EGFR inhibition in EGFR-mutant lung cancer cells perturbs innate immune signaling pathways in the tumor microenvironment

Abstract

Epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors (TKIs) elicit potent cell cycle arrest in EGFR-mutant non–small-cell lung cancer (NSCLC) cells. However, little is known about the mechanisms through which these drugs alter the tumor phenotype that contributes to the immune escape of EGFR-mutant cells. Using EGFR-mutant NSCLC cell lines and tissue samples from patients, we investigated the changes in immune checkpoints expressed in tumor cells following EGFR inhibition. Subsequently, we also analyzed the role of soluble factors from the dying tumor cells in the activation of immune signaling pathways involved in therapy resistance. Upon EGFR-TKI treatment, we found that EGFR-mutant cells upregulated the expression of innate immune checkpoint CD24 in vitro. We then analyzed biopsy samples from six patients who developed resistance to a first-generation EGFR-TKI without the acquired T790M mutation. Immunohistochemistry revealed that levels of tumor CD24 expression were increased upon treatment compared with those from pre-treatment samples. Monocyte-derived macrophages facilitated antibody-dependent cellular phagocytosis when EGFR-TKI-treated EGFR-mutant cells were incubated with anti-CD24 antibodies in vitro, suggesting that CD24 may be a therapeutical target for EGFR-mutant lung cancer. Moreover, EGFR inhibition accelerated the release of cell-free DNA (cfDNA) from dying tumor cells, which activated the type I interferon signaling pathways in human THP-1 monocytes in a stimulator of interferon genes-dependent manner. Our study indicates that EGFR inhibition in EGFR-mutant NSCLC cells fosters a tumor microenvironment associated with immune escape. Thus, CD24 targeted therapy and cfDNA monitoring may contribute to improved treatment outcomes in patients with EGFR-mutant NSCLC.

Abbreviations
• ADCP
• Antibody-dependent cellular phagocytosis
• APC
• Antigen-presenting cells
• cfDNA
• Cell-free DNA
• EGFR
• Epidermal growth factor receptor
• EMT
• Epithelial–mesenchymal transition
• GM-CSF
• Granulocyte–macrophage colony-stimulating factor
• IFN
• Interferon
• IL
• Interleukin
• IRF
• Interferon regulatory factor
• M-CSF
• Macrophage colony-stimulating factor
• NF-κB
• Nuclear factor kappa B
• NSCLC
• Non–small-cell lung cancer
• RIG-I
• Retinoic acid-inducible gene I
• STAT
• signal transducer and activator of transcription
• STING
• Stimulator of interferon genes
• TBK1
• TANK binding kinase 1
• TGF-β
• Transforming growth factor-β
• TKI
• Tyrosine kinase inhibitor
• TRIM32
• Tripartite motif containing 32
• VEGF
• Vascular endothelial growth factor

INTRODUCTION

Lung cancer is the leading cause of cancer-related death worldwide. The advent of next-generation sequencing technology has unveiled the underlying biology of lung cancer, leading to the development of gene-targeted therapies that have contributed to improved survival. Mutations in the gene encoding epidermal growth factor receptor (EGFR) are common in NSCLC and are therapeutic targets of small-molecule TKIs.¹ Overall, 30%–40% of Asian and 10%–20% of Caucasian patients with advanced NSCLC harbor activating mutations in the tyrosine kinase domain of EGFR.² Following the application of first- and second-generation EGFR-TKIs, the use of osimertinib, a third-generation EGFR-TKI that confers the acquired EGFR T790M resistant mutation, is the standard of care for patients with untreated advanced EGFR-mutant NSCLC.³ However, it remains challenging that the majority eventually become resistant even though the median progression-free survival reaches 18.9 months with osimertinib, as evidenced by a phase III FLAURA study.⁴ Genomic approaches have been extensively used to elucidate the mechanisms of resistance to osimertinib. These approaches have revealed acquired resistance mutations, amplifications, and fusions in approximately half of these patients.⁵ Ongoing research has focused on developing new agents and novel combination therapies targeting these pathogenic variants.

