Preclinical development of ABL206, a novel bispecific antibody-drug conjugate targeting B7-H3 and ROR1

ABSTRACT

Bispecific antibody-drug conjugates (BsADCs) represent a promising strategy to enhance tumor selectivity and overcome the limitations of monospecific ADCs. Here, we report the preclinical development of ABL206 (NEOK001), a novel BsADC targeting B7-H3 and ROR1, which are frequently co-expressed in various solid tumors. ABL206 was engineered in a 2 + 2 format and site-specifically conjugated with tavatecan, an exatecan-based linker-payload, utilizing GlycoConnect® technology to yield a highly homogeneous molecule with a drug-to-antibody ratio (DAR) of 4. In vitro, ABL206 demonstrated specific dual antigen binding, internalization, potent target-dependent cytotoxicity and a bystander killing effect. In vivo, ABL206 induced superior tumor regression, outperforming exact-matched monospecific ADCs at equimolar doses. Across 38 patient-derived xenograft (PDX) models spanning nine tumor types, ABL206 demonstrated broad translational potential with strong tumor regression. ABL206 consistently outperformed clinical-stage benchmark monospecific ADCs, ifinatamab deruxtecan (I-DXd) and zilovertamab vedotin, and effectively induced regression in I-DXd-pretreated regrowing tumors. Furthermore, ABL206 exhibited excellent in vitro plasma stability, a robust pharmacokinetic profile in rodents and non-human primates, and a favorable safety profile with the highest non-severely toxic dose (HNSTD) of 60 mg/kg in cynomolgus monkeys. Collectively, these data provide a comprehensive preclinical proof-of-concept for ABL206, supporting its advancement into a Phase 1 clinical trial for patients with advanced solid tumors (NCT07612176).

Introduction

Tumor heterogeneity remains a significant barrier to the clinical efficacy of targeted therapies based on monoclonal antibodies (mAbs). To overcome the limitations of monospecific targeting, bispecific antibodies (BsAbs) offer a promising approach by engaging two different, often co-expressed, tumor-associated antigens to enhance tumor specificity. Furthermore, combining this bispecific targeting strategy with a cytotoxic payload capable of a bystander effect allows for the eradication of adjacent antigen-negative tumor cells. Consequently, bispecific antibody-drug conjugates (BsADCs) have emerged as next-generation therapeutic agents well-positioned to tackle the highly heterogeneous tumor microenvironment.

Receptor tyrosine kinase-like orphan receptor 1 (ROR1) is an oncofetal protein that is minimally expressed in normal adult tissues^1,2 but broadly overexpressed in various hematological malignancies and solid tumors.^3–8 While clinical-stage anti-ROR1 ADCs, such as zilovertamab vedotin (ZV) and CS5001, have demonstrated significant clinical activity in hematological malignancies such as lymphoma, their efficacy in patients with advanced solid tumors has been limited.^9–13 For instance, the most clinically advanced candidate, ZV, yielded an overall response rate (ORR) of 1% in this patient population.¹² This clinical limitation highlights a critical need for novel targeting approaches to exploit the therapeutic potential of ROR1 in solid tumors.

B7 homolog 3 (B7-H3, CD276) is an attractive ADC target due to its limited expression in normal tissues and widespread overexpression across diverse cancer types.^14–20 In contrast to the limited efficacy of ROR1-directed agents in solid tumors, several B7-H3-targeted ADCs, mainly conjugated to topoisomerase 1 inhibitors, are currently in clinical development with highly promising outcomes. The frontrunner, ifinatamab deruxtecan (I-DXd), has shown robust objective responses in Phase 2 for solid tumors such as small cell lung carcinoma and has been advanced to a Phase 3 study.^21,22 However, alongside typical topoisomerase I inhibitor toxicities, interstitial lung disease (ILD), a known class effect of deruxtecan-based platforms often associated with off-target uptake by alveolar macrophages,²³ has been reported in over 10% of patients, presenting a significant dose-limiting safety challenge.²²

To address the limited solid tumor efficacy of ROR1-targeted ADCs and mitigate the potential side effects associated with B7-H3 ADCs, we rationally designed ABL206 (NEOK001), a novel BsADC targeting both antigens and conjugated to tavatecan via GlycoConnect® technology. In this study, we present the molecular design and comprehensive preclinical profile of ABL206. We hypothesized that this ADC design with attenuated FcγR binding would minimize off-target macrophage uptake, potentially reducing the risk of ILD,²⁴ while maximizing tumor-specific delivery. Furthermore, we demonstrate that the dual targeting of B7-H3 and ROR1 drives robust in vivo efficacy, significantly outperforming equimolar doses of exact-matched monospecific ADCs and clinical stage benchmark ADCs. We further validated its broad translational potential across an extensive panel of 38 patient-derived xenograft (PDX) models. In parallel, repeat-dose toxicology studies in non-human primates confirmed its excellent tolerability. These comprehensive data support the potential of ABL206 as a highly effective and tolerable clinical candidate, providing a strong rationale for the ongoing first-in-human Phase 1 trial.

Results

Molecular design and characterization of ABL206

To generate the constituent antibodies of ABL206, independent discovery campaigns were conducted for each target. ROR1- and B7-H3-specific single-chain Fvs (scFvs) were selected by ELISA from a synthetic human antibody phage library, using ROR1 or B7-H3 antigens.²⁵ The binding of hit clones to ROR1- or B7-H3-expressing cell lines was assessed using flow cytometry. Lead antibodies from each campaign were then advanced for bispecific format engineering.

Given that ROR1 and B7-H3 are frequently co-expressed across a broad spectrum of solid tumors (Supplementary Figure S1), dual targeting of both antigens provides a strong therapeutic rationale for maximizing tumor specificity. To realize this strategy, ABL206 was engineered as a novel bispecific antibody-drug conjugate (BsADC) targeting both ROR1 and B7-H3. Multiple formats differing in Fab/scFv stoichiometry (1 + 1, 1 + 2, 2 + 1, 2 + 2) and domain fusion orientation (IgG-scFv, scFv-IgG) were evaluated for biophysical stability, dual-target binding, and in vitro cytotoxicity activity. The 2 + 1 and 2 + 2 formats were further advanced to in vivo efficacy studies to secure the final optimal architecture. Among these, the 2 + 2 format – generated by fusing an anti-B7-H3 scFv to the C-terminus of an anti-ROR1 IgG1 – was selected to maximize bivalent avidity for both targets across heterogeneously expressing tumors. This configuration preserves the high binding capacity of the intact IgG1 to compensate for the relatively low surface expression and moderate intrinsic affinity of the anti-ROR1 domain, while incorporating the B7-H3-targeting arms as scFvs to moderate their intrinsic affinity.

