A Novel B7-H4xCD3 Bispecific T-cell Engager (PF-07260437) Synergizes with Breast Cancer Standard of Care and Immune Checkpoint Therapies Open Access

Breast cancer is the most common cancer among women, affecting 2.1 million women each year and causing the highest number of cancer-related deaths. In 2020, an estimated 685,000 women died from breast cancer worldwide, accounting for approximately 15% of all cancer deaths among women (1). Early detection and clinical intervention have significantly benefited patients (2); however, there is a critical need for novel therapies for those who relapse on current standard-of-care treatments (3).

B7-H4, encoded by the VTCN1 gene, belongs to the B7 family of immune regulatory ligands. Structurally, B7-H4 is a type I transmembrane protein with an extracellular domain that inhibits T-cell proliferation and cytokine production. Functionally, B7-H4 plays a role in immune evasion by tumors, making it a potential target for cancer immunotherapy (4, 5). B7-H4 expression is generally low in normal tissues but is upregulated in various tumors, including breast, ovarian, and pancreatic cancers (4, 6). B7-H4 is found in a significant subset of patients with breast cancer, particularly those with triple-negative breast cancer (TNBC) and HER2-positive breast cancer (6), as well as in gynecologic malignancies (7, 8). Studies show that B7-H4 expression is associated with poor prognosis and contributes to tumor progression by suppressing T cell–mediated immune responses (4, 5).

CD3 bispecific antibodies are a class of cancer immunotherapy designed to redirect T-cell killing to tumor cells expressing specific cell surface antigens (9). This class of molecules usually consists of two arms: one recognizing a tumor antigen and the other a subunit of CD3 in the T-cell receptor complex. The simultaneous binding of the bispecific to the tumor antigen and CD3 forms a tumor cell–bispecific T-cell trimer, creating an immunological synapse that initiates T-cell activation and tumor cell killing (10). This mechanism bypasses the need for cognate MHC–T-cell receptor interactions, making it useful for tumor indications with low tumor neoantigens and cognate T-cell clones, such as most breast cancers, especially HR⁺HER2⁻ and HER2⁺ subtypes. CD3 bispecific molecules have been clinically validated with approvals in hematologic malignancies (11, 12) and are being assessed in solid tumor indications (9, 13).

Clinical studies with CD3 bispecifics (9, 11–13) have highlighted the importance of targeting tumor antigens with low normal tissue expression to achieve efficacy without significant on-target normal tissue toxicity. The high expression of B7-H4 in breast cancer, combined with limited expression in normal tissues, makes it an attractive target for CD3 bispecific molecules in breast cancer treatment.

In this study, we characterized B7-H4 expression as a pan–molecular subtype tumor antigen for breast cancers with limited normal tissue expression. We demonstrated potent in vitro and in vivo pharmacologic activity of a novel anti–B7-H4/anti-CD3 bispecific molecule, PF-07260437, and the immunological processes it induces to initiate a T cell–mediated immune response in tumor models. Additionally, PF-07260437 synergized in combination with standard-of-care therapy (fulvestrant and palbociclib) and anti–PD-1 antibody treatment in vivo. To aid clinical evaluation, we used a CD8-PET biomarker (⁸⁹Zr-Df-IAB22M2C from ImaginAb Inc.) to show increased intratumoral CD8 T cells after treatment. Finally, PF-07260437 demonstrated a tolerated safety profile in cynomolgus monkeys, supporting further clinical investigation.

The human cell lines HEK-293 (Cat. # CRL-1573, RRID:CVCL_0045), MDA-MB-468 (Cat. # HTB-132, RRID:CVCL_0419), MDA-MB-453 (Cat. # HTB-131, RRID:CVCL_0418), HCC-1954 (Cat. # CRL-2339, RRID:CVCL_1260), HCC-1806 (Cat. # CRL-2335, RRID:CVCL_1258), HCC-1569 (Cat. # CRL-2330, RRID:CVCL_1259), HCC-1599 (Cat. # CRL-2331), ZR-75-1 (Cat. # CRL-1500, RRID:CVCL_0588), and T-47D (Cat. # HTB-133) were obtained from the ATCC. MX-1 cells were obtained from the NCI. SUM149PT cells were obtained from BioIVT Inc. (Cat. # HUMANSUM-0003004). MDA-MB-453 and T-47D were cultured in DMEM using GlutaMAX (Thermo Fisher Scientific) and 10% FBS (Gibco). HEK-293 and MDA-MB-468 were cultured in MEM using GlutaMAX and 10% FBS. HCC-1569, HCC-1599, HCC-1806, HCC-1954, ZR-75-1, and SUM149PT were cultured in RPMI 1640 with GlutaMAX and 10% FBS. MX-1 was cultured in a 1:1 mixture of Ham’s F-12 medium and DMEM with GlutaMAX and 10% FBS. Murine E0771 cells (CH3 BioSystems, Cat. # 94A001) were cultured in RPMI GlutaMAX medium with 10% heat-inactivated FBS and 1% penicillin/streptomycin (Gibco).

HEK-293-hB7-H4 cells and E0771-hB7-H4 were generated by transducing human HEK-293 and murine E0771 cells, respectively, with human VTCN1 lentiviral particles and selecting clones with stable expression. HCC-1954, HCC-1806, and MX-1 tumor cells were transduced with pantropic retrovirus produced from pMSCVpuro_LucSh or pMSCVneo_LucSh retroviral transfer vectors to introduce firefly luciferase, followed by the selection of drug-resistant pools with puromycin or geneticin. All cells were grown in a humidified chamber at 37°C under a 5% CO₂ atmosphere. All cells were authenticated using the short tandem repeat method at IDEXX BioAnalytics and tested for Mycoplasma using a MycoAlert kit (Lonza Bioscience) following the manufacturer’s protocol.

PF-07260437 is based on the IgG2 scaffold with an arm that binds human B7-H4 and an arm that binds the epsilon subunit of the human CD3 molecule. Mutations were introduced in the Fc to reduce binding of Fcγ receptors (EU numbering: D265A, A330S, and P331S; ref. 14). The correct pairing of the heavy and light chains was achieved by separate expression of the two arms and subsequent heterodimerization by an in vitro redox reaction facilitated by the following mutations (in EU numbering): C223E, P228E, and L368E mutations in the anti–B7-H4 arm and C223R, E225R, P228R, and K409R mutations in the anti-CD3 arm (15, 16). The sequence of PF-07260437 and the full experimental details for the preparation and characterization of the agent are described in patent US 2022/0017622 A1 (17). A control non-B7-H4–targeting CD3 bispecific antibody was generated by the same process.

Tumor samples were fixed in 10% neutral-buffered formalin and dehydrated overnight through a series of xylene, graded alcohols, and paraffin. Samples were embedded in paraffin, cooled, and sectioned to a five-micron thickness on a microtome. Sections were deparaffinized in xylene and rehydrated through a graded series of alcohols to deionized water.

