An Antibody-Tumor Necrosis Factor Fusion Protein that Synergizes with Oxaliplatin for Treatment of Colorectal Cancer

This article is featured in Highlights of This Issue, p. 2407

Metastatic colorectal cancer remains an important unmet medical need as currently available treatment options are typically unable to induce long-term remissions (1, 2). Immune checkpoint inhibitors remain inefficacious, with the exception of a small population of patients that have microsatellite instability and deficient mismatch repair (3). These features are linked to an upregulation of PD-1 and other immune checkpoint molecules, providing rationale treatment with PD-1 blockers such as nivolumab and pembrolizumab in some cases, and suggesting that colorectal cancer may be amenable to other immunotherapeutic approaches (4).

Engineered cytokine products are gaining importance as a novel class of biopharmaceuticals for cancer immunotherapy (5, 6), with a potential for combination with ionizing radiation (7, 8), cytotoxic agents (9–12), and immune checkpoint inhibitors (13). In our experiments, we explored antibody-based delivery of TNF to colorectal cancer, as an avenue to induce a rapid and selective tumor cell death within the neoplastic lesions while sparing normal organs. The approach has a sound platform; we and others have previously described TNF fusion proteins (14–16) and one of these products (L19-TNF) is currently being studied in pivotal clinical trials for the treatment of first-line metastatic soft-tissue sarcoma (NCT03420014; EudraCT number: 2016-003239-38).

For a target, we employed carcinoembryonic antigen (CEA), one of the best and most-validated tumor-associated antigens for malignancies of epithelial origin (17). Importantly, a substantial body of nuclear medicine data has confirmed that metastatic colorectal cancer lesions can efficiently be reached in vivo by anti-CEA antibodies, both as intact immunoglobulins (18) and in antibody fragment formats (19, 20). Sm3E, the targeting agent of choice, is an affinity-matured, humanized, anti-CEA single-chain Fv antibody fragment (scFv), with an extremely low dissociation constant (KD = 30 pmol/L) and established tumor-targeting potential in tumor-bearing mice (21). Antibody-TNF fusions specific to CEA have previously been described, but exhibited suboptimal biodistribution and therapy results (16, 22). We hypothesized that the use of a best-in-class antibody fragment and a mammalian cell expression system could improve in vivo performance, as Escherichia coli production of antibody therapeutics has previously been associated with suboptimal tumor-targeting results (23).

The organisms of choice when it comes to antibody production are mammalian cell lines, in particular Chinese hamster ovary (CHO) cells. Through the years, the production systems have been optimized and the use of CHO cells is nowadays a reliable antibodies production system. CHO cells expressing the antibody of interest are cultivated in large bioreactors, harvested by centrifugation, and the supernatant goes through a series of chromatographic-based purification steps, which allow for separation of aggregates and impurities, including endotoxins. Once the antibody has been purified to homogeneity, a suitable formulation is implemented to maintain the antibody in a stable conformation and ready for downstream analysis once all the quality controls have been performed (24, 25).

Here, we describe the generation and characterization of Sm3E-TNF, a mammalian-expressed fusion protein of Sm3E with murine TNF. We demonstrate selective targeting of Sm3E-TNF to CEA-transfected murine CT26 tumors in immunocompetent BALB/c mice models, leading to rapid hemorrhagic necrosis of the tumor mass. Moreover, our study showed that Sm3E-TNF exhibited potent anticancer activity as a single agent and was able to achieve complete tumor eradication in up to 80% of study animals when used in combination with oxaliplatin, an anticancer drug commonly used for the treatment of colorectal cancer (26).

CHO cells, LM fibroblasts, CT26 colon carcinoma cells, and HT29 colorectal adenocarcinoma cells were obtained from the ATCC. The C51 colon carcinoma was kindly provided by M.P. Colombo (Istituto Nazionale Tumori, Milan, Italy). Cell lines were received between 2010 and 2017, expanded, and stored as cryopreserved aliquots in liquid nitrogen. Cells were grown according to the manufacturer's recommendation. Authentication of the cell lines, including check of post-freeze viability, growth properties and morphology, test for Mycoplasma contamination, isoenzyme assay, and sterility, were performed by the cell bank before shipment.

