Moderator Kenneth Schaffner - 8:00am May 9, 1997 

Is the self-nonself distinction (still) important in immunology, and if so, what do these terms mean? If not, what are better terms (e.g., integrity, danger, etc.), and why are they better?  
Rationale:  

All postings thus far accept a major role for the immune system in detecting and eliminating pathogens, while not attacking the body or the immune system (i.e., self-destruction of the immune system). In recognizing some things as "to be eliminated" and others as not, is this tantamount to an implicit definition of the self-nonself distinction? If not, is there a better way to conceptualize this difference at a general level, or must the discussion proceed to mechanisms (cellular and molecular)? If some alternative to the self-nonself distinction is favored, why has self-nonself been so important for the past 50 years? 

Zlatko Dembic - 4:26pm May 9, 1997 (#1 of 32)  

The self-nonself discrimination as a principle of the immune system is an abstract matter and can be observed only by an outside viewer. The differences between self and nonself molecules are indistinguishable at the molecular level. Thus a biological control is required. This biological principle cannot be based on the self-nonself discrimination of the antigenic universe. The self-nonself concept lasted so long because it was a fairly satisfying approximation of the truth. With the increased scientific knowledge and development of molecular biology, it is becoming more obsolete. The self-nonself concept is now a virtual reality that blinds and hinders us from obtaining a better understanding of the immunologic phenomena. The "protection of integrity" and "danger" models gave us a hint of reality - there are signals that can be disrupted or set in motion, respectively. If we choose not to view the immune system through the eyes of the self-nonself concept, we might notice that the immune system will get "larger" and include nonspecific immunity (innate), which, I would argue, is the link between an integrated being (Day 1, #6) and specific T and B responses. Here lies the major difference between the self-nonself models and both integrity and danger models. Because of this, the explanation for the initiation of the immune response by the self-nonself model is weak (the mechanisms later). The preservation-of-integrity and danger models offer unique pathways for activation of specific immunity and bring into focus interesting novel signals that, evolutionarily speaking, predate other signals or "commands" (Day 1, #6). 

Becoming a space-faring species would have been impossible had humanity not changed its viewpoint of the universe from geocentric to heliocentric. I believe it is time to move away from the self-nonself discrimination viewpoint to a more appropriate one, be it danger, integrity, or any other, if we want to make further progress concerning autoimmunity and transplantation. 

Rod Langman - 5:56pm May 9, 1997 (#2 of 32)  

Yes. So long as the immune system can kill cells and eliminate extracellular proteins and carbohydrates in order to also eliminate pathogens, it must carry out a self-nonself discrimination if it does not exercise the same responses to kill the host (i.e., self). 

Please let it be clear that it is the immune system that we are discussing, not the brain, and not restriction enzymes, etc. Perhaps some of the heat could be diverted to power the national electrical grid if the "self-nonself" word pair was replaced by "lethal-nonlethal." The problem would not change, but the change of clothing might help refocus the issue. 

Let me pose an intermediate miniquestion. Does the restriction endonuclease in the bacterium make a self-nonself (lethal-nonlethal) discrimination at the level of DNA? Remember, however, that DNA is all the poor thing can see. You and I know that there are all kinds of proteins that we could categorize as self and nonself, but does the restriction enzyme get blamed because it has no idea what a protein is, let alone whether it is self or nonself?  

If you answer that the restriction enzyme does make a self-nonself discrimination, provided it is limited to what it can see and do, then I would ask whether the same criteria can be applied to the collection of "thingies" we call the immune system, again with the proviso that it applies only to the stuff that the immune system can actually see in real life.  

If you answer that the restriction enzyme does not make a self-nonself discrimination, is there at least a principle involved (if so, what is it?), or are we stuck with the one observation-one theory dictum? 

Zlatko Dembic - 8:10pm May 9, 1997 (#3 of 32)  

The preservation of the integrity is the principle. For example, the bacterium with its restriction enzyme indiscriminately cuts specific sites in linear DNA of any origin, with its own being protected, perhaps, by methylation. If methylation fails, the bacterium is gone, and this mutant is selected against. "Parasitic" DNA of linear origin can more easily integrate into the genome and damage or kill the host. Thus, the preservation-of-integrity principle and the self-nonself principle explain this example equally well. However, in more complicated systems, like the immune system of higher vertebrates, the consequences of applying each principle are different, and, I argue, more choices or more room is being offered by the integrity principle than by the self-nonself principle to theoretically explain the operating system of immunity. 

