Inhibition of LSD1 with Bomedemstat Sensitizes Small Cell Lung Cancer to Immune Checkpoint Blockade and T-Cell Killing

This article is featured in Highlights of This Issue, p. 4355

After decades with no notable change to the standard-of-care regimen for newly diagnosed extensive-stage small cell lung cancer (ES-SCLC), the IMpower133 (1) and CASPIAN (2) studies both demonstrated improved overall survival with the addition of an anti-PD-L1 antibody to induction platinum doublet chemotherapy. These outcomes led to a new standard of care in ES-SCLC. However, the observed magnitude of survival improvement was modest, at approximately 2 months, and durable responses were rare. By comparison to other tumor types that also harbor a high tumor mutation burden, such as non–small cell lung cancer (NSCLC) or melanoma, these results suggested considerable remaining opportunity to improve the efficacy of immune checkpoint blockade (ICB) in SCLC. It is important to acknowledge, however, that a small subset of patients with SCLC do seem to have durable responses and are therefore likely to strongly benefit from PD-1/PD-L1 axis inhibition (1, 2).

Multiple lines of evidence offer possible explanations for the modest effects seen with immune checkpoint inhibitors in SCLC. SCLC typically expresses low levels of antigen-presenting proteins (3–7), suggesting a mechanism by which mutation-induced neoepitopes may avoid recognition by surveilling immune cells. SCLC has also been associated with low levels of tumor-infiltrating lymphocytes (TIL; refs. 8, 9) and PD-L1 expression (10–12), raising the possibility that SCLC immune escape occurs prior to and/or independent of the PD-1/PD-L1 axis. Nevertheless, motivated, in part, by the observation of rare durable responses to ICB in SCLC, considerable effort has recently been directed to identifying biologically distinct subsets of SCLC, including subsets that benefit from ICB. Almost all SCLC harbors inactivating mutations or deletions in RB1 and TP53 whereas other frequent alterations in SCLC include amplification of MYCL, MYC, and MYCN; mutations in the PIK3CA/PTEN pathway; deletions of epigenetic regulators such as KMT2D, CREBBP, and EP300; and mutations in NOTCH family members (13–16). However, specific genetic alterations in SCLC have not yet been linked to differential responses to ICB. SCLC tumors also exhibit strongly differential activation of key transcription factors such as ASCL1, NEUROD1, and POU2F3 leading to the proposed transcription factor–defined SCLC subtypes (i.e., SCLC-A, -N, and -P; refs. 17, 18). Increased inflammation has also been identified in a subset of SCLC tumors associated with higher expression of MHC-I and lower expression of canonical neuroendocrine genes including ASCL1 (19, 20). Patients harboring tumors classified based on gene expression data as “inflamed” (termed SCLC-I) seemed to derive a greater clinical benefit from the addition of anti-PD-L1 antibody to chemotherapy in IMpower133, the clinical trial that resulted in the approval of atezolizumab for SCLC (19). A second study strengthened the link between suppression of neuroendocrine features and ICB responses in SCLC (20), whereas a third study found increased NOTCH activation in patients who benefitted from ICB (21). Features of SCLC transcription may emerge as biomarkers for ICB activity in SCLC, although prospective validation and the development of rigorous and reproducible classification methods are needed. Furthermore, given the rarity of durable responses observed in IMpower133 and CASPIAN and the paucity of alternative treatment options, therapeutic methods to overcome intrinsic ICB resistance would likely have a much greater impact on clinical outcomes than diagnostic methods to predict sensitivity.

Lysine-specific demethylase 1a (LSD1) encoded by KDM1A is a lysine demethylase active on mono- and dimethyl groups of histone H3 lysine 4 as well as nonhistone substrates (22). LSD1 also interacts with other key chromatin regulators such as histone deacytelase (HDAC) 1, 2, or 3; DNA methyltransferase 1 (DNMT1); and the neuroendocrine regulators INSM1 and RCOR1 (also known as CoREST). In so doing, LSD1 participates in multiprotein complexes that generally act to repress gene expression. Mouse genetic studies have revealed that LSD1 is important for cell lineage determination and differentiation in the developing neuroendocrine pituitary gland (23), and LSD1 inhibitors can inhibit growth of SCLC cells and in vivo patient-derived xenograft (PDX) models (24–26). We recently reported that inhibition of LSD1 using iadademstat (also known as ORY-1001) in PDX models of SCLC resulted in activation of NOTCH and suppression of ASCL1 and neuroendocrine transcriptional programs (24). Notch activation has been shown to suppress neuroendocrine programs during lung development (27, 28) and in SCLC (29). LSD1 inhibition with GSK2879552 has been tested as a single agent in patients with relapsed SCLC but not in combination with ICB (30). It has also recently been shown that LSD1 genetic suppression or pharmacologic inhibition can enhance response to PD-1 inhibitors in preclinical models of other tumor types (31–33). Given that NOTCH activation and suppression of neuroendocrine state was associated with clinical benefit from PD-L1 inhibition in patients with SCLC (19, 21), we asked if LSD1 inhibition could potentiate immune responses against SCLC.

C57BL/6NJ mice were purchased from the Jackson Laboratories (RRID:IMSR_JAX:005304). Nonobese diabetic severe combined immunodeficient IL2 receptor γ-deficient (NSG) mice were obtained from the Fred Hutchinson Cancer Research Center Co-Operative Center for Excellence in Hematology. Hemizygous OT-I mice on a CD45.1 background were generated in-house. Animal experiments were performed at the Fred Hutchinson Cancer Research Center with approval from the Institutional Animal Care and Use Committee (IACUC).

Generation of the FHSC04 PDX model has been described previously (24).

RP-48 and RP-116 C57BL/6 murine SCLC (mSCLC) cell lines derived from Rb1^lox/lox;Trp53^lox/lox mice treated with intranasal Adeno-Cre have been described previously (negative for mycoplasma in 2/2022 by MycoProbe Kit; R&D Systems; ref. 6). The G8545 cell line was derived from a lung tumor that developed following intratracheal administration of Adeno-Cre virus to a Rb1^lox/lox;Trp53^lox/lox;Crebbp^lox/lox genotype mixed-background (C57BL/6 and 129S) mouse as described previously (negative for mycoplasma in 4/2022 by MycoProbe Kit, R&D Systems; ref. 13). The FH-B6RP1 cell line was derived from a lung tumor that developed following intratracheal administration of Adeno-Cre virus to a Rb1^lox/lox;Trp53^lox/lox C57BL/6 mouse according to published methods (negative for mycoplasma in 4/2021 by MycoProbe Kit, R&D Systems; ref. 34).

