Cis association of leukocyte Ig-like receptor 1 with MHC class I modulates accessibility to antibodies and HCMV UL18


Full correspondence Dr. Deborah Burshtyn, Department of Medical Microbiology and Immunology, 659 Heritage Medical Research Centre, University of Alberta, Edmonton, AB T6G 2S2, Canada

Fax: +1-780-492-9828



Leukocyte Ig-like receptor (LIR) 1 (CD85j/ILT2/LILRB1) is an inhibitory receptor with broad specificity for MHC class I (MHC-I) and the human CMV MHC-I homologue UL18. LIR-1 can inhibit NK cells through the conventional interaction with MHC-I expressed on a target cell (in trans) but the nature and the effects of LIR-1 interactions with MHC-I in cis are not well understood. Here we show that MHC-I expressed in cis has an impact on the detection of LIR-1 with various antibodies. We found the cis interaction alters recognition by only one of two antibodies known to block functional trans recognition by LIR-1 on NK cells. Specifically, we observed an enhancement of recognition with GHI/75 in the presence of various MHC-I alleles on 721.221 cells. We found that blocking the LIR-1 contact site with anti-MHC-I antibodies decreased detection of LIR-1 with GHI/75. We also observed a decrease in GHI/75 following acid denaturation of MHC-I. Finally, disruption of LIR-1 cis interactions with MHC-I significantly enhanced UL18-Fc binding to NK92 cells and enhanced the relative inhibition of NK92 cells by HLA-G. These results have implications for LIR-1 function in scenarios such as infection when MHC-I levels on effector cells may be increased by IFNs.


NK cells are cytotoxic lymphocytes of the immune system, which provide innate protection against viral infection and tumor transformation [1]. NK cells eliminate altered self-cells through the direct recognition and subsequent lysis of engaged targets. The effector functions of these immune cells are dynamically regulated through the coordinated signaling of cell surface receptors that either activate or inhibit responses [2]. The stimulatory NK-cell receptors signal through coupled adaptor proteins that encode cytoplasmic ITAMs [3]. Conversely, the inhibitory receptors dampen NK-cell responses via long cytoplasmic tails encoding ITIMs, which recruit SH2 domain-containing phosphatases such as SHP-1 following tyrosine phosphorylation [4]. The inhibitory receptors typically recognize MHC class I (MHC-I) molecules, and therefore healthy self-cells expressing a full complement of MHC-I are protected from NK-cell lysis. NK cells thus sense the density of ligands on engaged cells, and it is the balance of signals received through these functionally opposing receptors that ultimately determines both the response of the effector cell and the fate of the target cell.

Inhibitory MHC-I receptors expressed on human NK cells include the killer cell Ig-like receptors (KIRs) and leukocyte Ig-like receptor-1 (LIR-1), which are encoded on chromosome 19q13.4 within the leukocyte receptor complex [5]. The LIR family encodes receptors with either two or four Ig domains and includes both inhibitory and activating members. While the ligands for all LIR family members have not yet been identified, the inhibitory members LIR-1 and LIR-2 bind MHC-I at the relatively conserved α3 domain and β2-microglobulin, allowing for broad specificity and recognition of both classical and nonclassical molecules [6, 7]. NK-cell expression of LIR-1 varies between individuals and mediates inhibition of NK cells in response to MHC-I, particularly HLA-G [8, 9]. LIR-1 is also targeted by a virally encoded MHC-I homologue known as UL18 [10], which shares approximately 25% amino acid sequence identity with MHC-I and associates with human β2-microglobulin [11, 12]. UL18 binds to LIR-1 with more than 1000-fold greater affinity than MHC-I molecules [7, 13, 14] and has been shown to mediate suppression of LIR-1+ NK cells in vitro [15].

Studies of LIR-1 function to date have primarily focused on the interaction of the receptor with its ligand in trans, such as inhibition of NK or T cells through recognition of a ligand expressed on a target cell or APC [16-21]. However, there is mounting evidence that the interaction of these immune receptors with ligands on the same cell, or in cis, influence how they function [22]. In the murine system, the cis interaction of MHC-I specific NK-cell receptors plays a key role in regulating NK-cell responses. The mouse inhibitory receptor Ly49A has been shown to interact with its MHC-I ligand in cis, and this interaction directly reduces its accessibility to ligands in trans, thereby lowering the threshold for activation [23, 24]. The murine orthologue of LIR-1, PIR-B, interacts with MHC-I in cis on the surface of mast cells and this interaction was suggested to influence cellular activation during allergic responses [25]. LIR-1 has also been shown to interact with MHC-I in cis on the surface of osteoclast precursors, with this binding suggested to contribute to the regulation of osteoclast development [26]. Whether this cis interaction occurs on the surface other hematopoietic cells, and how it affects ligand binding in trans has not been addressed. Here we present evidence that LIR-1 engages MHC-I in cis on the surface of NK cells, and that the cis interaction with MHC-I directly modifies the accessibility of LIR-1 to antibodies and binding to soluble UL18. Our data suggest the LIR-1 interaction with MHC-I in cis limits the availability of LIR-1 for the engagement of ligands in trans, implying the sensitivity of NK cells to LIR-1-mediated inhibition during an immune response may be regulated by the level of MHC-I on the surface of the NK cell itself. To our knowledge, this is the first evidence of a cis interaction regulating an inhibitory receptor on human NK cells.


