NK Cells, MHC Class I Molecules and the Missing Self


  • K. Kärre

    Corresponding author
    1. Microbiology and Tumor Biology Center, Karolinska Institute, Stockholm, Sweden
      Dr Klas Kärre, Microbiology and Tumor Biology Center, Karolinska Institute, 17177 Stockholm, Sweden. E mail: klas.karre@mtc.ki.se
    Search for more papers by this author

Dr Klas Kärre, Microbiology and Tumor Biology Center, Karolinska Institute, 17177 Stockholm, Sweden. E mail: klas.karre@mtc.ki.se


This article is based on a lecture presented at the Novartis Prize ceremony at the International Congress of Immunology in July 2001. It gives a personal and historical perspective on the research performed by the author and his colleagues during the development and pursuit of the model of ‘missing-self recognition’ for natural killer (NK) cells. This model is based on the idea that one important function of NK cells is to detect and eliminate cells because they fail to express normal self markers. Further mechanistic models predicted the existence of inhibitory major histocompatibility complex (MHC) class I specific receptors, later identified by the fellow Novartis laureates contributing in this issue. The article covers the first decade (1980–1990) of research on this concept. It discusses factors contributing to the formulation of a hypothesis, the use of predictions and experimental test models, the importance of international collaborations and reagent exchange, and several other aspects that allowed the progression of this research project. Finally, the perspective of today’s knowledge is used to discuss some surprising findings where the missing-self hypothesis made the wrong predictions, or at least failed to make the correct ones.

It is a great honour to have been awarded The Novartis prize for basic immunology. I wish to express my pride and my deep gratitude to Novartis and to the Jury, on behalf of myself and all the colleagues who have worked or collaborated with me. It was a special joy to share the prize with Wayne Yokoyama and Lorenzo Moretta, colleagues in the field of natural killer (NK) cells. I have admired their work and key discoveries for a long time, as I have also admired the work of Alain Fisher, the clinical prize winner. Let me add that it was a fantastic feeling to receive the prize during the International Congress of Immunology in Stockholm; in the presence of so many fellow immunologists, and on top of that, in my home town. Many of my present and past collaborators were actually present to share my pride.

In this article, I have been asked to briefly describe some key aspects of my work on NK cells. Many of today's immunologists have followed the exciting development in the research on molecular aspects of NK-cell recognition during the last decade. However, only a few may be familiar with the phenomena that prompted the formulation of the theoretical platform used to analyse NK-cell specificity, to discover, for example, inhibitory receptors, and the first in vivo experiments that supported it. I have therefore chosen to follow the same track as in the lecture at the prize ceremony, going back in time and describing the first ideas and experimental developments between 1980 and 1990, putting them in today's perspective. I will take the liberty to make some personal reflections and acknowledgments of sources of inspiration, collaboration and support.

The idea – missing-self recognition

The work really started with one single idea in the early fall of 1981 – the concept of missing-self recognition. The first version of this model was presented in my PhD thesis the same year [1]. Internationally, it was first published in the proceedings of the International Workshop on Natural Killer Cells 1984, in an article named ‘Role of target histocompatibility antigens in regulation of NK activity: a reevaluation and a hypothesis’[2]. What exactly did this hypothesis propose? The idea was that one central recognition mechanism of NK cells operated by detecting information that was missing in the target and present in the host rather than the opposite, which was (and still is) the accepted paradigm for how T cells recognize foreign or infected cells. More specifically, the hypothesis postulated that the absence or incomplete expression of host major histocompatibility complex (MHC) class I molecules in a normal cell would be sufficient to render it susceptible to NK cells, without any other change being required.

At the time, this was quite controversial, as NK cells were considered to work like T cells by recognizing the presence of foreign antigen, the only difference being that they were not influenced by MHC molecules of the target. The missing-self hypothesis postulated exactly the opposite: NK cells could work according to a fundamentally different principle, not requiring the presence of neoantigens; however, they were similar to T cells, in the sense that they too were strongly influenced by target MHC molecules!

Many predictions of this hypothesis have been confirmed, but like any scientific hypothesis, the missing-self concept has also turned out to be incomplete and not entirely correct. However, at the time it provided an alternative explanation for many intriguing observations that had been made in the past. For the future, it provided a theoretical platform for development of models, testable experimental predictions as well as for development of reagents, e.g. for novel type of receptors. The latter is readily illustrated by the articles in this issue by my fellow prize winners.