Exploring mechanisms in resistance other than genomic alterations, some studies have suggested the substantial role of the immune system either in growth inhibition or the promotion of EGFR-mutant NSCLC cells. Preclinical mouse models have shown a synergy between EGFR inhibition and T-cell checkpoint inhibitors.⁶ Nevertheless, neither concurrent nor sequential combination therapy is feasible in humans owing to safety concerns in clinical trials, including liver toxicity and interstitial pneumonitis.^7–9 Using biopsy samples prior to therapy, Takashima et al. reported that primary resistance to EGFR-TKIs is associated with the density of pre-existing tumor-infiltrating CD8⁺ T cells.¹⁰ These findings indicate that the immune escape of EGFR-mutant cells from T-cell immunity may contribute to resistance to EGFR-TKIs. In addition, innate immune cytokines, such as IFN-α and IFN-β, from EGFR-mutant NSCLC cells following EGFR inhibition reportedly enhance resistance to EGFR-TKI inhibition via STAT1 in an autocrine manner.¹¹ The RIG-I–TRIM32–TBK1–IRF3 innate immune signaling pathway in tumor cells is involved in the induction of type I IFNs.¹¹ Together with these adaptive and innate immune responses, proinflammatory cytokines contribute to the EMT in EGFR-mutant NSCLC cells, and promote their resistance to EGFR-TKIs.¹² Therefore, understanding immunological changes in the EGFR-mutant tumor microenvironment under EGFR inhibition is essential for the development of novel therapeutic strategies to overcome therapy resistance.

In this study, we investigated whether: (1) EGFR-TKIs intrinsically alter the expression levels of immune checkpoints, which promote the immune escape of EGFR-mutant NSCLC cells; and (2) certain exogenous factors from dying tumor cells retain the ability to stimulate immune signaling pathways associated with their therapeutic resistance following treatment.

MATERIALS AND METHODS Cell lines

The lung cancer cell line, PC9, was obtained from the European Collection of Authenticated Cell Cultures. H1975 and H522 cells were obtained from the ATCC. RERF-LC-Ad1 cells were obtained from the Japanese Collection of Research Bioresources Cell Bank. Cells were maintained in vitro in RPMI 1640 medium (Thermo Fisher Scientific) supplemented with 10% FCS (NICHIREI K.K.) and penicillin/streptomycin (FUJIFILM Wako Pure Chemical Corporation). All cell lines were routinely checked for mycoplasma contamination (LONZA, Rockland).

Patients

Tissue samples from 35 patients with NSCLC were analyzed using immunohistochemistry. EGFR mutations were identified in tumor tissues using the peptide nucleic acid-locked nucleic acid polymerase chain reaction clamp method or the Scorpion amplification refractory mutation system method. To measure cfDNA in the blood, plasma samples from four patients with EGFR-mutant NSCLC were analyzed. Patient characteristics are described in Tables S1, S2 and S4. All the participants provided written informed consent. This study was approved by the Institutional Review Boards of Hokkaido University Hospital (protocol codes: 019–0261, 019–0468, 020–0346) and Shinshu University Hospital (protocol code: 611).

Isolation of cfDNA and THP-1 monocyte experiments

Peripheral blood (20 ml) was collected in EDTA tubes. Blood samples were centrifuged at 400 g for 10 min at 20°C. Thereafter the plasma was then stored at −80°C. cfDNA in the culture supernatant and plasma was collected using a MagMAX Cell-Free DNA Isolation Kit (Thermo Fisher Scientific), according to the manufacturer's instructions. The amount of isolated cfDNA was measured using a Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific). To detect the activation of innate immune signaling pathways by cfDNA, we used THP1-Dual and STING-knockout THP1-Dual monocytes (InvivoGen, San Diego). Here, 1 × 10⁵ THP-1 monocytes were incubated with cfDNA at a concentration of 50 ng/ml for 24 h in a well of a flat-bottomed 96-well plate. The cfDNA was treated with Lipofectamine 3000 (Thermo Fisher Scientific) prior to incubation. The IRF and NF-κB activities were measured using GloMax (Promega K.K.) and a microplate reader set to 655 nm, respectively. Next, 3 × 10⁵ THP-1 monocytes were differentiated with PMA at a concentration of 150 nM for 24 h in a well of a flat-bottomed 24-well plate. Cells were pretreated either with lipofectamine, poly(dA:dT) (100 ng/ml), PC9 cfDNA (100 ng/ml)/lipofectamine, or IFN-α2a (20,000 U/ml) for 24 h and then treated with IFN-γ (100 ng/ml) in the presence of monensin for 24 h. The expression of CXCL9 and CXCL10 was analyzed by flow cytometry.

In vitro tumor growth inhibition by EGFR-TKIs and paclitaxel

On day 0, 2 × 10⁵ tumor cells were incubated in a T25 flask. Cells were treated either with 20 nM osimertinib (Cayman Chemical), 200 nM gefitinib (FUJIFILM Wako Pure Chemical Corporation), 400 nM afatinib (Selleck Chemicals LLC), or 3.5 nM paclitaxel (Cayman Chemical) for 72 h. Tumor cells and culture supernatants were collected and used for subsequent experiments.

