Pretargeted Radioimmunotherapy with the Novel Anti-oxMIF/HSG Bispecific Antibody ON105 Results in Significant Tumor Regression in Murine Models of Cancer

Radioimmunotherapy (RIT) holds promise for the treatment of localized and diffuse tumors. This technique employs monoclonal antibodies (mAbs) to deliver radionuclides to target antigens on cancerous cells or in the tumor microenvironment. RIT has achieved success in treating hematologic malignancies, leading to FDA/EMA approval for two radiolabeled CD20-targeting mouse monoclonal antibodies for non–Hodgkin lymphoma: ¹³¹I-tositumomab (Bexxar®, GlaxoSmithKline) and ⁹⁰Y-ibritumomab tiuxetan (Zevalin, Biogen Idec; ref. 1). However, its efficacy in solid tumor treatment has seen limited success, with only a few clinical trials advancing beyond phase II, and no EMA or FDA approved radioimmunoconjugate to date (2–4).

Unlike hematologic malignancies, solid tumors are less sensitive to radiation and demand significantly higher radiation doses to induce antitumor effects (5). This is a significant hurdle because elevated radiation doses are often related to unacceptable toxicity due to radiation-induced damage to healthy tissues. On the one hand, directly radiolabeled human antibodies used for conventional RIT clear slowly from circulation thereby exposing the bone marrow and other organs to continuous radiation, which can easily lead to dose-limiting myelotoxicity and tissue damage (6). On the other hand, smaller antibody formats clear faster from the circulation, but they also generally achieve lower tumor uptake and retention and may exhibit unfavorable renal accumulation (7). Thus, the pharmacokinetics and biodistribution of antibodies can make it hard to achieve sufficiently high therapeutic indexes to deliver curative radiation doses to solid tumors safely. Typically, a minimum tumor-absorbed dose of 50 Gy is considered necessary to achieve clinical benefits, a threshold largely unmet in most RIT studies targeting solid tumors (4, 5).

One strategy to reduce off-target irradiation and enhance efficacy is pretargeted RIT (PRIT; ref. 8). This sequential method involves the administration of a tumor-targeting bispecific antibody (or an alternative scaffold) prior to introducing a small, fast-clearing radioligand that exhibits high affinity for the bispecific antibody (bsAb). The delay between administrations allows the bsAb to accumulate within the tumor and to clear from circulation. Due to its small size, the radioligand can efficiently infiltrate solid tumors, binding to pretargeted antibodies, whereas unbound radioligand molecules are quickly removed from the body via renal excretion. By delaying the administration of the cytotoxic radioligand until the bsAb has substantially cleared from the circulation, radiation deposition is primarily localized to the tumor. As a result, pretargeting minimizes systemic radiation exposure and achieves high tumor-to-nontumor ratios thereby increasing the therapeutic index compared with conventional RIT. Consequently, this approach offers the potential to safely escalate administered radiation doses, as the majority will be selectively directed to the targeted tumor or quickly eliminated through renal excretion (9).

One notable pretargeting system that has undergone clinical evaluation involves a bispecific antibody–like molecule targeting both a tumor-associated antigen and the hapten histamine-succinyl-glycine (HSG). In this system, the bispecific targeting entity consists of three Fab domains artificially extended with self-assembling domains from the human type II A-kinase and the human A-kinase anchor protein, which allow the formation of a stable, covalently linked tri-Fab molecule (10). The radioligand (IMP288; Supplementary Fig. S1A) consists of a small peptide modified with two HSG moieties and conjugated with the chelator DOTA (1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid). In conjunction with the anti-CEA (carcinoembryonic antigen) x anti-HSG bispecific tri-Fab TF2, this DOTA-di-HSG peptide loaded with ¹⁷⁷Lu or ⁶⁸Ga has been investigated in patients with solid tumors for therapeutic or imaging purposes, respectively (11–15). Five phase I to II clinical studies showed that this pretargeting pair achieved overall selective tumor targeting and demonstrated the safety of the radioligand.

The modular pretargeting platform, PreTarg-it, described here, comprises two complementary components: a bispecific antibody specific to a tumor-associated antigen and a small HSG-based hapten-chelator conjugate that can be loaded with radioactivity for therapy or diagnosis. Here, we present the design, engineering and preclinical evaluation of our pilot PRIT pair from our PreTarg-it platform, consisting of the anti-oxidized macrophage migration inhibitory factor (oxMIF) × anti-HSG bsAb ON105, along with ¹⁷⁷Lu-labeled IMP288. OxMIF is the disease-related conformational isoform of MIF, which is generated in the tumor microenvironment (16). MIF has been described as a pleiotropic proinflammatory and protumorigenic cytokine implicated in the pathogenesis of cancer and inflammatory diseases (17–21). However, in a proinflammatory tumor microenvironment in which immune cells and ROS are present, MIF was shown to convert to oxMIF, an immunologically distinct disease-related structural isoform found in solid tumor tissues and corresponding metastases (16, 22, 23). The presence of oxMIF has been demonstrated in primary colorectal tumors, liver metastases, squamous cell carcinoma of the lung, pancreatic ductal adenocarcinoma (PDAC), and ovarian cancer (22).

