Over 95% of the world’s population is infected by the Epstein–Barr virus (EBV), an oncogenic γ-herpesvirus that is transmitted via saliva to naïve B cells in oral lymphoid tissues (1). EBV establishes a latent infection through stepwise restriction of viral protein and non-coding RNA (ncRNA) expression (from type III to 0 latency programs), hence persisting for life in memory B cells in which it may occasionally switch to a lytic replication phase, allowing virus reseeding and transmission to a different host (1, 2). EBV transition from latency to lytic replication is under the control of cellular factors, including signaling pathways, epigenetic regulation, and stress conditions, and of viral gene products, with the immediate-early BZLF1 protein having a key role in the transactivation of most lytic genes that allow the virus to replicate (3). Primary EBV infection normally occurs during infancy without causing symptoms, yet it may lead to infectious mononucleosis (IM), particularly when deferred into puberty; additionally, mostly in individuals with an immunodeficiency caused by co-infections, inborn error of immunity (IEI), or immunosuppressive therapies, EBV is associated with various lymphoproliferative disorders (LPDs) and malignancies affecting B cells or other atypical viral targets, including Burkitt’s lymphoma (BL), Hodgkin’s lymphoma (HL), NK or T cell lymphoma, nasopharyngeal carcinoma (NPC), and gastric carcinoma (GC), overall being responsible for approximately 1.5% of all human tumors (4). EBV-related pathologies arise from impaired or dysregulated immune responses against EBV, from the activity of various viral proteins and ncRNAs that promote tumorigenic transformation and immune evasion, and, possibly, from viral and host factors not yet identified (1, 4–6).

In healthy individuals, EBV-specific CD8 T cells are crucial for viral load suppression during primary infection, then persist as long-lasting memory cells (5). Before development of virus-specific T-cell responses, EBV is challenged by the encounter with innate immune cells, such as natural killer (NK) cells, which are abundant at the site of virus entry and function as the first line of defense against pathogens (7). NK cells are cytotoxic lymphocytes armed with the antigen-independent capacity to kill infected, tumor-transformed, and, in general, “stressed” cells via secretion of lytic granules; such innate response is controlled via the balance between opposite signaling pathways: on the one hand, a negative regulation is exerted by inhibitory receptors, including inhibitory killer immunoglobulin receptors (iKIRs), NKG2A, and LILRB-1, upon engagement with ligands belonging to the HLA-class I protein family; on the other hand, activating receptors, which comprise NKG2D, DNAM-1, natural cytotoxicity receptors (NCRs: NKp30, NKp44, and NKp46), NKG2C, and activating KIRs (aKIRs), mediate NK cell activation following recognition of ligands that are either virally encoded or upregulated on infected/stressed cells (8–10). NK cell-mediated killing also occurs via antibody-dependent cellular cytotoxicity (ADCC) triggered by the CD16 low-affinity Fc receptor or via apoptosis induced by death receptor ligands (TRAIL and CD95-L). Moreover, activated NK cells secrete cytokines, particularly IFN-γ and TNF, and chemokines that attract and stimulate T cells and innate cells, thus promoting broad immune responses (8, 10). Most receptors are expressed in NK cells in a random manner, generating a remarkable phenotypic and functional NK cell heterogeneity (11). In peripheral blood, two NK cell populations can be distinguished: a minor subset (~10%) consists of CD56^brightCD16^−/+ cells (referred to as CD56^bright), which are efficient cytokine producers and commonly considered as immature precursors of the most abundant highly cytotoxic CD56^dimCD16⁺ subset (referred to as CD56^dim) (12). Phenotypically, CD56^bright cells are characterized by the expression of NKG2A and various activating receptors, while the CD56^dim subset comprises a heterogeneous population of NK cells with different degrees of differentiation, which implies the gradual loss of NKG2A and sequential acquisition of iKIRs and the CD57 maturation marker (13, 14). Additionally, among CD56^dim cells, a population of NKG2C⁺ cells with features of adaptive immunity, also referred to as Memory-Like, has been identified in HCMV-seropositive individuals (15).

