Human Antigen-Specific Regulatory T Cells Generated by T Cell Receptor Gene Transfer

Background Therapies directed at augmenting regulatory T cell (Treg) activities in vivo as a systemic treatment for autoimmune disorders and transplantation may be associated with significant off-target effects, including a generalized immunosuppression that may compromise beneficial immune responses to infections and cancer cells. Adoptive cellular therapies using purified expanded Tregs represents an attractive alternative to systemic treatments, with results from animal studies noting increased therapeutic potency of antigen-specific Tregs over polyclonal populations. However, current methodologies are limited in terms of the capacity to isolate and expand a sufficient quantity of endogenous antigen-specific Tregs for therapeutic intervention. Moreover, FOXP3+ Tregs fall largely within the CD4+ T cell subset and are thus routinely MHC class II-specific, whereas class I-specific Tregs may function optimally in vivo by facilitating direct tissue recognition. Methodology/Principal Findings To overcome these limitations, we have developed a novel means for generating large numbers of antigen-specific Tregs involving lentiviral T cell receptor (TCR) gene transfer into in vitro expanded polyclonal natural Treg populations. Tregs redirected with a high-avidity class I-specific TCR were capable of recognizing the melanoma antigen tyrosinase in the context of HLA-A*0201 and could be further enriched during the expansion process by antigen-specific reactivation with peptide loaded artificial antigen presenting cells. These in vitro expanded Tregs continued to express FOXP3 and functional TCRs, and maintained the capacity to suppress conventional T cell responses directed against tyrosinase, as well as bystander T cell responses. Using this methodology in a model tumor system, murine Tregs designed to express the tyrosinase TCR effectively blocked antigen-specific effector T cell (Teff) activity as determined by tumor cell growth and luciferase reporter-based imaging. Conclusions/Significance These results support the feasibility of class I-restricted TCR transfer as a promising strategy to redirect the functional properties of Tregs and provide for a more efficacious adoptive cell therapy.


Introduction
Natural regulatory T cells (nTregs), defined by expression of the transcription factors FOXP3 [1] and more recently Eos [2], play a critical role in maintaining immune tolerance [3]. Recent interest has been directed at this population as a means for providing cellular therapy in settings of both autoimmunity and transplantation [4]. In comparison, relatively non-specific therapies such as anti-CD3, anti-thymocyte globulin, cytokines, and adhesion molecule based agents are associated with significant non-specific immunological effects and side-effects. Despite these limitations, important mechanistic findings have emerged from these immune therapies. First, short term modulation of T cells can elicit longterm effects on immune tolerance [5]. Second, many of these agents thought to provide long-lived efficacy are considered to do so through promoting Treg subsets [6].
Treg transfers have been shown to effectively prevent or even reverse disease, which is especially interesting in settings of autoimmunity where such disorders are thought to result from the genetic absence or loss of Tregs from the periphery [7,8]. A growing list of potential mechanisms of action for Tregs includes both cellcell contact dependent and independent mechanisms (reviewed in [9]). From these numerous pathways, a common theme has emerged that Tregs function by multiple mechanisms at the site of antigen presentation; both in draining lymphoid tissue and at the site of inflammation, to create a regulatory milieu that promotes bystander suppression and infectious tolerance [10,11]. In other words, activated Tregs are capable of suppressing cells recognizing common, as well as unrelated antigens in the local microenvironment and function to direct the development of activated conventional T cells (Tconv) toward a tolerant state. Discovering of the importance of Tregs in maintaining immune tolerance has opened a potentially new means of therapeutic intervention in immunology -namely, adoptive Treg therapy [12].
The concept of adoptive cell transfer initially emerged in the mid-1950s as a means to manipulate immune responses in the field of cancer therapy [13]. Cellular therapies have also been employed in patients with HIV using co-stimulated CD4 + T cells [14], as well as in the reconstitution of T cells following autologous and allogeneic stem cell transplantation [15]. While these treatments were designed to augment immune responses or stem cell recovery, a related approach involving transfer of T cells, specifically Tregs, has been employed to modify the immune response in autoimmune settings. The transfer of T cells from tolerized mice can prevent and/or reverse autoimmune disease [16]. The efficacy of transferred Tregs as a therapeutic modality has been demonstrated in animal models of systemic lupus erythematosus (SLE) [17], multiple sclerosis (MS) [18], inflammatory bowel disease (IBD) [19], oophoritis [20], and type 1 diabetes (T1D) [21,22,23], among others.
