KIR3DL1 Polymorphisms That Affect NK Cell Inhibition by HLA-Bw4 Ligand

Generation of KIR⁺ NK leukemia cell lines and cultured NK cells

PBMC samples were obtained from healthy human donors by informed consent under an Institutional Review Board approved protocol. 3DL1*007 (AF262973) was isolated as described (2) and subcloned into a retroviral vector pMX-puro (13), which is a gift from Dr. T. Kitamura (The Institute of Medical Science, University of Tokyo, Tokyo, Japan). NK leukemia (NKL) cells, a gift from Dr. M. Robertson (Department of Medicine, Indiana University, Indianapolis, IN) (14), were transduced to express KIR using the Phoenix Helper dependent protocol as described by Gary Nolan (〈www.stanford.edu/group/nolan/protocols/pro_helper_dep.html〉). KIR3DL1 mutants were generated using the Quick Change site-directed mutagenesis kit (Stratagene) according to the manufacturer’s instructions. The NKL 3DL1*002 cell line, which was produced by transduction of NKL with the 3DL1*002 (U31416) pMX retroviral vector, was generously provided by Dr. L. Lanier (University of California, San Francisco, CA).

PE-labeled, NK cell receptor-specific Abs anti-KIR3DL1 mAb DX9 (DNAX); anti-KIR2DL1 mAb EB6 (Corixa); anti-CD94 mAb HP3D9 (BD Biosciences); anti-NKG2A mAb Z199 (Corixa); and unconjugated anti-LILRB1 mAb HP-F1, a gift from M. López-Botet (Universitat Pompeu Fabra (DCEXS), Barcelona, Spain) with PE-labeled goat anti-mouse F(ab′)₂ were used to characterize the NKL cell lines by flow cytometry. KIR⁺ NKL cell lines were sorted for equivalent KIR expression with a FACStar cell sorter and clone-Cyt software and hardware (BD Biosciences).

Cultured NK cells were derived by the coculture of PBMC with irradiated RPMI 8866 cells as described (15). Following this expansion and subsequent enrichment by the MACS NK Cell Isolation kit (Miltenyi Biotec) per the manufacturer’s instructions, 3DL1*002⁺/LILRB1⁻ NK cells were sorted with a FACStar cell sorter and maintained in RPMI 1640 medium supplemented with 10% FBS, 2 mM glutamine (Invitrogen Life Technologies), 100 U/ml penicillin/streptomycin (Invitrogen Life Technologies), and 100 U/ml rIL-2 (National Institutes of Health, National Cancer Institute Preclinical Repository, Frederick, MD).

Cytolytic assays

Cell killing assays (i.e., 4-h ⁵¹Cr release assays) were performed as described (16) using KIR transduced NKL cell lines or polyclonal NK cell lines as effectors and ⁵¹Cr-labeled target cells: 721.221 (untransfected) and 721.221 (transfectants) expressing individual HLA class I allotypes that have been previously described (7). PE-conjugated, pan-HLA-specific mAb W6/32 (17) was used to assess and sort for equivalent class I expression on the 221 transfectants. Unconjugated Abs (anti-HLA mAb DX17 (DNAX), anti-LILRB1 mAb HP-F1, and anti-KIR3DL1 mAb DX9) were used as blocking reagents in cell killing assays at a concentration of 25 μg/ml per well.

Data analysis

Where appropriate, results were presented as specific lysis or the percentage of inhibition of lysis (IOL) relative to the lysis of 721.221 to normalize lytic activity between the cell lines and adjust for variation in the lysis of the 721.221 cell line. The percentage of IOL was calculated with the following formula: [1 − (percent-specific lysis of target cells)/(mean of percent-specific lysis of 721.221)] × 100. Additionally, some results were presented as the relative percentage of IOL compared with lysis by a control NKL cell line. All effectors lysed untransfected 221 target cells at levels exceeding 30% at an E:T ratio of 20:1, and the background spontaneous release for all targets was <15%.

A predicted model of 3DL1*002 (Accession no. U31416) was generated with Swiss Model (〈www.expasy.org/swissmod/SWISS-MODEL.html〉) (18). In this model the D0 domain was modeled separately from the D1 and D2 domains and the complete structure was assembled and viewed using Deep View and MolScript version 2.1 programs, respectively (〈www.avatar.se/molscript.html〉).

