Class II major histocompatibility complex mutant miceto study the germ-line bias of T-cell antigen receptorsDaniel Silbermana,b, Sai Harsha Krovib, Kathryn D. Tuttleb, James Crooksc, Richard Reisdorphd, Janice Whitea,James Grossa, Jennifer L. Matsudaa, Laurent Gapinb, Philippa Marracka,b,e,1, and John W. Kapplera,b,e,1

aDepartment of Biomedical Research, National Jewish Health, Denver, CO 80206; bDepartment of Immunology and Microbiology, University of ColoradoSchool of Medicine, Aurora, CO 80045; cDivision of Biostatistics and Bioinformatics, National Jewish Health, Denver, CO 80206; dPharmaceutical Sciences,University of Colorado School of Medicine, Aurora, CO 80045; and eHoward Hughes Medical Institute, National Jewish Health, Denver, CO 80206

Contributed by Philippa Marrack, July 6, 2016 (sent for review March 2, 2016; reviewed by Erin J. Adams and Martin Flajnik)

The interaction of αβ T-cell antigen receptors (TCRs) with peptidesbound to MHC molecules lies at the center of adaptive immunity.Whether TCRs have evolved to react with MHC or, instead, pro-cesses in the thymus involving coreceptors and other moleculesselect MHC-specific TCRs de novo from a random repertoire is alongstanding immunological question. Here, using nuclease-tar-geted mutagenesis, we address this question in vivo by generatingthree independent lines of knockin mice with single-amino acidmutations of conserved class II MHC amino acids that often areinvolved in interactions with the germ-line–encoded portions ofTCRs. Although the TCR repertoire generated in these mutants issimilar in size and diversity to that in WT mice, the evolutionarybias of TCRs for MHC is suggested by a shift and preferential use ofsome TCR subfamilies over others in mice expressing the mutantclass II MHCs. Furthermore, T cells educated on these mutant MHCmolecules are alloreactive to each other and to WT cells, and viceversa, suggesting strong functional differences among these rep-ertoires. Taken together, these results highlight both the flexibilityof thymic selection and the evolutionary bias of TCRs for MHC.

T-cell receptor | MHC | evolution | mutation | variable region

The genes for immunoglobulins (Igs), αβ T-cell receptors(TCRs), and antigen-presenting MHC proteins appeared at

least 450 million years ago in the cartilaginous fish and are presentin all modern vertebrates (1–3). The more primitive hagfish andlampreys lack these genes and have an adaptive immune systemcomprised of unrelated proteins (4). The main ligands for αβTCRs are short peptides derived from self and foreign proteins,captured in a specialized groove of MHC class I (MHCI) and classII (MHCII) molecules and presented to T cells (5, 6). FunctionalIgs and TCRs are created by very similar recombination mecha-nisms involving fusion of V, J, and sometimes D gene segmentswith additional variations at the junctions to create an enormouspotential repertoire of Igs and TCRs, suggesting a common, un-known evolutionary origin for these loci.These observations have raised several unanswered questions.

For example,why did a separate TCR-rearranging gene systemdevelop for lymphocytes recognizing peptide–MHC ligands? Howdid the extraordinarily polymorphic MHC genes stay functionallyconnected to TCR genes throughout 450 million years of evolu-tion? One long-standing hypothesis has been that certain featuresof TCRs and MHC molecules are evolutionarily conserved topromote their interaction (7–10). Like Igs, the antigen-recogni-tion portions of TCRs are partially encoded in the comple-mentary determining region (CDR) CDR1 and CDR2 loops ofgerm-line TCR Vα (TRAV) and Vβ (TRBV) genes and arepartially generated by somatic recombination processes thatform the CDR3 loops. This initial repertoire is culled dramati-cally during T-cell development in the thymus. First, only those Tcells whose TCRs have at least some minimal affinity for the self-peptide–MHC molecules expressed in the thymus are positivelyselected for further development (11, 12). The T cells in thispopulation whose TCRs have too high an affinity for these self-peptide–MHC molecules are eliminated by an apoptotic mech-

anism termed “negative selection” (13, 14). The remaining Tcells go on to mature and form the peripheral T-cell repertoire.The effect of positive and negative thymic selection on limiting

the T-cell repertoire has made it difficult to test directly whethergerm-line features of TCRs and MHC molecules have been con-served to promote their interaction. However, some data consistentwith this notion have accumulated over the past several decadesthrough sequencing, X-ray crystallographic, mutational, and de-velopmental studies. For example, random examination of mouseT cells before positive selection showed a high frequency of MHC-reactive cells (15–17). In mice constructed to allow positive selec-tion but incomplete negative selection, an even higher frequency ofgenerically MHC-reactive T cells was observed (18). Structural andsequencing studies of MHC molecules have shown that the greatmajority of their polymorphisms are within the peptide-bindinggroove, not on the tops of the MHC α- and β-chain helices thatinteract with TCRs (Table 1). The CDR1 and CDR2 loops ofTCRs are much less variable in length than those of Igs (19). In thedozens of structures of peptide–MHC/TCR complexes that havebeen solved, a diagonal orientation of the TCR is nearly alwaysseen. This orientation usually causes the somatically generatedCDR3s to be focused on the peptide and the germ-line–encodedCDR1 or CDR2 amino acids, especially those of CDR2, to bedocked on the conserved portions of the MHC helices (9).Mutation of these TCR amino acids impairs T-cell recognitionof the ligand and affects thymic development of the T cells invivo (8, 20–22). Some of these germ-line TCR amino acids canbe traced back to the TCRs of fish, and, despite their overall

Significance

The evolutionary hypothesis for T-cell antigen receptor–pep-tide major histocompatibility complex (TCR–pMHC) interactionposits the existence of germ-line–encoded rules by which theTCR is biased toward recognition of the MHC. Understandingthese rules is important for our knowledge of how to manip-ulate this important interaction at the center of adaptive im-munity. In this study, we highlight the flexibility of thymicselection as well as the existence of these rules by generatingknockin mutant MHC mice and extensively studying the TCRrepertoires of T cells selected on the mutant MHC molecules.Identifying novel TCR subfamilies that are most evolutionarilyconserved to recognize specific areas of the MHC is the firststep in advancing our knowledge of this central interaction.

Author contributions: D.S., S.H.K., L.G., P.M., and J.W.K. designed research; D.S., S.H.K.,K.D.T., R.R., and J.W. performed research; D.S., S.H.K., J.G., J.L.M., and J.W.K. contributednew reagents/analytic tools; D.S., J.C., and L.G. analyzed data; and D.S., S.H.K., L.G., P.M.,and J.W.K. wrote the paper.

Reviewers: E.J.A., University of Chicago; and M.F., University of Maryland.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. Email: marrackp@njhealth.org or kapplerj@njhealth.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1609717113/-/DCSupplemental.

E5608–E5617 | PNAS | Published online September 1, 2016 www.pnas.org/cgi/doi/10.1073/pnas.1609717113

Dow

nloa

ded

by g

uest

on

May

17,

202

0

weak sequence homology, substitution of fish V segments for themouse V segments preserves antigen recognition of the mousepeptide–MHC complex (23). Finally, although RAG-mediatedrearrangement makes the CDR3 more diverse, the CDR1 andCDR2 loops in TCRs, unlike those in Igs, do not undergo an-tigen-selected somatic mutation; thus they keep their germ-linesequence and antigen-driven responses throughout develop-ment, suggesting a conserved function (24, 25).Another model for the MHC restriction of TCRs has been put

forth. According to the selection model, MHC restriction is notintrinsic to TCR structure but imposed is by the CD4 and CD8coreceptors that promote signaling by delivering the tyrosinekinase Lck to TCR–MHC complexes through coreceptor bindingto MHC during positive selection (26).In the current study we assessed the importance of several

MHCII conserved docking sites for TCRs by introducing specificpoint mutations into mouse I-Ab MHCII α or β genes. In vitrothese mutations had little effect on the collection of self-peptidesbound by the mutant I-Ab but often disrupted the recognition ofpeptide plus MHC by T cells specific for a variety of foreign or self-peptides. In vivo, mice carrying these MHC point mutations de-veloped TCR repertoires that were similar in size to those of WTmice but with altered TRAV or TRBV gene use. Furthermore, invitro in mixed lymphocyte reactions, T cells from each of the WTand mutant mice responded strongly to antigen-presenting cells(APCs) from the other mice but not to their own cells. We discussthese results in relation to the current ideas and data about the roleof evolution vs. somatic selection in framing the T-cell repertoire.