Factors contributing to the formulation of a hypothesis

Why was the idea proposed, and what were the sources of inspiration? The most critical factor was the mystery of the NK cells themselves. They had been studied for at least half a dozen years, but their specificity remained a black box, in spite of many clues in the literature. I had the privilege to be able to speculate almost daily about this with my supervisor Rolf Kiessling, one of the discoverers of the NK cells [3, 4], and also quite frequently with George Klein, the head of the department. To work with such a problem, with supportive and open-minded mentors, represents in itself a major inspiration for the formulation of ideas. The most important set of observations that we struggled to explain was the intriguing phenomenon of ‘hybrid resistance’. This refers to the situation where an (A × B)F1 host, in defiance of conventional transplantation laws, rejects A or B grafts of parental origin (A and B referring to MHC genotypes).

Hybrid resistance, first described by Cudkowicz et al. [5], could be observed against tumour as well as bone marrow grafts, and it was clear already by the end of the 1970s that it was mediated by NK cells [6, 7]. But what did they actually recognize on the grafted cells? Several different theories were proposed. They were based on different assumptions, and agreed only in one aspect: the known MHC class I genes and their products could not explain the phenomenon. The missing-self hypothesis suggested the opposite, with the twist that the MHC class I molecules of the grafted cells exerted their crucial role in the graft that was accepted rather than in the graft that was rejected. It was thus proposed that NK cells could sense the presence of a complete set of MHC class I molecules in autologous grafts, which would prevent rejection. Conversely, the absence of some MHC class I genes in a parental to F1 graft (or all of them, as in a completely allogeneic graft A→B) would lead to rejection by NK cells.

Inspiration for this came from several sources. George Snell had proposed an inhibitory role in hybrid resistance for homotypic interactions between self-MHC molecules [8]. Investigators studying rapid elimination of allogeneic lymphocytes in rats discussed whether self-MHC matching could function as a possible ‘password’, allowing autologous lymphocytes to escape [e.g. 9]. The critical contributions of the missing-self hypothesis were its focus on NK cells, and its ability to use a novel principle allorecognition not only to explain but also make sense of NK cell reactivity against certain autologous cells. Allorecognition is often regarded as an artefact of experimental or clinical manipulation. The missing-self hypothesis argued that NK-cell mediated allorecognition reflected the ability of NK cells to detect infected and other abnormal cells by virtue of reduced MHC class I expression. This phenotype had been observed in certain viral infections as well as in many tumours, and it had been pointed out that it could lead to escape from T-cell recognition. It was also known that cells with this phenotype could be NK-sensitive [10], but nobody had made the causal connection to test whether the cells were killed by NK cells because they lacked MHC class I molecules. Similarly, nobody had made a causal connection between two known but seemingly unrelated effects of interferon (IFN) (in particular, IFN-7) on cells [1]: increase in MHC class I expression [2], reduced susceptibility to NK cells [11].

However, inspiration also came from the literature on organisms in which allogeneic transplantation is a part of the daily lifestyle. In the colonial tunicate Botryllus, fusion between different colonies is controlled by a single, highly polymorphic locus [12]. Self-recognition involving products of this locus appear central to allow fusion, i.e. inhibit rejection. There was no (and still is no) evidence for an involvement of MHC-like molecules, but the fascinating literature on tunicates and other invertebrate-incompatibility reactions [13, 14] provided one of many inspiring metaphors where the rate-limiting step in a complex reaction is mediated by an inhibitory signal. Another such metaphor was the strategy used in the early 1980s by the Swedish Navy Command to teach the general public how to distinguish a foreign submarine from a Swedish submarine. For the interested reader, a more detailed account of the parallels between NK cells, submarines and tunicates is given in the essay ‘How to recognize a foreign submarine’[15]. This essay also describes the critical moment for shaping of the missing-self idea. It came about during the struggle with a sentence in my doctoral thesis, when it occurred to me that it was easier to describe the characteristics of the cells that were resistant to NK cells rather than common properties of the susceptible ones. The first version of the idea was thus published in the thesis, and in this context I would like to make one of my central acknowledgements. As a graduate student, it is not evident that you should dare to propose and pursue a highly controversial hypothesis at a critical time in your career. However, my working environment, The Department of Tumor Biology of Karolinska Institute, provided an intensive, open and generous intellectual climate. With support from George Klein, the founder and chair of the department, and my advisor Rolf Kiessling, I presented the hypothesis in my thesis and at international meetings.

