Human NK cells: surface receptors, inhibitory checkpoints, and translational applications

Abstract

NK cells play important roles in innate defenses against viruses and in the control of tumor growth and metastasis. The regulation/induction of NK cell function is mediated by an array of activating or inhibitory surface receptors. In humans, major activating receptors involved in target cell killing are the natural cytotoxicity receptors (NCRs) and NKG2D. Activating receptors recognize ligands that are overexpressed or expressed de novo upon cell stress, viral infection, or tumor transformation. The HLA-class I-specific inhibitory receptors, including KIRs recognizing HLA-class I allotypic determinants and CD94/NKG2A recognizing the class-Ib HLA-E, constitute a fail-safe mechanism to avoid unwanted NK-mediated damage to healthy cells. Other receptors such as PD-1, primarily expressed by activated T lymphocytes, are important inhibitory checkpoints of immune responses that ensure T-cell tolerance. PD-1 also may be expressed by NK cells in cancer patients. Since PD-1 ligand (PD-L1) may be expressed by different tumors, PD-1/PD-L1 interactions inactivate both T and NK cells. Thus, the reliable evaluation of PD-L1 expression in tumors has become a major issue to select patients who may benefit from therapy with mAbs disrupting PD-1/PD-L1 interactions. Recently, NKG2A was revealed to be an important checkpoint controlling both NK and T-cell activation. Since most tumors express HLA-E, mAbs targeting NKG2A has been used alone or in combination with other therapeutic mAbs targeting PD-1 or tumor antigens (e.g., EGFR), with encouraging results. The translational value of NK cells and their receptors is evidenced by the extraordinary therapeutic success of haploidentical HSCT to cure otherwise fatal high-risk leukemias.

Introduction and general overview of human NK cells

Natural killer (NK) cells play major roles in first-line innate defenses against viral infections, tumor growth, and metastatic spread. Both the constitutive expression of efficient, ready-to-function lytic machinery, and the rapid release of interferon-gamma (IFN-γ) and tumor necrosis factor-alfa (TNF-α) following cell activation allow prompt intervention by NK cells, resulting in target cell killing and the initiation of an inflammatory response.

NK cells were discovered in the mid-70s;^1,2 however, the mechanism by which they discriminate between tumor or virus-infected cells and healthy cells remained a mystery for many years. The missing self-hypothesis, proposed in the late 80s by Karre and Ljunggren,³ represented a true milestone, inspiring subsequent groundbreaking discoveries in the early 90s. This hypothesis was based on the finding that murine NK cells could kill a lymphoma cell line that had lost major histocompatibility complex-class I (MHC-cl I) molecules, while parental MHC-cl I⁺ cells were resistant to lysis. Thus, NK cells appear to sense the absence of MHC-cl I on cells (i.e., the missing self).⁴ Research aimed to identify the molecular mechanism(s) involved in this phenomenon, leading to the discovery of inhibitory receptors specific for MHC-cl I molecules in both humans and in mice. Remarkably, while the first identified human receptors specific for human leukocyte antigen (HLA)-cl I molecules belong to the immunoglobulin (Ig) superfamily (and were subsequently named killer-cell immunoglobulin-like receptors, KIRs), the murine receptors (Ly49) are members of the lectin family. Thus, humans and mice developed two molecularly different receptors fulfilling the same function, i.e., the recognition of MHC-cl I molecules and the delivery of inhibitory signals to NK cells, resulting in NK cell inactivation.⁵ The discovery of KIRs was made possible by the availability of two important technologies, namely, monoclonal antibody technology and a high-efficiency lymphocyte cloning system. The latter technology allowed the clonal expansion of virtually 100% of human T lymphocytes.⁶ Thus, it has been possible to analyze the frequency of cells endowed with a given function (e.g., cytolytic activity) and to establish correlations between the surface phenotype and function of any T-cell population.⁷ After suitable adaptation of this cloning technique, NK cells were also revealed to be clonogenic and suitable for precise functional studies. NK cell clones were used for mouse immunization and the production of monoclonal antibodies (mAbs) that were selected for their ability to either inhibit or induce NK cell cytolytic activity or cytokine production. Two selected mAbs, originally defined as p58.1 and p58.2^8–10, recognized two highly homologous molecules expressed by partially overlapping NK cell subsets. These mAbs represented prototypes of HLA-cl I-specific inhibitory receptors, and were found to recognize allotypic determinants shared by the two main groups of HLA-C alleles.^10–13

Other HLA-cl I-specific receptors were identified (see below). Of relevance, the HLA-E-specific NKG2A,¹⁴ expressed earlier than KIR during NK cell maturation, was recently found to be of great potential interest as a target of checkpoint inhibitors in tumor immunotherapy^15,16(see below).

Remarkably, the pool of mature NK cells is equipped with at least one inhibitory receptor for self HLA-cl I antigens (whether KIR or NKG2A). Only very few peripheral blood (PB) NK cells lack such receptors and are anergic. The repertoire of inhibitory NK receptors is shaped during NK cell maturation and is the result of a process of selection termed NK cell “licensing” or “education”.^17–19 Accordingly, only NK cells expressing receptors for self HLA-cl I molecules acquire full functional potential, while the remaining are either deleted or anergic.

“Off” signals are necessary to prevent NK-mediated autoreactivity, implying the existence of “on” signals responsible for NK cell activation. Indeed, in the absence of inhibitory interactions, NK cells kill target cells and produce cytokines. Again, the use of NK cell clones and the selection of mAbs were fundamental for the discovery of prototypes that activate NK receptors, namely, NKp46,^20–22 NKp44²³ and NKp30,²⁴ collectively termed “natural cytotoxicity receptors” (NCRs).²⁵ mAbs directed to these surface molecules were selected based on their ability to modulate NK clone-mediated cytotoxicity against selected tumor targets. It became evident that NK cell activation is not due to a master receptor (such as TCR and BCR in T and B cells, respectively) but, in most instances, is the result of the combined action of several receptors and coreceptors (see below). Triggering of the activating NK receptors requires the expression of appropriate ligands on potential target cells. Such ligands are overexpressed or expressed de novo in stressed healthy cells and, more importantly, in cells that have undergone tumor transformation or viral infection.²⁶ Remarkably, tumors and certain viruses may induce the downregulation of activating NK receptors and/or their ligands on target cells, thus favoring escape from the NK-mediated control. This is evident in a relevant percentage of tumors, particularly neuroblastoma, in which not only HLA-cl I molecules but also ligands for most activating NK receptors may be downregulated.²⁷ This mechanism of tumor escape underscores the potential importance not only of T cells but also of NK cells in the immune response against neuroblastoma. Taken together, the expression of inhibitory and activating receptors in NK cells and their ligands on target cells revealed that NK cell activation/inactivation is controlled by different checkpoints.²⁸ This notion is important for translational applications that exploit NK cell function in therapy for leukemia or solid tumors (see below).

