Discovery and characterization of two anti-PD-1 antibodies with a unique binding mechanism to human PD-1

ABSTRACT

Targeting checkpoint inhibitors is an effective therapy for treating cancer, with human programmed cell death protein 1 (hPD-1) being one of the most successful targets for developing antibody-based drugs. In this work, we isolated a panel of anti-PD-1 single-chain variable fragments with different binding and functional profiles from a fully synthetic human phage display library. Conversion of the best clone to hIgG1LALA and hIgG4PE formats, called UDIZ-007 and UDIZ-008, respectively, resulted in antibodies that effectively blocked the PD-1:PD-L1/L2 interaction and were highly selective as they did not cross-react with CD28 receptor family members. Doses of UDIZ-007 or UDIZ-008 at 10 mg/kg every 3 days for a total of six intraperitoneal administrations eradicated MC38-hPD-L1 colon tumors in B-hPD-1 transgenic mice for hPD-1 at day 17, with no relapse until the end of the study at day 56. Importantly, these antibodies bind hPD-1 in a unique region compared to the anti-PD-1 antibodies of known structure, which might have an impact on novel oncology indications when used as a standalone therapy or in combination with currently approved anti-PD-1 therapeutic antibodies. Therefore, UDIZ-007 and UDIZ-008 seem to be promising candidates for the development of antibody-based drugs targeting checkpoint inhibitors as a treatment for cancer.

Introduction

Programmed death protein 1 (PD-1), also known as CD279, is a co-inhibitory receptor member of the CD28 receptor family.^1,2 It has a 15% identity with the amino acid sequence of CD28, whereas it exhibits 20% and 13% with cytotoxic T-lymphocyte antigen 4 (CTLA-4) and induced T-cell co-stimulator (ICOS), respectively.³ PD-1 is expressed on the surface of T cells, natural killer (NK) cells, and B cells, where it plays a crucial role in immune system downregulation and promoting self-tolerance upon binding to its ligands, PD-L1 (B7-H1) and PD-L2 (B7-DC).^4,5 In several malignancies, such as melanoma, non-small-cell lung cancer, breast cancer, squamous cell carcinoma, colon adenocarcinoma, and breast adenocarcinoma, PD-L1 and/or PD-L2 are upregulated, thereby inhibiting T cell proliferation, activation, cytokine secretion, and cytotoxic T lymphocyte (CTL) killing functions, resulting in tumor immune evasion.^6,7 Thus, therapies focused on blocking the interaction between PD-1 and PD-L1/L2 with monoclonal antibodies hold potential as efficacious treatments for a wide range of cancers.⁸

This potential was first realized in 2014, when the U.S. Food and Drug Administration (FDA) approved two anti-PD-1 therapeutic antibodies, pembrolizumab (Keytruda®) and nivolumab (Opdivo®).^9,10 Studies showing positive results for treatment with Keytruda® and Opdivo® over the past 10 years positioned Keytruda® as the best-selling drug of all medicines in 2024, with Opdivo® in seventh place.¹¹ Despite the proven efficacy and widespread success of these two anti-PD-1 antibodies in cancer therapy, only a subset of patients achieves durable responses.¹² This, combined with incomplete or partial responses in some cases, has underscored the need to improve the efficacy of anti-PD-1-based treatments.

In fact, new anti-PD-1 therapeutic antibodies, both as monotherapies and in combination with other treatments, have continuously expanded the efficacy of anti-PD-1 therapies to a broader range of oncology indications.¹³ As a result, additional approval of 12 anti-PD-1 therapeutic antibodies followed Keytruda® and Opdivo®, including cemiplimab (Libtayo®), dostarlimumab (Jemperli®), and retifanlimab (Zynyz®) approved by the FDA and the European Medicines Agency (EMA); toripalimab (Loqtorzi®), camrelizumab (AiRuiKa®), sintilimab (Tyvyt®), tislelizumab (Tevimbra®), penpulimab (Anniko®), serplulimab (Hetronifly®), pucotenlimab (Puyouheng™), and enlonstobart (Enshuxing®) approved by The National Medical Products Administration of China (NMPA); and prolgolimab (Forteca®) approved by the Russian Ministry of Health (Minzdrav).^14,15

Here, we describe the discovery and characterization of two anti-PD-1 antibodies, UDIZ-007 and UDIZ-008. These antibodies share the same variable (V) regions, but were engineered with two distinct isotypes, hIgG1LALA and hIgG4PE, respectively. The hIgG4PE and hIgG1LALA isotypes reduce binding to the Fcγ receptors (FcγRI, FcγRII, and FcγRIII) as well as to complement component C1q, leading to attenuated effector functions and, hence, avoid depletion of the T-cells.¹⁶ UDIZ-007 and UDIZ-008 demonstrated differential physicochemical profiles but similar biological properties, such as effectively blocking the interaction hPD-1:hPD-L1/L2 in vitro, did not cross-react with CD28 receptor family members, and cross-reacted with human and cynomolgus monkey PD-1. Further, both antibodies showed antitumor efficacy in C57BL/6-Pdcd1^tm1(PDCD1)/Bcgen mice inoculated with colon carcinoma, resulting in complete tumor eradication by day 17 after the first intraperitoneal administration, with no relapse up to the completion of the study at day 56. Importantly, the epitope on hPD-1 recognized by UDIZ-007, and by extension, UDIZ-008, were shown to be unique when compared to the epitopes of anti-PD-1 antibodies with known structure in complex with hPD-1. The implications of these findings are discussed.

Materials and methods

Antibody discovery

The variable (V) regions of UDIZ-007 and UDIZ-008 were isolated from a synthetic human phage display library constructed with the IGHV3-23 01 and IGKV1-39 01 gene segments, which are widely used in natural and therapeutic antibodies.^17,18 Three rounds of selection were performed using recombinant biotinylated hPD-1 protein (Acro Biosystems Cat. No. PD1-H82E4; Newark, DE, USA) at decreasing concentrations of 100, 10, and 1 nM as the selection progressed.