ResultsMutations of the I-Ab Conserved Amino Acids Affect the Presentationof Foreign Peptides to Antigen-Specific T-Cell Hybridomas. AlthoughMHC genes are extremely polymorphic, the amino acids on the αand β helices of MHCII that are frequently engaged by the TCRCDR1 and CDR2 loops are usually conserved, sometimes evenacross species (9). Ten of these amino acids tend to be almostmonomorphic in the mouse I-A alleles found in the majority oflaboratory strains (Table 1). A number are conserved as well inmouse I-E molecules and in the MHCII alleles of humans andother species (Table S1). These residues are located on the topsof the MHCII α-helices, in positions where they are less likely toaffect peptide binding and are more likely to affect interactionsof the MHCII protein with TCRs (Fig. 1A). Thus, we hypothe-size that these amino acids may have been conserved duringevolution to promote interactions with TCRs.To study the relative importance of these amino acids in TCR

recognition of peptide–I-Ab complexes, we mutated each of these10 residues separately. Nonalanine amino acids were replaced

with alanine (A), and alanines were replaced with glutamine (Q).Alanine was chosen as a neutral, frequently used mutational re-placement, and glutamine was chosen because it is already presentat the other positions on the helix and thus would not greatly alterthe chemistry at the surface of the protein. Genes encoding eitherthe mutant I-Ab α- or β-chain, paired with the corresponding WTI-Ab α or β gene, were transduced into an MHCII-deficient B-celllymphoma, M12.C3 (27, 28), to create APCs expressing the mu-tant I-Ab molecules. M12.C3 cells, derived from an H-2d mouse,lack an I-Ad β-chain but express a functional I-Ad α-chain from theoriginal M12 BALB/c lymphoma. This I-Ad α-chain can some-times pair with some other introduced I-A β-chains, including thatof I-Ab. For this reason we prepared M12.C3 cells transduced withonly the WT I-Ab β-chain to control for the possible activity of theI-Ad/b mixed molecule. M12.C3 cells transduced with both of theWT I-Ab genes served as a positive control, and M12.C3 cells withonly the WT I-Ab α-gene were also used as a negative control.All the M12.C3 transductants were cloned at limiting dilution,

and surface expression of I-Ab was confirmed by flow cytometry.Because each mutation might have affected the epitopes recog-nized by individual mAb differently, we stained the cells using avariety of anti–I-Ab

–specific mAbs. Fig. 1B shows data for the227 mAb, the antibody least affected by the mutations. With this

Table 1. Alignment of I-A haplotype helix residues

*The solvent-exposed residues of I-Aα or I-Aβ mutated in this study arenumbered.†Consensus sequence.‡Differences from consensus sequence.

Fig. 1. Screening and selection of I-Ab mutants that affect mature T-cellresponses. The WT and mutant MHC molecules are expressed at similar levelsin cell lines and present peptides. (A) The solvent-exposed residues that weretargeted for mutations are indicated for the α- (cyan) and β- (magenta) chains ofH-2 I-Ab. Residues were chosen for mutational analysis based on their conservationbetween H2 molecules and their predicted interaction with TCRs. (B) M12.C3 cellsexpressing the mutant constructs were stained with the 17/227 mAb, and theexpression of I-Ab was determined by flow cytometry. (C) Hybridomas werestimulatedwith cognate antigen presented by APCs expressingWT ormutant I-Ab.Activationwas determined by flow cytometry andwas defined as theMFI of CD69.The ability of each mutant to stimulate the hybridomas is displayed after nor-malization to WT responses. Data are representative of three or four biologicalreplicates per group; *P < 0.05 by a one-sample t test with a true value of 100. (D)Peptides were eluted from WT I-Ab and were analyzed by MS. Peptides withidentical HPLC retention times that were present in three separate WT sampleswere identified. Data indicate the percentage of peptides that were also identifiedin duplicate runs isolated from βT77A, βR70A, and αA64Q cells. (E) Peptides pre-sent in duplicate runs from each of the mutants were identified and comparedwith WT. Data are representative of multiple MS and MS-MS runs.

Silberman et al. PNAS | Published online September 1, 2016 | E5609

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

17,

202

0

mAb, the WT and mutant I-Ab cells all stained with a meanfluorescence intensity (MFI) 10- to 30-fold higher than that of thenegative controls. The MFI for the cells expressing the I-Ab/d mixedmolecule was much lower. Thus, all the mutants were expressed atabout the same level.Next, we devised a system in which the responses of many dif-

ferent T cells to antigen bound to the mutant MHCIIs could beassessed simultaneously. C57BL/6 mice were immunized separatelywith five different antigens (Table 2). Seven days later, T cells fromthe draining lymph nodes of the immunized mice were restimulatedwith their cognate antigens, expanded in vitro, and fused in bulk tothe TCR αβ− BW5147 thymoma cell line to create T-cell hybrid-omas. The preparations were named for their target MHC-II allele,I-Ab, and antigen (Table 2).The bulk T-cell hybridoma preparations were cultured with

M12.C3 cells expressing WT I-Ab, each of the mutant I-Abs, orWT I-Ad/b with or without the immunizing antigen. Activation ofthe T-cell hybridomas was assessed by up-regulation of CD69 onthe cells, as measured by flow cytometry (Fig. 1C). On average,about 50-fold more of the T-cell hybridomas in the bulk pop-ulations responded to their immunizing antigen plus M12.C3cells bearing WTαβ I-Ab than to control M12.C3 cells with onlyI-Ab WTβ. Nearly all the responses of the peptide or hen egg ly-sozyme (HEL)-specific bulk T-cell hybridomas were significantlyreduced when the mutant APCs were used instead of WT APCs,again with their immunizing antigen. The responses by bulk key-hole limpet hemocyanin (KLH)-specific T-cell hybridomas werealso reduced 1.5- to 16-fold in various β-chain mutants but not inα-chain mutants, perhaps because the large KLH protein may havemany potential I-Ab

–binding epitopes, and therefore, as group, Tcells specific for KLH may be less sensitive to any one MHCIImutation. Consistent with this relative lack of sensitivity to α-chainmutants, some of the bulk KLH-specific cells were also cross-re-active to KLH presented by APCs bearing the mixed I-A moleculein which the I-Ad α-chain replaced the I-Ab α-chain.These results confirmed and extended our previous studies

(29) because they showed that the conserved amino acids on theMHCII helices are not required for MHCII surface expression.However, in agreement with previous work, they are often impor-tant for TCR docking during CD4+ T-cell responses, leaving openthe possibility that their conservation might be required to ensuregerm-line–encoded favorable MHCII docking sites for TCRs.We selected mutants from this group for in vivo studies to find

out if they also affected T-cell thymic development. We con-sidered the four mutations that most consistently inhibited T-cellactivation: A64Q on the α-chain and R70A, T77A, and H81A onthe β-chain. αA64 is invariant in mouse and human MHCII andcreates a docking “cup” for TCR Vβs that contain a tyrosine (Y)at position 48 of CDR2 (9). βT77 and βH81 are adjacent on theI-Ab β-chain α-helix (Fig. 1A). βT77 is invariant in commonmouse I-A and I-E alleles and in human HLA-DR and HLA-DQalleles. In TCR/MHC structures, βT77 and βH81 are oftencontacted by the TRAV CDR1 loop (9). However, the highlyconserved βH81 has been implicated in the activity of mouseH-2DM and human HLA-DM, the proteins that catalyze endo-

somal peptide loading into MHC (30), and in TCR/MHC struc-tures often makes a surface-exposed H-bond to the peptidebackbone. Therefore we decided not to mutate this amino acid inour experiments. βR70 is nearly monomorphic in all mouse I-Aalleles (Table 1) but is not conserved in mouse I-E alleles or in theMHCII alleles of other species. In nearly all published TCR/I-Astructures it lies in the central region of the TCR footprintinteracting with the TCR CDR3s and therefore might be expectedto influence somatic CDR3 selection during thymic selection butperhaps not to have as strong an influence on germ-line Vα andVβ use. Therefore we choose αA64Q and βT77A as the primarymutations to test our hypothesis and βR70A as a potential control.