Human NK cells have been divided into two major subsets identified according to the surface density of the CD56 antigen.²⁹ CD56^bright cells represent only ~10% of PB NK cells, but they predominate in tissues. These cells express NKG2A, are poorly cytolytic but secrete cytokines, primarily IFN-γ and TNF-α, and efficiently proliferate in response to interleukin (IL)-2 or IL-15. CD56^dim cells largely predominate in PB (~90%), expressing NKG2A or KIR or both, and rapidly display potent cytolytic activity and cytokine production upon activation. Based on the expression of receptors/markers, CD56^dim NK cells are further subdivided into different stages of differentiation, characterized by a progressive decrease in proliferative capacity paralleled by an increase in cytolytic activity.³⁰ The most mature NK cells are KIR⁺ NKG2A^neg and express CD57 and (in individuals seropositive for cytomegalovirus, CMV) NKG2C, an HLA-E-specific activating receptor. Recent studies revealed that NKG2C⁺ NK cells display adaptive features, sensing different HLA-E-bound CMV peptides and undergo selective proliferation only in response to given HLA-E-peptide combinations.³¹ Such specific expansion, together with a number of adaptive characteristics, is compatible with the memory-like function of these cells, which can allow for prompt control of CMV reactivation.³²

While NK cells have been known for over 45 years, other cells that are developmentally related to NK cells were identified only 10 years ago and collectively termed innate lymphoid cells (ILCs).^33,34 Similar to NK cells, ILCs derive from CD34⁺ hematopoietic precursors and share with NK cells a common ID2⁺ lymphoid precursor.³⁵ Absent or infrequent in PB from healthy individuals, ILCs reside in mucosal tissues, skin, and lymphoid organs. Unlike NK cells, the other ILCs (termed “helper ILCs”) are noncytolytic and secrete typical sets of cytokines that differ for each subset, including IFN-γ (ILC1), IL-5, IL-13, and small amounts of IL-4 (ILC2), IL-22 (NCR⁺ ILC3), IL-17, and TNF-α (NCR^neg ILC3 or LTi-like cells).³⁶ Interestingly, helper ILCs mimic corresponding CD4⁺ T-cell subsets in terms of the type of cytokines produced. Therefore, these cytokines can shape not only adaptive but also innate immune responses, although with different timing. In general, ILCs contribute to innate defenses against various pathogens and are involved in tissue repair and homeostasis and in lymphoid organogenesis, particularly during fetal life. While they will not be discussed further in this review, which is devoted to NK cells, it is worth noting that ILCs may share some markers and the NCR activating receptors with NK cells.³⁷

NK cell traffic in the circulation and towards tissues and lymphoid organs is regulated by chemokines and chemokine receptors, directing the given NK cell subsets to specific sites.³⁸ Thus, while the expression of CD62L, CXCR3, and CCR7 will direct CD56^bright NK cells to secondary lymphoid organs (in response to CCL19 and CCL21), the expression of Chemerin R, CXCR1, CXCR2, and CX₃CR1 directs CD56^dim NK cells to inflammatory peripheral tissues in response to their corresponding chemokine ligands (Chemerin, IL-8, and Fraktalkine). Since CD34⁺ cells capable of generating NK cells have been detected in different tissues, including thymus,³⁹ tonsils,^40,41 decidua⁴², and liver,⁴³ it is conceivable that at least some tissue-resident NK cells may derive from these precursors, migrate from BM, and acquire unique properties as a consequence of their particular tissue microenvironment.⁴⁴

The following paragraphs will analyze in more detail the various NK receptors, recently identified inhibitory checkpoints, and their implications in the immunotherapy of tumors. In addition, we will outline the fundamental role of NK receptors in the cure of high-risk leukemias in allogeneic hemopoietic stem cell transplantation (HSCT).

Inhibitory and activating NK receptors and their roles in tumor cell killing

HLA-specific NK receptors

As mentioned above, important regulation of NK cell function is provided by inhibitory receptors specific for HLA-cl I molecules. Indeed, HLA-cl I molecule recognition on healthy cells by these inhibitory NK receptors prevents NK cell-mediated attack.

Human NK cells express two different classes of HLA-class I-specific inhibitory receptors: members of the KIR/CD158 family and the CD94/NKG2A (CD94/CD159a) heterodimer.^5,45 KIRs are type I transmembrane receptors specific for polymorphic HLA-A, B and C molecules,^8–10,12,13 whereas NKG2A is a type II transmembrane receptor that recognizes HLA-E, a non-classical HLA molecule characterized by limited polymorphism.¹⁴ To transduce inhibitory signals, both types of inhibitory receptors contain ITIM motifs in their cytoplasmic tail. In addition, activating forms of KIRs have been identified.⁴⁶ Different from inhibitory KIRs, activating KIRs lack ITIM motifs in their cytoplasmic tail and have a transmembrane domain carrying a charged amino acid residue⁴⁷ that mediates the association with the ITAM-bearing molecule KARAP/DAP12. The role of activating KIRs in the immune response is only partially known. Specificity for HLA-cl I molecules has been demonstrated for only a few of them.^46,48–54

Inhibitory KIRs are characterized by long cytoplasmic tails, whereas activating KIRs have short cytoplasmic tails (“L” or “S” in the nomenclature, respectively). Regarding Ig domain content, each KIR displays two or three extracellular Ig domains (“KIR2D” or “KIR3D” in the nomenclature, respectively). Two types of KIR2D can be determined according to their extracellular Ig domain content. KIR2Ds of the first type are composed of D1 and D2 domains and include the majority of KIRs (KIR2DL1/L2/L3 and KIR2DS1/S2/S3/S4/S5), whereas KIR2Ds of the second type are composed of D0 and D2 domains and include KIR2DL4/L5.⁵⁵ Notably, not only HLA class I but also KIRs are characterized by high levels of polymorphism, which may affect given KIR/HLA interactions. In addition, certain KIR/HLA combinations have been shown to correlate with protection or susceptibility to infectious, autoimmune, and reproductive disorders.

