KIR2DS2 recognizes conserved peptides derived from viral helicases in the context of HLA-C

See allHide authors and affiliations

Science Immunology  15 Sep 2017:
Vol. 2, Issue 15, eaal5296
DOI: 10.1126/sciimmunol.aal5296

Killing viral helicases

Recognition of evolutionarily conserved pathogen-associated molecules drives innate immune responses. Naiyer et al. report that a killer cell immunoglobulin-like receptor (KIR), KIR2DS2, promotes activation of natural killer (NK) cells by recognizing conserved peptides from flaviviral RNA helicases when presented by a particular human leukocyte antigen (HLA) allele, HLA-C*0102. They have identified two distinct peptide motifs—LNPSVAATL and MCHAT—that are sensed by KIR2DS2. The former is conserved across hepatitis C virus isolates; the latter is conserved in a number of flaviviruses including dengue, Zika, and yellow fever viruses. The study illustrates that a single KIR receptor has evolved to activate NK cells in response to multiple pathogenic viruses.


Killer cell immunoglobulin-like receptors (KIRs) are rapidly evolving species-specific natural killer (NK) cell receptors associated with protection against multiple different human viral infections. We report that the activating receptor KIR2DS2 directly recognizes viral peptides derived from conserved regions of flaviviral superfamily 2 RNA helicases in the context of major histocompatibility complex class I. We started by documenting that peptide LNPSVAATL from the hepatitis C virus (HCV) helicase binds HLA-C*0102, leading to NK cell activation through engagement of KIR2DS2. Although this region is highly conserved across HCV isolates, the sequence is not present in other flaviviral helicases. Embarking on a search for a conserved target of KIR2DS2, we show that HLA-C*0102 presents a different highly conserved peptide from the helicase motif 1b region of related flaviviruses, including dengue, Zika, yellow fever, and Japanese encephalitis viruses, to KIR2DS2. In contrast to LNPSVAATL from HCV, these flaviviral peptides all contain an “MCHAT” motif, which is present in 61 of 63 flaviviruses. Despite the difference in the peptide sequences, we show that KIR2DS2 recognizes endogenously presented helicase peptides and that KIR2DS2 is sufficient to inhibit HCV and dengue virus replication in the context of HLA-C*0102. Targeting short, but highly conserved, viral peptides provide nonrearranging innate immune receptors with an efficient mechanism to recognize multiple, highly variable, pathogenic RNA viruses.


Natural killer (NK) cells are critical determinants in the innate immune response to a number of globally important viral infections including HIV, hepatitis C virus (HCV), dengue virus, and Chikungunya virus (14). Their functions are controlled by nonrearranging immune receptors, many of which are well conserved, consistent with a role in innate immunity. The killer cell immunoglobulin-like receptors (KIRs) are a rapidly evolving family of NK cell receptors that exhibit substantial population diversity. They have peptide/MHC (major histocompatibility complex) class I specificity, and certain combinations of KIRs and their MHC class I ligands have been associated with the outcome of a number of viral infections including the inhibitory receptor KIR2DL3 and hepatitis C, and the activating receptor KIR3DS1 and HIV (4, 5).

The KIRs have evolved rapidly since the divergence of humans from chimpanzees, in a manner indicative of pathogen-mediated selection. One mechanism by which KIRs contribute to recognition of pathogens has been identified for inhibitory KIRs being related to the down-regulation of MHC class I (6). Conversely, there is little understanding of how activating KIRs can recognize pathogens. Recent work has shown that KIR3DS1, which shares sequence homology with human leukocyte antigen, class B (HLA-B)–specific inhibitory KIR3DL1, binds open conformers of HLA-F, and this may result in inhibition of HIV infection (7). However, the mechanisms by which the activating KIRs with putative HLA-C ligands recognize pathogens remain enigmatic (8).

Human KIRs bind subsets of HLA class I ligands and engage both the HLA class I heavy chain and the bound peptide. Furthermore, changes in the peptide content of HLA class I can abrogate KIR-mediated inhibition of NK cells, demonstrating a mechanism for KIR-associated NK cell activation that is distinct from HLA class I down-regulation (9). KIR2DL2 and KIR2DL3 are inhibitory KIRs with a predominant specificity for the group 1 HLA-C allotypes (HLA-C1). Both these inhibitory receptors have similar specificities for peptide/MHC, as could be predicted from the high sequence homology of their extracellular domains (10). The activating receptor KIR2DS2 also has ~98% amino acid identity in the ligand-binding domains to KIR2DL2 and KIR2DL3 and is in strong linkage disequilibrium with KIR2DL2. Despite this shared sequence homology, binding of KIR2DS2 to HLA-C has been difficult to identify. One of the key differences between KIR2DS2 and KIR2DL2/KIR2DL3 is the presence of a tyrosine (as compared with a phenylalanine) residue at position 45, and this is thought to substantially affect binding of KIR2DS2 to HLA-C (11). Low-level binding of KIR2DS2 to HLA-C in combination with an Epstein-Barr virus peptide was detected by surface plasmon resonance (12). More recently, KIR2DS2 has been demonstrated to have peptide-specific binding to HLA-A*1101, an HLA class I allele that binds a number of diverse KIRs including KIR2DS4 and KIR3DL2 (13, 14). KIR2DS2 has also been shown to bind cancer cell lines in a β2-microglobulin–independent manner (15). Nonetheless, on the basis of the high sequence homology of KIR2DS2 with KIR2DL2 and KIR2DL3, we sought to determine whether KIR2DS2 was also an HLA-C–restricted peptide-specific receptor.


KIR2DS2 recognizes the HCV peptide LNPSVAATL in the context of HLA-C*0102

Immunogenetic analyses by Thio et al. have identified HLA-C*0102 to be protective in the context of chronic HCV infection (16). On the basis of this observation, we used HLA-C*0102 as a template allele to identify peptides from HCV to determine how they may modulate NK cell function. We screened the HCV genotype 1b genome (AF313916.1) for potential HLA-C*0102 binding peptides using the algorithms ADT, NetMHCpan, and KISS, coupled with a manual scan using the known HLA-C*0102 motif xxPxxxxxL (17). We identified eight peptides for further study (table S1). We used the transporter associated with antigen processing (TAP)–deficient 721.174 cell line, which expresses HLA-C*0102 naturally, to present the HCV peptides in binding and functional assays. These cells are deficient in the TAP protein and hence can be used to present exogenous peptide to lymphocytes (18). Analysis of binding to HLA-C*0102 using peptide titrations demonstrated that four peptides (LLPRRGPRL, AQPGYPWPL, LSPHYKVFL, and RAYLNTPGL) had a low affinity for HLA-C*0102, not reaching saturation point at 100 μM; one [VLPCSFTTL (VLP)] had an affinity comparable to the endogenously presented peptide VAPWNSLSL, and the remainder [LSPRPVSYL, LNPSVAATL (LNP), and ARPDYNPPL] had an intermediate affinity for HLA-C*0102 (fig. S1).

