Induction of Higher-Avidity Human CTLs by Vector-Mediated Enhanced Costimulation of Antigen-Presenting Cells Free

Increasing evidence has shown that high functional avidity CD8⁺ T cells can mediate effective immunity to viral infection (1) and against tumors (2–10). However, no vaccine strategy has been developed to effectively enhance the functional avidity of human CTLs.

Several studies in cancer patients have shown that peptide vaccination can induce a heterogeneous peptide-specific CD8⁺ T-cell response; some CD8⁺ T cells have been shown to poorly recognize tumor cells endogenously expressing tumor-associated antigens (TAA; refs. 4, 9, 11, 12). However, the peptide-specific, tumor nonreactive CTLs were rendered tumor reactive once tumor cells were loaded with cognate peptide, indicating that the tumor nonreactive CTLs were of low avidity and not lytic defective (4, 7, 11). In fact, clinical trials have shown that enhanced levels of CD8⁺ T-cell responses following peptide vaccination were not associated with improvement in clinical outcome (12, 13). Many of these studies thus suggest that the efficacy of immune responses may depend on the avidity (quality) of the T cells induced as well as the magnitude (quantity) of T cells induced.

Two different strategies have been described to enrich higher-avidity antigen-specific CTL populations in vitro. One approach is based on the structural affinity of T-cell receptors (TCR) as determined by MHC-tetramer binding. Several studies have since reported that strong tetramer staining is correlated with enhanced tumor reactivity and the approach could be used to isolate “high-avidity” (stronger tetramer binding) CTLs from tumor patients for adoptive transfer therapy (6, 9, 14). However, other studies showed that some strong tetramer-binding CTLs have low or no tumor reactivity (15–19). Alternatively, other investigators (1–3) have shown that higher functional avidity murine CTLs can be enriched by using lower concentrations of peptide during in vitro stimulation (IVS) in a murine viral infection model and a murine tumor model. Although a lower dose of stimulating peptide has successfully expanded higher-avidity CTLs in mouse models (1–3), simply lowering the stimulating peptide dose did not elicit the enhanced avidity of human CTLs but only reduced the magnitude of CTL responses; this was shown in both peptide-vaccinated melanoma patients (5) and normal donors (4). All of the above murine and human studies were conducted by qualitatively or quantitatively altering signal 1 (i.e., antigen).

Previous murine preclinical studies showed that vaccination of mice with recombinant poxviruses containing the transgenes for a triad of costimulatory molecules (B7.1, intercellular adhesion molecule-1, and LFA-3, designated as TRICOM) and a TAA enhanced the level of antigen-specific CD8⁺ T cells generated (20). Oh et al. (21) have shown that vaccination of mice with peptide-pulsed murine B cells infected with a replication-defective avipox (fowlpox) TRICOM vector can lead to the generation of higher-avidity murine T cells. Hodge et al. (22) have shown recently that vaccination of mice s.c. with recombinant vaccinia viruses (rV-) containing transgenes for antigen [β-galactosidase or carcinoembryonic antigen (CEA)] and TRICOM (murine) led to the induction of higher-avidity CTLs than the use of recombinant vaccinia containing the transgenes for the antigens and one costimulatory molecule (B7.1) or just the antigens and no costimulatory molecules. For example, in CEA-transgenic mice, where CEA is a “self antigen,” s.c. vaccination with rV-CEA-TRICOM led to the induction of only a 1.3-fold increase in antigen-specific splenic CTLs than did vaccination with rV-CEA-B7.1 and 2.4-fold more CEA-specific CTLs than the use of rV-CEA. The avidity of the T cells produced, however, was 20-fold greater when employing rV-CEA-TRICOM rather than rV-CEA-B7.1 and 100-fold greater than the use of rV-CEA as the immunogen (22).

We have shown previously (23) that infection of peptide-pulsed human B cells with recombinant avipox (fowlpox, rF-) expressing human B7.1, intercellular adhesion molecule-1, and LFA-3 transgenes (designated rF-TRICOM) will enhance the quantity of antigen-specific human T cells compared with the use of uninfected or wild-type (WT) vector-infected peptide-pulsed B cells. We have also shown that the use of peptide-pulsed human dendritic cells infected with rF-TRICOM vectors enhances the level of human T cells better than the use of peptide-pulsed uninfected dendritic cells or control vector [fowlpox WT (FP-WT)]–infected dendritic cells (24). Neither of these studies (23, 24), however, addressed the avidity of the T cells generated. In the present study, we have investigated whether the use of peptide-pulsed human antigen-presenting cells (APC), infected with the replication-defective rF-TRICOM vector, would induce higher-avidity CTLs compared with uninfected or control vector-infected peptide-pulsed APCs. We employed immature dendritic cells [generated with granulocyte-macrophage colony-stimulating factor (GM-CSF) and interleukin (IL)-4] to simulate conditions when such recombinant vectors are used as i.d. or s.c. injected vaccines. We also employed dendritic cells matured with CD40L to determine whether similar results could be obtained with such APC populations. To our knowledge, these studies are the first to show a method to enhance the avidity of human T cells; they have implications for both the generation of more potent cancer vaccines and the in vitro generation of more effective T cells for adoptive transfer therapy regimens.

