A Novel Antibody–IL15 Fusion Protein Selectively Localizes to Tumors, Synergizes with TNF-based Immunocytokine, and Inhibits Metastasis

This article is featured in Highlights of This Issue, p. 761

IL15 is a proinflammatory cytokine, mainly produced by macrophages, dendritic cells, and monocytes. IL15 is a monomeric protein with a molecular mass of approximately 15 kDa, bearing a N-linked glycan on the position N127 (1, 2). IL15 is structurally related to IL2, the two cytokines share the same IL2/IL15 receptor β and γ subunits, which are displayed on the surface of natural killer (NK) cells and T cells (3). Both cytokines are able to stimulate the proliferation of T cells, promote the synthesis of immunoglobulin by B cells and the survival of (NK) cells (3). However, IL2 and IL15 have different roles during adaptive immune responses. IL2 is responsible for the maintenance of peripheral regulatory T cells (Tregs), which promotes the elimination of self-reactive T cells. In contrast, IL15 plays a crucial role in the survival of CD8⁺ memory T cells (2, 4). For these reasons, IL15 might be considered a better candidate than IL2 for cancer treatment (5).

Unlike engineered preparations of IL2, which have shown favorable pharmacokinetic profiles in mouse (6, 7) and man (8, 9), development activities for IL15-based products have been characterized by suboptimal tumor-targeting properties and ultra-rapid clearance from the bloodstream (10, 11).

Recombinant preparations of human IL15 have been considered for clinical and industrial development programs. A first-in-human trial was performed with recombinant human-IL15 (rhIL15) produced in Escherichia coli (NCT01021059). Bolus infusions of 0.3, 1.0, and 3.0 μg/kg per day of rhIL15 were administered for 12 days to patients with metastatic malignant melanoma or metastatic renal cell carcinoma. The study revealed a dramatic efflux of NK and memory CD8⁺ T cells upon rhIL15 administration and a 50-fold increase of inflammatory cytokines in sera. However, severe side effects, such as liver toxicity, hypotension, and thrombocytopenia were observed. The MTD of rhIL15 was determined at 0.3 μg/kg per day (12).

The Sushi Domain (SD) is the shortest region of the IL15 receptor α (IL15Rα) capable of binding to IL15. Recent studies with the complex SD-IL15 showed a prolonged serum half-life and an increased biological activity compared with IL15 alone (13, 14). Moreover, fusion proteins featuring IL15Rα−IL15 showed an increased anticancer activity compared with IL15 in mouse model of the disease (15, 16).

A product named XmAb24306 (17), consisting of an IL15/IL15Rα heterodimeric Fc-fusion featuring reduced in vitro potency, compared with wild-type IL15, and prolonged serum half-life is currently being investigated in phase I clinical trials, alone and in combination with atezolizumab in patients with locally advanced or metastatic solid tumors (NCT04250155).

Another IL15-based product, ALT-803 (18, 19), consisting of a novel IL15 superagonist complex bearing an asparagine to aspartic acid substitution (N72D) to increase binding affinity for IL15Rβ and IL15Rγ complex was investigated in patients with cancer. A first-in-human phase I trial with ALT-803 in patients with advanced solid tumor revealed a modest expansion of CD8⁺ T cells and a dramatic increase in NK cells. Repeated subcutaneous injections of ALT-803 were well-tolerated (20). In addition, ALT-803 was tested in combination with nivolumab in patients with metastatic non–small cell lung cancer. A phase IB trial showed a synergistic activity between the two products (21).

Tumor-homing antibodies have been considered as “vehicles” for the selective delivery of therapeutic cytokines to the tumor mass, with the goal to increase therapeutic activity and reduce toxicity (22, 23). In particular, antibodies specific to splice-isoforms of fibronectin may be considered suitable molecules for targeted delivery, because these antigens are virtually undetectable in normal healthy tissues (exception made for the placenta and the endometrium in the proliferative phase), while being strongly expressed in the majority of solid tumors and lymphomas, as well as other diseases (24, 25). mAbs (named F8 and L19) have been selected against the alternatively spliced extra domain A (EDA) and B (EDB) of fibronectin, respectively (26, 27). Because EDA and EDB are conserved from mouse to man, the L19 and F8 antibodies recognize their antigen with identical affinity, thus facilitating translational activities from preclinical models to clinical trials.

