HLA-C-restricted presentation of a conserved bacterial epitope to an innate NK cell receptor

The killer-cell Ig-like receptor (KIR) family, expressed mainly in natural killer (NK) cells, includes an activation receptor of unknown function, KIR2DS4. Here we show that KIR2DS4 is restricted by HLA-C*05:01 with a strong preference for tryptophan at position 8 of 9-mer peptides. ‘Self’ peptides with Trp8 eluted from HLA-C*05:01 are rare and only one out of 12 bound KIR2DS4. An HLA-C*05:01-peptide complex that bound KIR2DS4 was sufficient for strong activation of primary KIR2DS4+ NK cells, independently of coactivation by other receptors and of prior NK cell licensing. A highly conserved sequence in bacterial recombinase A, which is essential for DNA repair and survival, includes an epitope that bound to HLA-C*05:01 and activated KIR2DS4+ NK cells. Thus, in addition to their established role in defense against viruses and cancer, NK cells may have also evolved to detect and respond to hundreds of bacterial species through recognition of a conserved RecA epitope.


Introduction
MHC class I molecules (MHC-I) play critical roles in innate and adaptive immunity. MHC-I molecules present short peptides, commonly 8-11 amino acids in length, which are surveilled by alpha-beta T cell receptors expressed by CD8 T cells. MHC-I also serves as a critical regulator of natural killer (NK) cells, innate immune cytotoxic cells with the capacity to produce proinflammatory cytokines (Long et al., 2013;Vivier et al., 2011). Following the 'missing self' hypothesis, MHC-I binding inhibitory receptors expressed by NK cells detect loss of MHC-I leading to NK cell activation (Ljunggren and Karre, 1990). Additionally, interactions between inhibitory receptors and MHC-I dictate the effector potential of NK cells via a process known as 'education' or 'licensing' (Elliott and Yokoyama, 2011;Long et al., 2013). NK cells have established roles in immune defense against cancers and viral infections, where loss or downregulation of MHC-I is common (Morvan and Lanier, 2016;Vidal et al., 2011). The functions of MHC-I binding NK cell inhibitory receptors appear conserved across species and different families of receptors.
In humans, the major NK receptors for HLA-I (human MHC-I) are CD94:NKG2A, which binds HLA-E, and the killer-cell immunoglobulin-like receptors (KIR). There are 14 KIR genes which encode activating and inhibitory receptors. The ligands for inhibitory KIR are well defined as groups of HLA-A, -B or -C allotypes, each with a common epitope. All HLA-C allotypes carry either the C1 or C2 epitopes, which are ligands for the inhibitory receptors KIR2DL2/3 and KIR2DL1 respectively (Parham, 2005). KIR bind the peptide exposed face of HLA-I towards the C-terminal end of the peptide, incorporating peptide into the binding site and all HLA-C binding KIR studied to date demonstrate a degree of peptide selectivity (Boyington et al., 2000;Boyington and Sun, 2002;Fan et al., 2001;Rajagopalan and Long, 1997;Sim et al., 2017;Stewart et al., 2005). In contrast to the inhibitory KIR, definitive functional ligands for activating KIR are still lacking.
The KIR genes are organized into two broad haplotypes, KIR A and KIR B, which differ by gene content. The simpler KIR A haplotype contains only one activating receptor KIR2DS4, while the B haplotype is characterized by variable gene content including multiple activating receptor genes (Parham, 2005). The two haplotypes have a similar worldwide frequency, but differ between populations, such that KIR A homozygotes are not rare and for whom KIR2DS4 is the only activating KIR they carry. Due to variability of KIR haplotypes and the fact that HLA-I and KIR are on different chromosomes, individuals can express orphan receptors or express ligands without the corresponding KIR. Consequently, gene association studies have linked the presence or absence or KIR-ligand pairs with many disease processes including viral infections, autoimmunity and cancer (Bashirova et al., 2011;Boyton and Altmann, 2007;Hiby et al., 2004;Khakoo et al., 2004;Parham, 2005;Parham and Moffett, 2013). Additionally, activating KIR with the ability to bind HLA-C appear to have a protective role against disorders of pregnancy (Hiby et al., 2004;Kennedy et al., 2016;Nakimuli et al., 2015).
