We affirm that all research described herein complies with all relevant human and animal ethical regulations.
Directed evolution of human CD80
Yeast surface display libraries containing the randomly mutated IgV domain of human CD80 (residues 35-141) were generated by error-prone PCR as described21. Multiple selection strategies were employed to generate a diverse set of candidates. All intermediate selection outputs were assayed for CD28, CTLA-4, and PD-L1 binding by flow cytometry. The pool of mutants from which the lead candidate domain (H2393) was isolated and evolved as follows. A randomly mutated library (2.7 × 108 clones, avg 4.6 mutations/clone) was screened with 2 parallel flow sorting progressions: (A) 3 nM CTLA-4-Fc+ (produced at Alpine Immune Sciences) followed by two successive 300 nM CD28-Fc− (R&D Systems) sorts and (B) two successive 300 nM CD28-Fc− sorts followed by a 3 nM CTLA-4-Fc+ sort. Yeast plasmid DNA isolated from the A and B terminal sorts was pooled and used as template to build a subsequent error-prone PCR library. This second-round library was screened via the following sort progression: 50 nM PD-L1-Fc+ (R&D Systems) followed by 20 mM CTLA-4-Fc+, followed by 10 nM PD-L1-Fc+ followed by 25 nM CD28-Fc. Each sort step was followed by outgrowth in YPD media (Himedia) at 30 °C and expression induction in media derived from Benatuil et al. (2010), with the substitution of dextrose/galactose with dextrose only at 22 °C before progressing to the next sort.
Recombinant protein production and purification
CD80 vIgDs were expressed as soluble Fc-fusion proteins as described21. Recombinant vIgD Fc-fusion proteins and native IgSF receptor ECD Fc-fusion proteins were produced via transient expression in Expi293FTM cells (Invitrogen Thermo Fisher Scientific) following the manufacturer’s protocol. Protein was purified from conditioned media harvests by capture and elution from Protein A (MabSelect SuRe, GE Healthcare) followed by a Size Exclusion Chromatography (SEC) polishing step (HiLoad 16/600 Superdex 200 pg, 1.6 × 60 cm, GE Healthcare) to generate monomeric, highly purified material. Protein concentrations were based on 280 nm absorbance using extinction coefficients calculated from the predicted primary structure49. Purified proteins were formulated in 10 mM acetate, 9% sucrose, pH 5.0, and stored at −80 °C. A sample vial was thawed and assessed by analytical SEC to demonstrate stability after thaw. Analytical SEC was performed on a G3000SW XL column (7.8 mm × 30 cm, 5 µM, Tosoh Biosciences) at a flow rate of 0.5 mL min−1 using an Alliance 2695 (Waters) with 10 mM Acetate, 250 mM NaCl, pH 5 as running buffer. Purified proteins (50 μl) were applied at 0.5 mg ml−1. Additionally, material was tested for endotoxin (Charles River Laboratories LAL endotoxin kit). Endotoxin levels for all proteins were <1 EU/mg.
Crystallization of ALPN-202:PD-L1 complex
Monomeric ALPN-202 vIgD domain and PD-L1 ECD domain were prepared as carboxyl terminal His tagged recombinant proteins expressed in Expi293FTM cells, purified by affinity chromatography on IMAC columns (HisTrap Excel, GE Health Care) and buffer exchanged into 10 mM acetate, 9% sucrose, pH 5.0. The ALPN-202:PD-L1 complex was prepared by mixing ALPN-202-derived CD80 vIgD His and PD-L1 ECD His proteins at a 1.2:1 molar ratio and the CD80 vIgD:PD-L1 complex was isolated by SEC fractionation. The complex was concentrated, flash frozen in liquid nitrogen and stored at −80 °C until ready for crystallization.