Cell surface MHC-I expression modifies staining patterns with anti-LIR-1 antibodies

Based on the effects of cis MHC-I interaction on Ly49 receptors and the reported cis interaction of LIR-1 with MHC-I, we hypo-thesized that the presence of MHC-I in cis would affect detection of LIR-1 by antibodies. We further predicted that an interaction with MHC-I in cis would decrease detection by antibodies known to prevent LIR-1 recognition of MHC-I in trans. We first examined whether cell surface expression of MHC-I changes detection of LIR-1 using the human B-cell line 721.221 expressing several distinct MHC-I proteins compared with the parental cells, which lack most MHC-I, but still stains low with the anti-MHC-I mAb W6/32 (Fig. 1A). LIR-1 staining patterns were assessed using the GHI/75 and HP-F1 antibodies on 721.221 expressing HLA-B58 and HLA-G compared with 721.221 cells. Surprisingly, we observed that on cells expressing MHC-I, staining patterns with GHI/75 were enhanced compared with the parental cell line (Fig. 1B). This was in contrast to what was observed with HP-F1, where the staining of cells co-expressing LIR-1 and MHC-I was generally reduced. Similar results with GHI/75 and HP-F1 were observed with 721.221 cells expressing HLA-Cw15 (data not shown). The exception among the cell lines tested was 721.221 cells expressing HLA-B58, which exhibited enhanced staining with both GHI/75 and HP-F1, although the increase in HP-F1 was not as dramatic as that observed with GHI/75. Despite this, we still found that overall in the presence of MHC-I, the MFI of LIR-1 staining with GHI/75 was consistently increased relative to that observed with HP-FI (Fig. 1C). Together these data suggest that in the presence of cell surface MHC-I molecules that can engage LIR-1, recognition by the GHI/75 antibody is improved. Therefore, the GHI/75 antibody appears to preferentially recognize cis bound LIR-1 molecules relative to the free unbound receptor.

Figure 1.

721.221 cells expressing MHC-I exhibit modified LIR-1 staining with GHI/75 and HP-F1. (A) 721.221 cells and 721.221 cells stably expressing various HLA class I molecules were stained with the pan-MHC-I antibody W6/32 and analyzed by flow cytometry. (B) 721.221 cells shown in (A) were stained with the anti-LIR-1 mAbs GHI/75 and HP-F1. A reference line bisecting the LIR-1 histogram on the parental cell line is included for each panel. The MFI is indicated in each panel. (C) The ratio of the MFIs of GHI/75 to HP-F1 is shown. (D) CSFE-labeled 721.221 and unlabeled 721.221 cells expressing HLA-B58 were combined overnight prior to staining GHI/75. The gate employed for the GHI/75 staining on 721.221 cells in the mixed population is shown (left). Staining of both cell lines incubated alone is also shown (right). (A–D) Results shown are from one experiment representative of three experiments performed.

In order to examine whether the observed changes in staining on 721.221 cells with GHI/75 in the presence of MHC-I might be due to trans interactions through LIR-1, we cultured CSFE-labeled 721.221 cells and unlabelled cells expressing HLA-B58 in combination overnight. The cultured cells were then examined for LIR-1 staining patterns the next day. We found that cells cultured together overnight exhibited a comparable staining pattern as cells that were cultured overnight alone (Fig. 1D). In these assays the HLA-B58-expressing cells appeared to grow faster as they are removed from selection when placed in culture with 721.221 cells. Therefore, it does not appear that interactions on an opposing cell enhance the staining of LIR-1 observed with GHI/75 on cells co-expressing MHC-I.

Disruption of cis interactions by anti-MHC-I treatment modifies LIR-1 staining

The pan MHC-I monoclonal antibody W6/32 is able to prevent LIR-1 inhibitory signaling triggered by target cells expressing MHC-I, and hence is able to block ligand binding in trans. Therefore, we tested whether W6/32 could be used to disrupt a cis interaction as well, and whether this would lead to changes in anti-LIR-1 staining. The results are shown for NK92 cells, a human NK-cell line that expresses high levels of LIR-1 as well as a full complement of MHC-I (Fig. 2A). NK92 cells were incubated in the presence of excess W6/32 at 37°C to encourage the release of any cis bound LIR-1 molecules. NK92 were also incubated without antibody or in the presence of an isotype control. Cells were then stained for LIR-1 with directly labeled GHI/75 and HP-F1 at 4°C and analyzed by flow cytometry. Consistent with the hypothesis that GHI/75 preferentially binds cis interacting LIR-1, we observed a dramatic decrease in LIR-1 staining following treatment with W6/32, but not an isotype control antibody (Fig. 2B). Conversely, LIR-1 detection by HP-F1 was found to be slightly enhanced compared with that in controls. When comparing the MFI of LIR-1 staining between samples, both the decrease in GHI/75 and increase in HP-F1 following W6/32 incubation was found to be statistically significant compared with that in mock and isotype control samples (Fig. 2C). We also tested the effect of W6/32 incubation on LIR-1 staining with our 721.221 cell lines. As expected, the staining of parental 721.221 cells with GHI/75 was unaffected by prior incubation with W6/32, but all lines stably expressing MHC-I exhibited decreased staining, consistent with the results with NK92 (data not shown). To examine whether this effect was specific to W6/32, we tested the ability of DX17, another pan MHC-I monoclonal antibody, to disrupt LIR-1 cis binding on NK92. Similar to what was observed with W6/32, preincubation with DX17 significantly reduced GHI/75 staining while enhancing HP-F1 detection (Fig. 2D and E). These results suggest that a proportion of LIR-1 molecules on the cell surface of NK cells are bound to MHC-I in cis and that the same antibodies that block recognition of MHC-I by LIR-1 in trans can prevent this interaction. Furthermore, given that detection by HP-F1 increases following W6/32 treatment, the loss of cell surface GHI/75 staining is likely caused by changes in receptor conformation, and not due to changes in the surface expression of LIR-1. Finally, it appears that the epitope for GHI/75 on LIR-1 is more sensitive than that of HP-F1 to MHC-I expression.