Although the therapeutic potential for Treg-based therapies is now well-established in animal models, to date adoptive Treg therapy has not been directly applied to suppress autoimmunity in humans. A number of challenges exist that need to be overcome in order to employ these cells in clinical settings, including the need for robust methods to isolate and expand this rare population. We recently described a clinically-relevant FACS-based isolation and in vitro expansion procedure for generating Tregs from patients with T1D using the markers CD4, CD25 (IL-2 receptor a-chain), and CD127 (IL-7 receptor a-chain) [24]. This approach was based on the observation that CD127 is inversely correlated with FOXP3 expression in humans [25,26]. Utilizing this method, we demonstrated that isolation of the CD4 + CD127 lo/2 CD25 + T cell fraction yields an enriched population of Tregs that can be stably expanded in vitro without the need for additional selective agents [25].
As previously discussed, numerous examples exist that antigenspecific Tregs are more effective in several autoimmune syndromes when compared to polyclonal populations [21]. With this, a major challenge in antigen-specific Treg therapy in humans lies in the isolation of sufficient quantities of antigen-specific T cells -a challenge that is not easily overcome in humans where sampling is largely limited to peripheral blood. Limited but appreciable progress has recently been made to address this need including the development of reagents allowing for specific labeling of known antigen-specific T cells (e.g., peptide-MHC multimer staining of autoreactive cells); however, this approach is limited by several factors including the frequency of antigen-specific T cells in the periphery and the limited availability of appropriate peptide-MHC multimers. Adaptive Treg, such as those of the T R 1-or T H 3-type, can be generated from Tconv cell populations by culture under selective growth media conditions (e.g., rapamycin, retinoic acid, TGF-b or IL-10) [27]. Other efforts to overcome limited cell numbers have focused on delivering the transcription factor FOXP3 to Tconv cell populations [28]. However, both of these approaches to generate de novo Tregs are not without limitations. Specifically, adaptive Treg populations often lack the epigenetic profile of nTregs (i.e., they are often methylated at the FOXP3-Treg specific demethylation region (reviewed in [29]), bringing their stability in vivo into question. In the case of FOXP3 gene transfer, expression of the transcription factor is required for Treg development, however transfer of FOXP3 to Tconv cells does not recapitulate the full repertoire of transcripts and functional characteristics of nTregs [30]. Therefore, the approach we took was to initiate efforts with a fully mature polyclonal nTreg population isolated from peripheral blood and redirect cell specificity by viral TCR gene transfer. This approach has been employed previously as a cancer immunotherapy studies which showed that TCR a and b-chains could be introduced into CD4 + and CD8 + T cells in order to create a tumor-directed Teff cell population that upon adoptive transfer destroyed pre-existing tumor mass [31]. This approach has been shown to augment the therapeutic profile of mouse Tregs in an arthritis model [32], as well as redirect human Tregs using a chimeric immunoreceptor [33]. In this study, we demonstrate that the specificity of polyclonal human natural Tregs can be redirected by lentiviral TCR gene transfer. Moreover, we demonstrate in a proof-ofprincipal study that CD4 + Tregs can be engineered with class Ispecific TCRs capable of suppressing immunity in vivo.

Generation of a lentiviral expression vector expressing a tumor-specific TCR
To express de novo TCRs in Tregs, we first generated an expression vector employing a previously characterized tumorreactive TCR (TyrTCR) that recognizes the melanoma antigen tyrosinase in the context of the MHC class I molecule HLA-A*0201 [34]. The multicystronic expression vector allowed for equimolar TCR a and b chains by the inclusion of a T2A element [35] followed by an IRES element and eGFP reporter ( Figure 1A). This receptor functions independently of the coreceptor; allowing for productive TCR signaling in CD4 + T cells [34,36,37]. Lentiviral transduction of Jurkat T cells was detected by flow cytometric analysis based on GFP reporter and tyrosinasespecific TCR Vb12 expression ( Figure 1B). Signaling by the introduced TCR was further validated by upregulation of CD69 and production of IL-2 following stimulation with artificial antigen presenting cells (aAPC, K562 cell line expressing HLA-A2, CD80, and CD83) and tyrosinase peptide ( Figure 1C,D).