Screening for functional clones

Precultures individually picked out colonies from agar plates after the third round of panning were inoculated in Nunc^TM DeepWell plates (Thermo Fisher Scientific, Cat. No. 278743; Waltham, MA, USA) containing 2xYT medium supplemented with 1% glucose and 100 µg/mL carbenicillin per well. The plates were incubated overnight at 37°C with shaking at 225 rpm. The next day, the overnight precultures were transferred to wells containing 2xYT medium with 0.1% glucose and 100 µg/mL carbenicillin. Single-chain variable fragment (scFv) expression was induced with 1 mM IPTG with overnight incubation at 30°C and shaking at 225 rpm. The IPTG-induced culture supernatants were screened in three primary assays: (1) scFv expression level, (2) binding to hPD-1, and (3) blocking of the hPD-1:hPD-L1 interaction.

scFv expression level

The expression of the scFvs was evaluated by ELISA in Nunc Maxisorp 96-well plates (Thermo Fisher Scientific^TM Cat. No. 456537) coated with 1 µg/mL Protein L (ProSpec Cat. No. pro 1790-b; Rehovot, Israel) at 4°C overnight. The wells were blocked with 3% skim milk in phosphate-buffered saline (PBS)-Tween-20 0.1% (MPBST). Subsequently, 1:2 diluted IPTG-induced culture supernatants were added in each well and incubated for 1 h at room temperature (RT). Plates were washed with PBS-Tween-20 0.1% (PBST), and Protein A-HRP (Invitrogen^TM Cat. No. 101023; Carlsbad, CA, USA) was added at a 1:8,000 dilution and incubated at RT for 1 h. Finally, the plates were washed with PBST, and detection of the bound scFvs was performed by adding 3,3,’5,5’-tetramethylbenzidine (TMB) (BD OptEIA^TM, BD Biosciences Cat. No. 555214; San Diego, CA, USA). The reaction was stopped with Stop solution (Abcam Cat. No. ab171529; Cambridge, MA, USA), and absorbance was measured at 450 nm with a correction at 570 nm using a SpectraMax M3 microplate reader (Molecular Devices, LLC; San Jose, CA, USA).

Binding to hPD-1 by Intellicyt®

Binding to hPD-1 was evaluated using the Intellicyt® iQue3 system (Sartorius; Göttingen, Germany) as follows. Biotinylated hPD-1 (Acro Biosystems Cat. No. PD-1-H82E4) (40 ng/mL) was incubated with streptavidin-coated QBeads® (iQue QBeads® DevScreen, Sartorius) for 1 h at RT. The IPTG-induced culture supernatants were mixed at a 1:1 ratio with the biotinylated hPD-1–QBead complex, resulting in a twofold dilution. Protein L-PE (Sino Biological Cat. No. 11044-H07E-P; Beijing, China) was used to detect binding to hPD-1, and the medium fluorescence intensity (MFI) was measured using the Intellicyt® iQue3 flow cytometer.

Blocking of hPD-1:hPD-L1 and hPD-1:hPD-L2 interactions

Like the binding assay, the blockade of hPD-1 interaction with hPD-L1/hPD-L2 was evaluated using the Intellicyt® iQue3 System. Biotinylated hPD-L1 (Acro Biosystems Cat. No. PD-1-H82E5) or Biotinylated hPD-L2 (Acro Biosystems Cat. No. PD2-H82E8) at 2 µg/mL were incubated with SAv QBeads Streptavidin, and free hPD-L1/L2 was removed by washing the beads with PBS-bovine serum albumin (BSA) 1%. The IPTG-induced supernatants (or purified antibodies, see below) were mixed at a 1:1 dilution with biotinylated hPD-1. The SAV-biotinylated hPD-L1 or SAV-biotinylated hPD-L2 complex was added and incubated for 1 h at RT. The plates were washed and Streptavidin-PE (BD Pharmingen™, BD Biosciences Cat. No. 554061) was used to detect the bound biotinylated hPD-1, the samples were analyzed using Intellicyt® iQue3 flow cytometer software.

Conversion to IgG and expression in Chinese hamster ovary cells

V_H and V_L regions from selected scFvs were amplified by PCR and cloned into the TGEX mammalian expression vector (Antibody Design Labs; San Diego, CA, USA) containing the heavy chain constant regions of hIgG4PE (UDIZ-008), hIgG1LALA (UDIZ-007), or hIgG1 (D63E) and the kappa light chain constant region. Isotype controls were prepared with the V regions of the anti-hen egg white lysozyme (HEL) antibody (D1.3)¹⁹ and cloned into the vectors containing hIgG1LALA (D92C) and hIgG4PE (D55C). Plasmid DNA corresponding to the heavy and light chains of the anti-PD-1 and control antibodies were purified and co-transfected in Chinese hamster ovary (CHO) cells. Cultures were incubated in a 5% CO₂ environment at 37°C for 11 d.

IgG purification by Protein-A

Antibodies from the CHO cultures were purified through affinity chromatography using Protein A (HiTrap Protein A™ HP 5 mL Sure™; Cytiva, Sigma Aldrich Cat. No. 17–0403-01; Darmstadt, Germany). The column was equilibrated with five column volumes (CV) of the equilibrium solution (2.5 mM Na2HPO4, 25 mM NaH2PO4, 150 mM NaCl, at pH 7.4) until reaching the baseline. Antibody capture was performed at a flow rate of 2.5 mL/min, followed by column washing with 5 CV of the capture buffer or until reaching baseline. Elution was achieved with 0.1 M acetic acid at pH 2.8. The eluate was neutralized with 1M Tris-HCl pH 9.0, and the buffer was exchanged for PBS (pH 7.4).