Crystals of the CD80 vIgD in complex with PD-L1 were grown using the vapor diffusion method. The complex crystals were grown at 6.7 mg mL−1 from reservoir solution (0.03 M sodium nitrate, 0.03 M sodium phosphate dibasic, 0.03 M ammonium sulfate, 0.1 M sodium HEPES/MOPS (acid), pH 7.5, 20% v/v glycerol, 10% w/v PEG 4000). Crystals were frozen directly from the crystallization drop.
Datasets were collected at 100 K at station I03, Diamond Light Source, Didcot, England (λ = 0.9763 Å) equipped with a Pilatus3 6 M detector. Data were processed using XDS50 and merged using Aimless51. A data set for ALPN-202 CD80 vIgD in complex with PD-L1 was collected to 3.15 Å. Crystals belonged to space group P212121 and had cell dimensions: a = 59.9, b = 122.2, c = 152.7 Å; α, β, and γ = 90°.
The structure was determined by molecular replacement using the Molrep software52. Published CD80 (PDB: 1DR930) and PD-L1 (PDB: 5JDR53) structures were used as initial models and the binding surface and contact residues within 4 Å between the CD80 vIgD and the PD-L1 ECD were determined by using CONTACT of the CCP4 software package54. Two complexes of ALPN-202 and PD-L1 were found in the asymmetric unit. The structure was refined using Refmac555, then in Buster (Global Phasing Ltd.), and model building was carried out in Coot56. In the final refinement, the crystal structure of PD-L1 to 1.95 Å resolution (PDB: 4Z18), was used as a target for geometrical restraints. The structure of the CD80 vIgD in complex with PD-L1 includes amino acids 35–140 of the CD80 vIgD in chain A, 19–227 of PD-L1 in chain B, 35–57, 61–77, 80–122, and 126–140 of the CD80 vIgD in chain C, and 19–227 of PD-L1 in chain D. The side chains of Asn-53 and Asn-89 of CD80 vIgD (chains A and C), and Asn-35, Asn-192 and Asn-200 of PD-L1 were glycosylated. Thr-127 of PD-L1 in chain B was phosphorylated. Ramachandran plot quality was calculated using MolProbity v4.02b-46757. The data and model quality parameters are in Supplementary Table 2. The average area covered by the binding interactions was calculated using AREAIMOL (Supplementary Table 3). AIMLESS, Molrep, Coot, AREAIMOL, and CONTACT are all part of the CCP4 Software Suite v7.0.044 of the CCP4 software package.
Accession number
Structure coordinates of the ALPN-202 CD80 vIgD:PD-L1 ECD complex have been deposited in the worldwide Protein DataBase (wwPDB) with accession number 7TPS.
Modeling of trimeric CD28:ALPN-202:PD-L1 complex
The ALPN-202 CD80 vIgD:PD-L1 ECD structure was co-aligned with both the B7-1:CTLA-4 crystal structure (PDB: 1I8L32) and chain C from the CD28-Fab crystal structure (PDB: 1YJD51) to demonstrate availability of co-binding between the molecules. The ALPN-202 CD80 vIgD:PD-L1 structure was aligned with the PD-1:PD-L1 crystal structure (PDB: 4ZQK58) to demonstrate the blocking potential of ALPN-202 CD80 vIgD to the binding of PD-1 by PD-L1. Structure modeling was performed using The PyMol Molecular Graphics System Version 2.3.3 (Schrödinger, LLC).