The KIR2DS4 locus is not fixed and two major alleles exist that encode either the fulllength receptor (KIR2DS4-fl) or a version with a deletion (KIR2DS4-del). KIR2DS4-del encodes a 22-bp deletion leading to an early stop codon creating a truncated soluble protein with no recorded HLA-I binding Maxwell et al., 2002). KIR2DS4-fl is an HLA-I binding receptor and binds a subset of C1 and C2 HLA-C allotypes, in contrast to other KIR2D receptors, which dominantly bind C1 or C2 ). This previous report identified KIR2DS4 ligands via a binding assay using soluble KIR molecules and many HLA-A, -B and -C proteins bound to beads (Hilton et al., 2015a). This method has proved useful to screen many HLA-I allotypes at once, but the sequence and diversity of peptides presented on the beads is unknown. Furthermore, it is not clear whether HLA-C constitutes a functional ligand for KIR2DS4 or how peptide sequence contributes to KIR2DS4 binding. Indeed, the only known functional ligand for KIR2DS4 is HLA-A*11:02 .
Carrying KIR2DS4-fl is associated with protection from pre-eclampsia and glioblastoma, and with higher viral loads and faster progression to AIDS, in HIV infection (Dominguez-Valentin et al., 2016;Kennedy et al., 2016;Merino et al., 2014;Olvera et al., 2015). There is a clear need to define functional ligands for KIR2DS4 to fully understand its role in these disease processes and in the regulation of NK cells more generally. HLA-C*05:01, a C2 allotype, was reported to bind KIR2DS4-fl . Henceforth, we refer to KIR2DS4-fl as KIR2DS4 unless in direct comparison to KIR2DS4-del. The aim of this study was to establish whether HLA-C*05:01 is a functional ligand for KIR2DS4 and if peptide sequence influences KIR2DS4 binding. We find that KIR2DS4 binds HLA-C*05:01 in a highly peptide selective manner and that this binding potently activates KIR2DS4 + NK cells. Further, we link peptide specific recognition of HLA-C*05:01 by KIR2DS4 to a protein highly conserved among bacteria pathogenic in humans.
Due to the variegated expression of inhibitory NKG2A and KIR, KIR2DS4 + and KIR2DS4 -NK cells include both licensed and unlicensed NK cells. Licensing endows NK cell subsets with greater capacity to degranulate in the absence of MHC-I. However, activation of KIR2DS4 + NK cells requires expression of HLA-C, thus it is not clear how NK cell licensing may impact activation of KIR2DS4 + NK cells. To examine the role of NK cell licensing on activation of KIR2DS4 + NK cells, we gated CD56 dim KIR2DL1/S1 -NK cells into four subsets based on the expression of KIR2DS4 (S4) and other MHC-I binding receptors (KIR2DL2/L3/S2, NKG2A and KIR3DL1/S1), termed receptor positive (R + ) and receptor negative (R -) ( Fig. 2A). All donors carry the ligand for NKG2A (HLA-E), while donors vary for the presence of the ligands for KIR3DL1 and KIR2DL2/3. Thus, R + NK cells are expected to be licensed and degranulate well in response 221-C*05:01-ICP47 cells as they lack ligands for KIR2DL2/3, NKG2A and KIR3DL1. R -NK cells lack all licensing receptors and are expected to be degranulate weakly in response 221-C*05:01-ICP47 cells. As expected, licensed R + NK cells degranulated well in response to 221-C*05:01-ICP47 cells, while the R -NK cells degranulated poorly irrespective of KIR2DS4 expression (Fig. 2B,C).