To determine how these peptides modulate NK cell reactivity, we used the 721.174 cells. The CD158b epitope includes the KIR2DL2, KIR2DL3, and KIR2DS2 proteins and therefore marks NK cells with specificity for HLA-C*0102. We identified one peptide (LLPRRGPRL) that inhibited the activation of CD158b-positive (CD158b+) NK cells, but the remainder of the peptides had no effect on NK cell degranulation relative to the “no peptide” control (fig. S2). In the absence of peptide, there is activation of CD158b+ NK cells by 721.174 such that 25 to 30% NK cells express CD107a (Fig. 1, A and B). To identify specific activation of KIR2DS2-positive (KIR2DS2+) NK cells, we reduced the background activation of CD158b+ NK cells by using an inhibitory peptide (VAPWNSFAL) and testing the ability of the peptides to activate NK cells relative to this strong inhibitory signal. We used VAPWNSFAL because it was the strongest inhibitory peptide that we identified in a peptide screening assay for KIR binding and NK cell inhibition (9). From an initial screening assay using three unselected donors, we identified two peptides, LNP and VLP, which activated CD158b+ NK cells relative to VAPWNSFAL alone (Fig. 1A).

Fig. 1 KIR2DS2 recognizes the HCV peptide LNP.

(A) Peripheral blood mononuclear cells (PBMCs) were stimulated overnight with interleukin-15 (IL-15) and incubated for 4 hours with 721.174 cells that had been loaded with VAPWNSFAL or with VAPWNSFAL plus the indicated HCV peptides at a concentration of 5 μM. The mean CD107a expression plus 1 SD on CD3CD56+CD158b+ NK cells from three unselected donors is shown. (B to D) CD107a degranulation assays of NK cells from KIR2DL2+/KIR2DS2+ and KIR2DL3+/KIR2DS2 donors in response to 721.174 cells incubated with indicated peptides. Peptides were used at 5 μM each. (B) Flow cytometry histogram plots of CD107a expression gated on CD3CD56+CD158b+ NK cells from two donors of different KIR genotype. The results from five donors of each genotype are summarized for CD158b+ NK cells in (C) and for CD158b NK cells in (D). Means and SEs for each condition are shown. P values indicate comparison to VAPWNSFAL alone. ns, not significant. (E and F) NKL cells either untransfected or transfected with KIR2DL2 or KIR2DS2 were incubated with 721.221 cells transfected with HLA-C*0102 or with HLA-C*0102 + LNPSVAATL (C*0102-LNP), and the cytotoxicity of the 721.221 cells was measured by flow cytometry. (E) Flow cytometry histogram plots of the uptake of the LIVE/DEAD Fixable Aqua Dead dye by the target cells incubated with the indicated effector cells. The means and SEs from three independent cytotoxicity experiments are shown in (F). Effector-to-target (E/T) ratios are also shown. P values indicate the comparison of cytotoxicity between the HLA-C*0102 and the HLA-C*0102–LNP targets. Where shown, P values were determined by independent two-tailed t tests (*P < 0.05, **P < 0.01, ****P < 0.0001).

Because CD158b marks the inhibitory receptors KIR2DL2 and KIR2DL3 and the activating receptor KIR2DS2, we stratified the donors by KIR genotype: KIR2DL2+KIR2DS2+KIR2DL3 (KIR2DL2/S2 homozygous) and KIR2DL2KIR2DS2KIR2DL3+ (KIR2DL3 homozygous). VLP activated CD158b+ NK cells relative to VAPWNSFAL alone in both KIR2DL2+/S2+/L3 donors and KIR2DL2/S2/L3+ donors (Fig. 1B). In contrast, the effect of LNP on CD158b+ NK cells was present for KIR2DL2+/S2+/L3 donors but not for KIR2DL2/S2/L3+ donors (Fig. 1, B and C). Conversely, no differences were found for the CD158b-negative (CD158b) NK cells for either group of donors (Fig. 1D). This indicates that the increase in degranulation was related either to KIR2DL2 or to KIR2DS2. KIR2DL2+ NK cells are less responsive to changes in peptide content of MHC class I than KIR2DL3+ NK cells (19). Therefore, we reasoned that this change in degranulation was related to NK cell activation through engagement of KIR2DS2.

To map this specificity more closely, we cloned NK cells from a KIR2DL2+KIR2DS2+KIR2DL3 donor. We tested three KIR2DS2+ NK cell clones against the 721.174 cells (fig. S3). We found that although LNP augmented lysis of 721.174 cells by KIR2DS2+ NK cell clones, background levels of killing against targets in the absence of peptide were high. We therefore sought an alternative strategy using the KIR-negative cell line NKL. We transfected NKL cells with either KIR2DS2 or KIR2DL2 and used them as effector cells. We also generated 721.221 target cells stably expressing either HLA-C*0102 alone or HLA-C*0102 and LNP together. In cytotoxicity assays, NKL-2DS2 cells lysed 721.221:C*0102 cells to a level similar to that of untransfected NKL cells but lysed LNP-expressing 721.221:C*0102 cells significantly better than 721.221:C*0102 cells without LNP (Fig. 1, E and F). Thus, LNP specifically activates NKL cells expressing KIR2DS2 but has no effect on KIR2DL2+ NK cells. As there were similar levels of killing of 721.221:C*0102 cells by NKL-2DS2 and NKL cells, this observed increased level of lysis is not related to a generalized increase in baseline cytolytic activity of the NKL-2DS2 cell line against 721.221 cells but is related to the specific recognition of the combination of LNP and HLA-C*0102 by KIR2DS2.

