Antibody–cytokine fusion proteins can have the potential to increase the density and activity of subsets of leukocytes within the tumor mass. Here, we describe the design, production, and characterization of four novel antibody–cytokine fusion proteins directed against human carbonic anhydrase IX, a highly validated marker of hypoxia that is overexpressed in clear cell renal cell carcinoma and other malignancies. As immunomodulatory payloads we used TNF, IL2, IFNα2 (corresponding to products that are in clinical use), and IL12 (as this cytokine potently activates T cells and NK cells). Therapy experiments were performed in BALB/c mice, bearing CT26 tumors transfected with human carbonic anhydrase IX, in order to assess the performance of the fusion proteins in an immunocompetent setting. The biopharmaceuticals featuring TNF, IL2, or IL12 as payloads cured all mice in their therapy groups, whereas only a subset of mice was cured by the antibody-based delivery of IFNα2. Although the antibody fusion with TNF mediated a rapid hemorrhagic necrosis of the tumor mass, a slower regression of the neoplastic lesions (which continued after the last injection) was observed with the other fusion proteins, and treated mice acquired protective anticancer immunity. A high proportion of tumor-infiltrating CD8+ T cells was specific to the retroviral antigen AH1; however, the LGPGREYRAL peptide derived from human carbonic anhydrase IX was also present on tumor cells. The results described herein provide a rationale for the clinical use of fully human antibody–cytokine fusions specific to carbonic anhydrase IX.

Cytokines are immunomodulatory proteins, which play a central role in orchestrating immune responses by interaction with cognate cytokine receptors (1). Some recombinant human cytokines have gained marketing authorization for administration to patients with cancer. Granulocyte colony-stimulating factor and granulocyte-macrophage colony-stimulating factor are mainly used to counteract chemotherapy-induced neutropenia (2). IL2 has received marketing authorization for the treatment of patients with advanced melanoma or renal cell carcinoma, because a small proportion of patients treated with this product may enjoy a long-lasting complete remission (3, 4). IFNα2 has been approved not only for the treatment of certain viral conditions, but also for the treatment of patients with hairy cell leukemia, chronic myelogenous leukemia, multiple myeloma, and malignant melanoma (5). Finally, recombinant human TNF has received marketing authorization in Europe for the isolated limb perfusion of patients with inoperable sarcomas (6). The product is also used for the treatment of patients with inoperable in-transit metastases of melanoma (7). Unlike human IL2, IFNα2, and TNF, human IL12 has never received a marketing authorization.

Although cytokines may provide a benefit to patients with cancer, their clinical use is often hindered by severe toxicities, which can be manifested also at low (e.g., sub-milligram) doses (8). For this reason, several companies and research groups are currently involved in research activities, aimed at generating engineered cytokine products with improved therapeutic index. For example, in the last few months, PEGylated derivatives of IL2 and of IL10 have been the object of very large commercial agreements, involving Nektar/Bristol-Myers Squibb (9) and Eli Lilly/ARMO (10), respectively.

An alternative avenue for the generation of superior cytokine-based therapeutics consists in the fusion of cytokine payloads with antibodies, capable of selective localization at the site of disease (11). In mice, it has been clearly shown that the antibody-based targeted delivery of certain payloads (e.g., IL2, IL4, IL12, TNF) to neoplastic lesions leads to a substantial potentiation of the therapeutic activity, compared with the nontargeted murine cytokine counterpart (12, 13). A number of fully-human antibody–cytokine fusions have progressed to clinical trials (11).

Our group has extensively worked on the generation and in vivo testing of antibody–cytokine fusions, directed against components of the modified tumor extracellular matrix (e.g., splice variants of fibronectin and of tenascin-C; ref. 11). Other groups have focused on the antibody-based delivery of cytokine payloads to antigens expressed on the surface of tumor cells (14–17).

In this work, we focused on the development of novel antibody–cytokine fusions, directed against human carbonic anhydrase IX (CAIX). This homodimeric enzyme is expressed on the surface of the majority of clear cell carcinoma cells, but is otherwise undetectable in most normal adult tissues, exception made for certain gastrointestinal structures (18–20). CAIX is also strongly upregulated at sites of hypoxia and in certain other malignancies, including a proportion of patients with urothelial, colorectal, lung, stomach, pancreas, breast, head and neck, ovaries, brain, and cervix cancer (21). CAIX is one of the best validated targets for antibody-based pharmacodelivery applications. Radiolabeled preparations of the chimeric cG250 antibody have been administered to more than 200 patients with renal cell carcinoma, with excellent imaging results (22). Interestingly, the majority of clear cell renal cell carcinoma lesions could efficiently be reached in vivo by the antibody, whereas CAIX-positive structures in the stomach and in the intestine were minimally targeted by the antibody, presumably as a consequence of differences in vascular permeability between tumors and normal organs (22).

We have previously described three high-affinity fully-human monoclonal antibodies directed against human CAIX, called A3, CC7 (23), and XE114 (24). In this work, we have fused antibody fragments derived from XE114 with human IL2, murine TNF, murine IFNα2, and murine IL12. We used human IL2 because this payload is potently active in mice (25), whereas we had to use murine cytokines in the other cases in order to ensure reactivity with the cognate murine receptors (12, 26, 27). We chose to use antibody fragments rather than intact immunoglobulins as fusion partners, in order to achieve a selective localization at the tumor site and a rapid clearance from circulation. All four fusion proteins (termed XE-mTNF, XE-hIL2, XE-mIFNα2, and XE-mIL12) were potently active against CT26 tumors, which had been transfected with human CAIX, in immunocompetent mice. Although XE-mTNF mediated a rapid hemorrhagic necrosis of the tumor mass, the other fusions caused a slower regression of the tumor mass, which continued after the last injection. The antitumor activity of the fusion proteins was accompanied by an increased infiltration of CD8+ T cells into the cancer lesions, which were specific to AH1, a peptide derived from the envelope protein of a retrovirus (murine leukemia virus) which is endogenous in the mouse genome (28, 29).

