Targeting CD24 Activates Macrophages to Reduce Tumor Burden in Preclinical Models of Solid Tumors

Abstract

Purpose:

CD24 is a “don’t eat me” signal overexpressed across multiple solid tumors and contributes to immune evasion by suppressing macrophage-mediated phagocytosis. Targeting the CD24/SIGLEC-10 axis represents a novel immuno-oncology strategy to restore innate immune surveillance.

Experimental Design:

We developed PHST001, a humanized IgG4 monoclonal antibody targeting CD24, and evaluated its activity using in vitro phagocytosis assays, xenograft and immunocompetent syngeneic mouse models, and ex vivo systems incorporating human immune cells and tumor samples. Nonclinical safety parameters were assessed to evaluate translational feasibility.

Results:

PHST001 binds CD24 with high affinity and blocks SIGLEC-10 engagement, resulting in enhanced macrophage-mediated phagocytosis across multiple tumor indications and subtypes, inhibition of primary and metastatic tumor growth, and prolonged survival in preclinical models. PHST001 demonstrated a favorable nonclinical safety profile and exhibited antitumor activity as both monotherapy and in combination with standard-of-care treatments, including chemotherapy, radiotherapy, and antibody–drug conjugates. Antitumor responses were associated with engagement of tissue-resident macrophages and in syngeneic models, induction of tumor-reactive T-cell responses, supporting a role for CD24 in coordinating innate and adaptive immunosuppression.

Conclusions:

These findings establish CD24 as a critical regulator of tumor immune evasion and support the clinical development of PHST001 as a CD24-targeted immunotherapy. A phase I clinical study (NCT06840886) evaluating the safety and tolerability of PHST001 in adult patients with relapsed or refractory solid tumors is ongoing.

Translational Relevance.

PHST001 is a novel anti-CD24 antibody and exhibits robust antitumor activity in cancers with high unmet medical need, including ovarian cancer, triple-negative breast cancer, endometrial cancer, and cholangiocarcinoma, and enhances the efficacy of standard-of-care therapies, such as chemotherapy and antibody–drug conjugates. Using complementary xenograft and immunocompetent syngeneic mouse models, together with in vitro and ex vivo systems incorporating human immune cells and tumor samples, we define the mechanism of action of PHST001. Our findings establish CD24 as a key regulator of tumor-mediated innate immune evasion and demonstrate that CD24 blockade promotes macrophage activation and downstream tumor-reactive T-cell responses. These data provide a strong mechanistic and translational rationale for the clinical evaluation of CD24-targeted immunotherapy in solid tumors. A phase I clinical study (NCT06840886) evaluating the safety and tolerability of PHST001 in adult patients with relapsed or refractory solid tumors is ongoing.

Introduction

CD24 is an immunomodulatory sialoglycoprotein with characterized functions in regulation of both adaptive and innate immunity (1, 2). Its mature form is a 31–amino acid peptide anchored to the cell surface via a C-terminal glycosylphosphatidylinositol anchor. CD24 is heavily glycosylated, with 3 N-linked and multiple O-linked glycosylation sites. Variations in sialoglycosylation across tumor types can influence the binding of different anti-CD24 antibody clones (3). The sialic acid–binding immunoglobulin-type lectin (SIGLEC) family of receptors, and in particular SIGLEC-10, has been demonstrated to bind CD24 (2, 4). SIGLEC-10 is expressed on immune cells including macrophages, monocytes, Kupffer cells, activated CD4⁺ T cells, B cells, dendritic cells (DC), and Schwann cells (5–7).

In some tumor types, CD24 expression levels are markedly elevated compared with paired normal tissues, with up to ∼100-fold higher expression observed in ovarian cancer (8). This overexpression is most frequent in solid tumors, including cholangiocarcinoma and ovarian, breast, endometrial, colorectal, and lung cancers (8–11). Patients with tumors expressing high levels of CD24 may have poorer prognoses and worse overall survival (OS) compared with those with low CD24 expression; for example, a number of studies have identified CD24 as a poor prognostic marker in glioma, cholangiocarcinoma, non–small cell lung cancer, breast cancer, endometrial and bladder carcinoma, and ovarian cancer (8, 12–15). In a subset of patients with breast cancer, ovarian cancer, and lung squamous cell carcinoma, the CD24 locus is focally amplified, with the degree of amplification correlating with CD24 expression levels (16). For example, a multivariate analysis demonstrated that both focal amplification of CD24 and CD24 expression were significantly associated with poor prognosis in breast cancers to a more significant degree than classical risk factors such as TP53 mutation or breast cancer subtype (16).

The ability of tumors to evade the immune system is an established hallmark of cancer (17–19). In contrast to efforts targeting the adaptive immune system, few therapies have been successfully developed to stimulate the myeloid arm of the immune system to attack cancer (20, 21). The myeloid immune lineage consists primarily of granulocytes, monocytes, macrophages, and DCs. In many tumors, macrophages are the most abundant infiltrating immune cell type and are often found polarized into an immunosuppressive state, which can directly promote tumor growth and inhibit the activity of other infiltrating immune cells (22, 23). Recent work has demonstrated that CD24 serves as a “don’t eat me” signal that suppresses macrophages through its engagement with the inhibitory receptor SIGLEC-10 (8). Antibodies that block the CD24/SIGLEC-10 interaction can potently induce phagocytosis in vitro and are effective as a monotherapy in treating human tumor xenografts in vivo (8).

Here, we report the development and characterization of PHST001, a novel humanized monoclonal IgG4 antibody designed to bind and block all glycoforms of CD24 with high affinity. PHST001 shows potent antitumor efficacy in multiple preclinical models as both a monotherapy and in combination with standard-of-care drugs while posing a low risk of inducing toxic side effects. A phase I clinical study (NCT06840886) investigating the safety and tolerability of PHST001 in adult patients with relapsed/refractory solid tumors is ongoing.

Materials and Methods

PHST001 development

Full details of PHST001 antibody affinity maturation, humanization, and analyses are provided in Supplementary Methods S1.

Animal models

Female NOD.Cg-Prkdc^SCIDIl2rg^tm1Wjl/SzJ (NSG, The Jackson Laboratory, strain #005557, RRID: IMSR_JAX:005557) mice were used for all xenograft models; female B6(Cg)-Tyrc-2J/J (albino B6, The Jackson Laboratory, strain #000058, RRID: IMSR_JAX:000058) mice were used for MC38-huCD24 and ID8-huCD24 experiments. Female C57BL/6 (The Jackson Laboratory, strain #000664, RRID: IMSR_JAX:000664) and OT-I (The Jackson Laboratory, strain #003831, RRID: IMSR_JAX:003831) mice were used for in vitro experiments. All mice were 6 to 9 weeks of age in the beginning of the experiments. Albino B6 mice were used to enhance bioluminescence detection. Mice were acclimated and maintained at Pheast Therapeutics in a conventional, light-cycled facility for at least 5 days prior to use. Animals had free access to food and water. All animal procedures and housing were in accordance with our animal protocol (PHT-01-Y4) approved by Pheast Therapeutics Institutional Animal Care and Use Committee.

Tumor implantation and mouse survival follow-up

For subcutaneous or intramammary gland tumor implantation, the designated tumor cells [MC38-huCD24, 5 × 10⁴ cells; BT-474, 10⁶ cells; MFM-223, 10⁵ cells; OVCAR-3, 2 × 10⁶ cells; PACADD-137, 5 × 10⁶ cells; patient-derived xenograft (PDX) of ovarian cancer (TM00335, The Jackson Laboratory), 2 × 10⁶ cells; KKU-213, 2 × 10⁴ cells; SK-OV-3, 10⁶ cells; and DMS-53, 5 × 10⁵ cells] were resuspended in 75 μL of serum-free media/PBS and 25 μL of Matrigel (Corning; #356231) and injected into the right flank of mice under isoflurane anesthesia (2.5% vaporized in O₂). For intraperitoneal tumor implantation, the designated tumor cells (SK-OV-3, 10⁶ cells; SNGM, 10⁵ cells; and ID8-huCD24, 10⁵ cells) were resuspended in 200 μL of serum-free media/PBS and injected into the peritoneum of the mice. For intracardiac tumor implantation, MC38-huCD24 cells (10⁵ cells) or BT-474 cells (10⁶ cells) were resuspended in 50 μL of PBS and injected into the left ventricle of the heart of mice under isoflurane anesthesia (2.5% vaporized in O₂). For survival follow-up, death events were counted when mice were found dead, had a tumor larger than 2,000 mm³, showed >20% weight loss, or showed ulcerated tumors, whichever happened first.

Bioluminescence imaging

Bioluminescence in vivo imaging was performed on an IVIS Spectrum system (RRID: SCR_018621, auto exposure, binning 4, 8, or 16, field of view (FOV) 22.5 cm, f/stop1, open filter). Mice were subcutaneously injected with D-luciferin (150 mg/kg in PBS; Revvity) and imaged 10 minutes later under isoflurane anesthesia (2.5% vaporized in O₂). For analysis, total photon flux (photons/second) was measured from a fixed region of interest over the tumor area using Living Image 4.7.4 software (Revvity; RRID: SCR_014247).

Cell line culture

The following cell lines were purchased from ATCC and cultured per the manufacturer’s guidelines: BT-474 (cat. #HTB-20, RRID: CVCL_0179), SK-BR-3 (cat. #HTB-30, RRID: CVCL_0033), OVCAR-3 (cat. #HTB-161, RRID: CVCL_0465), SK-OV-3 (cat. #HTB-77, RRID: CVCL_0532), HT-29 (cat. #HTB-38, RRID: CVCL_0320), DLD-1 (cat. #CCL-221, RRID: CVCL_0248), NCI-H1563 (cat. #CRL-5875, RRID: CVCL_1475), NCI-H292 (cat. #CRL-1848, RRID: CVCL_0455), BxPC3 (cat. #CRL-1687, RRID: CVCL_0186), Hec1a (cat. #HTB-112, RRID: CVCL_0293), Hec1b (cat. #HTB-113, RRID: CVCL_0294), BT-20 (cat. #HTB-19, RRID: CVCL_0178), HCC-2157 (cat. #CRL-2340, RRID: CVCL_1261), MCF-7 (cat. #HTB-22, RRID: CVCL_0031), Raji (cat. #CCL-86, RRID: CVCL_0511), JeKo-1 (cat. #CRL-3006, RRID: CVCL_1865), PA-1 (cat. #CRL-1572, RRID: CVCL_0479), SW-837 (cat. #CCL-235, RRID: CVCL_1729), KLE (cat. #CRL-1622, RRID: CVCL_1329), and DMS-53 (cat. #CRL-2062, RRID: CVCL_1177). The following cell lines were purchased from Japanese Collection of Research Bioresources Cell Bank (JCRB) and cultured per the manufacturer’s guidelines: KKU-213 (cat. #JCRB1557, RRID: CVCL_M261), KKU-055 (JCRB, cat. #JCRB1551, RRID: CVCL_M258), Hec108 (cat. #JCRB1123, RRID: CVCL_2923), EMTOKA (cat. #JCRB1601, RRID: CVCL_E115), and SNGM (cat. #JCRB0179, RRID: CVCL_1707). The PACADD-137 cell line was purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ) (cat. #ACC711, RRID: CVCL_1850) and cultured per the manufacturer’s guidelines. The MFM-223 (cat. #98050130, RRID: CVCL_1408) and COR-L311 (cat. #96020721-1VL, RRID: CVCL_2412) cell lines were obtained from Sigma-Aldrich and cultured per the manufacturer’s guidelines. The MC38-Luc (cat. #YC-A002-Luc-P, RRID: CVCL_C8VZ) and ID8-Luc (cat. #YC-C103-Luc-P, RRID: CVCL_C8UV) cell lines were purchased from Ubigene and cultured per the manufacturer’s guidelines. Briefly, the following cell lines were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Thermo Fisher Scientific, cat. #11995073) with 10% fetal bovine serum (FBS; Sigma-Aldrich, cat. #F0926-500ML) and 1% penicillin–streptomycin (Pen–Strep, Gibco, cat. #15140-122): MFM-223, BT-20, KKU-055, KKU-213, Hec1b, and MC38-Luc. The BT-474 cells were cultured with DMEM, 10% FBS, and 1× insulin–transferrin–selenium (ITS-G; Life Technologies, cat. #41400045). The PACADD-137 cell line was cultured in a 1:1 mixture of DMEM and keratinocyte serum-free medium (SFM) (Thermo Fisher Scientific, cat. #17005042) with 20% FBS and 1% Pen–Strep. The following cell lines were cultured in RPMI 1640 (Thermo Fisher Scientific, cat. #11875119) with 10% FBS and 1% Pen–Strep: NCI-H1563, NCI-H292, HCC-2157, DLD-1, JeKo-1, BxPC3, Raji, EMTOKA, and COR-L311. OVCAR-3 cells were cultured in RPMI-1640, 20% FBS, and 1× ITS-G. The following cell lines were cultured in McCoy’s 5A (Thermo Fisher Scientific, cat. #16600108) with 10% FBS and 1% Pen–Strep: HT-29, SK-OV-3, SK-BR-3, and Hec1a. The KLE cell line was cultured in DMEM/F-12 (Thermo Fisher Scientific, cat. #10565042) with 10% FBS and 1% Pen–Strep. The SW-837 cell line was cultured in Leibovitz’s L-15 (Thermo Fisher Scientific, cat. #11415114) with 10% FBS, 1% Pen–Strep, 2 mmol/L L-glutamine (Gibco, cat. #A2916801), and no CO₂. After purchase, cells were not routinely tested for Mycoplasma.

