Why Is the Adaptive Immune System So Diverse and Specific?

Why is the adaptive immune system so diverse and speciﬁc?

What will you learn today?

Self tolerance is more complicated than what the textbook says

Repertoires are very diverse because lymphocytes have to be speciﬁc

Simple mathematical models help us thinking clearly

Polymorphism versus diversity of MHC First short refresh on tolerance Death by neglect 0.3

R/R0

0.2

0.1

0 1 2 3 4 MHC diversity (log M )

Figure 2: Positive and negative selection according to the avidity model [13]. The curve in (a) depicts the distribution of thymocyte avidities for self peptide–MHC complexes. In our model, the chance p to be positively selected by a single MHC type is the chance that the avidity between the thymocyte T cell receptor and any of the self peptide– MHC complexes exceeds threshold T1. Thymocytes with avidities for self peptide–MHC complexes exceeding the upper threshold T2 are negatively selected (with chance n per MHC type). Panel (b) depicts the size of the T cell repertoire as a function of MHC diversity. The number of clones in the functional repertoire R is plotted as a fraction of the total initial lymphocyte repertoire R0. Parameters are: p =0.01, and n =0.005.

2 MHC diversity within the individual

Since individual MHC diversity increases the presentation of pathogens to the immune system, one may wonder why the number of MHC genes is not much higher than it is. The argument that is mostly invoked is that more MHC diversity within the individual would lead to T cell repertoire depletion during self tolerance induction. This argument is incomplete, however, because more MHC diversity could also increase the number of clones in the T cell repertoire through positive selection. In order to be rescued in the thymus, lymphocytes need to recognize MHC–self peptide complexes with su⇤cient avidity. A high MHC diversity thus increases both the number of lymphocyte clones that are positively selected and the number of clones that are negatively selected. To calculate the net eect of these two opposing processes we develop a simple mathematical model [5].

Consider an individual with M dierent MHC molecules and an initial T lymphocyte repertoire consisting of R0 dierent clones. Let p and n denote the (unconditional) chances that a clone is positively selected by a single MHC type, because its avidity is higher than a threshold T1, or negatively selected because its avidity exceeds a higher threshold T2, respectively (see Fig. 2). By this deﬁnition, thymocytes can only be negatively selected by MHC molecules by which they are also positively selected, i.e. n

R = R (1 n)M (1 p)M , (7) 0 [5]. The functional repertoire R thus contains all T cell clones that⇥ fail to be negatively selected, minus the ones that also fail to be positively selected by any of the M dierent MHC molecules of the host.

Experimental estimates for the parameters of this model have recently become available. In mice, around 3% of the T cells produced in the thymus end up in the mature T cell repertoire, and at least 50% of all positively selected T cells have been shown to undergo negative selection in the thymus [17]. Thus, 94% of all thymic T cells fail to be positively selected by any of the MHC molecules in the host [17]. These estimates can be used to calculate the chances p and n of a T cell clone to be positively or negatively selected by a single type of MHC molecule. Taking into account that inbred mice are homozygous and therefore express 3 types of class I MHC and 3 types of class II MHC molecules, p and n follow from: (1 p)6 =0.94 and n = p/2. This yields p =0.01 and n =0.005.

Using these experimental estimates, the number of clones in the functional T cell repertoire R increases with

3 What is the diversity of the T cell repertoire? Parham cites estimate from Arstila [Science, 1999]: 2.5x107 Nowadays next generation sequencing (NGS):

The current study was designed explicitly to improve estima- A CD4 naive T cells CD8 naive T cells tion of the TCR repertoire richness of stringently defined naïve p=0.008 p=0.008 and memory T cells. By sequencing multiple replicate TCRB li- 2x108 braries of cells from each T-cell subset and applying nonparametric 1x108 statistical analysis, we find that human TCR repertoires are an order of magnitude more diverse than previously estimated. 5x107 Despite significant age-related decreases in richness, humans 7 maintain high diversity during healthy aging. Strikingly, we find 2x10 age-associated changes in the distribution of clonal sizes, in 1x107 particular in the naïve compartment, which may reflect un- Distinct TCRB sequences evenness of homeostatic expansion with clonal expansion of 20 - 35 70 - 85 20 - 35 70 - 85 2014 PNAS Qial. et some naïve T cells that equal or exceed clonal sizes of memory T B p=0.008Age p=0.008Age cells. The inequalities in clonal sizes could compromise the im- 4x107 mune response to the majority of antigenic epitopes while Note that this is β-chains only (nucleotides & Chao2)! causing an increased responsiveness to selected few epitopes. 2x107

sequences In mice almost every naive T cell unique TCR β Results 1x107 Gene Segment Use and CDR3 Features in Young and Elderly TCRB Rearrangements. In initially evaluating TCRB repertoires in 5x106 young and elderly subjects we compared the composition of TCRB gene rearrangements at the level of TCRBV and TCRBJ Distinct TCR 20 - 35 70 - 85 20 - 35 70 - 85 gene segment use and the features of the CDR3-encoding Age (years) Age (years) junctional nucleotides. Apheresis lymphocytes were obtained C CD4 memory T cells CD8 memory T cells from four 20- to 35- and five 70- to 85-y-old healthy adults who p=0.06 p=0.11 were regular platelet donors. Naïve CD4 and CD8 T cells were purified by cell sorting. We used a stringent definition of naïve 2x106 cells (CD3+CD4+ or CD8+CCR7+CD45RAhighCD28+) and very restrictive gating to ensure purity. Approximately 1.5–3 million sequence reads were obtained for each T-cell subset of each 1x106 donor (Table S1). TCRBV and TCRBJ gene segments were used at comparable frequencies in the repertoires of naïve CD4 and 0.0 CD8 T cells in young and old individuals (Fig. S1A), whereas the sequences TCRB Distinct 20 - 35 70 - 85 20 - 35 70 - 85 memory repertoires of CD4 and, particularly, CD8 T cells show variable and individual-specific gene segment use frequencies 6 p=0.06 p=0.10 (Fig. S1B). CDR3 sequences in the young and elderly were D 2x10 comparable in length and did not show any definitive age-related features (Fig. S1 C and D). 6

sequences 1x10 High Richness of Naïve CD4 and CD8 TCRB Repertoires in Young and β Elderly Adults. To obtain sequence data to estimate global TCRB repertoire richness we used the experimental design of analyzing 0.0 aseriesofreplicatelibrariesfromindependentcellaliquotsfrom 20 - 35 70 - 85 20 - 35 70 - 85 each T-cell subset in each individual. These replicates allowed us to Distinct TCR Age (years) Age (years) calculate repertoire richness by applying the “Chao2” estimator, anonparametricestimatorofunseenspecies(14).Theapproach Fig. 1. Age is associated with a modest decrease in diversity of the TCRB reper- allows estimation of the extent to which the full repertoire is cov- toire. TCRB sequences were obtained from replicate samples of naïve (A and B)and ered and use of this information to determine a lower bound of the memory (C and D)CD4andCD8Tcells.AlowerboundofTCRBrichnesswasesti- total number of species in the repertoire. Because the Chao2 estimated by applying nonparametric statistics using the Chao2 estimator. Results are mator requires only a binary characterization of presence or ab- shown for nucleotide (A and C)andderivedaminoacidsequences(B and D). – – sence of each clone in each replicate library, it circumvents the Estimates were compared by Wilcoxon Mann Whitney test. Increase in age is associated with a decline in richness of naïve CD4 and CD8 T cells; however, the challenges that arise in experimental designs using only a single repertoire in the elderly remains highly diverse. Richness in CD4 and CD8 memory library and avoids confusing the effects of PCR amplification with Tcellsmarkedlydiffered,whereastheimpactofagewasnegligiblysmall. the presence of expanded T-cell clones. To not count possible sequencing errors as independent sequences, we rejected single sequences as erroneous if a highly similar clone of greater frequency (15). Second, we estimated the confidence intervals using the ap- was identified in the same library (see Materials and Methods for the proach originally developed by Chao (16). The 95% confidence definition of similarity). intervals with both methods were very narrow (Table S2). The lower bounds on TCRB gene richness obtained with this Naïve TCRB repertoire richness declined significantly in the approach yielded higher estimates than previous studies (Fig. 1). 70- to 85-y-old adults to a lower bound richness of 8–57 million Young adults carried an estimated 60–120 million different different nucleotide sequences encoding ∼5–15 million TCR TCRB genes, both in the CD4 and CD8 naïve T-cell repertoires. β-chain amino acid sequences (P = 0.008, Fig. 1 A and B). In- This high diversity in nucleotide sequences was reflected in a terestingly, the estimates in elderly CD4 and CD8 naïve T cells large functional repertoire of TCR β chains with a lower boundary were similar despite the greater decline in CD8 compared with of ∼20 million different amino acid sequences. To determine the CD4 naïve T-cell numbers with aging (17, 18). robustness of our estimates, we used two approaches to estimate confidence intervals. We applied the BCa variant of bootstrapping CD4 and CD8 Memory T Cells Differ in TCRB Richness Independent of that is designed for obtaining confidence intervals when the un- Age. Memory cells have been selected from the naïve repertoire derlying bootstrap distributionisnotsymmetricaboutitscenter and clonally expanded in response to antigen and are therefore

13140 | www.pnas.org/cgi/doi/10.1073/pnas.1409155111 Qi et al. Immunity CD4+ Memory-Phenotype T Cells in Unexposed Adults

No evidence for clonal deletion?

A Figure 2. Frequency Analysis of the Human 100 + self antigens Antigen-Speciﬁc CD4 Lymphocytes

CD4) All blood samples were obtained from individuals

6 seronegative for HIV and CMV exposures. HSV 10 exposure is as indicated (naı¨ve versus exposed). (A) Frequencies of tetramer-tagged cells per million CD4+ T cells. Each symbol represents an 1 antigen-speciﬁc population from one individual, and the bar indicates the mean of experiments cells (per 10

+ performed independently with blood obtained at 0.1 different times. Number of Ag specific cellsspecific Ag of Number Tet (B) Comparison between T cells recognizing Antigen HIV CMV Fib FLU-HA FLU-PAtetanus gp100PPins a novel foreign antigen (pre: HIV-1, CMV, and Specificity: HSV-naive FLU-PB1 HSV-exposed HSV-naı¨ve), exposed foreign antigen (post: Flu, tetanus, and HSV-exposed), and self-antigen (self: BCFrequencies of tetramer-tagged cells per million CD4 T cells: 1-10D per 106 Fib, gp100, and PPins). The figure summarizes Each symbol representsns an antigen-specific populationns from one individualp .= 0.0192 data from all 26 individuals. <0.0001Fib (fibrogen), <0.0001 gp100, PPins (preproinsulin):16 self antigens. 16 (C) Precursor frequencies of self-specific lym- CD4) CD4) CD4) 8 90 8 6 phocytes that recognized gp100, Fib, or PPins in 6 6 60 4 Su et al. Immunity 20134 people ages <40 years (n = 7), 40–60 years (n = 13), 20 2 2 and >60 years (n = 6). 15 1 1 (D) Precursor frequency of self-specific lympho- 10 0.5 0.5 cytes that recognized gp100, Fib, or PPins in cells (per 10

cells (per 10 0.25 males (n = 17) and females (n = 9) (see also Fig-

cells (per 10 5 0.25 + +

+ ure S2). 0 0.125 0.125 Tet Tet <40 40-60 >60 MF

Tet pre post self Antigen Exposure Age Gender

The Unprimed CD4+ T Cell Repertoire Contains epitopes (HA and PB1), highlight the heterogeneity of a T cell Pre-existing Memory-Phenotype T Cells response in humans. Another notable feature in these data is To broadly survey antigen-specific CD4+ T cells, we selected the absence of a statistically significant difference between the self-peptides from gp100, fibrinogen (Fib), or preproinsulin frequencies of cells that recognize self-antigens (PPins, Fib, or (PPins) for their relevance in antitumor responses and autoim- gp100) and those that bind to a naı¨ve foreign antigen from munity (Nielen et al., 2005; Yuan et al., 2009; Zhang et al., HIV-1, CMV, or HSV in uninfected individuals (Figure 2B). These 2008). Foreign antigens from gag p24 from HIV-1, pp65 from self-antigen-specific T cells were examined for potential age- or CMV, and VP16 from HSV were selected because highly sensi- gender-dependent change, particularly because the risk of tive serologic tests are available for clearly distinguishing autoimmunity is generally higher in the elderly and in females. exposed individuals from unexposed individuals. The sensitivity Although we did not detect significantly more gp100-, Fib-, or of the HSV, HIV, and CMV serologic tests are 97%–100%, and PPins-specific T cells in older individuals (Figure 2C), we did with respect to HIV, the risk of a false negative is estimated to note a gender-specific difference—there were 1.7-fold more be less than 1 in 2.6 million blood donations nationwide (Lack- gp100-, PPins-, and Fib-specific lymphocytes in females than ritz et al., 1995). We also examined several epitopes from the in males (Figure 2D). influenza virus and tetanus toxin, to which most people have The absence of an obvious frequency difference between been exposed. Specific epitopes from each protein were self- and foreign-antigen-specific populations suggests that chosen from the literature on the basis of T cell stimulation negative selection is incomplete for the particular self-specific- and evidence that they are the product of natural processing ities examined. Alternatively, postthymic antigen stimulation and presentation (Alexander et al., 2010; Arif et al., 2004; Auger might have resulted in expansion of certain populations that et al., 2005; Bronke et al., 2005; Diethelm-Okita et al., 1997; masked a difference in their initial frequencies. To assess the Lamb et al., 1982; Norris et al., 2004; Novak et al., 2001; latter, we analyzed the memory versus naı¨ve phenotype of these Phan et al., 2003). tetramer-positive cells by using the classical memory marker We made tetramers by using peptide exchange to examine CD45RO (Appay et al., 2008). Unexpectedly, we found an abun- antigen-specific CD4+ T cells from 26 HLA-DR4+ healthy blood dance of HIV-1-, CMV-, and HSV-specific memory-phenotype donors (Figure S2). We found that the frequency of precursor T cells in individuals who were seronegative for those viruses cells recognizing a self-antigen or an unexposed viral epitope (Figure 3A). The fraction of these memory-phenotype T cells generally ranged from one to ten cells per million CD4+ T cells ranged from 12% to 93% between different individuals, and (Figure 2A). Notably, we found that previous exposure did not the average exceeded 50% (Figures 3B and 3C). These necessarily result in a detectable expansion of cells specific to memory-phenotype T cells positively correlate with T cell that antigen, as shown by the similarity in the frequencies of frequency (Figure 3D) and have similar frequencies of central HSV-specific T cells between seropositive individuals and those memory-phenotype cells by chemokine receptor CCR7 ex- who are HSV negative. These data, as well as large interindi- pression as influenza-specific T cells (Appay et al., 2008) vidual differences in T cells recognizing the two dominant flu (Figure S3).

