Programming tumor-reactive effector memory CD8⁺ T cells in vitro obviates the requirement for in vivo vaccination

Adoptive cell transfer (ACT) represents a promising approach for the treatment of patients with relapsed hematologic malignancies, posttransplantation viral diseases, retroviral-induced immunodeficiency, and advanced solid cancers.^1–6  In mice and humans, antigen (Ag) restimulation with vaccination after ACT significantly potentiates the in vivo antitumor efficacy of adoptively transferred self-/tumor-reactive CD8⁺ T cells.^7–12  Efficient translation of this approach to human trials has been limited by a number of technical challenges. For example, tumor-infiltrating lymphocyte (TIL) cultures often have specificity against multiple epitopes rather than a single, defined Ag.¹³  In addition, TILs may contain reactivity against uncharacterized, autologous tumor-specific Ags rather than well-defined tumor-associated Ags,^13–15  making the prospect of coadministering an Ag-specific vaccine impractical. Finally, nearly all T cells transferred in current human protocols contain a T-effector cell (T_EFF) or T-cell effector memory (T_EM) phenotypic population characterized by low or absent expression of the lymphoid trafficking molecules CD62L and CCR7.^16,17  In preclinical studies, transfer of CD62L^hiCCR7⁺ cells with a naive, central memory (T_CM), or stem cell memory CD8⁺ T cell (T_SCM) phenotype provide for superior antitumor treatment compared with T_EM or T_EFF cells when combined with in vivo tumor Ag vaccination.^18–23  Obtaining an Ag-specific population of CD62L^hiCCR7⁺ T cells for use in human trials has thus far proven elusive.^3,24 

One potential approach to overcome these limitations could be the programming of CD8⁺ T cells in vitro before ACT.²⁵  Experiments in mice have shown that application of a brief T-cell receptor (TCR) stimulus to naive CD8⁺ T cells is sufficient to initiate a highly stereotyped response leading to clonal proliferation,^26–31  acquisition of effector functions,^27–31  and ultimately formation of a memory population.^27,30,31  Initiation of a programmed response with an in vitro stimulus immediately before cell infusion has the theoretical benefit of minimizing in vitro differentiation before ACT. Furthermore, a nonspecific stimulus might be used to stimulate cells with diverse or unknown specificities, such as TIL. Lastly, because programmed cells do not require further activation cues to execute an effector response, it may be possible to bypass the previous requirement to obtain T_CM- or T_SCM-like cells to achieve optimal antitumor treatment when coadministering a vaccine.^18,23 

We sought to determine whether programming tumor-reactive CD8⁺ T_EM cells in vitro either with peptide-pulsed Ag-presenting cells (APCs) or combined anti-CD3/anti-CD28 stimulation before ACT could substitute for in vivo vaccination using a model based on the transfer of TCR transgenic pmel-1 CD8⁺ T cells reactive against the shared self/tumor Ag gp100.⁷  Our results demonstrate that provision of a brief (< 24 hours) programming stimulus to T_EM in vitro instructs cells to undergo a process of clonal expansion and acquisition of effector functions leading to the destruction of large, established tumors with near-equivalent efficacy as T_EM cells restimulated in vivo with vaccination. The programmed response was dependent on an interval of Ag-independent interferon-γ (IFN-γ) release that facilitated dynamic changes in tumor surface expression of the major histocompatibility complex class I molecule H-2D^b, subsequent tumor recognition, and ultimately cytolysis.

Pmel-1 TCR transgenic mice⁷  were crossed to B6.129S7-Ifng^tm1Ts/J³²  and B6.129S7-Ifngr1^tm1Agt/J³³  mice (all from The Jackson Laboratory) to derive pmel-Ifng^tm1Ts/J and pmel-Ifngr1^tm1Agt/J mice (henceforth described as pmel-Ifng^−/− and pmel-Ifngr1^−/−, respectively). Female C57BL/6 mice (The Jackson Laboratory) were used at 6 to 12 weeks of age as recipient mice in ACT experiments. All animal experiments were conducted with the approval of the National Cancer Institute Animal Use and Care Committee. B16 (H-2^b), a gp100⁺ spontaneous murine melanoma, was obtained from the National Cancer Institute tumor repository and maintained in complete media (CM) as described previously.⁷  The gp100⁻ cell lines MCA-205 and EL-4 (National Cancer Institute Tumor Repository) were also maintained in CM and used as irrelevant H-2D^b targets.