KIR3DL1 transduced NKL cell lines reproduce predicted HLA specificities

Retroviral transduction was used to make lines of NKL cells that stably express 3DL1*002 and 3DL1*007. Cell sorting was used to produce transductants expressing comparable amounts of 3DL1*002 and 3DL1*007 at the cell surface (Fig. 2⇓A). Using anti-KIR3DL1 Ab titration we observed no difference in Ab staining between the two KIR3DL1 variants (data not shown), suggesting equivalent Ab affinities. KIR transduced and untransduced NKL cells were compared for their capacity to kill 721.221 cells transfected with B*5101, a Bw4⁺ ligand for the KIR3DL1 receptor, and with B*1502, a Bw6⁺ allotype. As expected, B*5101 elicited stronger lysis inhibition from NKL cells expressing the KIR3DL1 variants than from untransduced NKL cells (Fig. 2⇓B). For NKL cells expressing either 3DL1*002 or 3DL1*007, stronger lysis inhibition occurred with B*5101 than with B*1502, although the latter was reproducible. Gumperz et al. (7) also found that some KIR3DL1⁺ NK cell clones were specifically inhibited by B*1502, but these same clones were significantly more inhibited by Bw4 allotypes (7). Thus, our data were consistent with previous descriptions of KIR3DL1 specificity for Bw4 HLA-B. Based on these data and previous descriptions of LILRB1 recognition of HLA-B allotypes (20, 21), we attributed inhibition of untransduced NKL cells to LILRB1 expression. However the observed differences in response between KIR3DL1 transduced and untransduced NKL cells were not due to differences in LILRB1 expression because LILRB1 expression was equivalent for transduced and untransduced NKL cells (Fig. 2⇓C).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

KIR3DL1 transduced cells express equivalent levels of KIR and are lysis inhibited by HLA-Bw4⁺ B*5101 similar to KIR3DL1⁺ peripheral blood NK cells. A, Expression of two KIR3DL1 allotypes on retrovirally transduced NKL cells in flow cytometry using the PE-conjugated KIR3DL1 specific mAb, DX9. B, KIR3DL1 allotypes, 3DL1*002 and 3DL1*007, reproduce predicted HLA specificities in a 4-h Cr release assay, which is representative of three independent experiments. Shown is the lysis of three target cells: 721.221 (♦), 721.221 cells transfected with Bw6⁺ HLA-B*1502 (▵) or Bw4⁺ HLA-B*5101 (▪) in the absence of anti-LILRB1 mAbs. All SDs are shown and where they are not apparent they are less than the size of the symbol. C, LILRB1 expression on KIR3DL1 transduced and untransduced NKL cells is equivalent. Untransduced NKL (dotted line), NKL 3DL1*002 (dashed line), and NKL 3DL1*007 (solid line), and untransduced NKL stained with an isotype control (filled histogram) are shown. D, Sorted and cultured NK cells isolated from the peripheral blood of one donor exhibit a 3DL1*002⁺/LILRB1⁻ phenotype. Freshly isolated, unsorted NK cells (thick line), cultured and sorted NK cells (filled histogram), and sorted NK cells stained with an isotype control (dotted line) are shown. E, 3DL1*002⁺/LILRB1⁻ NK cells are lysis inhibited by Bw4⁺ HLA-B allotypes similar to the 3DL1*002 transduced NKL cells. In this donor the 3DL1*002 allotype was the only KIR3DL1 protein expressed at the cell surface. Shown is the lysis of four target cells: 721.221 (♦), 721.221 cells transfected with Bw6⁺ HLA-B*1502 (▵), Bw4⁺ HLA-B*2705 (▴), or Bw4⁺ HLA-B*5101 (▪) in the absence of anti-LILRB1 mAbs. All SDs are shown and where they are not apparent they are less than the size of the symbol. These data are representative of two replicate experiments.

With this in vitro system, we replicated KIR3DL1 specificities previously described in the literature; but we had not established that KIR⁺ cell lines responded like NK cells derived from peripheral blood. For this reason, we tested KIR3DL1⁺ peripheral blood NK cells with representatives from the same panel of HLA transfectants as well as an additional Bw4⁺ HLA-B transfectant, B*2705. We derived polyclonal 3DL1*002⁺, LILRB1⁻ NK cells from the peripheral blood of a donor in whom 3DL1*002 was the only expressed 3DL1 receptor. Following in vitro expansion and sorting we verified the 3DL1*002⁺, LILRB1⁻ phenotype of the NK cells using flow cytometry (Fig. 2⇑D).