Figure 2.

Preincubation with anti-MHC-I antibodies reduces GHI/75 staining but enhances HP-F1 staining on NK92 cells. (A) NK92 cells were stained with W6/32 (left) and various HLA class I specific antibodies (right). (B) NK92 cells were stained with GHI/75 and HP-F1 following treatment with W6/32, control antibody, or media alone. The gray-line histogram represents staining with an isotype control antibody. (C) The average change in the MFI was calculated from three independent experiments for each antibody shown and is presented as mean + SE of three pooled samples. (D) NK92 cells were stained with GHI/75 and HP-F1 following treatment with DX17, control antibody, or media alone. The gray-line histogram represents staining with an isotype control antibody. (E) The average change in the MFI for GHI/75 and HP-F1 is shown as mean + SE of three samples pooled and is representative of three experiments. For these experiments, statistical significance was calculated using a two-sample t-test assuming equal variance (ns = not significant).

To further characterize the changes in the LIR-1 detection by GHI/75 caused by anti-MHC-I treatment, we tested the influence of time, temperature, and blocking energy metabolism on the changes produced in GHI/75 detection. To determine if longer incubations with W6/32 would lead to a further reduction, we extended the incubation time with W6/32 up to 3 h prior to LIR-1 staining. We observed no further effect on the decrease in detection with GHI/75 on both NK92 (Fig. 3A and B) and 721.221 cells expressing MHC-I (data not shown) beyond the first 30 min, although a shorter time period yielded a less than maximal reduction in staining. Therefore, changes in LIR-1 conformation occur quickly following antibody blocking of its binding region on MHC-I. However, the change in GHI/75 was minimal if the incubation with W6/32 was performed on ice at 4°C for 30 min (data not shown), suggesting a dynamic process was required. The most likely explanation is that at 37°C LIR-1 is binding and releasing from MHC-I more rapidly than at 4°C, providing access to the W6/32 antibody. Alternatively, an energy dependent cellular process could be involved. Therefore, to test this we performed the assay in the presence of sodium azide. Incubation of NK92 in azide-containing media had no effect on basal levels cell surface GHI/75 staining (Fig. 3C, left panel). Additionally, the presence of azide did not prevent the reduction in GHI/75 staining following W6/32 and DX17 treatment (Fig. 3C, right panels).

Figure 3.

W6/32 treatment decreases GHI/75 staining quickly and independent of ATP. (A) NK92 cells were incubated with W6/32 for the times indicated prior to staining with GHI/75. Control-treated cells were incubated for 180 min. Representative results for three experiments are shown. (B) MFI from samples stained and presented in (A). (C) NK92 cells were incubated with or without anti-MHC-I antibody for 30 min at 37°C in the presence or absence of azide prior to staining with GHI/75. Representative results from three experiments are shown.

Denaturation of surface MHC-I with citrate reduces GHI/75 staining

As an alternative to preincubation of NK92 with W6/32 antibody to disrupt cis interactions, we attempted to denature MHC-I on the cell surface by brief exposure to acidic buffer. Treatment of cells with low pH citric acid causes elution of peptides from MHC-I leading to denaturation of the surface complexes [27]. Therefore, we examined whether such denaturation of MHC-I on the cell surface would affect LIR-1 staining with various LIR-1 antibodies. Following incubation with citrate buffer, the loss of conformed MHC-I on the cell surface of NK92 was confirmed by reduced W6/32 staining4 (Fig. 4A, left panel). Coinciding with the loss of W6/32 staining on the cell surface, LIR-1 recognition by GHI/75 was also dramatically decreased following citrate exposure, consistent with the anti-MHC-I treatments (Fig. 4A, right panel). We also examined the effect of citrate treatment on LIR-1 staining with HP-F1 and two additional LIR-1 specific monoclonal antibodies, M405 and M402. While anti-MHC-I treatment was able to consistently and significantly increase HP-F1 staining on NK92, citrate treatment did not appear to enhance HP-F1 recognition of LIR-1. With the M405 antibody, we consistently observed reduced LIR-1 staining in repeat experiments, although the difference in the means did not reach statistical significance due to the variability between experiments. In assays using the M402 antibody, in contrast to M405 we observed slight yet consistent and significant increases in staining following acid exposure. In order to confirm that LIR-1 is not acid sensitive, we also treated 721.221 cells (LIR-1+ MHC-Ineg) with citrate buffer or PBS under the same conditions. The detection of LIR-1 on 721.221 cells was comparable between acid and PBS-treated cells with all antibodies tested and no statistically significant changes were observed despite reduced W6/32 staining (Fig. 4B). Therefore, the significant decrease in GHI/75 staining following acid exposure is likely due to the loss of native MHC-I at the cell surface and not to direct effects of acid treatment on LIR-1.