Effects of the βT77A, βR70A, and αA64Q Mutations on the PeptidesBound to I-Ab. Before proceeding to in vivo experiments withthese mutants, we considered the possibility that, despite thepredicted lack of a direct role for these I-Ab amino acids inpeptide binding, they might indirectly change the spectrum ofI-Ab

–presented self-peptides. Such changes would confound ourexperiments because positive selection involves reaction of theTCRs with both MHC and peptide, and we intended to examinethe effects of MHC mutations independent of changes in thebound peptide. To determine whether the MHCII mutations wecreated altered the spectrum of bound peptides, we comparedthe repertoire of peptides bound to WT I-Ab with those of thethree I-Ab mutants expressed in our M12.C3 transfectants. Onecaveat of these experiments is the thymus presents a different setof peptides in a cathepsin L-dependent fashion (31) and thusmay behave differently than the transfected M12.C3 cells.The WT and mutant I-Ab proteins were immunoprecipitated

from lysates of the transduced M12.C3 cells. Peptides wereeluted from these preparations and subjected to MS or MS-MSanalysis as previously reported (32) and as described in Materialsand Methods. A preliminary MS-MS analysis of the peptidesisolated from WT and mutant I-Ab showed that they had I-Ab

–binding motifs (Table S2) (33, 34). This finding served to vali-date our method of peptide isolation and suggested that theI-Ab mutations did not affect the I-Ab peptide-binding motif.To compare the peptides bound to WT vs. mutant I-Ab pro-

teins, immunoprecipitations and elutions for each sample wereperformed and analyzed with duplicate runs by MS. Limited MS-MS again confirmed the presence of the I-Ab

–binding motifin the peptides. A list of the peptides with identical HPLC re-tention times and calculated masses that were present in threeseparate WT I-Ab samples was compared with those in duplicateruns of mutant samples (Fig. 1D). Nearly all the total peptideintensities found in the WT I-Ab samples were also identified inall the mutant I-Ab samples. To determine if unique peptidesappeared only in the mutants, we first created a list of peptidesthat were found in duplicate MS runs of the same mutant sam-ple. Any peptide in this list that also appeared in any of the threeWT samples was also called present. Once again, most of thepeptides and nearly all the intensity from the mutant sampleswere found in the WT runs (Fig. 1E). Less stringent criteria, e.g.,requiring that a peptide be present in only two of the three WTor mutant I-Ab samples, identified even more peptides sharedamong the samples. This analysis does not identify peptidesbelonging to nested sets; therefore the similarity between sam-ples may be underestimated, because differently trimmed pep-tides were considered to be different by our analyses, althoughthe peptide they present to T cells is identical. Taken together,these experiments support the notion that βT77A, βR70A, andαA64Q MHC mutations do not notably alter the repertoire ofself-peptides bound to I-Ab. We therefore proceeded to test theeffects of these mutations on thymic selection in vivo.

Creation of Mice Bearing the βT77A, βR70A, and αA64Q Mutations.We produced mice expressing only the WT or mutant forms ofI-Ab. To ensure that the mutant and WT genes were expressedon the proper cells at the appropriate levels, we changed thecoding sequences of the genes in situ, using zinc finger nuclease

Table 2. Bulk hybridomas and their immunizing antigen

Bulk hybridomaProtein or peptide

antigen Sequence

BB5 Vaccinia B5 peptide FTCDQGYHSSDPNAV

BNP LCMV NP peptide SGEGWPYIACRTSIVGRA

BHEL Hen egg lysozyme Whole proteinBKLH Keyhole limpet

hemocyaninWhole protein

BMOG Myelin oligodendrocyteglycoprotein peptide

MEVGWYRSPFSRWHLYRNGK

Bulk hybridomas are named for the haplotype of the mouse (H-2b) andthe immunized antigen shown in this table.

E5610 | www.pnas.org/cgi/doi/10.1073/pnas.1609717113 Silberman et al.

Dow

nloa

ded

by g

uest

on

May

17,

202

0

(ZFN) technology to generate knockin point mutations directlyin fertilized C57BL/six eggs (35, 36).Custom ZFNs were designed (Sigma Aldrich) for both H2-Ab1

and H2-Ab2, the genes that encode the α- and β-chains of theonly MHCII molecule expressed in C57BL/6N mice. To reduceoff-target effects, the ZFNs were designed to ensure that no otherregion of the mouse genome had fewer than five DNA basemismatches to the sequence targeted by the ZFNs. A template forhomology-directed repair (HDR) was used to introduce our mu-tations into mice. This template had four components: (i) themutation of interest; (ii) a silent mutation to create a new re-striction enzyme site for screening of progeny; (iii) a silent mu-tation to disrupt the ZFN binding so that a subsequent insertion ordeletion event caused by nonhomologous end joining (NHEJ)would not occur after our mutation of interest had been in-troduced; and (iv) roughly 1,000 bp of homology on either side ofthe target of the ZFN (Fig. 2A) (37).DNA from the resultant mice was analyzed to identify chro-

mosomes bearing the desired mutation. The method was sur-prisingly robust, with NHEJ events identified in nearly all the miceand at least one chromosome with the correct mutation found in>10% of the mice overall. Mutant mice were crossed to WT miceand then intercrossed to create mice homozygous for each of thethree mutations. All mice showed equivalent levels of I-Ab cell-surface expression on peripheral cells (Fig. 2B).

Phenotypic Analysis of Thymic T Cells in the Mutant Mice. To de-termine whether any of our MHC mutations affected the de-velopment of CD4+ T cells, the thymus of each mouse strain wasanalyzed by flow cytometry. No significant difference in thenumber of thymocytes in the double-negative (DN), double-positive (DP), and single-positive (SP) populations was detectedbetween the thymi of the mutant mice vs. WT mice (Fig. 2 C–E).Analysis of CD5 and CD69 expression, markers of DP thymocyteactivation during positive selection, showed that the size of theexpressing population was not changed in the βR70A and βT77Amutant mice but was significantly reduced in the αA64Q mutantmice (Fig. 2 C, D, and F), suggesting that at least the αA64Qmutation reduced positive selection and MHCII reaction by TCRs.

Effect of the βT77A and βR70A Mutations on the Use of TRAVs. Wenext examined the TCR repertoire of peripheral T cells selectedin WT and mutated mice. Because the germ-line portions of theTCRs interacting with I-Ab βT77 were predicted to be those ofTRAV CDR1s, we predicted that the TRAVs used in the mutantmice would be more affected by these mutations than the TRBVs.Therefore, we compared TRAV use in the βT77A mice with thatin the WT mice, using the βR70A mutant mice as a possiblecontrol, because this amino acid most often interacts with randomlygenerated CDR3 regions rather than the germ-line encoded DR1and CDR2 regions. Anti-TRAV staining with the four availableanti-TRAV mAbs (TRAV14, TRAV9, TRAV12, and TRAV4)revealed a significant reduction in TRAV14 use in the mutant mice(Fig. 3 A–C). As expected, this reduction was seen only in CD4 Tcells, not in CD8 T cells. This TRAV14 shift was the first indicationof an altered TCR repertoire in the T77A mice.This analysis was limited by the small number of anti-TRAV–

specific mAbs available. Moreover, the TRAV antibodies thatare available might not distinguish between subfamily members ineach TRAV family (see below). To overcome this reagent limi-tation, we examined the TRAV repertoires of the MHCβ mutantmice in greater detail using deep sequencing. We used a set offorward primers specific for the TRAV families and common Cαreverse primers to generate a diverse PCR product that encodedthe TRAVs present in the naive CD4+ T cells from each strain ofmice. These fragments were sequenced with high-throughputmethods. Using software developed in house, we filtered outshort sequences and determined the TRAV, TRAJ, and CDR3used by each sequence. Although we designed our TRAV pri-mers to be family specific, of similar length, and with similarmelting temperatures, we expected our results to be quantitative

only in comparisons within TRAV families and to be just semi-quantitative in comparisons between TRAV families. Never-theless, the biases in analysis between each TRAV family shouldbe common to the different types of mice, so we believed com-parisons of TRAV use between mouse genotypes were justified.We compared the use of the TRAV family and the TRAV

subfamily among the mice. First, we looked at the average use ofthe 20 TRAV families present in the mice (Fig. 3D) in the WT vs.mutant mice. Use of the DESeq2 package (38), which is oftenimplemented in comparing mRNA expression in different cellpopulations, revealed no significant differences in the frequency ofTRAV use by the T cells in WT and R70A mice. However, therewere reproducible and statistically significant differences in TRAVuse by the T cells in T77A and WT mice. The TRAV 3, 6, and 11families were used more frequently, and the TRAV 5, 7, and 14families were used less frequently by T cells in T77A mice than byT cells in WT mice; the latter finding confirms our mAb-stainingresults (Fig. 3 A and B).However, TRAV family analysis does not compare the use of