The gene for CEA was cloned into the mammalian expression vector pcDNA3.1(+) containing an antibiotic resistance for G418 geneticin. CT26 wild-type cells were transfected with the pcDNA3.1(+) containing the human CEA gene using the TransIT-X2 Transfection Reagent (Mirus Bio) according to manufacturer's recommendations. Five days after the transfection, the medium was replaced with RPMI (10% FBS and 1% antibiotic-antimycotic) containing 0.5 mg/mL G418 (Merck) to select a stably transfected polyclonal cell line. A monoclonal cell line was generated from the stable polyclonal cell line by preparative single-cell sorting performed using a BD FACSAria III. The process was tracked by FACS analysis. Different clones were expanded and further tested for antigen expression. Clone CT26 1D6 was selected for further use on the basis of CEA expression, which was shown both by FACS and by immunofluorescence analysis. Similar procedures were used for the stable transfection of C51 murine adenocarcinoma cells with human CEA, but electroporation with Amaxa 4D-Nucleofactor (Lonza) with the SG Cell Line 4D-Nucleofector X Kit L (Lonza) was used for transfection. Retention of CEA expression over multiple passages was confirmed by flow cytometry analysis (Supplementary Fig. S1). Relevant features and sequences of the vector used for cell transfection can be found in Supplementary Fig. S2.

For cellular expression analysis of CEA, cells were detached with 50 mmol/L EDTA and 5 × 10⁵ cells were stained with rabbit anti-CEA antibody (Sino Biological, 500 nmol/L, 1 hour, 4°C) or Sm3E-TNF (500 nmol/L, 1 hour, 4°C) in a volume of 100 μL FACS buffer (0.5% BSA and 2 mmol/L EDTA in PBS). For signal amplification, the anti-CEA antibody was detected with an anti-rabbit AlexaFluor488-labeled antibody (Invitrogen, 1:200, 45 minutes, 4°C), whereas the Sm3E-TNF fusion protein was detected with a FITC-labeled protein L (ACROBiosystems, 1:200, 45 minutes, 4°C). In between and after staining, cells were washed with 100 μL FACS buffer and centrifuged at 300 rcf for 3 minutes. Stained cells were analyzed with a 2L-Cytoflex (Beckman Coulter). When looking for CEA expression, HT29 cells were used as positive control, while CT26 wild-type and C51 wild-type cells were used as negative control. FACS data were analyzed using the FlowJo 9 Software Suite (FlowJo LLC).

CEA-transfected CT26 cells were grown to 80% confluence and detached with Trypsin-EDTA 0.05% (Life Technologies). Cells were then washed once with Hank's Balanced Salt Solution (HBSS, pH 7.4, Gibco), counted, and resuspended in HBSS to a final concentration of 2.0 × 10⁷ cells/mL. Aliquots of 3 × 10⁶ cells, corresponding to 150 μL of the suspension, were injected subcutaneously in the right flank of female BALB/c mice (8–10 weeks of age, Janvier). Similar procedures were used for the CEA-transfected C51 cells, but aliquots of 1 × 10⁶ cells in 100 μL suspension were injected. Experiments were performed under a Project License granted by the Veterinäramt des Kantons (Zürich, Switzerland, 27/2015).

The VL and VH genes (US 2008/0003646 A1) of an affinity-matured humanized anti-CEA mAb (Sm3E; ref. 27) were fused through a G₄S-linker sequence. The resulting VH-(G₄S)₃-VL ScFv fragment was further fused at the N-terminus of the murine TNF gene through a S₄G-linker and the final construct, VH-(G₄S)₃-VL-(S₄G)₃-TNF, was then cloned into the mammalian expression vector, pcDNA3.1 (+) vector (Invitrogen). A VH-(G₄S)₃-VL ScFv fragment specific for hen egg lysozyme (KSF; ref. 28), used as negative control, was fused at the N-terminus of the TNF as described for the Sm3E construct. The sequences for both fusion proteins can be found in Supplementary Figs. S3 and S4.

