XMT-2056, a HER2-Directed STING Agonist Antibody–Drug Conjugate, Induces Innate Antitumor Immune Responses by Acting on Cancer Cells and Tumor-Resident Immune Cells

Abstract

Purpose:

Targeted tumor delivery may be required to potentiate the clinical benefit of innate immune modulators. The objective of the study was to apply an antibody–drug conjugate (ADC) approach to STING agonism and develop a clinical candidate.

Experimental Design:

XMT-2056, a HER2-directed STING agonist ADC, was designed, synthesized, and tested in pharmacology and toxicology studies. The ADC was compared with a clinical benchmark intravenously administered a STING agonist.

Results:

XMT-2056 achieved tumor-targeted delivery of the STING agonist upon systemic administration in mice and induced innate antitumor immune responses; single dose administration of XMT-2056 induced tumor regression in a variety of tumor models with high and low HER2 expressions. Notably, XMT-2056 demonstrated superior efficacy and reduced systemic inflammation compared with a free STING agonist. XMT-2056 exhibited concomitant immune-mediated killing of HER2-negative cells specifically in the presence of HER2-positive cancer cells, supporting the potential for activity against tumors with heterogeneous HER2 expression. The antibody does not compete for binding with trastuzumab or pertuzumab, and a benefit was observed when combining XMT-2056 with each of these therapies as well as with trastuzumab deruxtecan ADC. The combination of XMT-2056 with anti–PD-1 conferred benefit on antitumor activity and induced immunologic memory. XMT-2056 was well tolerated in nonclinical toxicology studies.

Conclusions:

These data provide a robust preclinical characterization of XMT-2056 and provide rationale and strategy for its clinical evaluation.

Translational Relevance.

This study describes the design strategy and preclinical characterization of XMT-2056, a clinical stage HER2-targeted antibody–drug conjugate (ADC) with a STING agonist payload. The data demonstrate that ADC-mediated tumor-targeted delivery of a STING agonist achieves improved efficacy and reduced systemic inflammation compared with a clinical benchmark small-molecule STING agonist. XMT-2056 induced immune-mediated antitumor activity against a broad range of tumor models with high and low HER2 expressions in immune-competent and immune-compromised mice. The antibody design strategy for XMT-2056 empowers combination treatments with other HER2-targeted agents in the clinical setting, and accordingly, preclinical studies demonstrated the benefit of combining XMT-2056 with trastuzumab, pertuzumab, and trastuzumab deruxtecan. Benefit was also observed from combining XMT-2056 and anti–PD-1, which is consistent with the mechanism of action of XMT-2056. The data presented in this study have broad implications for the clinical development of this ADC.

Introduction

Immune checkpoint inhibitors, such as anti–PD-1, CTLA-4, and PD-L1, modulate the adaptive immune response and have shown remarkable efficacy in oncology, including durable responses in the metastatic setting. However, they have limited or no benefit in a large patient population within approved indications, which may be explained by the lack of a preexisting inflamed phenotype (immune desert), an abundance of immune cells that are restricted from infiltration into the tumor cell mass (excluded tumors), or preexisting inflamed tumors replete with exhausted T cells (1, 2).

Activation of an innate immune response has the potential to address many of the limitations of checkpoint inhibitors. Innate immunity represents the body’s first-line defense and can lead to adaptive immunity, thus enabling short- and long-term responses (3, 4). Multiple pathways exist for activating the innate immune response, including Toll-like receptors, RIG-I–like receptors, and the cGAS/STING (stimulator of IFN genes) pathway (5, 6). Notably, the STING pathway has the potential to be activated in cancer cells and immune cells. STING pathway activation in the tumor microenvironment (TME) leads to type I IFN production, resulting in the recruitment and activation of antigen-presenting cells and consequently of CD8⁺ T cells (7–10). Experimental modulation of STING through ligand-independent gain-of-function mutations (11), mouse gene knockouts (12), and pharmacologic cyclic dinucleotide (CDN) and non-CDN agonists (13) has demonstrated that STING activation can generate an antitumor immune response.

Early clinical studies with STING agonists focused on intratumoral injection in an effort to reduce systemic inflammation. The demonstrated pharmacodynamic and clinical responses did not warrant further clinical development, possibly due to the technical challenges of intratumoral injection, the inability to inject into all tumor lesions, and/or the physicochemical properties of the molecules (14, 15). Intravenous (systemic) administration of STING agonists is also under clinical investigation (14). Although this approach may address some limitations of intratumoral injection, it may be more likely to cause systemic inflammation.

We hypothesized that a STING agonist–based antibody–drug conjugate (ADC) could overcome the limitations of free agonists. ADCs enrich the delivery of their payload to the tumor by targeting a tumor antigen; 13 ADCs based on cytotoxic payloads have been approved for oncology indications (16). Conceptually, a systemically administered STING agonist ADC would access all tumor lesions with minimal systemic inflammation and moreover would activate STING in the desired cell types (cancer cells via antigen binding and myeloid cells via antigen-dependent Fc receptor interaction; ref. 17) but not the undesired cell types (e.g., T and B cells, in which STING activation counters the antitumor immune response; refs. 18–20). A STING agonist ADC was recently described, with evidence of immune cell activation and macrophage repolarization (21).

Herein, we describe XMT-2056, a HER2-directed STING agonist ADC that induces greater antitumor activity and less systemic inflammation than a free STING agonist. XMT-2056 induced tumor regression in vivo in a variety of tumor models with high and low HER2 expressions. There was observed benefit when combining XMT-2056 with a variety of approved therapies: trastuzumab, pertuzumab, trastuzumab deruxtecan (T-Dxd) ADC, and anti–PD-1. Together with the nonclinical toxicologic evaluation, these results provide rationale for advancing XMT-2056 into clinical development.

Materials and Methods

Small molecules

The STING agonist was prepared as previously described (22). The drug-linker (reagent required for conjugation to the antibody) used to generate all ADCs except XMT-2056 and surrogate XMT-2056 was prepared as previously described (22). The drug-linker used to generate XMT-2056 and surrogate XMT-2056 was prepared using a modified synthetic route previously described (example 7-1, compound 31, in WO2021202984A1), which resulted in a mixture of L- and D-alanine being present. The diABZI STING agonist was either prepared as described by Ramanjulu and colleagues (23) or purchased from MedChemExpress (Cat. No.: HY-112921A). T-Dxd was sourced from McKesson, provided as a 100 mg/5 mL solution in sterile water, with a reported drug-to-antibody ratio of 7.7.

Antibodies

The following antibodies were used in the ADC generation or as unconjugated antibody controls: pertuzumab was obtained from Selleckchem (Cat. # A2008), and trastuzumab was obtained from STC Biologics (Cat. # STC101). Anti-human HER2 antibody HT19 was obtained from a library of fully human IgG1 antibodies generated using a naïve human synthetic yeast-based antibody discovery platform consisting of a unique library design that recapitulates the human preimmune diversity (Adimab LLC). The antibody was selected following a round of optimization geared toward improved ADC delivery and activity in vivo. Anti-rat/human HER2 antibody clone 7.16.4 used to generate XMT-2056 surrogate ADC was purchased from Bio X Cell (Cat. # BE0277). Anti-RSV human IgG1 (palivizumab) used as a nonbinding control antibody was produced by transient transfection in mammalian cell culture and by ProteinA purification. Fc-mutant HT19 with Fc regions engineered to abrogate Fcγ receptor binding was designed with three mutations in the heavy chain constant region, L234A, L235A, and P329G (AAG; Kabat Eu numbering; ref. 24), and generated through standard molecular biology procedures by a contract research organization.

Cell lines

The human cancer cell lines HCC1954 (CRL-2338), SKBR3 (HTB-30), MDA-MB-175-VII (HTB-25), MDA-MB-453 (HTB-131), SNU-5 (CRL-5973), SKOV3 (HTB-77), and EMT-6 (CRL-2755) were purchased from the ATCC. THP1 reporter cell lines, THP1-Dual (thpd-nfis), THP1-Dual KI-hSTING-H232 (thpd-h232), THP1-Dual KI-hSTING-R232 (thpd-r232), and THP1-Dual KO-STING (thpd-kostg) were purchased from InvivoGen. All cell lines were maintained in the supplier’s recommended medium in a humidified incubator at 37°C, 5% CO₂. Cells were routinely tested for Mycoplasma contamination and authenticated using short tandem repeat analysis on a quarterly basis (IDEXX BioAnalytics). All cells used in this study were negative for Mycoplasma. Primary human peripheral blood mononuclear cells (PBMC; frozen) were purchased from STEMCELL Technologies (Cat. # 70025.2). White blood cells (WBC) were isolated from fresh whole human blood (BioIVT, Cat. # 16140; Charles River Laboratories) using the red blood cell magnetic depletion reagent according to the supplier’s instructions (STEMCELL Technologies, Cat. # 18170). After enrichment, cells were stained with viability dyes (acridine orange and propidium iodide), and viable cells were counted on a Moxi GO II cell counter.

Generation of NucRed cancer cells

SKBR3 and MDA-MB-231 cells stably expressing nuclear-restricted (NR) mKate fluorescent red protein were generated by transduction with IncuCyte NucLight Red Lentivirus reagent (Sartorius, Cat. # 4476). Stably transduced cells (designated as SKBR3–NR or MDA-MB-231–NR) were selected in puromycin-containing media (2 µg/mL) for 2 to 3 days and expanded in their respective culture medium.

