Expression of ligands for activating natural killer cell receptors on cell lines commonly used to assess natural killer cell function

Natural Killer (NK) cells are a subset of lymphocytes that direct innate immune responses to kill stressed, virally-infected, and transformed cells [1]. NK cells belong to the group 1 innate lymphoid cells (ILCs) as do ILC1s [2–4]. NK cells can interact with target cells and lyse them directly via the release of cytotoxic granules containing perforin and granzyme [5, 6]. Activated NK cells can also secrete a broad range of cytokines and chemokines that can activate adaptive immune cells to lyse target cells, bridging the innate and adaptive immune systems [7–9]. Regardless of the method used by NK cells to lyse target cells, they must first be activated to elicit a response. The activation state of an NK cell results from the integration of different signals transmitted through its activating and inhibitory natural killer cell receptors (aNKR and iNKR) [9, 10]. Activation can result from the loss of inhibitory signaling, when there are no ligands for iNKR to engage, in conjunction with sustained aNKR signaling, or from the engagement of aNKR by their ligands that overwhelms inhibitory signaling through iNKR [8, 10, 11]. However, in vivo, NK cells interacting with target cells will receive a variety of signals through both receptor types and whether this results in activation or not depends on the number and strength of each type of signal transmitted.

A common method to activate NK cells is to co-culture them with human leukocyte antigen (HLA)-null cells. The most frequently used HLA-null cells include the myelogenous leukemia K562 and B-lymphoblastoid 721.221 (.221) cells lines, which do not express the major histocompatibility complex class I (MHC-I) HLA-A, -B, or -C antigens on their surface [12–14]. As these cells are incapable of engaging the inhibitory killer immunoglobulin-like receptors (KIR) on NK cells, which recognize subsets of HLA-A, -B, and -C, inhibitory signaling through these receptors is abrogated. As signals from these inhibitory receptors oppose those from aNKR, their removal allows the engagement of aNKR by their ligands to activate NK cells [15].

Previous work from our lab demonstrated that the K562 and .221 HLA-null cell lines stimulated NK cells differentially to secrete the cytokine interferon-γ (IFN-γ), the chemokine CCL4, and to express the degranulation marker CD107a [16]. Specifically, we observed that .221 activated a greater fraction of NK cells than did K562 and that stimulation with .221 preferentially induced IFN-γ and CCL4 secretion, while K562 more potently stimulated degranulation [16]. In the absence of the expression of the major ligands for iNKR on the surface of K562 and .221, it is likely that ligands to aNKR regulate NK cell activation and differences in their expression profiles might explain their different capacities to activate NK cells. Indeed, the expression of some ligands to aNKR was shown to differ on both cell lines, although these differences remain incompletely characterized [17–22].

CEM.NKr.CCR5 cells are commonly used to stimulate NK cells by antibody dependent cellular cytotoxicity (ADCC) [23, 24]. They were derived from CEM.NKr cells that were selected from the parental CEM T cell line for their resistance to direct NK cell lysis [25]. CEM cells express HLA-A, B and C antigens and the aNKR CD16, whose engagement is important for ADCC activity [25, 26]. The aNKR profile of this cell line is poorly defined.

There are two major classes of aNKR. The first is the C-type lectin NKG2D receptor that binds to a family of MHC-I like molecules expressed on healthy cells only after periods of cellular stress [17, 27]. These include the human cytomegalovirus UL-16 binding proteins (ULBP) and MHC-I related chain (MIC) proteins [19, 28, 29]. The second class of receptors includes the natural cytotoxicity receptors (NCR) NKp30, NKp44, and NKp46, which can bind to membrane-associated heparan sulfate glycosaminoglycans, viral hemagglutinin and β-1,3-glucan [30–40]. Despite these findings, the cellular ligands to NCRs remain incompletely defined.

In addition to the two major aNKR families, signaling through several other receptors can contribute to NK cell activation. These include cluster of differentiation (CD)244/2B4 and the NK-T-B cell antigen (NTB-A), which are CD2 family receptors that engage CD48 to trigger NK cell cytotoxicity [41]. Another receptor, expressed on virtually all NK cells, is the leukocyte adhesion molecule DNAX accessory molecule-1 (DNAM-1) that binds to CD112 (Nectin-2) and CD155 (poliovirus receptor) ligands [18, 42, 43]. Signaling through DNAM-1 can stimulate NK cells when it is co-expressed with the lymphocyte function-associated antigen 1 (LFA-1). LFA-1 can bind the integrins intercellular adhesion molecule (ICAM)-1 and ICAM-2 on target cells, bridging NK and target cells and forming the immunological synapse [44–48]. Additionally, target cells expressing the T cell co-stimulatory B7 molecules CD80 and CD86 can stimulate NK cells [49–51]. The nature of this signaling is still poorly understood, but it has been suggested that NK cell activation by CD80 and CD86 depends on a CD28 variant expressed on NK cells [50, 52].

