R EFLECTIONS ON S ELF:IMMUNITY AND B EYOND VIEWPOINT The Danger Model: A Renewed Sense of Self Polly Matzinger

For over 50 years immunologists have based their thoughts, experiments, the discovery that activated B lymphocytes and clinical treatments on the idea that the functions by hypermutate, creating new, potentially self- making a distinction between self and nonself. Although this paradigm has reactive cells. Realizing that often served us well, years of detailed examination have revealed a would be rare if immunity required the coop- number of inherent problems. This Viewpoint outlines a model of immu- eration of two cells, Bretscher and Cohn (14) nity based on the idea that the immune system is more concerned with added a new cell (the helper, later found to be entities that do damage than with those that are foreign. a ) and a new signal (help), proposing that the would die if it recognized First, do no harm. been incorporated into previous models [for antigen in the absence of help (Fig. 1B). In –Hippocratic oath example, why major histocompatibility com- 1975, Lafferty and Cunningham (15) dealt plex (MHC)–mismatched kidney transplants with the finding that T cells respond more Of all the mysteries in modern science, the from living donors often perform better than strongly against foreign cells of their own mechanisms of self versus nonself recognition MHC-compatible kidneys from cadavers (4); species than against cells of another species, in the immune system ranks at or near the why liver transplants are rejected less vigor- by adding another cell and another signal. top. ously than hearts; why women seem to be They proposed that T cells also need a second –D. E. Koshland Jr. (1) more susceptible than men to certain autoim- signal (named “costimulation”), which they As a graduate student, I was taught that mune diseases; why Rh disease of the new- receive from “stimulator” cells [now called the immune system functions by discriminat- born is a problem in the second pregnancy, antigen-presenting cells (APCs)], and sug- ing between self (defined early in life) and but not the first; why graft-versus-host dis- gested that this signal is species specific (Fig. nonself (anything that comes later), tolerating ease is less severe in recipients that have had 1C). self and attacking nonself. Although this el- gentle rather than harsh preconditioning treat- egantly simple idea seemed to make a lot of ments (5, 6); and so on] without adding The Infectious-Nonself (INS) Model sense, it had problems from the beginning special new situation-specific assumptions. The need for costimulation posed a major and has failed over the years to explain a This Viewpoint will first trace the history problem for SNS models. If, as they assumed, great number of findings. For example, what of the self-nonself (SNS) model, showing the decision to respond is made by antigen- happens when “self ” changes? How do or- how it had to be modified over the years to specific cells, and if self-reactive ones are ganisms go through puberty, metamorphosis, accommodate new data, then give a brief deleted, then immunity can be directed pregnancy, and aging without attacking new- description of the Danger model, and show against nonself. If, however, responses are ly changed tissues? Why do mammalian how it is leading us to a new way of thinking initiated by APCs, which are not antigen mothers not reject their fetuses or attack their about self-recognition. Those readers who specific (they capture all sorts of self and newly lactating breasts, which produce milk may already have encountered some of these foreign substances), then immunity cannot be proteins that were not part of earlier “self ”? concepts (2, 7–9) will find more details in the directed only against nonself. The concept of Why do we fail to make immune responses to supplementary Web material on Science On- costimulation was therefore essentially ig- vaccines composed of inert foreign proteins line (10). nored until it was rediscovered experimental- unless we add noxious substances, collective- ly by Jenkins and Schwartz in 1986 (16). In ly known as “adjuvants”? Why do we fail to The Self-Nonself Models 1989, Janeway offered an ingenious solution reject tumors, even when many clearly ex- Burnet’s original model (11) suggested (Fig. (17), suggesting that APCs have their own press new or mutated proteins? Why do most 1) that (i) each lymphocyte expresses multi- form of SNS discrimination and can recog- of us harbor autoreactive lymphocytes with- ple copies of a single surface receptor specif- nize evolutionarily distant . He pro- out any sign of autoimmune disease, while a ic for a foreign entity, (ii) signaling through posed that APCs are quiescent (a very impor- few individuals succumb? this surface antibody initiates the immune tant and previously unappreciated point) until To answer some of these questions, I pro- response, and (iii) the self-reactive lympho- they are activated