LNP engages KIR2DS2 but not KIR2DL2

To confirm the reactivity of KIR2DS2 with HLA-C*0102:LNP, we made a KIR2DS2 tetramer. In a binding assay, we detected low-level but consistent and significant binding of the KIR2DS2 tetramer to the peptide LNP (P < 0.001), as compared with the naturally eluted peptide VAPWNSLSL, as well as KIR2DS2 binding to the peptides VAPWNSATL and VAPWNAATL (fig. S4). This is in line with previous observations that positions 7 and 8 of the HLA class I peptide are important for KIR binding, as has been observed for the inhibitory KIR. Because the binding of the tetramer was at a low level (~1.5 times background), we used a clustering assay to determine whether a cell-cell contact assay could detect specific recognition of LNP by KIR2DS2. Transfectants of the Ba/F3 cell line expressing either KIR2DL2 or KIR2DS2 were used to investigate receptor clustering (20). These are murine B cells and hence do not express other human NK cell–activating receptors. The Ba/F3 transfectants were co-incubated with 721.174 cells pulsed with peptide. LNP induced clustering of KIR2DS2, but not of KIR2DL2, whereas the inhibitory peptide VAPWNSFAL induced clustering of KIR2DL2 only (Fig. 2, A and B).

Fig. 2 LNP activates KIR2DS2+, but not KIR2DL2+, NK cells.

(A and B) Ba/F3 cells expressing KIR2DS2 or KIR2DL2 were incubated with 721.174 cells loaded with the indicated peptides. Cells were fixed and stained with anti-CD158b before analysis by confocal microscopy (A). Peptides were used at a concentration of 100 μM. Arrowheads indicate clustering at the interface between the Ba/F3 and 721.174 cells. The intensity of staining of the Ba/F3 cells at the interface was compared with the membrane at a noncontact area and plotted as the fold increase above background (B). The results from three independent experiments with a total of 30 conjugates per condition are shown. The P value in comparison to 2DS2 with no peptide is shown. (C) NKL-2DS2 cells were incubated with 721.174 cells loaded with the indicated peptides, and DAP12 was coimmunoprecipitated with anti-2DS2 antibody from the cell lysates before Western blotting (WB) with anti-phosphotyrosine (pTyr) or anti-DAP12. DAP12 was also immunoprecipitated from NKL-2DS2 cells in the absence of 721.174 target cells (no target). Densitometry results of the pDAP12/DAP12 ratio from three independent experiments and P values in comparison to no peptide are shown. IP, immunoprecipitation. (D) NKL-2DS2 or NKL-2DL2 cells were incubated with 721.174 cells loaded with the indicated peptides and Western blotting for phospho-Vav1 (pVAv1) or Vav1 performed. “No target” lanes indicate immunoprecipitation from NKL-2DS2 or NKL-2DL2 cells alone. Densitometry results of the phospho-Vav1/Vav1 ratio from three independent experiments and P values in comparison to no peptide are shown. (E) 721.174 cells were cultured with 100 μM peptide (VAWPNSLSL or LNP) overnight, stained with the KIR2DL2-Fc fusion construct, and analyzed by flow cytometry. The histogram plots for the two peptides (filled histograms) and the median fluorescence of KIR2DL2-Fc, compared with no peptide (open histograms), are shown. For all panels, P values were derived using one-way analysis of variance (ANOVA), with Dunnett’s test for multiple comparisons (*P < 0.05, ****P < 0.0001).

To further determine whether the interaction of LNP with KIR2DS2 led to NK cell activation, we studied signaling in the NKL cell lines. Activating KIRs transduce positive signals through the adapter molecule DNAX-activation protein of 12 kDa (DAP12), and antibody cross-linking of KIR2DS2 has been shown to lead to DAP12 phosphorylation (20). We therefore tested whether LNP could lead to activation of NK cells via DAP12. Immunoprecipitation of KIR2DS2 from NKL-2DS2 cells cocultured with 721.174 cells in the presence or absence of peptide demonstrated LNP-induced enhanced tyrosine phosphorylation of the coimmunoprecipitated DAP12, consistent with NK cell activation (Fig. 2C). Moreover, the presence of LNP augmented tyrosine phosphorylation of the downstream guanine nucleotide exchange factor Vav1 in NKL-2DS2 cells but had no effect on KIR2DL2 signaling in NKL-2DL2 cells (Fig. 2D). We further tested whether KIR2DL2 could bind LNP in the context of C*0102 using a KIR2DL2-Fc fusion construct. Consistent with the functional experiments, there was no direct binding of a KIR2DL2 fusion construct to LNP-pulsed 721.174 cells (Fig. 2E). Together, these data demonstrate that LNP specifically activates KIR2DS2+ NK cells through the DAP12-Vav1 pathway but has no effect on KIR2DL2+ NK cells, establishing it as a peptide that discriminates an activating KIR from its closely related inhibitory counterpart.

KIR2DS2 recognizes LNP in the context of HCV infection

Having observed that KIR2DS2 recognized LNP in the context of HLA-C*0102 as both an exogenously and an endogenously loaded peptide, we wanted to test whether it could also recognize LNP in the context of HCV replication. To do this, we generated an HUH7 cell line that stably expresses both the HCV subgenomic replicon N17 (JFH1ΔE1E2-luc) and also HLA-C*0102. N17 contains the full-length HCV genome without the envelope proteins and replicates as a wild-type virus; hence, it is considered that it produces proteins to a level similar to that of a wild-type virus (21). It also has a luciferase gene incorporated within the HCV construct to allow assessment of replication by a luciferase assay. HUH7 is a hepatoma cell line that expresses HLA-A*1101, but not HLA-B or HLA-C, and is permissive for HCV replication (22). We used the transfected NKL cell lines to assess the potential for KIR2DS2 to specifically inhibit replication of N17 in this system. NKL-2DS2 inhibited HCV replication at low E/T ratios (0.01:1) in the HLA-C*0102–expressing HUH7 cells to a significantly greater extent than parental NKL cells or NKL-2DL2 cells (Fig. 3A, right). Conversely, in the parental HUH7 cell line, which is HLA-C–negative, NKL, NKL-2DL2, and NKL-2DS2 inhibited HCV replication to similar levels (Fig. 3A), suggesting that KIR2DS2 requires the presence of HLA-C to inhibit HCV replication.

