Trop-2 has limited presence on normal tissues but is highly expressed in diverse epithelial cancers. (E1)-3s is a T-cell–redirecting trivalent bispecific antibody (bsAb), comprising an anti-CD3 scFv covalently linked to a stabilized dimer of a Trop-2–targeting Fab using Dock-and-Lock. We show for the first time that bsAb-mediated bidirectional trogocytosis occurs between target and T cells and involves immunologic synapses. We studied the effects of interferon-α (INFα) on (E1)-3s–mediated T-cell killing of human gastric and pancreatic cancer cell lines. T-cell activation, cytokine induction, and cytotoxicity were evaluated ex vivo using peripheral blood mononuclear cells (PBMC) or T cells with NCI-N87 gastric cancer as target cells. In vivo activity was assayed with NCI-N87 and Capan-1 (pancreatic) xenografts. In the presence of target cells and PBMCs, (E1)-3s did not cause excess cytokine production. When combined with (E1)-3s, peginterferonalfa-2a—which alone did not increase T-cell activation or raise cytokine levels over baseline—increased CD69 expression but did not significantly increase cytokine induction. (E1) 3s mediated a highly potent T-cell lysis of NCI-N87 target cells in vitro. Inclusion of peginterferonalfa-2a or a more potent form of INFα, 20*-2b, significantly potentiated the activity of (E1)-3s by more than 2.5- or 7-fold, respectively. In vivo, combining peginterferonalfa-2a with (E1)-3s delayed Capan-1 growth longer than each single agent. Similarly, combination therapy delayed tumor proliferation of NCI-N87 compared with (E1)-3s or peginterferonalfa-2a single-treatment groups. (E1)-3s effectively induced T-cell–mediated killing of Trop-2–expressing pancreatic and gastric cancers, which was enhanced with INFα. Mol Cancer Ther; 13(10); 2341–51. ©2014 AACR.
This article is featured in Highlights of This Issue, p. 2251
Various formats of bispecific antibodies (bsAb) engineered to redirect cytotoxic T cells, most often via CD3 binding, have demonstrated therapeutic efficacy in preclinical and clinical settings by also binding tumor-associated antigens (TAA; ref. 1). The potential of this modality has been demonstrated preclinically for both hematopoietic (2, 3) and solid cancers (4–8). Tandem scFvs consisting of an anti-CD3 and an anti-TAA domain, termed bispecific T-cell engagers (BiTE), have advanced the furthest, with two agents under clinical investigation (9). The CD19xCD3 BiTE (blinatumomab or MT103) has been studied in phase I trials of lymphoma and leukemia (10), and in phase II trials for B-acute lymphoblastic leukemia (11). MT110, the EpCAM (epithelial cell adhesion molecule) antibody xCD3 BiTE (12), is currently undergoing a phase I study in various solid tumors, including lung, gastric, colorectal, breast, prostate, and ovarian cancers (ClinicalTrials.gov identifier NCT00635596).
Trop-2 [or EGP-1 (epithelial glycoprotein-1), GA733-1, or M1S1], is another TAA that could be highly effective for targeting various epithelial cancers. However, it has yet to be investigated in any bsAb format for T-cell–redirected therapy. Trop-2 is a 35 kDa transmembrane glycoprotein that is overexpressed relative to normal tissues in a variety of human cancers, including pncreatic and gastric carcinomas, in which increased expression correlates with poor prognosis (13–17).
We recently introduced a new platform for producing trivalent bsAbs for T-cell–redirected therapy (18). These constructs, which are assembled using the Dock-and-Lock (DNL) method (19), are designated (X)-3s, in which the code “(X)” denotes a stabilized TAA-specific Fab dimer that is fused site-specifically to an scFv of Okt3, indicated as “3s” (Fig. 1A). The (X)-3s were shown to mediate the formation of conjugates of T cells and cognate target cells, induce T-cell activation and proliferation in the presence of target cells, redirect T-cell–mediated killing of target cells in vitro, and inhibit growth of human tumor xenografts in vivo. One of the constructs exemplified in the report, (E1)-3s, for which (X) was derived from the humanized version of the original murine anti–Trop-2 mAb, RS7 (17), exhibited therapeutic efficacy in a human pancreatic cancer xenograft model (18).