Generation of stably transduced cell lines
Mammalian cells expressing cell surface CD28, CTLA-4, and PD-L1 were generated independently by cloning DNA encoding the full-length proteins into the lentiviral vector pRRL downstream of an MND promoter and Kozak sequence located downstream of the cPPT element. All inserts were preceded by native signal peptides to direct cell surface expression. To monitor lentiviral transduction efficiency and selection of stable integrants, protein expression was translationally coupled to EGFP or a puromycin resistance gene via an intervening T2A peptide59. Lentiviral vectors, and helper plasmids pMD2.G and psPAX2 were transfected into Expi293FTM (Thermo Fisher Scientific) cells to generate lentiviral particles. Viral particles were used to infect target cells for stable lentiviral expression60. For some experiments, the pLVX-EF1α-IRES-Puro lentiviral vector, Lenti-X Packaging Single Shots packaging plasmids and Lenti-X 293 X packaging cell line were used to generate viral particles per the manufacturer’s protocol (Clontech, Takara Bio). Viral supernatants were collected 48–72 h after transfection and in some cases concentrated using the Lenti-XTM Concentration kit (Clontech, Takara Bio) per the manufacturer’s protocol. Target cells were seeded in growth media in a 6-well plate (Corning) and 1 mL virus was added with Polybrene at a final concentration of 10 μg mL−1. The plate was spun at 2500 rpm for 30 min and incubated overnight at 37 °C in 5% CO2. Media was changed the next day and protein expression was monitored by flow cytometry.
Cell lines and primary cells
GloResponse™ IL-2-luc2P Jurkat cells (Promega) were cultured in RPMI 1640 supplemented with 10% fetal bovine serum, 1 mM sodium pyruvate and 200 µg mL−1 hygromycin B. aAPCs were K562 cells (ATCC #CCL-243) virally transduced to stably express a transmembrane anti-CD3 single chain Fv (scFv) and/or wild type PD-L1. Expi293™ cells (Thermo Fisher Scientific) were used to produce proteins or transiently express mouse CD28, CTLA-4, and PD-L1. ExpiCHO-S™ cells (Thermo Fisher Scientific) were virally transduced and selected for stable expression of human CD28, PD-L1, or CTLA-4. MC38 (originally from the laboratory of Dr James Allison), B16-F10 (ATCC #CRL-6475), and SCC152 (ATCC #CRL-3240) cells were virally transduced and selected under puromycin (2%) for stable expression of human PD-L1. CD28, PD-L1, or CTLA-4 expression was confirmed by flow cytometry.
Human whole blood as well as cryopreserved peripheral blood mononuclear cells and purified T cells were purchased from Bloodworks Northwest, an AABB and CLIA accredited agency. Monocytes were isolated from LRS chambers (Bloodworks Northwest) using the EasySep™ Human Monocyte Isolation Kit (StemCell Technologies). Blood and blood products (cells) were collected from heathly, consented adults under WIRB Protocol #20151321 (NML DONOR_BLOOD_002).
Binding affinity measurement by surface plasmon resonance (SPR)
Affinity determination was conducted on a Biacore 3000 (GE) optical biosensor equipped with a CM5 sensor chip prepared with goat anti-human IgG capture antibody (Jackson ImmunoResearch Laboratories). ALPN-202 or WT CD80 ECD-Fc was captured to a level of 120 RU and 160 RU, respectively. For kinetic assays, serial dilutions of recombinant monomeric human CD28, CTLA-4, and PD-L1 (Acro Biosystems) were prepared as follows: CD28, 500 nM to 6.17 nM; CTLA-4, 175 nM to 2.16 nM; and PD-L1, 1500 nM to 18.5 nM. Each analyte titration was run in triplicate in two separate experiments and multiple blank (buffer) injections were conducted to assess and subtract system artifacts. Association phases for all analyte concentrations were monitored for 240 s, while the dissociation phases were collected for 600 s, at a flow rate of 30 μL/min.
Data were aligned, double referenced, and fit using Scrubber v2.0® software (BioLogic Software Pty Ltd.). For kinetic data, association and dissociation phases were globally fit to the 1:1 binding model, to determine ka, kd, and Rmax values. KD was determined by the quotient of the dissociation rate to the association rate. For the weak CD28–WT CD80 and PD-L1–WT CD80 interactions, KD was estimated and expressed as a minimal KD by using the theoretical Rmax as a fixed parameter in the global fit.
