TCR ITAM multiplicity is required for the generation of follicular helper T-cells

The T-cell antigen receptor (TCR) complex contains 10 copies of a di-tyrosine Immunoreceptor-Tyrosine-based-Activation-Motif (ITAM) that initiates TCR signalling by recruiting protein tyrosine kinases. ITAM multiplicity amplifies TCR signals, but the importance of this capability for T-cell responses remains undefined. Most TCR ITAMs (6 of 10) are contributed by the CD3ζ subunits. We generated ‘knock-in' mice that express non-signalling CD3ζ chains in lieu of wild-type CD3ζ. Here we demonstrate that ITAM multiplicity is important for the development of innate-like T-cells and follicular helper T-cells, events that are known to require strong/sustained TCR–ligand interactions, but is not essential for ‘general' T-cell responses including proliferation and cytokine production or for the generation of a diverse antigen-reactive TCR repertoire.

T he T-cell antigen receptor (TCR) performs several essential cellular functions in T lymphocytes including conferring antigen recognition, initiation of activating signalling cascades in response to ligand binding, and regulation of key developmental steps required for intrathymic T-cell maturation. Subunits of the TCR complex can be classified into two distinct functional groups: ligand binding or signal transduction. The TCRs expressed by the majority of T lymphocytes contain clonotypic heterodimers composed of TCRa and TCRb chain proteins that are generated by V(D)J recombination of germline gene segments during early stages of T-cell development in the thymus. TCRa/b dimers confer ligand-binding specificity and associate non-covalently with dimers composed of the invariant signal transducing subunits: CD3g, CD3d, CD3e and CD3z. Although the exact subunit composition of mature abTCR complexes has not been unequivocally established, current data support an octameric structure with the following stoichiometry: TCRab, CD3ge, CD3de and CD3zz 1,2 .
Each of the invariant TCR subunits (CD3g, CD3d, CD3e and CD3z) contains one or more copies of a semi-conserved sequence, the Immunoreceptor Tyrosine-based Activation Motif (ITAM), within their cytoplasmic domains that are composed of two YxxL/I cassettes (Y ¼ tyrosine, L ¼ leucine, I ¼ isoleucine, x ¼ any amino acid) separated by 6-8 amino acids 3 . ITAMs operate at the apex of the TCR signalling cascade and ITAM tyrosine phosphorylation is the earliest detectable signalling event that occurs following TCR cross-linking or ligand binding 4 . TCR engagement by peptide-MHC (pMHC) complexes results in membrane-dissociation of ITAMs and rapid phosphorylation of ITAM tyrosine residues by Src family protein tyrosine kinases 5 . Recruitment and activation of the dual SH2 domain protein tyrosine kinase ZAP-70 to tyrosine phosphorylated ITAMs promotes ZAP-70-mediated phosphorylation of the cytosolic adapters LAT and SLP-76, leading to the recruitment and activation of multiple effectors including Sos, PLCg-1 and Vav1 that trigger Ras activation, calcium mobilization and cytoskeletal reorganization in T-cells, events that are essential for T-cell effector functions 6,7 . Although some TCR subunits contain additional conserved functional motifs, ITAMs are the predominant signal transducing sequences within the TCR complex [8][9][10] .
A longstanding and still unresolved question is why the TCR complex contains multiple signal transducing subunits and multiple ITAMs. CD3g, CD3d and CD3e each contain one ITAM, whereas CD3z contains three ITAMs, yielding a total of 10 ITAMs within a single octameric TCR complex. Mutagenesis experiments in which individual ITAMs within the CD3 signalling subunits were inactivated have shown that no single ITAM is essential for either T-cell maturation or T-cell activation indicating that TCR ITAMs are at least partially functionally redundant [11][12][13][14][15][16][17] . Several groups have independently examined the importance of ITAM multiplicity for TCR-mediated signalling by generating mouse models where transgene or retrovirus encoded ITAM-mutant CD3z chains were expressed in CD3z À / À mice 12,14,[17][18][19] . Data from each study documented a requirement for CD3z ITAMs in regulating the set-point for both positive and negative thymocyte selection. However, the impact of reducing the TCR signalling potential on mature T-cell responses was not examined extensively, and the results obtained were inconsistent, likely reflecting differences in the experimental mouse models.
