Research ArticleCancer Immunotherapy

Establishment of Antitumor Memory in Humans Using in Vitro–Educated CD8+ T Cells

See allHide authors and affiliations

Science Translational Medicine  27 Apr 2011:
Vol. 3, Issue 80, pp. 80ra34
DOI: 10.1126/scitranslmed.3002207


Although advanced-stage melanoma patients have a median survival of less than a year, adoptive T cell therapy can induce durable clinical responses in some patients. Successful adoptive T cell therapy to treat cancer requires engraftment of antitumor T lymphocytes that not only retain specificity and function in vivo but also display an intrinsic capacity to survive. To date, adoptively transferred antitumor CD8+ T lymphocytes (CTLs) have had limited life spans unless the host has been manipulated. To generate CTLs that have an intrinsic capacity to persist in vivo, we developed a human artificial antigen-presenting cell system that can educate antitumor CTLs to acquire both a central memory and an effector memory phenotype as well as the capacity to survive in culture for prolonged periods of time. We examined whether antitumor CTLs generated using this system could function and persist in patients. We showed that MART1-specific CTLs, educated and expanded using our artificial antigen-presenting cell system, could survive for prolonged periods in advanced-stage melanoma patients without previous conditioning or cytokine treatment. Moreover, these CTLs trafficked to the tumor, mediated biological and clinical responses, and established antitumor immunologic memory. Therefore, this approach may broaden the availability of adoptive cell therapy to patients both alone and in combination with other therapeutic modalities.


The diagnosis of melanoma with distant metastases carries a median survival of less than 1 year (1). However, recent clinical trials suggest that adoptive T cell therapy can induce long-lasting clinical responses and may prolong overall survival (2). Successful adoptive T cell immunotherapy necessitates the generation of tumor-specific T lymphocytes that have the capacity to eliminate or control the growth of cancer cells (38). Investigators have developed strategies to isolate and expand large numbers of CD8+ T lymphocytes (CTLs) that exhibit both antitumor specificity and effector function. Although these CTLs have been adoptively transferred to cancer patients without significant toxicity, biological and clinical activity were limited in early studies (912). Considerable evidence suggests that one of the mechanisms limiting their efficacy is the failure of these CTLs to persist in vivo (3, 10, 1315).

To address the failure of CTLs to persist when adoptively transferred, investigators have developed strategies to expand engrafted CTLs in vivo. Administration of interleukin-2 (IL-2) after adoptive T cell transfer significantly increases both T cell survival and biological activity (10, 12, 16, 17). Preinfusion lymphodepletion using myeloablative therapy combined with IL-2 administration further improves persistence of engrafted antitumor T cells and, moreover, has been associated with durable clinical responses (2, 13, 18). Lymphodepletion is thought to increase access to homeostatic cytokines such as IL-7 and IL-15, eliminate suppressive regulatory T cells, and provide T cells space to expand (2, 1821).

We have developed an alternative strategy to overcome the failure of adoptively transferred CTLs to persist that requires the generation of antitumor CTLs with a central memory and an effector memory phenotype and an intrinsic capacity to survive. Previously, we reported the development of a human cell–based artificial antigen-presenting cell (aAPC) genetically engineered to express HLA-A*0201 (A2), CD80, and CD83. These aAPCs expanded large numbers of CTLs restricted to various tumor-associated antigens in vitro from peripheral CD8+ T cells in the presence of IL-2/IL-15 (22, 23). These antigen-specific CTLs demonstrated a central memory and an effector memory phenotype and were remarkably long-lived in vitro, persisting more than a year without allogeneic feeder cells or cloning (23).

Here, we tested whether these unique antitumor CTLs generated with gene-engineered aAPC and IL-2/IL-15 could persist in humans. MART1-specific CTLs were generated in vitro from melanoma patients and then infused back without lymphodepletion or IL-2 administration. We chose the melanoma-associated antigen MART1 as our target because necessary immune assessment technologies to evaluate persistence and localization of infused MART1 T cells are widely available (10, 12). We report that CTLs with a memory phenotype generated using the aAPC-based system could be safely infused and functioned as memory T cells; these cells persisted long term, trafficked to tumors, and induced antitumor biological and clinical responses in humans.


Adoptive transfer of autologous MART1-specific CD8+ T cells generated in vitro using aAPC and IL-2/IL-15 was well tolerated

Nine patients with metastatic melanoma received a total of 17 infusions of autologous MART1-specific CTLs generated from peripheral CD8+ T cells using aAPC and IL-2/IL-15 over a 3-week period. The first infusion (28.0% MART1 multimer positivity, mean) was given on day 0, and the second infusion (30.7% MART1 multimer positivity, mean) was given on day 35 (Fig. 1A). MART1-specific CTLs (≥1.8 × 109) were successfully generated for all patients, and all but one patient received two infusions of cells (Table 1). The second graft was produced from CD8+ T cells harvested by leukapheresis 2 weeks after the first infusion (fig. S1). As previously published, autologous MART1-specific CTLs generated over a 3-week period displayed a central memory and an effector memory phenotype (CD45RA CD45RO+ CD62L+/−) (23) (Fig. 1B). Antigen-specific cytotoxicity and interferon-γ (IFN-γ) secretion was demonstrated against both peptide-pulsed targets and tumor cells, demonstrating their specificity and sufficient avidity (Fig. 1, C and D). Similar data were observed for all patients and grafts.