Generation of EMT-6–ratHER2 cells

EMT-6 mouse breast cancer cells were stably transduced with ratHER2 lentivirus (GeneCopoeia, Cat. # LPP-Rn14450-Lv156-200-GS) and selected with 2 µg/mL puromycin for 2 to 3 days. RatHER2-expressing cells were sorted using the anti-human/ratHER2 antibody (Bio X Cell, Cat. # BE0277).

Generation of XMT-2056 ADC

HT19 mAb was prepared at 80 mg/mL in 25 mmol/L sodium acetate, pH 5, further diluted to 5 mg/mL in 50 mmol/L HEPES and 1 mmol/L EDTA, pH 7, and reduced using 5 molar equivalents of TCEP-HCl (1 mg/mL in 50 mmol/L HEPES and 1 mmol/L EDTA, pH 7, 37°C, 90 minutes). Next, 10 molar equivalents of the drug-linker (16.4 mg/mL in dimethylacetamide) were added to the reduced mAb, and the reaction continued shaking at 37°C for 60 minutes before being quenched [15 molar equivalents of L-cysteine (1 mg/mL in 50 mmol/L HEPES and 1 mmol/L EDTA, pH 7, room temperature, 45 minutes)]. XMT-2056 was purified using a CHT II column (Bio-Rad PN: 12009257) and a step gradient, with the desired material being eluted using 10 mmol/L sodium phosphate and 2 mol/L sodium chloride, pH 6.7, for 5 column volumes. XMT-2056 was formulated into 50 mg/mL trehalose, 25 mmol/L citrate, 75 mmol/L sodium chloride, pH 5.5. XMT-2056 was analyzed by mass spectrometry, size-exclusion chromatography, and hydrophobic interaction chromatography. Size-exclusion chromatography was performed on a TSKgel G3000SWXL column (5 µm, 7.8 mm × 300 mm, Tosoh Bioscience PN: 0008541) at 35°C using isocratic conditions flow rate of 0.75 mL/minute for 25 minutes (mobile phase: 25 mmol/L sodium phosphate and 150 mmol/L sodium chloride). Hydrophobic interaction chromatography was performed on a TSKgel Butyl-NPR column (2.5 µm, 4.6 mm × 100 mm, Tosoh Bioscience PN: 0042168) at 35°C and eluted with a 25-minute gradient from 0% to 100% B at a flow rate of 1 mL/minute (mobile phase A: 1.5 mol/L ammonium sulfate in 25 mmol/L sodium phosphate, pH 7; mobile phase B: 25 mmol/L sodium phosphate, pH 7, and 10% isopropanol).

Generation of Fc Mutant XMT-2056 ADC

HT19 mAb with its Fc region engineered to significantly reduce Fc receptor binding was designed with three mutations in the heavy chain constant region, L234A, L235A, and P329G (AAG; Kabat Eu numbering; ref. 24), and generated through standard molecular biology procedures by a contract research organization. Fc-mutant HT19 was used to generate Fc-mutant XMT-2056 in the same manner as HT19 was used to generate XMT-2056.

Generation of isotype nonbinding control ADC

Anti-RSV human IgG1 (palivizumab) used as a nonbinding control antibody was produced by transient transfection in mammalian cell culture and by ProteinA purification according to the antibody sequence described by Johnson and colleagues (25). This antibody was used to generate an isotype control ADC in the same manner as HT19 was used to generate XMT-2056.

Surface plasmon resonance profiling of STING agonist

Surface plasmon resonance profiling was done using a Biacore 8K/8K⁺ (Cytiva) with a Series S Sensor Chip SA (cat. # 29104992, Cytiva). Chemically biotinylated recombinant human, cynomolgus monkey, and mouse STING proteins and a nonspecific binding control (biotinylated carbonic anhydrase) were immobilized at a level of 2,500 to 3,500 RU. Interaction between the STING agonist and biotinylated recombinant STING proteins was profiled by single-cycle kinetics analysis. For human, cynomolgus monkey, and mouse STING proteins, analytes were injected with 9-points serial concentration (dilution factor of 3), with the highest concentration of 1 µmol/L for 2 minutes followed by 30 minutes of dissociation at a flow rate of 30 µL/minute. Data processing was performed using Biacore Insight Evaluation Software. Solvent correction, baseline across the measurement, nonspecific binding to reference Fc, and sensorgram shape were visually checked as data quality control. After double referencing with reference Fc and blank cycle, data points were applied for a 1:1 kinetics fitting model to obtain kinetics parameters, including k_a(l/Ms), k_d(l/s), and K_D(M).

Epitope mapping of anti-HER2 antibody HT19

Epitope mapping was performed at a single amino acid resolution using shotgun mutagenesis technology at Integral Molecular. Briefly, a library of 640 mutant clones was generated in HEK-293 through alanine scanning mutagenesis of the human HER2 extracellular domain (residues 23–652 of sequence NM_004448). Clones were screened for binding to HT19 by immunofluorescence flow cytometry. Mean binding protein reactivities for all critical studies were identified relative to the wild-type (WT) protein and negative control.

Activity of STING agonist in THP1 cells

Cell suspensions of THP1 reporter cell lines (THP1-Dual (thpd-nfis), THP1-Dual KI-hSTING-H232 (thpd-h232), THP1-Dual KI-hSTING-R232 (thpd-r232), and THP1-Dual KO-STING (thpd-kostg; InvivoGen) were incubated in assay medium (RPMI-1640, 1% penicillin/streptomycin–10% heat-inactivated FBS) containing treatments and incubated at 37°C, 5% CO₂. Culture supernatants were collected at the indicated time points and subjected to either luciferase analysis representing IRF3 reporter activity or cytokine analysis. Luciferase activity was measured on a Spectramax M5 plate reader using ANTI-Luc luminescence assay reagent (InvivoGen, Cat. # rep-qlc). CXCL10 (IP-10) and IFN-β cytokine levels were determined using Human Magnetic Luminex Performance Assay Kit according to the kit’s instructions (R&D Systems, Cat. # FCSTM18-04). Dose–response curves were generated using GraphPad Prism software. EC₅₀ values were determined from four-parameter curve fitting in GraphPad Prism.

Activity of STING agonist in J774 mouse macrophages

Suspensions of murine J774-Dual cells (InvivoGen), derived from the J774.1 macrophage-like cell line, were incubated in assay medium (DMEM, 1% penicillin/streptomycin–10% heat-inactivated FBS) containing treatments and incubated for 24 hours at 37°C, 5% CO₂. Culture supernatants were collected, and CXCL10 levels were measured by ELISA using Mouse CXCL10/IP-10 DuoSet ELISA Kit (R&D Systems, Cat. DY466) following the manufacturer’s recommended procedures. Plates were analyzed using a SpectraMax M5 plate reader. Dose–response curves were generated using GraphPad Prism software. EC₅₀ values were determined from four-parameter curve fitting in GraphPad Prism.

SafetyScreen44

SafetyScreen44 was run at Eurofins Panlabs Discovery Services in New Taipei City, Taiwan. Briefly, the STING agonist was screened against a panel of 44 targets using enzyme or radioligand binding assays to assess off-target activity. Methods used in the study had been adapted from the scientific literature to maximize reliability and reproducibility. Reference standards were run to ensure the validity of the results obtained. Significant responses were considered to be ≥50% inhibition or stimulation within the measured assays.

Selection of HT19 Anti-HER2 antibody

Eight naïve human synthetic yeast libraries each of ∼10⁹ diversity were propagated as described previously (26). For the first two rounds of selection, a magnetic bead sorting technique utilizing the Miltenyi MACS system was performed as previously described (27). Sorting was performed using a FACS ARIA sorter (BD Biosciences). Sort gates were determined to select only HER2-binding clones for two rounds, and the third round was a negative sort to decrease reagent binders. After the final round of sorting, yeast cells were plated, and individual colonies were picked for characterization. Three rounds of affinity maturation cycles were performed using FACS sorting for selection.

Fcγ receptor I binding kinetics

Kinetics analysis of XMT-2056, HT19 unconjugated antibody, and Fc-mutant XMT-2056 ADC binding to Fcγ receptor I (Fcγ-RI) was performed by biolayer interferometry at 25°C on a ForteBio Octet QKe (Sartorius). All dilutions and baseline measurements were done in 1× kinetics buffer. Baseline measurements were performed at 60 seconds before and after the capture step. Ni-NTA biosensors (Sartorius) were used for the capture of 1 µg/mL of recombinant Fcγ-RI for 300 seconds. Association of the test articles to Fcγ-RI was measured at seven concentrations diluted two-fold from 100 nmol/L for 300 seconds followed by a dissociation step for 900 seconds. Binding constants were calculated using the 1:1 model and global fit average.

Phospho-Akt (Ser473) sandwich ELISA

HER2⁺ SKBR3 breast cancer cells were incubated in assay medium (RPMI-1640, 1% penicillin/streptomycin–10% heat-inactivated FBS) containing treatments at the indicated concentrations and incubated for 4 hours at 37°C/5% CO₂. Cells were lysed on ice with cell lysis buffer containing phosphatase/protease inhibitor cocktail. Cell lysates were processed using PathScan Phospho-Akt1 (Ser473) Sandwich ELISA Kit according to the kit’s instructions (Cell Signaling Technology, Cat. #7160). The optical density (450 nm) was measured immediately using a SpectraMax M5 plate reader. Phospho-AKT (% inhibition) was calculated relative to the untreated controls. Bar graphs were generated using GraphPad Prism software.