To address outstanding and unresolved questions concerning the role and importance of ITAM multiplicity for TCR-mediated signalling, we analysed two lines of 'knock-in' mice generated by gene targeting in embryonic stem cells: 6Y/6Y, which encodes a 'wild-type' CD3z chain, and 6F/6F, which encodes a CD3z protein where each of the 6 ITAM Y residues was mutated to Phenylalanine (F), rendering these ITAMs non-functional for signal transduction 20 (Fig. 1a). Both 'knock-in' alleles were placed under the control of endogenous CD3z regulatory sequences so that expression of the 6Y and 6F CD3z proteins mimics that of endogenous CD3z, both developmentally and quantitatively 20 . In  the current study, we used the 6Y/6Y and 6F/6F mouse models to  investigate the importance CD3z ITAMs, and by extension, TCR  ITAM multiplicity, for T-maturation and T-cell effector functions. Unexpectedly, we found that attenuation of the TCR signalling potential has an apparently negligible impact on generation of a broad antigen-reactive TCR repertoire or on 'general' T-cell responses such as proliferation and cytokine production. However, the maturation of innate-like T-cells (gd Tcells and iNKT T-cells) as well as the generation of T follicular helper (TFH) cells, events that are known to depend on TCR interactions that result in long dwell times and high signal intensity, were markedly impaired in 6F/6F mice. These results reveal that a key function of ITAM multiplicity is to facilitate developmental and functional responses that are dependent on strong or sustained TCR/ligand interactions.

Results
TCR ITAM reduction results in altered TCR-Va chain usage. The predicted subunit composition of abTCRs expressed in homozygous germline 6Y/6Y and 6F/6F 'knock-in' mice is depicted in Fig. 1a. Since the mature TCR complex (TCRab/ CD3ge/CD3de/CD3zz) contains a total of 10 ITAM signalling motifs (6 contributed by the CD3zz homodimer and 4 contributed by the CD3ge and CD3de dimers) the signalling potential of the TCR is theoretically reduced by 60% (6/10 ITAMs) in 6F/6F mice relative to 'wild-type' 6Y/6Y mice.
As demonstrated previously 20 T-cell development is only mildly impaired in 6F/6F mice; the main defect being a reduction in the number of mature CD4 þ CD8 À (CD4-single positive (CD4-SP)) and CD4 À CD8 þ (CD8-SP) thymocytes and peripheral CD4-SP and CD8-SP T-cells relative to 6Y/6Y control mice. Based on our previous findings, we suspected that reduction of the TCR signalling potential might alter the repertoire of TCRs that successfully navigate the selection process in the thymus 20 . Indeed, positive selection of 6F/6F thymocytes expressing a transgenic MHC Class I (MHC I)restricted (H-Y) or an MHC Class II (MHC II)-restricted (AND) abTCR transgene was almost completely blocked at the immature CD4 þ CD8 þ (double positive (DP)) stage (Fig. 1b,c). Moreover, CD8-SP thymocytes and T-cells in H-Y; 6F/6F mice and CD4-SP T-cells in AND; 6F/6F mice were almost exclusively T3.70 À or Va11 À , respectively, indicating that their maturation required the expression of endogenously encoded TCRa chain(s) ( Supplementary Fig. 1). We detected no significant difference in TCR-Vb chain usage by polyclonal CD4-SP or CD8-SP thymocytes or naïve (CD44 lo CD62L þ ) CD4-SP or CD8-SP T-cells in 6Y/6Y and 6F/6F mice (Fig. 1d). However, TCR-Va usage was significantly different in mature (CD24 lo ) CD4-SP and CD8-SP thymocytes and naïve (CD44 lo ) CD4-SP and CD8-SP peripheral T-cells from 6Y/6Y and 6F/6F mice ( Fig. 1e and Supplementary Fig. 2a). Moreover, when TCRb expression was fixed by introduction of a Vb3 transgene, the frequency of Vb3 pairing with its preferred Va partners (Va11 for CD4-SP and Va8 for CD8-SP thymocytes and T-cells) was significantly different in 6Y/6Y and 6F/6F mice ( Supplementary Fig. 2b).
repertoire mainly, if not entirely, through changes in TCRa chain usage.