Cell line. The human melanoma lines SKMEL24 and DM13 (both HLA-A2 positive, CEA negative), the colon cancer cell lines SW1463 and SW480 (both HLA-A2 positive, CEA positive), and the T2 cells (TAP deficient, HLA-A2-positive human lymphoblastoid cells) were used as targets in ⁵¹Cr release assay or stimulators for cytokine production. All cells were cultured in RPMI supplemented with 10% FCS (Invitrogen, Carlsbad, CA), antibiotics (100 units/mL penicillin, 100 μg/mL streptomycin, 0.25 μg/mL amphotericin B, 50 μg/mL gentamicin), 450 μg/mL l-glutamine, and 2.5 mg/mL sodium bicarbonate in 75 cm² T flasks (Costar, Cambridge, MA).

HLA typing. HLA-A*02 subtyping was done by the Blood Bank of the NIH with PCR using sequence-specific primers. The primers used were described previously by Bunce et al. (25) and Krausa et al. (26). All donors used in this study were HLA-A*0201 positive.

Recombinant viruses. The recombinant fowlpox virus rF-TRICOM containing the transgenes for the human costimulatory molecules B7.1 (CD80), intercellular adhesion molecule-1 (CD54), and LFA-3 (CD58) has been described previously (27–29).

Peptides. The HLA-A*0201-binding CEA agonist peptide, CAP1-6D (YLSGADLNL), has been described previously in detail and is designated here, unless otherwise specified, as CEA peptide (30, 31). The Flu M1 peptide (GILGFVFTL) was derived from influenza matrix protein (32). Both peptides were used to pulse dendritic cells or target cells as indicated. They were synthesized by SynPep (Dublin, CA), and their purity was >95%.

Generation of dendritic cells from peripheral blood mononuclear cells. Dendritic cells were generated from peripheral blood mononuclear cells (PBMC) as described by Romani et al. (33) with some modifications (34) by using GM-CSF (100 ng/mL, PeproTech, Rocky Hill, NJ) and IL-4 (20 ng/mL, PeproTech). On day 6, dendritic cells were either uninfected or infected with fowlpox-based vectors as described previously (24) and then loaded with peptide as APCs to generate CTLs. For some experiments, day 6 dendritic cells generated with GM-CSF/IL-4 were also matured by incubating with CD40L plus the cross-linking antibody Enhancer (each at 1 μg/mL, Alexis, San Diego, CA) for 24 hours; the dendritic cells were then left uninfected or infected with fowlpox-based vectors as described (24).

Antibody, tetramer staining, and flow cytometry assay. FITC-labeled anti-human CD8, CD58, CD80, CD83, HLA-A2 (BB7.2), HLA-DR, phycoerythrin (PE)-labeled CD11c, CD54, and Cy-labeled CD8 were used for staining cell surface molecules. All of the antibodies were purchased from BD PharMingen (San Diego, CA). FITC-labeled COL-1 (anti-human CEA) was prepared in the laboratory. PE-labeled HLA-A2 CAP1-6D tetramer, designated here as CEA tetramer, was provided by the NIH Tetramer Core Facility (Atlanta, GA) and HLA-A2 Flu tetramer was purchased from Beckman Coulter (San Diego, CA). For flow cytometric analysis of cell surface, 2 × 10⁵ to 5 × 10⁵ cells were incubated on ice with the appropriate antibodies for 30 to 45 minutes, washed twice, and analyzed on a FACSCalibur (BD Biosciences, Mountain View, CA). Background staining was assessed using isotype control antibodies. For tetramer staining, cells were stained with FITC- or Cy-labeled anti-CD8 and PE-labeled tetramer for 60 minutes on ice. Data were analyzed using CellQuest.

CTL generation. CTLs were generated using autologous dendritic cells as described previously (4). In brief, Pan T cells isolated using Pan-T kits (Miltenyi Biotech, Bergisch Gladbach, Germany) were stimulated with autologous dendritic cells pulsed with CEA peptide (20 μg/mL) at T cells/dendritic cells ratio of 20-30:1 for three to four cycles of IVS at 7- to 10-day intervals. IL-2 (20 IU/mL) was added 3 days after each IVS, except the first IVS. CTL activity was screened using T2 cells pulsed with native CEA peptide CAP1 (YLSGANLNL) in ⁵¹Cr release assay 7 days after three cycles of IVS.

Purification of tetramer-positive CTLs. Bulk CTL cultures were stained with PE-labeled CEA tetramer for 1 hour at 4°C 7 days following three to four cycles of IVS. Cells were washed twice and incubated with anti-PE-labeled beads for 15 minutes at 4°C to 8°C and tetramer-positive CTLs were isolated with AutoMACS (Miltenyi Biotech) according to the instructions provided by the manufacturer.