In contrast to HLA-A, -B, and -C, which signal through iNKRs, the non-classical HLA-E and -F can contribute to NK cell activation through the engagement of aNKRs. Much like its classical MHC-I counterparts, HLA-E interacts with the inhibitory C-type lectin-like receptor NKG2A, which heterodimerizes with CD94, and with the aNKRs NKG2E and -C [53–55]. HLA-F is a more recent addition to the family of ligands to aNKR and has been shown to stimulate degranulation and cytokine production through its binding to KIR3DS1 on NK cells [56–58].

To determine whether the differences observed in the frequency and function of NK cell subsets stimulated by K562, .221 and CEM.NKr.CCR5 cells are related to differences in their expression profiles of ligands to aNKRs, we analyzed the expression of a comprehensive panel of aNKR ligands on these cell lines by multi-parametric flow cytometry. As ligands to inhibitory KIR are not expressed on HLA-null cell lines, it is likely that expression of ligands to aNKRs plays a role in modulating differential NK cell responses to K562 and .221. Although others have proposed that the resistance of CEM.NKr.CCR5 cells to direct NK cell cytolysis is related to changes in cell surface marker expression that occurred when NK cell resistance to cytolysis was selected for, the ligands for aNKR have not been profiled on this cell line [25]. We report here a characterization of the aNKR ligand phenotype of three NK cell stimuli to contribute to a better understand the mechanisms governing their behavior as inducers of NK cell function and targets for NK cells cytolysis.

In this report we screened three cell lines that are frequently used to test NK cell functionality for expression of a panel of aNKR ligands. K562 and .221 cells are both HLA-null cell lines that should have a similar ability to abrogate NK cell inhibitory signals mediated by HLA binding to their iNKR. iNKR-HLA interactions determine NK cell education status and functional potential [60, 61]. Previous work showed that these HLA-null cell lines induced differential patterns of IFN-γ secretion, CCL4 secretion and CD107a expression by NK cells, suggesting that expression patterns for aNKR ligands may differ between these 2 cell lines and explain differential activation patterns. We found that K562 and .221 expression levels did not differ significantly for MIC-A, MIC-B, ICAM-2, HLA-E and NKp46 ligands. K562 expressed significantly higher levels of the NKG2D ligands ULBP-1, ULBP-2/5/6 and ULBP-3 and the ligands for NCRs NKp30 and NKp44. K562 differed from .221 by having significantly lower expression levels of CD48 and lower levels of CD80, and CD86 that did not achieve statistical significance using Dunn’s post tests. On the other hand, the NK resistant cells line CEM.NKr.CCR5, often used as a target cell in ADCC assays, expressed most of the aNKR ligands tested at low expression levels, with the exception of the ligands for CD48, ICAM-2, KIR3DS1 (HLA-F) and NKG2A/C (HLA-E).

Although prior studies have partially characterized the expression profiles of ligands to aNKRs on K562 and .221 cells, to our knowledge this has not been done for the CEM.NKr.CCR5 cell line that differs from its parent CEM cell line by being resistant to direct NK cell cytolysis [25, 62]. In this study, we included a larger and more comprehensive panel of mAbs/chimeric proteins detecting aNKR ligands than have previously been investigated.

The receptors implicated in triggering NK cell mediated cytolysis include NKG2D and the NCRs [62–65]. The ligands for NKG2D include ULBP-1, ULBP-2/5/6, ULBP-3, MIC-A and MIC-B [65]. NKG2D is a C type lectin receptor that associates with signaling molecules such as DAP10/KAP10 to initiate the cascade of events leading to cytolysis [17, 66–68]. The NCRs include NKp30, NKp44 and NKp46, among others [65]. They associate with different tyrosine kinase activating motifs bearing signal transducing polypeptides to mediate activating signals [31, 32]. The characterization of the ligands for these NCRs is incomplete, which is why chimeric proteins based on these receptors are used to probe for the presence of their ligands on target cells. Blocking the interaction of NCRs and NKG2D with these ligands also reduces cytolysis mediated by NK cells, highlighting the role of NKG2D engagement in NK cell killing [62].