Remarkably, the potent activation of KIR2DS4 + NK cells in response to cells loaded with P2-AW was similar between R + and Rsubsets. Thus, even unlicensed R -NK cells can respond strongly to KIR2DS4 stimulation overriding the lack of licensing.
There was a strong correlation (r 2 =0.9) between peptide dependent KIR2DS4-Fc binding to HLA-C*05:01 and the functional response of R -S4 + NK cells in response to the same peptides ( Fig. 2C,D). Activation of R -S4 + NK cells in response to 221-C*05:01-ICP47 cells loaded with P2-AW and P2-AY was dependent on peptide concentration (Fig. 2E). Concentrations of peptide P2-AW lower than 100 nM activated of R -S4 + NK cells and half-maximal activation was reached with 0.07µM of P2-AW and 1 µM of P2-AY (Fig. 2E). R -S4 + NK cells in response to 221-C*05:01-ICP47 cells loaded with P2-AW also reached a greater maximum than that was obtained with high concentrations of P2-AY (Fig. 2E). Peptide-specific activation of KIR2DS4 + NK cells also induced production of IFN-gamma and TNF-alpha, ascertained by intracellular cytokine staining (Fig. 2F,G and SFig. 2C).
Signaling by ITIM bearing inhibitory receptors dominates over signaling by NK cell activation receptors (Long et al., 2013). The potent activation of NK cells by KIR2DS4, which can override the lack of NK cell licensing, suggested it may be difficult to inhibit. To test whether NK cell activation by KIR2DS4 is regulated by inhibitory receptors, we first re-gated our data from previous degranulation experiments (Fig. 2C). We gated NK cells as CD56 dim then gated on R -NK cells without excluding the KIR2DL1/S1 + (EB6) population. These NK cells were then gated into four subsets based on expression of KIR2DS4 (S4) and KIR2DL1/S1 (L1/S1) (Fig. 2H). S4 + L1/S1 -NK cells displayed strong activation in response to 221-C*05:01-ICP47 cells loaded with P2-AF, P2-AY and P2-AW ( Fig 2H). In contrast, S4 + L1/S1 + NK cells displayed no activation even in the response to cells loaded with P2-AW. Because the mAb EB6 binds KIR2DL1 and KIR2DS1, we confirmed these results with two donors who were typed to only carry KIR2DL1 (SFig. 2D,E). As KIR2DS4 is generally expressed by a higher proportion of NK cells than KIR2DL1, many KIR2DS4 + NK cells will not be inhibited (SFig. 2F,G). Together, these data show that KIR2DS4 + NK cells are potently activated to degranulate and produce cytokines upon engagement of KIR2DS4 and HLA-C*05:01 in a highly peptide selective manner. Activation of KIR2DS4 + NK cells was limited to the KIR2DL1subset due to dominant inhibition by KIR2DL1 + NK cells. R -S4 + NK cells were unlicensed but potently activated by KIR2DS4 stimulation overriding a lack of NK cell licensing. Therefore, under physiological conditions where the ligands for many NK cell inhibitory receptors are expressed, it is likely that the R -S4 + NK subset is the major operative subset due to its lack of inhibitory receptor expression.
Activation of KIR2DS4 + cells was dependent on bead number (SFig. 3D) and expression of KIR2DL1/S1 on KIR2DS4 + NK cells inhibited NK cell activation (SFig. 3E), consistent with our previous results with peptide loaded 221-C*05:01-ICP47 cells (Fig. 2H). Next, HLA-C*05:01 conjugation to Dynabeads was titrated such that the number of HLA-C molecules per bead ranged from 0.5×10 3 to 10 5 (Fig. 3C, SFig 3F). The number of HLA-C molecules per bead was estimated using quantitative flow cytometry, by generating a standard curve from beads with known antigen densities (SFig 3F). S4 + L1/S1 -NK cells showed a sharp activation threshold, with half maximal responses at 2000 and 1000 HLA-C molecules per bead for donors 1 and 2, respectively ( Fig.   3E). In contrast, S4 + L1/S1 -NK cells were weakly stimulated with beads conjugated with HLA-C*05:01-P2-AV at antigen densities greater than 20,000 HLA-C molecules per bead (Fig. 3D, SFig. 3G). NK cells from a third donor showed a weaker response with a gradual increase in the activation of S4 + L1/S1 -NK cells with increasing HLA-C antigen density and a half maximal response at 8000 HLA-C molecules per bead (Fig. 3E). We concluded that HLA-C alone, immobilized to beads, is sufficient to activate KIR2DS4 + NK cells across a range of antigen densities in the absence of ligands for other co-activation receptors or adhesion molecules.

