Generation of Genetically Engineered Mouse Lung Organoid Models for Squamous Cell Lung Cancers Allows for the Study of Combinatorial Immunotherapy

There are limited lung squamous cell carcinoma (LSCC) mouse models that recapitulate the co-occurring human LSCC mutations in genes encoding proteins operative in TP53, SOX2, PI3K, and P16(INK4a) pathways. The study of cancer genes in mouse models has traditionally relied on genetically engineered strains made via gene targeting in embryonic stem cells. Such models take months to years to establish and require complicated breeding strategies when multiple genetic alterations are needed. Moreover, unlike human lung adenocarcinomas harboring EGFR-activating mutations or ALK fusions, for which targeted inhibitors have achieved objective responses in up to 80% of cases, no targeted therapies currently exist for patients with LSCC. The extent to which LSCC mutations in these pathways contribute to tumorigenesis, shape the tumor microenvironment, and affect therapeutic responses remains unclear. Here, we describe a new rapid approach using a CRISPR–Cas9 genome multi-editing system in lung organoids derived from adult transgenic mice to generate an immunocompetent syngeneic mouse model that furthers rational immunotherapeutic options for LSCC.

Targeting tumor immune suppression pathways represents a paradigm shift in the treatment of lung cancer, which is the second most common cancer type in the United States. Despite the promising clinical activity of immune checkpoint–blocking antibodies against programmed cell death protein 1/programmed death-ligand 1(PD-1/PD-L1) for non–small cell lung cancer (NSCLC), only a minority of patients (∼20%) show a durable response (1). Thus, there is an urgent need to improve objective response rates. One strategy is to combine anti–PD-1 (pembrolizumab) with chemotherapy, which has been approved for first-line treatment of patients with squamous NSCLC (2). Increasing evidence suggests that chemotherapy leads to immunologic effects such as reduced T-regulatory cell activity, induced PD-L1 tumor expression, and enhanced cross-presentation of tumor antigens (3, 4). Chemotherapeutic efficacy relies on DNA double-stranded break (DSB) formation followed by inflammatory cytokine production to drive the killing of tumor cells over several division cycles (5). Cyclin-dependent kinases 1 (CDK1) are fundamental drivers of the cell-cycle G₂–M checkpoint and are required for the progression of various cancers (5, 6). We and others have previously demonstrated that interference with the CDK1-negative regulator WEE1 via a selective small-molecule WEE1 kinase inhibitor activates CDK1, which potently induces DSB formation due to loss of control at the G₂–M checkpoint, leading to lung cancer cell death (6). Previous studies have also shown that CDK1 can activate STAT1 signaling during mitosis, increasing proinflammatory cytokine production (5). We hypothesized that dual targeting of tumor cell–intrinsic (WEE1 inhibition) and immune cell–intrinsic (anti–PD-1) pathways may potentiate superior antitumor activity compared with monotherapies.

Here, we show that CDK1 activation via WEE1 inhibition induces DNA damage that primes the endogenous type I IFN and antigen presentation system in primary mouse and human LSCC tumor cells. We show in two mouse models, including our novel organoid-derived LSCC model, that WEE1 inhibition can enhance the antitumor activity of anti–PD-1 monotherapy by promoting cytotoxic NK-cell–mediated clearance of tumor cells and decreasing immune-suppressive neutrophilic tumor infiltration.

All mice used in this study were housed in the pathogen-free animal facilities in Dana-Farber Cancer Institute (Boston, MA). The Rosa26R-lox-stop-lox-Sox2-IRES-GFP mice (hereafter referred as SOX2) have been described previously (7) and were generously gifted by Dr. Keith Ligon's laboratory (Boston, MA). The Hipp11-lox-stop-lox-Cas9 (hereafter referred as Cas9) mice were backcrossed to C57BL/6 background (Jackson Laboratory) and then bred with SOX2 to obtain a SOX2;Cas9 colony. All breeding and care procedures were approved by the Dana Farber Animal Care and Use Committee (Protocol number: 09-073) and carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the NIH (Bethesda, MD).

After rinsing the dissected mouse lung with Hank's Salt (HBSS) supplemented with Gibco antibiotic–antimycotic, we opened the main bronchi, placed the tissue in 1 mL of dispase and incubated the tissue in a thermomixer for 15 minutes at 37°C. We next peeled the epithelia from the submucosa and incubated the epithelia in 1 mL of 0.25% trypsin-EDTA at 37°C for 10 minutes, briefly vortexed, rinsed the tissue with 8 mL of soybean trypsin inhibitor, and repeated the trypsinization and wash cycle once more. We filtered the dissociated tissue through a 40-μm cell strainer using PBS and centrifuged the cells at 200 × g for 5 minutes before resuspending the cells in organoid media (advanced DMEM/F12 media supplemented with Glutamax, 0.15 mmol/L HEPES, Gibco antibiotic–antimycotic, N2 Supplement, B27 supplement, 1 μmol/L N-Acetylcysteine, 50 ng/mL human EGF, and 3% conditioned media from L-WRN cells containing Wnt3a, Noggin, and R-spondin). Using a 1:1 ratio of organoid media and growth factor reduced basement membrane matrix (Matrigel, Corning), epithelial organoids were maintained for successive passages (>30 passages) by hanging drop method. For subsequent passages, organoids were dissociated by incubation in 0.25% trypsin. The trypsinization process disrupted the spherical organoids into cell aggregates, which were then embedded in fresh Matrigel.