T_EM cells differentiated in vitro were produced as described elsewhere.^18,34–36  In vitro programming was accomplished in the following manner. In all experiments, day 7 pmel T_EM cells were frozen. For T_EM cells programmed with peptide stimulation, cells were thawed on days −1, −3, −5, and −7 before ACT, rested overnight in CM containing 5 ng/mL recombinant human interleukin-2 (rhIL-2; Chiron), and finally restimulated with 30 Gy irradiated C57BL/6 splenocytes pulsed with 1 μM human gp100_25-33 (hgp100_25-33) peptide at a ratio of pmel/irradiated-splenocytes of 1:10. All cell cultures were started with 10⁶ pmel T_EM cells per recipient mouse. The total number of cells after in vitro restimulation were adoptively transferred. For T_EM cells programmed in vitro with anti-CD3 and anti-CD28 monoclonal antibodies, 24-well plates were coated with 1 μg/mL anti-CD3ϵ (145-2C11) and 0.5 μg/mL anti-CD28 (37.51; both monoclonal antibodies from BD Biosciences) diluted in 1 mL sterile phosphate-buffered saline at 4°C overnight. The plates were washed with CM, and 2 × 10⁶ thawed pmel T_EM cells were added to each well. Cells were cultured overnight at 37°C in rhIL-2–containing CM and then adoptively transferred.

Where indicated, B16 was pretreated overnight with titrated doses of IFN-γ (PeproTech) before use in coculture, cytolytic, or fluorescence-activated cell sorter (FACS) analysis experiments. To assess dynamic changes in H-2D^b expression on B16 in response to soluble factors produced by pmel T_EM, a 24-well Transwell plate system (Corning) was used. A total of 10⁵ B16 cells/well were plated in the lower chamber of the Transwell plate and 10⁵ pmel or pmel-Ifng^−/− T_EM cells with or without restimulation with 10⁶ irradiated C57BL/6 splenocytes pulsed with 1 μM hgp100_25-33 peptide were plated in the upper chamber. B16 tumor cells were collected 48 hours later and evaluated for H-2D^b expression by FACS. ⁵¹Cr-release assays were performed as described in detail elsewhere,²²  except that the coculture was performed over 8 hours. Assessment of IFN-γ release by in vitro programmed pmel T_EM cells was accomplished by restimulating pmel T_EM with hgp100_25-33 peptide-pulsed irradiated splenocytes on day 0. Cells were harvested on days 1 through 6 after restimulation, isolated with Ficoll solution (GE Healthcare), and then plated at a 1:1 ratio of 10⁶ target-to-effector cells in an overnight coculture against indicated targets. A previously described H2-D^b–restricted epitope (βgal_96-103) was used as a negative control.³⁷  Supernatants were harvested and assessed for IFN-γ release as described elsewhere.⁷ 

Before ACT, all tumor-bearing recipient mice received sublethal irradiation (500 cGy) on the day of treatment. Cells prepared as described above in “In vitro programming and adoptive cell transfer” were adoptively transferred. All treated mice received rhIL-2 administered by an intraperitoneal injection twice daily at 36 μg per dose for a total of 6 doses. Where indicated, mice also received vaccination with a recombinant fowlpox virus expressing hgp100 (rFPhgp100).⁷  Enumeration of adoptively transferred cells was performed as described elsewhere.^19,38  Experiments were performed in a blinded, randomized fashion and performed independently at least twice with similar results.