TRAV subfamily members. In C57BL/6 mice, a portion of theTRAV locus has been triplicated. Herein genes in the supposed“original” [ImMunoGeneTics (IMGT) designation] mouse TRAVlocus are designated “A.” Genes in the second of the triplicationsare designated “D,” and genes in the third triplication are desig-nated “N.” Distinction between these subfamily genes is importantbecause often, but not always, the various subfamily members differin nucleotide and consequently in protein sequence, particularlyin their CDR1 and CDR2 sequences, which are of interest for ourexperiments (9, 19). Thus, each family is designated by a number(e.g., TRAV1, TRAV6, and so forth), and a second number andletter are used to designate the particular subfamily member (e.g.,TRAV6-5A, TRAV6-5D, and TRAV6-5N).We compared TRAV subfamily use for all the 88 TRAVs we

could distinguish with our sequencing by the CD4+ T cells in theWT, βT77A, and βR70A mice. The data for all the individualTRAV subfamily genes are contained in Fig. 3 E and F. TRAVsunderrepresented in the βT77A mice were 7-6A, 7-6D, 7-6N,8-1AD, 8-2D, 12-1AN, 13-2AN. and 14-3A. TRAVs overrepresentedin the βT77A mice were 3-3A, 4-4D, 6-2A, 6-3ADN, 6-4A, 6-4D,6-6N, 6-7A, 9-2D, and 11-1AD. Differences between WT and T77Acells for all the subfamilies and their significance scores can befound in Table S3. Many of these subfamily differences accountfor the differences in overall family use shown in Fig. 3D.

Comparison of the TCRα Repertoire Used by Naive CD4 T Cells fromWT and I-Abβ Mutant Mice. Sequencing identified not only theTRAV families and subfamilies used in the WT and mutant micebut also the complete sequences of the TCRα domains, includingthe TRAVs, TRAJs, and the somatically generated CDR3α re-gions. Thus, we analyzed the diversity of the entire TCRα se-quences among the naive splenic CD4 T cells in WT and mutantmice in several different ways. First, we examined the properties ofthe overall TRAV–CDR3–TRAJ repertoires. Initially, to measurethe richness and diversity in the population, we used a speciesaccumulation curve (39) in which a random sampling of ourpopulation along the x axis is shown on the y axis if each includedsequence adds a unique sequence to the total number of uniquesequences (Fig. 4A). This curve should plateau as the data ap-proach the saturation of all sequences present in the cDNA sample.Sequences of the TCRαs from the three types of mice have similarcurves and do not plateau even after analyzing 500,000 randomizedsequences. Therefore, the naive CD4 T cells in WT, βR70A, andβT77A mice all express similarly large, diverse TCRα repertoires.However, a different type of accumulation curve shows that

this large repertoire is not randomly dispersed, i.e., the frequencyof each sequence is not determined by a simple Poisson distri-bution (Fig. 4B). Despite the lack of saturation, the averagenumber of repeats of any given unique sequence in the sampleswas about five but ranged from 1 to more than 10,000. Using thisfrequency, we constructed a Poisson-predicted accumulationcurve that predicts the proportion of total sequences that should

Silberman et al. PNAS | Published online September 1, 2016 | E5611

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

17,

202

0

accumulate as we added sequences that occur from 1 to 20 times.This curve predicts that, if TCRα use is Poissonian, we shouldaccount for nearly all the sequences by the time we include thosethat occur 15 times or less, but the experimental accumulationcurve generated from the sequencing data shows that these se-quences account for only ∼50% of the total sequences. Likewise,more sequences were found fewer than three times than pre-dicted by the Poisson curve. Similar results were seen with thedata from the mutant mice. Thus, despite the great diversity ofsequences, their frequency was not as predicted by a Poissondistribution, a feature shared with previous repertoire analyses ofdifferent human T-cell populations (40). Some of these resultsmight be attributable to uneven efficiencies during the PCRswith the cDNA templates, but it is likely that both thymic andperipheral selective pressures also contributed.Finally, we examined the differences in overall sequences for

the three types of mice. Two types of analyses were done on thetotal TCRα (combining the TRAV subfamily, TRAJ, and CDR3)sequences. First, in Fig. 4C the overall data from the nine mice arerepresented as a three-component principal component analysis(PCA). PCA is a transformation of the data (in this case expressionvalues for all samples) into a new coordinate system whose axes

(the principal components) are defined by the variability in thedata. By construction, the first principal component is the linearcombination of TCRs that yields the highest variance in expressionlevels between samples. The second principal component is thenthe linear combination of TCRs that yields the highest variance inexpression levels subject to being perpendicular to the first prin-cipal component, and so forth. Often the first several principalcomponents explain the majority of the variance in the data. Onethen can plot the samples along the first few principal componentaxes to visualize high-dimensional expression data in terms of a setof simpler axes that represent the most important features of thesedata. In these plots, clear separation between and clusteringwithin genotype groups indicates that the genotype is drivingrepertoire-wide differences in expression patterns. The WT, βT77A,and βR70Amice clustered well and were separated from each otherfor two of the three components. Particularly well separated weretheWT and βT77A data. As a second analysis we directly comparedthe TRAV–CDR3–TRAJ combinations in the nine mice. Given thevery large number of comparisons being made, the bar for sig-nificance differences was set very high. To reduce the number ofcomparisons, we set a threshold of TCRαs sequenced at least 10times combined in all nine runs. As shown in the heat maps in

Fig. 2. Generation of MHCII mutant mice andcharacterization of the effects of mutation on thy-mic selection. (A) The schematic depicts the ZFN-targeting strategy used to generate MHC mutantmice. The DNA for HDR was designed to target exon3 of either H2-AB (R70A, T77A) or H2-AA (A64Q); thepositions of screening primers are indicated. Thestructure of the DNA for HDR included 1,000 bp ofhomology flanking at either end the ZFN recogni-tion site. Mutations in H2-AB are indicated in red,and mutations in H2-AA are shown in teal. The re-striction sites introduced to allow screening are in-dicated in green. The locations of ZFN recognitionsites are also indicated; these sites were disrupted bythe introduction of a silent mutation in the vector.(B) Splenocytes were stained for MHCII and markersof lymphocyte lineages and were analyzed by flowcytometry. Histograms depict the level of expressionof MHCII on TCRβ− cells in βR70A and βT77A (ma-genta traces), αA64Q (teal trace), and WT (dark grayshaded area) mice. (C and D) Thymocyte compositionis shown for WT, βR70A and βT77A (C), and for WTand αA64Q mice (D) as determined by expression ofCD4 and CD8. Thymocytes undergoing selection areidentified by the expression of CD5 and CD69.(E) Frequency of mature SP4 and SP8 thymocytes inthe indicated strains. (F) Frequency of thymocytesundergoing selection in the different mice. Data inC–F are representative of three or four independentexperiments containing 7–10 mice per group. Errorbars represent SEM; *P < 0.05.

E5612 | www.pnas.org/cgi/doi/10.1073/pnas.1609717113 Silberman et al.

