A2AR-mediated CXCL5 upregulation on macrophages promotes NSCLC progression via NETosis

Peripheral blood samples were obtained from healthy donors recruited from the Henan Red Cross Blood Center with informed consent. Lung tumor samples for immunofluorescent staining were collected from untreated patients with NSCLC and surgically resected the specimen at the First Affiliated Hospital of Zhengzhou University with Ethics Committee approval. Patients provided informed consent or relative's assent.

H460 and A549 NSCLC cell lines, obtained from the Chinese Academy of Sciences Shanghai Branch Cell Bank, were cultured in RPMI 1640 and DMEM-F12 with 10% FBS at 37 °C, 5% CO₂.

C57BL/6J mice (6–8 weeks) obtained from Beijing Vital River Biocytogen were housed under specific pathogen-free conditions. Humane care followed the Guide for the Care and Use of Laboratory Animals (National Institutes of Health publication 86–23, revised 1985). Lewis lung cancer (LLC) cells (1 × 10⁶) were subcutaneously injected for treatment evaluation, and tumor growth was monitored weekly using PerkinElmer IVIS spectrum until day 28. Tumor-infiltrating immune cells were analyzed after 2 × 10⁶ LLC cells were injected, and the mice were sacrificed on day 21.

Healthy donor-derived macrophages (5 × 10⁵/well) and NSCLC cells (H460/A549, 8 × 10⁵/well) were co-cultured in a Transwell device (0.4um). CPI-444 (Selleck, S6646), sodium metatungstate (POM-1) (Selleck, S5525), and JSH-23 (Selleck, S7351) were added 2 h before 24 h incubation. The controls included individual cultures at matching densities. Supernatants and cells were collected for protein and mRNA analysis.

RNA was extracted from human macrophages and T cells using RNAiso Plus (Takara, 9109). The mRNA sequencing and differential expression analyses were conducted at the Beijing Genomics Institute. The differential expression genes are represented in Supplementary Table 1.

CXCL5 concentration in the cell culture supernatant was quantified using an ELISA kit (BioLegend, CAT#440,904) according to the manufacturer’s protocol.

Cells were lysed in RIPA buffer and sonicated. Proteins were separated using 12% SDS-PAGE gel and transferred to nitrocellulose membranes. The membranes were blocked with 5% defatted milk for 1 h, incubated with primary antibodies (1:1000) overnight, and then with HRP-conjugated secondary antibodies for 1 h. Protein signals were visualized using ECL detection reagents.

Tumor tissues, fixed in 4% paraformaldehyde, were paraffin-embedded to generate 5 μm sections. Antigen retrieval was performed using citrate solution. Sections were permeabilized (0.1% Triton X-100), blocked (5% BSA), and incubated overnight at 4 °C with primary antibody. Sections were incubated with fluorescence-conjugated secondary antibodies (1 h at room temperature), mounted with DAPI-containing medium, imaged using Olympus microscope and Vectra Automated Multispectral Imaging system, and analyzed with ImageJ software. The dilution rates are represented in Table 1.

TMA (HLug-NSCLC150PT-01) from Shanghai Outdo Biotech contained 75 NSCLC and paired para-tumor tissues. Immunohistochemistry was performed by Wuhan Servicebio company. Clinical–pathological parameters are shown in supplementary Table 2.

Stable ShCD73-expressing H460 and A549 cell lines were generated by using lentiviral transduction and antibiotic selection. The shCD73 plasmid (hU6-MCS-Ubiquitin-firefly_Luciferase-IRES-puromycin) was purchased from GeneChem. JetPRIME^® Kit was used to transfect plasmids into HEK 293 T cells. After 48 h, the supernatant containing lentivirus was added to the A549 and H460 plates with 6 ng/ml Polybrene (Solarbio, H8761). Puromycin (MCE, HY-B1743A, 2 μg/ml) was added after 48 h for shCD73-expressing cell selection.