Defective clonal deletion in 6F/6F mice. In contrast to positive selection, which is mediated by relatively low affinity TCR-ligand interactions and low signal intensity TCR signalling, negative selection requires high intensity TCR signalling 21,22 . We previously demonstrated that negative selection is impaired in 6F/6F mice as assessed by a reduction in superantigen-mediated deletion; however, overt autoimmunity is not observed in 6F/6F mice 20 . In a recent report, it was show that in addition to clonal deletion, autoreactive thymocytes can also be negatively selected by 'developmental diversion' into abTCR þ CD4 À CD8 À (double negative (DN)) thymocytes 23 . Although clonal deletion normally predominates in wild-type mice, accumulation of developmentally diverted abTCR þ DN thymocytes can be observed in the absence of CD28 co-stimulation or if apoptosis is prevented by overexpression of the pro-survival factors Bcl-2 or Mcl-1 (ref. 23). Examination of 6F/6F DN thymocytes revealed that reduction of TCR signalling potential results in an increase in 'developmentally diverted' CD5 hi , PD-1 hi and CD69 þ abTCR þ DN thymocytes, even in the presence of CD28 costimulation ( Supplementary Fig. 2d). The increase in developmentally diverted thymocytes in 6F/6F mice is presumably due to the fact that these cells have escaped clonal deletion as a result of reduced TCR signalling potential.
High intensity TCR signalling is also required for maturation of innate-like T-cells, which include iNKT-cells and gdTCR þ T-cells (gd T-cells) that express invariant or semi-invariant TCRs and are thought to encounter their cognate antigens in the thymus [24][25][26][27] . We observed a significant reduction in the number of PBS-57:CD1d-tetramer þ thymocytes and peripheral iNKTcells in 6F/6F mice (Fig. 2a-c). Consistent with the results of a previous study 28 , a higher percentage of PBS-57:CD1d tetramer þ CD4-SP thymocytes in 6F/6F mice were immature (CD44 lo , TCR hi ) indicative of a block in iNKT-cell development 29 (Fig. 2b). We also detected a significant reduction of gdTCR þ thymocytes and peripheral gd T-cells in 6F/6F mice (Fig. 2d). These findings demonstrate that a full complement of TCR ITAMs is required for the generation of T-cell populations that express invariant TCRs and that are known to require relatively high affinity/high intensity intrathymic signalling for their complete maturation. Signalling responses are unequally impacted in 6F/6F mice. Analysis of signalling pathways downstream of the TCR revealed that, similar to what we had previously found in thymocytes 20 , activation of TCR-proximal signalling effectors including ZAP-70, SLP-76 and Cbl was reduced in 6F/6F peripheral T-cells relative to 6Y/6Y T-cells under the same stimulation conditions ( Supplementary Fig. 3a,b). However, the effect of reducing TCR signalling potential on activation of more distal effectors and responses was variable. For example, calcium mobilization was unaffected in 6F/6F T-cells, whereas Akt and Erk activation were strongly attenuated (Fig. 3a, Supplementary  Fig. 3a,b). In contrast to the results obtained with a different CD3z ITAM-mutant model system 19 , we found that CD3z ITAMs were not critical for either c-Myc or Notch activation in response to TCR engagement (Fig. 3b). From these results, we conclude that loss of CD3z-chain signalling does not equivalently impact the activation of signalling effectors downstream of the TCR, suggesting that there may be different thresholds for the activation of individual signalling intermediates and pathways.