Human SCLC cell lines NCI-H510, NCI-H1048, NCI-H841, and NCI-H146 were purchased from ATCC and were negative for mycoplasma in 5/2021 by MycoProbe Kit, R&D Systems.

Murine SCLC cell lines were grown in DMEM complete media [DMEM 1× (Thermo Fisher Scientific, 11965092), 15% FBS (Peak Serum, PS-FB2), 1 mmol/L sodium pyruvate (Thermo Fisher Scientific, 11360–070), 100 μmol/L β-mercaptoethanol (Thermo Fisher Scientific, 501147851), 1% penicillin–streptomycin (10,000 U/mL; Thermo Fisher Scientific, 15140122), and 10 μg/mL recombinant human insulin zinc solution (Thermo Fisher Scientific, 12585–014)]. Cell lines were incubated at 37°C in humidified 5% CO₂. Human SCLC cell lines NCI-H510, NCI-H1048, and NCI-H841 were grown in DMEM/F-12 (Thermo Fisher Scientific, 11320033) supplemented with insulin–transferrin–selenium (10, 5.5, and 0.0067 μg/mL, respectively; Thermo Fisher Scientific, 41400045), 30 nmol/L hydrocortisone (Sigma, H0135), 10 nmol/L β-estradiol (Sigma, E2257), 1% penicillin–streptomycin (10,000 U/mL; Thermo Fisher Scientific, 15140122), and 10% FBS (Peak Serum, PS-FB2). Human cell line NCI-H146 was grown in RPMI1640 (Thermo Fisher Scientific, 11875093) supplemented with 1% penicillin–streptomycin (10,000 U/mL; Thermo Fisher Scientific, 15140122) and 10% FBS (Peak Serum, PS-FB2). Cell lines were incubated at 37°C in humidified 5% CO₂.

Lysates from whole-cell extracts of cultured cell lines were prepared in diluted RIPA buffer (Cell Signaling, 9806, 10× concentration) with added protease/phosphatase inhibitor. Protein concentration was determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, 23227). Immunoblotting was done using standard procedures. The following antibodies were used: anti-REST (1:1,000; Thermo Fisher Scientific, 07–579, RRID:AB_310728), anti-NOTCH1 (1:1,000; Cell Signaling Technology, 3608, RRID:AB_2153354), anti-NOTCH2 (1:1,000; Cell Signaling Technology, 5732, RRID:AB_10693319), anti-MASH1 (ASCL1; 1:1,000; BD Pharmingen, 556604, RRID:AB_396479), anti-HES1 (1:1,000; Cell Signaling Technology, 11988, RRID:AB_2728766), anti-α-tubulin (1:1,000; Santa Cruz Biotechnology, sc-398103, RRID:AB_2832217), anti-RIG-1 (1:1,000; Proteintech, 20566–1-AP, RRID:AB_10700006), anti-PD-L1 for mouse (1:1,000; R&D Systems, MAB90781-SP, RRID:AB_2921258), anti-PD-L1 for human (1:1,000; Cell Signaling Technology, 13684, RRID:AB_2687655), anti-β2-microglobulin (1:1,000; Proteintech, 13511–1-AP, RRID:AB_2062735), anti-MDA5 (1:1,000; Proteintech, 21775–1-AP, RRID:AB_10734593), anti-β-actin (1:1,000; Proteintech, 66009–1-lg, RRID:AB_2782959), anti-LSD1 (1:1,000; Cell Signaling Technology, 2184, RRID:AB_2070132), anti-NOTCH1-ICD (1:1,000, Cell Signaling Technology, 4147, RRID:AB_2153348), and anti-STING (1:1,000, Cell Signaling Technology, 13647, RRID:AB_2732796).

To assay the dose–response to bomedemstat, RP-116 and RP-48 cells were seeded in clear-bottom, opaque-walled, 96-well culture plates with escalating dosages of bomedemstat. After 96 hours of co-culture, cell survival was evaluated using the CellTiter-Glo luminescent cell viability assay (Promega) with a BioTek plate reader to quantify luminescent signal.

Bomedemstat (Imago Biosciences) was resuspended in sterile water with 0.5% weight/volume Tween-80 and 0.5% weight/volume carboxymethylcellulose sodium salt. Bomedemstat vehicle was prepared as sterile H₂O with 0.5% weight/volume Tween-80 and 0.5% weight/volume carboxymethylcellulose sodium salt. Anti-PD-1 antibody (Bio × Cell BE0146) was diluted in cold PBS to a final concentration of 0.833 mg/mL.

SCLC cells in the desired number were precipitated via gentle centrifugation and resuspended in DMEM complete media. Per mouse, 100 μL of cells in media were then mixed 1:1 with 100 μL of Matrigel (Sigma, CLS354234). Two hundred microliters of cells were implanted, unilaterally, into the left flanks of NSG or C57BL/6NJ mice using a 28-gauge fixed needle syringe. Flank tumor length and width were measured twice each using digital calipers and measurements were averaged. Estimated tumor volume was calculated using the modified ellipsoid formula [(L × W²)/2]. Animals were entered into a treatment group when estimated flank tumor volume was ∼150 mm³. For the RP-48_OVA-LLO190 syngeneic flank tumor experiments, to facilitate time-sensitive analyses such as multiparameter flow cytometry, treatment start was synchronized in batches, with starting tumor size balanced across treatment groups to the extent possible. Mice were sacrificed and tumors were collected at planned experiment conclusion or when tumors exceeded maximum allowed size (∼2,000 mm³), or if unacceptable ulceration or other impairment to animal health developed.

Vehicle was administered via oral gavage to match bomedemstat administration schedule across all in vivo experiments. Bomedemstat was administered via oral gavage throughout. NSG mice were treated as follows. For molecular analyses, bomedemstat was administered at 25 mg/kg daily. For tumor growth kinetics, bomedemstat was administered at 40 mg/kg on days 1 to 5 of a 7-day cycle as a 4 mg/mL solution. C57BL/6NJ mice were administered bomedemstat at 45 mg/kg/day. Anti-PD-1 antibody was administered via intraperitoneal injection at a flat dose of 250 μg on days 1 and 4 of a 7-day cycle. Tumors were resected after euthanasia via cervical dislocation or carbon dioxide administration. Tumor fragments were snap-frozen and stored at −80°C.