A large body of clinical and experimental evidence supports the important role of NK cells in the control of EBV during primary infection and beyond, as we recently reviewed (16). First, discovery of six isolated NK cell deficiencies as well as impaired NK cell development or function in more than 50 different IEIs illustrated the critical role of these cells in anti-viral defense, particularly in the prevention of severe infection with EBV and other herpesviruses (17). Second, reduced NK cell frequency and/or cytotoxicity were associated with EBV⁺ LPDs and tumors by comparison with equivalent EBV⁻ diseases, pointing towards a key role of NK cells in restraining EBV-related pathogenesis (18–22). Furthermore, studies on acute EBV infection in IM patients and in a humanized mouse model of IM showed the expansion of early differentiated CD56^dimNKG2A⁺ NK cells with effective anti-EBV activity before initiation of virus specific T-cell responses (23, 24); additionally, NK cell depletion experiments in mice demonstrated that NK cells not only suppressed EBV viremia and prevented EBV-induced malignancies, but also reduced lymphocytosis and the consequent IM symptoms by cytotoxic clearance of activated CD8 T cells (23). Finally, studies in EBV-infected individuals pointed at the anti-viral potential of distinct NK cell subsets, specifically mature CD56^dimNKG2A⁺CD57⁺ cells (25) and tonsillar CD56^brightIFN^high cells (7).

The knowledge on NK cell biology has greatly increased in recent years, leading to the development of novel immunotherapies such as adoptive transfer of potentiated NK cells, which offer exciting promise for the treatment of EBV-associated diseases. However, for successful clinical applications, current understanding of the mechanisms that regulate the capacity of NK cells to kill EBV⁺ cell targets should be improved. Indeed, several experimental studies have shown the capacity of various EBV proteins and ncRNAs to modulate the expression of ligands for both inhibitory and activating NK receptors, though results are often controversial and little is known on the consequences for NK cell-mediated killing [reviewed in Ref (16).]. As for HLA-class I molecules, these are highly expressed on latently EBV-infected cells through the activity of viral factors such as LMP1 and then they are downregulated by various lytic proteins acting at different levels (from gene transcription to protein trafficking) once the virus has entered the productive replication program; hence, the interaction of HLA-class I on EBV⁺ cells with cognate iKIRs inhibiting NK cell activation could be either efficient or impaired depending on the latent or lytic phase, respectively. This model is in line with NK cell-mediated killing being more efficient against targets with lytic EBV replication reported in studies in vitro and in a mouse model (23, 24, 26), yet is at odds with the lack of KIR expression on those NK cell subsets that react against EBV in vivo during IM and upon in vitro challenge with EBV-infected targets (24). Another controversial matter concerns EBV modulation of ligands for NK cell activating receptors, particularly those recognized by NKG2D (MICA/MICB and ULBP1–6 family of proteins, referred to as NKG2DLs) and DNAM-1 (CD112 and CD155 proteins, referred to as DNAM-1Ls). Expression of cell-surface NKG2DLs, which is normally highly restricted but can be induced in response to infection, transformation, or other cellular stress (27, 28), was described on latently EBV-infected cells in some reports (24, 29, 30), even though there is no consensus on the type of ligand expressed (i.e., MICA, MICB, ULBP1, or ULBP4) and, in other studies, none of the NKG2DLs was found (26, 31); a few studies also analyzed NKG2DLs after entry in the EBV lytic phase, showing disparate results ranging from further MICA up-modulation (24), de novo induced ULBP1 expression (26), or complete absence of NKG2DLs (31). Analogously, contrasting evidence was provided for DNAM-1Ls that, depending on the report, were either absent in all phases of the EBV life cycle (31) or induced during lytic replication (24, 26). Furthermore, experimental studies demonstrated the capacity of individual EBV proteins or ncRNAs to either repress (e.g., mir-BART2/7, LMP2, and EBNA1) or induce (e.g., LMP1 and BZLF1) specific NKG2DL expression, without an apparent separation of inhibitory and stimulatory activities relative to the viral replication phases [reviewed in Ref (16).]. Overall, a growing number of disparate lines of evidence suggest that EBV has evolved strategies to modulate the expression of ligands for NK cell receptors but further investigation is clearly needed to understand which receptor/ligand axes effectively mediate the capacity of NK cells to control EBV infection throughout the viral life cycle.

