Positive selection of low responsive, potentially autoreactive T cells induced by high avidity, non-deleting interactions

International Immunology, Vol. 10, No. 7, pp. 999–1008 © 1998 Oxford University Press

Positive selection of low responsive,potentially autoreactive T cells induced byhigh avidity, non-deleting interactions

Ann P. Chidgey and Richard L. Boyd

Department of Pathology and Immunology, Monash Medical School, Commercial Road, Prahran 3181,Australia

Keywords: agonist peptide, antagonist peptide, CD8αα, peptide variants, positive selection

Abstract

Using a novel cell suspension model we investigated the relative abilities of nominal peptide andvariants thereof to modulate de novo positive selection of lymphocytic choriomeningitis virus(LCMV)-specific TCR transgenic T cells. Confirming our earlier findings intermediateconcentrations (10 –7 to 10–5 M) of the nominal agonist peptide, p33, induced CD8 co-receptordown-modulation at the level of the entire receptor and the CD8 β chain as a consequence of highbut non-deleting signal interactions. Agonist peptide variants caused down-modulation of the CD8 βchain but to a lesser degree. An antagonist peptide capable of inducing positive selection did notcause such modifications of the co-receptor. The positively selected TCR hiCD8αα and TCRhiCD8–

cells were functional but not as efficient as TCR hiCD8αβ cells, presumably due to lower avidityinteractions in the absence of the CD8 β chain or entire co-receptor. CD8 β mRNA was absent inthese cells and was not up-regulated when further stimulated with fresh antigen-presenting cellspulsed with 10 –5 M p33. Effectively our data suggest that it is not the agonist or antagonist natureof a peptide per se but the overall strength of signalling that determines whether a cell will bepositively or negatively selected, or die by neglect. Furthermore the agonist/antagonist propertiesof peptides defined at the level of mature T cell function do not unequivocally predict their effecton positive/negative selection. The ability of the T cell to down-modulate its CD8 co-receptor inresponse to high but non-deleting peptide interactions during positive selection allows the survivalof T cells with a broader range of affinities and represents a possible mechanism by which lowresponsive but potentially autoreactive cells may escape into the periphery.

Introduction

In mainstream thymopoiesis, positive selection of CD41CD81

cells to the CD8 lineage minimally requires the engagementof the TCR with MHC class I–peptide expressed on thymicepithelial cells. The precise role of peptide, however, is stillunclear with current theories being that a variety of selfpeptides, not exclusively related to the antigenic peptide,may be involved in the positive selection of a particular T cell(1,2). This is not unexpected given that the TCR is notexquisitely specific, being able to recognize a variety of MHC–peptide ligands, albeit with varying affinities (3,4). In supportof this, our model for positive selection using adult stromaand P14 TCR transgenic mice specific to the lymphocyticchoriomeningitis virus (LCMV) has shown positive selectionoccurring in the absence of nominal peptide—presumably byendogenous peptide(s) (5). These cells, fewer by comparison

Correspondence to: A. Chidgey

Transmitting editor: H. R. MacDonald Received 8 December 1997, accepted 31 March 1998

to selection occurring in the presence of low (10–10 M)concentrations of the p33 peptide, were phenotypically andfunctionally mature, exhibiting cytotoxicity and proliferationwhen challenged with the nominal p33 peptide.

T cells are able to interact via their TCR to a variety ofMHC–peptide ligands, with peptides ranging from singleamino acid variations of the agonist peptide to peptidesshowing little homology (3,6). These peptide variants, oftentermed altered peptide ligands, induce only partial activationof the T cell and can have a varied impact upon peripheralT cell response (7,8). How such peptides differentially affectthe TCR interaction in mature T cells is still unclear and evenmore so in immature thymocytes. It may be of a qualitativenature with agonist and antagonist peptides inducing differentdownstream signalling events in mature T cells (8–10). It may

1000 Influence of peptide variants on CD8β co-receptor down-modulation

be more quantitative, with antagonist peptides having fasterdissociation rates than agonist peptides (11,12), which in turnaffects their ability to induce a T cell response. Neverthelesssuch antagonist peptides are capable of inducing andblocking both positive and negative selection (13–16). Thus,the influence of antagonist peptides or altered peptide ligandson thymic selection remains inconclusive.

Regardless of the role of peptide, the TCR affinity and theavidity of T cell–thymic epithelial cell interactions are likely tobe important in determining whether a cell will undergopositive or negative selection. Herein lies the influence ofaccessory molecules such as the co-receptor, among others,which may influence thymic selection. The exact role of theCD8 co-receptor in positive selection of MHC class I-restrictedT cells, however, has yet to be clearly defined; whether itmerely contributes to the avidity of the TCR–MHC–peptideinteractions or whether its importance lies in intracellularsignalling, or both. In the absence of the CD8α chain,functional cytotoxic T cells do not develop (17,18) althoughmore recently it has been shown that positive selection ofCD8 T cells can be rescued in CD8α knockout mice transgenicfor the OT-1 TCR and for P14 TCR respectively, by increasingthe affinity of the TCR–MHC–peptide interactions (19,20). TheCD8β chain improves the efficiency of positive selection (21–23) though is not obligatory for the antigen-specific responsein mature CD8 cytotoxic T cells (21,24). Whilst the co-receptorincreases the affinity of the TCR–MHC–peptide interactionswith CD8αβ heterodimers being more efficient in this respectthan CD8αα homodimers (25) possibly due to a faster ‘on’rate (26), its configuration is also thought to be subject tomodulation through TCR–ligand interactions (27).

This apparent flexibility of co-receptor usage may be import-ant in the developmental fate of the T cell. We have shownrecently that during the process of positive selection, T cellscan modify the threshold for positive selection by altering theexpression of the CD8 co-receptor both quantitatively andqualitatively in response to high but non-deleting concentra-tions of agonist peptide (5). The down-regulation at the levelof the entire co-receptor or just the CD8β chain may allow abroader range of TCR affinities for a particular clonotype tobe positively selected. It may of course also allow the escapeof thymic tolerance of potentially autoreactive T cells.

