CD155 density on target cells drives divergent natural killer cell responses owing to DNAM-1 loss

Abstract

Activation of natural killer (NK) cells by transformed targets requires the engagement of activating receptors and commonly a concomitant loss of inhibitory signaling. While the coactivating receptor DNAM-1 is involved in NK cell recognition of tumors expressing the nectin-like molecule CD155, in many cancer types, high expression of CD155 is associated with a poor clinical prognosis. To assess the impact of ligand density on NK cell function, recognition of target cells expressing low and high levels of CD155 was compared. While low levels of CD155 on target cells augmented NK cell activation, most evident in elevated IFN-γ responses, high levels of CD155 drove rapid, activation-induced downregulation of DNAM-1 that did not result in enhanced IFN-γ responses. High levels of CD155 also drove TIGIT loss from the NK cell surface; however, gene editing of TIGIT from primary NK cells had no significant impact on target cell recognition. Although low levels of CD155 induced a degree of DNAM-1 loss, the improved activation was still mediated through DNAM-1, as assessed through selective disruption of the interaction. Similarly, NK cell recognition of targets expressing the nectin CD112, which also engages both DNAM-1 and TIGIT, resulted in modest loss of DNAM-1 expression and heightened IFN-γ responses. Together, these observations suggest that DNAM-1 recognition of its ligands is delicately poised such that low avidity receptor engagement augments NK cell activation, and in particular IFN-γ responses, while stronger engagement can result in the rapid downregulation of DNAM-1 without any significant enhancement of effector responses.

Graphical abstract.

Introduction

The capacity for natural killer (NK) cells to respond to aberrant cells without prior antigenic sensitization makes them an important component of the immune arsenal in the fight against cancer. This is accomplished by integrating signals received from a myriad of germ-line encoded activating and inhibitory receptors on NK cells. Under healthy conditions, inhibitory receptors primarily recognizing human leukocyte antigen (HLA) class I molecules prevent target cell–induced NK cell activation. However upon HLA downregulation, as can occur in cancer,¹ this inhibitory signal is lost and activating signals can cue destruction of the target cell and the secretion of both cytokines and chemokines.² Activating receptors recognize a broad array of ligands and vary in their signaling potential. The requirement for synergism between diverse coactivating receptors effectively exerts significant control over the generation of effector responses.³ Among these, variable contributions by DNAX accessory molecule 1 (DNAM-1; also known as CD226) and NKG2D, as well as the natural cytotoxicity receptors NKp46 and NKp30, have been implicated in the anti-tumor immune responses of NK cells, the importance of which varies in accordance with the profile of ligands expressed on target cells.⁴

The coactivating receptor DNAM-1 is expressed on T cells, NK cells, monocytes and subsets of B cells⁵ and augments NK cell activation through recognition of the nectin molecule CD112 (Nectin2) and the nectin-like (necl) molecule CD155, also known as the poliovirus receptor.^6,7 In the absence of ligands for NKG2D, DNAM-1 has been shown to play an important role in the elimination of tumors by NK cells in vivo⁸ as well as in NK cell–mediated killing of cancer lines including myeloma and ovarian carcinoma cells.^9,10 Notably, positive correlations have been observed between the level of ligand expression on neuroblastoma cells and acute myeloid leukemia cell lines, and their susceptibility to NK cell lysis.^11,12 Despite this observation, CD155 has also been reported to drive DNAM-1 downregulation on tumor-infiltrating NK cells, leading to impaired functional responses.¹³ Indeed, CD155-induced downregulation of DNAM-1 has been described as an immune escape mechanism, where CD155 engagement with DNAM-1 induces the phosphorylation, ubiquitination, and degradation of DNAM-1.¹⁴

Four other receptors share specificity for CD112 and/or CD155. The inhibitory receptor T cell immunoglobulin and ITIM motif (TIGIT) binds both ligands, along with CD113 and Nectin4, and is also expressed on NK cells and T cell subsets,^15,16 while the inhibitory poliovirus receptor–related immunoglobulin domain containing receptor (PVRIG or CD112R) specifically recognizes CD112.¹⁷ Both inhibitory receptors represent potential immune checkpoint targets due to their capacity to modulate T and NK cell function.^18,19 To date, in clinical settings, TIGIT blockade has had little impact when used as a monotherapy but has been shown to have some efficacy when used in combination with antibodies toward additional receptors such as PD-1. Critically, in these settings its impact has been dependent on the presence of DNAM-1, possibly reflecting the capacity of TIGIT to disrupt the interaction between DNAM-1 and CD155 owing to their overlapping binding sites.¹⁸ The inhibitory killer cell immunoglobulin-like receptor (KIR) 2DL5 also recognizes CD155, with evidence suggesting that it binds at a site distinct from DNAM-1 and TIGIT, and antibody blockade of KIR2DL5 can increase KIR2DL5⁺ NK cell anti-tumor responses.²⁰ Finally, the adhesion molecule CD96 (or T-cell–activated increased late expression [TACTILE]) also recognizes CD155, as well as the nectin CD111, but has been shown to potentiate both inhibitory and activating signals in different contexts.^21,22 Understanding how activating signals through DNAM-1 can be achieved or enhanced through this slew of coexpressed inhibitory receptors with overlapping ligand specificities in the setting of cancer may therefore be of importance to manipulate this axis therapeutically.

Overexpression of CD155 has been associated with poor prognoses in a variety of cancers including melanoma and pancreatic cancer.^23–25 This typically coincides with high levels of TIGIT expression on tumor-infiltrating lymphocytes^26,27 and decreased DNAM-1 expression.^13,14 TIGIT was originally reported to have a higher affinity for CD155 than DNAM-1,²⁸ leading to the long-held theory that binding to inhibitory TIGIT would outcompete binding to coactivating DNAM-1 on healthy cells, but upon transformation, CD155 could be upregulated along with other activating ligands to allow for costimulation via DNAM-1.²⁹ More recent studies, however, report DNAM-1 and TIGIT to share similar affinities for CD155,^30,31 suggesting that rather than direct ligand competition significantly regulating NK cell function, the level of receptor occupancy required for signaling or the requirement for additional receptors working in tandem are likely to be key factors.

To determine how the expression level of necl and nectin molecules impacts receptor expression on NK cells and subsequent NK cell function, target cells expressing low or high levels of ligand were generated. Here, high levels of CD155 expression were shown to rapidly downregulate DNAM-1 on NK cells, which functionally restricted their response to CD155-expressing targets. However, moderate downregulation of DNAM-1 as induced by lower levels of CD155 or by CD112 engagement was tolerated and increased IFN-γ responses from NK cells. Antibody blocking experiments and gene editing of primary NK cells confirmed this activation to be DNAM-1 dependent, suggesting that there exists a delicate window of CD155 expression that, while moderately down-regulating DNAM-1, still enables optimal receptor signaling.