Functional presentation of endogenous P2-AW peptide by HLA-C*05:01 to KIR2DS4 + NK cells
Our experiments thus far demonstrated that KIR2DS4 + NK cells can recognize peptides presented by HLA-C*05:01 in a highly selective manner. For these experiments, the peptides presented by HLA-C*05:01 have been dominated by single peptide sequences, in the form of recombinant HLA-C refolded with peptide or TAP-deficient cells loaded with peptide. While useful experimental tools, under physiological conditions HLA-C*05:01 does not present single peptides, but many peptides of different sequences (Di Marco et al., 2017;Kaur et al., 2017;Sim et al., 2017). To test whether KIR2DS4 + NK cells could recognize peptide in TAP sufficient cells where HLA-C*05:01 presents many different peptides, we expressed the peptides P2-AW or P2-AV in 221-C*05:01 cells using a retrovirus (Fig. 4A, SFig. 5A) (Gejman et al., 2018). Transduced cells were marked by expression of mCherry, and after drug selection both mini-genes were expressed in over 90% of cells (SFig. 5B). In degranulation assays, R -S4 + NK cells responded weakly to both 221-C*05:01 cells and those transduced with P2-AV (Fig. 4B). In contrast, R -S4 + NK cells exhibited enhanced degranulation in response to 221-C*05:01 cells expressing the P2-AW peptide (Fig. 4B). This activation was KIR2DS4 specific, as R -S4 -NK cells exhibited low responses to both 221-C*05:01 cells and those transduced with P2-AV or P2-AW. Thus, KIR2DS4 can detect a stimulatory peptide, presented by HLA-C*05:01 in TAP sufficient cells that present many different peptide sequences.
A 'self' peptide eluted from HLA-C*05:01 carrying Tryptophan at position 8 is a KIR2DS4 epitope We reasoned that peptides containing Trp at p8 may be enriched for KIR2DS4 binding peptides, as in the context of P2-AY, Trp at p8 conferred stronger binding of KIR2DS4 to HLA-C*05:01 than Tyr or Phe (Fig. 1). Peptides eluted from HLA-C*05:01 contain a low frequency of Trp (SFig. 5), which was similar to peptides eluted from other HLA-C allotypes (Di Marco et al., 2017;Kaur et al., 2017). Tryptophan is rare at p8 for all HLA-C allotypes, ranging from 0-2.6%, and is found at a frequency of 0.6% in HLA-C*05:01 peptides (SFig. 5B). The scarcity of peptides containing Trp at p8 presented by HLA-C*05:01 may explain the lack of KIR2DS4-Fc binding to 221-C*05:01 cells.
We synthesized and tested the 12 peptides with Trp at p8 that had been eluted from HLA-C*05:01 for KIR2DS4 binding and activation of KIR2DS4 + NK cells. One strong and one weak KIR2DS4 binding 'self' peptides were identified (Fig. 5). The KIR2DS4 weak binding peptide was TM9SF4323-311 (MSDVQIHWF). TM9SF4 is a member of the transmembrane superfamily 9, a highly conserved family across evolution with roles in cell adhesion, phagocytosis and autophagy (Paolillo et al., 2015). The KIR2DS4 strong binding 'self' peptide was HECTD11131-1139 (SNDDKNAWF). The Homologous to the E6-AP Carboxyl Terminus (HECT) domain E3 ubiquitin ligase 1 (HECTD1) has been linked to cholesterol export from macrophages (Aleidi et al., 2018) and suppression of epithelial-to-mesenchymal transition in cancer metastasis (Duhamel et al., 2018). Both TM9SF4 and HECTD1 are expressed in all tissues at the RNA and protein levels (Uhlen et al., 2015). That only two of 12 peptides with Trp at p8 bound KIR2DS4 emphasized the high peptide specificity of KIR2DS4 binding to HLA-C*05:01 and that Trp is not sufficient to identify KIR2DS4 binding peptides.