Hematoxylin and eosin staining was performed with chloramphenicol acetyltransferase hematoxylin followed by eosin. For single IHC stains, sections underwent heat-induced epitope retrieval in either Citra Plus AR (BioGenex) or Borg Decloaker. Endogenous peroxidase activity was quenched with peroxide. Nonspecific protein interactions were blocked with Background Punisher (Biocare Medical). Sections were incubated with the following primary antibodies: anti–B7-H4 (A57.1 Mouse mAb or clone D1M8I, Cell Signaling Technology, Cat. # 14572, RRID:AB_2750878), anti-CD3 (Abcam, Cat. # Ab135381), or anti–PD-L1 (Spring Bioscience, Cat. # M4420, RRID:AB_2861330), followed by labeled polymer MACH2 Mouse or Rabbit HRP. Color for all stains was developed with the chromogen Betazoid 3, 3'-diaminobenzidine (DAB). After a rinse in distilled water, the sections were counterstained for 10 seconds in Tacha’s hematoxylin. For sequential double immunolabeling, sections were incubated with primary antibodies, such as B7-H4, granzyme B (GzymB, Cell Signaling Technology, Cat. # 46890, RRID:AB_2799313), or anti-human IgG (Abcam, Cat. # ab109489, RRID:AB_10863040); secondary antibodies, such as anti-CD68 (Cell Signaling Technology, Cat. # 76437, RRID:AB_2799882), anti-GzymB, or anti-CD8; and for triple staining with anti-CD3. Reagents were sourced from Biocare Medical unless otherwise stated.

Slides were scanned using a Leica Aperio AT2 whole-slide digital scanner (Leica Biosystems) at a 20× magnification setting. Whole-slide image analysis was performed using Visiopharm software (Visiopharm). The tumor regions of interest were annotated by a board-certified pathologist, and regions containing necrotic tissue, normal tissue, and tissue folds were excluded. The intensity levels of the stain range from 0 to 3+: 0, negative; 1+, low expression (light brown); 2+, moderate expression (brown); and 3+, high expression (dark brown to black).

All image analysis was carried out using a custom application protocol packages (App) in Visiopharm software version 2022.02. An App was written to detect the viable tumor regions, which were manually verified and corrected wherever necessary. Then, a second App was executed to detect and count the number of CD8a-positive cells in the viable tumor region. The cell density was calculated as the ratio of CD8a-positive cells to the viable tumor area (in mm²).

Cells were cultured to 90% confluency and collected in 1× Cell Lysis Buffer (Cell Signaling Technology) with Halt Protease and Phosphatase Inhibitor Cocktails (Thermo Fisher Scientific). Protein concentration was determined using a Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Ten micrograms of protein was denatured with heat and reduced using 2-mercaptoethanol and run on a 4% to 12% Bis/Tris NuPAGE gel (Thermo Fisher Scientific) in 3-(N-Morpholino)propanesulfonic acid running buffer. After transfer to a nitrocellulose membrane with iBlot (Thermo Fisher Scientific), primary anti–B7-H4 antibody clone D1M8I (Cell Signaling Technology, Cat. # 14572, RRID:AB_2750878) was used at 1:1,000 in Tris-buffered saline containing 0.1% Tween 20 with 5% nonfat dry milk (Bio-Rad) overnight. Anti-GAPDH antibody (Cell Signaling Technology) was used at 1:5,000 as a control. Horseradish peroxidase–conjugated anti-rabbit secondary antibody was used for detection at 1:3,000 in Tris-buffered saline containing 0.1% Tween 20. Thermo West Pico substrate (Thermo Fisher Scientific) was applied, and the image was developed on Hyperfilm ECL film (GE Healthcare Life Sciences).

B7-H4–expressing human tumor cells were harvested and stained with serial dilutions of PE 1:1 conjugated anti–B7-H4 antibody (Pfizer-generated B7-H4 antibody clone 1D11, conjugated by eBioscience) and an isotype control antibody. Stained cells along with a Quantibrite PE Quantitation Kit (BD Biosciences) were acquired using a BD Biosciences LSRFortessa X-20. The averaged receptor density of B7-H4 per cell at the saturating concentration of antibodies was measured as antibody bound per cell, which was calculated using the standard curve from the Quantibrite PE kit.

Serial dilutions of PF-07260437 were incubated with multiple cell lines and primary human T cells isolated from healthy donor peripheral blood mononuclear cells (PBMC). Human pan T cells were isolated from PBMCs by negative selection using a T-cell enrichment kit (STEMCELL Technologies). The incubation was performed at 37°C for 2 hours with sodium azide functioning to inhibit antibody internalization. All cells were washed twice with Dulbecco's Phosphate Buffered Saline (DPBS; GE Healthcare Life Sciences) containing 1% BSA (Thermo Fisher Scientific) and 0.01% sodium azide (RICCA Chemical Company). The unbound secondary antibody was washed off, and bound PF-07260437 molecules were detected with a PE-labeled goat anti-human Fab secondary antibody (Jackson ImmunoResearch). Cells were resuspended in PBS containing 7-AAD Viability Staining Solution (BioLegend), and data were acquired on a BD LSRFortessa X-20. Data were analyzed using FlowJo (RRID:SCR_008520) and GraphPad Prism (RRID:SCR_002798).

Human PBMCs were isolated from healthy donor blood using a Histopaque-1077 gradient in ACCUSPIN tubes (Sigma-Aldrich). Pan T cells were isolated from PBMCs by negative selection using an EasySep Human T Cell Enrichment kit (STEMCELL Technologies). For the lactate dehydrogenase release assay, B7-H4–expressing human tumor cells were resuspended in RPMI medium with 10% FBS and 1% penicillin/streptomycin. T cells were added to tumor cells at an effector-to-target ratio of 2.5:1. The cells were treated with serial dilutions of PF-07260437 or a negative control bispecific. Lactate dehydrogenase released from compromised target cells was measured using a CytoTox 96 Non-Radioactive Cytotoxicity Assay kit (Promega) and a VICTOR microplate reader (PerkinElmer).

For luciferase-based viability assays, B7-H4–expressing human tumor cells transfected with a luciferase expression construct were resuspended in culture medium. T cells were added to tumor cells at an effector-to-target ratio of 2.5:1. The cells were treated with serial dilutions of PF-07260437 or a negative control bispecific. Luciferase signals were measured using a Neolite Reporter Gene Assay system (PerkinElmer) and a VICTOR microplate reader (PerkinElmer).

To evaluate cell line proliferation in vitro, cells were treated with serial dilutions of PF-07260437. Cell number was determined using the ATP-dependent CellTiter-Glo Luminescent Cell Viability Assay (Promega) and a VICTOR microplate reader.

To test PF-07260437–induced T-cell cytokine release, CTL assays were set up as described above. At the end of the CTL assay, supernatants were collected and analyzed using a MILLIPLEX MAP Human Cytokine/Chemokine Magnetic Bead Panel according to the manufacturer’s guidelines (MilliporeSigma), read on a Luminex 200 (Luminex xMAP Technology) with Luminex xPONENT software 3.1, and analyzed using MILLIPLEX Analyst version 5.1.0.0. The cytokines measured were human IFN-γ, IL-10, IL-2, and TNF-α.