Another HLA-specific inhibitory receptor is represented by LIR-1/ILT2/CD85j.⁵⁶ LIR-1 is a type I transmembrane protein belonging to the Ig-like receptor superfamily that can interact with classical (HLA-A, HLA-B, HLA-C) and non-classical (HLA-G) HLA-cl I molecules.^57–59 It can also bind UL18, a cytomegalovirus-encoded HLA-cl I homolog that is expressed on CMV-infected cells.⁵⁶ Notably, high LIR-1 expression correlates with the acquisition of NK cell memory in CMV⁺ donors.⁶⁰

Another HLA-specific activating receptor is represented by NKG2C, a receptor that, similar to NKG2A, binds HLA-E but with lower affinity.^61,62 Altogether with inhibitory KIRs, CD94/NKG2A prevents the response against cells with normal expression of HLA-I molecules, whereas CD94/NKG2C is involved in the response to human HCMV. Notably, NKG2A is primarily expressed by PB immature NK cells, whereas NKG2C is expressed only at late stages of NK cell maturation.⁶³ The terms “adaptive” or “memory-like” are currently employed to designate the human differentiated NKG2C^bright NK cell subset that is characterized by the CD56^dim CD57⁺ KIR⁺ NKG2A^neg phenotype and that is expanded in HCMV⁺ donors.^32,64,65 Similar to activating KIRs, NKG2C is coupled to the ITAM-bearing molecule KARAP/DAP12.

Activating NK receptors and coreceptors involved in tumor cell killing and their ligands

Human NK cells express several receptors that can trigger their function upon interaction with specific ligands on the surface of transformed, virus-infected, or stressed cells (Table 1).

Table. 1.

Human NK cell receptors and their ligands

Molecule	Ligand	CD56bright	CD56dim
Non-HLA-specific receptors
Coreceptors
CD59	LFA-2 (CD2)	All PB NK cells
NTB-A (CD352)	NTB-A (CD352)	All PB NK cells
NKp80	AICL	All PB NK cells
DNAM-1 (CD226)	Nectin-2 (CD112), PVR (CD155)	All PB NK cells
2B4 (CD244)	CD48	All PB NK cells
Inhibitory
PD-1 (CD279)	PD-L1 (CD274), PD-L2 (CD273)	−	Subsets
Siglec-7 (CD328)	Ganglioside DSGb5	Most of PB NK cells
IRP60 (CD300a)	α-Herpes virus, Pseudorabide virus, Phosphatidylserine, Phosphatidylethnolamine	All PB NKs
Tactile (CD96)	PVR (CD155)	All PB NK cells
IL1R8	IL-37	PB NK cells
TIGIT	PVR (CD155)	PB NK cells
TIM-3	Gal-9, PtdSer, HMGB1, CEACAM1	Subset of NK cells
Activating
NKp30 (CD337)	B7-H6, BAG6/BAT3	++	+
NKp44 (CD336)	21spe-MLL5 - Nidogen-1	Activated NK cells
NKp46 (CD335)	CFP (properdin), viral HA and HN, PfEMP1	++	+
NKG2D (CD314)	MIC-A, MIC-B, ULPBs	All PB NK cells
FcγRIII (CD16)	IgG	−/+	+/++
HLA-specific receptors
Inhibitory
NKG2A/KLRD1 (CD159a/CD94)	HLA-E	+	Subsets
KIR2DL1 (CD158a)	HLA-C2	−	Subsets
KIR2DL2/3 (CD158b)	HLA-C1, few HLA-Bb	−	Subsets
(CD158d)a	HLA-G	−	Subsets
KIR2DL5 (CD158f)	???	−	Subsets
KIR3DL1 (CD158e1)	HLA-A-Bw4, HLA-B-Bw4	−	Subsets
KIR3DL2 (CD158k)	HLA-A*03 and *11	−	Subsets
ILT2/LIR-1 (CD85J)	Different MHC-I alleles	.	Subsets
LAG-3 (CD223)	MHC-II	Activated NK cells
Activating
KIR2DS1 (CD158h)	HLA-C2	−	Subsets
KIR2DS2/3 (CD158j)	???	−	Subsets
KIR2DL4 (CD158d)a	HLA-G	−	Subsets
KIR2DS4 (CD158i)	HLA-A*11 and some HLA-C	−	Subsets
KIR2DS5 (CD158f)	???	_	Subsets
KIR3DS1 (CD158e1)	HLA-Bw4, HLA-F	−	Subsets
NKG2C (CD159a)	HLA-E	−	Subsets
Homing receptors
2° lymphoid tissues
CCR7 (CD197)	CCL19, CCL21	+	−
CXCR3 (CD183)	CXCL9, CXCL10, CXCL11	++	Subsets
L-Selectin (CD62L)	GLyCAM-1, MadCAM-1	++	Subsets
Inflammation sites
CXCR1 (CD181)	CXCL8 (IL-8)	−	+
CXCR2 (CD182)	IL-8-RB	−	+
CX3CR1	Fraktalkine	−	+
ChemR23	Chemerin	−	+
Others
CXCR4 (CD184)	CXCL2	Subsets of NK cells
CCR5 (CD195)	RANTES, MIP1α and MIP1β	Subsets of NK cells
S1P5	S1P	−	+
c-Kit (CD117)	SCF (KL)	+	−

^aKIR2DL4 has been shown to have both inhibitory and activating functions¹⁶⁴

^bKIR2DL2/L3 weakly recognize also HLA-C2 alleles and few HLA-B alleles that bear the HLA-C1 epitope (e.g., HLA-B*4601 and HLA-B*7301)

The NCRs²⁵ are among the major activating NK receptors and consist of three elements, called NKp46/NCR1/CD335,^20–22 NKp44/NCR2/CD336²³, and NKp30/NCR3/CD337.²⁴ These molecules were classically described as germline-encoded receptors and are important for inducing NK cell cytotoxic function against tumors and infected cells.