Given the recent clinical success of topoisomerase 1 inhibitor–based ADCs,^26–28 together with the favorable bystander effect and tolerability associated with this payload class, we selected tavatecan, an exatecan-based topoisomerase 1 inhibitor, as the payload for our ROR1 and B7-H3-targeting ADC.²⁹ Tavatecan is an exatecan-based branched linker-payload system comprised of a bicyclononyne group, hydrophilic sulfonamide group, Val-Ala-PAB dipeptide cleavable linker, and exatecan (Figure 1A). The linker-payload was subsequently conjugated to this BsAb using GlycoConnect® technology that enables the site-specific conjugation of the linker-payload to the conserved N297 glycan site in the constant region (Fc) of the antibody heavy chain (Figure 1A).^30,31 The DAR was determined to be approximately 4, consistent with the tailorable DAR achievable using branched linker-payload constructs in GlycoConnect® technology.³¹

Figure 1.

Characterization of ABL206. A, Schematic structure of ABL206. B, Binding activity of ABL206 and its naked antibody (the naked ABL206) to human ROR1 and B7-H3. C, Species cross reactivity of ABL206 against mouse, rat and human ROR1 or mouse, rat, monkey and human B7-H3. The binding was measured by ELISA. D, in vitro ADCC assay in PA-1 cell line. Trastuzumab was used as a positive control.

Binding affinity evaluated by ELISA showed that ABL206 binds to human ROR1 and B7-H3 with high affinity, comparable to that of the naked ABL206 (Figure 1B). This interaction was quantitatively confirmed by SPR analysis, which indicated a K_D of 132.8 nM for ROR1 and 15.0 nM for B7-H3 (Supplementary Table S3). There was no significant difference in antigen binding affinity between ABL206 and the naked ABL206, confirming that the drug conjugation process does not compromise antigen binding affinity. In addition, we confirmed that ABL206 exhibited no cross-reactivity with ROR2 (Supplementary Figure S2). Next, the species cross-reactivity of BsADC was assessed. Consistent with the high sequence homology of B7-H3 and identical extracellular domain (ECD) sequences for ROR1, binding affinities for cynomolgus monkey antigens were similar to those for human antigens. In contrast, the binding affinity for rat and mouse ROR1/B7-H3 was significantly lower (Figure 1C). These results were further corroborated by SPR analysis (Supplementary Table S3).

Finally, we investigated the impact of the conjugation on Fc-mediated functions, as N297 glycosylation within the antibody Fc domain is essential for binding to Fc-gamma receptors (FcγRs) and mediating immune effector functions. By utilizing GlycoConnect® technology, ABL206 was conjugated with the payload specifically at the N297 site, and the modification could reduce these interactions. Consistent with this design, ABL206 showed negligible binding to relevant FcγRs while the naked ABL206 retained binding affinity (Supplementary Table S4). In contrast to the naked ABL206, which demonstrated robust target cell lysis, ABL206 did not induce target cell lysis in an in vitro ADCC assay (Figure 1D). This confirms that the therapeutic activity of ABL206 is independent of immune-mediated effector functions.

Target engagement and internalization of ABL206

We next assessed the cell-binding activity of ABL206 across a dual-target overexpressing cell line and nine different cancer cell lines using flow cytometry (Figure 2A). ABL206 demonstrated specific, dose-dependent binding to both ROR1 and B7-H3, with negligible binding to the target-low/negative KATOIII cell line. In the dual-antigen overexpressing CHO-huROR1-huB7H3 cells, ABL206 showed markedly higher binding than either parental Ab, maintaining comparable activity to its naked BsAb. Furthermore, across most cancer cell lines with higher expression of B7-H3 compared to ROR1, ABL206 consistently exhibited stronger binding than the parental anti-ROR1 Ab and comparable binding to the parental anti-B7-H3 Ab. These results confirm that the 2 + 2 bispecific architecture effectively engages target cells and that the site-specific payload conjugation does not compromise its antigen-binding capacity.

Figure 2.

Target binding and internalization of ABL206. A, Binding of ABL206, the naked ABL206 and each parental antibody (anti-ROR1 Ab and anti-B7-H3 Ab) to target-expressing cancer cell lines, as determined by flow cytometry. B, Real-time internalization of Fabfluor pH-labeled ABL206, the naked ABL206, each parental antibody (anti-ROR1 Ab and anti-B7-H3 Ab) and isotype antibody in HCC1187 cells measured over 48 hours using the Incucyte S3 live-cell imaging system. Internalization was quantified as the red-to-phase area ratio. C, Confocal microscopy images showing cellular localization of ABL206 in HCC1187 cells after incubation for 0, 1, and 6 hours. Nuclei were stained with DAPI (blue), lysosomes with anti–LAMP-1 (red), and ABL206 was detected using an anti-human IgG secondary antibody (green). Scale bar, 5 µm.

Following target binding, efficient cellular internalization and subsequent lysosomal trafficking are prerequisite steps for the intracellular release of the ADC payload. We evaluated the internalization kinetics of ABL206 in target-expressing cell lines using real-time live-cell imaging and confocal microscopy. Real-time Incucyte live-cell imaging analysis of Fabfluor-pH-labeled antibodies demonstrated that the bispecific design of ABL206 effectively overcame the inherently poor internalization kinetics of the parental anti-ROR1 Ab (Figure 2B). To contextualize these internalization profiles, we determined the antibody binding capacity (ABC) of ROR1 and B7-H3 on HCC1187 cells by quantitative flow cytometry. ROR1 surface expression (25,398 antibody binding sites/cell) was approximately 5.8-fold lower than B7-H3 (147,800 antibody binding sites/cell), indicating that the attenuated internalization of the parental anti-ROR1 Ab reflects lower receptor density rather than intrinsically slow endocytic kinetics of the anti-ROR1 arm. Following surface engagement, confocal microscopy revealed that ABL206 strongly colocalized with the lysosomal marker LAMP-1 within 6 hours post-treatment (Figure 2C), whereas the isotype control showed negligible intracellular uptake (Supplementary Figure S3). Collectively, these findings demonstrate that the bispecific architecture effectively overcomes the poor internalization of the ROR1 arm, facilitating the robust lysosomal delivery necessary for potent ADC activity.