Cytotoxicity assays. Cultured CTLs were tested for cytotoxicity in a standard 4-hour ⁵¹Cr release assay (4). Tumor cells (1 × 10⁶−2 × 10⁶/mL) were labeled with 100 μCi sodium chromate for 1 hour at 37°C. Peptide-pulsed targets (1 × 10⁶/mL in the presence of peptide) were labeled with 100 μCi sodium chromate for 2 hours at 37°C. Target cells (5,000 targets per well) were added to wells containing effector CTLs. The percent-specific ⁵¹Cr release was calculated as described previously (4).

Cytokine induction and detection. T cells were cocultured with T2 cells pulsed with or without various concentrations of peptide for 24 hours. The T cells/T2 cells ratio was 10:1. Supernatants were collected at the end of culture and cytokine production was detected using fluorescence-activated cell sorting (FACS)–based Cytometric Bead Array (Human Th1/Th2 Cytokine Cytometric Bead Array kit was purchased from BD PharMingen).

Cytokine/chemokine detection by Luminex. Supernatants from T-cell cultures were collected 24 hours after stimulation with APC-pulsed peptide. A panel of 22 cytokines/chemokines was detected using a Human Cytokine/Chemokine Multiplex Immunoassay Kit from Linco Research (St. Charles, MO) on a Luminex¹⁰⁰ machine (Luminex Corp., Austin, TX) according to the instructions provided by the manufacturer. Data were analyzed using software MasterPlex QT2.0.

Avidity titration. Avidity of peptide-specific CTLs was titrated by cytolytic activity and IFN-γ production against T2 cells pulsed with various concentrations of peptide. Briefly, T2 cells (1 × 10⁶/mL) were pulsed with various concentrations of peptide as indicated at 37°C for 2 hours with (for lytic activity) or without (for IFN-γ production) ⁵¹Cr. T2 cells were washed twice before being used for lytic activity and IFN-γ production assay as described above. Avidity, expressed as MC₅₀ in moles, was defined as the concentration of peptide required to achieve 50% of maximal response and calculated using Microsoft Excel.

Statistical analysis. To assess the statistical significance of different treatment, the ANOVA test was done. The difference between two means was determined by Student's t test. To assess the statistical differences between two curves, a two-tailed t test was done using the asymptotic SE estimate of the difference between the two offset variables of the curves. The Ps were corrected for multiple comparisons by the method of Sidak (35).

rF-TRICOM-infected immature dendritic cells induce a greater magnitude of peptide-specific CTLs. Human dendritic cells, unlike murine dendritic cells, usually express low levels of CD80. The present and previous studies (24, 36, 37) have shown that dendritic cells generated from PBMCs, following GM-CSF and IL-4 treatment, were generally <20% positive for expression of CD80, with the majority of them being <10% CD80⁺. In the present study, adherent monocytes of PBMCs exposed to GM-CSF/IL-4 for 7 days gave rise to a cell population containing ∼90% of CD11c and HLA class II double-positive, CD14- and CD19-negative cells, indicating typical phenotypes of dendritic cells most likely encountered by vector-based vaccines when administered to humans i.d. and/or s.c. As shown in Fig. 1, infection of human dendritic cells with the rF-TRICOM vector resulted in up-regulation of CD80, CD54, and CD58 in terms of mean fluorescence intensity and percent cells positive. Control fowlpox virus infection of dendritic cells had no significant effect on surface molecule expression of HLA class I and II, CD11c, CD54, CD58, and CD80 compared with uninfected dendritic cells. No change was observed on dendritic cell maturation following either FP-WT or rF-TRICOM infection in terms of CD83 expression.

To investigate whether peptide-pulsed dendritic cells/TRICOM, versus peptide-pulsed dendritic cells, expanded CTLs to a greater number, dendritic cells generated from PBMCs of apparently healthy individuals were used either uninfected or infected with either FP-WT or recombinant rF-TRICOM and then pulsed with the CEA peptide. In initial experiments, CEA peptide-pulsed dendritic cells/TRICOM, peptide-pulsed dendritic cells/WT, or peptide-pulsed dendritic cells were used as APCs to stimulate autologous T cells from normal donors for the first cycle of stimulation (priming). The T cells were then further stimulated for two IVS using peptide-pulsed dendritic cells or peptide-pulsed dendritic cells/WT. T cells generated using dendritic cells/TRICOM as priming stimulation produced slightly higher levels of IFN-γ than T cells induced using dendritic cells/WT as priming stimulation (263 versus 161 pg/mL using 1 μg/mL peptide-loaded T2). There was, however, no difference in the avidity of T cells induced using the two methods of generating T cells. Thus, in subsequent studies, T cells were stimulated for all three or four IVS with peptide-pulsed dendritic cells/TRICOM versus dendritic cells/WT. After such three cycles of IVS, the bulk CTLs were stained with MHC-tetramer. As seen in Fig. 2A, the percentage of CEA tetramer-positive CD8⁺ T cells in CTLs elicited by dendritic cells/TRICOM was 6.83%, whereas that of tetramer-positive T cells in both dendritic cells and dendritic cells/WT groups was 2.9%. As a control, CTLs from all three groups were reacted with Flu-M1 tetramer and were negative (Fig. 2A, top).

