Table I). The KIR2DS2 target DU145 was genotyped to be A*03 and A*33. These alleles were shared with the nonresponsive targets WM9 (A*03/A*32) and T47D (A*33). Assuming surface expression, this would argue against HLA-A as the ligand for KIR2DS2 on DU145. Moreover, the targets DU145, PC-3 (A*01/A*24), and OVCAR-3 (A*02/A*29) did not share a single common HLA-A allele. As for HLA-B, we could not identify any correlation between genotype and reactivity with KIR2DS2/-2DL2/-2DL3.
KIR2DS2 reporter cells did not respond to cancer cell–derived HLA-C
We wanted to investigate whether the KIR2DS2 recognition of the cancer cells could be explained by cancer-related mutations at the HLA-C locus. We therefore cloned and sequenced multiple HLA-C cDNA clones for each cancer cell line. No mutations were identified, and the genotypes obtained by SSP typing were confirmed by cDNA cloning (Table I). To examine whether these allelic variants of HLA-C could interact with the KIR2DS2 reporters, we generated several independent stable transfectants for each HLA-C allele in the hamster cell line CHO-K1, expressing the transfected HLA-C variant at high levels together with human β2-microglobulin (Fig. 4A). In subsequent reporter assays, none of the tumor-derived HLA-C variants were able to activate the KIR2SD2 reporter cells. In conclusion, these data indicate that the observed KIR2DS2 reactivity with cancer targets was not due to interaction with a particular subset of HLA-C alleles. The KIR2DL3 reporters responded to all the tested HLA-C alleles belonging to the C1 group, but importantly they did not respond to the C2 allele (C*06:02:01) of the 2DL3-reactive cancer target DU145 (Fig. 4B). This suggested that, at least with DU145 targets, the tumor cell reactivity of KIR2DL3 was not due to interaction with normal HLA-C ligands, but instead possibly to the same ligand as KIR2DS2.
KIR reporter assay using target cells transfected with tumor-derived HLA-C. CHO cells were stably transfected with the indicated cancer cell derived HLA-C alleles or with HLA-A11. (A) Flow cytometry analysis of HLA surface expression (mAb W6/32). (B) The indicated HLA transfectants were coincubated overnight with the indicated KIR-CD3ζ reporter cell line, followed by flow cytometry analysis. Percentages of EGFP+ cells are shown in the upper right corners. PC-3 cells and untransfected CHO cells were used as positive and negative controls, respectively. One representative experiment out of three is shown.
We also tested the ability of CHO cells stably transfected with HLA-A*11:01:01 plus β2-microglobulin to activate the KIR2DS2 reporter cells. No KIR2DS2 or KIR2DL3 response toward A11 was observed in these assays (Fig. 4B).
Anti–HLA class I Abs did not block cancer cell activation of KIR2DS2 reporters
Although the 2DS2 reporters did not respond to the tumor-derived HLA-C variants directly, the response could possibly require cancer-specific peptides or some form of posttranslational modification associated with the cancer targets investigated. KIR2DS2 peptide specificity could also explain the lack of KIR2DS2 reporter response toward HLA-A*11:01:01 transfectants. This interaction may be highly peptide-specific, requiring peptides not sufficiently available in CHO cells. It has previously been demonstrated that binding of KIR2DL1, KIR2DL2, and KIR2DL3 to HLA-C can be blocked with certain anti–HLA class I Abs, such as the mAbs A6-136, 6A4, and B9.12.1. Including these, we applied a total of seven anti–HLA class I mAbs in blocking experiments. The cancer targets were incubated with the different Abs before and during incubation with the KIR reporter cells. None of the anti–HLA class I Abs tested were able to efficiently block the KIR2DS2 reporter cell response toward cancer targets (Fig. 5A).