Immunity 38, 373–383, February 21, 2013 ª2013 Elsevier Inc. 375 Michael B.A. Oldstone papers

Ohashi et al Cell 1991 and Oldstone et al Cell 1991

Induction of diabetes by viral infection of transgenic mice expressing LCMV protein in pancreas o mice are normal, no evidence for tolerance o infect mice with LCMV: virus is rejected

→ but later mice develop diabetes Polly Matzinger papers

Anderson et al J Immunol 2001

Graft male tissue onto female mouse lacking an immune system o graft is rejected when immune system reconstitutes o graft is accepted when female receives a male thymus

→ how to interpret this? Testing Time-, Ignorance-, and Danger-Based Models of Tolerance

Colin C. Anderson,1* Joseph M. Carroll,† Stefania Gallucci,* John P. Ridge,* Allen W. Cheever,‡ and Polly Matzinger*

© 2001 Nature Publishing Group http://medicine.nature.com his report addresses two currently held ideas. The first is riod.AR Therefore,TICLES tests for a tolerogenic period must be done earlier, the long-standing concept that there is an early period in at the inception of immune competence. T ontogeny during which the developing immune system is A numberIm ofm experimentsunity or to fittingleranc thesee: Op testpos criteriaite outc haveome showns of particularly susceptible to tolerance induction, such that it becomes that MHC or multiplemi minorcroch histocompatability-mismatchedimerism from skin grafts grafts tolerant of Ags that are present at that time. The second is the given before the development of immunocompetence can be re- newer conclusion that establishment of tolerance to skin cannot jected by a newly developingCOLIN C immune. ANDERSON & systemPOLLY MATZI (10–15).NGER However, occur in adult mice because T cell trafficking into the skin occurs it could be reasonablyGhost Lab, Sectio arguedn on T-cell Toler thatance andthese Memory, La experimentsboratory of Cellular and Mo arelecular I notmmunol aogy valid, NIAID/NIH Building 4 Room 111, 9000 Rockville Pike, Bethesda, Maryland 20892-0420, USA only during the neonatal period (1). We have tested both of these test of time-basedCorresp models.ondence should be Aaddres numbersed to C.C.A.; e-m ofail: can models,[email protected] includingov concepts and our data fit with neither. Bretscher’s, are based on the hypothesis that antigenic exposure Although evidence that newly developing immune systems are Solid organ transplants contain small numbers of leukocytes that can migrate into the host and early in ontogenyestablish lon isg-las tolerogenicting microchimerism. becauseAlthough such ofmicro thechimer lowism is of frequencyten associated with of tolerance-prone dates back to the experiments in newborn mice graft acceptance and tolerance, it has been difficult to demonstrate a true causal link. Using skin effector Thfr cellsom mutan int mic thee defici peripheryent for leukocyte s atubset thiss, we fou timend that d (9,onor T 16–18).-cell chimerism Theis a from Medawar and colleagues (2), recent evidence revealed that ‘double-edged sword’ that can result in very different outcomes depending on the host’s im- number of Agsmunol inogic MHCal maturity orand multiplethe antigenic dis minorparities inv mismatchedolved. In immunologica graftslly mature couldhosts, these experiments may have been misinterpreted. Newborn mice chimerism resulted in immunity and stronger graft rejection. In immature hosts, it resulted in be greater thantoleran thece to t numberhe chimeric T c ofells, tissue-specificbut not to graft antigens n selfot expr Agsessed by to the whichchimeric cell thes. are perfectly able to respond if given Ags in appropriate doses, Clinical efforts aimed at augmenting chimerism to induce tolerance must take into account the immune systemmaturati muston state of normally host T cells, the ty establishpe of chimerism p peripheralroduced by each or tolerancegan and the antige (19,nic with appropriate adjuvants, or on appropriate APCs (3–6). Fur- 20) and consequently,disparities involved,the lest th frequencye result be increased of reje Thction r cellsather tha againstn tolerance. these Ags thermore, adult mice can be rendered tolerant if given large doses would‘Passenge ber’ leu highkocytes, m enoughigrating from thatsolid org anyan tran tolerance-pronesplants, which the H-Y ant periodigen is on b wouldoth donor s bekin and T cells), of Ag-bearing cells (4) or smaller doses of cells from which pro- often enter the host circulation. In 1957, it was suggested that chimerism-induced tolerance resulted in graft acceptance. verythese short. migrants, Therefore,rather than the tis Bretschersues themselves, m suggestedight be Howeve anr, in experimentcases in which the sk inin d whichiffered from the host by fessional APCs have been removed (7, 57, 58). One interpretation the primary immunogenic cells1. It is now accepted that graft- multiple minor or major histocompatibility antigens, the toler- femalederived a recipientsntigen-presenting wouldcells (APCs) be can givenactivate ho APCsst T ance presenting was only specific fo ther the ch singleimeric cells mi-, and the skin grafts from these results is that newborn and adult immune systems are cells, and that depletion of donor APCs leads to enhanced ac- were rejected. These data establish a direct causal link between 2,3 nor© 2001 Nature Publishing Group http://medicine.nature.com cep Agtance o H-Yf solid org beforeans . More r theecently recipient, there has also b haseen c developedhimerism and immu immunocompe-ne responses to graft antigens. They affect both tolerizable and immunizable and that the decision to respond intriguing albeit indirect evidence that ‘passenger’ leukocytes future protocols for tolerance induction in transplantation, and tencemay a (9);lso indu Brestcher’sce tolerance4. For exa predictionmple, long-term o beingrgan re- i thatndicate sucha mecha earlynism by w presentationhich viruses (such as human im- or become tolerant must therefore be governed by features other cipients often have low levels of circulating donor hematopoi- munodeficiency virus) and other antigens produced in T cells ofet aic singlecells5–8, and minor-histocompatabilitypartial depletion of these donor cells by Agmay co wouldntrol immun induceity. tolerance than age or timing of Ag expression (8). However, another inter- ‘parking’, irradiation or antibody treatment is associated with ratherincrease thand rejection the9–11. Th immunityese findings led to r thatecent su wasggestions seenPassen whenger T cells e neonatalstablish chimerism females in immunodeficient hosts pretation, recently elaborated by Bretscher (9), is that the newborn that transplantation protocols should seek to enhance donor To determine whether ‘passenger’ leukocytes from skin can mi- werechime challengedrism to improve graf witht acceptan H-Yce, and s expressingtudies in mice gra APCste into the (4).host and Therefore, establish chimerism we, we grafted skin is too old and has already passed through the tolerance-prone pe- have shown some success. If such protocols are to be used in from male B6.Ly5a mice onto T cell-deficient B6 nude female graftedhumans, h maleowever, it andis impor femaletant to unders skintand the tocondi immunodeficienttions mice, which express Ly5 femaleb, a marker tha recipi-t enables the discrimina- in which donor chimerism leads to acceptance in contrast to tion of donor and host cells. Then, 11 months later (ample time ents,those allowedin which it lead thes to rap graftsid rejection to. heal for severalfor chime months,rism to develop then), we mea reconsti-sured chimerism in the blood To elucidate the ‘rules of the game’, we studied donor (Fig. 1a and Table 1). Unexpectedly, there were no donor-de- *Ghost Lab, Section on T cell Tolerance and Memory, Laboratory of Cellular and tutedchime therism fr miceom skin g withrafts in m fetalice. We c liverhose skin cells because orac- ariv fetaled B cell thymuss and seldom a andny APC followeds; the chimeric cells were al- ceptance is difficult to achieve with skin, and it has thus be- most entirely T-cell receptor (TCR)αβ+, CD4 or CD8 T cells Molecular Immunology, National Institute of Allergy and Infectious Diseases/Na- thecom fatee the of‘gold the stand grafts.ard’ of experimental transplantation. (called ‘chimeric T cells’ here) (Fig. 1b). The lack of donor-de- tional Institutes of Health, Bethesda, MD 20892; †Genetics Institute, Andover, MA Because irradiation, antibodies and drugs can have complicat- rived B cells was not simply due to competition with resident B ‡ iDifferentng effects on the models transplantedpredict organ, we to differentok a genetic ap outcomes- cells in the nud ofe hos thists, as rec experiment.ombination activation gene knock- 01810; and Biomedical Research Institute, Rockville, MD 20852 proach, using donors of skin grafts deficient in T cells or T and B out recipient mice (lacking both T and B cells; called RAG-KO Somecells. G time-basedraft-derived chimerism models, led to either inimmu whichnity or tole tolerancer- here) gave nea isrly i baseddentical res onults (d theata not low shown). Thus, even Received for publication May 15, 2000. Accepted for publication January 4, 2001. ance depending on the immunological maturity of the host. If in completely lymphocyte-deficient hosts, chimerism derived frequencythe immune sy ofstem helper was mature T, don cellsor cell ch earlyimerism i inndu ontogeny,ced from a skin g predictraft seemed to that be mai thenly the male result of migrating T- The costs of publication of this article were defrayed in part by the payment of page strong cytotoxic T-lymphocyte (CTL) responses and increased cell populations. charges. This article must therefore be hereby marked advertisement in accordance skin,the fr becauseequency of rejec ittion preexists,. However, when shouldthe immune besystem toleratedAll mice w byith he thealthy, newlylong-standin arisingg grafts showed some de- was allowed to regenerate from an immature state, the chimeric gree of chimerism, which increased with time. Their TCR reper- with 18 U.S.C. Section 1734 solely to indicate this fact. Tc cellsells migra (9,ted to t 16–18).he host thymus, Another where they indu recentced deletion time-basedtoires varied random model,ly and were u innexp whichectedly oligoclonal, with of CTLs. In cases in which the skin graft shared antigens with many mice lacking several Vβ families (Fig. 1c). For example, 1 Address correspondence and reprint requests to Dr. Colin C. Anderson, Ghost Lab, tolerancethe chimeric tocells skin (such a iss ma basedle grafts on onto fe ama particularle mice, in one propertymouse (Fig.1c, ci ofrcles neonatal) had no CD4 ce skinlls expressing Vβ2, 4, Laboratory of Cellular and Molecular Immunology, National Institute of Allergy and 80 NATURE MEDICINE • V OLU ME 7 • N U MBER 1 • JA N UARY 2001 Infectious Diseases/National Institutes of Health, Building 4, Room 111, 9000 Rock- that allows traffic of naive T cells, would predict that both the male ville Pike, Bethesda, MD 20892-0420. E-mail address: [email protected] graft and the normal adult female skin graft should be rejected (1).

Results reported that nude mice have a small number of mature T cells, Newly generated T cells reject healed male skin grafts which may have contributed to the rejection process. Therefore, we reperformed the grafting experiments in B10-RAG recipi- In our first series of experiments, we asked whether a healed graft ents by reconstituting with male or female fetal liver cells. In carrying a single, weak, minor histocompatability Ag, would in- addition, to test whether the maturity of the thymus itself has an duce tolerance, be ignored, or induce immunity in T cell popula- influence, we thymectomized a cohort of female B10-RAG tions that were new thymic emigrants. We assessed the response of mice and reconstituted them with a day-15 fetal B6 thymus plus newly generated T cells by giving female B6 nude mice a nude an injection of day-15 fetal liver cells (Fig. 1, C and D). Once male skin graft and 9 wk later, a female thymus. These thymuses again, in both groups, we found that the female mice reconsti- were from normal B6 donors (Fig. 1A), or from B10-RAG donors tuted with female fetal tissues rejected the long-standing male (Fig. 1B) that have no functional T cells and must be repopulated grafts. Thus, T cells generated from fetal precursors, in a fetal by stem cells from the nude host before any T cell development thymus, in hosts that (unlike nudes) have no preexisting T cells, can occur. The expectation was that newly maturing T cells from went on to reject a healed preexisting male skin graft, a sur- the female thymus should slowly seed the periphery in low num- prising result that was not predicted by any current model of bers, encounter the H-Y Ag either in the male skin graft or in its immunological tolerance. The mice did not reject grafts of fe- draining lymph node (25, 30), and become tolerant. As a control to male skin, showing that some aspects of peripheral tolerance ensure that central thymic tolerance was functioning properly in were intact, and those given male thymuses did not reject male these animals, we also grafted one cohort of nude females with a skin, showing that central tolerance was also intact. Rejection male B10-RAG thymus. of preexisting grafts was not peculiar to B6 or B10 mice, as Fig. 1, A and B, shows that newly generated T cell populations C.B-17 SCID mice reconstituted with day-12 BALB/c fetal rejected the long-standing male grafts, though the grafts had healed liver cells also rejected a preexisting B10.D2 minor-histocom- for 9 wk (or 14 wk, data not shown) and appeared as healthy as the patibility different skin graft (not shown and Ref. 59). host’s own skin. The rejection did not require the transfer of mature T cells with the thymus graft because the B10-RAG thymus grafts, like the normal B6 thymuses, lead to rejection. Recipients Lack of autoimmune responses in reconstituted mice reconstituted with a male thymus did not reject the male skin (Fig. The finding that the long-standing male grafts were rejected by 1B) showing that central tolerance mechanisms were functionally newly generated recent thymic emigrants counter the expectations intact and able to induce tolerance in the newly developing T cells, of most time-based models because these models hold that a pre- but peripheral tolerance to the grafts did not occur. Thus it ap- existing Ag should induce tolerance. They counter the ignorance peared that mature naive T cells and, more surprisingly, newly model because the graft Ags are peripheral Ags that should be developing T cells, were immunized rather than tolerized by the ignored. They counter the Danger model because the grafts had peripheral Ag on the healed grafts, despite the low frequency of healed. Healthy tissues should not immunize. responding cells to H-Y and the lack of cross-reactive environ- However, one recently elaborated time-based model did offer a mental Ags to drive the response (7, 24). possible explanation for the rejection. This model suggests that In the experiments above, the bone-marrow-derived precur- tolerance to peripheral Ags cannot be established in adult mice. It Testingsors were from adult Time-, marrow, and Ignorance-, the thymuses in which they andwas proposed Danger-Based that establishment of tolerance Models to skin, for example, of developed were from neonatal animals. It remained possible requires that naive T cells traffic into the skin, which occurs only Tolerancethat peripheral tolerance depends on some property intrinsic to in the neonatal period (1). If this were the case, we would expect fetal but not adult precursor cells or thymuses. It has also been to find general signs of peripheral autoimmunity, as has sometimes