The fusion proteins were expressed in CHO cells by transient gene expression, using procedures described previously (29), and were purified to homogeneity using either protein A (KSF-TNF, Sino Biological) or protein L (Sm3E-TNF, Thermo Fisher Scientific) chromatography. The purified proteins were characterized by SDS-PAGE (Invitrogen), Size exclusion Chromatography (Superdex200 10/300GL, GE Healthcare), and Protein Mass Spectrometry (UPLC-ESI-ToF-MS, Waters). The biological activity of Sm3E-TNF was assessed by incubation with mouse LM fibroblasts or CEA-transfected CT26 cells. In this assay, cells were incubated in 96-well titer plates containing medium supplemented with 2 μg/mL actinomycin D and varying concentration of fusion protein. After 24 hours at 37°C, cell viability was determined using a CellTiter Aqueous One Solution Kit (Promega).

The in vivo tumor targeting performance of the fusion proteins was assessed by quantitative biodistribution after radiolabeling following previously published experimental procedures (30). Briefly, approximately 7 μg of radioiodinated Sm3E-TNF or KSF-TNF was injected into the lateral tail vein of CEA-transfected CT26 tumor–bearing BALB/c (Janvier) mice (n = 5). Animals were sacrificed by CO₂ asphyxiation 24 hours after injection. Organs and tumors were excised, weighed, and the corresponding radioactivity was measured using a Cobra γ Counter (Packard). Biodistribution results were expressed as percent of injected dose per gram of tissue (%ID/g ± SEM).

Mice bearing CEA-transfected CT26 tumors were injected with a single dose of fusion protein alone or in combination with oxaliplatin, according to the therapy schedule, and subsequently sacrificed at variable timepoints after injection for microscopic analysis. Tumors were excised and embedded in Cryoembedding Medium (Thermo Fisher Scientific). Ice-cold acetone-fixed cryostat sections (10 μm) were prepared and stained with hematoxylin and eosin (H&E, Sigma-Aldrich) using routine methods, or subjected to immunofluorescence analysis. For immunofluorescence, the acetone-fixed sections were stained using the following antibodies: rat anti-CD31 (BD Bioscience, 553370), goat anti-CD31 (R&D Systems, AF3628), rabbit anti-activated-caspase-3 (Sigma, C8487), rat anti-NKp46 (BioLegend, 137601), rat anti-CD4 (BioLegend, 100423), rat anti-CD8 (BioLegend, 100702), and rat anti-FoxP3 (eBioscience, 14-5773-82), and detected with anti-goat AlexaFluor594 (Invitrogen, A11058), anti-rat AlexaFluor594 (Invitrogen, A21209), anti-rat AlexaFluor488 (Invitrogen, A21208), and anti-rabbit AlexaFluor488 (Invitrogen, A11008). Cell nuclei were counterstained with DAPI (Invitrogen; D1306). Slides were mounted with fluorescence mounting medium and analyzed with Axioskop2 Mot Plus Microscope (Zeiss).

Tumor growth was monitored daily and a caliper was used to determine the tumor volume (volume = 0.5 × length × width²). When tumors had reached a volume of approximately 100 mm³, mice were randomly subdivided into distinct study groups (n = 5). Sm3E-TNF (4 μg), dissolved in PBS, was administered every 48 hours by injection into the lateral tail vein. Vehicle control animals received the same volume of PBS alone. Chemotherapeutic agents [5-fluorouracil (5-FU, 50 mg/kg, Sandoz Pharmaceuticals AG) or oxaliplatin (7.5 mg/kg, Sandoz Pharmaceuticals AG)], dissolved in PBS, were injected into the lateral tail vein only once, 30 minutes prior to the first fusion protein administration.

The presence of AH1-specific T cells in draining lymph nodes and tumors of mice treated with saline, oxaliplatin, and Sm3E-TNF alone or in combination with oxaliplatin was assessed by flow cytometry analysis, following experimental procedures published previously (31).

Data were analyzed using Prism 7.0 (GraphPad Software, Inc.). Statistical significance was determined with a regular two-way ANOVA test with a Bonferroni posttest. Data represent mean ± SEM, and P < 0.05 was considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P ≤ 0.001; and ****, P < 0.0001).

