Exosomes miR-184 and miR-3913-5p are involved in Osimertinib resistance in non-small cell lung cancer patients with exon 21 L858R mutation

The third-generation of EGFR-TKI Osimertinib has become an important treatment option for patients with EGFR-mutant advanced NSCLC. In recent years, more and more studies have begun to pay attention to the ability of miRNAs in exosomes secreted by tumor cells to transmit resistance information. The mechanisms of exosomal miRNAs involved in Osimertinib resistance need to be studied. Methods We constructed an Osimertinib-resistant cell line H1975-OR, and collected the exosomes from the cells of the drug-resistant strain and the sensitive strain, and extracted respective RNA for sequencing. We also compared miRNAs in serum exosomes of three patients before and after resistance to Osimertinib. Enlarged samples were then validated in 64 NSCLC patients.


Background
Epidermal growth factor receptor (EGFR) mutations are the most common genetic mutations in non-small cell lung cancer (NSCLC) [1]. The PIONEER study (NCT01185314) [2] showed that the EGFR mutation rate in Chinese population reached 50.2%. Among them, exon 19 deletion (48.9%) and exon 21 L858R point mutation (45.4%) were more common [2,3]. 50-60% NSCLC patients with EGFR mutations usually develop acquired resistance to TKI after rst-and second-generation EGFR-TKI therapy [4], of which, p.Thr790Met (T790M) point mutation in exon 20 is the most common. The FLAURA study, which received wide attention at the 2019 European Society for Medical Oncology(ESMO) congress, con rmed that A total of 67 NSCLC patients were included in this study in the Department of Respiratory and Critical Care Medicine at Jinling Hospital from December 2018 to October 2019. Three pair of blood samples were collected before and after resistance to Osimertinib. Another 37 patients were resistant to Ge tinib but have not yet developed Osimertinib resistance, and 27 had Osimertinib resistant. In addition, 10 healthy people were recruited from the physical examination center of Jinling Hospital. All participants signed informed consent. This study was approved by the ethics committee of Jinling Hospital.

Exosomes isolation
In order to isolate the effect of serum, when the cells grew to 80% full in a petri dish, H1975 and H1975-OR cells were cultured in complete RPMI-1640 medium without added serum. After 24 hours, the cell culture supernatant was collected and exosomes were separated by differential ultracentrifugation as described in [13]. To remove cell debris, the supernatant was centrifuged at 300 × g for 5 min and then at 3000 × g for 10 min. The supernatant was then centrifuged at 10,000 × g for 30 min. The supernatant was passed through a lter with an aperture of 0.22 µm (Millipore, USA). The ltered supernatant was transferred to a clean ultracentrifuge tube and ultracentrifuged at 4 ℃, 120000 × g for 70 min. After discarding the supernatant, the pellet was resuspended in an appropriate amount of sterile phosphate buffered saline (1 × PBS). Extracted exosomes can be used for immediate downstream experiments or stored in a -80 °C refrigerator.
Serum samples were extracted using a commercial kit Total Exosome Isolation (from serum) (invitrogen, 4478360). The required volume of clear serum was transferred to a new test tube and 0.2 volume of total exosomal separation (from serum) reagent was added. Mix the serum/ reagent mixture, whether through eddy current or pipe up and down, to make the solution turbid. After incubation at 4 °C for 30 min, centrifuge at 10,000 × g for 10 min at room temperature. Aspirate and discard the supernatant. Exosomes are contained in granules at the bottom of the test tube. Exosomes were resuspended using half the volume of the initial serum volume of sterile phosphate buffered saline (1 × PBS).
Transmission electron microscopy, size distribution analysis and Western blot Exosome morphology was observed by transmission electron microscopy. The exosome suspension was mixed with an equal volume of 4% paraformaldehyde, and 10 µl of the mixture was placed on a clean copper grid (RT) at room temperature. Uranyl acetate staining was negative. The images were acquired by observation with a JEOL 1200EX TEMSCAN microscope. Exosomal suspensions were analyzed for particle size by dynamic light scattering (DLS) (Nanosizer ™ instrument, Malvern Instruments, UK).
The extracted exosomes were resuspended in cell lysate (Beyotime, Nantong, China) added with 1% PMSF. The protein concentration of exosomes was determined by Pierce BCA protein detection kit (Thermo Fisher Scienti c, Rockford, IL). The twelve alkyl sulfate polyacrylamide gel electrophoresis was prepared with 20 µg protein on each sample. Anti-CD63, anti-TSG101, anti-β-actin, and anti-GAPDH were purchased from (Abcam, Cambridge, UK).