A conserved KIR2DS4 epitope derived from Recombinase A is shared by hundreds of species of bacteria
To explore whether KIR2DS4 recognizes pathogen-derived peptides, we searched the proteomes of prokaryotes for sequences similar to P2-AW. P2-AW (IIDKSGAWV) showed remarkable homology to residues 283-291 of Recombinase A (RecA) of Helicobacter fennelliae (IIDKSGAWI; Fig. 6A). RecA is the prototypical DNA recombinase and is essential for DNA damage repair by homologous recombination (Lusetti and Cox, 2002). A previous study aligned the sequences of RecA proteins from 63 species of bacteria and found amino acid sequence similarity ranged from 43% to 100% (Karlin and Brocchieri, 1996). Residues 283-291 are highly conserved, and in particular, G288 is conserved in all 63 species, W290 is found in 61 out of 63 species and acidic residues at position 285 are found in 40 out of 63 species (SFig. 6A). A logo-motif for RecA283-291 from these 63 species shown in Figure 6B.
We tested 14 RecA283-291 peptides with different sequences, each containing Asp or Glu at position 3, a critical anchor residue for HLA-C*05:01. Due to the high level of conservation, these 14 peptides cover more than 14 species of bacteria; for example, RecA283-291 from Yersinia pestis is identical to that of Escherichia coli. RecA283-291 peptides with Phe or Leu at the C-terminus   , 2015). Taking a broader approach, we generated a RecA283-291 motif (xxD/ExxGxWp9) accounting for the essential role of acidic resides at position 3 for binding HLA-C*05:01, the high conservation of p6 G and the importance of p8 Trp for binding KIR2DS4 (Fig. 6G). We allowed for any amino acid at all other positions (x) expect p9. To focus our analysis on peptides predicted to be presented well by HLA-C*05:01, we counted only RecA283-291 sequences where Phe, Leu, Ile, Met or Val at p9, all common C-terminal anchors for HLA-C*05:01 peptides (SFig. 6B). Using this motif, we identified over one thousand different bacterial species that contain RecA283-291 sequences that have the potential to be presented by HLA-C*05:01 and bind KIR2DS4 (Fig. 6G). The majority of species containing RecA283-291 sequences predicted to bind KIR2DS4 were from Proteobacteria and within the Proteobacteria the most common Order was Rhizobiales (SFig. 6C,D). All bacterial species containing RecA sequences predicted to be presented by HLA-C*05:01 and recognized by KIR2DS4 are shown in Supplementary Table 1. Together our data suggest the possibility that HLA-C*05:01 + individuals expressing KIR2DS4 may have an evolutionary advantage in bacterial immunity, through the ability of their NK cells to recognize RecA epitopes presented by HLA-C*05:01.