A first-generation anti-oxMIF antibody, imalumab, was investigated in a phase I trial (NCT01765790) in patients with advanced solid tumors, demonstrating an acceptable safety profile and some antitumor activity (24). Recently, Rossmueller and colleagues introduced the second-generation anti-oxMIF antibody, ON203, with significantly improved physicochemical properties, including reduced surface hydrophobicity and aggregation propensity as well as strongly enhanced tumor accretion and retention, compared with imalumab (25). The bsAb ON105, presented here, features the improved oxMIF-binding Fab from ON203, alongside a single-chain variable fragment (scFv) newly designed based on HSG-binding variable domains (26). In this study, we evaluate a pretargeted intervention with ON105 and ¹⁷⁷Lu-loaded IMP288 in murine models of colorectal and pancreatic cancer. Our proof-of-concept studies show that the pretargeted treatment regimen is well tolerated and exerts potent therapeutic efficacy, leading to tumor regression or tumor growth arrest and significant survival benefit.

The sequence for the anti-oxMIF Fab was retrieved from monoclonal anti-oxMIF antibody ON203, which was recently found effective in murine xenograft models of solid tumors (25). For the anti-HSG arm, the variable region sequences of the light chain (LC) and heavy chain (HC) were newly designed based on HSG-binding variable domains (26). The structure of the antibody chains is shown in Supplementary Fig. S2A. The cDNAs of the bsAb full-length HCs and LC were de novo gene synthesized at GeneArt/Thermo Fisher Scientific (with sequence codon optimization for Cricetulus griseus) and subcloned into the pcDNA 3.4 TOPO mammalian expression vector (Invitrogen, catalog no. A14697) in-frame with the N-terminal signal peptide sequence METDTLLLWVLLLWVPGSTG.

The antibody was expressed using the ExpiCHO Expression system (Thermo Fisher Scientific, catalog no. A29133, RRID:CVCL 5J31) for transient protein expression. The manufacturer’s expression max titer protocol was strictly followed. The supernatants were harvested after 13 to 14 days of production by centrifugation, and the mAb was purified by Protein A chromatography using MabSelect Prism A columns (Cytiva, catalog no. 17549854, RRID:SCR_023581) with 25 mmol/L Tris, 25 mmol/L NaCl buffer (pH 7.1) for equilibration and 100 mmol/L glycine, and 50 mmol/L NaCl (pH 3.2) for elution. The eluates were immediately neutralized by adding 500 mmol/L 2-(N-morpholino) ethanesulfonic acid (MES; pH 5.8). As a polishing step, ON105 samples were loaded onto POROS GoPure XS prepacked cation exchange columns (Thermo Fisher Scientific, cat. no. 4481317, RRID:SCR_008452) and eluted stepwise by 50 mmol/L MES buffer (pH 6.0) containing NaCl. The antibody samples were buffer exchanged and sterile filtered before storing them at −20°C until further use.

Purified mAbs were analyzed on an ENrich SEC 650 10 × 300 column (Bio-Rad, cat. no. 7801650) with a mobile phase of 0.1 mol/L sodium phosphate, 0.2 mol/L arginine (pH 6.8) at a flow rate of 1.25 mL/minutes. Elution was monitored by measuring absorbance at 280 nm.

The anti-oxMIF mAb ELISA was performed as described previously (25).

A measure of 100 µL/well recombinant MIF [OncoOne, expressed and purified as described previously (25)] at 1 µg/mL diluted in DPBS (Thermo Fisher Scientific) was immobilized on Nunc MaxiSorp 96-well plates (Thermo Fisher Scientific) for 1 hour at room temperature (immobilization leads to a conformational change in MIF structure to expose anti-oxMIF mAb epitopes (27) and therefore functions as oxMIF surrogate). Wells were blocked with 250 µL 2% fish gelatine (w/v; Sigma-Aldrich) in TBST (50 mmol/L Tris, 150 mmol/L NaCl, 0.1% Tween20 (v/v), pH 7.5; referred to as dilution buffer) overnight at 4°C. Test antibodies in dilution buffer were added to the plates at 81 nmol/L and incubated at room temperature for 1 hour. Plates were washed with TBST. Biotinylated IMP288 (Supplementary Fig. S1B) serially diluted in dilution buffer was added to the plates and incubated at room temperature for 1 hour. Bound biotinylated IMP288 was detected by incubation with streptavidin-poly-conjugated horseradish peroxidase (Thermo Scientific, cat. no. 21140) diluted at 50 ng/mL in TBST and tetramethylbenzidine (Thermo Fisher Scientific) as substrate. The reaction was stopped with 50 µL/well 30% sulfuric acid (v/v), and the absorbance was measured at 450 and 650 nm using a Tecan Infinite M200 PRO microplate reader (RRID:SCR_019033). For data analysis, the absorbance values at 650 nm were subtracted from the values at 450 nm.

The surface plasmon resonance experiments were performed using a Biacore 3000 Instrument (Cytiva, RRID:SCR_018044) equipped with an SA sensor chip (the surface consists of a carboxymethylated dextran matrix preimmobilized with streptavidin; Cytiva, cat. no. BR100398). The ligand (Biotin-PEG₄-D-Tyr-D-Lys(HSG)-D-Glu-D-Lys(HSG)-NH₂, 1540 Da, >95% pure based on HPLC; Supplementary Fig. S1B), synthetized by Eurogentec (Belgium), was bound to the sensor chip by noncovalent capturing at a density of 416 RU. To collect kinetic binding data, the analyte (ON105, 124 kDa, >95% purity based on SEC) in 10 mmol/L HBS-EP buffer (HEPES, 150 mmol/L NaCl, 3 mmol/L EDTA, and 0.005% v/v P20, pH 7.4) was injected at concentrations of 167, 84, 42, and 21 mmol/L at a flow rate of 30 μL/minute and at a temperature of 25°C. The complex was allowed to associate and dissociate for 200 and 700 seconds, respectively. The surfaces were regenerated with a 60-second injection of 4 mol/L MgCl₂. Background binding control was performed after each binding cycle using antibody free HBS-EP buffer as analyte. Data were collected at a rate of 1 Hz. The background-subtracted data were fit to a Langmuir 1:1 interaction model using the BIAevaluation 4.1 software (RRID:SCR_015936). Fitted curves (global and local fitting) with χ² values <5% of R_max were used for the calculation of kinetic constants.