Predicting and testing – the target cell level

While the inspiration for the hypothesis came from F1 hybrid resistance and related allorecognition phenomena, the first critical predictions and test models focused on the autologous situation. This was important, because a model must not only provide explanation of available observations, but it should also present testable predictions, preferably unique in the sense that they should distinguish the model from competing ones. The obvious such unique prediction linking hybrid resistance to autologous NK-cell recognition was that an MHC class I-expressing, NK-resistant cell that somehow lost its MHC class I molecules should become NK-sensitive and be rejected in vivo in the autologous host (Β→Β). Conversely, restoration of MHC class I expression should lead to resistance against killing, allowing escape from NK cells in vivo.

When I ask the students in my classes today on how they would have attack this prediction experimentally, they answer –‘Simple, just make an MHC knockout mouse!’ However, at that time the first gene-targeted mice were still almost 10 years ahead. The most obvious way to generate MHC-deficient cells was to use Darwinian principles: to mutagenize and immunoselect in vitro-growing lymphoma cells. George Klein was sceptical about the project but promised to support it anyhow. He gave me the advice –‘never forget, the success of any selection is critically dependent on two factors: the strength of the selective pressure and the patience of the selecting investigator’. With these words ringing in my ears, I embarked on an ambitious project, starting with three different cell lines and, for each of them, several selection methods, including sorting by flow cytometry, a fairly new technique at the time. However, the selective pressure applied by sorting the 0.2–1% cells with lowest MHC class I expression after mutagenization was not strong enough. The successful technique, resulting in several variant cell lines, was instead the one that was based on the good old principle of treating cells with anti-MHC class I antibody and complement. In the first selection rounds, the proportion of cells killed by this treatment surpassed 99.95% (I gave up after counting 2000 dead cells in the microscope and had to rely on the hope that at least one had survived). The waiting for outgrowth from the cultures, which took on a greyish colour from all the debris of killed cells, seemed endless. Finally, after several weeks, cells grew out. The treatment was repeated, and for each selection round, the cell culture ‘came back’ earlier, and the cells eventually became completely resistant to antibody + complement treatment.

Several combination of variant wild-type lines were derived in this way, and we could rapidly confirm that the immunoselected variant cells had undetectable, or in some cases profoundly reduced (95%) but still detectable, MHC class I expression. The cell lines were first tested for susceptibility to autologous NK cells in vitro, with promising, yet quite uncertain, results. The variant cell lines were indeed significantly more susceptible to NK cells, but the general levels of killing were still low when NK cells from normal mice were used. However, the results became more interesting when we used activated NK cells from mice treated with IFN-inducers, and convincing when the cells were inoculated in vivo[16]. I can still remember the magical moment when my first graduate student Hans-Gustaf Ljunggren and I went to the animal department to inspect the first group of mice, inoculated two weeks earlier with low numbers of tumour cells. Only mice inoculated with wild-type control tumour had developed tumours, while the ones inoculated with MHC class I-deficient variant cells showed no palpable tumour! This difference persisted throughout this and many repeats of the experiment. The variant cells were rejected in all or the majority of autologous mice, while the control wild-type cells grew out in the majority of the mice [16]. The difference could be abolished by depletion of NK cells with the monoclonal anti-NK1.1 antibody [17]. The same pattern was seen in nude mice, thus excluding the possibility that NK T cells were the critical effector cells.

These results were controversial not only with respect to NK cells, but also in a broader tumour immunology perspective. Loss of MHC expression was now being reported in many tumours. The emerging paradigm, supported by several experimental models, was that this phenotype led to increased malignancy owing to escape from T-cell immunity. Our results showed the opposite: the MHC-deficient cells were less malignant than the MHC expressing counterparts. For the publication of the results, we therefore undertook an extended analysis, assuming that the outcome of the host–tumour interaction would depend on whether the experimental system was geared to predominantly detect T-cell or NK-cell reactivity. Indeed, we found a reversed picture when we designed our tests to overcome NK reactivity (by increasing the tumour cell dose) and increasing the potential for T-cell reactivity (by preimmunizing the mice with the tumour, or by using recipients that represented a minor histocompatibility barrier to the tumour). We could now observe that the wild-type control cells were rejected, while MHC class I-deficient variants escaped, thus behaving as being more malignant [16].