All procedures performed on animals were in accordance with regulations and established guidelines and were reviewed and approved by an Institutional Animal Care and Use Committee (Pfizer La Jolla/Pearl River sites) or through an ethical review process. Female NOD/SCID gamma (NSG) mice (The Jackson Laboratory, RRID:IMSR_JAX:005557) and female humanized CD3e transgenic mice (genOway) were housed in the Pfizer animal facility at ∼72°F with ∼50% humidity. Irradiated water and a standard diet of pelleted food (Purina irradiated 5V02) were provided ad libitum. For hormone-driven tumors, estradiol supplementation was provided, and mice were implanted with a 90-day release estradiol pellet for T-47D or were supplemented with estradiol (Sigma-Aldrich) at 8.5 µg/mL in drinking water for the ZR-75-1 model. Tumor fragments (4 × 4 mm) for patient-derived xenograft (PDX) models PDX-BRX-11380 and PDX-CRX-11201 (Asterand Bioscience), PDX-BRX-24301 (Champions Oncology), and PDX-BRX-26305 (MRL) were generated from 500 to 800 mm³ tumors, expanded, and then implanted subcutaneously in the flank of NSG mice. Cell line–derived xenograft (CDX) models were inoculated subcutaneously in the flank of NSG mice, whereas HCC-1954 and ZR-75-1 were orthotopically inoculated in the intramammary fat pad. Animals were euthanized if they crossed the Institutional Animal Care and Use Committee guidelines with respect to tumor size, body weight loss, or animal health. Mice were dosed with PF-07260437 or negative control bispecific via the subcutaneous or intravenous route in a 0.2 mL bolus injection weekly for up to 3 times. Palbociclib (Pfizer) was administered orally twice at 10 mg/kg, and fulvestrant (Tecoland, Cat. # 129453-61-8) was administered subcutaneously at 10 mg/kg per week after an initial 3× per week induction schedule.

For adoptive T-cell transfer in vivo studies, T cells were isolated from human PBMCs using the protocol described previously. On day 2, T cells were transferred to a G-Rex cell culture device (Wilson Wolf Manufacturing) for expansion, and IL-2 (Shenandoah Biotechnology Inc.) was added to the media and replenished after 3 days. T cells were harvested 1 week after activation/expansion. At the time of harvest, beads were removed with a magnet, and cells were resuspended in DPBS for in vivo inoculation. Cultured human T cells (2.5 × 10⁶) were inoculated intravenously 1 day after the first dose of PF-07260437 or negative control bispecific. For human PBMC transfer in vivo studies, 5 × 10⁶ untreated human PBMCs were injected intravenously 6 days prior to the first dose of PF-07260437 or negative control bispecific. For immunocompetent E0771-hB7-H4 tumor models, humanized CD3e transgenic mice were implanted orthotopically at a cell density of 5 × 10⁵ into the mammary fat pad. Mice were dosed subcutaneously with a single dose of 0.5 mg/kg of PF-07260437 and/or dosed intraperitoneally every three days for six times with 10 mg/kg of a murine anti–PD-1 surrogate. Tumor volumes were measured twice a week using calipers, and animals were monitored for changes in body weight, behavior, signs of toxicity, or a graft-versus-host response. Tumor growth inhibition (TGI) was calculated on the day all animals were in the study before they were removed because of excessive tumor volume using the following formula: 1 − (T_end − T_begin/C_end − C_begin), where T_end is the final average tumor volume of the treated group, T_begin is the average starting volume of the treated group, C_end is the final average volume of the vehicle group, and C_begin is the average starting volume of the vehicle group.

Male and female cynomolgus monkeys of Mauritian origin (>2.5 years of age) were obtained from Charles River Laboratories. Animals were administered 0 (1 animal/sex), 30 µg/kg/dose (1 male), 100 µg/kg/dose (1 animal/sex), or 300 µg/kg/dose (1 female) PF-07260437 by intravenous bolus once on days 1 and 8. On day 10 (2 days after the second dose), animals were humanely euthanized using an intravenous administration of a barbiturate followed by exsanguination. Assessments included clinical observations, body weights, body temperature, qualitative food consumption, clinical pathology parameters, pharmacodynamic (PD) biomarkers (cytokines and markers of T-cell activation), toxicokinetics (TK), and macroscopic and microscopic tissue pathology evaluations. In alignment with the three Rs (replace, reduce, and refine), the control animals were not necropsied, and microscopic comparisons were made to historic controls. Blood was collected for cytokine evaluation before the dose and at 3, 7, and 24 hours after the dose and on days 3 and 10 (IL-2, IFN-γ, IL-6, soluble IL-2-RA, IL-10, high-sensitivity C-reactive protein). Hematology (Siemens ADVIA 2120) and clinical chemistry (Siemens ADVIA 1800) were evaluated prior to the study start and on days 3 and 10. Coagulation parameters (Diagnostica STAGO Evolution) were evaluated prior to the study start and on day 10. Blood immunophenotyping samples were collected on days 3, 8, and 10 (CD4, CD8 T-cell subsets, B cells, and NK cells). TK collection on day 1 was before the dose and at 0.083, 2, 4, 7, 24, 48, 72, and 96 hours after dose administration, and on day 8, it was before the dose and at 0.083, 2, 4, 7, 24, and 48 hours after dose administration. Tissues were fixed in 10% neutral buffered formalin, except for the eyes, which were fixed in Davidson’s fixative. Microscopic evaluation of the tissues was performed by a board-certified veterinary pathologist.

Snap-frozen tumors were processed in a Precellys Lysing Kit tube and homogenized in 0.7 mL QIAzol (Qiagen) using the Precellys 24 Homogenizer for 30 seconds at 5,000 RPM. RNA was isolated using the Qiagen miRNeasy Mini Kit following the manufacturer’s protocol. cDNA was synthesized using the QuantiTect Reverse Transcription Kit (Qiagen) following the manufacturer’s protocol. TaqMan Fast Advanced Master Mix (Applied Biosystems, Cat. # 4444557) and probes (Thermo Fisher Scientific, VTCN1 Hs01552471_g1, IL2 Hs00174114_m1, IFNG Hs00989291_m1, CD3E Hs01062241_m1, CXCL9 Hs00171065_m1, CXCL10 Hs00171042_m1, and GAPDH Hs02786624_g1) were used for qRT-PCR amplification in the MyiQ2 thermocycler (Bio-Rad Laboratories) running iQ5 software (Bio-Rad Laboratories). Data were analyzed using the 2-ΔCt method (18), comparing threshold cycles to those of GAPDH expression.