B16 was stained with H-2D^b–FITC (KH95; BD Biosciences) as described.⁷  Samples were analyzed using a FACSCalibur flow cytometer and analyzed using FlowJo software (TreeStar).

The products of the perpendicular diameters for tumors were plotted as the mean plus or minus SEM. Tumor graphs and tissue partitioning were compared using the Wilcoxon rank-sum test. Differences in the absolute number of pmel cells after ACT and IFN-γ production were compared using a nonpaired t test. The correlation between D^b expression and cytolysis was calculated using log-regression statistics after the data were projected on a semilog plot.

We have previously shown in the pmel-1 model that optimal treatment of large, established B16 melanoma requires the tripartite regimen of adoptively transferred antitumor CD8⁺ T cells, exogenous γ_c signaling cytokine support (eg, IL-2, IL-7, IL-15, or IL-21), and either in vivo vaccination or myeloablation with hematopoetic stem cell rescue.^{7,9,22,39–41}  To test the hypothesis that provision of a brief (< 24 hours) programming stimulus in vitro could replace the in vivo requirement for tumor-Ag vaccination, we transferred pmel T_EM cells restimulated with hgp100_25-33 peptide-pulsed irradiated APCs for 24 hours before ACT. Pmel T_EM cells that did not undergo in vitro programming were transferred alone or in combination with rFPhgp100 as a negative and positive control for Ag restimulation, respectively. All treated mice received exogenous IL-2. To provide for an adequate treatment window, we intentionally transferred 10⁶ pmel T_EM cells, a dose previously established to be noncurative in our model.¹⁸ 

The combination of nonrestimulated pmel T_EM cells and IL-2 alone caused only a modest delay in tumor growth compared with untreated controls (Figure 1A; P < .01). By contrast, provision of pmel T_EM cells programmed in vitro with peptide before transfer significantly potentiated the antitumor response compared with nonrestimulated cells (P < .03). Notably, the benefit provided by in vitro programming was comparable with results obtained with the addition of rFPhgp100 vaccination to unprogrammed T_EM cells (P = .6).

To test whether an Ag-nonspecific programming stimulus was also sufficient to commit tumor-reactive cells to an effective antitumor program, we stimulated pmel T_EM cells with plate-bound anti-CD3/anti-CD28 for 24 hours before ACT. Again, we observed that in vitro programming of pmel T_EM cells could significantly augment their in vivo antitumor activity compared with nonrestimulated controls (Figure 1B; P < .05). Further, there was no significant difference in tumor treatment between mice that received anti-CD3/anti-CD28 programmed T_EM cells and the positive control of T_EM cells and rFPhgp100 vaccination (P > .9).

Finally, because the expansion of responding tumor-reactive CD8⁺ T cells in vivo has previously been correlated with antitumor treatment efficacy in both mice and humans,^{17–19,38,40,42}  we sought to evaluate the in vivo proliferative response of adoptively transferred in vitro–programmed CD8⁺ T cells compared with unprogrammed cells with or without in vivo rFPhgp100 vaccination. For groups that received pmel T_EM cells programmed with anti-CD3/anti-CD28 before transfer or pmel T_EM cells restimulated in vivo with rFPhgp100, we observed a massive expansion of self-/tumor-reactive CD8⁺ T cells in both the spleen (Figure 1C) and lymph nodes (supplemental Figure 1A, available on the Blood website; see the Supplemental Materials link at the top of the online article) of recipient mice peaking on days 5 and 6 after transfer. By contrast, pmel T_EM cells that did not receive either an in vitro or in vivo Ag restimulation did not expand to any measurable extent in either tissue after adoptive transfer. A similar pattern of in vivo proliferation was observed when cells were programmed with hgp100_25-33 peptide-pulsed irradiated APCs (data not shown). A comparison of the relative partitioning of transferred cells between the spleens and lymph nodes of recipient animals revealed no significant differences in the lymph node trafficking behavior of programmed compared with vaccine-stimulated pmel T_EM cells (supplemental Figure 1B). Depending on the individual experiment, there were relatively more T cells in either the in vitro programmed or the in vivo restimulated pmel T_EM groups; however, the overall kinetic pattern remained very similar across 3 independently performed experiments and matched previously reported results.^9,18,19,38 

Taken together, these data demonstrated that provision of a 24-hour in vitro programming stimulus to tumor-reactive CD8⁺ T_EM cells in either an Ag-specific (cognate peptide) or nonspecific (anti-CD3/CD-28) manner was sufficient to incite a program of in vivo clonal expansion and acquisition of effector functions leading to tumor destruction nearly equivalent to in vivo restimulation with vaccination.