Fig. 3 Endogenously presented LNP is recognized by KIR2DS2 in the context of HCV.

(A) HUH7 cells expressing the N17 (JFH1ΔE1E2-luc) replicon with (right) or without (left) HLA-C*0102 were co-incubated with NKL, NKL-2DL2, or NKL-2DS2 cells at the indicated E/T ratios. Luciferase activity from the cocultures was measured and normalized to expression in the absence of NKL cells. The relative inhibition of replication was then measured. The means and SEs of three experiments performed in duplicate are shown. Comparisons for NKL-2DS2 with NKL-2DL2 are indicated. (B) Replication of wild-type (WT) N17 (JFH1ΔE1E2-luc) replicon RNA or variants in which the LNP epitope had been mutated to aspartate at residues 3, 7, 8, and 9 (P3D, A7D, T8D, and L9D, respectively). As a nonreplicating control (Control), the N17 replicon carrying the lethal GND mutation in the viral NS5B protein was used. RLU, relative light units. (C) NKL, NKL-2DL2, or NKL-2DS2 cells were cocultured with HUH7-C*0102 cells expressing either wild-type or the L9D mutant HCV replicon. Luciferase activity was measured and plotted as percentage inhibition of viral RNA replication compared with HUH7 cells incubated in the absence of NKL cells. The means and SEs of three experiments performed in duplicate and P values for comparisons between NKL-2DS2 and NKL-2DL2 are shown. (D) NKL, NKL-2DL2, or NKL-2DS2 cells were cocultured with HUH7-C*0304 cells expressing either wild-type (left) or the L9D mutant HCV replicon (right). The percentage inhibition of replication compared with the no NKL control is shown. Statistical analyses for (A), (C), and (D) were performed using independent two-tailed t tests to compare NKL-2DS2 and NKL-2DL2. (E and F) 721:C*0304:ICP47 cells were loaded with the indicated peptides at saturating concentrations (100 μM) and then stained for HLA-C expression using the DT9 antibody (E) or the KIR2DS2–phycoerythrin (PE) tetramer (F). One representative histogram plot from three independent experiments is shown for each peptide, and the median fluorescence intensity of staining was indicated. Dark lines indicate peptide staining compared with the no peptide control (gray lines). (G) Western blot for phospho-Vav1 and Vav1 from 721.221:C*0304:ICP47 cells cultured with 20 μM of the indicated peptide and co-incubated with NKL-2DS2 cells at an E/T ratio of 1:1. One representative blot is shown together with densitometry of the phospho-Vav1/Vav1 ratio (mean ± SD) from three independent experiments. Statistical analysis was performed using one-way ANOVA, with Dunnett’s test for multiple comparisons (*P < 0.05, **P < 0.01, ***P < 0.001).

To investigate the role of the LNP peptide in the inhibition of HCV replication by KIR2DS2, we made nonconservative aspartate variants of the N17 replicon at the HLA-binding anchor residues P3 and P9 and the putative KIR binding residues P7 and P8 (17). Mutation of P9 retained HCV replication to wild-type levels, but mutation of the P3, P7, or P8 residue in this epitope profoundly inhibited replication of the N17 replicon viral RNA to a level similar to that of the nonreplicating control, which has a lethal “GND” polymerase mutation (Fig. 3B) (23). We therefore performed inhibition of replication assays using only the P9 leucine-to-aspartate N17 variant (L9D). In these assays, there was no specific effect of KIR2DS2 noted (Fig. 3C), suggesting that KIR2DS2 recognition of HCV in the context of HLA-C*0102 is dependent on the LNP epitope.

To test the HLA-C specificity of this response, we transfected HUH7 cells, expressing either the N17 replicon or the mutated N17-L9D replicon, with HLA-C*0304. NKL-2DS2 inhibited HCV replication to a significantly greater extent than NKL-2DL2 or the parental NKL cell lines at all E/T ratios tested (Fig. 3D). However, this was not related to the LNP peptide because NKL-2DS2 also inhibited replication of the N17-L9D replicon to a similar extent. LNP was unable to stabilize HLA-C*0304 in a peptide stabilization assay using 721.221:HLA-C*0304 cells transfected with ICP47, which blocks TAP function (Fig. 3E). Furthermore, using the same cell lines to present peptide, LNP did not engage KIR2DS2 in the context of HLA-C*0304 in tetramer binding or Vav1 phosphorylation assays (Fig. 3, F and G). To generate a positive control peptide for this experiment, we tested P7 and P8 derivatives of the naturally processed HLA-C*0304–binding peptide, GAVDPLLAL, guided by our current work and also that of Liu et al. (13, 24). GAVDPLLAL, GAVDPLAWL, and GAVDPLATL all stabilized HLA-C*0304, but only GAVDPLAWL induced binding of the KIR2DS2 tetramer to HLA-C*0304 and Vav1 phosphorylation (Fig. 3, E to G). Thus, KIR2DS2 has a broad peptide/HLA-C specificity, and it is likely that there are other HCV peptides that can be presented by HLA-C*0304 and recognized by KIR2DS2.

LNP is a highly conserved peptide

HCV is an RNA virus with high sequence variability due to a lack of a proofreading capability of its RNA polymerase. We therefore assessed whether there was a clinical correlate to the defective replication of replicons with LNP variants. We observed that the LNP epitope is highly conserved across HCV genotypes, which, in general, differ by up to 30% in sequence identity (table S2). Genotype 5 HCV, the only genotype with a consistent variation of this epitope, has a leucine-to-phenylalanine variant at P9, consistent with our finding in the replication assays that variation at P9 of the LNP epitope does not affect HCV replication. Furthermore, the LNP epitope is conserved across 878 (98.5%) of the 891 HCV NS3 sequences deposited in the Los Alamos HCV database ( Thus, a lack of natural variation in this region implies that, in vivo, mutation of LNP is relatively poorly tolerated and marks it as a viable target for a nonrearranging immune receptor. Furthermore, structural analysis demonstrates that LNP is located within the RNA helicase domain of the HCV NS3 protein in the Ia motif and that the LNP peptide is in the RNA binding region of this motif, consistent with its relative lack of variation (25). Thus, it is in a critical functional region of HCV.