Many cytokines can efficiently be delivered to tumors by fusion with suitable antibodies (e.g., IL2, IL4, IL6, IL9, IL10, IL12, TNF, IFNα), while other immunomodulatory payloads exhibit limitation in tumor targeting, preventing their pharmaceutical development (i.e., IL7, IL17, and IFNγ; refs. 27–30). For example, IFNγ, is a pleiotropic cytokine that plays a crucial role in the promotion of innate and adaptive mechanism of host defense. Because of its ability to directly inhibit the tumor growth, attempts to use IFNγ as a payload for tumor-targeting antibodies were explored. Unfortunately, these prototypes were not further developed due to in vivo receptor trapping (30). Quantitative biodistribution analysis, performed with radioiodinated preparation of L19-IFNγ in immunocompetent and IFNγ Receptor (IFNγR) knockout mice, revealed a clearly enhanced tumor targeting in mice lacking IFNγR, thus suggesting that the fusion protein is partially sequestered in vivo before binding to the EDB antigen expressed in the tumor.

An IL15 fusion protein with the L19 antibody in diabody format was cloned and expressed by Kaspar and colleagues (31). The immunocytokine was capable of inducing tumor growth retardation in mouse models of cancer. However, the biodistribution profiles of L19-IL15 were suboptimal compared with fusion proteins of the L19 antibody with other cytokines payloads.

It has been previously reported by Kermer and colleagues that a fusion protein comprising an anti-fibroblast activation protein (FAP) antibody in single-chain Fv format fused to an extended variant of the SD and to IL15 was able to prevent the formation of metastatic lesions in a B16-FAP lung metastasis mouse model (32). This molecule was further developed by adding costimulatory TNF-superfamily ligands (4-1BBL, OX40L and GITRL). The trifunctional antibody fusion protein was produced both as noncovalent homotrimer and as a single-chain polypeptide (33, 34).

In this study, we have generated two novel formats of IL15 fusion proteins (termed F8-F8-IL15 and F8-F8-SD-IL15), featuring the F8 antibody in single-chain diabody format, which have exhibited positive results in quantitative biodistribution studies and potent therapeutic activity in certain experimental models of cancer. The single-chain diabody format (22) is attractive as it allows to produce fusion proteins as single polypeptides, with two antigen-binding sites, leading to avid binding and efficient tumor-targeting properties. The results may be of clinical interest, because the EDA domain of fibronectin is extremely well conserved (only three amino acid mutations between mouse and man) and is overexpressed in the majority of solid and liquid tumors, while being virtually undetectable in normal adult tissues, exception made for certain structures of the female reproductive tract (35–37).

The gene encoding for the F8 antibody, human IL15 and human SD were PCR amplified, PCR assembled, and cloned into the mammalian expression vector pcDNA3.1(+) (Invitrogen) using NheI/BamHI/NotI restriction enzymes. The fusion proteins were expressed using transient gene expression (TGE) in CHO-S cells. Negative control constructs were cloned with similar methodologies using the KSF antibody, specific to hen egg lysozyme, an irrelevant antigen. For 1 mL of production, 4 × 10⁶ CHO-S cells in suspension were centrifuged and resuspended in 1 mL ProCHO4 (Lonza). 0.625-0.9 μg of plasmid DNAs followed by 2.5 μg polyethylene imine (PEI; 1 mg/mL solution in water at pH 7.0) per million cells were then added to the cells and gently mixed. The transfected cultures were incubated in a shaker incubator at 31°C for 6 days. Proteins were purified from the cell culture medium by protein A affinity chromatography and then dialyzed against buffer; 50 mmol/L Tris, 100 mmol/L NaCl pH 8.5 for F8-F8-IL15, F8-F8-SD-IL15, and PSB pH 7.4 for F8-F8-IL15-SD, IL15-SD-F8-F8, SD-IL15-F8-F8, F8-F8-SD, F8(Db)-IL15, F8(scFv)-SD-IL15, KSF-KSF-IL15 and KSF-KSF-SD-IL15. A detailed description of protein production and purification is present in the Supplementary Material.

SDS-PAGE was performed with 4% to 12% Bis-Tris gel (Invitrogen, NP0322BOX) under reducing and nonreducing conditions. Proteins were analyzed by size-exclusion chromatography using a Superdex 200 increase 10/300 GL column on an ÄKTA FPLC (GE Healthcare, Amersham Biosciences). Affinity measurements were performed by surface plasmon resonance using BIAcore X100 (BIAcore, GE Healthcare) instrument using an EDA coated CM5 chip. Samples were injected as serial-dilutions, in a concentration range from 125 nmol/L to 1,000 nmol/L. Regeneration of the chip was performed using HCl 10 mmol/L. Differential scanning fluorimetry was performed on an Applied Biosystems StepOnePlus RT-PCR instrument. Protein samples were diluted at 2 μmol/L in PBS in 40 μL and placed in PCR tubes, assay was performed in triplicates. 5× SYPRO ORANGE (Invitrogen, stock 5,000x) was added to samples prior to analysis. For thermal stability measurements, the temperature range spanned from 25°C to 95°C with a scan rate of 1°C/minute. Data analysis was performed in Protein Thermal Shift Software version 1.3 by calculating the derivative of the melting curve.

