Attenuated T Cell Responses to a High-Potency Ligand In Vivo

According to this study, the strongest T cell receptor ligands in vitro do not necessarily induce the strongest T cell responses in vivo, suggesting that vaccine designers may need to reconsider their strategies.


Introduction
Activation of T cells by peptides bound to major histocompatibility complex (pMHC) molecules on antigen presenting cells (APCs) initiates the development of adaptive immunity to pathogens and is characterized by relatively low affinity but highly sensitive interactions between TCRs and pMHC [1,2]. Our understanding of the biochemical requirements for T cell activation by pMHC ligands has been advanced by the use of peptides that vary from the natural ligands at TCR contact residues [3,4]. Since the initial report of a kinetic basis for TCR ligand discrimination [5], the affinity and half-life of TCR/pMHC interactions have been studied as important determinants of T cell activation potency. It has been generally accepted that pMHC ligands with greater affinity/slower off-rates result in a higher potency of T cell activation [6][7][8][9], although there are exceptions [10]. Insight into apparent discrepancies among studies that found the affinity [8,9,11] versus the half-life [6,10,12] of the TCR/ pMHC interaction to be the most influential determinant of T cell activation potency has been provided by a recent report in which affinity (K D ) was found to be the more closely associated with potency when the association rate (k a ) was large (.10 5 M 21 s 21 ), whereas half-life or dissociation rate (k d ) correlated well with potency when k on was small (10 3 M 21 s 21 ) [13]. In both cases (small or large k a ), the strongest interactions, whether defined by kinetic or equilibrium parameters, yielded the most potent T cell responses.
In addition to its influence upon the magnitude of T cell responses, TCR ligand potency is also thought to influence the outcome of in vitro CD4 + helper T cell differentiation [14,15], as well as T cell motility during early stages of in vivo T cell activation [16,17]. It is critical to understand how biochemical parameters of TCR recognition influence the entire course of an in vivo CD4 + T cell response, especially given that the goal of many vaccine strategies is to elicit strong, persistent T cell responses to pathogens and tumors. The importance of CD4 + T cell help in vaccination protocols is underscored by the suboptimal responses elicited by minimal CD8 + T cell epitopes in many cases [18].
We examined the effect of TCR ligand potency upon in vivo CD4+ T cell responses using adoptively transferred 5C.C7 T cells activated by immunization with lipopolysaccharide (LPS) and moth cytochrome c (MCC) peptide, or related ligands of varying potency of in vitro T cell activation [19][20][21][22][23], including one peptide initially identified as an in vitro superagonist ligand for the cytochrome c-reactive 2B4 TCR [24]. This strategy resulted in fully functional differentiation of effector and memory T cells and allowed us to evaluate the influence of TCR ligand potency on multiple stages of the T cell response, which is not possible using in vitro methods, in a system with few variables that is relevant to CD4 + T cell vaccine design [18]. In contrast to predictions from the in vitro hierarchy of ligand potency and the commonly held assumption that more potent T cell responses result from longerlived TCR/pMHC interactions during T cell recognition, we observed an optimal in vivo T cell response in the middle of the avidity spectrum.
A recent study of CD8 + T cell responses to Listeria monocytogenes expressing various OT-1 TCR ligands [25] showed a good correlation between ligand potency in vitro and the magnitude of the T cell response in vivo, i.e. the natural OT-1 TCR ligand SIINFEKL exhibited the most potent responses in vitro and in vivo. In our study of in vivo CD4 + T cell responses, we were able to examine a ligand that is stronger in vitro than the natural 5C.C7 TCR ligand MCC (88-103) [24], and the blunted responses we observed to this ligand in vivo were associated with attenuation of signaling, proliferation, and function at several points during the T cell response, rather than increased cell death, overactivation, or clonal exhaustion. These results point to an upper limit of in vivo TCR ligand potency, which could serve to protect against deleterious inflammatory effects during responses to strong TCR stimulation, and highlight the importance of considering the in vivo efficacy of TCR ligands as part of vaccine strategies that aim to promote high-avidity T cell recognition of tumors or pathogens [11].

Biochemical Determinants of 5C.C7 TCR Ligand Potency
We used 5C.C7 TCR transgenic T cells, which recognize a peptide from MCC in the context of the mouse MHC class II molecule I-E k [3], to study the effect of peptide/MHC ligand potency on in vivo CD4 + T cell activation. Amino acid substitutions of MCC (88-103) result in a panel of peptide ligands that bind to I-E k equally well [24] but exhibit a range of potency for activation of 5C.C7 T cells in vitro [19][20][21][22][23]. Surface plasmon resonance (SPR) was used to study the kinetics of the interaction of purified 5C.C7 TCR with immobilized pMHC complexes ( Figure 1). Figure 1A Figure 1B and C, respectively. The observed SPR curves (shown in Figure 1 as colored symbols) were fitted to a 1:1 Langmuir binding model (shown in Figure 1 as solid black lines) to calculate k a and k d constants, which were used to derive half-life and K D of the TCR/pMHC interactions ( Table 1). As shown in Figure 1A, the on-or off-rates of association of 5C.C7 with 102S/I-E k were too fast to be accurately measured by SPR. The data indicate that