To test the efficacy of the CTLs induced with peptide-pulsed dendritic cells to lyse target cells, lytic activity of CTLs was measured. As shown in Fig. 2B, CTLs induced by both uninfected dendritic cells and dendritic cells/WT showed specific cytotoxic activity toward T2 cells pulsed with CEA peptide. However, CTLs elicited by dendritic cells/TRICOM displayed a greater ability to lyse T2 cells pulsed with CEA peptide at all effector-to-target (E:T) ratios compared with dendritic cells (P = 0.0216, ANOVA) and dendritic cells/WT (P = 0.0386, ANOVA) groups. None of the CTLs could lyse T2 cells without peptide (Fig. 2B).

CTLs generated by rF-TRICOM-infected immature dendritic cells produce a higher level of cytokines and chemokines following peptide stimulation. The capacity of the different CTLs, induced by dendritic cells with or without rF-TRICOM infection, to produce cytokine following stimulation with peptide was compared. After 24 hours of stimulation with T2 pulsed with CEA peptide, CTLs induced by dendritic cells/WT released a low but detectable (<100 pg/mL) amount of IFN-γ (see Fig. 3A, inset). However, these CTLs did not produce detectable levels of IL-2 (<10 pg/mL; Fig. 3B). In contrast, CTLs induced by peptide-pulsed dendritic cells/TRICOM produced markedly increased levels of IFN-γ compared with CTLs induced by peptide-pulsed dendritic cells/WT (P < 0.0001, two-tailed t test; Fig. 3A) and IL-2 (P < 0.0001, two-tailed t test; Fig. 3B) in a dose-dependent manner following peptide stimulation. No detectable IL-4 and IL-10 (<5 pg/mL) was seen in the supernatants from CTL-elicited dendritic cells/TRICOM or dendritic cells/WT following peptide stimulation.

To further investigate cytokine and chemokine production following TRICOM costimulation, CTLs generated by peptide-pulsed autologous dendritic cells infected with FP-WT or rF-TRICOM further stimulated one more time in vitro with the corresponding dendritic cells pulsed with or without the CEA peptide. When stimulated with peptide-pulsed dendritic cells infected with rF-TRICOM, T cells dramatically increased production of IL-2, IL-4, IL-13, tumor necrosis factor-α, IFN-γ, GM-CSF, macrophage inflammatory protein-1α, and RANTES compared with peptide-pulsed dendritic cells infected with FP-WT control vector or other control groups (Table 1). Other cytokines/chemokines, such as IL-1α, IL-1β, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12, IL-15, IL-17, granulocyte colony-stimulating factor, eotaxin, and IP-10, were marginally increased or did not change following TRICOM stimulation (data not shown).

TRICOM stimulation facilitates the induction of high functional avidity CTLs: cytolytic assay and IFN-γ assay. Studies were then undertaken to determine whether TRICOM had any effect on the induction of higher-avidity CD8⁺ T cells. CTLs generated by uninfected peptide-pulsed dendritic cells, peptide-pulsed dendritic cells/WT, or dendritic cells/TRICOM were titrated for their ability to mediate lysis using different densities of peptide-MHC complexes on T2 target cells. As seen in Fig. 4A, CTLs induced by uninfected dendritic cells and peptide-pulsed dendritic cells/WT displayed the same lytic abilities to various peptide-MHC complex densities on T2 cells. However, CTLs induced by dendritic cells/TRICOM were more efficient in the lysis of T2 cells with lower peptide density compared with CTL-dendritic cells (P = 0.02, ANOVA) and CTL-dendritic cells/WT (P = 0.009, ANOVA) as shown by a shift of dose-response curve to the right (Fig. 4A). As shown previously in murine systems (21) using different CTLs generated by altering signal 1, this is more telling when the data are normalized to percentage of maximal cytolytic responses as shown in Fig. 4B. The avidity (21) of CTL-dendritic cells/TRICOM was 1.27 × 10⁻⁹ mol/L, whereas the avidities for CTL-dendritic cells and CTL-dendritic cells/WT were 3.21 × 10⁻⁸ and 3.32 × 10⁻⁸ mol/L, respectively.

In addition, avidity of the same CTL lines was also determined using IFN-γ production. The avidity of CTL-dendritic cells/TRICOM calculated by IFN-γ production was 1.32 × 10⁻⁹ mol/L, whereas that for CTL-dendritic cells/WT was 4.97 × 10⁻⁸ mol/L (Fig. 4C); these results are very similar to those determined by the cytolytic method for the corresponding CTLs (Fig. 4B).