IMP288 [DOTA-D-Tyr-D-Lys(HSG)-D-Glu-DLys(HSG)-NH2, (Supplementary Fig. S1A)] is a hapten peptide that contains two HSG moieties (28, 29), and another moiety capable of stable binding of a radionuclide, in this case 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA), for binding ⁹⁰Y, ¹⁷⁷Lu, and ¹¹¹In (30). In our study, IMP288 (molecular weight 1,452 Da) was synthesized and provided by Genepep, whereby a certificate of analysis from this manufacturer confirmed identity by MS and a purity of >97% (reverse-phase UPLC). The compound was radiolabeled with ¹⁷⁷Lu at a specific activity of 185 MBq/nmol as described below:

A measure of 370 MBq of ¹⁷⁷Lu chloride in 0.04 mol/L HCl (provided by ITM as EndolucinBeta; specific activity: >1,400 GBq/mg) were added to 1.97 nmol of IMP288 in 500 mmol/L MES (pH 5.5). After 15 minutes of incubation at 95°C, about 100 nmol diethylenetriaminepentaacetic acid (DTPA-Na₃) was added to the labeling reaction to chelate unbound ¹⁷⁷Lu. Lastly, NaCl and L-ascorbate were added to the solution at a final concentration of 0.9% and 10 g/L, respectively. Preparations for i.v. injection were formulated in 0.9% NaCl, 10 g/L L-ascorbate, 117 nmol/mL DTPA, and 500 µg/mL BSA. The final radioactive concentration was about 410 MBq/mL. Radiochemical purity of the preparations was determined by instant thin-layer chromatography (ITLC) and reverse-phase high-pressure liquid chromatography (RP-HPLC). ITLC was performed using 0.15 mol/L ammonium acetate (pH 5.5) in methanol (1:1 v/v) as the mobile phase and binderless, glass microfiber chromatography paper impregnated with silica gel (iTLC-SG, Agilent, cat. no. SGI001). After elution, ITLC strips were exposed on a phosphor screen (MS, BAS-IP, Fujifilm 2025) and revealed using Typhoon IP and associated software ImageQuant TL version 8.2 (RRID:SCR_014246). Final radiochemical purity was >90% (Supplementary Fig. S3A). Free ¹⁷⁷Lu and ¹⁷⁷Lu-DTPA (preincubated in the labeling buffer) were analyzed by the same systems for comparison. For RP-HPLC, the C18 column (Zorbax Rx-C18 3.5 µm 4.6 × 250 mm column; Agilent) was eluted with a linear gradient starting with a 99:1 (v/v) mixture of 0.1% trifluoroacetic acid in water and 0.1% TFA in acetonitrile solution and finishing after 30 minutes at a flow rate of 1 mL/minutes with a 5:95 (v/v) mixture of 0.1% trifluoroacetic acid in water and 0.1% TFA in acetonitrile solution. The radiochemical purity of all preparations exceeded 99% (Supplementary Fig. S3B).

A measure of 84 pmol of ¹⁷⁷Lu-IMP288 (220 MBq/nmol and 1.68 nmol/mL) was mixed with 10-fold molar excess of ON105 (840 pmol) and then incubated under gentle stirring at 37°C. The sample was analyzed by ITLC as described in the previous section or by size exclusion HPLC on a Superdex 200 Increase 10/300 GL column equilibrated in PBS (pH 7.4), coupled to a radioactivity detector and operated at a flow rate of 0.6 mL/minutes.

CT26 (RRID:CVCL_7254) mouse colorectal carcinoma cell line established from a female Balb/c mouse was purchased from ATCC (ATCC-CRL-2638). Cells were maintained at 37°C/5% CO₂ in RPMI1640 medium (Gibco, cat. no. 42401042) supplemented with 10% heat-inactivated FBS (Gibco, cat. no. 16140071), 2 mmol/L L-glutamine (Gibco, cat. no. 25030081), 100 U/mL penicillin, and 100 mg/mL streptomycin (Gibco, cat. no. 15140122). Subconfluent cultures (70%–80%) were split into new tissue culture flasks 1:2 to 1:10, that is, seeding at 1 to 5 × 10⁴ cells/cm², using TrypLE Express Enzyme (Thermo Fisher Scientific, cat. no. 12605010) twice a week. CFPAC-1 (RRID:CVCL_1119) pancreatic adenocarcinoma established from a 26-year-old male with a cystic fibrosis (liver metastasis) was purchased from ATCC (ATCC-CRL-1918). Cells were maintained at 37°C/5% CO₂ in Iscove’s modified Dulbecco’s medium (Thermo Fisher Scientific, cat. no. 12440053) supplemented with 10% heat-inactivated FBS (Gibco, cat. no. 16140071), 100 U/mL penicillin, and 100 mg/mL streptomycin (Gibco, cat. no. 15140122). Subconfluent cultures (70%–80%) were split into new tissue culture flasks 1:3 to 1:10, that is, seeding at 1 to 5 × 10⁴ cells/cm², using TrypLE Express Enzyme (Thermo Fisher Scientific, cat. no. 12605010) twice a week. Cells were routinely tested for Mycoplasma species by PCR at the respective cell bank collection (ATCC), at Charles River Laboratories Germany GmbH or at /Porsolt, and were used in experiments between passages 6 and 10. Cell morphology was confirmed by microscopy.

