Figure 3
Although initially overlooked, this stalk domain has been shown to play an important role in the receptor activity, as it increases ligand binding affinity (Hartmann et al., 2012). Initially considered to include 5 to 6 residues (Pende et al., 1999), the stalk domain considered here is a longer 15-residue domain (129KEHPQLGAGTVLLLR143) (Memmer et al., 2016). This stalk domain is very sensitive to point mutations, that lead to impaired binding and activity. The last residue of this domain is a cationic arginine. Arg143 is located on the membrane extracellular interface, but upon ligand binding it is translocated into the membrane, strengthening the interaction with the CD3ζ and FCεRIγ Asp residues (Li, Wang & Mariuzza, 2011; Memmer et al., 2016).
The initial structural elucidation of NKp30 identifies it to be monomeric in the crystal phase, in agreement with its behaviour in size exclusion chromatography and on the cell surface, but later work identifies it as a dimer (Joyce, Tran, Zhuravleva, Jaw, Colonna & Sun, 2011; Pende et al., 1999). However, the authors also point out that many similarities between NKp30 and the homodimeric NKG2D receptor suggest that NKp30 can exist as a homodimer in which the binding site is formed by the two identical subunits (Li, Wang & Mariuzza, 2011).
NKp30’s closest human homologues are CTLA-4 (Kaifu, Escaliere, Gastinel, Vivier & Baratin, 2011), a homodimeric member of the CD28 family, and PD-1, an immune checkpoint protein that guards against autoimmunity(Francisco, Sage & Sharpe, 2010; Joyce, Tran, Zhuravleva, Jaw, Colonna & Sun, 2011); a simple homology modelling approach based on the known structure of CTLA-4 originated a homodimeric NKp30 structure, in which the subunit interaction is dominated by a network of salt bridges and hydrogen bonds between Glu8, Arg10, Glu13, Arg106 and Glu110 of both monomers(Kaifu, Escaliere, Gastinel, Vivier & Baratin, 2011). However, this dimer is formed by a top-top approach of the subunits and is unlikely to be relevant in vivo. It has recently been shown that the formation of NKp30 ectodomain dimers leads to a higher affinity towards the B7-H6 ligand. This ectodomain dimerization appears to be driven by the association of the stalk domains of both polypeptide chains (Herrmann, Berberich, Hartmann, Beyer, Davies & Koch, 2014).
Superposition of the NKp30 and the NKp44 structures revealed common features as well as distinct conformations. The core regions of NKp30 and NKp44 are quite similar, but significant differences are found at the interacting region between the two β-sheets, which in the case of NKp30 presents six more H-bonds. An additional α-helix is found between strands in NKp30 that is absent in NKp44. Furthermore, NKp30 lacks the extensive positively charged region that is found in NKp44. Sequence homology is about 30% between NKp30 and NKp44 whilst there is no significant overlapping with NKp46 (Foster, Colonna & Sun, 2003; Joyce, Tran, Zhuravleva, Jaw, Colonna & Sun, 2011; Li, Wang & Mariuzza, 2011).
As previously mentiones, signal transduction occurs through the ITAM-bearing adaptor proteins CD3ζ and FCεRIγ. Like the other NCRs, association with the adaptor proteins is promoted by opposing charge residues. NKp30 and NKp46 contain positively-charged Arg residues that interact with Asp residues in CD3ζ and FCεRIγ. On the other hand, NKp44 contains positively-charged Lys residues that interact with Asp residues in DAP12 (Koch, Steinle, Watzl & Mandelboim, 2013). Interestingly, NKp30 has been shown to be expressed in several variants (NKp30a , b , c ,d , e , f ) resultant from alternative splicing. Cells transfected with the variants NKp30a and NKp30b were found to produce high amounts of IFN-γ upon NKp30 engagement. Conversely, cells transfected with the NKp30c variant produced small amounts of IFN-γ but high levels of IL-10, the immunosuppressor interleukin. This variant seems to interact poorly with the CD3ζ adaptor protein, in contrast with the a and b variants, which could justify the dual immunomodulatory effect of NKp30 (Delahaye et al., 2011; Siewiera et al., 2015). Moreover, one work has demonstrated that expression of higher levels of the c variant is directly correlated with poor prognosis in gastrointestinal tumour patients (Delahaye et al., 2011).
