An Integrin β6–Targeted Antibody–Drug Conjugate Optimized for Intravesical Delivery to Treat Non–Muscle-Invasive Bladder Cancer

Immunohistochemistry assessment of IB6 expression

Immunohistochemistry (IHC) for IB6 expression was performed on normal bladder and bladder cancer, including NMIBC subtypes, formalin-fixed, paraffin-embedded (FFPE) tissues. Commercially sourced bladder cancer samples included n = 3/subtype (unless otherwise noted) of Tis/CIS, Ta, T1, T2, T2a, T2b, T3, T3a, T3b (2), T4a, and T4 (1). The IHC procedures involved deparaffinization followed by antigen retrieval in the NxGen decloaking chamber (Biocare Medical) using the 110°C program with Diva Retrieval Solution. Slide staining using the intelliPATH automated stainer (Biocare Medical) involved pretreatment with peroxidase, avidin, biotin, and Sniper protein blocking agents, followed by the application of the primary rabbit anti-IB6 antibody (Seagen M6.2A1, lot #207713-06) at 1:5,000 (3 μL of antibody and 15,000 μL of Da Vinci Green diluent), a secondary HRP anti-rat polymer antibody (Envision), and then incubation with ImmPACT DAB (Vector). The slides were then counterstained with hematoxylin, rinsed, cleared, and cover slipped using the Prisma Plus Stainer. IHC evaluation included confirmatory IB6-positive control tissue (1223170B normal human placenta) and negative isotype control rabbit IgG (Jackson ImmunoResearch ChromPure, lot #011-000-003, RRID: AB_2337118) staining. Images were captured using the Aperio AT2 scanner (Leica, RRID: SCR_021256) and reviewed by an ACVP board–certified veterinary pathologist. IHC evaluation was based on a qualitative scoring assessment of stain intensity (0 = none; 1 = weak; 2 = moderate; and 3 = strong), stain frequency (0 = none; 1 = >0% to 25%; 2 = 26%–50%; 3 = 51%–75%; and 4 = >75% tumor cells), and stain localization (M = membranous; C = cytoplasmic; N = nuclear; A = apical; and G = Golgi). Specific IHC staining was determined based on the stain characteristics and in consideration of comparative negative and positive control samples.

IHC staining for IB6 expression on orthotopic bladder tumors was performed at Pfizer using rabbit anti-IB6 antibody clone SG44 20B8 A3 and Thermo Fisher CD147 clone 125 (RRID: AB_2784992). The slides were baked in an oven for 1 hour at 61°C prior to the IHC run. The IB6 SG44 20B8 A3 antibody was conjugated with Alexa Fluor 488 using the Invitrogen Alexa Fluor Antibody Labeling Kit (Thermo Fisher Scientific, 704060). Isotype rabbit IgG (Abcam, 172730, RRID: AB_2687931) was used as a negative control. The samples were run on the BOND RX Autostainer (Leica) at ambient temperature following the manufacturer’s instructions with the BOND Polymer Refine Detection Kit (Leica, DS9800). Images were captured using the Hamamatsu NanoZoomer S360 Digital Slide Scanner (RRID: SCR_023761) and uploaded into OMERO (RRID: SCR_002629) for image review.

RNA expression and protein abundance

The frozen tumors (n = 13) were obtained from BioIVT. Each tumor was split in half for DIA-MS proteomics and RNA-seq. RNA-seq was performed via GENEWIZ NGS Services from Azenta Life Sciences for paired-end sequencing. Proteomics analysis followed a standard global proteomics workflow. The samples were homogenized and lysed in 1% SDS/RIPA buffer with Precellys beads. Proteins were cleaned and digested following single-pot solid phase-enhanced sample preparation (SP3). Data acquisition was performed on a Zeno SWATH DIA platform (SCIEX ZenoTOF7600) coupled with a 30-minute LC gradient (Waters ACQUITY UPLC M-Class). Data analysis was performed in DIA-NN version 2.2 (RRID: SCR_022865) using an in silico spectral library. Gene-level quantification was reported with DIA-NN’s embedded QuantUMS algorithm.

Generation of IB6 antibodies, drug linker, and conjugation of PF-08052667

Detailed experimental procedures for the generation and humanization of 2A2 and the preparation of the drug linker have been previously described (24, 31, 32). PF-08052667 was prepared by full reduction of the antibody interchain disulfides followed by the addition of a 50% excess of maleimide–linker payload, as described previously (Supplementary Fig. S1A; refs. 31, 32). Representative analytical results for PF-08052667 are shown in Supplementary Fig. S2. The murine monoclonal 15H3 (m15H3) was generated using standard hybridoma technology following the immunization of mice with target antigen–expressing cells, followed by humanization via CDR grafting. The details are shown in the section below. The resulting h15H3 antibody binds human, cynomolgus monkey, rat, and mouse IB6 protein (Supplementary Table S1), whereas h2A2 binds only human and cynomolgus monkey IB6 (0.2 and 0.3 nmol/L, respectively; ref. (32). The rodent binding profile of h15H3 (0.3–0.4 nmol/L) allowed for the generation of a surrogate ADC with a comparable binding affinity to that of h2A2 that could be used for investigative studies in rodents.