Cells were stained with fluorescent-conjugated antibodies (15 min) or primary antibodies (30 min) at 4 °C in 1% FBS flow buffer. For intracellular staining, cells were fixed in 4% paraformaldehyde (Servicebio, CAT# G1101), permeabilized (BioLegend, CAT# 421,002; BD, CAT#562,574), and stained with antibodies (15 min) at 4 °C. Data were acquired with a Beckman Coulter DxFLEX cytometer. Analysis was performed using the CytExpert IDEAS Application, FlowJo v10, and CytExpert. The antibodies used are listed in Tables 2 and 3. Standard and imaging flow cytometry data were acquired and analyzed accordingly.

Total RNAs were extracted using TRIZOL (Takara, CAT#9101) following the manufacturer’s instructions. The RNA concentrations were measured using a NanoDrop 2000 (Thermo Fisher Scientific, USA). Next, 1 µg of RNA was reverse transcribed to cDNA using HiScript III RT SuperMix for qPCR (+ gDNA wiper) (Vazyme, CAT# R223-01). Real-time PCR (40 cycles, annealing temperature 60 °C) was performed using the ChamQ Universal SYBR qPCR Green Master Mix (Vazyme, CAT# Q711-02) on a CFX96 real-time system (Bio-Rad, USA). Relative gene expression was quantified by the 2 − ΔΔCT method, and human β-actin served as an internal control for each reaction. The primers used for qPCR are listed in Supplementary Table 3.

CD8⁺ T cells were activated 3 days before coculturing with NETs using CD3 (5 μg/ml) and CD28 (2.5 μg/ml) antibodies and rhIL2 (100 IU). CD8⁺ T cells (1 × 10⁶ per well) were treated with or without NETs for 3 days and collected for flow cytometry and RNA-seq.

293T cells were transfected with 0.25 μg luciferase plasmid pGL3-CXCL5, 0.25 μg pcDNA3.1-RELA reporter plasmid and 500 ng Renilla plasmid pRL-TK for 48 h. Cell lysates were collected and analyzed using the Dual-Luciferase Reporter Gene Assay Kit (YEASEN, Cat: 11402ES60) according to the manufacturer’s instructions. Luciferase and Renilla bioluminescence were detected using SpectraMax^® iD3 Multi-Mode Microplate Reader. Firefly luciferase activity was normalized to the Renilla luciferase activity.

Public gene expression data for The Cancer Genome Atlas (TCGA) lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC) were acquired to analyze gene correlation using the online websites GEPIA (http://​gepia.​cancer-pku.​cn) and TIMER [15] (https://​cistrome.​shinyapps.​io/​timer/​) and analyzed using Pearson’s correlation test. TISIDB was used to analyze the correlation between cell abundance and CXCL5 expression, and the correlation between immune subtype and CXCL5 expression [16]. The gene matrix used to identify neutrophils and macrophages was selected from the study by Charoentong et al. [17]. The adenosine signature dataset, previously reported in a renal cell carcinoma study [18], was used to analyze the RNA-seq data. The signature score was calculated as the mean log2 (TPM + 1) value of each gene in the differentially expressed gene (DEG) dataset. Heatmap and KEGG enrichment analyses were performed using the OmicShare Tool (GENE DENOVO, https://​www.​omicshare.​com). GSEA was performed using GSEA 4.3.2 and compared with the exhaustion signature (Supplementary Table 4) described in two previous studies. For survival analysis, dataset (GSE8894) was collected and analyzed using PrognoScan (http://​dna00.​bio.​kyutech.​ac.​jp/​PrognoScan/​).

Statistical analyses were performed using GraphPad Prism 9 software (GraphPad Software, USA). In the bar graphs, data were shown as mean ± SD and analyzed by two-tailed student’s t test or one-way ANOVA with Tukey’s test. Statistical significance was set at P < 0.05. P values are represented as follows: **** P < 0.0001, *** P < 0.001, ** P < 0.01, and * P < 0.05.