Cloning of immunocytokines

The antibody IgG(XE114) was described previously (30). For cloning of XE-mTNF in scFv format, a 42 bp linker was inserted between the DNA sequences of VH and VL of the XE114 antibody. The 5′ end of the scFv(XE) was linked to the DNA sequence of a mammalian signal peptide for transient gene expression in CHO-S cells and the 3′ end was linked to the DNA sequence of mTNF by a 45 bp linker.

For cloning of XE-hIL2 and XE-mIFNα2 in diabody format, VH and VL DNA sequences of the XE114 antibody were connected to each other with a 15 bp linker. The 5′ end of the XE diabody was linked to the DNA sequence of a mammalian signal peptide for transient gene expression in CHO-S cells and the 3′ end was fused to a 45 bp linker to add the DNA sequence of hIL2 or to a 27 bp linker to add the DNA sequence of mIFNα2.

For the forced diabody format of XE-mIL12, 2 diabodies were connected with a 45 bp linker. IL12p35 and IL12p40 DNA sequences were fused to the 5′ end using linkers of 16 bp and 45 bp, respectively. Further, the DNA sequence of a mammalian signal peptide was added to the 5′ end for transient gene expression in CHO-S cells. All cloned fragments were double digested using the restriction enzymes NheI and NotI and ligated into pcDNA3.1(+) previously digested with the same restriction enzymes.

Cell culture

CHO-S cells were cultured at 37°C in Power-CHO medium (Lonza) supplemented with 8 mol/L Ultraglutamin (Gibco), 4 mol/L HT (Gibco), and 1% antibiotic–antimycotic (Gibco) and kept between 0.5 and 6.0 × 106 cells per mL on a shaking incubator (170 rpm).

CT26.wt cells were previously transfected with a plasmid encoding human CAIX, leading to the monoclonal cell line CT26.3E10 expressing human CAIX. Quantification by FACS analysis has revealed that 71,900 copies of the CAIX protein are expressed on one CT26.3E10 cell (30). CT26.wt cells and CT26.3E10 cells were cultured in RPMI (Gibco) supplemented with 10% FBS (Gibco) and 1% antibiotic–antimycotic (Gibco) at 37°C and 5% CO2. Cells were trypsinized, passaged at a confluency of 80%, and kept in culture for no longer than 15 passages. The wild-type CT26 cell line was obtained from ATCC in 2015. Authentication of the cell line including check of post-freeze viability, growth properties and morphology, test for mycoplasma contamination, isoenzyme assay, and sterility test were performed by the cell bank prior to shipment.

Transient gene expression

For transient gene expression, CHO-S cells were diluted to a concentration of 4 × 106 per mL in Pro-CHO medium (Lonza) supplemented with 8 mol/L Ultraglutamin (Gibco), 4 mol/L HT (Gibco), and 1% antibiotic–antimycotic (Gibco). A total of 106 CHO-S cells were mixed with 0.9 μg of the respective plasmid and 2.5 μg of PEI and left 6 days for protein production at 31°C and 150 rpm. The protein was purified from the supernatant by protein A affinity chromatography, eluted with TEA or Glycine depending on the proteins pI and subsequently dialyzed into PBS pH 7.4 (XE-mTNF, XE-mIFNα2, and XE-mIL12) or 0.01 M NaHCO3 pH 8.0 (XE-hIL2). For quality control, the proteins were analyzed by SDS-PAGE under reducing and nonreducing conditions, by size exclusion chromatography using a Superdex 200 Increase 10/300GL column (GE Healthcare), by mass spectrometry, and by surface plasmon resonance analysis using a human CAIX coated CM5 sensor chip (GE Healthcare).

Activity assays

Bioactivity of the immunocytokines was evaluated using WEHI-164 cells and CTLL-2 cells, as previously described (26, 27, 31). In brief, for activity testing of XE-mTNF, 30,000 WEHI-164 cells per well were seeded in RPMI medium containing 10% FBS, 1% antibiotic–antimycotic, and 2 μg/mL actinomycin D. XE-mTNF was added in a serial dilution and incubated for 24 hours at 37°C. For activity testing of XE-mIFNα2, 5,000 WEHI-164 cells per well were seeded in RPMI medium containing 10% FBS and 1% antibiotic–antimycotic and a serial dilution of XE-mIFNα2 was added for 72 hours at 37°C. CTLL-2 cells were incubated without IL2 before seeding 20,000 CTLL-2 cells per well. A serial dilution of XE-hIL2 in RPMI medium containing 10% FBS, 1% antibiotic–antimycotic, 1 mmol/L sodium pyruvate, and 2 mmol/L l-glutamine was added and incubated for 48 hours at 37°C. After the incubation period, killing of WEHI-164 cells and proliferation of CTLL-2 was evaluated using cell titer aqueous one solution (Promega) and cell viability was expressed as percentage of cells treated compared with untreated controls.

FITC labeling

For FITC labelling of antibodies, IgG(XE114) and IgG(KSF) were dialyzed overnight into 0.1 mol/L NaHCO3 pH 9. Two mg per mL of protein were mixed with 50 μg of fluorescein isothiocyanate (Sigma) freshly dissolved in 50 μL anhydrase DMSO (Sigma). After 8 hours of incubation at 4°C, unbound FITC was separated using a PD-1 column (GE Healthcare) and aliquots of the labelled protein were collected in PBS pH 7.4.

Animal experiments

For tumor growth, a cell suspension of 5 × 106 CT26.3E10 cells or 5 × 106 CT26.wt cells in 150 μL HBSS (Gibco) was subcutaneously injected into the right flank of 9 weeks old female BALB/c mice (Charles River). Body weight and tumor growth were monitored daily until termination criteria were reached. In case of immunofluorescence or FACS studies, mice were sacrificed at earlier time points as described below. Tumor size was calculated as width × width × length × 0.5 (mm3).