Human embryonic kidney (HEK293T) cells, BEAS2B cells, and LSCC cell lines (NCI-H157, NCI-H226, and NCI-H520) were obtained from the ATCC. ATCC cells were cultured in either DMEM and RPMI1650 supplemented with 10% FBS and antibiotics_. The primary cultures of normal human bronchial epithelial HBE4 cells were previously derived (Supplementary Table S1) and cultured in keratinocyte serum free (KSF) medium (Gibco). Patient-derived xenografts were established from resected primary lung cancer without prior systemic therapy. Lung squamous carcinoma was confirmed by histology and p40 IHC, a diagnostic marker for carcinoma of squamous lineage. All patients provided written consent and all patient-derived xenograft (PDX) cell lines (PDX139, 267, and 277) were previously derived using protocols approved by the University Health Network Human Research Ethics and Animal Care Committee (8). KRAS, EGFR, and PIK3CA mutation status for PDX cell lines were detected using the Sequenom OncoCarta panel v1.0 as described previously (8) and on Supplementary Table S1. All cell lines were authenticated by short-tandem repeat fingerprinting and Mycoplasma tested by PCR.

LentiCRISPRv2 vector was obtained from Addgene (#52961). The cre-recombinase gene was first subcloned to replace the SpCas9 gene under the EFS promoter. Guide RNAs (gRNA) against mouse PTEN, CDKN2A/p16, and TP53 and E. coli LacZ gene were cloned into our modified lentiCRISPRv2-Cre vector and sequence verified (Supplementary Table S2). Lentiviral pLKO.1 vector containing shRNAs targeting WEE1 (TRCN0000226425 and TRCN0000226426) were purchased from Sigma-Aldrich. Transient transfections and virus preparation in HEK293T cells were performed using Fugene Reagents (Promega) as per the manufacturer's protocol. Lentivirus was prepared by transfecting two packaging plasmids into 293T cells using protocols from The RNAi Consortium (TRC; Broad Institute).

Organoids were isolated by digesting the Matrigel with 0.25% trypsin-EDTA in culture plates for 5–10 minutes at 37°C and washed twice with PBS. Once organoids were dissociated, cells were pelleted and resuspended in 250 μL lentiviral solution. Spinoculation was performed by transferring the suspension onto a 48-well plate and centrifuging the plate at 600 × g for 1 hour at 32°C. Plates were then incubated at 37°C for 6 hours before washing the suspension with fresh media and pelleting the cells to be embedded in fresh Matrigel media mixture by hanging drop method. Stable organoid cultures were selected by puromycin antibiotics (1–2 μg/mL).

Drugs were obtained commercially: AZD1775 (HY-10993, MedChemExpress), PD0166285 (HY-13925; MedChemExpress), Wee1_II (CAS 622855-50-9, Calbiochem), BX-795 (SML0694, Sigma-Aldrich), Carboplatin (NYU Pharmacy), InVivoMAb anti-mouse IFNAR-1 (clone MAR1-5A3, Bio X Cell), and anti-mouse CD8a (clone 53-6.7, Bio X Cell). Anti-mouse IgG and anti-mouse PD-1 (29F.1A12) were generously gifted by Dr. Gordon Freeman Laboratory (DFCI, Boston, MA).

Organoids (2 × 10⁶) were suspended in media, mixed with growth factor reduced Matrigel (Corning) at 1:1 ratio, and subcutaneously injected into female nude mice (Nu/Nu; Charles River Laboratories) in a final volume of 100 μL. Length and width were measured twice a week using calipers. Tumor volumes were calculated using the formula (Length × Width²)/2. All primary tumor cell lines were harvested from approximately 150–300 mm³ tumors, minced and digested in 4% collagenase for 30 minutes to 1 hour, then filtered through a 70 μmol/L cell strainer. Cell lines derived from mouse primary tumors were grown in advanced DMEM/F12 media supplemented with Glutamax (Invitrogen), 10% FBS (Thermo Fisher Scientific), and antibiotics. Serial passage of tumor cell lines in vivo was performed by subcutaneous implantation into the flank of female C57BL/6 mice (Jackson Laboratory) and harvested as described above. All mice included in the survival analysis were euthanized when criteria for disease burden or targeted end points were reached.

C57BL/6 and DBA/2J mice were obtained from Jackson Laboratory and subcutaneously inoculated with JH716 (3 × 10⁶) or KLN205 (0.5 × 10⁶) cells in the right flank of 7- to 8-week-old mice. Mice were examined every 3–4 days, and tumor length and width were measured using calipers. Tumor volumes were calculated using the formula (Length × Width²)/2. When tumor size reached an approximate mean of 150 or 300 mm³, mice were randomly stratified to treatment arms: vehicle, AZD1775 (20 mg/kg), anti–PD-1 (200 μg), anti-CD8a (200 ug), carboplatin (40 mg/kg), or combination therapies. JH716 mice were treated up to 30 days, while KLN205 mice were treated up to 16 days. All manipulations were performed under sterile conditions in a laminar flow hood, in accordance with procedures approved by the DFCI Animal Care and Use Committee. At sacrifice, portions of tumors were snap-frozen and stored in liquid nitrogen or were fixed in 10% buffered formalin for routine histopathologic processing.

Samples resected from mice were fixed in 10% formalin, and then stored in 70% ethanol before processing and embedding. IHC was conducted on 5-micron sections. Tissue sections were deparaffinized, rehydrated, and boiled for 15 minutes with a pressure cooker in 10 mmol/L pH6.0 Citrate buffer. Slides were incubated in 3% hydrogen peroxide for 30 minutes and then blocked in serum for 1 hour. Primary antibodies were added to the section and incubated overnight at 4°C (Supplementary Table S3). Representative sections of human and mouse lung tumors were reviewed by independent pathologists including a rodent pathologist with expertise in mouse lung pathology (Roderick T. Bronson). Patient slides were obtained from the Brigham and Women's Hospital (Boston, MA) Department of Pathology under the Institutional Review Board protocol 2014P001715 and reviewed by a pathologist (Lynette M. Sholl).