The interaction between the immune system and tumor cells is key for tumors to manipulate the TME and escape immune elimination. To explore the mechanisms of TAMs and tumor cell interactions in NSCLC, we mimicked the tumor microenvironment by co-culturing NSCLC cell line H460 with macrophages induced in vitro using Transwell devices for 24 h (Supplementary Fig. 1A, B). We conducted RNA-seq on macrophages and H460 cells before and after co-culture (Supplementary Fig. 1C, 2). We analyzed differentially expressed genes in macrophages (Supplementary Table 1) and observed that DEGs were enriched in cytokine-receptor-related KEGG pathways (Fig. 1a, b). The cytokine CXCL5 was the most varied gene, with a Log2FC of approximately 7 compared with the control after co-culture (Fig. 1c). We compared the CXCL5 concentrations in the co-culture, macrophage, and H460 cell supernatants and found the highest CXCL5 concentration in the co-culture supernatant (Fig. 1d). Analysis of TCGA data showed that macrophage abundance was associated with CXCL5 expression in tumor tissues (Supplementary Fig. 1E), suggesting an association between macrophages and CXCL5. RT-PCR confirmed that the relative expression of CXCL5 in macrophages after co-culture was much higher than that in H460 and A549 cells before and after co-culture (Fig. 1e), indicating that the upregulation of CXCL5 mainly originated from macrophages. Imaging flow cytometry of H460, A549 and macrophages after co-culture demonstrated that macrophages generated more CXCL5 than H460 and A549 cells (Fig. 1f). Furthermore, the detection of CXCL5 in tumor tissues from patients with NSCLC demonstrated that TAMs expressed higher levels of CXCL5 than other cells (Fig. 1g). These results demonstrated that CXCL5 expression is upregulated in macrophages, suggesting that the interaction between tumor cells and macrophages is mediated by tumor cell-derived factors.

As the results showed that macrophages were educated by tumor cells and consequently generated more CXCL5, we further analyzed RNA-seq data. By exploring the correlation between CXCL5 and the top 10 DEGs (Supplementary Table 5), we observed that CXCL5 was positively correlated with ADORA2A, G0S2, MARCO, MT1E, and SNAI1 in the TCGA database (Fig. 2a, b, Supplementary Fig. 3A, B). Since the co-culture device we used was built for indirect co-culture, we assumed that soluble factors might mediate the interaction between tumor cells and macrophages and that proteins localized on the cell membrane were more likely to be directly responsible for the upregulation of CXCL5 in macrophages. Therefore, MT1E, G0S2, and SNAI1 were excluded from consideration because studies have shown that their coding proteins are localized in the plasma and nucleus [19‐21]. Both MARCO and ADORA2A encode for membrane proteins. However, A2AR (encoded by ADORA2A) which is the receptor for extracellular adenosine has been reported to be associated with cytokine secretion in monocytes [18]. Further experiments confirmed that ADORA2A was upregulated after co-culture (Fig. 2c, d, Supplementary Fig. 4B, C). We then analyzed the adenosine signature that was upregulated after stimulating A2AR [18]. The adenosine signature score, which represents the activation of A2AR signaling, was higher in macrophages after co-culture, as was the cAMP metabolic process enrichment score (Fig. 2e, f, Supplementary Fig. 4A). Analysis of TCGA database revealed that the adenosine signature was positively associated with CXCL5 expression in NSCLC (Fig. 2g). We detected an increase in adenosine concentration in the co-culture supernatants (Fig. 2h). A2AR expression was upregulated according to the imaging flow cytometry results (Fig. 2i). By applying CPI-444, a selective A2AR inhibitor, to the co-culture medium, we observed the downregulation of CXCL5 (Fig. 2j). Treating macrophages with adenosine analog 5ʹ-N-Ethylcarboxamidoadenosine (NECA) significantly stimulated CXCL5 generation, while CPI-444 was able to reduce the CXCL5 expression  (Fig. 2k). These results suggest that tumor cell-induced A2AR signaling is responsible for CXCL5 upregulation in macrophages.