The biological activity of the fusion proteins was determined by their ability to stimulate the proliferation of murine cytotoxic T lymphocytes (CTLL2) (ATCC). In 96-well plates, cells (25,000-50,000 per well) were seeded in culture medium supplemented with serial dilutions of the fusion proteins. After incubation at 37°C for 72 hours, cell proliferation was determined with Cell Titer Aqueous One Solution (Promega). Results were expressed as the percentage of cell viability compared with untreated cells.

Biological NF-κB activity assay was performed as described previously (38). CTLL-2_ NF-κB reporter cells were starved in absence of IL2 for 6 to 9 hours. Cells were seeded in 96-well plates at 50,000 cells/well with serial dilutions of the fusion proteins. To assess luciferase production, 20 μL of supernatant was transferred in an opaque 96-well plate (PerkinElmer, Optiplate 96, white) with 80 μL of Coalenterazine (Carl Rath AG, 1 μg/mL). Luminescence was immediately measured at 466 nm and results were calculated and expressed by division with values obtained from untreated cells.

Antigen expression was confirmed on ice-cold acetone fixed 8-μm cryostat sections of F9, C51, and WEHI-164 stained with F8-F8-IL15, F8-F8-SD-IL15, KSF-KSF-IL15, and KSF-KSF-SD-IL15 (final concentration 0.01 mg/mL) and detected with Protein A-AlexaFluor488 (Invitrogen P11047). A rat anti-mouse CD31 (BD 550274) was used to detect blood vessels and detected with donkey anti-rat ALEXA 594 (Invitrogen A21209). Slides were mounted with fluorescent mounting medium containing DAPI (Dako Omnins GM304) and analyzed with Axioskop2 mot plus microscope (Zeiss). For ex vivo immunofluorescence analysis, 129/SvEv mice bearing F9 teratocarcinomas were injected intravenously with 100 μg of F8-F8-IL15, F8-F8-SD-IL15, KSF-KSF-IL15, and KSF-KSF-SD-IL15. After 24 hours mice were sacrificed, tumors and organs were excised and embedded in cryo-embedding medium (Thermo Scientific). Cryostat section of 8 μm were stained with Protein A-AlexaFluor488 (Invitrogen P11047) and donkey anti-rat ALEXA 594 to detect the blood vessels stained with rat anti-mouse CD31 (BD 550274). Slides were mounted with fluorescent mounting medium containing DAPI (Dako Omnins GM304) and analyzed with Axioskop2 mot plus microscope (Zeiss).

CHO cells, CTLL2, WEHI-164, F9, and C51 cells were obtained from the ATCC. Cell lines were received between 2018 and 2020, expanded, and stored as cryopreserved aliquots in liquid nitrogen. Cells were grown according the manufacturer's protocol and kept in culture for no longer than 14 passages. Authentication of the cell lines also including check of postfreeze viability, growth properties, and morphology, test for Mycoplasma contamination, isoenzyme assay, and sterility test were performed by the cell bank before shipment.

Six- to 8-week-old female 129/SvEv mice were obtained from Janvier Labs. 2 × 10⁷ F9 tumor cells were injected subcutaneously in the flank. When tumors reached a volume of 100 to 200 mm³, fusion proteins (100 μg) were radioiodinated with ¹²⁵I and Chloramine T hydrate, purified on a PD10 column and injected into the lateral tail vein as described before (39). Mice were sacrificed 24 hours after injection. Organs, blood, and tumors were weighed and radioactivity was detected using a Packard Cobra gamma counter. The immunocytokine uptake in blood, organs, and tumors was calculated and expressed as the percentage of the injected dose per gram of tissue (%ID/g ± SEM). Data were corrected for tumor growth (39).

