Engineered artificial antigen presenting cells facilitate direct and efficient expansion of tumor infiltrating lymphocytes

Patients were entered into an Institutional Review Board-approved clinical protocol and signed an informed consent prior to initiation of lymphocyte cultures. Generation of TILs was performed as described elsewhere [9]. Briefly, 2 mm³ tumor fragments were cultured in complete media (CM) comprised of AIM-V medium (Invitrogen Life Technologies, Carlsbad, CA) supplemented with 2 mM glutamine (Mediatech, Inc. Manassas, VA), 100 U/ml penicillin (Invitrogen Life Technologies), 100 μg/ml streptomycin (Invitrogen Life Technologies), 5% heat-inactivated human AB serum (Valley Biomedical, Inc. Winchester, VA) and 600 IU/mL rhIL-2 (Chiron, Emeryville, CA). TILs established from fragments were grown for 3-4 weeks in CM and expanded fresh or cryopreserved in heat-inactivated HAB serum with 10% DMSO and stored at -180°C until the time of study. Tumor associated lymphocytes (TAL) obtained from ascites collections were seeded at 3e6 cells/well of a 24 well plate in CM. TIL growth was inspected about every other day using a low-power inverted microscope. Each initial well was considered to be an independent TIL culture and was maintained accordingly. For enzymatic digestion of solid tumors, tumor specimen was diced into RPMI-1640, washed and centrifuged at 800 rpm for 5 minutes at 15-22°C, and resuspended in enzymatic digestion buffer (0.2 mg/ml Collagenase and 30 units/ml of DNase in RPMI-1640) followed by overnight rotation at room temperature.

KT64/BBL and KT32/BBL aAPCs were generated, cultured and prepared for co-culture as previously described [13, 15]. Briefly, Fc-binding receptors on KT64/BBL aAPCs were pre-cleared of serum immunoglobulins by culture in serum free AIM-V medium (SFM) overnight and then irradiated at 10,000 rad. Anti-CD3 (OKT-3) with or without anti-CD28 (clone 9.3) mAbs were loaded on aAPCs at 0.5 ug/10⁶ cells at 4°C for 30 minutes. Before use, aAPCs were washed twice with SFM. For KT32/BBL aAPCs, anti-CD3 and anti-CD28 antibodies were not washed out of culture medium, per established protocol [13, 15]. For expansion of IL-2 cultured TILs, an optimal 2:1 aAPC to TIL ratio was established and used in all experiments.

10⁶ heterogonous TILs or TALs were co-cultured with KT64/BBL or KT32/BBL aAPCs loaded with anti-CD3 with or without anti-CD28 antibody in one well of a 24 well plate. rhIL-2 (100 IU/ml) was added into co-cultures at day 2. Every other day the cell number was counted by on a Coulter Multisizer and adjusted to a concentration of 0.5-1 × 10⁶ cells/ml until day 8. Expanding cocultures were transferred into an appropriately sized flask and suspended in CM containing rhIL-2 100 IU/ml depending on total cell numbers. Confirmatory hemacytometer counts including Trypan Blue exclusion were performed. After day 9, phenotypes of expanded TILs or TALs were examined by flow cytometry. Final expanded products were uniformly comprised by CD3+ TILs, TALs or PBLs, without aAPC contamination, as verified by cell sizing, morphology and flow cytometry. The total duration of cell expansion culture was between 9 and 14 days. At the end of culture, all remaining cells were frozen in 90% HAB serum and 10% DMSO for continued analysis. For comparison to other methods of T cell expansion, TILs or TALs were cultured in three conditions: with rhIL-2 (600 IU/ml) in CM; with anti-CD3/CD28 magnetic beads (3:1 beads to T cells) in rhIL-2 (100 IU/ml) (Chiron); or in a "rapid expansion method" condition (200:1 allogeneic PBMC:TILs, 30 ng/ml of OKT-3 anti-CD3 mAb and 6000 IU/ml rhIL-2 in 20 mL of CM in a T75 flask). For stimulation of fresh tumor digests, 10⁶ total cells from tumor digested products were stimulated using an equivalent number of irradiated aAPC loaded with anti-CD3 mAb in media supplemented with 100 IU/mL IL-2.