Antibody competition assay by biolayer interferometry (Octet)

The assay was performed on a ForteBio Octet Red 384 system (Pall ForteBio Corporation) using a standard sandwich format binning assay. Trastuzumab or pertuzumab was loaded onto AHQ sensors, and unoccupied Fc-binding sites on the sensor were blocked with a nonrelevant human IgG1 antibody. The sensors were then exposed to 100 nmol/L of the HER2 extracellular domain, followed by HT19. Data were processed using ForteBio Data Analysis Software 7.0.

Flow cytometry analysis

Binding of XMT-2056 and HT19 to HER2 expressed on cancer cells was determined by flow cytometry. Cells were incubated with the ADCs, parental antibodies, and nonbinding control ADC for 1 hour in the presence of 6% goat serum in culture medium and washed with ice-cold PBS. After staining with the secondary antibodies (Alexa Fluor 647 goat anti-human IgG, 1 hour on ice), cells were run on a MACSQuant flow cytometer (Miltenyi Biotec). For the cell-binding competition assay, cells were incubated with anti-HER2 antibodies HT19–hIgG1 or trastuzamab–mIgG2a alone or in combination. Palivizumab–hIgG1 or palivizumab–mIgG2a were used as nonbinding control antibodies. After incubation for 1 and 5 hours, the cells were washed with ice-cold PBS and stained with secondary antibodies (Alexa Fluor 647 goat anti-human IgG or goat anti-murine IgG (Invitrogen, Cat. # A21445 or A21235, respectively) for 1 hour on ice. After washing with ice-cold PBS, cells were run on a MACSQuant flow cytometer (Miltenyi Biotec). Data analysis was performed using FlowJo software. Dose–response curves were generated using GraphPad Prism. EC₅₀ values were determined from four-parameter curve fitting.

STING agonist activity in WBCs

WBCs were obtained from fresh whole blood of three independent human donors as described above. Cell suspensions were subjected to treatments in assay medium (RPMI-1640, 10% FBS, and 1% penicillin/streptomycin) for 6 or 24 hours at 37°C, 5% CO₂. Culture supernatants were collected and subjected to CXCL10 (IP-10), IFN-β, IL-6, and TNF-α analysis using a magnetic Luminex assay kit according to the manufacturer’s instructions (R&D Systems, Human XL Cytokine Luminex Performance Panel, Cat. # FCSTM18-04). Belysa Immunoassay Curve Fitting Software was used for data analysis. Induction of CXCL10, IL-6, and TNF-α peaked at the 24-hour time point, whereas for IFN-β, the peak occurred at 6 hours; data obtained from the peak time points of each cytokine were used for plotting in GraphPad Prism.

THP1 IRF3 reporter cell assays

For cancer cell and THP1 IRF3 reporter cell coculture assays, SKOV3 cancer cells were seeded in 96-well tissue culture plates (15,000 cells/well) and allowed to attach overnight. Culture medium was replaced with assay medium (RPMI-1640, 10% FBS, and 1% penicillin/streptomycin), and after adding the indicated test articles, the plates were incubated for 20 minutes at 37°C. THP1-Dual IRF3 reporter cells (50,000 cells/well) were added, and the plates incubated for 24 hours at 37°C, 5% CO₂. For the plate-bound recombinant antigen assay, plates were first coated overnight with 1 µg/mL recombinant human HER2 protein (Sino Biological, Cat. # 10004-H08H4). The following day, plates were washed with 0.1% TBS with Tween-20 and blocked with 3% BSA in PBS (all solutions were filter-sterilized). After washing with assay medium 3× fresh assay medium containing treatments and 50,000 THP1 reporter cells/well were added to the plates and incubated for 24 hours at 37°C, 5% CO₂. For both cocultures and plate-bound antigen assays, supernatants were assayed for luciferase activity using ANTI-Luc luminescence assay reagent (InvivoGen, Cat. # rep-qlc) on a SpectraMax M5 plate reader. Concentration–response curves for all assays were generated using GraphPad Prism software. EC₅₀ values were determined from four-parameter curve fitting in GraphPad Prism.

Immunohistochemistry

IHC for human HER2 and murine CD68, CD45, CD11c, and PD-L1 was performed on formalin-fixed paraffin-embedded (FFPE) tumor xenografts. Manual antigen unmasking was done using heat-induced epitope retrieval with an electronic pressure cooker. Slides were immersed in EDTA pH 9.0 buffer (Vector Laboratories Cat.# H-3301-250) for HER2, mCD45, mCD11c, and mPD-L1 and in citrate pH 6.0 buffer (Vector Laboratories, Cat.# H-3300-250) for mCD68, heated to 99°C, and incubated for 20 minutes. Samples were blocked with peroxidase and then incubated for 30 minutes with the following primary antibody: rabbit polyclonal anti-HER2 (Agilent Technologies, Cat. No. A0485; final dilution 1:500), rabbit monoclonal anti-CD68 (clone EPR23917-164, Abcam, Cat. No ab283654; final dilution 1:800), rabbit monoclonal anti-CD45 (clone D3F8Q, Cell Signaling Technology, Cat. No 70257S; final dilution 1:250), rabbit monoclonal anti-CD11c (clone D1V9Y, Cell Signaling Technology, Cat. No 97585S; final dilution 1:350), or rabbit monoclonal anti–PD-L1 (clone D5V3B, Cell Signaling Technology, Cat. No 64988S; final dilution 1:100). For detection, slides were incubated with HRP-labeled anti-rabbit secondary antibody (Agilent Technologies, Cat. No. K400311-2) for 30 minutes followed by chromogenic stain with Liquid DAB⁺ (Agilent Technologies, Cat. No K346811-2). Samples were counterstained with hematoxylin (Abcam, Cat no. ab220365) and then dehydrated and cleared for mounting. Images were captured using an Olympus cellSens Entry 1.17 microscope camera (Olympus Corporation). Image analysis was conducted with HER2 algorithms built with HALO multiplex IHC software, version 3.2.3 (Indica Labs Inc.).

Gene expression analyses

Gene expression analysis of SKOV3 tumors harvested from CB.17 SCID mice was performed using NanoString. RNA was extracted from FFPE tumor tissue using Qiagen RNeasy FFPE Kit according to the kit’s instructions. For gene expression analysis of cocultures, cells were harvested after treatments, and RNA was extracted using Qiagen RNeasy Mini Kit. A total of 150 ng RNA per sample was analyzed on a NanoString nCounter Max system using the nCounter human PanCancer Immune Profiling code set (NanoString, XT-CSO-HIP1-12, Cat. # 115000132) or mouse PanCancer Immune Profiling code set (NanoString, XT-CSO-MIP1-12, Cat. # 115000142) and nCounter Standard Master Kit (NanoString, NAA-AKIT-048, Cat. # 100054). Data were analyzed using the nSolver Advanced Analysis Software (NanoString).

XMT-2056 pharmacokinetics in naïve CB.17 SCID mice

Female immunodeficient CB.17 SCID mice received a single intravenous administration of XMT-2056 which had been diluted with an appropriate volume of vehicle (sterile saline) to achieve a 0.04 mg/kg payload dose (1.14 mg/kg antibody dose) at 10 mL/kg (0.2 mL/20 g). Blood (40 µL) was serially collected from each animal via tail snip without anesthesia at 15 minutes, 24, 72, 168, and 240 hours postdose; full blood volume was collected via terminal cardiac puncture under anesthesia at 336 hours postdose (n = 4/group/time point). Blood was collected into K2EDTA Microcuvette tubes and processed for plasma: the samples were held on ice for ∼ 1 hour and centrifuged, and 15 to 20 µL of plasma (without red blood cells) were transferred to V-bottom tubes, snap-frozen, and stored at −80C. Neat plasma was diluted 10-fold in microsampling stabilizer buffer prior to assay for conjugated drug by LC/MS and total antibody by LBA. Pharmacokinetics (PK) data were analyzed with Phoenix 8.3, and PK concentration–time profiles were presented by analyte and test articles.

XMT-2056 Toxicokinetics in nonhuman primates

The study in cynomolgus monkeys was conducted at Charles River Laboratories. One male and one female animal were administered XMT-2056 via 45 minutes intravenous infusion on day 1 and day 22. Plasma was collected at 1, 6, 24, 48, 96, 168, 240, 336, and 504 hours following the first dose and 1, 6, 24, 48, 96, and 168 hours following the second dose. Plasma samples were analyzed for total antibody and conjugated drug.

IncuCyte cancer cell killing assay

Cancer cell death in monocultures, PBMC cocultures, or triple cultures was determined using an IncuCyte cancer cell killing assay. Cancer cells stably expressing mKate fluorescent red protein were seeded in 96-well tissue culture plates and allowed to attach overnight. The following day, the culture medium was replaced with fresh assay medium (RPMI-1640, 10% FBS, and 1% penicillin/streptomycin) containing treatments and incubated for 20 minutes at 37°C, followed by the addition of PBMCs (1:2–1:3 ratio). For triple culture assays, the HER2-expressing cancer cells (WT) were seeded with traced cancer cells expressing mKate fluorescent protein at the indicated ratios and treated as described above. Plates were then placed in an IncuCyte (Sartorious) live cell imaging instrument in an incubator (37°C, 5% CO₂) and scanned every 4 hours over 3 to 4 days. Red object confluency or area (cancer cells) over time was quantified using IncuCyte Zoom software. Percent viable cells were calculated relative to the average of the red object confluency/area of control wells. Concentration–response curves were generated using GraphPad Prism software. IC₅₀ values were determined from four-parameter curve fitting in GraphPad Prism.