Single-cell suspensions were made from minced tumors with an enzymatic mixture of collagenase IV (Worthington Biochemical) and 1,000 U/mL DNase I D (Worthington Biochemical) in Hanks' Balanced Salt Solution (Thermo Fisher Scientific) or by using the Miltenyi mouse tumor digestion kit (Miltenyi Biotec) according to the manufacturer’s protocol. Viability was determined, and cells were blocked with either 5% normal mouse serum (Jackson ImmunoResearch) or an unlabeled rat anti-mouse CD16/32 (Bio X Cell). Surface markers were stained with different panels of fluorochrome-conjugated mAbs. The panel for MDA-MB-468 tumor studies consisted of Brilliant Violet 570 anti-human CD4 antibody (BioLegend, Cat. # 300533, RRID:AB_10896788), Brilliant Violet 605 anti-human CD3 antibody (BioLegend, Cat. # 317321, RRID:AB_11126166), Brilliant Violet 785 anti-human CD8 antibody (BioLegend, Cat. # 344739, RRID:AB_2566201), PE/Cy5 anti-mouse CD45 antibody (BioLegend, Cat. # 103109, RRID:AB_312974), Alexa Fluor 532 anti-human CD45 monoclonal antibody (Thermo Fisher Scientific, Cat. # 58-0459-42, RRID:AB_11218673), and APC-R700 anti-human CD25 (BD Biosciences, Cat. # 565106, RRID:AB_2744339).

The panel for E0771-hB7-H4 tumor studies consisted of anti-mouse CD45 (BioLegend, Cat. # 103128, RRID:AB_493715), anti-mouse Thy 1.2 (BioLegend, Cat. # 140304, RRID:AB_10642812), anti-mouse CD3 (BioLegend, Cat. # 100214, RRID:AB_493645), anti-mouse CD8 (BioLegend, Cat. # 100734, RRID:AB_2075238), and LIVE/DEAD Aqua (Invitrogen, Cat. # L34957). After fixation and permeabilization with buffer (Thermo Fisher Scientific), intracellular and nuclear markers for MDA-MB-468 studies were stained with fluorochrome-conjugated mAbs (FITC anti-human/anti-mouse GzymB antibody; BioLegend). CountBright Absolute Counting Beads (Thermo Fisher Scientific, Cat. # C36950) were added to samples before acquisition on the Cytek Aurora spectral analyzer running SpectroFlo (Cytek Biosciences) for MDA-MB-468 studies. E0771-hB7-H4 samples were acquired on a BD LSRFortessa cytometer. Data were analyzed using FlowJo software.

RNA was extracted from formalin-fixed, paraffin-embedded tissue using the FormaPure XL Total kit (Beckman Coulter) according to the manufacturer’s protocol. Samples were prepared and assessed through the nCounter PanCancer IO 360 gene expression panel (NanoString) according to the manufacturer’s protocol.

NSG mice were inoculated with 5 × 10⁶ MDA-MB-468 tumor cells subcutaneously near the right scapula and engrafted with 5 × 10⁶ human PBMCs intraperitoneally 6 days prior to enrollment. Animals with average tumor sizes of approximately 200 mm³ were selected to be enrolled in the study (day 0). PF-07260437 or the negative control isotype-matched bispecific antibody was administered subcutaneously on days 0 and 7. Pharmacokinetics (PK) was evaluated on days 1 and 8 on a subset of animals to verify drug exposure. The study was divided into two cohorts (cohorts A and B) to assess CD8 T-cell infiltration at early and late time points, respectively. Animals in cohort A only received one dose of antibody treatment. On day 4, cohort A animals received approximately 75 µCi of zirconium-89 (⁸⁹Zr)–labeled anti-CD8 minibody (ImaginAb) at a 25 µg protein dose (i.e., 3 mCi/mg specific activity) intravenously. Twenty-two hours later, they were anesthetized by inhalation of a 2% to 3% isoflurane solution and imaged in the G8 PET/CT scanner (SOFIE Biosciences) using the following parameters: a ⁸⁹Zr peak-calibrated 7-minute static PET scan with a vendor provided 3D maximum likelihood estimation method algorithm reconstruction at 1.4 mm resolution. A two-minute co-registered CT scan followed each PET scan, with 50 kVp, 200 µA, and 75 microns on step-and-shoot (5°/step). Animals were kept at 37°C during the entire PET imaging procedure. Animals were euthanized immediately after imaging, and blood, tumor, spleen, liver, quadricep muscle, and kidneys were collected, weighed, and subjected to ex vivo gamma counting using the Wizard 2480 Gamma Counter (PerkinElmer). Tails were also collected and gamma counted to correct for the injected dose (ID). Cohort B animals received two doses of bispecific antibody treatment. They were injected with the imaging tracer on day 8 and imaged on day 9 following the same procedure. Cohort B tumors were also fixed in 10% neutral buffered formalin to enable subsequent histology and IHC analyses.

GraphPad Prism was used to calculate EC₅₀ values for in vitro cytotoxicity assays using four-parameter nonlinear regression analysis and EC₅₀ values for cytokine release assays by nonlinear regression plotting of cytokine secretion versus concentration of PF-07260437.

Tumor study mean ± SEM tumor volumes were calculated using GraphPad Prism. Proprietary Pfizer software (TGI analyzer) was used for in vivo analyses to calculate Bliss combination plots (19), one-tailed ANOVA for comparing treatments to the vehicle group, and two-tailed ANOVA for treatment combination comparisons.

All data are available from the corresponding author upon reasonable request.

B7-H4 has been described as an antigen highly expressed in breast cancer and gynecologic malignancies (6, 8). However, detailed data regarding expression in different molecular subtypes of breast cancer are still lacking. To assess the potential of B7-H4 as a pan-molecular subtype tumor antigen for the fragmented breast cancer indications, we characterized expression across human breast cancers. Sixty-one primary human breast cancer specimens were stained by IHC with an anti-human B7-H4 antibody and evaluated by a pathologist. B7-H4 was expressed in the majority of specimens (97% in HR⁺HER2⁻, 100% in HER2⁺, 90% in TNBC, and 95% overall) and at high-to-moderate levels in the majority of samples (H-score ≥100; 70% in HR⁺HER2⁻, 50% in HER2⁺, 71% in TNBC, and 67% overall), regardless of subtype (Fig. 1A–C; Supplementary Table S1). Expression heterogeneity within individual tumors varied across subtypes (Supplementary Fig. S1). Moreover, where both cancerous tissues and normal mammary gland tissues were present, B7-H4 was expressed at higher levels in tumor than in normal adjacent breast tissue (Fig. 1D), suggesting a significant tumor/normal tissue differential.

B7-H4 is not expressed on tumor-associated macrophages as previously claimed (20), and duplex staining of B7-H4 and the macrophage marker CD68 showed no overlap (Fig. 1E). Additionally, B7-H4 was absent in human PBMCs by Western blot (Fig. 1F), indicating no expression in monocytes, dendritic cells, or lymphocytes.

B7-H4 was found to be expressed in several human breast cancer models, including cell lines (Fig. 1F) and PDXs (Fig. 1G–I). The density of B7-H4 on cell lines used in xenograft models was measured by flow cytometry and presented as antibody bound per cell, demonstrating a range of B7-H4 expression levels across breast cancer subtypes (Supplementary Table S2).