Binding to human and mouse CTLA-4, CD28, and PD-L1 by flow cytometry
ExpiCHO cells stably expressing human or mouse CTLA-4, CD28 or PD-L1 were cultured in shake flasks containing ExpiCHO™ Expression Medium (Thermo Fisher Scientific) in a 37 °C, 8% CO2 humidified incubator. To measure binding to CD80 counter-structures, CHO cells were plated in a 96-well U-bottom plate and incubated 60 min on ice with serial dilutions of ALPN-202, WT hCD80-Fc, WT mCD80-Fc, or Fc control. Cells were washed and bound drug detected using PE or Alexa Fluor® 647-conjugated polyclonal anti-human Fc (Jackson ImmunoResearch). Cells were analyzed using a Becton Dickinson LSRII flow cytometer and data analyzed using FlowJo v7 software.
Blockade of PD-1–PD-L1, CD28–CD80 and CTLA-4–CD80
CD80-Fc and PD-1-Fc proteins (R&D Systems) were conjugated with Alexa Fluor® 647 (Molecular Probes® Alexa Fluor® Antibody Labeling Kit, Thermo Fisher Scientific) according to manufacturer’s instructions. To measure blockade of PD-1 – PD-L1, K562/PD-L1 cells were plated in a 96-well plate and incubated with serial dilutions of ALPN-202, anti-PD-L1 (atezolizumab), or Fc control. Cells were washed and incubated with PD-1-Alexa Fluor 647. Bound PD-1 was detected using a Becton Dickinson LSRII flow cytometer and data analyzed using FlowJo v7 software. To measure blockade of CD28 – CD80 and CTLA-4 – CD80, CHO cells stably expressing CD28 or CTLA-4 were plated in 96-well plates and incubated with serial dilutions of ALPN-202, anti-CD28 (clone 28.2, BioLegend), anti-CTLA-4 (ipilimumab) or Fc control. Cells were washed and incubated with CD80-Alexa Fluor 647. Bound CD80 was detected using a Becton Dickinson LSR II flow cytometer and data analyzed using FlowJo v7 software.
SHP-2 recruitment assay
40,000 PathHunter® Jurkat PD-1/SHP-2 cells (DiscoverX) were co-cultured with 60,000 K562/PD-L1 cells in the presence of ALPN-202, WT CD80 ECD-Fc, anti-PD-1 (nivolumab), anti-PD-L1 (atezolizumab), or Fc control. Test articles were serially diluted and added to cell plate for final concentrations of 200 nM to 0.003 nM. After 2 h, detection reagent (DiscoverX) was added to each well and luminescence measured.
Artificial APC T cell stimulation assay
K562 cells stably expressing membrane-anchored anti-CD3 single chain Fv (clone OKT3) and/or PD-L1 were plated at 12,500 cells/well in 96-well plate. 1 × 105 primary T cells (Bloodworks Northwest) and serial dilutions of each test article were added to each well. In some cases, titrations of ALPN-202 or monomeric CD80 vIgD were combined with 100 nM blocking anti-PD-L1 or anti-CD28. Plates were incubated 24 h at 37 °C. Secreted IL-2 was measured using a LEGEND MAX™ Human IL-2 ELISA kit (BioLegend).
In separate experiments, 1 × 105 primary T cells were co-cultured with 2 × 104 K562/PD-L1 cells and incubated with a 100 nM–0.02 nM ALPN-202, agonist anti-CD28 (clone 28.2, BioLegend), Fc control, anti-PD-L1 (atezolizumab), anti-CTLA-4 (ipilimumab), or a combination of anti-PD-L1 and anti-CTLA-4. Submaximal TCR stimulation was provided by 12.5, 2.5, 0.5, or 0 nM anti-CD3 (clone OKT3, BioLegend). After 16 h supernatants were collected and analyzed for secreted IL-2. Cells were then treated with brefeldin A/monensin for 6 h and characterized by flow cytometry for intracellular IL-2 and CD25 upregulation. Four donors were tested in two separate experiments with all samples run in triplicate.