To investigate whether KIR2DS2 was associated with the outcome of HCV infection on a genetic basis, we performed logistic regression analysis of putative activating KIR/HLA combinations (KIR2DS1:HLA-C2, KIR2DS2:HLA-A*11, KIR2DS2:HLA-C1, and KIR3DS1:HLA-BBw4) and control activating KIR/HLA combinations (KIR2DS1:HLA-C1 and KIR2DS2:HLA-C2), together with key protective (KIR2DL3:HLA-C1 homozygosity) and susceptibility (KIR2DS3) factors in our UK Caucasian population with resolved (n = 120) or chronic (n = 216) HCV infection (4). KIR2DS2, in combination with HLA-C1, was protective against chronic HCV infection [P = 0.033; odds ratio (OR), 2.06; 95% confidence interval (CI), 1.06 to 4.02], but no association was found between KIR2DS2 in combination with HLA-A*11 and the outcome of HCV infection (Table 1). We also noted that KIR3DS1 in combination with HLA-BBw4 was protective, but none of the control combinations were associated with the outcome of HCV infection. Thus, in HCV infection, group 1 HLA-C alleles are protective in combination with both KIR A (KIR2DL3) and KIR B (KIR2DS2) centromeric haplotypes.

Table 1 Genetic association of activating KIRs with their putative ligands and the outcome of HCV infection in 336 individuals exposed to HCV (216 chronically infected and 120 with resolved HCV infection).

Logistic regression was performed using SPSS v20.0 using the ENTER method. OR > 1 indicates protection against chronic HCV infection.

View this table:

A highly conserved peptide from multiple flaviviruses contains the “AT” KIR2DS2-binding motif

HCV is a member of the Flaviviridae family, which incorporates viruses within the genus Flavivirus that include globally important pathogens such as Zika virus (ZIKV), dengue virus (DENV), yellow fever virus (YFV), Japanese encephalitis virus (JEV), and West Nile virus (WNV). Similar to HCV, these viruses all encode an RNA helicase belonging to superfamily 2 as a C-terminal domain of the NS3 protein (26). However, because the LNP peptide is not conserved among these viruses, we sought additional HLA-C*0102–binding peptides that may engage KIR2DS2. We screened the viral genomes from these viruses using NetMHCpan 2.8 and a percentage rank cutoff for binding of 2. We identified putative HLA-C*0102–binding peptides from the helicases of these viruses that had both the HLA-C*0102–binding motif and the “AT” KIR2DS2-binding motif (table S3). These peptides all contained a highly conserved “MCHAT” sequence from the NS3 Ib motif that is completely conserved among 61 of 63 viruses within the genus Flavivirus regardless of the species tropism of the virus (Fig. 4 and table S4). Using a BLAST search (, this sequence is not found intact within the human genome.

Fig. 4 Protein sequence alignment of the NS3 helicase–encoding region of 63 flaviviruses demonstrates conservation of a KIR2DS2-binding peptide.

The box indicates the putative KIR2DS2-binding peptide. Numbering is from the start of the NS3 protein of DENV.

KIR2DS2 recognizes conserved peptides from flaviviral helicases

We then assessed KIR2DS2 binding to peptides containing this conserved MCHAT peptide motif. We identified representative peptides from key human pathogens including ZIKV and DENV [IVDLMCHATF: both ZIKV NS3255–264 (BAP47441.1) and DENV NS3256–265 (ACK57817.1)], YFV [VIDAMCHATL: YFV NS3255–264 (AIZ07887.1)], and WNV and JEV [IVDVMCHATL: both WNV NS3255–264 (AFI56984.1) and JEV NS3255–264 (ABU94628.1)]. We used these peptides to load 721.174 cells and demonstrated that at 200 μM in flow cytometry assays, these peptides induced binding to KIR2DS2 tetramers, with enhanced binding of KIR2DS2 by the ZIKV and DENV peptide IVDLMCHATF (Fig. 5A).

Fig. 5 KIR2DS2 recognizes conserved helicase peptides from multiple flaviviruses.

(A) KIR2DS2 tetramer (black lines) binding to 721.174 cells incubated with the indicated peptides at saturating concentrations (200 μM) as compared with no peptides (gray lines). The median fluorescence of tetramer staining of peptide is indicated (JEV). (B) NKL-2DS2 cells were incubated with 721.174 cells preloaded with the indicated Flavivirus peptides and then assayed for phospho-Vav1 and Vav1 by Western blotting. One representative blot of each is shown. (C) The peptide IVDLMCHATF (IVDL) was coexpressed with HLA-C*0102 in 721.221 cells and used as targets in cytotoxicity assays in which the effector cells were NKL, NKL-2DL2, and NKL-2DS2. Experiments were performed at an E/T ratio of 6:1, and comparisons were made between the HLA-C*0102 transfectant and the HLA-C*0102:LNP or HLA-C*0102:IVDL cells. (D to F) HEK, HEK-C*0102, or HEK:C*0304 cells expressing a DENV replicon (HEK:DENV, HEK:C*0102:DENV, or HEK:C*0304:DENV, respectively) or not were cocultured with NKL cells expressing KIR2DS2 or KIR2DL2, and cytotoxicity was assessed at 24 hours. The effect of DENV on KIR2DS2-mediated killing is shown in (D), and the effect of HLA-C*0102 on KIR2DS2-mediated killing is shown in (E) and fig. S5. A comparison of the effect of HLA-C*0304 on killing of DENV-expressing cells with that of HLA-C*0102 by NKL-2DS2 is shown in (F). Means and SEs of at least three independent experiments performed in triplicate are shown. (G) Comparison of HCV (left) and dengue NS3 helicases (right) showing contact of the LNP (HCV) and IVDLMCHATF (Dengue) epitopes with the nucleic acid within the helicase domain (yellow). Selected amino acids from the epitopes are illustrated. Images were derived from structures published by Gu and Rice (25) and Luo et al. (29) and rendered using PyMOL. Where shown, P values were determined by independent two-tailed t tests (*P < 0.05, **P < 0.01, ***P < 0.001).

Our binding and functional experiments all used the KIR2DS2*001 allele, and we therefore tested whether allelic diversity of KIR affected KIR2DS2-mediated peptide recognition. We tested the alleles KIR2DS2*007 and KIR2DS*008, which have polymorphisms in the ligand-binding domains compared with KIR2DS2*001 (27). Using a tetramer binding assay, we found that these allelic variants had a similar peptide specificity to KIR2DS2*001 (fig. S5). Thus, allelic diversity of KIR2DS2 did not influence recognition of HLA-C*0102–presented peptides by KIR2DS2.