This present paper addresses further the down-regulationof the CD8 co-receptor and CD8β chain in particular duringpositive selection, both in terms of the functional efficiencyof these subsets and the existence of similar co-receptormodulation in the presence of peptide variants of both agonistand antagonist nature. We have found that reduced signallinginteractions as in the case with low concentrations of nominalagonist peptide and intermediate to low concentrations ofpeptide variants will positively select CD8hi heterodimers(CD8αβ) in this LCMV transgenic system. However, strongsignalling interactions as in the case with intermediate to highconcentrations of the nominal agonist and moderate agonistvariant can induce co-receptor down-modulation, with theresultant positively selected cells being of the TCRhiCD8αβint,TCRhiCD8α homodimer or TCRhiCD8– phenotype. These cellsdo not up-regulate their CD8β chain or CD8 co-receptorupon further stimulation and hence represent a potentiallyautoreactive but low respondant population of cells that, within

Table 1. Peptides

Name Sequence Function Reference

A4Y KAVANFATM moderate agonist 29S4Y KAVSNFATM strong antagonist 29C4Y KAVCNFATM moderate antagonist Pam Ohashi (pers.

Commun.)G4Y KAVGNFATM weak antagonist 30,31AV SGPSNTPPEI adenovirus peptide 29

a normal healthy environment, would not be activated. Thesedata are consistent with the proposal that the overall strengthof signalling determines whether a cell will undergo positiveor negative selection, or die by neglect in the absence of astrong enough signal, rather than the inherent agonistic orantagonistic properties of the selecting peptide.

Methods

Mice

P14 αβ TCR transgenic mice specific to the LCMV, H-2Db

restricted (28), were bred onto a non-selecting background(H-2d) and were maintained in the Monash University animalhouse. Ly 5.1 congenic C57Bl6 mice were maintained in theMonash University animal house.

Peptides

The nominal LCMV agonist peptide, p33 (amino acids 33–41; KAVYNFATM), has the original cysteine at anchor position41 in the wild-type LCMV peptide replaced with a methionineto prevent dimer formation. The peptides listed in Table 1were a kind gift from Pamela Ohashi.

All peptides were synthesized at the Amgen Institute(Boulder, CO) by a solid-phase method using the Fmoc/tBu-based protocol on an ABI-431 instrument. The crude productswere purified on a reverse-phase preparative HPLC column(C4; Vydac, Hesperia, CA).

Co-cultures

The co-cultures were set up essentially as described inChidgey and Boyd (5). Briefly, thymic stromal cells from Ly5.1 congenic C57Bl6 mice (H-2b) were freshly prepared bygentle enzymic digestion (0.15% collagenase/0.1% DNase;Boehringer Mannheim, Mannheim, Germany) of lymphocyte-depleted thymuses and enriched to a lymphocyte to stromalcell ratio of ~2:1 by elutriation. All cells were resuspended inClicks medium (Gibco/BRL, Gaithersburg, MD) supplementedwith 10% heat-inactivated FCS (Gibco/BRL), 2m M L-glutamine(Flow, Irvine, UK), 0.05% benzyl penicillia (CSL, Melbourne,Australia), 0.05% streptomycin (Sigma, St Louis, MO) and5310–5 M 2-mercaptoethanol (Sigma) at a concentration of6.73105 cells/ml, and pulsed with peptide. E17 thymocytesfrom P14 αβ TCR transgenic mice (specific to the lymphocyticchoriomeningitis virus; H-2b restricted) (28) on a non-selectingH-2d background were prepared at a concentration of3.43106 cells/ml and mixed with the stromal cells at a ratioof 5:1. Cell suspensions were co-cultured as hanging drops

Influence of peptide variants on CD8β co-receptor down-modulation 1001

in inverted Terasaki plates at 37°C, 5% CO2 and harvestedgenerally at day 4. P14 transgenic thymocytes were distin-guished from stromal-associated lymphocytes by the expres-sion of Ly 5.2. Cells were generally stained with Ly 5.2–FITC(PharMingen, San Diego, CA), CD8–Biotin/TriColor (Phar-Mingen) and CD4–phycoerythrin (PE; PharMingen) andanalysed by flow cytometry.

Flow cytometry

Cells were analysed using a FACScan flow cytometer (Becton-Dickinson, Mountain View, CA) and Lysys II software (BectonDickinson). Cells were sorted on a FACStar Plus (BectonDickinson) using Lysys II software. Dead cells were excludedby a cell viability gate using forward versus side scatter.The FACS machines were calibrated with total thymocytesunstained and stained with Ly5.2–FITC, CD4–PE and CD8–biotin/TriColor.

Stimulation assay

Co-cultures of P14 H-2d transgenic thymocytes and freshlypurified H-2b stroma in the presence or absence of peptidewere prepared and analysed as described above; however,the purified stroma was irradiated at 3000 rad prior to co-culture to prevent proliferation of stroma-associated lympho-cytes. At day 4 the cells were counted and percentage ofviable transgenic cells determined by flow cytometry using aFACScan (Becton Dickinson). Alternatively, cells were stainedwith CD8α–PE (PharMingen) and CD8β–biotin (PharMingen)followed by streptavidin–FITC (PharMingen), and sorted on aFACStar Plus (Becton Dickinson). Dead cells were gated outusing a FSC/SSC-defined cell viability gate. The cells wereplaced into round-bottom wells of a 96-well plate (53104

transgenic cells/well) with IL-2 (25 units) and 53105 irradiated(3000 rad) H-2b splenocytes which had been pre-pulsed with10–5 M p33 peptide, unless specified differently, for a periodof 48 h.