The therapy study was started when tumors reached an average size of 90 mm3, mice were randomized into groups of 5 and received three intravenous injections of either 150 μL PBS as control, 3 μg XE-mTNF, 50 μg XE-hIL2, or 150 μg XE-mIFNα2 every 48 hours. Twelve micrograms of XE-mIL12 were injected every 72 hours. For tumor rechallenge, cured mice received a subcutaneous injection of 5 × 106 CT26.3E10 cells in 150 μL HBSS.

For ex vivo immunofluorescence studies, mice with an average tumor size of 300 mm3 received one intravenous injection of 150 μg FITC-labeled IgG(XE114), IgG(KSF), or XE-mIFNα2, 50 μg of XE-hIL2 or 3 μg of XE-mTNF. After 24 hours, tumor and organs were collected and frozen in cryoembedding medium.

For immunofluorescence and FACS studies, mice bearing CAIX-positive CT26 tumors with an average size of 100 mm3 received either one injection of PBS or of XE-mTNF (3 μg) or two injections of XE-hIL2 (50 μg) or XE-mIFNα2 (150 μg). Tumors and lymph nodes were excised 24 hours after the last injection.

Immunofluorescence staining

Cryosections of 10 μm were made of tumors and organs previously frozen in cryoembedding medium. Sections were fixed in ice-cold acetone, blocked with 20% FBS, in 3% BSA, and stained for CD4+ T cells (GK1.5; BioLegend), CD8+ T cells (53–6.7; BioLegend), activated caspase 3 (C8487; Sigma), and CD31+ endothelial cells (AF3628; R&D systems). Secondary antibodies used were goat-anti-rabbit-AF488, donkey-anti-rat-AF488, and donkey-anti-goat-AF594 (Thermo Fisher Scientific). For ex vivo immunofluorescence studies, FITC-labeled IgG was detected using rabbit-anti-FITC (Bio-Rad) and donkey-anti-rabbit-AF488 (Thermo Fisher Scientific) as secondary antibody. XE-hIL2, XE-mIFNα2, and XE-TNF were detected using anti-hIL2 (MQ1-17H12; eBioscience), anti-mIFNα2 (50525-T08; SinoBiological), and anti-mTNF (MP6-XT22; Invitrogen) antibodies, respectively. A rabbit-anti-rat antibody (ab102248; Abcam) was used in case of anti-hIL2 and anti-mTNF staining. Detection was performed using the Alexa Fluor-488 Tyramide SuperBoost Kit with goat-anti-rabbit-IgG (B40943; Thermo Fisher Scientific) according to the manufacturers' instruction. Nuclei were counterstained with DAPI (Thermo Fisher Scientific). Tumor sections were mounted with fluorescence mounting medium (Dako) and analyzed using an Axioscop Mot Plus Microscope (Zeiss).

FACS experiments

For detection of AH1-positive CD8+ T cells, tumor and lymph nodes were excised 24 hours after treatment of mice as described above. Tumor samples were digested for 2 hours at 37°C with 1 mg/mL collagenase II and 0.1 mg/mL DNase I in RPMI medium supplemented with 1% antibiotic–antimycotic. The filtered tumor cell suspension was diluted to 5 × 106 cells/mL. Lymph nodes were collected in sterile PBS and mechanically opened. The cell suspension was washed with Red Blood Cell Lysis buffer and isolated lymphocytes were diluted to 1 × 106 cells/mL. One hundred microliters of tumor cell or lymphocyte suspension were stained for 1 hour at 4°C with a PE-labeled antibody against CD90.2 (53–2.1; Biolegend), a FITC-labeled antibody against CD8+ T cells (53–6.7; Biolegend), a APC labeled H-2Ld tetramer loaded with AH1 peptide (28) and a live/dead fixable Near-IR Dead Cell Stain (Thermo Fisher Scientific). Following fixation and permeabilization, cells were analyzed on a CytoFLEX cytometer (Beckmann Coulter) and data were processed using FlowJo (v.10; TreeStar).

MHC binding prediction of human CAIX

The major histocompatibility complex (MHC) class I binding prediction analysis was performed to identify peptides with a length between 9 and 12 amino acids for the 3 BALB/c H-2 alleles Dd, Kd, and Ld following a procedure adapted from Fugmann and colleagues (32). In brief, a peptide binding score was calculated based on the positional-specific scoring matrix of Dd, Kd, and Ld derived from the investigation of the BALB/c MHC peptidome (28, 33). The score cut-off predictive of peptide binding was derived from a ROC curve analysis plotting the score distribution of eluted peptides (true positive dataset) versus one million random peptide sequences (true negative dataset) computed from the murine reference proteome, where the length distribution of in silico generated peptides followed the eluted peptide sequences. The score discriminating best between true positive and true negative dataset that defined peptide binding was found to be 4.27 for Dd, 4.25 for Kd, and 4.49 for Ld. Scores for peptides longer than 9 amino acids were computed by maximizing the score of 9 amino acid long sequence variants derived from the longer peptide considering three possibilities: (i) N-terminally elongated peptides, (ii) C-terminally elongated peptides, and (iii) peptides with one internal loop. Loops were allowed between 2 amino acid positions if at least one of them was not an anchor position or a terminal position. An anchor position was defined for an amino acid position in the positional-specific scoring matrix if at this position at least one amino acid had a frequency of more than 20%. Allowing only for one loop limited the number of stochastic possibilities and therefore also the probability to find random high-scoring sequence variants.