h15H3 antibody generation and humanization

The BALB/c mice were immunized 3 times with intraperitoneal injections of ∼5 × 10⁶ 3T3:huβ6 transfectants. Three days prior to fusion, the mice received a final injection of purified recombinant human αvβ6 that was given intravenously (6 μg) and intraperitoneally (30 μg). Lymphocytes harvested from the spleen and lymph nodes were fused to P3X63Ag8.653 myeloma cells using polyethylene glycol. Fused cells were recovered overnight in hybridoma growth media (Iscove's modified Dulbecco's medium containing 4 mmol/L of glutamine, 10% FetalClone I, 10% cloning factor, and penicillin/streptomycin). Following recovery, cells were spun down and then plated in semisolid media, consisting of CloneMatrix media supplemented with hybridoma growth media plus HAT for hybridoma selection and CloneDetect for IgG production. Hybridomas were incubated for 10 days at 37°C. At day 10, IgG-producing hybridoma clones were picked using a ClonePixFL (Molecular Devices) and transferred to 96-well plates containing IgG-depleted hybridoma growth media plus HT. Hybridoma culture supernatants were screened on 293F:huβ6 transfectants, and positive clones were identified using an Alexa Fluor 647–labeled secondary antibody for detection (Invitrogen, A21445, RRID: AB_2535862). The plates were read on an FMAT 8200 (Applied Biosystems). Hybridomas that bound to 293F:huβ6 and 293F:cynoβ6 but not 293F:vector were expanded for direct conjugation using the method described in International Application No. WO 2011/109308. The directly conjugated antibody panel was tested in binding and cytotoxicity assays. m15H3 was selected as the lead antibody because it showed cytotoxic activity as an ADC on multiple αvβ6-positive tumor cell lines, and it had comparable affinity to human and cynomolgus forms of the antigen. The specificity of mouse 15H3 and humanized 15H3 was confirmed in FMAT and FACS binding studies, in which the antibodies were shown to bind to 293F:huβ6 transfectants but not to the αvβ5-positive parental line (293F:vector). The binding specificity of m15H3 and h15H3 was also confirmed via ELISA, in which the antibodies were bound to recombinant human αvβ6 but not αvβ3 or αvβ8.

The murine 15H3 antibody was humanized using genomic sequences provided by VH1-46 and JH4 for the heavy chain and by VK2-30 and Jk2 for the light chain. In the first humanization round, 13 positions were identified in the heavy chain (H28, H30, H38, H40, H41, H48, H66, H67, H69, H71, H72, H93, and H108), and six positions were identified in the light chain (L21, L22, L36, L37, L45, and L63) at which the human acceptor sequence differed from the donor sequence and that may affect antibody binding as a result of contacting the antigen directly, affecting the conformation of CDRs, or affecting the packing between heavy and light chains. Eight humanized heavy chains and five humanized light chains were made, incorporating back mutations at different permutations of these positions. The remaining positions are occupied by the residues from the human acceptor sequence.

Cell lines and culture

The cancer cells were obtained from suppliers with lot numbers listed in Supplementary Table S2 and grown according to the recommendations provided. The cell lines were expanded upon receipt and cryo-frozen in aliquots between passages 3 and 5. The cells were authenticated and confirmed to be Mycoplasma-negative using IDEXX BioAnalytics, using STR-based DNA profiling and multiplex PCR. All cell lines were cultured at 37°C with 5% CO₂ in the relevant growth media, with assays performed between passages 3 and 22 or less than 6 weeks in culture.

Cross-species cell surface expression and flow cytometry assays

Flow cytometry studies to evaluate species cross-reactivity of the h2A2 antibody were conducted using the following antigen-expressing cell lines: 293F:hu β-6 transfectant, 293F:cyno β-6 transfectant (parental 293F RRID:CVCL_6642), FDC-P1 mouse β-6 transfectant (parental FDC-P1 RRID:CVCL_2039), and NBT-II rat bladder carcinoma cells (RRID: CVCL_0453). For transfected cell lines, the parental cells were included as controls. The cells were incubated with primary antibodies over a concentration range of 0 to 300 nmol/L for 30 minutes at 4°C. NBT-II and FDC-P1 cells were incubated with primary antibodies at 75 nmol/L for 30 minutes at 4°C. The cells were then washed twice with FACS buffer (PBS, 2% FBS, 0.02% NaN₃, 1 mmol/L of MnCl₂ for 293F and FDC-P1 cells; TBS, 2% FBS, 0.02% NaN₃, 1 mmol/L of MnCl₂ for NBT-II cells) before incubation with an Alexa Fluor 488–labeled secondary antibody for 30 minutes at 4°C. The cells were again washed twice and resuspended in FACS buffer before the acquisition of samples on an Attune NxT Flow Cytometer (Thermo Fisher, RRID: SCR_019590).

Saturation binding studies were conducted using the following cell lines: UM-UC-3 bladder cancer parental cell line (<1,000 receptor copies αvβ6 per cell, RRID: CVCL_1783), UM-UC-3 1A2 bladder–engineered IB6-expressing clonal cell line (125,000 receptor copies per cell), or HEK293 cell lines transfected with human and cyno IB6 orthologs (parental HEK293, RRID: CVCL_0045). The cells were aliquoted at 1 × 10⁵ cells per well into a 96-well V-bottom plate. The primary antibodies were added to the cells at concentrations ranging from 100 nmol/L to 0.3 pmol/L in FACS buffer (TBS, 2% FBS, and 1 mmol/L MnCl₂). The cells were incubated for 1 hour and then pelleted and washed 2 times with FACS buffer. The cells were pelleted and resuspended in a 10-μg/mL solution of an FITC-conjugated anti-human Fc secondary antibody (Jackson ImmunoResearch Labs, 109-096-098, RRID: AB_2337668) prediluted in FACS buffer. After a 45-minute incubation, the cells were pelleted and washed 2 times with FACS buffer. Upon resuspension in a 1% paraformaldehyde solution to 1 × 10⁶ cells/mL, fluorescence was detected using an Attune NxT Flow Cytometer. Fluorescence geometric mean was assessed via flow cytometry and calculated with FlowJo version 10.8 Software (RRID: SCR_008520). EC₅₀ was calculated using GraphPad Prism Software (RRID: SCR_002798).