In vitro cytotoxicity and bystander killing activity of ABL206

The in vitro cytotoxicity of ABL206 was evaluated across multiple target-expressing cell lines in both monolayers and 3D spheroids. All the cell lines were sensitive to exatecan, whereas the naked antibody showed no cytotoxicity (Figure 3B). ABL206 exhibited in vitro potency equivalent to the exact-matched monospecific ADCs (anti-ROR1-tavatecan or anti-B7-H3-tavatecan, generated by conjugating the identical linker-payload to the parental anti-ROR1 or anti-B7-H3 Ab at a DAR of 4, respectively) in both ROR1-only and B7-H3-only expressing cells (Figure 3A). Furthermore, ABL206 outperformed both parental ADCs in antigen-overexpressing cells, inducing potent, dose-dependent killing in dual-antigen expressing CHO-huROR1-huB7H3 (IC₅₀ = 8.8 nM, Figure 3A) and 3D PA-1 spheroids (IC₅₀ = 28.2 nM, Figure 3B). In contrast, in most cancer cell lines, ABL206 exhibited slightly better potency than the parental ROR1 ADC and comparable activity to the parental B7-H3 ADC, likely due to high basal B7-H3 expression (Figure 3B). The moderate activity and attenuated maximum cytotoxicity observed across treatment arms in these cell lines are consistent with the known limitation of 2D monolayer assays for camptothecin-based ADCs, in which minimal activity or non-sigmoidal dose–response curves are commonly observed, and 3D spheroid assays are generally considered more predictive of in vivo efficacy.³² The target contribution of this cytotoxicity was confirmed via a competitive assay using PA-1 spheroids, where the addition of recombinant ROR1 or B7-H3 antigens attenuated ABL206-mediated cell killing (Figure 3C).

Figure 3.

In vitro cytotoxicity and bystander killing effect of ABL206. A, Engineered dual-target expressing (CHO-huROR1-huB7-H3) and mono-target expressing (CHO-huROR1, KATOIII-huB7-H3) cell lines were treated with ABL206, parental ADCs (anti-ROR1 ADC or anti-B7-H3 ADC), or an isotype control ADC. B, Multiple target-expressing cancer cell lines (HCC1187, Calu-3, NCI-H446, and PA-1 spheroid) were treated with ABL206, the naked ABL206, exatecan, parental ADCs (anti-ROR1 ADC or anti-B7-H3 ADC), or isotype control ADC. Cell viability was measured using the CCK-8 assay after 6 days for monolayer cultures, and the CellTiter-Glo® 3D luminescence assay after 5 days for spheroids. C, Competitive cytotoxicity was also evaluated by co-treatment with recombinant ROR1 or B7-H3 antigens with ABL206 in PA-1 spheroids, confirming dual-target dependent activity of ABL206. D, Bystander killing effect of ABL206. The viability of GFP-positive target-negative CHO cells co-cultured with CHO-huROR1-huB7H3 cells at different ratios was quantified by measuring relative GFP area at 6 days post-treatment using Incucyte analysis. Iso; isotype control ADC, NT; no treatment.

Furthermore, because solid tumors exhibit highly heterogeneous antigen expression even within a single lesion,^33,34 we evaluated the bystander killing activity of ABL206. In co-culture assays, ABL206 significantly reduced the viability of adjacent antigen-negative CHO-GFP cells in a concentration-dependent manner, whereas minimal cytotoxicity was observed in antigen-negative cells cultured alone (Figure 3D). These findings demonstrate that ABL206 exerts bystander killing activity, potentially broadening therapeutic coverage beyond antigen-positive tumor cells.

These data suggest ABL206 retains the payload delivery capacity of each individual arm and maximizes potent cytotoxicity against dual-positive tumor cells, demonstrating strong potential to overcome intra-tumoral heterogeneity.

In vivo therapeutic superiority of bispecific targeting in tumor xenograft models

To elucidate the advantage of the bispecific format over monospecific targeting, we first performed an in vivo comparison using exact-matched ADCs. These control ADCs were constructed using the parental B7-H3 or ROR1 monoclonal antibodies, conjugated with identical tavatecan at a matched DAR of 4. Evaluation of these ADCs at equimolar doses revealed the superior antitumor efficacy of the bispecific format. While treatment with the monospecific anti-ROR1 ADC (6 mg/kg) resulted in only partial tumor inhibition and the anti-B7-H3 ADC (6 mg/kg) led to tumor stasis, ABL206 (8 mg/kg) induced robust and continuous tumor regression (Figure 4A).

Figure 4.

In vivo antitumor efficacy of ABL206 in xenograft mouse models. Tumor growth curves of the patient-derived xenograft (PDX) model of NSCLC (CTG-2803) established in athymic nude mice (n = 7/group), and cell line-derived xenograft (CDX) models of small-cell lung cancer (NCI-H446), ovarian cancer (PA-1), and human lung adenocarcinoma (calu-3) established in BALB/c nude mice (n = 8/group). Mice were treated with a single intravenous dose of ABL206, isotype control ADC, parental monospecific ADCs (anti–ROR1 ADC and anti–B7-H3 ADC), or benchmark ADCs (I-DXd and ZV) at the indicated doses. Tumor volumes were monitored for 28 or 29 days post-treatment. Due to differences in molecular weight, doses for the other ADCs were adjusted to equimolar amounts relative to ABL206. Baseline ROR1 and B7-H3 expression levels are shown as an inset of flow cytometry data for the NCI-H446 model, and as numerical immunohistochemistry (IHC) scores for the other models. Data are presented as mean ± SEM.

Having established the superiority of the bispecific design, we next benchmarked ABL206 against clinical-stage ADCs, ifinatamab deruxtecan (I-DXd) and zilovertamab vedotin (ZV), across multiple cell line-derived xenograft (CDX) models including NCI-H446, PA-1, and Calu-3. ABL206 exhibited superior and durable tumor growth inhibition compared to both I-DXd and ZV at equimolar dose levels across all tested CDX models (Figure 4B-D), underscoring its potent therapeutic profile. Throughout these in vivo studies, no significant body weight loss was observed in any treatment group (Supplementary Figure S5).