Cytokine analysis by bead-based multiplex or ELISA assays

Cytokine analysis in cell culture supernatants was performed using a magnetic bead-based 4-plex Luminex kit for CXCL10, IFN-β, IL-6, and TNF-α from R&D Systems (Human XL Cytokine Luminex Performance Panel, Cat. # FCSTM18-04) or analyzed using DuoSet ELISA Kit for CXCL10 alone from R&D Systems (Cat. # DY266). Cytokine analysis in mouse serum was performed using a multiplexed Luminex kit from EMD Millipore (MILLIPLEX Mouse Cytokine/Chemokine MAGNETIC BEAD Premixed 32 Plex Kit, Cat. # MCYTMAG-70K-PX32). Dose–response curves were generated using GraphPad Prism software. EC₅₀ values were determined from four-parameter curve fitting.

In vivo studies

In vivo studies were performed at Charles River Discovery Services (SKOV3, HCC1954, SNU-5, JIMT-1, EMT-6–rHER2, EMT-6, CT26, and NCI-N87 models) and Crown Bioscience, Inc. (mBR9013 model). Both facilities are accredited under the Association for Assessment and Accreditation of Laboratory Animal Care International, and all experiments were approved by the Institutional Animal Care and Use Committee of each facility. EMT-6–rHER2 cells were generated by engineering EMT-6 cells to express rat HER2 as described in the Supplementary Material. The mBR9013 murine breast cancer model was derived from a mouse mammary tumor virus (MMTV)-ERBB2 transgenic mouse which expresses rat HER2 under the direction of the MMTV promoter and was maintained as tumor fragments.

For efficacy studies, female CB.17 SCID (SKOV3, SNU-5, JIMT-1, and NCI-N87) or SCID beige (HCC1954) mice were subcutaneously inoculated in the right flank with 1 × 10⁷ cells in 50% Matrigel, and female BALB/c mice were inoculated in the right flank with 5 × 10⁶ EMT-6–rHER2 cells. For mBR9013, female FVB/NJ mice were inoculated in the right flank with 2- to 3-mm tumor fragments. Animals were randomized when tumors reached 100 to 150 mm³ (NCI-N87) or 60 to 100 mm³ (all other models) and treated according to the doses, schedules, and routes shown in the figures. Tumors were measured using calipers twice weekly, and tumor volumes were calculated using the formula width² × length/2. Tumor growth inhibition was defined as the percent difference in median tumor volumes between treated and control groups on the last day all control animals remained on study, with results analyzed for statistical significance using the Mann–Whitney U-test. A partial response was defined as at least three consecutive tumor volume measurements equal to 50% or less of the starting tumor volume and greater than 13.5 mm³. A complete response was defined as at least three consecutive tumor volume measurements less than 13.5 mm³. For the EMT-6–rHER2 rechallenge study, selected tumor-free animals or age-matched naïve BALB/c mice were inoculated with 5 × 10⁶ EMT-6 cells in the left flank and 3 × 10⁶ CT26 cells in the right flank.

For the pharmacodynamic study, female CB.17 SCID mice bearing SKOV3 xenografts were randomized into treatment groups (n = 10/group) when tumors reached a mean of 100 to 200 mm³. Animals were given a single, intravenous injection of treatments. Serum was collected at 6, 12, 24, and 72 hours following treatment (n = 5/time point). Tumors were collected at 12 and 72 hours following treatment, formalin-fixed, and paraffin-embedded (n = 5/time point).

Data availability

All data are contained within the presented figures and the supplementary figures. Any further reasonable data access requests can be submitted to the corresponding author, who can be reached by e-mail at: tlowinger@mersana.com.

Results

XMT-2056 elicits potent antigen-dependent and Fcγ-R–mediated activation of monocytes

XMT-2056 is a tumor cell–targeted ADC that delivers a STING agonist payload to HER2-expressing cancer cells and tumor-resident Fcγ-R–expressing immune cells, both in an antigen-dependent manner (Fig. 1A). XMT-2056 is comprised of a novel anti-HER2 human IgG1 antibody conjugated to a STING agonist payload via a cleavable linker with a drug-to-antibody ratio of ∼8 (Fig. 1B; Supplementary Fig. S1A–S1C). The optimization of the payload, linker, and scaffold used in XMT-2056 has been described elsewhere (22).

Figure 1.

XMT-2056 elicits potent antigen-dependent and Fc-R–mediated activation of monocytes in cocultures. A, Schematic of the XMT-2056 mechanism, which includes HER2-dependent ADC uptake into tumor cells and Fcγ-R–expressing myeloid cells. B, Chemical structure of XMT-2056. C, Cytokine induction as measured by a multiplex Luminex assay from supernatants of fresh human WBCs treated for 6 (IFN-β) or 24 (CXCL10, IL-6, and TNF-a) hours. Bars represent mean value of n = 2 data points shown as symbols. D, Competition with trastuzumab by biolayer interferometry (Octet): trastuzumab was loaded onto the sensor chip, and HER2 ECD or HT19 antibody associations are indicated by blue arrows. Additional binding by HT19 indicates noncompetitive binding. E, Competition with trastuzumab by cell-based flow cytometry. HT19 hIgG1 or mIgG2a formats detected with either Alexa Fluor anti-hIgG1 (h647) or Alexa Fluor anti-mIgG2a (m647) secondary antibodies in the presence or absence of trastuzumab–mIgG2a. Each point represents the mean and SD (n = 3). F, Binding of XMT-2056 or HT19 antibody to HCC1954 cells showing fluorescence intensities measured by flow cytometry. Each point represents the mean and SD (n = 2). G, CXCL10 cytokine induction in HCC1954 monocultures treated for 24 hours with XMT-2056, nonbinding control ADC, HT19, or the free payload. Each point represents the mean and SD (n = 2). H, Binding of XMT-2056 or Fc-mutant XMT-2056, HT19, and nonbinding control ADC to SKOV3 cells showing fluorescence intensities measured by flow cytometry. Each point represents the mean and SD (n = 3). I and J, IRF3 reporter activity of THP1 cells in coculture with SKOV3 cells (I) or cultured on recombinant HER2 antigen-coated plates (J) treated for 24 hours with XMT-2056 or Fc-mutant XMT-2056, nonbinding control ADC, or STING agonist payload. Each point represents the mean and SD (n = 3). When noted, EC₅₀ and B_max values represent mean of two independent experiments. gMFI, geometric mean fluorescence intensity; RLU, relative light units. (A, Created in BioRender. Cetinbas, N. [2025], https://BioRender.com/s16c825; G, Created in BioRender. Cetinbas, N. [2025], https://BioRender.com/s16c825; I and J, Created in BioRender. Cetinbas, N. [2025], https://BioRender.com/s16c825.)

The payload is a non-CDN designed and optimized for use in an ADC (22). Structurally related to the dimeric amidobenzimidazole chemical class, (23) the payload has been designed to have increased hydrophilicity, which improves the physicochemical properties of the ADC and reduces payload membrane permeability when released from the antibody. The payload is a subnanomolar binder to human, cynomolgus monkey, and mouse STING proteins (Supplementary Table S1) and showed specific activity in monocytic THP1 cells expressing each of the three most common STING haplotypes (28) but not in STING knockout cells (Supplementary Fig. S2A; Supplementary Table S1). The payload did not interact across a broad panel of 44 receptors/transporters at 1 µmol/L concentration (Supplementary Table S1). STING pathway cytokine production was concentration-dependent in THP1 and mouse macrophage–like J774 cells (e.g., CXCL10 and IFN-β; Supplementary Table S1) and in fresh human WBCs (e.g., CXCL10, IFN-β, IL-6, and TNF-α; Fig. 1C; Supplementary Fig. S2B).

We reasoned that an antibody that does not compete for binding with other HER2-targeted agents would offer advantages during clinical use, such as combination therapies with those agents. Accordingly, we used yeast display to identify a novel anti-HER2 mAb, named HT19, which does not compete for binding with trastuzumab or pertuzumab (Fig. 1D; Supplementary Fig. S2C–S2E). HT19 binds to an epitope including three required residues (E521, L525, and R530) on domain IV of human HER2. In the context of an auristatin-based dolaflexin ADC, HT19 outperformed trastuzumab (Supplementary Fig. S2E). HT19 partially inhibits HER2 signaling pathway relative to trastuzumab, as indicated by inhibition of AKT phosphorylation in SKBR3 cells (Supplementary Fig. S2F). HT19 is fully cross-reactive to cynomolgus monkey HER2 and does not bind to rat or mouse orthologs (Supplementary Table S2) and was found not to compete with trastuzumab for binding (Fig. 1E).

The binding profile of XMT-2056 to cancer cell lines was comparable with that of HT19 mAb (Fig. 1F and H; Supplementary Fig. S2G), which indicates that the bioconjugation process does not alter binding. HER2 status was defined by maximal fluorescence signal and found to be in good concordance with RNA sequencing HER2 expression levels obtained from the Cancer Cell Line Encyclopedia expression database (Supplementary Table S3; ref. 29).