Six- to 8-week-old female BALB/c mice were obtained from Janvier Labs. Tumor cells were implanted subcutaneously in the flank using 2 × 10⁶ cells (C51) and 5 × 10⁶ cells (WEHI-164). Mice were monitored daily and tumor volume was measured with a caliper (volume = length × width² × 0.5). When tumors reached a suitable volume (approximately 80–120 mm³), mice were randomized and injected into the lateral tail vein with the pharmacologic agents dissolved in saline solution, also used as negative control, every 48 hours. The study was conducted under blinding conditions where the researcher conducting the in vivo study received protein preparations labelled with a code from a second researcher. In the C51 tumor model the following doses were used: 5 μg/g F8-F8-IL15, 0.6 μg/g F8-F8-SD-IL15, 10 μg/g anti–PD-1 antibody (clone 29F.1A12, BioXCell, catalog No. BE0273). In the combination groups, the immunocytokine was injected first, followed after 24 hours by the anti–PD-1 antibody. In the WEHI-164 tumor model, the following doses were used: 5 μg/g F8-F8-IL15, 0.6 μg/g F8-F8-SD-IL15, 0.1 μg/g F8mTNF, 10 μg/g anti–PD-1 antibody (clone 29F.1A12, BioXCell, catalog No. BE0273). In the combination groups, the pharmacologic agents were injected simultaneously. On the basis of preliminary dose-finding studies (Supplementary Fig. S7) in the combination group F8-F8-SD-IL15 + F8mTNF the dose of F8-F8-SD-IL15 was reduced to 0.3 μg/g. Results are expressed as tumor volume in mm³ ± SEM and % mean body weight change ± SEM. For therapy experiments n = 5 mice/group. Euthanasia criteria adopted were body weight loss > 15%, ulceration of the subcutaneous tumor, tumor diameter > 1,500 mm and mice pain and discomfort. Mice were euthanized in CO₂ chambers.

Female BALB/c mice were injected intravenously with 10⁵ C51 cells. Twenty-four hours after, mice were divided into three groups (n = 6) and injected three times every 48 hours with PBS (100 μL), F8-F8-IL15 (5 μg/g), and F8-F8-SD-IL15 (0.6 μg/g). Body weight and breathing behavior, frequency and depth were monitored daily. At day 20, mice were sacrificed, lungs were removed, fixed in saline 4% formaldehyde, and examined with a Zeiss stereomicroscope. Results were expressed as numbers of metastatic foci per lung.

Three groups of mice BALB/c (n = 3), challenged subcutaneously with C51 or WEHI-164 tumor cells, were injected intravenously every 48 hours three times with PBS (100 μL), F8-F8-IL15 (5 μg/g), and F8-F8-SD-IL15 (0.6 μg/g). Twenty-four hours after the last injection, tumors and draining lymph nodes (dLN) were excised, processed and analyzed by FACS. Surface staining was performed with phycoerythrin (PE)-coupled AH1 tetramer produced as described previously (40) and anti-mouse fluorochrome–conjugated antibodies: CD3-APC/Cy7 (clone 17A2), CD4-APC (clone GK1.5), CD8-FITC (clone 53-6.7), NK1.1-PE (clone PK136), and MHCII(IA/IE)-BV421 (clone M5/114.15.2). Before analyses, cells were incubated for 5 minutes at 4°C with 7-AAD Viability Staining Solution (BioLegend). For the staining of the intracellular marker Foxp3, cells were stained for 15 minutes at room temperature with Zombie Red Fixable Viability Kit (BioLegend). After surface markers staining, cells were permeabilized with eBioscience Foxp3/Transcription Factor Staining Buffer Set, and incubated 30 minutes at room temperature with Foxp3-BV421 (clone MF-14). All the antibodies were purchased from BioLegend and diluted in FACS buffer 1:200. For the final analyses, CytoFLEX cytometer (Beckman Coulter) was used and the data were processed with FlowJo 10.4.

Mice were injected intravenously once with PBS (100 μL), F8-F8-IL15 (5 μg/g), and F8-F8-SD-IL15 (0.6 μg/g). Six hours after the injection, blood was taken from the heart, processed, and analyzed by FACS. Samples were treated three times with Red Blood Cell Lysis buffer (Roche). Surface staining was performed with and anti-mouse fluorochrome-conjugated antibodies: CD3-APC/Cy7 (clone 17A2), CD4-APC (clone GK1.5), CD8-FITC (clone 53-6.7), NK1.1-PE (clone PK136), and MHCII(IA/IE)-BV421 (clone M5/114.15.2). Before analyses, cells were incubated 5 minutes at 4°C with 7-AAD Viability Staining Solution (BioLegend). All the antibodies were purchased from BioLegend and diluted in FACS buffer 1:200. For the final analyses, CytoFLEX cytometer (Beckman Coulter) was used and the data were processed with FlowJo 10.4.