We cloned and expressed Sm3E-TNF in CHO cells. The fusion protein featured the antibody in scFv format (32), leading to the formation of a stable noncovalent homotrimeric structure, mediated by the murine TNF moiety (Fig. 1A). The product was purified to homogeneity using protein L chromatography, as assessed by SDS-PAGE analysis, size exclusion chromatography, and mass spectrometry (Fig. 1B–D). In parallel, we produced KSF-TNF, a negative control fusion protein comprising an scFv specific to hen egg lysozyme fused to murine TNF (Supplementary Fig. S5). Sm3E-TNF was able to kill TNF-sensitive LM1 fibroblasts (as well as CEA-transfected CT26 cells) in vitro at picomolar concentrations (Fig. 1E). Stably transfected murine CT26 and C51 colorectal cancer cell lines, expressing human CEA (the cognate antigen recognized by Sm3E) were produced by conventional methodologies, as described in the Materials and Methods section.

Specific binding of Sm3E-TNF (but not KSF-TNF) to these cell lines was confirmed by FACS analysis, using fluorescently labeled protein L as detection reagent (Fig. 2).

Twenty-four hours after intravenous administration to mice, Sm3E-TNF preferentially localized to CEA-expressing CT26 murine colorectal tumors in vivo, as assessed by a quantitative biodistribution analysis with radiolabeled protein preparations, where Sm3E-TNF had a 3.52 ± 0.38 %ID/g and tumor to blood ratio of 55.85 ± 10.42, whereas KSF-TNF had a 0.74 ± 0.28 %ID/g and tumor to blood ratio of 3.46 ± 1.34 (Fig. 3).

A first set of monotherapy experiments confirmed that the tumor-homing Sm3E-TNF product exhibited a more potent anticancer activity than KSF-TNF, with a significant difference (P ≤ 0.001) observed for slow growing CEA-expressing CT26 tumors. However, at the dose used (three injections of 4 μg of Sm3E-TNF), cancer could not be cured in a monotherapy regimen, which was associated with a body weight loss of approximately 5% (Fig. 4).

Chemotherapeutic strategies for the treatment of metastatic colorectal cancer in patients typically make use of 5-FU and/or oxaliplatin (26). For this reason, we performed a first set of experiments in CEA-expressing CT26 cells, using these drugs alone or in combination with Sm3E-TNF (Supplementary Fig. S6). Surprisingly, the combination of 5-FU (50 mg/kg) with Sm3E-TNF resulted in a worse therapeutic performance than single-agent Sm3E-TNF (P < 0.0001; Fig. 5A). In contrast, the combination with oxaliplatin (7.5 mg/kg) strongly potentiated the action of Sm3E-TNF, leading to cancer cures in 40% of the animals (Fig. 5B). Similar therapy experiments were also performed in CEA-transfected C51 tumors, with the aim to investigate whether Sm3E-TNF (alone or in combination with oxaliplatin) would be active in this model. The fusion protein was able to potently inhibit tumor growth when used as monotherapy. Importantly, when Sm3E was combined with oxaliplatin, 80% of tumor-bearing mice were cured (Fig. 5C). Cured mice rejected a subsequent challenge with tumor cells at day 50. Because in our experience, the take rate of CEA-transfected CT26 and CEA-transfected C51 cells in immunocompetent mice was 100%, the observation of complete rejection of a second challenge with CEA-transfected CT26 or CEA-transfected C51 cells, respectively, indicates the onset of a protective immunity.

From a mechanistic viewpoint, the antibody-based delivery of TNF caused a rapid and selective hemorrhagic necrosis in both CEA-transfected CT26 and C51 tumors (Fig. 6A), similar to that previously described by our group using different TNF fusion proteins directed against components of the tumor extracellular matrix (9, 33). A microscopic analysis of H&E-stained sections revealed that the tumor architecture was intact after treatment with saline or oxaliplatin, but was largely necrotic in both Sm3E-TNF monotherapy groups and in the combination groups (Fig. 6B). The induction of hemorrhagic necrosis was less efficient when KSF-TNF was used, leaving groups of living cells within the tumor mass (Supplementary Fig. S7).