NKp46 and NKp30 are expressed on nearly all resting human NK cells, upregulated on activated NK cells and downregulated on “adaptive” NK cells that are found in CMV⁺ individuals. Unlike NKp46 and NKp30, NKp44 is constitutively expressed only on CD56^bright NK cells, but it is acquired by essentially all NK cells after activation by cytokines.

NCRs are type I transmembrane molecules belonging to the immunoglobulin-like family.

These receptors were named in accordance with their molecular weights according to sodium dodecyl sulfate polyacrylamide gel electrophoresis (NKp46, NKp30, and NKp44). Their transmembrane domains contain a positively charged amino acid that allows their association with the transmembrane regions of the TCR-ζ and/or FcεRI-γ (for NKp30 and NKp46) or KARAP/DAP12⁶⁶ (for NKp44) adaptor proteins. In this regard, it is important to underline that the physical association between NKp44 and KARAP/DAP12 is essential for NKp44 surface expression⁶⁷ and that the decreased surface density of NKp30 and NKp46 on “adaptive” NK cells is associated with the lack of FcεRI-γ expression in these cells.⁶⁸

Different molecules can interact with the extracellular domains of NCRs. Some of them are virus-derived molecules that can activate NK cell function against infected cells (such as influenza virus-derived HA recognized by NKp46 or NKp44⁶⁹) or induce inhibitory signals (such as the HCMV-encoded pp65 protein recognized by NKp30.⁷⁰) Other NCR ligands (such as BAT3/BAG6, MLL5, and PCNA) are represented by intracellularly localized proteins that may reach the cell surface in response to stress or during tumor transformation. For example, BAT3/BAG6 is typically located in the nucleus;⁷¹ however, in cells exposed to heat shock, it can move to the plasma membrane and be secreted in exosomes by tumors and stressed cells.^71,72 Then, BAT3/BAG6-expressing exosomes can stimulate cytokine release by NK cells upon interaction with NKp30. MLL5 is expressed in the nuclei of normal cells, but 21spe-MLL5, an MLL5 isoform that functions as an NKp44 ligand, is located in the cytoplasm and at the cell surface.⁷³ Finally, the nuclear protein PCNA can function also as an NCR ligand. Indeed, in normal cells, PCNA is involved in the processes of DNA replication/repair and cell cycle control, but in tumor cells, it can be shuttled to the tumor cell surface and expressed on tumor-derived exosomes, functioning as an NKp44 ligand.⁷⁴

B7-H6, another NKp30 ligand,⁷⁵ is not expressed on healthy cells, but it is frequently present on the cell surface of many tumor types through a Myc-mediated mechanism.⁷⁶ Interestingly, its expression may also be acquired by normal cells upon TLR-mediated stimulation and in the presence of pro-inflammatory cytokines.⁷⁷

Recently, NKp46 and NKp44 have also been shown to recognize extracellular ligands. Indeed, NKp46 can bind to a soluble plasma glycoprotein called complement factor P/properdin, revealing cross-talk between two partners in innate immunity in the response to Neisseria meningitidis infections.⁷⁸ Moreover, NKp44 has also been shown to recognize an extracellular ligand called Nidogen-1 (NID1, also known as Entactin).⁷⁹ The NKp44/NID1 interaction results in reduced NKp44-mediated cytokine release by NK cells and induces relevant changes in the NK cell proteomic profile, suggesting an effect on different biological processes.

Importantly, it has been shown that tumors can orchestrate different mechanisms to impair NCR function. Thus, hypoxia or various soluble factors produced by tumor/tumor-associated cells (such as indoleamine 2,3 dioxygenase [IDO], tumor growth factor-beta [TGF-β], prostaglandin E2 [PGE2]), or inhibitory NCR ligands (such as the soluble form of BAT3 or B7-H6)^72,80 can induce decreases in NCR expression and function.⁸¹ Indeed, NCR^low NK cells can be detected in PB and particularly in the tumor site in patients affected by solid and hematologic tumors. Notably, reduced expression/function of NCRs can also be detected in NK cells from HIV-infected patients.⁸²

Another important activating NK receptor is NKG2D, a type II transmembrane and C-type lectin-like receptor, which may be expressed on cytotoxic T cells. NKG2D ligands are represented by ULBPs and MICA/B,⁸³ which are HLA-cl I structural homologs that are upregulated in infected, stressed, and tumor cells.^84,85 Notably, shedding of NKG2D ligands by tumor cells may represent a mechanism of tumor escape.

Other molecules, including 2B4,⁸⁶ NTB-A,⁸⁷ DNAM-1,⁸⁸ CD59,⁸⁹ and NKp80,⁹⁰ function primarily as coreceptors; indeed, they are capable of amplifying the NK cell triggering induced by NCRs or NKG2D. In addition, NK cells may express toll-like receptors (TLRs) that, after interaction with bacterial or viral products and in the presence of pro-inflammatory cytokines, induce potent NK cell activation.^91–94

Finally, the Fcγ receptor CD16, recognizing the Fc portion of IgG antibodies specific for unhealthy cells, can trigger antibody-dependent cell-mediated cytotoxicity (ADCC).⁹⁵ CD16^bright expression is restricted to mature CD56^dim KIR⁺ NK cells.