Cell line engineering

GFP-luciferase

Cells were infected with media containing polybrene (10 μg/mL) and the EF1a-FLuc-F2A-GFP lentivirus from Biosettia [cat. #GlowCell16, lot #21F28; multiplicity of infection (MOI) = 6] by centrifugation at 1,000 × g for 1 hour. Media was refreshed 48 hours after infection at which point cells were expanded prior to cell sorting. Cells were sorted at the Stanford Beckman Flow Cytometry Core utilizing the FACSAria Fusion Cell Sorter (RRID: SCR_025715). Sorted GFP⁺ cells were cultured as the parental cell lines, expanded, and frozen for future use.

Nuclear localized green

Cells were treated with 0.4 mL of media containing 8 μg/mL polybrene and approximately 100,000 infectious units (IFUs) (equal to MOI of 5) of the Incucyte Nuclight Green Lentivirus EF1a-Puro (Sartorius, cat. #4475) or the Incucyte Nuclight Green Lentivirus EF1a-Bleo (Sartorius, cat. #4477). Cells were then subcultured to expand the cell lines for purification using the Sony SH800S sorter (RRID: SCR_018066). Subsequent GFP⁺ cell lines were cultured, expanded, and frozen for future use.

CD24 knockout using CRISPR/Cas9 editing

Single-guide RNAs targeting either CD24 or Cd24a were purchased from Synthego [gene knockout (KO) kit] and complexed into a ribonucleoprotein (RNP) with SpCas9-2NLS (Synthego) by mixing and incubating at room temperature for 15 minutes. The RNP mixture was mixed with the target cells and nucleofected in a Lonza 2b transfection device (product code 13458999) according to user manual and preprogrammed Lonza protocols. Nucleofected cells were allowed to recover for at least 72 hours under standard culture conditions, expanded before staining with an anti-CD24 (clone: ML5, RRID: AB_2072735) or anti-CD24a (clone: M1/69, RRID: AB_439716) antibody conjugated with allophycocyanin, and sorted in a Sony SH800 cell sorter for the CD24⁻ population (Supplementary Fig. S1). The following guide RNA (gRNA) sequences were used to knock out human CD24: 5′ G*C*A*GGA​CGG​UCC​CCC​GGG​A 3′; 5′ A*G*G*GCCUCACCUGCGUGG 3′. The following gRNA sequences were used to knock out mouse Cd24a: 5′-A*G*G*GUC​UCA​CCU​GCG​UGG​GU-3′; 5′-U*G*U*UCG​CGC​GCG​CGU​AGG​AG-3′. All guide sequences also contained Synthego Modified EZ Scaffold.

Human CD24 overexpression

The ID8-luc mouse cell line was previously knocked out for Cd24a, as described previously (Supplementary Fig. S1). MC38-luc cells do not show detectable levels of mouse CD24a in the surface (Supplementary Fig. S1). MC38-luc and ID8-luc were infected with lentivirus carrying human CD24 gene (G&P Biosciences, cat. #LTV2801; MOI, 20) in the presence of 8 μg/mL polybrene. Human CD24⁺ cells were then sorted using the Sony SH800S sorter, generating custom MC38-huCD24 and ID8-huCD24 cells (Supplementary Fig. S1).

Cytoplasmic ovalbumin (OVA) overexpression

MFM-223 GFP-Luc cells were infected with lentivirus carrying a cytoplasmic OVA (cOVA) coding sequence (SignaGen Laboratories, custom order, LV-EF1a-cOVA-IRES-Puro; ref. 24) in the presence of 8 μg/mL polybrene. Cells were then selected by adding 2 μg/mL of puromycin. cOVA expression was confirmed by Western blot (anti-ovalbumin, cat. #PA1-196, Invitrogen).

Human samples

Human whole blood from healthy donors was purchased from Stanford Blood Center. Human tumor samples were purchased as whole tissue from commercial vendors (MT GROUP; Accio Biobank Online) or dissociated tumor cells (Discovery Life Sciences). Human ascites samples were obtained as discarded material from patients undergoing surgery at Stanford Hospital under a protocol approved by the Stanford University Institutional Review Board (protocol #42996). All participants provided written informed consent prior to sample collection in accordance with the Declaration of Helsinki. Primary human tissue microarrays were purchased from QuickArrays. All human samples were deidentified and handled according to Health Insurance Portability and Accountability Act regulations.

In vivo dosing

Treatments began 3 to 7 days after engraftment and randomization. PHST001 and antibody–drug conjugates (ADC) were diluted in 1× PBS; cisplatin was reconstituted in 0.9% saline. Cisplatin and ADCs were administered weekly via intravenous injection. PHST001 was administered via intraperitoneal injection 3 times per week for up to 18 doses. The dose levels are described on the legends.

Drug and ADC preparation for in vitro use

Carboplatin (MedChemExpress, cat. #HY-17393) and cisplatin (MedChemExpress, cat. #HY-17394) were purchased as dry powder and resuspended in water to prepare stock solutions at 20 and 10 mmol/L, respectively. All drugs were stored as single-use aliquots at −80°C prior to use in experiments. Trastuzumab emtansine (Kadcyla biosimilar, MedChemExpress, cat. #HY-P9921), trastuzumab deruxtecan (Enhertu biosimilar, MedChemExpress, cat. #HY-138298A), and mirvetuximab soravtansine (Elahere biosimilar, MedChemExpress, cat. #HY-P9921) were resuspended in 1× PBS to prepare stock concentrations of 1 mg/mL (trastuzumab emtansine and trastuzumab deruxtecan) or 10 mg/mL (mirvetuximab soravtansine), made into single-use aliquots, and stored at −80°C prior to use in experiments.

Human monocyte differentiation and polarization

Monocytes were thawed rapidly in a water bath at 37°C, washed in prewarmed macrophage media containing Iscove’s modified Dulbecco’s medium (IMDM; Thermo Fisher Scientific, cat. #31980097), 10% human serum (GeminiBio, cat. #100-512), and 1% Pen–Strep (Gibco, cat. #15140-122). Alternatively, monocytes were differentiated into macrophages by supplementation with X-VIVO 15 media (Lonza Bioscience, cat. #BE02-053Q), 50 ng/mL of human macrophage colony–stimulating factor (M-CSF; R&D Systems, cat. #216-MC-025/CF), and 1% Pen–Strep. Cells were pelleted for 5 minutes at 400 × g and 4°C. Pellets were resuspended in macrophage media, and cells were seeded onto non-tissue culture (TC)–treated 15-cm² bacterial Petri dishes (Sigma-Aldrich, P5856). Cells were maintained in culture at 37°C with 5% CO₂ for 7 days. Media was refreshed after 7 days in culture. For unpolarized (M0) macrophages, no additional components were added to the media. For M2-like macrophages, TGFβ (R&D Systems, cat. #240-B/CF) and IL10 (R&D Systems, cat. #1064-ILB) were added at a final concentration of 50 ng/mL to the macrophage media and respective media refreshed every 48 hours for an additional week to polarize the macrophages.

Flow cytometry and flow phagocytosis sample preparation

Full details on the preparation of samples from primary mouse tumors or human solid tumors are given in Supplementary Methods S1.

Flow phagocytosis

Human primary tumor cells were labeled with the CellTrace carboxyfluorescein diacetate succinimidyl ester (CFSE) cell proliferation kit (Thermo Fisher Scientific, cat. #C34570). Target cells were seeded into a 96-well ultralow attachment plate (Costar, cat. #7007). Antibodies were prepared in IMDM basal media to a final concentration of 50 μg/mL and added to respective wells with target cells. The target cells and antibodies were co-incubated at 37°C with 5% CO₂ for 20 to 30 minutes prior to the addition of macrophages. Macrophages were similarly lifted with TrypLE following 2 PBS washes and 1 TrypLE wash. Macrophages were incubated at 37°C with 5% CO₂ for 7 to 10 minutes and quenched with macrophage media (IMDM + 10% human serum + 1% Pen–Strep). Macrophages were centrifuged at 400 × g for 5 minutes at 4°C. The supernatant was aspirated, and the pellets were resuspended for counting. All reactions were carried out at an effector to target (E:T) ratio of 1:2. Coculture reactions were incubated for 90 to 120 minutes at 37°C with 5% CO₂. Phagocytosis reactions with primary human samples were incubated at 37°C with 5% CO₂ for 30 minutes. Cells were then pelleted at 900 × g for 2 minutes at 4°C. The supernatant was removed, and the cells were resuspended in CD11b (BioLegend, cat. #101218) prepared at the manufacturer’s suggested concentration in FACS buffer (PBS + 2% FBS). Following a 30-minute incubation in the dark on ice, cells were pelleted at 900 × g for 2 minutes at 4°C. The supernatant was removed, and the pellets were washed twice in 200 μL/well of FACS buffer. Cells were resuspended in FACS buffer containing 4′,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich, cat. #1310309) at 10 ng/mL and analyzed on the Agilent NovoCyte Quanteon (RRID: SCR_025831). All raw data were exported and analyzed in FlowJo V10.8.1 (RRID: SCR_008520) and plotted in GraphPad Prism (RRID: SCR_002798).

Flow cytometry

Following sample preparation, single-cell suspensions were resuspended in FACS buffer [PBS, 1% BSA, and 5 mmol/L ethylenediaminetetraacetic acid (EDTA)]. FcR was blocked with the appropriate reagent for 10 minutes [Mouse BD Fc Block, BD Biosciences, cat. #553142; Human BD Fc Block, cat. #564220; or anti–rat CD32 (BD Biosciences, cat. #550270) in FACS buffer containing 2% pig serum + 2% dog serum] and pelleted by centrifugation. Next, cells were labeled with the indicated pool of antibodies plus a viability dye for 30 minutes on ice (Supplementary Table S1). Intracellular staining for FoxP3 was conducted subsequently using the eBioscience Transcription Factor Staining buffer set, according to the manufacturer’s instructions. When needed, a fluorophore-conjugated secondary antibody was added after washing cells with FACS buffer and incubated for additional 30 minutes on ice. Data acquisition was performed on a NovoCyte 3000 (RRID: SCR_025890) or a NovoCyte Quanteon (RRID: SCR_025831; Agilent Technologies) flow cytometer. Data were analyzed using FlowJo software (RRID: SCR_008520) with bead-based compensation.

Cytokine release assay

To assess cytokine release, human whole blood was diluted (10%) in RPMI media. For plate-bound antibody stimulation, antibody coating was performed overnight using the indicated antibodies at 100 μg/mL. Next, the antibody solution was removed, and wells were washed once with PBS. Then, diluted human blood was added to precoated 96-well plates. For soluble antibody stimulation, the indicated antibodies were added at 5 μg/mL to diluted human blood samples in 96-well plates. Plates were incubated at 37°C for 48 hours. After the incubation period, cells were spun down, and the supernatant was collected and stored at −80°C. Cytokine release was measured based on Luminex xMAP using the MILLIPLEX Human High Sensitivity T Cell Panel (MilliporeSigma, cat. #HSTCMAG-28SPM×13) at the Human Immune Monitoring Center (RRID: SCR_018266) at Stanford University. Anti–human CD3 (clone: OKT3, RRID: AB_1107632), anti–human CD28 (theralizumab biosimilar, RRID: AB_2921570), and anti–human CD52 (alemtuzumab biosimilar, RRID: AB_2893935) were used as positive controls in this assay. The respective isotype controls were used as negative controls.

Immunogenicity assay

Human CD14⁺ monocytes were isolated from frozen peripheral blood mononuclear cell (PBMC) samples via positive magnetic bead selection (Miltenyi Biotec, cat. #130-097-052) and differentiated into immature DCs (iDC) using GM-CSF and IL4. iDCs were then harvested, washed, and loaded with each individual test protein/peptide for 4 hours at 37°C. A DC maturation cocktail containing TNFα and IL1β was then added for an additional 40 to 42 hours to activate and generate mature DCs (mDC). The expression of key DC surface markers (CD11c, CD14, CD40, CD80, CD83, CD86, CD209, and HLA-DR) at both the immature and mature stages was assessed via flow cytometry to ensure that the DCs were activated prior to T-cell interaction. A total of 1 × 10⁵ mDCs were then cocultured with 1 × 10⁶ autologous CD4⁺ T cells (isolated by positive magnetic selection) for 6 days at 37°C with 5% CO₂ in a humidified atmosphere. On day 6, autologous monocytes were isolated from PBMCs using positive magnetic bead selection (Miltenyi Biotec, cat. #130-097-052) and loaded with the same protein/peptide used to load the DCs. After incubation at 37°C with 5% CO₂ in a humidified atmosphere for 4 hours, 5 × 10⁴ monocytes/well were added to anti-IFNγ/anti-IL5 precoated FluoroSpot plates (Mabtech) along with the corresponding DC:CD4 coculture in quadruplicate (2.5 × 10⁵ CD4⁺ T cells/well). The FluoroSpot plates were incubated for 40 to 42 hours at 37°C with 5% CO₂ in a humidified atmosphere. After incubation, the FluoroSpot plates were developed using an in-house procedure and the spot-forming cells per well assessed for each cytokine under each test condition using an IRIS FluoroSpot reader (Mabtech, RRID: SCR_028448). The data analyzed were the spot-forming units per well (i.e., the number of cytokine-secreting cells per well) in response to each test condition. Data analysis was then carried out with the distribution-free resampling method (25). To assess the impact of each sample on the whole donor population, the stimulation index (SI) was calculated for each test condition in each donor. The SI was calculated by dividing the spot-producing units (SPU)/well in the test condition by the blank (assay medium only) to highlight the magnitude of the T-cell response in each donor. The geometric mean along with the median and interquartile range over the whole donor population was then calculated for each test condition.