Author Summary
As an important part of immune system, T cells fight infections by recognizing signs of foreign invaders. A molecule on the surface of these cells-called the T cell receptor-recognizes and binds to protein components (peptides) from bacteria, viruses, and other pathogens that are displayed on the surface of other cells. The T cells can then use this information to orchestrate the fight against infection. Vaccination involves injecting into the body foreign peptides that mimic a pathogen, therefore tricking it into raising a T cell response against that pathogen that will be protective in the event of a real infection. We studied T cell responses in mice injected with one of several peptides to which the T cell receptor binds more or less strongly. Contrary to expectations, we found that the peptide that interacted most strongly with the T cell receptor did not provoke the strongest T cell response. This may be nature's way of preventing harmful inflammatory damage due to excessively strong T cell activation. Our work shows that peptides that bind the T cell receptor with medium strength may be best to use for vaccines. Current vaccine strategies seeking to design peptides that bind with maximum strength to the T cell receptor may need to be reconsidered in light of our findings.    (Table 1).
To examine the binding of the various pMHC ligands to 5C.C7 TCR in the context of live T cells, we measured the half-life of tetramer binding to naïve transgenic 5C.C7 T cells using PElabeled I-E k tetramers bearing the weak agonist peptide 102S [19,20], the agonist peptides MCC and K3 [24], and the superagonist peptide K5 [22][23][24], as described in Materials and Methods ( Figure 1D, Table 1). At 25uC ( Figure 1D, left panel), 102S/I-E k tetramers have the shortest half-life of TCR binding (33s), consistent with the inability to measure the interaction by SPR ( Figure 1A) or by tetramer dissociation at 37uC ( Figure 1D, right panel). K3/I-E k tetramers have a shorter half-life of tetramer binding than MCC/I-E k tetramers at both 25uC and 37uC. K5/I-E k tetramers have the longest half-life of 5C.C7 TCR binding, consistent with the k d and half-life derived by SPR ( Table 1). The difference between K5/I-E k and MCC/I-E k tetramer dissociation is greatest at 37uC, with K5/I-E k tetramers binding almost twice as long ( Figure 1D, right panel). The half-lives of TCR/pMHC binding derived by tetramer dissociation are greater than those from SPR (Table 1), consistent with the interaction of multivalent versus monovalent pMHC complexes, but there is good agreement between the two methods when comparing relative off-rates of the various pMHC ligands ( Figure S1), in accordance with previously published data [24,26].
Although the SPR-derived K D value for the binding of 5C.C7 TCR to monomeric K5/I-E k complexes is only slightly lower than that for MCC/I-E k (Table 1), the K5/I-E k tetramer stains naïve 5C.C7 T cells more brightly at equilibrium ( Figure S2A). As described in Materials and Methods, titrations of the tetramers at equilibrium were done as part of apparent K D analysis ( Figure  S2B,C) [27], which shows that K5/I-E k tetramers bind to 5C.C7 T cells with relatively higher avidity than MCC/I-E k tetramers. The larger difference in the equilibrium K5 and MCC measurements seen with tetrameric pMHC and T cells ( Figure  S2) versus solution measurements with monomeric pMHC and TCR ( Figure 1, Table 1) could be explained by a greater effective affinity due to the two-dimensional confinement imposed by the T cell membrane [28][29][30]. k d is likely to be the more important determinant of pMHC ligand potency in this case [10], especially given the small association rates of the 5C.C7 TCR/pMHC complexes (k a ,10 3 M 21 s 21 ; Table 1) [13]. Taken together, the data indicate that the interaction of 5C.C7 TCR with K5/I-E k complexes is stronger than with MCC-I-E k .
Consistent with previously published data showing that fewer K5/I-Ek complexes are needed to induce calcium flux in 5C.C7 T cells [22,23], K5 peptide is a more potent inducer of T cell activation in vitro ( Figure 2). Naïve 5C.C7 T cells were stimulated with splenocytes pulsed in vitro with the various peptides, and proliferation was assessed by incorporation of 3 H-thymidine ( Figure 2A). Approximately four times as much 102S peptide and half as much K5 peptide, on average, is needed to induce the same amount of T cell proliferation as the natural ligand MCC (in vitro EC 50 , Table 1), which reflects their previous characterization as weak agonist and superagonist, respectively [19][20][21][22][23]. K3 peptide seems to activate 5C.C7 T cells better than its half-life would suggest (Table 1), which could be analogous to results previously described for the related cytochrome c-reactive 2B4 TCR, in which K3 binding induced a heat capacity change that is potentially indicative of structural rearrangements that contribute to T cell activation potency [24]. We also activated naïve 5C.C7 T cells in vitro with DCs purified from the spleens of B10.A mice immunized with the various peptide ligands ( Figure 2B-D), which shows that APCs pulsed with K5 peptide in vivo are the most potent stimulators of 5C.C7 T cells in vitro (see below).