Naive CD8⁺ T cells require as little as 2 to 3 hours of TCR-based instruction to commit to a program of Ag-independent proliferation and acquisition of effector functions.^26,28  However, the optimal programming duration to provide for maximal proliferation, IFN-γ release, and cytolytic function has been reported to range from 20 hours to 48 hours or longer.^29,31  Further, analogous experiments have yet to be conducted using memory phenotype CD8⁺ T cells in the context of a secondary recall response. We therefore sought to determine the optimal timing between initiation of a programming stimulus to pmel T_EM cells and ACT using tumor treatment efficacy as a read-out.

Pmel T_EM cells were in vitro programmed with hgp100_25-33 peptide-pulsed irradiated APCs for 12 hours, 24 hours, 3 days, 5 days, and 7 days before ACT. In each group, cultures were initiated with 10⁶ pmel T_EM cells per recipient mouse. To control for a loss of replicative potential of cells that had undergone division in vitro, we transferred the total yield of cells obtained at the conclusion of the programming stimulus. Provision of a brief (12-24 hours) programming stimulus to pmel T_EM cells significantly augmented their in vivo antitumor activity compared with nonprogrammed T_EM cells (Figure 2A; P < .01 for both groups). Furthermore, there were no significant differences between pmel T_EM cells peptide-programmed for 12 hours and 24 hours before transfer and the positive control of pmel T_EM cells restimulated in vivo with rFPgp100 vaccination (P > .3 for all comparisons). In contrast, there was a complete loss of antitumor treatment efficacy compared with nontreated controls for T_EM cells transferred after restimulation on day 3 or later (data not shown; P > .05 for all comparisons).

We next evaluated the persistence of nonstimulated pmel T_EM cells compared with cells that had undergone in vitro programming with peptide-pulsed irradiated APCs for either 24 hours or 72 hours before transfer to exclude the possibility that a relative detriment in engraftment efficiency was responsible for differences in tumor treatment. There was no decline in the absolute number of gp100-specific CD8⁺ T cells in mice that received programmed cells compared with mice that received unprogrammed T_EM cells (Figure 2B). Indeed, a significantly greater number of pmel CD8⁺ T cells were retrieved from mice that received cells programmed for 72 hours before transfer (P < .01), correlating with a higher yield of cells in vitro after the programming step. By contrast, there was not a significant difference in the absolute number of pmel CD8⁺ T cells between mice that received nonstimulated T_EM cells and cells programmed for 24 hours (P > .2). Collectively, these data suggested that a critical determinant of the programmed antitumor response was lost or exhausted between 24 hours and 72 hours after stimulation.

We next sought to determine what functional changes occurred in programmed pmel T_EM cells 72 hours after restimulation were responsible for their loss of in vivo antitumor activity. IFN-γ production by adoptively transferred CD8⁺ T cells has recently been shown to be an essential mediator in the recognition and subsequent destruction of self-tissues that express gp100, such as the eye.⁴³  Further, spontaneous IFN-γ release has been documented at early time points in naive CD8⁺ T cells executing an Ag-independent programmed response.¹⁷  We therefore evaluated the release of IFN-γ by tumor-reactive T_EM cells as a function of time after the initiation of a programming stimulus.