Flow cytometry

The following monoclonal antibodies, purchased from BioLegend, were used for flow cytometry: FITC-conjugated anti-CD80 (clone 2D10), phycoerythrin (PE)-conjugated anti-CD80 (clone 2D10), PE-conjugated anti-CD24 (clone ML5), PE-conjugated anti-CD47 (clone CC2C6), PE-conjugated anti-CXCL9 (clone J1015E10), PE-conjugated anti-CXCL10 (clone J034D6), PerCP/Cyanine5.5-conjugated anti-CD11b (clone M1/70), allophycocyanin (APC)-conjugated anti-CD11b (clone M1/70), APC-conjugated anti-Siglec-10 (clone 5G6), and APC-conjugated anti-CD163 (clone GHI/61). Dead cells were stained with Po-Pro-1 (Thermo Fisher Scientific) or Zombie Aqua (BioLegend). The numbers of cell surface CD24 and CD47 molecules were calculated using Quantibrite PE beads, according to the manufacturer's instructions (BD Biosciences, San Jose, CA). Briefly, BD QuantiBrite PE beads used in the study were designed for use with PE-conjugated monoclonal antibodies for the purpose of estimating antibodies bound per cell by flow cytometry. Stained cells were analyzed using FACS Canto II (BD Biosciences) and FlowJo (BD Biosciences).

Generation of monocyte-derived macrophages and phagocytosis assays

Peripheral blood mononuclear cells (PBMCs) were isolated from blood using Ficoll-Paque Plus (Sigma-Aldrich, Inc.) and stored in Bambanker medium (Nippon Genetics Co., Ltd.) in liquid nitrogen. CD14⁺ monocytes were isolated from PBMCs using CD14 MicroBeads (Miltenyi Biotec GmbH), according to the manufacturer's instructions. Their purity was >90%, as evidenced by flow cytometry. To generate monocyte-derived macrophages, isolated monocytes were cultured according to the protocols listed in Table S3. Differentiated macrophages were detached from the plate by incubating in PBS supplemented with 2.5 mM EDTA for 30 min on ice. For phagocytosis assays, the target PC9 tumor cells treated with 20 nM osimertinib for 72 h were stained with 150 nM Calcein-AM (DOJINDO LABORATORIES) for 20 min on ice in the dark. Thereafter, 0.1 × 10⁶ target tumor cells and 5 × 10⁴ M-CSF differentiated macrophages were incubated for 9 h at 37°C in a 1.5 ml Eppendorf tube with azide-free anti-CD24 antibodies (GeneTex, clone SN3) or isotype control antibodies (clone MOPC-21) at a concentration of 10 μg/ml. We used cytochalasin D (Sigma-Aldrich) to inhibit macrophage phagocytosis at a concentration of 5 μM.¹³ Macrophage phagocytosis of the target tumor cells was analyzed by flow cytometry.

Gene expression analysis

Total RNA was extracted using the Qiagen RNeasy Mini Kit (Qiagen GmbH, Hilden) and reverse-transcribed using the PrimeScript II 1st strand cDNA Synthesis Kit (Takara Bio Inc.). To analyze CD24 expression by qRT-PCR, a forward primer (5′-CTCCTACCCACGCAGATTTATTC-3′) and a reverse primer (5′-AGAGTGAGACCACGAAGAGAC -3′) were used. To analyze CD47 expression by qRT-PCR, a forward primer (5′-AGAAGGTGAAACGATCATCGAGC-3′) and a reverse primer (5′-CTCATCCATACCACCGGATCT-3′) were used.

For the analysis of microarray datasets, we obtained normalized expression data matrices from NCBI GEO (http://www.ncbi.nlm.nih.gov/geo). Differentially expressed genes were analyzed using the Bioconductor limma package in R. To identify changes in molecules expressed on the cell surface, we filtered the genes that matched those in the Cell Surface Protein Atlas¹⁴ or represented the cluster of differentiation. The processed data were visualized as volcano plots using ggplot2.

Estimation of the fraction of cell types in the tumor microenvironment of NSCLC

We used the processed cell fraction data from the TCGA-LUAD using the Kassandra deconvolution algorism.¹⁵ Cell types representing “B_cells,” “CD8_T_cells_PD1_high,” “CD8_T_cells_PD1_low,” “T_helpers,” “Tregs,” “NK_cells,” “Macrophages_M1,” “Macrophages_M2,” “Monocytes,” “Neutrophils,” “Endothelium,” “Fibroblasts,” and “Other” were subject to analysis. The data from normal tissues were removed.

CD24 transcriptional levels between EGFR-wild-type and EGFR-mutant NSCLC

We used the processed single nucleotide variation, gene expression, and clinical data from the TCGA-LUAD dataset.¹⁶ CD24 expression levels in stage III and IV patients were analyzed based on their EGFR mutation types.