Presence of oxMIF on the cell surface of murine carcinoma cell line CT26 or of human PDAC cell line CFPAC-1 was assessed by flow cytometry. Cells were dislodged from flasks by treatment with TrypLE Express Enzyme (Thermo Fisher, cat. no. 12605010) and were stained with ON105 or an irrelevant human IgG (palivizumab, AbbVie, cat. no. PZN-10974950) at 300 to 18.8 nmol/L in PBS supplemented with 5% FBS (PBS/5%FBS) for 45 minutes at 4°C. After washing with PBS/5%FBS, cells were stained with a polyclonal goat anti-human IgG (H + L) Alexa Fluor488–conjugated antibody (Thermo Fisher Scientific, RRID:AB_2534080) diluted 1:100 in PBS/5%FBS for 40 minutes at 4°C for the detection of bound mAbs. Cells were next washed with PBS and stained with eBioscience Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific, cat. no. 65-0865-14) diluted in PBS (1:2,000) for 20 minutes at 4°C. After washing with PBS/5%FBS, cells were measured on a CytoFlex-S flow cytometer (Beckman Coulter, RRID:SCR_019627), and data were analyzed by using FlowJo software (BD, RRID:SCR_008520).

This animal study was carried out at Charles River Laboratories Germany in strict accordance with the recommendations of the GV SOLAS in an Association for Assessment and Accreditation of Laboratory Animal Care International–accredited animal facility, and it was approved by the Committee on the Ethics of Animal Experiments of the regional council (permit number: G17/78) in Germany.

Pharmacokinetics and tumor uptake of ON105 were investigated in female BALB/c mice (age 9–11 weeks) with subcutaneous CT26 tumors. ON105 was labeled with IRDye 800CW using the IRDye 800CW Protein labeling kit (high MW, LI-COR Biosciences, cat. no. 929-70020) following the manufacturer’s instructions. The antibody concentration and the degree of labeling were determined by UV–visible spectroscopy. Mice received a unilateral subcutaneous injection of 3 × 10⁵ CT26 cells in 100 µL PBS under isoflurane anesthesia. Upon reaching the tumor volumes of approximately 150 to 300 mm³, three mice received each a single i.v. dose of IRDye 800CW-labeled ON105 at a dose of 5 mg/kg. Two mice received no injection and served as control. Mice were imaged under isoflurane anesthesia at 1, 8, 24, 48, 72, 96, and 168 hours after dosing using a LI-COR Pearl Trilogy imaging device (RRID:SCR_013430), except for one of the treated mice, which was imaged only until the 72-hour timepoint and thereafter sacrificed for the preparation of an FFPE tumor sample. Imaging was performed at 785 nm excitation and 820 nm emission wavelength. Image analysis was performed in the LI-COR Pearl Trilogy Software. Regions of interest (ROI) were drawn on the image to quantify the relative fluorescence per mm² (RFU/mm²) of the antibody in the tumor. The mean tumor RFU/mm² was plotted against imaging time. The Bateman equation was fitted to these data. In addition, 20 µL of blood was collected 6, 8, 24, 48, 96, and 168 hours postinjection by tail vein puncture. Blood was transferred into K₂-EDTA-coated vials containing 60 µL PBS and, after centrifugation, the supernatants (plasma diluted 1:4) were analyzed at OncoOne by a quantitative anti-oxMIF mAb ELISA (final plasma dilution 1:1,000–1:10,000) according to the oxMIF-binding ELISA method. The mean concentration of the antibodies per time point was plotted against collection time and half-lives were calculated by a two-phase decay model [SpanFast = (Y0-Plateau)^∗PercentFast^∗.01; SpanSlow = (Y0-Plateau)^∗(100-PercentFast)^∗.01; Y = Plateau + SpanFast^∗exp(-KFast^∗X) + SpanSlow^∗exp(-KSlow^∗X), in which Y is the antibody concentration, X is the collection time, KFast and KSlow are the two rate constants, expressed in reciprocal of the X, and Plateau the Y value at infinite time X] in GraphPad Prism (version 10; RRID:SCR_002798), using least squares regression and relative weighting, and constraining the Plateau to zero and the Y value at X = 0 to 155 µg/mL (theoretical c_max).

These animal experiments were performed at Chelatec in strict accordance with the European Union directive 2010/63/EU. All experimental procedures were approved by The French Ministry of Higher Education, Research and French local ethical committee C2EA-06 (Comité d’Ethique en Expérimentation Animale des Pays de la Loire).

