Memory T Cells Originate from Adoptively Transferred Effectors and Reconstituting Host Cells after Sequential Lymphodepletion and Adoptive Immunotherapy¹

Tcell memory is a crucial feature of the immune response to exogenous pathogens. The increased precursor frequency, heightened sensitivity to antigenic stimulation, accelerated kinetics of proliferation, and enhanced trafficking to tissues augment host protection upon exposure to previously experienced pathogens (1, 2). The selective pressure imposed by infectious pathogens has molded the features of the memory response, and its potency underpins the success of vaccination against viral diseases. Unfortunately, it has been difficult to effectively use the immune memory response for cancer therapy. Prophylactic vaccination for cancer is impractical because the shared tumor Ags defined to date are constituents of normal tissue (3, 4). Antigenic targets in commonly subverted signaling pathways, such as receptor tyrosine kinases and their downstream targets, are also shared with normal proliferating cells. Thus, it would be hard to maintain effective antitumor prophylaxis without the potential risk of adverse effects. An added difficulty arises with strategies for vaccination of patients with established tumors, because presentation of tumor Ags typically induces tolerance among T cells with the potential to recognize tumor Ags (5, 6). In addition, tumors lack the pathogen-associated molecular patterns that contribute to a robust immune response through activation of Toll-like receptors (7). As a consequence, the extant T cells detected with tumor Ag/MHC tetramer probes have functional defects (8, 9). Strategies to reverse tolerance through active immunotherapy have been partially successful, in that they are able to boost the frequency of tumor-reactive T cells. However, this has not translated into effective therapy of established metastatic disease, which remains the unsolved, clinically relevant problem.

We have used adoptive transfer of T cells obtained from mice with progressively growing tumors to define the requirements for optimal ex vivo activation of Ag-primed pre-effector T cells and to elucidate the mechanism of tumor regression. The Ag-primed T cells reside within the L-selectin^low subset in tumor-draining lymph nodes (LNs)³and are at a peak level between 9 and 12 days after tumor inoculation (10). The preferential trafficking of naive T cells through LNs and the short time period between tumor inoculation and T cell isolation are consistent with L-selectin down-regulation as a marker of recent Ag exposure (11, 12). Tumor sensitization alone, however, is insufficient to retard tumor growth in the primary host, and adoptive transfer of primed LN T cells does not mediate regression of even minimal tumor burdens in secondary recipients (13). This state of functional tolerance is reversible through ex vivo stimulation with anti-CD3 mAb and IL-2, which induces rapid proliferation, acquisition of Ag-specific IFN-γ production, and in vivo therapeutic efficacy against established tumor (14). Mice cured of established tumors by adoptively transferred effector T cells exhibit long term memory against tumor rechallenge, but the relative contributions of transferred cells, host cells, or Ab responses have not yet been defined.

Our previous studies demonstrated that although the tumors express only MHC class I molecules, adoptive transfer of either CD4⁺ or CD8⁺ T cells alone is sufficient to mediate tumor regression (15, 16). Indeed, cross-presentation of specific tumor Ags to effector T cells by bone marrow-derived APC within the tumor plays a key role in regression (17, 18). This raised the question of whether the process of immune-mediated tumor regression leads to the generation of a second wave of memory cells from host T cells. In this study we demonstrate that transferred effector T cells as well as host T cells regenerating in the context of immune tumor regression each contribute to the memory response.

Female C57BL6N (B6) mice were purchased from the Biologic Testing Branch, National Cancer Institute (Frederick MD), and female B6.PL.Thy1.1 mice were purchased from The Jackson Laboratory (Bar Harbor ME). They were maintained in a specific pathogen-free environment according to National Institutes of Health guidelines using an Institutional Animal Care and Use Committee approved protocol. The MCA-205 and MCA-207 fibrosarcomas, syngeneic to B6 mice, were serially passaged in vivo s.c. as described previously (19). Single-cell suspensions were prepared from minced s.c. tumor by enzymatic digestion for 3 h at room temperature with 0.1% collagenase type IV, 0.01% DNase I, and 2.4 U/ml hyaluronidase type V (Sigma-Aldrich, St. Louis, MO) in HBSS.