Library construction and sequencing
The total RNA in exosomes was extracted by using TRizol reagent (Invitrogen, Carlsbad, CA) according to the instructions. The NanoPhotometer spectrophotomete was used to detect RNA purity (OD260 / 280 and OD260 / 230 ratios). After using the Qubit2.0 Fluorometer to accurately quantify the RNA concentration, the Agilent 2100 bioanalyzer was used to accurately detect the integrity of the RNA. The total RNA of exosomes was reverse transcribed into cDNA. The starting RNA of the library was total RNA, and the total amount was ≥ 1 µg.
The small RNA library construction process is mainly summarized as follows: starting from the qualitytested RNA, 3 'end adaptor ligation, reverse transcription primer hybridization, 5' end adaptor ligation, reverse transcription into DNA, and then PCR ampli cation. After the library construction is completed, rst use Qubit2.0 for preliminary quanti cation, adjust the library to 1 ng/µl, and then use the Agilent 2100 to detect the length of the library insert. After meeting the expectations, use the Q-PCR method to determine the effective concentration of the library Accurate quanti cation (effective library concentration > 2 nM) to ensure library quality.
After the library was quali ed, different libraries were pooled according to the requirements of effective concentration and target o ine data volume for illumine sequencing. The basic principle of sequencing was sequencing by synthesis. Added four uorescently labeled dNTPs, DNA polymerases, and adaptor primers to the sequencing ow cell for ampli cation. When each sequencing cluster extended the complementary strand, each uorescently labeled dNTP could release the corresponding uorescence.
The sequencer obtained the sequence information of the fragments to be detected by capturing the uorescent signal and converting the optical signal into a sequencing peak by computer software.

Sequence read analysis
The raw data from small RNA-seq includes linker sequences and sequencing low-quality sequences. In order to ensure the accuracy of information analysis, the sequencing raw data needs to be ltered to obtain clean data, and subsequent bioinformatics analysis is performed based on clean data. The Qphred score (Qphred = 10log10 (e)) is used to represent the base quality value (Quality Score) to measure the quality of each base in the sequencing reads. Since miRNA is a small RNA with a length of about 22nt, the length of the sequence fragment mainly distributed around 22nt is selected as Clean reads data. The miRDeep2 software was used to analyze the miRNA expression abundance. According to the mapping position of read on the genome, it was necessary to verify whether the position of read and the known miRNA mature sequence were consistent. Cluster analysis and correlation analysis between samples (Pearson's Correlation Coe cient) were performed on the miRNA family in each sample. All sequencing raw data has been uploaded to the SRI database.

Target gene analysis
The analysis was performed by DESeq2 (no biological duplicate samples use DESeq or edgeR), and miRNAs with [logFoldChage] > 1 and p value < 0.05 were selected as miRNAs with signi cant differences. The volcano map of the results of miRNA differential analysis and the clustered heat map of miRNA expression of the samples were drawn. TargetScan was used to calculate a weighted context + + score to predict the target genes of different known miRNAs. These target gene functions were classi ed through a database established by the Gene Ontology Consortium, and GO enrichment analysis was completed.
At the same time, the rst 10 target genes of each sample were selected for KEGG Pathway enrichment analysis according to the parameters, so as to identify the downstream molecular metabolic pathways of these target genes.
Quantitative reverse transcription PCR As aforementioned, exosomes were extracted from the patient's serum, and then the total RNA of the exosomes was extracted by TRizol reagent. RNA was quanti ed and evaluated using NanoDrop® ND- All the above experiments were repeated twice.

Statistical analysis
The above data were mainly analyzed and mapped using SPSS 22.0 system (SPSS, Inc. Chicago, IL) and Graphpad Prism5. The differences in miRNAs levels between the two groups were mainly determined by the Mann-Whitney rank sum test for non-parametric data. The correlation between Exo-miR-184 and Exo-miR-3913-5p with clinicopathological features was determined by χ2 test. ROC curves were also used to determine diagnostic value. A P value less than 0.05 indicates a statistical difference.