Interferon-alpha (IFNα) has received approval for treatment of several neoplastic diseases (20). In oncology, the main indication of IFNα is for patients with resected stage II and III melanoma, in whom IFNα prolongs disease-free survival and shows a trend toward increased overall survival (21). For some tumors, IFNα can have a direct and potent antiproliferative effect through activation of STAT1 (22). IFNα has a pleiotropic influence on immune responses through effects on myeloid cells, NK cells, DCs, B cells, and T cells, and influences the production of numerous other cytokines, including IFNγ, TNFα, IL1, IL6, IL12, IL15, and IL18 (23). Several reports identify CD8+ T cells as direct targets of IFNα, affecting their IFNγ production, survival, activation, clonal expansion, and memory differentiation (24–31). IFNα provides a strong and direct signal to human CD8+ T cells, thereby resulting in upregulation of critical genes for cytotoxic T-cell activity, and is absolutely critical in the case of human-naive CD8+ T cells for effector function acquisition (26).
This study confirms our prior observations that (E1)-3s effectively induces T-cell–redirected killing of pancreatic and gastric cancer cell lines (18), validating Trop-2 as a choice candidate TAA for therapy of various epithelial-derived cancers using this modality. Here, we show for the first time transfer of membrane proteins by trogocytosis (32, 33) between target and T cells through immunologic synapses induced by a bsAb. Furthermore, we demonstrate that IFNα enhanced the therapeutic efficacy of (E1)-3s without increasing the production of other cytokines to levels that could induce cytokine release syndrome (CRS). The concept of adjuvant therapy with IFNα, or other cytokines, might be universally applicable for enhanced efficacy of T-cell immunotherapy.
Materials and Methods
Human cell lines of Capan-1 (pancreatic cancer), BxPC3 (pancreatic cancer), NCI-N87 (gastric cancer), and Raji (Burkitt lymphoma) were purchased from the ATCC. Cryopreserved vials from an early passage (≤5), were thawed and used within 50 passages for experiments. Samples of the cell lines were authenticated by the ATCC by Promega short tandem repeat authentication testing, in which each was identical with the respective database profile. Peripheral blood mononuclear cells (PBMC), T cells, and cell lines were maintained in RPMI-1640 containing 10% FBS (20% for Capan-1), 1% l-glutamine, 1% penicillin–streptomycin, and 1% Eagle's Minimum Essential Medium nonessential amino acids. All cell culture media and supplements were purchased from Life Technologies.
Peginterferonalfa-2a was obtained commercially. TF12, 20*-2b, and hRS7 were supplied by Immunomedics. Preparation of (E1)-3s, (19)-3s, and (20)-3s (Fig. 1A) were described previously (18). Details of the biochemical methods used for preparation and characterization of (E1)-3s are provided as Supplementary Methods. The bsAbs, (19)-3s (CD19XCD3) and (20)-3s (CD20XCD3), have the same design as (E1)-3s, with the stabilized (Fab)2 to Trop-2 replaced with a (Fab)2 to CD19 and CD20, respectively. TF12 (Fig. 1B), which is a bispecific Tri-Fab generated by DNL to comprise two Trop-2 (hRS7)–binding Fabs fused to a third Fab that binds the hapten, histamine succinyl glycine, and not to any human cells/tissue, has been described previously (34). The 19-3 BiTE, which has the identical deduced amino acid sequence as blinatumomab (CD19XCD3 BiTE), was produced recombinantly from a stable SpESF transfectant cell line, and purified from culture supernatant fluids with Ni-Sepharose.
Flow cytometric studies were performed on a FACSCalibur flow cytometer (BD Biosciences) with the data analyzed with FlowJo software (Tree Star). All antibody cell staining steps were for 30 minutes on ice, followed by two washes with 0.5% BSA–PBS. Fluorochrome-conjugated antibodies included anti–CD4-FITC (fluorescein isothiocynate), anti–CD4-APC (allophycocyanine), anti–CD8-PE (phycoerythrin), and anti–CD69-APC, all from BioLegend; and, goat anti-mouse IgG Fc-FITC (GAM-FITC; Santa Cruz Biotechnology). Detection of Trop-2 with MABC518 (Millipore) was not blocked by any of the bsAbs. Other cell-staining agents were PKH26 Red and PKH67 Green Fluorescent Cell Linker Kits from Sigma and 7-AAD (amino-actinomycin D). The number of cell surface Trop-2 molecules/cell was measured using anti-human Trop-2 PE, clone MR54 (eBiosciences), and QuantiBRITE PE beads (BD Biosciences), following the manufacturer's protocol. Each cell line was measured on three separate occasions.