Tumors were established in Thy1.1 mice by s.c. flank inoculation of 1.5 × 10⁶ MCA-205 cells. Twelve days after inoculation, the tumor-draining inguinal LNs and spleens were removed and mechanically disrupted to obtain a single-cell suspension. Splenocytes were subjected to ammonium chloride (ACK) lysis of erythrocytes. The lymphoid cells were incubated with 100 μl of anti-CD62L microbeads/10⁸ cells and applied to MACS columns (Miltenyi Biotec, Auburn, CA), and the flow-through fraction was collected as previously described (20). CD62L^low cells, containing ∼50% TCR-positive and 50% B220-positive subsets, were suspended in complete medium (CM) and incubated for 2 days at 4 × 10⁶/well in 24-well culture plates coated with anti-CD3 (145-2C11). Activated cells were washed, counted, and suspended at 0.5 × 10⁵/ml in CM with recombinant human IL-2 (4 U/ml) and recombinant mouse IL-7 (10 ng/ml; R&D Systems, Minneapolis, MN) and subsequently adjusted to 10⁵/ml on day 5 of activation.

Spleens were removed from recipients of activated Thy1.1 cells at 94 days after transfer for mice with persistent tumor, 144 days for cured mice, or 208 days for nontumor-bearing mice. ACK lysis and separation of CD62L^low cells are described above. The CD62L^low cells were then incubated with Thy1.2 microbeads and applied to MACS columns (Miltenyi Biotec) to segregate the Thy1.2⁺ cells. The flow-through fraction was incubated with PE-conjugated anti-CD90.1 (Thy1.1; BD Biosciences, Mountain View, CA) and subsequently with anti-PE microbeads and separated on a MACS column (Miltenyi Biotec). The Thy1.2 or Thy1.1 fractions were culture-activated with anti-CD3/IL-2/IL-7 as described above, but the culture period was extended to 15 or 16 days. Thy1.1 or Thy1.2 purification was repeated at the end of the culture period before adoptive transfer. For analysis of the therapeutic efficacy of fresh memory T cells, spleens from mice cured of s.c. tumors were harvested, treated with ACK, and immediately transferred to recipients with 3-day MCA-205 pulmonary metastases.

T cells were stimulated with a single-cell suspension of either MCA-205 or MCA-207 tumors at a 1:1 ratio or with immobilized anti-CD3. Brefeldin A was added at 5 h, and the cells were harvested at 14 h and stained for intracellular IFN-γ according to the manufacturer’s instructions (BD Biosciences). Abs for phenotype analysis were obtained from BD Biosciences and were used according to the manufacturer’s instructions. Cells were analyzed on a FACS using CellQuest software (BD Biosciences).

Mice were inoculated with MCA-205 or MCA-207 tumor cells (3 × 10⁵) i.v. to establish pulmonary metastases. Subcutaneous tumors were established by inoculation of 1.5 × 10⁶ cells. Mice bearing 10-day s.c. tumors were treated with 5 Gy of total body irradiation (TBI) delivered from a ¹³⁷Cs irradiator (J. L. Shephard, San Fernando CA) before i.v. transfer of T cells, whereas mice with pulmonary tumors were not irradiated. For pulmonary tumors, mice were sacrificed on day 20, the lungs were insufflated with India ink, and the number of surface tumor nodules was enumerated using a dissecting microscope. Subcutaneous tumors were measured in two perpendicular dimensions three times per week, and mice with progressive tumors were sacrificed when the product of dimensions exceeded 200 mm².

Treatment groups consisted of five individuals. Analysis of bidimensional s.c. tumor size was performed by the Mann-Whitney rank sum test. For pulmonary tumors, a t test was performed on paired samples, and p < 0.05 was considered significant.