Blocking experiments using anti–HLA class I Abs in KIR2DS2 or KIR2DL3 reporter assays against cancer cells. The indicated target cells were incubated with the indicated anti–HLA class I mAbs before and during overnight coincubations with KIR2DS2 or KIR2DL3 reporter cells. EGFP reporter cell responses were measured by flow cytometry. HLA C1/C2 genotype of the target cells is indicated. Column histograms display relative fractions of EGFP+ reporter cells, normalized with respect to control Ig. (A) Blocking of cancer cell recognition by KIR2DS2 reporters. (B) Blocking of target cell recognition by KIR2DL3 reporters. The average of three independent experiments is shown. A statistically significant reduction compared with control Ig was found. *p ≤ 0.05.
Some targets were only recognized by the KIR2DL3 reporters and not by KIR2DS2. These included the cancer targets WM9 (C1) and T47D (C1) as well as 721.221 cells transfected with the C1 group allele -C*03:04:01. The A6-136, 6A4, and B9.12.1 mAbs efficiently blocked KIR2DL3 reporter cell responses toward these targets (Fig. 5B and data not shown). Importantly, in contrast, KIR2DL3 reporter responses toward the cancer targets also recognized by KIR2DS2 could not be efficiently blocked (Fig. 5B). In summary, these results suggested that KIR2DL3 recognized two different ligands: one that could be blocked by anti–HLA class I Abs (normal HLA-C1) and a second ligand, not blocked by anti–HLA class I mAbs, which is also recognized by KIR2DS2.
Recognition of cancer cells ligand by KIR2DS2 and KIR2DL3 does not require β2-microglobulin
Responses to those cancer targets that activated both KIR2DS2 and KIR2DL3 were not blocked by anti–HLA class I Abs. However, the possibility still remained that KIR2DS2 binds to HLA-C in an altered conformation that does not allow binding of the Abs used, either as a result of posttranslational modifications, impaired folding, or a particular nature of the peptides loaded in the groove. In normal conformation, MHC class I molecules are expressed on the cell surface together with β2-microglobulin. To investigate whether KIR2DS2 reporter cells responded to HLA-A,-B, or -C on the cancer cells in a β2-microglobulin–dependent manner, we developed an efficient siRNA protocol that in some cell lines routinely achieved >97% knockdown of surface-expressed β2-microglobulin and similar levels of knockdown of HLA class I, measured by flow cytometry using several different Abs (Fig. 6A). KIR2DS2 and KIR2DL3 reporter responses toward the C2 tumor target DU145 were not reduced despite a two orders of magnitude reduction in surface expression of β2-microglobulin as well as HLA-A, -B, and -C. This indicates that β2-microglobulin is not required for KIR2DS2/-2DL3 reactivity with this target (Fig. 6B). β2-Microglobulin knockdown in the C1 target OVCAR-3 reduced the KIR2DL3 response partially, whereas the KIR2DS2 response showed no significant reduction (Fig. 6B). This result is in accordance with our findings that OVCAR-3 expressed the C1 allele C*07:02:01 (albeit at a low level), and indicated that KIR2DL3 binds to two different ligands on this target: one that is HLA-C, expressed in normal configuration, and a second ligand that does not require β2-microglobulin and cannot be blocked by anti–HLA class I Abs. β2-Microglobulin knockdown in T47D cells reduced the KIR2DL3 reporter response (Fig. 6B), in accordance with our finding that T47D was C1 (HLA-C*08:02:01) and that the anti–HLA class I Abs A6-136, 6A4, and B9.12.1 efficiently blocked the KIR2DL3 reporter response (data not shown). The mAb HC10 binds a conformation-independent epitope shared by HLA-B and -C alleles. Staining with this Ab was also reduced as a result of β2-microglobulin siRNA knockdown, but comparably less than with the other HLA-ABC–reactive mAbs. This suggested some degree of surface expression of HLA-C in open conformation (Supplemental Fig. 3 and data not shown). However, the KIR2DS2/-2DL3–reactive target SK-BR-3 did not stain with HC10 (Fig. 3). Although low expression levels of open conformer HLA on the target cells could theoretically be sufficient to activate reporter cells, these data overall suggest that KIR2DS2 does not recognize open conformers of HLA-B or -C.