Peptidome analysis

The peptidome analysis of CT26 cells stably transfected with human CAIX was performed essentially as described previously (28). In brief, cells were lysed at a density of 3.0 × 107 cells/mL (34) followed by purification of MHC class I complexes with M1/42 antibody (BioXCell) coupled resin. Analysis of MHC class I peptides by liquid chromatography coupled to the Q Exactive (Thermo Fisher Scientific) mass spectrometer was performed as described (34). Resulting spectra were processed and analyzed using the Proteome Discoverer software (Thermo Fisher Scientific, Version 1.4.1.14) considering all murine proteins (89′527 entries) of the UniProt database, downloaded on the March 22, 2018, spiked with the sequence of transfected human CAIX. The following analysis settings were used with SEQUEST: (i) no-enzyme (unspecific), (ii) precursor mass tolerance 4 ppm, (iii) fragment mass tolerance 0.02 Da, (iv) 1 variable modification (oxidation of methionine), (v) peptide length 6 to 50 amino acids. The identity of LGPGREYRAL was confirmed with a synthetic peptide ordered from JPT.

Statistical analysis

Data are represented as mean ± SEM and were analyzed using Prism 7 (GraphPad Software, Inc.). A two-way ANOVA test with a P-value <0.05 was performed for analysis of statistical significance towards the PBS control group (*P < 0.05, **P < 0.005, ***P < 0.0005, ****P < 0.0001).

We studied the tumor-targeting properties of the XE114 antibody in fully human IgG format, injecting it at high dose (150 μg) into immunocompetent BALB/c mice, bearing CT26 tumors transfected with human CAIX (30). An ex vivo immunofluorescence analysis performed 24 hours after intravenous administration revealed a preferential uptake of the antibody within the solid tumor mass, whereas no accumulation could be detected in normal tissues (stomach, intestine, kidney, liver, lung, heart, spleen, and muscle), using identical image acquisition conditions (Fig. 1). By contrast, an antibody of irrelevant specificity in the mouse (KSF, directed against hen egg lysozyme) did not exhibit a preferential uptake in CAIX-transfected CT26 tumors. Both IgG(XE114) and IgG(KSF) failed to localize to CT26 tumors that had not been transfected with CAIX, confirming the antigen specificity of the XE114 antibody (Supplementary Fig. S1).

We then fused the XE114 antibody with murine TNF, human IL2, murine IFNα2, and murine IL12, using different molecular strategies (Fig. 2; Supplementary Fig. S2). The homotrimeric TNF payload was fused to the antibody in scFv format, similar to previous TNF-fusions based on other antibodies (26), in order to prevent product polymerization. By contrast, the human IL2 and murine IFNα2 payloads were fused at the C-terminal end of the antibody in diabody format, leading to noncovalent homodimers (Fig. 2; refs. 27, 31). Murine IL12 was fused to the N-terminus of a single-chain tandem diabody of the XE antibody, as previously described for other fusion proteins of IL12 (Supplementary Fig. S2; ref. 12). Products were purified to homogeneity using protein A chromatography, as the XE114 antibody featured a heavy chain variable domain, derived from a germline gene that enables protein A binding (24, 35, 36). The SDS-PAGE analysis, size-exclusion profiles, BIAcore binding properties to the antigen, mass spectrometric characterization, and cytokine activity data are presented in Fig. 2 and Supplementary Fig. S2. The fusion proteins also strongly stained sections of CAIX-transfected CT26 tumors and XE-mIFNα2, XE-hIL2, and XE-mTNF localized to the tumor in vivo. Although XE-mIFNα2 showed a uniform staining, XE-hIL2 and XE-mTNF exhibited a patchy staining pattern, possibly reflecting extravasation from permeable blood vessels (Supplementary Fig. S3). The DNA and amino acid sequences of the fusion proteins are described in Supplementary Fig. S4.

Immunocompetent BALB/c mice, bearing subcutaneously-grafted human CAIX-positive CT26 tumors, received intravenous injections of XE-mTNF, XE-hIL2, and XE-mIFNα2 at day 7, 9, and 11, and of XE-mIL12 at day 10, 13, and 16, starting the therapy when the neoplastic masses had reached a volume of ∼90 mm3 (Fig. 3A; Supplementary Fig. S2). Similar to what had been previously described for fusion proteins based on TNF, IL2, IFNα2, and IL12, the products were administered at doses of 3 μg (26), 50 μg (31), 150 μg (27), and 12 μg (12), respectively. All mice in the XE-mTNF, XE-hIL2, and XE-mIL12 treatment groups were cured, while lasting complete remissions were only observed in 3 of 5 mice treated with XE-mIFNα2. All treatments were very well tolerated, even though mice in the XE-mTNF group exhibited a transient loss of 3% body weight (Fig. 3b). The therapeutic performance obtained by XE114 antibody-based delivery of mTNF, hIL2, and mIFNα2 was better than the one observed with similar fusion proteins, based on the KSF antibody, specific to hen egg lysozyme (an antigen of irrelevant specificity in the mouse; Supplementary Fig. S5A–F). Fig. 3C shows a photographic documentation of the tumor lesions at various time points. Although mice in the saline (PBS) treatment group exhibited progressively growing tumors, a rapid scab formation was visible as a result of XE-mTNF treatment, consistent with what had previously been reported for other antibody–TNF fusions. Importantly, scab formation could not be observed with the negative control KSF-mTNF fusion protein (Supplementary Fig. S5G; refs. 28, 37). XE-hIL2 and XE-mIFNα2 caused a slower regression of the tumor mass, which continued after the last injection. Interestingly, cured mice rejected subsequent challenges with CAIX-transfected tumor cells, thus providing evidence for the onset of protective immunity (Fig. 4A).

In parallel to the therapy study, additional mice which had received injections of the fusion proteins were sacrificed at different time points (Fig. 4B), in order to allow a microscopic analysis of tumor sections. Mice treated with saline showed no detectable staining for caspase 3, consistent with a low level of apoptosis in those tumors (Fig. 4C). By contrast, the majority of the tumor mass in the XE-mTNF treatment group exhibited a large area of cell death, with a strong and diffuse caspase 3 staining. XE-hIL2 and XE-mIFNα2 treated tumors showed the onset of patches with caspase 3-positive cells, mainly around CD31-positive blood vessels, both after the first and the second injection (Fig. 4C). For all three fusions, the increase in tumor cell death was paralleled by an increased density of CD4+ and CD8+ T lymphocytes (Fig. 4D and E).