In total, 1 × 10⁶ CT26 cells resuspended in 0.1 mL PBS were subcutaneously implanted in the right flank of 7- to 8-week-old female Balb/c ByJ mice (18–22 g). Tumor volumes were determined by measuring the length, the width and the depth of the tumor with a digital caliper and using the equation tumor volume = (length × width × depth) × 0.5. Once tumors had reached a volume of around 100 mm³, mice were randomized into 4 groups (10 mice per group), ensuring all cohorts had approximately equal average tumor volumes. Mice in the two treatment groups received ON105 at either 1 or 5 mg/kg (approximately 0.17 or 0.82 nmol per mouse). Three days later, these mice were given ¹⁷⁷Lu-IMP288 (radiolabeled at 185 MBq/nmol). The injected dose of ¹⁷⁷Lu-IMP288 was one 10th molar equivalents of the injected antibody dose, equivalent to an injected radioactive dose of either 150 or 750 MBq/kg (roughly 3.7 and 18.5 MBq per mouse, respectively). The two control groups received IMP288 vehicle buffer and 37 MBq ¹⁷⁷Lu-IMP288, respectively. Injections were done via the retro-orbital venous. All animals were monitored for 21–22 days following antibody administration, with body weight and tumor volume measured every 2 to 3 days. The animals whose tumors exceeded 1,100 mm³ were sacrificed.

The CT26 syngraft study was repeated with three modifications: (i) the pretargeting interval was extended to 5 days, (ii) the injected antibody doses were 2.5 and 5.0 mg/kg (accordingly, the injected radioactive doses were 375 and 750 MBq/kg, respectively), and (iii) the endpoint value for tumor volume was increased to 1,500 mm³.

This animal experiment was performed at Chelatec in strict accordance with the European Union directive 2010/63/EU. All experimental procedures were approved by The French Ministry of Higher Education, Research and French local ethical committee C2EA-06 (Comité d’Ethique en Expérimentation Animale des Pays de la Loire).

A measure of 5 × 10⁶ CFPAC-1 cells resuspended in 0.1 mL PBS were subcutaneously implanted in the right flank of 7 to 8-week-old female Balb/c nude (ByAnNRj-Foxn1^nu/nu) mice (17–22 g). Once tumors had reached a volume of 150 to 300 mm³, mice were randomized into 3 groups (10 mice per group). Mice in the two treatment groups received ON105 at either 1 or 5 mg/kg. Five days later, these mice were given ¹⁷⁷Lu-IMP288 (radiolabeled at 250 MBq/nmol). The injected dose of ¹⁷⁷Lu-IMP288 was one 10th molar equivalents of the injected antibody dose, which equaled an injected radioactive dose of either 150 or 750 MBq/kg. The control group received vehicle buffer instead of ¹⁷⁷Lu-IMP288. Injections were done via the retro-orbital venous. The animals were monitored for 28 days following antibody administration, with body weight and tumor volume (three-dimensional with digital caliper) measured every 2 to 3 days. The animals whose tumors exceeded 1,500 mm³ were euthanized.

Missing values from euthanized mice were imputed using the Last Observation Carried Forward method. Longitudinal analysis of tumor growth was carried out by linear mixed-effects modeling and type II ANOVA using the TumGrowth web tool (https://kroemerlab.shinyapps.io/TumGrowth/; ref. 31). Differences in the slopes between the control group(s) and all other study cohorts were analyzed by Tukey’s multiple comparisons test, and P values were adjusted with the Holm’s method. Cross-sectional analysis of tumor volumes across cohorts was performed with the same web tool. Pairwise comparisons of the control group(s) with all other study cohorts were done from day 5 or day 11 postinjection of ¹⁷⁷Lu-IMP288 for the syngraft or the xenograft studies, respectively. Analysis was performed with Wilcoxon rank sum test, with P values adjusted by the Holm’s method. Survival data were analyzed by Mantel–Cox log-rank test using GraphPad Prism (version 10; RRID:SCR_002798).

Data generated in this study are supplied within this article and supplementary data files. Raw data are available on request from the corresponding author.

The bispecific antibody ON105 (Fab-scFv-Fc; Fig. 1A; Supplementary Fig. S2A) consists of three different polypeptide chains, and features (i) an oxMIF-targeting Fab (25), (ii) an HSG-binding single-chain variable fragment (scFv), and (iii) a human Fc of the IgG1 isotype modified to significantly reduce binding to Fcγ receptors and to complement (32–34), and knob-into-hole mutations for HC heterodimerization (35).

ON105 was transiently expressed in ExpiCHO cells and purified from the cell culture supernatant by protein A affinity chromatography and cation exchange chromatography. The purity of ON105 was confirmed through SDS-PAGE and SEC (Supplementary Fig. S2). Under reducing conditions, the three constituent antibody chains migrated each as a discrete band and their relative intensities showed a molar ratio of 1:1:1, consistent with correct antibody assembly (Supplementary Fig. S2B). SEC chromatogram revealed one predominant peak (98.8%; Supplementary Fig. S2C) with an approximate molecular weight of 142 kDa, in line with the theoretical molecular weight of ON105 without glycosylation (124 kDa).

The affinity of ON105 to HSG was assessed by surface plasmon resonance spectroscopy using a biotinylated version of IMP288, in which the DOTA had been replaced with a biotin moiety, linked to the peptide core of IMP288 via a PEG₄ linker (Supplementary Fig. S1B). Surface plasmon resonance spectroscopy revealed a high affinity of ON105 to immobilized IMP288-biotin, with a k_on (±SD) of 7.2 × 10⁴ (±9.5 × 10³) mol/L⁻¹s⁻¹, a k_off (±SD) of 3.5 × 10⁻⁴ (±3.8 × 10⁻⁵) s⁻¹ and a K_D (±SD) of 5.1 (±0.9) nmol/L (Fig. 1B). Binding of ON105 to ¹⁷⁷Lu-IMP288 was further confirmed by ITLC (Fig. 1E; Supplementary Fig. S4A) and by size exclusion HPLC (Supplementary Fig. S4B).

