LncRNA APCDD1L-AS1 induces icotinib resistance by inhibition of EGFR autophagic degradation via the miR-1322/miR-1972/miR-324-3p-SIRT5 axis in lung adenocarcinoma

Epidermal growth factor receptor-tyrosinase kinase inhibitor (EGFR-TKI) resistance is the major obstacle in the treatment of lung adenocarcinoma (LUAD) patients harboring EGFR-sensitive mutations. However, the long non-coding RNAs (lncRNAs) related to EGFR-TKIs resistance and their functional mechanisms are still largely unknown. This study aimed to investigate the role and regulatory mechanism of lncRNA APCDD1L-AS1 in icotinib resistance of lung cancer. Molecular approaches including qRT-PCR, MTT assay, colony formation, RNA interference and cell transfection, RNA immunoprecipitation (RIP), dual luciferase reporter assay, RNA fluorescence in situ hybridization, TUNEL assay, flow cytometry, immunoblotting, xenograft model and transcriptome sequencing were used to investigate the mechanism of APCDD1L-AS1 in icotinib resistance. A novel lncRNA, APCDD1L-AS1 was identified as the most significantly upregulated lncRNA in icotinib-resistant LUAD cells by the transcriptome sequencing and differential lncRNA expression analysis. We found that APCDD1L-AS1 not only promoted icotinib resistance, but also upregulated the protein expression level of EGFR. Mechanistically, APCDD1L-AS1 promoted icotinib resistance and EGFR upregulation by sponging with miR-1322/miR-1972/miR-324-3p to remove the transcription inhibition of SIRT5. Furthermore, SIRT5 elevated EGFR expression and activation by inhibiting the autophagic degradation of EGFR, finally promoting icotinib resistance. Consistently, the autophagy initiator rapamycin could decrease EGFR levels and increase the sensitivity of icotinib-resistant LUAD cells to icotinib. APCDD1L-AS1 could promote icotinib resistance by inhibiting autophagic degradation of EGFR via the miR-1322/miR-1972/miR-324-3p-SIRT5 axis. The combination of autophagy initiator and EGFR-TKIs might serve as a potential new strategy for overcoming EGFR-TKIs resistance in LUAD patients.


Background
Lung cancer is one of the most common malignancies and the leading cause of cancer-related deaths worldwide [1,2]. In the past ten years, the development of targeted therapy has greatly improved the survival of lung cancer patients, especially those with lung adenocarcinoma (LUAD). Approximately 40% of LUAD patients in Asian carry the sensitive mutations in epidermal growth factor receptor (EGFR) [3]. EGFR-TKIs, such as gefitinib, erlotinib and icotinib, have been recommended worldwide as a standard first-line regimen. However, since the majority of the patients acquired drug resistance after initial response to EGFR-TKIs within 10-16 months [4], EGFR-TKIs resistance has become a major obstacle in the treatment of LUAD. The EGFR (T790M) mutation and gene amplification of MET are known to be the main molecular mechanisms of resistance to EGFR-TKIs, whereas the third-generation EGFR-TKIs (Osimertinib) or the combination of MET inhibitors (crizotinib) are wildly used for the therapy of these patients [5]. We and others have reported that the PI3K-AKT-mTOR or STAT3 signaling pathway also contributes to EGFR-TKIs resistance, and the mTOR or STAT3 inhibition could partially reverse this phenomenon [6][7][8][9]. In addition, we recently reported that Grb2 was highly expressed in icotinib-resistant LUAD cells, whereas and the combination of Grb2 inhibitor lymecycline could enhance the sensitivity to icotinib [6]. However, a large number of patients still have no effective strategy for overcoming EGFR-TKI resistance [10]. Therefore, it is urgent to further explore the mechanism of EGFR-TKIs resistance, and develop more effective therapeutic strategies in the treatment of LUAD.
With the development of next generation sequencing technologies, it has been well known that only < 2% of the human transcriptional products encode proteins, whereas the remaining 98% are non-coding RNAs (ncRNAs) lacking protein-coding potential [11]. Long non-coding RNAs (lncRNAs) are a novel class of ncRNAs with more than 200 nucleotides in length [12][13][14]. LncRNAs are known to be involved in various cellular processes such as cell differentiation, autophagy and apoptosis [15]. To date, numerous lncRNAs are recognized as novel biomarkers and therapeutic targets in the diagnosis and treatment of malignancies. Furthermore, mounting studies demonstrated that lncRNAs play a pivotal role in conferring anticancer drug resistance [16,17]. LncRNA GSTM3TV2 could promote gemcitabine resistance by sponging let-7 with concomitant upregulation of LAT2 and OLR1 in pancreatic cancer [18]; LncRNAH19 induced autophagy and this could contribute to tamoxifen resistance via H19/SAHH/DNMT3B axis in breast cancer [19]; MALAT1 contributed to gefitinib resistance by sponging miR-200a and enhancing ZEB1 expression [20]. However, the studies about EGFR-TKI resistance related lncRNAs are still limited, and their functional mechanisms remain largely unknown.
In the current study, we identified a novel icotinib resistance-related lncRNA, APCDD1L-AS1, and found that APCDD1L-AS1 could upregulateSIRT5 expression by sponging with miR-1322/miR-1972/miR-324-3pin LUAD cells. SIRT5 inhibited autophagic degradation of EGFR to induce icotinib resistance. An autophagy initiator could perfectly reverse icotinib resistance in LUAD cells. Therefore, this study revealed a novel lncRNAmediated mechanism of icotinib resistance, and provided a potential strategy for overcoming icotinib resistance in LUAD patients.