Fig. 1

Infused MART1-specific CTL grafts had a central memory and an effector memory phenotype and effector function. Seventeen MART1-specific CTL grafts were generated and administered to nine patients. (A) The MART1 multimer+ percentage of infused CTL grafts is shown separately for grafts 1 (n = 9) and 2 (n = 8) for all patients. Bars represent the mean values. (B) Percent expression for the indicated molecules on MART1 multimer+ CTLs is shown for all grafts (n = 17). (C and D) Representative example of functional assays for infused MART1-specific CTL grafts (subject 5). (C) Antigen-specific cytotoxicity was demonstrated for MART1 peptide–pulsed T2 (▪) versus control peptide–pulsed T2 targets (•), and the HLA-A2+ MART1+ melanoma line Malme-3M (□) versus the HLA-A2+ MART1 melanoma line A375 (○). (D) IFN-γ ELISPOT showed antigen-specific IFN-γ secretion using peptide-pulsed T2 cells and tumor cell line targets. Means ± SD of triplicates are shown. SFU, spot-forming unit.

Table 1

Patient characteristics, CTL infusions, and clinical status. WLE, wide local excision; LND, lymph node dissection; HD IL-2, high-dose IL-2; IFN, interferon-α; RT, radiation therapy; PD, progressive disease; SD, stable disease; CR, complete response; PR, partial response; MR, mixed response.

View this table:

We were able to generate targeted numbers of MART1 peptide–specific CTLs from all enrolled patients, and all patients received CTL infusions. Subjects 1 to 4 received 2 × 108 cells/m2 per infusion (dose level 1), whereas subjects 5 to 9 were assigned to 2 × 109 cells/m2 per infusion (dose level 2). Subject 1 alone did not receive a second infusion because of disease progression. No dose-limiting toxicities were observed in any patient. Possibly related grade 1 toxicities were limited to fatigue, pruritis [general or localized to delayed-type hypersensitivity (DTH) sites], and pain at sites of tumor. One patient experienced transient, asymptomatic peripheral blood eosinophilia (1100/μl; normal range, 0 to 400/μl) 3 weeks after the second infusion.

Adoptive transfer of MART1 CTLs with a memory phenotype induced long-term increases in numbers of circulating MART1-specific T cells

Because CTLs expanded in vitro with aAPC and IL-15 could be maintained in vitro for prolonged periods, we examined whether these CTLs persisted in patients upon adoptive transfer without further host manipulation. Multimer staining of peripheral CD8+ T cells revealed that immediately after infusion, the percentage of CD8+ T cells specific for MART1 increased in all patients (Fig. 2, A and B, and fig. S2). For patients receiving 2 × 108 cells/m2 per dose, the frequency of MART1-specific T cells increased by a median of 6.1-fold (range, 3.6 to 10.2), and for patients receiving 2 × 109 cells/m2, the MART1 T cell frequency increased by a median of 9.8-fold (range, 4.0 to 383.5). Previous reports have suggested that adoptive transfer of tumor-specific CD8+ T cells without preinfusion lymphodepletion or IL-2 administration persisted for less than a week (median) and was undetectable at 2 weeks (9, 10, 12). In contrast, we routinely observed increases in circulating MART1 multimer staining T cells 2 weeks after infusion of CTLs (Fig. 2C, P < 0.05). We also observed sustained increases in the frequency of MART1-specific T cells by more than twofold in four patients for 21 days after infusion 2 (Fig. 2D, left). Samples at later time points were available for six patients (subjects 2, 3, 5, 7, 8, and 9), and in three of these patients subjects 2, 7, and 8), we observed increased MART1-specific T cell frequencies on days 102, 258, and 358, respectively (Fig. 2D, right). Patients received no other therapy, including CTLA-4 blockade, during this period.

Fig. 2

Adoptive transfer induced sustained increases in the frequency of circulating MART1-specific CD8+ T cells. The frequency of MART1-specific T cells was determined by MART1 multimer staining of circulating CD8+ T cells before and after infusion without any in vitro expansion. Note that patients received no other therapies such as CTLA-4 blockade during the indicated time periods. (A) Representative MART1 multimer staining for subject 7 is shown. Day 0 and 35 analyses were performed on blood samples drawn 30 min after infusion of CTL grafts. (B) The frequency of MART1-specific CD8+ T cells over the course of the clinical study is shown for subjects 2 and 7. Mean values ± SD of quadruplicates are shown. (C) The frequency of MART1-specific CD8+ T cells for all patients is shown for time points before infusion and 14 days after infusion. The average of three preinfusion time points was used as the baseline. (D) A sustained increase in the frequency of MART1-specific T cells on day 56 was observed and is expressed as the ratio of multimer-positive CD8+ T cells on day 56/preinfusion baseline (left). Data for day 21 were included for subject 1, who did not receive a second infusion. The ratio of multimer-positive CD8+ T cells at later time points/preinfusion baseline is shown (right). Post-infusion samples were obtained for subjects 2, 3, 5, 7, 8, and 9 on days 102, 145, 134, 258, 358, and 133, respectively.