Peptide-programmed pmel T_EM cells were assessed for their ability to release IFN-γ in response to either B16 or hgp100_25-33–pulsed target cells on days 1 through 6 after restimulation. As a specificity control, we assessed IFN-γ release in response to the gp100⁻ tumor line EL-4 pulsed with βgal_96-103, an irrelevant D^b-restricted peptide. Immediately before in vitro programming, T_EM cells released IFN-γ only in response to cognate peptide and, to a more limited extent, to B16 (data not shown). In contrast, pmel T_EM cells released IFN-γ in an Ag-independent manner after receiving the programming stimulus (Figure 3A). This Ag-independent release of IFN-γ lasted for as long as 5 days after restimulation until ultimately fading to specificity by day 6 after restimulation such that only Ag-bearing targets (either hgp100_25-33 peptide-pulsed targets or B16) were recognized with significant IFN-γ release.

As previously reported by others,^44,45  unmanipulated B16 spontaneously expressed very low levels of the class I restriction element, D^b (Figure 3B-C). However, as has also been documented earlier, D^b could be dynamically up-regulated in response to titrated doses of IFN-γ (Figure 3B).^44,45  We therefore assessed whether D^b on B16 could be modulated through IFN-γ produced by in vitro programmed pmel T_EM cells. To evaluate the effect of soluble mediators, we used a Transwell system to physically isolate programmed T_EM cells from B16. Pmel T_EM cells were left unmanipulated or underwent in vitro programming with peptide. As a control, we used pmel T_EM cells genetically deficient in their ability to release IFN-γ (pmel-Ifng^−/−). After overnight coculture, B16 was isolated and assessed for surface expression of D^b by FACS analysis. D^b expression on B16 was significantly up-regulated after coculture with in vitro programmed WT but not IFN-γ–deficient T_EM cells (Figure 3C). An intermediate amount of D^b was expressed after coculture with nonrestimulated pmel T_EM cells (data not shown). These data indicated that provision of a programming stimulus to pmel T_EM cells incited an interval of Ag-independent IFN-γ release that provided for the dynamic up-regulation of D^b on the surface of B16 melanoma in a contact-independent manner.

Based on the observation that IFN-γ can cause a marked up-regulation of D^b on B16 melanoma, we next sought to determine whether these changes facilitated the recognition and ultimately destruction of tumor by pmel T_EM cells both in vitro and in vivo. First, we assessed the differential ability of nonprogrammed pmel T_EM cells to recognize B16 either with or without pretreatment with IFN-γ. An IFN-γ dose of 675 ng/mL was chosen because it was associated with the plateau portion of the dose-response curve for IFN-γ pretreatment and D^b expression on B16 in previous titration experiments (Figure 3B). In an overnight coculture, the recognition of B16 by pmel T_EM cells was significantly augmented by previous incubation with IFN-γ (Figure 4A; P < .001). Next, to test the hypothesis that increased D^b expression on B16 would lead to enhanced cytolysis by pmel T_EM cells, we pretreated B16 with titrated doses of IFN-γ for 48 hours before initiation of a cytolytic assay. We observed a direct correlation on a semilog plot between D^b expression on B16 tumor cells and the efficiency of cytolysis (Figure 4B; R² = 0.95, P < .001).

Given the requirement for IFN-γ to facilitate tumor recognition by pmel cells in vitro, we next tested whether the augmentation of in vivo antitumor function of in vitro programmed T_EM cells was the result of the release of IFN-γ. We transferred either WT or IFN-γ–deficient pmel T_EM cells programmed in vitro with plate-bound anti-CD3/anti-CD28 24 hours before transfer and compared their relative in vivo antitumor efficacy. As previously shown, adoptive transfer of in vitro–programmed pmel T_EM cells resulted in enhanced tumor destruction compared with unprogrammed pmel T_EM cells (Figure 5). The benefit of in vitro programming was lost by transfer of pmel-Ifng^−/− T_EM cells such that there were no significant differences between programmed pmel-Ifng^−/− cells and nonprogrammed WT pmel T_EM cells (P = .35).