Interestingly, the escape from T-cell immunity by one of the MHC-deficient variant lines, named RMA-S, later provided an interesting side project that developed in a research line of its own. The first observation was that RMA-S, although completely resistant to killing by MHC-restricted autologous T-cell, could still be killed by allospecific T cells from mice of another MHC type [18]. The next clue came in molecular studies performed during my post-doctoral studies with Madeleine Cochet and Philippe Kourilsky in Paris, as well in collaborative studies between H.-G. Ljunggren and Svante Pääbo in Uppsala. Surprisingly, we found that RMA-S, with 95% reduced expression of MHC class I molecules at the cell surface, transcribed and translated normal levels and forms of MHC class I heavy-chain and β2-microglobulin [19]. However, the association between the subunits appeared abnormal, and the explanation for this came in a further collaboration with Alain Townsend in Oxford. The low steady-state levels of MHC class I molecules could be increased by incubating the cell with peptides binding to either H-2Kb or Db [20]. This and other results suggested that the missing component in the cell had to do with the supply of MHC-binding peptides; as indeed confirmed later in the discovery of the function of the transporter associated with antigen processing (TAP) genes (one of which turned out to have a point mutation in RMA-S) by Jonathan Howard's group in Cambridge [21]. This sideline is mentioned here because it relates directly to the NK oriented feature of this article: after its European odyssey (Stockholm, Paris, Uppsala, Oxford, Cambridge), RMA-S came back to Stockholm with restored MHC class I expression after transfection with a correct version of the TAP-2 gene. Indeed, it had now become resistant to NK attack, in vitro as well as in vivo[22]! This was only one out of many systems where different groups could demonstrate direct, transfection-based evidence for a role of MHC class I molecules in NK recognition. Most importantly, several groups had now taken up the pursuit of the missing self in human systems. The group of Fradelizi [23] and the group led by Dawson and Creswell [24] were the first to provide molecular proof for MHC class I molecules in protection against attack by NK cells, and both groups used human effector and target cells.

Our own work included also tumour models such as the YAC-1 lymphoma and the B16 melanoma where endogenous expression of MHC class I was low but inducible by IFN-γ. We could demonstrate IFN-γ-mediated induction of melanoma metastasis correlating with increased MHC class I expression and reduced sensitivity to NK cells [25]. We also generated YAC-1 lymphoma variants where MHC class I surface expression went from ‘low’ to ‘undetectable’ owing to deficiency in β2-microglobulin. In contrast to the wild-type cells, they remained NK sensitive after IFN-γ treatment [26], and this could later be restored by transfection of the β2-microglobulin gene [27]. Follow-up studies with the RMA-S tumour revealed that target MHC class I molecules did not inhibit NK cells globally and permanently. When exposed to mixtures of MHC class I-expressing and -deficient targets, the NK cells always killed the latter, even if the former were in excess [28]. This has been illustrated nicely in recent videomicroscopic analysis, where NK cells forced to conjugate to several targets of different phenotypes simultaneously can kill one but leave the other intact [29]. We could also demonstrate that missing-self rejection operated in several different body compartments, with one notable exception: the brain, an organ where MHC class I expression is kept low in most cells [30].

Which were the critical success factors in the early ‘predicting and testing phase’? One crucial component was of course the fact that a high-risk project like this received support, intellectually and economically. In this context, I extend my warm thanks to my mentors mentioned above as well as The Karolinska Institute and The Swedish Cancer Society.

Equally important were the first young students and post-doctorates who dared to join the project: Hans-Gustaf Ljunggren and Gerald Piontek, and later Kazuto Taniguchi, Petter Höglund and Claes Öhlèn. Qualified technicians assisting with animal as well as in vitro work also contributed: Margareta Hagelin, Maj-Lis Solberg and Erene Eriksson.

Hybrid resistance revisited – predicting and testing at the host and effector cell level

The original model stressed a role for recognition of self rather than MHC class I molecules in general. Another important prediction was, therefore, that it should be possible to induce missing-self recognition by adding MHC class I molecules in the host rather than by subtracting them from the target. It became possible to test this prediction towards the end of the 1980s as the transgenic mouse technology became available. This allowed a return to the scenario where it all had started, the hybrid resistance phenomenon: if an (A × B)F1 host rejected a B graft because NK cells required to see a critical MHC class I molecule from A, then it should be possible to induce rejection of B-type cells in a B animal simply by introducing the novel, critical MHC class I gene in the host. This was tested and confirmed using H-2Dd transgenic B6 mice generated and generously provided by Gilbert Jay, first for rejection of B6 lymphoma cells by the student Petter Höglund [31], then for B6 bone marrow cells by the student Claes Öhlèn [32]. The transgenic technology also allowed a first demonstration that a self-MHC class I gene could protect normal cells from rejection by NK cell in vivo: H-2Dd transgenic B6 bone marrow cells became protected from rejection by NK cells in H-2Dd-expressing mice [32].