Integrity by SDS-PAGE

The integrity of UDIZ-007 and UDIZ-008 was assessed using polyacrylamide gel electrophoresis (SDS-PAGE) under non-reducing and reducing conditions. The Mini-PROTEAN® TGX Stain-Free™ Protein Gels (Bio-Rad, Cat. No. 4568125; Hercules, CA, USA) and the Mini-PROTEAN Tetra System electrophoresis chamber were used for separation. Gel images were captured using a ChemiDoc^TM MP imaging system (Bio-Rad).

Purity by size exclusion chromatography

The purity of UDIZ-007 and UDIZ-008 was evaluated in an Acquity UPLC BEH 200 size exclusion chromatography 1.7 µm column (4.6 × 150 mm) (Waters® Cat. No.186005225; Milford, MA, USA) at 30°C. Separation was carried outh with a 50 mM phosphate buffer solution containing 150 mM NaCl at pH 6.81 under isocratic flow conditions of 0.4 mL/min over 12 min.

Intact mass by mass spectrometry

The intact mass of UDIZ-007 and UDIZ-008 at a concentration of 1 mg/mL were analyzed using a Q-TOF mass spectrometer with electrospray ionization (Vion® Waters®), coupled with an Acquity class-I ultra-high-performance liquid (UPLC) chromatography (Waters®). A gradient of 0.1% v/v formic acid in water (line A) and 0.1% v/v formic acid in acetonitrile (line B) was applied, starting from 99% line A to 0% line A over 10 min. The stationary phase consisted of an Acquity UPLC BEH C4 column (1.7 μm, 300 Å, 2.1 mm × 100 mm) at 80°C. The samples were filtered through a 0.22 µm filter and placed in a glass vial for UPLC-MS injection. Five microliters of each sample were injected for acquisition, and the results were analyzed using UNIFI software (Waters®).

Thermal stability by Protein Thermal Shift™

Protein Thermal Shift™ assay²⁰ was performed with 8 µg/well of UDIZ-007 and UDIZ-008 incubated with SYPRO® Orange dye and analyzed over a range of temperature from 25°C to 99°C using the 7500 Fast Real-Time PCR system (Thermo Fisher Scientific). The resulting data were analyzed using the Protein Thermal Shift™ software (Thermo Fisher Scientific) to determine melting temperatures (T_m) and thermal stability profiles.

Dissociation constant (K_D) for hPD-1 and cynomolgus monkey (cPD-1) by surface plasmon resonance

The affinity of UDIZ-007 and UDIZ-008 for hPD-1 and cPD-1 was determined by surface plasmon resonance (SPR) using a Biacore T200 system (Cytiva®). A goat-derived anti-human Fc polyclonal antibody (R&D Systems, Cat. No. G-102-C; Minneapolis, MN, USA) was immobilized in a CM5 chip (Cytiva® Cat. No. BR100399) at 5 µg/mL using the amine coupling kit at pH 4.5 (Cytiva®) following the manufacturer’s instructions. UDIZ-007 or UDIZ-008, diluted in HBS-EP 1X pH 7.4 (Cytiva®), were flown over the sensitized chip at 25°C at a flow rate of 10 µL/min and a contact time of 30 s. Subsequently, different concentrations of hPD-1 (Acro Biosystems Cat. No. PD1-H522a) or cPD-1 (Acro Biosystems Cat. No. PD1-C52H5) at 50, 25, 12.5, and 6.25 nM, diluted in HBS-EP 1x pH 7.4, were flown at a rate of 10 µL/min with a contact time of 150 s and a dissociation time of 300 s. The resulting sensorgrams were analyzed using the 1:1 binding model in the BIAevaluation software (Cytiva®).

Affinity for Fcγ and human neonatal receptors by SPR

The affinity for Fcγ receptors (FcγRI (CD64), FcγRIIa (CD32a), and FcγRIIIa (CD16a)) and human neonatal receptor (hFcRn) was determined in a Biacore T200 (Cytiva®). A CM5 chip was coated with 5 µg/mL of an anti-His tag antibody using the amine coupling kit at pH 4.5, according to the manufacturer’s instructions. His-tagged FcγRI, FcγRIIa, FcγRIIIa isoform V158, and FcγRIIIa V158F polymorphic variant (R&D Systems, Cat. No.: 1257-FC, 1330-CD, 4325-FC and Acro Biosystems CD8-H52H4, respectively) were immobilized on the sensitized chip in HBS-EP running buffer at pH 7.4 to 50 mM and injected at a flow rate of 30 µL/min for 150 s. Following the immobilization of the receptors, UDIZ-007 or UDIZ-008, diluted in HBS-EP at pH 7.4, were flown over the chip in a concentration range of 62.5 to 1,000 nM, with a 300-s association phase and a 450-s dissociation phase at 30 µL/min. Between antibody concentrations, the chip was regenerated with glycine (pH 1.5) for 30 s at 10 µL/min. Sensorgrams were analyzed using the bivalent binding model in BIAevaluation software (Cytiva®).

The affinity for hFcRn was estimated at 5 µg/µL of hFcRn covalently immobilized onto a CM5 chip using the amine coupling kit at pH 4.5. UDIZ-007, UDIZ-008, and control antibodies were flown over the chip at a range of concentrations of 7.813 to 500 nM, with a contact time of 180 s and dissociation time of 300 s at a flow rate of 30 µL/min in PBS with 0.05% Tween 20 at pH 6.0. The results were analyzed using the bivalent binding model in the BIAevaluation software (Cytiva®).