Cytotoxicity screening

Fresh human whole bone marrow (BM) was fractionated to mononuclear cells (MNC) using density gradient centrifugation. A cell count and viability assays [by 7-aminoactinomycin D (7-AAD) dye exclusion using flow cytometry] were performed prior to cell use. The MNCs were then adjusted to the required working cell concentration and added to in vitro assay master mixes to produce the required culture master mixes. For detection of proliferation of the cell population using HALO-Tox HT (Hemogenix, RRID: SCR_028449) with 384-well plates, 25 μL of the culture master mix was plated into wells of 384-well plates to provide 5,000 to 10,000 cells/well suspensions. A measure of 2.8 μL of the appropriate compound dilution (or vehicle/diluent) was added directly to each replicate well. Plates were then transferred to a 37°C fully humidified incubator and cultured in an atmosphere containing 5% CO₂ and 5% O₂ for a duration of 6 days. This assay platform was used to detect potential hematotoxicity to SC-HPP2 primitive stem cells, SC-GEMM1 mature stem cells, and granulocyte–macrophage progenitor (P-GM), burst-forming unit–erythroid progenitor (P-BFU), megakaryocyte progenitor (P-MK), P-T-cell, and P-B-cell progenitor cells derived from human BM MNCs. Prior to processing the plates, an ATP standard curve was performed that calibrates and standardizes the ATP assay; three concentrations of ATP control samples were included. Cultures were processed by adding 25 μL of ATP enumeration reagent to each well using a liquid handler. The luminescence signal was detected in a plate luminometer.

Ex vivo phagocytosis assay

ID8-huCD24 cells were cultured as described above and lifted with TrypLE (Thermo Fisher Scientific, cat. #12604021) for 5 minutes at 37°C with 5% CO₂. Cells were quenched with appropriate media, pelleted at 450 × g for 5 minutes, and resuspended in Sartorius wash buffer. Cells were labeled with pHrodo Orange Cell Labeling Kit (Sartorius, cat. #4766) at a dilution of 1:500 per the manufacturer’s instructions. The final labeled cell pellets were resuspended at a concentration of 2 × 10⁷ cells/mL in a solution of either 400 μg/mL of PHST001 or human IgG4 (hIgG4) isotype control. A measure of 100 µL of cell + antibody suspension was injected intraperitoneally per mouse. After 24 hours, cells were harvested via peritoneal lavage and processed for flow cytometry.

Live-cell microscopy-based in vitro phagocytosis assay

Specifically for the radiation combination, target cells were irradiated using the CellRad X-Ray Irradiation System at iQ Biosciences prior to seeding on the same day. A total of 5 × 10³ target cancer cells were seeded into two 384-well plates in 40 μL media and allowed to recover overnight in the incubator at 37°C with 5% CO₂. On the day of the experiment, media was discarded and replaced with 44 μL of correspondent target cell media supplemented with 50 ng/mL M-CSF (R&D Systems, cat. #216-MC/CF), 50 ng/mL TGFβ (R&D Systems, cat. #240-B/CF), and 50 ng/mL IL10 (R&D Systems, cat. #217-IL/CF). A 10× stock of chemotherapy/ADC and a 10× stock of IgG4 (generated in-house) or PHST001 (generated in-house) were made in media supplemented with 50 ng/mL M-CSF, TGFβ, and IL10. Eight microliters of 10× IgG4 or PHST001 and 8 μL of 10× small molecule was dispensed onto cells. Twenty microliters of M2-like macrophages was added to treated cancer cells at appropriate E:T ratios (see figure legends). Coculture wells were imaged continuously every 4 hours for 72, 96, or 120 hours using Incucyte (RRID: SCR_026298) housed within a tissue culture incubator at 37°C with 5% CO₂. Normalized cell counting was calculated to compensate for radiation treatment effects in the absence of macrophages, according to the formula: normalized cell counting = (fold-change nuclear localized green (NLG)–positive objects on treatment condition with macrophages/fold-change NLG⁺ objects on treatment condition without macrophages).

OT-I T-cell stimulation assay

BM cells were harvested from a femur and a tibia of a 6-week-old C57BL/6 mouse and differentiated into macrophages by adding mouse CSF1 (50 ng/mL) in DMEM supplemented with 10% FBS and incubated for 4 days at 37°C in a humidified incubator with 5% CO₂. The BM-derived macrophages were further differentiated into M2-like macrophages by adding mouse IL10 and TGFβ (both at 50 ng/mL) and incubating for an additional 2 days at 37°C in a humidified incubator with 5% CO₂. Phagocytosis reaction followed the “Flow phagocytosis” protocol described above for 2.5 hours. After phagocytosis, target cells and mouse macrophages were spun down, and media was discarded. Then, splenic naïve OT-I T cells were labeled with CFSE (1 μmol/L) and added to the culture supplemented with RPMI/GlutaMAX (10% FBS, 1× non-essential amino acids, 10 mmol/L 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 1 mmol/L sodium pyruvate, 55 μmol/L 2-mercaptoethanol, and 1× Pen–Strep) and incubated for 72 hours in a humidified incubator (5% CO₂) at 37°C. OT-I CD8⁺ T-cell proliferation was assessed via flow cytometry, and cytokine production was assessed by Luminex (RRID: SCR_028024, panel #MDHSTC18, Eve Technologies).

Immunohistochemistry

MFM-223 xenograft tumors were fixed in 10% neutral buffered formalin (VWR, cat. #10790-714) for 24 hours, then transferred to ethanol (Sigma-Aldrich, cat. #1003723129) diluted to 70%, and stored at 4°C until shipment. Samples were embedded, sectioned, and processed by Ensigna Biosystems (RRID: SCR_028450). To assess CD24 expression in human tumors and healthy tissues, PHST001 was reformatted as mouse IgG1 (mIgG1) Fc and applied at 5 μg/mL. Antigen retrieval was performed at pH 9.5 for the detection of CD24. H-scoring was carried out by a board-certified pathologist, and data were analyzed in GraphPad Prism (RRID: SCR_002798). The list of antibodies used is provided in Supplementary Table S1.

Multiplex immunofluorescence

Tumors were fixed in 10% neutral buffered formalin (VWR, cat. #10790-714) for 24 hours, then transferred to ethanol (Sigma-Aldrich, cat. #1003723129) diluted to 70%, and stored at 4°C until shipment. Samples were embedded, sectioned, and processed by Ensigna Biosystems (RRID: SCR_028450). Briefly, paraffin was removed on slides containing tumor sections and then blocked with a peroxidase buffer. To assess macrophage infiltration, staining was done at pH 6.2 at 97°C for 20 minutes with anti-GFP (Cell Signaling Technology, cat. #2956) at 5 μg/mL and anti-F4/80 (Bio-Rad, cat. #MCA497GA) at 5 μg/mL. Horseradish peroxidase (HRP)–polymer was added following washes for 30 minutes at room temperature, and Opal 520 (for GFP detection) and Opal 690 (for F4/80 detection) were added for 10 minutes at room temperature following additional wash steps. To assess the presence of activated CD8⁺ T cells, staining was done at pH 6.2 at 110°C for 20 minutes with anti-CD8 (Cell Signaling Technology, cat. #98941) at 1.6 μg/mL and anti–granzyme B (Cell Signaling Technology, cat. #44153) at 0.23 μg/mL. HRP–polymer was added following washes for 30 minutes at room temperature, and Opal 520 (for CD8 detection) and Opal 620 (for granzyme B detection) were added for 10 minutes at room temperature following additional wash steps. DAPI was added for 10 minutes at room temperature, the slides were washed with deionized (DI) water, and coverslips were mounted with ProLong Gold Antifade mounting media. Slides were imaged using a ZEISS Axio Imager (RRID: SCR_018876) upright microscope, and data were analyzed using QuPath-0.5.1-x64 software (RRID: SCR_018257). Granzyme B⁺ CD8⁺ T cells were quantified using HALO image analysis software (Indica Labs, RRID: SCR_018350). Tumor core was defined as regions located >500 μm from the tumor edge, whereas tumor border was defined as regions located ≤500 μm from the tumor edge.

Retrogenix off-target screening

For library screening, 6,105 expression vectors, encoding both ZsGreen1 and a full-length human plasma membrane protein, secreted or a cell surface–tethered human secreted protein, plus an additional 400 human heterodimers were individually arrayed in duplicate across cell microarray slides (“slide-sets”). A custom hEGFR-ZsGreen1 expression vector was generated by Retrogenix and spotted in quadruplicate on every slide and was used to ensure that a minimal threshold of transfection efficiency had been achieved or exceeded on every slide. HEK293 (RRID: CVCL_0045) cells were used for reverse transfection/expression. PHST001 was added to each slide after cell fixation, giving a final concentration of 20 μg/mL. Detection of binding was performed using the same fluorescent secondary antibody as used in the prescreen [Alexa Fluor (AF) 647 anti–hIgG Fc, RRID: AB_2728444]. The fluorescence of each target spot was measured using appropriate scanning technologies and normalized against the secondary-only control value. Then, vectors encoding all interactions identified in the library screening, plus control vectors encoding CD20 (positive control) and EGFR (transfection and negative control), were arrayed and expressed in HEK293 cells on new slides. Confirmation screen slides and analyses were carried out as for the library screen either after cell fixation or in the absence of fixation. Slides were treated with 20 μg/mL of PHST001, 20 μg/mL of hIgG4 isotype control, 1 μg/mL of rituximab biosimilar (array positive control), or no test molecule (secondary only; negative control). Binding to target-expressing cells and untransfected cells was again assessed with fluorescence measurement using appropriate scanning technologies. Dose–response curves were performed for the targets that were confirmed on live cells. HEK293 cells were transfected with expression vectors encoding ZsGreen1 alone or together with CD24, AGRP, or NUCB1. Live-cell transfectants were incubated with a dose range (0.018–300 μg/mL) of PHST001. Cells were washed and incubated with the same AF647 anti–hIgG Fc detection antibody as used in the cell microarray screens. Cells were again washed and analyzed by flow cytometry using Accuri (BD Biosciences, RRID: SCR_019591). A 7-AAD live/dead dye was used to exclude dead cells, and ZsGreen⁺ (transfected) cells were selected for analysis.

ELISA

To prepare the ELISA plates (Corning, cat. #9018), wells were coated with 100 μL of the target protein at a final concentration of 10 μg/mL in PBS (Gibco, cat. #10010-023). The plates were sealed and left to incubate overnight at 4°C. The following day, the coated plates were cooled to room temperature and washed 3 times with 200 μL of PBS containing 0.05% Tween-20 (PBS-T; Thermo Fisher Scientific, cat. #28352). Nonspecific binding sites were then blocked by adding 100 μL of protein-free blocking buffer (Pierce, cat. #37573) to each well and incubating for 1 hour at room temperature. After discarding the blocking solution, a serial titration of the PHST001 or hIgG4 (Bio X Cell, cat. #CP147) or a serial titration of SIGLEC-10 wild-type (WT) biotin (KACTUS Bio, cat. #SIG-HM410B) or R119A biotin (KACTUS Bio, cat. #SIG-HM411B) was prepared in blocking buffer. Dilutions were applied to the coated wells and allowed to bind for 1 hour at room temperature. Specifically, for the PHST001 blocking assay, plates were washed 3 times with PBS-T, after which a fixed concentration of 80 nmol/L of biotinylated SIGLEC-10 WT (KACTUS Bio, cat. #SIG-HM410B) or biotinylated R119A (KACTUS Bio, cat. #SIG-HM411B) was prepared in blocking buffer, added to the wells, and incubated for 1 hour at room temperature. Plates were washed 3 times with PBS-T, after which 100 μL of HRP-conjugated anti–human IgG secondary antibody (Abcam, cat. #AB98624, RRID: AB_10673832) or HRP-conjugated streptavidin (Pierce, cat. #21130) were added to each well and incubated for 1 hour at room temperature. After additional washes, plates were developed with 100 μL of tetramethylbenzidine substrate (SeraCare, cat. #5120-0075) for 5 minutes at room temperature. Color development was stopped with 100 μL of 0.1 mol/L sulfuric acid (Ricca Chemical, cat. #8310-32), and the absorbance was measured at 450 nm using a microplate reader.