A constant feature of successful T cell adoptive immunotherapy is the development of long term memory, manifested as protection against tumor challenge. However, protection does not clearly define the requirements for cellular or humoral components, nor does it indicate whether the donor or host cells are critical. We harvested spleens from 10 naive mice and 10 cured mice, 70 days after successful T cell adoptive immunotherapy of 10-day s.c. MCA-205 tumors. The cured mice had no visible evidence of tumor at the original inoculation site or in other organs. The splenocytes were treated with ACK to eliminate erythrocytes and washed, and 7 × 10⁷ cells were immediately transferred to hosts bearing 3-day MCA-205 pulmonary metastases. Whereas control mice and mice treated with naive splenocytes had >250 pulmonary metastases, the recipients of splenocytes from cured donors had no metastases (p < 0.01). This indicates that a cellular component of spleens from cured mice could function without the requirement for ex vivo culture activation. This characteristic is similar to that of memory splenocytes derived from hyperimmune mice (21).

To distinguish between transferred or host T cells as the source of long term memory cells, Thy1.2⁺ mice bearing s.c. MCA-205 tumors were treated with adoptive transfer of congenic Thy1.1⁺ effector T cells. Although MCA-205 is a weakly immunogenic tumor that grows progressively, it nonetheless sensitizes T cells in draining LNs and spleen. We have previously demonstrated that tumor-sensitized T cells are highly enriched in the L-selectin^low subset and that such cells acquire potent effector function after ex vivo activation with anti-CD3 and IL-2 (10). Thy1.1 mice were inoculated in the hind flank with MCA-205 tumor cells, and 12 days later hyperplastic inguinal lymph nodes were removed, and L-selectin^low cells were activated with anti-CD3 for 2 days, followed by culture in medium containing IL-2 (4 U/ml) and IL-7 (10 ng/ml). On day 9 there was 95-fold numerical expansion of LN T cells and 33-fold expansion of spleen T cells. Thy1.2⁺ hosts bearing 10-day s.c. MCA-205 tumors were treated with 5 Gy of TBI, followed by adoptive transfer of activated tumor-reactive T cells. This conditioning regimen is nonmyeloablative, but provides transient profound lymphopenia.

The phenotype of the transferred cells (Fig. 1) demonstrates a mixture of CD8⁺ (46–49%) and CD4⁺ (25–28%) TCR⁺ cells that have uniformly low expression of CD62L and express Thy1.1, but not Thy 1.2. There are minimal phenotypic differences between the LN-derived (upper panels) and the spleen-derived (lower panels) effector T cells. Fig. 2 demonstrates the therapeutic response mediated by adoptive transfer of either LN- or spleen-derived effector cells. The doses of T cells used in this experiment were not uniformly curative against 10-day advanced s.c. tumors, with only three of five mice in each treatment group achieving complete tumor regression, but was statistically significant. In addition, the onset of tumor regression required several days, but was completed by day 33 in cured mice, similar to previous observations using hyperexpanded effector T cells (20).

Mice achieving complete tumor regression were observed for 144 days, and at the time of sacrifice no detectable tumor was present at the inoculation site or in other organs. L-selectin^low memory cells (26% of total splenocytes) were isolated from spleen and were further segregated into subsets of Thy1.2⁺ (10% of initial cell yield) and Thy1.1⁺ (2.8% of initial cell yield). A substantial number of Thy1.1 and Thy1.2 double-negative cells was initially present in the Thy1.1 culture, but did not proliferate, whereas Thy1.1⁺ cells rapidly expanded to predominate the culture. After activation with anti-CD3 and culture in IL-2/IL-7-containing medium for 15 days, there was 82-fold amplification in the Thy1.1⁺ cultures and 95-fold expansion in the Thy1.2⁺ cultures. The stimulated T cells were subjected to MACS selection to remove low level cross-contamination resulting in >99% purity (Fig. 3 A). There was a marked difference in the composition of the cultures, with the adoptively transferred Thy1.1 cells consisting of >90% CD4 and 6% CD8 cells that were L-selectin^low and TCR⁺. In contrast, the culture expanded host memory cells consisted of equal numbers of CD4 (37%) and CD8 (37%) cells, with a substantial fraction of CD4/8 double-negative (25%), L-selectin^low, TCR⁺ cells.