Immunohistochemistry

Formalin-fixed tissue sections (4 μm) were deparaffinized in xylene and rehydrated in a graded series of alcohol and distilled water. Antibodies against CD24 (clone SN3) were purchased from Thermo Fisher Scientific. Immunostaining was performed by VENTANA BenchMark ULTRA (Roche Diagnostics). Immunohistochemical staining for CD24 was assessed by the percentage of stained cells per slide (0%–100%) and by the staining intensity: negative, weak, intermediate, and strong. As inflammatory cells such as mononuclear cells and neutrophils frequently showed intermediate expression of CD24, immunostaining of CD24 was interpreted as strong when tumor cells were stained stronger than inflammatory cells. Immunostaining was interpreted as score 3 when more than 20% of tumor cells exhibited strong staining. Cases with >20% of tumor cells showing intermediate intensity to <20% of tumor cells showing strong intensity were evaluated as score 2. Cases with weak to intermediate CD24 expression below score 2 were interpreted as score 1. Cases with negative for CD24 were evaluated as score 0. The patients’ backgrounds and therapeutic histories were blinded to the expression assessment by a pathologist (U.T.).

Statistical analysis

Prism 9.0 (GraphPad Software, Inc.) was used for statistical analysis. For two group comparisons, Welch's t-test was used. For multiple group comparisons in Figure 1C,D and ; Figure S1B,C, p-values were adjusted by the two-stage step-up method of Benjamini, Krieger, and Yekutieli following Welch's t-test (i.e., q-values). In Figure 3D, three groups were analyzed using one-way ANOVA followed by Dunnett's test. In Figure 4E, the differences in CXCL9⁺ and CXCL10⁺ cells were analyzed using one-way ANOVA. For survival analysis, we used the log-rank test. A p-value <0.05 or a q-value <0.01 was considered statistically significant.

FIGURE 1. CD24 expression is induced in EGFR-mutant NSCLC cells upon EGFR-TKI treatment. (A) Differential gene expression between EGFR-TKI-treated versus -untreated EGFR-mutant NSCLC cells in GSE75308 (left) and GSE57156 (right) were analyzed. (B) EGFR-mutant PC9 and H1975 cells, and EGFR-wild-type RERF-LC-Ad1 and H522 cells were treated with osimertinib for 72 h in vitro. CD24 expression was analyzed by flow cytometry. Gray: isotype control; dotted line: untreated; line: treated. Data are representative of four independent experiments. (C) Numbers of CD24 molecules expressed on cells with or without EGFR-TKI treatment in vitro. Data are presented as mean ± SEM from four independent experiments. Numbers of the receptors were calculated using the BD QuantiBrite kit as described in Materials and Methods. (D) CD24 gene expression in tumor cells with or without EGFR-TKI treatment in vitro were analyzed by qRT-PCR. Data are presented as mean ± SEM of technical replicates from at least two independent experiments. (E) EGFR-mutant PC9 and EGFR-wild-type H522 cells were treated either with osimertinib, gefitinib, or afatinib for 72 h in vitro. CD24 expression was analyzed by flow cytometry. Data are representative of two independent experiments. *q [less than] 0.01.

RESULTS EGFR-mutant lung cancer cell lines upregulate CD24 upon EGFR inhibition

To oversee transcriptional changes, we first examined two public datasets in which EGFR-mutant lung cancer cells were treated with or without an EGFR-TKI. Briefly, EGFR-mutant PC9 cells were treated with gefitinib in GSE75308 and EGFR-mutant HCC2279 cells were treated with erlotinib in GSE57156. The genes coding for cell surface proteins were filtered using the Cell Surface Protein Atlas¹⁴ and the cluster of differentiation, and were analyzed thereafter. We found that transcription of innate immune checkpoint CD24 was significantly upregulated in EGFR-TKI-treated cells in both datasets (Figure 1A). Following the analysis of the public database, we assessed the changes in expression in EGFR-mutant PC9, H1975, EGFR-wild-type RERF-LC-Ad1, and H522 lung cancer cells by flow cytometry and qRT-PCR. In these experiments, we chose the third-generation EGFR-TKI osimertinib because H1975 cells harbored the acquired T790M mutation, which was resistant to first- and second-generation EGFR-TKIs. When these cells were treated with the drug in vitro, CD24 expression was significantly upregulated in EGFR-mutant cells but not in EGFR-wild-type cells (Figure 1B–D). Furthermore, both first- and second-generation EGFR-TKIs, such as gefitinib and afatinib, similarly induced CD24 expression in EGFR-mutant PC9 cells that harbored the deletion of exon 19 but not the T790M mutation (Figure 1E). Although we also analyzed changes in CD47 expression, another inhibitory innate immune checkpoint, we did not observe strong evidence of CD47 upregulation following EGFR inhibition in EGFR-mutant tumor cells, considering the limited effect size. (Figure S1A–C). These data demonstrate that blocking EGFR downstream signaling is associated with the upregulation of CD24 expression in EGFR-mutant NSCLC cells.