Materials and methods

Constructs, cell lines, and transfectants

Gene fragments encoding CD155, CD112, and CD111 were synthesized (GeneArt Gene Synthesis, Invitrogen) and cloned into the pEF6/V5-His-TOPO vector. Constructs were electroporated into the HLA-deficient 721.221 (221) B lymphoblastoid cell line, as well as K562 and 293T cells (CRISPR edited for endogenous CD155/CD112 expression), P815 cells, and 221 cells expressing HLA-B* 51:01³² at 250 V and 975 μF and placed under 20 μg/mL blasticidin (InvivoGen) selection. Transfected cells were sorted for the expression of low or high levels of CD155, CD112, and CD111 by staining with anti-CD155 (clone SKII.4, AF647/PE, BioLegend), anti-CD112 (clone TX31, PECy7/PE, BioLegend), or anti-CD111 (clone R1.302, APC, Miltenyi Biotec) or, for HLA, with anti-HLA-ABC (W6/32, AF647, BioLegend; or G46-2.6, PECy7, BD Pharmingen), along with Fixable Viability Dye eFluor 780 (eBioscience). To generate cell lines with inducible CD155, a doxycycline (DOX)–inducible promoter was incorporated by cloning CD155 through the pTRE-tight vector into pFUV1GFP (both generously provided by Macro Herold, Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia). Constructs were introduced into the 221 cell line via lentiviral transduction as previously described,³³ and sorted for GFP and CD155 expression upon DOX treatment. All 221, K562, and P815 cells, edited cells, and transfectants were maintained in RPMI 1640 media with 10% fetal bovine serum (FBS) and supplements with or without selection. Human embryonic kidney 293T cells were grown in DMEM with supplements and 10% FBS.

Primary NK cell isolation

Peripheral blood mononuclear cells (PBMCs) were isolated from healthy blood donors (Australian Red Cross Lifeblood) with the approval of the University of Melbourne Human Research Ethics Committee using Ficoll/Hypaque (Cytiva) density gradient centrifugation and cryopreserved in FBS with 10% DMSO. Following thawing, primary NK cells were purified using the EasySep Human NK Cell Enrichment Kit (Stemcell Technologies) and allowed to rest overnight in 100 U/mL recombinant human IL-2 (Mitenyi Biotec) before use. NK cells were then either used directly in functional assays or subjected to CRISPR/Cas9 gene editing and expansion before assaying.

Functional NK cell assays

Target cells and purified NK cells, or expanded CRISPR/Cas9-edited NK cells, were incubated at a 1:1 ratio in the presence of anti-CD107a (clone H4A3, PECy5/BV786, BD Pharmingen/BD Horizon) for 1 hour before the addition of monensin (GolgiStop, BD). Four hours later, NK cells were surface stained for CD56 (clone NCAM16.2, APC/BV421, BD/BD Horizon) and CD3 (clone SK7, APCCy7, BD Pharmingen), along with Fixable Viability Dye eFluor 780, and, where indicated, DNAM-1 (clone 11A8, FITC, BioLegend), TIGIT (clone A15153G, PECy7/PerCPCy5.5, BioLegend), NKG2D (clone BAT221, PE, Miltenyi Biotec), CD96 (clone NK92.39, PE, BioLegend), KIR2DL5 (clone UP-R1, PE, Miltenyi Biotec), and/or KIR3DL1 (clone DX9, FITC, BD Pharmingen). Cells were fixed with 4% paraformaldehyde before intracellular staining for IFN-γ (clone B27, AF700, BD Pharmingen) and, where indicated, CCL4 (clone D21-1351, PE, BD Pharmingen) and then analyzed by flow cytometry using a BD FACSCanto II or Fortessa. Data were analyzed with FlowJo software, gating on live, CD3⁻CD56⁺ NK cells and analyzed with GraphPad Prism software. Secreted IFN-γ was detected using the BD OptEIA Human IFN-γ ELISA set. In brief, purified NK cells were incubated with target cells at a 1:1 ratio for 0, 1, 3, 5, or 24 hours at 37 °C, 5% CO₂ in the presence of IL-2. Cells were pelleted and the supernatant collected at the indicated timepoints and applied to blocked plates coated with capture antibody followed by detector antibody. IFN-γ production was detected using the BD OptEIA TMB Substrate Reagent Set, measuring at 450 nm.

For pretreatment regimens, NK cells were incubated with dasatinib (Sigma) or Dynasore (Dynamin Inhibitor I, Sigma) at the indicated concentrations for 30 minutes at 37 °C, 5% CO₂ prior to the addition of target cells. For blocking assays, NK cell receptors were blocked with anti-DNAM-1 (clone DX11, BD Pharmingen), anti-TIGIT (clone MBSA43, Invitrogen), anti-NKG2D (clone 1D11, BD Pharmingen), or anti-KIR3DL1 (clone DX9, BD Pharmingen), or ligands on target cells blocked with anti-CD155 (clone SKII.4, BioLegend) or anti-HLA class I (clone W6/32, BioLegend) at 10 μg/mL for 30 minutes at 37 C, 5% CO₂ before coincubation with targets or NK cells, respectively. For redirected assays, P815 cells were incubated with purified anti-CD244 antibody (2B4, clone C1.7, BioLegend) at 2.5 to 5 μg/mL for 15 minutes at room temperature before the cells were pelleted and then coincubated at a 1:1 ratio with NK cells. To induce CD155 expression, lentivirally transduced targets were treated with 1 μg/mL DOX for at least 48 hours, after which cells were transferred to DOX-free media for 1 to 5 days to allow CD155 expression to progressively drop.

NK cell receptor expression recovery

Purified NK cells were incubated at a 1:1 ratio with target cells for 3 hours before sorting on live, CD56⁺CD3⁻ NK cells to remove target cells. NK cells were returned to culture in media containing 100 U/mL recombinant human IL-2 at 37 °C, 5% CO₂. Cells were stained for receptor expression immediately upon mixing with target cells and after the 3 hours coincubation, and then 18 or 90 hours postsorting. Cells were analyzed by flow cytometry using a BD Fortessa, assessing DNAM-1, NKG2D, and TIGIT expression on NK cells via FlowJo and GraphPad Prism software.