The stability of tetramer binding to TCR has been shown to correlate with functional avidity (17, 18). To test whether tetramer-TCR complexes on higher functional avidity CTLs elicited by dendritic cells/TRICOM were more stable, a tetramer dissociation assay was done as described (18). As seen in Fig. 4D, CTLs generated by both uninfected peptide-pulsed dendritic cells and peptide-pulsed dendritic cells/WT displayed very similar off-rates of tetramer with a half-time of dissociation close to 50 minutes. However, CTLs induced by peptide-pulsed dendritic cells/TRICOM showed a slower off-rate of 101.5 minutes, which is a 48% increase compared with that of CTL-dendritic cells and CTL-dendritic cells/WT (P = 0.021, two-tailed t test).

Although the cytolytic method is considered the gold standard for titration of functional avidity of CTLs (1, 3, 4, 6, 8, 15, 38), IFN-γ production has also been widely used (2, 3, 5, 38–41); this method also requires fewer cells and does not require radioactive materials. To test whether there was any difference in the avidity of CTLs as determined by the cytolytic method versus IFN-γ production, correlation analysis of lysis versus IFN-γ production employing various concentrations of peptide was done. As shown in Fig. 4E and F, respectively, IFN-γ release was directly correlated with cytolytic activity for CTLs induced by dendritic cells/WT (R² = 0.988) and CTLs induced by dendritic cells/TRICOM (R² = 0.979).

We then compared the functional avidity of CTLs induced by peptide-pulsed dendritic cells/WT and peptide-pulsed dendritic cells/TRICOM from four additional donors using the IFN-γ production method. Table 2 shows that functional avidity of CTLs induced by peptide-pulsed dendritic cells/TRICOM was at least 10-fold higher for four of the five donors than that of CTLs elicited by peptide-pulsed dendritic cells/WT from the same donors.

Comparison of the avidity of CTLs generated by CD40L-matured dendritic cells and rF-TRICOM-infected dendritic cells. Some studies have shown that mature dendritic cells are more potent to induce antigen-specific CTLs than immature dendritic cells (42–44). It was thus of interest to compare the avidity of CTLs induced by TRICOM-infected dendritic cells and by CD40L-matured dendritic cells. To do this, day 6 dendritic cells generated by GM-CSF/IL-4 were matured with CD40L as described in Materials and Methods, with or without infection with rF-TRICOM, and were then used to induce peptide-specific CTLs. As seen in Fig. 5A, exposure of dendritic cells to CD40L greatly increased the expression of CD83 and slightly increased the expression of the costimulatory molecules CD54, CD58, and CD80. Infection of CD40L-treated dendritic cells with either FP-WT or rF-TRICOM did not change the maturation status of dendritic cells in terms of CD83 expression (Fig. 5A). Expression of CD54, CD58, and CD80 on CD40L-treated dendritic cells following infection with rF-TRICOM, however, was greatly up-regulated compared with CD40L-matured dendritic cells in terms of both percentage of cells expressing each of the three costimulatory molecules in TRICOM and their mean fluorescence intensities (Fig. 5A).

The dendritic cells described above were pulsed with CEA peptide and used as APCs to induce peptide-specific CTLs from autologous human T cells. Avidity was titrated using the IFN-γ production method. As can be seen from Fig. 5B, there is ∼2.6- to 4.3-fold increase in IFN-γ production following various doses of peptide stimulation by CTLs induced by CD40L-matured dendritic cells versus nonmatured dendritic cells (P < 0.0001, two-tailed t test). IFN-γ production by CTLs elicited by dendritic cells/TRICOM was much higher than that by CTLs elicited by dendritic cells (P < 0.0001, two-tailed t test); IFN-γ production by CTL-dendritic cells-CD40L/TRICOM was much higher than that by CTL-dendritic cells-CD40L (P < 0.0001, two-tailed t test). The capacity of IFN-γ production by CTLs elicited by dendritic cells-CD40L/TRICOM was slightly superior to that by CTLs induced by dendritic cells/TRICOM (P < 0.001, two-tailed t test; Fig. 5B). In terms of vector controls, there was no significant difference between CTL-dendritic cells and CTL-dendritic cells/WT (P > 0.70, two-tailed t test) or between CTL-dendritic cells-CD40L and CTL-dendritic cells-CD40L/WT (P > 0.70, two-tailed t test) in IFN-γ production following peptide stimulation (Fig. 5B).

To compare the avidity of T cells, IFN-γ production was expressed as a percentage of maximal response (Fig. 5C). Avidity of CTLs induced by TRICOM-infected dendritic cells was 65-fold higher than CTLs induced by dendritic cells without TRICOM and 28-fold higher than CTLs induced by dendritic cells matured with CD40L. Avidity of CTLs, induced using dendritic cells matured with CD40L, was >2.3-fold higher than CTLs induced using dendritic cells treated only with IL-4 and GM-CSF. Finally, CTLs induced using dendritic cells matured with CD40L and infected with rF-TRICOM had an avidity over 100 times greater than that of CTLs induced by dendritic cells matured with CD40L (Fig. 5C; Table 3).