Tumor-bearing mice were sacrificed and tumors were minced and digested in collagenase D and DNase I in Hank's Balanced Salt Solution (HBSS) at 37°C for 30 minutes. After incubation, the digested tissue was filtered through a 70-μm cell strainer to obtain single-cell suspensions. Separated cells were treated with 1× RBC Lysis Buffer (BioLegend) to lyse red blood cells. Live cells were determined by LIVE/DEAD fixable Aqua Dead Cell Stain Kit (Molecular Probes). Fc receptors were blocked by incubation with the anti-mouse CD16/CD32 (Fisher Scientific). Cells were stained with antibodies listed in Supplementary Table S3 followed by fixation/permeabilization (BioLegend). Cells were read on LSR Fortessa (BD Biosciences) and analyzed with Flowjo Software (Tree Star) using gating strategies shown in Supplementary Fig. S9.

Multiplex assays were performed on the EMD Millipore mouse cytokine/chemokine 32-plex panel (MCYTOMAG-70K) utilizing the bead-based immunoassay approach. Conditioned media from ex vivo three-dimensional (3D) cultures were assayed neat. Mouse tumors were lysed in RIPA lysis buffer and protein levels were normalized in assays. Fold changes (FC) relative to the vehicle-treated tumors/Matrigel alone media were calculated and plotted as log₂FC. Lower and upper limits of quantitation (LLOQ/ULOQ) were imputed from standard curves for cytokines above or below detection.

Mouse tumoroids were processed and loaded onto microfluidic device as described previously (9). Live/Dead dual labeling was performed by loading microfluidic device with AO/PI Staining Solution (CS2-0106, Nexcelom). After 20-minute dye incubation at room temperature, images were captured on a Nikon Eclipse 80i fluorescence microscope equipped with Z-stack (Prior) and CoolSNAP CCD Camera (Roper Scientific). Image capture and analysis was performed using NIS-Elements AR software. Live and dead cell quantification was performed by measuring total cell area of each dye.

For 3D cell viability assays, tumoroids (40–100 μm) were filtered by cell strainers and embedded in 1:1 growth factor reduced Matrigel:media and seeded in 96-well spheroid microplates (Thermo Fisher Scientific) forming a hanging drop. Cell viability was measured at 7 days using CellTiter-Glo 3D (Promega) and the data were represented as the mean luminescence reading of 5–10 technical replicates relative to controls. Assays were repeated three times.

All animal experiments were performed in accordance with procedures approved by the DFCI Animal Care and Use Committee.

SOX2 is the most frequently amplified gene in LSCC (7). Lu and colleagues previously demonstrated that SOX2 acts as an oncogene in transgenic mouse models conditionally overexpressing SOX2 alone in the lung; however, penetrance in this model was only 50% and pathology of resulting tumors were lung adenocarcinoma (7). Reasoning that multiple genetic alterations are required to potentiate LSCC, we developed a transgenic mouse to enable combined overexpression of SOX2 with deletion of distinct tumor suppressors, starting by crossing mice with inducible SOX2 and Cas9 genes (two copies, ++, each; Fig. 1A). The SOX2 transgene at the Rosa26 locus and the Cas9 at the Hipp11 (H11) locus were both preceded by a stopper cassette (LSL) allowing the expression to be controlled via cre-recombinase in vitro and in vivo (Fig. 1A).

Given that LSCC is thought to originate from the basal progenitor cells, we next surgically harvested the tracheal bronchial epithelium from adult SOX2;Cas9 mice and derived organoid cultures after tissue dissociation (Fig. 1A). To determine the efficiency of conditionally overexpressing SOX2, we treated organoids with lentivirus containing cre-recombinase in vitro. As shown in Fig. 1B and C, treatment with cre induced protein expression of SOX2 and SOX2-fused GFP, as seen by immunoblot and IHC analyses. Transfected organoids express basal cell–specific markers such as cytokeratin 5 (CK5), p63, and EpCAM (Fig. 1C).

Using The Cancer Genome Atlas (TCGA) data from LSCC cohort, we next selected the top three tumor suppressor genes (TP53, PTEN, and CDKN2A/p16) most frequently comutated with SOX2 amplification in human LSCC (Fig. 1D and E). As shown in Fig. 1E, we constructed lentiviral expression vectors with sgRNAs targeting one to three distinct tumor suppressors to establish different genotypes of CRISPR-mutant organoids, herein defined as control (sgLacZ), double-mutant (sgp53;sgp16), and triple-mutant (sgPten;sgp53;sgp16). We recognized that alternative splicing of exon 1 in CDKN2A could lead to two isoforms, p16 and p19(ARF); therefore, we specifically designed our guide to disrupt both isoforms. Lentiviral infection of organoids did not alter the morphology of the organoid cultures compared with controls, but we did observe a modest growth advantage in the double- and triple-mutant organoids (Fig. 1F; Supplementary Fig. S2A). To reflect a realistic microanatomy of the tissue and to avoid potential clonal effects, our organoids were comprised of mixed multipotent progenitor cell types under puromycin selection. Sanger and whole-exome sequencing of organoids confirmed mutations in the targeted genes (Fig. 1G; Supplementary Fig. S2B; Supplementary Table S2) and immunoblotting confirmed that targeted proteins were indeed downregulated compared with control organoids (Fig. 1H). Furthermore, given that Pten is a negative regulator of PI3K signaling, we assessed whether loss of Pten resulted in upregulation of the PI3K pathway. As shown in Fig. 1H, immunoblotting showed increased phosphorylation of AKT at serine 473, indicating AKT pathway activation.

We next implanted transduced organoid cells into flanks of immunodeficient nude mice (Nu/Nu) and sequentially harvested tumors over a course of 96 days to evaluate tumor latency and disease progression (Fig. 1I and J). After 12 weeks in vivo, none of the lacZ control organoids formed tumors. The double-mutant organoids initially grew small tumors, but this growth eventually stalled. In early-stage cases, the triple-mutant organoids formed cilia, a feature seen in lung epithelia, and at week 3, tumors grew significantly and progressively larger compared with controls (Fig. 1J). Hematoxylin and eosin (H&E) staining confirmed that all the double-mutant tumors were benign, while triple-mutant tumors harvested after day 68 formed well-differentiated squamous cell carcinomas (Fig. 1J and K). SOX2 protein levels in triple-mutant tumor cell lines were indeed elevated relative to human LSCC tumor cell lines (Fig. 1L). We also observed some cases of adenosquamous cell carcinomas and well-differentiated adenocarcinomas arising from the triple mutants at earlier harvests (Supplementary Fig. S2C). We and others have previously observed that LKB1-deficient ADC can progressively transdifferentiate into SCC in mouse models (10, 11), but further studies will be needed to understand whether tumors are arising from multipotent cells that transition between ADC and SCC or through selective redifferentiation within squamous carcinoma.