RNA extraction and qRT-PCR
Total RNA was purified using Trizol reagent (Invitrogen, Carlsbad, CA, USA). PARIS™ Kit (Invitrogen, AM1921) was utilized to isolate nuclear and cytoplasmic RNA according to the manufacturer's instructions. RNA was reverse-transcribed to cDNA by the PrimeScript™ RT reagent kit (Takara, Tokyo, Japan) and the One Step Pri-meScript® miRNAcDNA Synthesis Kit's (TaKaRa, Japan) following the protocols. Quantitative real-time PCR was performed with the SYBR Premix ExTaq II kit (TaKaRa, Japan) and the Applied Biosystems 7500 Fluorescent Quantitative PCR system (Applied BiosystemsLife Technologies, Carlsbad, CA, USA). The 2 −ΔΔCt method was used to calculate the fold change of the relative mRNA expression. U6 or 18S were used as an internal control for normalization. The primer sequences were provided in the supplemental material (Additional file 1: Table  S1).

MTT assay for Icotinib-sensitivity and cell viability
The icotinib-resistant LUAD cells or their parental cells with or without transfection were seed into 96-well plates (2-3 × 10 3 cells/well), and treated with different concentrations of icotinib for 96 h, followed by the absorbance (570 nm) assessment using the 3-(4, 5dimethyl-2-thiazolyl)-2, 5-diphenyl-2-H-tetrazolium bromide (MTT, Sigma-Aldrich, USA). Then the 50% inhibition of growth (IC50) value for icotinb was calculated in each cell line as previously described [21]. For cell viability assay, the icotinib-resistant lung adenocarcinoma cells were treated with 10 μM of icotinib for 24 h, 48 h, 72 h and 96 h. The cell viability was also assayed using MTT assay. All experiments were performed in triplicate.

Colony formation
Cells were seeded into 12-well plate (600 cells/well) in standard medium. Next day, different concentrations of icotinib were added into the medium. Two weeks later, live cells were stained using crystal violet (Sigma-Aldrich, USA), and the number of colonies in each well was counted.

