PD-L1 co-stimulation contributes to ligand-induced T cell receptor down-modulation on CD8+ T cells

T cell receptor (TCR) down-modulation after antigen presentation is a fundamental process that regulates TCR signal transduction. Current understanding of this process is that intrinsic TCR/CD28 signal transduction leads to TCR down-modulation. Here, we show that the interaction between programmed cell death 1 ligand 1 (PD-L1) on dendritic cells (DCs) and programmed death 1 (PD-1) on CD8 T cells contributes to ligand-induced TCR down-modulation. We provide evidence that this occurs via Casitas B-lymphoma (Cbl)-b E3 ubiquitin ligase up-regulation in CD8 T cells. Interference with PD-L1/PD-1 signalling markedly inhibits TCR down-modulation leading to hyper-activated, proliferative CD8 T cells as assessed in vitro and in vivo in an arthritis model. PD-L1 silencing accelerates anti-tumour immune responses and strongly potentiates DC anti-tumour capacities, when combined with mitogen-activated kinase (MAPK) modulators that promote DC activation.


INTRODUCTION
The immune system must protect the organism against infectious diseases and cancer without provoking autoimmunity. T lymphocytes play a key role in the induction of protective and long-lasting immunity, but, if uncontrolled, can cause autoreactive disease. For this reason, T cell activation is regulated at multiple levels, particularly during antigen presentation. Understanding these mechanisms is essential for designing effective therapies for the treatment of cancer, infectious diseases and autoimmune disorders. T cells recognize specific peptides in association with major histocompatibility complex (MHC) molecules that are expressed on the surface of antigen presenting cells (APCs) through binding of their respective T cell receptor (TCRs). However, MHC-peptide recognition is not sufficient for full T cell activation, and a range of co-stimulatory ligand-receptor interactions is also required that can provide either positive or negative signals. For instance the CD80-CD28 interaction is stimulatory, while others such as programmed cell death 1 ligand 1 (PD-L1)-programmed death 1 (PD-1) association are inhibitory. The relative contribution of co-stimulatory/co-inhibitory signals determines the activation state of T cells, leading to T cell proliferation and acquisition of effector activities or differentiation into anergic or regulatory T cells. Thus, the overall integration of positive and negative signals during co-stimulation provides a checkpoint at which T cell responses are modulated (Nurieva et al, 2006). PD-L1 is a member of the B7 family of co-stimulatory/ inhibitory molecules, which is expressed in a wide range of cell types, including T cells and dendritic cells (DCs; Latchman et al, 2004;Sharpe et al, 2007). PD-1 is transiently up-regulated in activated T cells during antigen presentation, and its ligation to PD-L1 recruits src homology 2 domain-containing tyrosine phosphatases 1 and 2 (SHP 1 and 2) to its intracellular switch motif. SHPs dephosphorylate effector molecules associated with the TCR leading to termination of TCR signal transduction (Chemnitz et al, 2004;Sheppard et al, 2004).
Ligand-induced TCR down-modulation is another regulatory process of T cell activation at the level of antigen presentation. TCRs are removed from the T cell surface shortly after activation, limiting signal transduction and avoiding excessive responses (Holst et al, 2008;Naramura et al, 2002;San Jose et al, 2000;Schonrich et al, 1991;Shamim et al, 2007). The current view of ligand-induced TCR down-modulation is that intrinsic TCR signalling following antigen recognition is sufficient for TCR down-modulation. However, the exact mechanism by which this occurs is still under extensive research. Ligandinduced TCR down-modulation is a complex, multi-mechanistic process . It is well established that ligandengaged TCR complexes are quickly internalized (Cai et al, 1997;Dietrich et al, 1998;Huppa et al, 2010;Lauritsen et al, 1998;Valitutti et al, 1995). Afterwards, other TCR complexes including non-engaged ones, are down-regulated following signal transduction from the triggered TCRs (San Jose et al, 2000). Nevertheless, up-regulation of E3 ubiquitin ligases of the Casitas B-lymphoma (Cbl) family in T cells contributes to ligandinduced TCR down-modulation as demonstrated by studies with Cbl knock-out (KO) mice (Naramura et al, 2002;Shamim et al, 2007). Cbl KO T cells have reduced TCR down-modulation following antigen presentation, leading to sustained signalling and hyper-activation (Chiang et al, 2000;Naramura et al, 2002;Shamim et al, 2007). Interestingly, to date, no additional extrinsic signals provided by DCs in the immunological synapse, which are critical for ligand-induced TCR down-modulation, have been identified.
Counteracting negative signals transmitted to effector T cells is a promising approach to achieve therapeutic efficacy for cancer and infectious diseases. Thus, the disruption of PD-L1/ PD-1 signalling pathway, mainly by systemic administration of blocking antibodies, is receiving increasing interest in immunotherapy (Curran et al, 2010;Hirano et al, 2005;Hobo et al, 2010;Zhou et al, 2010). In addition, abrogation of Cbl ubiquitin ligases in effector T cells significantly enhances their antitumour/anti-viral activities (Bachmaier et al, 2000;Chiang et al, 2000;Naramura et al, 2002;Paolino et al, 2011), making them attractive therapeutic targets.