Persistent MART1-specific T cells after adoptive transfer were phenotypically and functionally memory T cells

We characterized the naïve/memory immunophenotype of peripheral MART1-specific T cells before and after CTL infusion. Infused MART1 CTLs harbored a central memory and an effector memory phenotype (CD45RA CD62L+/−) (Fig. 1B). In contrast, as seen for subjects 2, 7, and 8 (Fig. 3A, preinfusion), a large proportion of MART1 precursor T cells in healthy individuals’ and cancer patients’ circulation have a naïve phenotype (CD45RA+ CD62L+) (24). Therefore, we were able to distinguish infused MART1 CTLs from preexisting MART1 T cells by monitoring the difference in the phenotype of MART1-specific T cells. Immediately after infusion, we detected infused MART1 T cells with a memory phenotype in all patients, as shown for three representative examples (Fig. 3A). In subjects 2, 7, and 8, increases in the percentage of MART1-specific T cells with a central memory phenotype (CD45RA CD62L+) were demonstrated on days 102, 258, and 156, respectively (Fig. 3A). The frequency of MART1-specific T cells displaying a central memory phenotype was determined in all patients before and after infusion (Fig. 3B). Sustained increases were observed 5 weeks after transfer in seven of nine patients, suggesting that in vitro–generated CTLs with a memory phenotype engrafted upon adoptive transfer (P < 0.05). For subjects 7 and 8, persisting MART1-specific T cells on days 70 and 66, respectively (Fig. 3A), were also substantially positive for CD27, CD28, and CD127, and only a subpopulation was positive for CCR7 (Fig. 3C). This result suggests that CCR7+ MART1-specific T cells trafficked to lymph nodes and that primarily CCR7 MART1-specific T cells persisted in the circulation. Sufficient numbers of cells for subject 2 were not available for this analysis. Note that, during this period, patients received no other therapeutic or immune-modulating interventions including anti–CTLA-4 monoclonal antibody (mAb) treatment.

Fig. 3

Adoptive transfer increased the number of MART1-specific T cells with a central memory phenotype and memory function. (A) The phenotype of circulating MART1-specific T cells was assessed for all nine patients before and after infusion. In the depicted examples (subjects 2, 7, and 8), the CD45RA/CD62L phenotype of fresh, circulating MART1 multimer+ cells is shown below the multimer stain for each indicated time point. Note that patients received no other antimelanoma therapies such as anti–CTLA-4 mAb treatment during the indicated periods. (B) The percentage of peripheral CD45RA CD62L+ MART1 multimer+ CTLs is shown for all nine patients before and after infusion (P < 0.05). Data for day 35 were included for subject 1, who did not receive a second infusion. (C) The phenotype of MART1-specific CD8+ T cells for subject 7, day 70, and subject 8, day 66, was demonstrated. The phenotype of gated MART1 multimer staining cells is shown (open) with isotype control mAb staining (shaded). (D) Pre- and post-infusion MART1-specific recall responses are shown in IFN-γ ELISPOT assays for subjects 2, 7, and 8. Without any previous expansion, freshly purified, peripheral CD8+ T cells were stimulated with T2 targets pulsed with MART1 or control peptide (left). T cells were also incubated with melanoma cell lines A375 (HLA-A2+ MART1) and Malme-3M (HLA-A2+ MART1+) (right). Data represent means ± SD of triplicates.

To demonstrate that persisting peripheral MART1-specific CD8+ T cells with a memory phenotype were functionally memory T cells, we performed an IFN-γ enzyme-linked immunospot (ELISPOT) assay using peripheral blood freshly drawn from three patients shown in Fig. 3A. Purified CD8+ T cells were stimulated with target cells without any in vitro sensitization to strictly assess memory responses. In subjects 2, 7, and 8, we detected MART1 peptide–specific IFN-γ secretion on days 56, 70, and 66, respectively, but not before CTL infusion (Fig. 3D). IFN-γ secretion was also induced by HLA-A2+ MART1+ tumor cells but not by HLA-A2+ MART1 tumor cells. These results suggest that persisting peripheral MART1-specific CD8+ T cells with a memory phenotype were functionally memory T cells that had sufficient avidity to recognize tumor cells endogenously presenting MART1 peptide via HLA-A2 molecules.

We also conducted DTH testing by administering intradermal injections of MART1 peptide 3 weeks before and after infusion 1. We observed enhanced DTH reactions to peptide injections performed after infusion in subjects 1, 2, 5, and 7. In subject 7, histologic examination showed recruitment of mononuclear cells admixed with eosinophils accumulating around blood vessels. Immunohistochemistry analysis demonstrated recruitment of CD3+ and CD8+ lymphocytes, consistent with a memory recall response mediated by infused MART1-specific CD8+ T cells (fig. S3). Together, these in vivo data suggest that persisting MART1-specific T cells in the circulation were not only phenotypically but also functionally memory T cells.