T-cell proliferation is minimally impacted in 6F/6F mice. To determine if reducing the number of TCR ITAMs impairs mature T-cell functional responses, we first analysed in vitro TCR antibody-mediated proliferation and cytokine production. Notably, naïve CD4-SP and CD8-SP T-cells from 6F/6F mice exhibited only mildly impaired proliferative responses to either anti-CD3 induced stimulation or stimulation with anti-CD3 plus anti-CD28 (Fig. 4a, Supplementary Fig. 4a). We consistently observed a slight reduction in the proliferative response of 6F/6F T-cells with lower amounts of stimulating antibody, but proliferation was equivalent to that of 6Y/6Y T-cells when the amount of activating antibody was increased. Induction of CD69 and the interleukin-2Ra (IL-2Ra) chain (CD25) in response to TCR stimulation was also unaffected or only minimally impaired in 6F/6F T-cells ( Supplementary Fig. 4b,c). Polyclonal 6F/6F CD4-SP and CD8-SP T-cells upregulated CD40L or granzyme B, respectively, in response to TCR cross-linking (Fig. 4b,c). A higher percentage of naïve 6F/6F T-cells produced IL-2 in response to TCR þ CD28 stimulation although the total amount of IL-2 per cell was not increased (Fig. 4d,e). Interferon-g (IFNg) production by in vitro-activated naïve CD4-SP 6F/6F T-cells was also the same or greater than in 6Y/6Y T-cells, whereas IL-4 production by activated naïve 6F/6F CD4-SP T-cells was consistently reduced compared with identically treated naïve 6Y/6Y CD4-SP T-cells ( Supplementary Fig. 4d,e). As reported previously 20 , 6F/6F mice are lymphopenic and contain increased percentages of CD44 hi CD62L lo 'memory-phenotype' CD4-SP and CD8-SP T-cells ( Supplementary Fig. 5a). We observed that a higher percentage of ex vivo CD44 hi CD62L lo CD4-SP and CD8-SP T-cells in 6F/6F mice produced IL-2, IFNg and TNFa as assessed by intracellular staining (Supplementary Fig. 5b).
No overt gaps in the 6F/6F antigen-reactive TCR repertoire.
The differences in TCRa chain usage by positively selected 6F/6F T-cells relative to control 6Y/6Y T-cells raised the question of whether there are 'gaps' in the antigen-reactive T-cell repertoire in 6F/6F mice that would manifest by failure to respond to particular foreign antigens. To investigate this, we first immunized mice with the T-dependent antigen NP-Keyhole limpet haemocyanin (NP-KLH). As shown in Fig. 4f, 6F/6F mice mounted NP IgM antibody responses to NP-KLH that were similar to those of 6Y/6Y control mice. Similar results were obtained with another T-dependent antigen, NP-chicken gamma-globulin (NP-CGG). Next, we determined the total number of naïve T-cells that are specific for the I-A b MHC II-binding non-self-peptides, 2W, a variant of peptide 52-68 from the I-E alpha chain, and FliC, peptide 427-441 from the FliC protein of Salmonella typhimurium, or the K b MHC I-binding peptide from ovalbumin (OVA 257-264). Peptide-specific T-cells from spleen and total lymph nodes from each mouse were enriched with the relevant pMHC tetramers as described 30 and enumerated by flow cytometry. 6F/6F mice contained naïve 2 W-specific and FliCspecific CD4-SP T-cells as well as naïve OVA-specific CD8-SP Tcells that were present in numbers similar to those detected in 6Y/ 6Y mice ( Fig. 4g and Supplementary Fig. 5c). We also detected peptide-specific CD4-SP and CD8-SP thymocytes in 6F/6F mice by tetramer enrichment (Supplementary Fig. 5c). To evaluate the proliferative potential of the naïve CD4-SP T-cell populations, mice were injected intravenously with 2W and FliC peptides plus lipopolysaccharide to elicit a synchronous systemic antigenspecific T-cell response. Six days later, total spleen and lymph node peptide-specific CD4-SP T-cells were tetramer enriched and enumerated. As shown in Fig. 4g, proliferative expansion of naïve 2W-and FliC-specific T-cells in 6F/6F mice was as good or better than that observed in 6Y/6Y mice. Therefore, the number and expansion potential of naïve T-cells specific for these epitopes was not decreased in 6F/6F mice.