CRISPR/Cas9 editing

DNAM-1 in parental 221 and CD155-transfected cell lines, as well as CD155 and CD112 from K562 and 293T cells, was edited using CRISPR/Cas9 technology. In brief, ribonucleoprotein (RNP) complexes were formed between synthetic guide RNA (sgRNA) (CD226⁺69947029 and CD226^-69947061; PVR⁺44647235 and PVR^-44647223; NECTIN2^-44865272 and NECTIN2^-44865280 from Synthego) and Cas9 protein (Integrated DNA Technologies) and then added to 2 × 10⁵ cells resuspended in Nucleofector Solution SF and Supplement 1 (Lonza) before nucleofection with the 4D-Nucleofector X Unit, program CM130. Cells were sorted where necessary to achieve DNAM-1 or CD155/CD112 null cell lines. For editing of primary NK cells, 1 × 10⁶ NK cells resuspended in P3 Nucleofector media and Supplement 1 (P3 Primary Cell 4D-Nucleofector X Kit S, Lonza) were incubated with Cas9 protein and sgRNA before nucleofection with the 4D-Nucleofector X Unit, program CM137. The sgRNAs used for NK cells were DNAM-1: CD226⁺69947029 and CD226^-69947061; TIGIT: TIGIT⁺114295552 and TIGIT^-114295545; CD96: CD96⁺111545092 and CD96^-111545121; and NKG2D: KLRK1^-10386955 and KLRK1^-10386987. NK cells were allowed to recover for 4 days at 37 °C, 5% CO₂ with 100 U/mL IL-2 before sorting for receptor negative populations, staining for CD56, CD3, DNAM-1, TIGIT, NKG2D, and/or CD96, along with Fixable Viability Dye eFluor 780. Sorted NK cells were then plated onto irradiated PBMCs and 221 cells transfected with HLA-G, along with 100 U/mL IL-2 and 1.5 ng/mL phytohemagglutinin (Gibco Ltd) as described previously.³⁴ Cells were expanded for 2 to 3 weeks before use in functional NK cell assays.

Tetramer staining

Biotinylated DNAM-1, TIGIT, and CD96 recombinant proteins (Fc, Avitag) were purchased from ARCO Biosystems and tetramerized with streptavidin (SA)–PE (Invitrogen). The 221 transfectants were stained with 3-fold tetramer dilutions (or SA-PE only) at room temperature for 30 minutes along with Fixable Viability Dye eFluor 780, before fixing with 2% paraformaldehyde and analyzing by flow cytometry.

Results

DNAM-1 downregulation inversely correlates with CD155 expression

Previous studies have shown that recognition of CD155 can downregulate DNAM-1 expression on both NK and CD8⁺ T cells.^13,14 To investigate this in more detail, 721.221 (221) cells, which do not express endogenous CD155 or CD112, were transfected with CD155, and cell lines expressing low or high levels of the necl were established (Fig. 1A). Being of B-cell origin, parental 221 cells express DNAM-1 and therefore the gene was disrupted by transfection of sgRNA and Cas9 protein into both 221 and CD155 transfectants and the disruption of DNAM-1 confirmed by flow cytometry (Fig. S1A). Purified NK cells were then incubated with DNAM-1 knockout (DNAM KO) 221, 221 + CD155lo, and 221 CD155hi target cells for 5 hours and the expression of DNAM-1, TIGIT, and NKG2D on NK cells was assessed by flow cytometry (Fig. 1B; Fig. S2A). In the absence of target cells or with 221 DNAM KO cells, primary NK cells expressed similar levels of DNAM-1 (Fig. 1Bi). However, following incubation with target cells expressing low or high levels of CD155, there was a marked reduction in cell surface DNAM-1 expression, which was slightly more pronounced with higher levels of CD155 expression. The addition of an anti-CD155 monoclonal antibody (mAb), but not an irrelevant anti-class I antibody (W6/32), restored DNAM-1 expression. Expression of TIGIT was similarly impacted by the presence of CD155 on target cells where again greater downregulation was observed upon culture with 221 + CD155hi cells (Fig. 1Bii). The expression of NKG2D decreased when NK cells were incubated with 221 cells or any of the variants used here, and the extent of this was unaffected by the expression of CD155 (Fig. 1Biii).

Figure 1. Increasing CD155 expression on target cells drives DNAM-1 downregulation on NK cells.

721.221 (221) cells were transfected with CD155, and low- or high-expressing necl cell lines established, with endogenous DNAM-1 expression disrupted via CRISPR-Cas9 editing. (A) Transfectants were stained for CD155, HLA class I, and DNAM-1 surface expression and analyzed by flow cytometry (221 DNAM KO: orange, 221 + CD155lo DNAM KO: dark blue, 221 + CD155hi DNAM KO: light blue; unstained 221 DNAM KO: gray). (B) Transfectants were incubated with anti-CD155 (red), anti-HLA class I (orange) (10 μg/mL), or no antibody (blue) before coculturing with purified NK cells (1:1). The mean fluorescence intensity (MFI) of DNAM-1 (i), TIGIT (ii), and NKG2D (iii) expression on live, CD3⁻CD56⁺ NK cells was assessed by flow cytometry after 5 hours (6 donors across 3 independent experiments). (C) Purified NK cells were treated with increasing concentrations of dasatinib before coincubation with transfected target cells. Expression of DNAM-1 (i), TIGIT (ii), and NKG2D (iii) was assessed 5 hours later and normalized to levels seen on NK cells in the absence of targets (9 donors across 3 independent experiments). Error bars represent the SEM and data were analyzed via a 2-way ANOVA with Dunnett multiple comparisons test (****P < 0.0001, other significances as noted).

To determine whether loss of DNAM-1 from primary human NK cells in the presence of CD155 was dependent on activation, NK cells were treated with the tyrosine kinase inhibitor dasatinib before incubation with target cells. In the presence of 221 + CD155lo targets, dasatinib treatment prevented DNAM-1 downregulation while with high CD155-expressing targets, DNAM-1 remained at approximately 74% of that detected on NK cells in the absence of targets (Fig. 1Ci). TIGIT loss with low CD155-expressing targets was similarly prevented upon dasatinib treatment; however, TIGIT downregulation was not affected by dasatinib treatment in the presence of 221 + CD155hi targets (Fig. 1Cii). NKG2D expression on NK cells was also retained upon dasatinib treatment (Fig. 1Ciii). Loss of DNAM-1, TIGIT, and NKG2D was not a dynamin-dependent process as use of the dynamin inhibitor Dynasore had only marginal impact on receptor expression following incubation with CD155-expressing targets as previously reported,¹⁴ as did DMSO vehicle controls (Fig. S2B, C). Together these results demonstrate that DNAM-1 and TIGIT are downregulated on primary NK cells upon engagement with CD155 and that higher levels of CD155 increase the degree of receptor loss from the cell surface.

Low but not high levels of CD155 expression enhance NK cell activation

CD155 expression has been associated with both increased and decreased NK cell responses toward tumors.^11,13 To assess the extent to which the expression level of CD155 impacted NK cell function in vitro, the proportion of primary NK cells that degranulated and produced IFN-γ following incubation with 221 DNAM KO cells and those expressing CD155 was determined. While the 221 cells were able to activate NK cells owing to their lack of inhibitory HLA class I molecules, coculture with target cells expressing low levels of CD155 augmented this activation, with the enhancement being more noticeable for the IFN-γ response compared with degranulation (Fig. 2A; Fig. S3A). In contrast, this augmented NK cell activation was not observed following incubation of NK cells with target cells expressing high CD155 levels, where the responses resembled those seen with untransfected 221 cells. This effect was CD155-dependent as the addition of a CD155-specific mAb abrogated the enhanced response with 221 + CD155lo targets, and drove responses seen with 221 + CD155hi targets further below 221 activation levels. Blocking with anti-HLA class I had no effect on NK cell responses while dasatinib treatment prevented NK cell activation (Fig. S3B).