Isolation of PBMCs and T cells
Human PBMCs were prepared from heparinized whole blood (buffy coat) of healthy donors purchased from The Blood Center of New Jersey (East Orange, NJ) using UNI-SEPMAXI tubes (Novamed Ltd.), under a protocol approved by the New England Institutional Review Board. Total T cells or CD8+ T cells were isolated from PBMCs by negative selection using Pan T Cell or CD8+ T-Cell Isolation Kits (Miltenyi Biotec), respectively, according to the manufacturer's protocol.
PKH67 green–labeled NCI-N87 cells (1.25 × 105 cells/well) were added to 8-well chamber slides (ThermoWaltham) and allowed to attach overnight. The following day, media were removed, and PKH26 red–labeled CD8+ T cells (1 × 106 cells/well) were added in media containing 0.1 μg/mL of (E1)-3s, (19)-3s, or TF12. After 30 minutes at 37°C, slides were dipped in PBS to remove any unbound cells. Fluorescent images were captured with an Olympus BX66 microscope (Shinjuko) equipped with a Mercury-100W laser (Chiu Technical Corp.), using an Olympus 20X/0.75 air objective lens and a Kodak DC290 Camera with 115X zoom. The WB filter cube (dichroic mirror DM500, excitation filter BP450-480, barrier filter BA515) allowed simultaneous fluorescence of both red (appears as orange) and green fluorochromes. Images were captured and processed using Adobe Photoshop CS3 v.10 software with a Kodak Microscopy Documentation System 290 plug-in application.
BxPC3 cells were detached with trypsin (does not affect Trop-2) and mixed with purified T cells. Cell mixtures were treated with 0.1 nmol/L bsAbs at 37°C for 1 hour. Cells were stained with either: (i) anti–Trop-2 MABC518 followed by GAM-FITC, or (ii) anti–Trop-2–PE clone MR54 and anti–CD4-APC. Single BxPC3 and T cells were gated from cell conjugates by forward and side scattering, as well as Trop-2 and CD4 fluorescence.
In vitro cytotoxicity
Freshly isolated CD8+ T cells were incubated for 24 hours with 0.1 nmol/L peginterferonalfa-2a, 0.1 nmol/L 20*-2b (Fig. 1C), or media only. Treated or untreated T cells and PKH67 green fluorescent–labeled NCI-N87 cells were combined at a 5:1 ratio (5 × 104 target cells and 2.5 × 105 effector cells/well) in 48-well plates containing serial dilutions of (E1)-3s in triplicate. Peginterferonalfa-2a or 20*-2b were maintained at 0.1 nmol/L in the appropriate cell mixtures. Plates were incubated for 48 hours at 37°C. Suspension cells were removed and the attached cells were detached with trypsin-EDTA and combined with the corresponding suspension. Cells were washed and resuspended in 1% BSA–PBS containing 30,000 CountBright Absolute Counting Beads (Life Technologies) and 1 μg/mL of 7-AAD. Total live target cells (7-AAD−/PKH67+) were counted by flow cytometry. For each sample, 8,000 CountBright beads were counted as a normalized reference. The specific lysis (%) was calculated using the formula: [1 − (A1/A2)] × 100, where A1 and A2 represent the number of viable target cells in the test and untreated samples, respectively. Statistical significance (P ≤ 0.05) was determined for IC50 (the concentration resulting in 50% lysis), EC50 (50% effective concentration), and lysismax (maximal target cell lysis) by the F test on nonlinear regression (sigmoidal dose–response) curves with Prism software.
Cytokine release was measured ex vivo using 5 × 105 cells/0.5 mL/well of either NCI-N87, which were allowed to attach overnight, or Raji. Freshly isolated PBMCs (5 × 106 cells/0.4 mL/well) were added to each well. Treatments (100 μL, 10×) comprising (19)-3s, 19-3 BiTE, (E1)-3s, peginterferonalfa-2a, or (E1)-3s plus peginterferonalfa-2a were added to 0.1 nmol/L for each reagent. Alternatively, titrations ranging from 1 pmol/L to 10 nmol/L were used for dose–response studies. Following a 20-hour incubation at 37°C with gentle shaking, the supernatant fluid was diluted 1:2 (or greater when necessary) and the concentrations of TNFα, IFNγ, IL2, IL6, and IL10 measured using Single-Analyte ELISArray kits (Qiagen), following the manufacturer's protocol.