To determine the frequency of responding tumor-reactive T cells among the transferred or host memory cells they were incubated with MCA-205 tumor and assayed for intracellular IFN-γ production. Although there was minimal production of IFN-γ in the absence of stimulation, the majority of the Thy1.1⁺ subset produced IFN-γ specifically in response to MCA-205 tumor stimulation, but was nonreactive against the MCA-207 tumor (Fig. 4,A). Similarly, the Thy 1.2⁺ cells had minimal spontaneous IFN-γ production, but CD4 and CD8 cells reacted specifically against MCA-205 (Fig. 4,B). In keeping with the in vitro analysis of tumor reactivity, adoptive transfer of 10⁷ Thy1.1 or Thy1.2 cells cured established pulmonary metastases and was specific for MCA-205 (Fig. 5,A), but was without therapeutic effect against MCA-207 tumors (Fig. 5 B). The Thy1.1 cells demonstrated higher therapeutic activity on a per cell basis than Thy1.2⁺ cells at a lower cell dose (2 × 10⁶). This may reflect the presence of a higher percentage of memory cells with irrelevant Ag specificity among the host T cells.

Two recipients of adoptively transferred Thy1.1 effector cells developed incomplete tumor regression, followed by tumor growth that was much slower than in naive mice, but nonetheless progressive. Spleens were removed 94 days after adoptive transfer of Thy1.1 effector T cells to determine whether tumor progression resulted from depletion vs anergy of transferred immune cells. At the time of sacrifice, the s.c. tumors were 188 and 149 mm², but there was no detectable metastatic disease similar to numerous other observations with host bearing advanced s.c. MCA-205 tumors (Fig. 2,D). Despite the large s.c. tumor burden, the spleens were hyperplastic with 164 × 10⁶ cells per spleen. The L-selectin^low memory subset comprised 25% of the total splenocytes and was additionally segregated into Thy 1.²⁺ (6.25% of initial cell yield) and Thy 1.1⁺ (1.2% of initial cell yield). Isolated T cells were stimulated with anti-CD3 mAb and cultured in medium containing IL-2 and IL-7. T cells responded with vigorous proliferation over 16 days. As demonstrated in Fig. 3,B, CD4 T cells again comprised the majority of the Thy1.1 cultures, but the ratio of CD4:CD8 cells was 2:1, unlike the highly skewed CD4 predominance in the cured mice. The ratio of CD4:CD8 in the host Thy 1.2 cells was similar to the Thy 1.1 cells with a 2:1 predominance. Again, the cultures were overwhelmingly TCR⁺ and CD62L^low. Thy1.1 or Thy1.2 T cells were purified to >99% homogeneity before adoptive transfer to hosts bearing 3-day lung metastases. As shown in Fig. 5 C, a dose of 10⁷ Thy 1.1 cells was curative, and 2 × 10⁶ cells nearly completely eliminated metastases. This indicates that the transferred tumor-reactive T cells were not exhausted by prolonged exposure to Ag or tumor progression. Although the T cells were not effective in the tumor-bearing host, potent ex vivo activation signals could restore full effector function.

Thy1.1 effector T cells (1.5 × 10⁷) were transferred into 5 Gy of TBI-conditioned recipients that were nontumor bearing. The spleens were harvested 208 days later, and the L-selectin^low cells were fractionated as described above into the Thy1.1⁺ and Thy1.2⁺ subsets and culture-activated as previously described. The long term survival and ex vivo proliferation of CD4 and CD8 cells was equivalent in Thy 1.1 effector cells that had been parked in Ag-free recipients (Fig. 3,C). The phenotype of Thy 1.2 memory cells recovered from tumor-naive hosts showed slight predominance of CD8 cells, but was also CD62L^low and TCR⁺. Similarly to T cells recovered from either cured or tumor-bearing hosts, 24% of the Thy1.1 cells produced IFN-γ in response to MCA-205 tumor, but had minimal reactivity against MCA-207 tumor (Fig. 4,C). Adoptive transfer of 4 × 10⁶ Thy 1.1 eliminated MCA-205 pulmonary metastases, and 2 × 10⁶ cells had a substantial therapeutic effect (Fig. 5,D). As anticipated, because there was no source of tumor Ags during host lymphoid reconstitution, there was no detectable therapeutic activity among the Thy 1.2⁺ cells even at a dose of 10⁷ cells (Fig. 5 D).