To target B7-H4–expressing tumors with redirected T-cell killing, we designed a heterodimeric bispecific IgG molecule (PF-07260437, Fig. 2A, further described in “Materials and Methods”). PF-07260437 showed dose-dependent binding of B7-H4 on MX-1 tumor cells and transduced HEK-293 cells but not to B7-H4–negative HCC-1806 and parental HEK-293 cells (Fig. 2B), demonstrating specificity. Additionally, PF-07260437 showed dose-dependent CD3 binding to both human and cynomolgus T cells (Fig. 2B), indicating species cross-reactivity. The affinity for human CD3 was assessed at 63.5 nmol/L, and the affinity for cynomolgus CD3 was assessed at 66.75 nmol/L via surface plasmon resonance binding.

We next evaluated the capability of PF-07260437 to induce redirected T-cell killing via in vitro cytotoxicity assays. In co-culture with purified human pan T cells, PF-07260437 induced dose-dependent cell killing of tumor cell lines with varying levels of B7-H4 expression (Fig. 2C). Additionally, PF-07260437 induced T cell–mediated killing of breast cancer cell lines of all three major molecular subtypes, including HCC-1954 for HER2⁺ (Fig. 2D), T-47D for HR⁺HER2⁻, and MX-1, MDA-MB-468, and MDA-MB-453 for TNBC (Fig. 2C). The EC₅₀ values of PF-07260437 in each cell line were calculated using four-variable curve fitting (Supplementary Table S3). A strong negative correlation between B7-H4 receptor density and EC₅₀ was observed with a goodness-of-fit R² of 0.8045 (Fig. 2E), demonstrating the B7-H4–dependent pharmacology of PF-07260437.

Cytokine release was assessed as a surrogate measure of PF-07260437–induced T-cell activation. Supernatant from the cytotoxicity assay (Fig. 2C) was taken and assessed for IFN-γ, IL-10, IL-2, and TNF-α. PF-07260437 induced cytokine release with B7-H4–expressing cancer cell lines (Supplementary Table S4) but not in B7-H4–negative HCC-1806 cells, demonstrating target specificity (Fig. 2D). PF-07260437 in the absence of T cells did not have any significant effect on breast cancer cell line growth (Fig. 2F), indicating that the cytotoxicity observed depended on T-cell activation, not tumor-intrinsic effects. Combined, these data demonstrated that the cell-killing capability of PF-07260437 is dependent on both B7-H4 expression and the presence of T cells.

To correlate efficacy with B7-H4 expression, in vivo studies with PF-07260437 or an isotype-matched negative control bispecific were performed (Fig. 3A–G). In CDX tumor models, PF-07260437 demonstrated dose-dependent in vivo efficacy in three breast cancer subtypes (TNBC, HR⁺HER2⁻, and HER2⁺). In TNBC xenografts (MDA-MB-468) with human PBMCs, three weekly doses of PF-07260437 resulted in 133% TGI when dosed as low as 0.04 mg/kg (Fig. 3A; Supplementary Table S5). Significant TGI was also seen in HER2⁺ xenografts (HCC-1954; Fig. 3B; Supplementary Table S6). A limitation of NSG-human PBMC models is that human T cells engraft systemically and eventually mount a xenographic graft-versus-host response against mouse tissue, making it difficult to evaluate the durability of responses. Therefore, we used the adoptive transfer of in vitro expanded human T cells to test the in vivo efficacy of PF-07260437, which allowed for efficacy evaluation over a comparatively longer period. In HR⁺HER2⁻ xenografts (T-47D), PF-07260437 induced complete tumor regression in 10 of 10 mice at 0.5 mg/kg, with no tumor regrowth at 35 days (Fig. 3C).

PF-07260437 resulted in complete tumor regressions in the high B7-H4–expressing MX-1-Luc model of human TNBC at a dose as low as 0.05 mg/kg (Fig. 3D; Supplementary Table S7). The antitumor activity was dose-dependent as lower efficacy was observed at 0.01 and 0.0015 mg/kg of PF-07260437. Moreover, PF-07260437 treatment was efficacious following both intravenous and subcutaneous administration. The antitumor activity was sustained at the 0.05 and 0.5 mg/kg doses, with no tumors growing back by 35 days.

The in vivo efficacy of PF-07260437 was also tested in human PDX breast cancer models. PF-07260437 induced dose-dependent TGI in three different breast cancer PDXs: PDX-BRX-11380, PDX-BRX-24301, and PDX-BRX-26305 (Fig. 3E–G; Supplementary Tables S8–10). Similar to CDX models, PDX models showed >100% TGI with a 0.5 mg/kg dose of PF-07260437 with comparable efficacy following both intravenous and subcutaneous administration. B7-H4–negative PDX-CRX-11201 colorectal tumors did not respond (Fig. 3H; Supplementary Table S11).

To investigate the mechanism of action of PF-07260437 in vivo, a PD biomarker study was conducted in the MDA-MB-468 CDX model with human PBMC engraftment. Tumor-bearing mice received up to two doses of PF-07260437 or vehicle weekly at 0.5 or 0.05 µg/kg. Tumors were harvested on days 3, 5, 7 (before the second dose), and 14 and divided for analysis via orthogonal assays using IHC, qRT-PCR, and flow cytometry.

IHC analysis showed that PF-07260437 induced intratumoral CD3⁺ T-cell accumulation, starting at day 5 and accelerating by day 7 (Supplementary Fig. S2A). The production of the cytotoxic molecule GzymB suggests T-cell activation (Supplementary Fig. S2B). GzymB production was not restricted to CD8⁺ T cells, suggesting that PF-07260437 induced GzymB production in non-CD8 T cells, possibly CD4⁺ T cells. As T cells accumulated and were activated, B7-H4–expressing tumor cells were killed (Supplementary Fig. S2C), and by day 14, hematoxylin and eosin staining showed a large loss of cellularity (Supplementary Fig. S2D). The cytokine IFN-γ has been shown to be critical in inducing an antitumor T-cell response (21); its involvement is suggested by tumor cells upregulating the IFN-γ–inducible gene PD-L1 in response to T-cell activity (Supplementary Fig. S2E). This observation indicated that immune checkpoint pathways could be upregulated by tumor cells to dampen PF-07260437–induced T-cell responses. Additionally, co-staining of human IgG, CD3, and GzymB (Fig. 4A) illustrated (i) the biodistribution of PF-07260437 in the tumor, the target tissue; (ii) the deposition of PF-07260437 on tumor cells and T cells, the target populations; and (iii) the activation of T cells marked by GzymB production at the interface of T-cell/tumor cell engagement induced by PF-07260437.

qRT-PCR in tumor homogenates detected genes not easily assessed by protein-based methods (Fig. 4B and C). In addition to confirming the accumulation of T cells in tumors with a CD3E probe, T cell–specific cytokine genes, human IL2 and IFNG, were found to be induced in a dose- (Fig. 4B) and time-dependent manner (Fig. 4C) after PF-07260437 treatment. Human chemokines CXCL9 and CXCL10 were induced along with IFNG upregulation. CXCL9 and CXCL10 are IFN-γ–inducible T-cell chemokines not expressed by T cells. As the only human components in the xenograft tumors are T cells (Supplementary Fig. S3) and tumor cells, this result indicated that the human chemokines were most likely produced by tumor cells in response to IFN-γ produced from activated T cells.