Functional assays

For the proliferation assay, [3H]thymidine was added 14 hprior to harvesting and incorporated radioactivity measuredon a β-counter. For the cytotoxicity assay, target cells (EL4cell line) were incubated with 51Cr, with or without 10–5 M p33peptide, for 45 min. Effector cells were counted and incubatedwith target cells for 4 h, the supernatant removed and 51Crrelease measured on a γ-counter. Percentage specific lysisof target cells was calculated by: (c.p.m. test – c.p.m. minimumrelease)/ (c.p.m. maximum release – c.p.m. minimumrelease)3100.

RNA extraction and mRNA detection by PCR

RNA was extracted essentially by the method described inChomczynski and Sacchi (32). In brief, ~104 to 105 CD8αβ,CD8αα, CD8– cells were washed in HBSS, pelleted inEppendorf tubes and resuspended in GTC solution (4 Mguanidinium isothiocyanate, 0.25m M tri-sodium citrate, 0.1 Mβ-mercaptoethanol and 0.5% sodium lauryl sarkosinate). Thiswas followed by the addition of acidic phenol then chloro-form:isoamly alcohol and incubated for 30 min on ice. Aftercentrifuging, RNA was precipitated using isopropanol,washed in 70% ethanol and used for single-stranded cDNA

synthesis using an oligo(dT)15 primer. The cDNA was usedin PCR amplification with oligonucleotides recognizing CD8βsequences (33). The specific CD8β primers used were 59-CAAGATGCAGCCATGGCTCT-39 and 59-CGCACACAGTAAA-AGTAGAC-39. Amplification was carried out for 30 cycles withan annealing temperature of 55°C. The PCR products wereanalysed by electrophoresis on a 1% agarose gel.

Results

CD81 positively selected cells show an inverse correlationbetween the concentration of agonist peptide and theirfunctional efficiency

Using an adult stromal cell suspension model for in vitropositive selection we have previously found that transgenicTCRhiCD4–CD8αβhi cells can be selected by endogenouspeptide(s) expressed on thymic stromal cells (5,34). Thesecells can proliferate and specifically lyse target cells loadedwith the nominal p33 peptide. More efficient positive selection,in terms of absolute numbers of TCRhiCD8αβhi T cells,occurred in the presence of low concentrations of p33(10–10 M); these peptide-selected cells were also able toefficiently lyse p33-loaded target cells (5). However, in thepresence of higher concentrations of nominal peptide (10–8

to 10–5 M), fewer CD81 cells were selected and they lysedp33-loaded target cells with decreased efficiency (Fig. 1).This was also reflected in their reduced ability to proliferatewhen challenged with p33-loaded antigen-presenting cells(APC) (Fig. 2). This compromised functional efficiency correl-ated with increasing concentrations of agonist peptide in theco-cultures. In addition there was a progressive shift fromCD8αβ to CD8αα to CD8– cells.

When subsets of CD81 cells induced in the positive selec-tion co-cultures were sorted and stimulated with fresh APCloaded with the nominal agonist peptide, p33 (10–5 M), theCD8αβ cells clearly showed a greater proliferative capacityin comparison to the CD8αα subset (Fig. 3A); this was trueover several concentrations of peptide loaded onto the APC(Fig. 3B). Similarly, TCRint/hiCD4–CD8– cells consistently gavea similar or lower proliferative response than the CD8αα cells(Fig. 3A).

The CD8β chain is down-regulated at the level of the mRNAand is not reversible upon further stimulation with thenominal peptide

We have previously shown that the CD8αα and CD8– cellswere derived from CD41CD8αβ1 precursors (5). To determinewhether this down-regulation of the CD8 co-receptor chainwas solely at the level of surface expression, CD8– andCD8αα cells were sorted post-positive selection and mRNApurified. There was clearly no CD8β message (Fig. 4). Further-more the CD8– and CD8αα T cells do not up-regulate theirCD8β chain surface expression after being stimulated withp33-loaded fresh APC (data not shown).

Not all LCMV single amino acid peptide variants canpositively select

Peptide variants of the nominal agonist peptide were testedfor their ability to increase the efficiency of positive selection.


Fig. 1. (A) TCRhiCD4–CD81 cells which were positively selected in the absence of exogenous peptide are able to specifically lyse p33-loadedtarget cells. E17 TCR transgenic T cell precursors (non-selecting background) were cultured alone or co-cultured with freshly purified andirradiated thymic stromal cells (TSC) expressing the selecting MHC (H-2b) for 4 days in the absence of exogenous peptide. These were furtherstimulated for 48 h with p33 (10–5 M)-loaded fresh APC (irradiated) before testing for their ability to specifically lyse p33 target cells.(B) Reduced ability to specifically lyse target cells is evident when T cell precursors are co-cultured with thymic stroma in the presence ofincreasing concentrations of the nominal p33 peptide. E17 TCR transgenic T cell precursors (non-selecting background) were co-culturedwith freshly purified and irradiated TSC (H-2b) for 4 days in the presence of 10–10, 10–8 and 10–5 M p33 peptide respectively, then furtherstimulated for 48 h with p33 (10–5 M)-loaded fresh APC (irradiated). Transgenic T cell numbers were standardized prior to stimulation andagain prior to setting up the cytotoxic T lymphocyte assays, such that each culture contained the same number of transgenic cells. Theseresults are representative of three separate experiments.

These peptides have been described using various in vitroassays as having agonist or antagonist properties. The pep-tide variant A4Y (KAVANFATM) is a moderate agonist with noantagonist properties (29), S4Y (KAVSNFATM) is a strongantagonist with no agonist properties (29,30), C4Y (KAVCNF-ATM) is a moderate antagonist (Pam Ohashi, pers. commun.)and G4Y (KAVGNFATM) is a weak antagonist with no agonistproperties (30,31). The H-2b binding adenovirus peptide(SGPSNTPPEI) (29) was used as a negative control.