Broad translational potential in diverse patient-derived xenograft (PDX) models

To assess the translational potential and broad clinical applicability of ABL206, we conducted an extensive in vivo screening across 38 patient-derived xenograft (PDX) models spanning nine different solid tumor types. Using a 10 mg/kg screening dose to evaluate the maximal response potential across highly heterogeneous tumors, ABL206 induced significant tumor growth inhibition (TGI > 50%) in 84% (32/38) of the models and profound tumor regression below baseline in 53% (20/38) of the models (Figure 5A).

Figure 5.

In vivo antitumor efficacy of ABL206 in PDX mouse models. A, ABL206 was evaluated in 38 patient-derived xenograft (PDX) models derived from nine tumor types. Mice (n = 2–3 per group) received ABL206 (10 mg/kg, Q2W × 2 or single dosing) via intravenous administration, and tumor volume change (TVC, %) was calculated as defined in materials and methods. For models exhibiting tumor growth (TVC >0), TGI % values are displayed above the bars for those with TGI ≥50%. Note that all models demonstrating tumor regression (TVC <0) inherently meet the TGI ≥50% threshold. RNA expression levels for ROR1 and B7-H3 (log₂[TPM +1]) in each PDX model are shown as a heatmap below the bar graph. B–C, Representative tumor growth curves from PDX models treated with a single intravenous dose of ABL206 (6 mg/kg) compared with isotype control and benchmark ADCs I-DXd and ZV (4.5 mg/kg). ROR1 and B7-H3 expressions were determined by flow cytometry (gMFI) or IHC analysis. For CTG-2055 and CTG-0828 (flow cytometry): high >1,000 gMFI, med 100–1,000 gMFI, and low <100 gMFI. In the LU5264 model (C), a single dose of ABL206 (6 mg/kg) was administered on day 52 (red arrow) to the regrowing I-DXd group, resulting in tumor regression. Graphs show means of n = 7 (CTG models) or n = 5 (LU models) mice per group ± SEM.

To validate these findings at a clinically relevant dose, select PDX models were treated with a single 6 mg/kg dose of ABL206 and compared against benchmark ADCs. In these translational models, ABL206 consistently outperformed equimolar doses of I-DXd and ZV (Figure 5B). In the LU5264 PDX model, ABL206 maintained a durable complete response (CR) in all mice (5/5) up to day 38, whereas tumors in the I-DXd–treated group eventually relapsed (Figure 5C). Remarkably, when a single dose of ABL206 (6 mg/kg) was administered on day 52 to the mice bearing these I-DXd–pretreated regrowing tumors (tumor volume: 450–1760 mm³), ABL206 induced a rapid, significant tumor regression (Figure 5C and Supplementary Figure S4). No significant body weight loss was observed in any treatment group throughout the study period (Supplementary Figure S6). These data highlight the potency of ABL206 and its therapeutic promise, supporting its clinical evaluation in patients with advanced solid tumors.

In vitro stability, pharmacokinetics, and safety profile of ABL206

A critical attribute of a clinically viable ADC is robust systemic stability during circulation to prevent premature payload release and subsequent off-target toxicity. To assess linker-payload stability under physiological conditions, ABL206 was incubated in human, cynomolgus monkey, and rat plasma for up to 21 days. Payload release was subsequently measured by LC-MS, while changes in the drug-to-antibody ratio (DAR) were profiled using RP-UPLC. Payload release remained remarkably low across all matrices, with free exatecan release of only 4.0% in human plasma and 0.5% in monkey plasma on day 21 (Figure 6A). Consistent with this, DAR profiling demonstrated highly stable conjugation, with DAR distributions remaining completely unchanged across all tested matrices (Figure 6B).

Figure 6.

In vitro stability and pharmacokinetics of ABL206. A, in vitro plasma stability of ABL206 (200 µg/mL) in human, monkey, and rat plasma and in 1% BSA in PBS at 37°C for 21 days, shown as exatecan release (%). B, In vitro DAR stability of ABL206 during incubation in rat plasma and monkey plasma for 14 days, and human serum for 7 days, shown as DAR changes over time. C–D, Mean (±SD) concentration–time profiles of total antibody (solid lines) and ADC (dotted lines) in rats (C) and cynomolgus monkeys (D) following single IV bolus administration of ABL206 at 1, 3, or 10 mg/kg.

This in vitro stability translated directly to the in vivo pharmacokinetic (PK) profiles. Following a single intravenous administration at 1, 3, or 10 mg/kg in both rats and cynomolgus monkeys, ABL206 exhibited a biphasic decline with dose-dependent increases in systemic exposure (Figure 6C and 6D). Crucially, the plasma concentration-time profiles of the total antibody (TAb) and the ADC were highly parallel in both species. In cynomolgus monkeys, the terminal half-lives (T_1/2) were 71.2–126 hours for TAb and 106–163 hours for the ADC (Supplementary Table S5). Throughout the study, circulating free payload remained below the lower limit of quantification (BLQ <0.5 ng/mL) at all time points. These findings demonstrate sustained circulating ADC exposure and minimal systemic deconjugation, validating the robust in vitro and in vivo performance of the GlycoConnect® technology.

To comprehensively characterize the safety profile and establish the tolerable dose of ABL206, repeat-dose toxicity studies were conducted in rat and monkey, which are cross-reactive species (Supplementary Table S6). ABL206 was administered by intravenous infusion once every three weeks (Q3W) for a total of two doses at 80, 160, and 198 mg/kg in rats and 30, 60, and 90 mg/kg in monkeys, with dose-proportional systemic exposure confirmed by toxicokinetic monitoring throughout the study (Supplementary Figure S7). In rats, ABL206-related findings were minor such as minimal decreases in red blood cell count, hemoglobin, and hematocrit were noted at 198 mg/kg, all of which resolved during the recovery period. Therefore, 198 mg/kg was determined to be the highest non-severely toxic dose (HNSTD) in rats (Supplementary Table S6). In monkeys, ABL206-related clinical signs included intermittent abnormal feces at 90 mg/kg and skin erythema/hyperpigmentation at ≥60 mg/kg. Hematological changes including decrease in red blood cell count, hemoglobin, hematocrit, accompanied by a compensatory increase in reticulocytes and increased red cell distribution width were observed at ≥30 mg/kg (Supplementary Table S6 and Figure S8). These ABL206-related changes were dose-dependent and partially to fully reversible. Importantly, no test article-related macroscopic or microscopic findings were observed in the lungs at any dose level tested. Based on the tolerability of ABL206, the HNSTD in monkeys was determined to be 60 mg/kg.