KIR2DS2 recognition of cancer cell ligand does not require β2-microglobulin. siRNA knockdown of β2-microglobulin in cancer cells followed by KIR-CD3ζ reporter cell assays is shown. (A) Knockdown was measured at a surface expression level by flow cytometry using an anti–β2-microglobulin mAb and the anti–HLA-ABC mAb B9.12.1. Knockdown (kd) as percentage reduction in mean fluorescence intensity is indicated in the upper right corner. (B) Reporter assays using the indicated KIR-CD3ζ reporters and cancer target cells transfected with either β2-microglobulin siRNA or control siRNA. HLA C1/C2 genotype of the target cells is indicated. Column histograms display normalized relative fractions of EGFP+ reporter cells. The average of three or more independent experiments for each cell line is shown. A statistically significant reduction compared with control siRNA was found. *p ≤ 0.02.
Discussion
In this study, we describe the use of cellular reporters to detect a ligand for KIR2DS2 expressed by human carcinoma cell lines. We demonstrate, to our knowledge for the first time, a strong interaction of KIR2DS2 with a cancer cell–expressed ligand. All the cancer targets that were recognized by the KIR2DS2 reporters were also recognized by KIR2DL2 and KIR2DL3, independent of the C1/C2 genotype of the targets. Taken together, our results suggested that KIR2DL3 binds two different ligands on the cancer cells where one ligand was β2-microglobulin–dependent HLA-C1 and the other ligand was β2-microglobulin–independent and was also recognized by KIR2DS2.
Fig. 3), partly explaining the lack of detection with 2DL3–Fc fusion protein. Notably, these targets reproducibly activated the 2DL3 reporter cells.
We did not observe any cancer cells that were recognized by KIR2DS2 only and not by KIR2DL3, indicating that these two receptors could recognize the same β2-microglobulin–independent ligand. Moreover, our data indicate that the KIR2DS2/KIR2DL3 ligand on the cancer cells is not HLA-C expressed in its native form together with β2-microglobulin. HLA class I molecules can be expressed on the cell surface without peptide and β2-microglobulin, in so-called open conformation (53). In this conformation, many of the otherwise blocking Abs would not bind. Moreover, free HLA class I H chains can be surface expressed as dimers (53) and may form heterodimers, as previously reported for HLA-F (54). Free H chain dimers can serve as ligands for LRC-encoded receptors; KIR3DL2 can bind HLA-B27 dimers (55), and LILR1 is reported to bind several class I alleles in dimeric form (56). Because HLA-C molecules are well established as ligands for the inhibitory KIR2D receptors, one possible hypothesis is that HLA-C in open conformation, without β2-microglobulin, could represent the ligand for KIR2DL3 and KIR2DS2 on the cancer targets, as previously reported for KIR3DS1, KIR3DL2, and KIR2DS4 recognition of open conformation HLA-F (57–59). This would explain our observation that the anti–HLA class I Abs did not block the KIR2DS2 or KIR2DL3 responses toward the cancer targets. Surface staining with the mAb HC10, which binds HLA-B and HLA-C in open conformation, was markedly reduced upon siRNA knockdown of β2-microglobulin, without any reduction in the strength of the KIR2DS2 reporter response. The reduction in HC10 staining upon β2-microglobulin knockdown indicates that this mAb also binds normally assembled heterotrimers. The loss of HC10 binding likely corresponded to an asymmetrical loss of normal HLA class I with no reduction in open conformer expression (Supplemental Fig. 3). However, the SK-BR-3 cell line showed no HC10 binding, but activated the 2DS2 reporter, suggesting that the cancer cell reactivity with KIR2DS2 reporters is not due to surface expression of HLA-C in open conformation. The HCA2 mAb binds open conformers of HLA-A, and together with HC10 these two Abs should react with most free HLA-A, -B, and -C H chains. These mAbs have been reported to block KIR3DS1binding to open conformers of HLA-F (58), but in our experiments they did not block 2DS2 reporter response toward tumor cells (Fig. 5).