To confirm the phenotype of the LSCC triple-mutant tumors, we further performed IHC for markers used clinically to distinguish human lung SCC from ADC. Similar to the human SCC samples, the LSCC triple-mutant tumors displayed positive staining of p63, CK5, and SOX2 (Fig. 1M; Supplementary Fig. S2D). Triple-mutant tumors recapitulated the hallmarks and pathology of human LSCC: squamous cells grew in solid configurations with aberrant nuclear morphology, infiltrating neutrophils, and invading stroma with prominent keratin pearls (Fig. 1K and M). We also confirmed that our LSCC model recapitulates features of the human disease transcriptionally utilizing bulk RNA sequencing (RNA-seq). Prior TCGA studies on human LSCC have shown significant alternations in innate immune transcriptional networks, including HLA gene complex regulation, that are not observed in KRAS oncogene–driven lung adenocarcinoma (11–13). Intriguingly, we found significant enrichment for innate immune and infectious disease gene sets, including viral infection response and antigen processing, in SCC tumors relative to the well-established KRAS^G12C;TP53 lung adenocarcinoma model (Supplementary Table S5; Supplementary Fig. S3A).

Cytokine profiling of conditioned media from ex vivo 3D culturing of double- and triple-mutant tumors after 72 hours revealed increased levels of neutrophil regulators (CXCL1, G-CSF, and GM-CSF) in SCC tumors compared with benign double-mutant tumors, supporting our observation of neutrophil infiltration (Supplementary Fig. S3B).

We generated a stable LSCC cell line (JH716) and orthotopically implanted these lines back into littermates (Cas9;SOX2) via intratracheal injection to generate immunocompetent LSCC models. Immune profiling analysis demonstrated a similar percentage of tumor lymphocyte infiltration compared with our KRAS-driven genetically modified mouse models, but higher levels of myeloid CD11b⁺ cell populations (Supplementary Fig. S3C). LSCC phenotype was indeed preserved in tumors derived from these models and serially passaged in vivo to generate subsequent cell lines such as JH716-18 (Supplementary Fig. S3D). Taken together, these data indicate that the triple-mutant tumors (SOX2, TP53, PTEN, and CDKN2A/p16) strongly resembled human LSCC.

Given the recent clinical response and FDA-approval for the combination of anti–PD-1 with chemotherapy as a first-line treatment of patients with NSCLC, we reasoned that the DNA repair machinery is an underlying critical component that triggers anti–PD-1 response in lung cancer. We previously demonstrated that a selective small-molecule WEE1 kinase inhibitor potently induces DNA DSB formation leading to cell death in KRAS-mutant lung adenocarcinomas. We therefore first assessed the tumor-intrinsic cytotoxic impact of WEE1 inhibition (AZD1775) in our new squamous models (JH716) and patient-derived xenograft LSCC tumor cells (PDX139 and PDX267) along with matched normal mouse fibroblasts (JH716F) or nontumorigenic human bronchial epithelial (HBE4 and BEAS2B) cells derived from primary lung tissues (Fig. 2A and B). AZD1775 reduced cell proliferation in all lines tested but to a greater extent in tumor cells compared with normal cells (IC₅₀s: 330 nmol/L vs. >1,310 nmol/L). As expected, confocal microscopy confirmed that AZD1775 induced DSB formation as indicated by high levels of DSB marker γH2AX in treated-JH716 and PDX267 cells, supporting previous studies demonstrating cell death due to accumulation of DNA damage from mitotic catastrophe (Fig. 2C). Immunoblot analyses confirmed downstream target engagement (reduced phosphorylation of CDK1 at the inhibitory tyrosine 15 site and γH2AX upregulation in a dose- and time-dependent manner) in murine JH716 and human PDX267 cells treated with vehicle (DMSO) or AZD1775 (Fig. 2D; Supplementary Fig. S4A and S4B).

Inhibition of CDK1 has been shown to block mitotic entry and STAT1 activation (5). To investigate whether WEE1 inhibition increases STAT1 signaling by inducing CDK1 activation, we measured STAT1 phosphorylation at tyrosine 701 (pSTAT1 Y701) in mouse tumor cells treated with AZD1775 over a 24-hour time course. Activation of pSTAT1 Y701 was observed in the treated AZD1775 cells after 24 hours, and this was correlated with reduced phosphorylation of CDK1 on the inhibitory residue Y15 and elevated γH2AX, suggesting that WEE1 inhibition may be promoting DSB-induced STAT1 activation via CDK1 activation (Fig. 2E). Given that STAT1 is a critical downstream transcription factor for type I IFN signaling, we next examined the expression of IFN-stimulated genes (ISG), such as IFNB1, OAS1, CCL5, IFITM1, IFIT3, CDKN1A, CCL2, ISG54, and IFNA13. Consistent with STAT1 activation, expression of ISGs was significantly upregulated in both AZD1775-treated mouse and human LSCC lines (JH716, H157, and H520; Fig. 2F and G; Supplementary Fig. S5A and S5B) compared with controls. Similar trends were observed when cells were treated with additional small-molecule WEE1 inhibitors, Wee1_II (on JH716 cells; Fig. 2F), and PD0166285 (on H157 and H226 cells; Fig. 2G and H) suggesting that these were “on-target” effects. In addition, H226 and H157 cells were transduced with lentivirus containing shRNA against WEE1 to significantly downregulate WEE1 transcript levels by 51%–93% compared with cells transfected with control vector (Fig. 2I). RNA from cells was extracted 3 days after transfection to assess ISG expression levels. WEE1 downregulation significantly correlated with elevated ISGs as shown in Fig. 2I and Supplementary Fig. S5C.