ARTICLE
We also infected age-matched 6Y/6Y and 6F/6F mice with LCMV Armstrong and evaluated LCMV antigen-specific responses 8 days after infection. To identify virus-specific T-cells, splenocytes from LCMV-infected mice were stained with MHC I or MHC II tetramers bound to individual LCMV-derived peptides. 6F/6F mice mounted T-cell responses to each of the LCMV-specific antigens that were screened by tetramer staining (Fig. 5a,b). Moreover some LCMV peptide-specific T-cell responses (for example, MHC I-restricted NP396-404 (NP396) and MHC II-restricted GP61-80 (GP61)) were significantly better (as assessed by the number of tetramer þ spleen T-cells detected 8 days after LCMV infection) in 6F/6F mice than in 6Y/6Y mice (Fig. 5a,b). Following in vitro re-stimulation with peptide, the number and percentage of peptide-specific IFNg þ TNFa þ T-cells paralleled tetramer binding, indicating that 6F/6F T-cells are capable of acquiring a bona fide effector phenotype (Fig. 5c and Supplementary Fig. 6a).
Generation of memory T-cells is unimpaired in 6F/6F mice. We next evaluated the potential of 6F/6F T-cells to generate antigen-specific memory T-cells after LCMV Armstrong infection. Forty-five days after LCMV infection, memory T-cells reactive against each of the LCMV antigens that were screened by tetramer binding were detected in 6F/6F mice (Fig. 5d-f and Supplementary Fig. 6b). In general, memory T-cells in 6F/6F mice were composed of a higher percentage of CD62L À CD44 þ T effector memory (T EM ) and fewer CD62L þ CD44 þ central memory (T CM ) phenotype cells compared with 6Y/6Y controls ( Supplementary Fig. 7a). In accord with this finding, a higher percentage of tetramer-binding T-cells in 6F/6F mice, particularly GP61-reactive T-cells, expressed elevated levels of KLRG1 31 , (Supplementary Fig. 7b).
Defective generation of TFH cells in 6F/6F mice. Previous data have shown that strong TCR/pMHC interactions, particularly those with a long dwell time, favour the conversion of naïve CD4 T-cells into TFH 32 . Although this model predicts that the generation of TFH cells should also be dependent on stronger or more sustained TCR signals, a requirement for high intensity signalling has not been demonstrated experimentally. To evaluate   ARTICLE TFH-cell development, age-matched 6Y/6Y and 6F/6F mice were injected with sheep red blood cells (SRBC), and 7-10 days later splenocytes were harvested and analysed by flow cytometry. As shown in Fig. 6a, the percentage and total number of TFH (CXCR5 þ PD-1 þ ) CD4-SP T-cells was significantly reduced in 6F/6F mice compared with 6Y/6Y controls. The reduction in TFH cells was not due to a delayed development as a similar reduction in TFH numbers was also detected in 6F/6F 14 days after immunization. In addition to their numerical reduction, TFH cells in 6F/6F mice expressed reduced levels of the transcription factor Bcl6 (Fig. 6d). Consistent with impaired generation of TFH cells, 6F/6F spleens contained fewer germinal centres (GCs) and reduced numbers of (Fas þ GL-7 þ ) GC B-cells (Fig. 6b,c). A significant reduction in TFH cells and GC B-cells was also observed when 6F/6F mice were immunized with NP-CGG or NP-KLH. Analysis of NP-specific antibodies in immunized mice revealed impaired immunoglobulin class switching (Fig. 6e,f) and immunoglobulin affinity maturation (Fig. 6g) in 6F/6F mice also reflecting a defect in GC formation.