Human peripheral blood mononuclear cells (hPBMCs) were isolated from healthy donors with Ficoll solution, resuspended in X-VIVO 15 Medium, and stained (50,000 cells per well) with carboxyfluorescein diacetate succinimidyl ester (CSFE) using CSFE Cell Division Tracker Kit (BioLegend). Four different concentrations (0, 0.1, 1, 10 nmol/L) of the fusion proteins F8-F8-IL15 and F8-F8-SD-IL15 were added and proliferation induced at 37°C for 5 days. Flow cytometry analyses was performed after cell staining with anti-human CD3-APC/Cy7 (clone HIT3a), CD4-APC (clone A161A1), CD8-BV421 (clone SK1), NKp46-PE/Cy7 (clone 29A1.4). For the staining of the intracellular marker Foxp3, cells were stained for 15 minutes at room temperature with Zombie Red Fixable Viability Kit (BioLegend). After surface markers staining, cells were permeabilized with eBioscience Foxp3/Transcription Factor Staining Buffer Set, and incubated 30 minutes at room temperature with Foxp3-BV421 (clone 206D). All the antibodies were purchased from BioLegend and diluted in FACS buffer 1:200. For the final analyses, CytoFLEX cytometer (Beckman Coulter) was used and the data were processed with FlowJo 10.4.

Data were analyzed using Prism 7.0 (GraphPad Software, Inc.). Statistical significances between multiple groups were evaluated with the one-way ANOVA followed by Tukey as posttest. P < 0.05 was considered statistically significant. (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).

Experiments were performed under a project license (license number 04/2018) granted by the Veterinäramt des Kantons Zürich, Switzerland, in compliance with the Swiss Animal Protection Act (TSchG) and the Swiss Animal Protection Ordinance (TSchV).

A first fusion protein consisting of the F8 antibody, specific to the alternatively spliced EDA domain of fibronectin, in tandem diabody format (41), fused at the C-terminus by a peptide linker to human IL15 was cloned and produced (Fig. 1A). A second fusion protein, in which the IL15 moiety was fused at the N-terminus of the F8 antibody in tandem diabody format, was cloned (Supplementary Fig. S1A). However, the fusion protein was not able to induce cell proliferation in an in vitro assay as efficiently as the fusion protein with IL15 at the C-terminus.

A third fusion protein, in which the SD of IL15Rα was incorporated, was cloned and produced (Fig. 1B). In a scouting study, different formats, in which the orientation of cytokine payloads was changed, were cloned and produced. The best candidate was chosen based on production yield and biochemical properties (Supplementary Fig. S2).

The two fusion proteins, termed F8-F8-IL15 and F8-F8-SD-IL15 (full amino acid sequences in Supplementary Fig. S3), were expressed in CHO cells and purified to homogeneity exploiting the binding properties of the VH domain of the F8 antibody to Protein A resin (29), yielding 5mg/L for both fusion proteins. Figure 1A and B show the analytic characterization of F8-F8-IL15 and F8-F8-SD-IL15 by SDS-PAGE, gel filtration, and BIAcore analysis on an EDA-coated microsensor chip. An in vitro lymphocyte proliferation assay indicated that both F8-F8-IL15 and F8-F8-SD-IL15 retained the same IL15 activity compared with the positive control F8(Db)-IL15 (Supplementary Fig. S1B), a protein with similar structure as a previously described product (31). In contrast, F8-F8-SD (Supplementary Fig. S1C), a fusion protein produced with similar methodologies as for F8-F8-SD-IL15, but devoid of the IL15 moiety, was not able to induce cell proliferation. An additional in vitro assay based on a NF-κB reporter cell line (38) confirmed the biological activity of both fusion proteins (Supplementary Fig. S4). A denaturation profile was observed by differential scanning fluorimetry, with a transition at 50°C for F8-F8-IL15, whereas for F8-F8-SD-IL15 a first transition at 54°C and a second at 69°C was observed. An in vitro immunofluorescence analysis revealed that both F8-F8-IL15 and F8-F8-SD-IL15 products strongly stained the neovasculature of F9 teratocarcinomas, whereas analysis on C51 colorectal carcinomas and WEHI-164 fibrosarcoma revealed a more diffuse stromal staining pattern (Supplementary Fig. S5). When KSF-KSF-IL15 or KSF-KSF-SD-IL15 (Supplementary Fig. S1D and S1E; fusion proteins with identical format but specific to hen egg lysozyme) were used as negative control in immunofluorescence analysis, no detectable staining was observed in addition to the CD31 signal.