Here, we set up EBV-immortalized lymphoblastoid cell lines (LCLs) with stable expression of an inducible form of BZLF1 to challenge primary NK cells with targets in either the latent or lytic phase of EBV infection, and investigated killing efficiency, NK cell receptor/ligand interactions, and which subset of NK cells was most responsive.

EBV is usually acquired in young age and persists throughout life in a state of latency without causing harm, yet various EBV-related LPD and malignant diseases can be incurred, particularly in patients with primary or secondary immunodeficiency. For EBV, there is no preventing/therapeutic vaccine or effective drug available, at present, and EBV-induced diseases are treated with conventional strategies, such as surgery and chemotherapy, which are often burdened by complications. On the other hand, rapidly evolving immunotherapies have recently offered novel opportunities, with some strategies already having proved their efficacy in EBV-related malignancies, including adoptive transfer of EBV-specific T cells and use of antibodies targeting B cells or blocking immune checkpoint signaling (38–41). Another promising approach to treat EBV⁺ malignancies rely on NK cell-based therapies that include modified NK cells with engineered receptors and activation/expansion of NK cell populations for adoptive transfer; of note, NK cells can outperform T cells in adoptive cell therapy because they have a general lower risk of graft-versus-host disease or other adverse events, and display effector functions that are not sensitive to escape mutations in T-cell antigens, which often occur in EBV (42, 43). Importantly, accumulating evidence has demonstrated that the induction of EBV-induced malignancies is not an exclusive prerogative of viral latency proteins as lytic factors also contribute to tumorigenesis (6, 44, 45); hence, therapeutic targeting of both latent and lytic phases may be essential.

In the present study, to investigate the interaction of NK cells with EBV throughout its life cycle, we generated multiple inducible LCL-Z cell lines that, in the absence of stimuli, typically resemble EBV-associated B-cell lymphomas with type III latency, but then, upon HT-mediated activation of recombinant BZLF1 protein, they rapidly enter into the lytic replication phase. The lytic EBV replication, for which no permissive cell system is currently available, is usually investigated in those few cells among LCL or BL cell cultures in which the virus is spontaneously reactivated or, more often, after treatment with inducing agents such as phorbol esters, calcium ionophores, histone deacetylase inhibitors, or Ig crosslinkers. Conversely, in the LCL-Z cell system, the switch to the lytic replication phase occurs efficiently with a physiologic, i.e., BZLF1-initiated, modality, hence avoiding the use of EBV-reactivating drugs that profoundly alter gene expression of the host cell. By analyzing LCL-Z cells with and without lytic reactivation, we demonstrated that, among NKG2DLs and DNAM-1Ls, only MICB is significantly expressed at the cellular membrane during latency and to a higher level in cells that entered the lytic cycle, which translates into killing by NK cells in a manner that is dependent on the cognate NKG2D receptor. In addition, we detected increased intracellular levels of MICB as well as of ULBP2 and ULBP4 within lytic cells and accumulation in LCL-Z cultures of sMICB that was higher in cultures with lytic EBV replication. Apparently, infection with EBV results in up-modulation of NKG2DLs, particularly MICB, which is a common cellular defense against viruses mediated by activation of stress response pathways (46) while, analogously to other viruses, EBV has evolved the capacity to contrast cell-surface NKG2DL expression in order to avoid immune recognition, specifically retaining MICB, ULBP2, and ULBP4 into intracellular compartments and inducing the release of sMICB from the cell membrane through the activity of viral factors that are currently unknown. Our results are in line with some previous studies showing that latently infected LCLs express MICB along with, whenever tested, low levels of MICA and ULBP4, and they are susceptible to NKG2D-dependent killing by NK cells as well as EBV-specific CD8 T-cell clones or γδ T cells (29, 30, 47). Also in accordance with our data, it was shown that the