Dow

nloa

ded

by g

uest

on

May

17,

202

0

Fig. 4 D and E, 84 combinations were found to be significantlydifferent between WT and T77A mice. The figure gives a “ge-stalt” view of the data; the complete data for these sequences,including the TRAV, TRAJ, and CDR3 sequences and signifi-cance scores are contained in Table S4.In summary, although both WT and mutant mice develop

large diverse repertoires, significant changes have occurred inTRAV family and subfamily and in TCRα CDR3 sequences toaccommodate the mutations.

Effect of the αA64Q Mutation on the T-Cell TRBV13 Repertoire. Ouranalysis of the thymus in the αA64Q mice showed apparentlyreduced activation from positive selection in DP thymocytesbased on CD5/CD69 expression (Fig. 2 D and F). Because sub-stantial biological and structural data have shown that the sitethat includes αA64Q is often used as a docking site for βY48 ofthe CDR2 loop of the TRBV13-2 Vβ element and perhaps alsothe TRBV13-3 Vβ element (21), we focused our analysis of theeffects of this mutation on the repertoire of T cells using these

elements. We analyzed CD4 SP thymocytes and splenic CD4+

T cells from WT and αA64Q mice with a mAb that discriminatesTRBV13-2 from TRBV13-3 (Fig. 5 A and B). Flow cytometricdata showed a substantial, significant shift in use from TRBV13-2to TRBV13-3 in both populations in the A64Q mutant mice ascompared with WT mice.Next, with a strategy similar to that used in our analysis of the

TCRα repertoire in the βT77A and βR70A mice, we deep se-quenced the TRBV13 domains present in naive CD4+ T cells inthree WT and αA64Q mice. We created a PCR fragment with a5′-primer common to all three members of the TRBV13 family and a3′-primer within Cβ. Fig. 5C shows that the sequence data confirmedthe significant shift from TRBV13-2 to TRBV13-3 in the αA64Qmice, but there was no change in the use of the third family member,TRBV13-1. The heat map in Fig. 5D shows all the TRBV13/TRBJcombinations with an increased frequency in WT samples comparedwith A64Q, and Fig. 5E shows the combinations used more fre-quently in the mutant. The blue squares in Fig. 5 D and E indicatestatistical significance (Table S5), and these combinations group

Fig. 3. Differential use of TRAV families and subfamilies in T77A mutant mice. (A) Representative analysis via flow cytometry of TRAV14 expression in spleenCD4+ T cells fromWT, βR70A, or βT77A mice. (B and C) Frequency of TRAV14+ cells in CD8+ (B) and CD4+ (C) T cells in the spleen of the indicated mice. Data arerepresentative of three or four independent experiments containing 7–10 mice per group. Error bars indicate SEM. (C) Next-generation sequencing of theentire TCRα repertoire expressed in naive CD4+ T cells sorted from the spleen of WT, βR70A, or βT77A mice. (D) Frequencies indicate the proportion of anygiven TRAV family out of all valid sequences. (E and F) Heat maps representing the positive (E) or negative (F) fold change in the use of the individual TRAVsubfamily genes in CD4+ cells from the indicated strains. Green, yellow, and red indicate high, medium, and low use, respectively. Statistical significance (P <0.05) is indicated by an asterisk in C and by blue squares in D–F.

Silberman et al. PNAS | Published online September 1, 2016 | E5613

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

17,

202

0

almost perfectly with TRBV13 subfamily. Furthermore, lookingat the more commonly found TCRs, DESeq2 identifies indi-vidual TCRβs that are differentially expressed in the WT andA64Q mice (Fig. 5E and Table S6). Thus, these data support theprevious findings on a more global scale, with particularTRBV13-2–containing domains associating the TRBV13-2CDR2 loop with docking on the portion of the MHCII β1 helixcontaining with an evolutionary preference for MHCαA64.

Reciprocal T-Cell Recognition of the I-Ab Mutations. Our data clearlypoint to adjustments in the use of particular germ-line TRAVand TRBV elements driven by the mutations in the conservedamino acids on the α1 and β1 helices of I-Ab, but they do notreveal the functional consequences of the overall change in theTCR repertoire. To begin to address this question, we tested how“foreign” the WT and mutant I-Ab molecules appeared to CD4+

T cells from the various mice. We set up one-way mixed lym-phocyte reactions using all combinations of purified CD4+ Tcells and APCs from the WT and mutant mice. T cells and APCsfrom an I-Af mouse (B10.M) were used as a control. The results(Fig. 6) show that the CD4+ T cells did not respond to APCsfrom the same mouse but did respond to APCs from all the othermice, as measured by IL-2 production. The T-cell responses fromthe WT and mutant I-Ab mice were on the same order of mag-nitude as the alloresponses seen with the I-Af T cells and APCs.These results predict that differences in the TCR repertoiresamong the WT and mutant I-Ab mice should be similar to thoseamong mice of different MHCII haplotypes, indicating that thechanges in TCR repertoire had dramatic effects on the specificityand alloreactivity of the T cells. Furthermore, in addition to the

previously demonstrated influence of the varying peptide rep-ertoire among MHC haplotypes (41), these results provide evi-dence for the previously suggested (42) role of the germ-line biasof TCRs for MHC in alloreactivity. Thus, it is likely that both thepeptide repertoire and the germ-line–encoded bias contribute toalloreactivity to some extent.

DiscussionThe roots of our current thinking on the evolutionary conser-vation of interactions between TCR and MHC amino acids camefrom our studies of CD4+ T cells in mice expressing a single fixedpeptide–MHCII complex (18). These mice had impaired negativeselection because of the absence of a diverse set of self-peptidesbound to their MHCII. The T cells thus created reacted stronglyto self-MHC occupied by the normal complement of self-peptidesand also, surprisingly, to many different allo-MHCII alleles. Weconcluded that, although negative selection functions to removehigh-affinity self-specific T cells, in so doing it also eliminates alarge population of highly MHCII cross-reactive T cells.Based on our subsequent functional, mutational, and struc-

tural studies, we concluded that the high cross-reactivity of theseT cells was caused by dominant interactions of certain conservedamino acids in their TRAV and TRBV CDR1 or CDR2 loopswith conserved sites on the MHCII helices. In complexes be-tween the TCRs of T cells from normal mice and their activatingpeptide–MHCII ligands, reported by ourselves (20) and others(8), these conserved interactions were seen often, but they usu-ally were not so dominant. These findings have led us to ourcurrent hypothesis that random combinations of germ-line TCRα- and β-genes create, with high frequency, T cells reactive toMHCII, regardless of allele, and that, to escape negative selec-tion and contribute to the functional peripheral repertoire,T cells must bear TCRs whose somatically generated CDR3shave modulated this tendency away from generic MHC reactivityand toward peptide dependence. The results of direct mutagenesisexperiments by us and others are consistent with this idea.Our purpose in this present study was to determine how mu-

tations in MHCII I-Ab amino acids affect T-cell development andthe peripheral T-cell TCR repertoire. We chose I-Ab βT77 andαA64 for this study for several reasons: They are highly conservedamong MHCII molecules; they have been seen repeatedly as sitesof interaction with certain germ-line TRAV CDR1 and TRBVCDR2 amino acids; their mutation often disrupts the activation ofperipheral antigen-specific T cells in response to antigen; andthey do not participate directly in peptide binding. We choose I-Ab

Rβ70 as a control because it is less conserved and usually interactswith the TCR CDR3 loops.Our results show that none of the mutations prevented the

development of large, diverse peripheral CD4 T-cell populations.However, depending on the mutation, there were significantchanges in thymocyte subpopulations and changes in the pe-ripheral CD4 T-cell TCR repertoire. The subtlest changes wereseen with the βT77A mutation. There were no changes with thismutation in thymic cellularity or in the size of the thymic pop-ulation undergoing selection (CD4+CD69+). However, com-pared with WT mice, the βT77A mutation led to significant shiftsin TRAV family and subfamily use. In addition, this mutation ledto changes in the TRAV–CDR3–TRAJ repertoire, demonstratedby PCA that clearly separated the unique sequences in the mutantmice from WT mice and from each other. Our analysis of theαA64Q mice also showed normal thymic cellularity, but in this casethere was a significant reduction in the activation in thymocytesundergoing selection. In a more abbreviated peripheral repertoireanalysis, we compared the use of TRBV13-2 with the other twomembers of this family. The importance of the intimate interactionof evolutionarily conserved TRBV-Y48 in the CDR2 of TRBV13-2with the portion of MHCII α-chain helix containing A64 has beendocumented in numerous structural, functional, and thymic de-velopmental studies (1, 20, 21). The importance of this amino acidin the other family members is not as clear: It is present, but thereare other differences between the family members in their CDR1

Fig. 4. Differential use of complete TCRα in T77A mutant mice. (A) Species ac-cumulation curves for theWT, βR70A, and βT77A mice. Each curve represents theaverage of the accumulation curves of three mice of the indicated genotype. (B)A Poisson-predicted distribution based on the average number of repeats in anysample is comparedwith the accumulation of sequences based on its frequency inthe sample. (C) PCA showing that each individual run clusters by genotype on thefirst three components. (D and E) TCRα sequences differentially expressed in T77Aand WT cells are expressed in heat maps ordered by positive (D) or negative(E) fold-change. Green, yellow, and red indicate high, medium, and low use,respectively. Statistical significance (P < 0.05) is indicated by blue squares.