Allele frequency of full-length KIR2DS4 is inversely correlated with the frequency of HLA-C*05:01
The allele frequency of functional KIR2DS4-fl (KIR2DS4*001) is positively correlated with HLA-A*11, a previously identified KIR2DS4 ligand

Discussion
The concept of an innate receptor with specificity for one form of a highly variable antigen, such as an MHC-peptide complex, is counterintuitive. And yet, members of the killer-cell Ig-like receptor (KIR) family bind to HLA class I allotypes with various degrees of selectivity for peptides (Malnati et al., 1995;Rajagopalan and Long, 1997;Sim et al., 2017). In the case of the activating members of the KIR family, for which ligands are less well defined, a reasonable hypothesis was that they serve to recognize conserved features of modified 'self'. Here we describe an extreme case of peptide selectivity of an activating KIR, which is restricted by HLA-C*05:01. Peptides that promote binding of KIR2DS4 carried a Trp at position 8 of 9-mer peptides. Recognition of this peptide-HLA-C*05:01 complex by KIR2DS4 + NK cells induced potent activation of NK cells, and could even activate unlicensed NK cells. Interestingly, optimal peptides for activation of KIR2DS4 + NK cells identified in this study included an epitope that is highly conserved in the essential protein RecA expressed by hundreds of bacterial species, suggesting KIR2DS4 may have evolved to contribute to bacterial innate immunity.
RecA is essential for repair of damaged DNA by homologous recombination and is critical for the DNA damage response in E. coli (Lusetti and Cox, 2002). The epitope in the bacterial protein RecA for KIR2DS4 binding contains Trp at p8 and this position (residue 290 in E. coli) is conserved across bacteria, including many human pathogens. Tryptophan is a rare amino acid in general (1.3% of the human proteome), which is reflected in its low representation in peptides eluted from HLA-C (Di Marco et al., 2017;Kaur et al., 2017). Only 0.6% of peptides (n = 2035) eluted from HLA-C*05:01 contained Trp at p8. Out of 12 such peptides, only one bound KIR2DS4, providing an explanation for the lack of KIR2DS4 binding to, or activation of KIR2DS4 + NK cells by 221-C*05:01 cells. The low abundance of peptides including Trp at p8 provides the opportunity to detect similar but more abundant peptides, perhaps modified peptides or peptides derived from pathogens. We propose that KIR2DS4 is not specific for non-self but for 'rare self', which provides the opportunity to detect abundant foreign peptides carrying a 'rare self' epitope.
The major function of inhibitory KIR is recognition of 'self' to inhibit NK cell activation and dictate NK cell licensing (Long et al., 2013). Inhibitory KIR exhibit limited peptide selectivity and bind HLA class I in the context of many biochemically diverse peptides. However, differences in peptide selectivity exist between KIRs. Recently, we compared KIR2DL1 binding to HLA-C*05:01 and KIR2DL2/3 binding to HLA-C*08:02 in the presence of the same peptides. HLA-C*05:01 and HLA-C*08:02 differ only by the two amino acids that define the C1 and C2 epitopes and could be loaded with the same peptides. While KIR2DL1 bound HLA-C*05:01 in the presence of 24 out of 28 peptides, KIR2DL2 and KIR2DL3 bound HLA-C*08:02 in the presence of only 13 and 11 of the same peptides, respectively (Sim et al., 2017). Position 44 of KIR2DL1 (Met) and KIR2DL2/3 (Lys) dominates in determining KIR specificity for C2 and C1 allotypes, respectively (Boyington et al., 2000;Fan et al., 2001). However, Lys44 receptors exhibit cross-reactive binding to C2 allotypes (Moesta et al., 2008;Winter et al., 1998). KIR2DS4, which carries Lys44, has the capacity to bind a small subset of C1 and C2 allotypes. The high peptide selectivity of KIR2DS4 is reminiscent of cross-reactive KIR binding, where KIR2DL2/3 bound HLA-C*05:01 and KIR2DL1 bound HLA-C*08:02 in the presence of 5 and 2 out of 28 of the same peptides, respectively (Sim et al., 2017). KIR2DS4 was even more peptide selective than this, binding none of the same 28 peptides previously studied and only in the presence of the few peptides described here.