Flow cytometric analysis was performed on single-cell suspensions of tumor tissues (Supplementary Fig. S3), and cell numbers were enumerated with absolute counting beads and normalized to tumor weight. PF-07260437 induced time- and dose-dependent accumulation of tumor-infiltrating CD3⁺ T cells starting at day 5, consistent with IHC findings (Fig. 4D). CD8⁺ T cells presented an activated phenotype, showing time- and dose-dependent upregulation of the effector molecule GzymB (Fig. 4E) and a component of the high-affinity IL-2 receptor, CD25 (Fig. 4F). Apart from CD3⁺ T cells, no other human immune cells were found in the tumor-infiltrating human leukocytes (Supplementary Fig. S3).

To visualize intratumoral infiltration of T cells in vivo, ImaginAb’s ⁸⁹Zr-labeled anti-human CD8 minibody (⁸⁹Zr-Df-IAB22M2C) for PET imaging was utilized (22). MDA-MB-468 tumor-bearing mice humanized with human PBMCs were treated with escalating PF-07260437 doses and imaged by CD8 PET at early (day 4) and late (day 8) time points (Fig. 5A).

In vivo PET imaging (Fig. 5B) and ex vivo gamma counting quantitation (Fig. 5C) demonstrated that at the early time point, there was a clear, stepwise increase in tumor uptake of the CD8 minibody that corresponded with the PF-07260437 dose, with the low, medium, and high doses achieving 5.19, 5.93, and 6.89 %ID/g minibody uptake, respectively, compared with 5.14 %ID/g in the negative control bispecific (ex vivo data). Although the lower doses elicited only minimal change, differences between the high dose and negative control bispecific reached statistical significance (P < 0.05). The dose-dependent increase in tumor CD8 signal persisted, being even more pronounced at the later time point after two treatment doses, when low, medium, and high PF-07260437 doses, respectively, resulted in 6.16, 7.41, and 9.10 %ID/g of tumor uptake, compared with 5.86 %ID/g in the negative control. The differences in both the medium- and high-dose groups reached statistical significance at this late time point (P < 0.05 and P < 0.0001, respectively). The approximately 5 %ID/g tumor signal observed in the negative control groups was potentially caused by nonspecific retention of the minibody tracer due to tumor-associated enhanced permeability and retention. Such background levels of uptake are common with large molecule PET tracers (23).

Tumors from cohort B (the late time point) were examined by IHC, and the density of CD8a⁺ cells was quantified (Fig. 5D and E). Consistent with PET imaging and gamma counting, PF-07260437 induced a dose-dependent increase in intratumoral CD8 T-cell staining, with the difference between the high-dose and the negative control groups being statistically significant (P < 0.0001). Together, these data suggest that PF-07260437 was effective at increasing CD8 T-cell numbers in the tumor microenvironment.

The CD8-PET tracer also allows the evaluation of uptake at the whole body and tissue levels. We observed overall higher levels of tumor and organ distribution of the CD8 minibody in cohort B compared with cohort A, with the most noticeable increases detected in the spleen, liver (Supplementary Fig. S4), lungs, and the gastrointestinal tract (Fig. 5B). This can be attributed to the inherent expansion of human-derived T cells in human PBMC-engrafted mice, suggestive of the impending onset of GVHD, which is known to primarily target these organs (24). The 80 kDa anti-CD8 minibody is cleared through both the hepatic and renal routes (22). Therefore, the overall high tissue uptake in the liver and kidneys across treatment groups and time points was expected as part of the PK profile.

We evaluated the benefit of combining PF-07260437 therapy with standard of care (fulvestrant and palbociclib as monotherapies or in combination) in an HR⁺ model, ZR-75-1. Tumor-bearing mice were enrolled in the study 55 days after tumor implantation (day 0) and pretreated with fulvestrant, palbociclib, or both for 20 days. Subsequently, each of the pretreated groups was subdivided and received either PF-07260437 or a negative control at weekly intervals while continuing the same pretreatment regimen. The treatments were well tolerated with minimal effects on mouse body weight (Supplementary Fig. S5A) even after 40 days of continuous treatment. Palbociclib, fulvestrant, and their combination resulted in significant control of tumor growth versus vehicle control after 20 days (Fig. 6A; Supplementary Table S12, P > 0.001 for all comparisons). Subsequent PF-07260437 dosing decreased mean tumor volume at day 34 when all animals were still on study (Fig. 6B; Supplementary Table S13), indicating that PF-07260437 can drive tumor regression in mice exposed to prior lines of therapy. Combination treatment groups containing fulvestrant had decreased tumor volumes versus PF-07260437 monotherapy (Supplementary Table S13).

Bliss analysis uses a mathematical model to predict the combined effect of different treatment groups, resulting in a Bliss-predicted combination curve (19). If the actual effect of a treatment combination matches the Bliss line (nonsignificantly different), the treatments are considered to have an additive effect. If the combination results in significantly greater tumor control than the Bliss line, the treatments are considered synergistic. Bliss analysis demonstrated a synergistic combination benefit of PF-07260437 + fulvestrant (Supplementary Fig. S5B) and the PF-07260437 + fulvestrant + palbociclib triplet versus PF-07260437 monotherapy + fulvestrant + palbociclib combination treatment (Supplementary Fig. S5C).

PD evaluation using RNA and IHC readouts on study day 28 demonstrated that ZR-75-1 tumors treated with the triplet combination had greater levels of CD3 expression (Fig. 6C and D) and higher PD-L1 expression (Fig. 6E and F) relative to PF-07260437 monotherapy. Similar to PD-L1, the T-cell activation/dysfunction markers PD-1, CTLA-4, TIM-3, LAG-3, and TIGIT were elevated (Supplementary Fig. S5D), highlighting potential rational combination partners for improving response and overcoming resistance.

Given the upregulation of immune checkpoint markers observed, we explored whether combining PF-07260437 with an anti–PD-1 blocking antibody could be beneficial. We used an immunocompetent mouse model genetically engineered to express human CD3e (genOway) with the syngeneic TNBC mouse tumor cell line E0771, which was stably transfected to express human B7-H4 protein. This enabled both epitope-binding arms of PF-07260437 to bind to human protein in a mouse setting. E0771-hB7-H4 tumors were assessed by IHC for in vivo B7-H4 expression levels, showing a moderate level of expression (H-score = 121; Supplementary Fig. S6A).

A single subcutaneous dose of PF-07260437 induced 72% TGI, whereas the mouse anti–PD-1 surrogate monotherapy treatment, administered intraperitoneally every three days for 6 times, induced a minimal 5% TGI (Fig. 6G; Supplementary Table S14). Combining both treatments significantly increased antitumor activity to 106% TGI, with 9 of 15 mice having no tumor burden (Supplementary Fig. S6B). Bliss analysis of the two monotherapies demonstrated a significant combination benefit (Fig. 6G; P < 0.001). These data demonstrate an improved response with a combination treatment of PF-07260437 and an immune checkpoint inhibitor (ICI).

Additionally, PD evaluation assessed intratumoral T-cell infiltration at 9 days after treatment using flow cytometry. Significant increases in CD3 and CD8 T-cell infiltration were observed in the combination-treated groups (Fig. 6H) compared with the monotherapy or vehicle groups.