The tumor targets did not express significant levels of HLA-G (Supplemental Fig. 4). β2-Microglobulin knockdown efficiently reduced staining with the W6/32 mAb, which stains both classical and nonclassical class I molecules, including HLA-E and -F.
Most of the reduction in W6/32 was probably attributed to loss of HLA-A, -B, and -C. Based on the efficient knockdown of HLA-A, -B, and -C, together with the lack of HLA-G expression by the investigated tumor targets, we would expect surface expression of HLA-E to also be reduced. This was supported by reduced staining with DT9, which reacts with HLA-E and -C (data not shown).
A recent report found that KIR2DS2 interacts with HLA-A11 in a peptide-dependent manner (27). This could not explain our observations with the tumor targets. None of the cancer targets expressed HLA-A11. Moreover, CHO cells expressing HLA-A11 at high surface levels together with human β2-microglobulin did not trigger the KIR2DS2 or KIR2DL3 reporters in our hands. The peptides loaded in CHO cells will to some extent differ from human cells, and we cannot rule out that HLA-A11, in a cell that produces the appropriate peptides, would be recognized by KIR2DS2 reporter cells.
Our data showed that several independent cancer cell lines express the ligand for KIR2DS2/KIR2DL3, whereas others do not. Awaiting a molecular identification of the ligand (ongoing in our laboratory), one hypothesis is that the KIR2DS2/KIR2DL3 ligand is expressed or significantly upregulated during carcinogenesis, possibly as a marker of cellular stress parallel to the NKG2D ligands MHC class I polypeptide-related sequence A and B or UL16-binding protein 1–4. Alternatively, the ligand might be a surface Ag ordinarily expressed by epithelial cells, which is abnormally folded or undergoes unconventional posttranslational modifications in some cancer targets.
NK cell cytotoxicity toward target cells is regulated by a balance between activating and inhibitory receptors. Our results suggested that KIR2DL3 (and KIR2DL2) recognized the same ligand as KIR2DS2. To investigate the KIR-related selective forces that could have acted upon the ligand-positive tumors during tumor evolution, we KIR genotyped the investigated cancer cell lines. Two of the targets (PC-3 and SK-BR-3) came from KIR2DS2−, KIR2DL3+ individuals (Table II). In these two cases, de novo expression of a KIR2DL3 ligand on the tumor cell could offer protection from NK cell lysis (e.g., mediated by NKp30 and NKG2D), without the risk of activating KIR2DS2+ NK cells. Most human populations have a large proportion of group A KIR haplotype homozygous individuals. These lack KIR2DS2, but carry the KIR2DL3 locus. Thus, in group A/A individuals, this mechanism would provide an advantage for tumor cell survival. The same evasion strategy could hypothetically also operate in virus-infected cells. It is thus possible that KIR2DS2, an activating close relative of KIR2DL3, has evolved to counteract this putative evasion strategy. A similar example would be mouse Ly49H, which is thought to have evolved from the inhibitory receptor Ly49I to counteract the virus (murine CMV)–encoded decoy ligand m157. The DU145 and OVCAR-3 target cell lines came from KIR2DS2+ individuals that were KIR2DL3− but instead KIR2DL2+ (Table II). In these situations, cancer has developed although one would expect that a proportion of KIR2DS2+, -2DL2− NK cells would be able to kill the developing tumors. Moreover, depending on several factors, even KIR2DS2+, -2DL2+ NK cells could conceivably become activated to kill the cancer cells. KIR2DS2 and KIR2DL2 are in strong linkage disequilibrium; KIR2DS2+ haplotypes almost always carry KIR2DL2 instead of KIR2DL3. In our experiments, KIR2DL2 and -2DL3 reporter cells bind equally well to tumor cell ligand, although their reactivity toward individual C2 group alleles appears to differ (Supplemental Fig. 1A and data not shown) in accordance with previous findings (20). The ability to detect malignant cells by their expression of a ligand for KIR2DS2 would provide a selective advantage for those carrying this gene, contributing a selective pressure to maintain this receptor in the genome. The KIR2DS2 gene frequency varies greatly between populations, from <20% in some East Asian populations (60) to >80% in Australian aborigines (61). This difference seems more likely to be related to different microbial environments than to differences in incidence of neoplasms.