To generate RP-48_OVA-LLO190 cells, we used a retroviral plasmid based on the pMP71 backbone and encoding enhanced GFP fused to a peptide sequence from the listeriolysin O protein and a peptide sequence from chicken ovalbumin (OVA) that was a gift from Dr. Stan Riddell (pJV119-pMP71-mGFP-LLO-SIINFEKL). Retrovirus was produced in 293T cells with Lipofectamine 2000 (Thermo Fisher Scientific, 11668019) according to manufacturer's directions and using accessory plasmids pUMVC (Addgene, #8449) and pCMV-VSV-G (Addgene, #8454). RP-48 cells were transduced and expanded for 1 to 2 weeks in culture. GFP-positive cells were then isolated via FACS. For the generation of CRISPR-mediated LSD1 knockout RP-48_OVA-LLO190 cells, the following sgRNA sequences were cloned into the lentiCRISPRv2 vector (Addgene, #52961; a gift from Feng Zhang): Control-1 (5′-CGCTTCCGCGGCCCGTTCAA-3′), Control-2 (5′-CTAACAGCGCCGTTATCTAC-3′), Lsd1–1 (5′-ATATTCATCTTCTGAGAGGT-3′), Lsd1–2 (5′-TCTTCCTCAGGTGGGGCTTG-3′), and Lsd1–3 (5′-CCATGACCGAATGACCTCTC-3′). Lentivirus production with packaging plasmids psPAX2 (Addgene, #12260) and pMD2.G (Addgene, #12259) was performed as described above. RP-48_OVA-LLO190 cells were transduced and selected for 72 hours with puromycin (1 μg/mL, InvivoGen ANTPR1).

RP-48_OVA-LLO190 flank tumors were generated and collected as described above. Two cohorts were analyzed: a 16-day cohort which focused on analysis of immune cell subsets, and a 10-day cohort which focused on analysis of SCLC cell MHC-I (H-2K^b) expression.

For the 16-day timepoint cohort, tumor tissue was prepared and analyzed as follows. Fresh tumor tissue was minced using a razor blade in 1 to 2 mm fragments. Tumor tissue was dissociated into a single cell suspension by incubating for 45 minutes at 37°C in RPMI with HEPES (Thermo Fisher Scientific, #22400089) with 80 U/mL DNase (Worthington, #LS00206) and 300 U/mL Collagenase Type I (Worthington, #CLS-1), filtration through a 70 μmol/L strainer, and RBC lysis with ACK lysis buffer (Gibco) according to manufacturer's instructions. Following RBC lysis, tumors were resuspended as single-cell suspensions for viability and cell surface staining. Viability staining was performed with Fixable Viability Dye (eBioscience, 65–0865–14) according to manufacturer's instructions. For surface staining, cells were incubated at 4°C for 30 minutes in staining buffer (PBS, 2% FBS) with the following directly conjugated antibodies for murine proteins (from BioLegend unless otherwise specified): anti-CD45 (clone 30-F11; BD Biosciences, RRID:AB_2872789), -CD3 (145–2C11, RRID:AB_312670), -CD4 (RM4–5, RRID:AB_312714), -CD8 (53–6.7, BD, RRID:AB_2732919), -CD19 (6D5, RRID:AB_313646), -CD11b (M1/70, BD, RRID:AB_394491), -CD11c (N418, RRID:AB_11204262), -Ly6C (HK1.4, RRID:AB_10640119), -Ly6G (1A8, BD, RRID:AB_2737730), -F4/80 (BM8, RRID:AB_10901171), and -NK1.1 (PK136, BD, RRID:AB_2738002). Data were acquired on Canto II or Fortessa flow cytometers (BD Biosciences) and analyzed using FlowJo software (TreeStar).

For the 10-day timepoint cohort, tumor tissue was prepared and analyzed as follows. Fresh tumor tissue was minced using scissors into 1 to 2 mm fragments. Tumor tissue was dissociated into a single-cell suspension by incubating for 45 minutes at 37°C in RPMI with HEPES (Thermo Fisher Scientific, #22400089) with 80 U/mL DNase (Worthington, #LS00206) and 300 U/mL Collagenase Type I (Worthington, #CLS-1) filtration through a 100 μmol/L strainer, and RBC lysis with ACK lysis buffer (Gibco). Cells were stained using the Live/Dead Fixable Aqua Dead Cell Stain Kit (Invitrogen). For surface staining, cells were incubated at 4°C for 30 minutes in staining buffer (PBS, 2% FBS) with anti-CD45 (BioLegend, 30-F11, RRID:AB_2872789) and anti-H-2K^b (Biolegend, AF6–88.5, RRID:AB_313735) directly conjugated antibodies for murine proteins. Cells were fixed using Foxp3/Transcription Factor Staining Buffer Set (eBioscience) and stored at 4°C in the dark until analysis. Data were acquired on Symphony flow cytometers (BD Biosciences) and analyzed using FlowJo software (TreeStar).

RNA was extracted from cell pellets or tumor samples using TRizol reagent. RNA-seq libraries were generated from 500 ng of RNA using the NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs, E7530L). Sequencing was performed using an Illumina HiSeq 2500 instrument in 50 bp single end read configuration. Fastq files were filtered to exclude reads that did not pass Illumina's base call quality threshold. STAR v2.7.1 (35) with 2-pass mapping was used to align reads to mouse genome build mm10 and GENCODE gene annotation M23 (https://www.gencodegenes.org/mouse/). FastQC 0.11.9 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and RSeQC 3.0.0 (36) were used for QC including insert fragment size, read quality, read duplication rates, gene body coverage, and read distribution in different genomic regions. FeatureCounts (37) in Subread 1.6.5 (37) was used to quantify gene-level expression by unstranded read counting. Bioconductor package edgeR 3.26.8 (38) was used to detect differential gene expression between sample groups. Genes with low expression were excluded using edgeR function filterByExpr with min.count = 5 and min.total.count = 15. The filtered expression matrix was normalized by TMM method (39) and subject to significance testing using GLM LRT method and cell line as blocking factor (if applicable).

RPKM or FPKM values generated as described above were analyzed using Gene Set Enrichment Analysis v4.2.2 software (Broad Institute and Regents of the University of California) with default settings except that mouse to human gene mapping was performed using the supplied “Mouse_ENSEMBL_Gene_ID_Human_Orthologs_MSigDB.v7.5.1.chip” file and “gene_set” permutation was used for generation of statistics. For Hallmark pathway analysis, the provided “ftp.broadinstitute.org://pub/gsea/gene_sets/h.all.v7.5.1.symbols.gmt” file was used. For enrichment of ASCL1 gene targets in mSCLC cells, we used a list of 141 ASCL1 target genes that were shared between mouse and human chromatin occupancy analyses (40).