To determine whether these flaviviral peptides could activate NK cells, we tested Vav1 phosphorylation in NKL-2DS2 cells. 721.174 cells were loaded with peptide and cocultured with NKL-2DS2 cells. The flaviviral peptides that bound KIR2DS2 also induced Vav1 phosphorylation in NKL-2DS2 cells (Fig. 5B). We next generated a 721.221 transfectant stably expressing HLA-C*0102 and the DENV/ZIKV peptide IVDLMCHATF. We cocultured these cells with NKL, NKL-2DL2, and NKL-2DS2 cells. Endogenous expression of IVDLMCHATF with HLA-C*0102 resulted in augmented lysis of 721.221 transfectants by NKL-2DS2, as compared with lysis of 721.221 expressing HLA-C*0102 alone, but did not augment lysis by NKL or NKL-2DL2 cells (Fig. 5C).

To test the potential for KIR2DS2 to recognize native DENV, we used a dengue replicon system (28). The DENV genome replicates at a slightly lower level in human embryonic kidney (HEK) cells expressing the replicon than during a wild-type virus infection but with a slight overexpression of protein over time (fig. S6). HEK cells stably expressing the dengue replicon were transiently transfected with HLA-C*0102 or HLA-C*0304. NKL-2DS2 preferentially lysed DENV-replicating HEK cells transfected with HLA-C*0102, as compared with HEK:HLA-C*0102 not expressing DENV, as well as HEK cells expressing both DENV and HLA-C*0102 cells, as compared with those expressing DENV only (Fig. 5, D and E, and fig. S7). However, we did not observe KIR2DS2-specific lysis of HEK cells expressing DENV and HLA-C*0304 (Fig. 5F). Thus, in addition to HCV, KIR2DS2 also recognizes flaviviral helicase peptides in the context of HLA-C*0102.

Analysis of the crystal structures of the NS3 helicases of both HCV and DENV complexed with nucleic acids representative of the viral RNA indicates that similar to the HCV LNP epitope, the IVDLMCHATF peptide directly contacts the viral nucleic acid (Fig. 5G). Furthermore, the NS3 helicases of DENV, ZIKV, and JEV are all highly conserved, with this peptide occupying a similar position in all crystal structures (25, 2931). Thus, KIR2DS2 recognizes highly conserved peptide motifs within the helicases of the Flaviviridae family, which are directly involved in RNA binding and under extreme constraints on their ability to mutate.


These data demonstrate that KIR2DS2 functions as an antigen-specific receptor that recognizes conserved peptides from the Flaviviridae family of viruses. We have shown that KIR2DS2 recognizes a peptide derived from the helicase 1a region of HCV, which is highly conserved among HCV genotypes, and also a conserved peptide sequence from flaviviruses, including the DENV and ZIKV. All of the peptides have alanine and threonine at the C-terminal −1 and −2 positions, respectively, and within the flaviviral sequences, these two amino acids form part of an MCHAT motif within the helicase 1b region, which is conserved in 61 of 63 flaviviruses. The peptides can be endogenously present by HLA-C*0102, and the combination of the HCV peptide LNP and HLA-C*0102 is sufficient for KIR2DS2 to inhibit replication in vitro. Thus, KIR2DS2 has some functions analogous to a T cell receptor in recognizing peptide/MHC but has a broad specificity in that it also recognizes HLA-C*0304 and peptide (12). Binding to different peptides derived from diverse pathogens, as well as to different HLA class I molecules, is a likely prerequisite for KIR2DS2 to reach its current population frequency of about 50%. Our experiments suggest that an additional HCV peptide may be presented by HLA-C*0304 to KIR2DS2. However, further work is required to define the complete specificity of this receptor.

A broad specificity for KIR2DS2 is consistent with our understanding of how inhibitory KIRs engage HLA class I (10). The recognition of different peptides and HLA class I allotypes is facilitated by the motif-based recognition of HLA class I by KIR, in which the KIR/HLA interface is stabilized by salt bridges (32) and peptide selectivity is determined by the C-terminal −1 and −2 residues of the peptide. Thus, although the HCV and DENV peptides share only two residues, both are recognized by KIR2DS2 in the context of the same HLA class I molecule. This is consistent with the mode of binding observed for the HLA-C–specific inhibitory KIR. KIR2DS2 thus demonstrates similarities in binding properties to the inhibitory KIR but with distinct peptide selectivity from its inhibitory counterparts, KIR2DL2 and KIR2DL3. One concern is that a broad peptide specificity may lead to autoreactivity, and there is an association of KIR2DS2 with the autoimmune disease systemic sclerosis and rheumatoid vasculitis (33, 34).

We have studied ligands for KIR2DS2 in vitro, and analysis of our HCV population demonstrated an in vivo correlation of KIR2DS2 with the outcome of HCV infection. However, to date, we have not seen any studies that demonstrate an in vivo role for KIR2DS2 in flaviviral infection. In an immunogenetic study, there was an underrepresentation of KIR2DS2 in DENV-infected individuals (39.0%), as compared with healthy controls (64.8%) (35), suggesting a positive association of KIR2DS2 with DENV clearance. Furthermore, during acute DENV infection, NK cells are activated in vivo, but testing of KIR2DS2+ NK cells in this context is difficult because of the lack of specific reagents that distinguish KIR2DS2 from KIR2DL2/3. Thus, further in vivo work is required to determine whether KIR2DS2+ NK cells are specifically activated during acute flaviviral infections.

Nonetheless, higher KIR2DS2 gene frequencies are seen in populations from Central and Southern Africa, where flaviviral infections are more common (36). Furthermore, within the Amerindian population, both the KIR A and B haplotypes have been maintained (37). This balancing selection is thought to be due to the combined effects of the KIR B haplotype protecting against pregnancy-induced diseases and the KIR A haplotype protecting against infectious diseases. However, the genes associated with protection against pregnancy-associated diseases are located in the telomeric end of the KIR locus, so our data are consistent with a model in which the population frequency of the centromeric end of the KIR B haplotype is maintained through protection against infectious diseases. A broad specificity of KIR2DS2 for different HLA-C allotypes complexed with peptides derived from a pathogenic virus provides a rationale for the evolution of this family of receptors. Our observations, coupled with the recent discoveries of adaptive properties of NK cells, make them potential targets for novel therapeutic strategies against flaviviral infections.