Importantly, we observed increased expression of genes encoding antigen-presenting mouse major histocompatibility complex (MHC) class I molecules (H2K1 and B2M), peptide transporters (TAP1 and TAP2), peptide cleavers (ERAP1), and transporter-MHC interactors (TAPBP) in JH716 cells after WEE1 inhibition (Fig. 2F, right). AZD1775 treatment in vitro also significantly increased expression of orthologous genes in human lung cancer lines (H157, PDX277, and H520; Fig. 2G, right; Supplementary Fig. S5A and S5B). Flow cytometry analysis showed that 72-hour treatment of AZD1775 or PD0166285 at 600 nmol/L significantly elevated expression of H-2Kb (MHC-I) by approximately 2-fold but did not change PD-L1 or I-A/I-E (MHC-II) expression levels on the surface of JH716 tumor cells (Fig. 2J; Supplementary Fig. S5D). As further support, the JH715 tumor cell line was generated from another of the cohort of 8 malignant tumors in Supplementary Fig. S2C. We confirmed JH715 cells to be a separate clonal population as it contains a different cut site in p53 (c.809G>A/p.Arg270His) compared with the earlier-used JH716 cell line. Consistent with our results in JH716 cells, we also observed increased expression of H-2Kb upon AZD1775 treatment in vitro in JH715 cells (Supplementary Fig. S5E).

Given that our in vitro data demonstrated that WEE1 inhibition can increase expression of genes encoding antigen presentation molecules on lung squamous carcinoma models, we reasoned that combining AZD1775 with anti–PD-1 may enhance tumor cell killing by stimulating T-cell activity. To test this ex vivo, we first cultured fresh tumoroids from subcutaneous implants of JH716 cells and another lung SCC mouse model, KLN205, in 3D microfluidic culture systems. Our previous studies demonstrated that tumoroids retain the autologous tumor-infiltrating immune cells in this short-term culture system (9, 14). We observed that both single-agent AZD1775 and anti–PD-1 treatment were insufficient to induce significant cell death; however, combined anti–PD-1 and AZD1775 synergistically induced cell death in both JH716- and KLN205-derived lung squamous tumoroids by 6-day cotreatment (Fig. 3A and B).

To further investigate whether the observed cell death is mediated by type-I IFN signaling, we performed a CellTiter-Glo 3D cell viability assay using neutralizing antibodies specific for mouse IFN-α/β receptor subunit 1 (IFNAR) and a TBK1 small-molecule inhibitor (BX-795), a regulator of IFN transcription (15). The addition of anti-IFNAR and BX-795 to combined AZD1775 and immune checkpoint blockade (ICB) therapy (triple combination) significantly reduced cell death of JH716 tumoroids treated with AZD1775 and immune checkpoint blockade (ICB) therapy (P < 0.0001; Fig. 3C).

We next examined the effects of WEE1 inhibition on tumor burden in vivo using a low-AZD1775 dose to minimize potential off-target effects while achieving target engagement (Fig. 3D). AZD1775 alone was not sufficient to reduce tumor growth in JH716 mice (Fig. 3E; Supplementary Fig. S6A and S6B). However, combined AZD1775 and anti–PD-1 treatment led to significant tumor regression compared with single agents or vehicle-treated control mice, showing 97% tumor growth inhibition and nearly 60% (3/5) complete tumor remissions (Fig. 3E). JH716 tumors were sensitive to anti–PD-1 single-agent treatment with 59% tumor growth inhibition, but no complete regressions were observed. Depleting anti-CD8a was administered 7 days and 1 day prior to the first anti–PD-1 dose and at day 14 during treatment. Therapeutic efficacy of the combination treatment was partially lost when CD8⁺ T cells were depleted, suggesting that while the tumor elimination phenotype is partly T-cell dependent, there are other immune modulators contributing to complete tumor elimination (n = 7; Fig. 3E). We further validated whether AZD1775 complements PD-1 blockade in another myeloid-enriched syngeneic mouse model, KLN205, that is refractory to anti–PD-1 therapy and has lower tumor immune infiltrate (CD45⁺ cells) relative to JH716 tumors (Supplementary Fig. S6C). Similar to JH716 tumors, we observed KLN205 models were unresponsive to AZD1775 or anti–PD-1 alone but showed significant tumor growth inhibition in the combined treatment relative to anti–PD-1 or AZD1775 monotherapy, although we did not observe complete regressions (Fig. 3F).

Given the potent antitumor effect of combined AZD1775 and ICB therapy, we next evaluated how this compared with chemotherapy and ICB dual therapy, which is currently FDA-approved for NSCLCs. Using a secondary cell line derived from parental JH716 tumors (denoted JH716-18; Supplementary Fig. S3D), we repeated an independent efficacy study on JH716-18 tumors starting at a volume of approximately 300 mm³ (n = 10–13/treatment arm) with the addition of the chemotherapy (carboplatin) and a combination arm (carboplatin and anti–PD-1). Survival analysis up to terminal tumor volume (20 mm length) showed significant prolonged survival of mice (with 5/13 complete tumor regressions) in the AZD1775 and anti–PD-1 combination group compared with single-agent–treated arms (Fig. 3G). Similarly, combined carboplatin and anti–PD-1 also achieved significant survival benefit with 6/13 complete tumor regressions, suggesting that WEE1 inhibition may have comparable antitumor immunity as chemotherapy in our JH716-18 models. Hematologic toxicity assays and body weight analyses showed that both combination therapies were well tolerated in mice (Supplementary Fig. S7A and S7B). We also observed that combining WEE1 inhibition with ICB therapy markedly extended survival in treated KLN205 mice (Fig. 3H). Multiplexed cytokine profiling of JH716-18 tumors after 1 week of treatment revealed that cotreatment with AZD1775 and anti–PD-1 downregulated levels of G-CSF, GM-CSF, CXCL2, and CXCL1, which are neutrophil chemoattractants, and increased levels of CCL5, which may enhance NK- and T-cell recruitment, relative to vehicle controls (Fig. 3I).