Construction of Osimertinib-resistant strains and isolation of exosomes
In the commonly used lung cancer cell lines, H1975 carries EGFR-L858R-sensitive mutation or EGFR-T790M drug-resistant mutation. Therefore, we decided to use H1975 to construct a drug-resistant cell line for the study of Osimertinib resistance. The drug-resistant strain H1975-OR was established by increasing the concentration of Osimertinib. Six months later, MTT method was used to determine the cell viability of the sensitive strain H1975 and the resistant strain H1975-OR (Fig. 1A). The average IC50 values of these two cells were 4636 nM and 12101 nM, respectively, with signi cant differences (p = 0.0215) (Fig. 1B). RI was 2.61.
The circular vesicle-like exosomes were obtained from different cell supernatants by ultracentrifugation. After TEM identi cation (Fig. 1C) and DLS analysis (Fig. 1D), the average particle size was 97.79 nm. Meanwhile, the marker proteins CD63 and TSG101 of exosomes were identi ed by Western blot (Fig. 1E).
Screening for differential exosome miRNAs in H1975-OR and H1975 cell supernatant Exosomes were extracted and sequenced from the supernatants of H1975 and H1975-OR cells. The RNA-Seq data were analyzed by using the software of DESeq2 (no biological duplicate samples using DESeq or edgeR) and the corresponding FDR (p value corrected FDR value, the smaller the FDR value, the more signi cant the difference) with log2 (FC) (multiple change in expression level of the experimental group relative to the control group). The results showed that there were 1138 differential exosomal miRNAs in the two groups.

Functional enrichment analysis of differentially expressed miRNA target genes in exosomes
In order to further understand the biological functions of the exosomes of Osimertinib-resistant cells, the target genes of differential miRNAs were classi ed through the GO (Gene Ontology) database ( Fig. 2D), including biological process, molecular function and cell component. According to the selected target genes of different miRNAs, the rst 10 target genes of each miRNA are selected, and the hypergeometric distribution relationship between these target genes and a speci c branch (or branches) in go classi cation is calculated. GO analysis would return a p-value for each miRNA predicted target gene. A small p-value indicates that the target genes of differential miRNAs are enriched in this GO. Most genes involved in cell components, such as "extractor exosome", "extractor region" and "extractor space", once again veri ed that miRNAs are derived from exosomes. "Transcription, DNA templated", "positive regulation of gene expression" and other gene aggregations related to biological pathways also indicate that the miRNA of exosomes plays an important role in the resistance of Osimertinib (Table 1). KEGG pathway enrichment analysis can help us understand the molecular metabolic pathway information or complex biological activities involved in differential genes. According to the parameters, the rst 10 target genes of different miRNAs were selected for pathway enrichment (Fig. 2E). "Metabolic pathways" accounted for the largest proportion among the differential miRNA target genes, followed by "PI3K-Akt signaling pathway", "Ras signaling pathway", "Cytokine-cytokine receptor interaction", "Nonsmall cell lung cancer", etc. These are common pathways closely related to EGFR-TKI in NSCLC. Previous studies have shown that EGFR-independent mechanisms of Osimertinib resistance include bypass activation of PI3K pathway, which can be achieved through PIK3CA mutation/ampli cation and PTEN deletion. The sequencing results in the exosomes of H1975-OR cell supernatants con rmed that exosome PI3K/Akt pathway activation was involved in Osimertinib resistance.
In the dot plot (Fig. 2F), the most signi cant pathway enriched in Pathway analysis of known differential miRNA target genes is the Ras signaling pathway, and the Pathway analysis of novel differential miRNA target genes is concentrated in the MAPK signaling pathway ( Supplementary Fig. 1 Exosomes were also identi ed by TEM, DLS analysis and Western Blot ( Fig. 3A-C).
Based on exosome transcriptome data obtained from two RNA-seqs, Pearson's Correlation Coe cient (R2) was used to assess the correlation between the cell supernatant and different samples from the patient's serum, and the Correlation heat map matrix was drawn (Fig. 3D). The miRNAs in serum exosomes were compared before and after drug resistance in three patients. The miRNAs with signi cantly different expression were hierarchical clustered, and the clustering heat map was drawn (Fig. 3E). The rst patient listed in the gure had EGFR 21 L858R mutation, except for a large number of new differential miRNAs. Cluster analysis showed that miR-206, miR-200b-3p and miR-514b-5p were signi cantly up-regulated in patients with drug resistance to Osimertinib. However, the other two patients with EGFR 19 depletion signi cantly upregulated more miRNAs, including miR-25-5p, miR-184, miR-3913-5p and so on.
Increased miR-184 and miR-3913-5p in exosomes after Osimertinib resistant We performed the overlapping analysis of differential miRNAs obtained by sequencing exosomes in the cell supernatant (treat1) and three patients (treat2-4) by way of Wayne map (Fig. 4A, B). Among the exosome miRNAs up-regulated in the Osimertinib-resistant group compared with the sensitive group, no common differential miRNAs that overlapped in the four groups were found (Fig. 4A). However, in pairwise comparison, it was found that miR-184 overlapped in treat1 and treat4, miR-3913-5p was shared in treat3 and treat4, and miR-3614 co-occurred in treat1 and treat2 (Fig. 4A). These three miRNAs can be further studied. Similarly, miR-3614-5p, miR-4746-5p, miR-378i are the focus of miRNA in the down-regulation group (Fig. 4B).
In order to verify several signi cantly up-regulated Exo-miRNAs in the RNA-seq results, we decided to expand the sample size for experiments. Serum samples from 37 NSCLC patients with Ge tinibresistance and 27 patients with Osimertinib-resistance were collected as controls (GR) and experimental groups (OR) before and after Osimertinib resistance. QPCR veri ed that exosome miR-184 in the serum of drug-resistant patients was signi cantly higher (p = 0.0325) (Fig. 4C), and exosome miR-3913-5p was signi cantly increased in the Osimertinib resistant group (p = 0.0169) (Fig. 4D), which was consistent with the previous sequencing results. Combining these two exosome miRNAs, it was found that these two miRNAs in the serum exosomes of NSCLC patients were signi cantly up-regulated after Osimertinib resistance (p = 0.0092) (Fig. 4E, F).