The A4Y moderate agonist variant was able to inducepositive selection to a greater degree compared to the nominalp33 peptide, consistent with the findings of Sebzda et al.(29). In terms of absolute cell numbers (Fig. 5A) the A4Y

peptide at 10–7 M was able to induce a 2- to 3-fold increasein CD8hi cell number above background levels and a 1.5-foldincrease above that induced by the nominal p33 agonistpeptide. At this concentration the nominal peptide inducessome negative selection, particularly in the early stages ofco-culture, as illustrated in kinetics studies published in aprevious paper (5). In the presence of antagonist peptides,only C4Y substantially increased the efficiency of positiveselection at 10–7 M, inducing a 3-fold increase in CD8hi cellnumbers above background. The S4Y and G4Y variants hadno significant effect on positive selection nor had the unrelatedcontrol peptide AV. Interestingly, the strong antagonist S4Ypeptide variant did not antagonize the TCR interactions


Fig. 2. T cells positively selected in the presence of intermediate tohigh concentrations of nominal p33 peptide have a reduced abilityto proliferate when challenged with p33-loaded APC. Transgenic Tcells from co-cultures containing no exogenous peptide or lowconcentrations of the nominal p33 peptide (10–10 M) were functionallyefficient. T cells from co-cultures containing only transgenic cellsdid not proliferate. Proliferation is measured by incorporation of[3H]thymidine and indicated on the vertical axis (c.p.m.); theconcentration of nominal p33 peptide present in the positive selectionco-cultures is indicated on the horizontal axis. Cell numbers fromeach co-culture were standardized for the stimulation assay. Theseresults are representative of three separate experiments.

Fig. 3. CD8αβ cells are functionally more efficient in terms of proliferation when challenged with the nominal p33 peptide in comparison toCD8αα or CD8– cells. P14 transgenic precursors (E17; non-selecting MHC background) were co-cultured in the presence of selecting thymicstroma and 10–10 or 10–7 M of the nominal p33 peptide respectively. Post co-culture cells were sorted by FACS into subsets determined bytheir CD8α and CD8β expression and stimulated for 48 h in the presence of fresh APC loaded with (A) 10–5 M nominal p33 peptide or(B) 10–5, 10–7, 10–9 and 10–11 M nominal p33 peptide respectively. As a negative control, T cells from cultures containing only transgenic Tcell precursors (H-2d) which remained at the double-positive stage of development were used. [3H]Thymidine was added and incorporatedradioactivity measured on a β-counter 14 h later. These results are representative of three separate experiments.

involving endogenous peptide, with positive selection withthis peptide (21%) being similar to co-cultures containing noexogenous peptide; under the same conditions A4Y peptideincreased selection to ~30% (Fig. 5B).

CD8 T cells selected in the presence of antagonist peptidevariants do not down-regulate their CD8β chain

CD41CD8αβ1 transgenic T cells co-cultured with H-2b

expressing fresh thymic stromal cells loaded with high butnon-deleting concentrations of nominal peptide (10–7 M)induce the positive selection of predominantly CD8ααhi cells(Fig. 6). The A4Y agonist peptide variant at the same concen-tration induced a similar but reduced population of CD8ααhi

cells. Also present was a significantly greater population ofCD8αβint cells: 13% compared to 3% in the presence of thenominal p33 peptide and fewer CD8– cells: 38% comparedto 51%, consistent with the reduced agonist properties of thispeptide variant (data not shown).

The C4Y antagonist peptide at 10–7 M induced the positiveselection of predominantly CD8αβhi T cells which reflectsits reduced signalling properties compared to the agonistpeptides. In the presence of the S4Y and G4Y antagonistpeptide variants and the unrelated adenovirus control peptideAV, which did not enhance positive selection, cells of theCD8αβ phenotype were found at similar levels to that foundin co-cultures containing no exogenous peptide. These resultsare consistent with our model for positive selection describingthe quantitative and qualitative down-modulation of the CD8co-receptor with increasing avidity interactions/signallingwithin the window of positive selection (5).


Fig. 4. Down-regulation of the CD8 β chain in positively selectedCD8αα T cells is at the level of the mRNA. P14 transgenic T cellprecursors (E17; non-selecting MHC background) were cultured inthe presence of freshly purified thymic stroma (irradiated) and10–7 M nominal p33 peptide. After 4 days co-culture, transgenicCD8αα and CD8– cell populations were sorted on a FACStar Pluscell sorter. Cells from a CD8 cell line (B64.1) were used as a positivecontrol. mRNA was extracted from all cell types and used as templatesfor cDNA synthesis. The resultant cDNA from each cell type wasused for the PCR reaction to amplify transcripts from CD8β genes.Amplification products were analysed on an agarose gel. The arrowdenotes the 603 bp PCR product of CD8β-specific oligos. Lanes Aand B represent CD8αα cells, lanes C and D represent CD8–

cells, lane E represents CD8αβ cells and lane F represents thenegative control.

Functional capacity of CD81 T cells selected in the presenceof LCMV peptide variants

The proliferative and cytolytic capacities of cells selected inthe presence of LCMV peptide variants when further chal-lenged with fresh APC loaded with the nominal p33 peptideare shown in Fig. 7(A and B). Cells selected in the absenceof exogenous peptide or the presence of low concentrationsof the nominal peptide, predominantly of the CD8αβ pheno-type, are able to efficiently and specifically lyse p33-loaded(10–5 M) target cells and proliferate in response to p33-loaded(10–5 M) fresh APC. CD81 T cells selected in the presenceof the p33 peptide at relatively high concentrations (10–7 M),which are predominantly of the CD8ααhi cells, show a muchreduced capacity, cell for cell, to specifically lyse target cellsand proliferate. In comparison CD81 T cells selected in thepresence of the LCMV agonist peptide variant A4Y are betterable to specifically lyse target cells, however not as efficientlyas those selected at low concentrations of nominal peptide(10–10 M) or by the endogenous peptide. Presumably this isdue to the greater numbers of CD8αβint cells selected withthis peptide which appear to have a greater functional capa-city than the CD8ααhi cells.