Since Tregs are increased in 6F/6F mice 20 , we considered the possibility that this could explain the impaired TFH-cell response. To resolve this issue, we generated bone marrow chimeras by injecting lineage-depleted bone marrow cells from 6Y/6Y or 6F/6F mice (both CD45.2) together with an equal number of B6:CD45.1 lineage-depleted bone marrow cells into lethally irradiated B6:CD45.1 mice. Six weeks after bone marrow transfer, chimeric mice were immunized with SRBC and spleens were analysed 10 days later. As shown in Fig. 7a-c, a specific reduction in the development of 6F/6F TFH cells was observed in chimeric mice, demonstrating that the 6F/6F defect is T-cell intrinsic. In agreement with this conclusion, in vitro induction of CXCR5 þ PD-1 þ TFH-phenotype T-cells from   naïve 6F/6F CD4-SP precursors was also impaired relative to 6Y/6Y controls ( Supplementary Fig. 8a,b). 6F/6F T-cells were also defective in the generation of Foxp3 þ CXCR5 þ PD-1 þ 'Follicular Tregs', whereas the generation of conventional Tregs was enhanced as previously reported 20 (Fig. 7d). The reduction in 6F/6F TFH cells could not be attributed to a reduction in naïve CD4-SP precursors since TFH induction frequency was also significantly reduced compared with 6Y/6Y mice when calculated as per cent of total Foxp3 À cells (Fig. 7e). Finally, we considered the possibility that the defect in generation of TFH cells in 6F/6F mice might be due to differences in the naïve antigen-specific T-cell pool as a result of effects of the 6F mutation on positive selection. To address this issue, we infected 6Y/6Y and 6F/6F mice with Listeria monocytogenes (Lm) that expresses a 2W peptide-OVA fusion protein (Lm-OVA) since we had established that 6Y/6Y and 6F/6F mice contain similar numbers of naïve 2W peptide-reactive CD4 T-cells (Fig. 4g). One week after infection, spleens were harvested and 2W tetramer-positive T-cells were enumerated by flow cytometry. As shown in Fig. 7f, splenocytes from 6F/6F mice contained significantly fewer numbers of 2W-reactive TFH as well as significantly fewer 2W-reactive (CXCR5 hi PD-1 hi ) GC-TFH cells relative to 6Y/6Y controls. The number of listeriolysin O peptide (LLO) tetramer-positive TFH cells and GC-TFH cells was also reduced in Lm-OVA-infected 6F/6F mice ( Supplementary Fig. 8c). In contrast, the number of 2W-specific and LLO-specific IFNg þ (Th1) CD4-SP T-cells was not significantly different in Lm-OVA-infected 6Y/6Y and 6F/6F mice, confirming a selective defect in the generation of TFH cells (Fig. 7f, Supplementary Fig. 8c).

Discussion
More than two decades ago, the question of why the TCR contains (and presumably requires) multiple signal transducing subunits was first raised following initial reports of the multimeric subunit composition of the TCR. This query intensified on the discovery that TCR complexes contain as many as 10 copies of a semi-conserved (ITAM) signalling motif within their CD3 subunits. Despite continued investigation in the interim, a cogent resolution of the functional importance of TCR ITAM multiplicity has remained elusive, due in part to the finding that individual CD3 ITAMs are not assigned to couple the TCR to specific signalling pathways, and consequently are not indispensible for normal TCR signalling responses. Uncertainty regarding the purpose of TCR ITAM multiplicity can also be attributed in part to the inconsistent results obtained with different genetic reconstitution mouse models where the CD3z ITAMs were inactivated 12,14,[17][18][19] . Each model used heterologous gene-expression methods to reconstitute expression of wild-type or mutant TCR signalling subunits in CD3z À / À mice raising concerns that such non-physiological approaches might lead to confounding results, particularly in light of accumulated data revealing the complexity of the T-cell maturation process and its dependence on the precisely orchestrated expression of developmentally relevant genes. For this study, we used a gene 'knock-in' mouse model designed to place expression of 'wild-type' (6Y) CD3z chain or a non-signalling tyrosine to phenylalanine-mutated (6F) CD3z chain cassette under the control of the endogenous Cd3z gene regulatory sequences. In a previous report, we demonstrated that expression of the 6Y and 6F 'knock-in' alleles faithfully mimics that of endogenous CD3z 20 . Here we document that ab T-cell development is only mildly impaired in 6F/6F mice despite a theoretical 60% reduction in the signal transducing potential of the TCR. However, when the specificity of the TCR was fixed by introduction of an abTCR transgene, positive selection of 6F/6F thymocytes expressing the defined TCR was almost completely abrogated. The T-cells that were generated in these mice were dependent on expression of rearranged endogenous TCRa chains to complete their maturation in the thymus. From these results, we infer that most abTCRs that normally promote positive selection fail to transmit signals appropriate for positive selection in 6F/6F mice. Consequently, a different TCR repertoire is positively selected in 6F/6F mice. Our data also indicate that at least some of the TCR a/b heterodimers that are positively selected in 6F/6F mice bind with higher affinity to selecting self-ligands than most abTCRs that normally promote positive selection as evinced by the increased percentage of 'memory-phenotype' peripheral T-cells in 6F/6F mice even in a lympho-replete background 20 .