These experiments examine the development of memory T cells from adoptively transferred donor effector cells and from host lymphocytes that were regenerating coincident with an immune antitumor response. We used a broad polyclonal population of Ag-primed pre-effector T cells to study their in vivo development into memory cells in the settings of regressing tumor, progressing tumor, or absence of tumor Ag. The tumor model used in this study contributes to the body of in vivo work on memory cell development in that it examines weakly immunogenic tumor Ags rather than antiviral responses or systems with strong Ags and high frequencies of T precursors (22, 23).

There is a linear differentiation of T cells from naive, to Ag-primed, to effector, and then memory cells with different phenotypic and functional characteristics at each stage (24, 25). We chose to study the transition from Ag-primed pre-effector cells to long term memory cells because of the clear implications for clinical applications of cancer immunotherapy. To isolate Ag-primed pre-effectors, L-selectin^low cells were harvested 12 days after tumor inoculation. In mice, L-selectin is expressed at high levels on naive T cells and facilitates migration from the bloodstream into secondary lymphoid organs, but is rapidly down-regulated upon Ag activation and remains at low levels on most long term memory cells (11, 12, 26). The kinetics of tumor Ag sensitization is somewhat delayed compared with soluble Ags or virus inoculation because it takes several days for the tumor to become vascularized, establish lymphatic drainage, and recruit APC. However, L-selectin^low cells are highly enriched for tumor-sensitized T cells, whereas the reciprocal L-selectin^high fraction, which constitutes 80% of the T cells, is completely ineffective (27). Even though L-selectin is down-regulated on effector and memory cells, the inability of freshly isolated, tumor-draining LN cells to mediate tumor regression upon immediate adoptive transfer (G. E. Plautz, unpublished observations) (13) indicates that the L-selectin^low T cells were still at a pre-effector stage. It is likely that the enriched L-selectin^low cells from the initial Thy1.1 host also contained some passenger memory cells of various irrelevant specificities. However, the specificity of tumor immunotherapy exhibited by L-selectin^low T cells indicates that any pre-existing memory cells transferred as passenger cells do not have any detectable anti-tumor function and thus should not interfere with the functional analysis presented in this study (10). Even though the polyclonal population of effector cells precludes tracking with MHC/peptide tetramers, the highly focused immune response and functional competence allow monitoring of the persistence of antitumor memory cells.

The observation that adoptively transferred effector T cells can undergo further differentiation and persist in the recipient as memory cells was anticipated based on TCR transgenic models demonstrating linear differentiation of CD4⁺ or CD8⁺ effector cells to memory cells (24, 25, 26, 27, 28). Even though the transferred T cells were highly activated by anti-CD3 stimulation and rapidly proliferating, they were able to persist after Ag and IL-2 withdrawal in nontumor-bearing hosts. This is consistent with recent findings demonstrating the persistence of memory cells in MHC-deficient hosts, where Ag restimulation is not possible (29, 30).