The molecular basis of NK cell function

NK cell function is regulated by the activating and inhibitory receptors illustrated above. The fine balance of signaling that occurs through them determines if NK cells kill their target cells or remain inactive. As discussed, autoreactivity in human NK cells is controlled by KIRs and NKG2A inhibitory receptors specific for self HLA-cl I molecules. KIRs and NKG2A are clonally expressed or coexpressed in NK cell subsets, thus creating repertoires of NK cells with different phenotypic and functional characteristics that are capable of responding to different types of virally infected and tumor-transformed self cells. Under normal conditions, the interactions between these inhibitory receptors and their specific HLA-cl I ligands inactivate NK cells, thus preventing cytolytic activity against healthy cells. During cancer progression, the transformed cells decrease or even lose the expression of HLA-I on their surface. In addition, they express ligands for activating NK receptors: two events necessary for the induction of antitumor NK cell responses. Indeed, ligands for the activating NK receptors are generally absent or expressed in small amounts in normal cells, while they are expressed de novo or upregulated at the cell surface in stressed normal cells and, in particular, in virus-infected or tumor cells. Therefore, signaling through activating receptors can overcome the signaling mediated by inhibitory receptors on diseased cells.^25,96

NK inhibitory checkpoints and their ligands in solid tumors

Inhibitory checkpoints in NK cells

In addition to the HLA-cl I-specific inhibitory receptors that regulate NK cell function and prevent NK-mediated damage to healthy tissues, additional inhibitory checkpoints, responsible for maintaining immune cell homeostasis, have been described in NK cells, including PD-1, TIGIT, CD96, TIM-3, etc. Under pathological conditions, some of these checkpoint regulators, which are absent on resting NK cells, can be induced de novo and affect antitumor NK cell function upon interaction with their specific ligands that are frequently expressed on the tumor cell surface and facilitate tumor immune escape.^97–99

PD-1, a member of the immunoglobulin superfamily, is a major checkpoint regulating T-cell activation and ensuring peripheral tolerance.^100,101 However, when binding to its ligands (PD-L1 and PD-L2), which may be expressed on tumor cells, PD-1 may compromise the antitumor effector function favoring tumor immune escape. Notably, in mice, the blockade of PD-1 resulted in restoration of T-cell responses.¹⁰²

Although the expression of PD-1 has been described on T, B, and myeloid cells, recent studies have shown that, under pathological conditions, it is also expressed by NK cells.^103,104 PD-1⁺NK cells that can inactivate NK cell function were found in patients with CMV infections and cancer, including Kaposi sarcoma, ovarian carcinoma, and Hodgkin lymphoma.^63,104–107

A recent study by our group revealed the presence of a pool of PD-1 proteins and mRNA in the cytoplasm of healthy donor NK cells. The molecular mechanisms regulating the expression of PD-1 on human NK cells have not been defined so far. However, it is conceivable that signals delivered by cells and/or soluble factors in the tumor microenvironment may play a major role (Mariotti et al. in press).

In mice, PD-1 expression may be induced by glucocorticoids, underscoring the importance of specific signaling for its regulation.¹⁰⁸ Several murine tumor models showed that PD-1/PD-L1 interactions can strongly suppress NK cell-mediated antitumor immunity and that PD-1⁺ NK cells present at the tumor site display an exhausted phenotype.¹⁰⁹

In humans, in addition to PD-1, other inhibitory checkpoints are expressed not only by T lymphocytes but also by NK cells. They include the T-cell Ig and ITIM domain (TIGIT) and CD96 (tactile). These receptors are members of a group of Ig superfamily receptors that also include the activating receptor DNAM-1. Their corresponding ligands, usually upregulated in tumor cells, are the poliovirus receptor (CD155), nectin-2 (CD112), and nectin-like molecules.^110,111 TIGIT expression has been reported in different malignancies in peritumoral lymphocytes, often together with PD-1 and TIM-3, and is associated with CD8⁺ T-cell and NK cell suppression.¹¹² For example, TIGIT was upregulated in tumor-associated NK cells of patients with colon-rectal cancer. TIGIT knockout mice or the use of mAbs that block TIGIT could induce/potentiate NK and T-cell antitumor activity, prevent NK cell exhaustion and lead to the control of tumor growth. A recent study provided evidence that TIGIT targeting with specific mAbs may represent a new promising tool that can be used alone or in combination with other checkpoints.¹¹³ Thus, it has been shown that TIGIT blockade may induce antitumor immune activity in preclinical models, and its combination with PD-1/PD-L1 inhibitors is being explored.¹¹⁴

KLRG1, another inhibitory C-type lectin-like receptor containing one ITIM, is expressed by activated NK cells, ILC2s, T cells, mast cells, basophils, and eosinophils.¹¹⁵ The interaction between KLRG1 and its corresponding ligand E-cadherin has been shown to inhibit human ILC2 function in vitro; however, its actual effect in vivo remains to be established.^116–118 In non-small cell lung cancers (NSCLC) and colon-rectal carcinoma (CRC), KLRG1 expression was detected in tumor-associated ILC2s.^119,120

Other inhibitory receptors expressed by NK cells include LAG-3 (CD223) and TIM-3. LAG-3 is a negative costimulatory receptor homologous to CD4 that binds MHC-cl I molecules on antigen presenting cells (APC) with very high affinity. LAG-3 expression is associated with an exhausted profile of tumor-infiltrating-lymphocytes (TIL), and its blockade could restore antitumor immune function.^121–123 While its inhibitory activity in T-cell activation, proliferation and homeostasis has been defined, the effect of LAG-3 on NK cells remains poorly explored.¹²⁴ TIM-3, also known as Hepatitis A virus cellular receptor 2 (HAVCR2), is a type 1 glycoprotein expressed by subsets of T lymphocytes (Th17 and regulatory T cells), dendritic cells and mature NK cells. The main ligand of TIM-3 is Galectin-9,¹²⁵ but other ligands have been identified, such as phosphatidylserine (PtdSer), high mobility group protein 1 (HMGB1), and carcinoembryonic antigen related cell adhesion molecule 1 (CEACAM1).^126–128 Previous studies have shown that in TILs of lung,¹²⁹ gastric,¹³⁰ head and neck carcinomas¹³¹, and in melanoma,¹³² TIM-3 expression is upregulated. Moreover, the increased TIM-3 expression on NK cells in melanoma or lung adenocarcinoma was shown to be associated with impaired NK cell effector function.^133,134 TIM-3 is often coexpressed with PD-1 and is associated with T-cell exhaustion during chronic infection and cancer. Blockade of TIM-3 could restore T-cell effector function in preclinical models and result in increased NK cytotoxicity.^132,135

