Thirty patients with EGFR/ALK negative lung cancer with available clinical information including age, gender, stage, treatment history, and baseline plasma samples were enrolled in this study in Guangdong Provincial People’s Hospital from June 2017 to February 2019. Peripheral blood was collected from each patient on a regular basis for routine clinical care. Plasma sample was prepared within 2 hours of blood drawn and then stored at −80°C. In this study, plasma samples of patients with advanced EGFR/ALK wild-type (WT) NSCLC were collected before the administration of PD-1/PD-L1 inhibitors as baseline. The efficacy evaluation was conducted after three cycles of treatment. For every three cycles, plasma samples were collected from patients until the disease progressed. Patients who achieved partial response (PR) or complete response (CR) were included in this study as responders, compared with patients with progressive disease (PD) on treatment as non-responders. Flow chart for patient selection and exclusion criteria is shown in online supplementary figure 1. No CR was observed in this patient cohort. In total, five patients who achieved PR and four patients with PD at efficacy evaluation were included in this study. Plasma samples from seven healthy individuals were also collected as normal controls.

Online supplementary table 1 lists the baseline clinicopathological characteristics of all the patients enrolled in this study. formalin-fixed paraffin-embedded(FFPE) slides of the tumor samples were prepared for PD-L1 expression evaluation using immunohistochemistry with the following PD-L1 antibody clones: SP142 or SP263 (Roche, USA), and 28-8 (Abcam, USA). For the nine patients (responders and non-responders) included in this study, seven patients were analyzed using SP263 antibody, while the other two were analyzed with SP142 or 28-8, respectively. Patients were treated with different immunotherapy drugs targeting PD-1/PD-L1. The median follow-up time was 8 months, ranging from 1 month to approximately 21 months. Except one patient with lymphoepithelioma-like carcinoma, five patients were diagnosed with adenocarcinoma, while the other three patients were diagnosed with squamous cell carcinoma. The median age at diagnosis is 55, ranging from 47 to 68.

One milliliter of plasma sample was centrifuged at 10,000×g for 30 min at 4°C to remove any cell debris. The collected supernatant was then subjected for ultra-high-speed centrifugation at 150,000×g for 70 min at 4°C. The pellet containing exosome was resuspended in 200 µL Phosphate buffered saline(PBS) for downstream applications. Total RNA including miRNA was extracted from plasma-derived exosome using miRNeasy Serum/Plasma Kit (QIAGEN) following the manufacturer’s instructions. The quantification and size distribution of the extraction were analyzed by Qubit V.4.0 and Agilent Bioanalyzer 2100 (Agilent), respectively. Quantified RNA was subjected for sequencing library preparation using NEBNext Small RNA Library Prep Set for Illumina (NEB Biolabs) following the manufacturer’s instructions. Briefly, isolated total RNA was subjected for 3′ and 5′ adaptor ligation, followed by 17 cycles of PCR amplification. PCR products from library preparation were subjected for gel electrophoresis on 6% Novex TBE PAGE gel (Thermo Fisher Scientific) and DNA fragments between 140 and 150 bp were recovered from the gel. Purified small RNA cDNA library was quantified by Qubit V.4.0 and the size distribution was analyzed on Agilent Bioanalyzer 2100, followed by sequencing on Illumina HiSeq4000 platform.

For verification of purified exosomes using electron microscopy, purified exosomes suspended in PBS were dropped on copper-coated grids. After staining with 2% uranyl acetate, grids were dried at room temperature and visualized using a Hitachi H-7650 transmission electron microscope. miRNA-seq data analysis miRNA identification and reads counting in each miRNA were performed using miRDeep2.30 After trimming the 3′ adaptor sequence, all sequences ranging in length from 18 to 26 nt were recorded in a non-redundant file along with reads count. To identify known miRNAs, the miRNA tags were aligned against miRNA precursor sequences reported in the miRNA database “miRBase” (release 21) using the “quantifier.pl” script within miRDeep2. Differential expression (DE) analysis of miRNA sequence data was performed with the Bioconductor package edgeR.31 miRNAs with read counts per million mapped reads (CPM) ≥5 in at least 20% of all samples were identified as expressed miRNAs. DE between different groups was evaluated by fitting a negative binomial generalized linear model and then adjusting the p value for multiple testing using the Benjamini-Hochberg correction with a false discovery rate (FDR) of 0.1 and a minimum log₂(CPM) of 4.

All five patients who achieved PR (responders) and four patients with PD (non-responders) were enrolled in the exosomal miRNA differential expression analysis. Baseline clinicopathological characteristics of patients enrolled are listed in online supplementary table 1 with detailed treatment history of the patients listed in online supplementary table 2. Nine baseline pre-treatment plasma samples and the post-treatment plasma samples from four responders as well as normal control samples from seven healthy individuals were collected for exosome purification and exosomal miRNA profiling. We purified exosomes from plasma by differential centrifugation and verified them as small double-leaflet membrane particles (30–200 nm) by transmission electron microscopy (TEM) (figure 1).