We used tetramer technology in order to characterize the proportion of CD8+ T cells specific to the AH1 peptide (SPSYVYHQF), as this retroviral antigen has previously been shown to account for the immunogenicity of tumors in BALB/c mice (28). The percent of AH1-specific T cells was increased in the tumor bed, compared with tumor-draining lymph nodes (Fig. 5). The percentage of AH1-positive cells in the treatment groups was comparable to the one observed in the saline group, with exception made for the XE-hIL2 treatment group (Fig. 5). Because in other studies performed in our group with BALB/c derived tumors, the intralesional proportion of AH1-positive cells could exceed 50% of the total CD8+ T cells, we hypothesized that human CAIX-derived peptides could contribute to the immunogenicity of the stably-transfected tumor cells. This hypothesis was supported by a sequence alignment analysis, showing a moderate overall identity of 69.7% between human and murine CAIX. To clarify this issue, we studied MHC class I complexes on transfected CT26 tumor cells, using in silico predictions of peptide binding to the cognate MHC molecules. Several peptides derived from human CAIX were predicted to bind to the BALB/c MHC class I alleles H-2Kd, H-2Dd, and H-2Ld (Supplementary Fig. S6). A subsequent mass spectrometric characterization of peptides eluted from MHC class I complexes purified from CT26 cell lysates (34, 38) allowed the confident identification of a human CAIX-derived peptide (LGPGREYRAL), which was confirmed with a synthetic peptide (Fig. 6). This sequence was predicted to bind to the MHC class I H2-Dd allele and contained a single Q to R substitution compared to its murine counterpart (LGPGQEYRAL). None of the other 23 predicted binders derived from human CAIX could be identified by MS-assisted peptidome analysis. By contrast, the AH1 peptide could be identified from the MHC peptidome of CAIX-transfected CT26 cells, with an abundance that was comparable to the one observed in wild-type CT26 cells, as described previously (33).

We have described the production and potent anticancer activity of four novel fusion proteins, consisting of the XE114 antibody fused to different immunomodulatory payloads (human IL2, murine TNF, murine IFNα2, and murine IL12). For the preparation of these biopharmaceuticals, we used immunocytokine formats, which our group had successfully used with antibodies directed against splice variants of fibronectin (12, 26, 27, 31). Specifically, cytokine delivery to the EDA and EDB domain of fibronectin had tumor growth inhibitory activity in various immunocompetent models of cancer, but typically did not eradicate the disease when used as single agent (39). Similar cytokine-fusions, specific to hen egg lysozyme and used as negative control (12, 13, 27), had a substantially reduced activity, confirming the need for a targeted delivery of cytokine payloads to the tumor environment.

We decided to focus on CAIX, a tumor-associated antigen expressed on the surface of tumor cells, because this target has been extensively characterized using Nuclear Medicine procedures (22). In the REDECT trial, 195 patients with renal masses scheduled for resection received 124I-Girentuximab followed by PET/CT. Clear cell renal cell carcinoma was detected with a specificity of 85.9% and sensitivity of 86.2%, thus confirming the accuracy of the anti-CAIX antibody (22). Furthermore, these results are in line with the tumor-specific uptake of XE114 in CAIX transfected CT26 tumor bearing mice (Fig. 1).

IL2, IFNα2, TNF, and IL12 were investigated as payloads for fusions with the anti-CAIX antibody XE-114. High-dose IL2 has received marketing authorization for the treatment of patients of metastatic renal cell carcinoma and is still used for this indication, because a proportion of patients enjoys complete durable responses. However, the treatment is quite toxic and is often reserved only for younger patients, as severe capillary-leak syndrome may limit applicability with the high-dose regimen (40). In view of these considerations, targeted variants of IL2 may be attractive for the treatment of patients with metastatic kidney cancer. The rationale for the targeted delivery of IFNα2 is less strong. Treatment of patients with metastatic renal cell carcinoma using recombinant human IFNα led to a reduction in mortality after 1 year, but complete responses were only rarely observed. Severe adverse effects were more frequent in the IFNα2 group than in control groups (41). Systemic administration of recombinant human TNF was studied in a phase I clinical trial in 33 patients with different types of malignancies. The only complete response was observed in a patient with renal cell carcinoma, providing anecdotal evidence for a possible biological activity (42). Moreover, the targeted delivery of human TNF using fusions to Girentuximab in full IgG format was studied in a xenograft model of renal cell carcinoma and was found to synergize with IFNγ (43). Collectively, these data provide a rationale for the implementation of CAIX-targeted pharmaco-delivery applications, which have a potential to work in humans.

We had to use mouse tumor cells transfected with the human CAIX antigen in immunocompetent mice, because the XE114 antibody does not cross-react with murine CAIX. The system could, therefore, be more immunogenic than other murine tumors, due to the presence of a human protein, although the tumors were growing and staining with XE114 antibody demonstrated a homogeneous expression of the antigen on tumor cells (Fig. 1). A FACS analysis of tumor-infiltrating CD8+ T cells indicates that a large proportion of infiltrating lymphocytes are still specific to the AH1 peptide (Fig. 5). However, it is likely that a subset of CD4+ and CD8+ T cells may recognize peptides derived from human CAIX. Indeed, an immunopeptidome characterization showed that a peptide derived from human CAIX (LGPGREYRAL) strongly binds to the MHC class I H-2Dd allele and features a Q to R substitution compared with its murine counterpart (LGPGQEYRAL). Although both of the sequences are predicted to bind to H2-Dd with comparable affinity, it is possible that this neo-epitope may facilitate recognition events by cognate CD8+ T cells, because the mutated residue should project towards the T-cell receptor (44).

