Random mutagenesis screen.

(A) Soluble HLA-G (sG) is internalized with HA-tagged KIR2DL4 into endosomes, as shown by incubation for 2 h at 37°C with anti-HA coupled to Alexa-594 and soluble HLA-G coupled to Alexa-647. The loading assay was performed with 293T cells transfected with HA-tagged KIR2DL4. (B) Confocal microscopy showing the location of internalized KIR2DL4 (red) and soluble HLA-G (green). An overlay is shown in the right panel (Merge). As control, HLA-G was replaced by HLA-E, also coupled to Alexa-647 (middle panels). 293T cells transfected with HA-tagged receptor 2B4 were also incubated with HLA-G coupled to Alexa-647 (bottom panels). 2B4 (CD244) resides at the plasma membrane. (C) The loading assay was used to screen KIR2DL4 mutants and test for HLA-G binding. The main parameter measured was presence or absence of HLA-G and location of detectable HLA-G. Confocal images representative of the five categories of mutants are shown. A DIC image is shown in the upper right panel. The parameters indicated under the images include HLA-G presence and location, location of KIR2DL4, the number of mutants in each of the five categories, and an example of a mutation for each category.

Cysteine mutants in the D0 domain of KIR2DL4 and their impact on the distribution of KIR2DL4 and its interaction with HLA-G.

V = vesicular; PM = Plasma membrane

A stressed disulfide bond typical of an allosteric bond in the crystal structure of KIR2DL4.

(A) The KIR2DL4 D0 domain has a third Cys residue not found in other D0 domains. Sequence alignment of D0 domains with Cys residues indicated by an arrow and shown in red (B) Solvent accessibility of disulfide bonds visible in crystal structures of the first Ig domain of KIRs was determined by DSSP (define secondary structure of proteins) algorithm. The first Ig domain of KIR2DL4 and KIR3DL1 is a D0 domain, whereas the first Ig domain of KIR2DL4 is a D1 domain. (C) Position of the three Cys residues in the crystal structure of KIR2DL4 (3WYR). Cys10, Cys28 and Cys74 are on the A strand, B strand and F strands, respectively, are as indicated. (D) The disulfide bond configuration in the KIR2DL4 D0 domain is an –RHstaple.

Detection and quantification of disulfide bonds in KIR2DL4 isolated from cells.

(A) Disulfide bonds in KIR2DL4. The Cys10-Cys28 disulfide bond and the unpaired Cys74 in the D0 domain and the stable Cys123-Cys172 disulfide bond in the D2 domain are shown in yellow in the ribbon structure of KIR2DL4. PDB identifier is 3WYR. (B) Quantification of disulfide-linked peptides in recombinant KIR2DL4 by mass spectrometry. Peptide abundance was expressed relative to the peptide linked by Cys123 and Cys172 in the D2 domain. (C) Protocol to determine the redox state of KIR2DL4 disulfide bonds, as measured by differential cysteine alkylation and mass spectrometry. Reduced disulfide bond cysteines in KIR2DL4 were alkylated with 12C-IPA and the oxidized cysteine thiols with 13C-IPA following reduction with DTT. The ratio of 12C-IPA to 13C-IPA alkylation represents the proportion of the disulfide bonds in the population that are in the reduced state. (D) Quantification of cysteine redox states in recombinant KIR2DL4 by differential cysteine alkylation and mass spectrometry. (E) Human cell-derived KIR2DL4 was found in both Cys10-Cys28 and Cys28-Cys74 redox states. The redox state of disulfide Cys123-Cys172 is shown for comparison. (F) Incubation of purified KIR2DL4 with 3 different protein disulfide isomerases. Reduction of the Cys10-Cys28 bond occurred with PDI at a PDI:KIR2DL4 molar ratio of 2 and 10. The redox state of cysteines were quantified by differential cysteine labelling and mass spectrometry. The error bars (SD) were derived from measurements of 2 to 4 peptides. ***<0.005 as assessed by unpaired, two-tailed Students t-test.

Location of KIR2DL4 and HLA-G after inhibition of PDI.