CTLs generated by TRICOM lyse tumor cells more effectively. The functional difference of high- and low-avidity CTLs is that high-avidity CTLs should lyse tumor cells or targets with lower epitope density more efficiently. To compare the cytolytic activity of T cells toward tumor cells endogenously expressing CEA, CTLs induced by CD40L-matured dendritic cells infected with either FP-WT or rF-TRICOM and pulsed with CEA peptide were tested for their lytic activity against human colon carcinoma cells.

Bulk cultures of CTLs derived from a representative donor are shown in Fig. 6; 3.55% of cells induced by CD40L-matured dendritic cells were CD8⁺/tetramer positive, whereas 11.56% of CTLs induced by CD40L-matured dendritic cells/TRICOM were CD8⁺/tetramer positive (Fig. 6A). The avidity of the two CTL lines was titrated using a cytolytic assay (Fig. 6B) and normalized as percentage of maximal lysis (Fig. 6C) to calculate the avidity. As seen in Fig. 6B, CTLs generated by CD40L-matured dendritic cells/TRICOM lysed T2 targets with lower peptide density more efficiently compared with CTLs elicited by CD40L-matured dendritic cells (P < 0.001, two-tailed t test); avidity of CTL-dendritic cells was 4.6 × 10⁻⁸ mol/L and that of CTL-dendritic cells/TRICOM was 2.5 × 10⁻⁹ mol/L (Fig. 6C). The CTL lines were then used to define their efficacy to kill colon carcinomas endogenously expressing CEA. As shown in Fig. 6D, CTLs elicited by CD40L-matured dendritic cells showed marginal cytolytic activity toward colon carcinoma SW1463, which expresses both HLA-A2 and CEA. In contrast, CTLs induced by CD40L-matured dendritic cells/TRICOM showed more potent cytolytic effect to SW1463 at E:T ratios of 40:1 and 13:1 (P < 0.01 and P < 0.05, respectively, Student's t test). As a control, both cell lines did not kill a melanoma cell line SKMEL24 that is HLA-A2 positive and CEA negative (Fig. 6D).

Because peptide-specific CTLs, as determined by tetramer staining (Fig. 6A), in CTL-dendritic cells/TRICOM were ∼3-fold that in CTL-dendritic cells, one may argue that the failure of CTL-dendritic cells to kill SW1463 may be due to a lower number of peptide-specific CTLs. Therefore, tetramer-positive CTLs were isolated as described in Materials and Methods. As shown in Fig. 6E, purity of CD8⁺/tetramer-positive CTLs in both cell lines was >98%. The avidity of purified tetramer-positive CTLs was titrated and calculated as above for bulk CTL cultures. Purified tetramer-positive CTLs induced by CD40L-matured dendritic cells/TRICOM were again more efficient at lower peptide densities on targets compared with those induced by CD40L-matured dendritic cells (Fig. 6F); avidity of purified tetramer-positive CTLs induced by dendritic cells and dendritic cells/TRICOM was 6.1 × 10⁻⁸ and 2.2 × 10⁻⁹ mol/L, respectively, which was very similar to those of bulk CTL cultures (Fig. 6C and G). Cytolytic activity toward tumor cells by purified tetramer CTLs induced by CD40L-matured dendritic cells and dendritic cells/TRICOM was done at equal tetramer to target ratios. As shown in Fig. 6H, purified tetramer-positive CTLs elicited by CD40L-matured dendritic cells showed significant lysis toward SW1463. However, CTLs induced by CD40L-matured dendritic cells/TRICOM were still more efficient in the recognition of the endogenously presented antigenic epitope (P < 0.001 at an E:T ratio of 1:1, Student's t test) compared with CD40L-matured dendritic cells CTLs. In contrast, both purified antigen-specific CTLs did not lyse the HLA-A2-positive, CEA-negative melanoma SKMEL24 (Fig. 6H). In addition, the two CTL lines showed a positive lysis of another colon carcinoma SW480 (HLA-A2 positive/CEA positive) versus no lysis of another melanoma cell line DM13 (HLA-A2 positive/CEA negative; data not shown). Similar results were also obtained from PBMCs from another donor regarding percentage of tetramer-positive cells in bulk cultures (0.28% in CD40L-matured dendritic cell CTLs versus 0.62% in CD40L-matured dendritic cells/TRICOM-CTL), avidity of purified tetramer-positive T cells (1.03 × 10⁻⁷ versus 9.9 × 10⁻⁹ mol/L), and cytolytic activity of purified tetramer-positive T cells toward colon carcinoma SW1463 (28 ± 4.2% versus 51 ± 6.1% at an E:T ratio of 1:1, P < 0.01, Student's t test).

Recent studies in both animal models and clinical studies have shown that the functional avidity of CTLs can be a major determinant for T-cell–mediated tumor immunity (2–8, 10). Therefore, one strategy is to develop vaccines that can selectively induce and expand higher-avidity CD8⁺ T cells. In this study, we show for the first time that enhancing costimulatory signals on human APCs facilitates the induction of higher-avidity human T cells.