Figure 4.

Acid treatment modifies LIR-1 staining patterns on NK92 cells. (A) NK92 cells were treated with low pH citrate buffer for 3 min prior to staining with the indicated anti-LIR-1 antibodies or W6/32. The average MFI obtained from samples treated with PBS and citrate is shown as mean + SE of four samples pooled from four independent experiments, except for HP-F1 where three experiments were performed. (B) 721.221 cells were treated as in (A) and stained for LIR-1 and W6/32. Data are shown as mean + SE of three samples pooled from three experiments. Statistical significance determined by two-sample t-test assuming equal variance. (C) 721.221 cells expressing HA-tagged HLA-B27 were treated as in (A) and stained with the indicated antibodies. Data are shown as mean + SE of three samples pooled from three experiments. (D) Titration of GHI/75 on citrate-treated NK92 cells. (E) Titration of HP-F1 on citrate-treated NK92 cells. (F) The binding of various anti-LIR-1 antibodies at a concentration of 100 ng/mL to full-length LIR-1-Fc or D1D2-Fc was measured using an ELISA. (G) GHI/75, M405, and M402 binding to full-length LIR-1-Fc (solid lines) or D1D2-Fc (dashed lines) was measured by ELISA. The relative binding of the indicated antibodies at each concentration were calculated relative to a saturating amount of HP-F1. (D–G) The results shown are representative of three.

To further examine the effects of citrate buffer on cells, we treated 721.221 cells expressing HA-tagged HLA-B27 to allow us to track the denatured MHC-I following acid exposure. As seen with NK92 cells, citrate buffer treatment significantly reduced GHI/75 cell surface staining (Fig. 4C, left panel). To examine whether MHC-I was still present on the cell surface following citrate treatment, we compared the staining of anti-HA tag with W6/32. The staining with W6/32 was dramatically reduced following incubation in citrate buffer, as was seen with NK92 cells, but anti-HA tag staining was still largely present (Fig. 4C, middle panel). This indicates that following citrate treatment, the W6/32 epitope is denatured but MHC-I molecules are still present on the cell surface. For comparison, we also examined MHC-II, which is known for its acid resistance [28]. Interestingly, citrate treatment consistently enhanced the staining of MHC-II with the L243 antibody, although this increase was not statistically significant (Fig. 4C, right panel). On the other hand, surface proteins other than MHC-I can also be affected by citrate buffer treatment as seen for an epitope of LFA-1 (Fig. 4C, middle panel). Nonetheless, acid treatment of cells to denature MHC-I parallels treatment with W6/32, leading to a decrease in LIR-1 with the GHI/75 epitope.

Citrate treatment reduces LIR-1 accessibility to GHI/75

To examine whether the reduced detection of LIR-1 following citrate treatment was due to changes in antibody affinity or accessibility, we titrated both GHI/75 and HP-F1. LIR-1 antibody titrations were performed using purified GHI/75 and HP-F1 hybridoma supernatant followed by detection with fluorescently labeled secondary antibody and analysis by flow cytometry. While we again observed decreased LIR-1 intensity of staining with GHI/75 in citrate-treated samples compared with PBS treated, we interestingly observed a reduction at all antibody amounts tested (Fig. 4D). Even at saturating antibody concentrations, LIR-1 staining levels were still dramatically lower in acid-treated cells compared to control, suggesting that the change in LIR-1 conformation following MHC-I denaturation reduces the accessibility of GHI/75. When examining the HP-F1 titration, at high antibody amounts, we observed comparable levels of LIR-1 staining between both PBS- and acid-treated samples, consistent with our previous results suggesting that the HP-F1 epitope is less sensitive to interaction with MHC-I in cis (Fig. 4E). However, when staining with lower amounts of HP-F1, citrate-treated samples consistently stained higher for LIR-1 than PBS controls. This result is similar to what was observed with HP-F1 in our anti-MHC-I blocking assays. Therefore, while the loss of cis interactions with MHC-I directly limits the ability of GHI/75 to recognize LIR-1, HP-F1 likely requires free LIR-1 to bind.

The membrane proximal Ig-domains of LIR-1 are required for the GHI/75 epitope

In view of the differences in how the cis interaction affects recognition by the different LIR-1 antibodies, we wanted to investigate which LIR-1 extracellular domains are required for recognition by the various antibodies, and how the antibodies would behave in the absence of MHC-I. To assess this, we compared the binding of the four anti-LIR-1 antibodies with that of full-length LIR-1-Fc to a truncated LIR-1-Fc with only the ligand binding domains D1 and D2 using a capture ELISA detection method. As expected, HP-F1 bound comparably well to both the full length and D1D2-Fc fusion proteins (Fig. 4F). Interestingly, GHI/75, M405, and M402 demonstrated binding only to the full-length receptor, with no reactivity to D1D2, suggesting their epitopes are within the two membrane proximal Ig domains of LIR-1 (D3 and D4). Relative to HP-F1 binding, the hierarchy of binding to full-length LIR-1 is M405>GHI/75>M402 and aside from HP-F1, only M405 exhibited very weak binding to D1D2-Fc at high antibody doses (Fig. 4G). Therefore, all the antibody epitopes are present in LIR-1 in the absence of MHC-I, though only HP-F1 is able to recognize the D1D2 domains that bind MHC-I in trans.