E5614 | www.pnas.org/cgi/doi/10.1073/pnas.1609717113 Silberman et al.

Dow

nloa

ded

by g

uest

on

May

17,

202

0

and CDR2 regions. Our analysis showed that TRBV13-2 use byboth thymic and peripheral CD4 T cells is reduced in the mutantmice, with a concomitant rise in TRBV13-3 but no change inTRBV13-1 compared with the WT mice.The results of the present study clearly show that mutation of

either βT77 or αA64 alters the repertoire of developing CD4 Tcells. However, the magnitude of these effects was less thanthose we saw on the response of antigen-primed WT peripheralCD4 T cells to antigenic peptides presented by the mutantMHCII proteins. Likewise, mutation of conserved amino acids inthe CDR2 loop of TRBV13-2 had a much more profound effecton T-cell development than did the αA64Q mutation (21). Theseresults suggest that, during the development of the TCR reper-toire, adjustments not only in TRAV use but likely also in αβpairing and somatically generated CDR3 sequences can largelycompensate for the loss of a single conserved docking site onMHCII. However, once a T-cell has been selected by WTMHCII,it no longer can make these adjustments to the loss of the dockingsite. It also is worth noting that our previous results with mutationsin TRBV13-2 CDR2 were done with a transgenic TCR β-chainwith a fixed CDR3, thus limiting the possible adjustments inrepertoire to changes only in α-chain pairing. With the advent ofpaired TCRαβ sequencing from single cells, a future directioncould be to explore the entire TCRαβ pairs and elucidate anypotential compensation on the opposite TCR chain.It has been suggested that the great deal of latitude seen in the

docking angle of TCRs binding to MHC argues against the ideaof evolutionarily conserved amino acids in TCR–MHC inter-actions. However, the many structures of TCRs that includeTRBV13-2 bound to MHC show that conserved amino acids in

its CDR2 loop unfailingly react with related sites on the MHCIIα1 helix, even in the face of various docking angles of the TCRs.The set of structures available for analysis involving other TRAVand TRBV elements has not been as extensive, so analyses withthe other TRAVs and TRBVs are not currently possible. How-ever, because much of the tops of the MHC helices are con-served, it is possible that individual TRAV or TRBV elementsprefer docking to different conserved sites or can use alternativesto the preferred site.A recent study consistent with this idea comes from the Garcia

laboratory (43). They analyzed the structures of the same TCRbound to the MHCI allele, H2-Ld (Ld), engaged by many peptides.The results showed that, although the TRAV CDR1 and CDR2locations on the Ld α2 helix were very similar in the structures, theTRBV13-1 CDR1 and CDR2 loops had more than one dockingsite on the Ld α1 helix, altering the angle of engagement of theTCR with Ld. Interestingly there were discrete docking positions,not a continuous series. These results establish multiple discrete,conserved sites for TRBV13-1 docking on MHCI, the choice ofwhich is determined by the peptide. Therefore, the single aminoacid mutational approach used here may make it difficult toestablish completely the TRAV or TRBV partners for a par-ticular conserved site on the MHC helices.The results of the current study are not inconsistent with any of

the recent reports that have shown highly unusual MHC dockingmodes by some TCRs and non-MHC ligands for some TCRs. Forexample, natural killer T cells (NKT cells) and mucosal-associatedinvariant T cells (MAIT cells) have nonconventional MHC ligandsthat lack the conserved MHC docking sites (44, 45). The in-variant NKT and MAIT TCRs dock on their ligands in very

Fig. 5. Differential use of TRBV13 subfamilies in A64Q mutant mice. (A and B) Representative staining with mAb MR5-2 distinguishing TRBV13-2 andTRBV13-3 (A) and frequency, represented as a ratio of TRBV13-2 to TRBV13-3 in thymic and splenic CD4+ T cells (B). Data are representative of three in-dependent experiments containing seven mice per group. Error bars indicate SEM. (C) The frequency of the three genes that are members of the TRBV13family as determined by next-generation sequencing. Data are averages of three independent runs. Error bars indicate SEM. (D and E) Heat maps for theTRBV–TRBJ combinations are ordered by positive (D) or negative (E) fold change. (F) Individual TCRβ sequences that are differentially expressed in A64Q andWT cells. Statistical significance (P < 0.05) is indicated by asterisks in B and C and by blue squares in D–F.

Silberman et al. PNAS | Published online September 1, 2016 | E5615

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

17,

202

0

nonconventional ways. These specialized T cells and their ligandsarose evolutionarily after the development the conventional TCR–MHC system. One could consider that they have “hijacked” a partof system for another purpose, much as certain MHC-like mole-cules no longer function as ligands for T cells but have taken on newfunctions over evolutionary time.The set of conventional TCRs that deviate most in the orienta-

tions and location with which they interact with conventional pep-tide–MHC complexes comes primarily from autoreactive T cells.Their footprints on MHC can drift dramatically away from thoseseen with foreign peptide–MHC complexes and, in one case, evenreverse the orientation of the TCR on the ligand (46). These T cellsare the survivors of thymic negative selection and as such may needto venture into these unusual docking modes, not found in thethymus, to improve their affinity to achieve T-cell activation.Experiments aimed at the discovery of T cells that use non-

MHC ligands have turned up T cells that recognize other mole-cules. Most dramatically, one laboratory constructed a mouselacking MHCI, MHCII, CD4, and CD8 and introduced mutationsto uncouple essential downstream TCR-signaling molecules fromessential interactions (47–49). The mice developed a peripheralT-cell repertoire that contains T cells reactive to the surface proteinCD155. The authors conclude that these experiments show that theTCR repertoire need not be MHC dependent and that the usualspecificity for MHC is not inherent in the germ-line sequences ofthe MHC and TRAV/TRBV elements. Rather, they suggest that innormal mice MHC specificity arises by selection from a somaticallygenerated random repertoire of TCRs, yielding TCRs that cansatisfy the MHC-dependent geometry of the many components ofthe large TCR/coreceptor signaling complex.Our experiments do not argue against the generation of T cells

of these non-MHC specificities. In fact, given the recombinational

capacity of the thymus to generate an enormous number of uniqueTRAV and TRBV CDR3 loops, their existence is inevitable.However, if the initial, unselected TCR repertoire is random, thefrequency of T cells specific for any particular protein, such asCD155 or MHC proteins, will be very low. Subsequent culling ofthis scarce MHC-specific repertoire during the nonproliferativephase of T-cell development to make it both self-MHC restrictedand self-MHC tolerant will further reduce its size greatly, makingthe generation of the well-established, very large peripheral T-cellrepertoire very difficult. However, predisposing the preselectionTCR repertoire toward MHC recognition via embedded con-served amino acids in MHC and TCR proteins to promote theirinteraction should separate the “wheat” from the “chaff” duringselection much more efficiently. Evidence presented here and inprevious papers suggests that this idea is, to some extent, correctand that the preselected TCR repertoire is already skewed towardMHC reactivity (9, 15–17, 21, 23).