PD-L1 silencing in DCs inhibits CD8 TCR down-modulation
In this study, we investigated the consequences and therapeutic outcomes of silencing the co-stimulatory ligand PD-L1 in DCs during antigen presentation to T cells. Thus, we delivered a PD-L1-specific short hairpin (sh)RNA (termed p5) using a reported lentiviral platform (Arce et al, 2011). Transduction of bone marrow derived DCs (BM-DCs) resulted in the specific silencing of PD-L1 surface expression with no down-regulation of other classical DC maturation markers (Fig 1 of Supporting Information).
To test the effects of PD-L1 silencing on MHC class I and II antigen presentation, we co-delivered the shRNA p5 with IiOVA, a well-defined model antigen containing both MHC class I and class II epitopes (Fig 1A;Arce et al, 2011;Escors et al, 2008).
Thus, BM-DCs were transduced with control lentivectors expressing GFP alone, IiOVA-GFP or IiOVA-p5-GFP. Inclusion of the shRNA did not alter gene expression from the lentivector ( Fig 1A). BM-DCs transduced with IiOVA-GFP or IiOVA-p5-GFP activated proliferation of OVA-specific CD4 and CD8 T cells purified from transgenic OT-II and OT-I mouse strains, respectively (results not shown). However, while approximately 40% of CD8 T cells clustered around IiOVA-expressing DCs after 16 h of culture as assessed by microscopy, around 90% of CD8 T cells clustered around BM-DCs expressing IiOVA-p5 ( Fig 1B). By comparison, no differences were observed for DC-CD4 OT-II cell association, which was around 10% (results not shown). Therefore, in agreement with other studies (Fife et al, 2009), these results suggested that the DC-CD8 T cell association was enhanced when PD-L1 was silenced.
To investigate why increased DC-T cell clustering was observed upon down-regulation of PD-L1, we examined several potential factors. Interestingly, we found that TCR downmodulation was significantly inhibited when antigen was presented in the absence of PD-L1 (Fig 1C and D). Equivalent results were obtained by surface CD3 staining, the TCR component mediating signal transduction (Fig 1E and F), and by staining with OVA-specific class I pentamer (results not shown). Surface levels of the T cell lineage marker Thy1.2 remained unchanged, confirming the specificity of the TCR down-modulation ( Fig 1E). It is worth noting that TCR downmodulation reached a maximum on day 3-4, and TCR levels gradually recovered afterwards (Fig 2 of Supporting Information). PD-L1 silencing in DCs significantly delayed TCR downmodulation but did not completely abrogate it (Fig 2 of Supporting Information). Persistent TCR signal transduction was demonstrated by intracellular detection of phosphorylated extracellular signal-regulated kinase (ERK) in TCR high CD8 þ T cells (results not shown).
To confirm that a reduction in DC PD-L1 co-stimulation also reduced TCR down-modulation on activated T cells in vivo, we immunized Thy1.1 mice with BM-DCs transduced with the different lentivectors. Then, congenic Thy1.2 OT-I CD8 þ T cells were adoptively transferred the following day and analysed 7 days later. OVA-specific OT-I cells primed in vivo by DCs with silenced PD-L1 down-regulated TCR expression to a lesser extent than T cells primed in the presence of PD-L1 (Fig 2). Collectively, our results demonstrate that DC-derived PD-L1 signalling contributes to ligand-induced TCR down-modulation after antigen presentation.
PD-L1 silencing in DCs blocks expression of Cbl E3 ubiquitin ligases in CD8 R T cells Recent evidence demonstrates that expression of Cbl E3 ubiquitin ligases significantly contributes to ligand-induced TCR down-modulation and limits TCR signal transduction (Bachmaier et al, 2000;Chiang et al, 2000;Naramura et al, 2002). Cbl KO T cells exhibit a reduced TCR down-modulation after antigen presentation, exactly to the same extent as observed in our experiments (Fig 1; Shamim et al, 2007). Therefore, we tested whether PD-L1 silencing in antigenpresenting DCs inhibited up-regulation of the two major Cbl