via a set of germ line– posed the Danger model, which suggests that cytes are deleted early in life [based on encoded pattern recognition receptors (PRRs) the immune system is more concerned with Owen’s discovery that nonidentical cattle that recognize conserved -associated damage than with foreignness, and is called twins were mutually tolerant of each other’s molecular patterns (PAMPs) on bacteria. On into action by alarm signals from injured blood cells (12)]. This straightforward model activation, APCs up-regulate costimulatory tissues, rather than by the recognition of non- gained general acceptance when Medawar et signals, process the bacterial antigens, and self (2, 3). In the intervening 7 years, in al. found that adult mice would accept for- present them to passing T cells (Fig. 1D). The conversations with a wide variety of people, I eign skin grafts if they had been injected as PRRs, he wrote, allow APCs to discriminate discovered that this simple idea not only of- babies with cells from the donors (13). In between “infectious-nonself ” and “noninfec- fers answers to broad immunological ques- 1960, Burnet and Medawar shared the Nobel tious-self ” (18). tions, it also covers many details that had not Prize for their work, and the SNS discrimi- Although the essence of SNS recognition nation model has dominated the field ever was temporarily saved, Janeway’s infectious- since. nonself (INS) model created new complexi- Ghost Lab, Laboratory for Cellular and Molecular Im- The original SNS model has changed, ties while solving old problems. It could not munology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD however, to accommodate incompatible new explain why viruses stimulate immunity, why 20892, USA. E-mail: [email protected] findings. It was first modified in 1969, after transplants are rejected, what induces autoim-

www.sciencemag.org SCIENCE VOL 296 12 APRIL 2002 301 R EFLECTIONS ON S ELF:IMMUNITY AND B EYOND munity, why tumors are sometimes spontane- forth (Fig. 1e). Although purely theoretical at predictive power of this model provides insight ously rejected, or how nonbacterial adju- the time (20), many alarm signals have since into many of the things that the immune system vants, such as alum, work. Over the years, been empirically revealed (9). Alarm signals does right, as well as many of the things it Janeway’s model has been modified to ac- can be constitutive or inducible, intracellular seems to get wrong (21). count for some of these issues (see Viewpoint or secreted, or even a part of the extracellular by Medzhitov and Janeway in this issue) matrix. Because cells dying by normal pro- Danger Signals: Common Ground for (19), suggesting, for example, that viral dou- grammed processes are usually scavenged the INS and Danger Models? ble-stranded mRNA is a signature of foreign- before they disintegrate, whereas cells that Although they differ greatly in detail, both ness. What then of viruses that do not gener- die necrotically release their contents, any the INS and the Danger models assume that ate double-stranded RNA, adjuvants that do intracellular product could potentially be a resting APCs can be activated by signals not incorporate bacterial products, trans- danger signal when released. Inducible alarm from their immediate environment. The INS plants, and autoimmunity? Even with all of signals could include any substance made, or model has found support in the recent discov- the modifications that SNS/INS models have modified, by distressed or injured cells. The ery of the evolutionarily conserved membrane- undergone over the years, they still have dif- important feature is that danger/alarm signals bound Toll-like receptors (TLRs), which act as ficulty with some of these fundamental pro- should not be sent by healthy cells or by cells PRRs for components of bacteria and fungi, cesses. undergoing normal physiological deaths. and initiate immune responses in organisms Although this may seem to be just one more as distant as flies and mammals (22–26). The Danger Model step down the path of slowly increasingly com- There are presently 10 known mammalian Standing on the shoulders of the SNS models, plex cellular interactions, this small step drops TLRs, which bind a wide range of biological the Danger model added another layer of us off a cliff, landing us in a totally different molecules and awaken resting APCs (27). cells and signals (2), proposing that APCs are viewpoint, in which the “foreignness” of a The Danger model has been supported by activated by danger/alarm signals from in- pathogen is not the important feature that trig- the discovery of endogenous, nonforeign alarm jured cells, such as those exposed to patho- gers a response, and “self-ness” is no guarantee signals (9), including mammalian DNA (28), gens, toxins, mechanical damage, and so of tolerance. The surprising explanatory and RNA, heat shock proteins (Hsps), interferon-␣

Fig. 1. A history of immunological models.