Thus, we are not stuck with one theory. It is just that you see it your way and I see it my way. That already makes two. 

Rod Langman - 8:55pm May 9, 1997 (#4 of 32)  

The restriction endonucleases have DNA sequence-recognition sites (which is what makes them so useful to the DNA jockeys) that are specific in not recognizing host DNA sequences; but because there are viruses that did not exactly coevolve with that particular bacterium, this viral DNA does not necessarily lack the DNA sequences recognized by the restriction enzyme, and the enzyme snips the viral DNA, "killing" the virus. This has nothing to do with methylation or other such generalized phenomena. Thus, I come back to my original miniquestion: Does the restriction enzyme exercise a self-nonself discrimination or not? 

Ephraim Fuchs - 11:04pm May 9, 1997 (#5 of 32)  

The notion that the immune system discriminates self from nonself has been and will be useful to the field so long as it drives immunologists to propose novel and testable theories to account for the activation-tolerance decision of lymphocytes. The self-nonself model led Burnet and Fenner (1949) to postulate a critical period during ontogeny when any antigen would be regarded as self, which stimulated Lederberg (1959) to propose that immature lymphocytes would be inactivated upon encounter with antigen. This, in turn, led to the discovery of intrathymic clonal deletion of self-reactive T cells by Kappler and colleagues (1987), as well as by von Boehmer and colleagues (1988) and others. It was in response to a deficiency in Lederberg's model that Bretscher and Cohn formulated the two-signal model of lymphocyte activation (Bretscher and Cohn, 1970), which has now achieved the status of dogma within the community of immunology, and which richly deserves a Nobel Prize (although the first two-signal model of lymphocyte activation was proposed by Donald Forsdyke).  

It is easy to see how the self-nonself distinction was important during an era in which the major challenges to human health were viral and bacterial infections. Now, however, we have to deal with problems like autoimmune disease, cancer, and transplants (not that we have conquered infectious diseases entirely). It is difficult to see how a self-nonself paradigm would be of much assistance in understanding these phenomena. If the immune system discriminates self from nonself, why does autoimmune disease occur? Is cancer self or nonself? It now seems more useful to ditch the question of self vs. nonself and rather to flush out the fundamental determinants of T-cell activation vs. tolerance induction. The parameters of the activation/tolerance decision fall within the domains of time (the age or differentiation state of the lymphocyte), space (whether antigen is encountered in the thymus or periphery or on a dendritic cell, a B cell, or parenchymal tissue), and experience (the context of the last antigenic encounter). Only when immunologists can agree upon these parameters can we go back and assess whether the immune system really discriminates self from nonself antigens.  

Rod Langman - 12:06am May 10, 1997 (#6 of 32)  

I pass on a question of clarification:  

"I'm confused about your dismissal of Zlatko's suggestion about methylation failure (#3) in a bacterial host. If that bacterium produces a restriction enzyme that is methylation sensitive as well as sequence specific, the mutation of the host's methyl transferase could cause self destruction." 

My response:  

This kind of methylation defect, and any others like it, is not the reason that restriction enzymes are sequence specific, and unrelated to the host defense process. To argue that host DNA can be protected by methylation, but not viral DNA, is another problem that I'd rather not chase in detail other than to say that in the end whether to methylate or not, and how is this going to be decided. All I want to do is ask a very simple question about a very specific situation. I see no point to starting a discussion about the number of different host defense mechanisms and whether each has to undergo a self-nonself discrimination.  

I see no escape from the conclusion that all biodestructive protective mechanisms will have to make something that can be described as a self-nonself, or a lethal-nonlethal, or dangerous-undangerous, etc., discrimination based on specific recognition and exercise of the biodestructive consequence of recognition.  

Moreover, as I recall, methylation is never saturated in the DNA, and it tends to cluster according to function, which in part is why it is not generally used in bacteria. It would be tricky to have variable methylation for physiology and constant susceptibility to restriction endonuclease. 

