Table 1
NK cells are large granular lymphocytes with a diameter ranging from 7 to 12 µm, depending on the activation state. Upon staining, these cells are easily recognised among other circulating lymphocytes by their reniform nucleus and by the presence of azurophilic granules in the cytoplasm (Whiteside & Herberman, 1994). These characteristics are not exclusive of NK cells, as activated cytotoxic T cells may also present such phenotype. Therefore, for the correct identification of the NK phenotype, CD3 must be absent, and CD56 (neural cell adhesion molecule, NCAM) and CD16 (FcγRIII) must be present. Moreover, NK cells also lack surface immunoglobulins (Ig) and constitutively express IL-2 receptors (Nagler, Lanier & Phillips, 1990; Whiteside & Herberman, 1994). Many other surface markers are present in NK cells, as well in other lymphocytes. However, the unique combination of the CD3- CD56+ CD16+phenotype defines human NK cells (Abakushina, 2015). Additionally, two subsets of NK cells may be defined, depending on the relative expression of CD16 and CD56. These are called CD56brightCD16dim/- and CD56dimCD16+, where bright and dim are associated with high and low levels of expression, respectively. The first subset is outnumbered by the second in circulation but constitutes the majority of NK cells in lymphoid tissues. These subsets present different expression levels of other markers and receptors and different cytotoxic activities, with CD56bright being significantly less cytotoxic than CD56dim cells (Poli, Michel, Theresine, Andres, Hentges & Zimmer, 2009). It has been hypothesized that CD56bright cells are an immature form of NK cells that differentiate into CD56dim NK cells, which participate in natural and antibody-mediated cell cytotoxicity (Chan et al., 2007). On the other hand, CD56bright NK cells express higher levels of cytokine receptors, such as the IL-2 receptors αβγ and βγ (IL-2Rαβγ and IL-2Rβγ, respectively), higher levels of some activating and inhibiting receptors, and chemokine receptors. These cells produce higher levels of cytokines upon stimulation, namely of IL-10, an immunosuppressive cytokine, and therefore are thought to have immunoregulatory properties (Cooper et al., 2001).
NK cells also display a broad range of cell adhesion molecules (CAMs) essential for their cytotoxic activities, as these participate in interactions between NK cells, target cells, and accessory cells to generate effective immune responses. Some of these CAMs are upregulated upon activation, including lymphocyte function-associated antigen 1 (LFA-1; CD11a/CD18), LFA-3 (CD58) and intracellular adhesion molecule-1 (CD54). Interestingly, upon in vitro stimulation with IL-2, the cell surface levels of all these CAMs increase, as do the levels of CD56. Even more interesting is the fact that, upon this increase in CAM expression, the cytotoxic activity of NK cells towards NK-sensitive target cells also increases, and cells that were previously non-susceptible to NK-killing become targets (Robertson, Caligiuri, Manley, Levine & Ritz, 1990).

The cytotoxic response of NK cells

The cytotoxic response of NK cells upon activation is divided into 4 steps – formation of the immunological synapse, microtubule re-organization, lysosome docking to the membrane, and lysosome fusion with the membrane (Paul & Lal, 2017).
In the first, an interface between the NK cell and the target cell is formed, deemed immunological synapse. This occurs when the NK cell approaches the target, either accidentally or “intentionally” due to chemotactic signalling.
Upon contact, CD2 molecules present on the surface of NK cells recognise the presence of stage-specific embryogenic antigen 1 (SSEA-1, also known as CD15 or Sialyl-LewisX). At this point, if the target cell presents markers for NK cell inhibition, the formation of the immunological synapse is halted. Conversely, if activating receptors are present, tight adhesion between the two cells is promoted by receptor-ligand interactions of high affinity. LFA-1 and MHC-1 molecules present on NK cells bind ligands present on the target cell surface. The formation of these receptor-ligand complexes is enough to activate the cytolytic response, to some extent. However, these processes are considered more relevant in maintaining the immunological synapse than in triggering NK cell cytotoxicity. Full activation of the cytotoxic response depends on the engagement of specific receptors, such as the natural cytotoxicity receptors, present on the surface of NK cells. The immunological synapse is shaped in a way that both cells form a ring-shaped interface. Within this ring, cytotoxic granules and other cytotoxicity mediators are released directly and in a controlled fashion towards the target-cell surface (Orange, 2008; Stinchcombe & Griffiths, 2007). The secretory lysosome exocytosis requires reorganization of the cell cytoskeleton. In this step, the microtubule organizing centre (MTOC) becomes polarized and the secretory lysosomes are transported along the microtubules towards the synapse. Upon reaching the cell membrane, the granules dock and fuse, releasing the contents towards the target. This formally constitutes degranulation (Paul & Lal, 2017; Topham & Hewitt, 2009).