The response of these primary NK cells was consistent with the specificities of the 3DL1*002⁺ NKL cells (Fig. 2⇑E). The ranking of the inhibitory response elicited relative to 721.221 lysis was B*5101>B*2705>B*1502 from greatest to least lysis inhibition consistent with data from previous studies (7, 22). None of these class I allotypes (i.e., B*5101, B*2705, or B*1502) had a canonical signal sequence that was permissive for CD94:NKG2A ligand expression (19). Thus, we found that some HLA-B ligands for 3DL1*002 elicited stronger inhibitory responses than others.

Differences between 3DL1*002 and 3DL1*007 alter receptor strength

To distinguish inhibition due to KIR3DL1 from that due to LILRB1, assays were performed in the presence and absence of specific mAbs. In the absence of receptor blocking Abs, lysis of all three NKL cell lines was inhibited, though less so for NKL than for the two transductants. In the presence of anti-LILRB1 mAb a clear hierarchy of inhibition was seen: NKL expressing 3DL1*002 > NKL expressing 3DL1*007 > untransduced NKL cells (Fig. 3⇓). Thus, interaction of HLA-B*5101 with 3DL1*002 generates a stronger inhibitory response than interaction of HLA-B*5101 with 3DL1*007. This difference was most evident by the percentage of lysis inhibition relative to lysis of 721.221 (Fig. 3⇓B). Because the expression levels of the two KIR3DL1 allotypes were comparable, this difference demonstrates that one or more of the amino acid substitutions that distinguish 3DL1*002 from 3DL1*007 make it a stronger receptor for HLA-B*5101.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

KIR3DL1 and LILRB1 both contribute to inhibit NK cell function. The killing of 721.221 cells and 721.221-B*5101 cells by NKL cells and NKL cells transfected with either 3DL1*002 or 3DL1*007 was compared in the absence and presence of specific mAbs in 4-h Cr release assays at an E:T ratio of 20:1. The results are plotted as both a percentage specific lysis (A) and relative percentage of IOL (B) by 721.221-B*5101 targets in comparison to the lysis of untransfected 721.221 targets where the relative percentage of IOL is calculated using [1 − (percent-specific lysis of target cells)/(mean of percent-specific lysis of 721.221)] × 100. The combination of anti-LILRB1 and anti-KIR3DL1 mAbs abrogated the inhibition to an extent greater than either Ab alone. Ab specific for CD56 served as a negative control, Ab specific for HLA class I as a positive control. NKL (□), NKL 3DL1*002 (▪), and NKL 3DL1*007 (▦) are shown. The data presented are the average of three separate experiments. SEs are indicated. Statistical significance shown where ∗∗, p < 0.01 and ∗, p < 0.05 are by t test.

KIR3DL1 and LILRB1 both contribute to inhibit transduced NKL cells

When anti-KIR3DL1 mAb was added to the cytolytic assay, lysis inhibition caused by B*5101 binding to 3DL1*002 or 3DL1*007 was partially blocked, whereas there was no affect on inhibition in the untransduced NKL parental cell line (Fig. 3⇑). In contrast, when both anti-LILRB1 and anti-KIR3DL1 mAbs were included there was an equivalently strong blockade of inhibition of NKL cells expressing 3DL1*002 or 3DL1*007 and the parental cell line that was comparable to that observed in the presence of anti-HLA class I mAb. Thus, in the NKL cell transductants that express both LILRB1 and KIR3DL1, lysis inhibition is the consequence of both receptors recognizing the same HLA ligand. Individually, the inhibition mediated by LILRB1 is stronger than that mediated by KIR3DL1; however, when one of these receptors is blocked (e.g., LILRB1) the other receptor (e.g., 3DL1*002) can compensate to inhibit lysis. Functional differences between the KIR3DL1 variants were only apparent when the effects of LILRB1 were removed by the presence of anti-LILRB1 Ab. When LILRB1 function was not blocked (i.e., in the presence of a control Ab or no Ab) we did not observe differences between the KIR3DL1 variants. This outcome suggests that LILRB1 can compensate for functional differences between 3DL1*002 and 3DL1*007.

On interaction with HLA-Bw4, 3DL1*002 gives stronger inhibition of cytotoxicity than 3DL1*007

The removal of LILRB1-mediated inhibition revealed allotype differences in the inhibitory response to B*5101. Specifically, cells expressing 3DL1*002 were more inhibited than those expressing 3DL1*007. However, the question of whether this was a general phenomenon or specific to 3DL1*002-HLA B*5101 interactions remained unresolved. To address this issue, we compared the inhibitory responses generated by B*5101 to those generated by another Bw4⁺ HLA-B allotype, B*2705 (Fig. 4⇓). Again, we used anti-LILRB1 mAb to exclude the contributions of LILRB1 to lysis inhibition. As in previous experiments with B*5101, both 3DL1*002- and 3DL1*007-expressing cell lines were more inhibited than untransduced NKL cells in the presence of either a control Ab or no Ab (p < 0.01) (Fig. 4⇓A). By comparison, B*2705 elicited no difference between 3DL1*007-expressing and untransduced NKL cells, but did increase the inhibition through 3DL1*002. The levels of class I expression were equivalent between the B*5101 and B*2705 transfectants (Fig. 4⇓B). Thus, we attributed the differences in inhibitory response to intrinsic differences in the HLA-B ligands. We found that B*5101 elicited a stronger inhibitory response than B*2705. This finding confirms previous observations made by Cella et al. (22).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4.