Cis interaction with MHC-I restricts inhibition mediated by LIR-1

In order to assess whether the cis interaction of LIR-1 with MHC-I influences its inhibitory potential, we employed citrate-treated NK92 cells (LIR-1+ KIRneg) in a cytotoxicity assay with 721.221 cells and 721.221 cells expressing HLA-G as targets, and compared them with PBS-treated controls in the presence or absence of W6/32 to revert any inhibition through LIR-1. As expected, the expression of HLA-G on the target cells was sufficient to mediate inhibition of NK92 killing through LIR-1 (∼40% relative inhibition). This inhibition was further enhanced consistently to approximately 80% relative inhibition with citrate-treated NK92 cells. However, it should be noted that citrate treatment somewhat limited the maximal lysis of the HLA-G expressing cells in the presence of W6/32, and we therefore applied a correction by subtracting the relative inhibition in the presence of W6/32, bringing the relative inhibition to that shown in Figure 5A at ∼60%. Even with the harsh correction, it is clear that disruption of cis interactions between LIR-1 and MHC-I on NK cells allows for enhanced binding of its ligands on targets in trans. However, at lower E:Ts while the trend was the same, the differences were not statistically significant (data not shown).

Figure 5.

Disruption of LIR-1 associations with MHC-I in cis on NK cells enhances inhibition and UL18 binding in trans. (A) NK92 cells were treated with PBS or citrate buffer and used in a cytotoxicity assay against targets with or without HLA-G. NK92 cells were stained for W6/32 and GHI/75. Assays were performed in the presence of control antibody or W6/32 blocking antibody. The average relative inhibition for a E:T of 9:1 is shown as mean + SE for nine samples pooled from three experiments performed. (B) NK92 cells were incubated with W6/32 F(ab′)2 fragments prior to staining with GHI/75 (left) and UL18-Fc fusion protein (right). Representative results from three experiments are shown. (C) Summary of results presented in (B) comparing changes in MFI obtained with GHI/75 and UL18-Fc. Data are shown as mean + SE of three samples pooled from three experiments. (D) Antibody blocking of UL18-Fc binding. NK92 cells were preincubated for 15 min at room temperature with media alone, control antibody, or GHI/75 prior to staining as in (B). Antibody blocking was performed at a final concentration of 10 μg/mL. Data shown are representative of three experiments performed.

Cis interaction of LIR-1 and MHC-I restricts UL18 binding

Since the LIR-1 MHC-I cis interaction was able to directly modify LIR-1 recognition by “blocking antibodies” and influenced the degree of inhibition of cytolysis by HLA-G, we were interested in investigating its contribution to the binding of the viral MHC-I homologue and LIR-1 ligand UL18 when provided in trans, as it would be on an infected cell. To examine UL18 binding, we used an UL18-Fc fusion protein in a flow cytometry-based binding assay with NK92. LIR-1 cis interactions with MHC-I were disrupted on NK92 cells using W6/32 F(ab′)2 fragments to avoid any potential cross reactivity with the antihuman Fc secondary antibody used for UL18 detection. Incubation with W6/32 F(ab′)2 at 37°C resulted in a significant decrease in GHI/75 staining as well as a significant increase in UL18-Fc binding (Fig. 5B and C). Preincubation of NK92 cells with W6/32 F(ab′)2 had no effect on the binding of an Fc control protein (data not shown). Therefore, having LIR-1 engaged with MHC-I in cis on the surface of NK cells directly restricts the binding of its viral ligand UL18 in trans.

To confirm the ability of the GHI/75 antibody to “block” LIR-1 interactions in trans, we performed UL-18-Fc binding assays on NK92 cells in the presence of excess GHI/75 or an isotype control antibody. Consistent with previous reports of functional blocking of LIR-1, the binding of UL18-Fc was dramatically reduced in the presence of GHI/75 compared with control antibody (Fig. 5D). We also tested the ability of the M405 antibody to block UL18-Fc binding as this antibody binds to full-length LIR-1 but has minimal reactivity to D1D2, similar to GHI/75. However, in a single experiment, M405 was unable to affect UL18-Fc binding to LIR-1 relative to control (data not shown), suggesting it does not possess a similar blocking capacity for ligand interactions in trans.


Here we have presented evidence that the inhibitory receptor LIR-1 expressed on human NK cells is bound in cis with MHC-I molecules, and that this interaction modifies the binding of antibodies and ligands such as HCMV UL18 in trans and tempers the inhibition delivered in trans by HLA-G. While it has been reported that LIR-1 is capable of preventing NK-cell activation, the observation that the cis association with MHC-I can interfere with ligand recognition on target cells has implications on NK-cell activity during an immune response. Current models of NK-cell activation are largely based on the expression of ligands for NK-cell receptors on target cells [29]. In the “missing-self” model of NK-cell activation, it is the altered or loss of expression of MHC-I molecules on targets that renders them susceptible to NK lysis. In addition to this, the “induced self” model recognizes that enhanced expression of ligands for stimulatory receptors on target cells would also subject them to NK attack. However, given the growing evidence that some of these regulatory receptors are also subject to interactions within the same plane of the NK-cell membrane [22], it is becoming more clear that the density of expression of these ligands on the effector cells themselves is also a major contributor to setting the threshold for NK-cell activation.