The cohort of HCV-exposed individuals was recruited from UK hepatology clinics (4, 38). Two hundred thirty-three individuals had chronic HCV infection, and 128 had resolved infection. Overall, 239 (69%) were male, and 289 (82%) had acquired HCV through intravenous drug usage.

HCV replicon

The N17/JFH1 subgenomic replicon encoding the firefly luciferase reporter and the puromycin resistance marker has been described previously (21). Site-directed mutagenesis was performed using the QuikChange II XL kit (Agilent Technologies) with appropriate primers (sequences available upon request). The N17 replicon RNAs were generated in vitro using the T7 Megascript kit (Applied Biosystems). To measure replication of N17 SGR/JFH1WT, 4 × 106 naive HUH7 cells were electroporated with 10 μg of viral RNA, cells were seeded in triplicate and were lysed, and luciferase activity was measured using the Bright-Glo Luciferase assay system (Promega). HUH7 cells electroporated with the wild type or the L9D variant of the N17 replicon RNA were cultured in the presence of puromycin (2 μg/ml; Life Technologies), and the surviving cells were pooled and established as stable replicon-expressing cell lines.

Cell lines and transfectants

721.174 and HUH7 cells were cultured in R10 medium [RPMI 1640 supplemented with 1% penicillin-streptomycin (Life Technologies)] or Dulbecco’s modified Eagle’s medium (DMEM) and 10% fetal calf serum (FCS; Lonza). HLA-C*0102 or HLA-C*0304 were cloned into the pIB vector (39) and transfected into the HUH7.5 cells. The HEK:DENV replicon cells were grown in DMEM and 10% FCS supplemented with puromycin (2 μg/ml). The HLA-C*0304:ICP47 cell line was made by transfecting 721.221:C*0304 cells with the ICP47 gene cloned into pCDNA 3.1 using the Polyplus jetPRIME transfection system (Source Bioscience) and selected using hygromycin (500 μg/ml; Invivogen). The NKL cell line was transfected with the KIR2DL2, KIR2DS2, or KIR2DL3 cloned into the pIB vector, as previously described.

HUH7 cell lines stably expressing the wild-type N17 replicon or the L9D variant described above were transduced with a pIB-HLA-C*0102–expressing retroviral vector. The transduced cells were selected with puromycin (2 μg/ml) and blasticidin (1 to 6 μg/ml). HUH7:C*0102:replicon cells were cocultured with NKL cell lines for 24 hours at various E/T ratios for 24 hours, and then luciferase activity was assessed.

NK cell assays

CD107a degranulation

PBMCs were stimulated overnight with rHuIL-15 (human recombinant IL-15) (1 ng/ml; R&D Systems). Peptides were synthesized by GL Biochem Ltd or Peptide Protein Research. 721.174 cells were incubated with peptide at 26°C overnight, washed, and resuspended with the PBMCs at an E/T ratio of 5:1 and with anti–CD107a-AF647 (20 μl/ml). Cells were then incubated at 26°C for 4 hours with GolgiStop (6 μg/ml; BD Biosciences) added after 1 hour before staining and analysis by flow cytometry.

CellTracker Orange cytotoxicity assays

721.174 cells were incubated with peptide at 26°C for 16 hours and then with CellTracker Orange CMTMR (Life Technologies) and resuspended with NK cell clones at an E/T ratio of 10:1. Cocultures were incubated at 26°C for 4 hours, stained with LIVE/DEAD Fixable Aqua Dead Cell stain (Life Technologies), and fixed in 1% (w/v) paraformaldehyde (PFA) before analysis by flow cytometry.

LNP construct

A construct expressing HLA-C*0102 in continuity with the 2A self-cleaving peptide sequence from Thosea asigna virus, the E3/19K endoplasmic reticulum–targeting sequence, and the LNP or IVDLMACHATF peptide was synthesized from GeneArt (Life Technologies), cloned into the pIB vector, and transduced into the 721.221 cell line.

Inhibition of HCV replication assays

Target cells Huh7-J17-WT/HLA and Huh7-J17-R1/HLA were seeded. The next day, NKLs were cocultured with the target cells in duplicate for 24 hours. Plates were then centrifuged, and the cell pellet was lysed with Glo Lysis Buffer (Promega). Fifty microliters of each lysate was read using 50 μl of luciferase assay reagent (Promega) on a GloMax Discover luminometer.

HEK cell cytotoxic killing assay

HEK and HEK:DENV cells were transiently transfected with 2 μg of C*0102mCherry-pCDNA3 or C*0304mCherry-pCDNA3 using Polyplus jetPRIME (Source Bioscience). After 24 hours, cells were cocultured for 24 hours with NKL cell lines in triplicate at an E/T ratio of 6:1. Cells were stained with LIVE/DEAD Zombie Violet stain (BioLegend) for 1 hour, fixed with 1% PFA–phosphate-buffered saline (PBS) solution, and analyzed by flow cytometry. DENV-positive HEK cells were detected by the green fluorescent protein (GFP) expression of the DENV construct, and HLA-C*0102– or HLA-C*0304–transfected cell lines were detected by mCherry expression. Percentage cytotoxicity was determined by calculating the percentage of Zombie Violet–positive cells in the presence of effectors minus the percentage of Zombie Violet–positive cells in the absence of effectors.

KIR staining

The extracellular domain of KIR2DS2*001 or KIR2DL2*001 containing a C-terminal biotinylation tag was cloned into the pET23d+ vector. This was expressed as inclusion bodies in Rosetta 2 (DE3) bacteria (Merck Millipore), purified, denatured, and reduced in 6 M guanidine-HCl and 20 mM dithiothreitol. KIR2DS2 was purified by gel filtration chromatography using fast protein liquid chromatography. Biotinylation was performed using a BirA biotinylation kit (Avidity). Fluorescently labeled tetramers were produced by coupling biotinylated KIR2DS2 molecules to PE-conjugated streptavidin (Molecular Probes, Life Technologies). The KIR2DS2*007 and KIR2DS2*008 tetramers were made from a KIR2DS2*001 template using the QuikChange Lightning Site-Directed Mutagenesis kit (Agilent Technologies) (primer sequences are available upon request). The KIR2DL2-IgG fusion construct (KIR2DL2-Fc Chimera, R&D Systems) was conjugated with protein A Alexa Fluor 488 (Life Technologies).