302 12 APRIL 2002 VOL 296 SCIENCE www.sciencemag.org R EFLECTIONS ON S ELF:IMMUNITY AND B EYOND (an inducible protein often made by virus-in- cific for a bacterial LPS, cellular Hsp70 pro- nals, and the microbes subsequently evolved fected cells), interleukin-1␤, CD40-L (a surface tein, and hyaluron? mechanisms to use these receptors to enhance molecule on activated platelets and activated T One possibility is that we may be looking at their own survival. From this vantage point, it cells), and breakdown products of hyaluron the PRRs completely backwards (7). Perhaps may no longer be surprising that the TLRs (made when vessels are damaged). PRRs have not evolved to bind to pathogens at bind to so many different kinds of molecule, There is no reason why APCs should not all. Perhaps the pathogens have evolved to bind as each type of pathogen will have evolved its respond to both endogenous and exogenous to them! Many cell surface molecules involved own way of binding to a TLR. signals. Vertebrates and bacteria have shared in normal physiological functions are targeted S. Y. Seong has suggested the even more eons of evolutionary time and space, and thus by pathogens. Human immunodeficiency virus, intriguing possibility (34) that the same receptors for endogenous and exogenous sig- for example, binds to CD4, CCR5, and CxCR4, alarm signals may be used by many different nals may have evolved simultaneously. In- and Toxoplasma also seems to bind to CCR5 organisms. Because life evolved in water, any deed, there is evidence that these receptors (30), whereas Staphylococcus and Streptococ- hydrophobic portion (Hyppo) of a given mol- are often the same molecules. For example, cus bind to a conserved loop on T cell receptors ecule is usually buried in the depths of that TLR4 is a receptor for the bacterial product and to the Fc portion of antibodies. Coxsackie molecule, or hidden in the lipid membrane of lipopolysaccharide (LPS), the endogenous virus binds to intracellular cell adhesion mole- the cell, and could act as an alarm signal if cellular molecule Hsp70, and the extracellu- cule–1 (ICAM-1), rabies to N-CAM, and Ep- exposed (34). For example, the hydrophobic lar breakdown products of hyaluron; TLR2 stein-Barr virus to complement receptor 2, thus part of LPS is crucial for its immunostimu- binds bacterial lipoproteins and Hsp 60; and activating a B cell as it enters. No one suggests lating properties, yet LPS is normally an in- TLR9 binds to DNA CpG sequences (found that these molecules have evolved to act as tegral bacterial membrane molecule and its in all living creatures). Thus, it appears that receptors for pathogens. We assume instead Hyppos are hidden in the membrane. How- the TLRs can recognize both endogenous and that the pathogen’s ability to bind to these ever, when released by damaged or dead exogenous molecules. The binding character- molecules serves the pathogen’s agenda, not bacteria, the newly exposed Hyppos could act istics of a newly discovered family of intra- ours. Similarly, the PRRs may be misnamed. as a bacterial alarm signal (or perhaps a type cellular proteins, called nucleotide-binding For example, CD14, which recognizes ap- of quorum sensor (35), perhaps signaling the oligomerization domain (NOD) receptors, are optotic cells (31), has been called a PRR surviving bacteria to sporulate or otherwise not yet as well worked out, but it is beginning because it also binds to bacterial LPS (32). change their behavior. Plant and animal cells to appear that they too can respond to both However, mice lacking CD14 resist Gram- also have an abundant supply of hidden Hyp- injury/pathogen-related signals and normal negative bacteria more vigorously than pos in their membranes and cytoplasm. Dur- physiological signals involved with their normal littermates (33), suggesting ing protein synthesis, Hsps and other chaper- (29). Indeed, one