The tumor-targeting properties of radioiodinated preparations of F8-F8-IL15 and F8-F8-SD-IL15 were assessed by quantitative biodistribution analysis in immunocompetent mice, bearing subcutaneously grafted murine F9 teratocarcinomas. F8-F8-IL15 selectively localized to solid tumors with excellent tumor:organ and tumor:blood ratios 24 hours after intravenous administration (Fig. 2A). The biodistribution profile was found to be superior in terms of tumor uptake compared with F8(Db)-IL15 (Fig. 2B; Supplementary Fig. S1B), a fusion protein with a similar structure of the previously described L19(Db)-IL15 molecule (31). A similar profile was observed for L19-L19-IL15 (Supplementary Fig. S1F; Supplementary Fig. S6), thus confirming the superiority of the tandem diabody format over the noncovalent homodimer format. F8-F8-SD-IL15 selectively localized to solid tumors with an excellent tumor:blood ratio. An uptake was detectable in spleen, kidney, intestine, and liver 24 hours after intravenous administration (Fig. 2C). The biodistribution of F8(scFv)-SD-IL15 (Fig. 2D; Supplementary Fig. S1G), a molecule featuring only one antibody moiety (32, 33), revealed an inferior antibody uptake in the neoplastic lesions and increased uptake in healthy organs compared to the bivalent F8-F8-SD-IL15 format. The radioiodination process could, in theory, affect the binding properties or the performance of the antibody. We used chloramine T as oxidative agent to perform the radiolabeling reaction where ¹²⁵I is substituted for reactive hydrogen sites in the target molecules (42). Chloramine-T could indeed act not only as an oxidizing agent for the sodium iodide but also chemically modify the antibody (43). For these reasons, the tumor-homing properties of F8-F8-IL15 and F8-F8-SD-IL15 were further investigated in immunocompetent mice, bearing subcutaneously grafted murine F9 teratocarcinomas (39) using immunofluorescence procedures. Organs were examined 24 hours after intravenous administration of fusion proteins. A homogenous uptake to the tumor neovasculature was observed for both F8-F8-IL15 and F8-F8-SD-IL15, while no detectable antibody uptake could be seen in relevant normal organs. In contrast, no tumor and organs uptake could be observed for the KSF-KSF-IL15- and KSF-KSF-SD-IL15–negative control proteins (Fig. 2E).

In a first study, we compared the therapeutic activity of F8-F8-IL15 and F8-F8-SD-IL15 and the anti-mouse PD-1 antibody, either as single agents or as combination, in immunocompetent BALB/c mice bearing subcutaneously grafted C51 colorectal tumors. F8-F8-IL15 was administered at a dose of 5 μg/g as described before (31). The MTD of F8-F8-SD-IL15 was found to be 0.6 μg/g, as evidenced by preliminary dose-escalation studies (Supplementary Fig. S7A). The commercial anti–PD-1 antibody was used at the recommended dose of 10 μg/g (44, 45). A modest tumor growth retardation was observed in some (but not all) treated mice compared with the saline group. (Fig. 3A). All products were well tolerated as shown by body weight profiles (Supplementary Fig. S8A).

When injected intravenously, C51 tumor cells may give rise to lung metastases (31). We used this animal model to assess the antimetastatic potential of F8-F8-IL15 and F8-F8-SD-IL15. Immunocytokines were administered at the same dosage used in the subcutaneous C51 tumor model, three times (day 1, 3, and 5 following tumor cell injection) intravenously. Twenty days after tumor cell inoculation, mice were sacrificed and metastatic load was assessed by counting foci in lungs (Fig. 3B). Both immunocytokines displayed a clear antimetastatic activity compared with control mice treated with saline. F8-F8-SD-IL15 was strikingly more active (3 out of 6 mice free of metastases) compared with F8-F8-IL15 (Fig. 3C). Mice in the control group showed signs of respiratory distress, due to a higher metastatic load, which caused a reduction in body weight. Similar findings were not observed in treated mice (Supplementary Fig. S7B).

In a third study, we tested the activity of F8-F8-IL15 and F8-F8-SD-IL15 in immunocompetent BALB/c mice bearing subcutaneously grafted WEHI-164 fibrosarcoma tumors (Fig. 4A). Fusion proteins were tested both as a single agent and in combination with an anti–PD-1 antibody and F8mTNF (46). When used in combination with F8mTNF, F8-F8-SD-IL15 dosage was reduced to 0.3 μg/g based on preliminary toxicity study (Supplementary Fig. S7C). A tumor growth retardation was observed in mice treated with F8-F8-IL15 or F8-F8-SD-IL15 compared with the saline group. Treatment with anti–PD-1 antibody was found to be potently active against this tumor model leading to a complete response of 4 of 5 treated mice. Because of this strong anticancer activity, no synergistic effect was observed in combination with targeted IL15. Instead, a strong synergistic activity was observed in the combination groups with targeted TNF. In the monotherapy group F8mTNF was able to cure only one mouse out of five, whereas in both combination groups three of five mice showed a complete response phenotype. In this experimental setting, all the products were well tolerated as evidenced by body weight profiles (Supplementary Fig. S8B). All cured mice, from previous treatment, were rechallenged with WEHI-164 cells 70 days after primary tumor implantation and were found to have acquired a protective immunity against the tumor compared with naïve mice (Fig. 4B).