Colin C. Anderson,1* Joseph M. Carroll,† Stefania Gallucci,* John P. Ridge,* Allen W. Cheever,‡ and Polly Matzinger*

In this study, we present data showing that tolerance to Ags in the periphery is not determined by the time at which the Ag appears, or by special properties of tissues in newborn mice or newly developing immune systems. We placed male grafts onto immunoincompetent female mice, allowed the grafts to heal for up to 5 mo, and then repopulated the recipients with fetal liver stem cells. We found that the newly arising T cells were neither tolerant nor ignorant of the grafts, but promptly rejected them, though they did not reject female grafts, nor show any signs of autoimmunity. We also found that the H-Y Ag was continuously cross-presented on host APCs, that this presentation was immunogenic, not tolerogenic, and that it depended on the continuous presence of the graft. In searching for the stimulus that might activate the host APCs, we analyzed mRNA expression with a highly sensitive real-time quantitative PCR assay. By using two different “housekeeping” molecules for comparison, we analyzed the message levels for several stress and/or inflammatory molecules in the healed grafts. We found that the long-healed grafts were not equivalent to “normal” skin because the healed grafts expressed lower levels of GAPDH. Altogether, these data suggest that acceptance vs rejection of peripheral tissues is not attributable to ignorance, timing-based tolerance, or special circulation properties of naive T cells in neonatal tissues. It is more likely attributable to an aspect of the context of Ag presentation that remains toFIGURE be identified. 1. Well-healedThe skin Journal grafts induce of Immunology, immunity. A, B62001, nude female 166: mice 3663–3671. (n ϭ 6) were given a B6 nude male skin graft and 9 wk later were given a neonatal B6 female thymus graft. B, B6 nude female recipients received a B10-RAG male skin graft and, 9 wk later, a neonatal male (dashed line; n ϭ 4) or femalehis (solid report line; addressesn ϭ 4) B10-RAG two currently thymus. C,held B10-RAG ideas. female The recipients first is (n ϭ 4)riod. received Therefore, two skin tests grafts, for a B10-RAG a tolerogenic female period(dashed line) must and be done earlier, male skinthe graft long-standing (solid line). Eleven concept weeks later, that therethey were is an reconstituted early period with day-15 in B6at female the inception fetal liver of cells. immuneD, Thymectomized competence. B10-RAG female recipients receivedontogeny a B10-RAG during whichmale skin the graft developing and 7 wk later immune were reconstituted system is with day-15A B6 number female fetalof experiments liver cells and thymusfitting graft these (solid test line; criterian ϭ have shown T6). Controls (dashed line) were similarly treated except they received a male fetal thymus graft (n ϭ 1) or a female skin graft (n ϭ 2). particularly susceptible to tolerance induction, such that it becomes that MHC or multiple minor histocompatability-mismatched grafts tolerant of Ags that are present at that time. The second is the given before the development of immunocompetence can be re- newer conclusion that establishment of tolerance to skin cannot jected by a newly developing immune system (10–15). However, occur in adult mice because T cell trafficking into the skin occurs it could be reasonably argued that these experiments are not a valid only during the neonatal period (1). We have tested both of these test of time-based models. A number of models, including concepts and our data fit with neither. Bretscher’s, are based on the hypothesis that antigenic exposure Although evidence that newly developing immune systems are early in ontogeny is tolerogenic because of the low frequency of tolerance-prone dates back to the experiments in newborn mice effector Th cells in the periphery at this time (9, 16–18). The from Medawar and colleagues (2), recent evidence revealed that number of Ags in MHC or multiple minor mismatched grafts could these experiments may have been misinterpreted. Newborn mice be greater than the number of tissue-specific self Ags to which the are perfectly able to respond if given Ags in appropriate doses, immune system must normally establish peripheral tolerance (19, with appropriate adjuvants, or on appropriate APCs (3–6). Fur- 20) and consequently, the frequency of Th cells against these Ags thermore, adult mice can be rendered tolerant if given large doses would be high enough that any tolerance-prone period would be of Ag-bearing cells (4) or smaller doses of cells from which pro- very short. Therefore, Bretscher suggested an experiment in which fessional APCs have been removed (7, 57, 58). One interpretation female recipients would be given APCs presenting the single mi- from these results is that newborn and adult immune systems are nor Ag H-Y before the recipient has developed immunocompe- both tolerizable and immunizable and that the decision to respond tence (9); Brestcher’s prediction being that such early presentation or become tolerant must therefore be governed by features other of a single minor-histocompatability Ag would induce tolerance than age or timing of Ag expression (8). However, another inter- rather than the immunity that was seen when neonatal females pretation, recently elaborated by Bretscher (9), is that the newborn were challenged with H-Y expressing APCs (4). Therefore, we is too old and has already passed through the tolerance-prone pe- grafted male and female skin to immunodeficient female recipients, allowed the grafts to heal for several months, then reconsti- *Ghost Lab, Section on T cell Tolerance and Memory, Laboratory of Cellular and tuted the mice with fetal liver cells or a fetal thymus and followed Molecular Immunology, National Institute of Allergy and Infectious Diseases/Na- the fate of the grafts. tional Institutes of Health, Bethesda, MD 20892; †Genetics Institute, Andover, MA 01810; and ‡Biomedical Research Institute, Rockville, MD 20852 Different models predict different outcomes of this experiment. Received for publication May 15, 2000. Accepted for publication January 4, 2001. Some time-based models, in which tolerance is based on the low frequency of helper T cells early in ontogeny, predict that the male The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance skin, because it preexists, should be tolerated by the newly arising with 18 U.S.C. Section 1734 solely to indicate this fact. T cells (9, 16–18). Another recent time-based model, in which 1 Address correspondence and reprint requests to Dr. Colin C. Anderson, Ghost Lab, tolerance to skin is based on a particular property of neonatal skin Laboratory of Cellular and Molecular Immunology, National Institute of Allergy and Infectious Diseases/National Institutes of Health, Building 4, Room 111, 9000 Rock- that allows traffic of naive T cells, would predict that both the male ville Pike, Bethesda, MD 20892-0420. E-mail address: [email protected] graft and the normal adult female skin graft should be rejected (1).

A need therefore exists for a systematic study of tolerance within designated eGFPp), which was incorporated into an I-Ab tetramer used normal polyclonal CD4+ T cell repertoires specific for self peptides for the detection of CD4+ T cells with TCRs specific for this epitope16 with defined patterns of expression. Here we carried out such a study (Fig. 1a,b). This experimental system allowed us to determine the effects using cellular enrichment based on peptide–MHC class II tetramers. of different patterns of self-antigen expression on the same epitope- Intrathymic deletion was the mechanism used for T cells specific for specific T cell repertoire. Another advantage of this approach was that peptides from proteins expressed uniformly by thymic dendritic cells wild-type C57BL/6 (B6) mice, which express the I-Ab molecule but not and/or medullary thymic epithelial cells, whereas ‘ignorance’ was the eGFP, eYFP or CFP, could be used as controls with which to study the

mechanism- used for tolerance to peptides from cytosolic or nuclear relevant T cell repertoire in the absence of the self peptide. 187 , 3 proteins expressed only outside the thymus. Partial intrathymic We assessed tolerance in the eGFPp–I-Ab–specific CD4+ T cell .These 5 . 1 , 9 ,

deletion and impaired effector5 T cell potential but enhanced Treg cell repertoire in 14 different strains with transgenic expression of eGFP, , 1

potential among remaining 4 clones was the mechanism of tolerance for eYFP or CFP (Supplementary Table 1) by immunizing the mice with & . Other tissue- T cells was cells T 8 2 + , many peptides from tissue-restricted proteins that were also expressed eGFPp in complete Freund’s adjuvant (CFA). We obtained single- 7 ARTICLES

ARTICLES + of Department sparsely in the thymus. Thus, CD4 T cells2 tolerate self peptides with cell suspensions of secondary lymphoid organs from the immunized . This issue has come

4 b different expression1 patterns via different mechanisms. mice and stained these with the eGFPp–I-A tetramer labeled with T cells T 12– allophycocyanin, then with magnetic beads coated with antibody Diabetes Center, University of University Center, Diabetes

+ b 3 , Whitney E Purtha E Whitney , A need therefore exists for a systematic study of tolerance within designated eGFPp), which was incorporated into an I-A tetramer used 2 RESULTS to allophycocyanin. The suspensions underwent enrichment for + + 16

normal polyclonal CD4 T cell repertoiresrevealmicecansuchAnalysis of. specific for self peptides for+ the detection of CD4 T cells with TCRs specific for this epitope

A0 transgenic model of CD4 T cell tolerance to self tetramer-bound cells on magnetized columns, then were eluted and 1 , with defined patterns of expression. Here we8 carried out such a study (Fig. 1a,b). This experimental system allowed us to determine the effects , or may becomemay orlackanergic a to due 7 2 We used enhanced green fluorescent protein (eGFP), enhanced yellow flu- stained with antibodies designed to identify T cells. We assessed the using cellular enrichment based on peptide–MHC class II tetramers. of different patterns of self-antigen expression on the same epitope- b + , Kristin A Hogquist A Kristin , orescent protein (eYFP), or cyan fluorescent protein (CFP) as model self differentiation of eGFPp–I-A –specific CD4 T cells into the TH1, 3 Intrathymic deletion was the mechanism cells used for T cells specific for specific TARTICLES cell repertoire. Another advantage of this approach was that antigens in mouse strains with transgenic expression of these proteins and T 17, follicular helper T cell (T cell) and T cell subsets by intra- peptides from proteins expressedreg uniformly by thymic dendritic cells wild-type C57BL/6 (B6) mice, which express the HI-Ab molecule but not FH reg 7 T cells specific for green fluorescentgreen for specificcells T b + the MHC class II molecule I-A . These fluorescent proteins all contain an cellular staining of the lineage-defining transcription factors T-bet,

, You Jeong Lee Jeong You , and/or medullary thymic epithelial cells, whereas ‘ignorance’ was the eGFP, eYFP or CFP, could be used as controls with which to study the 1 I-Ab-binding peptide first identified in eGFP (HDFFKSAMPEGYVQE; RORGt, Bcl-6 and Foxp3, respectively (Fig. 1c).

mechanism used for tolerance T cells that encounter and recognize these epitopes to peptides from cytosolic or nuclear relevant T cell repertoire in the absence of the self peptide. proteins expressed only outside+ the thymus. Partial intrathymic We assessed tolerance in the eGFPp–I-Ab–specific CD4+ T cell

deletion and impaired effector T cell potential but enhanced6. Further, it is clear Tthat an abundance of T cellsreg with the cell repertoire in 14 different strains with transgenic expression of eGFP, 1 a 101 b c potential among remaining clones was the mechanism of tolerance*** for*** eYFP ****** or CFP (Supplementary*** NS Table 1) by immunizing the mice with doi:10.1038/ni.332 5 5 13.1 1.1 5 10 10 10 22.0 , Mark S Anderson S Mark , TH17 Treg many peptides from tissue-restricted proteinsMuch of the evidence cited above has been provided by studies of that were also expressed eGFPp in completeNS Freund’s adjuvant (CFA). We56 obtained single- 5 5 ****** *** + *** *** ** 4 4 T 1 Tolerance is established+ in polyclonal10 CD4 T cells 10 10 104 H mice with transgenic expression of monoclonalspecifictransgenicselfTCRswithmiceexpressionof for epitopes II class peptide–MHC ofcostimulation derived from the innate immune system secondarylymphoidorgans, specific cellsT would beretained thein repertoire in a fully responsive state of ignorance restrictedproteins that aresecreted arepresentor indying cells may reach secondary lymphoid organs, where they would be processed by antigen-presenting cells toCD4 epitopes. form self peptide–MHC class II to the fore in several studies as showingprominent a thatmechanism clonalin polyclonaldeletion T cell is repertoires,beennot reported asfor certain monoclonal has T cell populations tolerance mechanisms because of the large numbers of T cells avail Because limitations. several has approach this However, study. for able mousewithtransgenica expressionTCRsinglecontains froma TCR a larger repertoire, it can reveal only those thatapply to TCRs tolerancewith that affinity for a self mechanismspeptide–MHC class II epitope same TCR can create abnormal outcomes may differentiate into T into differentiate may studies, however, have been limited patternsofexpression. to epitopes with ubiquitous sparsely in the thymus. Thus, CD4 T cells tolerate self peptides with cell suspensions of secondary). lymphoid organs from the immunized 57.0 3 WT 10 WT 3 3 T