Antibodies against human CD3, CD4, CD8, CD16, CD25, CD32, CD64 and CD137 were purchased from BD Bioscience. 7-AAD antibody for viability staining was purchased from BD Bioscience (San Jose, CA). HER2:369-377 peptide (KIFGSLAFL) and MART-1:26-35(27L) peptide (ELAGIGILTV) containing HLA-A2010 tetramers were purchased from Beckman Coulter, Inc. (Brea, CA). Anti-FOXP3 antibody (clone 259D) was obtained from BioLegend (San Diego, CA). Fresh TILs or TALs were resuspended in FACS buffer consisting of PBS with 2% FBS (Gemini Bioproducts) at 10⁷ cells/ml and blocked with 10% normal mouse Ig (Caltag Laboratories) for 10 min on ice. A total of 10⁶ cells in 100 μl were stained with fluoro-chrome-conjugated mAbs at 4°C for 40 min in the dark. In some cases, cells were briefly stained with 7-AAD antibody for nonviable cell exclusion after washing twice and subsequently analyzed in a FACSCanto II (BD Biosciences). FOXP3 staining was performed using the eBioscience fixation and permeablization kits according to the manufacturer's instructions and cells stained with the anti-FOXP3 antibody from BioLegend. K562 aAPCs antibody loading was performed using anti-CD3 (OKT3) purchased from eBioscience (San Diego, CA) and anti-CD28 mAbs (clone 9.3). For cell division assays, TILs or PBLs were labeled with 128 nM of carboxyfluorescein succinimidyl ester (CFSE). CFSE labeled TILs or PBLs were expanded with aAPCs, CD3/28 beads, rhIL-2 (600 IU/ml) or REM as described above. At day 6, the cells were stained with anti-CD3, anti-CD4 and anti-CD8 and examined for CFSE division by FACS. Statistical significance of phenotypic differences was determined using paired two-tailed T-test.

Stimulation of TILs by tumor cells was assessed by IFN-γ secretion. 1 × 10⁵ TILs were cultured with 1 × 10⁵ target cells in triplicate overnight in a 96 well U bottom plate in 200 uL of CM containing 5% heat-inactivated human AB serum. Supernatants were harvested and analyzed for IFN-γ by ELISA, according to manufacturer's instruction (Biolegend, San Diego, CA). Values represent the mean cytokine concentration (pg/mL) ± SD of triplicate wells.

K562 cells expressing CD64, CD137L and CD28 ligands CD80 and CD86, pulsed with anti-CD3 antibody efficiently activate and expand CD8+ CD28- T cells and antigen-specific T cells from peripheral blood when co-cultured at a 0.5:1 aAPC to T cell ratio in the absence of exogenous IL-2 and in a CD137L dependent manner [15]. We therefore hypothesized that tumor infiltrating lymphocytes (TILs) derived from cancer lesions could be efficiently expanded to therapeutic treatment numbers using a K562 cell-based aAPC platform. To generate cell-based aAPCs, the parental K562 cell line was engineered to stably co-express the high-affinity Fc receptor CD64 and the costimulatory ligand CD137L (4-1BBL) by lentiviral gene transduction. Single cell clones (referred to as KT64/BBL) were isolated by flow-sorting and their CD64 and CD137L surface expression was confirmed by flow cytometry (Additional file 1Figure S1a). KT64/BBL aAPCs were cultured in the absence of serum to pre-clear CD64 of serum derived immunoglobulins, irradiated and then loaded with anti-CD3 and anti-CD28 agonist monoclonal antibodies (mAbs) for TIL expansion.

