Cytokines often display substantial toxicities at low concentrations, preventing their escalation for therapeutic treatment of cancer. Fusion proteins comprising cytokines and recombinant antibodies may improve the anticancer activity of proinflammatory cytokines. Murine IFNγ was appended in the diabody format at the C-terminus of the F8 antibody, generating the F8–IFNγ fusion protein. The F8 antibody is specific for the extra-domain A (EDA) of fibronectin, a tumor-associated antigen that is expressed in the vasculature and stroma of almost all tumor types. Tumor-targeting properties were measured in vivo using a radioiodinated preparation of the fusion protein. Therapy experiments were performed in three syngeneic murine models of cancer [F9 teratocarcinoma, WEHI-164 fibrosarcoma, and Lewis lung carcinoma (LLC)]. F8–IFNγ retained the biologic activity of both the antibody and the cytokine moiety in vitro, but, unlike the parental F8 antibody, it did not preferentially localize to the tumors in vivo. However, when unlabeled F8–IFNγ was administered before radioiodinated F8–IFNγ, a selective accumulation at the tumor site was observed. F8–IFNγ showed dose-dependent anticancer activity with a clear superiority over untargeted recombinant IFNγ. The anticancer activity was potentiated by combining with F8–IL4 without additional toxicities, whereas combination of F8–IFNγ with F8–TNF was lethal in all mice. Unlike other antibody–cytokine fusions, the use of IFNγ as payload for anticancer therapy is associated with a receptor-trapping mechanism, which can be overcome by the administration of a sufficiently large amount of the fusion protein without any detectable toxicity at the doses used. Cancer Immunol Res; 2(6); 559–67. ©2014 AACR.
Immunocytokines are fusion proteins consisting of a cytokine and a recombinant antibody. They represent a novel class of “armed” antibodies with considerable anticancer potential (1–4). Indeed, antibodies capable of selective accumulation at the tumor site may act as delivery vehicles and may substantially increase the therapeutic index of proinflammatory cytokines.
Various cytokines have been fused to the C-terminal extremity of full immunoglobulin G (IgG) antibodies, leading to products with considerable antitumor activity in mouse models of cancer, and therefore have progressed to clinical studies (1). We and others have fused cytokines to recombinant antibody fragments (e.g., scFv fragments and diabodies) devoid of the Fc portion to generate proteins that do not activate complement or bind to Fc receptors (1). We have focused on the use of antibodies that recognize tumor-associated antigens (TAA) found in the subendothelial extracellular matrix of solid tumors and lymphomas (5–8). In particular, we have used the F8 and L19 antibodies, specific to the alternatively spliced extra-domain A (EDA) and B (EDB) of fibronectin, respectively, for the construction and in vivo testing of several immunocytokines. These oncofetal isoforms of fibronectin are virtually undetectable in normal adult organs (except the placenta and endometrium during the proliferative phase; ref. 9), while they are expressed abundantly in the neovasculature and stroma of virtually all aggressive tumors in mice and humans (7, 10–13). Some proinflammatory cytokines [e.g., interleukin (IL)2, IL12, and TNF] that were fused to L19 or F8 have exhibited impressive anticancer activity and selective uptake at the tumor site (11–20). Other cytokines have shown limitations either in tumor targeting or in therapy experiments (e.g., IL7, IL17, IL15, and IL18; refs. 21–23), indicating that the immunocytokine format and the choice of payload have to be evaluated from case to case. At one extreme, anti-inflammatory cytokines (such as IL10) can be used as fusion partners with disease-homing antibodies with no inhibition of tumor growth (E. Trachsel and D. Neri; unpublished data) but with a substantial inhibition of the autoimmune and/or inflammatory conditions (9, 24).