Flow cytometry was used to quantify IB6 protein expression on the surface of cancer cell lines and PDOs using the Human IgG Calibrator Kit (Biocytex, CP010). PDOs and cells were dissociated with TrypLE (Thermo Fisher, 12605010) for 5 to 10 minutes at 37°C and then distributed into 96-well V-bottom plates (2 × 10⁵ cells/well). The cells were stained with h2A2 primary antibody (1 hour, 4°C), anti-IgG-FITC (45 minutes, 4°C), and DAPI (Miltenyi Biotec, 130-111-570; 5 minutes). Samples were acquired within 4 hours on a MACSQuant Analyzer 10 (Miltenyi Biotec, RRID: SCR_020268) or Attune NxT Flow Cytometer and analyzed using FlowJo version 10.8. Antigen-binding capacity (ABC) and specific ABC (SABC) were determined using calibration beads and isotype controls.

Internalization and intracellular trafficking of PF-08052667

Visualization of PF-08052667 on the plasma membrane of SW780 cells (RRID: CVCL_1728) and colocalization with lysosomes after internalization were performed via immunofluorescence. The cells were treated with 2 μg/mL of PF-08052667 and preincubated on ice for 30 minutes. The treated cells were washed with PBS to remove the unbound ADC and then immediately fixed and permeabilized (0-hour time point) or further incubated at 37°C for the 4-hour time point. The cells were stained with an anti-human IgG antibody (Invitrogen, A11013, RRID: AB_2534080) to detect PF-08052667, the anti-LAMP1 lysosome marker (BD Pharmingen, 562622, RRID: AB_2737684), and Hoechst (Invitrogen, 62249) to label nuclear DNA. Images were collected using the ImageXpress High-Content Confocal HT.ai system (Molecular Devices, RRID: SCR_020294).

Internalization was evaluated on IB6-expressing bladder cancer cell lines SW780, UM-UC-5 (RRID: CVCL_2750), UM-UC-3 1A2 IB6, and UM-UC-3 2F4 IB6 via automated fluorescence microscopy in a live cell kinetic assay using quenched fluorophore conjugates (Supplementary Fig. S1B). The cells were seeded in a 96-well plate, and quenched fluorophore (QF) antibody conjugates were added at 2 μg/mL. The antibody backbone of PF-08052667 (h2A2) was conjugated as a two-load to a fluorophore (sulfo-Cy5) and quencher (Tide Quencher %WS, AAT Bioquest) pair through a glucuronide linker (glucQF) to generate the antibody conjugate h2A2glucQF with fluorophore. A nonbinding mAb glucQF conjugate (h00glucQF) was used as a negative control. The fluorescence signal was measured by collecting images every 2 hours using the Incucyte SX5 (Sartorius, RRID: SCR_026298). The amount of Cy5 liberated from the antibody–quencher conjugate was quantified as the total integrated signal intensity and normalized to cell confluence for each time point. The data were graphed using GraphPad Prism software.

Cell viability assay to assess the cytotoxicity of PF-08052667

The cells were thawed from liquid nitrogen storage and allowed to recover until viability, determined via Vi-CELL XR (Beckman Coulter, RRID: SCR_019664), was above 90%. The cells were then plated in 384-well plates and allowed to adhere overnight. Serial dilutions of the ADCs were prepared in RPMI 1640 supplemented with 10% FBS, added to each cell plate in quadruplicate, and incubated for 96 hours. CellTiter-Glo luminescent assay reagent (Promega) was prepared according to the manufacturer’s protocol, and fluorescence was determined using the EnVision Multilabel Plate Reader (Perkin Elmer). The data were analyzed in GraphPad Prism using a nonlinear, four-parameter curve-fit model [Y = Bottom + (Top-Bottom)/(1 + 10^((LogEC₅₀^{− X) × HillSlope)})]. Cell viability was tested on BxPC3 (RRID: CVCL_0186), Detroit 562 (RRID: CVCL_1171), EBC-1 (RRID: CVCL_2891), FaDu (RRID: CVCL_1218), HPAF-II (RRID: CVCL_0313), SW780, UM-UC-5, A549 (RRID: CVCL_0023), A549 IB6 2B5, A549 3E8, UM-UC-3, UM-UC-3 IB6 1A2, and UM-UC-3 IB6 2F4.