EGFR-mutant NSCLC cells retain augmented CD24 expression in patients treated with EGFR-TKIs

Increased levels of tumor CD24 expression in NSCLC are associated with a poor prognosis¹⁷ and an attenuated response to immune checkpoint blockade.¹⁸ We first investigated whether CD24 expression prior to EGFR inhibition could stratify the treatment efficacy of EGFR-TKIs in patients with EGFR-mutant NSCLC in our cohort (Table S1). However, there was no statistical difference between patients with high CD24 expression and those with low expression in progression-free survival (Figure 2A,C). In the wake of in vitro observations, we next hypothesized that the use of CD24 expression prior to the therapy was inappropriate for stratification because the levels were upregulated upon EGFR inhibition. To compare the changes in tumor CD24 expression before and after EGFR inhibition, we analyzed specimens from six patients treated with first-line gefitinib or erlotinib in the cohort. All of them were negative for the acquired T790M mutation at relapse. We used tumor samples from three patients with EGFR-wild-type NSCLC before and after standard cytotoxic chemotherapy as controls (Table S2). Immunohistochemistry revealed that levels of CD24 expression in tumor cells were substantially increased after treatment compared with those in pre-treatment samples in five EGFR-mutant patients (83%) (Figure 2B,C). In contrast, the signal intensity remained unaffected in the EGFR-wild-type control cohort treated with cytotoxic chemotherapy (Figure 2B,C). Levels of CD24 expression between EGFR-wild-type and EGFR-mutant were comparable in advanced-stage NSCLC patients from the TCGA-LUAD dataset (Figure 2D). Therefore, in parallel with our observations in vitro, we showed that EGFR inhibition also resulted in upregulation of tumor CD24 expression in patients with EGFR-mutant lung cancer.

FIGURE 2. Tumor CD24 expression is augmented in patients with EGFR-mutant NSCLC treated with EGFR-TKIs. (A) Progression-free survival in patients treated with EGFR-TKIs is shown. Tumor CD24 expression score: None = 0, weak = 1, intermediate = 2, strong = 3. (B) Tumor CD24 expression scores in pre-treatment and post-treatment samples of patients (Pt1–Pt9) treated with first-generation EGFR-TKIs or chemotherapy are shown. (C) Representative tumor CD24 expression scores are shown. (D) CD24 transcriptional levels between advanced-stage EGFR-wild-type (n = 87) and EGFR-mutant NSCLC patients (n = 18) in the TCGA-LUAD dataset. ns, not significant.

CD24 induction in EGFR-mutant NSCLC cells is a target of antibody-dependent cellular phagocytosis

CD24 is an innate immune checkpoint that inhibits macrophage phagocytosis via Siglec-10 in humans, and Siglec-G in mice.¹⁹ CD24 antibody blockade enhances macrophage phagocytosis against CD24-expressing tumor cells in breast and ovarian cancer models.²⁰ In advanced-stage and recurrent NSCLC, tumor-associated macrophages (TAMs) are relatively abundant over other types of immune cells in the tumor microenvironment¹⁵ (Figure 3A; Figure S2A). To assess the relevance of Siglec-10 expression in TAMs in NSCLC, we re-analyzed the public scRNA-seq dataset.²¹ Our investigation revealed that SIGLEC-10 expression was substantially heterogeneous across human immune cell subtypes in the tumor microenvironment (Figure S2B). Macrophages with the M-0 phenotype (hMϕ1, hMϕ2, and hMϕ4) retain low levels of SIGLEC-10 compared with those with the M2-like gene signature (hMϕ8, and hMϕ9)²¹ (Figure 3B). In fact, using monocytes from three healthy donors, we demonstrated that human macrophages that were differentiated either with GM-CSF or M-CSF (M-0 macrophages) retained no Siglec-10 expression on the cell surface (Figure 3C). Moreover, Siglec-10 expression was restricted in macrophages exposed only to M-CSF, IL-10, and TGF-β in several differentiation protocols (Figure 3C; Figure S2C,D and Table S3). Taken together, only a subset of TAMs in NSCLC presumably express Siglec-10 on the cell surface, such as hMϕ8, and hMϕ9. Given the heterogeneity of Siglec-10 expression in TAMs in NSCLC, one might be concerned that patients with EGFR-mutant NSCLC would not benefit from an anti-CD24 antibody if its companion biomarker was only Siglec-10 expression in macrophages. To broaden the potential application of an anti-CD24 antibody in EGFR-mutant NSCLC, we hypothesized that CD24 could be utilized as a target of the antibody itself that elicits ADCP or that carries cytotoxic drugs, so-called antibody–drug conjugates. Accordingly, we investigated ADCP between EGFR-TKI-treated PC9 tumor cells and M-CSF differentiated human monocyte-derived macrophages, with or without anti-CD24 antibodies. The results showed that anti-CD24 antibodies enhanced the phagocytosis of EGFR-TKI-treated EGFR-mutant lung cancer cells (Figure 3D; Figure S3). Moreover, our results demonstrated that an anti-CD24 antibody exerted antitumor phagocytosis without the pre-existing CD24–Siglec10 immune escape axis because phagocytic macrophages differentiated only with M-CSF (M-CSF macrophages) lacked Siglec-10 expression (Figure 3C) but retained CD163 expression (Figure S2D).