Stability and functional capacity of TCR-redirected primary human CD4 T cells Given our ability to express a functional TCR in Jurkat T cells, we next sought to validate expression in primary human CD4 + T cells. Freshly isolated human CD4 + T cells were activated for 48 hr with anti-CD3 and anti-CD28-coated beads prior to transduction with lentivirus expressing the TyrTCR. GFP and TCR Vb12 co-expressing cells were isolated by fluorescence-based cell sorting (Figure 2A). Expression of surface TCR was stable following in vitro expansion ( Figure 2B) and the TyrTCRtransduced cells responded specifically to tyrosinase peptide in a dose-responsive fashion. No proliferation was observed from the native GFP 2 Vb12 + sorted population stimulated with tyrosinase peptide or from the GFP + Vb12 + T cell population incubated with an irrelevant HLA-A*201-restricted control peptide (MART-1) ( Figure 2C,D). Cytokine production (IFN-c, IL-10) by TyrTCRtransduced cells was also detected following co-culture with a melanoma tumor cell line expressing tyrosinase naturally processed and presented in the context of HLA-A*0201 (data not shown). The successful de novo expression of a functional TCR in primary human CD4 + T cells facilitated the subsequent targeting of this receptor to Tregs.

Generation of engineered human Treg and Tconv cells
We recently described a clinically-relevant FACS-based isolation and in vitro expansion procedure for generating Tregs enriched in FOXP3 expression in humans [25]. Utilizing this method, we demonstrated that isolation of the CD4 + CD127 lo/ 2 CD25 + T cells yielded a highly enriched population of Tregs that are greater than 90-95% FOXP3 + [24]. Subsequent studies suggested that the inclusion of CD45RA as a positive selection marker resulted in greater enrichment of FOXP3 + T cells and maximum stability following in vitro expansion [38,39] ( Figure  S1). The purity and stability of this sorted population was further validated by assessing FOXP3 protein and FOXP3 promoter methylation status following in vitro expansion. Prior reports have suggested that only bona-fide nTregs remain demethylated at the FOXP3 gene Treg-specific demethylation region (TSDR) [40]. CD45RA + Tregs were determined to be FOXP3 + by FACS (mean+SD, 92.3+9.6%, N = 8) and demethylated at the TSDR (91.0+15.3%) following a 14 day expansion period, demonstrating the specific expansion of the nTreg subset ( Figure S1G).
Next, the polyclonal population of purified nTregs was infected with lentivirus expressing a GFP reporter following activation as described in Materials and Methods. High levels of GFP expression were detected following in vitro expansion with no significant reduction in expansion capacity or suppressive function ( Figure S2). Given the high efficiency of transgene expression in Tregs, we subsequently transduced a population of purified naïve Tregs and Tconv cells with a lentiviral TyrTCR construct. Freshly sorted cells were activated for 48 hr in vitro prior to a single infection with lentivirus encoding the TyrTCR. Surface TCR and tetramer staining, along with GFP reporter expression were analyzed by flow cytometry at day nine of the in vitro expansion period. Initial transduction of Tregs resulted in a degree of heterogeneity in the level of TCR and GFP reporter expression. Flow cytometric analysis indicated 20.7616.6% of cells expressed eGFP, whereas 13.2614.5% of cells were GFP + Vb12 + following nine days of in vitro expansion (N = 10). Surface expression of TCR Vb12 might not indicate functional receptor expression, given the possibility for heterologous pairing with endogenous TCR a and b chains [41]. To further validate proper surface expression, HLA-A*201-tyrosinase tetramer staining of transduced Tregs revealed 33.5% of cells were GFP + TyrTCR tetramer + , whereas 16.8% of Tconv cells were GFP + TyrTCR tetramer + ( Figure S3B).