Diminished In Vivo T Cell Responses to High-Avidity TCR Stimulation
In vivo activation of small numbers of adoptively transferred 5C.C7 T cells by immunization with MCC peptide and LPS reproducibly results in T cell expansion, contraction, and maintenance ( Figure 3) [31]. At the peak of the expansion (6-7 d after immunization) the 5C.C7 T cells represented up to 8% of CD4 + T cells in peripheral blood, and after contraction were maintained at 1%-2% of blood CD4 + T cells ( Figure 3B). When The binding affinity of soluble I-E k for the various peptides was previously reported in [24]. The IC 50 value is the concentration of unlabeled peptide that prevented 50% of biotinylated peptide binding to I-E k . Errors represent s.d. from the mean. c Values were derived from SPR measurements obtained with immobilized peptide/I-E k complexes and soluble 5C.C7 TCR. Data in Figure 1 were fitted to a 1:1 Langmuir binding model to generate k a and k d . K D was calculated by k d /k a . Errors represent s.d. from the mean and are derived from two independent experiments. d The half-life (t 1/2 ) of I-E k monomer binding to 5C.C7 TCR is equal to ln2/k d . k d was calculated as described in note c. The half-life (t 1/2 ) of I-E k tetramer binding was calculated from linear regression plots ( Figure 1) as the time at which median tetramer fluorescence was equal to 50% of that at t = 0. Errors represent s.d. from the mean, calculated from three or more independent experiments, except in the case of 102S 25uC (two independent experiments). e EC 50 values were determined as the peptide concentration that yielded 50% of maximal 3 H-thymidine incorporation by 5C.C7 RAG22/2 lymphocytes in response to the indicated peptides presented by B10.A splenocytes ( Figure 2A). Values are averaged from four independent experiments. f Not determined. On-or off-rates were too fast to be measured accurately by SPR (see Figure 1). Thus, the half-life of 102S/I-E k monomer binding to 5C.C7 TCR cannot be calculated. The half-life of 102S/I-E k tetramer binding to 5C.C7 TCR was determined at 25uC but is too short to be measured at 37uC. doi:10.1371/journal.pbio.1000481.t001 we compared in vivo 5C.C7 T cell responses to the various peptide ligands, we saw that K3 and MCC agonist peptides induced greater CFSE dilution, peak expansion, and maintenance of 5C.C7 T cells than the weak agonist 102S peptide (Figures 4 and 5), consistent with their greater in vitro potency. Surprisingly, we found that the in vitro-defined superagonist K5 peptide induced the accumulation of fewer 5C.C7 T cells than MCC at every phase of the immune response (Figures 4 and 5). This correlated with slightly delayed CFSE dilution and decreased proliferative capacity of K5-stimulated cells at early times after immunization ( Figure 4A,B).
To examine the possibility that high-avidity interactions with APC make in vivo K5-stimulated 5C.C7 T cells more difficult to extract during homogenization of lymph node tissue prior to flow cytometric analysis [32], accounting for the decreased numbers of 5C.C7 T cells in lymph nodes after K5 immunization, we examined CD45.1 immunofluorescence on cryosections of day 6 lymph nodes from mice immunized with MCC or K5 peptide ( Figure S3). Figure S3A shows five representative images each from MCC and K5 immunizations. Using Metamorph software, the number of CD45.1 + 5C.C7 T cells was estimated within T cell areas by dividing the total PE fluorescence by the average individual 5C.C7 T cell area (defined by lack of B220 and CD19 staining) in MCC-or K5-stimulated samples ( Figure S3B and Materials and Methods). The immunofluorescence data are consistent with the flow cytometry results in Figure 4 and show that there are fewer K5-stimulated 5C.C7 T cells per mm 2 of T cell area ( Figure S3B). K5-stimulated 5C.C7 T cells appear to have, on average, a smaller individual cell area than MCCstimulated cells ( Figure S3C) and less CD45.1 ( Figure S3A and D), which was taken into account as part of the quantitation shown in Figure S3B.
Upon ex vivo restimulation of CD4 + T cells from day 6 lymph nodes, 5C.C7 T cells activated with any of the four peptides produced interleukin-2 (IL-2) and interferon-c (IFN-c) ( Figure 4A,C). For 102S, K3, and MCC peptides, the number of cytokine-producing 5C.C7 T cells correlates with in vitro ligand potency; however, fewer K5-stimulated 5C.C7s produced IL-2 and IFN-c when compared to MCC-stimulated cells ( Figure 4A,C), showing that the effector function of K5-stimulated cells is diminished in addition to proliferation. Thus, fewer 5C.C7 T cells accumulate in response to immunization with K5 peptide as compared to MCC peptide, and fewer of those secrete IFN-c in an ex vivo restimulation assay. The number of IFN-c + 5C.C7 T cells induced in response to immunization with the various ligands is shown in the lower right panel of Figure 4C. The number of IFN-c secreting 5C.C7 T cells is 60% reduced in K5-immunized mice,  when compared to MCC immunization. Such a reduction is likely to affect the inflammatory milieu, given that IFN-c is a key mediator of many aspects of immunity.
No significant differences were found in levels of canonical activation markers such as CD25, CD62L, and CD44 during the initial expansion of 5C.C7 T cells in response to the various peptides ( Figure S4A,C). However, a lower percentage of resting memory 5C.C7 T cells initially stimulated with 102S and K5 peptides have a CD62L lo phenotype ( Figure 5B), suggesting that, in addition to lower numbers of total memory cells in blood ( Figure 5A) and lymph nodes ( Figure 5C,D), these peptides may induce a lower fraction of effector memory cells (T EM ) [33][34][35] than MCC (although other possibilities exist, see Discussion). When the resting memory cells generated with the various peptides were restimulated with MCC peptide, the results were similar to those in the primary stimulation ( Figure S5A,B), which likely reflects precursor frequency [36]. Similar percentages of memory cells generated with K5 and MCC peptides are able to produce IFN-c upon ex vivo restimulation ( Figure S5C), suggesting that the memory cells generated with K5 peptide are as functional on a per-cell basis as those generated with MCC, even though they are maintained in lymph nodes at a much lower frequency ( Figure 5C,D). Thus, ligand potency affects the early in vivo division, accumulation, and cytokine production of a monoclonal population of CD4 + T cells. In contrast to what is predicted from biochemical data, the strong peptide ligand K5 induces a blunted in vivo 5C.C7 response that is of lower magnitude and function than the wild-type ligand MCC.

