Abstract

NK cells

Natural killer (NK) cells belong to the innate lymphoid cell (ILC) family.¹ The various ILC subtypes are characterized principally by their cytokine secretion profiles, which resemble those of CD4⁺ T helper cells; ILC1s produce the type 1 cytokine interferon-γ (IFN-γ), ILC2s produce the type 2 cytokines interleukin-5 (IL-5) and IL-13, and ILC3s produce the type 3 cytokines IL-17 and IL-22. NK cells are the predominant population of ILCs with cytolytic functions.

NK cells are present primarily in the peripheral blood, spleen and bone marrow, but they can infiltrate tissues and are also found in the liver, lungs, gut, lymph nodes and uterus.² Recruitment to inflamed tissues depends on the expression of several chemokine receptors, such as CX3CR1, CXCR2 and CXCR1. In mice, NK cells are defined as CD3⁻NKp46⁺CD127⁻.^1,3 In human peripheral blood, NK cells are defined as CD14⁻CD19⁻CD3⁻ and can be divided into two populations, CD56^bright and CD56^dim NK cells, on the basis of their level of expression of the CD56 antigen^4,5 (see Table 1). CD56^dim NK cells express CD16 (FcγRIIIA) and combine cytotoxicity and cytokine production as effector functions, while CD56^bright NK cells are more specialized in the secretion of cytokines. NK cell effector functions are tightly regulated by the balance between activating and inhibitory signals delivered by cell surface receptors.⁵ Accordingly, NK cells express activating and inhibitory receptors that bind to major histocompatibility complex (MHC) class I molecules, including killer cell immunoglobulin-like receptors (KIRs) in humans and Ly49 in mice.^4–6 These receptors are highly polymorphic. In contrast, the inhibitory receptor CD94/NKG2A and the activating receptor CD94-NKG2C are invariant. CD94-NKG2A dimers recognize human leukocyte antigen (HLA)-E in humans and Qa-1 in mice. The inhibitory receptors KIR, Ly49 and CD94-NKG2A detect self-MHC-I ligands on healthy cells and help prevent NK cell activation and cell killing. Several activating receptors, including CD16, NKG2D, SLAM-family members and the natural cytotoxicity receptors (NCRs), NKp30, NKp44 and NKp46, can be targeted to induce NK cell-mediated anti-tumor immunity.^5,7,8 NKp46 is highly conserved among mammalian species, whereas NKp30 and NKp44 are expressed only in humans. Human NK cells express NKp30 constitutively, whereas NKp44 expression is induced upon activation. The full activation of NK cells requires the coengagement of several different cell surface receptors.^9,10

Anti-tumor immunity of NK cells

NK cells form an immunological synapse with target cells, leading to the direct exocytosis of cytotoxic granules containing both perforin (a membrane-disrupting protein) and granzymes (a family of proteolytic enzymes), resulting in the specific lysis of target cells.¹¹ NK cells can also express membrane-bound tumor necrosis factor (TNF) and death-inducing ligands, such as FAS ligand and TNF-related apoptosis-inducing ligand (TRAIL), upon activation. Cytotoxicity is one of the keys to anti-tumor immunity, as shown by mice lacking perforin, which lose control of tumor growth.^12,13 In addition to their cytotoxic activities, NK cells also secrete cytokines, such as IFN-γ, TNF-α, growth factors (e.g., granulocyte–macrophage colony-stimulating factor) and chemokines (CCL3, CCL4 and CCL5), which shape the innate and adaptive immune responses.¹⁴ Chemokine secretion underlies NK cell colocalization with other hematopoietic cells, such as dendritic cells (DCs), in inflamed tissues.¹⁵ Furthermore, IFN-γ production helps to shape T cell responses in lymph nodes,¹⁶ activates myeloid cells and has effects on angiogenesis.¹⁷ The NK cell-mediated killing of target cells also affects T cell anti-tumor responses, possibly by promoting antigen cross-presentation to CD8⁺ cytotoxic T cells.¹⁸