Discussion

In this study, we detail the preclinical development and molecular engineering of ABL206, a novel bispecific antibody-drug conjugate (BsADC) targeting B7-H3 and ROR1. Despite the clinical validation of single-antigen targeting agents, the therapeutic efficacy of monospecific ADCs is frequently hindered by intra-tumoral heterogeneity and dose-limiting toxicities. ABL206 was rationally designed to overcome these challenges by employing an ROR1 or B7-H3 targeting. By utilizing a bispecific format, ABL206 not only demonstrates the potential to broaden the addressable tumor population by recognizing heterogeneously expressed targets but also achieves enhanced efficacy when both antigens are engaged, thereby significantly improving targeted payload delivery.

The fundamental advantage of this bispecific targeting strategy was demonstrated through our direct in vivo comparative studies. To date, it has been challenging to dissect whether the superior efficacy of a BsADC stems from the dual-targeting mechanism itself or merely the potency of the attached payload. By evaluating ABL206 against exact-matched monospecific ADCs carrying the identical tavatecan linker-payload, we confirmed that dual-antigen engagement at equimolar doses drives robust tumor regression, whereas monospecific blockade merely achieves tumor stasis. Consequently, ABL206 consistently outperformed clinical-stage benchmark ADCs, such as ifinatamab deruxtecan (I-DXd) and zilovertamab vedotin (ZV), across multiple cell line-derived and patient-derived xenograft (PDX) models, supporting its translational potential.

We further note that the payload chemistry itself may contribute to this difference. DXd is derived from exatecan by replacing the F-ring primary amine with a 2-hydroxyacetyl group, eliminating two hydrogen bonds that exatecan’s free amine additionally forms beyond the three shared bonds and correlating with weaker TOP1cc trapping and lower cytotoxic potency for DXd.³⁵ Consistent with this structural difference, an independent study directly comparing exatecan- and DXd-based antibody-drug conjugates in matched formats has described exatecan as a more potent topoisomerase I inhibitor than DXd, capable of overcoming resistance in tumors with low target expression or multidrug-resistance phenotypes.³⁵ We therefore cannot exclude the possibility that the intrinsic potency of the exatecan payload contributes, in addition to the bispecific targeting mechanism and potential differences in antibody pharmacokinetics, to the efficacy advantage of ABL206 over I-DXd observed in our study.

From a developability perspective, the structural homogeneity and systemic stability of an ADC are paramount. ABL206 leverages the GlycoConnect® technology for site-specific conjugation at the conserved N297 glycan, yielding a highly uniform molecule with a DAR of 4. This precise engineering translated into robust in vitro plasma stability and a highly favorable pharmacokinetic profile in both rodents and cynomolgus monkeys, characterized by sustained circulating ADC levels and minimal premature payload release. Given this robust stability in the bloodstream, the bystander killing effect of ABL206 is likely driven by the precise intracellular release of the membrane-permeable exatecan following target-mediated internalization, rather than nonspecific extracellular cleavage.

Crucially, ABL206 demonstrated a favorable safety profile in repeat-dose non-human primate toxicology studies. The observed adverse effects, including reversible hematologic and gastrointestinal changes, were entirely consistent with the known exatecan-driven toxicities.³⁶ ABL206 established the highest non-severely toxic dose (HNSTD) of 60 mg/kg in cynomolgus monkeys. This safety margin becomes particularly meaningful when contextualized with other clinical-stage ADCs. For instance, IBI-343, a CLDN18.2-targeting ADC utilizing the identical tavatecan and GlycoConnect® platform, exhibited a similar payload-driven toxicity profile with an HNSTD of 30 mg/kg in monkeys.³⁷ Furthermore, it was evaluated at doses up to 10 mg/kg in a Phase 1 human clinical trial, with 6 mg/kg determined as the recommended Phase 2 dose (RP2D).³⁸ Similarly, the benchmark I-DXd reported an HNSTD of 30 mg/kg in monkeys. While direct safety comparisons must be interpreted with caution due to differences in tested dose levels and administration schedules, achieving a two-fold higher HNSTD strongly positions the tolerability profile of ABL206 as highly promising for clinical dose escalation.

Immunohistochemical evaluation of ROR1 and B7-H3 in normal human tissue microarray cores (lung, head and neck, ovary, and breast) showed predominantly negative staining, with low-level positivity in only a minority of samples (Supplementary Table S7), consistent with previous reports of restricted normal-tissue expression of both targets.^1,2,14–20 The pharmacological relevance of cynomolgus monkey for both targets was established by ELISA cross-reactivity data (Figure 1C). Binding to the mouse and rat orthologs of ROR1 and B7-H3 was markedly lower than to human, whereas binding to the cynomolgus monkey ortholog of B7-H3 was comparable to human, consistent with 97.4% sequence identity, and was further corroborated by SPR (Supplementary Table S3). By contrast, binding to the cynomolgus monkey ortholog of ROR1 was not directly measured by either method but is inferred from the complete (100%) sequence identity of the extracellular domain between the two species. Given this pharmacological relevance, no antigen-dependent histopathological findings were observed in either rat or cynomolgus monkey at any dose level tested, consistent with the overall low normal-tissue expression of both targets. Nonetheless, as B7-H3 expression can be modulated by inflammatory cytokines in a tissue- and context-dependent manner,^39–41 its potential impact on ABL206 safety or efficacy under such conditions remains uncertain and warrants monitoring in future clinical development.