Another important inhibitory receptor, IL1R8 (also known as TIR8 or SIGIRR), is a member of the IL1 receptor family and a component of the human IL-37 receptor.^136,137 It has been shown to play a physiologic role in the regulation of responses to pathogens by dampening excessive inflammatory responses and tissue damage. A recent study has shown that IL1R8 is highly expressed in a subset of NK cells, in which it exerts major inhibitory control of cell activation and function. In a murine model, NK cells lacking IL1R8 could prevent the growth of a carcinogen-induced hepatocellular carcinoma. This study provided evidence that IL1R8 functions as an important inhibitory checkpoint in NK cell activation and function. Accordingly, its blocking can unleash NK cells and promote efficient antitumor activity.¹³⁸

PD-L1 and other checkpoint ligands

As illustrated above, PD-1 negatively regulates immune responses; under physiological conditions, these interactions lead to peripheral T-cell tolerance, while in cancer patients they may impair T-cell responses against tumor cells. In this context, immunotherapy with checkpoint inhibitors that disrupt the PD-1/PD-L1 interaction has proven to be highly effective in different tumor types, representing a true revolution in cancer therapy.¹³⁹ Since the approval of nivolumab and pembrolizumab in 2014, the number of PD-1/PD-L1 inhibitors has grown rapidly. Currently, there are six PD-1/PD-L1 inhibitors that have been approved by the Food and Drug Administration (FDA): three against PD-1 (pembrolizumab, nivolumab and cemiplimab) and three targeting PD-L1 (atezolizumab, durvalumab and avelumab). In parallel with this rapidly expanding arsenal, the spectrum of advanced malignancies for which each of these agents is indicated has broadened, including melanoma (pembrolizumab and nivolumab), non-small cell lung cancer (pembrolizumab, nivolumab, and atezolizumab), urothelial cancer (pembrolizumab, nivolumab, atezolizumab, durvalumab, avelumab), renal cell carcinoma (nivolumab), head and neck cancer (pembrolizumab and nivolumab), Hodgkin lymphoma (pembrolizumab and nivolumab), microsatellite instability or mismatch repair deficient (dMMR) cancers including dMMR colorectal cancer (pembrolizumab, nivolumab), gastric cancer (pembrolizumab), hepatocellular carcinoma (nivolumab), Merkel cell carcinoma (avelumab)¹⁴⁰, and cutaneous squamous cell carcinoma (cemiplimab).¹⁴¹ Although PD-1/PD-L1 inhibitors represent a revolution in the field of oncology, a significant proportion of patients do not benefit from this class of agents; therefore, predicting tumor responses to PD-1/PD-L1 blockade represents a major issue. So far, the most widely recognized parameter of tumor response is PD-L1 expression in tumor cells and/or immune cells, for which four assays have been approved to guide treatment decisions for different agents (Dako 22C3, Dako 28-8, Ventana SP142 and Ventana SP263).¹⁴² However, despite the clinical utility of PD-L1 expression as a biomarker of the response to PD-1/PD-L1 inhibitors, its predictive value still appears to be unsatisfactory, to the point that other parameters are being explored (including tumor mutational burden [TMB] and the presence and the numbers of TIL and genetic polymorphisms, among others). In this context, all possible limitations inherent to PD-L1 expression should be taken into account, including intra- and intertumor heterogeneity as well as technical issues represented by the use of different mAbs and different diagnostic materials (cytology versus diagnostic biopsies versus surgical specimens), in order to improve the predictive potential of PD-L1 immunohistochemistry.^{98,99,143–145} In any case, a significant proportion of patients would still be not eligible or not responsive to PD-1/PD-L1 axis inhibitors. For this reason, as illustrated above, ongoing research is aimed at finding additional checkpoints other than PD-1/PD-L1 and CTLA-4 to be exploited for targeting by therapeutic antibodies either alone or in combination.

Importantly, while the TCR-mediated antitumor activity of T lymphocytes correlates with TMB, this is not the case for NK cells. Therefore, therapeutic blockade of inhibitory checkpoint or their ligands may still be effective in tumors with NK cell infiltration. For this reason, there is a growing interest in the less-explored NK cell compartment for therapeutic purposes.^146,147 As indicated, promising inhibitory checkpoints are LAG-3, TIM-3, TIGIT, HLA-E, and HLA-G, among others.

In this context, two HLA-cl I b molecules expressed on tumors may be particularly relevant.

HLA-E is overexpressed in many tumors.¹⁴⁸ As illustrated below, a recent work demonstrated that monalizumab, a humanized anti-NKG2A antibody, enhanced NK cell activity against various tumor cells and rescued CD8⁺ T-cell function alone or in combination with PD-1/PD-L1 axis blockade.¹⁵ HLA-G is physiologically expressed by trophoblast cells, which are mainly involved in maternal-fetal tolerance.¹⁴⁹ HLA-G can interact with immunoglobulin-like transcript 2 (ILT2) expressed on NK, T and B lymphocytes, monocytes/macrophages, and dendritic cells and with ILT4, which is expressed only by myeloid cells, as well as KIR2DL4 expressed on NK cells. Under normal conditions, HLA-G is virtually absent in adult tissues; by contrast, most tumors express HLA-G at different levels, either on their cell membranes or released as soluble isoforms. Different studies have shown that HLA-G acts by impairing the cytolytic function and proliferation of peripheral and uterine NK cells and of cytotoxic T cells, the maturation and function of dendritic cells while inducing Tregs and myeloid suppressive cells, thus inhibiting different actors involved in antitumor responses.¹⁵⁰ Therefore, HLA-G plays a key role in the induction of immune tolerance, representing an important escape mechanism of tumor cells. However, the therapeutic role of HLA-G-based immunotherapy is still virtually unexplored. We speculate that the use of T or NK cells expressing chimeric antigen receptors (CAR-T or CAR-NK) targeting HLA-G could represent a promising approach for the adoptive immunotherapy of different tumors. Studies in this context are in progressing in our labs.