Figure 2. CD155-mediated DNAM-1 loss prevents enhanced NK cell activation.

(A) 221 and CD155-expressing 221 transfectants, CRISPR edited for DNAM-1 expression (DNAM KO), were incubated with anti-CD155 (red), anti-HLA class I (orange) (10 μg/mL), or no antibody (blue) and then mixed with purified NK cells in the presence of monensin. The degranulation (CD107a) (i) and production of IFN-γ (ii) by live, CD3⁻CD56⁺ NK cells was assessed by flow cytometry after 5 hours (6 donors across 3 independent experiments). Data were analyzed using a 2-way ANOVA with Dunnett multiple comparisons test (***P < 0.0001, other significances as noted). (B) Representative dot plots depicting NK cell production of IFN-γ and expression of DNAM-1 in the presence of CD155-expressing transfectants. (C) CD155 was lentivirally transduced into 221 DNAM KO cells under a doxycycline (DOX)–inducible promoter. Following DOX treatment, transduced cells were transferred to media in the absence of DOX for 1 to 5 days after which CD155 surface expression was assessed by flow cytometry and normalized to the maximal expression observed on 221 + CD155hi targets (i). Target cells, one to 5 days after DOX withdrawal, were incubated with primary NK cells and the resulting expression of DNAM-1 and TIGIT (ii) or degranulation and IFN-γ production (iii) was assessed by flow cytometry. Data from 6 donors across 2 independent experiments. Receptor surface expression was normalized to that observed on NK cells in the absence of targets, while functional responses are represented as the fold change to responses toward parental 221 cells (errors bars represent SEM, and data analyzed via an ordinary one-way ANOVA with Dunnett multiple comparisons test to 221 + LentiCD155 plus DOX, with only significances across the DOX timecourse noted).

The production of IFN-γ in the presence of 221 + CD155lo targets was confined to NK cells with downregulated DNAM-1 expression (Fig. 2B). Upon coincubation with 221 + CD155hi targets, DNAM-1 expression was lost without an associated increase in IFN-γ production. This difference in IFN-γ production was also detected in ELISA assays where IFN-γ secretion in the presence of low CD155-expressing targets over time outstripped that seen with parental or 221 + CD155hi targets (Fig. S3C). These results suggest that the events associated with cellular activation and receptor endocytosis may operate at different ligand thresholds.

The divergent effects of CD155 levels on NK cell activation suggested that the response may be Gaussian, based on CD155 expression. To examine NK cell recognition across a broader range of CD155 levels, a 221 DNAM KO cell line in which CD155 expression could be induced by culture with DOX was established. Treatment with DOX induced CD155 expression on these cells to 50% of that seen on stable 221 + CD155hi cells, which was approximately double the level observed on 221 + CD155lo cells (Fig. 2 Ci). When DOX was removed from these cells, the level of CD155 gradually dropped over 5 days. When incubated with primary NK cells, the loss of CD155 on target cells inversely correlated with the expression of DNAM-1 and TIGIT on NK cells (Fig. 2Cii). Although ligand expression on target cells 3 days post–DOX withdrawal resembled that of 221 + CD155lo cells and drove similar downregulation of DNAM-1 and TIGIT on NK cells, the IFN-γ response of NK cells was not augmented to the same degree as by 221 + CD155lo cells. This may reflect the breadth of CD155 expression at any given timepoint post-DOX withdrawal in the DOX-inducible system relative to the narrower and more homogeneous expression patterns on the cell lines that constitutively expressed CD155 (Fig. S3Di). Nevertheless, in comparison to cells where CD155 expression was maintained by the presence of DOX, those with decreased CD155 levels induced higher proportions of IFN-γ⁺ NK cells, particularly between 3 and 5 days post–DOX withdrawal (Fig. 2Ciii). These effects were not due to the presence of DOX, as treatment of untransduced cells with DOX had no effect on NK cell receptor expression or activation (Fig. S3Dii).

To determine whether the divergent effects of CD155 expression levels could be recapitulated with other target cells, CD155 expression was experimentally manipulated in the chronic myelogenous leukemia cell line K562 and the human embryonic kidney cell line 293T. As these cell lines endogenously express CD155 and CD112, the expression of both molecules was first eliminated via CRISPR-Cas9 gene editing. These cells were then transfected with plasmids encoding CD155 creating K562 + CD155lo and K562 + CD155hi cell lines as well as a 293T line that expressed high levels of CD155 (Fig. S4A). Upon coincubation, the expression of DNAM-1 and TIGIT on NK cells was significantly reduced in presence of high CD155-expressing targets when compared to CD155-deficient targets (Fig. S4B). Notably the expression of DNAM-1 and TIGIT was unchanged in the presence of K562 cells expressing low CD155 levels, whereas parental K562 cells decreased DNAM-1 expression, but not TIGIT. When the functional responses of NK cells were assessed, the presence of wild-type CD155 levels augmented IFN-γ production as for low CD155-expressing 221 cells, whereas higher levels of CD155 impeded such activation (Fig. S4C). Correlation analyses between CD155 expression and DNAM-1 expression or IFN-γ production confirmed that while DNAM-1 levels inversely mirror CD155, the production of IFN-γ does not follow this trend (Fig. S4D).

Since each target cell expresses additional activating ligands for NK cells, the impact of CD155 levels was also examined in a more reductionist redirected assay. The Fc-expressing murine P815 cell line was transfected with human CD155 to establish different expression levels (Fig. 3A) and then cocultured with primary NK cells preincubated with anti-receptor antibodies to crosslink coactivating receptors. As expected, the extent of DNAM-1 loss increased with CD155 levels whereas TIGIT was only downregulated with P815 + CD155hi targets (Fig. 3B). In the absence of costimulation, DNAM-1 engagement, either via CD155 or anti-DNAM-1 antibodies, failed to stimulate IFN-γ production or degranulation (Fig. 3C). In the presence of anti-2B4 antibodies, however, low and medium but not high CD155-expressing P815 cells activated NK cells, thus further demonstrating that CD155-dependent DNAM-1–mediated activation requires synergism with additional receptors such as the coactivating receptor 2B4. Together, these results indicate that there exists a complex nonlinear relationship between CD155 expression and its capacity to impact NK cell activity, where CD155 can augment NK cell function only when expressed at moderate levels and that these augmented responses are only evident across a relatively narrow range of CD155 expression.

Figure 3. High CD155 expression hinders NK cell activation in a reductionist setting.