We observed that increasing lentiviral titer led to only modest increases in the number of cells expressing functional receptors (data not shown). Therefore, we sought to enrich antigen-specific T cells while limiting non-specific growth, as previously described [42]. At day nine, cells expressing the TyrTCR were restimulated with aAPCs and tyrosinase peptide. This antigen-specific activation led to enrichment of the transduced cells with 66.1% of Treg and 50.3% of Tconv cells expressing both GFP and the TyrTCR ( Figure 3A,B). Mock transduced cells were restimulated by anti-CD3 and anti-CD28 coated beads. The expansion capacity of TyrTCR Tregs expanded on aAPC and peptide was dependent upon the initial transduction efficiency (N = 3; median 480-fold, range 240 to 1088-fold). Of note, we were able to obtain .5200fold expansion of TyrTCR + Tregs yielding 1.18610 9 cells (89.3% eGFP + , 91.7% FOXP3 + ), from an initial blood volume of 80 ml by extending the culture to day 20 (data not shown).
Following expansion, 96.4% of mock-transduced Tregs stained positive for FOXP3 protein by FACS with 100% of cells remaining demethylated at the FOXP3-TSDR for the sample shown ( Figure 3C  Functional capacity of engineered Tregs The capacity of expanded TCR-transduced Tregs to suppress proliferation of freshly isolated conventional CD4 + T cells was examined following stimulation with anti-CD3 and anti-CD28. GFP + TyrTCR + Tregs were FACS sorted following 14 days of in vitro expansion and mixed at indicated ratios of Tregs to Tconv cells. TyrTCR Tregs suppressed polyclonal T cell proliferation potently and in a dose-responsive fashion ( Figure 4A). We next tested the capacity of TyrTCR Tregs to suppress antigen-specific CD4 + CD127 + CD25 2 Tconv cell responses. Following expansion, Treg and Tconv GFP + TyrTCR + cells were sorted from expanded cultures and restimulated with tyrosinase peptide in the presence of aAPCs ( Figure 4B). TyrTCR-expressing Tregs were capable of suppressing Teff cell responses in an antigen-specific fashion.
The therapeutic efficacy of Tregs depends on their capacity to suppress an array of T cell responses in a bystander fashion, as well as antigen-specific fashion. In order to address this issue, we tested the capacity of TyrTCR-transduced Tregs to suppress prolifera- tion of hemagglutinin (HA) peptide-reactive Teff cells. To generate human HA-reactive Teff cells, we labeled PBMCs from a HLA-DR*0401 individual with CFSE and stimulated with HA peptide (306)(307)(308)(309)(310)(311)(312)(313)(314)(315)(316)(317)(318) to induce expansion of endogenous HA-specific T cells. Following 14 day expansion, CFSE-diluted cells (.13% HLA-DR*0401 HA-tetramer + ; data not shown) were used as responding Teff cells in subsequent assays. For suppression assays, TyrTCR Tregs were stimulated in the context of HLA-A*0201, while the responding Teff cells were stimulated in the context of HLA class II DR*0401 APCs incubated with HA peptide (306)(307)(308)(309)(310)(311)(312)(313)(314)(315)(316)(317)(318) . To test the specific activation requirements of Tregs in mediating bystander suppression, TyrTCR-transduced Tregs were stimulated with tyrosinase peptide or the irrelevant HLA-A*0201 restricted MART-1 peptide. This analysis indicated that TyrTCR-transduced Tregs were indeed capable of suppressing HA-reactive Teff cell proliferation in a bystander fashion and demonstrated that TyrTCR cells maintained optimal potency when activated specifically through the introduced TCR ( Figure 4C). This was particularly apparent at lower and more physiological ratios of Tregs to Teff cells (i.e. 1 / 4 :1, 1 / 8 :1, and 1 / 16 :1) ( Figure 4D). Despite this increased potency, we observed some antigen non-specific suppression by TyrTCR Tregs in vitro (particularly at higher Treg to Teff cell ratios, i.e. 1:1), suggesting that Tregs may not be fully rested following the activation received during the in vitro expansion protocol. We also analyzed the ability of TyrTCR Tregs to influence cytokine production (e.g. IFN-c, IL-2, and IL-10) when co-incubated with a tumor infiltrating cell line (TIL1235) and peptide pulsed APCs. Consistent with our prior investigations [24,44], TyrTCR transduced Tregs suppressed effector T cell cytokine production in a dose-responsive fashion for IFN-c and IL-2. Interestingly, the reciprocal cytokine production profile was seen for IL-10, wherein co-cultures with TyrTCR Tregs, but not mock Tregs produced the highest levels of this immunoregulatory cytokine ( Figure S4). While in vitro suppressive activity is an important indication of Treg function, it does not reflect the full activity of Tregs in vivo. Therefore, we sought to test whether TCR-transduced Tregs were capable of suppressing T cell-mediated tissue damage in vivo.