LMP2A latency protein of EBV, which activates signaling pathways to promote the proliferation of latently infected cells, induces MICA and MICB expression at the transcriptional level, while, on the other hand, reducing the cell-surface expression of both MIC proteins and ULBP4 through mechanisms that have not been identified as yet (30, 48). Additional lines of evidence indicating that EBV has developed means to contrast NKG2DL expression was provided by studies on viral ncRNAs showing that miR-BART2-5p and miR-BART7 impair translation of MICB and MICA mRNA, respectively (29, 49). These EBV activities have important functional consequences for NK cell responses, since inhibition of miR-BART2-5p in latent LCLs resulted in MICB upregulation and enhanced killing by NK cells (29), whereas overexpression of miR-BART7 in NPC lowered rates of NK cell-mediated lysis (49). Finally, studies in patients with mutated MAGT1 (referred to as “X-linked MAGT1 deficiency with increased susceptibility to EBV-infection and N-linked glycosylation defect”, XMEN) clearly pointed at a key role of the NKG2D/NKG2DL axis in the control of EBV-related diseases, which are the main cause of severe morbidity in these patients (50); specifically, because of the absence of MAGT1-mediated glycosylation, NKG2D is strongly downregulated on patients’ NK and CD8 T cells that, as a consequence, are poorly cytotoxic in vitro against autologous NKG2DL⁺ (i.e., positively stained by means of a recombinant NKG2D-Fc molecule) LCLs, unless NKG2D expression is previously restored. It should be mentioned, though, that some previous lines of evidence are discordant with our results; in particular, one study showed that cells with lytic EBV replication do not express various tested NKG2DLs (i.e., MICA, MICB, and ULBP2) or DNAM-1Ls (CD122 and CD155) and are killed by NK cells in a manner that is independent of NKG2D but requires DNAM-1, implying the engagement with a hypothetical DNAM-1L yet to be identified (31). We suggest that discrepancies are linked to differences in the experimental system employed, consisting in challenging an NK cell line (NKL) with lytic BL cells (AKBM treated with anti-IgG) or spontaneously reactivated LCLs.

To evade CD8 T-cell responses, EBV has developed a complex strategy targeting the antigen presentation machinery (APM) and HLA class I molecules that, on the other hand, may activate NK cells through the lack of inhibitory KIR/HLA class I interactions according to the “missing self” model of NK cell lysis (8). Specifically, by means of EBV activation in BL cell lines, transduction of individual EBV genes into EBV⁻ tumor cell lines, or EBV infection of primary B cells, it has been demonstrated that expression of AMP and HLA class I is stimulated by the latent LMP1 protein while the opposite effect is exerted by various lytic proteins and ncRNAs, with cell-surface HLA-A, -B, and -C molecules being drastically downregulated following entry into the lytic phase [reviewed in Ref (51).]. In the present study, we found that LCL and unstimulated LCL-Z cells express high HLA-A, -B, and -C levels in agreement with previous reports (24, 52), which reflects the nature of activated blasts and expression of LMP1; once entered into the lytic phase, HLA-A, -B, and -C levels were reduced both at the cell surface and intracellularly in a statistically significant manner, yet to a very modest extent. In accord with these results, antibody-mediated block of HLA molecules resulted in higher NK cell-mediated killing of both BZLF1⁻ and BZLF1⁺ targets, indicating that LCLs are protected from cytolysis by the interaction between HLA ligands with inhibitory KIRs on effector cells throughout the viral life cycle. Therefore, partial reduction of HLA-A, -B, and -C during lytic EBV replication is important for the virus to escape CD8 T-cell recognition, particularly via the downregulation of specific HLA-B molecules that present immunodominant EBV peptide during primary B-cell infection (53), but apparently is not sufficient to trigger NK cell activation. It should be noted, though, that NK cell inhibition via KIR/HLA-A, -B, and -C interactions shall be confirmed with EBV⁺ targets other than LCLs, for instance, EBV-associated tumors in which HLA class I molecules can be downregulated also by cellular oncogenic factors, as shown for c-Myc in the context of NPC (54).