Infused MART1 CTLs could traffic to tumor sites after adoptive transfer

To induce clinical responses, transferred CTLs must traffic to tumor sites and mediate effector responses within the tumor microenvironment. We therefore assessed whether infused CTLs trafficked to tumor by analyzing post-infusion tumor biopsies. Subject 5, who experienced an objective clinical response (fig. S4), underwent a tumor biopsy on day 5 of cycle 1. A preinfusion tumor biopsy showed minimal lymphocytic infiltration at four lymphocytes/mm2 (Fig. 4A, left). In contrast, after infusion, lymphocyte infiltration of tumor was brisk in both the central and the peripheral regions at 225 lymphocytes/mm2 (Fig. 4A, upper middle). Near-complete tumor destruction occurred with tumor/lymphocyte satellitosis, hemorrhagic cell necrosis, and fibrosis, consistent with a strong, overwhelming antitumor immune response. Almost all infiltrating T cells were strongly positive for CD8 (135 cells/mm2), with fewer stained by CD4 (56 cells/mm2). Foxp3 staining was negative, suggesting a lack of regulatory T cells (Fig. 4A, upper right and lower panels). Direct multimer staining of tumor-infiltrating lymphocytes (TILs) without any in vitro expansion identified the presence of MART1-specific T cells (Fig. 4B, left).

Fig. 4

Transferred CTLs trafficked to sites of antigen expression and mediated antitumor responses. Pathologic analysis was performed to determine whether transferred T cells were able to traffic to sites of antigen expression. (A) Infiltration of tumor by lymphocytes was morphologically assessed before and after infusion for subject 5. Immunohistochemical staining of the post-infusion biopsy for CD8, CD4, and Foxp3 is shown. Scale bars, 50 μm (upper right, applies to upper panels) and 100 μm (lower right, applies to lower panels). (B) Without in vitro expansion, MART1-specific T cells were identified in fresh TILs by MART1 multimer staining. Three MART1-specific clonotypes derived from TILs were identified. The presence of all three MART1-specific T cell clonotypes in both CTL grafts was shown by RT-PCR.

To evaluate whether CTLs mediating the antitumor response were present in the CTL grafts, we performed clonotypic analysis of tumor-infiltrating MART1-specific T cells using their CDR3 sequences as a molecular marker. Three MART1-specific CTL clonotypes—05Vb2A, 05Vb2B, and 05Vb14A—were identified from MART1-specific T cells expanded from the subject 5 tumor biopsy (Fig. 4B). CDR3-specific reverse transcription–polymerase chain reaction (RT-PCR) analysis showed that all three clones existed in both CTL grafts. Similar analysis was performed for subject 3, who experienced a mixed clinical response (fig. S5A). Five MART1-specific CTL clonotypes were identified from resected tumor samples, and four of the five clones were detected in both CTL grafts (fig. S5B). Because infused CTLs were not gene-marked, we cannot exclude the possibility that MART1-specific T cells at the tumor sites were derived endogenously and not from infused grafts. However, combined with the data that infused MART1 CTLs did persist and expand in patients upon transfer (Figs. 2 and 3), these data suggest that antitumor CTLs expanded in T cell grafts could traffic to sites of tumor and mediate antitumor functions after adoptive transfer.

Clinical responses occurred after CTL infusions and subsequent therapies

Patients received CTL infusions without any additional therapy including lymphodepletion, cytokine administration, or vaccination. Responses to CTL infusions were assessed by physical examination and radiographic measurements during the 10th week of the study (Table 1, status on day 70). On day 70, subject 5 had achieved a partial response by computed tomography (CT) scan, with a reduction in the size of tumor-involved lymph nodes by 33%. Follow-up on day 140 showed no evidence of fluorodeoxyglucose (FDG)–avid disease by positron emission tomography (PET)/CT with normalization in the size of lymph nodes previously involved by tumor (fig. S4). This response has been confirmed by multiple subsequent high-resolution CT and PET/CT scans, and, without any additional antitumor therapy, the patient has continued to remain without evidence of disease for more than 25 months. For the remaining patients on day 70, one had died without receiving a second CTL infusion, three had progressive disease, and four were stable by Response Evaluation Criteria in Solid Tumors (RECIST) criteria. Subject 3, who had stable disease overall, had a mixed response with reduction in the size of one metastatic pulmonary lesion (fig. S5A). Note that subjects 7 and 8 experienced clinical stabilization and did not require any other anticancer therapy for 11 and 12 months, respectively. Both experienced delayed mixed responses by PET/CT during this time period, a phenomenon that has previously been reported with other immune-based therapies (25). These clinical responses suggest that infused CTLs with a memory phenotype could traffic to and attack tumor.