The four fusion proteins had different mechanisms of action. On one hand, XE-mTNF rapidly induced hemorrhagic necrosis of the tumor mass, similar to what has previously been discussed for L19-TNF and F8-TNF (28, 37, 45). By contrast, the other cytokine payloads led to a slower but progressive tumor killing. Although the hemorrhagic necrosis induced by TNF was accompanied by recognition of AH1 (28), the low abundance of AH1 specific T cells in the tumor microenvironment after XE-hIL2 administration may suggest that IL2 requires a different mode of action. Interestingly, although IL2 and IL12 had previously shown potent therapeutic activity when fused to tumor-homing antibodies, IFNα2 had failed to induce a potent anticancer activity when fused to the F8 or L19 antibodies, specific to the alternatively-spliced EDA and EDB domain of fibronectin (Supplementary Fig. S7; ref. 27). At this moment in time, it is not clear why the targeted delivery of murine IFNα2 in CAIX-transfected tumors was more effective, leading to durable complete responses in 3 of 5 treated mice. Morrison and colleagues have previously described fusion proteins, consisting of intact antibodies against CD138 and IFNα2 as payload. The authors observed inhibition of tumor growth in a murine xenograft model of multiple myeloma with enhanced survival when mice received six instead of three treatment cycles with anti-CD138-IFNα2 (46). Similarly, scientists at TEVA have recently described potent anticancer activity with antibodies fused to IFNα mutants (Attenukines; ref. 14). It is thus possible that the targeted delivery of IFNα moieties may be particularly attractive in the context of antigens, located on the tumor cell surface, rather than for extracellular matrix components.

The IgG(XE-114) antibody exhibited a homogeneous targeting of tumor cells in the CT26 model transfected with human CAIX, whereas a perivascular accumulation of the same product was observed in SKRC-52, possibly due to differences in vascular permeability between the two models (24). Also, XE-hIL2 and XE-mTNF preferentially localized to perivascular tumor cells, although a more homogeneous distribution within the tumor mass was observed for XE-mIFNα2 (Supplementary Fig. S3B). Although formats of smaller size may have an improved tissue penetration, their increased clearance from the blood circulation may prevent a deeper penetration into less vascularized tumor areas. Slow infusion procedures may lead to a more homogenous product distribution profile within the neoplastic mass (47).

The article provides a strong rationale for the clinical development of XE-IL2 and for fully-human counterparts of XE-TNF, XE-IFNα2, or XE-IL12. Various antibody–cytokine fusions are currently being tested in clinical trials (11), some of which have the potential to synergize with chemotherapeutic agents (NCT03420014), external beam radiation (NCT02086721), and immunomodulatory drugs (48). The results of this article may also stimulate mechanistic investigations on how cytokines promote their biocidal activity. In most studies reported so far, CD8+ T cells and NK cells were found to play a crucial role, as evidenced by lymphocyte depletion experiments (28, 49). Although a higher density of lymphocytes was observed in mice treated with CAIX-targeting immunocytokines, it is still not clear whether the absolute number of leukocytes is sufficient to explain the ability to induce tumor regression and tumor protective immunity.

D. Neri is a board member at and has ownership interest (including stock, patents, etc.) in Philogen. No potential conflicts of interest were disclosed by the other authors.

All mice experiments were performed under the licenses ZH27/2015 or ZH04/2018 granted by the veterinary office of the Canton Zurich to Prof. Dario Neri.

Conception and design: B. Ziffels, D. Neri

Development of methodology: B. Ziffels, M. Stringhini, D. Neri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Ziffels, P. Probst, T. Sturm, D. Neri

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Ziffels, M. Stringhini, T. Fugmann, T. Sturm, D. Neri

Writing, review, and/or revision of the manuscript: B. Ziffels, D. Neri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Ziffels

Study supervision: D. Neri

The authors gratefully acknowledge funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (grant agreement 670603) and the Swiss National Science Foundation (grant agreement 310030B_163479/1).

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.