To characterize lymphocyte composition upon immunocytokine treatment, C51 tumors and tumor-draining lymph nodes were analyzed by flow cytometry (Material and Methods and gating strategy in Supplementary Fig. S9A–S9D). Fig. 5A shows the lymph node composition, in which we observed a significant increase in the ratio between CD8/CD4 in mice treated with both immunocytokines with a major effect in the prototype carrying the SD. No significant change was observed in the AH1-specific CD8⁺ T-cell population. As expected, the treatment with IL15-based products did not affect the proportion of regulatory T cells (Fig. 5B). A similar analysis was performed for immune infiltrate cells in tumors (Fig. 5C and D). A difference in the ratio CD8/CD4 was observed upon treatment with F8-F8-IL15.

The same investigation was performed in WEHI-164 tumor–bearing mice (Fig. 6). In addition, mice treatment with F8-F8-SD-IL15 and F8mTNF were also analyzed 24 hours after a single injection due to the induction of necrosis and hemorrhagic necrosis of TNF (40, 47). The lymph nodes composition (Fig. 6A) revealed a massive increase of CD8⁺ T cells after the treatments with F8-F8-IL15 and F8-F8-SD-IL15. An increase of NK cells and AH1-specific CD8⁺ T cells in the lymph nodes of mice treated with the combination F8-F8-SD-IL15 + F8mTNF was observed. Also, in this setting Treg cells were not affected (Fig. 6B). The analysis of the tumor composition (Fig. 6C) revealed only a difference in the proportion of antigen-presenting cells (APC) upon treatment with F8-F8-IL15 (Fig. 6E). The necrotic activity of the F8mTNF combined with F8-F8-SD-IL15 can be detected in the percentage of total living cells present in the tumor (Fig. 6D). In addition, healthy mice were injected with saline, F8-F8-IL15, and F8-F8-SD-IL15 and sacrificed 6 hours after intravenous administration. Serum analysis revealed an increase in the percentage of CD4⁺ T cells, CD8⁺ T cells, APC, and NK cells (Supplementary Fig. S10).

Finally, the ability of F8-F8-IL15 and F8-F8-SD-IL15 to stimulate hPBMCs (Fig. 7) was assessed through an in vitro proliferation assay followed by flow cytometry analysis (Materials and Methods and gating strategy in Supplementary Fig. S9E and S9F). We tested four different concentrations of immunocytokines (0, 0.1, 1 and 10 nmol/L), the analysis revealed no proliferation in the CD4⁺ T cells compartment, but a significant increase of CD8⁺ T cells with the highest concentration of F8-F8-IL15 (10 nmol/L) and with F8-F8-SD-IL15 already at 1 nmol/L concentration. The most striking proliferative activity was observed among natural killer (NK) cells. The incubation with F8-F8-IL15 at 10 nmol/L promoted the division up to 68% of NK and even higher (83%) with F8-F8-SD-IL15 already at 1 nmol/L concentration. Incubation with F8-F8-IL15 induced a proliferation in the percentage of Treg cells, whereas no expansion was observed when PBMCs were stimulated with F8-F8-SD-IL15.

In this work, we have described the generation and the characterization of six different immunocytokine formats based on an anti-EDA antibody fragments fused to human IL15 and to the SD of IL15Rα. Among these prototypes only two products, named F8-F8-IL15 and F8-F8-SD-IL15, had encouraging biochemical properties suitable for further in vivo validation.

Favorable biodistribution profiles were observed for two of the novel immunocytokine formats, with a substantial improvement compared with previously described work on antibody-IL15 fusions (31–33). These findings suggest that the relative spatial arrangement of antibody fragments and of the cytokine moiety may have a direct impact on the ability of the fusion proteins to efficiently extravasate and remain at the site of disease. The use of a diabody as a single-chain polypeptide has previously worked well with other cytokine payloads (41, 45).

The therapeutic activity of F8-F8-IL15 and F8-F8-SD-IL15 was evaluated in immunocompetent mice bearing subcutaneously grafted C51 colon carcinomas and WEHI-164 sarcomas. The MTD of F8-F8-SD-IL15 was found to be almost 10 times lower compared with the one of F8-F8-IL15, thus confirming the potentiation of IL15 activity by the SD (13–16). Both products were not able to induce complete responses in these subcutaneous tumor models either alone or in combination with check point inhibitors. In contrast, the combination with targeted TNF in a murine sarcoma model (WEHI-164) was found to be more efficacious compared with single-agent treatments, thus providing a rationale for future combinatorial studies.