Cultured SCLC tumor cells were seeded with bomedemstat (1 μmol/L) or equal volume of vehicle (DMSO) added to the culture media. Varying concentrations of recombinant murine IFNγ (Peprotech 315–05 for RP-48/RP-48_OVA-LO190 or Abcam ab9922 for FH-B6RP1) or human IFNγ (Sigma, I17001) were added 24 or 48 hours before collecting cells. Cells were incubated at 37°C in humidified 5% CO₂ for a total of 72 or 96 hours after plating. Cells were then harvested for flow cytometry analysis. Briefly, cells were stained using the Live/Dead Fixable Aqua Dead Cell Stain Kit (Invitrogen) according to the manufacturer's protocol. For surface staining, cells were incubated at 4°C for 30 minutes in staining buffer (PBS, 2% FBS) with anti-H-2K^b (murine; BioLegend, AF6–88.5, RRID:AB_313735) or anti-pan-MHC-I (human; BioLegend, 311405, RRID:AB_314875) directly conjugated antibodies. Cells were fixed using Foxp3/Transcription Factor Staining Buffer Set (eBioscience) and stored at 4°C in the dark until analysis. Data were acquired on Symphony flow cytometers (BD Biosciences) and analyzed using FlowJo software (TreeStar).

RP-48_OVA-LLO190 tumor cells were seeded at a density of 4 × 10⁵ cells in 3 mL of DMEM complete media. On the day of tumor cell plating, bomedemstat (1 μmol/L) or equal volume of DMSO with or without 0.5 μmol/L gamma-secretase inhibitors dibenzazepine (DBZ; Selleckchem, S2711) or RO4929097 (RO4; Selleckchem S1575) was added to the culture media. Cells were passaged on days 4 and 7. Recombinant murine IFNγ (10 ng/mL; Peprotech, 315–05) was added for the final 48 hours (days 8–10). Cells were then harvested for flow cytometry analysis as above. FH-B6RP1 cells were processed identically except that 2 × 10⁶ cells were seeded, RO4 was omitted, cells were treated for a total of 96 hours, and IFNγ was added for the final 24 hours (72–96 hours).

RP-48_OVA-LLO190 tumor cells were seeded in 6-well tissue-culture–treated plates at a density of 1 × 10⁵ cells in 3 mL of DMEM complete media and treated with 1 μmol/L bomedemstat or vehicle for 72 hours, with recombinant IFNγ (10 ng/mL; Peprotech, 315–05) added for the last 24 hours of culture. In parallel, CD8⁺ T cells were isolated by negative selection from spleens of OT-1 transgenic or wild-type B6 mice using a Negative Isolation Kit (Stemcell, 19853) and stimulated with 1 mg/mL each of plate-bound anti-CD3 and anti-CD28 (Biolegend, 145–2C11 and 37.51, respectively) for 48 hours in a 37°C, 5% CO₂ incubator in complete RPMI (RPMI1640, 10% heat inactivated FBS, 1 mmol/L sodium pyruvate, 1 mmol/L HEPES, 100 U/mL penicillin/streptomycin, 50 μmol/L β-mercaptoethanol) supplemented with 50 U/mL recombinant murine IL2 (Peprotech, 212–12). Seventy-two hours after tumor cell plating and 48 hours after T-cell isolation, T cells and tumor cells were co-cultured in 96-well U-bottom plates at a ratio of 2:1 (50,000 tumor cells and 25,000 T cells) in complete RPMI. Co-cultures were incubated for 24 hours in a 37°C, 5% CO₂ incubator. Cells were then harvested for flow cytometry analysis. A defined number of counting beads (Polybead, 18328–5) were spiked into each sample. Cells were stained with Live/Dead Fixable Aqua Dead Cell Stain Kit (Invitrogen) and surface stained with anti-CD8 (BioLegend, 53–6.7, RRID:AB_2732919) and anti-CD45 (BioLegend, 30-F11, RRID:AB_2872789) directly conjugated antibodies. Cells were fixed using Foxp3/Transcription Factor Staining Buffer Set (eBioscience) and stored at 4°C in the dark until analysis. Data were acquired on Symphony flow cytometers (BD Biosciences) and analyzed using FlowJo software (TreeStar). Percent survival was calculated as: 100–100*[[(# of live tumor cells in CD8 co-culture) – (# of live tumor cells in OT-1 co-culture)]/(# of live tumor cells in CD8 co-culture)].

RP-48_OVA-LLO190 tumor cells were seeded in 6-well tissue-culture–treated plates at a density of 1 × 10⁵ cells in 3 mL of DMEM complete media and treated with bomedemstat (1 μmol/L) or vehicle for 72 hours, with recombinant IFNγ (10 ng/mL; Peprotech, 315–05) added for the last 24 hours of culture. In parallel, CD8⁺ T cells were isolated as described for the co-culture assay above. Seventy-two hours after tumor cell plating and 48 hours after T-cell isolation, T cells and tumor cells were co-cultured in 96-well U-bottom (FB012931) plates at a ratio of 2:1 (100,000 T cells and 50,000 tumor cells) in complete RPMI with PE conjugated anti-CD107a antibody (1 μg/mL; Biolegend, 1D4B). For T-cell independent stimulation, PMA (50 ng/mL) and ionomycin (1 μg/mL) were added in 0.2 mL of complete RPMI. Co-cultures were incubated for 24 hours in a 37°C, 5% CO₂ incubator. Cells were then harvested for flow cytometry analysis. Cells were stained with Live/Dead Fixable Aqua Dead Cell Stain Kit (Invitrogen) and surface stained with anti-CD8 (BioLegend, 53–6.7, RRID:AB_2732919) and anti-CD45 (BioLegend, 30-F11, RRID:AB_2872789) directly conjugated flow antibodies. Cells were fixed using Foxp3/Transcription Factor Staining Buffer Set (eBioscience) and stored at 4°C in the dark until analysis. Data were acquired on Symphony flow cytometers (BD Biosciences) and analyzed using FlowJo software (TreeStar).

Protein expression in immunoblots was compared using unpaired t tests. Flank tumor growth kinetics were compared using estimation of individual mixed effects models comparing bomedemstat+anti-PD-1 to each other treatment group. RP-48_OVA-LLO190 flank tumor immune cell subset frequencies, MHC-I positivity, and geometric mean fluorescence intensity (gMFI) were compared using one-way ANOVA with follow-up testing using Dunnett multiple comparison test, where follow-up tests were performed comparing the bomedemstat+anti-PD-1 group to the other three treatment groups. SCLC cell line H-2K^b/pan-MHC-I positivity and gMFI were compared using multiple unpaired Welch t tests. All statistical analysis was done in GraphPad Prism 9.

Experimental design graphics were designed on biorender.com.

The gene expression data generated in this study are publicly available in Gene Expression Omnibus (GEO) at GSE208614.