Research Article
PD-L1 co-stimulation down-modulates the TCR on CD8 þ cells

Research Article
Katarzyna Karwacz et al.  proteins, c-Cbl and Cbl-b, in CD8 þ T cells (Fig 3A and B). Our data show that, while antigen presentation by IiOVA-DCs strongly up-regulated both Cbl proteins, PD-L1 silencing in IiOVA-p5-DCs did not result in significant expression of Cbl-b or, to a lesser extent, c-Cbl ( Fig 3B). We found that lack of Cbl-b expression rather than c-Cbl was consistent with the inhibition of CD3 internalization (Fig 3C and D).
To confirm that PD-L1/PD-1 co-stimulation was a regulator of TCR down-modulation and also exclude off-target shRNAmediated mechanisms, we repeated our DC-T cell co-cultures in the presence of well-described PD-L1 or PD-1 blocking antibodies. Blockade of either the ligand or receptor reproducibly replicated our findings using PD-L1 silencing in DCs, including significant inhibition of TCR down-modulation and Cbl-b expression ( Fig 3E and F and data not shown). In fact, CD3 internalization and Cbl-b expression were reduced in a dose-dependent manner (data not shown). A nearly complete abrogation of CD3 internalization was achieved with the highest concentration of PD-1 blocking antibody, while surface levels of Thy1.2 remained unchanged ( Fig 3E). Collectively, these data demonstrate that the PD-L1/PD-1 interaction contributes to ligand-induced TCR down-modulation and that it depends on PD-1 expression in T cells after activation in our experimental system.
To further demonstrate the requirement for PD-1 expression in T cells, we silenced PD-1 in human T cells by delivering shRNAs using lentiviral vectors. We chose human cells due to their susceptibility to transduction in the absence of activation. Thus, PD-1-specific shRNAs and GFP were co-delivered under the control of the U6 and phosphoglycerate kinase (PGK) promoters. To overcome the necessity of specific antigen presentation, we stimulated human CD4 þ T cells with