K5 pMHC Complexes Are Not Limiting In Vivo
Since endogenous T cells have been shown to compete with adoptively transferred T cells for cognate pMHC-bearing APCs [37], we examined responses to the various peptides by MCC/I-E k tetramer staining of lymph node cells from B10.A mice (containing no adoptively transferred 5C.C7 T cells). Seven days after immunization, we found low percentages of endogenous responder T cells to all four peptides (,0.1% of total CD4 + T cells; Figure  S6). A more competitive endogenous response is therefore unlikely to be the explanation for the decreased responses of 5C.C7 T cells to K5 peptide in vivo. Also, titration of peptides in vivo showed that there was no peptide dose at which K5 peptide induced more proliferation than MCC (unpublished data).
We addressed potential differential in vivo antigen persistence by injecting MCC or K5 peptide at various times prior to transfer of CFSE-labeled 5C.C7 T cells ( Figure S7). In this experiment, the extent of division of the transferred T cells serves as a readout of amount of antigen present. The data show that MCC and K5 peptides persist equally and for a few weeks in vivo. We then considered the possibility that the blunted responses were occurring because the amount of K5 pMHC complexes was limiting in vivo. In order to directly examine the stimulatory function of in vivo-pulsed APCs, we used irradiated splenic DCs isolated from mice immunized with the different peptides to activate naïve 5C.C7 T cells in vitro, without the addition of exogenous peptide ( Figure 2B,C). The rank order of T cell activation potency of the peptides in this assay was the same as in the in vitro proliferation assay shown in Figure 2A, which demonstrates that presentation of K5 peptide is not diminished in vivo. Transfer of 5C.C7 T cells does not affect the results of the experiment ( Figure 2B). These results also show that K5 pMHC complexes assembled in vivo are high avidity 5C.C7 T cell ligands and that the blunted in vivo responses to K5 peptide are not due to attenuation of APC maturation or function, since the APCs from mice immunized with K5 peptide are equally capable of presenting antigen ( Figure 2D). Thus, the diminished responses to K5 peptide in vivo are not recapitulated in vitro when in vivopulsed APCs are used, which suggests that T cell activation is attenuated at some point after initial high-avidity ligand recognition in vivo. Indeed, at early time points after activation, 5C.C7 T cells stimulated in vivo have higher levels of CD5 ( Figure  S4B), which is consistent with their experience of a stronger TCR signal in vivo [38][39][40][41].
We also did not find evidence that the blunted responses to K5 peptide are due to preferential suppression of the 5C.C7 effector T cells by regulatory T cells. When CD4 + CD25 + Foxp3 + cells were depleted by injection of anti-CD25 antibody ( Figure S8), numbers of MCC-and K5-stimulated 5C.C7 T cells in lymph nodes at the peak of the response are indistinguishable from controls ( Figure  S5E). The levels of regulatory T cell depletion observed after injection of anti-CD25 antibody (one third of initial levels in lymph nodes; Figure S8D) are sufficient to significantly enhance tumorreactive T cell responses in mice [42,43]. Subcutaneous immunization with K5 peptide in the presence of LPS does not result in expression of Foxp3 by 5C.C7 T cells ( Figure S9). This is in contrast to what we observe during intravenous injection 5C.C7 T cell ligands in the absence of LPS [44]. In summary, the lower capacity for expansion, cytokine production, maintenance, and effector memory differentiation displayed by the K5-stimulated cells is not explained by competitive K5-reactive endogenous T cell responses, antigen dosage or persistence effects, lower in vivo stability of the K5 peptide (or pMHC complexes), or enhanced suppression of high-avidity T cell responses by regulatory T cells.
In Vivo, K5-Stimulated 5C.C7 T Cells Have Decreased Levels of Phosphorylated Signaling Intermediates Compared to Those Stimulated with MCC When we further investigated the phenotype of K5-stimulated 5C.C7 T cells, we found no evidence of clonal exhaustion as defined by increased expression of PD-1 ( Figure 6A) or increased amounts of apoptosis ( Figure 6B) compared to MCC-stimulated cells. In addition, the possibility that K5-stimulated 5C.C7 T cells divide more rapidly than MCC-stimulated cells and are subsequently deleted [45] is not consistent with the delay in CFSE dilution exhibited by K5-stimulated cells at early time points ( Figure 4A). Taken together, these results indicate that K5stimulated 5C.C7 T cells are not likely to be abortively overactivated in vivo. Our data suggest that T cell responses to high-avidity ligands are attenuated in vivo. CTLA-4 is a well-characterized negative regulator of T cell activation that has been suggested to preferentially restrict T cell responses to strong TCR signals [46]. However, we found no evidence that inhibition by CTLA-4 is involved in the blunted T cell responses we observe to K5 peptide in vivo (E.C. and J.P.A., manuscript in preparation).
To gain insight into the molecular mechanisms of attenuation that occur in response to stimulation with K5 peptide, we fixed lymph nodes immediately after sacrifice (60 h after immunization) and analyzed the in vivo activated 5C.C7 T cells for levels of phosphatases and phosphorylated signaling intermediates ( Figure 6C and Figures S10 and S11). The SH2-domain containing tyrosine phosphatase (SHP-1) is thought to participate in a negative feedback loop that functions in TCR ligand discrimination [47], but we did not observe significant differences in levels of either SHP-1 or phosphorylated ERK between MCCand K5-stimulated cells at the time points examined ( Figure S10). This could be because regulation of TCR ligand discrimination by SHP-1 levels is restricted to the boundary between low-affinity agonists and non-agonists/antagonists, as has been suggested [48]. We also did not detect increased levels of the lipid phosphatase PTEN, which negatively regulates the PI3 kinase pathway [49], in K5-stimulated 5C.C7 T cells ( Figure S10).
Strikingly, when the in vivo-activated 5C.C7 T cells were analyzed for levels of S473-phosphorylated Akt (pAkt), which reflects optimally active Akt kinase [50], and phosphorylated Stat3 (pStat3), which is induced in response to multiple cytokines [51] ( Figure 6C, Figure S11), fewer K5-stimulated 5C.C7 T cells were positive for pAkt and pStat3, compared to MCC-stimulated cells ( Figure 6C). The decreased amounts of signaling intermediates in two important pathways in the K5-stimulated 5C.C7 T cells suggest that their activation state is negatively regulated on an intracellular level in response to high-avidity TCR stimulation in vivo. Upon in vitro stimulation, more K5-than MCC-stimulated 5C.C7 T cells are positive for pAkt and pStat3 ( Figure S12), consistent with the greater in vitro potency of K5 peptide ( Figure 2). These results reveal a clear difference between the in vivo and in vitro regulation of the T cell response to a high-avidity ligand.