A comprehensive analysis of antibody payloads, which completely abrogate the in vivo accumulation of the parental antibody at the tumor site, revealed a number of possible mechanisms. For example, a fusion of the L19 antibody with highly charged molecules [such as calmodulin, TAT peptide, or murine VEGF-164 (but not the less-charged VEGF-120 isoform)] abolished tumor targeting in vivo. Similarly, excessive glycosylation (25) or too large molecular weight of the fusion protein (17) can lead to immunocytokines with targeting preference of the parental antibody in vitro, but they do not accumulate at the tumor site in vivo.
IFNγ has long been considered as a payload for immunocytokine development (26, 27), as high concentrations of this protein at the tumor site can mediate a potent influx and activation of leukocytes (28). The expression of IFNγ-based immunocytokines is complicated by the presence of many cysteine residues, which lead to disulfide-linked high-molecular-weight aggregates. We have shown that the deletion of cysteine residues or their mutation to serines leads to the production of immunocytokines with acceptable pharmaceutical properties (29). One of these mutants, L19-IFNγmut4, was found to exhibit measurable antitumor activity, which could be potentiated by combination with L19–IL2. The tumor-targeting activity of L19-IFNγmut4 was substantially higher in mice with deletion of the gene encoding the receptor for IFNγ than in wild-type mice (29).
In this study, we investigated novel IFNγ-based immunocytokines, using the F8 antibody, whose cognate antigen is the EDA domain of fibronectin, a TAA highly expressed in murine (30, 31) and human tumors (11–13). Furthermore, compared with the work of Ebbinghaus and colleagues (29), we explored different strategies for the fusion of IFNγ, which resulted in the production and purification of immunocytokines with pharmaceutically acceptable profiles in SDS-PAGE, gel-filtration, surface plasmon resonance (SPR) analysis, and cytokine activity assays.
We found that F8–IFNγ did not exhibit the expected preferential localization on tumors in vivo. However, when 20 μg of unlabeled F8–IFNγ was administered before the intravenous injection of radioiodinated F8–IFNγ, a selective accumulation of the immunocytokine at the tumor site was observed (tumor:blood ratio = 20 at 24 hours), indicating the presence of a receptor-trapping mechanism that could be saturated. The biodistribution information was used for therapy experiments at relatively high doses (200 μg every 3 days), demonstrating potent antitumor activity with no detectable toxicity for immunocytokine F8–IFNγ. The anticancer activity of F8–IFNγ was superior to that of KSF–IFNγ, a fusion protein consisting of the anti-hen egg lysozyme KSF antibody, which served as a negative control in this study.
Materials and Methods
Cell lines and tumor models
Chinese hamster ovary (CHO) cells, WEHI-279 cells, and murine tumor cell lines F9 teratocarcinoma, WEHI-164 sarcoma, and Lewis lung carcinoma (LLC) were purchased from American Type Culture Collection-LGC (ATCC-LGC). All cells were cultured according to the supplier's protocol, and no additional authentication was performed. All animal experiments were performed under a project license granted by the Veterinäramt des Kantons Zürich, Switzerland (42/2012), and animals were sacrificed when the tumor volume reached a maximum of 2,000 mm3. Female 129/SvEv and male DBA/1 mice were obtained from Charles River Laboratories. Female Balb/c and C57BL/6J mice were obtained from Janvier. For therapy studies and biodistribution experiments, mice were implanted subcutaneously in the flank with 25 × 106 (F9), 3 × 106 (WEHI-164), or 1 × 106 (LLC) tumor cells.
Cloning of IFNγ-based antibody fusion proteins
The fusion proteins F8–IFNγ and KSF–IFNγ contain the antibody F8 (specific to the alternatively spliced EDA domain of fibronectin; ref. 8) or the antibody KSF (specific to egg lysozyme) in the diabody format sequentially fused to murine IFNγ (aa 23-155, Cys155 to Ser155; cDNA from Source BioScience) by a 9-amino acid (aa) linker. Genes encoding the F8 antibody, the KSF antibody, and murine IFNγ were PCR amplified, PCR assembled, and cloned into the mammalian cell expression vector pcDNA3.1(+) (Invitrogen) by a HindIII/NotI restriction site (New England BioLabs; for nucleotide and amino acid sequence, see Supplementary Data S1).