The role of the positively charged Arg143 residue in the end of the stalk domain in the association of the receptor with CD3ζ, together with the observation that CD3ζ increases cell membrane expression of NKp30, has led to the proposal that this receptor exists in two different assemblies – a signalling-incompetent one and a ligand-induced signalling-competent complex (Memmer et al., 2016), which is consistent with the licensing hypothesis for NK cell response (Yokoyama & Kim, 2006).
It is also important to mention that NKp30 is present in the extracellular surface of the cell as a glycoprotein and that different glycosylation patterns affect the affinity towards its ligands. For instance, glycosylation of Asn42 increases the binding affinity of BAT-3 but severely hampers the interaction with B7-H6 (Hartmann et al., 2012). These structural features, which may vary between individuals, may be the cause of some pathologies and the reason why some patients develop aggressive malignancies even when their NK cell population is normal.
The NKp30 receptor is the product of the NCR3 gene expression, which can be transcribed as four exons that undergo alternative splicing(Delahaye et al., 2011; Shemesh, Brusilovsky, Kundu, Ottolenghi, Campbell & Porgador, 2018). Up to this date, only isoforms NKp30 a to c, resulting from alternative splicing of exon 4 and corresponding to different intracellular terminal domains, have been studied. NKp30a and NKp30b present a stronger association to the intracellular adaptor CD3ζ (and also to FcεRI-γ), while NKp30c exhibits a lower association to either (Delahaye et al., 2011; Kaifu, Escaliere, Gastinel, Vivier & Baratin, 2011).
NKp30a and NKp30b are typically associated with an immunostimulatory activity of NK cells, and are associated with overall improved survival rate and with better prognosis in gastrointestinal stromal tumours (Delahaye et al., 2011; Rusakiewicz et al., 2017) and in hepatocellular carcinoma (Mantovani et al., 2019), in contrast to a higher morbidity observed in patients that predominantly express isoform NKp30c. A similar trend is observed in the case of some viral infections, namely in chronic hepatitis virus C infection (Mantovani, Mele, Oliviero, Barbarini, Varchetta & Mondelli, 2015), while no clear association between NKp30 expression status and outcome could be found in HIV-1 patients (Prada et al., 2013).
This isoform-prognostic correlation is particularly evident in paediatric neuroblastoma patients, where the pattern of isoform expression is highly correlated to a 10-year event-free survival in multiple cohorts (Semeraro et al., 2015; Semeraro, Rusakiewicz, Zitvogel & Kroemer, 2015). However, this requires further study, as a later clinical study addressing paediatric neuroblastoma response to imatinib treatment failed to identify any correlation between NKp30 isoforms and clinical response (Morandi et al., 2018).
While isoforms NKp30a, NKp30b and NKp30c present a V-type Ig-like extracellular domain, splice variants NKp30d, NKp30e and NKp30f encode a different C-type Ig extracellular domain, coming from an exon 2 alternative splicing, with aminoacid residues 66 to 90 being absent (Hollyoake, Campbell & Aguado, 2005; Kaifu, Escaliere, Gastinel, Vivier & Baratin, 2011; Neville & Campbell, 1999; Shemesh, Brusilovsky, Kundu, Ottolenghi, Campbell & Porgador, 2018). The intracellular domains, resulting from exon 4 expression, are shared between isoforms a and e, between isoforms b and d, and between isoforms c and f (Shemesh, Brusilovsky, Kundu, Ottolenghi, Campbell & Porgador, 2018). A seventh isoform, computer-generated, that matches experimental evidence at protein level, but corresponds only to the extracellular Ig domain of NKp30, has been proposed (Uniprot accession number A0A0G2JKT7) but has not been confirmed yet.

NKp30 polymorphisms in disease

The outcome variability associated with immunotherapy regimens is most likely of multifactorial origin, and a contributing factor is the variability of the immune cell receptors involved in recognition and signalling. A large number of single nucleotide polymorphisms (SNP) of the NKp30-coding NCR3 gene have been identified, but only a very small number of them have been studied, particularly regarding population resistance to parasite infection (Delahaye, Barbier, Fumoux & Rihet, 2007; Hermann et al., 2006).