of the puzzling features of that the LPS-CD14 interaction is more fa- ones bind to the Hyppos of nascent proteins TLRs and NODs is that each one can bind to vorable to the bacterium than to the host. to prevent their aggregation. Should a cell be many different kinds of molecules. How can Thus, perhaps TLRs and NODs originally disrupted, the Hyppos of both the nascent one receptor—TLR4, for example—be spe- evolved as receptors for injury-related sig- proteins and their chaperones would be ex- posed. Future evidence may show that our Fig. 2. Partitioning the immune systems may thus be using TLRs and universe of antigens. other receptors to respond to truly ancient SNS models split all multipurpose signals of distress that cross antigens into two sets: species barriers. self and nonself (sets a and b). The INS model divides anti- Dangerous Self and Harmless Foreign: gens into “noninfec- The Uncommon Ground Between the tious self ” (set a) and Two Models “infectious nonself ” Although the INS version of the SNS models (set f ), suggesting the and the Danger model have some common existence of patho- gen-associated molec- features, their basic assumptions about what ular patterns (PAMPs) initiates immunity are fundamentally differ- that are evolutionarily ent. Is it microbial nonself or is it danger? As conserved on patho- nonself is sometimes dangerous, the defini- gens that are evolu- tions overlap, but they are not identical (Fig. tionarily very distant 2). For dangerous foreign pathogens (Fig. 2, from their hosts, and that the host APCs can sets d and e) or harmless self (Fig. 2, set a), therefore have germ the two models make the same predictions. line–encoded pattern However, some things (Fig. 2, sets b and f) recognition receptors are foreign but harmless (e.g., fetuses), (PRRs) to detect them. whereas others (Fig. 2, set c) are self but It tends to ignore the harmful (e.g., some mutations). For these en- subsets b and f. The Danger model parti- tities lying outside the overlapping sets, the tions antigens into INS and Danger models make different pre- those associated with dictions, and these are therefore the interest- dangerous entities or ing test cases. Below I will briefly cover a harmless ones, defin- few of these (10). ing as dangerous anything that induces stress or nonphysiological death of a cell. Dangerous Foreign entities that are not associated with entities may be self (set c), such as mutations that lead to stress or inappropriate cell death or inefficient scavenging; or nonself, such as pathogens (set e), environmental toxins (set d), and such. microbes include transplants and fetuses. Why Set f would contain evolutionarily distant organisms that have PAMPs, but that are not dangerous should the former be rejected and the latter not? (e.g., symbiotic organisms, well-adapted viruses). Although the INS model would suggest that

www.sciencemag.org SCIENCE VOL 296 12 APRIL 2002 303 R EFLECTIONS ON S ELF:IMMUNITY AND B EYOND neither should be rejected because they are not teresting aspect of the Danger model has been a because they respond to stress-induced self mol- associated with microbial stimulators, and the deep-rooted shift in thinking that it inspired. The ecules rather than to the foreign entities seen by old SNS models would suggest that both should shift came in two phases that abruptly expanded the “adaptive” lymphocytes (45). These cells be rejected because they are nonself, the Dan- the model’s explanatory range. Originally con- have no place in the SNS and INS models, and ger model suggests that healthy fetuses should ceived to answer the first question the immune they have remained on the fringe of respectabil- not be rejected because they do not send alarm system must consider when faced with a poten- ity. But, seen from the standpoint that immunity signals. Transplants, however, cannot be per- tial