(A to D) Confocal microscopy images of 293T cells transfected with HA-KIR2DL4 (anti-HA coupled to Alexa 594, red) and loaded with labeled HLA-G (green) for 2 h in the presence of DTNB (A), pCMPS (B), Rutin (C) and anti-PDI antibody (D), at the indicated concentrations. The time 0 image in (A) and (B) is the same, and so is the time 0 image shown in (C) and (D). The distribution of KIR2DL4 in vesicles or on the cell surface was determined by scoring at least 200 cells per condition. (E) Confocal microscopy images of primary resting human NK cells loaded with labeled HLA-G (green) for 2 h in the presence of 1 mM DTNB or 30 μM Rutin. The number of HLA-G positive cells were scored in multiple fields (11-18 fields/condition). (F) Enlarged image of Fig. 4E (bottom left) with vesicular localization of HLA-G. (G) Quantification of data obtained from analyzing experiments using NK cells from 3 different donors.

Cysteine mutants in KIR2DL4 tested for their impact on KIR2DL4 distribution.

(A) Distribution (vesicles or plasma membrane) of WT and of the indicated cysteine mutants of KIR2DL4 in transfected 293T cells. (B) Distribution of HLA-G in 293T cells transfected with KIR2DL4 wild-type and with mutations at Cys10. HLA-G was scored for presence or absence in KIR2DL4+ vesicles. (C) Close-up view of the trio of cysteines in the D0 domain of KIR2DL4, as reported in the crystal structure (3WYR, left) and as predicted by AlphaFold (right). (D) Quantitative PCR analysis of IFI44L expression in response to HLA-G stimulation of 293T cells with stable expression of WT, C10S or C74S KIR2DL4. Baseline IFI44L transcription in untreated cells was set at 1. Data are represented as mean ± SD and are from 8 to 15 independent experiments depending on the cell line. ****p≤ 0.001; ns, p>0.05 by Wilcoxon test. (E) Reactivity of the antibodies anti-HA tag, mAb #33 and mAb 2238 with wild-type and the 3 cysteine mutants of KIR2DL4. (F) Flow cytometry analysis of primary NK cells pretreated with DTNB (1 mM) or Rutin (30 μM) for 1 h at 37°C and stained for surface expression of KIR2DL4 using mAb 2238 at 4°C in the presence of inhibitor for 30 min, followed by staining goat anti-mouse secondary antibodies coupled to APC at 4°C for 30 min. Data are represented as median fluorescence units and are from NK cells from 5 donors tested separately.

SPR affinity measurements.

(A) Representative sensorgrams from single dilution series of KIR2DL4, KIR2DL4 (C10L) and KIR3DL1 binding to HLA-GVPLH (top panel) and interaction analysis of this binding (bottom panel). (B) Analysis of KIR2DL4 and KIR2DL4(C10L) interaction with non-binding epitopes. Error bars denote standard deviation. Equilibrium dissociation constants KD were determined using a 1:1 binding model from two independent experiments performed in duplicate.

Predicted outcome of an allosteric disulfide switch on the conformation of KIR2DL4 and its predicted interaction with HLA-G.

(A) Overlay of the crystal structure (3WYR; purple) and the structure predicted by AlphaFold DB (blue). The allosteric disulfide Cys10-Cys28 is in yellow and the Cys28-Cys74 structural disulfide is in green. The region indicated with a blue arrow corresponds to the loop highlighted in B. (B) D0 domain shown with the Pro46-Pro48 loop facing up and pointing toward the predicted HLA-G binding site. The position of Val45 and Pro46 in the crystal structure (C10-C28; yellow) is overlayed with that in the AlphaFold-predicted structure (C28-C74; green). The distance between Val45 and Pro46 in the crystal structure and their counterpart in the predicted structure of a Cys28-Cys74 KIR2DL4 is 5.3 and 5.4 Å, respectively. (C) Docking of the KIR2DL4 crystal structure (3WYR, pink) with HLA-G (blue), as predicted by AlphaFold 3. (D) Docking of the predicted KIR2DL4 structure by AlphaFold 3 (green) with HLA-G (blue). In the Cys28-Cys74 KIR2DL4 bonded form predicted by AlphaFold for wild-type KIR2DL4, the Val45-Pro48 loop appears to be in a different orientation, suggesting potential contacts with Met76 and Glu19 of HLA-G.