Seven patients received additional therapy upon disease progression (Table 1). Five were treated with the anti–CTLA-4 mAb ipilimumab. Three patients (subjects 2, 3, and 9) achieved partial responses that were durable, lasting up to 1 to 2 years, and two (subjects 7 and 8) had stable disease, lasting 5 to 6 months. Notably, subject 3 achieved near-complete resolution of pulmonary and adrenal metastases and is considered to have a partial response only because of subcentimeter radiographic abnormalities detected on CT. Yet, even though anti–CTLA-4 therapy was halted because of autoimmune toxicity, she has remained free of disease progression for more than 31 months.

Anti–CTLA-4 mAb therapy may promote the expansion of infused CTLs as memory T cells

As described above, subject 2, who had progressive disease after CTL infusion, was treated with multiple cycles of CTLA-4 blockade and achieved a partial response. In this patient, before CTL infusion, all MART1 multimer+ CD8+ T cells had a naïve phenotype (Fig. 5A, left). However, MART1 multimer+ T cells with a memory phenotype persisted and demonstrated memory function after CTL transfer (Figs. 3D and 5A). After CTLA-4 blockade, there was a marked further expansion of peripheral MART1 multimer+ T cells, which continued to display a central memory and an effector memory phenotype (Fig. 5A, middle). On day 537, MART1 T cells increased to 7.8% of circulating CD8+ T cells and were oligoclonal, consisting of Vβ 14+ and Vβ 14 clones (Fig. 5A, right). Furthermore, these circulating day 537 MART1-specific T cells were able to recognize MART1+ tumor cells and demonstrated potent antigen-specific recall responses without in vitro sensitization (Fig. 5B). A similar effect of CTLA-4 blockade was observed for subject 8 (fig. S6).

Fig. 5

Adoptively transferred CTLs expanded in a patient treated with anti–CTLA-4 (ipilimumab) therapy. After two infusions of MART1 CTLs, subject 2 began anti–CTLA-4 mAb therapy on post-infusion day 103. (A) The frequency of peripheral MART1-specific T cells is shown before and after infusion of CTLs and after initiation of anti–CTLA-4 mAb therapy. The CD45RA/CD62L phenotype of multimer-positive cells is shown below the multimer stain for each time point. TCR Vβ 14 staining of markedly expanded MART1 multimer+ T cells on day 537 is shown (right). (B) Without any previous in vitro expansion, MART1-specific recall responses on day 537, but not before infusion, were demonstrated with the IFN-γ ELISPOT assay using PBMCs incubated with control or MART1 peptide (left). Fresh day 537 PBMCs were also incubated with the melanoma cell lines A375 (HLA-A2+ MART1) and Malme-3M (HLA-A2+ MART1+) (right). Data represent means ± SD of triplicates. (C) MART1-specific CTLs were expanded from the memory fraction (CD45RA) of CD8+ T cells on day 56, before CTLA-4 blockade, to identify 10 MART1-specific clonotypes. The presence of 7 of these 10 clones in CTL grafts 1 or 2, including Vβ 14 clones 02Vb14B and 02Vb14C, is shown. (D) Identification of the Vβ 14 clone 02Vb14B exclusively in the CD45RA memory subpopulation only is shown on day 74, before CTLA-4 blockade (left), and on day 537, after CTLA-4 blockade (right).

To molecularly track persisting MART1-specific T cells in subject 2, we isolated 10 circulating individual MART1-specific T cell clonotypes from the CD45RA CD8+ T cell memory fraction on day 56 (Fig. 5C, left). CDR3-specific RT-PCR revealed that 7 of 10 of these clonotypes were present in the infused CTL grafts, including a Vβ 14 clone, 02Vb14B (Fig. 5C, right). On day 74, after CTL transfer and before CTLA-4 blockade, we detected the Vβ 14 clone 02Vb14B in the CD45RA memory fraction, but not in the CD45RA+ CD62L+ naïve or the CD45RA+ CD62L terminally differentiated fractions (Fig. 5D, left). Intriguingly, on day 537 after CTLA-4 blockade, the Vβ 14 clone 02Vb14B was again detected exclusively in the CD45RA memory fraction (Fig. 5D, right). These data suggest that 02Vb14B CTLs were included in Vβ 14+ MART1 tetramer+ cells, with a memory phenotype shown in Fig. 5A (right). These results provide strong molecular evidence that MART1 CTL clonotypes with sufficient avidity to attack tumors could be selected in vivo from adoptively transferred polyclonal MART1 CTLs and then persist and expand as memory T cells. Furthermore, CTLA-4 blockade may augment antitumor immunological memory and responses established by adoptive transfer of MART1 CTLs.


In this “proof-of-concept” clinical trial, we tested the hypothesis that antitumor CTLs educated in vitro using our gene-engineered aAPC would establish antitumor immunological memory in patients when adoptively transferred without lymphodepletion or cytokine therapy. We successfully showed that infused CTLs with a memory phenotype were able to persist as memory T cells for long periods of time, thereby establishing antitumor immunological memory. Furthermore, persisting CTLs had antitumor function, and molecular evidence indicated that CTLs trafficked to and mediated biological and clinical responses. Adoptive transfer of CTLs generated using the aAPC-based system enables the establishment of antitumor memory that alone, or in combination with other immune modulators, can promote antitumor immunity.