threat—namely, whether to respond—the is governed by the tissues, these self-reactive formed without surgical and/or ischemic dam- model now also offers a suggestion for the next cells do not seem so odd (7), and Bendelac calls age. Thus, to induce the acceptance of trans- question—having decided to respond, what this “autoimmunity by design” (45). For exam- plants without lifelong immunosuppression, we kind of response should it make? How does the ple, the dendritic epidermal T cells (␥␦ T cells) should mimic the body’s own way of inducing immune system know whether to generate killer found in mouse and bovine skin all express tolerance, i.e., by blocking the endogenous T cells to eliminate a virus or immunoglobulin E exactly the same receptor (46), arise late in fetal alarm and/or costimulatory signals. Most of the (IgE) antibody to catch a worm? In immunolog- life, and emigrate in one wave to the skin, where current immunosuppressive drug protocols en- ical terms, how does it determine the effector they settle quietly. When stimulated by the ap- deavor to block Signal One (antigen recogni- “class” of the response? pearance of stress-induced molecules on kera- tion). Although this effectively blindfolds the The first shift came from the realization tinocytes (47), they produce epidermal cell lymphocytes, it also prevents them from be- that the immune system may not be the ulti- growth factor, IL-2, and IFN-␥. These cells are coming tolerant, and consequently the drugs mate controller of immunity. Like most im- clearly not the kind of lymphocytes we are must be given for life. In contrast, studies in munologists, I had thought that immunity is accustomed to. They seem to be there to pro- rodents and monkeys (36–38) have shown that controlled by the cells of the “adaptive” im- duce cytokines that heal damaged skin by in- short-term treatment with costimulation block- mune system (lymphocytes) or the more an- ducing cell growth and nudging local immunity ers, or blockers of alarm signaling (39), can cient “innate” immune system (such as mac- toward a DTH. In human gut, T cells expressing lead to long-term graft acceptance in the ab- rophages, dendritic cells, and the complement self-reactive V␦6 receptors also respond to sence of long-term immunosuppression. Per- system). I now believe that the ultimate pow- stress-induced molecules (48). Many other ␥␦ T haps the lower extent of damage can explain er lies with the tissues. When healthy, tissues cells may be similar, responding to endogenous why kidneys from living donors are accepted induce tolerance. When distressed, they stim- stress signals rather than to foreign antigens. more easily than those from cadavers (4). Sim- ulate immunity, and (continuing down this T cells bearing ␣/␤ receptors can also be ilarly, fetuses should not elicit immunity, in path) they may also determine the effector usefully self-reactive. The thymus, bone mar- spite of being foreign, as long as they are class of a response. Although it has long been row, and liver contain NK1 T cells specific for healthy and do not send alarm signals (10). thought that the effector class is tailored to the ancient MHC-like molecule, CD1, which is Tumors are entities for which both the INS the targeted pathogen (e.g., virus or worm), I expressed by activated but not resting APCs and the Danger models have the same predic- now think that it is tailored to the tissue in (49). Activated NK1 T cells from the thymus tion, namely, that tumors should not stimulate which the response occurs. produce copious amounts of IL-4, a cytokine immunity, either (INS) because they are not Different tissues seem to have different that skews local immune responses away from associated with microbial stimulators, or (Dan- means of determining the effector class of a a DTH and toward the production of IgG1 and ger) because they are healthy growing cells that response. For