Binding to hPD-1 and CD28 family receptors by ELISA

The ability of UDIZ-007 and UDIZ-008 to bind hPD-1 and members of the CD28 receptor family (CD28, ICOS, and CTLA-4) was evaluated by ELISA. In brief, Nunc Maxisorp 96-well plates were coated with 25 ng/well of hPD-1 or 50 ng/well of recombinant human CD28 (Sino Biological Cat. No. 11524-HCCH), CTLA-4 (Sino Biological Cat. No. 11159-HNAH) or ICOS (Sino Biological Cat. No. 11559-H08H). Coated wells were washed with PBST and blocked with 3% BSA-PBS. UDIZ-007, UDIZ-008, and control antibodies were added in serial dilutions in a range of 5x10⁻⁵ to 1 µg/mL for hPD-1 or 5x10⁻⁴ to 10 µg/mL for CD28 family receptors and incubated for 1 h at RT. The plates were washed with PBST, and detection was performed by adding 100 µL/well of goat anti-human IgG Fc – HRP conjugate (Abcam, Cat. No. ab97225) diluted 1:15,000 followed by incubation for 1 h at RT. After a final wash, TMB substrate was added. The absorbance was measured at 450 nm with correction at 570 nm using a SpectraMax M3 microplate reader. The data were fitted to a four-parameter dose – response model using GraphPad Prism version 10.2.0.

Binding to hPD-1 on recombinant Jurkat cells

Human T lymphocyte cells stably overexpressing PD-1 (PD-1/NFAT Reporter Jurkat Recombinant Cell, BPS Bioscience Cat. No. 60535; San Diego, CA, USA) were used to assess antibody binding. A total of 5x10⁴ cells were incubated at 4°C for 1 h with UDIZ-007, UDIZ-008 or control antibodies at concentrations ranging from 5x10⁻⁴ to 10 µg/mL. Antibody binding was detected by flow cytometry using goat anti-human IgG Fc secondary antibody conjugated to PE (Invitrogen^TM Cat. No. 12–4998-82) and measured using IntelliCyt® iQue3 system.

hPD-1:hPD-L1 blocking by recombinant CHO-K1 and Jurkat cells

Recombinant CHO-K1 cells constitutively expressing hPD-L1 (PD-L1/TCR Activator CHO Recombinant cell line, BPS Bioscience Cat. No. 60536) (3.5x10⁴ cells/well) were cultivated on Thaw Medium 3A (BPS Bioscience Cat. No. 60186) overnight at 37°C/5% CO₂. In parallel, increasing concentrations of UDIZ-007, UDIZ-008, and control antibodies from 5x10⁻⁴ to 10 µg/mL were pre-incubated for 30 min at 37°C/5% CO₂ with PD-1/NFAT Reporter Jurkat Recombinant Cells (4x10⁴ cell/well). This mix was added (100 µL/well) to the CHO-K1 cells and incubated for 6 h at 37°C/5% CO₂. Finally, One Step^TM Luciferase Assay (BPS Bioscience Cat. No. 60690) was added, and luminescence was recorded using SpectraMax M3 microplate reader.

Efficacy in a hPD-1 transgenic mouse colon carcinoma model

Five groups of six female C57BL/6-Pdcd1^tm1(PDCD1)/Bcgen transgenic mice per group of 7–10 weeks of age were utilized in the study. The study was conducted at Biocytogen Boston (Boston, MA, USA). The general animal care and housing procedures were in accordance with the standards of the Biocytogen Institutional Animal Care and Use Committee. Mice were subcutaneously inoculated with 1x10⁶ MC38-hPD-L1 tumor cells. Tumor growth was monitored for 12 d post-inoculation, and animals with tumor size between 75 and 125 mm³ were enrolled for treatment. Additionally, six animals without tumor implantation served as a control group. The treatments consisted of Keytruda® (Pembrolizumab), UDIZ-007, UDIZ-008, and isotype controls D55C and D92C. All treatments were administered intraperitoneally at a dose of 10 mg/kg on days 0, 3, 7, 10, 14, and 17. The study lasted 56 d, during which the animals were monitored twice weekly. Monitoring included clinical observations, tumor volume measurements, and tumor progression assessment.

Epitope mapping

The epitope of UDIZ-007 on hPD-1 was identified using chemical cross-linking mass spectrometry (XL-MS), conducted at CovalX.²¹ Briefly, UDIZ-007 was mixed at a 2.8:1 molar ratio with hPD-1 (Acro Biosystems Cat. No. PD1-H522a). UDIZ-007:hPD-1 complexes were incubated with deuterated cross-linking reagents and subjected to multienzyme digestion. After enrichment, the mixture was analyzed using high-resolution mass spectrometry (nLC-Q-Exactive^TM).

Results

Selection of anti-hPD-1 scFvs

Specific anti-hPD-1 scFvs were obtained by solution panning of a synthetic human phage display library with recombinant hPD-1 protein. After three rounds of selection, 315 clones were screened by ELISA to assess scFv expression and binding specificity to hPD-1 and BSA as a negative control (Figure 1). Out of these clones, 215 were positive and specific for hPD-1, with 60 resulting in unique sequences. Functional screening of the unique clones identified 39 scFvs capable of blocking the hPD-1:hPD-L1 interaction. The six scFvs with the best hPD-1 binding and hPD-1:hPD-L1 blocking activity were converted to hIgG4PE, expressed in CHO cells, and purified by protein A chromatography for further assessment of hPD-1 binding and hPD-1:hPD-L1 blocking (Table 1). One of the clones, D6E (in bold in Table 1), renamed as UDIZ-008, showed superior binding and blocking signals and was converted to hIgG1LALA and called UDIZ-007.

Figure 1.

Selection strategy and screening process for identifying functional anti-PD-1 antibodies.

Table 1.

Binding and blocking properties of six best scFv converted to hIgG4PE.