Exploiting NK cells for translational applications

The majority of current tumor immunotherapies are based on the use of mAbs that target tumor antigens or unlock inhibitory checkpoints expressed by T lymphocytes or the generation of T cells expressing CARs specific for tumor antigens. However, recent contributions reassessed or confirmed the importance of NK cells in tumor therapy. Indeed, as illustrated under the headings below, NK cells can fulfill unique antitumor activities such as ADCC, thanks to the expression of CD16 or the killing of HLA-cl I-deficient tumor cells that are undetectable by cytolytic T lymphocytes. In addition, masking the HLA-cl I-specific KIR or NKG2A inhibitory receptors may induce NK cell-mediated cytolytic activity against HLA-cl I⁺ target cells. Finally, the generation of CAR-NK cells with extensive proliferative capacity is now available and may be implemented with CAR-T cell therapy, exploiting the homing properties and functional characteristics typical of NK cells.

Targeting NKG2A alone or in combination with other therapeutic mAbs

As underlined above, blocking monoclonal antibodies specific for inhibitory checkpoints have revealed unprecedented potential in tumor therapy, representing a true revolution in the prognosis of a subset of otherwise highly aggressive cancers, including NSCL cancer, metastatic melanoma, kidney tumors and others.¹³⁹ In spite of this major progress, the majority of patients do not respond to treatment with mAbs targeting the CTLA-4 or PD-1-mediated inhibitory pathways. This partial failure may be due to one or more mechanisms of tumor escape. Thus, as mentioned above, in many instances, tumor progression is accompanied by a partial or total loss of classical HLA-cl I antigens, which are necessary for tumor antigen presentation to cytolytic T lymphocytes. In addition, the tumor microenvironment may greatly contribute to tumor escape. Indeed, an important mechanism is the release of cytokines (e.g., TGF-β) and soluble inhibitory factors (e.g., kynurenine, PGE2), not only from tumor cells but also from tumor-infiltrating cells (e.g., myeloid-derived suppressor cells, M2 macrophages, fibroblasts/MSC). Additionally, the hypoxic environment typical of many tumors may play a substantial negative role.¹⁵¹ Of note, all of these mechanisms impair both T- and NK cell function by targeting receptor-associated polypeptides that are crucial for the transmission of activating signals. Although different therapeutic approaches are being explored to antagonize such soluble inhibitors, a particularly relevant approach is to identify other inhibitory checkpoints expressed by T or NK cells that, when blocked, may unleash the antitumor effector activity of these cells and possibly favor their proliferation.

Most promising progress towards this goal has recently been reported by André et al.,¹⁵ who showed that NKG2A may represent an important checkpoint in tumor therapy. Indeed, NKG2A targeting, alone or in combination with other checkpoint inhibitors or with mAbs specific for tumor antigens, led to impressive results both in preclinical and preliminary clinical studies in highly aggressive human tumors (Fig. 1).

Fig. 1.

Effects of blocking NKG2A alone or in combination with other therapeutic mAbs disrupting the PD-1/PD-L1 axis or directed to tumor-specific antigens. (1) Cytolytic NK cells expressing the NKG2A inhibitory receptor acquire the ability to kill HLA-E⁺ tumor cells upon mAb-mediated masking of NKG2A. (2) NKG2A⁺ NK cells that also express PD1 are rescued from their anergic state and kill HLA-E⁺ PD-L1⁺ tumor cells upon disruption of the two inhibitory pathways with mAbs specific for NKG2A and PD-L1 (or PD-1), respectively. (3) The therapeutic activity of mAbs directed to tumor-specific antigens (in this case EGFR) is mainly mediated via CD16 (FcγRIII) expressed by highly mature cytolytic NK cells. However, in the case of HLA-E⁺ tumors, the coexpression of NKG2A may represent a major brake on NK-mediated function. Accordingly, masking of NKG2A may restore NK cell function and allow killing of tumor cells targeted by specific mAbs

As discussed above, specific mAb-mediated blocking of NKG2A and/or KIR can restore NK cytotoxicity, which is inhibited by the interaction with HLA-cl I. Importantly, both inhibitory receptors, while constitutively expressed by mature NK cells, have also been detected in cytolytic T lymphocytes.^152,153 The expression of CD94/NKG2A can be induced by prolonged antigenic T-cell stimulation¹⁵⁴ or exposure to TGF-β¹⁵⁵, a cytokine released in the tumor microenvironment. Such ex novo NKG2A expression leads to the impairment of T-cell function.^154,155 Thus, blocking of NKG2A was expected to unleash not only NK cells but also T lymphocytes with potential antitumor activity. This was indeed proven in elegant preclinical studies in suitable murine models. Subsequently, by studying human cancers, André et al.¹⁵ showed that HLA-E (the NKG2A ligand) is expressed in many highly aggressive tumors (including those of the pancreas, colon, lung, head and neck, stomach, and liver). Moreover, most tumor cells examined were HLA-E⁺. Analysis of lymphoid infiltrates showed that NKG2A was expressed in NK cells and in CD8⁺ T lymphocytes, which could also coexpress PD1. Experiments in vitro confirmed that IFN-γ production and the cytolytic activity of NK and T cells against HLA-E⁺ tumors were recovered, at least in part, by blocking anti-NKG2A mAbs, and, in the case of coexpression with PD-1, better functional restoration was documented in combination with therapeutic mAbs against PD1/PD-L1 (Fig. 1). This combined treatment could also restore T-cell proliferation, suggesting the possible generation of tumor-specific T-cell memory.