PDO establishment and culturing

Bladder tissue samples for organoid establishment were collected during surgery with approval from the ethics committees of the University Medical Centre Utrecht (Biobanking protocol HUB-Cancer, TCBIO, 12-093). Samples were obtained from 3 hospitals: University Medical Centre Utrecht, Meander Medical Centre Amersfoort, and Erasmus University Medical Centre Rotterdam. Written informed consent was obtained from all patients. Most bladder PDOs originated from low-grade NMIBCs (pT0, pTis, pTa, and pT1), as determined by pathologic assessment. The tumor tissue was cut into small pieces (1–2 mm) using a surgical blade and digested with Collagenase II in the bladder organoid medium supplemented with ROCK inhibitor (5 μmol/L). After centrifugation, the pellet was embedded in 80% BME and plated in prewarmed six-well plates. Tumor cells collected from urine samples were washed and plated directly. Once the BME solidified, the bladder organoid medium was added. The organoids were passaged every 14 to 21 days using mechanical dissociation by pipetting. ROCK inhibitor (10 μmol/L) was added after passaging to prevent anoikis. The medium was refreshed 3 times per week. The organoids were cryopreserved in Recovery Cell Culture Freezing Medium (Thermo Fisher, 12648010). Organoid identity was confirmed using a single-nucleotide polymorphism (SNP) array targeting 60 SNPs with TaqMan OpenArray technology and the QuantStudio 12K Flex Real-Time PCR System at the Utrecht Sequencing Facility. The organoids were routinely tested for Mycoplasma contamination, and they consistently tested negative.

Bladder organoid media

The human bladder organoid medium is composed of Advanced DMEM/F-12 (Thermo Fisher, 12634010), HEPES buffer (10 mmol/L, Lonza, 17737E), penicillin/streptomycin (50 U/mL, Thermo Fisher, 15070-063), GlutaMAX (2 mmol/L, Thermo Fisher, 35050038), FGF10 (100 ng/mL, PeproTech, 100-26), FGF7 (25 ng/mL, PeproTech, 100-19), FGF2 (12.5 ng/mL, PeproTech, 100-18B), B27 supplement (1x, Thermo Fisher, 17504001), A83-01 (5 μmol/L, Tocris, 2939), N-acetylcysteine (1.25 mmol/L, Sigma-Aldrich, A9165), nicotinamide (10 mmol/L, Sigma-Aldrich, N0636), ROCK inhibitor (5 μmol/L, Y-27632, AbMole, M1817-50), and Primocin (50 μg/mL, InvivoGene SAS, ANT-PM-2).

PDO viability assay to assess the potency of PF-08052667

Twenty-nine bladder cancer PDOs were screened for drug sensitivity. For screening, PDOs were mechanically dissociated and replated 2 to 6 days prior, depending on growth. On days 2 to 6, PDOs were harvested by incubating with Dispase II (1 mg/mL; Thermo Fisher, 17105041) for 30 minutes at 37°C to digest the basement membrane extract (BME; Bio-Techne, 3533-010-02). The PDOs were resuspended in bladder organoid medium supplemented with ROCK inhibitor (Y-27632, 5 μmol/L; AbMole Bioscience #M1817) and 5% BME. Using a Multidrop Combi dispenser, 375 organoids were plated in 384-well ultralow-attachment plates (Corning, 4588). A nine-dose drug concentration range was added using a Tecan Fluent liquid handler. Cell viability was measured on day 6 using CellTiter-Glo 3D (Promega, G9681) on a Tecan Spark multimode microplate reader (RRID: SCR_021897); baseline viability was measured on day 0 in untreated controls. Screens were performed in technical triplicates. Percentage viability and growth rate (GR) inhibition metrics were calculated (33), and IC₅₀, GR₅₀, and GR_AOC values were determined.

ATP release assay

The SW780 cells were plated at 2 × 10⁴ cells/well in the 96-well plates and treated with 1 μg/mL of ADC, antibody, or 10 nmol/L of MMAE. The extracellular ATP was measured 70 hours after treatment using the RealTime-Glo Extracellular ATP Assay (Promega) according to the manufacturer’s protocol, and luminescence was measured using the Cytation 5 Cell Imaging Multimode Reader (Agilent BioTek, RRID: SCR_019732). ATP release is reported as fold change over untreated cells, normalized to background signal.

High-mobility box 1 release assay

The SW780 cells were plated at 2 × 10⁵ cells/well in the 12-well tissue culture–treated plates and treated with 1 μg/mL of ADC, antibody, or 10 nmol/L of MMAE. High-mobility box 1 (HMGB1) release was measured 96 hours after treatment via standard ELISA according to the manufacturer’s protocol (Tecan, 30164033). The optical density was measured at 450 nm with an Envision Multimode Plate Reader (Perkin Elmer, RRID: SCR_018038). HMGB1 release is reported as fold change over the signal of untreated cells.

Cell surface calreticulin

The SW780 cells were plated at 2 × 10⁵ cells/well in the 12-well tissue culture–treated plates and treated with 0.15 μg/mL of ADC, antibody, or 1.5 nmol/L of MMAE. Then, 120 hours after treatment, the cells were harvested using nonenzymatic dissociation buffer. The cells were spun down at 300 × g for 3 minutes and resuspended in staining buffer consisting of FACS buffer [(DPBS + 2% FBS) + Calreticulin AF647 (1:50), Abcam, AB196159, RRID: AB_2819061]. The cells were incubated on ice for 30 minutes, protected from light. They were washed twice with FACS buffer and then resuspended in staining buffer consisting of 1x binding buffer + Annexin V-PE (1:20) and 7-AAD (1:20; BioLegend, 64093). The cells were incubated for 15 minutes at room temperature, protected from light, and then analyzed on an Attune NxT Flow Cytometer. Calreticulin was measured in live Annexin V⁺7-AAD^−/lo gated cells using FlowJo version 10.8.

Animal use ethics statement

Studies with cell line–derived xenografts were performed at Pfizer. Pfizer is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. All experiments involving animal handling and study designs were approved under Institutional Animal Care and Use Committee guidelines and appropriate animal research approvals. Studies were conducted in compliance with facility standard operating procedures and in accordance with the Department of Agriculture Animal Welfare Act and the National Research Council’s Guide for the Care and Use of Laboratory Animals. All animals were purpose-bred and experimentally naïve at the outset of the studies.