The predominantly CD8αβ cells selected in the presenceof the moderate antagonist peptide variant C4Y show afunctional capacity similar to those cells selected in thepresence of low concentrations of nominal peptide (10–10 M)

or endogenous peptide, both in terms of specific lysis andproliferation.

Discussion

We have found previously that positive selection of TCRhiCD8hi

T cells from TCR transgenic mice specific to the nominalLCMV agonist peptide, p33, can occur in the presence ofendogenous peptide(s), albeit inefficiently. This is consistentwith recent publications (1,2) suggesting that peptides notnecessarily related to the nominal peptide are able to positivelyselect T cells. The efficiency of positive selection, however,could be increased in the presence of low concentrations(10–10 M) of the nominal peptide, as reflected by the increasednumbers of TCRhiCD4–CD8αβhi T cells induced (5,35). Cellsselected in the presence of high but non-deleting doses ofagonist peptide were shown to be predominantly of theTCRhiCD8ααhi and TCRhiCD8– phenotype, these cells havingescaped deletion by down-regulating their CD8β chain orentire CD8 co-receptor (5).

The small population of cells selected by endogenouspeptide or low concentrations of the nominal antigenic peptidep33 (10–10 M) were functionally responsive both in terms ofproliferation when challenged with APC loaded with p33(10–5 M) and in their capacity to specifically lyse p33-loadedtarget cells. In the presence of increasing concentrations ofthe p33 peptide during the process of positive selection,the positively selected T cells become correspondingly lessfunctionally efficient. This is most likely due to the increasedpresence of positively selected CD8– and CD8αα cells which,by modulating their co-receptor, were able to escape deletion.This was more apparent when these cells were sorted intoindividual subsets post co-culture with the CD8– and CD8ααphenotypes showing reduced proliferative capacity in com-parison to CD8αβ cells. In these subsets the CD8β chain wasdown-regulated at the level of the mRNA and was not reversedwhen activated with fresh APC loaded with the nominal p33peptide. This reduced functional capacity of CD8αα T cellswas consistent over a range of peptide concentrations pre-sented by APC. Although they did not reach the functionalcapacity of CD8αβ T cells when challenged with APC loadedwith high concentrations (10–5 M) of p33, CD8αα T cells mayhave important relevance in autoimmune disease states whenantigen is presented in abnormally high concentrations,enough to activate these normally low responsive T cells. Thecorrelation of reduced functional efficiency with an increasedpercentage of cells of the CD8– and CD8αα phenotype isconsistent with the findings that CD8α homodimers bind theMHC with lower affinity than CD8αβ heterodimers (25,26).

Given the controversial role of agonist and antagonistpeptide variants in positive selection, we investigated theirability to positively select in the LCMV system and theirinfluence on CD8 co-receptor and CD8β chain down-modula-tion. The peptide variant A4Y has a single alanine substitutionat position 4, which is thought to effect a TCR contactresidue and has been described as having moderate agonistcharacteristics and no antagonist properties (29). Positiveselection was enhanced in the presence of A4Y (10–7 M) witha 2- to 3-fold increase in TCRhiCD8hi T cells compared to thoseselected in the absence of exogenous peptide, consistent with


Fig. 5. Positive selection is enhanced in the presence of 10–7 M of the nominal peptide p33, the moderate agonist A4Y and the moderateantagonist C4Y. (A) Transgenic CD8hi cell numbers recovered from these co-cultures were standardized as number per million of the originalE17 transgenic cells placed into culture and are indicated on the vertical axis. The types of peptide present in the co-cultures are indicatedon the horizontal axis. Cell numbers represent an average of three separate experiments. (B) P14 transgenic T cell precursors (E17; non-selecting MHC background) were co-cultured in the absence or presence of freshly purified thymic stroma and the nominal p33 peptide orpeptide variants thereof (10–7 M) with the adenovirus peptide (AV) used as a control. Typical CD4 versus CD8α profiles of the transgenic cellsafter 4 days co-culture are illustrated.

the findings of Sebzda et al. (29). This was above that ofthe nominal p33 peptide at the same concentration, whichenhanced TCRhiCD8hi cell numbers only by ~1.5-fold, mostlikely because the p33 peptide at this concentration alsoinduces negative selection. CD8 T cells selected in thepresence of A4Y are functional both in terms of proliferationand cytotoxicity when challenged with the 10–5 M nominalpeptide p33. They do, however, show a slightly reducedcytolytic capacity compared to those cells selected in thepresence of low concentrations of the agonist peptide(10–10 M) or endogenous peptide. This illustrates the impactpeptide variants can have on different T cell effector functions,consistent with previous findings (7,8). Regardless, the T cellsselected in the presence of A4Y have a greater functionalefficiency in comparison to those cells selected in the pres-ence of similar concentrations of the nominal p33 peptide(10–7 M). This difference in functional capacity correlates withthe phenotype of the CD8 cells positively selected. The A4Ypeptide was found to induce down-modulation of the CD8co-receptor and CD8β chain, as with the nominal peptide,but to a lesser degree. We found an increased percentageof CD8αβint cells selected in the presence of this moderateagonist. Also the ratio of CD8ααhi to CD8αβhi cells was lessin the presence of A4Y in comparison to cells selected in thepresence of the nominal p33 peptide at the same concentra-tion (10–7 M), these being predominantly CD8αα or CD8–

cells. Thus the increased presence of CD8αβint cells appearsto improve the functional capacity of the selected T cells

Fig. 6. CD8 cells positively selected in the presence of intermediateconcentrations (10–7 M) of the nominal agonist peptide (p33) showa high CD8α to CD8β ratio in the CD8hi cells selected. This ratioexists but to a lesser extent in CD8hi cells selected in the presenceof the moderate agonist variant (A4Y). CD8hi cells selected in thepresence of the moderate antagonist C4Y or no exogenous peptideare predominantly CD8αβ cells.

compared to CD8ααhi cells but not to the same degree asCD8αβhi cells.