Simultaneous CD28 and PD-L1 binding to ALPN-202 by SPR
ALPN-202 was captured to the sensor chip as described previously and monomeric analytes (1500 nM CD28, PD-L1, or CTLA-4) were sequentially co-injected to saturate binding with a 180 s association phase for each injection at a flow rate of 30 μL min−1. Single analyte injections were paired with blank buffer injections as controls. Samples were run in duplicate and sensorgram overlays depicted in Fig. 2e.
Generation of human E6 TCR transgenic T cells
Cryo-preserved purified primary human T cells (Bloodworks Northwest) were thawed and incubated 6 h in media containing 100 IU/mL IL-2 and anti-CD3/anti-CD28 beads (Dynal) at a 2:1 bead to cell ratio. Titered viral stocks were added to attain a multiplicity of infection of 4 and incubated overnight at 37 °C. Viral supernatants were removed and fresh media including 100 IU/mL IL-2 was added. Cultures were incubated 2 days at 37 °C. On day 3, beads were removed by magnetic selection and cells were replated in media containing IL-2 as above. Cultures were incubated for an additional 7 days with media changes every two days including IL-2 as above. T cells were harvested on day 10 and transduction efficiency was evaluated by flow cytometry for the presence of the mouse Cβ constant region included in the E6 TCR.
T cell and M2c macrophage coculture assays
M2c macrophage polarization was performed in 96-well plates by differentiating 50,000 monocytes/well in complete X-Vivo 15 media (Lonza) with 100 ng/mL of M-CSF for 5 days. Differentiated macrophages were polarized to an M2c phenotype with 50 ng/mL of IL-10 for 2 days. Alternatively, differentiated macrophages were polarized to an M1 phenotype with 100 ng/mL of IFNγ and 1 ng/mL of LPS for a positive control. Macrophages were characterized by flow cytometry using 1:100 dilutions of antibodies specific for human CD80, CD86, CD163, CD206, MHC class II, and PD-L1 (Supplementary Fig. 5). (Fig. 4a–d) Polarized macrophages in each well were co-cultured with 5 × 105 autologous CellTrace™ Violet (CTV; Thermo Fisher Scientific)-labeled T cells with different concentrations of test articles and 100 ng/mL anti-CD3 (clone OKT3). Cell culture supernatants were harvested at 24 h for IL-2 quantitation by HTRF (Cisbio). Dilution of CTV was used to measure T cell proliferation at 72 h. Supernatants were also harvested at end of assay to measure IFNγ, IL-2, TNFα, GM-CSF and IL-21 using a multiplex bead-based magnetic MAGPIX™ assay (Luminex). (Fig. 4e) M2c polarized macrophages, E6 TCR+ T cells, and PD-L1+ SCC-152 tumor cells (25,000 each) were co-cultured in triplicate wells for 24 h in the presence of test articles. Supernatants were collected and soluble IFNγ, IL-2, and TNFα were quantitated.
Jurkat IL-2 reporter assay
K562/PD-L1 cells expressing membrane-anchored anti-CD3 scFv (clone OKT3) were plated at 25,000 cells/well in a white, 96-well plate. 1 × 105 GloResponse™ IL-2-luc2P Jurkat cells (Promega), serial dilution of monomeric CD80 vIgD, and 100 nM anti-CD28, anti-PD-L1 or Fc control was added to each well. Plates were incubated 5 h at 37 °C. After incubation, an equal volume of BioGlo™ Assay Reagent (Promega) was added to each well and luminescence measured.