Expression and in vitro characterization of IFNγ-based immunocytokines
Fusion proteins were expressed by the clonal, stably transfected CHO cell line. G418-resistant clones were screened for protein expression by ELISA using recombinant EDA or hen egg lysozyme. For protein production, cells were adapted for suspension growth, and cell culture supernatant was purified to homogeneity by protein A chromatography (GE Healthcare; ref. 32). The purified proteins were analyzed by SDS-PAGE (NuPage 4%–12% Bis–Tris Gel, MOPS running buffer; Invitrogen), size exclusion chromatography (Superdex200 10/300GL; GE Healthcare), and SPR analysis (BIAcore; GE Healthcare) on an EDA-coated sensor chip (8, 33). Cytokine activity was analyzed using a cytostasis assay with WEHI-279 lymphoma cells (20,000 cells/well). Cells were incubated in triplicates for 72 hours with different concentrations of F8–IFNγ, KSF–IFNγ, or recombinant murine IFNγ (produced in Escherichia coli; Merck Millipore), and cell viability was determined with CellTiter AQueous One Solution (Promega).
Immunofluorescence on tumor sections
Freshly frozen cryostat sections (10 μm) of untreated tumors were fixed in ice-cold acetone and stained with biotinylated SIP(F8), SIP(L19), SIP(KSF) or F8–IFNγ, and rat anti-CD31 (BD Biosciences). Streptavidin–Alexa Fluor 488 (Biospa) and donkey anti-rat IgG Alexa Fluor 594 were used for detection. Slides were mounted with fluorescent mounting medium (Dako) and analyzed with an Axioskop2 mot plus microscope (Zeiss).
Quantitative biodistribution experiments
To assess the targeting activity, F8–IFNγ and KSF–IFNγ proteins were labeled with 125-iodine as described previously (23), and the radioiodinated fusion proteins were injected intravenously into the tail vein of tumor-bearing mice. Mice were sacrificed after 1 or 24 hours, and the organs were excised and weighed. Radioactivity content was measured using a Packard Cobra γ counter and expressed as a percentage of injected dose per gram of tissue (%ID/g ± SE).
Blood binding assay
Fresh murine blood obtained from DBA/1 mice was incubated for 10 minutes with the radioiodinated protein preparations at different concentrations in lithium heparin containing Microtainer tubes (BD Biosciences). Blood cells and liquid content were separated by centrifugation (3 minutes; 3,000 × g) and the radioactivity was measured. Results are expressed as a percentage of radioactivity dose.
Tumor therapy studies
For the assessment of antitumor activity, mice were randomly grouped (n = 5) when tumors reached the size of 100 mm3 and treatment was started by intravenous injection into the lateral tail vein according to therapy schedule. The combination agents F8–TNF and F8–IL4 are described elsewhere (30, 31). Mice were monitored daily and tumor volume was measured with a digital caliper. Tumor volume was calculated using the formula: volume = 0.5 × length × width2.
Immunofluorescence analysis of tumor-infiltrating cells
For ex vivo detection of targeting, mice were treated according to the therapy schedule (3 injections, every 72 hours, 200 μg) and tumors were excised 1 day after the final injection. Tumors were embedded in optimum cutting temperature (OCT) medium (Thermo Scientific) and 10-μm cryostat sections were stained using the rat anti-murine IFNγ antibodies (eBioscience) and anti-rat Alexa Fluor 488–coupled secondary antibodies (Invitrogen). For the assessment of tumor-infiltrating immune cells, sections were stained using rat anti-CD45 (leukocytes; BD Biosciences), rat anti-CD4 (CD4+ T cells; BioXCell), rat anti-CD8 (CD8+ T cells; BioXCell), rat anti-F4/80 (macrophages; Abcam), rabbit anti-Asialo GM1 [natural killer (NK) cells; Wako Pure Chemical Industries], rat anti-CD45R (B cells; eBioscience), rat anti-Foxp3 (eBioscience), and rabbit anti-CD25 (Santa Cruz Biotechnology) antibodies, detected with Alexa Fluor 488–coupled secondary antibodies (Invitrogen). For vascular costaining, goat or rat anti-CD31 (Santa Cruz Biotechnology; eBioscience) and anti-goat or anti-rat IgG Alexa Fluor 594–coupled secondary antibodies (Invitrogen) were used. Slides were mounted with fluorescent mounting medium (Dako) and analyzed with an Axioskop2 mot plus microscope (Zeiss).