For these studies we employed bomedemstat (also known as IMG-7289), an orally available, irreversible LSD1 inhibitor that is currently being tested in clinical trials for the treatment of myeloproliferative neoplasms, but whose activity in SCLC has not been well studied. We first evaluated the activity of bomedemstat in a PDX model of SCLC previously shown to strongly respond to another LSD1 inhibitor, iadademstat (also known as ORY1001; ref. 24). Similarly to iadademstat, bomedemstat at maximum tolerated dose induced complete regression of FHSC04 flank tumors (Supplementary Fig. S1), and intermediate dose treatment with bomedemstat led to NOTCH activation and suppressed neuroendocrine signaling, with reduced expression of ASCL1 (Fig. 1A; Supplementary Fig. S2).

We then tested the activity of bomedemstat in cell lines derived from Rb1/Trp53-deleted murine SCLC (mSCLC) models. Bomedemstat did not substantially reduce cell growth at a dose of up to 10 μmol/L (Fig. 1B). However, Western blot and RNA-seq analyses revealed Notch pathway activation and suppression of a gene program driven by the neuroendocrine master regulator ASCL1 (Fig. 1C and D; Supplementary Fig. S3). Gene Set Enrichment Analysis (41, 42) focusing on a set of 141 ASCL1 target genes (40) revealed broad suppression of ASCL1 targets with bomedemstat treatment (Fig. 1E). Thus, LSD1 inhibition with bomedemstat leads to Notch activation and suppression of neuroendocrine signaling in a PDX model of SCLC and in mSCLC cell lines.

We next asked whether bomedemstat would enhance responses to PD-1 inhibition in vivo. Low mutational burden and neoantigen load in genetically engineered mouse models of SCLC (43) might obscure the effects of bomedemstat on antitumor immunity. Thus, we engineered an mSCLC cell line derived from a C57BL6 genetic background to express two model murine antigens, the SIINFEKL peptide derived from chicken OVA and the listeriolysin O 190–201 amino acid peptide (LLO190; ref. 44), which we termed RP-48_OVA-LLO190 (Fig. 2A). We then assessed whether LSD1 inhibition with bomedemstat would enhance antitumor immune responses to RP-48_OVA-LLO190 flank tumors grown in syngeneic C57BL6 recipient animals. Treatment with bomedemstat or anti-PD-1 antibody alone exerted minimal growth suppressive effect, whereas treatment with bomedemstat in combination with anti-PD-1 antibody significantly suppressed growth (Fig. 2B), suggesting the possibility that LSD1 inhibition was augmenting antitumor immunity. Thus, although PD-1 inhibition alone did not suppress tumor growth in this model, strong tumor growth inhibition was seen when combined with LSD1 inhibition.

To investigate whether bomedemstat alters the immune tumor microenvironment, we performed flow cytometric immunophenotyping of treated RP-48_OVA-LLO190 flank tumors. We observed a selective increase in the prevalence of cytotoxic CD8⁺ T cells in tumors treated with bomedemstat and anti-PD-1 antibody (Fig. 2C), without a clear change in all other immune cell subsets examined, with the exception of neutrophils (Fig. 2D–G; Supplementary Fig. S4). Although neutrophils were relatively infrequent across all groups, they were present at especially low frequencies in the bomedemstat-containing treatment groups (Fig. 2G). Of note, neutropenia is an expected side-effect of LSD1 inhibitor administration as LSD1 is required for terminal differentiation of myeloid lineage hematopoietic cells (45). Together, these data suggest that bomedemstat alters the composition of the immune tumor microenvironment in favor of cytotoxic CD8⁺ T cells.

We then performed RNA-seq to further characterize the molecular features of treated RP-48_OVA-LLO190 flank tumors. Bomedemstat and anti-PD-1 combination therapy led to marked increase in inflammatory gene expression pathways as revealed by Gene Set Enrichment Analysis (41, 42) when compared with bomedemstat alone, and both inflammatory and NOTCH pathways when compared with anti-PD-1 antibody alone (Fig. 3A and B; Supplementary Fig. S5). Inspection of selected differentially expressed genes in pathways of interest in individual tumors confirmed systematic variation by treatment group (Fig. 3C). These data showed increased expression of antigen presentation machinery, including components of MHC-I, as well as chemokines such as CXC3L1 and CXCL14, which can mediate recruitment of T cells and dendritic cells, respectively. Activation of NOTCH pathway components and suppression of neuroendocrine transcripts was also observed when bomedemstat was combined with PD-1 treatment (Fig. 3C). Growth kinetic, flow cytometric, and transcriptomic analyses therefore coordinately identified a signal of increased antitumor immunity when LSD1 inhibition with bomedemstat was combined with anti-PD-1 antibody treatment.

A described mechanism of immune evasion in SCLC is epigenetic silencing of antigen presentation machinery gene expression leading to very low levels of surface MHC-I (7). Moreover, it has emerged recently that MHC-I re-expression can potentiate antitumor immunity in SCLC (6, 20, 21). To explore the possibility that LSD1 inhibition with bomedemstat induces MHC-I expression, we generated a second cohort of RP-48_OVA-LLO190 flank tumors and again observed a clear trend of suppressed tumor growth in animals treated with the combination of bomedemstat and anti-PD-1 antibody, although statistical significance was not reached at the early timepoint used for this cohort (Supplementary Fig. S6). Flow cytometric analysis of tumor cells revealed a higher fraction of MHC-I–expressing cells in tumors treated with a bomedemstat-containing regimen relative to vehicle as well as increased surface MHC-I expression on MHC-I–positive cells when treated with both agents (Fig. 4A–C; Supplementary Fig. S6).

We then sought to confirm that this observation was a direct effect of bomedemstat on SCLC cells by using flow cytometry to determine surface MHC-I expression in cultured RP-48_OVA-LLO190 cells. Echoing the effects observed in flank tumors, we found that bomedemstat alone or in combination with varying concentrations of IFNγ led to an increase in the proportion of MHC-I–positive cells as well as increased surface expression of MHC-I in MHC-I–positive cells (Fig. 4D and E; Supplementary Fig. S7). To exclude the possibility that this was a cell line–specific effect or a consequence of expression of an exogenous class I model antigen, we then tested the effect of bomedemstat treatment on surface MHC-I expression in an additional Rb/p53-deleted murine cell line, FH-B6RP1. We again found that bomedemstat increased the proportion of cells expressing surface MHC-I as well as the intensity of surface expression in positive cells (Fig. 4F; Supplementary Fig. S7). These results confirm that bomedemstat reproducibly induces MHC-I across multiple murine models of SCLC and independently of exogenous expression of a model antigen.