Discussion
In contrast to the assumption that more potent T cell responses result from higher-avidity TCR recognition, here we report diminished T cell responses to a strong TCR ligand in vivo. We studied the in vivo activation of 5C.C7 T cells by immunization with LPS and a panel of peptides with a range of in vitro T cell activation potency, which results in fully functional effector and memory responses. Interestingly, we observed optimal T cell responses to the peptide ligands with intermediate TCR/pMHC half-lives. Contrary to what is predicted from in vitro data, the K5 ligand, which has the longest TCR/pMHC half-life, induces less in vivo division, accumulation, cytokine production of 5C.C7 T cells during the effector phase of the response, as well as persistence of fewer 5C.C7 T cells during the memory phase.
Given that we observe fewer CD62L low memory cells in lymph nodes after immunization with K5 peptide, we cannot rule out that K5 favors T EM development and greater homing to tissues [34], thus accounting for the fewer numbers of memory 5C.C7 T cells in lymph nodes prior to exit ( Figure 5B). The data in Figure 5B would also be consistent with accelerated T CM development, or even impaired T EM development if the CD62L low population present in the MCC response is considered to be lymph node T EM cells ( Figure 5B). Increased T EM development and tissue homing induced by K5 peptide could potentially account for the decreased numbers of K5-stimulated T cells at later time points, but it is unlikely that it could explain the delayed CFSE dilution and decreased accumulation in lymph nodes at the peak of the primary response, observed at day 2.5 and day 6, respectively ( Figure 4 and Figure S4C), especially given the lack of evidence for increased egress of K5-stimulated cells into the blood at these earlier time points ( Figure 5).
We found that the blunted response to K5 peptide is not due to a lack of in vivo availability of peptide/MHC complexes, dose effects, or competition by endogenous responses. In addition, the 5C.C7 T cells stimulated with K5 peptide in vivo do not show signs of apoptosis or clonal exhaustion, making it unlikely that they are abortively over-activated. Finally, the K5-stimulated 5C.C7 T cells are not preferentially inhibited by regulatory T cells or by the T cell inhibitory receptor CTLA-4. Interestingly, we observed attenuation at the level of intracellular signaling intermediates associated with Akt kinase activation and cytokine signaling, which we measured directly ex vivo using antibodies to pAkt and pStat3. Fewer K5-stimulated 5C.C7 T cells were positive for pAkt, indicating decreased activity of the mammalian target of rapamycin (mTOR)-containing complex mTORC2 in these cells [50]. We also observed fewer pStat3-positive K5-stimulated 5C.C7 T cells, which may indicate that the cells are less responsive to cytokines which activate the Stat3 pathway [51].
In a recent study of in vivo CD8 + T cell activation, in vitro ligand potency correlated with the magnitude of the in vivo T cell response, i.e. the highest-affinity ligand (which was the natural OT-1 ligand SIINFEKL) induced the most robust response [25]. The reason we observed attenuation at the high end of the ligand potency spectrum may be because we were able to examine T cell responses to a superagonist peptide, which possibly exceeds a ligand potency threshold at which T cell attenuation is observed. Attempts to find other strong 5C.C7 TCR ligands, which would facilitate examination of this hypothesis, have been unsuccessful thus far. It is also possible that the attenuation we observed in vivo in response to high-avidity TCR stimulation is unique to CD4 + T cell responses.
The effect of adjuvant upon the avidity of TCRs in the polyclonal CD4 + T cell response to pigeon cytochrome c (PCC) protein was recently examined by Malherbe et al. [52]. These authors found that immunization in the presence of non-depot forming adjuvants containing toll-like receptor 4 or 9 agonist resulted in accumulation of high-avidity PCC-reactive T cells at the peak of the response, based on their previous characterization of the avidity of TCRs bearing specific CDR3 sequences [53], as well as levels of pMHC II tetramer binding. The presence of highavidity CD4+ T cell clones in the Malherbe et al. study is not inconsistent with attenuation of the monoclonal CD4 + T cell population that we tracked in response to strong TCR ligand. It is hard to compare the avidities of the responding TCRs since both the TCRs and the peptide