FIGURE 3. CD24 induction in EGFR-mutant NSCLC cells is a target of antibody-dependent cellular phagocytosis. (A) The fraction of immune cell types in the tumor microenvironment of NSCLC from the TCGA-LUAD dataset (B) Arbitrary gene expression levels of SIGLEC-10 and MRC1 across macrophages in the tumor microenvironment of NSCLC (C) Siglec-10 expression was analyzed by flow cytometry. Gray: isotype control, line: Siglec-10. Human monocyte-derived macrophages were differentiated according to the protocols shown in Supplementary Table 3. Data are a representative of at least two independent experiments from three healthy donors. (D) M-CSF differentiated macrophages and EGFR-TKI-treated PC9 cells were incubated for 9 h with anti-CD24 antibodies or isotype control antibodies. Cytochalasin D was used to inhibit macrophage phagocytosis as described in Materials and Methods. Data are presented as the mean ± SEM of technical replicates from at least two independent experiments. ***p [less than] 0.001.

EGFR-TKIs accelerate the release of cell-free DNA activating type I IFN response in a STING-dependent manner

In parallel with the intrinsic cellular CD24 upregulation, we subsequently investigated whether EGFR-mutant lung cancer cells gave rise to soluble factors that exogenously stimulated the local immune microenvironment. EGFR-TKIs elicit potent growth arrest in EGFR-mutant cells, so that the response benefits the majority of patients. Indeed, in the phase III FLAURA trial, 80% of the patients showed objective responses when treated with osimertinib.⁴ These clinical observations suggest that dying tumor cells under EGFR inhibition actively release certain cellular components into the local tumor microenvironment. To address these points, we used our in vitro culture system to recapitulate the immediate tumor growth inhibition by EGFR-TKIs and cytotoxic agents such as paclitaxel (Figure 4A). Using EGFR-mutant PC9 and H1975 cells, we found that the release of cfDNA in the culture supernatant was substantially accelerated upon EGFR-TKI treatment compared with paclitaxel (Figure 4B). Extracellular double-stranded DNA elicits an immunological response via the DNA sensor stimulator of interferon genes (STING).²² Accordingly, we examined whether cfDNA from dying EGFR-mutant cells retained a similar mechanism. Human THP-1 monocytes activated the IRF pathway significantly in a STING-dependent manner when they took up cfDNA from EGFR-TKI-treated tumor cells (Figure 4C). The NF-κB pathway, which is partially involved with DNA sensing, was also marginally activated (Figure 4C). We then interrogated the role of the IRF-type I interferon signaling pathway in immune inhibition. Intriguingly, THP-1 differentiated macrophages no longer produced CXCL9 upon IFN-γ stimulation when they were pretreated either with poly(dA:dT), tumor-derived cfDNA, or IFN-α (Figure 4D). CXCL10 production remained unchanged in these macrophages. Thus, our data showed that tumor-derived cfDNA retained the ability to trigger the innate immune signaling pathways in neighboring immune cells. The activation leads to the production of type I IFNs that enhance tumor resistance to EGFR inhibition¹¹ or inhibit CXCL9 production in macrophages.

FIGURE 4. EGFR-TKIs accelerate the release of the cell-free DNA activating type I IFN response in a STING-dependent manner. (A) EGFR-mutant tumor cells were treated with either osimertinib (EGFR-TKI) or paclitaxel. Data are presented as the mean ± SEM of technical triplicates from at least two independent experiments. (B) cfDNA concentrations in the supernatant were measured as described in Materials and Methods. Data are presented as the mean ± SEM of technical triplicates from at least two independent experiments. (C) THP-1-Dual and THP-1-Dual STING-knockout (KO) cells were cultured either with LPS (500 ng/ml), IFNa2a (10,000 U/ml), tumor-derived cfDNA (PC9 and H1975 cfDNA)/lipofectamine, or lipofectamine alone. IRF (left) and NF-κB (right) activities were analyzed as described in Materials and Methods. (D) PMA-differentiated THP-1 monocytes were pretreated either with lipofectamine, poly(dA:dT) (100 ng/ml), tumor-derived cfDNA (PC9) (100 ng/ml)/lipofectamine, or IFNa2a (20,000 U/ml) followed by IFN-γ stimulation (100 ng/ml). (E) Expression levels of CXCL9 and CXCL10 were analyzed using flow cytometry. *p [less than] 0.05, **p [less than] 0.01, ****p [less than] 0.0001, *q [less than] 0.01. ns, not significant.