T cells b 10 FH

by distinct mechanisms, according to self-peptide 10 + . - - - different expression- patterns via different mechanisms. mice and stained these with the eGFPp–I-A tetramer labeled with

1 2 10 10 t 0 0 expression patterns allophycocyanin, then with magnetic0 beads coated with antibody CD4 CD44 85.8 +

RESULTS to allophycocyanin. The suspensions underwent enrichment forROR T-bet 1 1,6 3 1 2 3 3 4 5 3 3 4 5 3 4 5 . Some. T Deepali Malhotra , Jonathan +L Linehan , Thamotharampillai Dileepan Tet 10 , You Jeong Lee , Whitney E Purtha , 0 10 10 10 –10 0 10 10 10 0 10 10 10 6 A transgenic model of CD4 hi Jennifer V Lu3, Ryan W Nelson 1T, Brian cell T Fifetolerance4, Harry T Orrto 5self, Mark S Anderson3, Kristintetramer-bound A Hogquist2 & cells on magnetized columns, then were eluted and , Harry T Orr T Harry , 2– Tet Foxp3 Bcl-6 4 We used enhancedMarc K Jenkins green1 fluorescent protein (eGFP), enhanced yellow flu- stained with antibodies designed to identify T cells. We assessed the 2 [email protected] CD44 T cells have revealed several mechanisms of self-tolerance; however, which self-tolerance;however, of mechanisms severalrevealed have cells T 10 b + + , Thamotharampillai Dileepan Thamotharampillai , orescent Studiesprotein of repertoires (eYFP), of ormouse cyan monoclonal fluorescent CD4+ T cells protein have revealed (CFP) several as mechanisms model self of self-tolerance; differentiation however, which of eGFPp–I-A –specific2016 FEBRUARY CD4 T cells into the TH1, 6 ,

+ Minneapolis,School,Medical Minnesota of University Immunology, for CenterMedicine, of Department e

1 mechanisms operate in normal repertoires is unclear. Here we studied polyclonal CD4 T cells specific for green fluorescent 4 antigens in mouse strains.Low-affinity inter with transgenic expression of these proteins and T 17, follicular helper T cell (T cell) and T cell subsets by intra- 9.8 2.3 protein expressed1 in various organs, which allowed us to determine the effects of 1specific expression patternsH on the same FH reg 33 18.8 10 TH17 Treg ) thymic epithelial cells cells epithelial thymic ) epitope-specific T cells. Peptides T cells in the thymus b presented uniformly by thymic antigen-presenting cells were tolerated by clonal deletion, T 1

+ H the MHC class II molecule I-A+ . These fluorescent proteins all contain an cellular staining of the lineage-defining transcription factors T-bet, whereas peptides excluded from the thymus were ignored. Peptides with limited thymic expressionCFP induced partial clonal WT Cluster 1: Cluster 1: b eGFPeGFPeYFPeGFP eGFPeGFPeGFPG eGFPeYFPeGFPeGFPeGFPeGFP 53.1 I-A -bindingdeletion peptide and impaired first effector identified T cell potential in eGFPbut enhanced (HDFFKSAMPEGYVQE; regulatory T cell potential. These mechanisms ROR weret, alsoBcl-6 active and for Foxp3, respectivelyeGFP (Fig. 1c). eGFP TFH AireCdh1 Ins2 + Ins1 Ins1 Ins1 , Brian T Fife T Brian , T cell populations specific for endogenously expressed self antigens. Thus, the immunotoleranceUBC Itgax ofChat polyclonal CD4 T cells was ACTB Pcp2 Foxp3Foxd1 2 NUMBER 1 Myh11 CD207Zfp423

maintained by distinct mechanisms, according to self-peptide expression patterns. cells in the thymus might not

+ 87.9 Self-reactive CD46 + T cells are generated in the normal course of rear- secondary lymphoid organs, specific T cells would be retained in the 10 1 17 rangementa of the T cell antigen receptor (TCR)1. Low-affinity interNS - repertoirebd in a fullyT reg TresponsiveFH T H T H state of ignorance7,8. cOther tissue- *** *** ****** *** WT 1 1.5 57.5 actions between the TCR and ligands composed of self peptide and restricted proteins that 5are secretedeGFP or are present in dying cells may 5 13.1 1.1 5 11.2 10 10 10 22.0 TH17 Treg major histocompatibility (MHC) class II (self peptide–MHC class II) reach secondary lymphoid organs,Ins1 where they would be processed TH17 Treg 4.5 5 ****** *** *** *** ** NS eGFP56 T 1 TH1 10 + 1 104 Foxd1 4 4 H

are essential for the positive selection of CD4 T cells in the thymus . by antigen-presenting cells to form self peptide–MHC class II 10 VOLUME 17 eGFP Cluster 2: 10 Cluster 2: 39.4 + 57.0 However, high-affinity interactions between the TCR and self epitopes. CD4 T cells 3that encounterIns2 and recognize2 these epitopes WT eYFP WT 3 eGFP eGFP TFH , Jonathan L Linehan L Jonathan , 10 3 T T cells CD2072 10 Ins2 10 FH Ins2

1 peptide–MHC class II can cause autoimmunity. Therefore, tolerance may differentiate into T cells or may become anergic due to a lack + reg 4 cell) if its TCR binds strongly to a self peptide– 2 eGFP mechanisms10 have evolved to limit T cells with TCRs with high affinity of costimulation derived10 fromMyh11 the innate immune system1,9. t eGFP 0 reg 1 0 for self peptide–MHCrear normalthecoursegeneratedof in cells are T class II ligands. Much of the evidence0 citedFoxp3 above has been provided by studies of CD4

, Ryan W Nelson W Ryan , eGFP CD44 + 85.8 41.0

+ Zfp423 3 Different tolerance mechanisms may be engaged depending on mice with transgenic expression of monoclonal TCRs specific for self eGFP ROR T-bet 3 Pcp27,8,10 Minneapolis,School,Medical Minnesota of Neuroscience, University Translational for Institute Pathology, and MedicineLaboratory of Department Bethesda,Health, of InstitutesNationalDiseases, Infectious and Allergy of InstituteNationalDiseases, Parasitic of Laboratoryaddress:Present where a self peptide–MHC class II epitope is displayed. A develop- peptide–MHC class II epitopes . Analysis3 4 of5 such mice can reveal5 6 3 3 4 5 3 4 5

Tet 10 eGFP 0UBC 10 10 10 3 –10 0 10 10 10 0 10 10 10 ing thymocytehi may die by apoptosis, or it may differentiate into a tolerance mechanisms because of theeYFP large numbers of T cells avail- 105 5 7.1 64.3 5 Itgax Foxp3 Bcl-6 10 T 17 T 10 0 regulatory T cell (Treg cell) if its TCR binds strongly to a self peptide– able for study. However, thisTet approacheGFP has several limitations. Because 0.06 H reg Chat 4 4 4 TH1 MHC class II2 epitope displayed on particular antigen-presenting a mouse with transgenic TCR expressionCFP contains a single TCR from 10 10 10 CD44 10 Cdh1 Cluster 3: Cluster 3: 0 cells1,2. Epitopes in this class may be derived from proteins expressed a larger repertoire, it can reveal only those tolerance mechanisms 14 strains expressedderivedproteinsbefrom thismayclass in Epitopes . eGFP 3 3 3 T + 2 Aire eGFP 10 eGFP FH or taken up by MHC class II–positive (MHCII ) thymic epithelial cells that apply, to TCRs with that affinity eGFPfor a self peptide–MHC class II UBC UBC 10 10 e 1 ACTB or dendritic cells. However, MHCII+ cells in the thymus might not epitope11. Further, it is clear that an abundance of T cells with the 9.8 2.3 t 0 1 Different tolerance mechanisms may be engaged depending on 33 T 17 T 18.8 display epitopes10 from all host proteins. Thus, T cells expressing TCRs same TCR can create abnormal outcomes12–14. This issue has come H reg 0 0 Department of Microbiology and Immunology, Center for Immunology, University of Minnesota Medical School, Minneapolis, Minnesota, USA. Minnesota, Minneapolis, School, Medical Minnesota of University Immunology, for Center Immunology, and Microbiology of Department T 1 display epitopes from all host proteins. Thus, T cells expressing TCRsexpressingcells T Thus, proteins. hostall from epitopes display with high affinity for certain self epitopes may not be deleted in the thymusandthusexitmay secondaryto lymphoid organs Minnesota, USA. Minnesota, rangementof the T cell antigen receptor (TCR) mechanisms operate in normal repertoires is unclear. Here we studied polyclonal CD4 polyclonal studied we Here unclear. is repertoires normal in operate mechanisms USA. Minnesota, ( M.K.J. to addressed be should Correspondence USA. Maryland, 2016; January 4 online published 2015; October 21 accepted 2015; August 21 Received Tolerance is established in polyclonal CD4 Malhotra Deepali CD4 monoclonal mouse of repertoires of Studies CD4 Self-reactive 1 IMMUNOLOGYNATURE protein expressed in various organs, which allowed us to determine the effects of specific expression patterns on the same the on patternsexpressionspecific of effects the determine to us allowed which organs,various in expressedprotein deletion,clonal by toleratedwere antigen-presenting cellsthymic by uniformlypresentedPeptides cells. epitope-specificT clonal partialinduced expression thymic limited with Peptidesignored. were thymus the from excludedpeptides whereas for active also weremechanisms Thesepotential. cell T regulatoryenhanced but potentialcell T effectorimpaired and deletion CD4 polyclonal of immunotolerance the Thus, antigens. self expressed endogenously for specific populations cell T MHC class II epitope displayed on cells particular antigen-presenting (MHCII II–positive class MHC by up taken or Laboratory Medicine and Pathology, Center for Immunology, University of Minnesota Medical School, Minneapolis, Minnesota, USA. Minnesota, Minneapolis, School, Medical Minnesota of University Immunology, for Center Pathology, and Medicine Laboratory California San Francisco, San Francisco, California, USA. California, Francisco, San Francisco, San California by distinct mechanisms, according to self-peptideto accordingmechanisms,distinct by patterns expression However, high-affinity interactions peptide–MHCTherefore,classcausecanIIautoimmunity.tolerance between the affinity high with TCRs with cells T TCRlimit to evolved have mechanisms and self forself peptide–MHC class II ligands. where a self peptide–MHC class II epitope is displayed. A develop ing thymocyte may die by apoptosis, or it may differentiateregulatorycellT(T into a or dendritic cells. However, MHCII maintained by distinct mechanisms, according to self-peptide expression patterns. expression self-peptide to according mechanisms, distinct by maintained cells in this class may be specific for epitopes from proteins that are expressedlocationssecretedinpoorarewithdrain lymphatic notor age.Since these proteins would notreach antigen-presenting cells in Jennifer V Lu MarcKJenkins actions between the TCR and ligands composed of self peptide and histocompatibilitymajor (self(MHC)classIIpeptide–MHC classII) essentialpositivetheselectionare forCD4 of H 28.6 with high affinity for certain self epitopes may not be deleted in the to the fore in several studies showing that clonal deletion is not 10–3

CFP WT 0 CD44 ROR T-bet eGFPeGFPeYFPeGFP eGFPeGFPeGFPeGFPeYFPeGFPeGFP2–eGFP6 eGFP Cluster 1: 96 Cluster 1: 53.1 thymus and thus may exit to secondary lymphoid organs . Some T as prominent a mechanism192288 in383 polyclonal479575670 T cell repertoires, as has 3 4 5 T 3 4 5 3 4 5 eGFP 1,8383,0054,1735,3406,5087,6764,5,15eGFP 0 10 10 10 FH 0 10 10 10 0 10 10 10 cells in this classUBC mayItgax beAire specificCdh1Chat for epitopesIns2 from proteinsIns1 that are beenIns1 reported for certain monoclonal T cell populationsIns1 . These ACTB Pcp2 not secreted or are expressed in locationsMyh11 withCD207 poorZfp423Foxp3 lymphaticFoxd1 drain- studies, however, have beenCD44 limitedhi toTet epitopes+ cells with ubiquitous Tet Foxp3 Bcl-6 age. Since these proteins would not reach antigen-presenting cells in patterns of expression. 87.9