With the transit into the next decade, the introduction of the gene targeting technology made the story complete. It now became possible also to delete MHC class I expression in normal cells and in the host. The group led by David Raulet and our group used two differently derived strains of β2-microglobulin-deficient mice, generated by R. Jaenisch and Beverly Koller, respectively, to test and confirm a series of predictions. Deficiency in MHC class I expression was sufficient to render also normal cells sensitive to NK mediated killing [33, 34]. Furthermore, NK cells that had developed in a MHC class I-deficient host were incapable of missing-self recognition, and it could further be shown that this capacity depended on MHC class I expression in the haematopoetic system [34]. Mice generated by the transgenic and gene targeting technology since then have served, and today continue to serve, as useful tools to study how NK cells are educated to the MHC phenotype of the host. This issue, which includes the concept of NK tolerance, is still much of a black box.

Coming back to the discussion on critical factors for the testing and development of a hypothesis, I would like to emphasize the importance of international collaboration, allowing exchange of ideas, reagents and techniques. This was a critical aspect for the work on molecular analysis and restoration of MHC class I-deficient variants described in the previous section, as well as for the work on genetically engineered mice described in this section. The generosity of colleagues in different countries, providing us with reagents, models as well as practical training when one of us visited their laboratories, allowed us to proceed by applying the most recent technology and critical test tools.

The most important prediction – effector inhibition within a multiple choice model

It was important from the start to introduce possible mechanistic models for missing-self recognition [2, 16, 35]. The effector inhibition model postulated an NK receptor that transduces a negative signal after recognition of the appropriate MHC class I molecule or another structure dependent on it for its expression. This required that other molecular interactions established effector–target contact, and the initial triggering of the effector cell, which would be allowed to proceed to lysis unless the negative signal was received. The alternative target interference model was based on a triggering-target structure whose exposure or transport to the cell surface would be prevented by MHC class I molecules.

Each of the models provided unique testable predictions. These have been a most useful platform in the search for receptors. Such mechanistic studies followed in the second decade of research, once the work during the first 10 years had provided conclusive evidence for the principle of missing-self recognition at the operational level. In particular, fundamental discoveries by the fellow Novartis laureates Wayne Yokoyama and Lorenzo Moretta served to open the research field on MHC-recognizing NK receptors. This exciting development is covered in their articles in this issue and will therefore not be discussed here, except to make an additional acknowledgement. If a model is to remain vital, and develop further, it must be taken up for critical testing by several groups in the scientific community. This allows for a broader use of reagents and test models, and a healthy discussion leading to modifications of the hypothesis. The effector inhibition model had predicted that it should be possible to identify a receptor that mediates an ‘off signal’ after binding of an MHC class I molecule or an MHC class I-dependent ligand. Blocking of this receptor should cancel the difference between MHC class I-positive and -deficient targets, making both NK cells sensitive. This experimental set-up was exactly the one used by the groups of Yokoyama [36] and Moretta [37] when they discovered the first inhibitory receptors.

Recent research has identified several activating receptors on NK cells, and also demonstrated that these may override the inhibitory MHC recognition. Although several of our suggestions relating to the missing-self hypothesis have turned out to be incomplete or wrong, it should be noted that the emerging role of activating receptors is not an example of this. The first version of the hypothesis [2] stated explicitly that one (not the) recognition strategy of NK cells was based on detecting the missing rather than the present. It was thus clearly recognized and discussed that NK cells could very well work with several parallel systems to detect the abnormal, some working via activating receptors. Again, this differed from the contemporary thinking about lymphocytes, where clonal aspects of selection and expansion required one type of major, rate-limiting recognition event per cell. For a cell of the innate immune system, I argued that the ‘adaptation of the receptor repertoire is likely to have occurred during phylogeny … This should have made it possible to evolve an economical solution based on many different receptors in one cell rather than many different cells with one receptor’[38]. This ‘multiple choice model’[38, 39] had several implications. For example, it postulated that there is no single interaction that is necessary and sufficient for effector function, and the search for absolute correlations between target sensitivity (or resistance) and single molecules therefore becomes meaningless. Each effector–target combination must be studied with respect to the rate-limiting steps defined as single ‘gates’, and it may be wise to choose targets with moderate NK sensitivity in the analysis. The likelihood that only one ‘gate’ controls the outcome of the effector–target interaction is then increased, and this is desirable if one wishes to interfere with it experimentally. The possibility of a multiple recognition strategy of NK cells was indeed discussed by several authors in contemporary articles.