Positive selection was also enhanced in the presence of


Fig. 7. CD8 cells positively selected either in the absence of exogenous peptide or low concentration of the nominal peptide (10–10 M) orpresence of 10–7 M of p33 and peptide variants A4Y and C4Y are able to (A) proliferate when challenged with p33 (10–5 M) loaded APC and(B) specifically lyse target cells loaded with p33 (10–5 M). CD8 cells selected in the presence of 10–7 M p33 show reduced functional abilityboth in terms of proliferation and cytotoxicity. These results are representative of three separate experiments.

the moderate antagonist peptide C4Y which has a singlecysteine substitution at position 4. The T cells selected werepredominantly of the CD8αβhi phenotype and showed strongfunctional responsiveness. This is not surprising in the lightof evidence that antagonist peptides have faster dissociationrates than agonist peptides (11). The overall reduced level ofsignalling produced by this TCR–MHC–antagonist peptideinteraction does not appear to impact upon the modificationof the co-receptors as do the stronger TCR–MHC–agonistpeptide interactions. Whether this is due primarily to thereduced avidity of the interactions and subsequent reducedstrength of overall signalling, to an alteration in the signallingpathway itself due to properties of the antagonist peptide ordue to conformation changes induced in the TCR has not yetbeen determined. However, the outcome is consistent withour model suggesting that T cells, in response to the strengthof their TCR–ligand interactions occurring within the windowof positive selection, can modulate their co-receptors to avoiddeletion. In the case of this antagonist peptide, the strengthof signalling was not great enough to induce deletion nordown-modulation of the CD8 co-receptor or CD8β chain.

The other peptide variants tested, S4Y, a strong antagonistand G4Y, a weak antagonist, did not show any significantinfluence on positive selection. Interestingly the strongantagonist S4Y did not significantly reduce background selec-tion, with similar CD8 cell numbers selected as those in theco-cultures containing no exogenous peptide. This may bedue to the endogenous peptide having a greater affinity tothe MHC than S4Y; it may be due to the degradation of S4Yin culture with some CD41CD81 cells not coming into contactwith this peptide or the S4Y peptide may have induced

incomplete TCR signalling which did induce some positiveselection but was not great enough to enhance positiveselection beyond that of background levels. An alternativepossibility is that the CD41CD81 cells which express multipleTCR may be exposed to a number of TCR–peptide–MHCinteractions within the window of positive selection and it isthe sum of these interactions/signals which determines theoutcome of whether a cell will be positively selected, assuggested in a review by Williams et al. (36).

Effectively our data suggest that it is not the agonist orantagonist nature of a peptide per se but the overall strengthof signalling that determines whether a cell will be positivelyor negatively selected, or die by neglect. Furthermore, theagonist/antagonist properties of peptides defined at the levelof affecting mature T cell function do not unequivocally predicttheir effect on positive/negative selection. Some antagonistpeptides might inhibit T cell function at the level of MHCcompetition and thus are unlikely to directly affect positiveselection, whereas other antagonists might negatively signalthrough the TCR—these could influence positive selection.The varied results found in different TCR transgenic systemscan be reconciled by the avidity model for positive selection.In the LCMV TCR transgenic model, low concentrations ofnominal peptide and of certain peptide variants both of anagonist and antagonist nature induce low signalling/low avidityinteractions and result in positive selection. High concentra-tions of nominal peptide induce strong signalling/high avidityinteractions and result in negative selection. In the F5 TCRtransgenic model, it was found that strong antagonist peptideswere able to block negative selection (16). This suggests thatthe antagonist peptide used was acting at the level of the


MHC–peptide ligand, blocking any signalling through theTCR, as by definition they block activation of mature cells inthe periphery. The antagonist peptides that were able toinduce positive selection in OT-1 TCR transgenic mice specificto the OVA peptide, consisted of a mixture of antagonist andpartial agonists (13). The strong antagonist used may beacting at the level of downstream TCR signalling by transduc-ing only a partial signal (8,9) or a similar effect may resultfrom a fast on–off rate in binding to the MHC or TCR (11,12).The end result would be a low signalling/reduced avidityinteraction, hence inducing positive selection. The nominalagonist peptide in this system delivered only a deleting signaleven at low concentrations (37), presumably due to thestrength of signalling and higher affinity TCR compared tothat in the LCMV system. A recent paper by Girao et al. (38)has suggested that only weak agonist peptides could inducethe positive selection of functional CD81 cells in fetal thymicorgan cultures from P14 TCR transgenic mice on a TAP-1–/–background. In their system high concentrations of peptidedid induce some positive selection of CD81 cells; however,they suggest that these cells were not functional due to somepartial negative selection. We would suggest that their resultsare consistent with our data, that high but non-deletingconcentrations of agonist peptide do positively select cellsbut induce the down-modulation of the CD8 co-receptorand/or CD8β chain which translates into reduced functionalcapacity of these T cells. These latter two alternatives needto be tested in their system.