Antibody name	Binding to hPD-1 EC50 (µg/mL)	Blocking hPD-1:hPD-L1 IC50 (µg/mL)
D3E	0.043	0.852
D4E	0.082	0.535
D5E	0.060	0.406
D6E	0.011	0.095
D8E	0.056	1.839
D9E	0.066	0.541
3C5	—	—

Note: The antibody 3C5, an anti-VEGFR3 antibody,²² with an hIgG4PE isotype, expressed and Protein-A purified side-by-side with the anti-PD-1 antibodies, was used as a negative control.

Biophysical characterization

The biophysical profiles of UDIZ-007 and UDIZ-008 were assessed using SDS-PAGE, SEC-UPLC, Intact Mass by MS, and thermal stability profiling using the Protein Thermal Shift^TM assay (Figure 2A-D). SDS-PAGE analysis (Figure 2A) under non-reducing conditions showed bands of 129–135 kDa corresponding to an IgG. Under reducing conditions, two bands were observed, one at 47.6–49.4 kDa corresponding to the heavy chain, and another at 24–26 kDa, corresponding to the light chain, also consistent with the typical pattern of an IgG.

Figure 2.

Physicochemical characterization of UDIZ-007 and UDIZ-008. (A) SDS-PAGE electrophoretic profiles under reducing (R) and non-reducing (NR) conditions. (B) Monomeric content of UDIZ-007 and UDIZ-008. (C) Intact mass profiles of UDIZ-007 and UDIZ-008. All signals were identified with a maximum mass error of 100 ppm. (D) Thermal stability profiles of UDIZ-007 and UDIZ-008.

The mass distribution determined by SEC-UPLC showed main peaks for UDIZ-007 and UDIZ-008 of 171.91 kDa and 167.59 kDa, respectively, in agreement with the expected molecular weight of an IgG. The monomeric content was estimated to be 99.62% for UDIZ-007 and 98.55% for UDIZ-008 (Figure 2B).

Intact mass analysis of UDIZ-007 and UDIZ-008 (Figure 2C) revealed a predominant peak corresponding to the G0F/G0F glycoform, with observed masses of 146,849 Da and 146,741 Da, respectively. These main peaks differed from the calculated mass by −6.05 Da for UDIZ-007 and −4.66 Da for UDIZ-008. In addition to this principal species, both antibodies exhibited glycoforms commonly found in FDA-approved therapeutic monoclonal antibodies.²³ These included the glycoform combinations Man5/G0F (observed at 146,641 Da in UDIZ-007 and 146,534 Da in UDIZ-008), G1F/G1F (147,172 Da and 147,062 Da, respectively), and G0F/G1F (147,025 Da and 146,907 Da, respectively).

Thermal transitions of UDIZ-007 and UDIZ-008 are presented in Figure 2D. The first thermal transition (Tm1) corresponded to the C_H2 domain, which is the least thermally stable region of IgGs,²⁴ while the second transition (Tm2) is attributed to the Fab region. UDIZ-007 Tm1 was 69.32°C, higher by 6.0°C than that of UDIZ-008, consistent with the fact that hIgG1 is more stable than hIgG4. Tm2 for UDIZ-008 was 73.38°C, slightly lower by 2.0°C than UDIZ-007, probably due to the destabilizing effect of the hIgG4 isotype.

UDIZ-007 and UDIZ-008 affinity for hPD-1 and cPD-1

The K_Ds of UDIZ-007 and UDIZ-008 for hPD-1 and cPD-1 were determined by SPR (Table 2). For hPD-1, UDIZ-007 exhibited a K_D of 3.5 nM (Figure 3A-B), while UDIZ-008 showed a slightly higher affinity with a K_D of 2.8 nM. Both dissociation constants were similar to the K_D of Keytruda® (2.3 nM), determined under the same conditions, and very close to the K_D reported in the literature²⁵ for Keytruda® (2.5 nM).

Table 2.

Affinity of anti-PD-1 antibodies to hPD-1 and cPD-1 measured by SPR.

Antibody name	Ligand	ka (1/Ms)	kd (1/s)	KD (nM)
UDIZ-007	hPD-1	6.0 x 105	2.1 x 10−3	3.5
UDIZ-008		8.6 x 105	2.4 x 10−3	2.8
Keytruda®		2.5 x 105	3.0 x 10−3	2.3
UDIZ-007	cPD-1	1.7 x 105	3.6 x 10−3	21.5
UDIZ-008		1.3 x 105	5.0 x 10−3	38.5

Figure 3.

Binding kinetics of UDIZ-007 and UDIZ-008 to hPD-1 (A–B) and cPD-1 (C–D) Sensorgrams showing concentration-dependent binding of UDIZ-007 (A) and UDIZ-008 (B) to immobilized hPD-1. (C–D) Binding profiles of UDIZ-007 (C) and UDIZ-008 (D) to immobilized cPD-1.

For cPD-1, the K_D values were 21.5 nM for UDIZ-007 and 38.5 nM for UDIZ-008 (Figure 3C-D). Both dissociation constants were lower than the K_D of Keytruda® reported in the literature²⁶ (3.6 nM). However, the K_D difference of UDIZ-007 between hPD-1 and cPD-1 was less than one order of magnitude (~6x), giving us confidence that the eventual toxicological studies of UDIZ-007 in non-human primates (NHP) could be translated to humans.

Binding to hPD-1 on recombinant Jurkat cells

The capacity of UDIZ-007 and UDIZ-008 to bind hPD-1 on Jurkat cells surface (Figure 4) was performed by flow cytometry using Jurkat cells stably transfected to overexpress hPD-1. Both UDIZ-007 and UDIZ-008 bound hPD-1 with similar EC₅₀ values, only departing 3x from Keytruda®.

Figure 4.

Binding of UDIZ-007 and UDIZ-008 to hPD-1 in Jurkat cells. Keytruda® was used as a reference. EC₅₀s are shown for UDIZ-007 and UDIZ-008 in comparison to Keytruda®. The negative controls D92C and D55C, which are not related to hPD-1, correspond to the hIgG1LALA and hIgG4PE isotypes, respectively.