Differences in targeting, tumor volume, and survival data were compared using the repeated-measures (mixed-model) ANOVA analysis and Mantel–Cox test, respectively, of GraphPad Prism.
Cloning, production, and characterization of fusion proteins
The immunocytokines F8–IFNγ (specific to the EDA domain of fibronectin, a marker for tumor angiogenesis) and KSF–IFNγ (specific to hen egg lysozyme) were cloned and expressed by stable transfection in CHO cells. The immunocytokine comprised the antibody in a stable noncovalent homodimeric diabody format (i.e., 5-amino acid linker between VH and VL domains), with the IFNγ moiety appended to the C-terminal extremity via a flexible 15-amino acid linker (Fig. 1A and B). F8–IFNγ and KSF–IFNγ were purified from cell supernatant by protein A chromatography, yielding a protein preparation (Fig. 1C and E) containing different glycosylation forms similar to the naturally occurring IFNγ (34), as indicated by the two molecular weight bands in SDS-PAGE (Fig. 1C). The product retained high affinity for the cognate antigen, as revealed by SPR analysis (Fig. 1F), and displayed a higher biologic cytokine activity than the IFNγ from recombinant E. coli, as measured in a cytostasis assay with WEHI-279 cells (Fig. 1D).
Biodistribution studies and single-agent therapeutic activity
To study the tumor-targeting properties and the therapeutic activity of immunocytokines F8–IFNγ compared with the negative control KSF–IFNγ, we used three murine models of cancer, grafted onto immunocompetent mice. The F9 teratocarcinoma grows in Sv129 mice, the LLC grows in C57BL/6 mice, and the WEHI-164 fibrosarcoma grows in Balb/c mice. Figure 2A shows that EDA is strongly expressed in the vasculature and in the interstitium of LLC and WEHI-164 tumors, while antigen expression is preferentially found around the angiogenic blood vessels of F9 tumors. Following intravenous injection of 200 μg of either F8-IFNγ or KSF–IFNγ, only the F8–IFNγ immunocytokine could be detected by immunofluorescence to preferentially accumulate in all three types of tumors (Fig. 2B).
A quantitative biodistribution analysis was performed using radioiodinated protein preparations in immunocompetent mice bearing F9 tumors. Although the parental F8 antibody in small immune protein (SIP) format exhibited selective tumor uptake in line with previously published results (8), F8–IFNγ did not accumulate in the tumor when used at a dose of 15 μg (Fig. 2C). However, pre-administration of 20 μg unlabeled F8–IFNγ, followed 15 minutes later by the injection of radioiodinated F8–IFNγ, resulted in a biodistribution profile similar to that of the parental antibody (Fig. 2C), indicating that a trapping interaction could be blocked and saturated at the doses used. It is unlikely that this interaction took place at the level of circulating leukocytes, because an ex vivo incubation of radiolabeled F8–IFNγ (at 1 μg/mL and at 100 μg/mL) with mouse blood followed by centrifugation led to the recovery of the immunocytokine in the supernatant (Fig. 2D).