TIL cultures for expansion were outgrown from solid ovarian cancer fragments for 3-4 weeks in culture media (CM) containing 600 IU/mL rhIL-2 cytokine, as described [4, 9], and were comprised of >95% CD3+ T cells and <1.5% NK cells. To test the capacity of antibody-loaded aAPCs to mediate ex vivo expansion of TILs, aAPC were co-cultured with TILs at aAPC to TIL ratios ranging between 0.5 and 10 to 1 in the continued presence of IL-2 (100 IU/ml). Peak TIL expansion was achieved at the 2:1 aAPC to T cell ratio (Figure 1a), which contrasts the 200:1 feeder to T cell ratio commonly used in REM-based TIL expansion [9]. The 2:1 aAPC to T cell ratio was therefore used for the experiments detailed below. The contribution of CD137L to TIL expansion was confirmed using control KT64 aAPCs lacking CD137L expression, which mediated diminished TIL expansion compared to KT64/BBL (Additional file 1Figure S1a,b), consistent with our prior study using antigen-experienced T cells [15]. Since our first generation of K562 based aAPC (referred to as KT32/BBL) relied upon the low affinity Fc receptor CD32 for anti-CD3 antibody loading and demonstrated the capacity to expand circulating T cells [15], we evaluated the relative efficiency of CD32 and CD64-expressing aAPCs for expanding TILs. KT64/BBL aAPCs were superior to KT32/BBL aAPCs, and therefore used in all further experiments (Additional file 2Figure S2).

To investigate the impact of CD28 costimulation and IL-2 on aAPC-mediated TIL expansion, KT64/BBL aAPCs were loaded with anti-CD3 mAb +/- anti-CD28 mAb and used to stimulate TILs in the presence or absence of 100 IU/ml of IL-2 (Figure 1b). In the absence of IL-2, TILs underwent minimal expansion after stimulation with aAPCs loaded with anti-CD3 mAb with (11-fold) or without anti-CD28 mAb (9-fold), albeit more than when continually grown in IL-2 (3-fold). By comparison, addition of IL-2 to aAPC-based expansion induced vigorous numerical growth of TILs (>170-fold) in the presence or absence of anti-CD28 mAb, and the level of TIL expansion was similar whether or not anti-CD28 mAb was loaded onto the aAPCs. These results demonstrate that cell-based aAPC-mediated TIL expansion is largely independent of CD28 signaling when 4-1BBL is provided on aAPC, but dramatically improved by addition of IL-2 cytokine to culture.

The limited contribution provided by anti-CD28 mAb to the expansion of TILs in the absence of IL-2 counters that previously observed for peripheral blood T lymphocytes (PBLs) from healthy donors where CD28 costimulation in concert with TCR signaling induces robust proliferation [13, 15]. We therefore evaluated the contribution of CD28 in the expansion of TILs and PBLs collected from the same patient with ovarian cancer. In paired comparison, measurement of CD28 expression on matched TILs and PBLs from the same patients revealed a higher relative expression of surface CD28 by T cells from the circulation than by T cells from tumor in all cases (Additional file 3Figure S3). Among CD3+ TILs, more CD4+ TILs expressed CD28 than CD8+ TILs (76.5 ± 32.9% vs. 34.7 ± 12.2%, respectively; p = 0.003). CD3+ T cells from the blood were heterogeneous in differentiation state and comprised of naïve (CD45RO- CD62L+), central memory (CD45RO+ CD62L+), and effector memory (CD45RO+ CD62L-) cell subsets; TILs however were comprised primarily of cells with a more differentiated, effector memory phenotype (representative examples are shown in Additional file 3Figure S3).