Research Article
PD-L1 co-stimulation down-modulates the TCR on CD8 þ cells monocytes and staphylococcus enterotoxin-B (SEB) superantigen, following a published protocol (Horgan et al, 1990). Accordingly, CD3 surface down-modulation was significantly inhibited following PD-1 silencing in stimulated GFP þ CD4 þ T cells, although not completely abrogated (results not shown). We reproduced these results with shRNAs targeted to three different PD-1 regions.
PD-L1 silencing in DCs leads to hyperactivated pro-inflammatory TCR high CD8 R T cells To investigate whether PD-L1-induced dysregulation of TCR internalization altered the function of primed T cells, we initially compared proliferation in T cells incubated with BM-DCs transduced with IiOVA or IiOVA-p5 lentivectors. As expected, an inverse correlation between surface expression of TCR chains and proliferation was observed (Fig 4A-C). In contrast, T cells stimulated in the presence of reduced PD-L1 signalling proliferated more extensively in vitro, even though they had significantly increased TCR levels (Fig 4B and C). These T cells produced increased levels of IFN-g and IL-17, suggesting an enhanced pro-inflammatory phenotype (Fig 4D and E). Detection of annexin V in DC-T cell co-cultures indicated that apoptosis was not increased in T cells primed in the absence of PD-L1 (data not shown). Therefore, these data suggested that

Research Article
Katarzyna Karwacz et al. (bottom). ÃÃÃ , highly significant differences. D. IFN-g and IL-17 expression in OT-I cells co-cultured with lentivector-transduced DCs (top). Percentage and mean fluorescent intensities are shown within the gates. E. As in (D), plotting mean percentages and error bars from three experiments, including IL-2 expression. Ã , significant differences. F. Knee inflammation after intra-articular challenge with OVA in groups of five mice. One day before challenge, OT-I cells were co-cultured with lentivector-transduced DCs (legends within the graph) and intravenously transferred. Means are plotted with error bars (standard deviations). ÃÃ , very significant differences. G. Percentage of OVA-specific endogenous CD8 þ T cells in draining lymph nodes stained with class I OVA-pentamers, on day 7 after vaccination with the indicated lentivectors used for DC transduction (bottom  (Chiang et al, 2000;Naramura et al, 2002), which spontaneously develop lethal autoimmune disorders (Naramura et al, 2002). Therefore, we tested whether T cells primed in the absence of PD-L1 co-stimulation could cause immune pathology in a well-established inflammatory arthritis model using OVA as a surrogate autoantigen (Arce et al, 2011; Fig 4F). Transfer of T cells primed by DCs transduced with IiOVA-p5 resulted in severe knee inflammation after intraarticular challenge with OVA compared to mice that had received T cells primed with DCs expressing IiOVA, which induced much less severe inflammation. Therefore, these results show that T cell priming without PD-L1 co-stimulation leads to the development of hyperactivated T cells, which cause significant immune pathology in vivo if directed towards an autoantigen ( Fig 4F).
Our conclusions were further reinforced by studying the endogenous CD8 þ T cell expansion after subcutaneous vaccination with lentivector-modified DCs. Effector CD8 þ T cell expansion was significantly increased 7 days after immunization when PD-L1 was silenced in DCs as observed by pentamer-staining in draining lymph nodes ( Fig 4G) and IFNg ELISPOT in splenocytes (Fig 4H). Numbers of pentamerpositive CD8 þ T cells after vaccination with PD-L1-silenced IiOVA-DCs remained higher than in non-silenced vaccination controls up to 3 weeks after vaccination, with high pentamer fluorescent intensities that probably reflect high TCR surface levels (Fig 5). These results suggested an accelerated or uncontrolled CD8 þ T cell activation/expansion in agreement with our in vitro studies using OT-I cells.
PD-L1 silencing in DCs accelerates anti-tumour immune responses Our observation that lentivector delivery of antigen with PD-L1 silencing induced potent inflammatory T cells suggested that these T cells could be a useful means of enhancing immunity in the context of anti-tumour therapy. Thus, we tested the therapeutic efficacy in an EG7 mouse lymphoma model (EL4 cells expressing OVA as a surrogate tumour antigen). Subcutaneous administration of EG7 cells results in aggressive solid lymphomas, which can be effectively treated by vaccination with IiOVA-expressing DCs . DCs were transduced with lentivectors expressing GFP alone, IiOVA-GFP or IiOVA-p5-GFP and were injected subcutaneously when tumours were detectable (on day 4 or 5; Fig 6A). The therapeutic immunization with as few as 10 5 IiOVA-p5-GFP transduced DCs significantly reduced tumour growth compared to non-silenced controls ( Fig 6B). As a consequence, a significant increase in lifespan of tumour-bearing mice was observed ( Fig 6C). Surprisingly, overall survival after 1 month was similar when compared to non-silenced DC counterparts (Fig 6D). Thus, our data strongly suggested that PD-L1 silencing inhibited tumour growth and prolonged survival, but did not increase cure rates. All surviving tumour-free mice pooled from three independent