A quantitative biodistribution analysis performed 1 hour after intravenous injection of radioiodinated protein preparations revealed that, in contrast with the parental F8 antibody in SIP format, the immunocytokine F8–IFNγ mainly localized to the liver and the spleen (Fig. 2E). The pre-administration of unlabeled F8–IFNγ reduced the liver and spleen uptake, whereas F8-B7.2, a highly glycosylated F8-based fusion protein (25) that did not contain the IFNγ moiety, showed no inhibitory effect. These data further support the hypothesis that in vivo trapping of F8–IFNγ at low doses is associated with the presence of the IFNγ moiety.
In a therapy experiment, the dose of F8–IFNγ could be escalated from 7.8 (3-μg equivalents of IFNγ) to 78 μg (30-μg equivalents of IFNγ), with a clear dependence of tumor growth inhibition on the administered dose (Fig. 3A). None of the regimens used was toxic to mice (Fig. 3B).
Biodistribution experiments were repeated in the WEHI-164 and LLC models, using pre-administration of unlabeled immunocytokine and comparing the tumor-targeting activity of F8-IFNγ with the corresponding negative control protein, KSF–IFNγ. In all cases, a preferential tumor accumulation was observed for F8–IFNγ. The tumor uptake (measured as %ID/g) was greater in F9 tumors, compared with the other two tumor models (Fig. 4A, C, and E). In all three syngeneic tumor models tested, the tumor growth inhibition of the targeted F8–IFNγ immunocytokine was significantly greater (P < 0.0001; for statistical analysis of therapy data, see Supplementary Data S2), compared with KSF–IFNγ (Fig. 4B, D, and F). At the chosen dose of 200 μg of F8–IFNγ (corresponding to 78-μg equivalents of IFNγ), no signs of toxicity were observed (Supplementary Data S3). A histologic analysis of tumor sections following immunocytokine treatment, staining for vascular structures (CD31), CD45+ leukocytes, CD4+ and CD8+ lymphocytes, FoxP3 (a marker expressed in regulatory T cells), CD25+ lymphocytes, CD45R+ cells (mainly B cells), Asialo GM1+ cells (mainly NK cells), and F4/80+ cells (mainly macrophages) revealed an increased infiltration of leukocytes in the targeted IFNγ treatment groups, while the infiltration of FoxP3+ cells was decreased (Fig. 5).
Combination studies with immunocytokines
C57BL/6 mice bearing subcutaneous LLC tumors were used to test whether the therapeutic activity of F8–IFNγ could be improved by combining it with other immunocytokines. When tumors were clearly palpable, mice were grouped and therapy started by the injection of either PBS (buffer vehicle), F8–IFNγ (200 μg), F8–TNF (2 μg), or a mixture of F8–IFNγ and F8–TNF (administered together at the same dose used as single agents). Although F8–IFNγ had a more potent tumor growth retardation effect than F8–TNF and the single-agent treatment was not toxic (Supplementary Data S4), all mice in the combination treatment group died after the second injection (Fig. 6A and B). Figure 6C shows the therapy results obtained with 3 injections of F8–IFNγ (200 μg), F8–IL4 (90 μg), or the combination of both. Substantial tumor growth retardation was observed for the combination regimen compared with the single agents (P < 0.0001; for statistical analysis of therapy data, see Supplementary Data S2), but the treatment had no curative results (Fig. 6D). No signs of toxicity were observed (Supplementary Data S4).
While studying the tumor-targeting properties of immunocytokines, we observed that some fusion proteins based on certain cytokines or growth factors (e.g., IL2, IL12, IL15, TNF, GM-CSF, and VEGF-120) retain the tumor-targeting properties of the parental antibody, while others (e.g., those based on TAT peptides, calmodulin, or dual cytokine fusions) completely abrogate tumor targeting in vivo, even though they were fully immunoreactive in vitro (1). We have reported that antibody–IFNγ fusions do not selectively localize to tumors in wild-type mice, but retain the tumor-targeting ability when injected into mice defective for the IFNγ receptor (29). In this study, we show that injecting unlabeled IFNγ-based immunocytokine before injecting labeled fusion protein could overcome the trapping by the IFNγ receptor, reduce the uptake of radiolabeled F8-IFNγ in the liver and the spleen, and allow the immunocytokine to selectively accumulate at the tumor site. The clear dependence of the tumor growth inhibition profile on the dose of the immunocytokine (Fig. 3) indicates that the IFNγ receptor has to be saturated before the development of the antitumor effect. Our data suggest that the receptor trapping of F8–IFNγ can be overcome by the administration of sufficient amounts of the immunocytokine without any additional toxicity at the doses used.