1 17 reg FH H H hi + b + d T T T T Figure 1 Three main patterns of tolerance to self antigens. (a) Total CD44 Tet (eGFPp–I-A -tetramer (Tet)-binding) CD4 T cells from pooled 1Department of Microbiology andWT Immunology, Center for Immunology,1 University of Minnesota Medical School, Minneapolis, Minnesota, USA. 2Department of 1.5 57.5 3 11.2 Laboratory Medicine and Pathology, CentereGFP for Immunology, University of Minnesota Medical School, Minneapolis, Minnesota, USA. Diabetes Center, University of TH17 Treg Ins1 4 secondary lymphoid organs of wild-type4.5 (WT) and transgenic mice 13–14 d after immunization with 100 Mg eGFPp emulsified in CFA. Each symbol California San Francisco, San Francisco, eGFPCalifornia, USA. Department of Medicine, Center for Immunology, University of Minnesota Medical School, Minneapolis, TH1 Minnesota, USA. 5Department ofFoxd1 Laboratory Medicine and Pathology, Institute for Translational Neuroscience, University of Minnesota Medical School, Minneapolis, eGFP represents Clusteran individual 2: mouse (colors matchCluster those 2: of clusters identified in d); small horizontal39.4 lines indicate the geometric mean. (b) Flow cytometry Minnesota, USA. 6Present address:Ins2 Laboratory of Parasitic2 Diseases, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, eGFP eGFP T Maryland, USA. CorrespondenceCD207 should beeYFP addressed to M.K.J. ([email protected]). Ins2 Ins2 + FH eGFP analyzing CD44 and tetramer (Tet) staining in tetramer-enriched CD4 T cells from pooled secondary lymphoid organs of wild-type (WT) mice primed Received 21 August 2015; acceptedMyh11 21 October 2015; published online 4 January 2016; doi:10.1038/ni.3327 + − − + Foxp3eGFP 14 d earlier with eGFPp in CFA. (c) Expression of RORGt and Foxp3 by Tet cells (left) and of T-bet and Bcl-6 by Foxp3 RORGt Tet cells (right) among eGFP 41.0 Zfp423 hi + + hi + + eGFP the CD44 Tet CD4 T cells identified in b. (d) Unsupervised hierarchical clustering of CD44 Tet CD4 T cells (geometric mean cell number and NATURE IMMUNOLOGY VOLUMEPcp2 17 NUMBER 2 FEBRUARY 2016 187 UBCeGFP 3 eYFP phenotype) in 15 mouse5 strains (right margin), grouped5 into7.1 three64.3 main clusters5 (far left and right). (e) Flow cytometry analyzing CD44 and tetramer Itgax 10 10 T 17 T 10 0 eGFP 0.06+ H reg T 1 Chat staining of tetramer-enriched4 CD4 T cells from pooled4 secondary lymphoid4 organsH of three eGFP-expressing mouse strains in clusters 1–3 identified CFP 10 10 10 Cdh1 Cluster 3: Cluster 3: + hi + 0 eGFP in (d) (left), and transcription3 factor expression as in c by the CD4 CD44 Tet cells identified at left (right). Numbers adjacent to outlined areas Aire eGFP 10 eGFP 3 3 TFH eGFP UBC 10 10 ACTB UBC (b,c,e) indicate percent cells in each. NS, not significantt (P > 0.05); *P < 0.05, **P < 0.005 and ***P < 0.0001, versus wild-type B6 mice (one-way 0 0 0 analysis of variance (ANOVA) of log -transformed values). 28.6Data are pooled from 12 independent experiments with 3–23 mice per group (a,d) or are 10–3 10

0 CD44 ROR T-bet 96 192288383479575670 3 4 5 3 4 5 3 4 5 1,8383,0054,1735,3406,5087,676 representative of 12 experiments0 10 10 (b10,c) or eight experiments0 (10e). 10 10 0 10 10 10 CD44hi Tet+ cells Tet Foxp3 Bcl-6

hi + b + Figure 1 Three main patterns of tolerance to self188 antigens. (a) Total CD44 Tet (eGFPp–I-A -tetramer (Tet)-binding) CD4 T cellsVOLUME from pooled 17 NUMBER 2 FEBRUARY 2016 NATURE IMMUNOLOGY secondary lymphoid organs of wild-type (WT) and transgenic mice 13–14 d after immunization with 100 Mg eGFPp emulsified in CFA. Each symbol represents an individual mouse (colors match those of clusters identified in d); small horizontal lines indicate the geometric mean. (b) Flow cytometry analyzing CD44 and tetramer (Tet) staining in tetramer-enriched CD4+ T cells from pooled secondary lymphoid organs of wild-type (WT) mice primed 14 d earlier with eGFPp in CFA. (c) Expression of RORGt and Foxp3 by Tet+ cells (left) and of T-bet and Bcl-6 by Foxp3−RORGt−Tet+ cells (right) among the CD44hiTet+ CD4+ T cells identified in b. (d) Unsupervised hierarchical clustering of CD44hiTet+ CD4+ T cells (geometric mean cell number and phenotype) in 15 mouse strains (right margin), grouped into three main clusters (far left and right). (e) Flow cytometry analyzing CD44 and tetramer staining of tetramer-enriched CD4+ T cells from pooled secondary lymphoid organs of three eGFP-expressing mouse strains in clusters 1–3 identified in (d) (left), and transcription factor expression as in c by the CD4+CD44hiTet+ cells identified at left (right). Numbers adjacent to outlined areas (b,c,e) indicate percent cells in each. NS, not significant (P > 0.05); *P < 0.05, **P < 0.005 and ***P < 0.0001, versus wild-type B6 mice (one-way analysis of variance (ANOVA) of log10-transformed values). Data are pooled from 12 independent experiments with 3–23 mice per group (a,d) or are representative of 12 experiments (b,c) or eight experiments (e).

188 VOLUME 17 NUMBER 2 FEBRUARY 2016 NATURE IMMUNOLOGY A simple mathematical model A ﬁrst model (DeBoer.prsl93)

S number of self epitopes evoking tolerance (105) R0 potential repertoire (before tolerance) R “functional”A repertoire ﬁrst model after(DeBoer.prsl93) tolerance (> 108) p lymphocyte recognition probability (precursor frequency)

S number of self epitopes evoking tolerance (105) Probability of responding to foreign epitope: R0 potential repertoireA simple (before mathematical tolerance) model S number of self epitopes evoking tolerance (10R5) [Burroughs.i04] 8 “functional” repertoireP =1 after(1 tolerancep) 12 ( 10 ) R R0 Thepotential “optimal”repertoirei (before tolerance) speciﬁcity (huge: 10 cells)> R “functional” repertoire after tolerance (> 108) [Qi.pnas14] wherep lymphocyte recognitionp recognition probability probability (precursor freq. 10-5) (precursor[Blattman.jem02, Su.i13] frequency) Size of functional repertoire: We had S R = R0(1 p) S Probability of responding toR foreign0(1 p) epitope: Pi =1 (1 p) Probability of responding to a foreign epitope: R Pi =1 (1 p) AccommodatesThis probability two of problems:responding,Pi, recognition has its maximum and at deletion. Taking the derivative of Pi to p gives the optimum where 1 5 13 p = 10 S ⇥ R = R (1 p)S 1993 Soc Royal Proc Perelson & Boer De 0 Thus speciﬁcity seems determined by number of self epitopes (DeBoer.prsl93,Whitaker.jtb93,Nemazee.it96,Borghans.ji99). Accommodates two problems: recognition and deletion.

14 13 A simple mathematical model

1 Too Too much Pi specific deletion 0.8 1/S

0.6

0.4

0.2

0 -14 -12 -10 -8 -6 -4 -2 0 SPECIFIC DEGENERATE Cross-reactivity (log(p))

Wide range of cross-reactivities for which functional repertoire is complete. True optimum at p =1/S.

15 Lymphocytes are speciﬁc AND recognize many peptides

Repertoire: 108 clones Peptides: 1013

5 p = 10

5 13 8 10 10 = 10 peptides per clone ⇥

Because there are many more peptides (2010 ≃ 1013) than clones, we need sufﬁcient cross-reactivity [Mason Imm. Today 98].

But each peptide is recognized by very few clones: p ≃ 10−5 1 1 1

1 1 1

1 0.8 1 0.8 1 0.8

0.8 0.8 0.8

0.8 0.6 0.8 0.6 0.8 0.6 1 1 1 Pi Pi Pi 0.6 0.6 0.6 0.6 0.6 0.6 Pi 0.8 0.4 Pi0.8 0.4 0.8 Pi 0.4 Pi Pi Pi

0.4 0.4 0.4 0.4 0.4 0.4 0.6 0.6 0.6 0.2 0.2 0.2

Pi Pi Pi 0.2 0.2 0.2 0.2 0.2 0.2 0.4 0.4 0.4

0 –10 –8 –6 –4 –2 0 –10 –8 –6 –4 –2 0 –10 –8 –6 –4 –2

0 log[p] 0 log[p] 0 log[p] 0 –10 –8 0.2 –6–10 –4–8 –2 –6 0 –4 –10 –2 –80.2 –6 –10 –4 –8–2 –60 –4–10 –2–80.2 –6 –4–10 –2 –8 –6 –4 –2 log[p] log[p] log[p] log[p] log[p] log[p]

0Figure 1: The probability of mounting0 an immune response Pi from0 Eq. (2) as a function of the specificity p of the –10 –8 –6 –4 –2 5 –10 –8 9–6 –4 –2 –10 –8 –6 –4 –2 Figure 1: The probabilitylymphocytes. of mounting Parameters an immuneS response= 10 Pandi fromR0 Eq.= 10 (2). as Panel a function (b) depicts of the specificity the eectp ofof decreasing the the initial repertoire size Figure 1: The probabilitylog[p] of mounting an immune responselog[p] Pi from Eq. (2) as a functionlog[p] of the specificity p of the 5 9 95 8 97 lymphocytes. ParametersfromS R=0 10= 10and, RR00== 10 10., Panel to R0 (b)= depicts 10 . Panel the e (c)ect depicts of decreasing the e theect initial of incomplete repertoire size tolerance induction, i.e., f = 1 and lymphocytes.9 8 Parameters7 S = 10 and R0 = 10 . Panel (b) depicts the eect of decreasing the initial repertoire size from R0 = 10 , R0 = 10 ,9 to R0 = 10 8. Panel (c) depicts7 the eect of incomplete tolerance induction, i.e., f = 1 and from R0 =f 10=0,.8R in0 = Eq. 10 (6)., to R0 = 10 . Panel (c) depicts the eect of incomplete tolerance induction, i.e., f = 1 and f =0.8 in Eq. (6).Figure 1: The probability of mounting an immune response Pi from Eq. (2) as a function of the specificity p of the f =0.8 in Eq. (6). 5 9 lymphocytes. Parameters S = 10 and R0 = 10 . Panel (b) depicts the eect of decreasing the initial repertoire size 9 8 7 from R0 = 10 , R0 = 10 , to R0 = 10 . Panel (c) depicts the eect of incomplete tolerance induction, i.e., f = 1 and ⇧pPi of Eq. (3)f and=0.8 solving⇧p inP Eq.i of⇧ (6).p Eq.Pi = (3) 0 to and find solving that that⇧pP thei = maximum 0 to find is that at Pi that=1/S the. This maximum optimum is suggests at Pi =1/S. This optimum suggests that the lymphocyte⇧pPi of Eq.that specificity (3) the and lymphocyte is solving largely⇧ determinedpP specificityi = 0 to by find is the largely that number that determined of the self maximum epitopes by the the is number immune at Pi =1 system of/S self. This has epitopes optimum the immune suggests system has to be tolerantthat to. the Thus,to lymphocyte be the tolerant specificity specificity to. is Thus, not determined is the largely specificity determined by the recognitionis not by determined the of number pathogens, by of the self but recognition by epitopes the demand the of immune pathogens, system but by has the demand to remain tolerantto be⇧pP toleranti toof a Eq. large to. (3) number and Thus, solving theof self specificity⇧pP epitopes.i = 0 to is find Once not that lymphocytes determined that the maximum by are the specific recognition is at thePi repertoire=1 of/S. pathogens, This has optimumto be but suggests by the demand that theto lymphocyte remain tolerant specificity to a is large largely number determined of self by epitopes. the number Once of self lymphocytes epitopes the immune are specific system the has repertoire has to be su⇤cientlyto diverse remain tosu guarantee tolerant⇤ciently to recognition diverse a large to number of guarantee foreign of epitopes self recognition epitopes. (Fig. 1b). of Once foreign lymphocytes epitopes (Fig. are specific 1b). the repertoire has to be su⇤tociently be tolerant diverse to. to Thus, guarantee the specificity recognition is not of determined foreign epitopes by the recognition (Fig. 1b). of pathogens, but by the demand to remain tolerant to a large number of self epitopes. Once lymphocytes are specific the repertoire has to be su⇤ciently diverse to guarantee recognition of foreign epitopes (Fig. 1b). 1.1 Incomplete tolerance 1.1 Incomplete1.1 Incomplete tolerance tolerance Although there1.1 is promiscuous Incomplete expression tolerance of self antigens in the thymus, it remains unlikely that self tolerance is complete.Although HealthyAlthough there individuals is promiscuous there do harbor is promiscuous lymphocytes expression expression of that self can antigens recognize of self in antigens the self thymus, epitopes. in the it To thymus, remains study how unlikelyit remains the that unlikely self tolerance that self tolerance results changeis complete. whenAlthoughis tolerance complete. there Healthy is is promiscuous incomplete individuals Healthy we expression individuals define do harbor a of new self do lymphocytes parameter antigensharbor lymphocytes inf thefor that thymus, the can fraction recognize thatit remains of can self self unlikelyrecognize epitopes epitopes. that that self self To epitopes. tolerance study how To studythe how the manage to induce tolerance. For f = 1 the new model should be identical to the previous one. For a foreign resultsis complete. changeresults when Healthy change tolerance individuals when tolerance is incomplete do harbor is incomplete lymphocytes we define a we that new define can parameter recognize a new parameterf selffor epitopes. the fractionf for To the study of fraction self how epitopes the of self that epitopes that epitope we now require that it is recognized, but that none of the clonotypes recognizing the foreign epitope manageresults tomanage change induce when to tolerance. induce tolerance tolerance. For is incompletef = 1 For the wef new= define 1 model the a new new should parameter model be should identicalf for the be to fraction identical the previous of self to the epitopes one. previous For that a one. foreign For a foreign also recognize onemanage of the to (1 inducef)S tolerance.self epitopes For f that= 1 fail the to new induce model tolerance. shouldIncomplete be Otherwise identical tolerance the to the clone previous will be one. held For a foreign responsibleepitope forepitope auto-immunity. weepitope we now now require we require Following now that that require Borghans it it is is that recognized,et it al. is[4] recognized, but we but thatlet that nonebe none but the of fraction the that of the clonotypes none clonotypesof clonotypes of recognizing the clonotypes recognizing recognizing the foreign recognizing the foreign epitope the epitope foreign epitope also recognizealso recognize one of the one (1 off the)S self (1 epitopesf)fS S numberself that epitopes of self fail epitopes to that induce evoking fail tolerance. to tolerance induce (0 tolerance.< Otherwise f <= 1) the Otherwise clone will the be clone held will be held at least one ignoredalso recognize self epitope, one i.e., of the (1 f)S self epitopesR0 potential that failrepertoire, to induce R “functional” tolerance. repertoire, Otherwise p precursor the clone freq. will be held (1 f)S responsibleresponsibleresponsible for for auto-immunity. auto-immunity. for auto-immunity. Following Following=1 (1 Borghans Following Borghansp) et. Borghanset al. al.[4][4] we we letet let al.be[4] thebe we fraction the let fraction ofbe clonotypes the of fraction clonotypes(4) recognizing of clonotypes recognizing recognizing at leastat least oneat one least ignored ignored one self ignored self epitope, epitope, self i.e., i.e., epitope,Fraction i.e., of clones recognizing at least one ignored self epitope: The chance that the system remains tolerant when stimulated with a foreign(1 f(1)S f epitope)S (1 is thef)S probability that =1=1(1(1p)=1p) .(1 . p) . (4) (4) (4) none of the clones in the functional repertoire R will respond (with chance p) and is potentially auto-reactive (with chanceThe),The chance i.e., chance that that the the system system remains remains tolerant tolerantSize whenof when functional stimulated stimulated repertoire: with with a foreign a foreign epitope epitope is the probabilityis the probability that that The chance that the systemR remains tolerantfS when stimulated with a foreign epitope is the probability that nonenone of the of the clones clones inP in thet the= (1 functional functionalp) repertoirewhere repertoireRR=RwillRwill0(1 respond respondp) (with. (with chance chancep) andp) is and potentially is potentially(5) auto-reactive auto-reactive (withnone chance of), the i.e., clones in the functional repertoire R will respond (with chance p) and is potentially auto-reactive Now the chance(with of chance a “successful), i.e., immune response” is theProbability probabilityR to remain that tolerant: the systemfS remains tolerant and (with chance ), i.e., Pt = (1 p) R where R = R0(1 p) . fS (5) responds to the foreign epitope, which is the chancePt = to (1 remain p) tolerantwhere minusR R the= R chance0(1 top) not. respondfS at all: (5) Pt = (1 p) where R = R0(1 p) . (5) Now the chance of a “successful immune response” is the probability that the system remains tolerant and Now the chance of a “successful immune response”R is the probability that the system remains tolerant and responds to the foreign epitope,P whichs = Pt is the(1Probability chancep) , to of remaina successful tolerant immune minus response: the chance to not(6) respond at all: respondsNow to the the foreign chance epitope, of a “successful which is the immune chance to response” remain tolerant is the probabilityminus the chance that the to not system respond remains at all: tolerant and responds to the foreign epitope, whichP = isP the(1 chancep)R , to remain tolerant minus the chance(6) to not respond at all: where the functional repertoire R is given in Eq. (5). To studys howt incomplete tolerance aects the results we P = P (1 p)R , Borghans et al JI 1999 (6) plot Eq. (6) for f =0.8 and f = 1 in Fig. 1c. s t R where the functional repertoire R is given in Eq. (5). To studyPs = howPt incomplete(1 p) tolerance, aects the results we (6) plot Eq. (6) for f =0.8 and f = 1 in Fig. 1c. Fig. 1c demonstrateswhere the that functional the eect repertoire of incompleteR is tolerance given in is Eq. enormous. (5). To The study region how of incomplete specificity values tolerance where aects the results we the chanceplot of a Eq. successful (6)where for response thef =0 functional.8P andapproachesf = repertoire 1 in one Fig. isR 1c. muchis given narrower. in Eq. Moreover (5). To study the optimum how incomplete has shifted tolerance aects the results we Fig. 1c demonstrates thats the eect of incomplete tolerance is enormous. The region of specificity values where leftwards, i.e., towardsplot a specificity Eq. (6) for muchf =0 smaller.8 and thanf =p 1=1 in/S Fig.. Thus 1c. the p =1/S estimate [8] is an upper the chance of a successful response Ps approaches one is much narrower. Moreover the optimum has shifted bound for theFig. lymphocyteleftwards, 1c demonstrates i.e., crossreactivity: towards that a the specificity when eect the of much incomplete initial smaller repertoire than tolerancep is=1 su/S⇤ isciently. enormous. Thus large the p the The=1 immune/S regionestimate systemof specificity [8] is an upper values where operates eventhe betterbound chanceFig. when for of the 1c a lymphocytes successful demonstrates lymphocyte response are crossreactivity: that more theP specifics approaches eect when [4]. of the The incomplete initialone conclusion is repertoire much tolerance remains narrower. is su is⇤ that enormous.ciently Moreover lymphocytes large The the the areimmune region optimum of system specificity has shifted values where specific to avoid auto-immunity, and not to recognize many pathogens. leftwards,operatesthe i.e., even chance towards better of when a a specificity successful lymphocytes much response are smaller morePs specificapproaches than p [4].=1 The one/S. conclusion is Thus much the remains narrower.p =1/S thatestimate Moreover lymphocytes [8] the is are optimum an upper has shifted boundspecific forleftwards, theto avoid lymphocyte auto-immunity, i.e., towards crossreactivity: and a specificity not to recognize when much the many smaller initial pathogens. repertoire than p =1 is/S su.⇤ Thusciently the largep =1 the/S immuneestimate system [8] is an upper operatesbound even better for the when lymphocyte lymphocytes crossreactivity: are more specific when [4]. the The initial conclusion repertoire remains is su⇤ciently that lymphocytes large the immune are system specific tooperates avoid auto-immunity, even better when and not lymphocytes2 to recognize are many more pathogens. specific [4]. The conclusion remains that lymphocytes are 2 specific to avoid auto-immunity, and not to recognize many pathogens.