An unexpected finding in the current study was that despite the profound effects of TCR ITAM reduction on positive selection, the extent of TCR-Vb CDR3 diversity in the pool of naïve CD4-SP T-cells from 6Y/6Y and 6F/6F mice is no greater than it is between 6Y/6Y and B6 mice. Thus, the differences in the abTCR repertoire in 6Y/6Y and 6F/6F mice can primarily be attributed to differences in TCRa chain usage by a similar repertoire of TCRb chains. These results are consistent with the idea that TCR-Vb genes have been evolutionarily selected to react with MHC proteins 33 . It is currently unknown if the differential TCR-Va usage by mature 6F/6F T-cells is primarily the result of selection from the pool of TCR a/b heterodimers initially expressed on DP thymocytes or is due to iterative sampling of TCRa chains through processive TCRa locus rearrangement. In any event, it is notable that the main consequence of reducing TCR signalling potential is skewing of the mature abTCR repertoire through alternative TCRa chain usage rather than failure of positive selection.
Although it might have been predicted that significant differences in TCRa chain usage by positively selected abTCRs would skew the potential antigen-reactive repertoire of the mature T-cell pool, we failed to identify any obvious 'gaps' in the TCR repertoire, which would be revealed by non-responsiveness to particular foreign antigens. Although by no means comprehensive, these results suggest that there is a striking degree of plasticity in the selection process that ensures preservation of the antigen-specific TCR repertoire despite attenuation of the TCR signalling potential. These findings raise the interesting possibility that, similar to 'superantigen' recognition, ligand recognition is primarily dictated by the TCRb chain with TCRa functioning to stabilize TCRb/ pMHC interactions or adjust ligand-binding affinity 34 .
gd T-cell and iNKT-cell development were impaired to a greater extent in 6F/6F mice than was ab T-cell development. By definition, immature gd T-cells and iNKT-cells, which express invariant or semi-invariant TCRs, are less prone to 'revision' of TCR-ligand binding affinity by pairing of TCRg or TCRb chains with alternate TCRd and TCRa chains, respectively. Unlike conventional ab T-cells, maturation of gd T-cells and iNKT cells also apparently requires interaction with their cognate ligands in the thymus and is thought to be contingent on relatively high intensity TCR signalling [25][26][27] . Negative selection of conventional abTCR þ thymocytes by clonal deletion, which also requires high intensity TCR signalling, was likewise impaired in 6F/6F mice. However, we also found that 'clonal diversion' of autoreactive thymocytes to the abTCR þ DN fate is increased in 6F/6F mice, similar to the effect observed when CD28 co-receptor stimulation is removed or when pro-survival factors are overexpressed in DP thymocytes. The fact that thymocytes are still negatively selected, albeit by clonal diversion rather than clonal deletion, helps to explain the absence of autoimmune disease in 6F/6F mice 20 .