Discussion
Osimertinib (AZD9291) is the rst third-generation EGFR-TKI to be approved by the FDA and EMA for the treatment of NSCLC [5]. It is selective for EGFR-TKI sensitization and T790M resistance mutations, and has less effect on WT-EGFR [14]. Regardless of rst-line or second-line use of Osimertinib, similar to other EGFR-TKIs, patients still inevitably develop resistance after receiving Osimertinib, which greatly limits the long-term clinical bene ts of this targeted drug [15]. The mechanism of Osimertinib resistance depends on the high tumor heterogeneity of NSCLC, which is divided into two aspects: EGFR-dependent and EGFRindependent [16]. Moreover, previous studies have shown that the mechanism of resistance after rst-line or second-line use of Osimertinib varies with clonal evolution [17]. Our study focused on patients with advanced NSCLC who used non-rst-line Osimertinib. Patients enrolled in the sensitive group (GR) were those who were sensitive to Osimertinib after Ge tinib resistance, while patients in the resistant group (OR) were those who had developed clinical resistance to Osimertinib. This could ensure the homogeneity of these two groups of samples.
The most common drug-resistance mechanism that dependent on EGFR mutations is C797S mutation occurring on exon 20, which has previously been reported to account for 10-26% of second-line Osimertinib resistance [18]. G796 mutation adjacent to C797S, in addition, there are multiple mutation sites such as L792, L718, G719, G724 and EGF overexpression [19]. However, in our study, of the 27 patients with Osimertinib resistance, only 3 (11.11%) had the C797S mutation con rmed by molecular testing. Of course, there were also a large number of patients who did not perform genetic testing again due to economic reasons.
Among the mechanisms of drug resistance independent of EGFR, mainly are activation of bypass signals, abnormalities in downstream pathways, and histological transformation. The most common in activated bypass pathways are MET ampli cation [20] and HER2 ampli cation [21]. Preclinical studies have shown that Osimertinib combined with Crizotinib, a c-MET inhibitor, can overcome MET ampli cation after Osimertinib resistance [22]. The abnormality of RAS-MAPK pathway is an important mechanism of Osimertinib resistance. Ortiz-Cuaran et al. [23]con rmed that when acquired resistance to second-line Osimertinib was developed, re-biopsy of the tumor revealed a KRAS G12S mutation. Kim et al. [24]reported a case of MAPK1 mRNA overexpression in a patient who received second-line treatment with Osimertinib in advanced stages. This is consistent with our sequencing results of exosomes. In Fig. 2F, the most obvious pathway enriched by the known differential miRNA target gene of exosomes in the supernatants of drug-resistant (H1975-OR) and sensitive (H1975) strains is Ras signaling pathway. As shown in supplementary Fig. 1, MAPK signaling pathway is the most abundant one among the novel differential miRNA target genes with the largest number of genes involved. This indicates that the mechanism of exosomal miRNA involvement in Osimertinib resistance mainly affects the abnormality of the RAS-MAPK pathway. In addition, the PI3K pathway has a place in bypass activation. It is currently believed that PIK3CA mutation or ampli cation and PTEN deletion can lead to PI3K pathway activation [24]. E545K, E542K, R88Q, N345K, and E418K mutations were related to second-line Osimertinib resistance. E545K mutation was the most common and was veri ed in vitro [25]. In our study, the PI3K-Akt signaling pathway was found in both the cell supernatant and the exosome miRNA target gene enrichment pathway in patient serum. Exosomal miRNA can convey Osimertinib resistance information to affect PI3K pathway activation. Osimertinib resistance is also related to changes in cell cycle genes, including cyclin D1, cyclin D2, cyclin E1, cyclin-dependent kinase (CDK) 4 and CDK6 [26]. In the sequencing results, exosome miR-6087 increased signi cantly in the drug-resistant group. Its target gene was CCND1 (Table 4), which encodes the cyclin D1 protein. This again demonstrates that exosomes participate in