3DL1*002 is a stronger Bw4-specific inhibitory receptor than 3DL1*007. A, NKL cell lines expressing either 3DL1*002 or 3DL1*007 were compared in their lysis of HLA B*5101 and HLA B*2705 expressing 721.221 cells in 4-h Cr release assays at an E:T ratio of 20:1. In this comparison, a control (untransduced NKL cells) was also included. The data are plotted as relative lysis inhibition compared with the lysis of class I-negative 721.221 cells. As done previously, anti-LILRB1 Ab was added to eliminate LILRB1-mediated lysis inhibition and anti-CD56 Ab was used as a control. The data presented are the average of three separate experiments. NKL (□), NKL 3DL1*002 (▪), and NKL 3DL1*007 (▦) are shown. Statistical significance shown where ∗∗, p < 0.01 is by t test. B, Class I expression on HLA transfectants using anti-HLA Ab, W6/32 PE-labeled. 721.221 (filled histogram), 721 B*5101 (solid line), and 721 B*2705 (dashed line).

In the presence of anti-LILRB1 Ab, we observed more marked differences among the KIR3DL1 variants in their inhibitory response to B*5101 and B*2705. Both B*5101 and B*2705 elicited significant inhibition through 3DL1*002 (p < 0.01); however, B*2705 did not elicit inhibition through 3DL1*007 (Fig. 4⇑A). Thus, 3DL1*007 either does not recognize or responds poorly to B*2705. Although 3DL1*007 was inhibited by B*5101, KIR3DL1*002 was more inhibited by this class I allotype. Therefore, 3DL1*002 was the stronger inhibitory receptor with more than one Bw4⁺ HLA-B ligand, and this was most evident in the response to HLA-B*2705.

The functional difference between 3DL1*002 and 3DL1*007 is determined by polymorphism at position 238 in the D2 domain and position 320 in the transmembrane region

The 3DL1*002 and 3DL1*007 allotypes differ at three positions in the amino acid sequence: 238 in the D2 domain, 320 in the transmembrane region at the junction of the extracellular and transmembrane domains, and 373 in the cytoplasmic tail (Fig. 5⇓A) (9). To determine the contribution of each position to the phenotypic difference we made a panel of six mutants, three from 3DL1*002 and three from 3DL1*007, in which each mutant had one position changed to the residue present in the other KIR3DL1 allotype (Fig. 5⇓B). These mutants were transduced into NKL cells and the transductants sorted to obtain cells having comparable levels of KIR3DL1 expression at the surface (Fig. 5⇓C). The mutant and the wild-type KIR3DL1 were then compared for their inhibitory capacity in cytolytic assays using 721.221-B*2705 target cells and inclusion of anti-LILRB1 mAb. The lysis obtained with transductants expressing KIR3DL1 variants and mutants was compared with that obtained with untransduced NKL cells.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5.