Recent clinical trials have shown that the level of immune responses following TAA peptide vaccination is not always consistent with clinical outcome (12, 13). Many factors (e.g., loss of HLA and/or TAA by the tumor or inefficient infiltration of T cells into tumor masses) may be responsible for the failure of clinical response. The relatively low avidity of induced T cells may be another important factor. Several studies have shown that TAA-specific CTL populations in melanoma and carcinoma patients and normal donors appear to be very heterogeneous (4–6, 10, 14, 17, 39, 45, 46). Some peptide-specific CTLs generated from both cancer patients and normal donors could not recognize tumor cells that endogenously present antigen, although these T cells could efficiently mediate lysis of peptide-pulsed targets (4–8). Studies (4, 7, 10, 11) suggest that the majority of TAA peptide-specific T cells are of low avidity and that only a fraction of these cells are relatively higher-avidity T cells, which mediate effective lysis of tumor cells.

It has been assumed that the intensity of tetramer staining would directly correlate with functional avidity of CTLs. Initial attempts to apply this concept to sort higher tetramer-binding CTLs from heterogeneous populations have met with some success (9, 14). However, discrepancy between tetramer binding and functional avidity has also been observed (15–19). In addition, the relative efficiency of staining with the corresponding fluorescent MHC class I-peptide tetramer complexes can vary considerably with staining conditions and does not necessarily correlate with the avidity of antigen recognition (10, 15, 18, 19). It has also been reported that the activation status of CTLs can also affect tetramer binding (39).

Higher functional avidity CTLs have been successfully generated from spleen cells of vaccinated mice by using lower doses of stimulating peptide using both virus- and TAA-derived peptides (1, 2). However, reduction of peptide concentration to activate human T cells resulted in the decreased magnitude of peptide-specific CTL response in bulk cultures without a significant change in functional avidity from PBMCs of both peptide-vaccinated melanoma patients (5) and normal donors (4). These results underscore the differences that can be observed in the activation of murine versus human T cells. Although Oh et al. (21) and Hodge et al. (22) showed in animal studies that TRICOM-based vaccines increased the avidity of antigen-specific CTLs, it was still unclear if human CTLs of enhanced avidity could be generated in vitro. The studies reported here show for the first time that enhanced costimulation using vectors containing a triad of T-cell costimulatory molecules has the capacity to preferentially induce and expand higher-avidity CTLs from human PBMCs. In the present study, we investigated the avidity of CTLs generated by TRICOM-infected dendritic cells from seven normal donors (Figs. 5 and 6; Table 2); six of seven donors showed substantial increases in CTL avidity following dendritic cells/TRICOM stimulation and one donor (Table 2, donor 4) showed a slight increase in CTL avidity. In addition, these studies show that human T cells stimulated with immature dendritic cells infected with rF-TRICOM also produce higher levels of certain important chemokines and cytokines. For example, GM-CSF, macrophage inflammatory protein-1α, and RANTES were dramatically increased following TRICOM and peptide stimulation. GM-CSF is a well-known dendritic cell activation factor, and macrophage inflammatory protein-1α and RANTES are important mediators of acute and chronic inflammation. Increased production of these chemokines will attract more dendritic cells, macrophages, and monocytes to the immunization sites and thus potentially enhance antigen presentation and T-cell activation and expansion. The potential consequence of enhanced secretion of IL-13 by T cells as a consequence of dendritic cell/TRICOM stimulation is not known as this time. IL-13 is a pleomorphic cytokine produced mainly by T cells. Evidence exists that IL-13, among other activities, is an anti-inflammatory cytokine that can enhance monocyte survival and MHC class II and CD23 expression (47) and can generate an “alternatively activated” phenotype in macrophages (48). It has also been indicated that immunosurveillance may be negatively regulated via CD4⁺ natural killer T cells possibly mediated via IL-13 (49).

The infection of human professional APCs with TRICOM vectors may seem counterintuitive at first. In previous publications using murine dendritic cells (50) and human dendritic cells (24), we have shown that when dendritic cells that are already expressing CD54, CD58, and CD80 (the three T-cell costimulatory molecules in TRICOM) are infected with TRICOM vectors, they then express more of these molecules on their cell surface. This, in turn, has been shown to correlate with their ability to enhance the quantity of activated T cells. In the studies reported here, we have shown that dendritic cells treated with CD40L only moderately up-regulate CD54, CD58, and CD80, whereas infection with rF-TRICOM substantially up-regulates each of these three molecules (Fig. 5). CD40L, on the other hand, is shown to up-regulate CD83, whereas rF-TRICOM does not. We show here that the combination of CD40L and rF-TRICOM infection further up-regulates CD54, CD58, CD80, and CD83 (Fig. 5) and results not only in the generation of more CTLs but also in the generation of higher-avidity CTLs (Figs. 5 and 6). The purpose of the studies reported here is to provide a rationale that immature human dendritic cells, which are likely to be encountered by vectors when patients are injected i.d./s.c. with vector-based vaccines, will better facilitate the generation of higher-avidity CTLs. The studies reported here thus complement those recently completed in vivo in mice (22).