Osimertinib resistance by causing bypass activation. As mentioned above, exosomes contain a large number of proteins, nucleic acids, and lipids, which transmit information between cells [27]. Tumor-derived exosomes can be detected in patients' blood or other body uids. They carry a large number of tumor-derived molecules [10], which can change the molecular phenotype to promote tumor progression and affect tumor microenvironment reconstruction [28]. This also shows that tumor-derived exosomes can be objects of liquid biopsy, re ecting the tumor burden and drug resistance. It has been proved that exosomes can affect the therapeutic response and induce drug resistance of tumor cells [29]. Recent research have suggested that drug-resistant cells transfer drug resistance to drug-sensitive cells through miRNAs and drug e ux mercury [28].
MicroRNAs (miRNAs), as a short non-coding RNA, have been thoroughly studied in the eld of tumors [30]. It can inhibit or relieve protein expression by binding with target mRNA, so as to regulate the occurrence, transformation and drug response of tumor. When miRNAs are loaded into exosomes, they can be protected from degradation by RNases [31]. An interesting surprise was the discovery that exosome miRNAs could assist in the diagnosis of non-small cell lung cancer. We compared the levels of exosomes miR-184, miR-3913-5p, miR-3614-5p and miR-4746-5p in the serum of NSCLC patients and healthy people, and found signi cant differences between the two. The AUC of miR-184 was 0.803 (95% con dence interval: 0.701-0.905), and the area under the curve (AUC) was greater than 0.75, indicating that miR-184 of serum exosomes of lung cancer patients could be used as a biomarker for the diagnosis of NSCLC (Supplementary Fig. 3).
In the study of exosomal resistance, Chen [32] et al. found that miR-222-3p in exosomes can be used as a biomaker for gemcitabine resistance. Qin [33] et al. found that exosomes control DDP resistance by transmitting miR-100-5p. Exosome miRNAs and EGFR-TKI resistance have recently been published. Yan Zhang [34]et al. con rmed that exosomes miR-214 from Ge tinib-resistant PC9-GR could inhibit apoptosis in vitro, in vivo inhibit tumor growth and consequently gain Ge tinib resistance in EGFR-mutant lung cancer. Liu [29] et al. found that miR-522-3p in exosomes from H1975 cells can increase the resistance of Ge tinib to PC9 cells.
However, no published literature has discussed on the relationship between three-generation EGFR-TKI resistance and exosomes. Our study used next-generation sequencing to compare exosomes miRNAs in H1975 and H1975-OR cells in vitro, combined with miRNAs from three pairs of serum exosome of patients before and after drug resistance, and veri ed in another 64 NSCLC patients' serum samples. We found that miR-184 and miR-3913-5p were signi cantly elevated in exosomes after Osimertinib resistance. Previous studies have shown that EGFR-independent mechanisms of Osimertinib resistance include bypass activation of PI3K pathway, which can be achieved through PIK3CA mutation/ampli cation and PTEN deletion. The sequencing results in the exosomes of H1975-OR cell supernatants con rmed that exosome PI3K/Akt pathway activation was involved in Osimertinib resistance (Fig. 6).
In the clinicopathological features, LDH has been proved to be closely related to clinical prognosis in a variety of malignant tumors [35]. NSCLC patients with higher LDH levels have a worse prognosis and shorter survival [36]. Although exosome miR-184 found in our study was related to LDH level (p = 0.018), due to the time limit of this experiment, we could not continue to follow up to obtain survival data, whether exosome miR-184 could be a prognostic indicator is not known. CEA is the most common biomarker of lung adenocarcinoma [37]. It has been con rmed that the increase of CEA level during TKI treatment for EGFR mutation patients may be a more sensitive predictor of the explosive progression of lung adenocarcinoma [38]. Platelet count (PLT) is often associated with platelet-to-lymphocyte ratio (PLR) [39]. Studies have suggested that the preoperative PLT-PLR score could be of great signi cance in predicting the prognosis of patients with surgically resected NSCLC [40]. In this study, we found that Exo-miR-3913-5p was related to TNM stage (p = 0.045), PLT (p = 0.024), CEA (p = 0.045), distant metastasis (p = 0.049) and bone metastasis (P = 0.03). Moreover, AUC in the ROS curve was greater than 0.75, which further suggested that exosome miR-3913-5p level was associated with advanced progression of lung adenocarcinoma in patients with EGFR mutations during the use of TKI.