Consistent with their disparate differentiation phenotypes, peripheral blood T cells and TILs from the same patient demonstrated a relative difference in expansion in response to aAPC stimulation. The expansion of TILs in response to stimulation with aAPCs loaded with anti-CD3 mAb with or without CD28 agonist mAb co-loading was modest and similar (62-fold v. 63-fold, respectively), but was substantially augmented by the addition of IL-2 to culture (182-fold; Figure 1c). PBLs in parallel culture exhibited greater expansion in response to anti-CD3 mAb loaded aAPC stimulation compared to TIL, whether or not CD28 signaling was intact, however, PBL expansion was substantially elevated when the aAPCs were also loaded with CD28 agonist mAb (254-fold), relative to anti-CD3 mAb alone (95-fold). In the absence of CD28 costimulation, robust PBL expansion could be restored by addition of exogenous IL-2 cytokine (187-fold). Although PBL expansion in the condition of CD28 costimulation out-performed the addition of IL-2 at day 9 (Figure 1c), IL-2 supplementation was superior to CD28 costimulation by day 11 of PBL culture (737-fold v. 340-fold, respectively); at this time point, TIL cultures were unchanged in expansion hierarchy with a 287-fold expansion in the CD3/IL-2 condition. Consistent with previous findings[15], PBLs stimulated with anti-CD3 and anti-CD28 mAb loaded aAPCs expanded better than those stimulated with magnetic beads coated with anti-CD3 and CD28 mAbs to crosslink endogenous CD3 and CD28 (254-fold v. 56-fold, respectively; Figure 1c). TILs stimulated with CD3/CD28 beads did not undergo robust expansion (18-fold).

Supplement of TIL cultures with IL-2 cytokine, but not CD28 costimulation, during aAPC-induced stimulation dramatically improved TIL expansion, while PBLs showed improved expansion in response to aAPC with addition of either IL-2 or CD28 costimulation. This suggests that PBLs, which express elevated levels of CD28 relative to TILs, may produce and secrete more IL-2 when costimulated than their CD28^low TIL counterparts, thus supporting T cell expansion. Consistent with this notion, cytokine secretion analysis performed on supernatants from TILs or PBLs stimulated overnight with anti-CD3 mAb loaded aAPCs +/- anti-CD28 mAb revealed that TILs produce little to no IL-2 when stimulated with aAPC either with or without CD28 costimulation, or with CD3/CD28 beads (Figure 1d). By contrast, PBLs secreted high levels of IL-2 in response to aAPC which was augmented by the addition of CD28 agonist mAb loading. CD3/CD28 bead stimulation of PBLs resulted in an even greater level of IL-2 production than that achieved with aAPC. Both TILs and PBL secreted IFN-γ and TNF-α in response to aAPC and bead stimulation (not shown), indicating that the lack of IL-2 production by TILs was not a result of functional anergy.

To date, clinical preparation of TILs has largely relied upon expansion by IL-2 alone [4, 36] and, more recently, by the "rapid expansion method" (REM) of anti-CD3 antibody, allogeneic feeder cells and IL-2 [5, 8, 9]. For polyclonal expansion of peripheral blood T lymphocytes, CD3/CD28 beads have been used [10], however their application for TIL expansion has not been reported. We compared the relative effectiveness of KT64/BBL aAPCs and other established culture methods of TIL expansion. TIL cultures outgrown in IL-2 containing CM and primary PBLs were either continually cultured in 600 IU/mL IL-2, or activated with CD3/CD28 beads, REM or KT64/BBL aAPCs. PBLs that were cultured in the presence of IL-2 did not divide, but underwent significant cell division in response to CD3/CD28 beads, although a fraction of cells remained undivided (Figure 2a). CD3/CD28 bead-induced cell division by PBLs was suboptimal and similar in level to that observed after activation with KT64/BBL aAPCs loaded anti-CD3 mAb at the 0.5:1 aAPC to T cell ratio. By comparison, all PBLs divided extensively after stimulation with aAPCs at aAPC to T cell ratios of 2:1 and 5:1, or after expansion by REM.