Levels of cfDNA are decreased in the peripheral blood of patients who respond to osimertinib

According to our in vitro data, the tumor response to EGFR-TKIs represents the accelerated release of cfDNA, which stimulates inflammatory immune signaling pathways. However, in patients, systemic inflammatory symptoms, such as pyrexia, are not as common as one might expect, given the high response rate to osimertinib.⁴ To compare the relevance between clinical outcomes and cfDNA concentration levels in the blood, we analyzed four patients with durable tumor responses to osimertinib (Table S4). Interestingly, along with changes in tumor markers, cfDNA levels in the blood also decreased significantly (Figure 5). In addition, none of the patients experienced pyrexia during the observation period. Given that cfDNA degrades quickly, our clinical observations suggest that distinct cfDNA regulatory mechanisms between the tumor microenvironment and the peripheral circulating blood system may play an important role in controlling innate immune responses via DNA sensors. Furthermore, cfDNA is considered an alternative surrogate biomarker if conventional tumor markers such as carcinoembryonic antigen (CEA) and cancer antigen 19–9 (CA19-9) are not available in some patients.

FIGURE 5. Levels of cfDNA are decreased in the peripheral blood of patients who respond to osimertinib. Patients (Pt) 10–13, with EGFR-mutant NSCLC, were treated with first-line osimertinib on day 1 and thereafter. Computed tomography images on the left show tumor responses after treatment. Changes in cfDNA levels in the blood and tumor markers are shown on the right. TM, tumor marker (CEA or CA19-9).

DISCUSSION

Patients with EGFR-mutant NSCLC benefit from EGFR-TKIs, with durable responses. However, EGFR inhibition is rarely curative, and residual dormant cells eventually acquire resistance, resulting in tumor relapse.²³ In this study, we identified CD24 and tumor-derived cfDNA as potential targets for eradicating dormant EGFR-mutant cells under EGFR inhibition.

CD24 is a ligand of Siglec-G in mice, and Siglec-10 in humans.¹⁹ CD24 expression is associated with shorter survival and an impaired response to anti-PD-1/L1 immune checkpoint blockade in NSCLC.^17,18 In addition to NSCLC, adverse effects of CD24 on survival have been reported in a variety of cancer types.²⁴ However, the role of CD24 in the context of immunity remains controversial. CD24-deficient mice were resistant to experimental autoimmune encephalomyelitis.²⁵ These results indicated that CD24 promotes cytotoxic T-cell responses. In contrast, graft-versus-host disease was exacerbated in Siglec-G-deficient mice.²⁶ This study revealed that the interaction between Siglec-G on host APCs and CD24 on donor T cells plays a key role in inhibiting T-cell activation. CD24 also inhibits tumoricidal macrophage phagocytosis through its interaction with Siglec-10.²⁰ Therefore, CD24, as well as CD47,²⁷ is defined as a “do not-eat-me” molecule. Although this background supports the development of a therapeutic anti-CD24 antibody aimed at inhibiting the Siglec-10–CD24 axis in cancer, we found that Siglec-10 expression was restricted to M-CSF monocyte-derived macrophages with IL-10 and TGF-β. Intriguingly, in contrast with the data in the literature,²⁰ IL-4 did not induce Siglec-10 in M-CSF macrophages in three healthy individuals in our study. In addition, GM-CSF macrophages, even with IFN–γ and LPS, did not upregulate Siglec-10 expression. Given the temporal and spatial plasticity of macrophages, targeting CD24 by antibodies may have an additional role rather than blocking the interaction with Siglec-10. In our study, CD24 induction in EGFR-mutant lung cancer cells treated with EGFR-TKIs was shown to be a therapeutic target that promoted ADCP. With the clinical success of antibody–drug conjugates such as trastuzumab deruxtecan²⁸ and sacituzumab govitecan,²⁹ an anti-CD24 antibody conjugated with cytotoxic drugs may also exert efficacious antitumor activity in EGFR-mutant NSCLC when combined with EGFR-TKIs. Our study also suggests that tumor CD47 in EGFR-mutant NSCLC plays a limited role in adaptively acquired immune resistance as its expression levels were comparable before and after EGFR-TKI treatment. Although CD47 is reportedly not associated with the treatment efficacy of immune checkpoint blockade in patients with NSCLC,¹⁸ the role of CD47 expression before EGFR inhibition in the primary resistance to EGFR-TKIs remains to be elucidated in the future.