Furthermore, interstitial lung disease (ILD) remains a prominent clinical concern for topoisomerase I inhibitor-based ADCs. As demonstrated in nonclinical studies of trastuzumab deruxtecan (T-DXd), this toxicity is known to be mediated by alveolar macrophages and occurs even in the absence of target expression in the alveolar tissue.⁴² While this target-independent uptake poses a class-wide risk among topoisomerase I inhibitor-based ADCs, and the weak expression of B7-H3 in the lung presents an additional potential liability,⁴³ ABL206 was rationally designed to mitigate these challenges. Specifically, the N297 site-specific conjugation in ABL206 effectively reduces Fc-gamma receptor interactions. We hypothesize that this modification could minimize off-target binding to alveolar macrophages, thereby potentially lowering the clinical risk of ILD. In addition to the reduced Fc-FcγR interaction conferred by N297 glycan conjugation, differences in linker-payload chemistry between ABL206 and DXd-based ADCs may also contribute to the favorable pulmonary safety profile observed. Premature cleavage of the antibody-drug linker, with resultant systemic diffusion of the released payload, has been described as one of several established mechanisms underlying ADC-associated pulmonary toxicity, alongside immune-mediated inflammation and nonspecific pulmonary uptake.⁴⁴ Consistent with this, exatecan-based ADCs employing the same GlycoConnect® platform as ABL206 have shown favorable pulmonary safety in both clinical (IBI343) and preclinical (MGC028) settings.^38,45

Consistent with this rational design, no pulmonary toxicities were observed in our monkey studies. However, further clinical evaluation is warranted to confirm whether this design effectively translates to a reduced risk of ILD in humans. This hypothesis is further corroborated by emerging clinical data from IBI-343, a CLDN18.2-targeting ADC utilizing the identical GlycoConnect® platform, which has reported no incidents of ILD to date.³⁸

In conclusion, ABL206 represents a structurally optimized, highly stable, and potent BsADC that demonstrates the therapeutic superiority of dual B7-H3 and ROR1 targeting. Its comprehensive preclinical proof-of-concept, combined with a favorable safety profile, strongly supports its clinical translation. The first-in-human Phase 1 clinical trial was initiated, with the first patient dosed in 2026 (NCT07612176).

Materials and methods

Antibodies and ADCs

Antibodies, including anti-ROR1 and anti-B7-H3 monoclonal antibodies (mAbs), were produced in CHO cells and purified through affinity chromatography and additional ion exchange chromatography. The parental anti-B7-H3 mAb was generated against the scFv domain of ABL206, while the parental anti-ROR1 mAb was developed against the Fab region of ABL206, serving as parental antibodies for comparative studies. The ADCs, including anti-ROR1 ADC, anti-B7-H3 ADC, and isotype control ADC (human IgG1 with irrelevant antigen specificity), were generated based on previously described methods.³¹ Briefly, prior to conjugation, the N-glycan site on the heavy chain was enzymatically remodeled using endoglycosidase, UDP-azido sugar, alkaline phosphatase, and glycosyltransferase to introduce an azido handle. This N-glycan remodeling protocol was consistently applied across all antibodies, including ABL206 and the aforementioned anti-ROR1 and anti-B7-H3 mAbs. Site-specific ADCs were generated using metal-free click chemistry, reacting the azido-modified antibodies with tavatecan in the same manner as employed for ABL206 conjugation. This reaction yielded site-specifically conjugated species with a drug-to-antibody ratio (DAR) of 4 for all ADC variants. I-DXd (Cat# HY-P3371) and ZV (Cat# HY-P99956) used in this study were purchased from MedChem Express.

Cell lines

Human cancer cell lines were obtained from the following repositories: NCI-H446, MDA-MB-231, DU-145, PA-1, A549 and KATOIII from ATCC, Calu-3 and HCC1187 from KCLB, and NCI-H82 from DSMZ. CHO-huROR1 cells were obtained from BPS Bioscience. CHO-GFP cells were obtained from GenTarget. The dual-target positive CHO-huROR1-huB7H3 and mono-target positive KATOIII-hB7H3 cell lines were generated in-house at ABL Bio. Cells were maintained at 37°C in a humidified incubator containing 5% CO₂. All cell lines were authenticated by short tandem repeat (STR) profiling, tested negative for mycoplasma and used within 20 passages after thawing. Specific culture conditions and catalog numbers for each cell line are detailed in Supplementary Table S1.

Binding affinity and target specificity of ABL206 by ELISA

To evaluate binding affinity and specificity (including ROR2), high-binding 96-well plates were coated overnight at 4°C with recombinant antigens in phosphate-buffered saline (PBS, Gibco, Cat# 10010–023). The coating concentrations were 100 ng/well for human ROR1 (Sino Biological, Cat# 13968-H08H), human B7-H3 (R&D Systems, Cat# 2318-B3/CF), mouse ROR1 (Sino Biological, Cat# 5A0666-M08H), rat ROR1 (Acro Biosystems, Cat# RO1-R5221), and human ROR2 (Sino Biological, Cat# 16133-H08H). 20 nM for mouse B7-H3 (Sino Biological, Cat# 50973-M08H), rat B7-H3 (Sino Biological, Cat# 80380-R08H), and cynomolgus B7-H3 (Sino Biological, Cat# 90806-C08H).

After washing with 0.05% Tween-20 in PBS (PBST) using a plate washer (Tecan, Hydrospeed), the plates were blocked with 1% bovine serum albumin (BSA, Gibco, Cat# 30063–572) in PBS for 2 h at 37°C. Serial dilutions of the test antibodies were prepared in blocking buffer and added to the plates (100 µL/well), followed by incubation for 2 h at 37°C. Bound antibodies were detected using a horseradish peroxidase (HRP)-conjugated goat anti-human IgG Fc cross-adsorbed secondary antibody (1:30,000 dilution; Pierce, Cat# 31413) for 1 h at 37°C. After washing, 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (Sigma, Cat# T0440) was added, and the reactions were stopped with 0.5 N H₂SO₄ (Samchun, Cat# S1410). The absorbance at 450 nm, with a reference wavelength of 650 nm, was measured using a microplate reader (Molecular Devices, SPECTRA MAX 190).

Antibody-dependent cell-mediated cytotoxicity (ADCC) bioassay

ADCC activity was determined using the Promega ADCC Reporter Bioassay (Promega Corp., Cat# G7102) following the manufacturer’s guidelines. PA-1 target cells were plated in 96-well plates at a density of 1.5 × 10⁴ cells per well and cultured overnight at 37°C in a humidified atmosphere containing 5% CO₂. Serial dilutions of test antibodies were prepared in ADCC assay buffer (RPMI-1640 supplemented with 0.5% ultra-low IgG FBS), starting from a maximum concentration of 100 nM. Effector cells were resuspended in the same buffer and co-cultured with target cells at an effector-to-target (E:T) ratio of 10:1. After 24 h of incubation, luciferase activity was quantified using the Bio-Gl^TM reagent and measured with a PHERAstar plate reader. Dose-response relationships and EC₅₀ values were calculated by nonlinear regression using GraphPad Prism version 10.4.1.