(A) Representative histograms depicting the surface expression of CD155 on P815 cells transfected to generate low (lo), medium (med), and high (hi) CD155-expressing targets. (B, C) P815 cells and transfectants were incubated with anti-receptor antibodies (αDNAM-1 and/or α2B4) before coincubation with primary NK cells. After 5 hours, receptor surface expression was assessed (B), staining for DNAM-1 (i), TIGIT (ii), and NKG2D (iii), and normalizing their expression to that seen on NK cells in the absence of targets, and degranulation (CD107a) (i) and production of IFN-γ (ii) by NK cells detected (C). Data represent 6 donors across 3 independent experiments, analyzed with an ordinary one-way ANOVA with Šídák multiple comparisons test, comparing to either 221 DNAM KO targets or P815 targets in the absence of antibody (****P < 0.0001, other significances as noted). Error bars represent the SEM.

KIR2DL5⁺ NK cells are inhibited through engagement with CD155

The inhibitory receptor KIR2DL5 has also been reported to bind CD155 and prevent NK cell activation.²⁰ To determine whether KIR2DL5 engagement contributed to the lack of NK cell augmentation seen with 221 + CD155hi targets, IFN-γ production by KIR2DL5⁺ and KIR2DL5⁻ NK cells upon target cell encounter was examined (Fig. 4Ai, Aii). Notably, DNAM-1 expression in these targets was not edited; however, functional analyses demonstrated that endogenous expression of DNAM-1 on 221 cells had no impact on NK cell recognition when compared with DNAM-1–edited versions of the target cell lines (Fig. S1B). KIR2DL5⁺ NK cells responded robustly to 221 cells, but their activation was inhibited by the expression of CD155. In contrast, KIR2DL5⁻ NK cells again demonstrated increased activation upon stimulation with cells expressing low levels of CD155. Despite the potent inhibition observed through KIR2DL5, only a small subpopulation of NK cells expressed this inhibitory KIR (approximately 4%) such that while KIR2DL5⁺ NK cells may have contributed to NK cell inhibition in the presence of CD155, their impact on the overall activation or inhibition of the total NK cell population in most donors is likely minor (Fig. 4Aiii). Notably, upon coincubation with high CD155-expressing targets, KIR2DL5 surface expression was also decreased. Recognition of HLA class I interaction through more canonical inhibitory KIR also overrode CD155-dependent activation, with coexpression of HLA-B* 51:01 and low levels of CD155 efficiently inhibiting KIR3DL1⁺ NK cells, with KIR3DL1⁻ NK cells again having elevated responses to target cells expressing low levels of CD155 (Fig. 4B). These findings affirm that inhibitory signaling via KIR, either through KIR2DL5 binding to CD155 or more classical binding of KIR to cognate HLA, can override DNAM-1–mediated NK cell activation, and that the variable expression of KIR2DL5 can further diversify NK cell responses to CD155 across different donors.

Figure 4. Inhibitory KIR engagement overrides DNAM-1–mediated activation signals.

(A) Purified NK cells from donors expressing KIR2DL5 were coincubated with transfected target cells. Representative gating of KIR2DL5⁺ NK cells from live CD3^-CD56⁺ NK cells, and the comparative production of IFN-γ and degranulation (CD107a) by KIR2DL5⁻ and KIR2DL5⁺ subsets upon coincubation with transfected targets (i). The subsequent production of IFN-γ by KIR2DL5⁻ (purple) and KIR2DL5⁺ (orange) subsets (ii) and the percentage of KIR2DL5⁺ NK cells (iii) was assessed by flow cytometry (5 donors across 2 independent experiments). Error bars represent the SEM, analyzed via a 2-way ANOVA with Dunnett multiple comparison test. (B) 221 cells were cotransfected with HLA-B* 51:01 and CD155, at high and low levels. (i) Representative histograms depicting the expression of CD155 and HLA class I on transfected cell lines. (ii) Fold change in the production of IFN-γ by KIR3DL1^- (blue) and KIR3DL1⁺ (red) NK cell subpopulations upon coincubation with target cell panel normalized to responses against parental 221 cells (6 donors across 2 independent experiments). Error bars depict the SEM and data was analyzed via a 2-way ANOVA with Tukey multiple comparisons test (****P < 0.0001, other significances as noted).

Nectin and necl expression levels differentially impact NK cell function

In addition to CD155, DNAM-1 and TIGIT both recognize CD112.^7,15 To determine whether varied expression of CD112 also resulted in divergent NK cell responses, 221 cells expressing low and high levels of CD112 were generated (Fig. 5Ai). Since CD111 does not bind either DNAM-1 or TIGIT, control 221 transfectants expressing low and high levels of the nectin CD111 were also established.³⁵ These cells were not CRISPR edited for DNAM-1 and consequently endogenous cell surface levels of DNAM-1 on 221 cells varied when the nectin and necls were coexpressed, with the lowest levels observed on 221 + CD155hi targets, followed by high CD112-expressing cells and low CD155-expressing cells (Fig. 5Aii). This decreased DNAM-1 expression may be driven by trans interactions with adjacent cells or via cis interactions on the same cell, where structural analyses of CD155 and DNAM-1 suggest such interactions are possible.³¹

Figure 5. Differential response of NK cells to nectin and necl levels.

(A) 221 cells were transfected with CD155, CD112, or CD111 and high- or low-expressing cells established. Representative histograms following staining for CD155, CD112, and CD111 on transfectants (i). The 221 transfectants were stained for DNAM-1 (ii) and the mean florescence intensity (MFI) was compared using a 2-way ANOVA with Dunnett multiple comparison test (4 independent experiments). SEM depicted by the error bars. (B) Purified NK cells were coincubated with 221 and 221 transfectants and the resulting chemokine and cytokine production, and degranulation was assessed by flow cytometry. Total CCL4 (i), CD107a (ii), and IFN-γ (iii) production was normalized to responses seen with parental 221 cells and analyzed via an ordinary one-way ANOVA with Šídák multiple comparisons test (6 donors across 3 independent experiments). (C) The proportion of NK cells producing each functional output in combination. (D) Transfectants were stained with increasing concentrations of tetramerized DNAM-1, TIGIT, or CD96 proteins conjugated to SA-PE and the mean fluorescence intensity (MFI) of their binding was measured by flow cytometry (3 independent experiments; error bars depict the SEM). Orange, 221; dark blue, 221 + CD155lo; light blue, 221 + CD155hi; dark pink, 221 + CD112lo; light pink, 221 + CD112hi; dark green, 221 + CD111lo; light green, 221 + CD111hi.

Upon incubation with primary NK cells, expression of CD111 at low or high levels or low-level expression of CD112 did not significantly enhance NK cell degranulation, IFN-γ production, or production of the chemokine CCL4, compared to untransfected 221 cells (Fig. 5B). In contrast, target cells expressing high levels of CD112 augmented NK cell activation to a similar level as that seen with 221 + CD155lo targets. To determine whether the nectin and necl proteins differentially impacted the quality of the NK cell responses, the proportion of cells producing IFN-γ or CCL4, or degranulating (CD107a) alone or in combination was assessed (Fig. 5C). Unstimulated NK cells expressed low levels of CCL4 as previously reported,³⁶ while each target stimulated similar CCL4, CD107a, and combined CCL4 and CD107a responses. Differences in the response to nectin-transfected target cells were primarily seen in the CD107a/CCL4/IFN-γ triple-positive population. Thus, the additional engagement of DNAM-1 during NK cell recognition of 221 targets, by virtue of introducing CD155 or CD112, drove differences primarily in the production of IFN-γ.