We also performed a time series of microscopic investigations of CEA-transfected CT26 tumors following treatment, to assess structural variations by H&E staining and by immunofluorescence analysis (Fig. 6). The tumor mass was almost completely necrotic 24 hours after Sm3E-TNF therapy. Proximal tissue sections were studied by immunofluorescence and areas corresponding to rectangles in the H&E sections. The most striking differences between saline, oxaliplatin, and Sm3E-TNF (monotherapy or combination) treatment regimens were related to the ability of targeted TNF to induce hemorrhagic necrosis and to promote an increased intratumoral density of natural killer (NK) cells and of T cells (Supplementary Table S1).

The AH1 retroviral antigen is frequently the most relevant rejection antigen in BALB/c-derived tumors (9, 31) and we hypothesized that AH1 might play a role in the tumor response. However, although we detected AH1-specific T cells in the CEA-transfected tumors using tetramer reagents (Supplementary Fig. S8), there was no apparent link to a treatment regime.

Sm3E-TNF is a novel fusion protein, capable of specific recognition of CEA and of selective homing to CEA-expressing colorectal tumors in vivo. The product was able to induce a potent tumor growth retardation, which was enhanced by combination with oxaliplatin. Mice that were cured as a result of the combination treatment, rejected subsequent challenges of the same tumor cells at day 50, indicating the onset of protective anticancer immunity.

From a pharmacodynamic viewpoint, the most visible effect of Sm3E-TNF on CEA-positive tumors was the rapid induction of hemorrhagic necrosis, leading to the death of large tumor portions after a single injection of the product. Living cells, however, typically survived at the tumor rim, with an enhanced, but patchy, infiltration by T cells and NK cells. Interestingly, in the therapy experiment performed on mice bearing CEA-transfected C51 tumors, KSF-TNF seemed to have a therapeutic effect. This may be explained by the fact that the C51 tumors grow faster than CT26, leading to a leaky neovasculature, which facilitates the extravasation of KSF-TNF. The therapeutic effect of KSF-TNF was less prominent in the CT26 model, which grows slowly and has a tighter vasculature (34, 35)

We have previously shown that the targeted delivery of TNF to colorectal tumors and to other cancer types in immunocompetent mouse models potently synergizes with other therapeutic modalities, including other cytokines (e.g., IL2; ref. 36 and IL12; ref. 37), cancer vaccines (31), and immune checkpoint inhibitors (38). In this article, we studied the combination of Sm3E-TNF with 5-FU and oxaliplatin, as these drugs are commonly used in chemotherapeutic strategies for the treatment of patients with metastatic colorectal cancer (39). Surprisingly, 5-FU worsened the therapeutic effect of Sm3E-TNF, while the combination with oxaliplatin improved anticancer activity.

Previous reports have shown that 5-FU could only provide a partial tumor growth retardation in the CT26 model (40) We could not find published examples on the combination of 5-FU with TNF-based pharmaceuticals. It is possible that 5-FU antagonizes the action of Sm3E-TNF by being toxic to lymphocytes (41).

Oxaliplatin has been shown not only to induce direct tumor cell death by apoptosis and secondary necrosis, but also to potentiate anticancer immunity through immunogenic cell death (42, 43). In a previous vaccination experiment, CT26 cells were treated in vitro with oxaliplatin for 24 hours and inoculated subcutaneously in immunocompetent BALB/c mice, which upon rechallenge rejected a new subcutaneous injection of living CT26 tumor cells. In a further experiment, it was shown that immunocompetent tumor-bearing mice treated with oxaliplatin responded better to the treatment than immunocompromised tumor-bearing mice, demonstrating that the anticancer effects were, at least, in part, mediated by the immune system (42, 44). CT26 and C51 tumors have previously been described as rather immunogenic tumors, by virtue of a higher proportion of infiltrating T cells and NK cells (45, 46). Tumors of BALB/c origin exhibit an unusually high infiltration by CD8⁺ T cells, which recognize the AH1 peptide, derived from the aberrantly expressed envelope protein of murine leukemia virus (31, 47). Recent studies have indicated that noncoding regions (including retroviral antigen sequences) may represent the main source of tumor rejection antigens (48).