In the 1980s, several studies reported a higher incidence of cancers in individuals with defective NK cell function.^19,20 During the same period, mutant mice with impaired NK cell activity and mice treated with an NK cell-depleting antibody were reported to display higher rates of tumor growth and metastasis.^21,22 Many studies have since reported poor NK cell function in cancer patients,^23–25 and case reports have been published on NK cell deficiencies²⁶ that are characterized by an absence of NK cells or NK cell function or are caused by mutations of genes such as GATA2²⁷ or MCM4,²⁸ suggesting a link to higher rates of malignancy. However, these immunodeficiencies also affect cells other than NK cells, and the relationship between a high risk of cancer and low levels of NK cell activity in humans remains purely correlative. Recent studies showed that the incidence of cancers was not higher in NK cell-deficient patients than in the general population, but a compensatory mechanism involving redundant leukocyte populations could not be excluded.^29,30 In addition, the considered cohorts were too small, and the time of follow-up was insufficient to properly analyze the involvement of NK cells in tumor immunosurveillance. In mice, NK cells have been shown to control the growth and metastasis of transplantable tumors in studies based on the antibody-mediated depletion of NK cells.³¹ However, in many of these models, mouse NK cells were depleted with an antibody against NK1.1 that can also target ILC1s, subsets of activated T cells and, at least in theory, invariant natural killer T cells. Other studies used polyclonal anti-asialo GM1 serum, but asialo GM1 was detected on a subpopulation of NKT, CD8⁺ T and γδ T cells^32,33 and on some activated CD4⁺ T cells, macrophages, eosinophils and basophils under certain experimental conditions.^34–37 NKp46 was recently identified as the cell marker displaying the highest degree of restriction to NK cells, but even this marker is also expressed by discrete subsets of T cells (in particular γδ T cells), ILC1s and a fraction of ILC3s.¹ Therefore, there is no formal proof that NK cells are absolutely required for immunosurveillance against tumor formation under natural conditions. However, NK cells may have an antileukemic effect in patients undergoing allogeneic hematopoietic stem cell transplantation if donor NK cells express inhibitory KIRs mismatched with the patient’s cells for MHC class I molecules because they recognize recipient leukemia cells as foreign cells.^38–41 In addition, the adoptive transfer of genetically modified NK cells, checkpoint inhibitors and bispecific antibodies recently developed to target NK cells and tumor antigens are promising immunotherapeutic strategies that take advantage of the anti-tumor activities of NK cells to eliminate cancer.⁴²

The presence of NK cells in solid tumors

Several studies have shown that the natural cytotoxicity of peripheral blood is significantly lower in patients with various types of solid tumors than in healthy individuals.⁴³ The downregulation of the activating receptors NKp46, NKp30 and NKG2D at the cell surface of peripheral blood NK cells, and their magnitudes, were correlated with tumor progression in patients with cervical cancer.⁴⁴ However, little information is available concerning NK cell infiltration into tumors.

An early study reported a correlation between the risk of local recurrence after tumor resection and preoperative levels of NK cells in patients with colorectal carcinoma; low levels of these cells were associated with a higher risk of local recurrence than high levels of these cells.⁴⁵ More recently, colorectal tumors have been shown to be almost devoid of NK cells, despite efficient T cell infiltration.⁴⁶ Similarly, melanoma skin lesions display poor CD56⁺ NK cell infiltration (<1% of the infiltrating cells).⁴⁷ However, CD56^dimKIR⁺CD57⁺ NK cells, which are highly cytotoxic, were present in tumor-infiltrated lymph nodes,⁴⁸ and the levels of NKp30 and NKG2D expression were inversely correlated with the number of tumor cells in infiltrated lymph nodes adjacent to metastatic melanomas,⁴⁹ suggesting a possible role for NK cells in tumor control.

In breast cancer patients, tumor-infiltrating NK cell counts have been associated with the response to neoadjuvant treatments, including anti-HER2 mAbs,^50,51 and therapies with anti-HER2 trastuzumab–docetaxel and the antibody drug conjugate trastuzumab emtansine (T-DM1).^52–54 Recently, baseline NK tumor-infiltrating lymphocytes (TILs) were defined as independent biomarkers with great predictive value for complete response to anti-HER2 antibody-based treatment.⁵⁵ NK cell infiltration of breast tumors is thus an indication of good prognosis.