Both the human B7-H4 and human CD3 arms of PF-07260437 do not cross-react with rodents, preventing the formation of a trimer of tumor/drug/T cell in mice. Due to this lack of binding and CD3 functional activity in rodents, PF-07260437 was tested in cynomolgus monkeys, the only pharmacologically relevant toxicology species available with similar low nanomolar binding to both human and cynomolgus T cells (Fig. 2B). Nonclinical safety and TK of PF-07260437 were evaluated in cynomolgus monkeys in an exploratory toxicity study (once weekly intravenous administration for two doses on days 1 and 8) at 30 (1 male), 100 (1 male and 1 female), and 300 (1 female) µg/kg/dose (Supplementary Fig. S7A).

TK analysis of PF-07260437 in all dose groups showed that the serum concentrations followed biphasic kinetics (Supplementary Fig. S7B), consistent with human IgG data in nonhuman primates (25). Key TK measures, including Cmax (peak observed concentration) and AUC, indicated that systemic exposures of PF-07260437 increased dose proportionally. The apparent terminal half-life over the 0- to 168-hour interval was up to 8 days. In cynomolgus monkeys, similar to mice, PF-07260437 PK was dose-proportional with a terminal half-life of 6 to 11 days. In both species, PF-07260437 exhibited typical human IgG–like, dose-proportional PK, suggesting typical clearance via nonspecific pinocytosis and catabolism.

All dose levels were tolerated. The main clinical signs in PF-07260437–dosed animals were emesis, salivation, being cool to the touch, and decreased body temperature. These clinical signs were transient and were observed only after the first dose administration at 300 µg/kg/dose, consistent with cytokine release. Anticipated PD effects of PF-07260437 were observed at ≥100 µg/kg/week, demonstrated by increased percentages of activated (CD69⁺) CD8⁺ T cells in peripheral blood (Supplementary Fig. S7C) and elevated serum cytokines such as IFN-γ, IL-2, IL-10, and IL-6, peaking at 7 hours after the dose (Supplementary Fig. S7D). Cytokine induction observed after the first dose was dampened after the second dose, indicating a priming effect.

Liver function tests in the toxicity studies showed no observable liver signals (e.g., alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase, or total bilirubin) at any dose level. Additionally, the expression of B7-H4 protein was undetectable in the liver. Microscopic observations at all doses included mononuclear cell infiltration and/or epithelial degeneration/regeneration in multiple organs (kidney, urinary bladder, stomach, duodenum, lung, mammary gland, skin of the injection site, pituitary gland, thymus, bone marrow, and/or spleen), generally consistent with B7-H4 expression, with many findings following a dose response. The limited severity (rarely above mild) of these findings would have no expected impact on the survivability of cynomolgus monkeys, indicating the doses evaluated were tolerated from a nonclinical perspective.

Over the past decade, ICIs have revolutionized oncology treatment (26). However, in breast cancer, ICIs have had limited success (27), except for TNBC (28). The low frequency of tumor neoantigens (29) and fewer cognate T-cell clones that ICIs can enhance to eliminate tumor cells may contribute to this limited success. Redirected T-cell killing with a CD3 bispecific molecule targeting a tumor-associated antigen (TAA) in an MHC-independent manner could be a promising strategy to overcome the shortcomings of ICIs. We have previously demonstrated the utility of CD3 bispecifics with other tumor-targeting antigens for treating various cancers, including solid tumors (13, 30–33).

Our data and other studies (6, 34–36) show that B7-H4 is a promising TAA for targeting breast cancers. B7-H4 is highly expressed across all three major breast cancer molecular subtypes (HR⁺HER2⁻, HER2⁺, and TNBC). As breast cancer treatments become more fragmented, B7-H4’s broad expression offers a unique opportunity to develop novel therapeutics benefiting a wide patient population. Although B7-H4 is expressed in some normal tissues (6), its lower expression compared with tumor tissue makes it an attractive target. In an exploratory toxicity study, the B7-H4xCD3 antibody PF-07260437 showed a clinically manageable toxicity profile. These data highlight the potential of targeting B7-H4 in humans with a CD3 bispecific molecule.

Since its discovery (37), B7-H4, due to its structural similarities with PD-L1, was thought to suppress T-cell activation by binding to an inhibitory receptor on T cells (20, 37). However, despite advancements in proteomics over the past decade, the binding partner of B7-H4 on T cells remains unidentified (20). Conflicting reports exist on the role of human B7-H4 on T cells (38, 39). Our study did not detect B7-H4 expression in human tumor–associated macrophages or major leukocyte populations as previously claimed (40). Given that CD3 bispecifics are dosed at low levels in the clinic because of safety concerns, even if B7-H4 has immunosuppressive effects on T cells, a B7-H4 targeting CD3 bispecific is unlikely to reach critical receptor occupancy for immune checkpoint blockade at clinical doses.

PF-07260437 demonstrated significant in vivo efficacy in multiple CDX and PDX models. In CDX models, some lines (MDA-MB-468, T-47D, and MX-1-Luc) had complete regression at higher doses, whereas others (HCC-1954) showed stasis. This differential response may be due to varying effector T-cell infiltration, immunosuppressive tumor microenvironments, intratumoral PF-07260437 exposure levels, or unidentified tumor intrinsic mechanisms. Further studies examining PK and PD (e.g., flow cytometry, mass cytometry, RNA sequencing, or SomaScan assays) in these models could clarify these mechanisms.

Higher doses of PF-07260437 were needed to achieve in vivo efficacy in PDX models compared with CDX models despite comparable B7-H4 tumor expression. This may be due to the more complex structural and stromal components in PDX models, which hinder T-cell infiltration. As T cells are essential for PF-07260437–induced tumor killing, higher levels of drug exposure may be required in PDX models. Dedicated biomarker studies comparing the kinetics of T-cell infiltration in PDX and CDX models would allow further testing of this hypothesis.

Biomarker studies showed that PF-07260437 triggers a cascade of activities leading to an antitumor immune response. These findings suggest that PF-07260437 may induce a self-perpetuating and self-amplifying loop (proposed model illustrated in Supplementary Fig. S8). Key steps include the following: (i) PF-07260437 binds B7-H4 on tumor cells and CD3 on T cells, forming an immune synapse and activating T cells; (ii) activated T cells produce IL-2 and upregulate its high-affinity receptor (41) and hence are able to proliferate further; (iii) T cells express the master transcriptional factor T-bet and produce IFN-γ (41), which induces tumor cells to produce the chemokines CXCL9 and CXCL10 (42); (iv) these chemokines recruit more T cells into the tumor; (v) T-cell activation, proliferation, and recruitment lead to T-cell accumulation in the tumor; and (vi) accumulated T cells produce functional GzymB to kill tumor cells (42). These biomarker studies showed that PF-07260437 was able to induce T-cell accumulation in tumors that had low baseline infiltration of T cells, as very low numbers of T cells could be identified early in the tumor or in contemporary tumors that were not treated with CD3 bispecific. Additionally, these data indicated that once T cells started to infiltrate, they kick-started a self-perpetuating and self-amplifying immune response, suggesting that PF-07260437 could induce activity in poorly infiltrated human tumors.