We next tested whether upregulation of MHC-I also occurs with bomedemstat treatment in human SCLC cell lines. As LSD1 inhibition leads to loss of neuroendocrine features, we chose two human SCLC-A models, NCI-H510 and NCI-H146, that express high levels of ASCL1 and other neuroendocrine genes and relatively low levels of the MHC-I subunit B2M, according to publicly available RNA-seq data (46). We again observed bomedemstat-induced MHC-I upregulation across a range of IFNγ concentrations (Fig. 4G and H; Supplementary Fig. S7). We also treated NCI-H1048, a SCLC-P model, and NCI-H841, a SCLC-Y model, with bomedemstat. Unlike in the two SCLC-A models, MHC-I expression was observed in nearly all cells at baseline and bomedemstat exposure did not increase the percentage of MHC-I–positive cells or the intensity of cell surface staining (Supplementary Fig. S8).

To support the notion that upregulation of MHC-I by bomedemstat is not an off-target effect of this compound, we used CRISPR to delete LSD1 in RP-48_OVA-LLO190 cells. We first sought to confirm expected changes in LSD1 and NOTCH pathway levels in cells expressing one of three independent sgRNAs targeting LSD1 relative to cells expressing one of two nontargeting control sgRNAs. Transduction with LSD1-targeted sgRNAs led to modest but consistent reduction in LSD1 protein levels and strong upregulation of NOTCH pathway members (Fig. 4I). We then subjected these cells to flow cytometry to quantify surface MHC-I, finding that CRISPR-mediated LSD1 knockout increased the proportion of MHC-I–positive cells (Fig. 4J) and increased surface MHC-I intensity in positive cells (Supplementary Fig. S7). Thus, decreased LSD1 activity, either via genetic or pharmacologic methods, induces MHC-I upregulation across a range of SCLC preclinical models.

We then sought to determine mechanistic links between LSD1 inhibition and antigen presentation machinery upregulation. Given the central role of LSD1 in transcriptional regulation, we hypothesized that LSD1 inhibition might increase transcription of key antigen presentation machinery components. We therefore performed transcriptomic profiling of RP-48_OVA-LLO190 cells that had been treated with bomedemstat for an extended duration (10 days), with or without IFNγ for the last 48 hours. Regardless of IFNγ exposure, we observed increased expression of antigen presentation machinery components B2m (a required MHC-I subunit), H2-D1 (a murine MHC-I locus), and Tapbp (a mediator of antigen loading onto newly formed MHC-I complexes; Supplementary Fig. S9), supporting the hypothesis that transcriptional reprogramming is involved in upregulation of antigen presentation machinery.

We also queried whether PD-L1 expression was altered in response to LSD1 inhibition in our system, as has previously been shown in other cancer types (31). Bomedemstat treatment increased PD-L1 protein and transcript levels in the presence of IFNγ sensitization in RP-48_OVA-LLO190 cells (Fig. 5A; Supplementary Fig. S10), and CRISPR knockdown of LSD1 in RP-48_OVA-LLO190 mSCLC cells was sufficient to induce PD-L1 levels even in the absence of IFNγ (Fig. 5B; Supplementary Fig. S10). In the human SCLC-A models tested, which exhibited widely varying PD-L1 expression at baseline, we did not observe PD-L1 increase, despite molecular evidence of induced antigen presentation machinery, that is, increased B2M (Fig. 5C and D; Supplementary Fig. S10). We also searched for evidence of activation of cytoplasmic nucleic acid sensing pathways, which have previously been implicated in LSD1i-induced immunogenic signaling in other cancer models (31) and SCLC tumor cell intrinsic immunogenic signaling in response to other targeted therapies (47). In RP-48_OVA-LLO190 cells, we observed increased transcript expression of the dsRNA sensors Ddx58 (Supplementary Fig. S9), which encodes RIG-I, and Ifih1 (Supplementary Fig. S9), which encodes MDA5, irrespective of IFNγ treatment. Using immunoblotting, we also detected increases in STING, which is involved in cytoplasmic DNA sensing, in RP-48_OVA-LLO190 cells harboring LSD1 knockout and H510 cells treated with bomedemstat (Fig. 5A–D; Supplementary Fig. S10). STING protein was not detected in H146 cells. Together, these data provide rationale for the potentiation of anti-PD-1 treatment by bomedemstat.

We also explored whether the increased chemokine transcript abundance detected in tumor tissue (Fig. 3C) was a tumor cell–intrinsic phenomenon using the transcriptomic profiling data from RP-48_OVA-LLO190 cells treated with extended duration bomedemstat with or without IFNγ. Focusing on a curated consensus list of 133 immune system–related cytokine genes (48), we identified 12 genes that were differentially expressed, eight of which were significantly upregulated by bomedemstat irrespective of IFNγ presence (Supplementary Fig. S9). We observed increased transcript abundance of the chemoattractive cytokines CX3CL1 (49) and CXCL14 (Supplementary Fig. S9; ref. 50), which had also been observed in bomedemstat-treated flank tumors, as well as CXCL12 (51) and IL33 (52), among other genes.

LSD1 inhibition also strongly induces NOTCH signaling and NOTCH induction has been implicated in antigen presentation machinery upregulation in SCLC (21). We therefore tested whether suppression of NOTCH signaling was sufficient to alter the degree to which bomedemstat treatment increased MHC-I surface expression. Treatment with either of two γ-secretase inhibitors led to substantial reduction in NOTCH pathway activation as evidenced by decreased induction of NOTCH1 intracellular domain and HES1 but did not reduce MHC-I upregulation in the two mSCLC models tested (Supplementary Fig. S11). These data suggest that LSD1 inhibition leads to MHC-I upregulation through one or more NOTCH-independent mechanisms.

Finally, we sought to determine whether bomedemstat treatment of mSCLC cells was sufficient to alter cytotoxic T-cell function, as evidenced by cytotoxic degranulation and target cell survival. To test whether bomedemstat treatment of SCLC cells led to altered cytotoxic activation of antigen-specific T cells, we performed in vitro co-culture of RP-48_OVA-LLO190 cells with OT-I (antigen-specific) or bulk CD8⁺ (antigen-nonspecific) T cells in the presence of the cytotoxic degranulation-marking antibody anti-CD107a. Pretreatment of SCLC cells with bomedemstat and IFNγ led to a modest but significant increase in the fraction of CD107a-positive OT-I antigen-specific T cells compared with IFNγ alone, and a trend was also observed in the absence of IFNγ, although the difference was not statistically significant when corrected for multiple testing (Fig. 5E). In contrast, no effect on CD107a positivity was observed when RP-48_OVA-LLO190 cells were co-cultured with antigen-nonspecific bulk CD8⁺ T cells, confirming the requirement for T-cell recognition of a specific MHC-I–antigen complex (Fig. 5E). To establish that increased cytotoxic degranulation was associated with functional consequences, we employed the same experimental system and measured viable RP-48_OVA-LLO190 cell numbers after bomedemstat exposure and co-culture with antigen-nonspecific CD8 T cells or OVA-specific OT-I T cells. We found that bomedemstat pretreatment, with or without IFNγ, significantly reduced SCLC cell survival in the antigen-specific (“OT-I”) co-culture relative to the antigen-nonspecific (“CD8”) co-culture (Fig. 5F). These data establish that the molecular changes induced by bomedemstat in target SCLC cells are sufficient to increase cytotoxic CD8⁺ T-cell activation in an antigen-dependent manner, and that increased antigen-specific CD8⁺ T-cell activity is associated with decreased target cell survival.