In our previous studies comparing the functional profile induced in NK cells by K562 and .221 cell stimulation, we found that while both HLA null cell lines were able to induce CD107a expression, the functional NK subsets that included CD107a expression contributed to a higher proportion of the total NK cell response when stimulated by K562 than by .221 cells [16]. The higher expression level of NKG2D and NCR ligands on K562 than on .221 cells reported here and the lower levels of expression of other ligands for aNKR such as CD48, CD80 and CD86 would be consistent with an aNKR ligand profile that favors NK cell stimulation towards CD107a expression, which is a marker for degranulation, a step in the pathway toward cytolysis. NKG2D signaling has been shown to play a crucial role in NK cell cytotoxic granule polarization, degranulation, and cytotoxicity [69]. The ability of K562 to induce signaling through this pathway may explain why this cell line preferentially stimulates NK cell degranulation, rather than cytokine or chemokine secretion [16]. Unfortunately, no antibody currently exists that can discriminate between ULBP-2, − 5, and − 6 and we are incapable of determining which combinations of these ligands are expressed on K562. However, it is possible that K562 cells express ligands for all three of these receptors, in addition to ligands for ULBP-1 and -3. The consequence of this may be induction of more potent signaling through NKG2D of K652 than .221 cells, which only express MIC-A and MIC-B with a modest MFI. Moreover, expression of NKp30 on NK cells has been correlated with both perforin expression and degranulation. Engagement of this aNKR by K562 may also contribute to their induction of degranulation [70]. CD112 and CD155 bind to DNAM-1 [18, 28, 29]. Thus, K562 has a ligand profile that does not only stimulate NK cells through NKG2D, NKp30 and NKp44 but also through DNAM-1.

Our finding that .221 cells express the ULBP-1, ULBP-2/5/6 and ULBP-3 ligands for NKG2D at a low MFI that is not much above background levels is in line with a report by Pende et al. [62]. Our findings differ for MIC-A, which we found was expressed over background by .221 cells while Pende et al. found it not to be expressed over background. These discrepant results may be due to differences in the reagents used to detect the NKp30 ligand or to other technical issues relating to staining and analysis methods. These findings are consistent with previous reports that ULBP ligands are preferentially expressed on K562 and with the observation that K562 cells express the tumor ligand B7-H6, which is a ligand for NKp30 [21, 62].

The higher expression levels of CD48, CD80, CD86 and ICAM-1 on .221 than K562 cells may explain why .221 cells stimulated a higher frequency of functional NK cells, particularly those secreting IFN-γ and CCL4 than did K652 [16]. When K562 expressed a higher MFI of aNKR ligands than .221 cells, as was in the case of CD112, CD155 and HLA-F, the differences were more modest than were the between-cell differences in CD48, CD80, CD86 and ICAM-1 expression on .221 versus K562 cells. The elevated expression level of these ligands on .221 cells may provide these cells with a more potent activating signal. Engagement of the aNKR 2B4 by its ligand CD48 on .221 has been reported to induce low levels of IFN-γ production and concurrent signaling through 2B4; signaling through other aNKRs can drive high levels of cytokine and chemokine production [9, 20]. Although, the exact NKRs that recognize the co-stimulatory molecules CD80 and CD86 have not been identified, expression of these ligands on target cells can trigger NK cell-mediated cytolysis [50]. Together, the ligands that are expressed on .221 are capable of engaging receptors that are important for cytokine and chemokine production, which may explain why .221 cell stimulate larger numbers of IFN-γ and CCL4 secreting NK cells than do K562 cells [16].

Our findings confirm that HLA-E, the ligand to the aNKRs NKG2E and NKG2C and the iNKR, NKG2A, on NK cells, is expressed by K562 and .221 cells with a low MFI that was nevertheless above background staining levels [71]. On the other hand, CEM.NKr.CCR5 cells express cell surface MHC-I antigens and higher HLA-E levels than do K562 and .221. Stable cell-surface expression of HLA-E requires binding to one of a set of nonamer peptides derived from the leader sequences of MHC-Ia molecules or HLA-G, which are absent on the HLA-null .221 cell surface [72, 73]. It is unlikely that HLA-F expression by K562 and .221 cell lines contributes to HLA-E expression as the signal sequence of HLA-F does not have a nonamer peptide able to bind HLA-E [55]. However, the HLA-E peptidome was recently shown to be less restricted than previously thought. HLA-E can also bind to an array of self-peptides in the absence of HLA class I signal peptides, permitting its stable expression and induction of NK cell cytotoxicity [74, 75]. In addition, HLA-E is also capable of presenting EBV-derived peptides, such as BZLF1, which would be expected to be present in the EBV transformed .221 cell line [76, 77]. HLA-E was recently shown to present a cytomegalovirus derived signal peptide important in driving the expansion of adaptive-like NKG2C+ NK cells [78]. HLA-E/BZLF1 complexes are poorly recognized by NK cells and it is likely that HLA-E molecules presenting non-canonical self, make greater contributions to .221-induced NK cell activation [76].