The ability of XMT-2056 to induce STING pathway activation was first evaluated in HER2-expressing HCC1954 cancer cells, which respond to STING agonism in contrast to most cancer cell lines (17). XMT-2056 induced CXCL10 production with an EC₅₀ value of 2.67 nmol/L payload (Fig. 1G). XMT-2056 exhibited >75-fold greater activity than the free payload, demonstrating an advantage of the targeted delivery. Neither HT19 nor the nonbinding control ADC induced CXCL10; the nonbinding control ADC consists of the anti-RSV antibody palivizumab (25) conjugated to the same linker-payload as XMT-2056. These results demonstrate that XMT-2056 directly activates STING in antigen-expressing cancer cells.

In the hypothesized mechanism of action of XMT-2056, its internalization into myeloid cells requires antigen binding on neighboring cancer cells (Fig. 1A; ref. 17). To evaluate this hypothesis, THP1 monocytes with an IRF3-based reporter were treated with XMT-2056 while in coculture with HER2-expressing cancer cells or on plates coated with recombinant HER2 protein. An Fc-mutant version of XMT-2056 was generated by introducing mutations into HT19 (L234A, L235A, and P329G; Kabat Eu numbering; ref. 24) and as expected did not bind to Fcγ-RI (Supplementary Fig. S2H) yet retained binding to HER2 (Fig. 1H). XMT-2056, but not the Fc-mutant ADC, induced STING activity in THP1 cells when cultured with HER2-expressing SKOV3 cells (Fig. 1I) or with recombinant HER2 protein (Fig. 1J). These data indicate that the uptake of XMT-2056 through Fc–Fc $\gamma$ R interactions contributes to the observed activity. The antigen-dependent nature of the Fcγ-R–mediated uptake was evidenced by the lack of activity of the nonbinding control ADC (Fig. 1I and J) or in the absence of HER2 protein (Supplementary Fig. S2I). Together, these data demonstrate that XMT-2056 productively delivers payload to cancer cells and myeloid cells in an antigen-dependent manner.

XMT-2056 is active against HER2-high and HER2-low cancer cells and causes concomitant immune-mediated killing of HER2-negative cancer cells

The ability of XMT-2056 to induce cancer cell killing was assessed by coculture of PBMCs from healthy human donors with cancer cells that stably express NR mKate fluorescent red protein such that cancer cell death could be specifically measured. XMT-2056 induced cancer cell death, with IC₅₀ and maximal cell killing values of 0.023 nmol/L payload and 92.2% for SKBR3–NR (high HER2; Fig. 2A) and 0.38 nmol/L and 62% for MDA-MB-175-VII–NR (low HER2; Fig. 2B), respectively. Free payload showed significantly lower potency. HT19 and the nonbinding control ADC showed minimal activity (Fig. 2A and B; Supplementary Fig. S3A), which is consistent with the therapeutic hypothesis that the pharmacology is driven by STING pathway activation and is dependent on antigen binding.

Figure 2.

XMT-2056 induces killing of HER2-high and HER2-low cancer cells and induces concomitant immune-mediated killing of HER2-negative cancer cells in vitro. A and B, Cancer cell death induced by XMT-2056, nonbinding control ADC, or STING agonist payload, shown as percent viable SKBR3–NR (A) or MDA-MB-175-VII–NR (B) cells in PBMC coc ultures (84 hours time point). Each point represents the mean and SD (n = 3). C and D, Cytokine induction by XMT-2056, nonbinding control ADC, or STING agonist payload in supernatants of SKBR3–NR (C) or MDA-MB-175-VII–NR (D) cells in PBMC cocultures (24 hours time point). Each point represents the mean and SD (n = 3). Note that the y-axis scales vary. E, Schematic of concomitant antigen-dependent, immune-mediated killing of HER2-negative cancer cells. F, Flow cytometry of HER2 expression on SKBR3 and nuclear red expressing MDA-MB-231–NR. Each point represents the mean and SD (n = 4). G, Cancer cell death mediated by XMT-2056, nonbinding control ADC, or STING agonist payload, shown as percent viable MDA-MB-231–NR cells in coculture with PBMCs and HER2-positive SKBR3 at the indicated ratios or in the absence of SKBR3 (84 hours time point). Each point represents the mean and SD (n = 3). (E, Created in BioRender. Cetinbas, N. [2025], https://BioRender.com/k38t564.)

Consistent with STING activation, XMT-2056 led to the induction of CXCL10, IFN-β, IL-6, and TNF-α, with EC₅₀ values between 0.17 and 0.46 nmol/L payload for SKBR3–NR (Fig. 2C) and 0.3 to 0.76 nmol/L for MDA-MB-175-VII–NR (Fig. 2D) after 24 hours. In contrast, cytokine induction by the free payload was modest and sometimes absent at the highest concentration. HT19 and the control ADC led to minimal induction (Fig. 2C and D; Supplementary Fig. S3B).

We hypothesized that XMT-2056 would elicit immune-mediated killing of all cancer cells in the vicinity of immune cell activation regardless of their antigen expression level while still initially dependent on antigen binding (Fig. 2E). We developed a triple coculture assay consisting of PBMCs, unlabeled HER2-positive SKBR3 cells, and nuclear red-labeled HER2-negative MDA-MD-231–NR cells (Fig. 2F). Neither of these cancer cell lines is affected by XMT-2056 in the absence of immune cells (Supplementary Fig. S3C). In the presence of HER2-positive SKBR3 cells and PBMCs, the HER2-negative MDA-MB-231–NR cells were efficiently killed by XMT-2056 treatment, with an IC₅₀ value of 0.25 nmol/L at a 1:1 ratio of HER2-positive:HER2-negative cells and an IC₅₀ value of 0.34 nmol/L at a 1:4 ratio (Fig. 2G). XMT-2056 had no effect on HER2-negative cells in the absence of HER2-positive cells (Fig. 2G). These data demonstrate that XMT-2056 induces immune-mediated killing of HER2-negative cancer cells specifically in the presence of HER2-positive cancer cells and immune cells.

XMT-2056 induces tumor regressions with contributions from STING activation in cancer cells and immune cells

A single intravenous administration of XMT-2056 elicited antigen-dependent, dose-dependent activity against SKOV3 tumor xenografts, resulting in complete and sustained tumor regressions at doses of 1 mg/kg (0.04 mg/kg payload) or higher without any changes in clinical condition or substantial body weight loss (Fig. 3A; Supplementary Fig. S4A and S4B; Supplementary Table S4). We hypothesized that the Fc-mutant ADC described above would have significant yet reduced activity in comparison with the Fc-WT ADC. Indeed, the Fc-mutant ADC resulted in tumor growth inhibition relative to vehicle (P ≤ 0.001) yet was less active than XMT-2056 (P ≤ 0.001; Fig. 3B).

Figure 3.

Intravenous administration of XMT-2056 elicits tumor-specific cytokine changes, immune cell tumor infiltration, and antitumor activity. A, Antitumor activity of XMT-2056 in SKOV3 xenograft model. Tumor-bearing CB.17 SCID mice were intravenously administered a single dose (black arrowhead) of XMT-2056, nonbinding control ADC, HT19 antibody, or STING agonist payload, or three doses (orange arrowheads) of diABZI STING agonist. Each point indicates the mean tumor volume and SEM (n = 10). The free payload dose of 0.128 mg/kg is equivalent to the payload dose of 3 mg/kg XMT-2056. B, Antitumor activity of XMT-2056 and Fc-mutant XMT-2056 at the indicated doses in the SKOV3 xenograft model. Each point represents the mean tumor volume and SEM (n = 10). C, Antitumor activity of XMT-2056 in SKOV3 after 3 weekly administrations of lower doses; the fractions of animals with partial response (PR) or complete response (CR) are indicated. D, Cytokines/chemokines measured in serum of SKOV3 tumor–bearing mice after a single intravenous injection of XMT-2056, nonbinding control ADC, or diABZI STING agonist using a 32-plex Luminex assay. Sampling was performed at 6, 12, 24, and 72 hours after administration. Each point represents the mean and SEM (n = 5). E, Mouse gene signature scores for tumors 12 hours after treatment as measured by NanoString analysis. DC, dendritic cell. F, Normalized counts for mouse mRNA or human mRNA of individual tumor cytokine/chemokines in the xenografts sampled at 12 hours. Each point represents the mean and SD (n = 2). G,CD68 and CD45 murine mRNA counts in tumors at the indicated time points. Each floating bar represents the mean and minimum/maximum values (n = 2). H, Representative IHC images of tumors collected 72 hours after treatment stained for the macrophage marker CD68. Scale bar, 100 µm.

Three weekly administrations of the Fc-WT ADC at doses as low as 0.067 mg/kg resulted in partial and complete responses (Fig. 3C). In contrast, an intravenously administered diABZI STING agonist evaluated at the published dose regimen (23), which is cumulatively 100-fold higher than the payload dose at 1 mg/kg XMT-2056, elicited only modest antitumor activity (Fig. 3A). Minimal activity was observed with HT19 or the nonbinding control ADC, and no activity was observed with the free payload at the dose equivalent to the highest dose of XMT-2056. These studies were conducted in immunodeficient CB.17 SCID mice, which have intact innate immune compartments but no adaptive immunity (no B or T cells); thus, the observed activity is consistent with an innate antitumor immune response. The pharmacokinetic profile was characterized by slow clearance and high ADC stability, as indicated by similar kinetics of antibody and conjugated drug (Supplementary Fig. S5A).