Research Article
PD-L1 co-stimulation down-modulates the TCR on CD8 þ cells Figure 5. PD-L1 silencing in DCs accelerates expansion of endogenous antigen-specific CD8 R T cells. As in Fig 4G, flow cytometry analyses of endogenous OVA-specific CD8 þ T cells expanded after vaccination with the indicated transduced DCs (top of the graphs). T cells from draining inguinal lymph nodes were stained with a class I-specific OVA pentamer, and percentages of pentamer-specific OT-I cells, with their corresponding mean fluorescent intensities are shown in each graph. Analyses were performed 1 week after vaccination (top row, indicated on the right of the graphs), 2 weeks (middle row) and 3 weeks (bottom row). Vaccination groups consisted of between 3 and 5 mice. Plots correspond to individual mice. experiments were resistant to re-challenge with EG7 cells demonstrating effective recall responses (Fig 6E).

PD-L1 silencing potentiates the adjuvant capacities of molecular DC activators
To test whether anti-tumour immunity could be further improved, we combined PD-L1 silencing with the expression of selected DC molecular activators (Arce et al, 2011;Escors et al, 2008). PD-L1 shRNA p5 was co-delivered with IiOVA and a constitutive activator of mitogen-activated kinase (MAPK) p38 (MKK6EE) or the ERK inhibitor MEK1 DNES AA (Fig 7A). We chose these MAPK modulators because we have previously demonstrated that p38 activation and ERK inhibition using these modulators increase expression of co-stimulatory and adhesion molecules in DCs such as CD80, CD40 and ICAM I (Arce et al, 2011;Escors et al, 2008). Vaccination with a low dose of transduced DCs (10 4 /mouse) showed a synergistic increase in CD8 þ responses by IFN-g ELISPOT when p5 was co-delivered with MAPK modulators (Fig 7A). Immunization with as few as 10 4 PD-L1-silenced DCs activated with MAPK modulators resulted in effective control of tumour growth compared to mice receiving DC with IiOVA alone (Fig 7B). The combination of ERK inhibition with PD-L1 silencing was particularly effective in reducing tumour growth ( Fig 7B). Interestingly, PD-L1 silencing with either ERK inhibition or p38 activation significantly increased tumour-free survival (Fig 7C). This suggests that whilst PD-L1 inhibition can generate a T-cell response superior at controlling tumour growth, additional co-stimulatory signals (via DC activation) are required to optimize T-cell functionality for efficient tumour killing (Fig 7C). These data demonstrate that in the absence of PD-L1 silencing, as few as 10 4 activated DCs are sufficient to prime potent CD8 þ T cell responses, which can control tumour growth. We further show that by improving the potency of DCs, in conjunction with removal of PD-L1-based regulation, we can generate significantly enhanced T cell immunity, capable of controlling tumour growth in a therapeutic setting.