Xenograft studies
The icotinib-resistant or parental lung adenocarcionma cells (150 μl, approximately 1.0 × 10 7 cells) were injected subcutaneously into female nude mice (4-week old, Beijing Vital River Laboratory Animal Technology Co., Ltd., China). Two weeks later, the tumor-bearing mice were orally treated with icotinib (50 mg/kg) in 1% Tween-80 saline solution every three days, and the tumor volumes were assessed (0.5 × length×width 2 ) per week for 3 consecutive weeks. After all mice were euthanized, the tumors were surgically removed and photographed.
To explore the function of APCDD1L-AS1 in vivo, the icotinib-resistant lung adenocarcinoma cells (150 μl, approximately 1.0 × 10 7 cells) with stable Lv-sh RNA-APCDD1L-AS1 or Lv-NC were injected subcutaneously into mice. The dosage of icotinib, treatment method and duration were same as above. Partial tumor tissues were immediately frozen in liquid nitrogen for RNA extraction, and others were fixed in 10% buffered formalin for immunohistochemistry, immunofluorescence and Tunel assay. The antibodies for immunohistochemical staining were EGFR (1:100, Santa, Clara, CA, USA), SIRT5 (1: 100, Sigma-Aldrich, USA). The antibody for immunofluorescence staining was LC3B (1:200, Cell Signaling Technology, Danvers, MA, USA). The images were quantified using the default 'Analyze particles' plugin in image analyser (Image Pro Plus 6.0, Media Cybernetics), and results were presented in average LC3B puncta numbers per each cell according to the guidelines of autophagy [22]. Tunel staining was performed on paraffin section using one-step tunnel apoptosis assay kit (Byotime, Jiangsu, China) according to the manufacturer's instructions. The image analyser (Image Pro Plus 6.0, Media Cybernetics) was used to score all the staining sections of Tunel, and results were presented in mean gray value [23]. This study was performed in animal laboratory center of China Medical University and was granted by the Institutional Review Board of China Medical University (IACUC Issue No: 2019067).

RNA immunoprecipitation (RIP)
The procedure was performed using Magna RNAbinding protein immunoprecipitation kit (Millipore, Billerica, MA, USA) as per the manufacturer's protocol. Human anti-AGO2 antibody (Abcam, Cambridge, USA) and anti-rabbit IgG (Millipore, USA) were used for RIP. The immunoprecipitated RNAs were isolated and detected by qRT-PCR. The gene-specific primer sequences used for the qRT-PCR of the RIP assay were listed in the supplemental material (Additional file 1: Table S1).

RNA fluorescence in situ hybridization
Cy3-labeled APCDD1L-AS1 was purchased from Servi-ceBio (Wuhan, China). RNA FISH were performed as described previously using Fluorescent in Situ Hybridization kit (ServiceBio, China) following the manufacturer's instructions. The slides were observed for immunofluorescence with a fluorescence microscope (Eclipse Ti-SR, Nikon). U6 and 18S probes (ServiceBio, China) were used as internal reference of nuclear and cytoplasmic RNA.

Flow cytometry for the detection of apoptosis
Icotinib-resistant LUAD cells were treated with icotinib (5 μM or 10 μM) for 72 h. Then both floating and attached cells were harvested, and staining with Annexin V-FITC Apoptosis Detection kit. Cells apoptosis were analyzed by FACS cytometry (BD Biosciences Inc., Franklin, NJ, USA).

Statistical analysis
The data were shown as mean ± standard deviation (SD). The paired Student's t test or One-Way ANOVA was used to compare the differences among two or multiple groups. A two-tailed P < 0.05 was deemed to be statistically significant while P < 0.01 was very significant. All data analyses were carried out with GraphPad Prism software 5.0 (GraphPad Software, Inc., San Diego, CA, USA) and SPSS 16.0 (IBM, SPSS, Chicago, IL, USA).