Materials and MethodsMutant MHC I-Ab α- and β-Chains. Plasmids encoding MHCII I-Ab α- andβ-chains were previously used (29). MHC mutations were cloned by over-lapping primers using engineered restriction sites. The I-Ab α-chain wascloned into a murine stem cell virus (MSCV)-based retroviral plasmid with aninternal ribosome entry site plus Thy1.1 as a reporter. The I-Ab β-chain wascloned into a similar MSCV vector with a GFP reporter. These MSCV vectorswere also available in the J.W.K./P.M. laboratory.

Retroviral Packaging. Retroviral plasmids were cotransfected into Phoenixcells with pCLEco accessory plasmid using Lipofectamine 2000 (Invitrogen)according to the manufacturer’s instructions. Retrovirus-containing super-natants were collected 48–72 h after transfection and were filtered througha 0.45-μm filter to remove cell debris.

MHC-Expressing Cell Lines. MHC constructs were expressed by retroviraltransduction of an APC line, M12.C3. M12.C3 cells are derived from a BALB/CB-cell lymphoma that was selected for loss of I-A expression (27), althoughthey contain a functional I-Ad α-chain. For retroviral infection of M12.C3cells, 105 cells were spin-infected with retroviral supernatants containing8 μg/mL of Polybrene (Sigma-Aldrich) for 90 min at 37 °C. Cells were ex-panded in culture and subsequently were cloned by limiting dilution; clonesof equal MHC expression were chosen.

Preparation and Stimulation of Bulk T-Cell Hybridomas. Antigen-specific T-cellhybridomas were generated by immunizing mice with the desired antigenemulsified in complete Freund’s adjuvant. The para-aortic lymph node cellswere isolated 7 d later, expanded in culture for 3 d with the same antigenwith which the mice had been immunized, and cultured in IL-2 for 5 d. Afterthis in vitro culture, activated T cells were fused to BWα−β−, a variant of thefusion partner BW5147 generated to lack both TCR α- and β-chains (50).

For stimulations, 5 × 104 to 1 × 105 hybridomas were cultured with differentstimuli for 4–24 h in 200 μL culture medium in 96-well microtiter plates. Hy-bridoma responses were measured by CD69 expression and IL-2 production. IL-2ELISAs were done using the anti–IL-2 antibody JES6-1A12 (eBioscience) to cap-ture and the biotinylated antibody clone JES6-5A4 (eBioscience) with streptavi-din conjugated to HRP (Jackson ImmunoResearch) to detect the bound antibody.

MS. WT and mutant I-Ab proteins were immunoprecipitated from lysates ofroughly 109 of the transduced M12.C3 cells using antibody clone Y3P. Pep-tides were eluted in 2.5-M acetic acid and were separated from beads,antibodies, and MHCII molecules by passage through a 10,000-Da cutoffultrafiltration unit (Millipore) and were subjected to MS or MS-MS analysisas previously reported (32). Peptides were analyzed via LC/MS-MS or LC/MSon an Agilent Q-TOF instrument (model 6520) as described in detail in SIMaterials and Methods.

Flow Cytometry. Cells, either ex vivo or hybridomas, were preincubated withsupernatant from the anti-CD16/CD32 producing hybridoma, 2.4G2. Cellswere stained under saturating conditions with antibodies to mouse TCRβ(clone H57-597), CD4 (clone GK1.5), CD8 (clone 53-6.7), CD25 (clone PC61),CD44 (clone IM7), CD5 (clone 53-7.3), CD69 (clone H1.2F3), CD24 (clone M1/69), B220 (clone RA3-6B2), CD11b (clone M1/70), γδ TCR (clone GL3), CD62L(clone MEL-14), Vβ8.x (clone F23.1), Vβ8.2 (clone F23.2), Vβ8.3 (clone 1B3.3),Vβ8.1/2 (clone MR5-2), and Vα2 (B20.1), purchased from eBioscience or BD

Fig. 6. Functional TCR repertoire differences identified by reciprocal T-cellrecognition of the I-Ab mutations. One-way mixed lymphocyte reactions usingall combinations of purified CD4+ T cells and APCs from WT and mutant mice.T cells and APCs from an H-2f haplotype mouse (B10.M) were used as a control.Data shown are from three independent experiments. Error bars indicate SEM.

E5616 | www.pnas.org/cgi/doi/10.1073/pnas.1609717113 Silberman et al.

Dow

nloa

ded

by g

uest

on

May

17,

202

0

Pharmingen or generated in house. Cells were analyzed by flow cytometryon a FACScan, LSR II, or LSRFortessa system (BD Biosciences).

Generation of Knockin MHC Mutant Mice. As described in detail in the SIMaterials and Methods, embryos were isolated from superovulated femalemice (51) and pronuclear injections performed with ZFN mRNA (Sigma-Aldrich) to introduce point mutations in the genes encoding IAb. All animalswere housed and maintained in the Biological Resource Center within NJH inaccordance with the research guidelines of the National Jewish Health In-stitutional Animal Care and Use Committee.

Sequencing of TCR Repertoire. Naive CD4 T cells were stained as describedabove and sorted at the National Jewish Health Flow Cytometry Core Facility.RNAwas isolated using the RNeasy Kit (Invitrogen). cDNAwasmade using the

SuperScript VILO Kit (Invitrogen). Details on the PCRs used to generate thesequencing library and the sequences of the primers can be found in SIMaterials and Methods.

Statistical Analysis of TCR Repertoires. Differential expression analyses wereperformed using the DESeq2 package (v1.8.1) (38) in the R language (v3.2.2)(52). Details of these analyses can be found in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Francis Crawford and Ella Kushner, RandyAnselment and Thomas Danhorn of the National Jewish Health Center forGenes, Environment and Health, Josh Loomis and Shirley Sobus of the Na-tional Jewish Health Cytometry Core for technical assistance, and Dr. GregKirchenbaum for assistance with some of the experiments in this paper. Thiswork was supported by NIH Grants AI-18785 (to P.M.), AI092108 (to L.G.),AI103736 (to L.G.), and T32 AI007405.

1. Yin L, Scott-Browne J, Kappler JW, Gapin L, Marrack P (2012) T cells and their eons-oldobsession with MHC. Immunol Rev 250(1):49–60.

2. Sunyer JO (2012) Evolutionary and functional relationships of B cells from fish andmammals: Insights into their novel roles in phagocytosis and presentation of partic-ulate antigen. Infect Disord Drug Targets 12(3):200–212.

3. Criscitiello MF, et al. (2012) Shark class II invariant chain reveals ancient conservedrelationships with cathepsins and MHC class II. Dev Comp Immunol 36(3):521–533.

4. Cooper MD, Alder MN (2006) The evolution of adaptive immune systems. Cell 124(4):815–822.

5. Babbitt BP, Allen PM, Matsueda G, Haber E, Unanue ER (1985) Binding of immuno-genic peptides to Ia histocompatibility molecules. Nature 317(6035):359–361.

6. Buus S, Sette A, Colon SM, Jenis DM, Grey HM (1986) Isolation and characterization ofantigen-Ia complexes involved in T cell recognition. Cell 47(6):1071–1077.

7. Jerne NK (1971) The somatic generation of immune recognition. Eur J Immunol 1(1):1–9.8. Feng D, Bond CJ, Ely LK, Maynard J, Garcia KC (2007) Structural evidence for a

germline-encoded T cell receptor-major histocompatibility complex interaction ‘co-don’. Nat Immunol 8(9):975–983.

9. Marrack P, Scott-Browne JP, Dai S, Gapin L, Kappler JW (2008) Evolutionarily con-served amino acids that control TCR-MHC interaction. Annu Rev Immunol 26:171–203.

10. Garcia KC, Adams JJ, Feng D, Ely LK (2009) The molecular basis of TCR germline biasfor MHC is surprisingly simple. Nat Immunol 10(2):143–147.

11. Fink PJ, Bevan MJ (1978) H-2 antigens of the thymus determine lymphocyte speci-ficity. J Exp Med 148(3):766–775.

12. Zinkernagel RM, Doherty PC (1974) Restriction of in vitro T cell-mediated cytotoxicityin lymphocytic choriomeningitis within a syngeneic or semiallogeneic system. Nature248(5450):701–702.