In vivo activity of TyrTCR expressing Tregs
We employed a tumor model system to assess the in vivo function of TyrTCR-redirected Tregs. EL-4-HLA-A2/K b tumors expressing tyrosinase were transferred into HLA-A2/K b transgenic mice. Mice were then separated into groups receiving in vitro expanded murine TyrTCR Tregs, mock vector-transduced Tregs, or control vector-transduced Teff cells via intravenous injection. All mice received antigen-specific Teff cells expressing the TyrTCR. These TyrTCR Teff cells could be tracked in vivo by a firefly-Luciferase (fLuc) reporter element encoded within the expression construct ( Figure 5A). The murine TyrTCR expression construct was modified for optimized surface expression and improved tetramer reactivity (R. Koya, unpublished observations). Specifically, the TyrTCR construct was modified to express murine TCR constant regions and human TCR variable regions [34]. Additional modifications consisted of introducing cysteine residues to the TCR a and b chains to induce disulfide linkages, 2A peptide linker sequences, and leucine zipper motifs between TCR chains as described in Materials and Methods. The modified TyrTCR could be expressed on murine CD4 + T cells with expression by expanded Tregs verified by HLA-A*201-Tyrosinase (368-376) tetra-mer staining following in vitro expansion with anti-CD3 and anti-CD28-coated microbeads and IL-2 ( Figure 5B).
We monitored the expansion and distribution of TyrTCRexpressing Teff cells using an in vivo luciferase reporter in treatment groups receiving TyrTregs, mock-vector transduced Tregs, or mock-vector transduced Teff cells. Imaging during the experimental time-course revealed broad TyrTCR Teff cell distribution and tumor infiltration in mice receiving either polyclonal Tregs or Teff cells ( Figure 5C). In contrast, TyrTCR Teff cell expansion was restricted in mice receiving antigen-specific TyrTCR Tregs. This was most apparent at day eight post transfer with 5.6610 5 , 3.1610 6 , and 3.9610 6 photons/sec/cm 2 /sr detected for TyrTCR Treg, polyclonal Treg, and Teff cells, respectively ( Figure 5D). No significant difference in luciferase reporter was detected between Teff and polyclonal Treg treated mice.
TyrTCR-redirected Teff cells are highly effective at facilitating tumor regression upon transfer. We tested the capacity of antigenspecific Tregs to block tumor immunity in mice receiving TyrTCR Tregs, polyclonal Tregs, or control Teff cells. Mice receiving TyrTCR-redirected Tregs effectively blocked TyrTCR Teff cellmediated tumor rejection as determined by tumor size ( Figure 5E). In contrast, a partial suppressive effect was observed in mice treated with polyclonal Tregs, with near complete tumor rejection observed in mice receiving Teff cells alone. These data argue that TCR-redirected Tregs are indeed capable of suppressing Teff cell responses in vivo, and further demonstrate that antigen-specific Tregs are more potent than polyclonal Tregs in limiting Teff-cell mediated tissue destruction.

Discussion
Here we describe a novel method to generate therapeutically relevant numbers of antigen-specific human Tregs. These TCRredirected Tregs can be selectively enriched in vitro by reactivation with APCs and peptide to reach clinically viable cell numbers (i.e., in the range of 10 9 cells) [12]. Lentiviral infected Tregs expand with similar kinetics to mock-infected Tregs and maintain the phenotypic and functional characteristics of nTregs in vitro. Specifically, antigen-specific Tregs suppressed Teff cells recognizing a common antigen (tyrosinase), in addition to bystander Teff cell responses. Importantly, the suppressive phenotype of human Tregs expressing the TyrTCR was recapitulated in vivo using murine Tregs and antigen-specific Teff cells.