Moreover, our results show an important role for the NKG2A/HLA-E interaction in the capacity of NK cells to kill LCL targets. In line with some previous studies in BL cells (26, 55), we found that latent LCLs efficiently express HLA-E, non-classical HLA class I molecules that present peptides derived from leader sequences of HLA-A, -B, -C, and -G molecules; then, upon EBV switch to lytic replication, HLA-E is drastically downregulated. HLA-E/peptide complexes normally suppress NK cell function by binding to CD94 in the heterodimeric NKG2A/CD94 receptor, which results in NKG2A-mediated inhibitory signaling; this important immune checkpoint can be targeted by blocking antibodies that are currently in clinical development to treat HLA-E⁺ tumors (56). Unexpectedly, we found that blocking of NKG2A via a specific antibody had no effect on the capacity of NK cells to kill latent LCLs, notwithstanding their HLA-E expression, and clearance of lytic BZLF1⁺ cells that are nearly devoid of HLA-E was rather decreased. This apparent discrepancy, however, can be explained by considering some peculiar aspects of HLA-E expression in EBV-infected cells as well as of NKG2A biology. First, peptides derived from both latent and lytic EBV proteins have been shown to associate with HLA-E molecules and prevent NKG2A triggering, thus unleashing the cytotoxic responses of NKG2A⁺ NK cells (57, 58), a phenomenon that may account for the lack of NKG2A-mediated inhibition of LCL killing in our cell system. Second, while KIR-mediated inhibition of NK cells follows a linear relationship with the levels of HLA-A, -B, and -C on targets, high HLA-E levels are required for NKG2A to inhibit activated NK cells (59). Moreover, inhibitory signals delivered by KIRs and NKG2A are needed for NK cells to become fully functional through a process referred to as “education” (also “arming” or “licensing”) in which NKG2A has the strongest impact (60). Accordingly, it was reported that tumors with low HLA-E levels are killed more efficiently by NKG2A⁺ than NKG2A⁻ NK cells and independently of KIR/HLA-A, -B, and -C interactions (61), indicating that better education by NKG2A can prevail on its inhibitory signaling, at least when HLA-E is poorly expressed on target cells. Therefore, we suggest that NKG2A cannot inhibit NK cell activation against LCLs due to the association of decoy viral peptides to HLA-E that impair engagement of NKG2A throughout the EBV life cycle as well as to insufficient HLA-E expression during the lytic phase. On the other hand, NKG2A identifies the population of NK cells that react against LCLs with lytic EBV replication, since the use of anti-NKG2A antibody reduced killing of BZLF1⁺ cells in our lysis assay. We assume that, in the absence of NKG2A/HLA-E association, engagement with anti-NKG2A antibody may trigger receptor-mediated inhibitory signals and, ultimately, inhibit cytotoxicity of NKG2A⁺ NK cells against BZLF1⁺ targets. Accordingly, by investigating the phenotype of NK cells that degranulated against LCL-Z cells with or without EBV reactivation, we showed that NKG2A⁺ NK cell subsets, specifically CD56^bright and NKG2A⁺KIR⁺CD56^dim cells, are those that better respond against cells with lytic EBV replication. Our results are largely in agreement with previous in vivo and in vitro evidence, including recruitment in the tonsils of EBV⁺ individuals of a CD56^bright NKG2A⁺ NK cell subset with a strong IFN-γ-mediated capacity to suppress EBV replication in vitro (7) and expansion during IM of circulating NKG2A⁺CD56^dim NK cells that are proficiently activated when challenged with lytic BL cells (23, 24). In addition, in vitro experiments using IL-2-activated NK cells as effectors showed that CD56^dimNKG2A⁺ cells are preferentially