Considering the possibility that the tumor cell ligand is HLA class I in open conformation, unable to present peptides to T cells but serving as a ligand for the inhibitory NK receptors KIR2DL2 and KIR2DL3, this could hypothetically represent a tumor mechanism to avoid NK cells as well as T cells. We find it unlikely that a ligand with these characteristics would be germline encoded, because it is hard to envisage what evolutionary advantage it would provide, but would not like to exclude the possibility that this ligand could have a viral genetic origin.
Alternatively, defects/alterations in the posttranslational processing of normal, germline-encoded surface molecules could have produced a ligand with these characteristics, providing a selective advantage over other cells in the tumor population. The ligand-bearing carcinoma lines all expressed high levels of HLA class I, presumably capable of providing sufficient NK cell inhibition through interaction with KIR2DL/KIR3DL receptors or CD94/NKG2A. Although this point would speak against the hypothesis, it cannot be excluded that cancer-related modifications of class I/peptide presentation in tumor cells (altered self) could have hampered also interaction with these inhibitory receptors.
We and others have previously discussed the origin and evolution of pairs of activating and inhibitory NK cells receptors (62, 63). It seems likely that KIR2DS2 might have evolved from a KIR2DL2/3-like ancestor, and one hypothesis suggests that activating NK receptors in evolution have undergone cycles of birth and death, always arising from an inhibitory ancestor by recombination events (62). Although not yet completely understood, most if not all pairs of activating/inhibitory NK cell receptors seem to be under a selective pressure to maintain some similarity in their ligand-binding domains, occurring by exchange of genetic material (gene conversion or unequal crossing-over) between activating and inhibitory receptor loci (63). Although incompletely understood in the absence of identified ligands for activating KIR, this suggests that the functions of activating and inhibitory receptors such as KIR2DS2 and -2DL2/-2DL3 are interrelated, in some instances possibly regulating NK cell responses to the same ligand.
The significance of the KIR2DS2/KIR2DL2/KIR2DL3 ligand is yet undetermined. Whether the ligand is germline encoded and upregulated during carcinogenesis, or represents aberrant posttranslational modifications, it could to some extent be expressed by healthy cells, albeit at low levels. This could explain the need for a balanced sampling of ligand expression, where activating signals through KIR2DS2 would be kept in check by inhibitory signals from KIR2DL2 and KIR2DL3. How this balance would tip in favor of NK cell activation and killing of the cancer cell remains hypothetical. To this end, ongoing work in our laboratory is aiming to disclose the molecular nature of the ligand for KIR2DS2 as well as the biological mechanism behind its expression.
Disclosures
The authors have no financial conflicts of interest.
Acknowledgments
We thank Wendi Jensen and Hildegunn Dahl for technical assistance and Jodie P. Goodridge and Karl-Johan Malmberg for sharing reagents.
Footnotes
This work was supported by the Norwegian Cancer Society and by the Research Council of Norway.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- CHO
- Chinese hamster ovary
- EGF
- epithelial growth factor
- EGFP
- enhanced GFP
- KIR
- killer cell Ig-like receptor
- siRNA
- small interfering RNA
- SSP
- sequence-specific primer.
- Received May 31, 2016.
- Accepted January 22, 2017.
- Copyright © 2017 by The American Association of Immunologists, Inc.