The binding to the tumor-associated antigen oxMIF was evaluated by direct-binding ELISA to plate-coated oxMIF (Fig. 1C). The apparent binding affinity of ON105 (EC₅₀ 2.6 nmol/L) was in line with an internal benchmark antibody for monovalent binding to oxMIF.

The ability of ON105 to bind to HSG and to oxMIF simultaneously was examined by a bridging ELISA, in which it was shown that ON105 bound to both plate-coated oxMIF and in-solution IMP288-biotin at the same time (Fig. 1D). The internal negative control C0123—with one Fab arm specific for oxMIF and a second arm with irrelevant specificity (Supplementary Fig. S5)—did not show any binding in this setting.

In summary, the individual target-binding arms of ON105 (anti-HSG and anti-oxMIF) retain the high binding affinity of their respective parental arms, and both arms can be occupied simultaneously by their respective targets.

Biodistribution and pharmacokinetics (PK) of ON105 was evaluated in a murine syngraft model of CT26 colorectal cancer known to produce oxMIF (Supplementary Fig. S6A and S6B) with the aim of assessing its biodistribution and retention in the subcutaneous tumor mass and determining the clearance rate of the antibody from the circulation. Mice received a single i.v. injection of 5 mg/kg IRDye 800CW-labeled ON105 (dye-to-antibody ratio: 1.5:1) when the tumors had reached a volume of about 140 mm³. The intratumoral accumulation of ON105 peaked at about 8 hours postinjection, and tumor retention was evident until at least 7 days postinjection (Fig. 2A). The clearance of the antibody from the circulation followed a two-phase decay, with an initial half-life of 1.4 hours and a terminal half-life of 55.5 hours (Fig. 2B).

Efficacy of our PRIT approach was evaluated in two syngraft murine models of colorectal cancer (schematic in Fig. 3A and E). Previous pretargeted interventions have delayed the administration of the radiolabeled hapten until the antibody in the circulation had decreased to about 1% to 3% of its initial concentration or lower (14, 36, 37). In our PK study, plasma concentration of ON105 dropped from 1% to 3% 72 to 120 hours (3–5 days) postinjection of the theoretical maximum plasma concentration of c_max = 155 µg/mL. Thus, we followed this paradigm and tested two pretargeting intervals: 72 and 120 hours. The molar ratio of injected ON105 to injected IMP288 was 10:1 in all treatment groups. This ratio was found appropriate in a similar study in mice testing a different antibody (38) and is within the range of molar ratios implemented in clinical studies evaluating two-step pretargeting interventions with IMP288 (11, 14). Control groups received either vehicle (buffer of ¹⁷⁷Lu-IMP288) or 37 MBq ¹⁷⁷Lu-IMP288 without prior antibody administration.

The tumor growth curves for both the treated and control groups are shown in Fig. 3B and F. At the administered dose of 37 MBq, ¹⁷⁷Lu-IMP288 exerted no effect in the absence of ON105 and was identical to the vehicle group (Fig. 3B). This is fully in line with prior experience demonstrating that the radiolabeled hapten is rapidly eliminated from the body through renal excretion (39).

About the treatment groups, a clear dose-dependent effect on tumor growth was evident. The highest treatment dose (5 mg/kg ON105 in combination with 18.5 MBq ¹⁷⁷Lu-IMP288) led to significant tumor regression when administered with a pretargeting interval of 3 days and to significant tumor growth inhibition when applied with a 5-day pretargeting interval, whereas the lower treatment doses (1 and 2.5 mg/kg ON105 in combination with 3.7 or 9.2 MBq, respectively) slowed down tumor growth. This antitumor effect translated into a significant survival benefit for all ON105-treated groups (Fig. 3C and G). In particular, the treatment with 5 mg/kg ON105 in combination with 18.5 MBq ¹⁷⁷Lu-IMP288 (administered with either pretargeting interval) proved especially effective. In the 3-day study, tumors in this treatment cohort experienced a 50% reduction in volume by day 17 from its maximum on day 7 and subsequently remained stable until at least day 22 p.i., at the end of the observation period. The treatment resulted in 100% survival, as opposed to the control groups, in which all mice had been sacrificed between day 8 and 16 because of excessive tumor burden. When the pretargeting interval was extended to 5 days, tumors in the control group had tripled in volume by day 7 and all mice in this group had been euthanized by day 12 because of excessive tumor burden. By contrast, the highest treatment dose (5 mg/kg ON105 in combination with 18.5 MBq ¹⁷⁷Lu-IMP288) arrested tumor growth on day 10 and thereafter remained stable until the end of the observation period on day 21, with two mice showing further tumor regression. Tumor growth was accompanied by a modest increase in body weight. This is evident from the control groups, in which body weight and tumor size increased concomitantly. The treatments were well tolerated, suggesting no acute toxicity at doses equal or lower than 5 mg/kg ON105 and 18.5 MBq ¹⁷⁷Lu-IMP288 (Fig. 3D and H). Only minor average body weight losses were noted in the treatment cohorts, but this decline never exceeded 5% from the baseline and it was transient, with all mice regaining their baseline body weight level prior to the end of the study.