NK cells invade clear cell renal tumors, and this invasion was associated with a favorable prognosis.⁵⁶ NK cells were distributed throughout the tumor mass, rather than being restricted to the surrounding stroma.^57,58 In this pathology, infiltrating NK cells were abundant but only weakly cytotoxic. Infiltrating NK cells had a CD56^bright phenotype and overexpress NKG2A, but they exhibited lower levels of KIRs and leukocyte immunoglobulin-like receptor subfamily B member 1 (LILRB1) than autologous peripheral NK cells.

Head and neck (H&N) cancers are infiltrated with NK cells (Fig. ​(Fig.1),1), and this infiltration was associated with longer survival. Poor peripheral blood NK cell function was correlated with the development of metastases in H&N cancer⁵⁹ and pharyngeal cancer.^25,60 NK TILs displayed a decreased frequency in the mature CD56^dimCD16^hi population, with a concomitant decrease in expression of the maturation marker CD57 (Figs. ​(Figs.22 and ​and3).3). The frequencies of NK TILs expressing the inhibitory KIRs, KIR3DL1 and KIR2DL1/2/3, were lower among infiltrating than among circulating NK cells, but we observed an increase in the frequency of NK TILs expressing the inhibitory receptor NKG2A (Fig. 2 and see ref.⁶¹).

Open in a separate window

Fig. 1

NK infiltration in head and neck squamous cell cancer and lung small cell cancer samples Number of NKp46⁺ cells per mm² measured by IHC in tumor samples and normal adjacent tissue (NAT) and percentage of tissue slice with NK cells are indicated (A and B, left panels). NK cells frequencies defined as CD45⁺, CD3⁻, CD56⁺, CD4⁻ and/or CD14⁻ among CD45⁺ leukocytes were measured in blood, NAT and tumor samples was by flow cytometry. The box and whiskers represent the maximum, 75 percentile, median, 25 percentile, and minimum. Each dot indicates a sample

Open in a separate window

Fig. 2

Contour plots showing the positivity thresholds for indicated markers. NK cells were defined by flow cytometry as CD45⁺, CD3⁻, CD56⁺, CD4⁻ and/or CD14⁻ and represented by red contour plots. T cells defined as CD45⁺, CD3⁺, CD56⁻ are represented in blue contour plots. Representative expression of several markers, on blood or tumor samples from head and neck squamous cell cancer and non-small cell lung cancer patients, exemplifying the staining and the positioning of the thresholds are shown

Open in a separate window

Fig. 3

Expression profile of surface markers on NK cells from head and neck squamous cell cancer samples. NK cells were analyzed by flow cytometry and defined as CD45⁺, CD3⁻, CD56⁺, CD4⁻ and/or CD14⁻ in blood, normal adjacent tissue (NAT) and tumor samples. Frequencies of NK cells expressing indicated surface marker are shown. Box and whiskers represent maximum, 75 percentile, median, 25 percentile, and minimum. Each dot indicates a sample. For each marker the Wilcoxon test is used when matched data for the three kinds of samples were available, otherwise the Kruskal-Wallis test is used. Significant p-values below 0.05 are shown

Checkpoint blockade immunotherapy targeting the programmed cell death protein 1/programmed death-ligand 1 (PD-1/PD-L1) inhibitory axis produced remarkable results in the treatment of several types of cancer.^62–67 PD-1 is mostly expressed by T cells, but NK cells with an activated and more responsive phenotype can also express PD-1.^68,69 We detected a few NK TILs expressing PD-1 in H&N tumors (Fig. ​(Fig.33 and see ref.⁶¹). CD96 and TIGIT are both checkpoints currently under consideration for possible use in immunotherapy for cancer. We observed an increase in the frequency of NK TILs expressing CD96 but not TIGIT,⁷⁰ which are coinhibitory receptors recognizing nectin and nectin-like ligands overexpressed in various cancer cell types.⁷¹