Statistical analyses

All statistical analyses were carried out using GraphPad Prism software (RRID: SCR_002798), with appropriate statistical tests. EC₅₀ and IC₅₀ were calculated using a sigmoidal four-parameter logistic curve fit in which “X” was used as the concentration. P values < 0.05 were considered statistically significant for all studies.

RNA sequencing and single-cell RNA sequencing analyses

Full details are given in Supplementary Methods S1.

Results

PHST001 binds to CD24 expressed on solid tumors

PHST001 was generated from a murine anti–human CD24 clone by grafting the original complementarity-determining region (CDR) residues together with Vernier zone residues onto a human IgG variable domain scaffold. Affinity maturation was performed using yeast surface display of a randomly mutagenized heavy-chain variable domain library (Supplementary Methods S1). After three rounds of selection, sequencing identified one mutation in hypervariable CDR (HCDR) 1 and one in HCDR2 enriched relative to the input sequences (Fig. 1A). Each mutation independently improved binding affinity compared with the parental antibody (Fig. 1B).

Figure 1.

PHST001, a humanized anti-CD24 antibody, binds with high affinity to cell lines and primary human tumors. A, Positional enrichment after selection for improved CD24 binding. Affinity-enhancing mutations were focused on CDRs H1 and H2. B, Effect of individual CDR1 and CDR2 mutations in an on-yeast CD24-binding assay. C, OASis humanization assessment of PHST001 in comparison with other clinical-stage antibodies and correlation with ADA. D, ELISA measurement of PHST001 binding affinity and specificity against CD24, aglycosylated CD24, and an aglycosylated peptide consisting of a randomized CD24 sequence (n = 2, technical replicates). E, Binding of PHST001 to BT-474 CD24⁺ parental cells (pink line) and to BT-474 CD24 KO cells (blue line) was assessed by flow cytometry. Gray line shows hIgG4 isotype binding to BT-474 CD24⁺ parental cells. F, Left, schematic depicting the off-target screening pipeline. Graph displays the mean fluorescence intensity (MFI) of live HEK293 cells expressing the indicated membrane or tethered protein when incubated with PHST001 (n = 2, technical replicates). G, Correlation of PHST001 (50 μg/mL) MFI (fold change relative to isotype) assessed by flow cytometry and CD24 mRNA expression on human tumor cell lines from different indications. Spearman r = 0.8824; ****, P < 0.0001. Data fit with nonlinear regression. H, Left, PHST001 surface binding to EpCAM⁺ tumor cells derived from primary tumor samples at a concentration of 50 μg/mL. Each point represents a distinct biological donor. Right, representative contour plots for each cancer type shown in the left. I, Quantification (H-score) of malignant ovarian tumor (n = 191, biological replicates), normal tissue (of the same origin; n = 6, biological replicates), and benign ovarian tumor (n = 12, biological replicates) cores stained by IHC with PHST001-mIgG1k. Welch t test [t_(27.2) = 8.273; ****, P < 0.0001] and representative images of stained tissue cores. J, Quantification (H-score) of malignant breast tumor (n = 172, biological replicates), normal tissue (of the same origin; n = 5, biological replicates), and benign breast tumor (n = 10, biological replicates) cores stained by IHC with PHST001-mIgG1k. Welch t test [t_(24.23) = 8.397; ****, P < 0.0001] and representative images of tissue cores. K, Quantification (H-score) of malignant colon tumors (n = 92, biological replicates), normal tissues (of the same origin; n = 2, biological replicates), and benign colon tumors (n = 3, biological replicates) cores stained by IHC with PHST001-mIgG1k. Welch’s t test [t_(17.12) = 6; ****, P < 0.0001] and representative images of stained tissue cores. L, Quantification (H-score) of malignant lung tumor (n = 67, biological replicates) and normal tissue (of the same origin; n = 5, biological replicates) cores stained by IHC with PHST001. Welch t test [t_(10.68) = 7.127; ****, P < 0.0001] and representative images of stained tissue cores. COAD, colon adenocarcinoma; DAB, 3,3′-diaminobenzidine; HGS, high-grade serous subtype; TNBC, triple-negative breast cancer; TPM, transcripts per million.

We then sought to maximize the humanness of a clone by building a yeast library to humanize remaining framework residues. Two additional Vernier zone residues were humanized without loss of binding affinity. We estimated the humanness of the resulting composite antibody using OASis (26), which placed PHST001 above the 60th percentile among clinical-stage antibodies, consistent with a low risk of anti-drug antibody (ADA) in patients (Fig. 1C).

PHST001 binds to both a synthetically produced CD24 peptide and CD24 protein produced in mammalian cells but not a synthetic peptide of a randomized sequence with the same amino acid composition as CD24 (Fig. 1D). This indicates that PHST001 binds glycosylated and aglycosylated forms of CD24. Specific binding of PHST001 to live cells was confirmed using flow cytometry. PHST001 bound to parental BT-474 breast cancer cells and not CD24 KO cells (Fig. 1E). To further investigate PHST001 binding specificity, we used live-cell screening Retrogenix cell microarray technology (27). CD24 was the primary specific binding partner, with only low-affinity interactions detected for AGRP and NUCB1 (Fig. 1F). Overall, PHST001 displayed specificity to CD24 with minimal off-target interactions.

Next, we assessed PHST001 binding affinity to relevant polymorphic variants of human CD24. A C > T single-nucleotide polymorphism (SNP) leads to a replacement of alanine (A) by valine (V) at position 57 (28). Peripheral blood cells from healthy donors were screened to identify A/A, A/V, or V/V SNP donors. PHST001 showed similar binding affinity (EC₅₀) across all genotypes, although differences in maximal binding were observed and are not directly comparable across experiments (Supplementary Fig. S2A).

SN3 is an antibody that has previously been used as a binding and blocking clone against CD24 (8). PHST001 binds CD24 with higher affinity than SN3 across multiple CD24-high tumor lines (Supplementary Fig. S2B and S2C). CD24 mRNA is elevated in many tumors versus normal tissues, supporting therapeutic targeting (Supplementary Fig. S2D; ref. 9). PHST001 surface binding correlated with CD24 mRNA expression across 23 tumor lines from 8 indications (Fig. 1G).

Next, we investigated PHST001 binding to primary human cancer cells across multiple indications. PHST001 bound >90% of epithelial cell adhesion molecule (EpCAM)–positive cells in all breast and ovarian tumor samples (n = 6, each) and showed high binding to EpCAM⁺ cells from endometrial and colorectal cancer samples (Fig. 1H). Consistent with CD24 expression patterns, PHST001 bound CD19⁺ B cells derived from tumor samples (tumor-infiltrating B cells) but not CD11b⁺ myeloid cells or other major immune cell populations in the tumor microenvironment (TME; Supplementary Fig. S2E and S2F). PHST001 bound malignant ovarian, breast, colon, and small cell lung cancer (SCLC) tissues more strongly than normal or benign tissues (Fig. 1I–L). Collectively, these data demonstrate robust and specific recognition of CD24 by PHST001 across different indications and subtypes of solid tumors.

PHST001 blocks CD24/SIGLEC-10 interaction and induces phagocytosis of tumor cells

To interrogate PHST001 as a potential therapeutic, we examined its ability to block the interaction between CD24 and SIGLEC-10. PHST001 caused dose-dependent blockade of SIGLEC-10 binding (Fig. 2A); mutant SIGLEC-10 lacking the ability to bind sialic acid served as a negative control (Supplementary Fig. S3A). We then generated macrophages from PBMCs and polarized them to an M2-like (IL10/TGFβ) state. M2-like (IL10/TGFβ) macrophages expressed significantly more SIGLEC-10 than unstimulated M0 macrophages (Fig. 2B). Genes previously associated with tumor-associated macrophages (TAM), including immunosuppressive TAM programs (e.g., SIGLEC-10, CD163, MARCO, SPP1, FN1, and CCL18; ref. 29), were upregulated in M2-like (IL10/TGFβ) macrophages in comparison with M0 macrophages, whereas genes associated with inflammatory responses (e.g., CD80, CD86, HLA-DRA, and HLA-DRB1) remained unchanged (Supplementary Fig. S3B and S3C). Accordingly, myeloid-derived cells from ovarian cancer expressed higher levels of SIGLEC-10 than normal tissue (Supplementary Fig. S3D). To test PHST001 in macrophage-mediated phagocytosis of cancer cell lines, we used a flow cytometry–based coculture assay (Supplementary Fig. S4A). PHST001 increased phagocytosis of cancer cells by M2-like (IL10/TGFβ) macrophages versus M0 macrophages lacking SIGLEC-10 (Fig. 2C). PHST001 induced a 4.6- to 11-fold increase in phagocytosis of breast and ovarian tumor cell lines, exceeding the activity of SN3, an established functional anti-CD24 antibody (Fig. 2D).

Figure 2.

PHST001 blocks CD24/SIGLEC-10 interaction and induces phagocytosis of human cell lines and primary tumors in vitro. A, PHST001-induced blockade of SIGLEC-10 binding to CD24 was assessed by ELISA. CD24 was immobilized on ELISA plates, and a constant concentration of SIGLEC-10 or SIGLEC-10 R119A (80 nmol/L) was applied with a titration of PHST001 (n = 2, technical replicates). B, Macrophages differentiated from donor PBMCs (n = 8, biological replicates) were cultured in human serum-containing media (M0) or further polarized with TGFβ and IL10 (M2-like). SIGLEC-10 expression was assessed by flow cytometry on M2-like (IL10/TGFβ) macrophages and matched M0 macrophages for each donor. Paired t test: T₍₇₎ = 9.328; ****, P < 0.0001. C, PHST001-induced (50 μg/mL) phagocytosis of BT-474 cells by M2-like (IL10/TGFβ) or M0 macrophages was assessed by flow cytometry after 2.5 hours of incubation (n = 4 donors, biological replicates; ****, P < 0.0001). Two-way ANOVA was performed with multiple comparisons. D, Tumor cells were cocultured with M2-like (IL10/TGFβ) macrophages in the presence of antibodies at 50 μg/mL. Phagocytosis was assessed by flow cytometry (MCF-7 PHST001, n = 3; MCF-7 SN3, n = 3; SK-OV-3 PHST001, n = 3; SK-OV-3 SN3, n = 8; MFM-223 PHST001, n = 3; and MFM-223 SN3, n = 3; isotype control shows the combination of the respective isotype controls; all biological replicates). Ordinary one-way ANOVA was performed with multiple comparisons. MCF-7 [F_(2,9) = 46.49; ****, P < 0.0001; ***, P < 0.001], SK-OV-3 [F_(2,16) = 36.16; ****, P < 0.0001], and MFM-223 [F_(2,6) = 1.732; ****, P < 0.0001; ***, P < 0.001]. E, Phagocytosis of BT-474 cells by M2-like (IL10/TGFβ) macrophages was assessed by flow cytometry upon PHST001, trastuzumab, or magrolimab treatments (all antibodies were used at 50 μg/mL; PHST001, magrolimab, and IgG4 isotype n = 6 macrophage donors, biological replicates; trastuzumab and IgG1 isotype n = 3 macrophage donors, biological replicates). Ordinary one-way ANOVA with multiple comparisons [F_(4,21) = 14.26; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001]. F, Correlation of PHST001 (50 μg/mL) MFI (fold change relative to isotype) by flow cytometry and PHST001-induced phagocytosis by M2-like (IL10/TGFβ) macrophages on human tumor cell lines from multiple indications (color-coded; n = 7 for BT-474; n = 4 for PACADD-137, HT-29, BT-20, MFM-223, HCC-2157 and SW-837; n = 3 for NCI-H292, NCI-1563, BxPC3, OVCAR-3, SK-OV-3, DLD-1, MC7, SK-BR-3, KKU-055, KKU-213, HEC1a, HEC1b, COR-L311, and JeKo-1; n = 2 for KLE and Raji; all biological replicates). Spearman r = 0.5958; **, P < 0.01. Data fit with nonlinear regression. G, Live-cell microscopy-based phagocytosis assay of BT-474 cells (NLG⁺) in coculture with M2-like (IL10/TGFβ) macrophages upon PHST001 treatment (10 μg/mL). Data are shown as fold change relative to time = 0. Left, fold change in NLG⁺ counts at 72 hours. Paired t test was performed. Right, representative cell growth curve over time. **, P < 0.01. E:T ratios = 1:1. n = 3 macrophage donors (biological replicates) per target cell line. n = 2 technical replicates per condition; data represent mean ± SD H, Representative images of the assay shown in G. BT-474 cells were cocultured with M2-like (IL10/TGFβ) macrophages (E:T = 1:1) in the presence of 10 μg/mL isotype (IgG4) or PHST001. I, Schematic representing workflow for processing primary tumors to isolate EpCAM⁺ cells for functional assays. J, Phagocytosis of primary human tumors cocultured with M2-like (IL10/TGFβ) macrophages was assessed by flow cytometry upon PHST001 treatment (50 μg/mL) after 30 minutes of incubation. Data are presented as fold change relative to isotype control. High-grade serous ovarian cancer [paired t test, n = 4 macrophage donors, biological replicates; t₍₃₎ = 6.34; **, P < 0.01], endometrial endometrioid [paired t test, n = 3 macrophage donors, biological replicates; t₍₂₎ = 5.552; *, P < 0.05], colorectal cancer [CRC; paired t test, n = 3 macrophage donors, biological replicates; t₍₂₎ = 3.584; P = 0.0698], triple-negative breast cancer [TNBC; paired t test, n = 4 macrophage donors, biological replicates; t₍₃₎ = 7.565; **, P < 0.01], and an intrahepatic cholangiocarcinoma [paired t test, n = 4 macrophage donors, biological replicates; t₍₃₎ = 3.207; *, P < 0.05]. MFI, mean fluorescence intensity. [I, Created in BioRender. Therapeutics, P. (2026) https://BioRender.com/k0vah0a.].