Extracellular adenosine is a ligand of A2AR, which can be generated from ATP hydrolysis driven by CD39-CD73 catalysis. We detected higher CD39 expression in macrophages and higher CD73 expression in tumor cells by comparing the two groups using flow cytometry and RNA-seq (Fig. 3a–c); flow cytometry demonstrated that CD39 and CD73 were only expressed in macrophages and tumor cells, respectively (Fig. 3a, b). These results suggested that cooperation between macrophages and tumor cells is necessary to promote extracellular adenosine generation. As expected, CXCL5 concentration decreased after CD39 inhibition (Fig. 3d). We then constructed CD73-knockdown NSCLC cell lines H460-shCD73 and A549-shCD73 and selected the sh1 RNA-knockdown cell line for further experiments (Fig. 3e). The CXCL5 concentration decreased when macrophages were co-cultured with shCD73 cells instead of shNC cells (Fig. 3f). These results indicated that cooperation between macrophages and tumor cells is required for extracellular adenosine accumulation and CXCL5 upregulation (Fig. 3g).

We identified transcription factors involved in A2AR-induced CXCL5 upregulation. According to the RNA-seq data, certain transcription factors were upregulated in macrophages (Supplementary Table 6). We cross-checked this gene list against the transcription factor list of genes predicted by the online tool PROMO (https://​alggen.​lsi.​upc.​es/​cgi-bin/​promo_​v3/​promo/​promoinit.​cgi?​dirDB=​TF_​8.​3) (Supplementary Table 6), which had binding sites located in the CXCL5 promoter sequence and found seven transcription factors in both lists (STAT4, ETS2, REL, RELA, JUNB, ETS1, and CEBPB). We examined these transcription factors using JASPAR (https://​jaspar.​genereg.​net), and RELA (encoding P65, a subunit of NFκB) had the highest relative score (Supplementary Table 6). Other studies have reported that RELA is a transcription factor for CXCL5 [22]. Therefore, we used RT-PCR and western blotting to verify the upregulation of RELA and phosphorylated P65 in macrophages after co-culture (Fig. 4a, f). By adding the NFκB inhibitor JSH-23 to the co-culture medium, we confirmed that NFκB regulated CXCL5 concentration (Fig. 4b). The dual-luciferase reporter assay showed that RELA was bind to the promoter region of CXCL5 (Fig. 4c, d). We also observed that phosphorylated P65 (RELA) was upregulated in post-co-cultured macrophages and translocated from the cytoplasm to the nucleus of macrophages after A2AR agonist stimulation. However, the A2AR inhibitor inhibited this upregulation and translocation (Fig. 4e, f). These results indicated that transcription factor NFκB mediated A2AR-regulated CXCL5 upregulation.

Studies have demonstrated that CXCL5 can recruit neutrophils that express high levels of the receptor CXCR2. In addition, CXCL5 expression was positively associated with neutrophil abundance in the TCGA LUSC and LUAD datasets (Fig. 5a). Considering that neutrophil NETosis, which releases chromatin outside the cell, is a significant phenomenon reported to be associated with tumor progression [23], we stained neutrophils with the nucleic acid dye SYTOX Green and observed a stained web-like structure under a fluorescent microscope after rhCXCL5 treatment (Fig. 5b). Therefore, we detected citrullinated histone H3, a marker of NETs, in rhCXCL5-treated neutrophils using immunofluorescence and observed a CitH3 fluorescence signal after rhCXCL5 treatment (Fig. 5c). NETs play an important role in tumor metastasis and promote immune evasion, but their influence on immune cells has not been fully elucidated. We further explored their influence on CD8⁺ T cells by co-culturing these cells with freshly isolated NETs. RNA-seq analysis demonstrated that exhaustion-associated genes were upregulated in CD8⁺ T cells after treatment with NETs (Fig. 5d). We observed increased expression of TOX, a transcription factor that regulates immune checkpoint expression, and upregulation of TIM3 and LAG3 (Fig. 5e). However, IFN-γ, TNF-α, and IL2 expression were significantly decreased (Fig. 5f). GSEA and GO enrichment analysis showed a significant upregulation of the cytosolic DNA-sensing pathway and its downstream type I interferon-associated pathway (Fig. 5g, h). STING1 is pivotal in the cGAS-STING cytosolic DNA sensing pathway, and its expression was increased in NETs-treated T cells. Inhibiting STING using its inhibitor C176 significantly upregulated IFN-γ, TNF-α, and IL2 expression but slightly suppressed TIM3 and LAG3 (Fig. 5e, f), suggesting a more complex mechanism that regulates NETs-induced T cell dysfunction. These results indicate that CXCL5 stimulates NETosis, which subsequently promotes CD8⁺ T cell dysfunction, partly through the STING-mediated pathway.