Mouse tumor studies
All mouse studies were performed under Alpine Immune Sciences (AIS)-affiliated Institutional Animal Care and Use Committee (IACUC) protocol number 5305-01. During the development and evaluation of therapeutics in mouse tumor models (including the human PD-L1 transduced MC38, B16-F10, and SCC152 models in this paper), we conducted power calculations to estimate minimum sample sizes that are appropriate for determining statistically significant differences in tumor volumes over time (power = 0.80; alpha = 0.05). The results of those calculations (9–10 animals per group for tumor studies) are used for all studies in which the same model and Fc control treatment are used. For all MC38 and B16-F10 tumor studies, female C57BL/6NJ mice (8–10 weeks old at study start) were purchased from the Jackson Laboratory. For the SCC152/hPD-L1 tumor model using human E6 TCR Tg T cells, female NOD-scid IL2Rgammanull (NSG) mice (aged 7–8 weeks) were purchased from Jackson Laboratory. All mice were housed under specific pathogen-free conditions at 24 ± 2 °C on a 12-h light/12-h dark cycle at the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-approved animal facility affiliated with AIS. All mice had free access to food and water at all times. Tumor progression was monitored twice weekly by caliper measurements and tumor volume was calculated as length × (width2)/2. Though maximum allowable tumor volume was 2000 mm3, mice were euthanized when tumor volumes reached 1500 mm3 or appeared ulcerated. No tumors reached the maximum 2000 mm3. Data were analyzed and presented as mean tumor growth ± SEM. Statistical significance was determined using repeated measures two-way ANOVA.
MC38 and MC38/hPD-L1 tumor studies
1.5 × 106 MC38 cells were implanted subcutaneously (SC) into the flanks of C57BL/6NJ mice. After 7 days (mean tumor volumes were 53–104 mm3), mice were randomly assigned to treatment groups and treatments initiated either the same day or the following day via intraperitoneal (IP) administration. For repeat-dose studies comparing ALPN-202 to WT CD80 ECD-Fc, anti-hPD-L1 (durvalumab; Catalent), anti-mCTLA-4 (clone 9D9; BioXcell), anti-mPD-1 (clone RMP1-14; BioXcell) or anti-mCD28 (clone E18; mouse IgG2b Fc silent™; Absolute Antibody), mice were treated with one or three doses (depending on the study) of 100 μg of ALPN-202, 100 μg of the appropriate antibody (anti-hPD-L1, anti-mCTLA-4, anti-mPD-1, or anti-mCD28 as noted above), 130 μg of WT CD80 ECD-Fc or 75 μg of Fc control every 2 or 3 days as indicated. For the dose-ranging study (Fig. 5f, g, and Extended Data Fig. 6), single IP injections of 20, 100, 500, or 1500 μg ALPN-202 or 75 μg of Fc control were administered on day 8. Serum was collected from subsets of 3 mice per group per time point at various time points after dosing (1–840 h) for pharmacokinetic (PK) analysis. Serum concentrations of ALPN-202 were determined by an ELISA developed at AIS and PK parameters estimated using Phoenix WinNonlin v6.4 software (Certara). In studies that included ex vivo analysis of tumors by flow cytometry (Figs. 5, 6) or RNA-Seq (Fig. 5), mice were randomized to the ex vivo subgroups at the time of staging on day 7. The ex vivo subgroups had similar mean tumor volumes as the mice used to monitor tumor growth. To evaluate the effect of blocking PD-L1 or CD28 antibodies on ALPN-202 responses in vivo, ALPN-202 was co-administered with either anti-hPD-L1 (durvalumab) or an anti-mCD28 blocking antibody (clone E18) administered as 100 μg IP injections on days 8, 11, and 14.
B16-F10 tumor studies
Female C57BL/6NJ mice were implanted SC with 0.5 × 106 B16-F10/hPD-L1 cells into the right flank on Day 0 and sorted into groups of 13. On days 6, 8, and 11, mice received IP injections of 75 μg Fc control, 100 μg ALPN-202, 100 μg anti-mPD-1 (clone RMP1-14; rat IgG2a; BioXcell), or a combination of ALPN-202 and anti-mPD-1.