The studies reported here show that the higher-avidity CTLs elicited by peptide-pulsed dendritic cells infected with rF-TRICOM showed more stable TCR-MHC-tetramer complex as determined by the tetramer dissociation assay (Fig. 4). Previous studies (17, 18, 24) suggest that increased stability of MHC-peptide and TCR complexes may be a general indicator, or a variable, of high functional avidity of CTLs, although some exceptions may exist. In other words, CTLs with slower off-rates of MHC-peptide complexes from a TCR are not necessarily higher functional avidity CTLs and vice versa.

CTLs induced by TRICOM vector-infected dendritic cells also displayed higher recognition efficacy for tumors endogenously expressing TAA. This was shown by tumor lysis not only by bulk CTL cultures but also by purified tetramer-positive, antigen-specific CTLs, which further indicated that killer cells elicited by TRICOM were of higher avidity (Fig. 6). In addition, CTL avidity calculated in bulk CTL cultures and in purified tetramer-positive T cells was very similar, indicating that the purity of antigen-specific CTLs does not significantly affect titration of CTL avidity and the enhanced avidity in bulk CTLs induced by TRICOM is not simply due to the larger number of antigen-specific CTLs in the bulk cultures.

It has been shown that more mature dendritic cells enhance the level of induction of peptide-specific CTLs in vitro. The studies reported here show that dendritic cells infected with rF-TRICOM (with or without CD40L maturation) enhanced CTL avidity. The present study is not consistent with some previous reports, which have shown that “mature” dendritic cells are required to induce peptide-specific CTLs (42–44). However, the induction of CTL reactivity via dendritic cell peptide presentation in different laboratories shows a highly variable requirement for dendritic cell maturation or CD83 expression (51–54). For example, Kuniyoshi et al. (55) showed that an enhanced CTL response was observed after a single IVS with CD40L-treated dendritic cells compared with non-CD40L-treated dendritic cells, but after an additional IVS with CD40L-treated and nontreated dendritic cells induced comparable peptide-specific CTL reactivity. In contrast, Wurtzen et al. (54) showed that immature dendritic cells were slightly more efficient in inducing peptide-specific CTLs than CD40L-treated dendritic cells following a single IVS; however, after a second IVS, both types of dendritic cells induced equal levels of CTL responses. Another study, reported by Terheyden et al. (52), showed that although CD40 ligation did not change the phenotype of dendritic cells (including CD83), CD40-activated dendritic cells were superior to nontreated dendritic cells in inducing peptide-specific CD8⁺ T cells. Moreover, Zarling et al. (53) showed that the induction of peptide-specific CTLs was primarily donor dependent and peptide dependent and did not reflect the maturation status of the dendritic cells. Taken together, the discrepancy among these studies on the relationship between the maturation stage of dendritic cells and CTL response is probably due to differences among individual donors, different peptides used, culture conditions, and timing for CTL assay.

There may be more than one mechanism underlining the enhanced avidity of CTLs by increased costimulation. Membrane compartmentalization between rafts and nonrafts is required for efficient T-cell activation (56). It was reported that CD28 costimulation induced recruitment of Lck and lipid rafts as well as their accumulation at the immunologic synapse (57, 58). Cawthon et al. (59) found that high-avidity CTLs colocalized substantially more TCR with CD8 compared with low-avidity CTLs. The ability of high-avidity CTLs to respond functionally to fewer TCR engagement events than low-avidity CTLs is directly related to integrating lipid rafts on their surface. In addition, our previous study showed that enhanced expression of costimulatory molecules on target cells infected with rF-TRICOM led to the formation of stable and a greater number of conjugates/synapses between targets and T cells (60). The enhanced interaction between T cells and rF-TRICOM-infected targets also led to enhanced signaling through Lck, ZAP70, and signal transducers and activators of transcription-1 in CD8 T cells (60). Taken together, the results suggest that clustering of membrane and intracellular kinase-rich lipid rafts at the site of TCR engagements and enhanced synapses between T cells and targets/APCs induced by enhanced costimulation may be attributed to the enhanced avidity of CTLs observed in the present study.

In summary, the present study shows for the first time a method in which one can preferentially induce higher-avidity human CTLs from heterogeneous populations as judged by cytolytic activity, IFN-γ production, tetramer-TCR complex stability, and recognition efficacy of tumors endogenously expressing TAA. These results also suggest that vectors expressing multiple costimulatory molecules may be used toward the development of more efficient antitumor vaccines and in vitro to expand higher functional avidity human CTLs for adoptive transfer therapy.

We thank Drs. James Hodge, Helen Sabzevari, and Douglas Grosenbach (Laboratory of Tumor Immunology and Biology, Center for Cancer Research, National Cancer Institute) for critical discussions, Drs. Seth Steinberg and David Venzon (Biostatistics and Data Management Section, National Cancer Institute) for assistance in statistical analysis, and Debra Weingarten for editorial assistance in the preparation of the article.