We then compared the potency of PHST001 with other clinical-stage monoclonal antibodies that engage with macrophages. PHST001 induced greater phagocytosis than anti-HER2 opsonizing antibody, trastuzumab, or the anti-CD47 antibody magrolimab (Fig. 2E). Across 24 tumor cell lines from various tumor indications, PHST001 binding correlated with phagocytic activity (Fig. 2F).

To understand the kinetic response of human cancer cell lines to PHST001, we used live-cell imaging to quantify the loss of GFP⁺ tumor cells over time in coculture with M2-like (IL10/TGFβ) macrophages. Although CD24 has been previously reported to modulate apoptosis and proliferation (30), PHST001-mediated blockade of CD24 did not affect tumor cell accumulation over time in the absence of macrophages (Fig. 2G; Supplementary Fig. S4B). PHST001 induced near-complete clearance of GFP⁺ tumor cells in macrophage coculture (Fig. 2G and H; Supplementary Fig. S4C), with picomolar EC₅₀ values (Supplementary Fig. S4D).

Primary tumor samples obtained from patients with cholangiocarcinoma and ovarian, colorectal, breast, and endometrial cancers were processed for PHST001 activity against EpCAM⁺ tumor cells with M2-like (IL10/TGFβ) macrophages (Fig. 2I). Following dissociation of the human tumors, CD45⁺ cells were removed and EpCAM⁺ tumor cell enrichment was confirmed by flow cytometry (Supplementary Fig. S4E). PHST001 induced phagocytosis of EpCAM⁺ tumor cells isolated from all patient-derived tumor samples tested, including ovarian ascites (Fig. 2J; Supplementary Fig. S4F–S4H). Monocytes from a patient with treatment-naïve colorectal cancer were differentiated into M2-like (IL10/TGFβ) or M0 macrophages. Monocyte-derived M2-like (IL10/TGFβ) macrophages from a patient with colorectal cancer expressed high levels of SIGLEC-10 and demonstrated greater PHST001-induced phagocytosis than M0 macrophages (Supplementary Fig. S4I). These data demonstrate that PHST001 blocks CD24/SIGLEC-10 interaction, induces macrophage phagocytosis, and promotes tumor cell clearance in vitro.

PHST001 presents a low risk of hematopoietic toxicity and cytokine release

To assess the potential adverse effects of CD24 blockade by PHST001, we sought to determine CD24 expression across healthy human tissues. In human tissues, CD24 is expressed on normal cell populations, including immune cells, endocrine tissues, kidneys, and the gastrointestinal tract (5–7). Accordingly, CD24 protein was strongly detected by immunohistochemistry (IHC) in the esophagus, colon, and kidney (Supplementary Fig. S5A and S5B). To further explore potential toxicities related to hematologic, pulmonary, and hepatic effects, we analyzed publicly available single-cell RNA sequencing (scRNA-seq) datasets. scRNA-seq analyses revealed that B cells, lung ciliated cells, and cholangiocytes express CD24, whereas CD24 is not expressed on lung macrophages, Kupffer cells, or erythroid cells (Supplementary Fig. S6A–S6C).

Next, we assessed PHST001 binding to peripheral immune cells from healthy human donors via flow cytometry. Consistent with CD24 mRNA expression, PHST001 bound robustly to peripheral B cells (Fig. 3A). Although CD24 mRNA expression was not detected in neutrophils, we observed high PHST001 on-cell binding via flow cytometry (Fig. 3A). Notably, neutrophils are known to have low RNA and high RNase content, representing a technical challenge for scRNA-seq analysis (31). Importantly, PHST001 did not bind to peripheral T cells, NK cells, red blood cells (RBC), or platelets, as observed with a commercial anti–human CD24 antibody (Supplementary Fig. S7A). Together, these data indicate that peripheral neutrophils and B cells can be targeted by PHST001.

Figure 3.

PHST001 has a low risk of causing hematopoietic toxicity and cytokine release. A, PHST001 binding to peripheral blood cells from five healthy human donors was assessed by flow cytometry. B, PHST001-induced phagocytosis of CFSE-labeled human neutrophils was assessed by flow cytometry using M2-like (IL10/TGFβ) macrophages in the presence or absence of 50% human serum/IgG-free human serum after a 2.5-hour incubation period. PHST001 EC₅₀ = 1.79 nmol/L (n = 1, representative of two independent repeats). C, Cytokine (IL2, TNFα, IL6, and IFNγ) release was assessed by Luminex after incubating diluted whole blood from 10 healthy donors (biological replicates) with the indicated antibody. Antibody treatment was given as plate-bound or soluble. Anti-CD3, anti-CD28, and anti-CD52 were used as positive controls. Wilcoxon test; *, P < 0.05; ns, not significant. D, Donor-matched MoDCs and CD4⁺ T cells from 30 healthy donors (biological replicates) were cocultured in the presence of KLH, PHST001, or KLH + PHST001. KLH was used as the positive control. E, Hematotoxicity of blood cell progenitors was assessed by measuring intracellular ATP upon PHST001 treatment. Primitive lympho-HSCs (HPP2) and primitive HSCs (GEMM1) were derived from human BM MNCs (n = 8, technical replicates). The 4-parameter logistic curve fit is shown for each treatment. Omeprazole and cycloheximide were used as negative and positive controls, respectively.

PHST001 failed to bind to the CD24 peptide from the rat, mouse, pig, dog, and cynomolgus monkey (Supplementary Fig. S7B). Accordingly, PHST001 did not bind to peripheral CD45⁺ blood cells from any of these species (Supplementary Fig. S7C). Due to the lack of cross-reactivity in relevant toxicology species, potential adverse effects of PHST001 were evaluated using in vitro assays with human CD24–expressing cells.

To further investigate whether peripheral blood cells could be targeted by macrophages upon PHST001 binding, we performed phagocytosis assays using peripheral human neutrophils as targets. Consistent with PHST001 binding to human neutrophils, we observed increased macrophage-mediated phagocytosis of human neutrophils upon PHST001 treatment (Fig. 3B). When we recapitulated the peripheral neutrophil environment by adding human serum, PHST001 failed to induce phagocytosis of neutrophils (Fig. 3B). The addition of IgG-depleted human serum did not protect neutrophils from PHST001-induced phagocytosis (Fig. 3B), suggesting that free IgG may protect peripheral blood cells from PHST001-mediated phagocytosis in the circulation.

Disruption of immune checkpoint signaling axes can lead to immune activation and, in some cases, immune-related adverse events, including systemic inflammatory responses (32). Whole-blood samples from healthy human donors were stimulated with PHST001 in plate-bound and soluble configurations (Fig. 3C; Supplementary Tables S2 and S3). Anti-CD3, anti-CD28 (theralizumab), and anti-CD52 (alemtuzumab) antibodies were used as positive controls (33). Plate-bound PHST001 stimulation did not induce cytokine release, unlike anti-CD3 and anti-CD28 stimulation (Fig. 3C; Supplementary Tables S2 and S3). Soluble PHST001 stimulation induced only a 1.32-fold increase in TNFα secretion, whereas anti-CD3 and anti-CD28 stimulation induced 25.54- and 3.01-fold increases in TNFα, respectively (Supplementary Tables S2 and S3). Other relevant inflammatory cytokines, such as IL2, IL6 and IFNγ, were not affected by PHST001 treatment (Fig. 3C; Supplementary Tables S2 and S3). Accordingly, PHST001 did not affect T-cell activation and cytokine production in mixed lymphocyte reactions (Supplementary Fig. S8A and S8B). CD24 is not expressed by T cells upon activation (Supplementary Fig. S8C and S8D) nor in other effector lymphocyte populations found in the breast TME, such as cytotoxic NK cells (Supplementary Fig. S8E). These data suggest a low potential for PHST001 to induce cytokine release.

Computational analysis of PHST001 suggested a low risk of inducing ADAs (Fig. 1C; ref. 23). To further assess PHST001 immunogenicity, monocyte-derived DCs (MoDC) were pulsed with PHST001 and cocultured with CD4⁺ T cells from the same human donors. IFNγ and IL5 secretion was assessed by FluoroSpot after restimulation with PHST001-pulsed monocytes obtained from the same donors, and the frequency of responders was calculated as an indication of immunogenicity. Moderately immunogenic proteins induce cytokine response in 10% to 20% of donors. Highly immunogenic proteins induce cytokine response in more than 20% of donors. PHST001 induced IFNγ and IL5 responses in 10% and 7% of total donors (n = 30), respectively, whereas >70% of donors secreted IFNγ and IL5 when stimulated with a known immunogen, keyhole limpet hemocyanin (KLH; Fig. 3D). PHST001 did not affect cytokine response of CD4⁺ T cells when cocultured with KLH-pulsed MoDCs (Fig. 3D; Supplementary Table S4). These findings suggest a low potential for sequence-based immunogenicity and reflect T cell–dependent immunogenicity.

Hemolysis is a potentially severe adverse effect associated with drugs that bind RBCs. PHST001 does not bind RBCs (Fig. 3A) and accordingly did not induce hemolysis in an in vitro assay (Supplementary Fig. S8F). Additionally, PHST001 did not induce toxicity in human hematopoietic stem cells (HSC) and progenitor cells, including SC-HPP2 (lympho-HSCs), SC-GEMM1 (primitive HSCs), P-BFU1 (erythropoietic progenitors), P-GM1 (granulocyte–macrophage progenitors), P-Mk1 (megakaryopoietic progenitors), P-T, and P-B cells (Fig. 3E; Supplementary Fig. S8G). These data suggest a low potential for hematotoxicity associated with PHST001.

PHST001 inhibits tumor growth in vivo

We hypothesized that PHST001 would enable phagocytosis of xenotransplanted solid tumor cells, resulting in tumor growth inhibition (Fig. 4A). PHST001 inhibited the growth of BT-474 in vivo, as evidenced by bioluminescence imaging (Fig. 4B). PHST001 showed dose-dependent antitumor activity and increased survival of BT-474–bearing mice in all doses tested (Fig. 4B). In addition, PHST001 was more efficacious than CD47 blockade with magrolimab (Supplementary Fig. S9A). PHST001 displayed antitumor effect despite the presence of polyclonal IgG (Supplementary Fig. S9B), indicating that the inhibitory effect of IgG present in human serum (Fig. 3B) is not a barrier to its efficacy in the TME. Survival of mice treated with PHST001 at 1 and 5 mg/kg persisted for 168 days following treatment cessation (Fig. 4B). Similarly, PHST001 showed antitumor activity against MFM-223, a triple-negative breast cancer model, with survival extending to 200 days following treatment cessation (Fig. 4C). Ovarian cancer xenotransplantation models were established by engrafting OVCAR-3 cells subcutaneously and the ascites-derived SK-OV-3 line intraperitoneally (Fig. 4D and E). PHST001 inhibited the growth of ovarian cancer cells in vivo (Fig. 4D and E) and extended the survival of mice bearing ovarian tumors (Supplementary Fig. S9C and S9D). In a pancreatic cancer xenograft model generated using subcutaneous PACADD-137 cells, PHST001-treated mice showed significantly lower pancreatic tumor burden and extended survival (Fig. 4F; Supplementary Fig. S9E). To further assess PHST001 efficacy against endometrial cancer and cholangiocarcinoma, we generated xenograft models with the SNGM and KKU-213 lines, respectively. PHST001 significantly inhibited the growth of SNGM (Fig. 4G) and KKU-213 tumors (Fig. 4H) in vivo and increased the survival of SNGM- and KKU-213–bearing mice (Supplementary Fig. S9F and S9G). In DMS-53 SCLC xenograft model, PHST001 showed sustained efficacy and kept tumor burden below the inferior limit of bioluminescence detection for more than 70 days after the last dose (Fig. 4I). Together, these data demonstrate that PHST001 can inhibit the growth of various solid tumors in vivo.

Figure 4.