We investigated changes induced by XMT-2056 in the periphery and in SKOV3 tumors at a dose that resulted in tumor regression (1 mg/kg). First, we measured levels of serum (peripheral) cytokines. Transient elevations were observed for cytokines associated with STING signaling, including CXCL10 (IP-10), IL-6, MIP-1β, RANTES, and CXCL9 (Fig. 3D). Strikingly, the diABZI STING agonist induced 1.5- to 10-fold higher elevations of serum cytokines than XMT-2056 at the dose that yielded modest antitumor activity (Fig. 3D). These results highlight the therapeutic rationale of a STING agonist ADC: improved antitumor activity and reduced systemic inflammation relative to a free IV STING agonist.

Second, we assessed molecular and cellular changes induced by XMT-2056 in the TME. Conducting this analysis with human tumor xenografts in mice enabled the discrimination of gene expression in tumor cells vs. host cells using human- and mouse-specific NanoString panels. Pathway analysis of mouse gene expression revealed significant upregulation of key immune-related signatures in the tumor within 12 hours of treatment in an antigen-dependent manner (Fig. 3E).

XMT-2056 induced STING activation in tumor cells and murine host cells in an antigen-dependent manner, with significant changes observed after a dose of 0.3 mg/kg XMT-2056 (Fig. 3F; Supplementary Fig. S4C). STING pathway activation in both cell types was consistent with the results in vitro (Fig. 1G–J; ref. 17) and in vivo, whereas the Fc-mutant ADC only activated STING in the tumor cells (Fig. 3B).

Cellular changes in the TME were also consistent with STING activation. Murine mRNA levels of immune cell markers CD68 and CD45 were elevated in tumors harvested after 72 hours (Fig. 3G), a finding confirmed by IHC staining which demonstrated an increase in CD68⁺ cells and CD45⁺ cells in the same samples (Fig. 3H; Supplementary Fig. S4D). Consistent with an antitumor immune response, IHC staining also revealed an increase in CD11c⁺ cells and an increase in mPD-L1 staining in the XMT-2056–treated tumors (Supplementary Fig. S4E). It was not possible to investigate T-cell infiltration because the tumors were grown in CB.17 SCID mice, which lack T cells.

XMT-2056 induces antitumor activity in a diverse panel of tumor models

Further in vivo studies were conducted in tumor xenograft models with a range of HER2 expression: HCC1954, JIMT-1, and SNU-5 (Fig. 4A; Supplementary Table S3; ref. 29). XMT-2056 demonstrated dose-dependent antitumor activity in these models. In HCC1954 and SNU-5, a single dose administration at 1 or 3 mg/kg dose (0.04 and 0.13 mg/kg payload dose) resulted in complete and sustained tumor regressions, whereas in the trastuzumab-refractory JIMT-1, tumor growth delay was observed (Fig. 4B–D; Supplementary Table S4). In all cases, XMT-2056 outperformed the diABZI STING agonist at a 100-fold lower payload dose equivalent than the small molecule. Overall, neither HT19 nor the nonbinding control ADC showed substantial activity. Thus XMT-2056 induced antigen-dependent activity in tumor models with a broad range of HER2 expression levels.

Figure 4.

XMT-2056 elicits antitumor activity in a broad range of tumor models. A, HER2 expression by IHC in tumor xenograft models (human HER2) and syngeneic tumors (rat HER2); see Supplementary Table S3 for quantitation. Scale bar, 20 µm. B–D, SCID beige mice bearing subcutaneous HCC1954 xenograft tumors (B) or CB.17 SCID mice bearing subcutaneous SNU-5 (C) or JIMT-1 (D) xenograft tumors were intravenously administered a single dose (black arrowhead) of XMT-2056, nonbinding control ADC, or HT19 antibody, or three doses of the diABZI STING agonist (orange arrowheads). Each point represents the mean tumor volume and SEM (n = 10). E, FVB/NJ immune-competent mice bearing syngeneic mBR9013 subcutaneous tumors derived from a spontaneous tumor in MMTV-ERBB2 FVB mouse expressing rat HER2 were intravenously administered a single dose of XMT-2056 surrogate ADC targeting rat HER2, nonbinding control ADC, or three doses of the diABZI STING agonist. Each point represents the mean tumor volume and SEM (n = 10). F, BALB/c immune-competent mice bearing syngeneic EMT-6 subcutaneous tumors that were engineered to express rat HER2 (EMT-6–rHER2) were treated as described in E. Each point represents the mean tumor volume and SEM (n = 10).

XMT-2056 was evaluated in two syngeneic murine mammary carcinoma models: EMT-6, engineered to express rat HER2, and mBR9013, which was derived from a transgenic mouse that expresses rat HER2 under the MMTV promoter and is refractory to anti–PD-1 (Supplementary Fig. S6). IHC confirmed the expression of rat HER2 (Fig. 4A). Because the HT19 antibody does not recognize rat or mouse HER2 orthologs (Supplementary Table S2), these studies required a surrogate ADC based on a different antibody (which recognizes rat and human HER2) conjugated to the same linker-payload as XMT-2056. The XMT-2056 surrogate showed dose-dependent antitumor activity in both models, inducing complete and sustained tumor regressions with a single administration at ≥1 mg/kg (0.04 mg/kg of payload) and outperforming the diABZI STING agonist (Fig. 4E and F; Supplementary Table S4). These results demonstrate the efficacy of XMT-2056 in immune-competent animals and in a PD-1–refractory tumor.

XMT-2056 improves antitumor activity of HER2-directed agents in combination

XMT-2056 was evaluated in combination with HER2-targeted agents trastuzumab, pertuzumab, and T-Dxd in order to leverage the noncompetitive binding profile of HT19. The combination of XMT-2056 and trastuzumab showed improved activity over single agents in SKOV3, JIMT-1, and SNU-5 (Fig. 5A–C; Supplementary Table S4). Similar results were observed with pertuzumab (Supplementary Fig. S7A and S7B; Supplementary Table S4). The combination of XMT-2056 with T-Dxd resulted in durable regressions in JIMT-1 (Fig. 5D; Supplementary Table S4); the single agent activity of T-Dxd was consistent with previous reports (30). No changes in clinical condition or substantial body weight loss were observed.

Figure 5.

Observed benefit in combining XMT-2056 with other HER2-targeted agents. A–C, Combination of XMT-2056 and trastuzumab. CB.17 SCID mice bearing subcutaneous SKOV3 (A), JIMT-1 (B), or SNU-5 (C) xenograft tumors were administered in 3 weekly doses (A and B) or a single dose (C) of XMT-2056, nonbinding control ADC, trastuzumab, or the combinations indicated. ADCs were administered intravenously, whereas trastuzumab was administered intraperitoneally. Each point represents the mean tumor volume and SEM (n = 10). D, Combination of XMT-2056 and T-Dxd. CB.17 SCID mice bearing subcutaneous JIMT-1 xenograft tumors were intravenously administered XMT-2056, nonbinding control ADC, T-Dxd, or a combination of XMT-2056 and T-Dxd. STING agonist ADCs were administered as a single dose, whereas T-Dxd was administered twice, 1 week apart (red triangles). Each point represents the mean tumor volume and SEM (n = 10).

Combination of XMT-2056 surrogate with immune checkpoint inhibitor anti–PD-1 enhances immunological memory

PD-L1 was upregulated in SKOV3 tumors after treatment with XMT-2056 (Fig. 6A; Supplementary Fig. S4E), which highlights the rationale for combining XMT-2056 with anti–PD-1. EMT-6–rHER2 tumor-bearing mice were treated with a mouse anti–PD-1 (5 mg/kg, biweekly × 2), XMT-2056 surrogate (0.3 mg/kg, single dose), or the combination. Additionally, anti–PD-1 was dosed in combination with the control ADC. The combination of XMT-2056 surrogate and anti–PD-1 resulted in complete and sustained responses in 8 of 10 mice, compared with 5 of 10 for XMT-2056 surrogate, and 0 of 10 for anti–PD-1 alone or in combination with the control ADC (Fig. 6B and C; Supplementary Fig. S8A). No changes in the general condition or body weight of the mice were noted (Supplementary Fig. S8B).

Figure 6.

Combination of XMT-2056 with immune checkpoint inhibitor anti–PD-1 elicits tumor clearance and immunologic memory in the EMT-6–rHER2 syngeneic model. A, Normalized counts for PD-L1 human mRNA and mouse mRNA in SKOV3 xenograft tumors. B, Study design including efficacy and rechallenge. Tumor implantations are indicated by circles on the left and right flanks of the mice. C, BALB/c immune-competent mice bearing syngeneic EMT-6–rHER2 engineered tumors were treated with XMT-2056 surrogate ADC targeting rat HER2, anti-mouse PD-1 (clone RPM1-14), or the indicated combinations. The ADCs were administered as single doses, whereas the mouse anti–PD-1 was administered twice weekly for 2 weeks, as indicated by the red triangles. Each point represents the mean tumor volume and SEM (n = 10). D, Tumor-free animals from the XMT-2056 surrogate ADC (blue) and the combination (green) groups were implanted with EMT-6 parental cells on the left flank (opposite the original EMT-6–rHER2 implantation) and CT26 cells on the right flank. Age-matched untreated mice were included as controls. Right, tumor growth in individual mice in each group after implantation on day 0 plotted relative to age-matched naïve mice. CR, complete response; LF, left flank; RF, right flank. (B, Created in BioRender. Cetinbas, N. [2025], https://BioRender.com/k61r850.)