Finally, we sought to establish whether IFN-γ is acting only on tumor cells or whether it may also be activating responding CD8⁺ T cells in an autocrine or paracrine manner. We therefore transferred programmed pmel T_EM cells genetically deficient in the IFN-γ receptor (pmel-Ifngr1^−/−) and compared their antitumor treatment efficacy to programmed WT pmel T_EM cells. Transfer of either IFN-γ receptor–deficient or competent T_EM cells resulted in significantly delayed tumor growth compared with untreated controls (Figure 5; P < .01). There was no significant difference in tumor treatment between the 2 treatment groups (P > .1), demonstrating that the effect of IFN-γ was independent of the ability of adoptively transferred CD8⁺ cells to perceive IFN-γ. Collectively, these data indicated that production but not consumption of IFN-γ was a critical determinant of the in vivo antitumor efficacy of in vitro–programmed T_EM cells.

We found that provision of a brief programming stimulus to CD8⁺ T_EM cells led to clonal expansion and acquisition of effector function in a manner analogous to results seen previously with programmed naive CD8⁺ T cells.^26–30  Similar results were obtained whether an Ag-specific stimulus with cognate peptide or an Ag-nonspecific stimulus with plate-bound anti-CD3/anti-CD28 were used. Thus, Ag-experienced CD8⁺ T cells, like naive cells,^28,29  were capable of executing a highly stereotyped effector response with as little as 12 hours of stimulation. Although work from the Schoenberger laboratory has demonstrated that programming of naive CD8⁺ T cells can significantly potentiate their in vivo effector function after adoptive transfer in a model of T cell–mediated autoimmune diabetes,²⁹  a direct comparison of the relative efficacy of in vitro–programmed CD8⁺ T cells versus in vivo–stimulated T cells had previously not been made. Our findings demonstrated that in vitro programming of Ag-experienced CD8⁺ T cells could replace the requirement for in vivo vaccination with near-equivalent efficacy.

Previous inquires into the duration of a programming stimulus for naive CD8⁺ T cells have found that anywhere between 20 hours to more than 48 hours of stimulation are required for an optimal response^29,31 ; however, requirements for programming Ag-experienced cells have not been previously evaluated. We therefore sought to determine the optimal duration between in vitro programming of T_EM cells and cell transfer using tumor treatment as a read-out. We found that the enhanced antitumor efficacy of in vitro–programmed CD8⁺ T_EM cells was entirely abrogated if cells were transferred more than 72 hours after programming. Although our experiments did not determine the minimal requirements for optimal programming of Ag-experienced cells, we found that as little as 12 hours was sufficient to significantly potentiate the in vivo antitumor response.

The loss of the programming benefit when cells were transferred more than 72 hours after stimulation was probably not the result of an impairment in proliferative potential as we transferred all cells generated in vitro after the programming stimulus. Indeed, there were a significantly greater number of adoptively transferred cells present in the spleens of mice that received T_EM cells programmed for 72 hours before transfer compared with mice receiving cells programmed for 24 hours or nonprogrammed T_EM cells. We found instead that a functional alteration in IFN-γ secretion was a critical determinant of the programmed antitumor response. During early time points after programming, T cells underwent an interval of Ag-independent IFN-γ release capable of inducing dynamic changes in the surface expression of the major histocompatibility complex class I restriction element D^b on B16 melanoma and subsequent tumor recognition and destruction. Similar to results from Rivoltini et al using human melanoma cell lines, we observed a significant correlation between the magnitude of class I expression on tumor cells and the efficiency of cytolysis by antigen-specific T cells.⁴⁶  The functional significance of IFN-γ was further confirmed by experiments using programmed CD8⁺ T cells genetically deficient in IFN-γ in which the antitumor benefit of programming was entirely abrogated.

