NK cells, a class I act

Colin Brooks

 Department of Immunology, The Medical School, Newcastle NE2 4HH, U.K.

        In recent years natural killer (NK) cells have come to the forefront of immunological attention through the realization that not only do they reside at a critical evolutionary and functional interface between the innate and adaptive immune systems, but they possess recognition capabilities of a far more complex and intriguing nature than was ever imagined.  In the words of a recent review article, we have gone “From no receptors to too many” [1].

        The receptors that have attracted most interest are those that bind to MHC class I molecules [Figure 1], thereby conferring upon NK cells the capacity for MHC-restricted recognition.  The largest group of such recep-tors, currently comprising about 30 members, constitutes a sub-group of the Ig superfamily [IgSF] known as the KIR family.  Which of these molecules are allomorphs and which the products of distinct genes is still unclear.  The production of soluble forms of these molecules has led to the unambiguous demonstration that they do indeed bind to MHC class I molecules [2], and to the first crystal structure which revealed a previously unsuspected homology to hemopoietic growth factor receptors, the extracellular Ig domains being held at an acute angle to each other with the class I binding region at the apex of this structure [3].  Remarkably, a single amino acid within this region can determine class I specificity [4].  Surprisingly and uniquely amongst IgSF members, the extracellular domains are chelated to Zn atoms which are required in some way for signal transmission but not for surface ex-pression or class I binding [5].

        The second group of class I receptors on NK cells belongs to the C-type lectin superfamily.  One sub-group, the Ly49 family, first described in mice, is encoded by at least 14 distinct genes [6], with evidence of addi-tional variation generated by polymorphism and alternate splicing.  Ly49 molecules have now been identified in rats and humans, although the one human molecule found so far appears to be non-functional [7]. Surprisingly, NK cells from fetal mice lack expression of all known Ly49 molecules except Ly49E suggesting that this particular Ly49 member may play a critical role during NK cell development [8].  The finding that these immature NK cells are inhibited by class I expression on target cells further suggests either that Ly49E is a universal class I receptor or that other class I receptors are present.

        A candidate for the latter would be receptors belonging to the second subgroup of C-type lectin receptors on NK cells, the CD94 family.  These have now been identified, at least at the RNA level, in mouse NK cells [9].   In the human, they exist as heterodimeric structures containing a common CD94 chain associated with an NKG2 molecule; six forms of the latter have been identified at the RNA level, being the products of both alternate genes and alternate splicing.  Based on analysis of the killing of supposedly class I -ve lymphoblastoid cells transfected with individual class I molecules, it had been widely believed that CD94-based receptors had a broad specificity for HLA-A, B, and C molecules.  However, in one of the most remarkable developments in the NK field in recent years, it has now been shown that what the protective transfected HLA molecules had in common was a nonamer peptide in their leader sequences that promoted the assembly and expression of another, endogenous, non-classical class I molecule, HLA-E.  By using soluble tetrameric HLA-E constructs, it was shown directly that HLA-E could bind to CD94-containing complexes on NK cells [10].

        A striking feature of each of the NK class I receptor families is the presence of both inhibitory and activa-tory members [Figure 1].  It has been known for some time that the former have cytoplasmic domains containing ITIMs that upon phosphorylation [by as yet unknown kinases] become associated with intracellular dephosphorylation enzymes, such as SHP1 [that act on as yet unknown substrates].  By contrast, the activatory recep-tors lack ITIMs, and have transmembrane domains that contain a positively charged amino acid.  This latter feature is found in several other positive signalling molecules, including the TCR and certain Fc receptors, each of which signals via its association with CD3z and/or FcRg.  It has now been demonstrated that signalling by, and in many cases surface expression of, activatory NK cell receptors is dependent on association with a novel member of the CD3z/FcRg family named DAP12 [11,12].  Interestingly, mouse CD94, unlike human CD94, has a positively charged amino acid in its transmembrane domain [9], suggesting that murine CD94 complexes may function dif-ferently to their human counterparts.

        Given the current obsession of NK cell biologists with inhibitory receptors, the finding of activatory receptors of any kind has come as a great relief.  However, the fact these particular receptors bind class I molecules potentially undermines the “missing self” paradigm of NK recognition which has provided such a satisfying rationale for the function of NK cells.  A simple way out of this dilemma would be to propose that all NK cells have at least one inhibitory receptor for self class I molecules, that inhibitory signals are dominant, and that the role of class I activatory receptors is to promote recognition of cells that have lost expression of only some class I molecules.   However, this leaves unresolved the nature of the activatory receptors required to recognize class I negative cells.  The idea that NK cells utilize a variety of widely expressed adhesion/costimulatory molecules for the purpose has gained further support this year from the finding that expression of B7.1/B7.2 [13] or CD40 [14] molecules on target cells can greatly enhance their sensitivity to NK cell lysis.  Intriguingly, at least in the case of  ICAM2, it appears that it is the precise location and/or distribution of the molecule on target cells, rather than its total quantity, that is critical for NK cell triggering [15].  

        Understanding the cellular and genetic events that control the development of the NK cell repertoire continues to be a major intellectual challenge.  Data obtained from the first Ly49 transgenic mice [16] support a three-rule model in which inhibitory receptors are expressed at random, NK cells are positively selected only if they express at least one inhibitory receptor for self class I, and cells expressing multiple inhibitory receptors are disfavoured by the expression/selection process.  As far as can be judged from an exhaustive analysis of NK clones the same rules of repertoire development apply in man [17].  The finding that NK clones derived from immature progenitors in fetal mice undergo continuous and extensive diversification in the expression of certain surface molecules, including in some strains of Ly49 molecules [18], may provide a tool for analysing the mechanisms controlling receptor expression.  The notion that NK cells can adapt to their environment by adjusting their expression of surface molecules [receptor calibration] has been supported by studies in a second line of Ly49 transgenic mice [19].

        The next few years are likely to bring continued rapid progress in our understanding of NK recognition.  Major projects currently underway to map and sequence the KIR gene complex in humans and the NK complexes of humans and mice have already borne fruit [20, 21] and will eventually provide revelatory information on the number, organization, and evolution of class I receptors, together with a solution to the conundrum of whether KIR-type molecules are expressed in rodents and Ly49 molecules in humans.  The importance and impact of soluble forms of both class I receptors and class I molecules will continue to be felt, but additional reagents are also required, especially in the mouse, where the dearth of mAbs against Ly49 and CD94 receptors is inhibiting progress in understanding the in vivo function of class I receptors and indeed of NK cells themselves.  In the next year or two we can probably look forward to crystal structures of receptor-class I complexes, a resolution of whether Qa1 in the mouse is a functional as well as a structural homologue of HLA-E, and the creation of CD94 and DAP12 knock-out mice.   The recent discovery of a large family of IgSF receptors homologous to the KIRs but much more widely expressed in the immune system [22, 23] has potentially momentous implications for our understanding of immunoregulation.  In an influential article entitled “Genetics of the MHC, the final act” published in 1983, Klein and colleagues argued that the true function of  MHC class I and class II molecules had finally been established, namely that of presenting peptides to T cells [24].  Although few immunologists would wish to abandon this concept, perhaps it is time to pause for thought.  If nothing else it would appear that class I molecules are giving us a dramatic encore, and those who left the theatre early may have missed the most exciting part of the performance.

This article is reproduced with the kind permission of the British Society for Immunology .