Recent studies suggest that Tregs may exert direct tissue protective effects, in addition to their ability to limit Teff cell responses. Mechanistically, Tregs are thought to accomplish this through enzymatic pathways that reduce oxidative stress and produce anti-inflammatory byproducts [45,46,47,48]. Here we show that Tregs can be directed to recognize MHC class Irestricted antigens allowing direct tissue recognition by transfer of a high-avidity TCR. This capability allows Treg activity to be directed to specific tissue microenvironments and target tissues deficient in APCs expressing MHC class II molecules. While not specifically addressed here, the use of TCR gene transfer to Tregs should facilitate studies addressing the optimal affinity of receptors as well as the lineage stability of Tregs in humanized animal model systems [49].
This technology opens up avenues for additional transgenes to be transferred in addition to the TCR as therapeutic gene cassettes. This approach could extend the function of engineered Tregs beyond that of even native Tregs. For example, the desired TCRs may be expressed along with a combination of immunoregulatory cytokines (IL-10 or TGF-b), chemokine receptors or tissue-homing receptors, or even tissue regenerative growth factors if necessary.
The use of viral gene transfer in Treg therapy settings confers the ability to address several important safety concerns. First, while a high purity of FOXP3 + Tregs was detected following our expansion procedure, the concern remains that an autoreactive TCR could potentially be expressed in a contaminating Teff cell or a Treg cell that has lost regulatory function [50]. Our approach could reinforce the regulatory program by the expression of FOXP3, or other regulatory genes in our expression cassette. Second, our system allows for the inclusion of inducible suicide genes or depletion markers should situations involving leukemia or pan-immunosuppression result.
One potential concern with this approach is the possibility for mismatched pairing of the introduced a and b TCR chains with the endogenous TCR subunits, resulting in reduced surface expression and biological activity. Mismatched pairing of receptors could potentially also generate harmful reactivities when expressed in effector T cell subsets. Cohen and colleagues previously addressed this issue by adding a single cysteine on each receptor chain to promote the formation of an inter-chain disulfide bond [51]. These cysteine-modified receptors retain their function and lead to preferential pairing, thus interfering with heterologous chain pairing between introduced and endogenous receptors [51]. Strategies such as codon optimization of TCR a and b chains have also been used to optimize de novo TCR expression over the endogenous receptors (reviewed [52]). The tyrosinase TCR used in our animal model experiments incorporated many of these modifications resulting in improved pairing and in vivo function.
In summary, novel gene therapy approaches, in concert with the previously outlined cellular therapeutics may offer new opportunities to generate antigen-specific tolerance by targeting the regulatory properties of Tregs to specific antigen targest. Prior studies employing adoptive cell therapy in the cancer setting imply that large numbers of transferred cells may be required in order to elicit a therapeutic effect [12,53]. Here we demonstrated a strategy to efficiently generate therapeutic numbers of antigen-specific Tregs from low starting peripheral blood volumes. Applications of this strategy to engineer Tregs capable of recognizing autoantigens targeted in T1D are currently underway, with the goal of developing robust and targeted immunotherapies for the treatment of autoimmune diseases and transplantation.

Patient Population
Patient samples were collected from normal healthy adult subjects from the general population (N = 10; 5M/5F, median age 28.58 yr, range 20.71 to 34.71).

Sample processing
Fresh peripheral blood was collected in sodium heparinized vacutainer tubes (BD; Franklin Lakes, NJ) and processed immediately for isolation of lymphocyte subpopulations. Whole blood was pre-enriched with the RosetteSep CD4 + negative selection cocktail and processed according to manufacturer recommendations (StemCell Technologies, Vancouver, BC, Canada). As an additional purification step, CD4 + T cells were incubated with CD45RO-biotinylated antibody (Leinco Technologies, St. Louis, MO; 0.1 mg/ml per 1610 ' 6 cells). CD45RObiotin labeled CD4 + T cells were then bound with BioMag streptavidin microbeads (50 ml/10 8 cells) followed by a 5 minute incubation and subsequent depletion on the EasySep magnet (StemCell). Untouched CD4 + CD45RA + T cells were then resuspended at 1610 8 cells per ml and stained with antibodies in PBS containing 10% FBS for FACS isolation.