DISCUSSION
We have provided evidence that PD-L1 in antigen presenting DCs contributes to ligand-induced TCR down-modulation. Ligand-induced TCR down-modulation is a complicated process in which more than one mechanism is involved . Firstly, serially ligand-engaged TCRs internalize quickly

Research Article
Katarzyna Karwacz et al. Statistical significance was tested with the log-rank test. Ns, no significant differences. E. Kaplan-Meier survival plot-in surviving tumour-free mice from experiments as in (D), re-challenged with EG7 lymphoma cells. A control group with naive mice was used.
(from seconds to hours) after ligand-MHC binding (Huppa et al, 2010;Valitutti et al, 1995). Secondly, non-engaged TCR complexes can also be down-regulated following signal transduction from previously engaged (and endocytosed) TCR complexes (San Jose et al, 2000). It is still under debate whether TCR down-modulation is caused by internalization followed by degradation or by preventing TCR recycling back to the membrane San Jose et al, 2000;Valitutti et al, 1995). Interestingly, we have observed that TCR downmodulation progressively continues up to 3 days after initial antigen presentation by DCs. Subsequently, TCR surface levels recover again. This TCR down-modulation may be caused by either direct internalization or just a shift in recycling towards intracellular retention of TCR complexes (San Jose et al, 2000;Valitutti et al, 1995). Our results implicate PD-L1/PD-1 costimulation in the second, slower phase and they are in agreement with the following working model for regulation of T cell activation during antigen presentation by DCs (Fig 8): Firstly, antigen recognition (with additional co-stimulatory interactions) leads to TCR signal transduction. Following T cell activation, PD-1 is exposed to the T cell surface where it engages with PD-L1 on the surface of the APC. PD-1 signal transduction recruits SHP phosphatases and, importantly, leads to Cbl-b expression by a yet undefined pathway. Cbl-b ubiquitinates and inactivates key TCR signal transduction mediators, while removing TCRs from the T cell surface. Whether this removal is triggered by active internalization or a change in the dynamics of TCR recycling is yet unclear. Our data supports PD-1 as an early brake that fine-tunes T cell activation during antigen presentation after TCR signal transduction. In any case, our model can provide an explanation for results reported in several unrelated studies. Firstly, TCR/CD28 signal transduction per se is not required for full ligand-induced TCR down-modulation, as observed in Cbl-b KO CD8 þ T cells (Shamim et al, 2007). Secondly, PD-1 blocking antibodies inhibit Cbl-b transcriptional up-regulation, amongst other negative regulators of T cell activation as shown by polymerase chain reaction (after reverse transcription) (RT-PCR; Nurieva et al, 2006). Thirdly, human T cells activated in vitro with anti-CD3/anti-CD28 antibodies maintain high surface TCR levels when compared to antigen presentation by APCs (Yang et al, 2010). And fourthly, old human CD8 þ T cells, which maintain low surface TCR levels, significantly increase their surface expression after in vitro activation with anti-CD3/anti-CD28 antibodies only in the presence of PD-L1 blocking antibodies (Mirza et al, 2010). This last observation would implicate PD-L1/PD-1 homotypic T cell