This study is the first evidence that a receptor on human NK cells is regulated by a cis interaction with MHC-I. In mice, the inhibitory NK-cell receptor Ly49A constitutively associates with its MHC-I ligand H-2Dd in cis, and this interaction directly limits the number of Ly49A binding sites available for H-2Dd and H-2Dk in trans [23]. Furthermore, having the receptor bound in cis results in reduced inhibitory capacity through Ly49A, thereby lowering the threshold for activation. Interestingly, in the murine system the interactions of Ly49A and MHC-I occur with identical specificity in both cis and trans, implying that the region of Ly49A that makes contact is the same in either case. Moreover, Ly49 constructs with shortened stalks appear able to function only in trans, suggesting the extension by the stalk is required to reach the MHC-I in cis, and that the cis and trans interactions are mediated by two distinct conformations of the receptor [30]. The long extracellular stalk region provides Ly49A with the flexibility necessary to mediate interactions with MHC-I in cis. Therefore, structurally how LIR-1, which lacks a long stalk, might be able to fold toward the membrane in order to engage MHC-I on the same cell remains a question of interest. For the two membrane distal Ig domains to be able to engage the conserved region of MHC-I on the same membrane, LIR-1 could adopt a U-shaped “horseshoe” conformation where the receptor folds back upon itself, perhaps with the region between D2 and D3 acting as a hinge. This conformation would be comparable to the structure of the four N-terminal Ig domains of the Drosophila melanogaster protein Dscam, where a five-residue linker region between the second and third domains provides flexibility [31]. Whether or not the same sites on LIR-1 and MHC-I are used for both the cis and trans interactions remains unclear. We have shown that two antibodies that are broadly reactive with MHC-I, and block LIR-1 function in trans, also affect detection of the GHI/75 epitope. These results suggest that LIR-1 does bind to MHC-I in cis at a site similar to when it interacts in trans. As the two N-terminal Ig domains of LIR-1 possess the known ligand binding domains, perhaps the third and fourth domains are important for allowing the cis interactions in a manner similar to the stalk of Ly49. However, given that the decrease we observe in GHI/75 staining following MHC-I antibody blocking and low pH treatment is not proportional to the increase in staining with HP-F1, there may be differences in how this receptor engages its ligand in the two conformations.

It has been previously reported that the GHI/75 antibody is capable of reversing inhibition mediated through LIR-1 by target cells expressing HLA-G [32]. Consistent with this report, we observed that GHI/75 is also able to prevent trans interactions with UL18. As our data indicate that GHI/75 preferentially recognizes LIR-1 engaged in cis and does not interact with the two membrane distal Ig domains that mediate MHC-I binding, it is curious that it also possesses such functional activity, unless the manner in which LIR-1 recognizes MHC-I in trans is indeed different. However, alternative explanations could be that GHI/75 is able to prevent LIR-1 inhibition by promoting cis interactions with MHC-I, or locking the receptor in a “closed” conformation on the surface of NK cells, thereby impeding its ability to engage MHC-I in trans. Another interesting possibility raised by the results of our study is that IFN-driven changes in MHC-I levels could modulate the function of LIR-1 through the cis interaction. Whether the cis interaction with MHC-I is capable of signaling inhibition has not been directly examined. While lymphocytes typically have very high levels of MHC-I proteins as we show for the cell line NK92, monocytes and DCs, which also express LIR-1, upregulate MHC-I particularly during inflammatory responses when IFN is produced.

We observed that the cis association of LIR-1 and MHC-I directly restricts binding of the viral MHC-I homologue UL18 in trans. UL18 is expressed on the surface of HCMV infected cells to avert NK-cell responses. This suggests that the engagement of MHC-I molecules by LIR-1 in cis may provide for improved NK-cell responses during HCMV infection. This may be of particular importance considering the significantly higher affinity with which UL18 binds to LIR-1 compared with its endogenous ligand, and would therefore be able to outcompete MHC-I for binding. Interestingly, in addition to MHC-I and UL18, LIR-1 has also been reported to bind to a variety of bacterial species [33], however the role of these interactions during infection and whether this binding would also be sensitive to cis associations as well remains to be determined. It is important to note however that even in the presence of the cis interaction with MHC-I, binding to UL18 is limited but not absent. This indicates that only a proportion of the LIR-1 receptor molecules on the surface of NK cells are bound to MHC-I in cis and there remains free unbound receptor available to engage ligands in trans. Therefore, the possibility exists that in situations where MHC-I expression is upregulated, LIR-1 may be sequestered at the cell surface, limiting its ability to negatively regulate NK cells and thus enhancing sensitivity.