PHST001 induces clearance of solid tumors and tumor metastases in vivo. A, Schematic of in vivo study design. Mice were treated 3 times per week for a total of 18 doses intraperitoneally. B, Left, total tumor flux (mean ± SEM) of NSG mice bearing BT-474 (subcutaneous, right flank) tumors treated with PHST001 (5, 1, or 0.2 mg/kg) or isotype control over time (n = 10 mice per group; Kruskal–Wallis multiple comparisons followed by Dunn test; ****, P < 0.0001). The dotted line depicts the limit of bioluminescence detection, as established with non–tumor-bearing mice. Right, Kaplan–Meier survival curves of BT-474 tumor-bearing mice treated with PHST001 or isotype control. Mice were assessed for survival for up to 244 days after engraftment (201 days after treatment stopped). ***, P < 0.001; ****, P < 0.0001. C, Left, total tumor flux (mean ± SEM) of NSG mice bearing MFM-223 (subcutaneous, right flank) tumors treated with PHST001 (20, 2 and 0.2 mg/kg) or isotype control over time (n = 10 mice per group; Kruskal–Wallis multiple comparisons followed by Dunn test; ***, P < 0.001). Right, Kaplan–Meier survival curves of MFM-223 tumor-bearing mice treated with PHST001 or isotype control. Mice were assessed for survival for up to 245 days after engraftment (200 days after treatment stopped). *, P < 0.05; ****, P < 0.0001. D, Total tumor flux (mean ± SEM) of OVCAR-3 (subcutaneous, right flank) tumor-bearing NSG mice treated with PHST001 20 mg/kg or isotype control over time (n = 10 mice per group; Student t test; ***, P < 0.001). E, Total tumor flux (mean ± SEM) of SK-OV-3 (intraperitoneal) tumor-bearing NSG mice treated with PHST001 20 mg/kg (n = 10) or isotype control (n = 9) over time (Student t test; **, P < 0.01). F, Tumor volume (mean ± SEM) of PACADD-137 (subcutaneous, right flank) tumor-bearing NSG mice treated with 20 mg/kg of PHST001 or isotype control over time (n = 8 mice per group; Student t test; ****, P < 0.0001). G, Total tumor flux (mean ± SEM) of SNGM (intraperitoneal) tumor-bearing NSG mice treated with PHST001 20 mg/kg or isotype control over time (n = 10 mice per group; Student t test; ***, P < 0.001). H, Total tumor flux (mean ± SEM) of KKU-213 (subcutaneous, right flank) tumor-bearing NSG mice treated with PHST001 20 mg/kg (n = 10) or isotype control (n = 9) over time (Student t test; *, P < 0.05). I, Total tumor flux (mean ± SEM) of DMS-53 (subcutaneous, right flank) tumor-bearing NSG mice treated with PHST001 20 mg/kg or isotype control over time (n = 9 mice per group; Student t test; ***, P < 0.001). J, Representative images of BT-474 tumors harvested 3 weeks into treatment (42 days after engraftment) and stained for nuclear DAPI (blue), F4/80 (red), and GFP to detect tumor cells (green). Isotype scale bars, 400 μm. PHST001 scale bars, 200 μm. Tumor core indicated by white outline. K, Left, bioluminescence of BT-474 spontaneous metastases at 56 days after engraftment from NSG mice that were engrafted orthotopically in the mammary fat pad and treated with PHST001 20 mg/kg or isotype control (n = 10 mice per group; Student t test; *, P < 0.05). Right, bioluminescence images of tumor burden in mice with BT-474 spontaneous metastases in the axillary lymph nodes. Primary tumors engrafted on the right mammary fat pad were covered to maximize the bioluminescence signal of the metastatic tumors (56 days after engraftment). L, Schematic of experimental design for intracardiac injection–based metastatic models used in L and M. Mice were treated 3 times per week for up to 18 doses intraperitoneally. M, Left, total tumor flux of BT-474 metastatic tumor-bearing NSG mice over time. BT-474 cells were implanted intracardially, and once metastases were established, mice were treated with PHST001 20 mg/kg or isotype control (n = 9 mice per group; Student t test; ****, P < 0.0001). Right, representative bioluminescence images of metastatic BT-474 tumors after treatment with PHST001 or isotype. N, Left, total tumor flux on albino C57BL/6 mice (n = 9 mice per group) bearing metastatic MC38-huCD24 tumors at 13 days after engraftment. Tumors were implanted subcutaneously in the right flank, and mice were treated via an intraperitoneal injection. Nonparametric t test. **, P < 0.01. Graphs show mean ± SEM. Right, individual growth curves for mice bearing MC38-huCD24 metastatic tumors and treated with 20 mg/kg of PHST001 or hIgG4 isotype over time. mpk: mg/kg. [A and L, Created in BioRender. Therapeutics, P. (2026) https://BioRender.com/gmssmja.].

To assess the efficacy of PHST001 in models of advanced solid tumors, NSG mice were implanted with BT-474 subcutaneously, and tumors were allowed to grow for 21 days before treatment. PHST001-treated mice showed reduced tumor burden over time (Supplementary Fig. S9H). PHST001-treated tumors showed sustained growth suppression for up to 2 months following treatment cessation (Supplementary Fig. S9H). Tumors were harvested and stained with anti-F4/80, a mouse macrophage marker. PHST001 treatment induced macrophage infiltration at the margins and within the core of BT-474 and MFM-223 tumors, as shown by immunofluorescent multiplex staining (Fig. 4J) and IHC (Supplementary Fig. S9I), respectively. To further confirm macrophage involvement in PHST001-mediated inhibition of tumor growth, we used clodronate liposomes to deplete phagocytes in the SK-OV-3 orthotopic tumor model. Successful depletion of F4/80⁺ macrophages was confirmed via flow cytometry (Supplementary Fig. S9J). PHST001 activity was abrogated in clodronate-treated mice when compared with isotype control under the same condition (Supplementary Fig. S9K), supporting macrophages as key mediators of PHST001 antitumor activity.

We hypothesized that PHST001 could prevent tumor metastases by limiting the spread of circulating tumor cells or reducing growth of recently seeded tumors. We noticed that BT-474 tumors spontaneously metastasized to the axillary lymph nodes when engrafted orthotopically in the mammary fat pad of NSG mice (34). When treated with PHST001, 4 of 10 mice presented lymph node metastases, whereas 8 of 10 mice presented lymph node metastases in the isotype control group (Fig. 4K). PHST001-treated mice showed lower tumor burden of lymph node metastases (Fig. 4K), suggesting that PHST001 may prevent tumor metastasis. Besides preventing metastatic spread, to investigate the efficacy of PHST001 against existing advanced metastatic tumors, we engrafted BT-474 cells intracardially, generating a disseminated metastatic model (Fig. 4L). PHST001 showed antitumor activity against established metastatic tumors, with PHST001-treated mice presenting lower tumor burden (Fig. 4M; Supplementary Fig. S9L). To investigate PHST001 activity against established metastatic tumors in a fully immunocompetent model, we generated murine MC38 colorectal cancer cells expressing human CD24 (MC38-huCD24) and implanted intracardially. PHST001 treatment significantly reduced metastatic tumor growth over time (Fig. 4N). Altogether, these data indicate that PHST001 can both prevent metastatic spread and treat existing primary and metastatic solid tumors of different indications, suggesting that PHST001 can be effective against hard-to-treat models.

PHST001 potentiates standard-of-care cancer therapies against solid tumors

Radiotherapy can combine with “don’t eat me” signal blockade, such as CD47, to reduce tumor growth (35, 36). We observed a dose-dependent contribution of radiation to potentiate PHST001-induced phagocytosis of cancer cells, increasing the maximum phagocytic effect and sensitizing tumor cells to lower doses of PHST001 (Fig. 5A and B).

Figure 5.

PHST001-induced phagocytosis is enhanced in combination with chemotherapy, ADCs, and radiation. A, Effect of X-ray irradiation on PHST001-induced phagocytosis of OVCAR-3 cells. Graphs show normalized cell counting at 116 hours, and normalization is described in “Patients and Methods”. Dashed lines show hIgG4 isotype effect at 10 μg/mL. E:T = 2:1. B, Effect of X-ray irradiation on PHST001-induced phagocytosis of PACADD-137 cells over time (0–120 hours). Normalized cell counting over time on each treatment condition. PHST001 was used at 0.001 μg/mL, and isotype antibody was used at 10 μg/mL. E:T = 1:1. C, Data represent fold change (FC) in cancer cell growth (NLG⁺ objects) in the presence (pink dots) or absence (gray dots) of M2-like (IL10/TGFβ) macrophages at 24 hours of treatment for MFM-223 or 72 hours of treatment for BT-474 with 0.5 μg/mL of PHST001 or isotype in the presence or absence of 10 μmol/L carboplatin (Carbo) or cisplatin (Cis). Data (mean ± SD) represent three healthy donor–derived M2-like (IL10/TGFβ) macrophages, each with two technical replicates. E:T ratios were as follows: BT-474, E:T = 1:1; MFM-223, E:T = 1:2. D, Left, tumor volume (mean ± SEM) on NSG mice bearing PDX ovarian tumor measured using a caliper. Tumor cells were engrafted subcutaneously (right flank) and treated with PHST001 20 mg/kg, cisplatin 2 mg/kg, or isotype control (Kruskal–Wallis multiple comparisons followed by Dunn test; ***, P < 0.001). Mice were treated with antibodies 3 times per week for a total of 18 doses intraperitoneally. Cisplatin was dosed once a week for a total of five doses intraperitoneally. The blue rectangle depicts the treatment period. Right, Kaplan–Meier survival curves of PDX ovarian tumor–bearing mice treated with PHST001 20 mg/kg, cisplatin 2 mg/kg, or isotype control (n = 10 mice per group). **, P < 0.01; ***, P < 0.001. E, Effect of trastuzumab emtansine or trastuzumab deruxtecan treatment in the presence or absence of 10 μg/mL of PHST001 Fc-inert in BT-474 at 96 hours. Data represent n = 2, technical replicates; E:T = 1:1. F, Total tumor flux (mean ± SEM) of NSG mice bearing BT-474 tumors engrafted subcutaneously (right flank) and treated with PHST001 5 mg/kg, trastuzumab deruxtecan 1 mg/kg, or isotype control (n = 10 mice per group; Kruskal–Wallis multiple comparisons followed by Dunn test; ****, P < 0.0001). The blue rectangle depicts the treatment period. Mice were treated with PHST001 or isotype control once a week for a total of three doses intraperitoneally. Trastuzumab deruxtecan was dosed once a week for a total of three doses intravenously. G, Effect of mirvetuximab soravtansine treatment in the presence or absence of 10 μg/mL of PHST001 Fc-inert in OVCAR-3 at 96 hours. Data represent n = 2 technical replicates; E:T = 3:1. H, Left, total tumor flux (mean ± SEM) of NSG mice bearing SK-OV-3 tumors engrafted intraperitoneally and treated with PHST001 20 mg/kg (n = 10), mirvetuximab soravtansine 15 mg/kg (n = 9), isotype control (n = 9), or a combination of PHST001 20 mg/kg and mirvetuximab soravtansine 15 mg/kg (n = 10; Kruskal–Wallis multiple comparisons followed by Dunn test; ****, P < 0.0001; *, P < 0.05). The blue rectangle depicts the treatment period. Mice were treated with PHST001 or isotype control 3 times a week for a total of 18 doses intraperitoneally. Mirvetuximab soravtansine was dosed once a week for a total of five doses intravenously. Right, Kaplan–Meier survival curves of SK-OV-3–bearing mice treated with PHST001 20 mg/kg, mirvetuximab soravtansine 15 mg/kg, or isotype control; ****, P < 0.0001. A, B, E, and F, Donor numbers represent the healthy donor ID of primary monocytes that were used to generate M2-like (IL10/TGFβ) macrophages. w/ macs, with macrophages; w/o macs, without macrophages.

We hypothesized that genotoxic chemotherapies would increase “eat me” signals on the surface of cancer cells and enhance phagocytosis with PHST001. We selected a low dose of PHST001 (0.5 μg/mL) to induce minimal phagocytosis on its own. When cancer cells were cocultured with M2-like (IL10/TGFβ) macrophages, PHST001 induced significant removal of cancer cells (Fig. 5C). Carboplatin and cisplatin monotherapies were ineffective but in combination with PHST001 induced a significant increase in cancer cell clearance (Fig. 5C). To assess the combinatorial effect of PHST001 and chemotherapy in vivo, we identified a PDX model of ovarian cancer that is resistant to single-agent PHST001 and cisplatin monotherapy. Combination therapy with PHST001 and cisplatin was significantly efficacious in controlling tumor growth (Fig. 5D). In this treatment-resistant model, combination therapy significantly improved OS (Fig. 5D).