We next sought to determine how WEE1 inhibition and combined treatment may affect the quantity and subsequent recruitment of tumor-associated immune cell populations. Using flow cytometry, we analyzed the tumor-infiltrating lymphoid and myeloid cell populations at 14 days posttreatment in JH716 mice (Fig. 4A–I) and 9 days posttreatment in KLN205 mice (Fig. 4J–Q). In JH716 tumor-bearing mice, we observed a significant increase in CD8⁺ tumor-infiltrating lymphocytes (TIL) in anti–PD-1–treated tumors compared with AZD1775 alone or vehicle control (Fig. 4B). Combined treatment did not further elevate CD8⁺ TIL levels relative to anti–PD-1, but did significantly increase tumor-infiltrating NK (CD45⁺/CD3⁻/CD49b⁺) cells relative to vehicle controls (Fig. 4C). We also observed increased Ki67⁺-proliferating NK cells in tumors in the combination group compared with controls (Fig. 4D). In addition to its effects on the lymphoid populations, combined therapy significantly decreased the CD11b⁺ cells, including the tumor-associated neutrophils (CD11b⁺/MHCII⁻/Ly6G⁺/Ly6C^lo; TANs) compared with AZD1775 and isotype controls (Fig. 4E and F). However, combined therapy did not significantly change the proportions of tumor-associated monocyte (CD11b⁺/MHCII⁺/F4/80⁻/Ly6C⁺), macrophage (CD11b⁺/MHCII⁺/F4/80⁺/Ly6C⁻; TAMs) or dendritic cell (CD11b⁺/MHCII⁺/F4/80⁻/Ly6C⁻/CD11c⁺) populations in the tumor (Fig. 4G–I). In KLN205 tumor-bearing mice, we observed significantly increased CD8⁺ TILs in the combination-treated tumors compared with anti–PD-1 and isotype controls (Fig. 4J). In addition, we found significantly more proliferating NK cells in the combination group relative to the controls (Fig. 4K and L). In fact, AZD1775 alone increased NK infiltration (Fig. 4K). Consistent with JH716 tumors, combined therapy had no significant effects on inflamed monocytes or dendritic cells, but significantly reduced TANs in KLN205 tumors relative to isotype controls (Fig. 4N–Q). Taken together, these data suggest that combining WEE1 inhibition with ICB therapy enhances antitumor activity by reducing tumor-associated neutrophils in myeloid-enriched tumors, which is correlated with higher levels of cytotoxic CD8⁺ T and NK cells.

In vivo interrogation of the function of genes implicated in tumorigenesis is limited by the resource-intensive need to generate and cross mice bearing multiple, distinct engineered inducible mutant alleles. Furthermore, these models require tissue-specific cre-drivers, which can pose challenges for squamous cancers, because tamoxifen-inducible cre recombinase can lead to premature study terminations when mice develop skin lesions before developing tumors (16).

Since the Clevers group described the successful derivation of epithelial stem cell–based organoid long-term cultures from single Lgr5⁺ stem cells in 2009, a plethora of different tissue organoid culture methods have been developed (17). Emerging data suggest that 3D organoid cultures can recapitulate cellular composition, structure, and signaling characteristics better than traditional two-dimensional culture (18). Here, we demonstrate a successful organoid culture approach where LSCC is induced by CRISPR–Cas9–based multiplex editing of TP53, PTEN, and CDKN2A (p16) tumor suppressor genes in lung epithelial cells derived from a SOX2-conditionally expressing transgenic mouse. Subsequent transplantation to immunocompetent mice reproduced LSCC that phenotypically mimic the human disease. This system enabled rapid in vivo characterization of driver mutations essential to induce squamous differentiation and enabled further interrogation of immunotherapeutic combinations on a genetically defined preclinical model. Given the capacity to generate organoid models from a variety of distinct squamous tissues (19, 20), we believe this general approach will prove robust for generating squamous cell models with varied genomic features, which will powerfully enable both fundamental mechanistic and preclinical studies.

There is a clear need to develop immune-competent murine models of lung and other SCCs. Historically, squamous NSCLC treatment was mostly limited to cytotoxic chemotherapy due to the lack of targetable aberrations. In the past 5 years, clinical trials have demonstrated that inhibitors of PD-1 are effective in the treatment of squamous and nonsquamous NSCLC (21–23). In the CheckMate 017 trial, the overall survival for previously treated patients with squamous cell NSCLC was significantly better with nivolumab than with docetaxcel, regardless of PD-L1 expression levels (21).

Most recently, the landmark KEYNOTE-407 trial has revealed efficacious antitumor responses in metastatic, squamous NSCLC treated first-line with pembrolizumab and platinum-based chemotherapy combination, leading to FDA approval of the treatment (2). Accumulating evidence suggests that chemotherapy can stimulate immunosurveillance in several ways by acting on cancer cells to: (i) increase antigen presentation, (ii) induce danger signals that stimulate innate immune responses, or (iii) increase susceptibility of tumor cells to be recognized and killed by immune effector cells. Cisplatin, etoposide, and paclitaxel have all been shown to induce breast cancer cells to produce IFNβ, which in turn operates as an autocrine factor to stimulate MHC class I expression (24). Similar to the chemotherapeutics' reliance on DSB formation to drive tumor cell killing, we and others have previously shown that selective inhibition of WEE1, a negative regulator of a CDK1, leads to potent DSB formation due to exacerbated mitotic catastrophe in deregulated cancer cells (25). As such, we hypothesized that selective inhibition of WEE1 may similarly elicit tumor cell immunogenicity. Indeed, we show that AZD1775 and WEE1i treatment in vitro potently increases expression of antigen-processing and presentation genes in both mouse and patient-derived LSCC cell lines, which correlates with DSB as measured by increased γH2AX foci formation in tumor cells.