Subsequently, the efficacy of the two-step PreTarg-it regimen was evaluated in a murine xenograft model of human pancreatic adenocarcinoma, in which the findings observed in the CT26 colorectal cancer syngraft studies were recapitulated (Fig. 4). The selected cell line, CFPAC-1, was verified to present oxMIF on its surface (Supplementary Fig. S6C and S6D). Mice bearing CFPAC-1-derived subcutaneous tumors first received a single i.v. injection of ON105 (either 1.0 or 5.0 mg/kg) followed 5 days later by 3.7 or 18.5 MBq ¹⁷⁷Lu-IMP288, respectively. The intervention had a dose-dependent impact on tumor progression, wherein the highest dose significantly arrested tumor growth and the low treatment dose slowed down tumor progression (Fig. 4B). Given the very slow tumor growth, even in the absence of treatment, only four mice had to be removed from the study because of excessively large tumors. At any dose, the treatment resulted in a significant survival benefit (Fig. 4C). In addition, it was well tolerated, as no overt signs of toxicity were noted (Fig. 4D).

In this study, we introduce the design, engineering, and preclinical assessment of our pilot PRIT pair from our PreTarg-it platform. This pair comprises ON105, a bispecific monoclonal antibody selective for HSG and for oxMIF, and the clinically validated ¹⁷⁷Lu-labeled DOTA-di-HSG peptide IMP288. Compared with directly radiolabeled antibodies, PRIT decouples the administration of the tumor-targeting entity (the antibody) from the delivery of the radioactive effector compound. This makes it possible to delay the administration of the radioligand until the antibody has cleared from the circulation and accumulated in the tumor. In addition, PRIT employs small, fast-clearing radioligands that are quickly eliminated by renal excretion. Consequently, pretargeting can significantly minimize background radiation deposition and enhance the therapeutic index (9).

Although various pretargeting technologies have been developed, none has yet gained approval for human use. A therapeutic strategy comprising an anti-DOTA × anti-disialoganglioside GD2 fusion antibody and ¹⁷⁷Lu-loaded DOTA is currently in phase I (NCT05130255). This antibody consists of two tandem single-chain variable domains fused to a self-assembling and disassembling domain (40). Also, in phase I clinical investigation for the first time is a pretargeting pair relying on click chemistry for in vivo ligation (NCT05737615). The pretargeting pair consisting of the tri-Fab TF2 and the DOTA-di-HSG peptide IMP288 has already undergone extensive clinical scrutiny, with five trials up to phase II demonstrating overall selective tumor targeting (11–15). Moreover, the radiolabeled IMP288 was found to clear quickly from the circulation, with serum initial and terminal half-lives of ∼3 and ∼29 hours, respectively, and to undergo rapid elimination via renal excretion (11). Nevertheless, a significant number of patients showed high neutralizing anti-tri-Fab TF2 antibody titers after TF2 injection (11–14).

Two alternative approaches evaluated in murine models consist of a tetravalent bsAb in which both LCs have been extended with a DOTA-binding scFv (41), or an antibody in which Fc has been extended with a V_H and a V_L domain, enabling subpicomolar binding to lead-loaded DOTAM (42). Both antibodies have been tested in three-step pretargeting schemes, utilizing a clearing agent to rapidly reduce the level of the pretargeting antibody from the circulation before radioligand administration. Although this represents a valid strategy, it involves the development and clinical validation of an additional compound, adding complexity to an already intricate regimen.

Our approach is based on a compact, 1 + 1 human Fab-scFv IgG1 format with knob-into-hole Fc modifications for HC heterodimerization, as found in various bispecific antibodies currently in clinical trials [vibecotamab (43), zanidatamab (44)]. This conventional format should minimize the risk of immunogenicity. Our antibody relies on a Fab for the monovalent binding to the tumor target with single digit nanomolar affinity. The hapten-binding scFv uses a generic flexible glycine-serine linker, and our data show that it features high affinity to HSG. The Fc incorporates validated mutations for strongly reduced binding to Fcγ receptors and to complement, which provides an additional layer of specificity and safety. In addition, our technology benefits from the “affinity enhancement system” proposed by Le Doussal and colleagues: because the radioligand features two hapten moieties, each of which can independently bind to an antibody molecule, stronger (bivalent) binding occurs at sites with high bsAb concentrations, i.e., within the tumor tissue. By contrast, lower bsAb concentrations in the blood favor weaker, monovalent interactions, which facilitates dissociation and thus, in turn, allow the renal clearance of the radioligand (28). The modularity of IMP288 is also of note: because binding to the antibody is mediated by the HSG moiety, the chelator can be loaded with various radioisotopes without affecting the binding to the antibody (30, 45).

When tested in murine models of colorectal and pancreatic cancer, our PreTarg-it approach with ON105 and ¹⁷⁷Lu-IMP288 demonstrated a potent therapeutic efficacy. The highest administered dose (5 mg/kg ON105 and 18.5 MBq ¹⁷⁷Lu-IMP288) led to significant tumor regression and halted tumor growth when applied with a pretargeting interval of 3 or 5 days, respectively, and resulted in 100% survival. In the CT26 syngraft model using the 5-day interval, efficacy was remarkable because the tumors had reached a substantial volume of ∼450 mm³ prior to the injection of the radioligand in this study. Likewise, in the CFPAC-1 model, the maximum of 100% survival was obtained for PreTarg-it treated mice.