We have described therapy results obtained with the combination of IL2– and IL12–based immunocytokines (20), IL2– and TNF-based immunocytokines (35), IL4– and IL2–based immunocytokines, and IL4– and IL12–based immunocytokines (30). The additive effects of F8–IL4 plus F8–IFNγ are surprising in terms of the opposite T-cell polarization properties of the two cytokines. However, a possible synergy between Th1 and Th2 responses has been predicted previously (28, 36).
We have described recently that receptor-trapping mechanisms of certain immunocytokines can be conveniently tested in vitro by incubation of radiolabeled protein preparations with fresh blood, followed by centrifugation and counting. In the case of F8–IFNγ, however, this method was not predictive for in vivo activity (Fig. 2D), possibly because of receptor expression in the endothelium (37). Indeed, a quantitative biodistribution analysis performed at an early time point revealed an unusually high uptake of the immunocytokine in the liver and the spleen, a process dependent on the presence of the IFNγ moiety in the fusion protein.
The doses of F8–IFNγ used in the mouse to achieve efficient tumor targeting and tumor growth inhibition were in the 10-mg/kg dose range. It is interesting to note that human IFNγ (Actimmune in the United States, Vidara Therapeutics; Imukin, outside the United States, Canada, and Japan, Boehringer Ingelheim) is administered to patients at a recommended dose of 1.5 μg/kg. This discrepancy may reflect different activities of the cytokines among species, making clinical development more difficult. Indeed, while we have demonstrated a strong single-agent tumor growth inhibition with F8–IFNγ in the three syngeneic immunocompetent mouse models of cancer tested, it remains to be considered whether the fully human F8–IFNγ immunocytokine or the F8–IL12 immunocytokine would be the better candidate for clinical development (20, 38). Indeed, F8–IL12 also targets tumors efficiently at low doses (∼0.5 mg/kg) and is able to potently induce IFNγ overexpression at the tumor site (18, 20). On the other hand, clinical development of F8–IFNγ as a “me better” (biosuperior) product may be facilitated by the fact that recombinant IFNγ is an approved biopharmaceutical, while the industrial development of recombinant IL12 has been stopped at the level of several phase II clinical trials. Irrespective of future industrial plans, the lessons learned in this article shed light on the nature of immunocytokine receptor-trapping mechanisms in vivo and suggest possible solutions, such as the use of high doses or the use of mutants with decreased affinity to the receptor.
Disclosure of Potential Conflicts of Interest
T. Hemmerle is a consultant/advisory board member for Philochem AG. D. Neri has an ownership interest (including patents) and is a consultant/advisory board member for Philogen AG.
Conception and design: T. Hemmerle, D. Neri
Development of methodology: D. Neri
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Hemmerle, D. Neri
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Hemmerle, D. Neri
Writing, review, and/or revision of the manuscript: T. Hemmerle, D. Neri
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Neri
Study supervision: D. Neri
This study was financially supported by the Swiss National Science Foundation, the ETH Zurich, the Commission for Technology and Innovation (CTI) Switzerland, the Swiss Cancer League, and the European Union (FP7 Project PRIAT).
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.
The authors thank Phillipp Probst for his help with the cloning procedure and protein preparation.
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).
- Received October 16, 2013.
- Revision received February 27, 2014.
- Accepted February 28, 2014.
- ©2014 American Association for Cancer Research.