Harding and colleagues previously showed that CDK1i prevented STAT1 activation in PARPi-treated BRCA1-mutated ovarian cancer cells, suggesting that mitotic entry is a gateway to inflammatory signaling for diverse DSB-inducing stimuli (5). Consistent with this, we demonstrate that activation of CDK1 by AZD1775 treatment not only induces STAT1 phosphorylation at Y701 in a time-dependent manner, but also increases ISG expression in tumor cells, likely explaining their enhanced capacity for antigen presentation (24). Along similar lines, Goel and colleagues recently showed that CDK4/6 inhibition induced cell-cycle arrest, which resulted in global upregulation of an IFN-driven transitional program and viral mimicry on breast cancer cells (26). Although CDK1 is best known for its role in controlling the G₂–M cell-cycle checkpoint, a novel function of the kinase in immune response has emerged. For instance, Cribier and colleagues identified that CDK1 phosphorylates SAMHD1 at residue Thr592, directly inhibiting its antiviral restriction activity (27). SAMHD1 is a deoxynucleotide triphosphate (dNTP) triphosphohydrolase that reduces the intracellular pool of dNTPs to restrict HIV-1 synthesis (27). Depletion of SAMHD1 leads to enhanced antigen presentation and chronic type I IFN secretion upon HIV-1 infection and facilitates T-cell responses in coculture models (28). Further studies will be necessary to carefully interrogate whether SAMHD1 or other interacting partners of CDK1 may be mediating the IFN response upon AZD1775 treatment, rendering the tumor microenvironment more permissive to T- or NK-cell infiltration.

Nearly all humans with LSCCs have a history of tobacco use that drives squamous metaplasia and the development of an inflammatory microenvironment and chronic immune suppression (29). TANs and TAMs comprise a significant proportion of the inflammatory infiltrates in both mouse tumor models and human cancers (29). Here, we found that our orthotopic JH716 tumors were indeed enriched for tumor-infiltrating CD11b⁺ myeloid cells and inflammatory cytokines compared with a well-established KRAS^G12D lung adenocarcinoma mouse model (Supplementary Fig. S3C). Our in vivo efficacy studies demonstrate that cotreatment with anti–PD-1 and AZD1775 significantly reduces tumor growth and TAN infiltration in JH716 and KLN205 tumors. We found that treatment with AZD1775 in JH716 mice resulted in reduced levels of neutrophil chemoattractant G-CSF. Notably, the addition of anti–PD-1 therapy with AZD1775 further reduced intratumoral levels of G-CSF and CXCL2. As type I IFNs have been shown to downregulate G-CSF expression and G-CSF is known to be a major regulator of neutrophil development and mobilization and CXCL2 secretion, we speculate that reduced neutrophil chemoattractants may be partly due to DSB-stimulated type I IFN signaling upon AZD1775 treatment (30).

Previous studies have shown that TANs can suppress T-cell responses and inhibit NK-cell function, establishing a protumor microenvironment that facilitates metastatic formation (31). Consistent with these results, we observed an inverse correlation between TAN abundance and increased CD8⁺ T and NK cells in our JH716 and KLN205 tumors treated with dual therapy. In agreement with our data, De Henau and colleagues showed that targeting myeloid cells via PI3Kγ inhibition enhanced immunotherapeutic effect of anti–PD-1 in resistant tumors (32). Given that emerging data suggest that neutrophils contribute to anti–PD-1 resistance, further studies are needed to delineate the mechanistic basis of how combined anti–PD-1 and AZD1775 reduce TAN infiltration.

Previous studies have demonstrated that CDK inhibition prevents death of terminally differentiated neurons, suggesting that noncycling cells such as neutrophils may require activation of CDK for apoptotic induction (33, 34). In contrast, others have shown that human neutrophils express CDK1, CDK2, and CDK5, and that specific inhibitors of CDKs (namely, R-roscovitine) induce caspase-dependent apoptosis (35). CDK activity is clearly important to neutrophil survival (34, 35). Thus, it would be interesting to examine whether AZD1775-induced DNA damage directly mediates neutrophil homeostasis by either deregulating signaling of its trafficking receptors (i.e., CXCR2) or apoptotic response.

In addition to the reduction of TANs, we also observed that AZD1775 combined with ICB therapy led to increased NK-cell infiltration. Our cytokine profiling data showed elevated intratumoral CCL5 levels in cotreated JH716 samples, which may partially explain the heightened NK-cell influx. Recently, Böttcher and colleagues demonstrated that cDC1 accumulation in mouse tumors often depends on NK cells that produce cDC1 chemoattractants CCL5 that augment immunotherapeutic response (36). Indeed, we also observed a positive correlation between intratumoral CCL5 chemokine levels and NK infiltrates. However, while we observed a trend of increased tumor-associated DCs (CD11b⁺/MHCII⁺/F4/80⁻/Ly6C⁻/CD11c⁺) compared with controls in JH716 tumors, this did not reach significance. Further NK-cell depletion studies are needed to delineate whether NK cells are directly mediating the antitumor effects of combined therapy or indirectly stimulating recruitment of cDC1. It is also tantalizing to speculate that we may be observing NK-cell–mediated antitumor immunity due to WEE1 inhibitor–mediated DNA damage and subsequent STING and type I IFN pathway activation, as has been recently shown (37). Such a mechanism would be consistent with our observed NK-cell accumulation and increased proliferation in our AZD1775-treated tumor models.