Another promising immunotherapy is based on the combined use of anti-NKG2A and tumor antigen-specific mAbs (e.g., an anti-EGFR mAb). The antitumor effects of these mAbs are, at least in part, dependent on the action of mature NK cells recognizing Ig-opsonized tumor cells via their CD16 to mediate ADCC (Fig. 1). In these in vitro experiments, the combined use of anti-NKG2A and anti-EGFR enhanced NK cell-mediated cytotoxicity against antibody-opsonized tumor cells. In this context, in preliminary clinical studies, the use of monalizumab (anti-NKG2A) combined with the anti-EGFR cetuximab in an ongoing phase II trial provided encouraging results in patients with squamous cell carcinoma of the head and neck (SCCHN). Notably, this study also emphasized the importance of harnessing NK cell antitumor activity, while immunotherapeutic strategies so far have been focused on potentiating T-cell responses. When unlocked from NKG2A blocking, NK cells may not only potentiate their natural cytotoxicity but also kill tumor cells via ADCC in the presence of tumor-specific therapeutic mAbs.

Role of NK cells in haploidentical HSCT for the cure of high-risk leukemia

HSCT is the life-saving therapy for acute leukemias with either adverse cytogenetic or molecular characteristics or poor response to chemotherapy or relapse. However, only 2 out of 3 patients requiring HSCT can find a suitable HLA-compatible donor. A novel transplantation approach has been attempted to rescue such patients for whom no other valuable therapeutic option are available. Thus, haploidentical HSCT, developed in the late 90s, is based on the infusion of “megadoses” of highly purified CD34⁺ haemopoietic precursors isolated from a parent or a sibling. Given the high degree of HLA incompatibility, deep T-cell depletion is mandatory to avoid severe, life-threatening graft versus host disease (GvHD). In this transplantation setting, NK cells of the donor may express KIRs that are mismatched with the HLA-cl I alleles of the patient. The first wave of lymphoid cells in patient PB occurs at ~2 weeks after transplant and is composed almost exclusively of NK cells. However, these NK cells are CD56^bright and express NKG2A but not KIR. More mature KIR⁺NK cells are detectable only after additional 4-6 weeks. In the absence of T cells, NK cells have been shown to play a prominent role in the anti-leukemia effect.¹⁵⁶ While NK cells per se may kill leukemia cells that express low levels of HLA-cl I molecules, the anti-leukemia effect has been primarily related to NK cell alloreactivity consequent to a KIR-HLA mismatch (in the donor vs recipient direction).¹⁵⁶ Studies by our group revealed a direct correlation between the size of the alloreactive population and the degree of anti-leukemia cytotoxicity in vitro as well as the survival probability of the patient.^157,158 Thus, clinical data in over 80 pediatric patients receiving haplo-HSCT indicated the treatment was particularly good for high-risk ALL with an ~70% 5-year survival probability in the presence of NK alloreactivity and ~40% in the absence of alloreactivity. The overall survival was ~60%. The outcome was less favorable for AML (~40% and ~20% with and without NK alloreactivity, respectively, and 30% overall survival). Although all these patients would have died in the absence of HSCT, the overall survival was not satisfactory, particularly for AML patients. Thus, a novel transplantation setting has been applied in an attempt to fill the gap between the time of HSCT and the generation of mature KIR⁺ alloreactive NK cells. Indeed, most of the deaths occurred in the first weeks after HSCT due to leukemia relapse or infections. This approach is based on the selective depletion of T cells expressing TCRαβ, which are responsible for GvHD, and B cells (to avoid B cell malignancies in immunocompromised patients). Thus, the infused mononuclear cells contained, in addition to CD34⁺ progenitors, mature KIR⁺ (possibly alloreactive) NK cells and TCRγδ T cells,¹⁵⁹ plus different myeloid cells. Notably, not only NK cells but also TCRγδ T cells can kill leukemia blasts.¹⁶⁰ The prompt availability of effector cells capable of attacking leukemia and exerting control of viral reactivation greatly improved the clinical outcome. Indeed, the overall survival probability was ~70% for AML and over 70% for ALL. Altogether, these data clearly indicate that NK cells exert strong anti-leukemia activity. The ability of CD34⁺ cells from different sources to give rise to ILC3s, including NCR^negILC3s with LTi-like activity, which contribute to tissue repair and regeneration of lymphoid tissues damaged by chemo/radiotherapy during conditioning, should also be carefully evaluated.¹⁶¹

Concluding remarks

A large body of experimental evidence, particularly from preclinical studies of both in vitro and in vivo animal models, indicates that NK cells play relevant roles in antitumor defenses. This is due to their direct cytolytic activity against tumor cells and their ability to produce cytokines that shape effective (Th1) downstream adaptive responses. However, different mechanisms occurring primarily in the tumor microenvironment may greatly impair NK cell function by compromising their antitumor activity, and they may even play an unwanted role in tumor promotion. This may occur when the NK cells present at the tumor site are few and/or their function is compromised. The result may be NK-mediated tumor editing (i.e., selection of the most resistant tumor cells)¹⁶² or the induction of epithelial/mesenchymal transition (that favors tumor spread and metastasis).¹⁶³ In this context, any therapeutic approach that may favor/reconstitute NK cell function and expansion may be useful in tumor therapy. Thus, mAb-mediated masking of HLA-cl I inhibitory receptors may allow the direct attack of NK cells towards HLA-cl I⁺ tumor cells. Importantly, our group demonstrated that such inhibitory receptors, particularly NKG2A, also can be expressed on CD8⁺ CTLs following antigen- or cytokine-induced proliferation or exposure to TGF-β.^152,154,155 These data offered clues to interpret recent studies by Andrè et al.¹⁵ showing the efficacy of blocking NKG2A in tumor therapy. On the other hand, the recent finding that not only T but also NK cells may express PD-1 revealed a new perspective, especially for the therapy of HLA-cl I-negative tumors (invisible to T lymphocytes).

It should be underscored that the major therapeutic success of haploidentical HSCT in the cure of high-risk leukemia is mostly related to NK cell activity. Great expectations for the therapy of both hematologic malignancies and solid tumors are based on the use of NK cells expressing CARs specific for tumor antigens. These CAR-NK cells may complement/substitute CAR-T lymphocytes in view of their potent cytolytic activity and their particular homing capacity.

Competing interests

The authors declare no competing interests.