SCC152 xenograft tumor studies
Female NOD-scid IL2Rgammanull (NSG) mice (aged 7–8 weeks) were SC implanted in the right flank with 4 × 106 SCC152/hPD-L1 cells prepared in Matrigel (Corning). On day 12, when tumors were 80–120 mm3, animals were SC dosed with 10 mg hIgG to reduce FcR-mediated effects of therapeutic treatments. On day 14, animals were randomized into groups of 10 mice each. On day 15, 5 × 106 human E6 TCR+ T cells/mouse were transferred via retro-orbital injection. On day 16, test articles were administered as follows via IP dosing: Fc control (75 μg/dose), every 3 days for 4 doses (Q3Dx4); ALPN-202 (100 μg/dose), Q3Dx4; ALPN-202, every 7 days for 3 doses (Q7Dx3); anti-hPD-L1 (durvalumab) (100 μg/dose), Q7Dx3.
Flow cytometry of MC38 tumors ex vivo
MC38 tumors from 5 mice/group were harvested either 72 h after a single injection (Fig. 5g) or 48 h after the second injection of indicated treatments (Fig. 6c–e). Tumors were enzymatically digested at 37 °C for 42 min using a Mouse Tumor Digestion Kit (Miltenyi Biotec) and a gentleMACS™ Octo Dissociator with Heaters (Miltenyi Biotec) according to manufacturer’s instructions for soft and medium tumors. For each tumor 0.5–1 × 106 live cells were stained with a viability stain (LIVE/DEAD Fixable Aqua, Life Technologies Corp.) according to manufacturer’s instructions, followed by resuspension of the tumor pellets in 5 μg mL−1 mouse Fc block (clone 2.4G2; BD Biosciences) and a 45 min incubation on ice with a cocktail of fluorescently-labeled anti-mouse antibodies against CD45 (1:100 dilution; clone 30-F11; BioLegend), CD3ε (1:50 dilution; clone REA606; Miltenyi Biotec), CD4 (1:100 dilution; clone GK1.4; BioLegend), CD8α (1:100 dilution; clone KT15; Invitrogen), and CD11b (1:100 dilution; clone M1/70; BioLegend). In the study that included a PE-conjugated mouse class I p15E tetramer to identify antigen-specific CD8+ T cells (Fig. 5g), 7.5 μl tetramer (1:13 dilution; MBL International) was pre-incubated with each tumor cell suspension sample for 30 min at 25 °C prior to addition of the flow antibody cocktail. For granzyme B intracellular staining, surface-stained cells were fixed and permeabilized using an intracellular fix/perm buffer set (BioLegend) prior to a 45 min incubation at 25 °C with anti-mouse granzyme B (1:20 dilution; clone QA16A02; BioLegend). Stained samples were washed and collected onto a Becton Dickinson LSRII flow cytometer. Cell subsets were identified and quantified using FlowJo v10 software using the gating scheme shown in Extended Data Fig. 7.
RNA-Seq analysis of MC38 tumors ex vivo
Mice implanted with MC38/hPD-L1 cells were allowed to reach tumor volumes of 60 mm3 (~7 days). Mice were randomized and treated with a single IP injection of control Fc protein (75 µg), ALPN-202 (100 µg), or anti-PD-L1 (durvalumab) (100 µg) on day 8. Four tumors per treatment group were harvested after 72 h and RNA was prepared from tumor sample lysates.
Total RNA (0.5 ng) was added to lysis buffer from the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Takara), and RT-PCR to generate full-length amplified cDNA was performed. Sequencing libraries were constructed using the NexteraXT DNA sample preparation kit (Illumina) to generate Illumina-compatible barcoded libraries. Libraries were pooled and quantified using a Qubit® Fluorometer (Life Technologies). Dual-index, single-read sequencing of pooled libraries was carried out on a HiSeq2500 sequencer (Illumina) with 58-base reads, using HiSeq v4 Cluster and SBS kits (Illumina) with a target depth of 5 million reads/sample. Base calls were processed to FASTQs on BaseSpace (Illumina), and a base call quality-trimming step was applied to remove low-confidence base calls from the ends of reads. The FASTQs were aligned to the mouse reference genome, using STAR v.2.4.2a and gene counts were generated using htseq-count. QC and metrics analysis was performed using the Picard family of tools (v1.134).