The capacity of the transfectants to directly bind DNAM-1 was next assessed by staining with soluble recombinant DNAM-1 tetramers. 221 + CD155lo and 221 + CD112hi targets had similar DNAM-1 binding profiles, whereas low CD112-expressing targets only weakly bound the recombinant protein and high CD155-expressing targets strongly bound DNAM-1 (Fig. 5D). These same binding patterns were recapitulated when the transfectants were stained with recombinant TIGIT tetramer. In addition to DNAM-1 and TIGIT, CD155 also binds CD96²¹; however, CD96 tetramers were only observed to appreciably bind 221 + CD155hi transfectants. Notably, CD111-expressing transfectants did not bind CD96 tetramers, which may be reflective of its low affinity for the receptor.³⁵

NK cell activation is retained upon moderate loss of DNAM-1

The retention of NK cell activation with low CD155-expressing targets, despite some downregulation of DNAM-1, suggested that activation may depend upon the degree or timing of DNAM-1 loss. Therefore, to understand the relationship between the kinetics of DNAM-1 loss and NK cell activation, low and high CD155- and CD112-expressing targets were coincubated with primary NK cells for 5 hours, 3 hours, or 1 hour, or mixed directly prior to staining, and the surface expression of DNAM-1 and TIGIT was assessed by flow cytometry (Fig. S5A). The 221 + CD155hi targets drove seemingly instantaneous loss of DNAM-1 upon target and NK cell encounter (diminishing by around 60%) and dropped further over time (Fig. 6Ai; Fig. S5B). DNAM-1 downregulation in the presence of 221 + CD155lo and 221 + CD112hi targets occurred in a similar manner, steadily declining over the 5 hours to around 70%. Only a minor decrease in DNAM-1 expression was observed when coincubated with low CD112-expressing targets. The expression of TIGIT was similarly affected upon encounter with 221 + CD155hi cells, immediately dropping to 25% (Fig. 6Aii). A progressive loss in TIGIT expression was also seen with 221 + CD155lo targets; however, 221 + CD112hi targets did not alter the expression of TIGIT, despite equivalently binding the receptor (Fig. 5D), as similarly observed with wild-type K562 cells (Fig. S4B).

Figure 6. DNAM-1 signals at low density. Purified NK cells were incubated with 221 and 221 transfectants for 1, 3 or 5 hours, or mixed immediately before staining.

(A) Live, CD3^-CD56⁺ NK cells were analyzed by flow cytometry and the mean fluorescence intensity (MFI) of DNAM-1 (i) or TIGIT (ii) normalized to their expression on NK cells in the absence of targets was measured. Data were analyzed via a 2-way ANOVA with Dunnett multiple comparison test, with differences from 221 cells at 5 hours depicted (**P =0.0064, ****P < 0.0001). (B) NK cells were concurrently assessed for degranulation (CD107a expression) (i) and IFN-γ production (ii). Data were analyzed via a 2-way ANOVA with Dunnett multiple comparison test, with transfectant differences only from 221 cells at 5 hours depicted (CD107a: 221 + CD155lo **P = 0.0015, 221 + CD112hi **P = 0.0028; IFN-γ: 221 + CD155lo **P = 0.0028, 221 + CD112hi **P = 0.0041). (C) The normalized, average DNAM-1 MFI (i) or TIGIT MFI (ii) correlated with IFN-γ production. Data derived from 5 donors across 2 independent experiments. Gray, NK only; orange, 221; dark blue, 221 + CD155lo; light blue, 221 + CD155hi; dark pink, 221 + CD112lo; light pink, 221 + CD112hi. Error bars depict the SEM.

DNAM-1 loss has been observed to be reversible upon removal of CD155-expressing target cells.¹⁴ To ascertain how exposure to low or high CD155-expressing targets impacts the recovery of DNAM-1, NK cells were separated from targets 3 hours post-coincubation via fluorescence-activated cell sorting and then allowed to rest for 1 to 4 days before receptor expression levels were assessed. Due to a general increase in receptor expression over 4 days with IL-2, receptor mean fluorescence intensity was assessed relative to that seen on NK cells in the absence of target cell experience at each timepoint. NK cells in the presence of 221 + CD155lo cells showed a 52% decrease in DNAM-1 expression after 3 hours, which recovered to 64.56% after 4 days in the absence of the target cell line (Fig. S6). In contrast, DNAM-1 expression was reduced to 32.52% with 221 + CD155hi cells and failed to recover over the 4 days. No differences were observed for NKG2D while TIGIT regained greater expression than DNAM-1.

NK cell degranulation and IFN-γ production were also measured over the 5-hour time course (Fig. S5C). NK cells quickly began to degranulate within 1 hour of coincubation with all targets, with nectin and necl transfected targets slightly outstripping degranulation stimulated by parental 221 cells (Fig. 6 Bi). NK cell degranulation diverged at 5 hours with 221 + CD155lo and 221 + CD112hi targets eliciting a greater response than NK cells stimulated with 221 parental cells. IFN-γ production was evident at 3 hours and had further increased by 5 hours where the proportion of IFN-γ⁺ NK cells segregated by target cell, with 221 + CD155hi targets eliciting a similar response to parental 221 targets, 221 + CD155lo and 221 + CD112hi targets stimulating the highest proportion of IFN-γ⁺ cells, and 221 + CD112lo targets stimulating more modest activation (Fig. 6Bii). When the IFN-γ response was correlated with DNAM-1 expression, increased IFN-γ was seen with initial reductions in DNAM-1, but this enhanced response was not evident in cells that had more extensive DNAM-1 downregulation (Fig. 6 Ci). The correlation between TIGIT and IFN-γ production, however, did not follow such a trend, with wild-type levels of TIGIT expression essentially retained in the presence of 221 + CD112hi targets despite efficient cytokine production (Fig. 6Cii). These results suggest that NK cell responses to nectin and necl-expressing targets may be more dependent on DNAM-1–mediated activation than TIGIT-mediated inhibition, and that there exists a balance between the extent of DNAM-1 downregulation and the capacity for NK cells to respond to targets.