A very potent anticancer activity was observed in a mouse model of lung metastasis. F8-F8-SD-IL15 completely prevented the formation of metastatic lesions, whereas treatment with F8-F8-IL15 led only to a modest reduction in metastasis load. These findings suggest that treatment with IL15-based products may be efficacious when used in a preventive setting, such as the adjuvant use for the management of minimal residual disease. Debulking of larger lesions was not efficient in our hands. These findings are in line with the ones recently reported by other groups with other types of IL15-based therapeutics (11, 17, 48). The therapeutic activity described in the lung metastasis model may rely on the observed expansion of circulating CD4⁺ T cells, CD8⁺ T cells, and NK cells in blood.

The analysis of lymphocyte composition in lymph nodes and tumor mass from mice provided insight on IL15 biology. Both F8-F8-IL15 and F8-F8-SD-IL15 were able to induce a pronounced proliferation of CD8⁺ T cells compared with saline treatment. Importantly, the Treg population was not expanded upon pharmacologic treatment. However, in the PBMCs proliferation assay F8-F8-IL15 induced an expansion in the percentage of Treg cells. Interestingly, when PBMCs were stimulated with F8-F8-SD-IL15 the Treg population was not affected. This feature may be relevant when considering the therapeutic use of IL2- and IL15-based products. SD-IL15–based cancer therapy may be preferable, because IL2 is known to expand the Treg population in certain settings (2, 49).

We focused on AH1-specific CD8⁺ T cells as an important subset of T cells responsible for the tumor rejection in BALB/c mice. We observed that combination treatment of F8-F8-SD-IL15 with F8mTNF was able to boost the expansion of AH1-specific CD8⁺ T cells. AH1 is a peptide derived from gp70 envelope protein of murine leukemia, which is endogenously present in the genome BALB/C mice. The gp70 protein is strongly expressed in many tumors of BALB/c origin, but absent in healthy tissues and the AH1 peptide has been described as the major tumor rejection antigen in CT26- and WEHI-164–derived mouse models (40, 50). These results strengthen experimental evidence from our group (40, 50).

Results obtained in mice experiments were in line with in vitro proliferation assays on hPBMCs, both F8-F8-IL15 and F8-F8-SD-IL15 were able to boost CD8⁺ T cells and NK cells from healthy donors, thus giving a rationale for further clinical applications.

IL15-based products are currently being investigated in clinical trials against several malignancies, either as single agents or in combination with immune check point inhibitors (NCT04250155; ref. 21). In particular, products in which the SD of IL15Rα is incorporated are gaining clinical relevance due to the unique ability of the SD to stabilize the IL15-IL15R complex and to increase the biological activity of the cytokine. The results presented in this study reinforce the use of targeted IL15 for cancer therapy and provide a rationale for the clinical evaluation of the described prototypes in combination with targeted TNF for the treatment of patients with soft tissue sarcoma. L19-TNF is currently being investigated in combination with doxorubicin [phase III (EU), EudraCT 2016-003239-38 and phase IIb (UC), NCT03420014] and in combination with DTIC (EudraCT 2018-004104-19). Novel combinations (i.e., with targeted IL15) may provide additional benefits in the treatment of sarcoma.

R. Corbellari reports a patent for PCT/EP2020/066344 pending; in addition, R. Corbellari is an employee of Philochem AG, daughter company of Philogen acting as discovery unit of the group. J. Mock reports grants from Swiss National Science Foundation (SNF) during the conduct of the study. A. Villa reports a patent pending; in addition, A. Villa is an employee of Philochem AG, daughter company of Philogen acting as discovery unit of the group. D. Neri is CEO, CSO, cofounder, and shareholder of Philogen R. De Luca reports a patent pending; in addition, R. De Luca is an employee of Philochem AG, daughter company of Philogen acting as discovery unit of the group. No disclosures were reported by the other authors.

R. Corbellari: Conceptualization, data curation, validation, investigation, methodology, writing–original draft, writing–review and editing. M. Stringhini: Validation, investigation, methodology. J. Mock: Validation, investigation, methodology. T. Ongaro: Investigation, methodology. A. Villa: Validation. D. Neri: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. R. De Luca: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We would like to thank Lisa Nadal, Alessandro Sannino, and Marco Catalano for their help with experimental procedures. We gratefully acknowledge funding from ETH Zürich and the Swiss National Science Foundation (grant No. 310030_182003/1). This project has received funding from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (grant agreement 670603).

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.