example, the class of response IgE. Furthermore, T cells specific for brain do not send alarm signals. Thus, to eradicate a that occurs most often in the skin (e.g., after proteins can lessen the secondary damage that tumor, we should infect it (40), or cause it exposure to poison ivy, TB tests, or subcutane- follows neural injury (50). repeated damage to alert the local APCs [as Bill ous vaccinations), called “delayed type hyper- All of these tissue-localized cells appear Coley did in the late 1800s (41, 42)], or we sensitivity” (DTH), is characterized by swell- to be useful self-reactive cells involved in should vaccinate repeatedly with a tumor vac- ing, redness, an influx of macrophages, and the local immunity, and there may be many other cine that stimulates immunity. production of tumor necrosis factor (TNF) and localized and/or circulating cells doing simi- For autoimmunity, the Danger model of- interferon-␥ (IFN-␥). Unlike skin, however, lar jobs. For example, a lot of effort has gone fers a unique suggestion that would not arise both the gut and the eye can be destroyed by into the search for the foreign ligands recog- from the SNS or the INS models. Starting DTH responses, and the most common re- nized by circulating ␥␦ T cells. After more with the view that “bad” death or cell stress sponse in these organs is the production of IgA, than a decade, very few have been found, and can elicit an immune response, the model an antibody found at high levels in tears, saliva, these include such ubiquitous cellular mole- suggests that some autoimmune diseases may milk, and gut secretions. To ensure that IgA is cules as polyprenyl pyrophosphate (51) and be caused by mutations in genes governing made, and TNF and IFN-␥ are not, the cells of phosphorylated nucleotides (52). the normal physiological death and clearance the anterior chamber of the eye produce vaso- Perhaps, if we move from the idea that processes, or by environmental pathogens or active intestinal peptide (VIP) and transforming every lymphocyte should be directed against toxins that cause cellular stress or death. In growth factor–␤ (TGF-␤), two cytokines that non-self antigens whose appearance stimulates these cases, the immune system is not at fault; are also made by the gut and that promote a the response, and consider instead the possibil- it is doing its job of responding to alarm switch to IgA and suppress the DTH response ity that immunity is controlled by an internal signals (but, in these cases, to the detriment (43, 44). Thus, local tissue cells strongly influ- conversation between tissues and the cells of of the host). If we could pinpoint these mu- ence the local immune response. the immune system (53), we may regain a tations or environmental agents, we might be The second shift came from the realization renewed sense of the self that we have lost. able to reduce the incidence of autoimmune that tossing out the idea that the immune system diseases. uses SNS discrimination to decide whether to References and Notes respond leaves us free to use self-recognition in 1. D. E. Koshland Jr., Science 248, 1273 (1990). 2. P. Matzinger, Annu. Rev. Immunol. 12, 991 (1994.) A Renewed Sense of Self: Expanding a positive way to control other aspects of the 3. The model arose from 2 years of discussions with the Horizons of the Danger Model response. Many organs harbor special popula- Ephraim Fuchs (54). There is inherent beauty in a model that uses tions of lymphocytes that appear to be evolu- 4. A. M. D’Alessandro et al., Surgery 124, 604 (1998). 5. P. J. Heidt, J. M. Vossen, J. Med. 23, 161 (1992). very few assumptions to explain a wide variety tionarily old, often have limited receptor diver- 6. G. R. Hill et al., Blood 90, 3204 (1997). of phenomena. However, for me, the most in- sity, and have been called “innate lymphocytes” 7. P. Matzinger, Semin. Imunol. 10, 399 (1998).