The molecular mechanisms of target-cell killing

As referred before, contrarily to T cells, NK cells do not rely on the somatic rearrangement of receptor genes to accommodate the expression of a variety of receptors. Instead, NK cell receptors are germ-line-encoded and consistently expressed (Biassoni, 2008; Whiteside & Herberman, 1994). The homeostasis of NK cell activity is ensured by a specific set of receptors with activating and inhibitory activities, some of which overlap. Inhibitory receptors contribute to self-tolerance of NK cells, preventing the lysis of normal healthy cells. On the other hand, activating receptors trigger the lytic activity of these cells, prompting the destruction of cells that present activating ligands.
Upon activation, NK cells release a series of lytic enzymes through the degranulation process already described, namely perforins and granzymes. Granzymes induce target-cell apoptosis, but their actions depend on being appropriately delivered by perforins (Boivin, Cooper, Hiebert & Granville, 2009). Together with these enzymes, a wide variety of cytokines is also released, including interferon γ (IFN-γ), tumour-necrosis factor α (TNF-α), granulocyte-macrophage colony-stimulating factor (GM-CSF), IL-10, IL-5 and IL-13, chemokine macrophage inflammatory proteins 1α and 1β (MIP-1α, MIP-1β), IL-8, and chemokine (C-C motif) ligand 5 (CCL5, also known as RANTES) (Paul & Lal, 2017).
Besides the direct effect of granzymes (and perforins), cytokines such as IFN-γ also play an important role in inducing target-cell death. IFN-γ, as seen before, is an important activator of macrophages, as well of APCs. Stimulation of the latter, as shown before, upregulates the expression of cytokine IL-12, as well as IL-18 and co-stimulatory molecules CD86, that enhance Th1 differentiation and CTL function. IFN-γ can also exert antiproliferative effects on tumour cells by enhancing the expression of the cell cycle inhibitor proteins p27Kip, p16 or p2 (Ni & Lu, 2018). TNF-α is also implicated in the collapse of tumour vasculature, but can additionally induce tumour regression by triggering apoptosis, T cell activation (by Tregs’ blockage), and neutrophil and monocyte chemoattraction to the tumour region, and by downregulating the immunosuppressive phenotypes of tumours (Josephs et al., 2018).
GM-CSF is a part of the inflammatory cascade, recruiting monocytes and inducing the differentiation, proliferation, and migration of granulocytes (neutrophils, eosinophils and basophils) to the inflammation site. It is also essential for the differentiation of dendritic cells that, as seen before, are responsible for processing and presenting of tumour antigens and for activating cytotoxic T lymphocytes (Yan, Shen, Tien, Chen & Liu, 2017).
IL-5 and IL-13 are associated with B cell growth, eosinophil activation, and regulation of inflammatory and immune responses (Minty et al., 1993). IL-10, on the other hand, seems to have a pleiotropic effect on the immune response. It downregulates Th1 responses, cytokine secretion, such as TNF-α, IFN-γ and IL-12, and CD4+ T cell activation (de Waal Malefyt, Abrams, Bennett, Figdor & de Vries, 1991; de Waal Malefyt et al., 1991). However, IL-10 can also induce the secretion of granzymes and perforins by CD8+ T cells and potentiate TCR-dependent IFN-γ secretion (Emmerich et al., 2012). Lastly, MIP-1α/1β, IL-8 and CCL-5 have chemotactic activities towards granulocytes, neutrophils and T cells, eosinophils and basophils (Kohidai & Csaba, 1998; Wolpe et al., 1988).
Through the secretion of these cytokines and chemokines, NK cells are able to recruit other immune cells, namely dendritic cells that can infiltrate the tumour tissue and trigger robust and sustained immune responses (Bottcher et al., 2018).

Balance between activation and inactivation of NK cells

NK cells express a wide variety of receptors with activating and inhibiting functions that allow for the fine discrimination between healthy and ailing cells in a matter of seconds; the activating and inhibitory signalling pathways involved are depicted in Figure 1 . Most of the inhibitory receptors detect the absence of MHC I molecules, while activation receptors probe for the presence of specific ligands that flag the target cells as harmful (Leung, 2014). Because NK cells are extremely cytotoxic when active and can very quickly deliver their lytic response without prior sensitization, contrarily to other lymphocytes, a complex activation/inactivation system tightly regulates their responses. Activation depends on tipping the balance between stimulatory and inhibitory signals, an equilibrium that is dictated by the engagement of different receptors. After cell-to-cell contact, NK cells integrate, within seconds, the signals from their activating and inhibiting surface receptors. Normal cells, expressing normal MHC I molecules, will engage the inhibitory receptors, rendering NK cells unresponsive, without compromising their functionality (licensing ). Cells lacking MHC I and presenting surface activating ligands will trigger the cytolytic response. A dynamic equilibrium is reached when the target cells present both MHC I molecules and activating ligands. In this case, no response is triggered as the positive and negative feedback loops cancel each other. If more activating ligands are engaged, the activation signal dominates and target-cell lysis occurs (Figure 2 ) (Vivier, Ugolini, Blaise, Chabannon & Brossay, 2012).
The engagement of different receptors is translated into activating and inactivating signals through ITIM and ITAM motifs, present in either the receptors themselves or in adaptor proteins. The ITAM and ITIM signalling cascades are combined intracellularly to determine the response of NK cells. Upon binding of activating ligands, the ITAM motifs are phosphorylated, triggering the recruitment of Syk-family kinases and the subsequent activation of cytolytic responses. However, engagement of inhibitory receptors also leads to the phosphorylation of ITIM motifs on these receptors, a process that is promoted by the same Src kinase responsible for ITAM phosphorylation. These now phosphorylated ITIMs recruit the tyrosine phosphatases SHP1 or 2, which can in turn terminate intracellular signals emanating from ITAM signalling receptors via their phosphatase activity, rendering NK cells inactive (Figure 2 )(Linnartz-Gerlach, Kopatz & Neumann, 2014).