Research Article
PD-L1 co-stimulation down-modulates the TCR on CD8 þ cells Statistical comparisons were performed using the Kruskall-Wallis non-parametric test. Ã , significant differences. C. As in (B), but representing the Kaplan-Meier survival plots. Vaccination groups are indicated within the plots. Statistical significance was tested with the logrank test. Ã , significant differences; ÃÃÃ , highly significant differences.
interactions (T cell-T cell) either in trans or in cis as suggested by others (Fooksman et al, 2010). The interpretation of all these results is, therefore, straightforward if PD-L1/PD-1 binding contributes to ligand-induced TCR down-modulation through Cbl-b expression. Unexpectedly, we did not observe significant ligand-induced TCR down-modulation after 3-4 days co-culture of OVA-specific CD4 þ T cells with IiOVA-expressing BM-DCs, whether PD-L1 was silenced or not. Accordingly, we were unable to test in the mouse model if PD-L1/PD-1 co-stimulation was involved in CD4 þ TCR down-modulation. It is worth noting that unlike CD8 þ T cells, only a minor proportion of OVA-specific OT-II T cells strongly associated to IiOVA-expressing DCs. This is in agreement with other reports in which naive CD4 þ T cells detach quickly from APCs, but they get fully activated later by multiple, short-lived contacts (Celli et al, 2005). However, it is likely that PD-L1/PD-1 interaction may also be involved in CD4 TCR downmodulation as we did observe inhibition of TCR downmodulation in PD-1-silenced human CD4 T cells, but in the context of superantigen stimulation.
PD-L1/PD-1 interaction is receiving an increasing interest as a therapeutic target. While most of the studies use systemic administration of blocking antibodies, a minor proportion have silenced B7 inhibitory molecules in DCs (Breton et al, 2009;Hobo et al, 2010). In any case, therapeutic outcomes vary considerably and are quite dependent on the specific experimental system. In many instances, combination with other strategies has to be applied to achieve effective therapy (Breton et al, 2009;Curran et al, 2010;Hobo et al, 2010;Pilon-Thomas et al, 2010). In our experimental system, PD-L1 silencing in DCs alone did not improve overall cure rates compared to nonsilenced DCs. PD-L1 silencing had to be combined with DC molecular activators to achieve increased therapeutic activities. Curiously, this phenomenon is observed in many other published studies and may account for the relative lack of efficacy of PD-L1/PD-1 blocking antibodies unless they are given in combination therapies. Thus, the anti-tumour efficacy of PD-L1/PD-1 blocking antibodies is enhanced by lentivector vaccination (Sierro et al, 2011;Zhou et al, 2010). In these two last reports, an increased recruitment of tumour infiltrating CD8 þ T cells was observed after systemic administration of PD-L1/PD-1 antibodies. In agreement with this, we also observed a tendency to increased intra-tumour CD8 þ T cell numbers in mice vaccinated with IiOVA-expressing PD-L1-silenced DCs. However, we could not demonstrate significance in relative numbers or TCR surface levels compared to vaccination with non-silenced DCs. In contrast to systemic administration of blocking antibodies, PD-L1 silencing in our experimental model is taking place in a limited number of DCs. Secondly, we could not differentiate a specific recruitment of effector CD8 þ T cells from just simply an overall increase in CD8 þ T cell numbers (Figs 4 and 5). However, in light to previous reports, it is likely that also in our case, there is an enhanced, active intra-tumoural recruitment of CD8 þ T cells.
Summarizing, our results demonstrate that PD-L1 silencing in DCs may speed-up T cell expansion and cytokine responses (which is important for tumour immunotherapy) but other factors also contribute to the effectiveness of the immune response, for instance the functionality of activated T-cells which is a direct consequence of a range of co-stimulatory signals during antigen presentation. Further co-stimulation could enhance the expression of cytotoxic molecules in effector CD8 þ T cells such as granzyme B and perforin, which could be responsible for the observed increase in therapeutic activity.
Our model highlights the contribution of PD-L1/PD-1 costimulation to ligand-induced TCR down-modulation during 'physiological' antigen presentation by DCs. This is a very complex process that is regulated at multiple levels, and there

Research Article
Katarzyna Karwacz et al. are other pathways that inhibit effector T cell activities without modifying TCR levels (Barber et al, 2006;Choi & Schwartz, 2007;Wherry, 2011). Defining the mechanism by which PD-1 engagement leads to Cbl-b up-regulation and modulates TCR levels is a key question in T cell physiology and a potential target for therapeutic intervention.

Cells and mice
293T, EG7 cells and BM-DCs were grown as described . Approval for the animal studies was obtained from the University College London Animal Ethics Committee. The following mouse strains were used: C57BL/6 mice, OT-I and OT-II mice. The last two strains express a transgenic TCR specific for MHC-class I and class II OVA epitopes, respectively. DC-T cell co-cultures were performed as described (Arce et al, 2011). T cells were added to DCs 3 days after transduction at a ratio of 10:1, and co-cultured for a minimum of 2 days and a maximum of 1 week. When indicated, isotype antibody control, purified anti-mouse PD-L1 clone MIH5 or anti-mouse PD-1 antibody clone RMP1-14 were added to DCs after lentivector transduction, at the concentrations indicated in the text. These last two antibodies are well-characterized blocking antibodies (Freeman et al, 2000;Kanai et al, 2003). Intracellular cytokine detection by flow cytometry was performed as described (Arce et al, 2011). Immunoblots using HA-specific antibodies were performed as described .