ligands in question are different. Therefore, it is possible that the responses to the strongest ligand in our study exceeds an avidity threshold beyond which T cell attenuation occurs, and that this threshold is not reached by any of the TCRs in the Malherbe et al. study. It is also possible that the naïve repertoire in the Malherbe et al. experiments contains T cells with even higher avidity for PCC antigen than the clones favored by immunization with the TLR agonists. These clones may then be attenuated in response to high-avidity TCR engagement, such as we observed in our study. This could result in the underrepresentation of such clones at the peak of the primary response, and thus they may not be present during the CDR3 sequence analysis performed by Malherbe et al. [52].
At the low potency end of the ligand spectrum, our results were in agreement with the CD8 + T cell study mentioned above [25]; the lowest affinity ligand we examined (102S) was able to induce a functional response. Interestingly, we saw that the lowermagnitude T cell response induced by 102S correlates with low levels of pAkt and pStat3 before the expansion phase; however, it is possible that the accumulation of these phosphorylated intermediates occurs at a later time in response to the weak agonist peptide [54]. The observation of sustained signaling intermediates such as pAkt in the in vivo T cell response to antigen is not necessarily predicted by in vitro studies [55] and may point to the importance of the cumulative activity of signaling pathways in defining the magnitude of in vivo T cell responses. Thorough examination of the timing of engagement and sustained activity of signaling pathways downstream of in vivo T cell activation is likely to lend further insight into such questions, as well as the biochemistry of in vivo T cell activation in general.
We propose that negative regulation of the extent of T cell activation occurs in response to high-avidity in vivo TCR stimulation, potentially in the context of the long-lived T cell-DC contacts that are a hallmark of in vivo antigen recognition [56], and that this is characterized by but not necessarily limited to a decrease in the phosphorylated signaling intermediates we measured in this study. Although we did not find increased levels of the inhibitory phosphatases SHP-1 and PTEN in the K5stimulated cells, cell-intrinsic attenuation of activation could be certainly mediated by other intracellular negative regulators of TCR signaling [57] and/or increased internalization and degradation of cell surface receptors [58] in response to strong sustained in vivo TCR engagement. We did not observe increased TCR downregulation by K5-stimulated cells (unpublished data). The strong in vivo TCR signal could also influence (directly or indirectly) induction of or responsiveness to cytokines, resulting in attenuated T cell responses. Although we did not find a role for regulatory T cells in the blunted response to K5 peptide in experiments where CD4 + CD25 + Foxp3 + cells were depleted by injection of anti-CD25 antibody ( Figure S8), nor did we observe Foxp3+ induction by 5C.C7 T cells in response to subcutaneous immunization with K5 ( Figure S9), other cell-extrinsic mechanisms of attenuation cannot be excluded.
Attenuation of in vivo T cell responses to high-avidity pMHC ligands may function to prevent detrimental inflammatory effects that could occur in response to high-avidity TCR recognition of either self or foreign peptides. Our results are contrary to the common assumption that higher-avidity TCR recognition promotes a more potent T cell response and point to an upper limit of T cell ligand potency in vivo. Interestingly, a study of a panel of peptide ligands with variable affinity for a tumor-reactive TCR showed that elicitation of the most potent tumor-reactive CD8 + T cells in vivo and the best tumor-free survival occurred after vaccination with peptide ligands that displayed intermediate affinity and in vitro T cell activation potency [11]. Although it is unclear if the tumor-reactive T cells in vivo bear the same TCR as the clone on which ligand affinity was characterized, the results point out an important discrepancy between the in vitro and in vivo efficacy of TCR ligands. This discrepancy, along with the attenuation we observed in vivo while tracking the responses of monoclonal CD4 + T cells to a strong ligand, suggests that vaccination strategies that aim to elicit maximum-avidity TCR recognition with engineered peptide ligands may warrant reconsideration [11].