NK cell infiltration per se has no impact on clinical outcome in non-small-cell lung cancer (NSCLC) (Fig. ​(Fig.11 and see refs. ^72,73). However, NK cells are the least abundant immune cell lineage in lung adenocarcinoma early lesions compared to normal lung tissue.⁷² NK cells infiltrated lung tumors (Fig. 1 and see refs.^72,73) and were found in the stroma on NSCLC biopsy, rather than in direct contact with tumor cells.⁷³ Nevertheless, the clinical outcome was more dependent on the NK cell phenotype and function than on the NK cell density in patients with NSCLC. Indeed, tumor-infiltrating NK cells had a profoundly impaired cytotoxic potential despite their expression of several activation markers, including NKp44, CD69 and HLA-DR (Fig. ​(Fig.44 and see refs.^72,73). However, other activating receptors, such as NKp30 and NKp80, were strongly downregulated on intratumoral NK cells, and this may impair cytotoxicity.⁷⁴ Furthermore, NSCLC samples had higher levels of CD56^hiCD16⁻CD57⁻ NK cells than the surrounding normal lung tissues, and in addition to NKG2A, NK cells from these samples expressed inhibitory KIRs that are normally expressed by CD56^lowCD16⁺ NK cells (Fig. ​(Fig.44 and see ref.⁷³). We observed frequencies of NK TILs expressing higher levels of CD96 than those in the surrounding tissues and on circulating NK cells (Fig. 4). We also detected a few NK TILs expressing PD-1 in lung tumors (Fig. 4). Overall, these results indicate that NK cells can invade some solid tumors. When NK cells infiltrate solid tumors, high numbers of NK cells may be associated with better survival. However, in most cases, anti-tumor properties of NK cells are attenuated at the tumor bed. Thus, the targeting of NK cells by immunotherapy is an attractive anti-cancer strategy.

Open in a separate window

Fig. 4

Expression profile of surface markers in natural killer (NK) cells from lung small-cell cancer samples. NK cells were analyzed by flow cytometry and defined as CD45⁺, CD3⁻, CD56⁺, CD4⁻ and/or CD14⁻ in blood, normal adjacent tissue (NAT) and tumor samples. Frequencies of NK cells expressing the indicated surface markers are shown. Box and whiskers represent the maximum, 75 percentile, median, 25 percentile and minimum. Each dot indicates a sample. For each marker, the Wilcoxon test was used when matched data for the three types of samples were available; otherwise, the Kruskal–Wallis test was used. Significant p values below 0.05 are shown

The tumor microenvironment represses NK cell recruitment and/or anti-tumor NK cell functions

Some tumors are poorly permeable to NK cells. Tumor cells themselves may affect the ability of NK cells to infiltrate the tumor mass via several mechanisms. For example, neuroblastoma cells have been reported to modulate the NK cell chemokine-receptor repertoire by producing transforming growth factor-β (TGF-β).⁷⁵ Neuroblastoma cells induce the upregulation of CXCR4 and CXCR3 expression on all NK cells and the downregulation of CX₃CR1 by the CD56^dim subset. Another mechanism by which tumors evade NK cell attack is based on the prevention of cell contact. Ovarian tumor cells interfere with the formation of immunological synapses with NK cells by expressing the mucin 16 (MUC16) glycoprotein, which has antiadhesion properties.⁷⁶ Tumor cells may also display altered patterns of activating or inhibitory receptor ligand expression, thus preventing NK cell activation. For example, melanoma cells become resistant to NK cell attack after multiple contacts through an upregulation of their surface expression of both classical and nonclassical HLA class I molecules.⁷⁷ Another well-known mechanism of NK cell-mediated lysis evasion involves the modulation of NKG2D ligands at the tumor cell surface or the secretion of the soluble NKG2D ligand MICA or soluble NKp30 ligands B7H6 and HLA-B-associated transcript 3/BCL2-associated athanogene 6 (BAT3/BAG6).^78,79 Finally, NK cell activities may also be downregulated by the engagement of the NKp44 receptor with proliferating cell nuclear antigen, which is overexpressed in various tumor types.^80,81