Interestingly, CEA-expressing cell lines are not rejected by the mouse immune system after subcutaneous injection. This immunologic tolerance has also been observed in previous studies, where antigens of human origin were transfected in murine cell lines that have been subsequently used for tumor implantation in syngeneic immunocompetent mice (49, 50). In one of these studies, the ex vivo analysis of CEA-transfected tumor-bearing mice showed the presence of anti-CEA antibody in the blood; however, the tumors were still able to grow and the role of these antibodies in the inhibition of the tumor growth was not identified (50). One of the reasons for which the murine immune system fails to clear a potentially immunogenic human antigen–transfected cell line might be due to the tumor immunosuppressive microenvironment. This unfavorable microenvironment could prevent the host immune system from rejecting the tumor, for instance, by tumor-induced impairment of antigen presentation, where the immunosuppressive microenvironment causes the antigen-presenting cells to present tumor antigens that lead to the induction of T-cell tolerance (51).

Various immunologic approaches have been attempted for the treatment of metastatic colorectal cancer, but unfortunately clinical success has been limited. A bispecific antibody, directed against CEA and CD3, showed promising results in preclinical models, but did not mediate objective responses in patients with colorectal cancer (52). Similarly, anti-CEA antibodies have been used to deliver a human IL2 mutant to colon tumors (53). The targeted delivery of TNF to CEA-positive lesions may represent an alternative and potentially more efficacious strategy, because of the rapid induction of tumor cell death.

The debulking of the tumor mass triggered by Sm3E-TNF in mouse models was impressive, but it is still unclear whether similar effects could be observed in patients. Isolated limb perfusion (ILP) of patients with an advanced melanoma and soft-tissue sarcoma of the limbs, who were candidates for amputation, using TNF in combination with melphalan and mild hyperthermia, typically induces a hemorrhagic necrosis of the neoplastic mass (54–56), but TNF is administered at doses which would not be compatible with systemic administration. The targeted delivery of TNF to the tumor site, mediated by fusion with the L19 antibody (specific to the alternatively spliced EDB domain of fibronectin), has recently been shown to induce complete responses and rapid hemorrhagic necrosis in the ILP setting, at doses which are more than 10-times lower than those conventionally used with recombinant human TNF (56). There is an urgent need to perform similar studies in patients with disseminated disease, following intravenous administration of antibody-TNF fusions. Perfusion MRI (57) or ¹⁸F-FDG-PET (58) procedures may be considered as imaging modalities to visualize vascular shutdown of the tumor, immediately after the administration of targeted TNF therapeutics.

In the preclinical studies presented, the humanized Sm3E antibody was fused to murine TNF because the human cytokine homologue is only partially reactive with murine TNF receptors (59). It should be straightforward, however, to fuse Sm3E with human TNF, in full analogy to that described previously for L19-TNF (56, 60). A human Sm3E-TNF fusion protein might be expected to effectively deliver TNF to target, as CEA-specific antibodies have been extensively validated for their ability to selectively localize to metastatic colorectal cancer lesions using nuclear medicine procedures (19, 20, 61, 62). Our results indicate that a fully human Sm3E-TNF product could have potential to help debulk cancer lesions, and facilitate the management of patients with metastatic disease, in combination with oxaliplatin.

K. Chester reports a patent for US7232888B2 issued. D. Neri reports personal fees from Philogen (board member and shareholder) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

D. Bajic: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. K. Chester: Writing-original draft. D. Neri: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing.

D. Neri and D. Bajic gratefully acknowledge financial support from the ETH Zürich, the Swiss National Science Foundation (grant number, 310030B_163479/1), and the European Research Council (under the European Union's Horizon 2020 research and innovation program, grant agreement, 670603). We would like to thank Drs. Roberto De Luca (Philochem AG) and Mattia Matasci (Philochem AG) for their help with experimental procedures.

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.