The contribution of cis interactions in the regulation of NK cells is an emerging topic of study. Inhibitory receptors bound in cis could potentially serve as a rheostat-like control mechanism to regulate the inhibitory input received by an NK cell in order to optimize activation efficiency. In mice it was reported that antibody-mediated sequestration of Ly49A, which mimics receptor sequestration by MHC-I in cis, can enhance the function of mature NK cells triggered via activating receptors [30]. During NK-cell development, a rheostat model has been proposed, where the inhibitory input an NK cell receives during education directly adjusts the efficiency of activation pathways in signaling responses [34, 35]. Additionally, the cis interaction of inhibitory receptors was demonstrated to significantly contribute to NK-cell development in mice, as a Ly49A variant that is unable to engage MHC-I in cis failed to educate NK cells, despite maintaining the ability to signal in trans [30]. However, it was reported that LIR-1 single positive NK cells did not exhibit enhanced missing-self responses relative to KIR positive subsets, and were functionally comparable to “receptor-null” cells, suggesting LIR-1 expression alone is not sufficient for education [36]. To date there is no evidence that KIR is capable of recognizing MHC-I in cis. This may be due to differences in the site of engagement with MHC-I, as KIR binds to the top portion of the molecule across the α1 and α2 domains, while LIR-1 and Ly49 bind to regions on MHC-I closer in proximity to the cell membrane [37].

In the present study, we have examined the contribution of the cis interaction between LIR-1 and MHC-I in cells that express uniform high levels of this receptor. While primary human monocytes and B cells express LIR-1 to a similar extent, LIR-1 expression on peripheral NK-cell populations is quite different. We have observed quite a large range of LIR-1 expression on donor NK-cell populations both in terms of the frequency and level of expression for both GH1/75 and HP-F1. Therefore, it is possible that the cis interaction influences trans NK-cell ligand recognition and binding in individuals expressing higher or lower levels of LIR-1 and with various polymorphisms [8]. Although the effect of D1D2 polymorphisms on HLA-class I binding in trans was reported to be minimal in a previous study [38], polymorphisms could alter the cis interaction. Gaining a better understanding of the mechanisms involved in regulating LIR-1 signaling and function will provide further insight into the susceptibility of individuals to infections such as HCMV.

Materials and methods

Cell lines

The human NK-cell line NK92 was obtained from ATCC (Manassas, VA, USA). NK92 cells were cultured in αMEM medium containing 12.5% characterized FBS (Thermo Fisher Scientific, Waltham, MA, USA), 12.5% horse serum (Invitrogen, Carlsbad, CA, USA), supplemented with 25 μM 2-mercaptoethanol and 1 mM L-glutamine (Invitrogen), and 100 U/mL human recombinant IL-2. The MHC-Ineg transformed B-cell line 721.221 cells were maintained in Iscoves medium with 10% FBS (Sigma, Oakville, ON, Canada) and 2 mM L-glutamine. The 721.221 stably transfected cell lines 221.B58, 221.Cw15, and 221.G [39] were obtained from Dr. Eric Long (National Institutes of Health, Rockville, MD, USA) and maintained in 0.5 mg/mL geneticin. For the generation of HA-tagged HLA-B27, HLA-B*2705 was amplified by RT-PCR from RNA extracted from cells stably expressing HLA-B27 (provided by Dr. Eric Long). The pDisplay signal sequence and HA tag were fused in frame to the beginning of the mature HLA-B27 sequence. The sequence was confirmed, the construct moved into pMX and transduced into 721.221 cells using the method previously described [40]. The cells were selected in 1 μg/mL puromycin and stable clones established by single cell sorting.

Antibodies, Fc fusion proteins, and flow cytometry

PE-Cy5 mouse anti-human CD85j (GHI/75), purified mouse antihuman CD85j (GHI/75), mouse antihuman HLA-ABC (DX17) were purchased from BD Biosciences (Mississauga, ON, Canada). Allophycocyanin mouse antihuman CD85j (HP-F1) was purchased from eBioscience (San Diego, CA, USA). For directly coupled antibodies, isotype-matched controls were also purchased from BD Biosciences (PE-Cy5 IgG2b) and eBioscience (allophycocyanin-IgG1). HP-F1 hybridoma supernatants were kindly provided by Dr. Miguel Lopez-Botet (University Pompeu Fabra, Barcelona, Spain). The mouse antihuman LIR-1 antibodies M405 and M402 were obtained from Amgen (Seattle, WA, USA). Mouse antihuman HLA class I (W6/32), mouse antihuman CD8 (51.1), mouse anti-HLA-DR (L243), and mouse antihuman LFA-1 (HB-203) were purified using protein-A affinity chromatography from hybridoma supernatants (ATCC). Anti-HA-tag was purchased from Cedarlane (Burlington, ON, Canada). W6/32 F(ab′)2 fragments were generated using the Pierce F(ab′)2 Preparation Kit (Thermo Fisher Scientific) [21]. The UL18-Fc fusion protein was produced according to previously published protocols [41]. Briefly, the UL18-Fc fusion protein was transiently expressed in 293FT cells and purified from the cell culture supernatant using a HiTrap™ Protein A HP column (GE Healthcare, Uppsala, Sweden) followed by reconstitution with a 10 M excess of recombinant human β2m was well as the actin-derived peptide ALPHAILRL. Properly folded UL18-Fc dimers were isolated by size exclusion chromatography on a Superdex 200 10/300 GL column (GE Healthcare) and their purity verified by SDS-gel electrophoresis. Full-length LIR-1-Fc was purchased from R&D Systems (Minneapolis, MN, USA). The D1D2 Ig domains of LIR-1 were amplified using the forward primer 5′-G CTA GCG GGG CAC CTC CCC AAG CCC ACC-3′ and reverse primer 5′-CGC TAG CCC TAG GAC CAG GAG CTC CAG GAG-3′. LIR-1 D1D2 was inserted into Cd51neg1 vector (provided by Dr. Eric Long) with NheI. LIR-1 D1D2-Fc and control Fc protein were affinity purified on Protein G Plus-Agarose (Calbiochem, San Diego, CA, USA) from serum-free supernatants of transfected COS-7 cells and dialyzed into PBS with Amicon centrifugal filters (Millipore, Billerica, MA, USA) essentially as described [42]. Purity was verified by SDS-PAGE and coomassie staining under nonreducing or reducing conditions. Protein concentrations were determined using the Micro Bicinchonic Acid assay (Pierce). PE goat antihuman IgG (Fc gamma-specific) was purchased from eBioscience for detection of Fc fusion protein binding. All samples were analyzed using a FACSCanto II or LSR II flow cytometer (BD Biosciences). Data analysis was performed using FACSDiva Software (BD Biosciences) and FlowJo (Tree Star Inc.).