Lentiviral vectors for expression of TCRs
The generation of the tyrosinase TCR (TyrTCR) construct was previously described [34,54]. Expression of the tyrosinase-reactive TCR a and b-chain genes in Treg and Tconv cells was achieved by cloning the TCR construct into the pSICO lentiviral vector [35]. This receptor is derived from a CD4 + T cell but functions independently of the co-receptor [34]. Equal molar expression of TCR a and TCR b-chain genes was made possible by the inclusion of a T2A expression element between the TCR a and b chain genes [35]. A 1857 bp EcoRI-XhoI TyrTCR fragment was generated by PCR from the original vector and ligated with a 1303 bp XhoI-NotI internal ribosome entry site (IRES) and eGFP reporter gene into the pSICO backbone. The pSICO construct was driven off the murine leukemia virus-derived MND promoter. All constructs were validated by restriction enzyme digest and sequence analysis.
In vitro expansion procedure FACS isolated cells were plated at 2.5610 5 Tregs or Tconv cells per well in a 24-well plate (Costar; Cambridge, MA) and activated with anti-CD3/anti-CD28 coated microbeads (Invitrogen; Carlsbad, CA) at a 1:1 bead to cell. At 48 hr, cultures were split into mock transduced and lentiviral TCR transduced populations. Cultures were grown in the presence of IL-2 following day 2 and continuing throughout the 14-day expansion period (300 IU/ml, Proleukin; Chiron Therapeutics, Emeryville, CA). Cells were resuspended and fresh media and IL-2 (300 IU/ml) added at days 5, 7, 9, and 12 assuming consumption of IL-2. On day 9, cells were restimulated with fresh anti-CD3/anti-CD28 coated beads for mock transduced cells at a 1:1 ratio or with artificial APCs (aAPC), a K562 cells line expressing HLA-A*0201, CD80, CD83 (A kind gift of Lee M. Nadler, Boston, MA) and soluble tyrosinase peptide (0.1 mg/ml, [Asp 370 ]-Tyrosinase (368-376) ; AnaSpec, Inc., San Jose, CA). Cultured cells were harvested at day 14 for analysis by FACS and functional studies.

Lentiviral production and transduction of T cells
Lentivirus was produced as previously described [54]. Activated T cells were transduced 48 hr following polyclonal activation with anti-CD3 and anti-CD28 coated microbeads (Dynal, Invitrogen). 2.5610 5 cells per well were transduced in one ml total volume of fresh culture media in a 24-well plate containing lentivirus (20 TU/cell) and protamine sulfate (8 mg/ml; Sigma-Aldrich, St. Louis, MO). Cells were spinnoculated by centrifugation at 10006g for 30 min at 32uC. Transgene expression was assessed no earlier than 72 hrs post-transduction.

Phenotypic analysis of expanded human Treg populations
Freshly expanded cells were evaluated for continued expression of CD4 (SK3), eGFP, TCR Vb12 (clone VER2.32.1) or HLA-A*0201 tyrosinase (YMDGTMSQV) tetramer (Beckman Coulter, Fullerton CA). Anti-human FOXP3 (clone 206D) was purchased from BioLegend (San Diego, CA) and intracellular staining performed with the FOXP3 staining kit according to manufacturer's instructions and modified as previously described [25]. Flow cytometric data was collected on a FACSCalibur cytometer (BD Biosciences) and analyzed with FlowJo software (version 7.2.2, TreeStar, Ashland, OR). Epigenetic analysis of the FOXP3-TSDR was conducted by a methylation-sensitive real time PCR method developed by Epiontis Corporation (Berlin, Germany).