304 12 APRIL 2002 VOL 296 SCIENCE www.sciencemag.org R EFLECTIONS ON S ELF:IMMUNITY AND B EYOND

8. P. Matzinger, Scand. J. Immunol. 54, 4 (2001). gen in house dust mite, is a protease that attacks the 43. J. Stavnezer, J. Immunol. 155, 1647 (1997). 9. S. Gallucci, P. Matzinger, Curr. Opin. Immunol. 13, surface of B cells and lung epithelium. Likewise, bee 44. H. Kimata, M. Fujimoto, Eur. J. Immunol. 24, 2262 114 (2001). venom is not an innocuous substance. From the stand- (1994). 10. Supplementary material is available on Science On- point of the Danger model, it is not surprising that the 45. K. Benlagha, A. Bendelac, Semin. Immunol. 12, 537 line at www.sciencemag.org/cgi/content/full/296/ immune system responds to allergens. What is not clear, (2000). 5566/301/DC1. however, is why some individuals make IgG whereas 46. R. Boismenu, W. L. Havran, Curr. Opin. Immunol. 9, 11. F. M. Burnet, The Clonal Selection Theory of Acquired others make IgE (the antibody associated with allergy). 57 (1997). Immunity (Vanderbilt Univ. Press, Nashville, TN, 22. B. Lemaitre, E. Nicolas, L. Michaut, J. M. Reichhart, 47. M. Girardi et al., Science 294, 605 (2001). 1959). J. A. Hoffmann, Cell 86, 973 (1996). 48. V. Groh et al., Science 279, 1737 (1998). 49. A. Bendelac et al., Annu. Rev. Immunol. 15, 535 (1997). 12. R. Owen, Science 102, 400 (1945). 23. R. Medzhitov, P. Preston-Hurlburt, C. A. Janeway Jr., 50. M. Schwartz, J. Mol. Med. 78, 594 (2001). 13. R. E. Billingham, L. Brent, P. B. Medawar, Nature 172, Nature 388, 394 (1997). 51. C. T. Morita et al., Res. Immunol. 147, 347 (1996). 603 (1953). 24. R. Medzhitov, C. Janeway Jr., Trends Microbiol. 8, 452 14. P. Bretscher, M. Cohn, Science 169, 1042 (1970). 52. P. Constant et al., Science 264, 267 (1994). (2000). 53. The Network model, which proposed that “self ” is 15. K. J. Lafferty, A. Cunningham, Aust. J. Exp. Biol. Med. 25. D. A. Kimbrell, B. Beutler, Nature Rev. Genet. 2, 256 defined in a positive way, also emphasized the idea of Sci. 53, 27 (1975). (2001). connectedness. The idea was that lymphocytes react 16. M. K. Jenkins, R. H. Schwartz, J. Exp. Med. 165, 302 26. A. Aderem, R. J. Ulevitch, Nature 406, 782 (2000). (1987). 27. S. Akira, K. Takeda, T. Kaisho, Nature Immunol. 2, 675 against each other’s antigen-specific receptors and 17. C. A. Janeway Jr., Cold Spring Harbor Symp. Quant. (2001). maintain a balance of self-reactive and foreign-reactive Biol.࿜࿜࿜࿜54, 1 (1989). 28. K. Ishii et al., J. Immunol. 167, 2602 (2001). cells (55). The proponents of the network have been 18. , Immunol. Today 13, 11 (1992). 29. N. Inohara, G. Nunez, Oncogene 20, 6473 (2001). arguing for years that the study of single lymphocytes 19. R. Medzhitov, C. A. Janeway Jr., Science 296, 298 (2001). 30. J. Aliberti et al., Nature Immunol. 1, 83 (2000). is an inappropriate way to study the immune system, 20. Like physicists, who deduced the need for a new particle 31. A. Devitt et al., Nature 392, 505 (1998). but that we should study the connectivity between cells based on the behavior of the system, Bretscher and 32. J. Pugin et al., Immunity 1, 509 (1994). (56) Finally, after years of finding the model intriguing, Cohn (14) and Lafferty and Cunningham (15) postulat- 33. A. Haziot et al., Immunity 4, 407 (1996). but narrow, I agree. However, I think that we should not ed cells and/or signals for which, at the time, there was 34. S. Y. Seong, personal communication. limit the study to interactions between lymphocytes no evidence. Later experiments showed resoundingly 35. C. Fuqua, M. R. Parsek, E. P. Greenberg, Annu. Rev. but expand it to include their conversations with all the that they were correct. In a similar vein, Janeway pos- Genet. 35, 439 (2001). bodily tissues, which have the ultimate say. tulated a new state for a previously known cell, the APC. 36. C. P. Larsen et al., Nature 381, 434 (1996). 54. M. Matzinger, E. Fuchs, J. NIH Res. 8, 35 (1996). Up to that time, APCs were thought to be constitutively 37. D. J. Lenschow et al., Transplantation 60, 1171 (1995). 