The predominant isolation of CD4⁺ cells from among the surviving adoptively transferred Thy1.1 cells was somewhat surprising considering that most of the transferred effector cells were CD8⁺. This suggests either that CD4⁺ effector cells were better suited to withstand activation-induced cell death upon encountering tumor in vivo, or that long term surviving CD8 cells are preferentially located in other anatomic compartments. Studies of the anatomic distribution of murine CD8⁺ memory cells responding to viral or bacterial Ags have, in fact, demonstrated their preferential localization in nonlymphoid tissues (31, 32). In addition, the rapid ex vivo proliferation of CD4⁺ compared with CD8⁺ memory cells after anti-CD3/IL-2/IL-7 activation was unanticipated because the Ag-primed L-selectin^low CD8⁺ cells from LNs proliferate much faster than CD4⁺ cells. It is possible that the optimal cytokine requirements for expansion of CD8 memory cells were not used in these experiments. Several recent studies of CD8⁺ memory cells have documented the importance of either IL-7 or IL-15 for acute homeostatic proliferation and IL-15 for maintenance through basal proliferation with inhibitory effects of IL-2 (33, 34). It is important to note that basal or homeostatic proliferation is independent of strong TCR/CD3 activation induced by Ag challenge or in our system by anti-CD3 mAb. Under the activation conditions used in this study, IL-2 does promote the proliferation of CD8 T cells that produce IFN-γ and mediate effector function. In future studies we will compare the effects of IL-15 with IL-2 on the proliferation of memory CD8 cells after anti-CD3 stimulation.

The finding that a new cohort of host memory T cells arose from reconstituting T cells has relevance to clinical applications of T cell immunotherapy for cancer. It is evident that the process of tumor regression allowed tumor Ags to be acquired by DC under activating, not tolerizing, conditions. Sensitized host T cells were competent to complete their differentiation to long term memory cells that were harvested >90 days after the disappearance of tumor Ag. Vaccination with tumor Ag during lymphoid reconstitution has been demonstrated to augment T cell sensitization in other systems (35, 36). It remains to be tested whether regression of established tumor provides a quantitatively or qualitatively different immune response in regenerating T cells (37). Host lymphodepletion with nonmyeloablative chemotherapy is an important component of several clinical protocols of T cell adoptive immunotherapy (38, 39). Presumably, lymphodepletion enhances the survival of transferred T cells as its predominant effect, but the data in this study suggest that such a maneuver could also enhance the emergence of a host antitumor response. This may be highly relevant for protocols that involve adoptive transfer of clonal T cells reactive against a single tumor Ag because a broadened host response could decrease the potential for escape variants (40).

The importance of immune memory for protection against re-exposure to exogenous pathogens is obvious. Relevance of memory for tumor immunotherapy, where the target is exclusively confined to the host is less clear. The success of adoptive immunotherapy is dependent on the dose of effector cells and presumably long term memory is irrelevant if a sufficient dose of effectors is able to rapidly and completely eliminate all tumor cells. However, there are situations where it might be beneficial to optimize the survival of memory T cells. The first is for protection against tumor cells that evade effector T cells in immunoprivileged sites. Our previous studies demonstrated that adoptive immunotherapy could cure established tumors in an immunoprivileged site, the CNS (19, 41, 42). However, it is not known whether detached single tumor cells or small metastatic foci can evade detection during the relatively short-lived effector phase. For example, in our previous clinical trials of adoptive immunotherapy for malignant glioma we observed delayed tumor recurrence in the contralateral hemisphere in a responding subject despite control at the initial site of bulk disease (43).

The second scenario where a memory response is likely to be important is protection against recurrent disease arising from neoplastic stem cells. It is becoming increasingly obvious that many types of cancer arise from neoplastic stem cells that give rise to more differentiated progeny comprising the bulk of the tumor cells. This phenomenon is well described for leukemia (44, 45). More recently, transformed stem cells have been described in breast carcinoma and brain tumors (46, 47). There is evidence that leukemic stem cells are susceptible to elimination in vivo by CD8 T cells directed at a minor histocompatibility Ag, so properly directed T cell responses are likely to be curative (48). However, immune responses generated by vaccination using tissue-restricted Ags, for example, melanocyte or prostate Ags, might allow evasion by the root neoplastic cell. In such situations, the ability to regenerate large numbers of effector cells from tumor-reactive memory cells would be clinically important in controlling, if not curing, disease. The observation that memory cells can be rescued even from hosts with progressive tumors and expanded ex vivo to high numbers indicates that this may be a useful strategy for tumor immunotherapy.