KIR genotyping

KIR genotyping of the donors was performed by quantitative polymerase chain reaction (qPCR), as previously described (4).

NK cell clones

NK cells (CD3 CD56+ CD158b+) were single cell–sorted from PBMCs from a KIR2DL2+S2+L3 donor into stem cell growth medium (CellGenix) containing 5% heat-inactivated human serum and IL-12 (10 ng/ml; PeproTech), IL-15 (20 ng/ml; R&D Systems), and IL-18 (100 ng/ml; R&D Systems). Clones were fed weekly with irradiated PBMCs (1 × 106/ml), IL-2 (250 IU/ml), and phytohemagglutinin (2.5 μg/ml; Thermo Fischer Scientific). Clones were typed for expression of KIR2DS2 by reverse transcription qPCR. RNA was extracted from the NK clones using TRIzol (Life Technologies) and reverse-transcribed using a high-capacity reverse transcriptase kit (Applied Biosystems). One nanogram of complementary DNA was used for qPCR analysis. KIR2DS2, KIR2DL2, and KIR2DL3 primers and probes were as previously described (40). Each sample was read in triplicate, and gene expression was determined using the ΔCt method.

Conjugate formation and staining

721.174 cells were incubated overnight at 26°C in the absence or presence of 100 μM peptide. Ba/F3-KIR2DS2 or Ba/F3-KIR2DL2 cells (1 × 105; a gift from L. Lanier) were co-incubated with the 721.174 cells at an E/T ratio of 2:1. Cells were fixed in 4% (w/v) PFA in PBS and then stained with CD158b-PE (GL183) antibody (20 μg/ml; Beckman Coulter). Cells were imaged using a Leica SP5 resonance scanning microscope (Leica Microsystems). Transmission images and PE emission were collected in different channels. Data were processed using Leica Imaging software and ImageJ software (

Immunoprecipitation and Western blotting

721.174 cells were cultured with 20 mM peptide overnight at 26°C. NKL-2DS2 and 721.174 cells were co-incubated at an E/T ratio of 1:1 for 5 min at 37°C and lysed in 20 mM tris-HCl (pH 7.6), 150 mM NaCl, 1 mM EDTA, 1 mM sodium orthovanadate, and 0.5% Triton X-100. KIRs were immunoprecipitated with anti-CD158b (clone GL183, AbD Serotec) and analyzed by Western blotting. Antibodies recognizing phospho-Vav1 (EP510Y, Abcam), Vav1 (Cell Signaling Technology), anti-phosphotyrosine (4G10 Platinum, Millipore), and DAP-12 (clone D7G1X; Cell Signaling Technology) were used with horseradish peroxidase–conjugated secondary antibodies (Cell Signaling Technology). Membranes were stripped using the Western Blot Recycling kit (Alpha Diagnostics). Protein bands were detected by chemiluminescence (SuperSignal West Pico Chemiluminescent Substrate, Perbio Science) using the ChemiDoc-It Imaging system with VisionWorks software (UVP) and quantified with ImageJ software.

Peptide binding analysis

The online resources ADT, NetMHCpan, and KISS were used to search for potential HLA-C*0102–binding epitopes (4143).

Statistical analysis

Statistical analyses (t tests and ANOVA) were performed using GraphPad Prism version 6, and logistic regression analysis was performed using SPSS v20.0 (IBM).


Fig. S1. Peptide stabilization of HLA-C*0102 by HCV peptides.

Fig. S2. The effect of single-HCV peptides on degranulation of CD158b NK cells.

Fig. S3. Cytotoxicity of KIR2DS2+ NK cell clones to peptide-loaded 721.174 cells.

Fig. S4. KIR2DS2 tetramer binding to HLA-C*0102 and HCV peptides or VAPWNSLSL peptide derivatives.

Fig. S5. Flow cytometry plots comparing binding of KIR2DS2*001, KIR2DS2*007, and KIR2DS2*008 to HLA-C*0102 and peptide.

Fig. S6. Analysis of viral RNA and protein production in DENV replicon–containing cells compared with DENV-2–infected cells.

Fig. S7. Flow cytometry plots illustrating gating strategy and killing of HEK cells expressing GFP-tagged DENV replicon by NKL-2DS2 cells.

Table S1. Summary of HCV peptides identified to bind HLA-C*0102.

Table S2. Protein sequence alignment of HCVs.

Table S3. Flavivirus NS3 HLA-C*0102–binding peptides as determined by NetMHCpan.

Table S4. Accession numbers of Flavivirus sequences used to compile the alignment in Fig. 4.

Data file S1. Raw data for Figs. 1 to 5.

Data file S2. Western blot gels for Figs. 1 to 5.


Acknowledgments: We thank L. Lanier for the reagents; S. Gadola, L. Tereza, and J. Vivian for helpful discussions; and A. Al-Shamkhani for the critical reading of the manuscript. Funding: This work was supported by grants from the Wellcome Trust (WT089883MA to S.A.C., WT093465MA to M.A.P., and WT076991MA to S.I.K.) and the Medical Research Council (G1001738 to S.I.K., G0401586 to A.D.D., and MC_UU 12014/2 to A.H.P.). Author contributions: M.M.N., S.A.C., A.M., V.C., H.K., B.M., S.M., A.H.P., M.A.P., S.I.K., and K.C. designed the experiments. M.M.N., S.A.C., A.M., V.C., K.C., H.K., B.M., S.M., P.R., and K.C. performed the experiments. A.D.D., A.M., L.J.F., and F.H.J.C. provided the key reagents. M.M.N., A.H.P., M.A.P., A.D.D., and S.I.K. wrote the manuscript. M.M.N., S.A.C., H.K., and S.I.K. performed data analysis. M.M.N., S.A.C., H.K., S.H., and S.I.K. performed statistical analysis. Competing interests: S.I.K., S.A.C., M.M.N., and M.A.P. have applied for a patent on the use of peptides for NK cell therapy. All other authors declare that they have no competing interests.
View Abstract

Navigate This Article