We next explored whether PHST001 enhances the efficacy of ADCs. First, we assessed the role of macrophages in ADC-induced antitumor activity and observed a notable increase in trastuzumab emtansine, trastuzumab deruxtecan, and mirvetuximab soravtansine antitumor activity in the presence of M2-like (IL10/TGFβ) macrophages (Supplementary Fig. S10A and S10B). As these ADCs have an active IgG1 Fc, these molecules may promote phagocytosis and enhance cancer cell removal by macrophages, potentially via Fc-mediated opsonization. To deconvolute the phagocytosis effect of an ADC combination with PHST001, we engineered an Fc-inert version of PHST001, bearing a LALAGANA mutation (PHST001 Fc-inert; ref. 37), which blocked CD24 and induced less phagocytosis than PHST001 (human IgG4) in macrophage coculture (Supplementary Fig. S10C). We found that blockade of CD24 with PHST001 Fc-inert greatly enhanced the antitumor potency of trastuzumab emtansine and trastuzumab deruxtecan (Fig. 5E). Accordingly, combination of PHST001 and trastuzumab deruxtecan induced potent tumor reduction and durable antitumor response that led to complete elimination of tumors in the BT-474 xenograft mouse model (Fig. 5F). We observed a similar boost in the potency of mirvetuximab soravtansine in OVCAR-3 cells (Fig. 5G). In vivo, PHST001 improved the antitumor effect of mirvetuximab soravtansine against SK-OV-3 implanted orthotopically (Fig. 5H). These data demonstrate that CD24 blockade broadly potentiates standard anticancer therapies.

PHST001 engages TIM4⁺ macrophages and activates CD8⁺ T cells in the TME

PHST001 was efficacious and increased survival of tumor-bearing mice in the SK-OV-3 orthotopic model (Fig. 4E; Supplementary Fig. S9D). However, SK-OV-3 tumors grown in a subcutaneous environment did not respond to PHST001 monotherapy (Supplementary Fig. S11A). This prompted us to analyze the TME in the subcutaneous and intraperitoneal spaces and investigate the immune cells involved in PHST001-induced efficacy (Fig. 6A). We hypothesized that the orthotopic model may leverage a relevant, tissue-resident population that further enhances the early antitumor response (38). Indeed, upon flow cytometry analysis, we found that the majority of the CD45⁺ cells in the orthotopic TME were large peritoneal macrophages expressing the receptor TIM4 (F4/80^hi/TIM4⁺), consistent with a known population of tissue-resident macrophages, whereas the subcutaneous TME was largely comprised of infiltrating Ly6C⁺ monocytes and TIM4⁻ monocyte-derived macrophages (Fig. 6B; ref. 38). Further characterization revealed that TIM4⁺ tissue-resident macrophages expressed higher levels of SIGLEC-G, the mouse ortholog of SIGLEC-10 (2), than the TIM4^– monocyte-derived macrophages (Fig. 6C). Accordingly, TIM4⁺ macrophages from primary human ascites samples expressed significantly higher levels of SIGLEC-10 than the TIM4⁻ macrophages (Fig. 6D). Because tissue-resident macrophages of the peritoneal cavity express higher levels of SIGLEC-G (Fig. 6C), this macrophage population may be more responsive to CD24 blockade. We tested this hypothesis by developing an ex vivo phagocytosis assay in which the murine ID8 ovarian cancer cell line was engineered to overexpress human CD24 (ID8-huCD24) and labeled with a pH-sensitive dye (pHrodo orange) prior to co-injection with isotype or PHST001 into the peritoneal cavity of albino C57BL/6 mice. Peritoneal lavage was performed 24 hours after injection, and cells were analyzed via flow cytometry (Fig. 6E). PHST001 increased phagocytosis of ID8-huCD24 cells by peritoneal macrophages (Fig. 6F). When we discriminated the phagocytosing macrophage population (F4/80⁺pHrodo⁺) by either a tissue-resident (F4/80^hi/TIM4⁺) or monocyte-derived (F4/80^mid/TIM4⁻) phenotype, we found that tissue-resident macrophages presented higher levels of tumor phagocytosis than monocyte-derived macrophages upon PHST001 treatment (Fig. 6F). Moreover, PHST001 further enhanced phagocytosis, especially by tissue-resident macrophages, consistent with higher SIGLEC-G expression in this population (Fig. 6C). These data demonstrate the ability of PHST001 to leverage tissue-resident macrophages in the early stages of antitumor response.

Figure 6.

PHST001 engages tissue-resident macrophages and activates CD8⁺ T cells. A, Experimental schematic for the analysis of tissue-resident macrophages versus monocyte-derived infiltrating macrophages. B, Flow cytometry analysis of immune subsets from solid tumors subcutaneously engrafted (left) or orthotopically engrafted (right) with SK-OV-3, treated with PHST001 or isotype, and harvested after 19 days. Subsets are defined as Ly6G⁺ neutrophils, Ly6C⁺ monocytes, F4/80^hiTIM4⁺ tissue-resident macrophages, and F4/80^midTIM4⁻ monocyte-derived infiltrating macrophages. n = 5 mice per group. C, Expression of SIGLEC-G on F4/80^hiTIM4⁺ tissue-resident macrophages and F4/80^midTIM4⁻ monocyte-derived infiltrating macrophages harvested from subcutaneous solid tumors or peritoneal lavage of orthotopic tumors (n = 7 mice). D, Flow cytometry analysis of SIGLEC-10 expression in TIM4⁺ and TIM4⁻ macrophages from primary human ascites samples. Macrophages were defined as CD11b⁺CD64⁺ subset following gating to exclude CD11c⁺ DCs and CD15⁺ neutrophils. Paired t test; n = 3, biological replicates; t₍₂₎ = 5.369; *, P < 0.05. E, Experimental schematic of ex vivo phagocytosis assay in albino C57BL/6 mice engrafted with ID8-huCD24 cells overexpressing human CD24 and labeled with pHrodo. Tumor cells were treated with PHST001 or isotype prior to intraperitoneal implantation. After peritoneal lavage, macrophages and tumor cells from the mouse TME were analyzed via flow cytometry. F, Left, PHST001 induced phagocytosis of pHrodo-labeled ID8-huCD24 24 hours after intraperitoneal injection. Phagocytosis is defined as the percent pHrodo⁺ cells of total F4/80⁺ cells. Data are shown as ± SEM. Unpaired t test [t₍₈₎ = 3.991; **, P < 0.01]. Right, pHrodo orange mean fluorescence intensity (MFI) signal on F4/80^hiTIM4⁺ and F4/80^midTIM4^– macrophages harvested 24 hours after injection of ID8-huCD24 with PHST001. MFI was normalized as a fold change relative to background signal of labeled, but uneaten, target cells. n = 3 isotype-treated mice; n = 7 PHST001-treated mice. Data are shown as ± SEM. Two-way ANOVA with multiple comparisons. **, P < 0.01; ****, P < 0.0001. G, Effector (CD44⁺CD62L⁻) CD4⁺ T cells and CD8⁺ T cells, Treg frequency, and effector CD8⁺ T-cell:Treg ratio in the peritoneum of mice implanted with ID8-huCD24 cells were assessed by flow cytometry. Animals were treated for 2 weeks with isotype control or PHST001 at 20 mg/kg. Student t test; ***, P < 0.001; *, P < 0.05; ns, not significant. Graphs show mean ± SEM. H, MC38-huCD24 cells were subcutaneously implanted in the right flank of albino C57BL/6 mice (n = 10/group) and treated with isotype control or PHST001 at 20 mg/kg via intraperitoneal injection. Kaplan–Meier survival curves for mice treated with PHST001 or isotype control. *, P < 0.05. I, Mice (19 weeks) that cleared MC38-huCD24 tumors (shown in I) were rechallenged by implanting MC38-huCD24 cells in the contralateral site of the original tumor implantation. Naïve mice (6 weeks) that had not been implanted with MC38-huCD24 previously were implanted as a control. Tumor growth was assessed by bioluminescence imaging. J and K, OT-I CD8⁺ T cells were stimulated for 72 hours with mouse M2-like (IL10/TGFβ) macrophages previously fed with MFM-223 cOVA in the presence of PHST001 (50 μg/mL) or isotype control (50 μg/mL). J, Left, Prior to the addition of OT-I T cells, phagocytosis of MFM-223 cOVA cells by mouse M2-like (IL10/TGFβ) macrophages in the presence of PHST001 was assessed by flow cytometry (Student t test; ****, P < 0.0001). Right, proliferation of OT-I CD8⁺ T cells was measured by assessing CFSE dilution by flow cytometry (n = 4, technical replicates; Student t test; **, P < 0.01). K, Supernatants from OT-I T cell cocultures were harvested and analyzed by Luminex for cytokine production analysis (n = 3, technical replicates; log-normal t test; *, P < 0.05; ****, P < 0.0001). gMFI: geometric mean fluorescence intensity [A, Created in BioRender. Therapeutics, P. (2026) https://BioRender.com/b0ssgfm; E, Created in BioRender. Therapeutics, P. (2026) https://BioRender.com/pzizc5v.].

Phagocytosis of cancer cells by macrophages can lead to antigen presentation and activation of T cells (24). Hence, we assessed T-cell infiltration in the peritoneum of albino C57BL/6 mice implanted with ID8-huCD24 cells. PHST001 increased the frequency of effector CD8⁺ T cells and reduced the frequency of regulatory FoxP3⁺CD4⁺ T cells (Treg; Fig. 6G; Supplementary Fig. S11B). We also observed an increased ratio of effector CD8⁺ T cells to Tregs (Fig. 6G), indicating a decrease in immunosuppression upon PHST001 treatment. To assess this phenotype in a different syngeneic mouse model, we implanted MC38-huCD24 cells subcutaneously in albino C57BL/6 mice. PHST001 improved the OS of MC38-huCD24–bearing mice and completely cleared the tumors in 3 of 10 mice (Fig. 6H). A separate group of mice implanted with MC38-huCD24 cells subcutaneously were used to analyze the impact of PHST001 in the immune cell infiltration in the TME. The frequency of effector CD8⁺ T cells increased in the TME after two doses of PHST001 (Supplementary Fig. S11C). No differences in the frequency of effector CD4⁺ T cells were observed (Supplementary Fig. S11C). PHST001 and isotype groups showed similar amounts of total CD4⁺, CD8⁺ T cells, NK cells, and Tregs in the TME (Supplementary Fig. S11D). In addition, granzyme B⁺ CD8⁺ T cells were detected within both the tumor border and core of PHST001-treated tumors, demonstrating that activated CD8⁺ T cells are not restricted to the tumor periphery (Supplementary Fig. S11E). We hypothesized that the increased frequency of effector CD8⁺ T cells upon PHST001 treatment would contribute to the generation of immunologic memory in mice that have cleared tumors after treatment (Fig. 6H). We then rechallenged the three tumor-free mice in the PHST001 group by subcutaneously implanting MC38-huCD24 cells in the contralateral flank. Mice that were cured by PHST001 did not develop tumors after rechallenge, whereas naïve mice grew tumors when implanted with the same cells (Fig. 6I), suggesting that PHST001 generates immunologic memory. This is notable because PHST001 does not directly act on CD8⁺ T cells and is consistent with adaptive immune stimulation by macrophages following PHST001 treatment.

To further explore this hypothesis, we modified a PHST001-responsive human cell line (MFM-223) to express cOVA (MFM-223 cOVA). As illustrated in Supplementary Fig. S11F, MFM-223 cOVA cells were incubated with PHST001, cocultured with mouse BM–derived M2-like (IL10/TGFβ) macrophages to induce phagocytosis, and subsequently exposed to OVA-specific OT-I CD8⁺ T cells to assess T-cell activation. PHST001 induced phagocytosis of MFM-223 cOVA cells by mouse M2-like (IL10/TGFβ) macrophages (Fig. 6J). This was associated with increased proliferation of OVA-specific CD8⁺ T cells (Fig. 6J), indicating activation of tumor antigen–specific T cells. Consistently, PHST001 treatment resulted in elevated secretion of IFNγ, TNFα, and IL2 compared with isotype control (Fig. 6K). Collectively, these data indicate that PHST001 promotes tumor antigen cross-presentation, leading to functional activation of CD8⁺ T cells.

Discussion

Macrophages are key effectors of the innate immune system with a wide range of functional roles in the context of cancer (39). Although proinflammatory macrophages contribute to antigen presentation, direct tumor cell killing, and inflammatory cytokine production, TAMs consist mostly of anti-inflammatory macrophages that assist tumor growth by a variety of mechanisms (39). TAMs are often found in high numbers within tumors, and the degree of macrophage infiltration correlates with poor prognosis across many different types of cancer (40–46). Strategies to leverage TAMs have shown limited success. Here, we show that PHST001 specifically blocks human CD24 on the surface of cancer cells and directs these cells to be killed by macrophage-mediated phagocytosis (Figs. 1 and 2). High CD24 expression is associated with worse outcomes in ovarian and breast cancers (8). CD24 protein levels were upregulated in malignant versus normal and benign tissues in the human breast, ovary, colon, and lungs (Fig. 1I–L), indicating that CD24 is a prominent tumor-associated target. We also showed that PHST001 blocked CD24 interaction with SIGLEC-10 which is highly expressed by M2-like (IL10/TGFβ) macrophages (Fig. 2A and B).