Ethics Statement
All mice were maintained in microisolator cages and treated in accordance with NIH and the American Association of Laboratory Animal Care regulations. Experiments in this study were approved by the Memorial Sloan-Kettering Cancer Center Institutional Animal Care and Use Committee.

Purification of Soluble 5C.C7 TCR and SPR
Coding regions for the 5C.C7 TCR a and b chains were cloned by PCR into a modified version of the transfer vector pACGP67A (Pharmingen), which placed the 5C.C7 a and b in frame with c-jun (a) and c-Fos (b) heterodimerization motifs. Both chains were linked at the C-terminus to a 6XHis tag through a 3C-proteasecleavable site. TCR was expressed in insect cells as previously described [23], isolated from the supernatant by Ni-NTA purification, cleaved with 3C protease, and further purified by ion-exchange and size exclusion chromatography. Protein concentration was determined using an extinction coefficient (280 nm) of 1.3 ml?mg 1 ?cm 1 . All SPR data were collected at 25uC using a BIAcore 3000 TM instrument (BIAcore Inc.). Purified biotinylated pMHC complexes were immobilized on a streptavidin chip (BIAcore Inc.) to a level of 250-500 RU. Free biotin-binding sites were blocked with 10 mM biotin. Binding of 5C.C7 TCR was measured in PBS, pH 7.8 with 0.005% Surfactant P20 (BIAcore Inc.) by injecting a range of concentrations (0.9-14.3 mM) at a flow-rate of 30 ml/min. All data were background subtracted using a biotin-blocked streptavidin surface as a reference surface. Data were fitted to a Langmuir binding model using global fitting (Biaevaluation, BIAcore Inc.) to calculate k a and k d constants. To prevent aggregation artifacts, all proteins were purified no more than 12 h before analysis without subsequent concentration steps.

Mice
5C.C7 TCR transgenic RAG22/2-mice were purchased from Taconic Farms and bred to B10.A CD45.1 (provided by W. Paul via the NIAID contract facility at Taconic Farms) to generate 5C.C7 RAG22/2 CD45.1 mice, which were used as donor cells in adoptive transfer experiments. Male B10.A recipient mice were purchased from Taconic and were used to initiate experiments at 6-9 wk of age.

Tetramers, Tetramer Dissociation, and Apparent K D Analysis
Tetramer production, dissociation assays, and data analysis were done as described [27]. Briefly, PE-labeled I-E k tetramers with MCC and variant peptides were produced. Naïve 5C.C7 T cells were incubated with tetramers (50 mg/mL) for 2 h at RT in the presence of fresh 0.2% NaN 3 , equilibrated to 37uC or 25uC, and measurement of dissociation was initiated with the addition of anti-I-E k antibody. Aliquots were removed in 10-20 s intervals and fixed immediately in 1% PFA. Levels of tetramer staining were assessed by flow cytometry. Apparent K D analysis was done as previously described [27]. Briefly, varying concentrations of tetramers were used to stain the 5CC7 T cells as described above, and the median intensity of tetramer staining was subjected to Scatchard analysis. Since tetramers were always in excess, free tetramer concentration is equal to staining concentration.