In vivo studies
All animal studies were approved by the Rutgers School of Biomedical and Health Sciences Institutional Animal Care and Use Committee. Female 4- to 8-week-old NOD/SCID mice (Charles River Laboratories) were injected s.c. with a mixture of 5 × 106 tumor cells (Capan-1 or NCI-N87) and T cells (2.5 × 106) combined with an equal volume of Matrigel. Therapy began 1 hour later by i.v. injection, as per the BiTE methodology (35). Treatment regimens, dosages, and number of animals in each experiment are described in the Fig. 6 legend. Tumor volume was determined twice weekly by measurements in two dimensions using calipers, with volumes defined as: L × w2/2, where L is the longest dimension of the tumor and w the shortest.
Statistical analysis of tumor growth was based on area under the curve (AUC). Profiles of individual tumor growth were obtained through linear-curve modeling. An F test was used to determine equality of variance between groups before statistical analysis of growth curves. A critical Z test on the survival data identified any outliers within a given treatment group with P ≤ 0.05 censored from the final data analysis. A two-tailed t test was used to assess statistical significance between the various treatment groups and controls, except for the untreated control, in which a one-tailed t test was used. In addition, efficacy was determined by log-rank using Prism software on Kaplan–Meier curves using survival surrogate endpoints as time for tumor progression to 1.0 cm3. Significance was considered at P ≤ 0.05 for all comparisons.
Affinity purification of (E1)-3s using KappaSelect and Ni-Sepharose resins produced a homogeneous product that resolved as a single protein peak of the expected retention time by size exclusion HPLC (high-performance liquid chromatography; Supplementary Fig. S1) and comprising only the three constituent polypeptides (hRS7 Fd-DDD2, hRS7 kappa, and Okt3-scFv-AD2) by reducing SDS–PAGE (Supplementary Fig. S2). LC/MS analysis of (E1)-3s identified as a single RP-HPLC peak having a deconvoluted mass of 135,400.3 Da, which is within 25 ppm of the calculated mass of 135,396.9 Da from its deduced amino acid sequence, including the predicted amino-terminal pyroglutamates on the Okt3scFv-AD2 and each of the two CH1–DDD2–hRS7 Fd chains (Supplementary Fig. S3).
Trop-2–binding affinity measured by surface plasmon resonance (Biacore) for (E1)-3s (KD = 1.03 ± 0.19 nmol/L) was not significantly different (P = 0.1995) from hRS7 IgG (KD = 0.86 ± 0.05 nmol/L; Supplementary Fig. S4).
Cell binding and trogocytosis
(E1)-3s induces the formation of immunologic synapses between T cells and target cells. This was shown previously using Capan-1 pancreatic carcinoma cells (18). Here, addition of 0.1 μg/mL (E1)-3s to a mixture of purified CD8+ T cells and NCI-N87 gastric carcinoma cells, which were membrane-labeled with red and green fluorescence, respectively, resulted in the formation of conjugates evident by fluorescence microscopy (Fig. 2A and B). No conjugates were observed in the presence of (19)-3s (Fig. 2C) or TF12 (Fig. 2D), which bind only T cells or NCI-N87, respectively. Dunking the slides in saline washed off the vast majority of T cells in wells containing (19)-3s or TF12, whereas many T cells remained bound to the adherent NCI-N87 cells in the wells treated with (E1)-3s.
Treatment of BxPC3 (500,000 Trop-2/cell) and purified T-cell mixtures with (E1)-3s specifically induced trogocytosis, whereby Trop-2 was transferred from BxPC3 to T cells Fig. 2E and F). Whereas (E1)-3s treatment resulted in 40% Trop-2+ T cells, <5% of the T cells were counted in the Trop-2+ gate following treatment with control bsAbs binding only Trop-2 (TF12) or CD3 [(20)-3s], or with (E1)-3s in the absence of BxPC3 cells. The uptake of Trop-2 by T cells coincided with its reduction on BcPC3 cells (Fig. 2G and H). During the short incubation time, the T cells (97.5% live) and BxPC3 (94.5% live) remained at high viability, indicating that the T cells acquired the tumor antigens by trogocytosis and not by adhering to membrane fragments of dead cells (Supplementary Fig. S5). Trogocytosis mediated by (E1)-3s was bidirectional, because T-cell membrane components were transferred to BxPC3 cells, as shown for CD4 (Supplementary Fig. S6).