To examine the effects of A2AR on CXCL5, neutrophils, and tumor growth in vivo, we constructed a mouse model bearing subcutaneous LLC tumors (1 × 10⁶ cells per mouse). We treated the mice with CPI-444 10 mg/kg (A2AR inhibitor) daily from day 1 after inoculation and SB225002 10 mg/kg (a CXCR2 inhibitor) every three days from day 4 after inoculation until day 28. Tumor growth was monitored weekly. The results demonstrated that the A2AR and CXCR2 inhibitors significantly inhibited tumor growth after treatment (Fig. 6a, b). Some mice were tumor-free after 14 days of treatment. To evaluate the tumor-infiltrating immune cells, we subcutaneously inoculated 2 × 10⁶ LLC tumor cells into 6-week-old mice and treated them with CPI-444 or SB225002 for 20 days. Tumor tissues were analyzed using immunofluorescence and flow cytometry. We found that CXCL5 expression decreased after CPI-444 treatment (Fig. 6c). Neutrophil infiltration and NETs area significantly reduced after CPI-444 and SB225002 administration (Fig. 6d). We tested CD8⁺ T cell function in tumors and found that treatment reduced PD1⁺ TIM3⁺ T cells and rescued IFN-γ and IL2 expression (Fig. 6e). Ki67⁺ T cells increased after treatment (Fig. 6e). These data suggest that A2AR mediates CXCL5 expression in LLC tumors. Blocking A2AR significantly reduced neutrophil infiltration and NETosis.

To evaluate the relationship between CXCL5 and clinical data, we used TISIDB (http://​cis.​hku.​hk/​TISIDB/​index.​php) to analyze CXCL5 expression in NSCLC tissues classified by clinical stage in the TCGA database and found that CXCL5 expression was positively associated with clinical stage in LUAD but not LUSC (Fig. 7a, b). Additionally, we analyzed NSCLC tissue microarrays and observed higher CXCL5 expression in tissues with pathologically grades II and III (Fig. 7c). Thorsson et al. characterized tumor immune microenvironment and identified six immune subtypes: Wound Healing (C1), IFN-γ Dominant (C2), Inflammatory (C3), Lymphocyte Depleted (C4), Immunologically Quiet (C5), and TGF-β Dominant (C6) [24]. The C6 was the least favorable of the six subtypes in terms of prognosis and expressed higher levels of CXCL5 (Fig. 7d, e). Furthermore, the C6 immune subtype of NSCLC which has been reported to have a negative correlation with immunotherapy expressed higher levels of CXCL5 (Fig. 7d, e). Kaplan–Meier analysis generated from PrognoScan (http://​dna00.​bio.​kyutech.​ac.​jp/​PrognoScan/​) demonstrated that high expression of CXCL5 was associated with decreased survival of NSCLC patients (Fig. 7f). This suggests that NSCLC with high CXCL5 expression is associated with a poor prognosis.