It could be seen from the Venn diagram analysis that miR-184 was found in the overlap of Treat1 (cell) and Treat4 (A patient with EGFR exon 19 del). MiR-3913-5p was found in the overlap of Treat3 and Treat4, and both patients were patients with EGFR exon 19 deletion. However, we found in these subgroup analyses that both miRNAs were more signi cant in patients with EGFR exon 21 L858R mutations.
Previous studies have found that exon 19 deletion mutations (55.0%) have a higher rate of T790M resistance mutations than exon 21 L858R point mutations (37.3%) [41]. Recent studies have shown that the hazard ratio of survival bene t for Asian and L858R mutant populations is close to 1.00 in all people receiving Osimertinib treatment [6]. Our study found that these two exosomal miRNAs changed signi cantly in the L858R mutant population, which might indicate that exosomal miRNAs were mainly involved in Osimertinib resistance for patients with EGFR 21 exon mutation .
Previous studies have reported that nearly half of patients would lost T790M mutation at the time of progression on Osimertinib [42], and this loss may be related to the early resistance of Osimertinib. The loss of T790M mutation is not conducive to prognosis [43]. Plasma T790M levels may predict acquired resistance [44]. However, considering that Osimertinib is selective for EGFR sensitivity and T790M mutation, scholars believe that the emergence of T790M under Osimertinib treatment is not a drug resistance mechanism [16]. Our study found that exosomes miR-3913-5p changed signi cantly in T790Mpositive patients, indicating that exosomes miR-3913-5p may be involved in the drug resistance mechanism of T790M-positive patients. Exosomes miR-184 and miR-3913-5p are likely to be important molecules for resistance transmission of Osimertinib.
There are also some limitations and de ciencies in our research. The number of patients validated in this study was 64, and the amount of collected specimens was insu cient. Only patient serum samples were used to extract exosomes, and no further humoral exosomes were used for veri cation. The clinicopathological characteristics of the patients were collected, but no survival analysis was performed due to the limitation of follow-up time. The predicted target genes and pathways will be veri ed in our subsequent experiments.
Our experiments further explain the mechanism by which exosomes are involved in Osimertinib resistance. This study helps to further expand the application of liquid biopsy in the eld of clinical lung cancer drug resistance. It will provide research directions for the design of a new generation of targeted drugs and exosome drug-loaded therapy for advanced NSCLC patients with EGFR mutations in the future.

Conclusions
Exosomes miRNAs derived from Osimertinib resistant strains are signi cantly different from sensitive strains. The mechanism of exosomes miRNAs involved in Osimertinib resistance is due to the activation of the bypass pathways (RAS-MAPK pathway abnormality, PI3K pathway activation). After patients obtained Osimertinib resistance, exosomes miR-184 and miR-3913-5p increased signi cantly in serum.

Declarations
Ethics approval and consent to participate The research protocol was reviewed and approved by the Ethical Committee and Institutional Review Board of the Jinling Hospital a liated to Nanjing University School of Medicine, and written informed consent was obtained from each patient included in the study.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
All the authors do not have any possible con icts of interest.   Exosomal miRNAs participate in Osimertinib resistance mainly through bypass activation mechanisms in NSCLC Supplementary Files