2 2 Regulatory T cells K. Saeki et al. / Journal of Theoretical Biology 375 (2015) 4–12 7

A T cell scans a DC

The DC activates a TN cell The DC activates a TS cell A TR cell deactivates the DC No interaction p = y (1 - (1 - p)n ) p = y (1 - (1 - p)n + m (1- f ) ) p = y (1 - (1 - p)n + m (1- f (1 - α)) ) N N S S R R pU = 1- pN - pS - pR

Fig. 3. The four possible scenarios when a DC is scanned by a random T cell: (i) the encounter leads to the activation of a non-self-reactive T cell, (ii) the encounter leads to the activation of a self-reactive T cell, (iii) the DC is recognized by a regulatory T cell and deactivated, or (iv) the T cell does not recognize any of the presented epitopes and remains inactive. The occurrence of events (i)–(iii) requires both that the randomly selected T cell is of the appropriate type and that it recognizes at least one of the epitopes presented on the DC. This is reﬂected in the two terms constituting the probability of these events. Saeki et al JTB 2015

The occurrence of event (i) requires that a randomly encoun- tolerant to self, can be calculated as tered T cell is non-self-reactive and that it recognizes at least one k 1 of the n foreign epitopes on the DC. Therefore, the probability that k À i Pt pN pU pR ∑ pN pU a single T cell-DC encounter leads to activation of a non-self- ¼ð þ Þ þ i 0 ð þ Þ ¼ reactive T cell (and therefore an immune response to the patho- k gen), pN, is given by the product of these two probabilities, 1 p p p p k p Àð N þ U Þ ; 4 n N U R p y 1 1 p : 3a ¼ð þ Þ þ 1 pN pU ! ð Þ N ¼ Nð Àð À Þ Þ ð Þ Àð þ Þ To consider the probability of event (ii), we assume that all self- where the first term indicates the probability that the DC does not epitopes are presented on a DC with equal probability. Thus, a DC activate any Tregs or self-reactive T cells during its k encounters, is expected to present 1 f m different self-epitopes not pre- ð À Þ while the second term provides the probability that the DC is shut sented in the thymus, and the number of potential ligands of a down by a regulatory T cell before mounting an autoimmune self-reactive T cell on a DC is 1 f m n. Because the probability ð À Þ þ response. In the same manner, the probability that none of the of recognition for one epitope is p, the probability that a random T encounters leads to any immune response can be calculated as cell-DC encounter leads to activation of a self-reactive T cell is given by k 1 pU k P0 p p À ; 5 1 f m n U R p y 1 1 p ð À Þ þ : 3b ¼ þ 1 pU ð Þ S ¼ Sð Àð À Þ Þ ð Þ À According to the same assumption, the number of potential again including the probability that no recognition takes place in ligands of a regulatory T cell is expected to be n foreign epitopes all k encounters (first term), and the probability that none of the and 1 1 α f m self-epitopes, since Tregs potentially recognize ð Àð À Þ Þ encounters leads to activation of the T cell until the DC is shut any self-epitope that does not induce apoptosis in the thymus. down by a Treg (second term). Hence, pR, the probability that a randomly encountered T cell is Now, suppose that there are q activated DCs present in the regulatory and recognizes at least one epitope on the DC, which draining LN. We define an immune response as “successful” if leads to deactivation of the DC, is none of these DCs elicits an autoimmune response, while at least 1 1 α f m n one DC evokes an immune response to at least one foreign p y 1 1 p ð Àð À Þ Þ þ : 3c R ¼ Rð Àð À Þ Þ ð Þ epitope. Then the probability of a successful immune response Finally, the probability of no cognate recognition in an encoun- can be written as ter between a DC and a randomly selected T cell in the draining LN, p , is P Pq Pq: 6 U suc ¼ t À 0 ð Þ p 1 p p p : 3d U ¼ À R À S À N ð Þ i.e. the probability of staying tolerant to self, minus the probability As soon as a DC is recognized by a Treg, the DC will be shut that no reaction at all is mounted. To investigate the influence of down and lose its capacity to activate naïve T cells. Let k be the regulatory T cells on the immune system, this probability Psuc was average number of different T cell clones that scan a single studied as a function of T cell specificity p and Treg induction activated DC. In this model we set k R T T T , assuming likelihood parameter α. Other parameter values were set to ¼ ¼ R þ S þ N that scanning is quick and a DC is scanned by the whole peripheral S 105 self-epitopes (Burroughs et al., 2004), R 1010 initial T ¼ 0 ¼ repertoire of T cell clones. The probability that no self-reactive T cell clones (Borghans et al., 1999), f 0.9, m 90 self-epitopes, ¼ ¼ cell is activated in all these encounters, i.e. that the animal stays n 10 self-epitopes and q 1 DC. ¼ ¼ MHC: very polymorphic but not diverse.

Hundreds of alleles in every population Almost everyone heterozygous at every locus 6 major loci: 12 different MHC molecules per host

Why not many more? 0.3

R/R0

0.2

0.1

0 1 2 3 4 MHC diversity (log M )

2 MHC diversity within the individual

Since individual MHC diversity increases the presentation of pathogens to the immune system, one may wonder why the number of MHC genes is not much higher than it is. The argument that is mostly invoked is that more MHC diversity within the individual would lead to T cell repertoire depletion during self tolerance induction. This argument is incomplete, however, because more MHC diversity could also increase the number of clones in the T cell repertoire through positive selection. In order to be rescued in the thymus, lymphocytes need to recognize MHC–self peptide complexes with su⇤cient avidity. A high MHC diversity thus increases both the number of lymphocyte clones that are positively selected and the number of clones that are negatively selected. To calculate the net eect of these two opposing processes we develop a simple mathematical model [5].

Consider an individual with M dierent MHC molecules and an initial T lymphocyte repertoire consisting of R0 dierent clones. Let p and n denote theMHC (unconditional) diversity within an individual chances that a clone is positively selected by a single MHC type, because its avidity is higher than a threshold T1, or negatively selected because its avidity

p0.3 probability of positive selection exceeds a higher threshold T2, respectively (see Fig.R/R 2).n0 probability By of thisnegative selection deﬁnition, thymocytes can only be negatively Note0.2 p > n selected by MHC molecules by which they are also positivelyM number of MHC selected,molecules per hosti.e. n

Using these experimental estimates, the number of clones in the functional T cell repertoire R increases with

3 0.3

R/R0

0.2

0.1

0 1 2 3 4 MHC diversity (log M )

2 MHC diversity within the individual

Since individual MHC diversity increases the presentation of pathogens to the immune system, one may wonder why the number of MHC genes is not much higher than it is. The argument that is mostly invoked is that more MHC diversity within the individual would lead to T cell repertoire depletion during self tolerance induction. This argument is incomplete, however, because more MHC diversity could also increase the number of clones in the T cell repertoire through positive selection. In order to be rescued in the thymus, lymphocytes need to recognize MHC–self peptide complexes with su⇤cient avidity. A high MHC diversity thus increases both the number of lymphocyte clones that are positively selected and the number of clones that are negatively selected. To calculate the net eect of these two opposing processes we develop a simple mathematical model [5].

R = R (1 n)M (1 p)M , (7) 0 0.3 ⇥ [5]. The functional repertoire R thus containsR/R0 all T cell clones that fail to be negatively selected, minus the ones that also fail to be positively selected0.2 by any of the M dierent MHC molecules of the host.