Thus it is the sum of the TCR–ligand interactions and otherco-stimulatory molecules which we believe is the determiningfactor for positive selection rather than the intrinsic propertiesof an agonist or antagonist peptide. This is supported by theability of a T cell to alter its CD8 co-receptor expression toavoid deletion by reducing the overall signalling strength ofthe TCR–ligand interaction. These interactions, important indetermining the fate of a developing T cell, are optimal in thepresence of peptides specific to the TCR, with particularpeptide variants similarly able to induce positive selection.Cortical epithelial cells in the thymus are exposed to a vastarray of self peptides. During positive selection, TCR on thedeveloping thymocyte may bind to a selection of peptidevariants, inducing varying degrees of signalling through theTCR. The overall signalling over a period of time (avidity) willinduce either negative selection if the signal is above aparticular threshold, as in the presence of high concentrationsof agonist peptide, or positive selection in the presence ofreduced signal interactions, such as that in the presence ofsome moderate agonists and antagonists. Positive selectionwith CD8 co-receptor down-modulation or CD8β chain down-modulation occurs in the presence of relatively high but non-deleting signalling interactions as in the case of intermediateconcentrations of peptides of an agonist nature. Effectivelythis would allow the survival of T cells with a broader rangeof affinities. Important to this model but as yet undefined isthe influence of co-stimulatory molecules expressed on thymicepithelial cells.

Clearly the cellular and molecular interactions occurringduring positive selection are crucial to the balance andmaintenance of a healthy broad spectrum yet non-auto-reactive repertoire of T cells, a mechanism which possibly

breaks down in states of defective presentation or otherthymic abnormalities which occur in some autoimmunediseases (39,40).

Acknowledgements

The authors would like to thank Pam Ohashi for providing the LCMVpeptide variants, Frank Carbone for providing the B64.1 cell line,Brendan Classon for the CD8β primers and Claerwen Jones forassistance with the mRNA extraction. This research was supportedby grants from the National Health and Medical Research Councilof Australia.

Abbreviations

APC antigen-presenting cellLCMV lymphocytic choriomeningitis virusPE phycoerythrin

References

1 Pawlowski, T. J., Singleton, M. D., Loh, D. Y., Berg, R. and Staerz,U. D. 1996. Permissive recognition during positive selection. Eur.J. Immunol. 26:851.

2 Hogquist, K. A., Tomlinson, A. J., Kieper, W. C., McGargill, M. A.,Hart, M. C., Naylor, S. and Jameson, S. C. 1997. Identificationof a naturally occurring ligand for thymic positive selection.Immunity 6:389.

3 Evavold, B. D., Sloan-Lancaster, J., Wilson, K. J., Rothbard, J. B.and Allen, P. M. 1995. Specific T cell recognition of minimallyhomologous peptides: evidence for multiple endogenous ligands.Immunity 2:655.

4 Tallquist, M. D., Yun, T. J. and Pease, L. R. 1996. A single T cellreceptor recognizes structurally distinct MHC/peptide complexeswith high specificity. J. Exp. Med. 184:1017.

5 Chidgey, A. P. and Boyd, R. L. 1997. Agonist peptide modulatesT cell selection thresholds through qualitative and quantitativeshifts in CD8 co-receptor expression. Int. Immunol. 9:1527.

6 Kersh, G. J. and Allen, P. M. 1996. Essential flexibility in the T-cellrecognition of antigen. Nature 222:495.

7 Evavold, B. D., Sloan-Lancaster, J. and Allen, P. 1993. Ticklingthe TCR: selective T cell functions stimulated by altered peptideligands. Immunol. Today 14:602.

8 Sloan-Lancaster, J. and Allen, P. M. 1996. Altered peptide ligand-induced partial T cell activation: molecular mechanisms and rolein T cell biology. Annu. Rev. Immunol. 14:1.

9 Sloan-Lancaster, J., Steinberg, T. H. and Allen, P. M. 1996.Selective activation of the calcium signalling pathway by alteredpeptide ligands. J. Exp. Med. 184:1525.

10 La Face, D. M., Couture, C., Anderson, K., Shih, G., Alexander,J., Sette, A., Mustelin, T., Altman, A. and Grey, H. M. 1997.Differential T cell signaling induced by antagonist peptide–MHCcomplexes and the associated phenotypic responses.J. Immunol. 158:2057.

11 Lyons, D. S., Lieberman, S. A., Hampl, J., Boniface, J. J., Chien,Y., Berg, L. J. and Davis, M. M. 1996. A TCR binds to antagonistligands with lower affinities and faster dissociation rates that toagonists. Immunity 5:53.

12 Alam, S. M., Travers, P. J., Wung, J. L., Nasholds, W., Redpath,S., Jameson, S. C. and Gascoigne, N. R. J. 1996. T-cell receptoraffinity and thymocyte positive selection. Nature 381:616.

13 Hogquist, K. A., Jameson, S. C., Heath, W. R., Howard, J. L.,Bevan, M. J. and Carbone, F. R. 1994. T cell receptor antagonistpeptides induce positive selection. Cell 76:17.

14 Page, D. M., Alexander, J., Snoke, K., Appella, E., Alessandro,S., Hedrick, S. M. and Grey, H. M. 1994. Negative selection ofCD41CD81 thymocytes by T-cell receptor peptide antagonists.Proc. Natl Acad. Sci. USA 91:4057.


15 Spain, L. M., Jorgensen, J. L., Davis, M. M. and Berg, L. J. 1994.A peptide antigen antagonist prevents the differentiation of T cellreceptor transgenic thymocytes. J. Immunol. 152:1709.

16 Williams, O., Tanaka, Y., Bix, M., Murdjeva, M., Littman, D. R. andKioussis, D. 1996. Inhibition of thymocyte negative selection byT cell receptor antagonist peptides. Eur. J. Immunol. 26:532.

17 Fung-Leung, W. P., Louie, M. C., Limmer, A., Ohashi, P. S., Ngo,K., Chen, L., Kawai, K., Lacy, E., Loh, D. Y. and Mak, T. W. 1993.The lack of CD8 alpha cytoplasmic domain resulted in a dramaticdecrease in efficiency in thymic maturation but only a moderatereduction in cytotoxic function of CD81 T lymphocytes. Eur. J.Immunol. 23:2834.