Cell Culture and In Vitro T Cell Activation
Cells were cultured in a 37uC humidified chamber with 5% C0 2 in complete RPMI1640 (supplemented with 10% FCS, 2 mM glutamine, 100 U/mL penicillin and streptomycin, 2 mM 2mercaptoethanol). Single-cell suspensions were prepared from lymph nodes harvested from 5C.C7 RAG2-deficient mice (routinely .90% Va11+Vb3+ by flow cytometry) and stimulated in triplicate in 96-well round-bottom plates with irradiated (2,000 rad) B10.A splenocytes and the indicated peptides for 60 h. Proliferation was monitored by the addition of 3 H-methylthymidine (1 mCi/well). Cells were harvested onto glass-fiber filters using a Tomtec harvester, and filters were counted using a MicroBeta scintillation counter (Perkin-Elmer).

Flow Cytometry and Antibodies
Flow cytometry was done on a BD LSRII and data were analyzed with Flowjo software (Treestar). Antibodies to surface markers were from BD Pharmingen, eBioscience, or BioLegend. Antibodies to phospho-Akt, phospho-Stat, phospho-Erk, and PTEN were Alexa 647 conjugates from Cell Signaling Technology. SHP-1 antibody was from Santa Cruz Biotechnology.
In Vivo T Cell Activation 1-5610 4 5C.C7 RAG22/2 CD45.1 T cells were transferred into B10.A recipients by tail vein injection, and the mice were immunized the next day with 20 mg peptide and 10 mg LPS at two sites on either side of the base of the tail. In some experiments the cells were CFSE-labeled prior to transfer. Proliferative capacity was calculated from CFSE profiles as described [59]. Total lymph nodes were pooled for all experiments. Cytokine production was assessed by ex vivo restimulation for 4-5 h with B10.A DCs and peptide and intracellular cytokine staining using BD Biosciences reagents.
Immunofluorescence 5C.C7 RAG22/2 CD45.1 T cells were activated in vivo as described above. Six days after immunization with MCC or K5 peptide (5 mice per peptide), mice were sacrificed and the lymph nodes were embedded directly into OCT and frozen using dry ice and methanol. After temporary storage at 280uC, 10 mm sections were cut through the entire block (containing 50 lymph nodes) using a Leica Cryostat and stained with anti-CD45.1-biotin/ streptavidin-PE to mark 5C.C7 T cells and anti-B220-FITC and anti-CD19-FITC to mark B cell areas. The stained sections (each containing 10 or more lymph nodes) were scanned with a Mirax slide scanner (epifluorescence). The number of 5C.C7 T cells was quantified from the T cell areas of 20 independent lymph nodes using Metamorph software. Total fluorescence was divided by the average individual T cell area to estimate the number of 5C.C7 T cells per mm 2 of T cell area.

CD4+ Purification from Lymph Node Suspensions
To enrich for 5C.C7 T cells originating from low precursor frequency before flow cytometry, CD4+ T cells were purified from lymph node suspensions by Dynal negative selection (Invitrogen).

Analysis of Phospho-Intermediates
Lymph node suspensions were fixed immediately after harvest and then negatively selected for CD4+ T cells. After staining with antibodies to CD4 and CD45.1, cells were permeabilized with methanol and stained with phospho-antibodies.

In Vivo-Pulsed APCs
CD11c + cells were purified by positive selection (Miltenyi) from spleens of mice adoptively transferred and immunized as described, irradiated, and used to stimulate naïve 5C.C7 RAG22/2 T cells in vitro. Proliferation was assessed by 3 H-thymidine incorporation as described.

Statistical Analysis
Data were analyzed for significance using unpaired Student's t test analysis with Prism software. Figure S1 Correlation of 5C.C7 TCR/peptide-I-E k halflife derived by tetramer dissociation and SPR. Plot of the half-life of interaction of purified 5C.C7 TCR with I-E k complexed with K3, MCC, and K5 peptides derived by SPR ( Figure 1A-C) versus the half-life derived by tetramer dissociation ( Figure 1D). Figure S11 Specificity controls for phosphoantibodies. B10.A mice containing adoptively transferred 5C.C7 T cells were immunized with LPS alone or LPS+MCC peptide. 2.5 d after immunization, lymph node samples were fixed immediately after harvest, methanol permeabilized, and stained with the indicated phospho-antibodies. In the right panels, the antibodies were incubated with a peptide corresponding to the reactive phosphoepitope (according to the manufacturer's protocol) prior to staining of lymph node samples.