Zlatko Dembic - 1:19am May 10, 1997 (#7 of 32)  

Rod: The methylation of DNA is a property of some bacteria and could be how eukaryotic cells might inactivate some genes. Whether the example that I have mentioned exists or has existed in the past and was lost in the course of evolution is not relevant for the case. The coevolution of viruses does not change the answer that I have already given: The self-nonself discrimination or the fight for keeping the integrity fits equally well as an explanation, whether you like it or not. In contrast, there is little evidence for the self-nonself discrimination when the bacteria perform one of the most usual habits of nature: conjugation. Here the problem is that bacteria are different in one aspect of this process - they conjugate by directly penetrating the cells, thus transferring their genetic material. If the genetic material is dangerous to their integrity, it gets destroyed. If it comes from the same species, it is accepted. So here, the self-nonself distinction gets blurred. I suggest that we now return to immunology. 

I would very much agree with Ephraim Fuchs (#5) in what the self-nonself concept has brought to immunology over the past years. Without this theory, we would not be able to claim the progress we have observed. It is to the greats of the field that we should pay respect, from Burnet, Medawar, and Lederberg to Cohn and Bretscher, to mention but a few. Rod Langman's contribution should be appreciated too. It is my wish to extend further the ground explored by those men and perhaps find some new important information. In my discussions, I have never wanted to disregard the intellectual benefit that allowed me to propose an extension to their thoughts.  

William O. Weigle - 2:19am May 11, 1997 (#8 of 32)  

To diverge from the ongoing debate concerning DNA splitting by endonucleases, I would like to address epitope specificity in self-nonself recognition in the immune response. It appears that we are treating the entire issue of self tolerance as one single event that is covered by one general mechanism. I am sure that most of us would accept the view that self tolerance is not that simple. Although this unitarian approach may readily lend itself to making and defending theories, it will not give us a complete picture of discrimination of self in the immune system. Is it possible that self-nonself recognition can be at play at one level of self tolerance while at a second level other mechanisms may be responsible? As pointed out by Zlatko, the antigenic properties of self are continually changing throughout life. Furthermore, sporadic release of sequestered antigens obviously occurs periodically throughout life.  

Thus, the mechanisms for the induction and maintenance of tolerance in the periphery are required, not merely as fail-safe mechanisms, but as a necessity throughout life to keep self antigens below the threshold required for disease-producing autoimmune reactivity. It is my belief that it is at this level of peripheral tolerance that one has to question the mechanism of self-nonself recognition at the epitope level. Such tolerance can readily be induced, in the absence of the thymus, to heterologous serum proteins that mimic self tolerance in CD4+ T-cell subsets (Gahring and Weigle, 1989). Because it can be shown that the animal host can fail to respond to epitopes on at least some foreign proteins but can be tolerized to them, at what site is self-nonself discrimination, if it does exist at this level? Are there non-immunogenic sites present on self as well as heterologous serum proteins that allow them to be recognized as self in the same manner that endonucleases recognize target sites on DNA? Or can self tolerance in the periphery be explained by nonspecific signals to T cells in the immediate microenvironment? I am certain that opportunities will arise in the next several days to permit the presentation of possible solutions to this and other dilemmas concerning self-nonself discrimination. I can see no reason for discarding the term "self-nonself recognition" as long as we recognize its limitations and where it does and does not apply.  

Rod Langman - 2:40am May 11, 1997 (#9 of 32)  

I hate to sound like I'm beating a dead horse, but I think it is important to get straight what it is we are trying to solve. Zlatko and Ephraim are clearly opposed to the use of words like "self" and "nonself." I would be happy to use any labels imagined to describe the concept.  

As a pure thought experiment, without prejudice, let us simply have a Greenian collection of "thingies." Starting with a host thingie (HT) that can be killed if infected with a pathogenic thingie (PT), one day the HT comes across in its bag of mutations something special that we can call a biodestructive thingie (BDT). This imaginary BDT has a host that makes it and, because the BDT can kill the PT, the HT keeps the BDT tucked into its genome. However, PT and HT are made up of the same basic stuff, nucleic acids, proteins, carbohydrates, etc., and so the BDT can actually kill the HT if the BDT gets out of control. Thus the BDT has to be able to tell the difference between the stuff belonging to the HT and the stuff belonging to the PT.  