To evaluate immunologic memory, tumor-free mice were rechallenged with the parental EMT-6 cells on the opposite flank of the original implant and the unrelated CT-26 colon cancer cells on the original flank (Fig. 6B). As a control, age-matched naïve mice were implanted with both tumor cell lines. In mice that had been treated with XMT-2056 surrogate, two of five mice rejected the parental EMT-6 tumors, and all mice showed slowed tumor growth compared with naïve mice (Fig. 6D). In mice that had been treated with XMT-2056 surrogate + anti–PD-1, all four mice rejected the implanted EMT-6 cells (Fig. 6D). The rejection of parental (antigen-negative) EMT-6 is consistent with an adaptive immune response against the cancer cell following the initial antigen-dependent activation of innate immunity by the ADC. This mechanism could enable the ADC to suppress disease and recurrence in an antigen-independent manner. As expected, the immunologic memory is specific to the EMT-6 cells: the CT-26 tumors were not rejected because the mice had not been exposed to them in the first phase of the study. These results demonstrate the tumor-specific immunologic memory induced by surrogate XMT-2056.

Toxicologic assessment

XMT-2056 was evaluated in repeat-dose studies in cynomolgus monkeys in order to understand the safety and toxicokinetics profiles. Importantly, the STING agonist payload binds with high affinity to cynomolgus monkey STING (Supplementary Table S1), and the antibody HT19 is fully cross-reactive to cynomolgus HER2 (Supplementary Table S2). No adverse clinical or histopathologic signs were observed at any of the dose levels assessed, and there were no findings attributable to the expression of HER2 in normal tissues. The toxicokinetics profile in monkeys was characterized by slow clearance and high stability, based on similar kinetics for total antibody and conjugated drug (Supplementary Fig. S5B), and was consistent with the pharmacokinetics profile in mice (Supplementary Fig. S5A).

Discussion

XMT-2056 was designed to overcome the challenges encountered in clinical development with intratumoral or intravenous injection of small-molecule STING agonists. In this preclinical study, XMT-2056 simultaneously exhibited increased antitumor activity and reduced elevation of systemic cytokines compared with a clinical benchmark diABZI STING agonist (Fig. 3A and D).

XMT-2056 delivers STING agonist to antigen-expressing cancer cells and Fcγ-R–expressing myeloid cells both in an antigen-dependent manner. The findings in this study and our related study (17) are consistent with recent reports that cancer cell–intrinsic STING can contribute to antitumor responses (31, 32). The cancer cell contribution to the antitumor activity of XMT-2056 was demonstrated in vivo by gene expression changes in cancer cells and by the activity of Fc-mutant ADC, which is not internalized into immune cells (Fig. 3B and F). The activation of STING in both cell types has important implications for the clinical development strategies of STING agonists, including patient selection biomarkers, pharmacodynamics biomarkers, and the selection of tumor types; it also distinguishes the STING pathway from other innate immune pathways such as TLR7/8, which have not been found active in cancer cells (33).

ADC delivery to myeloid cells is consistent with studies that revealed Fcγ-R–mediated internalization into innate immune cells (34, 35). The IgG₁ isotype used in XMT-2056 is also used in most approved ADCs; thus, it can be inferred that cytotoxin-based IgG₁ ADCs also deliver payload to tumor-resident myeloid cells. Whereas the delivery of certain cytotoxins to myeloid cells may confer benefit such as the activation of dendritic cells (36, 37), these benefits are likely incremental compared with the effect of direct delivery to cancer cells and might be counterproductive in some cases.

XMT-2056 conferred immune-mediated killing of HER2-negative cancer cells in the presence of HER2-positive cells, including when the negative cells outnumbered the positive cells by 4:1 (Fig. 2G). Based on our related studies with STING agonist ADCs, this killing can be mediated by soluble factors released in immune cell/cancer cell cocultures following antigen-dependent STING activation in monocytes by the ADC (17). Thus, this mechanism is distinct from the classical ADC “bystander effect” in which a cytotoxic payload delivered to antigen-positive cells diffuses or is transported out of those cells and into neighboring cells (38, 39). Conceptually, immune-mediated cell killing could be more effective than the bystander effect, as it is not limited by payload quantity. The results suggest that XMT-2056 has the potential to elicit meaningful clinical activity in tumors with heterogeneous antigen expression.

XMT-2056 activated multiple immunologic pathways in mouse-derived cells (e.g., immune cells), including those involved in inflammation, dendritic cell function, macrophage function, and antigen processing. XMT-2056 induced inflammatory cytokine/chemokine expression in both immune (mouse) and cancer (human) cells and the tumor infiltration of CD68⁺ macrophages. These observations parallel the increases in CD68 positivity in tumor biopsies from patients following treatment with intratumoral STING agonist; the CD68 increases were associated with increases in CD8 T-cell positivity (40).

Immunologically “cold” or “non–T cell–inflamed” tumors tend to be abundant in immunosuppressed myeloid cell populations such as tumor-associated macrophages and myeloid-derived suppressor cells (41, 42). XMT-2056 delivers the STING agonist to these myeloid cells (Fig. 1I and J) and thus has the potential to overcome the immune-suppressive phenotype (4, 12, 43). The antitumor activity observed in immune-deficient mice that lack B and T cells demonstrates the effectiveness of an innate immune response, whereas the induction of tumor-specific immunologic memory in immune-competent animals demonstrates how an adaptive immune response can augment the innate immune response. Importantly, the mechanism of ADC delivery avoids the undesirable delivery of the STING agonist payload to B and T cells, in which STING activation can be deleterious (18–20). Thus, delivery of a STING agonist via ADC enables an adaptive immune response without the counterproductive activation in B and T cells.

We hypothesized that a tumor-directed approach to deliver a STING agonist payload to cancer cells has the potential to reduce off-tumor inflammatory effects while maintaining potent antitumor activity. Indeed, the clinical benchmark diABZI STING agonist resulted in up to 10-fold higher induction of systemic cytokines than XMT-2056 at the same doses at which the diABZI also had inferior antitumor activity (Fig. 3). Safety evaluation in nonhuman primates showed that XMT-2056 is tolerated at a range of dose levels with no adverse clinical or histopathologic observations, no findings attributable to the expression of HER2 in normal tissues, and a toxicokinetic profile demonstrating high stability of the ADC in circulation. Immunogenicity was not interrogated in nonclinical species due to confounding factors when administering a human protein to a nonhuman subject; immunogenicity is being monitored in the ongoing clinical trial.

XMT-2056 induced antitumor activity in a broad range of tumor models, including HER2 high, medium, and low; the observed potent activity against HER2-low tumors and the potential for concomitant immune-mediated killing of HER2-negative cells suggest that XMT-2056 may be effective against tumors that evaded prior HER2-directed therapies by many of the reported mechanisms (44). Therapeutic resistance based on altered oncogenic signaling will likely also be overcome by STING pathway activation as a distinct mechanism for antitumor activity.

The deliberate choice of an antibody that does not compete for binding with trastuzumab or pertuzumab expands the potential for therapeutic combinations with XMT-2056. Interestingly, the benefit of combining XMT-2056 with the HER2-directed agents trastuzumab and pertuzumab was most impactful in the trastuzumab-refractory JIMT-1 model (Fig. 5B; Supplementary Fig. S7A). We speculate that potential mechanisms underlying the benefit of this combination include enhanced receptor clustering which promotes Fc-mediated myeloid cell uptake (45), increased binding and internalization into antigen-expressing cells (46–48), and enhanced ADCC activity of trastuzumab via STING-mediated recruitment of NK cells to the TME (49–52). By extension, the benefit of combining XMT-2056 with T-Dxd (Fig. 5D) may be explained by the above mechanism(s) driven by antibody, a mechanism mediated by the immune-stimulating aspects of the cytotoxic payload, such as immunogenic cell death (53–55), and/or the observation that trastuzumab-resistant cancer cells suppress the STING pathway (56).

The biological rationale for combining STING activation with immune checkpoint inhibitors is supported by the demonstrated benefit of such combinations in several settings (7, 14, 57–59). In a phase 1b study, combination of intratumorally administered STING agonist with the anti–PD-1 spartalizumab in patients with solid tumors showed higher overall response rate (13.4%; ref. 60) than observed with single-agent STING agonist (2.1%; ref. 40). Consistent with these reports, the preclinical data reported in this article demonstrate that the combination of XMT-2056 and anti–PD-1 antibody enhances antitumor activity as well as immunologic memory (Fig. 6B–D; Supplementary Fig. S8).

In summary, we have leveraged the ADC modality to deliver a STING agonist to the TME and in particular to cancer cells and tumor-resident myeloid cells but not B or T cells. In these preclinical studies, XMT-2056 resulted in improved antitumor activity and reduced systemic inflammation relative to a clinical benchmark–free STING agonist across a broad range of tumor models. XMT-2056 leverages the unique biological features of the STING pathway, resulting in contributions to antitumor activity from both cancer and myeloid cells. A phase 1 dose escalation study with XMT-2056 in patients expressing HER2⁺ tumors is currently underway (NCT05514717).