Intravesical dosing and translation across species

Intravesical dosing uses a transurethral catheter to empty residual urine and deliver test materials directly into the bladder. In small animals, especially male rodents, this can be challenging due to complex urethral anatomy, which may increase procedural effects and variability. Larger animals allow for easier administration with fewer procedural effects. However, through optimization of the dosing method and the use of flexible urinary catheters, these procedural effects can be reduced, making female rodent models feasible for studies on orthotopic activity and tissue permeation.

Supplementary Table S3 displays the cross-species translation of intravesical doses based on the total dose/bladder surface area (mg/cm²) utilized in the development program. Dose scaling on body surface area is not appropriate due to limited systemic exposure and the local administration dose route; instead, the total dose has been normalized to bladder tissue surface area (34). Bladder surface area was derived from the volume of the bladder in which V = (4/3)πr³ and SA = 4πr² with the estimates based on the average static bladder size: specifically V_mouse = 0.15 mL and SA_mouse = 1.37 cm², V_rat = 1 mL and SA_rat = 4.83 cm², V_NHP = 61.8 mL and SA_NHP = 75.43 cm², and V_Human = 500 mL and SA_Human = 304.19 cm².

Bladder tissue penetration studies using a rodent cross-reactive IB6 ADC surrogate

To assess the effects of bladder prewash prior to ADC administration on MMAE levels and penetration into the bladder tissue, a rat cross-reactive surrogate IB6 ADC, h15H3-5088(8), was administered via a single intravesical dose with 3Fr closed-end feline urinary catheters (Tom Tiddle, HB02221A). Only female rats were used as urethra anatomical differences in male rodents make catheterization challenging, causing procedural damage. After dosing 1.2 mg/cm² of the surrogate ADC for a 60-minute dwell time ± bladder prewash of 0.1% n-dodecyl-β-D-maltoside (DDM), the bladder tissue was collected after 1, 6, 48, 120, or 168 hours (n = 2/time point) after dose. The tissue was flash-frozen for bioanalytical analysis of the total unconjugated MMAE from homogenized bladder tissue and/or fixed for immunohistochemical staining with an anti-MMAE antibody to examine the penetration of the drug within the bladder tissue layers. When both IHC and bioanalytical methods were to be performed, removed bladders were halved at the midline and fixed or frozen as appropriate. To assess the effect of dwell time ± bladder prewash, surrogate ADC was administered as a single intravesical dose of 0.8 mg/cm² for dwell times of 15, 30, 60, or 90 minutes. The bladder tissue was collected 24 hours after the dose (n = 3/group) for bioanalytical analysis. The total unconjugated MMAE in the bladder tissue was quantitated via solid-phase extraction and reverse-phase LC–MS/MS following tissue homogenization and protein precipitation. Systemic exposures were evaluated in rat plasma for total antibody; however, the observed exposures were highly variable between individuals, and in the 7-day kinetic study, they represented less than 6.2% of the administered dose.

IHC for MMAE was performed on FFPE rat urinary bladder tissues from an exploratory rat pharmacokinetic study involving intravesical treatment with the rat cross-reactive IB6 ADC surrogate, h15H3-5088(8), at 1.2 mg/cm². The IHC procedures involved deparaffinization followed by antigen retrieval in the NxGen Decloaking Chamber (Biocare Medical) using the 110°C program with Diva Retrieval Solution. Slide staining using the intelliPATH automated stainer (Biocare Medical) involved pretreatment with peroxidase, avidin, biotin, and Sniper protein blocking agents, followed by the application of the primary mouse anti-MMAE mAb (Seagen 15.22.F4.1B10.1C10) at 2 μg/mL, a secondary biotinylated goat anti-mouse antibody (Jackson Immuno, 715-065-151, RRID: AB_2340785) at 1:1,000, and then incubation with Vector ABC Elite HRP and DAB. The slides were then counterstained with hematoxylin, rinsed, cleared, and cover slipped using the Tissue-Tek Prisma Plus Stainer. IHC evaluation included confirmatory MMAE-positive control tissue (rat liver) and negative isotype control rabbit IgG (Abcam 172730) staining. Images were captured using the P1000 slide scanner (3DHISTECH) and reviewed by an ACVP board–certified veterinary pathologist. IHC evaluation was based on a qualitative scoring assessment of stain intensity (0 = none; 1 = weak; 2 = moderate; and 3 = strong), stain frequency (0 = none; 1 = >0% to 25%; 2 = 26%–50%; 3 = 51%–75%; and 4 = >75% tumor cells), and stain localization (M = membranous; C = cytoplasmic; N = nuclear; A = apical; and G = Golgi). Specific IHC staining was determined based on the stain characteristics and in consideration of comparative negative and positive control samples.