In vitro cytotoxicity
To evaluate redirected T-cell killing of Trop-2–expressing tumor cells, CD8+ T cells were mixed with NCI-N87 cells in the presence or absence of IFNα2 (0.1 nmol/L peginterferonalfa-2a or 20*-2b) along with titrations of (E1)-3s (Fig. 3). Considerable variability in T-cell potency was observed among donors. With a donor of very active T cells, (E1)-3s mediated a highly potent (IC50 = 0.37 pmol/L; lysismax = 77.1%) T-cell lysis of NCI-N87 cells, and inclusion of peginterferonalfa-2a enhanced its activity, improving the IC50 (0.14 pmol/L; P = 0.0001) by more than 2.5-fold and increasing lysismax (84.0%; P < 0.0001; Fig. 3A). NCI-N87 was only weakly sensitive to the direct actions of IFNα (peginterferonalfa-2a IC50 = >10 nmol/L; Supplementary Fig. S7), and inhibited <10% by 0.1 nmol/L peginterferonalfa-2a in the absence of (E1)-3s. The more potent form of IFNα, 20*-2b, consisting of 4 IFNα molecules fused to a bivalent anti-CD20 mAb by DNL (36, 37), enhanced the potency of (E1)-3s by more than 7-fold (IC50 = 0.05 pmol/L; P < 0.0001). At 0.1 nmol/L, 20*-2b inhibited NCI-N87 by 12.6% in the absence of (E1)-3s. The 20*-2b was included only to show enhanced activity with another (more potent) form of IFNα, and that the effect is not restricted to peginterferonalfa-2a. The anti-CD20 mAb moiety is not functional in this experiment. In a similar assay using very weak donor T cells, (E1)-3s was considerably less potent (EC50 = 39 pmol/L; lysismax = 21%); however, addition of peginterferonalfa-2a enhanced the potency by >25-fold (EC50 = 1.4 pmol/L; P = 0.0008; Fig. 3B). Potent (E1)-3s–mediated T-cell killing also was observed for the human pancreatic cancer line, BxPC3 (IC50 = 0.4 pmol/L); however, the effect of adding IFNα was not evaluated with this cell line (Supplementary Fig. S8).
A Trop-2XCD3 BiTE (or equivalent) was not available for comparison with (E1)-3s. However, the availability of both (19)-3s, which has the same (X)-3s molecular configuration as (E1)-3s, and 19-3 BiTE, which has the identical amino acid sequence as the CD19XCD3 BiTE, blinatumomab, enabled a head-to-head comparison to evaluate the relative cytokine-inducing potency of the two bsAb formats. Initially, titrations of (19)-3s and 19-3 BiTE were added to mixtures of PBMCs (two independent donors), and Raji NHL cells and the levels of TNFα, IFNγ, and IL6 were measured after 20 hours (Supplementary Fig. S9). Minimal cytokine levels were detected from PBMCs alone, even with the addition of a bsAb. However, because of a mixed lymphocyte reaction occurring between Raji and the donor PBMCs (stronger for donor A), cytokine levels in untreated cell mixtures were elevated for each TNFα (200 and 50 pg/mL), IFNγ (600 and 200 pg/mL), and IL6 (190 and 220 pg/mL). The levels of TNFα and IL6 were increased above those of untreated only at ≥1 nmol/L (19)-3s. Apparently, (19)-3s inhibited TNFα and IL6 production at lower concentrations. Alternatively, TNFα and IL6 were elevated to >1,000 pg/mL at all concentrations of 19-3 BiTE tested (≥1 pmol/L). The levels of IFNγ were not increased significantly by (19)-3s, whereas 19-3 BiTE showed a dose-dependent increase to >2,000 pg/mL. For all further comparisons, agents were tested at 0.1 nmol/L, which is approximately what has been used in similar studies with BiTE (38). We compared the levels of TNFα, IFNγ, IL2, IL6, and IL10 induced by 0.1 nmol/L (19)-3s or 19-3 BiTE from Raji mixed with PBMCs, using 4 different donors (Fig. 4A). With each of the 4 donors, the levels of each of the five cytokines were significantly higher with 19-3 BiTE, compared with (19)-3s. The mean TNFα concentration with 19-3 BiTE (2,284 ± 1,483 pg/mL) was 8-fold higher (P = 0.0001) than that with (19)-3s (280 ± 188 pg/mL). Treatment with 19-3 BiTE, compared with (19)-3s, resulted in levels of IFNγ (3,002 ± 560 pg/mL vs. 416 ± 169 pg/mL), IL2 (13,635 ± 2,601 pg/mL vs. 1,024 ± 598 pg/mL), IL6 (981 ± 364 pg/mL vs. 168 ± 96 pg/mL), and IL10 (4,006 ± 2,520 pg/mL vs. 493 ± 242 pg/mL) that were 7-, 13-, 6-, and 8-fold higher for 19-3 BiTE, respectively (P < 0.0001 for each). These results indicate that the (X)-3s bsAb format is a considerably less potent inducer of cytokine release, compared with the BiTE format.