Concluding remarks – the missing-self hypothesis in today's perspective

In this article, I have tried to describe the origins of the missing-self hypothesis and my own group's work for the first 10 years of research. Space restrictions make it impossible to review the further development during the 1990s, and I will therefore conclude here by a giant leap into today's perspective. It then seems appropriate to discuss some aspects where the missing-self hypothesis made the wrong predictions, or at least failed to make the correct ones. Although activating receptors were not only possible but actually required within the framework of missing-self recognition, it was not predicted that such receptors could also have MHC class I molecules as ligands. This has now been clearly demonstrated in the rat, mouse and the human. Even when our own group reported one of the first formal proofs of this (rejection of MHC class I Dd transgenic B6 bone marrow by B6 NK cells in vivo, [32]), I let my own bias prevent us from discussing the correct conclusion in a straightforward manner. It was mentioned but not highlighted in the paper, and the data were rather discussed in terms of indirect mechanisms that were consistent with missing-self recognition. A good example of the dangers involved in pursuing your own hypothesis!

The use of MHC class I molecules as triggering ligands is still an enigma. Does it conceal an NK-cell function based on positive recognition of altered MHC or even alloMHC? Or is it just an example of the notion that activating NK receptors can be allowed to use any ubiquitously expressed ligand, as long as the ultimate decision is taken downstream by inhibitory receptors? In this context, it should be emphasized that even though humans and mice have evolved entirely different families of receptors for missing-self recognition, both are based on the complicating principle of distinguishing between subsets of MHC class I molecules, as defined by polymorphisms. This allows the discrimination between autologous and nonautologous cells also via missing-self recognition [2, 40]; but is this really an important purpose of the system?

The philosophical discussion of ‘the purpose’ of the system leads to the role played by NK cells in the overall immune defence. Although the work of my own group has evolved from and within tumour and transplantation immunology, I have never been convinced that NK cells evolved primarily to reject tumours, and certainly not transplanted cells. This, however, cannot be excluded, as infected and malignant cells may actually be transmitted between individuals (e.g. in pregnancy and sexual contacts), and a mechanism that rapidly eliminates such cells may be of value. Still, it was natural to propose killing of autologous virus-infected cells as a probable driving force for evolution of missing-self recognition [2, 3]. Downregulation of MHC expression by certain viruses had been reported already in the early 1980s, and this concept has now evolved as a research field in itself, with many surprising and exciting molecular mechanisms. The idea that NK cells use inhibitory receptors mainly to identify and eliminate virus-infected cells certainly has intellectual appeal. However, there is to date no conclusive evidence that this mechanism is used by NK cells in combat of viral infections in vivo. The emerging evidence rather points at a critical role for activating NK receptors for discrimination between infected and healthy cells (e.g. [41]), even in the case of cytomegalovirus, which possesses several mechanisms to interfere with MHC class I expression. Is this another example where the missing-self hypothesis has suggested the wrong direction? Or is it too early to say, given that we do not know the ligands of the activating receptors involved in NK recognition of virally infected cells, and other possible viral mechanisms interfering with NK cells?

A third, final example of a surprising finding that was not predicted by the missing-self hypothesis is the expression of inhibitory MHC receptors by many activated T cells. The role of such receptors is still unclear. Does their presence mean that activated T cells can play two parts, acting also as NK cells? Not necessarily, as T cells still must rely on their specific activating T-cell receptor (TCR), requiring presence of the specific peptide antigen. One possibility is that the inhibitory receptors are important to rapidly modulate T-cell effector function and/or development of memory.

With these examples where the missing-self hypothesis turned out incomplete or wrong but nevertheless contributed something to the development, I would like to once again express my gratitude to group members, mentors, close and distant collaborators, and a scientific community which has so generously interacted by discussion, constructive criticism, supply of reagents, new hypotheses and experimental models. This has provided me with a continuous sense of reward for the past 20 years. I now gratefully accept the Novartis prize as a concrete award, and as an inspiration to continue the work with old and new colleagues, in the search for novel basic mechanisms in NK-cell recognition and in immunology.


The work of the author has been supported by Karolinska Institute, The Swedish Cancer Society, The Swedish Foundation for Strategic Research, The Göran Gustafsson Foundation, The European Molecular Biology Organization, and The Tobias Foundation. The author wishes to express his gratitude to all collaborators in the past and present: graduate students, undergraduate students, postdoctoral fellows, mentors, and collaborators in other groups in many different countries.