Cells from the tumor microenvironment may also have suppressive activities against NK cells, such as myeloid-derived suppressor cells, tumor-associated macrophages, which can be isolated from patients with hepatocellular carcinoma,⁸² or cancer-associated fibroblasts.^77,83,84 The physical status of the tumor microenvironment, including characteristics such as hypoxia and the production of various soluble factors by the tumor cells or other cells, may also impair NK cell function. For example, TGF-β, indoleamine 2,3-dioxygenase (IDO), IL-4 and prostaglandin E2 (PGE2) can interfere with NK cell activation.^77,85–87

NK cell-based immunotherapies for solid tumors

Efforts should be made to improve NK cell recruitment to the tumor bed. Tumor NK cell infiltration is an important parameter controlled by chemokines and adhesion molecules that may differ from those regulating T cell migration.^46,88 An antibody-based NK cell-homing protein, the NK cell-recruiting protein-conjugated antibody (NRP-body), has recently been developed.⁸⁹ The NRP-body increased NK cell infiltration along a CXCL16 gradient in a murine primary and metastatic pancreatic ductal adenocarcinoma tumor model. NRP-body administration increased the infiltration of transferred NK cells into tumor tissues and reduced tumor burden relative to that in controls, thereby increasing the overall survival. The combination of the NRP-body with NK cell therapy may, therefore, be of potential therapeutic value.

Other strategies consist of promoting NK cell activation. This is, for example, achieved using cytokine treatments. NKTR-214 is a PEGylated IL-2 that preferentially activates IL-2Rβ.⁹⁰ In preclinical models, NKTR-214 increased the expansion of the NK cell population and strongly increased the CD8⁺ T cell to regulatory T cell ratio. NKTR-214 is currently in phase I/II clinical trials for various solid tumors (NCT02983045, NCT02869295, NCT03138889, NCT03435640 and NCT03282344).

As mentioned above, activated NK cells have been shown to express PD-1 in certain circumstances.^91,92 NK cells from patients with multiple myeloma express PD-1, and the anti-PD-1 antibody (CT-011) could potentially be used to restore NK cell-mediated anti-tumor activity.⁹³ A combination of anti-PD-1 and anti-CTLA4 mAbs has recently been shown to be effective for the treatment of melanoma.^65,94 Monalizumab is an IgG4-blocking mAb directed against NKG2A, which, in combination with the anti-PD-L1 antibody, promotes NK cell effector functions and antibody-dependent cell-mediated cytotoxicity (ADCC) and effector T cell responses.⁶¹ A combination of monalizumab and cetuximab (anti-epidermal growth factor receptor (EGFR)) gave positive results in a phase II clinical trial in patients with previously treated squamous cell carcinoma of the head and neck.⁶¹ Monalizumab is being investigated in several studies, particularly in combination with durvalumab, an anti-PD-L1 antibody, in solid tumors (phase I, NCT02671435).

Another approach to direct NK cell-mediated cytotoxicity against tumors was achieved through chimeric antigen receptor (CAR) expression, which is an interesting strategy for treating refractory cancers. Various tumor antigens, including EGFR, epithelial cell adhesion molecule, human epidermal, ganglioside GD2, growth factor receptor 2 (HER2; also known as ERBB2), the inactive tyrosine-protein kinase transmembrane receptor ROR1 and Wilms’ tumor protein (WT1), have been targeted by CAR NK therapy with exciting clinical results.⁹⁵ Another approach based on NKG2D-expressing CAR NK cells promotes anti-tumor NK cell activity in an osteosarcoma mouse model.⁹⁶ These cells have been tested in one phase I study in patients with metastatic solid tumors.⁹⁷

Finally, bispecific killer cell engagers have been designed to promote the NK cell-mediated lysis of tumor targets. Two bispecific CD16-based antibodies (AFM22 and AFM24) bind to the EGFR variant III (which is expressed by several types of solid tumors, including glioblastoma) and wild-type EGFR, respectively, and are in preclinical development. Other bispecific antibodies have been generated and are in preclinical tests (Gauthier et al., unpublished data). These NK cell engagers (NKCEs) target the activating NK cell receptor NKp46 together with a tumor antigen and an Fc fragment to promote ADCC via the activating receptor CD16, which is expressed on NK cells. These NKCEs were more potent in vitro than clinical therapeutic antibodies targeting the same tumor antigen. In vivo, these antibodies efficiently controlled tumor growth in mouse models of solid and invasive tumors. Interestingly, these antibodies promoted the infiltration/proliferation of NK cells within tumors.