UDIZ-007 and UDIZ-008 selectivity for hPD-1

UDIZ-007 and UDIZ-008 did not bind other members of the CD28 family, including CTLA-4, ICOS, and CD28 (Figure 5). Since these members of the CD28 family are involved in CD28 and ICOS T cell activity enhancement, while BTLA, CTLA-4, and PD-1 function as suppressors,²⁷ binding of UDIZ-007 and UDIZ-008 to hPD-1, but not to members of the CD28 family, is expected to minimize the risk of adverse effects due to binding to other molecules involved in the T-cell immune regulation.

Figure 5.

Binding of UDIZ-007 and UDIZ-008 to CD28 family molecules. Specific antibodies against each protein. (A) CTLA-4, (B) ICOS, and (C) CD28 were used as positive controls.

Blockade of hPD-1 interaction with hPD-L1 and hPD-L2

The functionality of UDIZ-007 and UDIZ-008 in blocking hPD-1:hPD-L1 and hPD-1:hPD-L2 interactions was assessed by flow cytometry (Figure 6). Both antibodies effectively inhibited the binding of hPD-1:hPD-L1 and hPD-L2.

Figure 6.

Flow cytometry of blockade of hPD-L1/L2 by anti-PD-1 antibodies. (A) Blocking of hPD-1 binding to hPD-L1 ligand. (B) Blocking of hPD-1 binding to hPD-L2 ligand. Keytruda® was used as the reference. The antibody 3C5, an anti-VEGFR3 antibody,²² was used as the negative control.

hPD-1: hPD-L1 functional blockade in the cellular context

The ability of UDIZ-007 and UDIZ-008 to functionally block hPD-1:hPD-L1-mediated immunoinhibitory activity was evaluated in a cell-based co-culture assay (Figure 7). Both antibodies blocked the interaction between hPD-1 and hPD-L1 with EC₅₀ values of 0.1229 and 0.1372 µg/mL, respectively.

Figure 7.

In vitro hPD-1:hPD-L1 functional blocking assay performed with UDIZ-007 and UDIZ-008. Commercial Keytruda® and Opdivo® antibodies were included as reference controls. D92C (hIgG1LALA) and D55C (hIgG4PE) antibodies were used as negative isotype controls.

Binding to Fcγ receptors

Having characterized the biophysical and in vitro functionality of UDIZ-007 and UDIZ-008, their potential toxicity due to binding to the Fcγ receptors (FcγRs) and thus triggering antibody-dependent cellular cytotoxicity (ADCC) and antibody-dependent cellular phagocytosis (ADCP)²⁸ was assessed by SPR. MabThera® (Rituximab), an IgG1 antibody that binds FcγRs, was used as a positive control. Neither UDIZ-007 nor UDIZ-008 bound the FcγRs.

Table 3 summarizes the ka, kd, and K_D values of UDIZ-007, UDIZ-008, and the control antibody D63E (same V regions as UDIZ-007 and UDIZ-008 but with a hIgG1 isotype) to the FcγRs. In contrast to D63E, UDIZ-007 and UDIZ-008 did not bind to FcɣRI, FcγRIIa, and FcγRIIIa. Human FcγRIIIa is expressed on innate immune cells such as macrophages and NK cells and is the most relevant receptor for ADCC of therapeutic antibodies. Further, the two polymorphic forms of the receptor, FcγRIIIa 158 V and FcγRIIIa 158F have been associated with higher cytotoxic activity. Therefore, it is expected that UDIZ-007 and UDIZ-008 have low toxicity due to low or no binding to the FcγRs assayed.

Table 3.

Affinity values of D63E, UDIZ-007, and UDIZ-008 for FcγRI, FcγRIIa, FcγRIIIa 158 V, and FcγRIIIa 158F receptors.

Parameter	FcγRI (CD64)			FcγRIIa (CD32a)			FcγRIIIa 158V & 158F (CD16a)
	D63E	UDIZ-007	UDIZ-008	D63E	UDIZ-007	UDIZ-008	D63E	UDIZ-007	UDIZ-008
ka (1/ms)	2.5 x 104	ND	ND	7.8 x 102	ND	ND	3.0 x 103 (158V)	ND	ND
kd (1/s)	8.6 x 10−4	ND	ND	5.1 x 10−3	ND	ND	7.2 x 10−4 (158V)	ND	ND
KD (M)	3.4 x 10−8	—	—	6.5 x 10−6	—	—	2.4 x 10−7 (158V)	—	—

ND: Not Detectable.

Binding to hFcRn

The binding of UDIZ-008 and UDIZ-007 to the human neonatal receptor (FcRn) at pH 6.0 was also assessed by SPR (Figure 8A-B). The K_D of UDIZ-007 was 665 nM, which is in the same order of magnitude as other IgG1s, such as adalimumab (672 nM) and infliximab (727 nM).²⁹

Figure 8.

Evaluation of binding to FcRn. The affinity of UDIZ-007 (A) and UDIZ-008 (B) for the hFcRn receptor was assessed by SPR.

Efficacy in a colon carcinoma model

Having characterized the developability, in vitro functionality, and potential toxicity of UDIZ-007 and UDIZ-008, their efficacy was evaluated using the MC38-hPD-L1 colon cancer model in B-hPD-1 transgenic mice (C57BL/6-Pdcd1^{tm1(PDCD1)/Bcgen}) according to the scheme shown in Figure 9A. The results showed that UDIZ-007 and UDIZ-008 (and Keytruda®) did not significantly modify the body weight gain during the study compared to the isotype controls (D92C and D55C) (Figure 9B), suggesting that the antibodies were well tolerated.

Figure 9.