In general, (E1)-3s in the presence of PBMCs and target cells caused even less cytokine production than (19)-3s, because there is no mixed lymphocyte reaction to elevate the baseline levels (Fig. 4B). Levels remained low for the proinflammatory cytokines INFγ (<100 pg/mL), TNFα (<100 pg/mL), and IL2 (<250 pg/mL) with 4 of 5 donors. IL6 was low (<400 pg/mL) in 3 of 5 donors, and moderate (800–1,100 pg/mL) in donors D-2 and D-5. Donor D-2 also responded to the (E1)-3s more than the others for IFNγ (1,000 pg/mL) and TNFα (190 pg/mL). IL10, an anti-inflammatory cytokine, was significantly (P < 0.0001) elevated by (E1)-3s to >1,200 pg/mL in 3 of 5 donors. Of note, donor D-2, who had a uniquely potent proinflammatory response, produced relatively low levels of IL10 (230 pg/mL) after treatment with (E1)-3s. Peginterferonalfa-2a alone did not increase the level of any cytokine over background. Addition of peginterferonalfa-2a to (E1)-3s consistently increased IFNγ (∼1.5–3-fold) over (E1)-3s alone. For the rest of the cytokines, there was an apparent trend for a moderately increased production with the combination; however, a consistent effect was not observed.
Addition of 0.1 nmol/L peginterferonalfa-2a increased CD69 upregulation on T cells treated with (E1)-3s moderately, but significantly. For (E1)-3s dose–response experiments measuring %CD69+ T cells, the EC50 was lowered from 26 to 16 pmol/L (P < 0.0001) for CD4+ T cells, and from 11 to 6 pmol/L (P = 0.0204) for CD8+ T cells in the presence of IFNα (Fig. 5A). Peginterferonalfa-2a combined with (E1)-3s resulted in more CD69+ cells (Fig. 5B and C, P < 0.0001), and also, the activated cells had significantly higher CD69 expression with IFNα (Fig. 5B and D; MFI = 907 vs. 726; P < 0.0001). Peginterferonalfa-2a induced minimal CD69 expression in the absence of (E1)-3s. Likewise, (E1)-3, either alone or in combination with peginterferonalfa-2a, did not activate T cells in the absence of target cells.
In vivo antitumor activity of (E1)-3s is augmented with IFNα
In vivo efficacy with human pancreatic cancer was evaluated with Capan-1 xenografts. In the first study, treatment with a combination of (E1)-3s and peginterferonalfa-2a [median survival time (MST) >59 days] was superior to all other treatments (P < 0.0007, log-rank), including (E1)-3s (MST = 50 days) or peginterferonalfa-2a (MST = 53 days) alone (Fig. 6A). Even with the omission of T cells, peginterferonalfa-2a extended survival (MST = 45 days, P = 0.0059 vs. saline, log-rank), indicating direct action on the tumor cells. However, peginterferonalfa-2a was more effective in the presence of T cells (P = 0.0260, AUC), suggesting stimulation of T cells by IFNα. TF12, which binds target but not T cells, did not affect tumor growth or survival. A repeat experiment, using T cells from a different donor, confirmed the results of the first study (Fig. 6B). The second study continued until all groups reached their MST. As in the initial experiment, the combination of (E1)-3s and peginterferonalfa-2a (MST = 119.5 days) was superior to all other groups in terms of both tumor growth inhibition and overall survival [P = 0.0475 vs. (E1)-3s alone; P < 0.0001 vs. all other groups; log-rank]. (E1)-3s (MST = 68 days) was superior (P = 0.0373, AUC over 29 days) to peginterferonalfa-2a with T cells (MST = 53 days) and to T cells alone (MST = 37.5 days; P = 0.0014 log-rank).
For the NCI-N87 gastric cancer xenograft model (Fig. 6C), the combination of (E1)-3s and peginterferonalfa-2a (MST = 100 days) was superior to (E1)-3s alone (MST = 49 days; P = 0.0007, log-rank). Compared with the control group with only T cells (MST = 32 days), peginterferonalfa-2a alone with T cells provided only a minor, but significant, survival advantage (MST = 35 days; P = 0.0276). (E1)-3s plus peginterferonalfa-2a without T cells did not improve survival significantly.