In contrast to PBLs, a portion of TILs underwent IL-2 induced cell division, likely due to their pre-conditioning in IL-2; however a substantial number of TILs in these cultures did not divide. TILs cultured with aAPCs at the 2:1 ratio underwent extensive cell division, which was similar to that observed in TILs stimulated by the REM, and consistent with T cell counts (Figure 2a). Nearly all TILs stimulated with CD3/CD28 beads or aAPCs at the 0.5:1 ratio divided, albeit at a moderate level. At the 5:1 ratio, most TILs had undergone an intermediate level of cell division, consistent with cell counts (Figure 1b), likely resulting from overcrowding due to space limitations in culture vessels. After 9 days of culture, TILs stimulated by REM or KT64/BBL aAPCs had undergone significant cell expansion, relative to continued IL-2 culture (p < 0.05 by paired t-test; Figure 2b). TILs underwent a mean fold expansion of 205 ± 77 (mean ± SEM) when stimulated with the REM, and a 114 ± 54 fold-expansion by aAPC, a difference which was not statistically significant (p = 0.15). Expansion of TILs with CD3/CD28 beads was not robust, resulting in an 18.8 ± 7.3 mean fold expansion, and was not significantly different from continuous IL-2 culture (21.8 ± 11.9-fold, p = 0.32) or media alone control (4.8 ± 2.2-fold; p = 0.12). To evaluate their continued expansion potential, TILs that had expanded less than 100-fold after a single-round of aAPC stimulation were restimulated with aAPC. After restimulation, TILs underwent further robust expansion, reaching 10,000-fold growth over 25 days (Figure 2c).

Flow cytometric analysis was performed to determine the impact of expansion by the various methods on TIL phenotype. Prior to stimulation, CD4 T cells dominated TIL cultures at a CD4: CD8 ratio of 2.05 ± 0.30 (mean ± SEM; n = 6). After expansion, aAPC stimulated TILs had a low CD4:CD8 T cell ratio (0.77 ± 0.21) that was statistically similar to that observed after REM or IL-2 based expansion (Figure 3a). TILs stimulated with CD3/CD28 beads were largely comprised of CD4 T cells with a CD4:CD8 ratio that was higher than those observed in all other conditions (p < 0.04), likely due to the CD8+ TIL subset containing a much higher proportion of CD28- cells than the CD4+ subset. Although a favorable CD4:CD8 ratio (<1) was seen at the 2:1 aAPC:TIL ratio, higher aAPC:TIL ratios resulted in increased CD4:CD8 ratios following stimulation and culture (Figure 3b). CD16+ NK cells, which were detectable at levels <1.5% of starting IL-2 cultured TIL samples, were not detectable after aAPC-based expansion (not shown). Among CD4+ T cells in TIL cultures, the frequency of FOXP3+ CD4+ T cells was highest in TILs that had been expanded with CD3/CD28 beads, which was significantly greater than in TILs expanded with aAPC (p < 0.05; Figure 3c). Since in vitro activation of T cells can induce transient FOXP3 upregulation [37], analysis was performed only after TILs had rested down as defined by a return of cells to their pre-expansion size, measured using a Multisizer 3 Cell Sizing device, and a lack of spontaneous proinflammaorty cytokine release. The level of FOXP3+ CD4+ T cells was similar among TILs expanded with aAPC, REM or continuous IL-2 culture. The differentiation phenotype of TILs after expansion was not significantly different when stimulated with KT64/BBL, beads, REM or IL-2 with a predominant CD28^int CD27^low CD45RA^neg CD45RO^pos CCR7^low CD62L^int phenotype (not shown).