Nonclinical safety evaluation

PF-08052667 was evaluated for toxicologic effects in an exploratory non-GLP single-dose study (n = 1/sex/group) and a weekly for 6-weeks (Q1Wx6, n = 3/sex/group) GLP-compliant repeat-dose toxicology study in the pharmacologically relevant cynomolgus monkey model (Macaca fascicularis). Doses up to 1.3 mg/cm² of PF-08052667 were administered via intravesical administration with or without a 0.1% DDM prewash. DDM was evaluated in a manner consistent with the intended clinical use, as a mucosal barrier prewash given as a short-duration intravesical instillation (up to 10 minutes) alone. Following the prewash, when applicable, vehicle or ADC, PF-08052667, was administered intravesically. A previously performed non-GLP intravenous single-dose toxicity study was also leveraged to identify potential toxicities secondary to systemic exposure (n = 2/sex/group). However, due to the lack of systemic exposure observed upon intravesical administration, this study is not considered necessary for development but was incorporated as previously available. Animals were assessed for evidence of toxicity, including mortality, clinical observations, body weights, and clinical pathology (hematology, coagulation, clinical chemistry, and urinalysis). At study termination, macroscopic evaluations, organ weights, and tissue samples were collected for microscopic evaluation. Systemic exposures were evaluated for total antibody in plasma via ELISA or Gyrolab immunoassay (Gyros Protein Technologies) with lower limits of quantitation (LLOQ) between 2 and 100 ng/mL, and for the total unconjugated MMAE in plasma via LC–MS/MS with an LLOQ of 0.01 ng/mL. In the 6-week GLP study, the bladder tissue was collected 1 week after the last dose when terminal collections for histopathology were planned, and local tissue exposure was evaluated for the total unconjugated MMAE via LC–MS/MS with an LLOQ of 0.1 ng/g.

Orthotopic murine models

The SW-780 bladder cancer cells were engineered to express red luciferase, and UM-UC-3 cells were engineered to express red luciferase and IB6 due to a lack of endogenous target expression. Clonal cell lines were selected and characterized for human IB6 surface expression via qFACS and the optical activity of cell lines via bioluminescence imaging on the IVIS SpectrumCT. For these experiments, sixty 12- to 16-week-old female NSG (RRID: IMSR_JAX:005557) or SCID (RRID: IMSR_CRL:236) mice were instilled with 2 million SW780-RL⁺ or UMUC3-IB6⁺RL⁺ cells via a 24-gauge Terumo Surflo IV catheter, following a pretreatment regimen of 0.1% DDM (Sigma, 69227-93-6) and 0.1% poly-L-lysine (Sigma, P8920). After the 45-minute cell inoculation period, the catheters were removed, and the mice were closely monitored until recovery from anesthesia. Two to three weeks later, the mice were injected with 150 mg/kg of Luciferin IP and imaged after 15 minutes to quantify the bioluminescence signal in the bladder region. The mice were randomized into groups of 5 to 7 to achieve a consistent mean bioluminescence signal. The treated mice received 4 weekly doses of 0.2 or 0.6 mg/cm² of PF-08052667 or vehicle in 50 μL for 60 minutes via a soft, beveled catheter to reduce procedural trauma (Instech C20PU-RCA1935) and escape of test agents into circulation. The select groups received a bladder prewash of 100 μL of 0.1% DDM for 10 minutes immediately prior to PF-08052667 or vehicle. Additionally, treatment with PF-08052667 along with prewash for 15- versus 60-minute dwell times was compared in these studies. Tumor growth was monitored 1 to 2 times weekly using bioluminescence imaging, and body weights were collected at each imaging time point. Survival events were recorded in UMUC3-IB6⁺RL⁺ using clinical endpoints of hematuria, palpable bladder, body weight loss, and/or bladder signal greater than 1 × 10¹⁰ photons/second and followed for 71 days after dose initiation. The SW780-RL⁺ study was terminated 55 days after enrollment due to slow progression of disease.

Statistical analysis

A normalized and robust area under the curve (AUC) metric was used for measuring efficacy. The AUC metric captures the average slope of all data points collected regardless of the endpoint and is considered a more efficient and statistically supported endpoint metric for comparing data with log-linear tumor growth kinetics (35). A complete response (CR) is defined when the bioluminescence signal in the bladder region is quantified at or below the baseline signal of 1 × 10⁵ photons/second ±20% (based on historic data generated in non–tumor-bearing mice) for the final three consecutive bioluminescence measurements. The statistical significance comparing the differences in AUC was assessed via one-way ANOVA followed by the Tukey HSD test. Survival events are defined as tumor-bearing mice reaching clinical endpoints indicative of high tumor burden (e.g., severe hematuria, palpable bladder, and/or bladder signal >1 × 10¹⁰ photons/second). The median survival was compared across groups, and the Mantel–Cox log-rank test was used to determine statistical significance.

Expression of IB6 in NMIBC and normal bladder tissue

IB6 expression was evaluated via IHC staining on normal human urinary bladder, cynomolgus monkey urinary bladder, and human bladder cancer tumor samples using an anti-αvβ6 antibody. IHC evaluation revealed strong stain intensity in bladder cancer (NMIBC and MIBC) samples and normal human urinary bladder and moderate stain intensity in normal cynomolgus urinary bladder, with high stain frequency and target accessibility (membranous) in both normal urinary tissues and bladder tumor cells. Across the bladder cancer subtypes tested, there was no difference in the intensity or frequency of IB6 expression. Representative bladder cancer images and a summary of expression in bladder cancer or normal bladder are shown in Fig. 1A–C.

Additionally, RNA and protein expression were quantified in 12 matched frozen NMIBC tumor samples. The RNA transcripts and protein abundance were compared across select genes targeted in bladder cancer therapy (36–38), revealing discordance between RNA expression and protein abundance of IB6. IB6 demonstrated low RNA transcript levels but high protein abundance relative to the comparators (Fig. 1D and E).