Experimental estimates for the parameters0.1 of this model have recently become available. In mice, around 3% of the T cells produced in the thymus end up in the mature T cell repertoire, and at least 50% of all positively 0 1 2 3 4 selected T cells have been shown to undergo negativeMHC diversity selection (log M ) in the thymus [17]. Thus, 94% of all thymic T

Parameters p=0.01 and n=p/2=0.005 [Borghans Eur J Imm 2003] cells fail to be positivelyFigure 2: Positive and selected negative selection by according any to the of avidity the model MHC [13]. The curve molecules in (a) depicts the distribution in the host [17]. These estimates can be used to of thymocyte avidities for self peptide–MHC complexes. In our model, the chance p to be positively selected by a calculate the chancessingle MHC typep isand the chancen thatof the a avidity T between cell the clone thymocyte to T cell be receptor positively and any of the self peptide– or negatively selected by a single type of MHC MHC complexes exceeds threshold T1. Thymocytes with avidities for self peptide–MHC complexes exceeding the upper threshold T2 are negatively selected (with chance n per MHC type). Panel (b) depicts the size of the T cell repertoire molecule. Takingas a function into of MHC account diversity. The that number of clones inbred in the functional mice repertoire areR is homozygous plotted as a fraction of the total and therefore express 3 types of class I MHC and 3 types of classinitial lymphocyte II MHC repertoire R0 molecules,. Parameters are: p =0.01,p andandn =0.005.n follow from: (1 p)6 =0.94 and n = p/2. This yields p =0.01 and n =0.005. 2 MHC diversity within the individual

Since individual MHC diversity increases the presentation of pathogens to the immune system, one may wonder why the number of MHC genes is not much higher than it is. The argument that is mostly invoked is that more Using these experimentalMHC diversity within the estimates, individual would lead the to T cell number repertoire depletion of clonesduring self tolerance in the induction. functional T cell repertoire R increases with This argument is incomplete, however, because more MHC diversity could also increase the number of clones in the T cell repertoire through positive selection. In order to be rescued in the thymus, lymphocytes need to recognize MHC–self peptide complexes with su⇤cient avidity. A high MHC diversity thus increases both the number of lymphocyte clones that are positively selected and the number of clones that are negatively selected. To calculate the net eect of these two opposing processes we develop a simple mathematical3 model [5]. Consider an individual with M dierent MHC molecules and an initial T lymphocyte repertoire consisting of R0 dierent clones. Let p and n denote the (unconditional) chances that a clone is positively selected by a single MHC type, because its avidity is higher than a threshold T1, or negatively selected because its avidity exceeds a higher threshold T2, respectively (see Fig. 2). By this deﬁnition, thymocytes can only be negatively selected by MHC molecules by which they are also positively selected, i.e. n

Using these experimental estimates, the number of clones in the functional T cell repertoire R increases with

3 REVIEW

Control of adaptive immunity by the innate immune system

Akiko Iwasaki & Ruslan Medzhitov

Microbial infections are recognized by the innate immune system both to elicit immediate defense and to generate long-lasting adaptive immunity. To detect and respond to vastly different groups of pathogens, the innate immune system uses several recognition systems that rely on sensing common structural and functional features associated with different classes of microorganisms. These recognition systems determine microbial location, viability, replication and pathogenicity. Detection of these features by recognition pathways of the innate immune system is translated into different classes of effector responses though specialized populations of REVIEW dendritic cells. Multiple mechanisms for the induction of immune responses are variations on a common design principle wherein the cells that sense infections produce one set of cytokines to induce lymphocytes to produce another set of cytokines, which in turn + + 60,61 75,76 activate effector responses. Here we discuss theseCD103 emergingCD11b DCs principles (IRF4 dependent) of innate. All four control subsets of tissue-of adaptivetion with antigenimmunity. and complete Freund’s adjuvant , which demon- resident DCs migrate into the draining lymph nodes and contribute in strates the key role of this DC subset in TH1 cell-mediated responses a distinct manner to priming responses of the adaptive immune system. to bacterial and fungal pathogens (Fig. 2). However, the exact iden- + In addition to the migrant DCs, within lymphoid organs CD8a DCs tity of the TH1-priming DCs remains controversial. While one report Innate control of adaptive immunity is now a well-established(Batf3 dependent) paradigm. and CD11b + DCsof are the constitutively innate present. control Studies of of adaptivehas shown aimmunity. requirement for For CD103 example,+CD207+ DCs the in TH1 type priming 2 76, First introduced by Charles Janeway Jr. in 1989 (ref. 1),mice it states with specific that deletions recog- of subpopulationsimmune ofresponse DCs have revealed induced an anotherby parasitic has shown worms that TH1 primingand allergens occurs normally appears in the absenceto emerging view of the unique capacity of DC subpopulations to generate of those DCs75 following immunization with myelin oligodendrocyte nition of conserved features of microbial pathogens bydifferent the innate classes immuneof T cell responses be (Table largely 1). independent of PRRs,glycoprotein perhaps in complete because Freund's multicellular adjuvant. parasites The clearance of viral infection relies on cytotoxic lymphocytes The TH17 subset of helper T cells is specialized in eliminating extra- system is mediated by pattern-recognition receptors (PRRs), which detect lack molecular structures that are both conserved across77,78 different groups (CTLs), whose differentiation follows the engagement of nucleic acid cellular bacteria and fungal pathogens . The priming of TH17 cells conserved pathogen-associated molecular patterns (PAMPs),sensors, which such leads as to bac the -productionof parasites of type 1 interferons and distinct and IL-12. from begins the with host the organism.engagement of Similarly,CLRs such as allergensdectin-1 and aredectin-2 terial and fungal cell-wall components and viral nucleicBatf3-dependent acids. Detection CD103+ DCs foundnot within microbial or under various in origin, mucosal lack expressed conserved by DCs, Langerhans structural cells andfeatures macrophages and79–81 induce. In addition epithelial cells, as well as the Batf3-dependent CD8a+ DCs in lymphoid to inducing dectins, the phagocytosis of bacteria-infected apoptotic of PAMPs by PRRs leads to the induction of inflammatoryorgans, are required responses for cross-presentation type 2 and immune the activation responses of CD8+ T cells through induces the mechanisms production of TGF- thatb and remain IL-6 by DCs, largely which pro- 62–64 5 82 and innate host defenses. In addition, the sensing ofcells microbes. These by DCs PRRs express RLRs,unknown NLRs and TLR3. Anotherbut not TLR7 (refs.big puzzlemotes T isH17 the cell differentiationapparent ability. Several ofstudies the support immune the notion 65,66) and produce type 1 interferons and IL-12 in response to the recog- that CD11b+CD103+ DCs that develop under the control of IRF4 have 60,61,83,84 expressed on antigen-presenting cells, particularly dendriticnition of viruses cells and (DCs), virus-infected system cells (Fig. to2). Thesedistinguish cells express between the a key rolebeneficial in TH17 cell commensal induction in vivo microorganisms. The cytokines criti - leads to the activation of adaptive immune responses.C-type Several lectin DNGR-1 classes (CLECL9A), of and which pathogenic detects F-actin microorganisms. exposed by cal for T BothH17 responses, types IL-6of microbes and IL-23, are express both selectively PAMPs secreted necrotic cells for the uptake and cross-presentation of necrotic cells67,68. by the CD11b+CD103+ DCs60,61. In addition to the CD11b+CD103+ PRRs have now been identified and characterized inIn some addition detail. to requiring These CD103 + DCs,and infection are detectable with fungal pathogens by PRRs; DCs, however, Langerhans the cells outcomeare necessary ofand their sufficient recognition for induction of the + 69 69 include Toll-like receptors (TLRs), nucleotide-bindingrequires oligomerization Batf3-dependent CD207 candermal depend DCs for CTL on responses additional. The characteristics,TH17 cell response to suchfungal pathogensas invasiveness in the skin and. We are pro tempted- CD141hiCLEC9A+ DCs found in the dermis, lung and liver interstitia of to speculate that these DCs co-engage TLRs and CLRs to selectively + domain (Nod)-, leucine-rich repeat–containing receptorshumans (NLRs), share a transcriptome RIG-I- signatureduction similar ofto that toxins. of mouse Recent CD8a induceprogress IL-23 andin suppressunderstanding IL-12 for optimal the T Hcomplexity17 cell induction in and CD103+ DCs and are also superior to other DCs in cross-priming vivo85. Human CD1c+ DCs, the counterpart of mouse CD11b+ DCs, like receptors (RLRs), C-type lectin receptors (CLRs) and AIM-2+ like recep70 - and diversity+ of commensal microbiota suggests that perhaps the clas- CD8 T cells ex vivo . In addition, human CD1a Langerhans cells also express IRF4, secrete IL-23 and promote TH17 cells in response to tors, as well as a family of enzymes that function as intracellularinduce functional sensors maturation of of CD8sic +notion T cells superior of ‘pathogens’ to that induced is Aspergillusin fact applicable fumigatus60 (Fig. only 2). to rare outliers of the nucleic acids, including OAS proteins and cGAS2–4. Theby otherphysiological skin DC subsets role ex vivo 71,72entire. spectrum of the microorganismsAlthough sensors that in the can innate colonize immune systema mammalian that detect prote - Defense by the immune systemREVIEW against intracellular bacteria and pro- ase allergens and helminthes are not yet known, studies have dem- + of different PRRs in the sensing of microbes and the inductiontozoa requires of responses adaptive by CD8 host. T cells andHowever, T helper type this 1 cells does (TH 1not onstrated mean that that induction the immune of the TH2 systemcell response is concernedto both requires the cell), which develop as a consequence of the engagement of TLRs by PD-L2+CD301b+ DC population86–88. The CD301b+ DCs comprise the immune responses continues to be investigated, but inPAMPs general, that leads the to theavail production- only of IL-12. with CTL theseresponses few to bacterial bad apples;majority evenof dermal the DCs normal that are CD207 commensal-, and these cellsinhabit require- IRF4 able evidence unequivocally supports the view that PRR-mediatedand protozoan infection sensing are ‘preferentially’ ants need induced to by be Batf3-dependent continuously for monitored their ability to promoteand somehow TH2 cell-mediated ‘managed’ immunity. by Of the note, in CD103+ DCs that produce IL-12 (refs. 73,74). 6–8 instructs the adaptive immune responses: specifically, PRRsT 1 cell -determinemediated immunity the to fungalimmune patho- system to prevent their outgrowth and mischief . H Lymphocyte Pathogen Sensor PRR Cytokines origin of the antigens recognized by the antigen receptorsgens depends expressed on Batf3-dependent on dermalMany DCs discoveries made in the field over the past decade calldifferentiation for a T cells and B cells, as well as determine the type of infectionthat express encountered, CD103 and CD207 (ref.more 69). completeThe and nuanced picture of innate instruction of the adap- GM-CSF-dependent DC population is required RLR Type I TLR3 and instruct lymphocytes to induce the appropriate effectorfor the induction class of ofTH 1the cells aftertive immuniza immune- responses. Here we review some of the interferonsrecent develop-CTL CDS IL-6 Batf3-dependent NLR Controlimmune response. of adaptive immunity by the innatements in studies of innateViruses control+ of +adaptive immunity. We highlight IL-1β Figure 2 The instruction of various classes of CD8α ,CD103 DCs AIM2 (mouse) Studies over the past decade have also revealed Tseveral cell responses important by DCs. The development some and emerging concepts that expandCD141+ DCs the pattern-recognition paradigm. function of DCs in vivo requires transcription (human) immuneaspects of immune system system’s function that require factorsan expanded that are critical viewfor each stageFinally, of their we discuss some of the major gaps and unknowns.IL-12 TLR IL-6 T 1 differentiation program. Studies probing the Type 1 Bacteria H NLR IL-1 Akiko Iwasaki & Ruslan Medzhitov requirement or sufficiency for DC subsets have immunity β cell suggested a relationship between the DC lineage Howard Hughes Medical Institute, Department of Immunobiology, Recognition by thePr innateotozoa immuneBatf3-dependent system and T cell programming. CTL responses are + + Microbial infections are recognized by the innate immune system both to elicit immediate defense and to generate long-lasting CD207 CD103 dDCs Yale University School of Medicine, New Haven, Connecticut,induced USA. by lymphoid organ-resident, Batf3- adaptive immunity. To detect and respond to vastly different groups of pathogens, the innate immune system usesMicrobial several recognition targets of recognition by PRRs are structurally diverse and dependent CD8a+ DCs and tissue CD103+ DCs IL-23 Bacteria Dectins systemsCorrespondence that rely on sensing should common be addressedstructural and tofunctional A.I. ([email protected] features associated(mouse) with and different the )human classes CD141 of microorganisms.+ DC counterpart. These IL-6 include complex polysaccharides, glycolipids,TLR lipoproteins, nucleo-T 17 recognition systems determine microbial location, viability, replication and pathogenicity.T 1 cell mediated Detection immunity of these requires features stimulation, by recognition IL-1β H or R.M. ([email protected]). H - Fungi NLR cell pathways of the innate immune system is translated into different classes of effectorin a GM-CSF-dependent responses though manner, specialized by tidesthe populations and nucleicof acids. SeveralLangerhans families cells (mouse), of PRRs detect theseTGF-β structures IRF4-dependent dendritic cells. Multiple mechanisms for the induction of immune responses are variations+ on+ a common design principle wherein + + CD207 CD103 DC subset, a minor throughpopulation of the use of distinct ligand-recognitionCD103 CD11b DCs domains, including leu- the cells that sense infections produce one set of cytokines to induce lymphocytesdermal to produce DCs (mouse). another T 17 set responses of cytokines, are induced which in turn (mouse) H + + by mucosal IRF4-dependent CD103+CD11b+ DCs CD1c CD11b activateReceived effector 26 responses. January; Here accepted we discuss 10 these February; emerging principles published of innate online control 19 of March adaptive 2015; immunity. cine-rich repeats, C-type lectin domains(human) and various nucleic acid–bind- as well as by skin Langerhans cells, depending on doi:10.1038/ni.3123 Innate control of adaptive immunity is now a well-established paradigm. of the innatethe control location of of adaptive the pathogen immunity. (mouse). Foring The example, humandomains. the type In2 addition to being characterized by their structure and + ? ? T 2 CD1c DCs prime TH17 cell responses. Finally, Type 2 H First introduced by Charles Janeway Jr. in 1989 (ref. 1), it states that recog- immune response induced by parasitic worms and allergens appears to cell nition of conserved features of microbial pathogens by the innate immune be largely independentTH2 cell-mediated of PRRs, immunity perhaps requiresbecause multicellularIRF4- parasitesimmunity dependent CD301b+CD11b+ DCs (mouse). These system is mediated by pattern-recognition receptors (PRRs), which detect lack molecular structures that are both conserved across different groups IRF4-dependent DCs comprise the majority of dermal DC (dDC) CD301b+CD11b+ DCs REVIEW conserved pathogen-associated molecular patterns (PAMPs), such as bac- of parasites and distinct from the host organism. Similarly, allergens are Helminth (mouse) NATURE IMMUNOLOGY VOLUME 16 NUMBER 4 APRILpopulation 2015 and are distinct from CD207+ CD103+ 343 terial and fungal cell-wall components and viral nucleic acids. Detection not microbial in origin, lack conserved structural features and induce Allergens Langerhans cells dermal DCs. In contrast, human Langerhans cells Venoms (human) of PAMPs by PRRs leads to the induction of inflammatory responses type 2 immune responses through mechanisms that remain largely induce TH2 cytokine secretion ex vivo. and innate host defenses. In addition, the sensing of microbes by PRRs unknown5. Another big puzzle is the apparent ability of the immune expressed on antigen-presenting cells, particularly dendritic cells (DCs), system to distinguish between beneficial commensal Level microorganisms 1 Level 2 Pathogen Sensor LymphocyteEffector Response leads to the activation of adaptive immune responses. Several classes of and pathogenic microorganisms. Both types of microbescytokine express PAMPs cytokine NATURE IMMUNOLOGY VOLUME 16 NUMBER 4 APRIL 2015 347 PRRs have now been identified and characterized in some detail. These and are detectable by PRRs; however, the outcome of their recognition ILC1 include Toll-like receptors (TLRs), nucleotide-binding oligomerization can depend on additional characteristics, such as invasiveness and pro- ILC3 domain (Nod)-, leucine-rich repeat–containing receptors (NLRs), RIG-I- duction of toxins. Recent progress in understanding the complexity TH1 like receptors (RLRs), C-type lectin receptors (CLRs) and AIM-2 like recep- and diversity of commensal microbiota suggests that perhaps the clas- IFN-γ Macrophage IL-1β T 17 Lysis of pathogens tors, as well as a family of enzymes that function as intracellular sensors of sic notion of ‘pathogens’ is in fact applicable only to rare outliers of the H IL-17 Neutrophil nucleic acids, including OAS proteins and cGAS2–4. The physiological role entire spectrum of the microorganisms that can colonizeIL-12 a mammalian CTL Tight junction IL-22 Epithelia of different PRRs in the sensing of microbes and the induction of adaptiveType 1 host. However, this does not mean that the immune systemIL-23 is concerned NK Antimicrobial NKT immune responses continues to be investigated, but in general, immunitythe avail- only withViruses these few bad apples;Macrophage even the normal commensalIL-6 inhabit- peptides able evidence unequivocally supports the view that PRR-mediated sensing ants need to be continuously monitored and somehow ‘managed’ by the γδT Bacteria DC Proliferation instructs the adaptive immune responses: specifically, PRRs determine the immune system to prevent their outgrowth and mischief6–8. Fungi IFN-γ IgG1 (mouse) origin of the antigens recognized by the antigen receptors expressed on Many discoveries made in the field over the past decade call for a B cell TFH1 IgG2 (mouse) T cells and B cells, as well as determine the type of infection encountered, more completeProtozo anda nuanced picture of innate instruction of the adap- IL-21 IgG4 (human) and instruct lymphocytes to induce the appropriate effector class of the tive immune responses. Here we review some of the recent develop- IgG1 (human) immune response. ments in studies of innate control of adaptive immunity. We highlight IgG3 (human) Studies over the past decade have also revealed several important some emerging concepts that expand the pattern-recognition paradigm. aspects of immune system’s function that require an expanded view Finally, we discuss some of the major gaps and unknowns. IL-1α ILC2 IL-4 Basophil Howard Hughes Medical Institute, Department of Immunobiology, Type 2 Recognition by the innate immune system TSLP TH2 IL-5 Eosinophil Worm expulsion Yale University School of Medicine, New Haven, Connecticut, USA. Microbial targets of recognition by PRRs are structurally diverse and T 9 immunity IL-25 H IL-13 Macrophage Tissue repair Correspondence should be addressed to A.I. ([email protected]) include complex polysaccharides, glycolipids, lipoproteins, nucleo- NKT IL-33 AREG Epithelia or R.M. ([email protected]). tides and nucleic acids. Several families of PRRs detect these structures Mucus secretion IL-9 throughHelminth the use of distinct ligand-recognition domains, including leu- IgG1 (mouse) Received 26 January; accepted 10 February; published online 19 March 2015; cine-rich repeats, C-type lectin domains and various nucleic acid–bind- IL-4 IgG4 (human) Allergens T 2 B cell doi:10.1038/ni.3123 ing domains. In addition to being characterized by their structure and FH IgE Venoms Epithelia IL-21