1.
Murphy
K
,
Weaver
C
.
Janeway's immunobiology
. 9th ed.
New York, NY:
Garland Science/Taylor & Francis Group
,
LLC
; 
2017
. p.
904
.
2.
Mehta
HM
,
Malandra
M
,
Corey
SJ
. 
G-CSF and GM-CSF in neutropenia
.
J Immunol
2015
;
195
:
1341
9
.
3.
Weide
B
,
Eigentler
TK
,
Pflugfelder
A
,
Leiter
U
,
Meier
F
,
Bauer
J
, et al
Survival after intratumoral interleukin-2 treatment of 72 melanoma patients and response upon the first chemotherapy during follow-up
.
Cancer Immunol Immunother
2011
;
60
:
487
93
.
4.
Chow
S
,
Galvis
V
,
Pillai
M
,
Leach
R
,
Keene
E
,
Spencer-Shaw
A
, et al
High-dose interleukin2 - a 10-year single-site experience in the treatment of metastatic renal cell carcinoma: careful selection of patients gives an excellent outcome
.
J Immunother Cancer
2016
;
4
:
67
.
5.
Paul
F
,
Pellegrini
S
,
Uze
G
. 
IFNA2: the prototypic human alpha interferon
.
Gene
2015
;
567
:
132
7
.
6.
Verhoef
C
,
de Wilt
JH
,
Grunhagen
DJ
,
van Geel
AN
,
ten Hagen
TL
,
Eggermont
AM
. 
Isolated limb perfusion with melphalan and TNF-alpha in the treatment of extremity sarcoma
.
Curr Treat Options Oncol
2007
;
8
:
417
27
.
7.
Testori
A
,
Ribero
S
,
Bataille
V
. 
Diagnosis and treatment of in-transit melanoma metastases
.
Eur J Surg Oncol
2017
;
43
:
544
60
.
8.
Baldo
BA
. 
Side effects of cytokines approved for therapy
.
Drug Saf
2014
;
37
:
921
43
.
9.
Garber
K
. 
Cytokine resurrection: engineered IL-2 ramps up immuno-oncology responses
.
Nat Biotechnol
2018
;
36
:
378
9
.
10.
Lilly
E
.
Lilly announces agreement to acquire ARMO biosciences
; 
2018
.
11.
Hutmacher
C
,
Neri
D
. 
Antibody-cytokine fusion proteins: biopharmaceuticals with immunomodulatory properties for cancer therapy
.
Adv Drug Deliv Rev
2018 Sep 7
. [
Epub ahead of print
].
12.
Pasche
N
,
Wulhfard
S
,
Pretto
F
,
Carugati
E
,
Neri
D
. 
The antibody-based delivery of interleukin-12 to the tumor neovasculature eradicates murine models of cancer in combination with paclitaxel
.
Clin Cancer Res
2012
;
18
:
4092
103
.
13.
Gutbrodt
KL
,
Schliemann
C
,
Giovannoni
L
,
Frey
K
,
Pabst
T
,
Klapper
W
, et al
Antibody-based delivery of interleukin-2 to neovasculature has potent activity against acute myeloid leukemia
.
Sci Transl Med
2013
;
5
:
201ra118
.
14.
Pogue
SL
,
Taura
T
,
Bi
M
,
Yun
Y
,
Sho
A
,
Mikesell
G
, et al
Targeting attenuated interferon-alpha to myeloma cells with a CD38 antibody induces potent tumor regression with reduced off-target activity
.
PLoS One
2016
;
11
:
e0162472
.
15.
Trinh
KR
,
Vasuthasawat
A
,
Steward
KK
,
Yamada
RE
,
Timmerman
JM
,
Morrison
SL
. 
Anti-CD20-interferon-beta fusion protein therapy of murine B-cell lymphomas
.
J Immunother
2013
;
36
:
305
18
.
16.
Siegemund
M
,
Schneider
F
,
Hutt
M
,
Seifert
O
,
Muller
I
,
Kulms
D
, et al
IgG-single-chain TRAIL fusion proteins for tumour therapy
.
Sci Rep
2018
;
8
:
7808
.
17.
Neal
ZC
,
Yang
JC
,
Rakhmilevich
AL
,
Buhtoiarov
IN
,
Lum
HE
,
Imboden
M
, et al
Enhanced activity of hu14.18-IL2 immunocytokine against murine NXS2 neuroblastoma when combined with interleukin 2 therapy
.
Clin Cancer Res
2004
;
10
:
4839
47
.
18.
Pastorek
J
,
Pastorekova
S
. 
Hypoxia-induced carbonic anhydrase IX as a target for cancer therapy: from biology to clinical use
.
Semin Cancer Biol
2015
;
31
:
52
64
.
19.
Neri
D
,
Supuran
CT
. 
Interfering with pH regulation in tumours as a therapeutic strategy
.
Nat Rev Drug Discov
2011
;
10
:
767
77
.
20.
McDonald
PC
,
Winum
JY
,
Supuran
CT
,
Dedhar
S
. 
Recent developments in targeting carbonic anhydrase IX for cancer therapeutics
.
Oncotarget
2012
;
3
:
84
97
.
21.
Lv
PC
,
Roy
J
,
Putt
KS
,
Low
PS
. 
Evaluation of nonpeptidic ligand conjugates for the treatment of hypoxic and carbonic anhydrase IX-expressing cancers
.
Mol Cancer Ther
2017
;
16
:
453
60
.
22.
Divgi
CR
,
Uzzo
RG
,
Gatsonis
C
,
Bartz
R
,
Treutner
S
,
Yu
JQ
, et al
Positron emission tomography/computed tomography identification of clear cell renal cell carcinoma: results from the REDECT trial
.
J Clin Oncol
2013
;
31
:
187
94
.
23.
Ahlskog
JK
,
Schliemann
C
,
Marlind
J
,
Qureshi
U
,
Ammar
A
,
Pedley
RB
, et al
Human monoclonal antibodies targeting carbonic anhydrase IX for the molecular imaging of hypoxic regions in solid tumours
.
Br J Cancer
2009
;
101
:
645
57
.
24.
Cazzamalli
S
,
Dal Corso
A
,
Widmayer
F
,
Neri
D
. 
Chemically defined antibody- and small molecule-drug conjugates for in vivo tumor targeting applications: a comparative analysis
.
J Am Chem Soc
2018
;
140
:
1617
21
.
25.
Mosmann
TR
,
Yokota
T
,
Kastelein
R
,
Zurawski
SM
,
Arai
N
,
Takebe
Y
. 
Species-specificity of T cell stimulating activities of IL 2 and BSF-1 (IL 4): comparison of normal and recombinant, mouse and human IL 2 and BSF-1 (IL 4)
.
J Immunol
1987
;
138
:
1813
6
.
26.
Hemmerle
T
,
Probst
P
,
Giovannoni
L
,
Green
AJ
,
Meyer
T
,
Neri
D
. 
The antibody-based targeted delivery of TNF in combination with doxorubicin eradicates sarcomas in mice and confers protective immunity