We therefore constructed a framework for antitumor immune potentiation by the LSD1 inhibitor bomedemstat (Fig. 5G). Bomedemstat likely acts via both LSD1 demethylase activity inhibition as well as disruption of protein complexes involving LSD1, thereby altering chromatin organization and function and eventually resulting in more effective antitumor immunity and reduced tumor growth (Fig. 5G). To test the hypothesis that bomedemstat-mediated LSD1 inhibition can potentiate antitumor immunity, we have initiated a clinical trial for patients with extensive-stage SCLC. This trial will test the combination of bomedemstat and standard-of-care maintenance atezolizumab in those patients with stable or reduced disease after four cycles of induction platinum, etoposide, and anti-PD-L1 antibody (Fig. 5H). The primary outcomes of this single-arm study are safety and tolerability of the combination of bomedemstat and atezolizumab and 6-month progression-free survival measured from the initiation of maintenance bomedemstat and atezolizumab relative to the historical control of IMpower133 (1). Additional endpoints will include overall survival and correlative biomarker assessments of circulating cell-free DNA. Our in vitro studies and the results using an immunocompetent syngeneic mouse model taken together serve as proof-of-concept evidence that epigenetic reprogramming of SCLC cells with the LSD1 inhibitor bomedemstat can overcome SCLC-intrinsic immunosuppressive mechanisms including MHC-I downregulation and sensitize SCLC tumors to immune recognition and killing.

Most SCLC tumors exhibit what is commonly referred to as an immunologically “cold” tumor and tumor microenvironment, with low expression of antigen processing and presentation machinery including MHC-I and low T-cell infiltration. Recent studies, however, have identified a subset of human SCLC with high expression of antigen processing and presentation machinery genes associated with improved clinical responses to ICB (19–21). In one study, high MHC-I expression in clinical samples was associated with reduced ASCL1 and the neuroendocrine marker chromogranin, and MHC-I high status was strongly enriched in patients with more durable responses to ICB (20). In a second study, transcriptional analyses revealed a subset of SCLC that the authors termed SCLC-I (“Inflamed”), with low expression of neuroendocrine genes such as ASCL1 and high expression of the neuronal repressor REST (19). The SCLC-I subset experienced greater clinical benefit from the addition of ICB based on IMpower133 clinical outcomes data (19). In a third study, both low neuroendocrine differentiation and high antigen presentation machinery expression correlated with clinical responses to ICB in SCLC patient samples, but the most significant correlate of response to ICB was activation of NOTCH signaling (21).

Pharmacologic approaches that drive activation of key pathways associated with sensitivity to ICB could theoretically overcome intrinsic resistance to ICB in a substantial fraction of patients with SCLC and thereby considerably improve the clinical benefit associated with ICB use. EZH2 inhibition, for example, was recently found to increase MHC-I expression in SCLC and may represent a promising approach to improving ICB efficacy (6, 20). Our previous finding (24) that inhibition of another epigenetic regulator, LSD1, activated NOTCH, decreased expression of ASCL1 in SCLC, and reduced neuroendocrine characteristics, that is, features associated with clinical responses to ICB, led us to test LSD1 inhibition as a potential mechanism to improve immune responses in SCLC. Indeed, we observed striking improvements in tumor growth inhibition when LSD1 inhibition was added to PD-1 inhibition in an immunocompetent syngeneic mouse model of SCLC.

Our previous work with the LSD1 inhibitor iadademstat (ORY1001) showed that while most SCLC PDX models exhibit modest tumor growth inhibition with LSD1 suppression, one PDX model, FHSC04, derived from a patient with SCLC that had relapsed following front-line chemotherapy, exhibited a striking and durable regression with iadademstat monotherapy (24). Importantly, the activation of NOTCH and suppression of ASCL1 was a feature of all ASCL1-positive PDX models studied ex vivo (24). This finding is consistent with our current studies of bomedemstat, which did not exert substantial cytotoxic effect against the mSCLC cells tested but did induce NOTCH and suppress an ASCL1 gene program. Thus, although there may be potential for a rare exceptional response to LSD1 inhibition alone, the general ability of LSD1 inhibition to activate NOTCH and suppress neuroendocrine features of SCLC points to potential activity across a much broader subset of patients with SCLC.

Potentiation of the antitumor efficacy of PD-1 blockade by LSD1 in an immunocompetent model was associated with increased CD8 infiltration, suggesting a key role of antigen-specific cytotoxicity. Consistent with previous studies targeting the SCLC epigenome (6, 20), tumors treated with LSD1i and PD-1 blockade also displayed increased expression of antigen presentation machinery by both transcriptional and cell surface protein analysis, leading us to explore this phenomenon further using a panel of cell culture models. Bomedemstat treatment alone and in combination with IFNγ increased cell surface MHC-I expression across a panel of murine and human SCLC-A models, and this effect was phenocopied by CRISPR knockout of LSD1, supporting LSD1’s central role. However, LSD1 inhibition did not induce MHC-I expression in either of the two ASCL1-low human models tested, which also expressed high MHC-I levels at baseline. Future investigations should determine whether subtype as defined by transcription factor expression, with or without other features, can reliably predict LSD1i-induced immunogenicity.

We also explored potential roles for NOTCH (21) and cytoplasmic nucleic acid sensor signaling (31, 47) in LSD1i-induced MHC-I upregulation. Substantial reduction of NOTCH activation did not attenuate MHC-I upregulation in our studies, suggesting that LSD1i may retain immunogenic activity in the considerable fraction of SCLC tumors that harbor inactivating NOTCH mutations (15). We did however detect cytoplasmic nucleic acid sensor upregulation in multiple models, raising the question of whether LSD1i treatment will lead to similar immunogenicity in NOTCH-high tumors, where it has recently been shown that STING signaling is suppressed by NOTCH activity (53).