My question: Does the BDT discriminate between HT and PT? There is no implication as to how the BDT does its deadly work, just that it destroys biopolymers so well that it destroys life. What is the conceptual framework that describes the relationships among these thingies, HT, PT, and BDT? Can the BDT avoid being endowed with some kind of specificity that allows it to tell the difference between a HT and a PT?  

Doug Green - 5:11am May 11, 1997 (#10 of 32)  

It's somewhat interesting to me that, in response to the question of whether self-nonself discrimination is still important in immunology, so much discussion is focusing on bacterial physiology. Most of us who seek to understand immunology do so from the context of the vertebrate immune system. The existence and basis of defense mechanisms in nonvertebrate (or even non-eukaryotic) organisms can certainly be interesting in their own right (if one is interested in those organisms) or if it gives us insight into vertebrate immunity. I think I'm not being closed-minded here, I'm being practical - biological space is huge, and some of us simply don't have time to deal with every problem in biology. The adaptive immune system is, as far as we know, unique to vertebrates; there is no really compelling evidence for adaptive immunity in other types of organisms. Whether the concept of self-nonself discrimination is important to our understanding of bacteria, diatoms, cedars-of-Lebanon, or stagmoss may not (I suggest) be germane to the question at hand. That is, unless immunology is redefined as the study of self-nonself discrimination in any biological organism, in which case it is something different from what I study.  

Doug Green - 5:12am May 11, 1997 (#11 of 32)  

My answer (somewhat late I'm afraid) to this second question is likely to be inflammatory (pun intended), but here goes. The concept of self-nonself discrimination is no longer useful to the study of immunology (as in the study of the vertebrate adaptive immune system). These terms don't have practical meaning, except in the context of what they are supposed to be explaining (e.g., "self" is anything that the immune system doesn't make a response to, unless it's during a defective anti-"self" response?). I would argue that we will get closer to a mechanistic explanation of the function of the immune system when we eliminate the use of terms that invoke poetic but hard-to-pin-down images such as "self," "nonself," "antigen," "immunogen," and perhaps "danger" and move toward descriptions of concrete entities and processes.  

What could effectively replace these terms? As a first approximation, for example, the process of central tolerance (negative selection) creates an illusion of self-nonself discrimination because any protein that is always available for presentation to developing lymphocytes will eliminate those cells from the population (e.g., Green et al., 1992). Thus, in the constantly developing immune system subject to negative selection, responses can only occur against proteins that are only sometimes available for presentation to the lymphocytes. If a new protein appears for any reason in the body, there are likely to be lymphocytes that are capable of responding to it. Whether they do respond is dependent upon additional factors, such as the presence or absence of tissue damage, proinflammatory molecules, and many others, all of which can be defined (in actuality or in principle) without resorting to broad concepts of self and nonself. 

I agree with those discussants who have suggested that the terms "self" and "nonself" were useful in the past. The fundamental principles of clonal selection and deletion, as well as several newer concepts, depended upon considerations of self-nonself discrimination. I argue, however, that the time has come for us to at least try to describe the immune system in terms of what it actually does, what the cells and their receptors actually respond to, how these ligands form, and the influences of other receptor-ligand interactions on the outcome of recognition of these ligands. It's a big job, but I think that it will go much further toward giving us an understanding of the operation of the vertebrate immune system than will any analysis of territorial competition in sea anemones (despite the fact that territorial competition in sea anemones is extremely interesting in its own right). Thus, I cast my vote with those calling for dispensing with the terms "self" and "nonself" in immunology, albeit for different reasons. 

Doug Green - 5:21am May 11, 1997 (#12 of 32)  

Rod: I think your thought experiment (#9) sums up the problem in good abstract terms, but I reckon that we'll come back to the idea that concepts such as self-nonself are only useful when they describe something we cannot define in more specific (and concrete) terms. But, as a small point, if you really want to create a "Greenian" entity, then the acronym should spell something cute (it's more fun).