Gene Transcripts Per Million (TPM) were calculated using the gene lengths from NCBIM37 as downloaded via BioMart (ensembl.org) for the May 2012 version. The counts for each gene were divided by gene length and multiplied by 1000. Then all values were divided by the number of million reads in that sample.
Evaluation of differential gene expression between ALPN-202 and Fc control treated tumors was performed using the R-based Needle Genomics Expression Atlas (Needle Genomics). Gene signature analysis was performed by identifying the 200 most differentially regulated genes comparing ALPN-202 with Fc-control treated tumors using false discovery rate (FDR) and focusing on genes upregulated >2x, which yielded a list of 124 genes. Signature transcripts from the indicated immune cell populations were identified using the ImmGen Population Comparison application [https://www.immgen.org/ImmGenpubs.html]. Pathway analysis with this gene set was carried out using the online Enrichr Pathway Analysis tool [https://maayanlab.cloud/Enrichr/].
Cytokine release assay
For soluble assay, human PBMC from three separate donors (Bloodworks Northwest) were thawed in X-VIVO 15™ media (Lonza) supplemented with antibiotics (100 U mL−1 penicillin, 100 µg mL−1 streptomycin) and GlutaMAX™ (Thermo Fisher Scientific) and plated at 1.0 × 106 cells/well in quadruplicate in a round-bottom 96-well plate. Test articles were added to appropriate wells incubated for 24 h at 37 °C prior to collection of supernatants. For the solid phase assay format, test proteins were diluted in PBS and aliquoted into 96-well round-bottom plates and incubated overnight at 4 °C. The next day, PBMC were thawed as described above and rested for 3 h at 37 °C at high density. Rested PBMC were added to assay plates and incubated for 24 h at 37 °C. Supernatants were assayed for secreted IFNγ, IL-2, IL-4, IL-6, IL-8, IL-10, and TNFα using a custom MILLIPLEX™ MAP Human Magnetic Bead array kit (Millipore) and following the manufacturer’s instructions. Data were collected on the Millipore MAGPIX™ System.
Statistics and reproducibility
For in vitro assays, mean or median ± SD or SEM was calculated using GraphPad Prism™ v8. Experiments were independently repeated two or three times and representative data shown. Between three and five donors were used for in vitro assays using human cells to account for donor-to-donor variability. No data were excluded from the analyses.
We conducted power calculations to estimate minimum sample sizes appropriate for determining statistically significant differences in tumor volumes over time (power = 0.80; alpha = 0.05). The results of those calculations (9-10 animals per group for tumor studies) are used for all studies in which the same model and Fc control treatment are used. While not officially blinded, tumor volumes were measured by caliper in a predetermined, consistent manner by individuals with no/limited knowledge of the function of the test articles. Tumor growth curves over time were evaluated for significant differences between treatment groups by repeated-measures two-way ANOVA for ‘treatment’ effects. Ex vivo tumor flow cytometry and RNA transcript data were analyzed by one-way ANOVA with Dunnett’s multiple comparisons test versus Fc control or ALPN-202 (Figs. 5–7). Significant differences between groups for all statistical analyses (P < 0.05) were calculated using GraphPad Prism 8. For Enrichr Pathway Analyses, P-values were calculated using Fisher’s exact test. The adjusted P-value was calculated using the Benjamini-Hochberg method for correction of multiple hypotheses testing. The Odds Ratio was computed using a modification to Fisher’s exact test to indicate a risk of deviation from the expected rank. Combined score was calculated by multiplying the ln(P-value) by the inverse of the Odds Ratio.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