activated against autologous LCLs, although targets with lytic EBV replication were not tested (34). When in-depth analysis of NKG2A⁺CD56^dim cells reacting in vitro against EBV⁺ targets has been performed, this showed not only high expression of NKG2D and loss of CD16 in line with our results (34, 35, 62), but also lack of KIR expression, which is at odds with our finding that NKG2A⁺KIR⁺CD56^dim cells are those that better respond against lytic LCLs. However, we think that the present study differs from others in that it is the first one to investigate the overall distribution into maturation subsets of NK cells that have reacted against autologous LCLs with either latent or lytic EBV replication. Interestingly, the simultaneous presence of NKG2A and KIRs is indicative of highly educated CD56^dim cells provided with full responsiveness to target cells (63).

It may seem counterintuitive that EBV has evolved means to impair NKG2A-mediated inhibition of NK cells and elicit the antiviral effector function of NKG2A⁺ NK cells. However, two distinct viruses, HIV-1 and SARS-CoV-2, also encode for HLA-E binding peptides that abrogate inhibition of NKG2A⁺ NK cells, so that infected cell targets (CD4 T and epithelial lung cells, respectively) are killed in vitro more efficiently by NKG2A⁺ NK cells (both CD56^bright and NKG2A⁺ CD56^dim) than NKG2A⁻ counterparts (64, 65). It is plausible that viral HLA-E binding peptides correspond to conserved sequences critical for the replication fitness of EBV, HIV-1, and SARS-CoV-2 viruses that, on the other hand, have evolved other immune evasion strategies (e.g., downregulation of ligands for activating receptors) to counterbalance missing-self activation of NKG2A⁺ NK cells.

This study has limitations because host genetic factors that fine-tune NK cell functionality have not been considered, including HLA polymorphisms that influence HLA-E stability and education of NKG2A⁺ NK cells (66) and HLA alleles and KIR haplotypes associated with EBV infection and disease (67). Another aspect that deserves to be further investigated is the capacity to clear EBV⁺ targets of adaptive NK cells that express the inhibitory CD94/NKG2C receptor for HLA-E and display superior effector functions in the context of HCMV infection (15). In general, all our results should be confirmed in clinical samples. In particular, it should be investigated whether MICB is upregulated in various EBV-associated tumors and eventually accumulate in a soluble form in patients’ plasma. Likewise, the expression and function of NKG2D in NK cells of patients with EBV-related disease require investigation since chronic binding to cell-surface or soluble NKG2DLs can lead to receptor “detuning” and overall desensitization of NK cells (68). These studies will shed light on the potential of the NKG2D/MICB axis as a therapeutic target in EBV-driven malignancies. At present, several bioengineered molecules harnessing the NKG2D pathway against cancer have been developed such as antibodies that block MIC protein shedding, the extracellular NKG2D domain fused to immune activating components, and the NKG2D-based chimeric antigen receptor (CAR.NKG2D) [reviewed in Ref (69).]. Moreover, we speculate that the development of allogenic HLA mismatched NK cells with the NKG2A⁺KIR⁺ phenotype and enhanced expression of NKG2D may harbor great potential for adoptive cell therapies in patients with EBV⁺ malignancies. Finally, the efficacy of NK cell-based therapies may be further increased by the combination with the “lytic induction” approach (70), which implies reactivation of EBV with histone deacetylase inhibitors or other epigenetic drugs that are well-known for their potent capacity to induce NKG2DL expression among various biological effects (71).