18 Bachmann, M. F., Oxenius, A., Mak, T. W. and Zinkernagel, R.1995. T cell development in CD8–/– mice. J. Immunol. 155:3727.

19 Goldrath, A. W., Hogquist, K. A. and Bevan, M. J. 1997. CD8lineage commitment in the absence of CD8. Immunity 6:633.

20 Sebzda, E., Choi, M., Fung-Leung, W. P., Mak, T. W. and Ohashi,P. S. 1997. Peptide-induced positive selection of TCR transgenicthymocytes in a co-receptor-independent manner. Immunity6:643.

21 Crooks, M. E. C. and Littman, D. R. 1994. Disruption of Tlymphocyte positive and negative selection in mice lacking theCD8 β chain. Immunity 1:277.

22 Fung Leung, W. P., Wallace, V. A., Gray, D., Sha, W. C., Pircher,H., Teh, H. S., Loh, D. Y. and Mak, T. W. 1993. CD8 is needed forpositive selection but differentially required for negative selectionof T cells during thymic ontogeny. Eur. J. Immunol. 23:212.

23 Itano, A., Cado, D., Chan, F. K. M. and Robey, E. 1994. A role forthe cytoplasmic tail of the β chain of CD8 in thymic selection.Immunity. 1:287.

24 Fung-Leung, W. P., Kundig, T. M., Ngo, K., Panakos, J., De,S. H. J., Wang, E., Ohashi, P. S., Mak, T. W. and Lau, C. Y. 1994.Reduced thymic maturation but normal effector function of CD81

T cells in CD8 beta gene-targeted mice. J. Exp. Med. 180:959.25 Renard, V., Romero, P., Vivier, E., Malissen, B. and Luescher, I. F.

1996. CD8b increases CD8 co-receptor function and participationin TCR-ligand binding. J. Exp. Med. 184:2439.

26 Garcia, K. C., Scott, C. A., Brunmark, A., Carbone, F. R., Peterson,P. A., Wilson, I. A. and Teyton, L. 1996. CD8 enhances formationof stable T-cell receptor/MHC class I molecule complexes.Nature 384:577.

27 Luescher, I. F., Vivier, E., Layer, A., Mahiou, J., Godeau, F.,Malissen, B. and Romero, P. 1995. CD8 modulation of T-cellantigen receptor–ligand interactions on living cytotoxic Tlymphocytes. Nature 373:353.

28 Pircher, H., Burki, K., Lang, R., Hengartner, H. and Zinkernagel,

R. M. 1989. Tolerance induction in double specific T-cell receptortransgenic mice varies with antigen. Nature 342:559.

29 Sebzda, E., Kundig, T. M., Thomson, C. T., Aoki, K., Mak, S.-Y.,Mayer, J. P., Zamborelli, T., Nathenson, S. G. and Ohashi, P. S.1996. Mature T cell reactivity altered by peptide agonist thatinduces positive selection. J. Exp. Med. 183:1093.

30 Bachmann, M. F., Oxenius, A., Speiser, D. E., Mariathasan, S.,Hengartner, H., Zinkernagel, R. and Ohashi, P. S. 1997. Peptide-induced T cell receptor down-regulation on naive T cells predictsagonist/partial agonist properties and strictly correlates with Tcell activation. Eur. J. Immunol. 27:2195.

31 Bachmann, M. F., Ohteki, T., McKall-Faienza, K., Zakarian, A.,Kagi, D., Speiser, D. E. and Ohashi, P. S. 1997. Altered peptideligands trigger perforin-rather than fas-dependent cell lysis.J. Immunol. 159:4165.

32 Chomczynski, P. and Sacchi, N. 1987. Single-step method ofRNA isolation by acid guanidinium thiocyanate–phenol–chloroform extraction. Anal. Biochem. 162:156.

33 Nakauchi, H., Shinkai, Y. and Okumura, K. 1987. Molecular cloningof Lyt-3, a membrane glycoprotein marking a subset of mouse Tlymphocytes: molecular homology to immunoglobulin and T cellreceptor variable and joining regions. Proc. Natl Acad. Sci.USA 84:4210.

34 Chidgey, A., Pircher, H., MacDonald, H. R. and Boyd, R. 1997.An adult stromal cell suspension model for in vitro positiveselection. Dev. Immunol., in press.

35 Sebzda, E., Wallace, V. A., Mayer, J., Yeung, R. S. M., Mak, T. W.and Ohashi, P. S. 1994. Positive and negative thymocyte selectioninduced by different concentrations of a single peptide.Science 263:1615.

36 Williams, O., Tanaka, Y., Tarazona, R. and Kioussis, D. 1997.The agonist–antagonist balance in positive selection. Immunol.Today 18:121.

37 Hogquist, K. A., Jameson, S. C. and Bevan, M. J. 1995. Strongagonist ligands for the T cell receptor do not mediate positiveselection of functional CD81 T cell. Immunity 3:79.

38 Girao, C., Hu, Q., Sun, J. and Ashton-Rickardt, P. G. 1997. Limitsto the differential avidity model of T cell selection in the thymus.J. Immunol. 159:4205.

39 Takeoka, Y., Whitmer, K. J., Chen, S.-Y., Ansari, A. A., Boyd, R. L.,Shultz, L. D., Suehiro, S. and Gershwin, M. E. 1995. Thymicepithelial cell abnormalities in (NZB3H-2u) F1 mice. Clin. Immunol.Immunopathol. 76:297.

40 Watanabe, Y., Naiki, M., Wilson, T., Godfrey, D., Chiang, B.-L.,Boyd, R., Ansari, A. and Gershwin, M. E. 1993. Thymicmicroenvironmental abnormalities and thymic selection inNZB.H-2bm12 mice. J. Immunol. 150:4702.

Documents

Positive selection of low responsive, potentially autoreactive T cells induced by high avidity, non-deleting interactions