The aAPC-based system was specifically designed to be feasible and transferable and to educate CTLs so that they could be long-lived in vitro and persistent in vivo. First, we ectopically expressed CD83 on aAPC in conjunction with HLA-A2 and CD80. We and others previously demonstrated that CD83 is critical for the longevity of lymphocytes both in vitro and in vivo (22, 26). Second, we added IL-15 to cultures for T cell expansion. Unlike IL-2, IL-15 has been reported to favor the expansion of antigen-specific T cells with a central memory phenotype (27). Furthermore, it has been shown that in vitro exposure to IL-15 enables T cells to be long-lived in vivo (28). These unique attributes of the aAPC-based system may have enabled the generation of antitumor CTLs that were capable of prolonged persistence and survival as memory T cells after adoptive transfer.

The percentage of peripheral MART1-specific T cells with a CD45RA CD62L+ central memory phenotype increased after transfer, even though only a minority of infused CTLs expressed CD62L in some grafts. This result supports that CD45RA CD62L+ CTLs generated in vitro can in fact behave as central memory T cells in vivo and thus preferentially persist, whereas MART1 CTLs with CD45RA CD62L effector memory and CD45RA+ CD62L terminally differentiated effector phenotypes may not engraft or expand. Persistent MART1 T cells with a memory phenotype also demonstrated recall responses without any in vitro expansion, suggesting that they were functionally memory T cells. These findings are in accordance with animal studies showing that CTLs with a central memory phenotype are more likely to retain memory T cell properties that enable superior persistence and expansion in vivo (27, 29). Our findings further support that CTLs with a central memory phenotype can mediate memory responses and can persist after adoptive transfer without lymphodepletion or IL-2 administration.

In this early-phase study, the degree of persistence of MART1 T cells in the peripheral blood was not strictly correlated with clinical responses. Indeed, other factors may contribute to the success of adoptive T cell transfer, including the removal of negative regulatory cells. However, we did observe the induction of inflammatory infiltrates within tumors and detected the presence of identical clonotypic CTLs in both infused grafts and tumor T cell infiltrates. Combined with data that persistent CTLs have antitumor function, this strongly indicates that transferred CTLs trafficked to tumor and modulated the host antitumor immune response. No patient experienced any untoward side effects, and we did observe beneficial biological and clinical responses. Subject 5 had a partial clinical response on day 70 and continues to have complete resolution of disease for more than 25 months. Three other patients experienced mixed responses or long-lasting disease stabilization. One patient showed a mixed response with marked reduction of a pulmonary lesion. Two other patients, who both showed increases in circulating MART1 T cells (Fig. 2D), had mixed clinical responses in the months after CTL transfer and did not require further treatment for more than 300 days. With the exception of subject 5, who has not needed any further therapy, CTL infusion alone, however, was not sufficient to prevent eventual tumor progression.

As shown in Table 1, upon disease progression, five patients were treated with ipilimumab, which blocks CTLA-4 and potentiates antitumor T cell responses (1, 30, 31). Among five patients, three achieved partial responses (subjects 2, 3, and 9) and two other patients had stable disease (subjects 7 and 8). Although we cannot draw definite conclusions because of the small number of patients and the potential for selection bias, we found these responses noteworthy because the reported overall response rate for ipilimumab is less than 16% (1, 32). Moreover, cellular and molecular evidence indicates that CTLA-4 blockade can induce a marked expansion in vivo of adoptively transferred antitumor CTLs with memory phenotype and function. Either adoptive cell therapy or CTLA-4 blockade alone may be ineffective for most advanced melanoma patients. However, our data suggest that by first establishing antitumor memory responses with adoptive transfer, anti–CTLA-4–induced immune activation can result in enhanced biological responses.

By intensive analysis of infused MART1 CTLs at both the cellular and the molecular levels, we demonstrated that antitumor CTLs with a memory phenotype educated in vitro using gene-engineered aAPC and IL-2/IL-15 can persist in vivo as memory T cells. Although antitumor biological effects were ultimately insufficient to prevent progression for most patients, establishment of antitumor memory by engraftment of infused CTLs may shift the balance between tolerance and antitumor immunity. Thus primed, antitumor immune responses could be enhanced by CTLA-4 blockade, resulting in the observed tumor regressions seen in patients with previously progressive disease. Further improvement in clinical activity may be achieved by combining CTL transfer with approaches such as lymphodepletion, which can eliminate immune-suppressive elements such as regulatory T cells. Adoptive transfer of CTLs generated with aAPC therefore provides a platform for establishing antitumor memory, which can then be combined with immune modulators such as checkpoint inhibitors, lymphodepletion, cytokines, or vaccination to improve tumor regression and patient prognosis.