Cell mixing assay

721.221 cells (1 × 107 cells) were labeled with 0.5 μL of 5 mM CFSE for 20 min at 37°C, quenched for 5 min at 37°C by addition of complete medium, and then washed twice in culture medium. CFSE-labeled 721.221 cells and unlabelled 221.B58 were cultured alone or in combination at a concentration of 1 × 105 cells/mL in culture media overnight. Cells were washed and stained with PE-Cy5 GHI/75. Cells were fixed and analyzed by flow cytometry.

MHC-I antibody blocking assay

NK92 cells were used from mid-log cultures 2–3 days after splitting. A total of 1 × 105 NK92 cells were incubated in a volume of 100 μL in the presence of purified W6/32 or DX17 or isotype-matched control mAb (10 μg/mL) or W6/32 F(ab′’)2 (20 μg/mL) or without antibody in their usual medium for a minimum of 30 min at 37°C and then washed on ice with cold FACS buffer (PBS/2% FBS/5 mM EDTA). Sodium azide was added to medium at a final concentration of 80 mM where indicated. Cells were then immediately stained with directly coupled anti-LIR-1 antibodies for 30 min on ice and in the dark. Cells were washed and fixed prior to analysis by flow cytometry.

Citrate treatment

NK92 or 721.221 cells were incubated on ice in citrate buffer (0.122 M citric acid and 0.066 M Na2HPO4, pH 3.2) or PBS for 3 min at a concentration of 1 × 106 cells/mL. Cells were then washed twice with an excess of PBS and once with FACS buffer, and stained with W6/32 to assess the loss of conformed surface MHC-I, and various anti-LIR-1 antibodies and analyzed by flow cytometry.


A capture-based ELISA was used to assess the binding of various antibodies to LIR-1. Plates were coated with 25 μg/mL goat antihuman IgG Fc in 0.1 M NaHCO3 (pH 9.6) at 4°C overnight, followed by 1 h at room temperature with LIR-1 Fc fusion proteins and washed. All washes were done three times with 0.05% Tween-20 in PBS. Samples were then incubated with HP-F1, GHI/75, M405, or M402 at room temperature for 1 h, washed, and detected with AP-conjugated F(ab′)2 goat antimouse IgG (1:10 000 dilution) and the PNPP substrate (Pierce).

Cytotoxicity assay

NK92 cells were treated with PBS or citrate buffer at room temperature for 3 min, washed twice in excess PBS, and resuspended in warm assay medium (Iscoves medium with 5% FBS and 2 mM L-glutamine). Cytolysis was measured in a standard chromium release assay. Target cells were labeled with 51Cr sodium chromate (NEN) for 1 h at 37°C with 5% CO2, washed three times in warm assay medium, plated with NK92 cells in triplicate at an effector to target ratio of 1:1, and incubated at 37°C with 5% CO2 for 4 h. NK92 cells were preincubated at room temperature with control (51.1) or blocking (W6/32) antibody at 20 μg/mL for 15 min prior to adding targets. Chromium release was quantified for 25 μL of supernatant and analyzed in a 1450 Microbeta Trilux (Wallac). 51Cr release was calculated as percent lysis using the formula: 100 × (specific 51Cr release – spontaneous release)/(maximum release – spontaneous release) for each sample. Relative inhibition was calculated as (100% – (%lysis 221-G 51.1/%lysis 221 51.1)) – (100% – (%lysis 221-G W6/32/%lysis 221 W6/32)).


The authors would like to thank Dr. Miguel Lopez-Botet for providing the HP-F1 hybridoma supernatants, Amgen for providing the M405 and M402 antibodies, and members of the Burshtyn and Stafford labs for helpful discussions. We also thank Kinola Williams for assistance with preparing the figures. This work was supported by awards from the Canadian Institutes of Health Research (MOP12357) and the Alberta Heritage Foundation for Medical Research awarded to DNB and by grants from the Swedish Research Council and the Swedish Cancer Society awarded to A.A.

Conflict of interest

The authors declare no financial or commercial conflict of interest.


leukocyte Ig-like receptor


killer cell Ig-like receptor


MHC class I