Research Article
PD-L1 co-stimulation down-modulates the TCR on CD8 þ cells The paper explained PROBLEM: Our study addresses a fundamental immunological mechanism for the control of T cell activation during antigen presentation, TCR down-modulation. This process limits TCR signal transduction and prevents T cell hyperactivation after antigenpresentation by dendritic cells. Current understanding of this process is that intrinsic TCR/CD28 signal transduction leads to TCR internalization. However, to date, no additional extrinsic signals have been identified which contribute to TCR downmodulation.

RESULTS:
We show here that binding of PD-L1 on dendritic cells to PD-1 on T cells is implicated in ligand-induced TCR down-modulation, by promoting Cbl-b E3 ubiquitin ligase expression. This finding expands the role of PD-L1 in immune regulation to a critical step in T cell activation.
We have further analysed the consequences of preventing TCR down-modulation, in vitro and in vivo in two very different disease models; antigen-specific inflammatory arthritis and a T cell-derived lymphoma. By combining PD-L1 silencing with modulators of selected signalling pathways (MAPKs) in dendritic cells, we have increased their anti-tumour activities, obtaining therapeutic effects with doses 100-to-1000-fold lower than those currently used in experimental cancer models and in human clinical trials.

IMPACT:
Our study will be of interest to a broad range of basic researchers in cell biology, immunology, rheumatology as well as clinicians. Our findings will be of particular importance for clinical application in autoimmune disorders and cancer, as they highlight TCR trafficking and its interference as a potential key therapeutic target.
Anexin V staining kit following the instructions of the manufacturer (BD Biosciences). Pentamer staining was performed as described (Arce et al, 2011).

Vaccination, T cell responses and tumour experiments
All vaccination experiments were repeated independently at least three times, using groups of five mice, as described (Breckpot et al, 2010;Escors et al, 2008;Goold et al, 2011). IFN-g ELISPOTs were performed as described . Therapeutic tumour experiments were performed as described before, using 10 mice per vaccination groups . Lentivectortransduced DCs at the indicated doses were injected subcutaneously when tumours were detectable (on day 4 or 5). Tumour growth was monitored daily and mice were sacrificed when tumour surface was higher than 140-150 mm 2 .
T cell purification, CFSE labelling, adoptive transfer and antigen-induced arthritis Untouched CD4 þ T or CD8 þ T cells were purified from mouse spleens using the Dynal Mouse CD4 or CD8 Negative Isolation Kit (Invitrogen). T cells were labelled with CFSE as described (Arce et al, 2011). Purified OT-I cells were adoptively transferred intravenously (between 10 6 and 10 8 cells/mouse) into groups of five mice, when indicated. After transfer, antigen-induced arthritis was triggered by intra-articular injection of 100 mg of purified OVA (Sigma) in 10 ml of PBS. Knee inflammation was monitored daily. The experiment was repeated twice.

Statistical analysis
ELISPOT data was analysed as described (Arce et al, 2011;Escors et al, 2008). Normally, a minimum of three independent experiments was performed with five mice per group. For tumour experiments, 10 mice per group per experiment were used. Mean fluorescence intensities from surface or intracellular staining were analysed as described (Arce et al, 2011;Escors et al, 2008). Survival data from tumour experiments was compared using the log-rank test as described before . CD3 surface expression levels, tumour size between groups and lifespan of tumour-bearing mice were compared using the non-parametric Mann-Whitney U-test or Kruskall-Wallis test for multicomparisons.

Author contributions
All authors were involved in drafting the article or revising it critically for important intellectual content. KK, MC, CLB and DE responsible for study conception and design. KK, CB, DMD, FA and DE performed data acquisition and interpretation.