55. N. Jerne, Ann. Immunol. (Paris) 125C, 373 (1974). active, but a seemingly small glitch in the behavior of 38. A. D. Kirk et al., Proc. Natl. Acad. Sci. U.S.A. 94, 8789 56. A. Coutinho, Scand. J. Immunol. 42, 3 (1995). the system (the need for adjuvant) led him to suggest (1997). 57. I thank past and present members of the ghost lab (S. that they were normally quiescent and needed to be 39. J. C. J Thomas et al., Transplantation 68, 1660 (1999). Gallucci, J. Wright, S. Wolf, K. Abdi, E. Bachelder, O. activated. These insights showed that theoretical biol- 40. J. Lindenmann, P. A. Klein, J. Exp. Med. 126,93 Alpan, P. Rohwer-Nutter, A. Perez, R. Massey, and D. ogy and physics may have more in common than is (1967). Culp), as well as H. Arnheiter and Y. Rosenberg for sometimes thought. 41. W. B. Coley, JAMA (20 August 1898), p. 389. comments on the manuscript. I would also like to 21. Allergy, for example, is a partial conundrum. Many aller- 42. S. S. Hall, A Commotion in the Blood (Holt, NY, 1998), send a kiss to the anonymous referee who placed a gens are dangerous substances. Der-p1, the major aller- p. 198. heavy but appropriate boot in the right place.

VIEWPOINT Recognition and Rejection of Self in Plant Reproduction June B. Nasrallah

Plant self-incompatibility (SI) systems are unique among self/nonself ments that are tolerated and do not elicit a recognition systems in being based on the recognition of self rather than response. In SI, self is the condition that nonself. SI in crucifer species is controlled by highly polymorphic and elicits the response and is inhibited, where- co-evolving genes linked in a complex. Self recognition is based on as nonself is the condition that is ignored allele-specific interactions between stigma receptors and pollen ligands and does not elicit a response. that result in the arrest of pollen tube development. Commonalities and differences between SI and other self/nonself discrimination systems are A Variety of Plant SI Systems discussed. As an advantageous outbreeding device, SI is widely distributed in flowering plants The concept of self/nonself discrimination outcome, namely the inhibition of self-related (2). It evolved independently in several was elaborated by Burnet (1) as a way to pollen tube development and, consequently, lineages, and the SI systems adopted by describe specificity in the immune response the prevention of sperm cell delivery to the different plant families vary with respect to and is most often associated with the field of ovules. site and mechanism of self inhibition. In . It is perhaps less well known SI systems are said to discriminate be- self-incompatible species of the crucifer that, in the plant kingdom, sophisticated self- tween self and nonself because they pro- family (e.g., Brassica species and close recognition systems have evolved that allow duce different outcomes in self- and cross- relatives of Arabidopsis thaliana), SI dis- plants with perfect (hermaphroditic) flowers pollinations. Specificity in SI is typically rupts hydration and germination of a pollen to avoid inbreeding. These intraspecific controlled by one or more highly polymor- grain on the stigma epidermis, thus pre- prefertilization mating barriers are collective- phic genetic loci. In the context of SI, self venting growth of pollen tubes into the ly known as self-incompatibility (SI). This and nonself mean, respectively, genetic subepidermal tissues of the pistil. In other term encompasses several systems that are identity and nonidentity at the SI locus (or families, SI acts after pollen germination mechanistically distinct but have the same loci) in pistils and pollen. The outcome of and pollen tube ingress into the pistil, either this discrimination is the converse of that of within the stigmatic zone (as in the poppy Department of Plant Biology, Cornell University, the immune response, in which case self family), or later, within the style (as in the Ithaca, NY 14853, USA. *E-mail: [email protected] has been classically defined as those ele- tobacco, rose, and snapdragon families).

www.sciencemag.org SCIENCE VOL 296 12 APRIL 2002 305