The antigen density measured for NCI-N87 [247,000 (±65,000) Trop-2/cell] and Capan-1 [157,000 (±37,000) Trop-2/cell] was not significantly different. Compared with NCI-N87, Capan-1 cells were >5-fold more sensitive (IC50 = 2 nmol/L vs. >10 nmol/L) to direct inhibition by peginterferonalfa-2a in vitro (Supplementary Fig. S7). (E1)-3s does not cross-react with mouse Trop-2 or CD3 (not shown), and NOD-SCID mice are T-cell deficient.
Previously, we described the potential of the (X)-3s bsAb format for redirecting T-cell–mediated therapy of both hematopoietic and solid tumors using several example constructs, including (E1)-3s, (19)-3s, and (20)-3s (18). In one in vivo experiment from that study, in which Capan-1 xenografts were treated with (E1)-3s, we included groups with peginterferonalfa-2a, because prior (unpublished) data showed that Capan-1 was inhibited by IFNα. Because the aspect of IFNα was preliminary and beyond the scope of the initial report, we presented only the results for the groups without peginterferonalfa-2a. However, the striking enhancement observed with the addition of IFNα spurred further investigation, leading to the current report. The previously reported data, as well as three additional groups, including peginterferonalfa-2a (not published), are included in this report in Fig. 6A. Herein, we also include a new Capan-1 study, which used a different T-cell donor and was extended until all groups reached their MST, confirming that IFNα can enhance the in vivo efficacy of T-cell killing of an IFNα-sensitive cell line. We followed with a study showing that IFNα also can enhance T-cell–mediated killing of a cell line that is weakly sensitive to the direct actions of IFNα. These in vivo studies were performed following methods, including dosing and schedules, typically used with BiTE constructs (35).
Flieger and colleagues demonstrated that in vitro killing by CD3+CD56+ NK-T cells, which were expanded ex vivo and redirected with an EpCAMxCD3 BiTE (MT110), was enhanced with either IFNα or IL2 (39). However, even in the absence of the bsAb, IFNα significantly inhibited the target cells. Because a control to evaluate potential direct effects of IFNα on target cells was lacking, the extent to which the enhanced cytotoxicity was because of IFNα-stimulating NK-T cells, compared with direct inhibition of target cells, could not be determined. Therefore, we measured the sensitivity to IFNα for both target cells and included groups with peginterferonalfa-2a only, both in the presence and absence of pan-T cells. For Capan-1 tumors, which were more sensitive to IFNα in vitro, peginterferonalfa-2a improved survival in the absence of T cells, and even more so with T cells, indicating that IFNα acted on both Capan-1 as well as T cells in this model. In the absence of T cells, peginterferonalfa-2a did not improve survival of mice bearing NCI-N87 xenografts, which were weakly sensitive to IFNα in vitro, indicating that the enhancement with IFNα was due primarily to its actions on T cells. The mechanism of the observed T-cell enhancement by IFNα is unclear. The increase in CD69 expression attributed to IFNα was moderate, but significant, suggesting that the cytokine may potentiate T-cell activation induced with the bsAb. In addition, IFNα specifically increased (up to 3-fold) the release of IFNγ, which is considered the chief cytotoxic cytokine produced by cytotoxic T cells, whereas none of the other cytokines measured increased consistently.
To our knowledge, combination therapy with IFNα and a T-cell–redirecting bsAb has not been investigated clinically, or even in animal models. However, IL2 was combined with an F(ab′)2 fragment of an anti-CD3/EpCAM quadroma in a clinical trial (40), but treatment was limited because of considerable toxicity most likely caused by induction of secondary cytokines, known as CRS or cytokine storm. Systemic administration of IL2 is known to induce a cytokine storm (41), and the severity of adverse events associated with CRS, such as with the TGN1412 catastrophic trial, is correlated with IL2 release (42). Although it is not without side effects, immunotherapy with IFNα, which is not produced by T cells, is not typically associated with cytokine storm.