Evaluation of the efficacy of UDIZ-007 and UDIZ-008 in an in vivo model of colon carcinoma. C57BL/6-Pdcd1^tm1(PDCD1)/Bcgen mice were implanted with MC38-hPD-L1 tumor cells subcutaneously. They were then administered 10 mg/kg of each antibody. The mice were monitored over a 56-day period, during which time the following were evaluated. (A) Experimental design; (B) Body weight. (C) Individual mouse tumor growth treated with UDIZ-007, isotype control D92C, and Keytruda® as a reference; (D) Individual mouse tumor growth treated with UDIZ-008, isotype control D55C, and Keytruda® as a reference. (E) Percentage tumor growth inhibition ± SEM.

The isotype control groups showed an exponential growth of the tumors, in sharp contrast with UDIZ-007, UDIZ-008, and Keytruda® groups, where the tumors started to shrink after the first treatment with the antibodies (Figure 9C-D) and disappeared around day 17. The potency in inhibiting tumor growth (IC₅₀ expressed in days) was estimated by non-linear regression analysis in 8.12, 7.72, and 5.4 d for UDIZ-008, UDIZ-007, and Keytruda®, respectively (Figure 9E). No tumor relapse was observed after day 17 up to the end of the experimental period (56 d).

UDIZ-007 epitope

The UDIZ-007 epitope on hPD-1 was determined by XL-MS. Since UDIZ-007 has the same V regions as UDIZ-008, it was assumed that both antibodies bind the same epitope. The epitope was located at residues 80–115 (Figure 10) and encompassed the C’D loop, and β-strands D and E. The N-terminal region of hPD-1 (residues 80–104) was involved in the interaction with the three HCDRs, whereas the C-terminal region (residues 105–115) was bound by HCDR2, LCDR1, and LCDR2. LCDR3 was not identified as in contact with hPD-1.

Figure 10.

Epitope mapping of UDIZ-007 on human PD-1. The mapped epitope is delineated by amino acid residues within the extracellular domain of hPD-1, specifically in the region encompassing residues 80 to 115.

Interestingly, the interaction of UDIZ-007 with hPD-1 is mostly located on the opposite side of the PD-L1 binding site (Figure 11A-B) hPD-1 binds hPD-L1 with the front β-sheet face built by CC’FG strands. UDIZ-007 epitope did not overlap with the residues involved in the hPD-1:hPD-L1 interaction (Figure 11C). This suggests that UDIZ-007 (and by extension UDIZ-008) could be an indirect blocking mechanism where the size and shape of the antibody create a physical barrier that prevents hPD-L1 from accessing its binding site on hPD-1, rather than binding directly to the residues on hPD-1 involved in the interaction with hPD-L1, impairing its function via direct competition.

Figure 11.

UDIZ-007 epitope mapped onto the hPD-1 structure (PDB id: 4ZQK). (A) Ribbon representation of hPD-1 showing the regions that UDIZ-007 binds to in hPD-1 (magenta and light blue, see Figure 10). (B) Solvent accessible surface view of hPD-1 in the same orientation and color-code as in panel A. (C) Solvent accessible surface view of hPD-1 showing the hPD-L1 binding site (green) and UDIZ-007 epitope. As can be seen, although the PD-L1 binding site and UDIZ-007 epitope are adjacent, there is no overlap between hPD-L1 binding site and UDIZ-007 epitope. The hPD-L1 binding site was determined with PDBePISA (https://www.ebi.ac.uk/pdbe/pisa/) and the PDB structure 4ZQK. Structural images were generated using BIOVIA discovery studio 2020.

Discussion

In the previous sections, we reported the isolation and characterization of two fully human anti-PD-1 antibodies, UDIZ-007 and UDIZ-008. These antibodies have the same V regions, isolated from a synthetic human scFv phage display library, combined with two distinct isotypes, hIgG4PE (UDIZ-008) and hIgG1LALA (UDIZ-007). Both isotypes minimize effector functions by abrogating binding to the FcγRs as shown in Table 3. IgG4 has been the predominant isotype in anti-PD-1 therapeutic antibodies, with 13 of 15 approved molecules being IgG4P. The mutation S228P in the hinge region of IgG4 abolishes the heterogeneity of human IgG4 antibodies.^30,31 We added the mutation L235E in the C_H2 domain to further reduce the effector functions of the IgG4.³² The isotype IgG1LALA (L234A/L235A), on the other hand, has been the isotype of choice in two approved anti-PD-1 therapeutic antibodies, prolgolimab and penpulimab, and has frequently been used in other therapeutic antibodies.^14,15

The physicochemical profile by SDS-PAGE, SEC-UPLC, MS, and thermal stability resulted in similar UDIZ-007 and UDIZ-008 profiles except for the thermal stability. The first thermal transition of UDIZ-007 occurred at 69.3°C, which is h 6.0°C higher than that of UDIZ-008, which occurred at 63.2°C. The second transition in UDIZ-007 was 76.0°C whereas UDIZ-008 had 73.2°C. The thermal transitions of Keytruda® measured side-by-side with UDIZ-007 and UDIZ-008 gave a Tm1 of 67.0°C and a Tm2 of 69.6°C. The higher stability of UDIZ-007 with respect to UDIZ-008 (and Keytruda®) is consistent with the superior stability of the hIgG1 isotype when compared to hIgG4,³³ which may translate into enhanced manufacturability of UDIZ-007.

From a functional point of view, the binding affinity of UDIZ-007 and UDIZ-008 for hPD-1 as determined by SPR was in the low nanomolar range. Both antibodies blocked the interaction with PD-L1 and PD-L2 and were selective for hPD-1, as they do not bind to related members of the CD28 receptor family members. Importantly, both antibodies eradicated tumors in B-hPD-1 mice at 10 mg/kg at around day 17, with no relapse. Additional efficacy studies in the same animal model (data not shown) suggest that lower doses of 5.0 and 2.5 mg/kg have a similar effect to 10 mg/kg, indicating that the latter could be a good starting point for designing therapeutic doses of UDIZ-007/008 in clinical studies.