We have pioneered the development of ex vivo genomic editing of mouse 3D organoid cultures to generate a rapid, novel LSCC mouse model that exhibits characteristic histopathology and biomarker expression similar to human LSCC. We generated two cell lines, JH715 and JH716, harboring SOX2;PTEN;CDKN2A;TP53 alterations. JH715 cells differ from JH716 by the TP53 cut site, but also exhibited increased expression of H-2Kb upon AZD1775 treatment in vitro, which we believe supports the WEE1-specific immunologic mechanism described in significant depth utilizing JH716 cells. We have also applied this system to explore different genotypes of human LSCC, but have observed that single or double alterations of TP53, CDKN2A, or KDM6A in concert with SOX2 overexpression was not sufficient to form LSCC tumors, therefore limiting generation of additional mouse cells from these specific genotypes (Fig. 1I; Supplementary Fig. S8A–S8D). Our work illustrates the challenges in understanding which cooperating mutations potentiate LSCC, because not all genotypes lead to LSCC in mice. Consistent with our data, Bern's group has also demonstrated that various combinations of genetic lesions can determine different cell fates and showed that SOX2;PTEN;CDKN2A transgenic mice gave rise to LSCC tumors in 7–9 months with a 73% penetrance, while FGFR;PTEN;CDKN2A mice had 19% SCC penetrance (38). Sequencing of JH716 tumors also revealed mutations in genes not known to drive LSCC (data not shown). We believe that the combined effects of deleting genes in the DNA damage repair/cell-cycle machinery during long term in vivo culturing leads to accumulation of passenger mutations that help facilitate rapid LSCC tumorigenesis. Yet another possibility is the potential of Cas9-induced immunity or off-target effects, which continues to be actively investigated in this field because the full impact of preexisting immunity has yet to be fully determined (39).

Using this mouse model, we explored a rational therapeutic combination to enhance response to PD-1 blockade by stimulating immunosurveillance with a selective DSB-inducing agent. We demonstrated for the first time that CDK1 activation via WEE1 inhibition can induce DSB-stimulated type I IFN signaling and antigen presentation in mouse and patient-derived LSCC tumors. Combined WEE1 and PD-1 inhibition significantly induced antitumor efficacy both ex vivo and in vivo, consistent with the efficacious response observed in mice treated with carboplatin and anti–PD-1 cotreatment. Although we did observe some inhibitory activity in nontransformed cells treated with AZD1775, we believe that cancer cells with high replicative stress are more vulnerable, and therefore we speculate that selectively activating CDK1 by AZD1775 may minimize the off-target side effects associated with chemotherapy while augmenting antitumor immunity. The clinical challenge will be to assess dose-limiting toxicity associated with the promiscuous nature of kinase inhibitors such as AZD1775. In fact, phase I trials (NCT00648648) of AZD1775 and chemotherapy showed responding patients with advanced solid tumors were mildly enriched for p53 mutations (response rate of 21% and 12% in p53 mutant and p53 wild-type, respectively), supporting the notion that patient selection may be necessary to obtain the maximum benefit from AZD1775 and ICB combination therapy (40). Yet another possibility is that relapsed p53-mutant tumors relying on G₂–M checkpoint may be more responsive to AZD1775 and ICB combination therapy than tumors with intact G₁ checkpoint. Future studies may also evaluate whether triple combinations such as AZD1775, PARPi, and ICB therapies targeting both G₁, G₂, and immune checkpoints may reach improved efficacy. Our observations illuminate a clinical hypothesis for combining anti–PD-1 with DNA damage–inducing therapies in the treatment of LSCC. As with all preclinical model systems, prospective clinical trials will be needed to substantiate the hypothesis of our work.

L.M. Sholl is an employee/paid consultant for AstraZeneca, LOXO Oncology, Foghorn Therapeutics, and EMD Serono. A.J. Bass reports receiving commercial research grants from Novartis, Bayer, and Merck. K.-K. Wong is a founder and equity holder of G1 Therapeutics and he has consulting/sponsored research agreements with AstraZeneca, Janssen, Pfizer, Array, Novartis, Merck, Takeda, Ono, Targimmune, and Bristol Myers Squibb. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Hai, A.J. Bass, K.-K. Wong

Development of methodology: J. Hai, J. Zhou, Z. Wu, T. Chen, J.B. Liu, A.J. Bass

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Hai, H. Zhang, J. Zhou, Z. Wu, E. Papadopoulos, C.M. Dowling, V. Pyon, Y. Pan, H. Silver, P.H. Lizotte, M.-S. Tsao, C. Thakurdin, K.-K. Wong

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Hai, H. Zhang, C.M. Dowling, Y. Pan, R.T. Bronson, H. Silver, P.H. Lizotte, J. Deng, J.D. Campbell, L.M. Sholl, K.-K. Wong

Writing, review, and/or revision of the manuscript: J. Hai, H. Zhang, P.H. Lizotte, L.M. Sholl, A.J. Bass, K.-K. Wong

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Hai, T. Chen, Y. Pan, J.B. Liu, C. Ng

Study supervision: J. Hai, A.J. Bass, K.-K. Wong

Others (pathology for this study): R.T. Bronson

The authors thank Mei Zhang (BWH) for help with IHC work, Yanxi Zhang (DFCI) and Xiaoen Wang (DFCI) for technical animal assistance, Dr. Elena Ivanova (Belfer Center) for technical help and protocols for staining 3D cultures, and Dr. Israel Canadas (DFCI) for thoughtful discussions about innate pathway signaling. The authors thank Drs. Joe Wang, Limei Ouyang, and Elinor Sun from the BGI Group for whole-exome sequencing and RNAseq analysis.

This work was supported by the NCI [U01 CA233084-01 (co-principal investigator), CA213333 (principal investigator), R01 CA219670-A1 (co-principal investigator), CA216188-01A1 (co-principal investigator), CA222218-01A1 (co-investigator), CA205150 (co-investigator), CA201049 (co-investigator), CA197329 (co-investigator), CA140594 (principal investigator), CA166480 (co-investigator), P01 CA098101 (principal investigator), and CA154303 (co-investigator)] to K.-K. Wong and CA187119 to A.J. Bass.

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.