.
Br J Cancer
2013
;
109
:
1206
13
.
27.
Frey
K
,
Zivanovic
A
,
Schwager
K
,
Neri
D
. 
Antibody-based targeting of interferon-alpha to the tumor neovasculature: a critical evaluation
.
Integr Biol (Camb)
2011
;
3
:
468
78
.
28.
Probst
P
,
Kopp
J
,
Oxenius
A
,
Colombo
MP
,
Ritz
D
,
Fugmann
T
, et al
Sarcoma eradication by doxorubicin and targeted TNF relies upon CD8(+) T-cell recognition of a retroviral antigen
.
Cancer Res
2017
;
77
:
3644
54
.
29.
Huang
AY
,
Gulden
PH
,
Woods
AS
,
Thomas
MC
,
Tong
CD
,
Wang
W
, et al
The immunodominant major histocompatibility complex class I-restricted antigen of a murine colon tumor derives from an endogenous retroviral gene product
.
Proc Natl Acad Sci U S A
1996
;
93
:
9730
5
.
30.
Cazzamalli
S
,
Ziffels
B
,
Widmayer
F
,
Murer
P
,
Pellegrini
G
,
Pretto
F
, et al
Enhanced therapeutic activity of non-internalizing small-molecule-drug conjugates targeting carbonic anhydrase IX in combination with targeted interleukin-2
.
Clin Cancer Res
2018
;
24
:
3656
67
.
31.
Frey
K
,
Schliemann
C
,
Schwager
K
,
Giavazzi
R
,
Johannsen
M
,
Neri
D
. 
The immunocytokine F8-IL2 improves the therapeutic performance of sunitinib in a mouse model of renal cell carcinoma
.
J Urol
2010
;
184
:
2540
8
.
32.
Fugmann
T
,
Sofron
A
,
Ritz
D
,
Bootz
F
,
Neri
D
. 
The MHC class II immunopeptidome of lymph nodes in health and in chemically induced colitis
.
J Immunol
2017
;
198
:
1357
64
.
33.
Probst
P
,
Stringhini
M
,
Ritz
D
,
Fugmann
T
,
Neri
D
. 
Antibody-based delivery of TNF to the tumor neo-vasculature potentiates the therapeutic activity of a peptide anti-cancer vaccine
.
Clin Cancer Res
2019
;
25
:
698
709
.
34.
Gloger
A
,
Ritz
D
,
Fugmann
T
,
Neri
D
. 
Mass spectrometric analysis of the HLA class I peptidome of melanoma cell lines as a promising tool for the identification of putative tumor-associated HLA epitopes
.
Cancer Immunol Immunother
2016
;
65
:
1377
93
.
35.
Hoogenboom
HR
,
Marks
JD
,
Griffiths
AD
,
Winter
G
. 
Building antibodies from their genes
.
Immunol Rev
1992
;
130
:
41
68
.
36.
Silacci
M
,
Brack
S
,
Schirru
G
,
Marlind
J
,
Ettorre
A
,
Merlo
A
, et al
Design, construction, and characterization of a large synthetic human antibody phage display library
.
Proteomics
2005
;
5
:
2340
50
.
37.
Ziffels
B
,
Pretto
F
,
Neri
D
. 
Intratumoral administration of IL2- and TNF-based fusion proteins cures cancer without establishing protective immunity
.
Immunotherapy
2018
;
10
:
177
88
.
38.
Ritz
D
,
Gloger
A
,
Neri
D
,
Fugmann
T
. 
Purification of soluble HLA class I complexes from human serum or plasma deliver high quality immuno peptidomes required for biomarker discovery
.
Proteomics
2017
;
17
.
doi: 10.1002/pmic.201600364
.
39.
Pretto
F
,
Elia
G
,
Castioni
N
,
Neri
D
. 
Preclinical evaluation of IL2-based immunocytokines supports their use in combination with dacarbazine, paclitaxel and TNF-based immunotherapy
.
Cancer Immunol Immunother
2014
;
63
:
901
10
.
40.
Rosenberg
SA
. 
Interleukin 2 for patients with renal cancer
.
Nat Clin Pract Oncol
2007
;
4
:
497
.
41.
Koneru
R
,
Hotte
SJ
. 
Role of cytokine therapy for renal cell carcinoma in the era of targeted agents
.
Curr Oncol
2009
;
16
:
S40
4
.
42.
Creaven
PJ
,
Brenner
DE
,
Cowens
JW
,
Huben
RP
,
Wolf
RM
,
Takita
H
, et al
A phase I clinical trial of recombinant human tumor necrosis factor given daily for five days
.
Cancer Chemother Pharmacol
1989
;
23
:
186
91
.
43.
Bauer
S
,
Oosterwijk-Wakka
JC
,
Adrian
N
,
Oosterwijk
E
,
Fischer
E
,
Wuest
T
, et al
Targeted therapy of renal cell carcinoma: synergistic activity of cG250-TNF and IFNg
.
Int J Cancer
2009
;
125
:
115
23
.
44.
Chen
DS
,
Mellman
I
. 
Elements of cancer immunity and the cancer-immune set point
.
Nature
2017
;
541
:
321
30
.
45.
Danielli
R
,
Patuzzo
R
,
Di Giacomo
AM
,
Gallino
G
,
Maurichi
A
,
Di Florio
A
, et al
Intralesional administration of L19-IL2/L19-TNF in stage III or stage IVM1a melanoma patients: results of a phase II study
.
Cancer Immunol Immunother
2015
;
64
:
999
1009
.
46.
Yoo
EM
,
Trinh
KR
,
Tran
D
,
Vasuthasawat
A
,
Zhang
J
,
Hoang
B
, et al
Anti-CD138-targeted interferon is a potent therapeutic against multiple myeloma
.
J Interferon Cytokine Res
2015
;
35
:
281
91
.
47.
Neri
D
. 
Antibody-cytokine fusions: versatile products for the modulation of anti-cancer immunity
.
Cancer Immunol Res
2019
;
7
:
348
54
.
48.
De Luca
R
,
Neri
D
. 
Potentiation of PD-L1 blockade with a potency-matched dual cytokine-antibody fusion protein leads to cancer eradication in BALB/c-derived tumors but not in other mouse strains
.
Cancer Immunol Immunother
2018
;
67
:
1381
91
.
49.
De Luca
R
,
Soltermann
A
,
Pretto
F
,
Pemberton-Ross
C
,
Pellegrini
G
,
Wulhfard
S
, et al
Potency-matched dual cytokine-antibody fusion proteins for cancer therapy
.
Mol Cancer Ther
2017
;
16
:
2442
51
.

Supplementary data