Nectin and necl-expressing target recognition is driven by DNAM-1

To test for the involvement of DNAM-1 and TIGIT in CD155-expressing target cell recognition, antibodies specific for DNAM-1, TIGIT, or NKG2D, or KIR3DL1 as a control, were used to block cognate receptor/ligand interactions by coincubating them with NK cells prior to the addition of target cells. As observed previously, in the absence of blocking antibodies, cells expressing low levels of CD155 stimulated increased IFN-γ responses from NK cells relative to either untransfected cells or those expressing high levels of CD155. As expected, the responses from cells preincubated with anti-KIR3DL1 further recapitulated these observations (Fig. 7A). Blocking DNAM-1 also had no effect on the recognition of 221 + CD155hi targets or untransfected 221 cells but led to reduced IFN-γ production in the presence of 221 + CD155lo targets, suggesting that the elevated IFN-γ response was dependent on DNAM-1. Blocking of NKG2D impacted recognition of all targets, although target cells expressing low levels of CD155 still stimulated greater IFN-γ production than those that completely lacked it or expressed it at high levels. The addition of a TIGIT-specific mAb modulated recognition of both CD155-expressing target cell lines, demonstrating that a degree of TIGIT inhibition still occurs whenever CD155 is present and is consistent with the reduction in receptor expression. However, this TIGIT-mediated inhibition in the presence of 221 + CD155hi targets did not increase NK cell responses beyond those observed with untransfected 221 cells, possibly due to the to lack of reciprocal DNAM-1–mediated activation.

Figure 7. Recognition of nectin and necls is DNAM-1 mediated.

(A) Purified NK cells were incubated with antibodies toward DNAM-1, TIGIT, NKG2D, or KIR3DL1 for 30 minutes prior to coculture with transfected target cells (not edited for DNAM-1 expression) and the resulting production of IFN-γ examined by flow cytometry (5 donors across 3 independent experiments). (B) Purified NK cells were gene edited to disrupt surface expression of the receptors DNAM-1, TIGIT, CD96, or NKG2D, sorted for receptor loss and expanded. Representative histograms compare wild-type (dark gray) and knockout (KO, colored) expression of targeted receptors on expanded NK cells (unstained: light gray). (C) Wild-type and knockout expanded NK cells were coincubated with 221, 221 transfectants, or 293T cells for 5 hours and the production of IFN-γ by CD3⁻CD56⁺ NK cells was assessed by flow cytometry (3 donors across 3 independent experiments). Total percentage of NK cells expressing IFN-γ upon incubation with parental target cells (i) and fold change in IFN-γ production compared to responses against 221 cells when wild-type and knockout NK cells were incubated with 221 transfectants (ii). Data were analyzed by a 2-way ANOVA with Dunnett multiple comparisons test (****P < 0.0001, other significances as noted) where error bars depict the SEM.

To further validate the role of DNAM-1 in the differential recognition of CD155 and CD112-expressing targets, the genes encoding DNAM-1, along with TIGIT, CD96, and NKG2D, were disrupted in primary NK cells via CRISPR/Cas9 gene editing. NK cells were then sorted for loss of receptor expression and expanded before use in functional assays. Staining for receptor expression confirmed efficient disruption/KO of DNAM-1, CD96, and NKG2D expression, while some residual TIGIT expression remained (Fig. 7B). The capacity for CRISPR/Cas9-edited cells to recognize targets cells was first compared between parental 221 cells and 293T cells, with endogenous CD155 and CD112. As expected, disruption of DNAM-1, TIGIT, and CD96 expression had no impact on NK cell recognition of 221 cells, while NKG2D deletion reduced their activation (Fig. 7 Ci). In contrast, NK cell production of IFN-γ when incubated with 293T cells decreased upon DNAM-1 editing and increased with loss of TIGIT, consistent with their activating and inhibitory roles upon CD155/CD112 engagement. Notably, editing of CD96 had no impact on NK cell recognition of 293T cells, whereas NKG2D disruption decreased activation.

The CRISPR/Cas9-edited NK cells were then assessed for their capacity to recognize the low and high nectin and necltransfected 221 cell lines. When normalized to IFN-γ production in the presence of untransfected 221 cells, the enhanced activation seen with 221 + CD155lo targets was significantly diminished upon DNAM-1 editing, an effect similarly seen with 221 + CD112hi targets and, to a lesser degree, 221 + CD112lo targets (Fig. 7Cii). Disruption of DNAM-1 expression had no impact on the recognition of high CD155-expressing targets. Editing of TIGIT had no impact on the recognition of any transfected target cell. The lack of effect of TIGIT seen here in comparison to the antibody blocking assays may reflect the use of ex vivo versus IL-2 expanded NK cells or residual TIGIT expression. Disruption of CD96 similarly had no impact on recognition of any of the targets. Curiously, incubation of expanded NK cells with 221 + CD111hi targets resulted in diminished IFN-γ production compared with 221 + CD111lo targets, suggesting that CD111 may mediate other interactions leading to reduced activation at higher levels of expression. Together, disruption of DNAM-1 interactions by blocking or receptor editing affirmed the prominent role of this receptor in mediating NK cell activation when encountering low CD155-expressing targets and that its loss precludes the possibility of a response toward high CD155-expressing targets.

Discussion

The downregulation of receptors from the surface of cells is thought to limit the strength or duration of signaling,³⁷ making it an important mechanism by which immune activation can be regulated. Here, the extent to which the coactivating receptor DNAM-1 was downregulated from the surface of primary human NK cells was observed to inversely correlate with the expression level of its ligands, CD155 and CD112, on target cells. However, while high levels of CD155 rapidly drove near-complete DNAM-1 loss and precluded NK cells from responding to the ligand, lower levels of CD155 augmented NK cell function despite still substantially decreasing the surface expression of DNAM-1. This suggests that there may exist a window of CD155 expression that is able to augment NK cell IFN-γ responses toward tumor targets, and that signals driving receptor signaling and downregulation may operate at different thresholds.

Multiple activating and coactivating receptors on NK cells are observed to undergo ligand-induced downregulation with a subsequent attenuation of NK cell function, including 2B4, NKp46, and NKG2D.^38–40 In the case of NKG2D, the ligand MICA was observed to drive greater receptor internalization than ULBP2, and potentially via a different pathway, with one such possible explanation being the affinities/avidities of their interactions.^41,42 Here, while increasing CD155 levels drove more extensive DNAM-1 downregulation to the point where it did not augment NK cell activation, higher CD112 levels resulted in increased NK cell IFN-γ responses without evidence of functional impediment. This difference could potentially stem from the presence of additional receptors that target CD112 such as PVRIG counterbalancing activating signals from DNAM-1 and potentially attenuating the extent of its internalization and degradation.¹⁷ Alternatively, these differences may be due to affinity/avidity differences in the interaction between DNAM-1 and CD112 compared with CD155. CD112, unlike CD155, has been reported to form dimers in solution through strong homophilic interactions, which may impact the avidity of their interaction with receptors.⁴³ Indeed, binding of TIGIT to CD112 is thought to require the dissociation of these CD112 dimers,⁴⁴ whereas CD155 monomers may be more readily available, potentially resulting in CD155 being a more potent ligand. While CD112 has been reported to bind DNAM-1 with marginally lower affinity than CD155,⁴⁴ differences in their spatial distribution may also modulate both their capacity to recruit sufficient receptors to transduce activation signals and drive receptor downregulation.