Materials and Methods


This Dana-Farber/Harvard Cancer Center phase I clinical protocol (NCT00512889) received approval from the local institutional review board and biosafety committees, the National Institutes of Health Recombinant DNA Advisory Committee, and the U.S. Food and Drug Administration. All patients had stage IV metastatic melanoma with baseline biopsies positive for MART1 staining by immunohistochemistry (Covance). MART1 expression was >90% for subjects 3, 7, 8, and 9; 50 to 90% for subjects 1, 2, and 5; and 10 to 50% for subjects 4 and 6. All patients were HLA-A*0201+ by high-resolution human leukocyte antigen (HLA) DNA typing (American Red Cross). Patients received no other anticancer therapy, including lymphodepletion, IL-2 administration, or antibody treatment, within 4 weeks of the leukapheresis and during study participation. Primary study endpoints were to define the feasibility and toxicity of administering MART1-specific CTLs at two dose levels (2 × 108 and 2 × 109 cells/m2). Secondary endpoints included analyzing the phenotype and function of in vitro–expanded CTLs, evaluating the in vivo frequency and function of MART1-specific CTLs after adoptive transfer, and evaluating whether transferred CTLs can traffic to sites of tumor.

Adoptive transfer of MART1-specific CTLs generated in vitro using aAPC and IL-2/IL-15 to patients with advanced melanoma

Patient peripheral blood mononuclear cells (PBMCs) were leukapheresed and CD8+ T cells were selected with CliniMACS (Miltenyi). The generation of clinical-grade aAPC has previously been reported (23). Purified CD8+ T cells were stimulated with clinical-grade aAPC pulsed with MART1 peptide (ELAGIGILTV) obtained from Clinalfa (23). Between stimulations, cultures were supplemented with IL-2 at 10 to 50 IU/ml (Novartis) and IL-15 at 10 to 50 ng/ml (PeproTech). Stimulation and expansion of CTLs were performed in gas-permeable fluorinated ethylene propylene (FEP) culture bags to allow consistent and reproducible production of CTL grafts (American Fluoroseal Corp.). After three weekly stimulations, CTLs were harvested and infused to patients on day 0 at either dose level 1 (2 × 108 cells/m2) or dose level 2 (2 × 109 cells/m2) without any other therapies such as lymphodepletion, IL-2 administration, or vaccination. Two weeks after the first CTL infusion, patients underwent a second leukapheresis to generate a second CTL graft. MART1-specific CTLs for the second CTL graft were similarly generated and were infused on day 35. The date of the first infusion is defined as day 0 throughout the article.

Phenotype analysis of MART1-specific T cells and CTL grafts

To determine the phenotype of in vitro–expanded MART1-specific T cells, we stained CTLs from grafts with HLA-A2/peptide multimers (ProImmune) as previously described (23). Multimer-positive T cells were costained with the following anti-human mAbs: CD8, CD28, CD45RO, and CD62L (Beckman Coulter); CD27 and CD45RA (Invitrogen); CD127 (BD Biosciences); and CCR7 (R&D Systems).

Functional assays

IFN-γ ELISPOT and standard chromium release assays were performed as described elsewhere (22, 23). In these in vitro assays, the native MART1 peptide (27AAGIGILTV35) and the control HIV pol peptide (476ILKEPVHGV484) were used (New England Peptide). T2 cells, A375 (HLA-A2+ MART1), and Malme-3M (HLA-A2+ MART1+) were obtained from the American Type Culture Collection. Assays were performed with a portion of infused CTL grafts and with CD8+ T cells or PBMCs freshly drawn from patients without any in vitro stimulation.

Phenotypic analysis of peripheral and tumor-infiltrating MART1-specific CD8+ T cells before and after CTL infusion

Without any in vitro expansion, peripheral CD8+ T cells were isolated from freshly drawn PBMCs and stained with HLA-A2/MART1 peptide multimer (ProImmune) as previously described (23). Positive staining was confirmed by concurrently staining samples with control HLA-A2/HIV pol peptide multimer. Where indicated, multimer-positive cells were costained with anti-CD45RA, CD62L, CD27, CD28, CD127, CCR7, and isotype control mAbs. T cell receptor (TCR) Vβ subtype analysis was performed on MART1 multimer+ T cells with the Beta Mark kit and anti–Vβ 14 mAb (Beckman Coulter) without any in vitro expansion.

Molecular analysis of clonotypic MART1-specific T cells

MART1 multimer+ T cells were highly purified with flow cytometry–guided sorting from short-term expanded TILs or CD45RA CD8+ T cells. On the basis of the TCR Vβ subtypes of MART1 multimer+ cells determined by the Beta Mark kit, CDR3 regions of positive Vβ subtypes were amplified by RT-PCR with SuperScript reverse transcriptase (Invitrogen) and Phusion high-fidelity DNA polymerase according to the manufacturer’s instruction. Sequences of Vβ primers are published elsewhere (33). Amplified fragments were cloned to plasmid vectors, and their sequence was determined.

For the detection of circulating MART1-specific clonotypic T cells, without any in vitro culture, fresh CD45RA+ CD62L+ naïve, CD45RA memory, and CD45RA+ CD62L terminally differentiated effector CD8+ T cells were isolated by flow cytometry–guided sorting at the time points indicated. CDR3-specific RT-PCR was performed with SuperScript One-Step RT-PCR (Invitrogen) according to the manufacturer’s protocol to determine the existence of clonotypic MART1 T cells in each subpopulation. To detect MART1-specific clonotypic T cells in each CTL graft, we purified MART1 multimer+ CTLs and performed CDR3-specific RT-PCR similarly. A housekeeping gene, UbcH5B, was used as an internal control and its primer sequences are as follows: 5′-TCTTGACAATTCATTTCCCAACAG-3′ (sense) and 5′-TCAGGCACTAAAGGATCATCTGG-3′ (antisense).