A susceptibility locus for mild malaria, a major cause of morbidity and mortality in many developing countries, is located in the MHC region in chromosome 6p21, where TNF polymorphisms have been associated with mild malaria. The NCR3 gene is located just 15 kb distal toTNF , and early studies have shown that theNCR3  ‑412G>C SNP is associated with increased risk and frequency of mild malaria (Delahaye, Barbier, Fumoux & Rihet, 2007), when compared to the standard NCR3 ‑412G allele. A later study confirmed this, but also showed that this association only occurs for mild malaria cases, and not for severe malaria cases (Thiam et al., 2018). Noticeably, STAT4, a transcription factor essential for IL-12 mediated cytotoxicity and IFN-γ production in mouse and human NK cells, binds this promoter region with higher affinity for the G allele (Baaklini et al., 2017). In agreement with this, purified NK cells were found to lyse Plasmodium falciparum -parasitized erythrocytes, through a direct interaction of NKp30 with P. falciparumerythrocyte membrane protein–1 (Pf EMP-1) (Mavoungou, Held, Mewono & Kremsner, 2007). A more recent study with healthy human volunteers found that parasitemia levels and NK cell activation follow the same trend during the first 18 days post-challenge by the bites of five P. falciparum 3D7 strain-infected Anophelesmosquitoes, in what is the first study identifying the role of the NKp30 protein against P. falciparum (Walk & Sauerwein, 2019).
When considering the role of NKp30 on Trypanosoma cruziinfections, Hermann et al. found that cord blood NK cells from newborns congenitally infected with T. cruzi had a reduced expression of NKp30, which could be due to earlier NK cell activation or to NK cell activity down-regulation by T. cruzi itself (Hermann et al., 2006). Later work has shown that NK cells were activatedin utero , during the first exposure to the parasite, and that the fetuses response to T. cruzi is an adult-like one, based on IFN-γ production by CD8+ T cells and through an IL-12 dependent monocyte pathway (Guilmot, Bosse, Carlier & Truyens, 2013; Hermann et al., 2006; Hermann et al., 2010; Sathler-Avelar et al., 2003).
Two other NKp30 SNPs in the promoter region have been studied in healthy Japanese, the ‑201G>A and ‑163G>C, but no transcription binding factors are known at these sites; a second set of SNPs were found in the same population, c.111G>A and c.156C>T, but these are synonymous substitutions, and so the phenotype remains unchanged(Sato et al., 2001).

Ligands for NKp30

The NKp30 receptor may be engaged by several ligand types. Interestingly, and even though NKp30 is considered an activating receptor, some ligands have been associated with the opposite response. This is the case of Poxvirus haemagglutinin (HA) that was found to bind NKp30, acting as antagonist. On the other hand, these same antigens are able to trigger the NKp46-mediated response (Jarahian et al., 2011). Human cytomegalovirus tegument protein pp65 (HCMV pp65) is also able to block the triggering signal of NKp30 engaging. This protein interacts directly and specifically with NKp30 without blocking the interaction of this receptor with its activating ligands. Despite being able to form NKp30-activating ligand complexes, the killing action of NK cells is suppressed by pp65, which suggests a downstream effect of this protein on the stimulatory signal transduction. It was demonstrated by Arnon and co-workers that pp65 disrupts the interaction of NKp30 with the adaptor protein CD3ζ, directly reducing the activation state of NK cells (Arnon et al., 2005). The other known ligands for NKp30 have activation effects and their engaging results in cytokine production and perforin and granzyme release. For instance, it was found that molecules expressed on the surface of Plasmodium falciparum -infected erythrocytes can be recognised by NKp30 and NKp46. The P. falciparum erythrocyte membrane protein 1 (PfEMP-1) was identified as a strong activating ligand of NKp30 and NKp46 that, upon engagement, results in the lysis of the infected cells (Mavoungou, Held, Mewono & Kremsner, 2007).