The non-classical MHC-I antigen, HLA-F, is a KIR3DS1 ligand, which is expressed by K562, .221 and CEM.NKr.CCR5 cells [56–58]. HLA-F has also been reported to interact with KIR3DL2 and KIR2DS4 [79]. Although the interaction of HLA-F with KIR3DL2 was confirmed by others, its interaction with KIR2DS4, which is structurally related to KIR3DL2 due to a gene conversion event has not been confirmed [56, 80]. HLA-F is also expressed on HIV infected cells [56]. Other investigators did not observe HLA-F on K562 [56, 57]. Here, we used both the mAb 3D11 and KIR3DS1-Fc to stain K562 cells for cell surface HLA-F. Both reagents generated concordant results for the presence of HLA-F on this cell line.

KIR3DS1 homozygotes were more frequent in a population of HIV exposed seronegative than in HIV susceptible individuals and KIR3DS1 homozygotes remained uninfected for longer time intervals despite HIV exposure than those with other KIR3DL1/S1 genotypes, suggesting that KIR3DS1 HLA-F interactions may provide protection from HIV infection [81, 82]. The global distribution of KIR3DS1 varies from one population to another [83, 84]. For example, it is rare in sub-Saharan African populations [83]. It is interesting to speculate on whether HLA-F/KIR3DS1 or /KIR3DL2 or possibly /KIR2DS4 combinations can influence HIV control mediated by NK cells and whether this could account for between-individual or -population differences in HIV susceptibility or the rate of HIV disease progression.

For the purpose of this study, the ligands analyzed were included on the basis of their ability to stimulate NK cell responses through the engagement of aNKRs. However, it is important to consider that several of these ligands are capable of engaging both aNKRs and iNKRs. CD112 and CD155, which signal through the activating DNAM-1, can also bind to the iNKR, T cell immunoreceptor with immunoglobulin and ITIM motifs (TIGIT) [85, 86]. While both DNAM-1 and TIGIT are widely expressed on NK cells, the affinity of CD155 for TIGIT is greater than for DNAM-1 and TIGIT expression can reduce DNAM-1/CD155 interactions in a dose-dependent manner [87–89]. TIGIT has also been shown to compete with DNAM-1 for the binding of CD112. Furthermore, when transfected into the NK cell line YTS, TIGIT greatly limits NK-mediated cytotoxicity by disrupting cytotoxic granule polarization [89, 90]. Considering this, it is possible that CD112, which is exclusively expressed on K562, and CD155 which is expressed at higher levels on K562 than .221 cells contributes more to NK cell inhibition than activation and may be an additional reason why K562 activates a smaller fraction of NK cells, compared to .221 [16]. Another aNKR ligand, HLA-E, similarly contributes to both NK cell activation and inhibition. HLA-E binds to the CD94/NKG2 family of NK cell receptors, which includes the activating NKG2E and -C and the inhibitory NKG2A and -B NKRs [53, 54]. Interactions between NKG2A, which is expressed on the majority of NK cells, and HLA-E have been shown to predominate over interactions with NKG2C and surface expression of HLA-E is sufficient to rescue those cells from lysis by NKG2A⁺ NK cells [53, 54, 91]. Despite this, work assessing NK cell stimulation by autologous HIV-infected CD4 T cells, which express HLA-E, found that expression of NKG2A on NK cells was associated with improved activation [92]. It is plausible that, while interactions between HLA-E on .221 and NKG2A on NK cells can tune down NK cell activation, signaling through the other aNKR engaged by .221 ligands can compensate for this inhibitory input.

CEM.NKr.CCR5 cells expressed few aNKR ligands. Ligands for CD48 and ICAM-2 were present on this cell line as well as low levels of ligands for NCRs. The interaction of ICAM-2 with its ligand may contribute to the formation of NK cell CEM.NKr.CCR5 conjugates [65]. Less is known regarding the consequences of CD48 ligand expression. It is interesting to speculate that the presence of few other cell surface ligands for aNKR contributes to the resistance of this cell line to direct NK cell cytolysis in the absence of an antibody bridging target and effector cells [25]. Pende et al., tested the CEM.NKr.CCR5 parental cell line, CEM, for cell surface expression of ULBP-1. ULBP-2, ULBP-3 and MIC-A and found that CEM cells expressed ULBP-2 and ULBP-3 [62]. The absence of these aNKR on CEM.NKr.CCR5 is suggestive that the process of selecting for CEM.NKr resistance to direct NK cell cytolysis led to loss of several ligands for aNKR [25].

The two HLA-null cell lines K562, .221 and the ADCC target cell CEM.NKr.CCR5 differed in their expression of ligands to aNKR. The data presented here provide a systematic assessment the stimulatory potential of three cell types commonly used to study NK cell activation. The different aNKR ligand expression profiles of K562, .221 and CEM.NKr.CCR5 are associated with the induction of qualitative differences in the NK cell responses to these cell lines. CEM.NKr.CCR5 cells, that are resistant to direct NK cell killing express few ligands for aNKR. This work provides a basis for examining the specific contribution of each ligand-aNKR pair to different stages of NK cell activation.