Examples of pathogenic bacteria that carry an epitope within RecA for HLA-C*05:01restricted binding of KIR2DS4 include: Helicobacter pylori, an infectious agent associated with peptic ulcers and gastric cancers (Suerbaum and Michetti, 2002); bacteria of the Brucella genus, which cause a zoonotic infection called brucellosis (Skendros et al., 2011); Campylobacter jejuni, a major cause of gastroenteritis with an incidence as common as that caused by Salmonella infections (Marder et al., 2017); and Chlamydia trachomatis, which is the most common sexually transmitted infection with over 90 million cases annually worldwide (Brunham and Rey-Ladino, 2005). There are bacterial species with a RecA sequence that lacks an Asp or Glu at position 3, which is required for peptide binding to HLA-C*05:01 (Di Marco et al., 2017;Kaur et al., 2017;Sim et al., 2017). Such species include Mycobacterium tuberculosis which carries the conserved Trp at p8 but has an Arg at position 3 (LIRKSGAWF). Additionally, other species such as E. coli and Yersinia pestis carry a Tyr at p9, which is a poor C-terminal anchor for HLA-C*05:01. It is possible that other HLA-C allotypes may also present these RecA peptides by accommodating other amino acids at p3 or p9, expanding the coverage of pathogenic bacteria detected by KIR2DS4 is a strong activator of NK cells and was sufficient on its own, to elicit functional responses by KIR2DS4 + NK cells. Using an assay with recombinant HLA-C*05:01 carrying an optimal Trp8 peptide conjugated to beads, we determined that the number of HLA-C molecules required for half maximal activation of KIR2DS4 + NK cells to be from 1000-8000 suggesting the number of molecules required to activate KIR2DS4 + NK cells may be within a physiological range.  (Maxwell et al., 2002;Middleton et al., 2007;Robinson et al., 2015). This is the first evidence for balancing selection between an activating KIR and its ligand, which is reminiscent of examples of balancing selection in populations with high frequencies of inhibitory KIR ligands (Parham et al., 2012 likely to contribute to clearance of bacterial pathogens. The first line of defense against invading bacteria are phagocytic, antigen-presenting cells (APC) such as monocytes, macrophages and dendritic cells. NK cells co-operate with these cell types to produce IFN-g via cell contactdependent mechanisms and IL-12 production (Ferlazzo et al., 2003;Gerosa et al., 2002;Walzer et al., 2005). It is possible that KIR2DS4 + NK cells could directly recognize bacterially infected cells, indeed C. trachomatis, C. jejuni and B. abortus are intracellular bacteria (Brunham and Rey-Ladino, 2005;Skendros et al., 2011;Watson and Galan, 2008). An additional scenario is that KIR2DS4 + NK cells could be activated upon interaction with APCs that present RecA epitopes on HLA-C to produce IFN-g early during infection, and facilitate a Th1 response necessary to clear the bacterial infection. In this case, the RecA epitope would be presented by HLA-C through crosspresentation, the process whereby exogenous antigens enter the MHC-I pathway (Blander, 2018). The DC subset associated with efficient cross-presentation consists of classical DC1 NK cell functional assays. NK cell degranulation assays were carried out largely as described (Sim et al., 2017). Primary, resting NK cells were isolated by negative selection from PBMCs and were greater than 95% CD56 + , less than 5% CD3 + (EasySep NK cell Isolation Kit, Stem Cell, USA). Donors were screened for the presence of KIR2DS4 with the mAb JJC11.6 (Miltenyi Biotec). 10 5 resting NK cells were mixed at 1:1 with 221 cell lines or peptide loaded target cells in presence of 1 µl BV421 anti-CD107a mAb for 2 hr at 37°C (H4A3, Biolegend 328626). NK cells were then stained with mAbs to identify subsets based on the expression of KIR2DL1/S1 (Beckman Coulter, EB6), Receptor (R; Beckman Coulter, Z199, GL183, CD158e1/2) and KIR2DS4 (S4; Miltenyi Biotec, JJC11.6). For intracellular cytokine staining, NK cells were mixed with targets for 6hr at 37°C, Golgi Plug (BD, 555029) was added after 1 hr. Cells were then fixed and permeabilized with Cytofix/Cytoperm (BD, 554714) and stained for IFN-g (BD, B27) and TNFa (BD, 554512). Cells were washed with PBS three times and data acquired by flow cytometry.