CD47, a relevant myeloid checkpoint, limits macrophage-mediated phagocytosis through binding to SIRPa, an inhibitory receptor expressed on macrophages (47). Although blockade of CD47 has failed in clinical trials against hematologic malignancies (48), evorpacept (SIRPa-Fc silent) is still being tested in solid tumor indications (NCT04675333, NCT04675294, NCT05002127, NCT05027139, and NCT05167409). The limited success to date of CD47 blockade in human trials could be partially explained by the CD47 expression profile in normal cells (48, 49). CD24 is different from CD47 and has limited expression in normal cells. Notably, CD24 is not expressed by human RBCs or platelets (Fig. 3A), which were key targets of several anti-CD47 agents. Although CD24 was detected on the surface of peripheral neutrophils and B cells (Fig. 3A), peripheral neutrophils were not targeted by macrophages in the presence of serum (Fig. 3B). In addition, PHST001 was engineered with an IgG4 Fc backbone rather than an IgG1 to minimize Fc-mediated cytotoxicity, including antibody-dependent cell-mediated cytotoxicity and complement activation. Because its mechanism of action is based on blocking the CD24/SIGLEC-10 axis to enhance macrophage-mediated phagocytosis, limiting Fc effector function may reduce off-target depletion of CD24-expressing nonmalignant cells, particularly neutrophils, while preserving therapeutic activity. Although these findings suggest that CD24-expressing peripheral cells may be less susceptible to PHST001-mediated phagocytosis, further evaluation in more complex in vivo settings will be required to further characterize this effect.

Targeting myeloid checkpoints to leverage macrophage infiltration has the potential to circumvent the poor immune infiltration often associated with the lack of response to immunotherapy in solid tumors (50). In this study, we demonstrated that PHST001 restrained tumor growth in various models of solid tumors including cholangiocarcinoma and breast, ovarian, endometrial, colorectal, lung, and pancreatic cancers (Figs. 2 and 4). Although cross-reactivity between human CD24 and mouse SIGLEC-G was not directly assessed in this study, prior work has demonstrated interaction between human CD24 and mouse SIGLEC-G, supporting conservation of the CD24/SIGLEC axis across species (2). Consistent with this, PHST001 promoted phagocytosis of human tumor cells by mouse BM–derived macrophages (Fig. 6J), providing functional evidence of cross-species activity. However, direct validation of this interaction would further strengthen interpretation of xenograft model findings. Metastasis is the primary reason for failure of local therapies, such as surgery or radiotherapy (51), representing a major therapeutic challenge. PHST001 not only inhibited the growth of primary site tumors but also prevented the occurrence of spontaneous metastases in the lymph nodes (Fig. 4K) and eliminated previously established metastases (Fig. 4M and N). These data suggest that PHST001 may be effective against both primary and metastatic diseases.

Cancer treatments often leverage induction of genotoxic stress, which can activate the immune system through recognition of antigens resulting from immunogenic cell death (ICD; ref. 52). Macrophages trigger this type of immunologic response because of their phagocytic capacity. Cancer cells treated with various chemotherapy agents and radiation display “eat me” signals that are recognized by macrophages and DCs (52). Here, we showed that chemotherapy agents, radiation, and ADCs enhanced PHST001-mediated phagocytosis in multiple models of solid tumors (Fig. 5) in vitro and in vivo, demonstrating how myeloid checkpoint therapy can combine with standard-of-care cancer therapy. Although the mechanisms underlying standard-of-care–enhanced phagocytosis were not directly assessed in this study, both Fc-mediated opsonization and ICD-associated signals represent plausible, nonmutually exclusive contributors (53, 54). Further studies will be required to determine the relative contribution of these mechanisms to PHST001 activity. This approach has been shown to improve radiotherapy response when combined with CD47 blockade in preclinical models of SCLC and glioma (35, 36). Moreover, CD24 overexpression has been recently reported as a putative mechanism of therapy evasion in pancreatic tumors treated with Kras G12C and Kras G12D inhibitors (55), suggesting that the concomitant CD24 blockade and Kras inhibition is a promising therapeutic approach.

Our histologic analyses show that CD24 blockade by PHST001 increased macrophage infiltration into the tumor (Fig. 4J; Supplementary Fig. S9I), potentially reprogramming an immunologically “cold” TME. Besides macrophages, other immune cell populations can be involved as part of the mechanism of action of PHST001, as previously demonstrated when targeting CD47 (56–58). Although our data support a central role for macrophage-mediated phagocytosis in the antitumor activity of PHST001, the contribution of other immune populations cannot be fully excluded. Here, we provide evidence that tissue-resident macrophages may play a major role in PHST001-induced antitumor activity (Fig. 6) and that PHST001 preferentially leverages TIM4⁺ macrophages to phagocytose tumor cells (Fig. 6F). TIM4 is a scavenger receptor that binds to phosphatidylserine to mediate phagocytosis of apoptotic cells (59), and studies have shown that TIM4⁺ macrophages are required to mediate cross-presentation of tumor antigens and are associated with higher T-cell infiltration in various cancer types (38, 60).

Modulating macrophage-mediated phagocytosis is a promising strategy that promotes tumor antigen presentation to T cells (24). Although immune checkpoint therapies (ICT) have drastically improved outcomes in many oncology indications, ICT has limited efficacy in many solid tumor types. TAMs are a major barrier that limits ICT response in solid tumors (39). Several approaches are under investigation to shift macrophage phenotypes from immunosuppressive toward inflammatory states (61), thereby improving ICT response in solid tumors. In accordance with observations that phagocytes are required to achieve full anti–PD-1 therapeutic response in a glioma mouse model (62), we showed that CD24 blockade with PHST001 increased the frequency of effector CD8⁺ T cells in two different syngeneic models (Fig. 6G; Supplementary Fig. S11C) and led to immunologic memory against the tumor in a syngeneic mouse model of colorectal cancer (Fig. 6H and I). Accordingly, PHST001-mediated phagocytosis promoted functional activation of tumor antigen–specific CD8⁺ T cells, increasing T-cell proliferative response and cytokine production (Fig. 6J and K). Because immune populations were analyzed as frequencies within the live-cell compartment, absolute changes in other TME cell populations cannot be excluded. Thus, CD24 blockade may initiate a robust innate immune response that promotes durable adaptive antitumor immunity.

PHST001-mediated enhancement of CD8⁺ T-cell activation and reduction in Tregs support a shift toward a more immunostimulatory TME. By increasing macrophage-mediated phagocytosis, CD24 blockade may enhance antigen presentation and promote expansion of tumor-reactive T cells, complementing immune checkpoint blockade. Consistent with this, therapies targeting innate immune suppression have demonstrated synergy with immune checkpoint inhibitors in preclinical models (63). Together, these findings provide a strong rationale for combining PHST001 with PD-1/PD-L1 blockade, as well as with other standard-of-care therapies.

In the present study, we show that PHST001 is efficacious across multiple solid tumor models. These data support ongoing clinical evaluation in patients with relapsed or refractory solid tumors (NCT06840886) and the development of biomarker-driven strategies, including CD24 expression and immune contexture. Overall, our findings indicate that targeting the CD24/SIGLEC-10 axis may activate both innate and adaptive immunity and could provide clinical benefit, particularly in combination with standard-of-care therapies.

Supplementary Material

Supplemental Figures 1-11

Supplemental figures and legends from Supp. Fig. 1-11.

Supplemental Table 1

Supplemental Table 1 shows antibody information.

Supplemental Table 2

Supplemental Table 2 shows CRA data.

Supplemental Table 3

Supplemental Table 3 shows raw CRA data.

Supplemental Table 4

Supplemental Table 4 shows raw cytokine data from immunogenicity assay.

Supplemental Methods 1

Supplemental methods show additional methodological information and relevant references.

Data Availability

All analyzed scRNA-seq datasets have been published and are available at European Genome-phenome Archive (RRID: SCR_004944; accession numbers: EGAS00001004809, EGAD00001006608, EGAD00001005064, and EGAD00001005065), ArrayExpress (RRID: SCR_002964; accession numbers: E-MTAB-8107, E-MTAB-9221, and E-MTAB-10553), and Gene Expression Omnibus (GEO; RRID: SCR_005012; accession numbers: GSE130148 and GSE154600) databases. RNA-seq datasets for M2-like (IL10/TGFβ) and M0 macrophages are available at the GEO database (accession number: GSE335752). Additional raw data supporting the findings of this study, including data underlying the figures, are available from the corresponding author upon reasonable request.

Authors’ Disclosures

D.V. Faget reports personal fees and other support from Pheast Therapeutics, Inc. during the conduct of the study. R.E. Brewer reports personal fees and other support from Pheast Therapeutics, Inc. during the conduct of the study. J. Sampaio reports other support from Pheast Therapeutics, Inc. during the conduct of the study. G. Blacker reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study. G.C. Forcina reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study. S. Qazi reports nonfinancial support and other support from Pheast Therapeutics Inc. during the conduct of the study. P.R. Malusare reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study. S. Ludwig reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study, as well as a patent for US 19/670,779 pending. K.E. Hart reports nonfinancial support and other support from Pheast Therapeutics, Inc. during the conduct of the study. A. Beans reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study. R. Virani reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study. O. Dorigo reports grants from Genentech, AstraZeneca, TORL BioTherapeutics, A2 Biotherapeutics, Pheast Therapeutics, Inc., and NIH and personal fees from Eli Lilly and Company, Corcept Therapeutics, and Immatics during the conduct of the study. R. Rousseau reports current employment with Pheast Therapeutics, Inc. with stock options and equity ownership in the company. P.-J. Ko reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study. J.Y. Cao reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study, as well as a patent for WO2025101970A1 pending. J.S. Burg reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study, as well as a patent for P 24 01 03075 pending, a patent for AU2024376811 pending, a patent for BR112026011105-4 pending, a patent for CA3310683 pending, a patent for EA202691601 pending, a patent for EP24813291.2 pending, a patent for IL328281 pending, a patent for IN202617066751 pending, a patent for JP 2026-527380 pending, a patent for KR 10-2026-7017629 pending, a patent for MX/a/2026/005611 pending, a patent for TBD pending, a patent for NZ832543 pending, a patent for SG11202602929Q pending, a patent for TW113143038 pending, a patent for US 19/670,779 pending, and a patent for ZA2026/04632 pending. R. Majeti reports other support from Pheast Therapeutics, Inc. outside the submitted work; being a member on the advisory boards at Kodikaz Therapeutic Solutions, Prelude Therapeutics, Mubadala Capital, Aculeus Therapeutics, Sequentify, Bristol Myers Squibb, and Bectas Therapeutics; and being a co-founder and equity holder of MyeloGene. I.L. Weissman reports other support from Pheast Therapeutics, Inc. during the conduct of the study; other support from Pheast Therapeutics, Inc. outside the submitted work; and a patent for CD24 Discovery pending and licensed by Stanford University to Pheast Therapeutics, Inc. A.A. Barkal reports nonfinancial support and other support from Pheast Therapeutics, Inc. during the conduct of the study; a patent for this work issued, licensed, and with royalties paid from Pheast Therapeutics, Inc. and being a co-founder, board member, and shareholder of Pheast Therapeutics, Inc. R.L. Maute reports personal fees, nonfinancial support, and other support from Pheast Therapeutics, Inc. during the conduct of the study, as well as patents (all pending) for P 24 01 03075, AU2024376811, BR112026011105-4, CA3310683, EA202691601, EP24813291.2, IL328281, IN202617066751, JP 2026-527380, KR 10-2026-7017629, MX/a/2026/005611, NZ832543, SG11202602929Q, TW113143038, US19/670,779, ZA2026/04632, and PI 2026002689. S.A. Kahn reports a patent for P 24 01 03075 pending, a patent for AU2024376811 pending, a patent for BR112026011105-4 pending, a patent for CA3310683 pending, a patent for EA202691601 pending, a patent for EP24813291.2 pending, a patent for IL328281 pending, a patent for IN202617066751 pending, a patent for JP 2026-527380 pending, a patent for KR 10-2026-7017629 pending, a patent for MX/a/2026/005611 pending, a patent for NZ832543 pending, a patent for SG11202602929Q pending, a patent for TW113143038 pending, a patent for US 19/670,779 pending, and a patent for ZA2026/04632 pending. No disclosures were reported by the other authors.

Authors’ Contributions

D.V. Faget: Data curation, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing. R.E. Brewer: Data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. J. Sampaio: Formal analysis, investigation, methodology, writing–original draft. G. Blacker: Formal analysis, investigation, methodology. G.C. Forcina: Formal analysis, investigation, methodology, writing–original draft. S. Qazi: Formal analysis, investigation, methodology. P.R. Malusare: Formal analysis, investigation, methodology. S. Ludwig: Formal analysis, investigation, methodology. K.E. Hart: Investigation, methodology. J.M. Hansen: Formal analysis, investigation, methodology. A. Beans: Investigation, methodology. R. Virani: Formal analysis, investigation, methodology. O. Dorigo: Resources, writing–review and editing. R. Rousseau: Writing–review and editing. P.-J. Ko: Formal analysis, supervision, investigation, methodology. J.Y. Cao: Data curation, supervision, project administration, writing–review and editing. J.S. Burg: Data curation, supervision, writing–original draft, project administration, writing–review and editing. R. Majeti: Conceptualization, writing–review and editing. I.L. Weissman: Conceptualization, writing–review and editing. A.A. Barkal: Conceptualization, writing–review and editing. R.L. Maute: Conceptualization, data curation, supervision, project administration, writing–review and editing. S.A. Kahn: Conceptualization, data curation, supervision, writing–original draft, project administration, writing–review and editing.