Retroviral transduction of murine T cells and adoptive cell transfer (ACT)
High-titer retrovirus stocks were prepared by transient cotransfection of Phoenix-Eco cells as previously described [56]. Retroviral constructs were based on a MSCV vector backbone containing a hybrid TyrTCR with complete murine constant region substitutions, and addition of cysteine residues on each chain as described [51]. Leucine zipper motif sequences were added to the 39 terminus of each chain to improve pairing and 2A sequences inserted between the TCR a, b chain, and marker transgenes [35]. For effector T cell generation, cells from spleen and lymph nodes of 8 weeks-old HLA-A2/K b transgenic mice were harvested and RBC-depleted with Red Blood Cell Lysing Buffer (Hybri-Max, Sigma, St. Louis, MO). Cells were then incubated in anti-CD3 and anti-CD28-coated plates (BD Biosciences) for activation of T cells. At 48 hours post-activation, cells underwent spinoculation with retrovirus supernatants (10 multiplicity of infection) in retronectin (Takara-bio, Shiga, Japan)coated plates at 10006g, 120 minutes, 32uC, in a Beckman CS-6R centrifuge. The transduction procedure was repeated again following 24 hours and incubated overnight. Cells were then washed and re-suspended in culture media for analysis or in PBS for injection. Eight-week old HLA-A2/K b transgenic mice were injected subcutaneously (s.c.) with EL4-A2/K b cells expressing tyrosinase protein were used for adoptive cell transfer when the tumors reached 5 mm in diameter. Mice were conditioned with myeloablation (900 cGy total-body irradiation), followed by rescue with intravenous injection of 10 6 total bone marrow cells. On the next day (day 0), mice were randomized into control groups with Treg cells and T effector cells transduced with control mock vector and treatment group with Treg cells transduced with Tyr-TCR/ TK/GFP vector for intravenous injection. Dendritic cells (DC) were differentiated from bone marrow progenitor cells obtained from HLA-A2/K b mice by in vitro culture in murine granulocyte macrophage colony-stimulating factor (GM-CSF, 50 ng/ml) and murine IL-4 (50 ng/ml; R&D Systems, Minneapolis, MN) as described [57]. DCs were harvested as loosely adherent cells, washed twice in PBS (Mediatech) and pulsed with tyrosinase (368-376) peptide at a concentration of 10 mM in serum-free media for 90 minutes at room temperature. Each mouse received 10 5 pulsed DCs given s.c. on day 0. On day 2, all mice received T effector cells transduced with Tyr-TCR/firefly (f)Luciferase vector intravenously. Recombinant human IL-2 (250,000 I.U.) was injected intraperitoneally on days 0, 1 and 2. HLA-A2/Kb transgenic mice were originally obtained as a kind gift from Dr. Linda Sherman (Scripts, La Jolla, CA).

Luciferase-based in vivo Bioluminescence Imaging (BLI)
BLI was performed with a Xenogen IVIS 200 Imaging System (Xenogen/Caliper Life Sciences, Hopkinton, MA). Mice were anesthetized with 2% isoflurane and injected with D-luciferin (Xenogen; 1,500 mg per animal) for imaging of fLuc-transduced T cells. Imaging was performed 5 minutes later and analyzed with Living Image 2.50 software (Xenogen). The time of acquisition varied from 3 to 5 minutes. The mean number of photons per square centimeter per second per steradian in the region of interest (ROI) was determined on tumor-implanted sites.
Cytokine Analysis. Supernatant was collected from TCR transduced Jurkat T cells and analyzed for IL-2 production by standard sandwich ELISA according to manufacturer recommendations (BD Biosciences).

Statistics and methods of analysis
Multiple group comparisons were conducted by ANOVA with Bonferroni post tests. Data analyses utilized GraphPad Prism 5.00 software (GraphPad, San Diego, CA, USA) and values at P,0.05 were deemed significant. Cytokine concentrations were determined using SOFTmaxH PRO software (Molecular Devices, Sunnyvale, CA) with four-parameter data analysis. Figure S1 Representative FACS plots showing sorted human CD4+CD25+CD45RA+ Treg and CD4+CD25-CD45RA+ Tconv T cells. Human peripheral blood was collected in sodium heparin vacutainer tubes. CD4+ T cells were enriched by incubating whole blood (80 ml) with the CD4 negative selection cocktail (50 ml/ml) followed by ficoll density gradient centrifugation. Following CD4 enrichment, cells were incubated with antihuman CD45RO-biotinylated antibody (0.1 mg/16106 cells) followed by magnetic depletion with streptavidin-coated microbeads. The resulting negatively selected CD4+CD45RA+ T cell population was stained for CD4, CD25, CD127, and CD45RA and sorted on a FACS Aria II cell sorter. Shown are pre-and postsort expression of (A-C) CD4 (x-axis) and CD45RA (y-axis) and