Cell binding assay by flow cytometry

Cells suspended in FACS buffer (1% BSA in PBS) were incubated with serially diluted test antibodies for 1 h at 4°C. Following washes, cells were incubated with a FITC-conjugated anti-human Fc secondary antibody (1:500; Sigma Aldrich, Cat# F9512) for 1 h at 4°C in the dark. Fluorescence data were acquired using a BD LSRFortessa X-20 flow cytometer (BD Biosciences) and analyzed using FlowJo software version 10 (BD Biosciences). Antibody binding was quantified as the geometric mean fluorescence intensity (gMFI) of the FSC/SSC-gated cell population. Data were plotted using GraphPad Prism version 10.6.0 (GraphPad Software).

Internalization assay

For live-cell imaging, cells seeded in 96-well plates (50 μL/well) were incubated overnight at 37°C. Test articles were labeled with Incucyte Human Fabfluor-pH dye (Sartorius, Cat# 4722) at a 1:3 molar ratio for 15 min at 37°C in the dark. Labeled articles (50 μL) were added to the cells in duplicate. Real-time internalization was monitored using an Incucyte Live-Cell Analysis System (Sartorius), acquiring phase-contrast and red fluorescence images (10× objective) every 1 h. Internalization was directly quantified as the red-to-phase area ratio using Incucyte S3 software version 2019B Rev2.

To assess the internalization and lysosomal trafficking of the antibodies, cells were incubated with test samples (100 nM) for 1 h at 4°C for surface binding, followed by 1 or 6 h at 37°C to permit internalization. Cells were then fixed with 4% paraformaldehyde for 15 min at 4°C, permeabilized with 0.1% Triton X-100 (LPS Solution) for 15 min at room temperature (RT), and blocked with 5% normal goat serum (Vector Laboratories, Cat# S-1000–20) for 30 min at RT. Cells were incubated overnight at 4°C with a mouse anti-LAMP-1 primary antibody (5 μg/mL, Abcam, Cat# ab25630). Test articles and lysosomes were subsequently detected using Alexa Fluor 488-conjugated goat anti-human IgG (1:500, Invitrogen, Cat# A11013) and Alexa Fluor 568-conjugated goat anti-mouse IgG (1:1000, Invitrogen, Cat# A11004), respectively, for 1 h at RT in the dark. Slides were mounted using a DAPI-containing medium (Vector Laboratories, Cat# H-1200). Images were acquired using an LSM 900 confocal microscope and analyzed using Zen software (Carl Zeiss).

Cytotoxicity assay

For 2D cytotoxicity assays, cells were seeded into 96-well plates (0.2–2×10⁵/mL) and incubated for 6 h at 37°C with 5% CO₂. Cells were then treated with serially diluted test articles for 6 days. Viability was assessed using the Cell Counting Kit-8 (CCK-8) (Dojindo, Cat# CK04). Following a 2- to 24-h incubation, absorbance was measured at 450 nm (650 nm background subtraction) using a SPECTRA MAX 190 microplate reader (Molecular Devices). For 3D spheroid assays, cells (2,000 cells/well) were seeded into 96-well round-bottom ultra-low attachment (ULA) plates, centrifuged (125 × g for 10 min), and incubated for 3 days at 37°C with 5% CO₂. Spheroids were treated with serially diluted test articles for 5 days. Viability was determined using the CellTiter-Glo 3D Cell Viability Assay (Promega, Cat# G968A) and a PHERAstar FS plate reader (BMG LABTECH). For both assays, relative cell viability (%) was calculated by normalizing treated sample signals to untreated controls. Half-maximal inhibitory concentration (IC₅₀) values were determined using a sigmoidal four-parameter non-linear regression model in GraphPad Prism version 10.6.0 (GraphPad Software).

Bystander effects

Target-positive (CHO-huROR1-huB7H3) and target-negative (CHO-GFP) cells were co-cultured at 1:1 or 3:1 ratio (2,000 total cells/well) in 96-well plates using 50 μL of DMEM/F-12 containing 10% FBS and 1% AA. After overnight incubation at 37°C with 5% CO₂, 50 μL of ADC or isotype control was added. Real-time bystander cytotoxicity was monitored using an Incucyte S3 Live-Cell Analysis System (Sartorius), acquiring phase-contrast and green fluorescence images (10× objective) every 1 h for 6 days. Target-negative cell viability was determined by quantifying the green fluorescence area using Incucyte S3 software version 2019B Rev2. The relative green area (%) was calculated by normalizing treated samples to untreated controls. Data were plotted as mean ± standard deviation (SD) using GraphPad Prism version 10.6.0 (GraphPad Software).

Animal studies

Cancer cell line derived xenograft studies

All animal experiments performed in this study were approved by the Institutional Animal Care and Use Committee at ABL Bio Inc. (ABL16-2024-IA0017, IA0023 and ABL16-2025-IA0017).

To establish xenograft models, 5 × 10⁶ NCI-H446 and PA-1 cells, and 7.5 × 10⁶ Calu-3 cells, were prepared in a 1:1 mixture of Matrigel (Corning, Cat# CLS356237) and 1x PBS (200 µL/head). These suspensions were then subcutaneously injected into the right flank of 8-week-old female BALB/c nude mice (CAnN.Cg-Foxn1^nu/CrlOri) (OrientBio Inc., Korea), respectively.

When the average tumor volume reached approximately 200–300 mm³, the mice were randomly allocated to each vehicle control group and ABL206 or other test substances (n = 8 mice/group) were administered via a single intravenous injection on Day 1.

The vehicle control group and the test substance diluent consisted of the HSP formulation buffer containing 20 mM histidine, 8% (w/v) sucrose and 0.02% (w/v) polysorbate 80, pH 6.0. During the study period, tumor volume was measured twice a week. The tumor volume was subsequently calculated using the formula: Tumor volume = 0.5 ×Length ×Width², where Length (L) is defined as the longest axis and Width (W) is defined as the shortest axis of the tumor.