These findings parallel recent results demonstrating that IFN-γ is a critical determinant in the reactivity of adoptively transferred cells against nontransformed self-tissues, such as the eye, that also express the melanocyte differentiation Ag gp100.⁴³  Further, these data also correlate with a recently completed clinical trial from the National Cancer Institute Surgery Branch where detectable serum levels of IFN-γ could be found immediately after adoptive transfer of peripheral blood mononuclear cells transduced with TCRs against either MART-1 or gp100.⁴⁷  Approximately 30% of the patients in this trial who received T cells transduced with a high-affinity receptor against the MART-1 melanoma Ag⁴⁸  obtained an objective clinical response consistent with standardized oncologic criteria,⁴⁹  suggesting a possible relationship between serum levels of IFN-γ and the potency of T cell–mediated antitumor treatment efficacy.

Based on these and other results in our preclinical animal model⁴³  as well as recent clinical experience using TCR gene–transduced peripheral lymphocytes reactive against the shared melanoma/melanocyte Ags gp100 and MART-1,⁴⁷  we propose a model of “immunologic sonar” in which programmed adoptively transferred cells initially release IFN-γ in an Ag-independent manner as an immunologic “ping” to facilitate target recognition. By nonspecifically releasing IFN-γ, T cells can up-regulate target Ag presentation to facilitate an “immunologic echo” or return signal. When transferred cells permeate tissues expressing cognate antigen, a virtuous cycle of positive feedback may be established in which Ag-specific T cells recognize their cognate Ag, leading to further IFN-γ release, greater target recognition, and ultimately target killing in a second phase characterized by specificity. We are currently working to confirm this model using dynamic in vivo imaging with 2-photon microscopy.

In vitro expansion techniques using beads coated with anti-CD3/anti-CD28 have successfully been used to generate large numbers of autologous and allogenic T cells for the treatment of HIV,⁵⁰  refractory non-Hodgkin lymphoma,⁵¹  relapsed hematologic malignancies after stem cell transplantation,⁵²  and other hematologic malignancies.^53–55  However, accumulating evidence has demonstrated that prolonged in vitro expansion of T cells can cause a progressive loss of replicative capacity and functional exhaustion limiting their in vivo antitumor efficacy.^19,56  To limit the detrimental impact of prolonged in vitro culture, we used a time-constrained exposure to anti-CD3/anti-CD28 to deliver an instructional program of expansion and acquisition of effector functions for T cells to execute in vivo after transfer. This approach not only provided for enhanced in vivo antitumor function but has the practical advantage of limiting the cost and labor of expanding large number of T cells for infusion.

In conclusion, provision of a brief programming stimulus to CD8⁺ T_EM cells in either an Ag-specific (peptide) or nonspecific (anti-CD3/anti-CD28) manner before ACT caused cells to execute a response of in vivo clonal expansion and tumor destruction nearly equivalent to that of vaccine-stimulated T_EM cells. This response was dependent on an interval of Ag-independent IFN-γ release by programmed cells that facilitates the recognition and subsequent cytolysis of B16 tumor cells. Future studies will be focused on the translation of these findings in mice to human TIL- and TCR-transduced peripheral blood mononuclear cell cultures with the goal of improving the efficacy of current adoptive cell transfer protocols.

The authors thank S. A. Rosenberg and C. S. Hinrichs for critical review of the manuscript and T. Fojo for providing the precious resource of time to complete this project.

This work was supported in part by the intramural program of the National Cancer Institute, National Institutes of Health (Bethesda, MD) and the Howard Hughes Medical Institute (C.A.K.).

This study was done in partial fulfillment of a PhD in Biochemistry (to D.C.P.) at The George Washington University, Washington, DC.

National Institutes of Health

Contribution: C.A.K., Z.Y., and N.P.R. designed research; C.A.K., Z.Y., L.N.H., L.G., and D.C.P. performed research and collected data; C.A.K., Z.Y., and N.P.R. analyzed and interpreted data; and C.A.K., L.G., and N.P.R. wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Christopher A. Klebanoff or Nicholas P. Restifo, National Cancer Institute, National Institutes of Health, 9000 Rockville Pike, Bethesda, MD 20892; e-mail: klebanoc@mail.nih.gov or restifo@nih.gov.