Supplementary Material

Supplementary Figure S1

Figure S1. Characterization of XMT-2056

Supplementary Figure S2

Figure S2. A. STING mediated IRF3 activity of THP1 reporter cells expressing the indicated human STING1 haplotypes or STING1 knockout (KO) after treatment with STING agonist at the indicated doses. Each point represents the mean and SD (n=2). B. Cytokine induction as measured by a multiplex Luminex assay from supernatants of fresh human white blood cells treated for 6 (IFN-β) or 24 (CXCL10, IL-6, TNF-α) hours with indicated concentrations of STING agonist. Bars represent mean value of n=2 data points shown as symbols. C. Structure of the human HER2 extracellular domain showing HT19 epitope mapped to domain IV through shotgun mutagenesis using an Alanine Scanning Mutagenesis. D. Lack of competition with pertuzumab by bio-layer interferometry binning assay (Octet). Pertuzumab was loaded onto the sensor chip, and HER2 ECD and HT19 associations indicated by blue arrows. Additional binding by HT19 antibody indicates non-competitive binding. E. HER2+ NCI-N87 gastric cancer xenograft model treated with Dolaflexin (DF)-based ADCs including conjugates of trastuzumab or non-binding isotype control ADC at the indicated doses. Each point represents mean tumor volume and SEM (n=10). F. Bar graphs of % inhibition in phospho-AKT in SKBR3 cells for HT19, trastuzumab, and lapatinib. Each bar represents mean and SD of 2 independent experiments. G. Binding of XMT-2056, HT19, and hIgG1 isotype control antibody to SKBR3, JIMT-1, MDA-MB-453, SNU-5, and MDA-MB-175-VII cells showing fluorescence intensities measured by flow cytometry. Each point represents mean and SD (n=3). H. Graphical traces of binding kinetics of XMT-2056, HT19, or Fc mutant XMT-2056 to human Fcγ-RI recombinant protein, showing association and dissociation phases of binding by Octet. I. IRF3 reporter activity of THP1 cells in monoculture (Ieft) or cultured in plates in the absence of HER2 antigen (right) treated for 24 hours with XMT-2056, Fc mutant XMT-2056, non-binding control ADC, or free payload. Each point represents mean and SD (n=3).

Supplementary Figure S3

Figure S3. A. Cancer cell death induced by XMT-2056, HT19 antibody, or non-binding control ADC, shown as percent viable SKBR3-NR (left) or MDA-MB-175-VII-NR (right) cells in PBMC co-cultures (84 hr time point). Each point represents mean and SD (n=3). B. Cytokine induction by XMT-2056, HT19, or non-binding control ADC measured in supernatants of SKBR3-NR (top) or MDA-MB-175-VII-NR (bottom) cells in PBMC co-cultures (24 hr time point). Each point represents mean and SD (n=3). C. Cancer cell death by XMT-2056, non-binding control ADC, or STING agonist payload shown as percent viable cells in monocultures of MDA-MB-231-NR (left) and SKBR3-NR (right) (84 hr time point). Each point represents mean and SD (n=3).

Supplementary Figure S4

Figure S4. A. HER2 expression by IHC in SKOV3 xenograft tumors; scale bar, 20 µm. B. Percent changes in body weights of SKOV3 tumor-bearing CB.17 SCID mice administered a single dose (black arrowhead) of XMT-2056, non-binding control ADC, HT19, or STING agonist payload, or 3 doses (orange arrowhead) of the diABZI STING agonist. Each point represents the mean change in body weight and SEM (n=10). C. Normalized counts for mouse mRNA and human mRNA of individual tumor cytokine/chemokines in SKOV3 xenografts harvested 12 hours after treatment. Each point represents the mean and SD (n=2). D. Representative images of IHC for leukocyte marker CD45 and macrophage marker CD68 in SKOV3 tumors after treatment with vehicle, 1 mg/kg XMT-2056, or 1 mg/kg control ADC, and collected 72 hr after treatment. Scale bar, 3 mm. Images are provided for 3 animals per group. The red boxes indicate the tumors shown at higher magnifications in main Figure 3H. E. Representative images of IHC for dendritic cell marker CD11c and murine PD-L1 in SKOV3 tumors collected 12 hrs (CD11c) or 72 hrs (mPD-L1) after treatment. Scale bar, 100 µm.

Supplementary Figure S5

Figure S5. A. PK profile of XMT-2056 in non-tumor bearing CB.17 SCID mice as a function of total antibody (Total Ab) and conjugated drug concentrations in plasma after a single intravenous administration. Each line represents an individual animal (n=4). B. PK profile of XMT-2056 in cynomolgus monkey as a function of total antibody and conjugated drug concentrations in plasma during Q3W repeat-dose administration indicated by the black arrowheads. 1 male and 1 female animal were dosed; each line represents data from one animal.

Supplementary Figure S6

Figure S6. FVB/NJ immune competent mice bearing syngeneic mBR9013 subcutaneous tumors were intravenously administered anti-mouse PD-1 (clone RPM1-14) 10 mg/kg twice weekly for 3 weeks (red triangles). Each point represents the mean tumor volume and SEM (n=5).

Supplementary Figure S7

Figure S7. A. CB.17 SCID mice bearing subcutaneous JIMT-1 xenograft tumors were administered 3 weekly doses of XMT-2056, a combination of non-binding control ADC and pertuzumab, or a combination of XMT-2056 and pertuzumab. B. CB.17 SCID mice bearing subcutaneous SNU-5 xenograft tumors were administered a single dose of XMT-2056, non-binding control ADC, pertuzumab, or a combination of XMT-2056 and pertuzumab. ADCs were administered intravenously while pertuzumab was administered intraperitoneally. Each point represents the mean tumor volume and SEM (n=10).

Supplementary Figure S8

Figure S8. A, B. BALB/c immune competent mice bearing syngeneic EMT-6-rHER2 tumors were treated with XMT-2056 surrogate ADC, anti-mouse PD-1 (Clone RPM1-14), non-binding control ADC in combination with anti-mouse PD-1, or XMT-2056 surrogate ADC in combination with anti-mouse PD-1. The ADCs were administered once while the mouse anti-PD-1 was administered twice weekly for 2 weeks as indicated by the red triangles. A. Tumor volumes of the individual mice, with the number of complete responders (CR) indicated. B. Percent change in body weight (BW). Each point represents the mean change in BW and SEM (n=10).

Supplementary Table S1

Table S1. Summary of binding (SPR) and functional (EC50) activities of the STING agonist. IRF3 reporter and cytokine assays performed at 24 hours unless otherwise indicated based on signal peaks

Supplementary Table S2

Table S2. EC50 values for binding of XMT-2056 or HT19 antibody to human and nonclinical species HER2 extracellular domain. Highest dose tested is 100 nM by antibody. Values represent mean of 2 independent experiments.

Supplementary Table S3

Table S3. HER2 expression in cell lines and tumor xenografts, including gene expression by RNAseq (log2 transformed); binding EC50 and maximal fluorescence intensity (Bmax FL) of XMT-2056 by flow cytometry, and HER2 immunohistochemistry in tumor xenografts. EC50 and Bmax values represent mean of 2 independent experiments.

Supplementary Table S4

Table S4

Authors’ Disclosures

R.A. Bukhalid reports a patent for Mersana Therapeutics and holds patents and patent applications relevant to this work pending, issued, and licensed to various institutions. J.R. Duvall reports personal fees from Mersana Therapeutics outside the submitted work; in addition, J.R. Duvall has a patent for ADCs comprising STING agonists pending. K. Lancaster reports other support from Mersana Therapeutics outside the submitted work. K.C. Catcott reports other support from Mersana Therapeutics outside the submitted work and is employed by Mersana Therapeutics. N. Malli Cetinbas reports personal fees and other support from Mersana Therapeutics during the conduct of the study; in addition, N. Malli Cetinbas has a patent for Mersana Therapeutics pending and issued. J.D. Thomas reports personal fees from Mersana Therapeutics outside the submitted work; in addition, J.D. Thomas has a patent for Mersana Therapeutics and holds patents and patent applications relevant to this work pending. L. Xu reports a patent for W02021202984A1 issued and licensed to Mersana Therapeutics. D. Toader reports a patent for certain pending patent applications and granted patents pending. M. Damelin reports a patent for Mersana Therapeutics and holds patents and patent applications relevant to this work pending. T.B. Lowinger reports other support from Mersana Therapeutics outside the submitted work; in addition, T.B. Lowinger has various patents pending and issued to Mersana Therapeutics. No disclosures were reported by the other authors.

Authors’ Contributions

R.A. Bukhalid: Conceptualization, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. J.R. Duvall: Conceptualization, supervision, investigation, writing–original draft, writing–review and editing. K. Lancaster: Data curation, formal analysis, supervision, visualization, writing–original draft, writing–review and editing. K.C. Catcott: Investigation, methodology, writing–original draft. N. Malli Cetinbas: Supervision, investigation, methodology. T. Monnell: Investigation, methodology. C. Routhier: Investigation, methodology, writing–original draft. J.D. Thomas: Investigation. K.W. Bentley: Investigation. S.D. Collins: Supervision, investigation, visualization, methodology, writing–original draft. E. Ditty: Investigation. T.K. Eitas: Investigation, methodology. E.W. Kelleher: Investigation. P. Shaw: Investigation, methodology. J. Soomer-James: Investigation, visualization, methodology, writing–original draft. E. Ter-Ovanesyan: Investigation. L. Xu: Supervision, investigation. J. Zurita: Investigation. D. Toader: Supervision, investigation, writing–original draft. M. Damelin: Conceptualization, supervision, investigation, writing–original draft, writing–review and editing. T.B. Lowinger: Conceptualization, resources, supervision, writing–original draft.