NATURE IMMUNOLOGY VOLUME 16 NUMBER 4 APRIL 2015Figure 4 The two-tiered design of the immune responses. Immune responses343 are orchestrated by three categories of cells that function as ‘sensors’ that detect pathogens and secrete ‘level 1’ cytokines, the tissue-resident lymphocytes that respond to ‘level 1’ cytokines to secrete ‘level 2’ cytokines, and the effector cells that respond to ‘level 2’ cytokines to carry out effector functions to eliminate the pathogen. DCs and macrophages serve as sensors for the type 1 immune response and epithelial cells and mast cells for the type 2 immune response. ‘Level 1’ cytokines act on terminally differentiated lymphocytes, including ILCs, ILLs, TFH cells and TRM cells. In response to ‘level 1’ cytokines, these lymphocytes produce ‘level 2’ cytokines. The ‘level 2’ cytokines act on the effector cells of the immune response. These include macrophages, neutrophils, epithelial cells, eosinophils and basophils, B cells as well as sensory neurons, endothelium and smooth muscle cells. These cells perform diverse effector functions, including barrier defenses, killing and expulsion of pathogens, antibody production and tissue repair.

and virulence activities induce responses of progressively higher cost. 2 immune response. The different modes of recognition by the innate Thus, macrophages activated by PRRs alone induce effector response of immune system described above operate in these cell types to induce lower cost (secretion of cytokines and antimicrobial peptides) than those production of the first level of cytokines. The ‘level 1’ cytokines include induced by macrophages activated by PRRs and interferon-g (IFN-g) IL-12, IL-23, IL-6 and IL-1b for the type 1 immune responses, and TSLP, (enhanced microbicidal activities and collateral damage) (Fig. 3). This IL-25, IL-33 and IL-1a for the type 2 immune responses115,116. These strategy should optimize the immune response so as to minimize its cost cytokines act on the second category of cells, which is made up of various while providing sufficient protection. In addition to immunopathology, populations of lymphocytes, including ILCs, innate-like lymphocytes immune responses can have other costs, most notably metabolic costs. (ILLs; including NKT cells, MAIT cells and epithelial gd T cells), TFH Therefore, immune responses are probably calibrated on the basis of cells and TRM cells. In response to ‘level 1’ cytokines, these lympho- weighted contribution of different costs. cytes produce ‘level 2’ cytokines. These include IFN-g, IL-17, IL-22 for type 1 immune response, and IL-4, IL-5, IL-9, IL-13 and AREG for type Design principle of immune responses 2 immune response. IL-21 produced by TFH cells is common to both Over the past decade, many exciting discoveries have been made about type 1 immune responses and type 2 immune responses. The ‘level 2’ the functions and regulation of ILCs108–110, tissue-resident memory T cytokines act on the third category of cells, which are effectors of the 111–114 cells (TRM cells) and epithelial cells, as well as the signals that immune response. These include macrophages, neutrophils, epithelial activate these cell types and the cytokines they produce. These discov- cells, eosinophils and basophils, and B cells (specifically B-2 cells), as eries paint a complex picture of multiple cell types connected by cyto- well as sensory neurons, endothelium and smooth muscle cells. These kines that perform distinct functions. However, behind this complexity, cells perform diverse effector functions, including barrier defense, kill- there is a simpler pattern that reveals a common design principle of the ing and expulsion of pathogens, antibody production, and tissue repair. immune response, with many details being a variation on a common Some cell types can function both as sensors and as effectors (Fig. 4). theme. Macrophages function as sensors when activated by the PRRs and as The commonalities become apparent with the consideration that all effectors when activated by ‘level 2’ cytokines produced by lymphocytes. the cell types involved in the initiation and execution of the immune However, as discussed above, the responses elicited by these stimuli are response fall into three categories (Fig. 4). The first category is made distinct: as sensors, macrophages secrete ‘level 1’ cytokines to elicit up of cell types that function as sensors of infection or damage. These responses from local lymphocytes. In contrast, macrophages stimu- are the cell types that express various PRRs and other sensing machin- lated as effectors by a ‘level 2’ cytokine (for example, IFN-g) become ery to detect microorganisms, macroparasites, allergens, toxins and highly activated effectors, induce robust antimicrobial and phagocytic venoms and any other stimuli that elicit immune responses. The cell responses, including the production of reactive oxygen and nitrogen types that function as sensors include DCs and macrophages for the species. Similarly, epithelial cells serving as sensors of helminth infection type 1 immune response, and epithelial cells and mast cells for the type produce the ‘level 1’ cytokines IL-33, IL-25 and TSLP, whereas epithelial

NATURE IMMUNOLOGY VOLUME 16 NUMBER 4 APRIL 2015 349 A simulation model: Borghans, Int. Immunol. 2002

bacterial peptides

viral peptides ....

toxin peptides

self peptides Repertoire

Set of peptides for each antigen group. Clones will recognize peptides from all sets with probability p.

21 Play an immune system game

1. Make a pathogen by drawing e epitopes from its set 2. Determine phenotype of all clones responding 3. If all naive: switch them to memory of corresponding mode 4. If memory available: switch all to that mode • determine succes • 5. Goto 1.

22 A simple example

Clone numbers: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 ... Ro

Initial types: 0 0 0 0 0 0 1 0 0 0 0 0 1 0 ... 0

Antigen 1, type 7: 7 0 0 7 0 0 1 7 0 0 0 0 1 0 ... 0 Zero score

Antigen 2, type 5: 7 5 0 7 5 0 1 7 0 0 0 5 1 0 ... 0 Zero score

Antigen 3, type 5: 7 5 5 7 5 0 1 7 0 0 0 5 1 0 ... 0 Positive score

Antigen 4, type 9: 7 5 5 7 5 5 1 7 0 0 5 5 1 0 ... 0 Negative score

23 10e3 epitopes per set, 10e6 clones

1 response positive negative 0.8 autoimmunity

0.6

0.4 Fraction immune responses 0.2

0 10−8 10−6 10−4 10−2 100 Lymphocyte cross−reactivity (p)

24 528 Diversity and speci®city in the immune system

Fig. 3. The performance of a diverse immune system. The fractions Fig. 4. The performance of a less diverse immune system. The 4 of different immune responses have been plotted for different repertoire consists of R0 = 10 clonotypes; for other parameters and degrees of cross-reactivity (p), after challenge with 103 different the interpretation of the different curves, see the legend of Fig. 3. pathogens. The thick curve denotes the fraction of challenges that This repertoire has the largest chance of making an immune induce any immune response at all. The fraction of challenges in response at p ~ 10±4. With such a high cross-reactivity, however, which pre-existing effector/memory clones induce the correct type of many mistakes are made. immune response is denoted by the thin curve. The fraction of challenges in which pre-existing effector/memory clones induce an inappropriate immune response is denoted by the dashed curve. example of a small simulation is given in Fig. 2. The above The fraction of challenges leading to autoimmunity, caused by gives a full description of our model simulations. The C-code of ignorant self-speci®c clones that are triggered by foreign antigens, our program is available upon request. is denoted by the dash-dotted curve. This repertoire is most functional at a cross-reactivity of p ~ 10±6, because the chance of immunity is then close to 1, and the net contribution of pre-existing effector/memory cells is high. There are e = 6 different epitopes per Results pathogen, pathogens come from m ± 1 = 8 different groups, each consisting of N = 103 different epitopes, a fraction f = 0.8 of all S = Somatic learning requires speci®city 4 6 10 self epitopes induces tolerance and there are R0 = 10 clonotypes. We have studied how the performance of an immune system that is challenged with 103 different pathogens depends on the cross-reactivity p of its lymphocytes (Fig. 3). The probability of immunity (i.e. the probability that all 103 pathogens with the corresponding effector mechanism. Even if an are recognized by at least one non-tolerant clone, see the thick inappropriate type of response is induced, naive lymphocytes line in Fig. 3) is close to 1 in a wide range of intermediate cross- acquire the corresponding (incorrect) effector mechanism. reactivities. A too cross-reactive immune repertoire fails to Memory clonotypes involved in a response to an antigen do respond to foreign antigens because the majority of lympho- not switch effector type (15,61). Pathogens never kill their cytes has been rendered non-functional during self-tolerance hosts, i.e. the simulations are continued even if an inappro- induction (53,54,62). If lymphocytes are too speci®c, on the priate response is induced. other hand, the immune repertoire frequently fails to recognize The performance of the model immune system is followed a foreign antigen. Thus, the simulations con®rm that suf®cient by counting the fractions of infections in which the presence of cross-reactivity is required to ensure an immune response effector/memory clones helps or hinders the induction of an against any pathogen (63). appropriate immune response. In the default situation, the Within the range of cross-reactivities yielding a high chance pathogen is only recognized by naive clonotypes, and the of immunity, effector/memory clones only tend to make correct effector type of the responding clonotypes is determined by decisions if they are suf®ciently speci®c (see the thin line in the innate immune system and the immunological context of Fig. 3). At a relatively high cross-reactivity, i.e. p ~ 10±3, the the pathogen. All cases in which pre-existing effector/memory positive contribution of effector/memory clones is largely clones establish the correct type of response against a coincidental. Even if there is no structural relationship between pathogen (without being responsive to any self antigens) the pathogens, such a protective effect is observed because contribute to the protection of the individual. The cases in of random cross-reactions (not shown). Since there are eight which effector/memory clones establish an incorrect type of different responsive modes in our simulations, the probability response against a pathogen are detrimental. We also count with which effector/memory clones coincidentally induce the the number of autoimmune responses, which are induced right type of immune response is 1/8. If lymphocytes are when naive clonotypes that are ignorant of their self epitopes suf®ciently speci®c, this randomness disappears and the are triggered into one of the responsive modes (49±52). An fraction of immune responses in which the presence of pre-