In vitro MOA characterization of PF-08052667

The intracellular delivery of cytotoxic payloads by ADCs involves antigen binding on target cells, internalization, and trafficking to lysosomes. Consistent with this MOA, we have previously shown that the parental antibody for PF-08052667, h2A2, binds exclusively to the integrin dimer αvβ6 via ELISA (32). Although h2A2 does not bind to mouse or rat orthologs, the binding affinity of h2A2 to human and cyno IB6, as determined by saturation binding using transfectant 293F cell lines, showed highly concordant EC₅₀ values of 0.3 and 0.2 nmol/L for human and cyno IB6, respectively (32). To ensure that the conjugation of the linker-payload did not inhibit h2A2 binding to its target, the h2A2 antibody and PF-08052667 were evaluated for binding to UM-UC-3 parental cells and UM-UC-3 cells engineered to express human IB6. PF-08052667 and the h2A2 antibody exhibited comparable binding curves in this assay, with an EC₅₀ of 0.08 and 0.10 nmol/L, respectively (Supplementary Fig. S3).

To confirm internalization, PF-08052667 was first detected on the plasma membrane surface of the IB6-positive human bladder cancer cell line SW780 following a brief incubation on ice to allow ADC binding [Fig. 2A (left); Supplementary Fig. S4]. After a 4-hour incubation at 37°C, most of the PF-08052667 signal was shown to internalize and colocalize with LAMP1-stained vesicles [Fig. 2A (right); Supplementary Fig. S4], indicating trafficking to lysosomes. The PF-08052667 signal on the plasma membrane also progressively decreased over the 4 hours, indicating internalization of the ADC [Fig. 2A (right); Supplementary Fig. S4].

The ADC cell surface binding and intracellular trafficking kinetics were assessed by an orthogonal assay that used fluorescence to visualize internalization and drug linker cleavage in IB6-expressing bladder cancer cell lines (Supplementary Table S4). The h2A2 antibody backbone of PF-08052667 was conjugated to a dye and quencher pair using the same glucQF as in PF-08052667 (Supplementary Fig. S1B). The quencher prevents the Cy5 dye from emitting fluorescence until the antibody is internalized and catabolized, leading to the release of the dye from the antibody and quencher. The internalization of antibody–glucQF conjugates h2A2glucQF and nonbinding control h00glucQF was detected by a fluorescent signal due to the unquenching of the fluorophore upon linker cleavage by intracellular glucuronidase. Fluorescence was then quantified by imaging cells every 2 hours and normalizing the fluorescence intensity to cell confluence. The fluorescent signal of h2A2glucQF steadily increased over time, indicative of internalization and linker cleavage (Fig. 2B).

The ability of PF-08052667 to elicit cytotoxicity consistent with the proposed MOA was evaluated in a 96-hour assay with a panel of cancer cell lines endogenously expressing or engineered to express IB6 (Supplementary Table S4). PF-08052667 exhibited cytotoxic activity in all IB6-expressing cell lines (Fig. 2C). In 10 of 11 cell lines tested, the absolute EC₅₀ (X₅₀) values were below 100 ng/mL, ranging from 0.38 to 81.81 ng/mL (Supplementary Table S4). In addition, all cell lines were sensitive to the unconjugated MMAE payload, whereas no activity was observed with the nonbinding ADC control with the same MMAE payload or on IB6-negative cell lines. Overall, these results show that PF-08052667 elicits in vitro cytotoxicity on a range of IB6-expressing cancer cell lines.

Finally, we tested the cytotoxic effect of PF-08052667 on PDOs (Supplementary Table S5). The IB6 expression level for 29 bladder cancer PDOs (5 MIBCs and 24 NMIBCs) was quantified via SABC (Supplementary Fig. S5A) before each PDO was incubated with PF-08052667, SV, or a positive or negative ADC control at nine concentrations for 6 days. A summary of responses, represented by a heatmap of GR₅₀ and GR_AOC metrics determined by an ATP-based viability assay (Fig. 2D and E), demonstrated a wide breadth of activity of PF-08052667 across IB6 expression levels, and responses were comparable across MIBC and NMIBC subtypes. The GR₅₀ values and representative cell kill curves are shown in Supplementary Table S6 and Supplementary Fig. S5, respectively.

PF-08052667 induces hallmarks of ICD

ADCs delivering MMAE with the valine–citrulline linker (vedotin ADCs) have been shown to induce ICD, a distinct type of cell death that can lead to adaptive immune responses and the establishment of immunologic memory (27, 28, 39). The ability of PF-08052667 to induce molecular hallmarks of ICD, such as the exposure of calreticulin on the cell surface and the release of ATP and HMGB1 (39), was compared with that of the unconjugated IB6 mAb as a negative control and to SV and unconjugated MMAE as positive controls in vitro. At 70 hours after PF-08052667 treatment, SW780 cells demonstrated significant cellular release of ATP comparable with SV and MMAE controls (P < 0.05; Fig. 2F), and a similar effect was observed with HMGB1 release 96 hours after PF-08052667 treatment (P < 0.05; Fig. 2G). Translocation of calreticulin was measured in live Annexin V⁺7-AAD^−/lo gated cells 120 hours after PF-08052667 treatment and demonstrated a significant increase in surface calreticulin comparable with that observed with SV and unconjugated MMAE (P < 0.005; Fig. 2H).