CRS is a risk associated with immunotherapy using any T-cell–directed mAb (e.g., Okt3) or bsAb, including BiTE (43). However, not all bsAb formats necessarily have the same risk. Brandl and colleagues reported cytokine induction with blinatumomab, in which response levels of IL2, IL6, IFNγ, and TNFα were variable among donors and typically peaked at >1 ng/mL, with some donors reaching levels as high as 5 ng/mL (38). We lacked a suitable BiTE, or equivalent construct, for direct comparison with (E1)-3s. However, we were able to compare the relative cytokine-inducing potency between the (X)-3s and BiTE formats, using a CD19XCD3 BiTE (identical sequence as blinatumomab) and (19)-3s made by DNL. The 19-3 BiTE induced similar cytokine levels as reported by Brandl and colleagues under similar conditions (38). The levels of the five cytokines measured were 7- to 13-fold higher for 19-3 BiTE, compared with those of (19)-3s. The use of foreign lymphoma cells (Raji) caused a mixed lymphocyte reaction, which increased the baseline cytokine levels, particularly for IL2. BiTE, but not (19)-3s, increased the cytokine levels well above the mixed lymphocyte baseline level. Using NCI-N87 gastric carcinoma cells as the target for (E1)-3s did not increase baseline cytokine levels. We observed an expected variability in donor response to (E1)-3s; however, the resulting cytokine levels were even lower than those induced by (19)-3s, particularly for TNFα and IFNγ, which were <100 pg/mL. Nevertheless, 1 of 5 donors had elevated levels (∼1 ng/mL) of IFNγ and IL6. Addition of IFNα (peginterferonalfa-2a) to (E1)-3s increased IFNγ 2- to 3-fold, but did not consistently affect the levels of the other cytokines. These results suggest that compared with other constructs, such as BiTE, the (X)-3s bsAb format may be less likely to induce CRS, and the addition of IFNα to a therapeutic regimen is not likely to increase this risk. These experiments were designed only to compare the two bsAb formats and to evaluate the potentially added risk that might accompany combination therapies, including IFNα. These results are not predictive of actual serum levels expected in the clinic.
We observed considerable variability in the potency of donor T cells. The in vitro results shown in Fig. 3 represent the most and least active T cells that we have tested, with a 100-fold difference in potency (IC50 = 0.37 pmol/L vs. 39 pmol/L) for killing NCI-N87; however, an IC50 = 1 to 5 pmol/L is most representative (>10 donors) and the low-activity T cells were atypical. Notably, lysis with the weaker T cells was augmented by IFNα more than with the potent T cells.
EpCAM is a widely exploited TAA that is overexpressed in many carcinomas. However, the heterogeneous expression of EpCAM in carcinomas and the fact that EpCAM is not tumor-specific, because it is expressed on most normal epithelia, raises concerns that immunotherapy directed toward EpCAM could have severe side effects (44, 45). Like EpCAM, Trop-2 is highly expressed in diverse carcinomas, but its expression in normal tissues is under debate. Several reports indicate that, in contrast with tumor cells, somatic adult tissues show little or no Trop-2 expression, which is invariably upregulated in tumors, regardless of baseline expression in normal tissues (46, 47). However, recent evidence indicates expression of Trop-2 on epithelia of several normal tissues (48). Nonetheless, expression of Trop-2 in Cynomolgus monkeys did not result in toxicities after administrations of reasonably high doses of hRS7 (humanized anti–Trop-2) conjugated with SN-38 as an antibody–drug conjugate (ADC; ref. 49). Furthermore, in clinical studies with this anti–Trop-2 ADC, no increased normal organ toxicity other than manageable neutropenia and diarrhea, expected from the drug (a metabolite of irinotecan), was observed at therapeutic doses (50). Thus, immunotherapy, including T-cell–redirected therapy, using Trop-2 for tumor targeting, is expected to have a similar, or greater, therapeutic index compared with similar regimens targeting EpCAM.
To our knowledge, this is the first report of trogocytosis between target tumor and T cells mediated by a bsAb. This finding demonstrates that the target/T-cell conjugates induced with (E1)-3s have functional immunologic synapses. We observed a similar bidirectional trogocytosis between B and T cells, which was mediated by (19)-3s (unpublished data), and suspect that this is likely a common phenomenon with T-cell–redirecting bsAbs that warrants further investigation.
Disclosure of Potential Conflicts of Interest
D.M. Goldenberg has ownership interest (including patents) in and has provided expert testimony for Immunomedics, Inc. No potential conflicts of interest were disclosed by the other authors.
Conception and design: E.A. Rossi, T.M. Cardillo, C.-H. Chang, D.M. Goldenberg
Development of methodology: E.A. Rossi, D.L. Rossi
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.A. Rossi, D.L. Rossi, T.M. Cardillo
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.A. Rossi, D.L. Rossi, T.M. Cardillo, C.-H. Chang
Writing, review, and/or revision of the manuscript: E.A. Rossi, D.L. Rossi, T.M. Cardillo, C.-H. Chang, D.M. Goldenberg
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.A. Rossi, D.M. Goldenberg
Study supervision: E.A. Rossi, T.M. Cardillo, C.-H. Chang, D.M. Goldenberg
This work was supported by Immunomedics, Inc.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.