Pathologic evaluation of tumor biopsies

When tumor was accessible, biopsies were performed during the first week of cycle 1. Tissues were fixed in 10% neutral-buffered formalin and embedded in paraffin. Sections (5 μm) were stained with hematoxylin-eosin, and immunohistochemical staining was performed on paraffin sections with antibodies against CD3, CD4, and CD8 (Ventana Medical Systems) and Foxp3 (Abcam). Tumor lymphocyte reactivity was quantitatively assessed according to criteria previously described (34).

Pathologic evaluation of DTH reactions

DTH tests were performed by intradermal injections of 50 μg of MART1 peptide (ELAGIGILTV) on day −21 immediately after leukapheresis and on cycle 1 day 21. Injection sites were inspected 2 days later, and biopsies were performed. Tissues were fixed in 10% neutral-buffered formalin and embedded in paraffin. Sections (5 μm) were stained with hematoxylin-eosin for pathologic analysis. Immunohistochemical staining was performed on paraffin sections with antibodies against CD3 and CD8 (Ventana Medical Systems). DTH reactions were assessed according to criteria previously described (35).

Statistical analysis

Welch’s t test was used for two-sample comparisons, and the Wilcoxon signed-rank test was used for paired comparisons. All P values are two-sided and considered significant at the 0.05 level.

Supplementary Material

Fig. S1. Clinical protocol time line.

Fig. S2. Verification of HLA-A2/MART1 peptide multimer staining.

Fig. S3. Adoptive transfer increased delayed-type hypersensitivity reaction to injected MART1 peptide.

Fig. S4. Clinical response to adoptive transfer of MART1-specific T cells in subject 5.

Fig. S5. Trafficking of transferred CTLs to pulmonary lesions, subject 3.

Fig. S6. Expansion of adoptively transferred MART1 CTLs in subject 8 treated with anti–CTLA-4 (ipilimumab) therapy.


  • Citation: M. O. Butler, P. Friedlander, M. I. Milstein, M. M. Mooney, G. Metzler, A. P. Murray, M. Tanaka, A. Berezovskaya, O. Imataki, L. Drury, L. Brennan, M. Flavin, D. Neuberg, K. Stevenson, D. Lawrence, F. S. Hodi, E. F. Velazquez, M. T. Jaklitsch, S. E. Russell, M. Mihm, L. M. Nadler, N. Hirano, Establishment of Antitumor Memory in Humans Using in Vitro–Educated CD8+ T Cells. Sci. Transl. Med. 3, 80ra34 (2011).

References and Notes

  1. Acknowledgments: We thank the Dana-Farber Cancer Institute Connell-O’Reilly Cell Manipulation Core Facility for the generation of CTL grafts. We thank J. Daley and S. Lazo-Kallanian at the flow cytometry core for expert technical assistance, and S. Conley for photography. Funding: Funding was provided by the Immunotherapy Fund 1 (L.M.N.), the S. Craig Lindner Fund for Cancer Research (L.M.N.), the Rudolf E. Rupert Foundation for Cancer Research (L.M.N.), the Cancer Research Institute/Ludwig Institute for Cancer Research Cancer Vaccine Collaborative (M.O.B., L.M.N., and N.H.), Friends of the Dana-Farber Cancer Institute (M.O.B. and N.H.), and Dunkin’ Donuts Rising Stars Program (M.O.B. and N.H.). N.H. was funded by NIH grants K22CA129240 and R01CA148673 and the American Society of Hematology Scholar Award. Preclinical scale-up was supported by the Center for Human Cell Therapy Boston (NIH 5U24HL074355). Subjects received ipilimumab on clinical trials supported by Bristol-Myers Squibb (F.S.H.). Author contributions: M.O.B., N.H., and L.M.N. conceived the study and wrote the paper. M.O.B. performed the clinical trial as the principal investigator. P.F., L.D., L.B., M.F., D.L., F.S.H., M.T.J., and S.E.R. assisted with conducting the clinical trial. D.N. and K.S. performed statistical analysis. M.O.B. and N.H. designed and conducted laboratory experiments. M.I.M., M.M.M., G.M., M.T., A.P.M., A.B., and O.I. helped to perform experiments. E.F.V. and M.M. analyzed histopathological specimens. All authors approved the manuscript. Competing interests: The Dana-Farber has filed a patent related to aAPC, application number 10/850,294, entitled “Modified Antigen-Presenting Cells,” on which M.O.B., L.M.N., and N.H. are named as inventors. F.S.H. is a nonpaid consultant to Bristol-Myers Squibb. The other authors declare that they have no competing interests.

Stay Connected to Science Translational Medicine

Navigate This Article