Bladder prewash enhances delivery of MMAE to bladder tissue in a rat model

PF-08052667 is designed for intravesical instillation into the bladder for delivery of MMAE to cells expressing IB6. Additionally, a mucosal barrier prewash with 0.1% DDM has been developed as a permeation enhancer to increase local exposure for other modalities, which we hypothesized would also apply to ADCs (40). The effects of the bladder prewash on MMAE concentration and penetration into bladder tissue were assessed with a rat cross-reactive surrogate ADC. Based on IHC, MMAE staining in the urinary bladder in animals with and without prewash was localized to the urothelium and into deeper submucosal and muscular layers (Fig. 3A). In animals treated with bladder prewash, MMAE staining intensity and frequency in urinary bladder tissues were more pronounced, with penetration in submucosal and muscular layers that persisted longer than in animals without prewash [Fig. 3A (top vs. bottom)]. In agreement with IHC findings, bladder prewash increased tissue concentrations of the total unconjugated MMAE up to 1 week after the dose (Fig. 3B). In addition, the effect of ADC bladder instillation dwell time on MMAE exposure was investigated in this model. Typical dwell times in patients vary from 1 to 2 hours, and dwell times ranging from 15 to 90 minutes in the rat studies resulted in no meaningful differences in the total unconjugated MMAE, but with consistently higher concentrations observed with bladder prewash compared with ADC alone, further confirming that 0.1% DDM prewash enhances the delivery of MMAE to bladder tissue (Fig. 3C).

PF-08052667, with and without prewash, is well-tolerated with minimal systemic exposure and no systemic toxicity

The antibody component of PF-08052667 has comparable binding to cynomolgus and human IB6 orthologs (32), and thus GLP toxicity studies of the ADC, with and without prewash, were performed in this pharmacologically relevant intravesical model. Similar to the results in rats, the bladder prewash enhanced MMAE delivery in nonhuman primate bladder tissue. In the GLP study, the total unconjugated MMAE in the treated bladder tissue 1 week after the last dose was 42-fold higher with prewash than in animals treated with the same dose of PF-08052667 alone despite not affecting systemic exposure and being assessed a full week after the last dose (Fig. 4A). Importantly, plasma concentrations of the total antibody and total unconjugated drug were generally undetectable regardless of bladder prewash, with the estimated systemic absorption of <0.1% of the administered dose based on total antibody measurements (Fig. 4B).

In concordance with the general lack of systemic exposure, there were no systemic PF-08052667-related toxicities observed. PF-08052667-related findings were limited to local effects on the lower urinary tract and were consistent with the local administration of MMAE and the presence of typical procedural trauma related to the route. Local findings were nonadverse but included minimal to mild vacuolation, inflammation, and edema of the urothelium, lamina propria, and/or bladder muscularis in animals treated with PF-08052667, with or without prewash. These findings were considered PF-08052667-related as they tended to increase in incidence/severity with increasing dose although the effects overlapped with those that occur secondarily from procedural trauma. Similarly, administration of bladder prewash demonstrated no effects distinguishable from procedural and/or background findings in local tissues or systemically when administered without PF-08052667. This result is consistent with the well-characterized safety profile of alkylsaccharide excipients, including n-dodecyl-β-D-maltoside (DDM), which have been incorporated into multiple approved intranasal drug formulations as an inactive ingredient, with an established safety profile generally recognized as safe (41–43). The well-tolerated nature of PF-08052667, specifically designed for intravesical administration, is consistent with the route and low turnover rate of the urothelium, providing resistance to the effects of MMAE. This is in contrast to the systemic toxicities observed with intravenous administration, in which classical MMAE-related bone marrow toxicity and target-mediated findings occurred in rapidly dividing epithelial tissues.

PF-08052667 demonstrates robust antitumor activity in orthotopic mouse xenograft models after a bladder prewash

The impact of the mucosal barrier prewash, ADC dose, and instillation dwell time of PF-08052667 was further evaluated in CDX orthotopic bladder cancer models endogenously expressing or engineered to express human IB6 (SW780-RL⁺ and UMUC3-IB6⁺RL⁺, respectively). These tumor cells engraft and proliferate in the bladder mucosa, mimicking clinical NMIBC, and show differing IB6 expression levels and tumor growth kinetics (Fig. 5A). In the slow-growing SW780-RL⁺ model, vehicle control or single-agent PF-08052667 did not affect tumor growth compared with untreated tumors; however, 0.2 and 0.6 mg/cm² of PF-08052667 with DDM prewash significantly reduced tumor burden (Fig. 5B; Supplementary Table S7; Supplementary Fig. S6). In this model, the 15-minute dwell time produced a similar antitumor response compared with the standard 60-minute incubation, indicating that shorter dosing periods may drive similar efficacy (Fig. 5C). These results were confirmed in a second, more aggressive model, UMUC3-IB6⁺RL⁺, in which both doses of PF-08052667 with DDM prewash significantly reduced tumor burden, and shorter dwell time did not affect activity (Fig. 5C; Supplementary Table S8). Additionally, single-agent PF-08052667 drove a significant survival benefit (22 vs. 48-day median) compared with vehicle control, which was further enhanced by the prewash (endpoint not reached). The improved durability of response to PF-08052667 with prewash resulted in 4/5 CRs compared with 0/5 CRs in the dose-matched single-agent group (Supplementary Fig. S7). Due to the technically challenging nature of intravesical dosing in mice, procedural complications caused a small number of mice to be excluded from the study (noted in Supplementary Table S8). PF-08052667 and DDM prewash were well tolerated, and body weight was not significantly affected while on treatment. These data demonstrate that intravesical administration of PF-08052667 with DDM prewash is effective in delivering MMAE to cancer cells engrafted in the urothelium, and this effect is maintained at shorter dwell times.