Fuzzy frontiers between NK cells and ILC1s

NK cells and ILC1s have different developmental paths, but they share several features; however, the properties differentiating NK cells from ILC1s are still poorly defined, particularly in humans. NK cells are cytotoxic cells that circulate in the bloodstream, whereas ILC1s are tissue-resident cells.^98,99 ILC1s are generally weakly cytotoxic or noncytotoxic and act as a first line of defense against infections with viruses and certain parasite and bacteria, such as Toxoplasma gondii¹⁰⁰ or Clostridium difficile.¹⁰¹ Both of these cell types mainly produce IFN-γ. The phenotypic characterization of ILC1s is often problematic. In both humans and mice, ILC1s preferentially express CD49a and TRAIL, but the specificity of these markers is tissue dependent, and their expression is often lost upon cell activation. ILC1s share some phenotypic markers with NK cells and ILC3s, such as NK1.1 in mice, and NCRs, such as NKp44 in humans and NKp46 in both humans and mice, in at least some organs (Table 1). In humans, CD127 may not be an absolute ILC1 marker because a majority of CD56^bright CD16⁻ NK cells in peripheral blood express CD127, as ILC1s. In contrast, highly cytotoxic CD56^dim CD16⁺ NK cells do not express this marker. CD200R expression has been shown to distinguish ILC1s from NK cells in mice.¹⁰² Another marker of human NK cells that is not expressed on ILC1s is NKp80.¹⁰³ Human ILC1s are, thus, negatively selected as Lin⁻CD127⁺CRTH2⁻cKit⁻NKp80⁻ cells in flow cytometry, and specific positive staining is not currently possible in immunohistochemistry.

Recent studies have shown that NK cells can be converted into ILC1s (converted ILC1s),¹⁰⁴ ILC3s can acquire an ILC1-like phenotype (ex-ILC3)^105,106 and markers previously used to differentiate ILC populations are often specific to particular immune contexts or tissues. These observations are of particular interest because converted ILC1s may facilitate tumor growth in the TGFβ-rich environment of mouse models of solid tumors, as TGFβ induces the conversion of CD49b⁺CD49a⁻EOMES⁺ NK cells into CD49b⁻CD49a⁺EOMES⁻ ILC1s.¹⁰⁴ Tumor ILC1s express higher levels of inhibitory receptors, including NKG2A, killer cell lectin-like receptor subfamily G member 1 (KLRG1), cytotoxic T lymphocyte antigen 4 (CTLA4) and lymphocyte activation gene 3 protein (LAG3), and they have low levels of DNAM1 and increased levels of TIGIT, potentially enabling tumors to escape immune surveillance by the innate immune system.^104,107 Given the potential pro-angiogenic role of ILC1s, further work in the phenotypic characterization of these cells is clearly required.

Conclusions

Unlike T cells or dendritic cells, NK cells are often underappreciated at the tumor bed, leading to speculation that these cells might only be involved in the control of hematologic malignancies and tumor metastasis. However, the cytolytic activities of NK cells promote CD8⁺ T cell cross-priming, and NK cell-derived cytokines and chemokines attract and activate DCs, macrophages, and neutrophils. NK cells can, therefore, make a significant contribution to anti-tumor immune responses. Remarkably, high levels of NK cell infiltration have been associated with a better prognosis, at least for some tumors. Thus, it is of interest to promote NK cell infiltration of tumors and to develop immunotherapeutic approaches designed to harness NK cell functions in the tumor bed, including those that aim to reinforce conventional anti-tumor therapies to increase the chances of successful treatment.

Acknowledgements

Competing interests

G.H., P.A. and E.V. are employees of Innate Pharma. The other authors declare no competing interests.

References