Biomarker evaluation after invasive tumor biopsies or systemic evaluation using whole blood has inherent limitations. Noninvasive PET imaging–based biomarkers provide whole-body information, revealing heterogeneity across tumor nodules, lymph nodes, and any untoward on- or off-target toxic effects. CD8-PET imaging can demonstrate the interlesion heterogeneity, whole-body distribution of T cells and immune-related adverse events after ICI therapy (43). We used ⁸⁹Zr-Df-IAB22M2C, a CD8 minibody, to visualize the distribution of CD8⁺ leukocytes in the tumor and whole body (33). CD8-PET quantified the increase in CD8 T-cell infiltration in a time- and dose-dependent manner in the tumors after PF-07260437 treatment. The progressive uptake in T cell–rich tissues highlights potential utility as a noninvasive PD biomarker in clinical studies.

The heterogeneous expression of TAAs, including B7-H4, poses a significant challenge for targeted immune therapies. Combining these therapies with standard-of-care treatments that enhance T-cell recognition of cancer cells or induce immunogenic cell death could widen the therapeutic response of CD3 engagers (44). Endocrine therapy, with or without a CDK4/6 inhibitor combination, is the standard of care for HR⁺ and HER2⁻ breast cancer (45), whereas checkpoint inhibitors have shown promise in TNBC (46). To enhance clinical translatability and to address whether PF-07260437 has combinatorial benefit in the standard-of-care setting or with ICI therapy, we performed combination studies. We found that PF-07260437 synergistically combined with fulvestrant and palbociclib, with the triple combination showing the best efficacy. PF-07260437 was also efficacious in a fulvestrant and palbociclib pretreated setting. PD readouts demonstrated upregulation of immune checkpoint genes that could potentially lead to resistance. Thus, agents and treatments that have the potential to enhance T-cell infiltration or activation in the tumor could enhance the efficacy of PF-07260437 in the clinic. In a humanized syngeneic model, PF-07260437 combined with anti–PD-1 therapy induced stronger TGI than the monotherapies alone and was associated with higher levels of intratumoral T-cell infiltration. Chemotherapy and radiotherapy could also be part of a treatment sequence to induce the generation of neoantigens and inflict tissue damage to initiate immune infiltration. Anti-angiogenic agents, by normalizing tumor-associated vasculature (47, 48), have also been shown to enhance T-cell infiltration in the tumor. Preclinical (13) and clinical studies (47, 48) have illustrated their effects on T-cell infiltration into the tumor, and the combination of such agents with PF-07260437 could be investigated further.

In summary, we demonstrate that B7-H4–positive tumors can be preferentially targeted with an anti–B7-H4/anti-CD3 bispecific. Our lead clinical candidate, PF-070260437, showed potent single-agent antitumor efficacy across a range of in vitro and in vivo breast cancer models. The efficacy of PF-070260437 was not affected by prior standard-of-care therapies, and encouragingly, therapeutic synergy was observed with the fulvestrant and palbociclib combination in vivo. Activity could be further enhanced in vivo with anti–PD-1 immune checkpoint blockade with an associated increase in intratumoral T-cell infiltration. Collectively, these findings support pursuing further evaluation of PF-070260437, along with combination agents, in clinical studies to examine their ability to reduce tumor burden in patients with breast cancers.

L. Wu reports personal fees from Pfizer during the conduct of the study and has a patent for WO2022013775A1 pending. M. Nocula-Lugowska reports other support from Pfizer, Inc. during the conduct of the study and has a patent for WO_2022_013775_A1 pending. E. Rosfjord reports employment with AstraZeneca and ownership of Pfizer and AstraZeneca stock. D. Mathur reports a patent for US11434292 (B2) issued. K. Maresca reports personal fees from Pfizer, Inc. outside the submitted work. A. Giddabasappa reports employment with Pfizer. C. Lees reports other support from Pfizer outside the submitted work and employment with Pfizer and ownership of Pfizer shares. A.T. Hooper reports other support and personal fees from Pfizer, Inc. during the conduct of the study as well as personal fees from Regeneron Pharmaceuticals outside the submitted work and has a patent for WO2022/013775 A1 pending. No disclosures were reported by the other authors.

K. Abayasiriwardana: Data curation, investigation, methodology, writing–review and editing. L. Wu: Conceptualization, investigation, writing–original draft. H. Laklai: Investigation, writing–original draft, writing–review and editing. M. Nocula-Lugowska: Conceptualization, resources, formal analysis, supervision, investigation, writing–original draft, writing–review and editing. L. Tchistiakova: Conceptualization, resources, supervision, writing–review and editing. J. Narula: Investigation, methodology, writing–review and editing. A. Jackson-Fisher: Data curation, formal analysis, investigation, methodology, writing–review and editing. J. Golas: Formal analysis, investigation, writing–review and editing. M.-H. Lam: Investigation, writing–review and editing. V. Grinstein: Investigation. J.W. Kang: Investigation. J.C. Kearney: Formal analysis, investigation. C. Hosselet: Investigation. E. Upeslacis: Investigation. L. Lemon: Investigation. Y. Zhang: Investigation, methodology, writing–review and editing. C. Ji: Investigation, methodology, writing–review and editing. B.S. Buetow: Methodology, writing–review and editing. M.B. Finkelstein: Methodology, writing–review and editing. N. Marshall: Investigation, writing–review and editing. S. Bisulco: Investigation. E. Rosfjord: Conceptualization, supervision, methodology, writing–original draft. D. Mathur: Conceptualization, supervision, writing–original draft, project administration. J. Athanacio: Formal analysis, investigation. A. Thomas: Investigation. A. Trageser: Investigation. D. Fernandez: Formal analysis, methodology, writing–review and editing. Z.K. Jiang: Formal analysis, investigation, methodology, writing–original draft, writing–review and editing. S. Ram: Formal analysis, investigation, methodology, writing–review and editing. E. Cabral: Investigation. L. Manzuk: Investigation. K. Maresca: Methodology, writing–review and editing. A. Giddabasappa: Formal analysis, supervision, methodology, writing–review and editing. C. Lees: Conceptualization, formal analysis, supervision, project administration, writing–review and editing. A.T. Hooper: Conceptualization, supervision, writing–original draft, project administration. P. Sapra: Conceptualization, supervision, writing–original draft, project administration. S. Chintharlapalli: Supervision, project administration, writing–review and editing.

The authors would like to extend their gratitude to the following individuals for their contributions: Douglas Armellino, Joanalyn Bartlome, Laird Bloom, Vlad Buklan, Nicholas Collette, Roger Conant, Mike Damore, Mike Eisenbraun, Lori Horton, Sibo Jiang, Keith Kobylarz, Meng Li, Xingzhi Tan, Steven Tran, Thomas Wolfe, Johnny Yao, and the Pfizer Comparative Medicine Technical Staff. The authors would also like to thank ImaginAb Inc. for access to the CD8 PET tracer technology and their continued support and collaboration.