TILs outgrown from ovarian cancer can recognize and respond to stimulation with autologous tumor as well as known tumor antigens ex vivo [28‐34], although the prevalence of tumor-reactive TILs in ovarian cancer is low. To evaluate whether TILs with specific tumor reactivity are maintained in aAPC expanded cultures, we selected TILs isolated and expanded in IL-2 from melanoma fragments, where tumor antigen-specific T cells are frequently detected for expansion with KT64/BBL aAPCs (Figure 4a). More than a 220-fold expansion was observed over 10-12 days culture in independent assays. MART-1:27-35 peptide-specific CD8+ TILs from HLA-A2+ patients, which were readily detected in pretreatment TILs, were also observed post-expanded TIL populations (Figure 4b). Control HER2:369-377 tetramer staining was negative in these melanoma TIL cultures. In co-culture assays, aAPC-expanded TILs containing MART-1-specific CD8+ T cells retained the ability to recognize and respond to the HLA-matched, MART-1 expressing melanoma cells line 624, but not when stimulated with HLA-A2^neg MART-1⁺ melanoma (938), or HLA-matched MART-1^neg (OVCAR5) or HLA-A2^neg MART-1^neg (SKOV3) ovarian cancer cell lines (Figure 4c), indicating maintenance of anti-tumor reactivity by aAPC expanded TILs.

Extended culture of human T cells results in progressive T cell differentiation and loss of replicative potential which impairs in vivo T cell persistence and anti-tumor responses following adoptive cell transfer [20, 24, 25, 38]. We therefore tested whether so-called "young" TILs could be generated via direct aAPC-based expansion of TILs. We modified the approach of TIL generation, using primary co-cultures of collagenase-digested tumor specimens rather than IL-2 outgrown microcultures derived from solid tumor fragments. Following enzymatic digestion, tumor specimens were comprised of EpCAM+ tumors cells, and a CD45+ leukocyte population that contained CD14+ monocytes and CD3+ T cells, as well as a CD14- CD3- leukocyte subset (Figure 5a). The frequency of CD3+ T cells in the starting digested tumor specimens was low, ranging from 0.76% to 15.68% of all viable cells (mean 6.3 ± 2.1%, n = 7). Stimulation of 1 million total cells from tumor digested products with an equivalent number of irradiated aAPC loaded with anti-CD3/28 antibodies in media supplemented with IL-2 yielded on average a 75-fold numerical expansion of total cells after 11 days, which was substantially higher than that achieved by IL-2 culture alone (mean of 5.6-fold; Figure 5b). Stimulation of the heterogenous tumor cell product resulted in the rapid, preferential expansion of CD3+ CD45+ T cells, which dominated the final cell product (Figure 5a). CD64+ CD137+ aAPCs were not detectable in the final TIL preparation and no viable aAPCs were observed in independent parallel cultures of aAPC alone after day six. Longitudinal enumeration of CD3+ TILs during expansion revealed that TILs, which were a relatively small portion of the starting digested tumor cell product, underwent a robust 1,500-fold mean expansion over 11 days in culture (Figure 5c). "Young" TILs that expanded to modest levels (185-fold mean expansion) were also amenable to secondary expansion with KT64/BBL aAPC, reaching an average total level of ~25,000-fold expansion 8 days after restimulation (Additional file 4Figure S4). Phenotypic analysis revealed that "young" TILs that had been expanded directly from solid tumor digests with aAPC trended toward having increased CD8+ T lymphocyte composition (Figure 5d), a higher frequency of T cells expressing the costimulatory molecules CD27 and CD28 (Figure 5e,f), and reduced frequencies of CD4+ T cells expressing FOXP3, relative to TILs cultured in IL-2 in parallel (Figure 5g), although not to the level of statistical significance. Young ovarian TILs that had been expanded directly from fresh enzyme-digested tumor specimens exhibited autologous tumor reactivity ex vivo (IFN-g secretion >200 pg/mL and twice background) that was statistically similar to the reactivity of TILs that had been outgrown in parallel IL-2 cultures (p = 0.95; n = 4; Figure 5h). Reactivity to MHC-mismatched ovarian cancer cell lines was not observed (not shown) Thus, TILs can be vigorously expanded directly from enzyme-digested tumor specimens ex vivo with KT64/BBL aAPCs, and display favorable phenotypic and functional attributes for the application of adoptive immunotherapy of cancer.