Patient derived xenograft (PDX) model studies

All animal studies were approved by the Institutional Animal Care and Use Committees (IACUC) at Crown Bioscience (Taicang, Jiangsu, China), Champions Oncology (Rockville, MD, USA), and LIDE Biotech (Xi’an, Shaanxi, China).

Subcutaneous PDX models representing diverse solid tumors were established in immunocompromised mice (GemPharmatech (Nanjing, China), Charles River (Wilmington, MA, USA), or Beijing Vital River (Beijing, China)).

When mean tumor volumes reached approximately 100–300 mm³, mice were randomized into treatment groups (n = 2–7 per group) based on tumor size. To comprehensively evaluate the in vivo efficacy of ABL206, three distinct sets of PDX studies were conducted. To compare with exact-matched monospecific ADCs, mice received a single dose of ABL206 or monospecific ADCs. For broad efficacy screening across 38 PDX models, mice were treated with ABL206 or a vehicle control, administered as a single dose or Q2W depending on the specific study design for each indication. Specific details regarding cancer types, mouse strains, demographics, and dosing schedules are provided in Supplementary Table S2. To compare with clinical-stage benchmark ADCs in five selected models, mice received a single dose of ABL206, benchmark or isotype ADCs. Notably, in one of these models, tumors regrowing after initial I-DXd treatment were re-treated with ABL206 on day 52.

To account for molecular weight differences between the bispecific and monospecific formats, the treatments were administered at molar equivalent doses.

Tumor volumes and body weights were monitored twice weekly. Tumor volume was calculated using the formula: 0.5 (or 0.52) ×length ×width². Mice were euthanized when individual tumor volumes exceeded 1,500–2,000 mm³, if body weight loss exceeded 20%, or if other humane endpoints (e.g., severe tumor ulceration or necrosis) were reached.

Antitumor activity was quantified as percentage of tumor growth inhibition (%TGI = [1 – (ΔT_t /ΔC_t)] x 100) and percentage tumor volume change (%TVC = 100 * (T_n-T_i)/T_i)), where ΔT_t and ΔC_t represent the tumor volume change between dosing and day 22 to 28 for the treated (T_t - T₀) and vehicle control (C_t - C₀) groups, respectively. Tn is the smallest tumor volume measured at time of the best response or 7 days post dosing and T_i is the tumor volume of the first dosing date.

In vitro stability in rat, monkey, and human plasma

The in vitro stability of ABL206 was assessed by evaluating the payload release rate and monitoring the DAR in biological matrices. The release rate of free exatecan from 200 µg/mL ABL206 in rat, cynomolgus monkey, and human plasma, and 1% BSA in PBS was evaluated at 37°C for up to 21 days. Free exatecan was quantified by LC-MS/MS (QTRAP 6500+, AB Sciex), and the payload release (%) was calculated relative to the theoretical exatecan content based on the DAR value. Additionally, the DAR of 0.1 mg/mL ABL206 in rat (Biochemed) and cynomolgus monkey plasma (BIOIVT, Cat# NHP01PLK2-0101676) and human serum was evaluated at 37°C for up to 14 days or 7 days. Samples were purified using anti-human IgG beads (for plasma, ACRO Biosystems, Cat# MPC-A-004) or human ROR1-coupled beads (for human serum, ACRO Biosystems, Cat# MBR-K009), and the DAR was determined by reversed-phase UPLC (ACQUITY Premier UPLC, Waters).

Pharmacokinetics of ABL206 in rat and monkey

All pharmacokinetic (PK) and toxicology studies were conducted in accordance with Good Laboratory Practice (GLP) regulations and approved by the Institutional Animal Care and Use Committee (IACUC) at Charles River Laboratories (Mattawan, MI, USA).

The pharmacokinetics (PK) of ABL206 were assessed following a single intravenous (IV) bolus administration to Sprague-Dawley rats and cynomolgus monkeys at 1, 3, and 10 mg/kg (n = 3/group). Blood samples were collected for up to 28 days post-dose. Total antibody (TAb) concentrations in serum were measured using an electrochemiluminescent (ECL) immunoassay (Meso Scale Discovery), while plasma concentrations of unconjugated payload (free payload) and antibody conjugated payload (ac-payload) were quantified using a LC-MS/MS method. ADC concentrations were subsequently calculated from the ac-payload measurements. Pharmacokinetic parameters were determined using noncompartmental analysis (NCA) with Phoenix® WinNonlin software version 8.3 (Certara).

Toxicity study of ABL206 in rat and monkey

Sprague-Dawley (SD) rats and cynomolgus monkeys were randomly assigned to receive either a vehicle control or ABL206. SD rats were assigned to groups receiving 80, 160 mg/kg (n = 20; 10/sex/group), or 198 mg/kg (n = 30; 15/sex/group). Cynomolgus monkeys were administered 30, 60 mg/kg (n = 6; 3/sex/group), or 90 mg/kg (n = 10; 5/sex/group). All the animals were treated with ABL206 via intravenous infusion once every 3 weeks for a total of two doses. Throughout the study, the animals were monitored for clinical signs, body weight, food consumption, clinical pathology, and safety pharmacology. Necropsy was performed 3 weeks after the second administration for all the main study groups. The highest dose groups in both species were maintained for an additional 4-week recovery period to evaluate the reversibility of any findings. Histopathology was performed for all animals following their scheduled necropsy.

Statistical analysis

All statistical analyses and graph generation were performed using GraphPad Prism version 10 (GraphPad Software) except for pharmacokinetic concentration-time profiles, which were plotted using Python. Data are presented as the mean ± standard error of the mean (SEM) or mean ± standard deviation (SD), as indicated in the figure legends. EC₅₀ and IC₅₀ values were determined using a four-parameter logistic (4PL) regression model.

Funding Statement

The author(s) reported there is no funding associated with the work featured in this article.

Disclosure statement

JAK, SL, JHK, BMY, HJP, IR, JYK, JHE, YGS, DY, BS, MY, AS, JY, JYP, JJ, JA, WKY, and MJK are employees and shareholders of ABL Bio Inc. SHL is CEO and founder of ABL Bio Inc.

Data availability statement

The data generated in this study are available within the article and its supplementary data files.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/19420862.2026.2731283