13. Kappler JW, Roehm N, Marrack P (1987) T cell tolerance by clonal elimination in thethymus. Cell 49(2):273–280.

14. Kisielow P, Blüthmann H, Staerz UD, Steinmetz M, von Boehmer H (1988) Tolerance inT-cell-receptor transgenic mice involves deletion of nonmature CD4+8+ thymocytes.Nature 333(6175):742–746.

15. Zerrahn J, Held W, Raulet DH (1997) The MHC reactivity of the T cell repertoire priorto positive and negative selection. Cell 88(5):627–636.

16. Merkenschlager M, et al. (1997) How many thymocytes audition for selection? J ExpMed 186(7):1149–1158.

17. McDonald BD, Bunker JJ, Erickson SA, Oh-Hora M, Bendelac A (2015) Crossreactive αβT cell receptors are the predominant targets of thymocyte negative selection. Immunity43(5):859–869.

18. Ignatowicz L, Kappler J, Marrack P (1996) The repertoire of T cells shaped by a singleMHC/peptide ligand. Cell 84(4):521–529.

19. Lefranc MP, et al. (2009) IMGT, the international ImMunoGeneTics information sys-tem. Nucleic Acids Res 37(Database issue):D1006–D1012.

20. Dai S, et al. (2008) Crossreactive T cells spotlight the germline rules for alphabetaT cell-receptor interactions with MHC molecules. Immunity 28(3):324–334.

21. Scott-Browne JP, White J, Kappler JW, Gapin L, Marrack P (2009) Germline-encodedamino acids in the alphabeta T-cell receptor control thymic selection.Nature 458(7241):1043–1046.

22. Rubtsova K, et al. (2009) Many different Vbeta CDR3s can reveal the inherent MHC re-activity of germline-encoded TCR V regions. Proc Natl Acad Sci USA 106(19):7951–7956.

23. Scott-Browne JP, et al. (2011) Evolutionarily conserved features contribute to αβ T cellreceptor specificity. Immunity 35(4):526–535.

24. Davis MM (1985) Molecular genetics of the T cell-receptor beta chain. Annu RevImmunol 3:537–560.

25. Davis MM, Bjorkman PJ (1988) T-cell antigen receptor genes and T-cell recognition.Nature 334(6181):395–402.

26. Rangarajan S, Mariuzza RA (2014) T cell receptor bias for MHC: Co-evolution or co-receptors? Cell Mol Life Sci 71(16):3059–3068.

27. Glimcher LH, McKean DJ, Choi E, Seidman JG (1985) Complex regulation of class IIgene expression: Analysis with class II mutant cell lines. J Immunol 135(5):3542–3550.

28. Cohn LE, et al. (1986) Identification of functional regions on the I-Ab molecule by site-directed mutagenesis. Proc Natl Acad Sci USA 83(3):747–751.

29. Huseby ES, et al. (2005) How the T cell repertoire becomes peptide and MHC specific.Cell 122(2):247–260.

30. Doebele RC, Busch R, Scott HM, Pashine A, Mellins ED (2000) Determination of theHLA-DM interaction site on HLA-DR molecules. Immunity 13(4):517–527.

31. Nakagawa T, et al. (1998) Cathepsin L: Critical role in Ii degradation and CD4 T cellselection in the thymus. Science 280(5362):450–453.

32. Lippolis JD, et al. (2002) Analysis of MHC class II antigen processing by quantitation ofpeptides that constitute nested sets. J Immunol 169(9):5089–5097.

33. Liu X, et al. (2002) Alternate interactions define the binding of peptides to the MHCmolecule IA(b). Proc Natl Acad Sci USA 99(13):8820–8825.

34. Dongre AR, et al. (2001) In vivo MHC class II presentation of cytosolic proteins re-vealed by rapid automated tandem mass spectrometry and functional analyses. Eur JImmunol 31(5):1485–1494.

35. Carbery ID, et al. (2010) Targeted genome modification in mice using zinc-fingernucleases. Genetics 186(2):451–459.

36. Rémy S, et al. (2010) Zinc-finger nucleases: A powerful tool for genetic engineering ofanimals. Transgenic Res 19(3):363–371.

37. Urnov FD, et al. (2005) Highly efficient endogenous human gene correction usingdesigned zinc-finger nucleases. Nature 435(7042):646–651.

38. Love MI, Huber W, Anders S (2014) Moderated estimation of fold change and dis-persion for RNA-seq data with DESeq2. Genome Biol 15(12):550.

39. Fisher RA, Corbet AS, Williams CB (1943) The relation between the number of speciesand the number of individuals in a random sample of an animal population. J AnimEcol 12(1):42–58.

40. Qi Q, et al. (2014) Diversity and clonal selection in the human T-cell repertoire. ProcNatl Acad Sci USA 111(36):13139–13144.

41. Morris GP, Ni PP, Allen PM (2011) Alloreactivity is limited by the endogenous peptiderepertoire. Proc Natl Acad Sci USA 108(9):3695–3700.

42. Bevan MJ (1984) High determinant density may explain the phenomenon of allor-eactivity. Immunol Today 5(5):128–130.

43. Adams JJ, et al. (2016) Structural interplay between germline interactions andadaptive recognition determines the bandwidth of TCR-peptide-MHC cross-reactivity.Nat Immunol 17(1):87–94.

44. Reantragoon R, et al. (2012) Structural insight into MR1-mediated recognition of themucosal associated invariant T cell receptor. J Exp Med 209(4):761–774.

45. Gadola SD, et al. (2006) Structure and binding kinetics of three different humanCD1d-alpha-galactosylceramide-specific T cell receptors. J Exp Med 203(3):699–710.

46. Beringer DX, et al. (2015) T cell receptor reversed polarity recognition of a self-anti-gen major histocompatibility complex. Nat Immunol 16(11):1153–1161.

47. Van Laethem F, et al. (2007) Deletion of CD4 and CD8 coreceptors permits generation ofalphabetaT cells that recognize antigens independently of theMHC. Immunity 27(5):735–750.

48. Tikhonova AN, et al. (2012) αβ T cell receptors that do not undergo major histocom-patibility complex-specific thymic selection possess antibody-like recognition specificities.Immunity 36(1):79–91.

49. Van Laethem F, et al. (2013) Lck availability during thymic selection determines therecognition specificity of the T cell repertoire. Cell 154(6):1326–1341.

50. White J, et al. (1989) Two better cell lines for making hybridomas expressing specificT cell receptors. J Immunol 143(6):1822–1825.

51. Nagy A, Gertsenstein M, Vintersten K, Behringer R (2006) Collecting two-cell tocompacted morula-stage embryos. CSH Protoc 2006(1).

52. Development Core Team (2005) R: A Language and Environment for StatisticalComputing (R Foundation for Statistical Computing, Vienna).

53. Kostic AD, et al. (2012) Genomic analysis identifies association of Fusobacterium withcolorectal carcinoma. Genome Res 22(2):292–298.

54. Decock A, et al. (2012) Genome-wide promoter methylation analysis in neuroblas-toma identifies prognostic methylation biomarkers. Genome Biol 13(10):R95.

55. Taiwo O, et al. (2013) DNA methylation analysis of murine hematopoietic side pop-ulation cells during aging. Epigenetics 8(10):1114–1122.

56. Stark R, Brown G (2013) DiffBind: Differential Binding Analysis of ChIP-Seq Peak Data.Available at https://bioconductor.org/packages/devel/bioc/vignettes/DiffBind/inst/doc/DiffBind.pdf. Accessed August 25.

57. Ross-Innes CS, et al. (2012) Differential oestrogen receptor binding is associated withclinical outcome in breast cancer. Nature 481(7381):389–393.

58. Benjamini Y, Hochberg Y (1995) Controlling the false discovery rate: A practical andpowerful approach to multiple testing. J R Stat Soc B 57(1):289–300.

Silberman et al. PNAS | Published online September 1, 2016 | E5617

IMMUNOLO

GYAND

INFLAMMATION

PNASPL

US

Dow

nloa

ded

by g

uest

on

May

17,

202

0