Bomedemstat treatment of SCLC cells in culture also increased markers of antigen-specific T-cell activation and sensitivity to antigen-specific T-cell killing. Our studies focused on MHC-I derepression as a likely mechanism by which LSD1 inhibition sensitizes SCLC to PD-1 blockade, but the complete mechanism in the fully immunocompetent system we utilized is likely to be more complex. Bomedemstat treatment of tumors was associated with lower neutrophil levels, and whereas myelosuppression is an expected effect of LSD1 inhibitor therapy (45), neutrophils have previously been implicated in promoting an immunosuppressive immune microenvironment in lung cancer (54). Whether neutrophil depletion from the TME plays a role in the antitumor immunity induced by bomedemstat remains to be determined. We also observed increased mRNA expression of potentially immunostimulatory chemokines such as CX3CL1 and CXCL14 in LSD1i+anti-PD-1–treated tumors, and these transcripts were again upregulated by LSD1i in cultured mSCLC cells along with other potential chemoattractants such as IL33 and CXCL12. The role of LSD1i-induced chemokine secretion will also be important to delineate. Moreover, although our studies have been tumor cell focused, it has recently been shown that LSD1 can promote the terminal differentiation of T cells, and that LSD1 suppression can increase the population of progenitor-exhausted CD8⁺ T cells, which is the CD8⁺ T-cell population most responsive to PD-1 inhibition (32).

On the basis of our observations that LSD1 inhibition can potentiate responses to PD-1 axis inhibition in vivo, we reasoned that the addition of an LSD1 inhibitor to standard maintenance anti-PD-L1 antibody in front-line treatment of ES-SCLC might overcome intrinsic ICB resistance in a meaningful fraction of patients, leading to improved clinical outcomes. Importantly, however, a phase I clinical trial of the LSD1 inhibitor GSK2879552 administered as monotherapy in patients with relapsed SCLC was terminated (30). Bomedemstat has been administered to more than 170 patients with hematologic malignancies at doses of up to 6 mg/kg/day (55–57), that is, doses up to 10 times higher than those planned in the SCLC clinical study. We therefore expect that bomedemstat will be generally well tolerated in patients with SCLC and that combination therapy with bomedemstat and ICB merits clinical exploration. To this end, a clinical trial has been initiated in which bomedemstat is added to maintenance atezolizumab in extensive-stage SCLC following induction treatment with platinum/etoposide chemotherapy and PD-L1 inhibitor (Clinicaltrials.gov identifier: NCT05191797). We hypothesize that bomedemstat will sensitize intrinsically ICB-resistant SCLC and therefore substantially improve progression-free survival. Additional, exploratory endpoints will include analysis of tumor tissue and circulating cell-free DNA for biomarkers of sensitivity to bomedemstat and atezolizumab.

After decades of incremental progress in SCLC research, a recent proliferation of promising discoveries has occurred (3, 58). Prominent among these are (i) a new standard of care for patients with untreated ES-SCLC that clearly, albeit modestly, improves overall survival for all comers (i.e., chemotherapy with ICB), (ii) proposal of an epigenetically driven taxonomy of SCLC subtypes, (iii) the association of NOTCH pathway activation and neuroendocrine feature loss with benefit from ICB, and (iv) the identification of targeted therapies that can induce SCLC epigenetic reprogramming (6, 20). A critical next step will be to determine how to best integrate epigenetic markers and therapies into routine SCLC care, including how to achieve optimal benefit with ICB. Because only about 15% of patients appeared to derive more durable benefit from the addition of ICB to platinum doublet (1, 2), it is less likely that epigenetic features will be employed as clinical biomarkers to select patients for ICB use. Instead, therapeutic manipulation of the SCLC epigenome to sensitize intrinsically resistant SCLC tumors to ICB treatment represents an alternative and potentially more broadly beneficial approach. We have found that LSD1 inhibition in SCLC drives epigenetic reprogramming including NOTCH pathway activation and neuroendocrine gene suppression, increases MHC-I presentation and T-cell killing and reduces tumor growth when combined with PD-1 axis blockade. Of note, our findings are highly consistent with a very recently published study also describing the immunostimulatory role of LSD1 blockade in SCLC (59). LSD1 inhibition in combination with ICB is therefore an especially promising example of clinical translation of the emerging paradigm of epigenetic heterogeneity and plasticity in SCLC biology, with the potential to substantially improve on the outcomes that ICB can achieve for patients with SCLC.

J.B. Hiatt reports grants from NCI and National Heart, Lung, and Blood Institute during the conduct of the study. T.J. Friesen reports grants from Fred Hutchinson Cancer Research Center during the conduct of the study, as well as personal fees from Alpine Immune Sciences and Century Therapeutics outside the submitted work. K.D. Sutherland reports grants from NHMRC during the conduct of the study. K.D. Sutherland also reports personal fees and other support from AGIOS Pharmaceuticals; grants from Pfizer Pharmaceuticals, Victorian Cancer Agency Mid-Career Fellowship, and NHMRC; and other support from The Peter and Julie Alston WEHI Centenary Fellowship outside the submitted work. H.Y. Rienhoff Jr reports other support from Imago BioSciences during the conduct of the study, as well as other support from Imago BioSciences outside the submitted work; in addition, H.Y. Rienhoff Jr has a patent 20170502 pending. S. Srivastava reports grants from Lyell Immunopharma, as well as other support from Lyell Immunopharma outside the submitted work; in addition, S. Srivastava has a patent for PCT/US2018/056796 licensed and with royalties paid from Bristol Myers Squibb and a patent for PCT/US2018/049812 licensed and with royalties paid from Lyell Immunopharma. D. MacPherson reports grants from NIH/NCI during the conduct of the study, as well as grants from NIH/NCI and Janssen outside the submitted work. No disclosures were reported by the other authors.

J.B. Hiatt: Conceptualization, data curation, supervision, investigation, visualization, writing–original draft, writing–review and editing. H. Sandborg: Investigation, writing–review and editing. S.M. Garrison: Validation, investigation, visualization, writing–review and editing. H.U. Arnold: Validation, investigation, visualization, writing–review and editing. S.-Y. Liao: Validation, investigation, visualization, writing–review and editing. J.P. Norton: Investigation, writing–review and editing. T.J. Friesen: Investigation, writing–review and editing. F. Wu: Investigation, writing–review and editing. K.D. Sutherland: Writing–review and editing. H.Y. Rienhoff Jr: Supervision, writing–review and editing. R. Martins: Conceptualization, supervision, writing–review and editing. A.M. Houghton: Supervision, writing–review and editing. S. Srivastava: Supervision, investigation, writing–review and editing. D. MacPherson: Conceptualization, supervision, funding acquisition, writing–original draft, project administration.

This work was supported by the NIH (Fred Hutch NIH Lung Spore award P50CA228944 to D. MacPherson and A.M. Houghton, training grant awards T32HL007093 and K12CA076930 to J.B. Hiatt, University of Washington/Fred Hutch Cancer Center Support Grant P30CA015704); a Conquer Cancer Foundation Young Investigator Award to J.B. Hiatt. We acknowledge support from the Fred Hutch Flow Cytometry, Genomics and Bioinformatics and Comparative Medicine Shared Resources.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.