The striking loss of DNAM-1 upon encounter with high CD155-expressing targets without the associated production of IFN-γ suggests that while the processes of receptor downregulation and signaling are both dependent on ligand engagement, downregulation can occur without significant augmentation of functional responses. Whether this hypofunctionality occurs with other NK cell receptors warrants further investigation. Activation signals transduced through DNAM-1 have been shown to be dependent on phosphorylation of Tyr322 within its immmunoreceptor tyrosine tail–like motif by Src kinases, which allows for the coupling of Grb2 and the downstream activation of Vav-1, phosphatidylinositol 3^’ kinase (PI3K), and phospholipase C-γ1 (PLCγ1).⁴⁵ Phosphorylation of this same tyrosine residue was also found to be partially required for the downregulation of DNAM-1 on NK cells, and mediated the ubiquitination of DNAM-1 at its C-terminal lysine residues by the E3 ubiquitin ligase Cbl-b.¹⁴ The incomplete restoration of DNAM-1 expression seen with high CD155-expressing targets upon dasatinib treatment here is consistent with previous observations and suggests that additional mechanisms may contribute to DNAM-1 downregulation.¹⁴ Previous work dissecting T-cell receptor (TCR) downregulation and activation found that while knockdown of the Src kinases Lck and Fyn impaired T-cell activation, they only partially impacted TCR downregulation whereas dasatinib treatment abrogated both processes.⁴⁶ Notably, the PI3K regulatory subunit p85, which can bind the SH3 domain of Grb2, was found associated with phosphorylated DNAM-1.⁴⁵ Although the actions p85 are best understood with regard to its regulation of the p110 catalytic subunit of PI3K, it has been found to mediate functions independent of the molecule’s kinase activity.⁴⁷ For example, upon erythropoietin (Epo) binding, endocytosis of the Epo receptor was driven by ubiquitination of p85 by Cbl.⁴⁸ Internalization of 2B4 is also reported to be independent of PI3K kinase activity.³⁸ This raises the potential for ligand engagement by NK cell receptors to drive dual or divergent functions. Although the “high” levels of CD155 expressed here may be beyond that which is commonly observed in vivo, the rapid downregulation of DNAM-1 in the absence of activating signaling induced by these cells may provide a model with which to investigate the relationship between signals for endocytosis and activation as well as their points of divergence.

The augmented response of NK cells in the presence of low CD155-expressing targets primarily manifested in IFN-γ production, rather than degranulation or chemokine expression. Cytokine production by NK cells is under stringent control, typically requiring the engagement of multiple coactivating receptors, and occurs later than degranulation or chemokine production following stimulation.³⁶ The recognition of parental 221 cells by NK cells involves 2B4 interaction with CD48, and likely NKp46 interactions,³⁸ as well as NKG2D engagement, as evident here through the decreased NK cell responses in the presence of anti-NKG2D antibodies or upon CRISPR editing of NKG2D from NK cells. While synergism between 2B4, NKG2D, and NKp46 may therefore dictate NK responses to parental 221 cells, the addition of CD155 to the targets and the subsequent engagement of DNAM-1, which also synergizes with 2B4,³ may alter or enhance downstream signaling pathways to drive greater cytokine responses. Such modulation of the IFN-γ response was also evident with K562 and 293T cells, which do not express CD48, but where additional DNAM-1 engagement may enhance activation through synergism, for example, with NKp46 ligands.³ Indeed, different activation thresholds for degranulation and cytokine production have been described for cytotoxic T cells, with the formation of a more mature immune synapse required for cytokine responses.⁴⁹ Given the role of DNAM-1 in adhesion, along with LFA-1, and its co-operation with other natural cytotoxicity receptors to kill tumors,^5–7 the additional DNAM-1 engagement here may function to enhance synapse formation and thereby cytokine production.

Despite binding to both CD155 and CD112, the inhibitory receptor TIGIT had a very modest capacity to modify NK cell responses in this system. Although antibody blocking of TIGIT on ex vivo NK cells slightly increased IFN-γ responses to both low and high CD155-expressing targets, CRISPR editing of the receptor from expanded NK cells had no impact on transfected 221 target cell recognition. TIGIT also exhibited rapid downregulation from the cell surface in the presence of high levels of CD155; however, unlike DNAM-1, this downregulation was not rescued by dasatinib treatment. While the effectiveness of TIGIT blockade on T cells reportedly depends upon the presence of DNAM-1,¹⁸ the rapid loss of DNAM-1 on NK cells here may decrease the significance of any actions of TIGIT. Notably, coexpression of DNAM-1 and TIGIT on tumor-infiltrating lymphocytes is rare and therefore the relativity availability of each receptor for CD155 binding may dictate their responses.⁵⁰ Indeed, the observed upregulation of TIGIT on tumor-infiltrating NK cells, despite the presence of high levels of CD155, may relate to their exhausted phenotype,²⁶ with a greater TIGIT to DNAM-1 ratio observed on tumor-infiltrating NK cells in metastatic melanoma⁵¹ and TIGIT expression seen to increase following prolonged NKG2D stimulation.⁵² Unlike DNAM-1, the mechanism of TIGIT downregulation is not yet defined. Here, TIGIT was less sensitive to downregulation in the presence of CD112 and its expression was only affected by high levels of CD155 on K562, 293T, or P815 cells. Furthermore, the impact of TIGIT engagement has been reported to differ with cell type, with IFN-γ production inhibited in CD8⁺ T cells but not in NK cells.⁵³ Similarly, a role for CD96 was not apparent here, with CRISPR/Cas9-editing of the receptor having no impact on NK cell response to targets, consistent with previous observations blocking CD96 with antibodies.⁹ In contrast, recognition of CD155 by KIR2DL5 potently inhibited NK cell activation. However, given that a significant proportion of individuals either lack this gene completely, express allotypes that are not well expressed at the cell surface or, that even when present, it is typically expressed on only a small subset of NK cells,⁵⁴ KIR2DL5 recognition of CD155 is not likely to have a central role in modulating bulk NK cell responses.

Together, the data here show that expression of CD155 on target cells has the potential to both augment and inhibit NK cell function, with DNAM-1 playing a central role in enhancing IFN-γ production. Critically, the data also suggest that the potential of CD155 to augment NK cell activation through DNAM-1 is somewhat juxtaposed with its capacity to downregulate the receptor. This ultimately creates an “expression window” where low levels of CD155 augment activation but above which receptor downregulation is likely so rapid and extensive as to prevent the propagation of DNAM-1–dependent activation signals. This may be important in tumor settings with heterogenous CD155 expression as the presence of a modest number of cells expressing high levels of the ligand may be sufficient to prevent augmented responses to cells expressing lower ligand levels. Thus, in vivo variation of CD155 expression levels is likely to have marked impacts on how NK cell activation is regulated by receptors such as DNAM-1, TIGIT, CD96, and KIR2DL5 and indeed the extent to which this network can be effectively targeted by immunotherapeutic approaches.

Supplementary Material

Supplementary materialis available at The Journal of Immunology online.

Data availability

Data are available upon reasonable request from the authors.