Substitution at position 238 in the D2 domain and position 320 in the transmembrane domain make 3DL1*002 a stronger inhibitory receptor than 3DL1*007. A, Sequence alignments of the mature proteins reveal that 3DL1*002 and 3DL1*007 differ at three positions, one in each of the extracellular (D2), transmembrane (TM), and cytoplasmic (CYT) domains, respectively. The predicted boundary between extracellular and transmembrane domains is indicated by an arrowhead, where the extracellular region is to the left of the arrowhead and the transmembrane region is to the right. B, Six point mutants of 3DL1*007 and 3DL1*002 were made. Each had a residue swapped at one of the three positions, where the natural wild-type (WT) allotypes differ: 238 in the D2 domain (D2), 320 in the transmembrane region (TM), and 373 in the cytoplasmic tail (CYT). C, Expression of mutated KIR3DL1 allotypes is equivalent on transduced NKL cells: untransduced NKL (filled histogram), NKL 3DL1*002 WT (green line), NKL 3DL1*007 WT (pink line), NKL 3DL1*007 238R (orange line), NKL 3DL1*007 320I (purple line), NKL 3DL1*007 373E (navy blue line), NKL 3DL1*002 238G (black line), NKL 3DL1*002 320V (yellow line), and NKL 3DL1*002 373Q (light blue line). D, The capacity of NKL cells transfected with the point mutant KIR3DL1 to be inhibited by HLA-B*2705 (▪) in the presence of anti-LILRB1 mAb was compared with untransfected NKL cells. Lysis of 721.221 cells (□) in the presence of anti-LILRB1 mAb was also included for comparison. E, In addition, the data from D were normalized to lysis inhibition of NKL cells where the relative percentage of IOL is calculated using [(percent lysis inhibition of effector − percent lysis inhibition of NKL)/(percent lysis inhibition of NKL)] × 100. F, Three double mutants of 3DL1*002 were made, each of which had two of the three difference residues swapped from 3DL1*007. A triple mutant was also made, which converted 3DL1*002 to 3DL1*007. G, The capacity of NKL cells transfected with the double and triple mutant KIR3DL1 to be inhibited by HLA-B*2705 was compared with untransfected NKL cells. H, The data from G were also normalized to lysis inhibition of NKL3DL1*007 where the relative percentage of IOL was calculated by [(percent lysis inhibition of effector − percent lysis inhibition of NKL3DL1*007)/(percent lysis inhibition of NKL3DL1*007)] × 100. All cytolytic assays were performed at an E:T ratio of 20:1 in the presence of anti-LILRB1 mAb. Data presented are the average of five (D and E) and three (G and H) separate 4-h Cr release assay experiments with SEs as indicated.

Target cells expressing HLA-B*2705 were selected for this assay based on their ability to differentially inhibit the wild-type KIR3DL1 variants relative to untransduced cells. HLA-B*2705 is in general a weaker inhibitor of KIR3DL1 than B*5101, and 3DL1*007 is a weaker receptor than 3DL1*002. Whereas 3DL1*002 engages B*2705 to inhibit lysis, the combination of HLA-B*2705 and 3DL1*007 gave no inhibition (Figs. 4⇑A and 5⇑, D and E). For a detailed comparison of the effects of the mutant and wild-type KIR3DL1 allotypes, the data were normalized to untransduced NKL cells (Fig. 5⇑E). Mutants of 3DL1*007, with substitution at either position 238 or 320, engage B*2705 to give some inhibition. Conversely the reciprocal mutations in 3DL1*002 reduce its inhibitory capacity. Mutation at position 373 in either 3DL1*002 or 3DL1*007 had no functional effect.

As none of the mutants having a single residue change altered the phenotype completely, four further mutants of 3DL1*002 were made: three had the various combination of two amino acid substitutions, the fourth had all three substitutions and served as a control (Fig. 5⇑F). As done previously, the mutants were sorted for equivalent KIR3DL1 expression (data not shown). To determine the relative impact of these double and triple substitutions in 3DL1*002 the data were normalized to wild-type 3DL1*007 (Fig. 5⇑, G and H). As predicted from the results obtained from the single residue mutants, the double mutant of 3DL1*002 with substitution at positions 238 and 320 gave no inhibition, like 3DL1*007 and the triple mutant, whereas the two other double mutants retained some inhibitory capacity. In conclusion, substitution at position 238 in the D2 domain and position 320 in the transmembrane region contribute to 3DL1*002 being a stronger inhibitory receptor than 3DL1*007, with each position making a comparable contribution to the difference.

To further understand how these residues influenced inhibitory function we generated a predicted molecular model of 3DL1*002 based on established KIR2D crystal structures (23, 24). In the KIR2D/HLA co-crystal structures ligand binding is localized to KIR residues in the interdomain hinge region. In our model we found that residue 238 in the D2 domain was distant from the putative ligand-binding site in a loop opposite the interdomain hinge region (Fig. 6⇓). Thus, we propose that the enhanced inhibitory effects of these residues are due to indirect influences on ligand recognition, rather than direct ligand interactions.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 6.

Residues 238 and 320 that give functional difference to 3DL1*002 and 3DL1*007 are not within the predicted ligand binding site. A model of KIR3DL1 based upon the known structures for KIR2D is shown. The ligand binding site for HLA-Bw4 is predicted to be formed by interaction of the D1 and D2 domains as indicated by the arrow. Residue 238 is situated in a β-strand loop of the D2 domain but is distant from the ligand binding site; it is shown (boxed) as the arginine present in 3DL1*002. Residue 320 is in the transmembrane domain, which is not shown in the schematic.