Raw reads of miRNA-seq from all plasma samples were normalized to CPM and 706 miRNAs were retained as expressed miRNAs in plasma exosomes (online supplementary table 3). Using hierarchical clustering, healthy individuals could be distinguished from patients with lung cancer based on expressed miRNAs profile in plasma-derived exosomes (figure 2). We further analyzed differentially expressed miRNAs (DE miRNAs) between healthy control samples and patient samples. As shown in online supplementary table 4, 155 DE miRNAs were identified and the top 10 upregulated and top 10 downregulated miRNAs are listed in online supplementary table 4a. Some of these top regulated miRNAs were highly related to immune pathways. Among them, hsa-miR-146a has been well studied in autoimmune disease pathogenesis and is related to NF-κB pathway involved in immune response.32 hsa-miR-221 and hsa-miR-224 have been reported to be related to anti-TNF treatment in inflammatory bowel disease.33 hsa-miR-224 has been proven to induce cell proliferation by target Smad4 through SMAD/TGF-β pathway.34 Meanwhile, three of the top regulated miRNAs which were downregulated in patients with lung cancer have been reported as tumor suppressors in different cancer types, including hsa-miR-144,35–37 hsa-miR-363-3p,38 and hsa-miR-150.39 40

Kyoto Encyclopedia of Genes and Genomes(KEGG) pathway analysis was performed with 155 differentially expressed exosomal miRNAs using DIANA tool (online supplementary table 4b). Cancer-related pathways were significantly enriched in lung cancer patient group including TGF-β, MAPK, Hippo, and Ras signaling pathways. Immune-related pathways including TNF signaling pathway, B-cell and T-cell receptor signaling pathways were also enriched in the patient group. In conclusion, plasma-derived exosomes’ miRNA expression profile was altered in patients with EGFR/ALK negative lung cancer and differently expressed miRNAs were enriched in cancer-related and immune-related pathways compared with healthy control.

To identify the potential predictors of anti-PD-1 therapy in patients with lung cancer, we divided patient samples into PR group and PD group. Then we compared the miRNA profile between PR group and PD group using exact test. With cut-off at logCPM>4, p<0.05, FDR≤0.1, hsa-miR-320d was considered as the most significant DE miRNA followed by hsa-miR-642a-3p, hsa-miR-320c, and hsa-miR-320b with FDR around 0.1 (table 1).

Next, we examined the miRNA expression level of hsa-miR-642a-3p, hsa-miR-320c, hsa-miR-320b, and hsa-miR-642a-3p in overall patient cohort and healthy population (H). We separated all the samples from patients into three groups: PR pre-treatment samples (PR-pre), PD pre-treatment samples (PD-pre), and PR post-treatment samples (PR-post). As shown in figure 2, hsa-miR-320d, hsa-miR-320c, and hsa-miR-320b belonged to hsa-miR-320 family and all showed significant difference in mRNA expression level between PD-pre group and PR-pre group. Moreover, compared with healthy control and PR group (PR-pre and PR-post), hsa-miR-320d, hsa-miR-320c, and hsa-miR-320b showed upregulation in PD-pre group (figure 3). hsa-miR-643a-3p was eliminated due to its low expression level and no significant difference in miRNA expression level among these three groups (figure 3D). In conclusion, high level of hsa-miR-320d, hsa-miR-320c, and hsa-miR-320b might correlate with unfavorable response to anti-PD-1 treatment and hsa-miR-320d, hsa-miR-320c, and hsa-miR-320b might be potential predictors of anti-PD-1 therapy.

Paired pre-treatment and post-treatment plasma samples were collected from the four patients who achieved PR on anti-PD-1 therapy (online supplementary table 1). To identify potential exosomal miRNAs influenced by anti-PD-1 therapy during the treatment course, we compared the exosomal miRNA profiles before (PR-pre) and after (PR-post) anti-PD-1 treatment. With cut-off at logCPM>4, p<0.05, FDR≤0.1, only hsa-miR-125b-5p was differentially expressed between PR-pre and PR-post samples (table 2 and figure 4A), which was dramatically downregulated in the PR-post samples and suggested that the level of exosomal hsa-miR-125b-5p might be used as a potential indicator to monitor the efficacy of anti-PD-1 treatment. Interestingly, plasma samples from the PD group at baseline exhibited a trend of increased level of hsa-miR125b-5p compared with PR-pre group and healthy controls (figure 4A).

This study has demonstrated that hsa-miR-125b-5p could induce γδ T-cell apoptosis and downregulate γδ T-cell activation and cytotoxicity to tumor cells by decreasing degranulation and inhibiting the secretion of IFN-γ and TNF-ɑ41 (figure 4B). An increased exosomal hsa-miR-125b-5p level in PD group at baseline might indicate impaired γδ T-cell function and cytotoxicity to tumor cells that could confer resistance to immunotherapy. On the other hand, a significantly reduced exosomal hsa-miR-125b-5p level could indicate the activation of T cells and cytotoxicity to tumor cells in response to immunotherapy.