Despite the importance of these virus- and parasite-infected cell destruction activities, researchers have focused primarily on NK cell tumour-targeting. One crucial finding stimulating this research field was the discovery of heparin and heparan sulphates as activating ligands for NCRs. An initial research paper described, in 2004, that membrane-associated heparan sulphate proteoglycans were involved in the recognition of cellular targets by NK cells. In fact, the NCR NKp44, as referred before, bears an extensive positively charged surface that may constitute a binding site for sulphated carbohydrate structures (Bloushtain et al., 2004). Heparin and heparan sulphate sequences have formal negative charges (in their sulphate groups) that would account for strong interactions. Binding specificities of NKp30, 44 and 46 to these carbohydrates were further investigated, revealing that different receptors bind different monomer sequences (Hecht et al., 2009). However, since NKp30 does not possess a positively charged region, contrary to NKp44, and, as stated before, there is no structure homology between the receptors, the existence of other binding sites in the receptors or the cooperation of ligands to induce the activation response, seems plausible. In fact, some evidences indicate that heparan sulphates are able to bind NCRs but unable to induce cytolysis on their own. On the other hand, the diversity of heparan sulphate molecules with different sequences and chemical moieties may account for the inconsistent results that have been presented, as these sequences might also interact with inhibitory receptors (Hershkovitz et al., 2008). Besides the wide diversity of heparan sulphate sequences, it is also worth to mention that different NCR glycosylation states may play an important role in the binding of different molecules, especially large molecules that assume secondary and tertiary conformations, like polysaccharides (Brusilovsky, Radinsky, Yossef, Campbell & Porgador, 2014; Hershkovitz et al., 2008).
Other molecules that are also expressed by tumour cells and act as activating ligands of NKp30 are the proteins BAT3 and B7-H6, as seen before. BAT3 (HLA-B associated transcript 3, also identified as BAG-6) is overexpressed by tumour cells in response to stress signals. Under these conditions, the protein is translocated to the cell membrane and released from the cells, engaging the receptor NKp30. Interestingly, cytokine release is only experimentally attained using tumour-released BAT3 and no activation of NK cells is observed when a recombinant protein is used. This suggests that BAT3 only induces the cell response when it is part of vesicles, such as exosomes (Pogge von Strandmann et al., 2007). On the other hand, the other known tumour-derived ligand of NKp30, B7-H6, is able to induce cytokine release in its soluble or membrane-bound form (Gutierrez-Franco et al., 2018; Phillips, Romeo, Bitsaktsis & Sabatino, 2016). Membrane expression of B7-H6 has been widely implicated in the sensitivity of cancer cells to NK cell-mediated cytolysis (Cao, Wang, Zheng, Wei, Tian & Sun, 2015). Conversely, NK cells chronically stimulated with soluble forms of B7-H6 have been shown to downregulate the expression of NKp30, contributing to tumour immune escape (Pesce et al., 2015). It seems that, besides the normal regulation mechanisms presented by NK cells, there is yet another layer of regulation dictating the activity of NK cells. Nevertheless, among all the known ligands for NKp30, B7-H6 has been regarded as a possible target for novel cancer therapies. The basis of the B7-H6-NKp30 recognition has been widely studied in several works, in an effort to highlight the molecular mechanisms behind the strong NK cell response elicited by tumours presenting B7-H6. Protein glycosylation also plays a role, as it has been shown that single, double and triple mutations, where residues Asn42, Asn68 or Asn21 were mutated to Gln, lead to increased KD values for B7-H6 binding, with the exception of the Asn121Gln, which practically does not affect KD values (Hartmann et al., 2012).

NKp30’s B7-H6 binding site

To understand the basis of tumour recognition by NKp30, several authors have tried to accurately describe the interactions between NKp30 and its natural ligands. As there is no significant sequence homology between the known NKp30 ligands, or similarity in their predicted structure folding, it is not possible to determine a plausible binding site by identifying common features in ligand structures. Site-directed mutagenesis was used to alter the surface charge and/or hydrophobicity of specific areas of NKp30, allowing the identification of a relatively small region where at least one of the known ligands – B7-H6 – interacts (Joyce, Tran, Zhuravleva, Jaw, Colonna & Sun, 2011). The B7-H6 binding region of NKp30 is comprised mostly of non-polar amino acid residues, that account for the hydrophobic interactions between the two proteins. Two hydrogen bonds are also involved, as well as a salt bridge. B7-H6 affinity is extremely sensitive to changes in the protein structure, as the interactions of the two proteins are restricted to a small region of NKp30 (residues Ile50-Val53, Leu80, Ser82, Phe85, Leu86, Glu111, Leu113 and Gly114), as seen in Figure 4 . In fact, selective mutation in a few residues results in marked changes in the receptor affinity towards the ligand – Ser52Arg and Ser52Ala mutants display a 150 and 14 fold affinitiy reduction, respectively, and Ile50Ala mutants display a 14 fold affinitiy reduction; also, the Phe85Ala and Leu86Ala mutants display a 53 fold reduction in the binding affinitiy (Li, Wang & Mariuzza, 2011).