In transplantable mouse tumor models, tumor-infiltrating APCs take up tumor-derived DNA and produce type I IFNs in a STING-dependent manner.³⁰ Type I IFNs are essential for eliminating tumor cells^31,32 and have been used clinically to treat patients with renal cell carcinoma.^33,34 However, they also suppress T-cell responses during chronic viral infection.^35,36 These findings indicate that the role of type I IFNs is context dependent. In the present study, we found that EGFR-TKIs accelerated the release of cfDNA from EGFR-mutant cells in vitro, which activated the type I IFN signaling pathways. We showed that tumor-derived cfDNA abrogated CXCL9 production in macrophages stimulated by IFN-γ. In addition, type I IFNs and proinflammatory cytokines directly enhance tumor resistance by bypassing EGFR inhibition and promoting the EMT signaling pathways.^11,37 Conceivably, these innate immune pathways may contribute to therapy resistance when cfDNA is secreted by tumor cells under EGFR inhibition.

The use of cfDNA contributes to the prediction of treatment response and disease progression across organ types.^38–40 In these studies, cfDNA with a specific mutation was directly analyzed as a fraction or an amount. However, it remains unclear whether analyzing bulk amounts of cfDNA in the blood without targeting a specific mutation is similarly predictive. Some studies have pointed out that levels of bulk cfDNA do not correlate with tumor burden.^41,42 In this study, our data suggest that bulk cfDNA may be useful to predict the response, at least in EGFR-targeted therapy. We may then avoid designing mutation-specific PCR or performing droplet digital PCR, which could save our labor and financial costs.

In conclusion, our study shows that EGFR-TKIs foster the tumor microenvironment associated with tumor immune escape via innate immune signaling pathways. Therefore, we propose the importance of targeting the innate immune responses to overcome the resistance of EGFR-mutant NSCLC cells to EGFR inhibition.

AUTHOR CONTRIBUTIONS

A. S., T. N., and H. D-A. were involved in all aspects of this study, including planning and performing experiments, analyzing and interpreting data, and writing the manuscript. S. A and I. K. designed the in vitro tumor growth inhibition assays and interpreted the data. U. T. and Y. M. planned and performed immunohistochemistry, interpreted the data, and wrote the manuscript. Y. T., J. S, and N. S. assessed the clinical data of patients treated with EGFR-TKIs. Y. O., J. T., S. T., Y. S., and T. K. participated in blood sampling and cfDNA isolation, and interpreted the data. T. N. oversaw all the work performed.

ACKNOWLEDGMENTS

We are grateful to Professor S. Taniguchi and Associate Professor S. Joshita at Shinshu University School of Medicine for their constructive criticisms and comments. We thank Y. Akahane, M. Miyagawa, S. Komiya, and K. Ohnishi for their technical support in blood processing and cfDNA isolation. We also thank T. Goda and T. Amano at the Department of Medical Oncology, Hokkaido University School of Medicine for their discussions, and K. Miyazaki for assistance with immunohistochemistry. Aspects of study at Hokkaido University were performed with assistance from the Department of Concentrated Facilities at Hokkaido University Graduate School of Medicine.

FUNDING INFORMATION

This study was supported by the Grants-in-Aid for Scientific Research [Young Scientists Grant no. 19 K16827 (T.N.)] from the Ministry of Education, Culture, Sports, Science and Technology of Japan, a research grant from the Akiyama Life Science Foundation (T.N.), and the internal research grants from Shinshu University Hospital and Shinshu Igaku Shinkokai (T.N.).

CONFLICT OF INTEREST

H Dosaka-Akita received commercial research grants from Taiho Pharmaceutical Co., Ltd., Ono Pharmaceutical Co., Ltd., Eli Lilly Japan, Chugai Pharmaceutical Co., Ltd., and Eisai Pharmaceutical Co., Ltd.; received honoraria from Chugai Pharmaceutical Co., Ltd., AstraZeneca Pharmaceutical Co., Ltd., and Ono Pharmaceutical Co., Ltd. No potential conflicts of interest were disclosed by the other authors.

ETHICS STATEMENT

Approval of the research protocol by an Institutional Reviewer Board: This study was approved by the Institutional Review Boards of Hokkaido University Hospital (protocol codes: 019–0261, 019–0468, 020–0346) and Shinshu University Hospital (protocol code: 611).

Informed Consent: All the participants provided written informed consent.

Registry and the Registration: N/A.

Animal Studies: N/A.