Random mutagenesis screen.

(A) Strategy for the generation of mutants. (B) Examination of mutants by confocal microscopy. WT = wild-type. (C) In the same assay as in Figure 1C, mutants C10R and P48S were tested in the absence of soluble HLA-G (sG). Confocal images of anti-HA-Alexa-594 revealed their location. A DIC image is shown in the right panel. PM = plasma membrane. (D) Flow chart of mutants distributed among 4 categories distinct from WT. Several mutants from each category were also tested in the absence of soluble HLA-G (sG). HLA-G often dictates the location of KIR2DL4. For example, C10R internalization depended on HLA-G, and P48S was not internalized into vesicles in the presence of soluble HLA-G.

Analysis of disulfide-linked peptides in recombinant KIR2DL4 protein by mass spectrometry.

Tandem mass spectra of the peptides linked by disulfide bond between Cys10 and Cys28 and between Cys28 and Cys74 are shown. The accurate mass of Cys10-Cys28 and Cys28-C74 peptides are shown in the insets (Cys10-Cys28: [M + 3H]3+ = m/z 762.03039 and expected [M + 3H]3+ = m/z 762.0299; Cys28-Cys74 : [M + 3H]3+ = m/z 405.18726 and expected [M + 3H]3+ = m/z 405.1870).

Flow cytometry profiles of 293T cells stably expressing WT, C10S or C74S mutants of HA-tagged KIR2DL4.

Cells were stained with anti-HA tag antibody coupled to Alexa 488. Median fluorescence intensity (MFI) for the red profiles is indicated in the upper right corner.

Flow cytometry profiles of 293T cells expressing HA-tagged KIR2DL4 WT and, in parallel, the 3 Cys to Ser mutants loaded for 2 h at 37°C with an antibody to the HA-tag and two different antibodies to KIR2DL4.

Staining with the control anti-HA antibody coupled with Alexa 488 is shown on the x axis versus no antibody (top), whereas mAb 2238-Alexa 647 (center) and mAb #33-APC (bottom) are on the y axis. The lower left quadrant was set to include 98% of unstained control cells.

Flow cytometry profiles of primary NK cells from two different donors stained at the cell surface with mAb #33 (blue profiles) or mAb 2238 (red profiles) followed by APC-coupled secondary antibodies.

Control staining with secondary antibodies alone is also shown (black profiles).

Size exclusion chromatography (SEC) profiles.

Overlay of size exclusion profile with elution volume of purified KIR2DL4 wt (blue), KIR2DL4 C10L (orange) and KIR2DL4 C74S (green).

Potential interaction of KIR2DL4 with HLA-G, as predicted by AlphaFold (A) Docking of the C74R KIR2DL4 (beige) with HLA-G (blue), as predicted by AlphaFold 3.

The structure of KIR2DL4 in this Cys10-Cys28 configuration is similar to that of the crystal structure (Fig. 7C). (B) C10R KIR2DL4 (orange) with HLA-G (blue). The predicted structure of this Cys28-Cys74 KIR2DL4, with the Val45-Pro48 loop closer to HLA-G, resembles that of WT KIR2DL4 (Fig. 7D). (C) Predicted folding of the D2 domain in the four KIR2DL4 isoforms (3WYR crystal structure of WT, and AlphaFold predictions of WT, C74R, and C10R) examined here, indicate a conserved D2 domain folding (shown as an overlay). Several predicted contacts of KIR2DL4 with HLA-G (blue) are similar to those of KIR3DL1 D2 domain with HLA-B, such as Ser128–Arg145 (3.1 Å, left), Asp130–Arg145 (2.7 Å, center), and Ser179–Lys146 (2.7 Å, right). As KIR2DL4 lacks the 100 amino acid-long D1 domain of KIR3DL1, the Ser128, Asp130, and Ser179 of KIR2DL4 correspond to Ser228, Asp230, and Ser279 of KIR3DL1.

KIR2DL4 peptides analysed to determine disulfide bond redox state.

Peptides were detected by Byonic analysis software, confirmed by MS/MS and have errors <6 ppm. Only peptides with peak areas >10 million for a given Cys were included in the analysis.

Summary of SPR measurements