Remarkably, the epitope of UDIZ-007 (and by extension UDIZ-008) does not overlap with the residues involved in the hPD-1:hPD-L1 interaction. hPD-1 binds hPD-L1 with the front β-sheet face built by CC’FG strands (Figure 11C) whereas hPD-1 binds UDIZ-007 with the C’D loop, and β-strands D and F. These β-strands are on the opposite side of hPD-L1 binding site, suggesting that UDIZ-007 MOA is not by directly outcompeting hPD-L1 interaction with hPD-1. The C’D loop, on the other hand, is highly flexible,^34,35 and in fact is unstructured and missed in some experimental structures.^36–38 Also, it has different conformations free or in complex with diverse ligands, and important differences have been found at the PD-L1 binding site due to single amino acid substitutions in hPD-1, leading to notable differences between the human and the mouse ortholog in the C’D loop.³⁴ Thus, an alternative (or complementary) explanation for the UDIZ-007/008 MOA, could be that the binding of UDIZ-007 to the C’D loop of hPD-1 stabilizes a conformation that prevents the interaction with hPD-L1.

A comparison with the epitopes recognized by anti-PD-1 antibodies of known structure in complex with hPD-1 (Figure 12) shows that the UDIZ-007 epitope is unique. Overall, three regions can be identified in the epitopes of anti-PD-1 antibodies: (1) the N-loop and β-strands A’ and A (residues 25–43); (2) the C’D loop and C and C’ β-strands (residues 59–100); and (3) the FG loop (residues 121–144). Of the 16 different antibodies from 21 crystallographic structural data, four (Nivolumab, GY-5, D12, and Toripalimab) bind regions 1, 2 and 3, one antibody (NBO1a) binds regions 1 and 3, and most of the antibodies (11 structures) bind regions 2 and 3 only. The UDIZ-007 epitope partially overlaps with region 2 and binds a region of hPD-1 mainly located in the β-strand between regions 2 and 3. None of the antibodies in Figure 12 has epitopes that map onto this region of hPD-1.

Figure 12.

The hPD-1 epitopes recognized by 16 anti-PD-1 antibodies with known structures in comparison with the UDIZ-007 epitope. Only those residues (25–147) of hPD-1 where at least one anti-PD-1 antibody binds hPD-1 are shown. On top, the residues identified as antagonist (gray) or agonist (red) epitopes as reported at Kensuke Suzuki et al.³⁹ blue arrows indicate β-strands and squares the loops connecting them in the hPD-1 secondary structure. The residues in contact with PD-L1 are highlighted in green. The epitope of UDIZ-007 is indicated with the same color code as in Figure 11. Using the crystallographic structural data from 21 structures, the hPD-1 residues in the interface zone with each antibody were determined by PDBePISA server (http://www.Ebi.ac.uk/pdbe/prot_int/pistart.Html) and are highlighted in yellow. Blue boxes (1–3) enclose the epitope regions recognized by these antibodies. Three structures of pembrolizumab, two of nivolumab, two of cemiplimab, and two of tislelizumab solved under different crystallization conditions are included to show slight differences in the epitope recognized by the same antibody. PDB IDs are included in parentheses next to each antibody name.

The epitope specificity plays an important role in defining the anti-PD-1 antibody functions. For instance, the evaluation of antibodies targeting hPD-1 in a functional in vitro assay revealed that pembrolizumab is a slightly more effective PD-1 blocker than nivolumab.⁴⁰ As shown in Figure 12, pembrolizumab and nivolumab bind different hPD-1 epitopes. Further, a recently systematic analysis³⁹ of a panel of anti-PD-1 antibodies binding hybrids of hPD-1/mPD-1 indicated that the membrane-proximal extracellular region of PD-1 has agonist activity, in contrast to the binding of the membrane-distal region by antagonistic antibodies (see the top panel of Figure 12). Therefore, it is reasonable to expect that the UDIZ-007 unique epitope could lead to differences in its efficacy compared to the currently approved anti-PD-1 antibodies. For example, UDIZ-007 could be efficacious in different oncology indications or demonstrate improved efficacy where approved antibodies have shown limited application.

In summary, we have isolated and characterized two fully human anti-PD-1 antibodies with a unique interaction with hPD-1. Both antibodies effectively eradicated tumors at a dose of 10 mg/kg in a transgenic mouse model expressing hPD-1. Although the fine details of UDIZ-007/008 binding to hPD-1 must be elucidated by higher-resolution techniques such as X-ray crystallography or cryo-electron microscopy to identify the specific residues of hPD-1 in contact with UDIZ-007/008, determine the orientation of the antibody with respect to hPD-1, and assess how the possible steric hindrance of the antibody prevents the interaction between hPD-1 and hPD-L1 (and/or visualize conformational changes in the C’D loop of hPD-1 upon binding of UDIZ-007), the atypical UDIZ-007/008 epitope defines a unique MOA. Therefore, UDIZ-007/008 alone, in combination with other approved anti-PD-1 antibodies, or as part of multispecific therapeutic formats linking UDIZ-007/008 with antibodies binding non-overlapping epitopes to enhance the specificity of current anti-PD-1 therapeutic antibodies, could be a new valuable therapeutic option to treat cancer.

Funding Statement

The work was supported by the Laboratorios Columbia S.A. de C.V.

Disclosure statement

The anti-PD-1 antibodies reported in this publication were discovered and are in development with funds from GlobalBio, Inc., Unidad de Desarrollo e Investigación en Bioterapéuticos (UDIBI) and Laboratorios Columbia SA de CV. Provisional patents claiming this invention have been filed and some of the authors have commercial interest in the products related to this research.