Although a histologic examination and quantification of markers of kidney function and other potentially affected organs are necessary to evaluate the potential toxicity of our PRIT comprehensively, our proof-of-concept studies revealed no obvious side effect from the treatment in the mice carrying CFPAC-1 tumors, and only minimal acute side effects in some CT26 tumor-bearing mice that received the highest treatment dose.

A future dosimetry study could allow to evaluate tumor-to-nontumor absorbed dose ratios, providing a complementary assessment of safety and efficacy. To achieve a curative response in solid human tumors, benchmark requirements have been put forward by Larson and colleagues: an effective PRIT should aim to deliver 100 Gy to the tumor while keeping the radiation absorbed by the kidney, the bone marrow and colon mucosa below 1,500, 150 and 250 cGy, respectively. This amounts to tumor-to-nontumor absorbed dose ratios of ∼7 to 10, ∼40 to 100, and ∼40 to 60, respectively (7). In this respect, it is noteworthy to mention two clinical trials evaluating the pretargeting pair tri-Fab TF2 and IMP288. These included one pretherapeutic imaging cycle with ¹¹¹In-loaded IMP288 which served to determine the radiation delivered to critical nontarget tissues, thus making it possible to adjust the ¹⁷⁷Lu dose administered per patient in the subsequent therapeutic cycle(s) (11, 14).

The tumor target of ON105, oxMIF, is generated from the pool of extracellular, secreted MIF by locally restricted, oxidative modifications (16). Targeting of oxMIF with optimized monospecific antibody ON203 was recently found effective in murine xenograft models of solid tumors (25), and another optimized monospecific antibody, ON104, was shown to be highly disease-modifying in a collagen-induced arthritis model (46), underscoring the dual role of oxMIF in oncology and chronic inflammation and the potential of oxMIF neutralization. In our PreTarg-it approach, we utilize the strict tumor specificity of oxMIF. In contrast to conventional tumor-associated antigens, which usually are cell surface-exposed transmembrane proteins, oxMIF strongly binds to the cell surface of cancerous cells, thus providing a means for oxMIF-specific antibodies to flag these cells.

PreTarg-it is a two-component system with increased complexity in regulatory and manufacturing compared with single agents, such as monoclonal antibodies. We consider hard-to-treat cancers, such as pancreatic cancer, gastric cancer, or head-and-neck cancer, as potential applications that justify the additional effort to provide treatment options for these devastating diseases. PDAC accounts for more than 90% of pancreatic cancers and is increasing worldwide (47). Most patients diagnosed with PDAC present unresectable tumors (80%–90%). Available treatment options for these patients are limited, and the overall life expectancy has an average 5-year survival rate of less than 10% (48, 49). Immunohistochemistry data indicate that oxMIF is present in tumor tissue of PDAC and can be detected even at early stage of the disease (22). Our proof-of-concept data demonstrating strong anticancer activity in the murine models of pancreatic cancer underscore the potential of ON105 as a promising new PRIT candidate for further development. This offers a new venue to safely deliver sufficiently high radiation doses to hard-to-treat solid tumors with unmet medical need.

A.A. Puchol Tarazona reports grants from Austrian Research Promotion Agency (FFG) and from Austrian promotional bank (aws) during the conduct of the study, as well as personal fees from OncoOne Research & Development GmbH outside the submitted work. A. Schinagl reports grants from Austrian Research Promotion Agency (FFG) and Austrian promotional bank (aws) during the conduct of the study; personal fees from OncoOne Research & Development GmbH outside the submitted work; a patent for PCT/EP2022/074449 pending; and being a shareholder of OncoOne Research & Development GmbH. I. Mirkina reports grants from Austrian Research Promotion Agency (FFG) and from Austrian promotional bank (aws) during the conduct of the study, as well as personal fees from OncoOne Research & Development GmbH outside the submitted work. G. Rossmueller reports grants from Austrian Research Promotion Agency (FFG) and Austrian promotional bank (aws) during the conduct of the study, as well as personal fees from OncoOne Research & Development GmbH outside the submitted work. R.J. Kerschbaumer reports grants from Austrian Research Promotion Agency (FFG) and Austrian promotional bank (aws) during the conduct of the study; personal fees from OncoOne Research & Development GmbH outside the submitted work; a patent for PCT/EP2022/074449 pending; and being a shareholder of OncoOne Research & Development GmbH. F. Bachmann reports grants from Austrian Research Promotion Agency  (FFG) and Austrian promotional bank (aws) during the conduct of the study, as well as personal fees from OncoOne Research & Development GmbH outside the submitted work. M. Thiele reports grants from Austrian Research Promotion Agency (FFG) and the Austrian promotional bank (aws) during the conduct of the study; personal fees from OncoOne Research & Development GmbH outside the submitted work; a patent for PCT/EP2022/074449 pending; and being a shareholder of OncoOne Research & Development GmbH.

A.A. Puchol Tarazona: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing—original draft, writing—review and editing. A. Schinagl: Conceptualization, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing—review and editing. I. Mirkina: Conceptualization, formal analysis, validation, investigation, methodology, writing—review and editing. G. Rossmueller: Formal analysis, investigation, methodology. R.J. Kerschbaumer: Conceptualization, funding acquisition, project administration, writing—review and editing. F. Bachmann: Project administration, writing—review and editing. M. Thiele: Conceptualization, funding acquisition, project administration, writing—review and editing.

The authors are indebted to Verena Hoeld for her support in the recombinant expression and purification of the antibody. The authors would also like to thank biolution GmbH for the design of the graphical abstract.