Similar to what we had observed in 6F/6F thymocytes 20 , the impact of inactivating CD3z ITAMs on TCR signalling responses varied considerably depending on the signalling pathway and the proximity of the downstream effectors to the TCR. TCRmediated activation of proximal effectors including ZAP-70, SLP-76 and Cbl was reduced in 6F/6F T-cells, whereas the effect on downstream responses was variable. Activation of some intermediates such as Erk and Akt was strongly attenuated, whereas calcium mobilization and activation of NF-kB, Jnk and p38 were unimpaired or only mildly affected 20 . These results are most consistent with a kinetic proofreading model for TCR signalling where downstream effectors/pathways exhibit different activation thresholds, with some responses appearing to be more 'analogue' whereas others are more 'digital' 35 . T-cell proliferation in vitro in response to TCR cross-linking or in vivo in response to LCMV infection was unaffected, slightly impaired, or in some cases enhanced in the absence of CD3z ITAMs. These results are concordant with those obtained when the cytosolic adapter SLP-76 was deleted or mutated to reduce TCR signalling potential which failed to observe an effect on T-cell proliferation 36,37 but contrast markedly with those obtained using a different 6F/6F model system where it was found that a full complement of ITAMs is crucial for T-cell proliferation 19 . As discussed above, these discrepancies, which we believe can be attributed to differences in the model systems employed, further underscore the importance of retaining normal gene regulation and cell specificity in the design of mouse models of TCR signal attenuation.
In a previous study, we documented that the differential effects of ITAM reduction on downstream signalling pathways, most notably attenuation of Akt activation and relative sparing of the calcium and NF-kB pathways favors Treg development and results in an increase in the number of both thymus derived and peripherally derived Tregs in 6F/6F mice 20 . In this report, we show that the same reduction in the number of TCR ITAMs significantly impairs the generation of TFH cells in vitro in response to TCR cross-linking or in in vivo in response to antigenic stimulation and that this defect is T-cell intrinsic. It is worth noting that, similar to iNKT-cells and gd T-cells, TFH cells have been shown to require high avidity TCR interactions and/or sustained TCR signalling for their development 32,38 . We found that in comparison to 6Y/6Y mice, induction of Bcl6, which is essential for the generation of TFH cells, was reduced in TFH cells in 6F/6F mice. In addition, we observed that IL-2 production in response to TCR engagement is not impaired in 6F/6F T-cells. Since IL-2 receptor signalling inhibits Bcl6 expression 39 , the lack of a defect in IL-2 production by 6F/6F T-cells along with attenuated TCR signalling potential provides an explanation for the reduced efficiency of TFH induction. It is also interesting to note that 45 days after LCMV infection, 6F/6F mice exhibited a higher T EM /T CM ratio of peptide-specific CD4-SP and CD8-SP T-cells, a result consistent with recent data demonstrating that IL-2 and Bcl6 exert opposite effects on the induction of T EM and T CM cells 40 .
What insights do the current results provide regarding the 'raison d'être' for a multi-ITAM TCR signalling structure? The selective requirement for CD3z ITAMs, and by extension, ITAM-mediated signal amplification, for T-cell developmental and functional responses that are regulated by high avidity TCR/pMHC interactions suggests that it is important to regulate these events by linking them to high signalling thresholds. A high signalling threshold for triggering negative selection is logically consistent with the idea that this event should be tightly restricted to avoid limiting the potentially useful TCR repertoire and with the notion that high intensity signalling is coupled to initiation of specific cellular events that induce apoptosis 22 . Likewise, the threshold for TFH-cell induction is presumably intentionally set high (high affinity ligand binding, high signal intensity) to limit this event to 'full blown' antigen/pathogen responses, particularly as inadvertent TFH-cell induction by self-peptides might lead to severe autoimmunity. One way to adjust the activation threshold is through TCR signal amplification (mediated by multiple ITAMs) in response to high affinity/long dwell time interactions. According to this idea, only strong interactions result in strong signals through the activation (phosphorylation) of multiple TCR ITAMs. Notably, reduction of the TCR signalling potential does not appear to significantly restrict the antigen-reactive mature T-cell repertoire, nor does it substantially impact several 'general' T-cell responses, including proliferation and cytokine production. Thus, ITAM-mediated signal amplification is selectively required for and linked to specific effector responses. These findings contribute to the objective of accurately predicting the effects of TCR signal attenuation on both T-cell maturation and T-cell functional responses as a means of evaluating potential targeted and off-target effects of drug treatments for human immune disorders.