APCDD1L-AS1 was significantly up-regulated in icotinibresistant LUAD cells
Firstly, the icotinib sensitivity was compared in parental (PC9 and HCC827) and icotinib-resistant cell lines (PC9/IcoRL, PC9/IcoRH, HCC827/IcoRL and HCC827/ IcoRH). MTT assays showed that icotinib sensitivity was much lower inicotinib-resistant cells than their parental cells in a dose-and time-dependent manner ( Fig. 1a-b). The colony formation assay also showed that the colonies of the parental cells were significantly smaller and less than those of icotinib-resistant cells after icotinib treatment (Fig. 1c). Similarly, subcutaneous xenograft tumors of PC9 cell group showed smaller size than those of PC9/IcoRL cell group after icotinib treatment (Fig. 1d). Additionally, the protein and the phosphorylation levels of EGFR were significantly higher in icotinib-resistant cells than parental cells (Fig. 1e). All these results confirmed the icotinib-resistant phenotype in the LUAD cells used in our study.
To screen the lncRNAs related to icotinib resistance, we performed the global expression profiles analysis of lncRNA in PC9, PC9/IcoRL and PC9/IcoRH cells. Among the differentially expressed lncRNAs, APCD D1L-AS1 was the most up-regulated lncRNA in both PC9/IcoRL and PC9/IcoRH cells compared to PC9 cells ( Fig. 1f-g). Similar results were also verified by qRT-PCR (Fig. 1h). Due to the significant upregulation of APCD D1L-AS1, we decided to further investigate its potential role in icotinib resistance.
APCDD1L-AS1 contributed to the icotinib resistance of LUAD cells APCDD1L-AS1, located at chromosomal 20q13.32 and composed of seven exons with a full length of 2099nucleotide (nt) (Fig. 2a), has been identified as an lncRNA in the LNCipedia database (version 5.2). The prediction result of Coding Potential Assessment Tool further confirmed that APCDD1L-AS1 only possessed a very low coding potential (Fig. 2b). To explore the subcellular localization of APCDD1L-AS1, we searched the database lncLocator and the results predicted the predominant cytoplasmic localization of APCDD1L-AS1 with minor nuclear localization (Fig. 2c). This result was further verified by the nuclear/cytoplasmic RNA separation analysis using qRT-PCR and RNA fluorescence in situ hybridization (FISH) assay in icotinib-resistant LUAD cells (Fig. 2d-e; Additional file 1: Figure S1A and S1B).  were calculated using CPAT database. c The online software lncLocator was used to predict the subcellular localization of APCDD1L-AS1. d Relative expression of APCDD1L-AS1 in cytoplasm or nucleus of the icotinib-resistant cells (PC9/IcoRH, HCC827/IcoRH) was determined by qRT-PCR. e The localization of APCDD1-AS1 in the PC9/IcoRH cells was detected by RNA-FISH. Blue, DAPI-stained nuclei; Red, Cy3-labeled positive hybridization signals (scale bar, 100 μm). U6 and 18S were used as positive controls. f The effect of APCDD1L-AS1 KD on the IC50 value of icotinib was evaluated in icotinib-resistant cells (PC9/IcoRH, HCC827/IcoRH) by MTT assay. g The level of EGFR and p-EGFR in APCDD1L-AS1-KD or -OE icotinib-resistant cells (PC9/IcoRH, HCC827/IcoRH) was determined by western blot. h Kaplan-Meier analyses of the correlations between APCD D1L-AS1 expression (classified into high and low expression groups according to the median of APCDD1L-AS1 expression) and OS in 672 lung adenocarcinoma patients using Kaplan-Meier Plotter online database. Log rank test was used to calculate P values. i-j The apoptosis of APCDD1L-AS1-KD icotinib-resistant cells (PC9/IcoRL, PC9/IcoRH) induced by icotinib (10 μM) was analyzed using flow cytometry (i), and apoptosis-related protein PARP in PC9/IcoRL and PC9/IcoRH cells was detected by western blot (j). GAPDH was used as the internal control. The mean ± SD of triplicate experiments were plotted, *P < 0.05, **P < 0.01, ***P < 0.001 To determine whether APCDD1L-AS1was involved in icotinib resistance, we transiently knocked down or overexpressed APCDD1L-AS1 in icotinib-resistant cells (Additional file 1: Figure S1C and S1D). The result showed that APCDD1L-AS1 knock down (KD) could significantly enhance the icotinib sensitivity ( Fig. 2f; Additional file 1: Figure S1E) and decrease the protein and phosphorylation levels of EGFR ( Fig. 2g; Additional file 1: Figure S1F). In contrast, APCDD1L-AS1 overexpression (OE) increased the expression and activation of EGFR ( Fig. 2g; Additional file 1: Figure S1F). The similar results were also obtained with gefitinib treatment (Additional file 1: Figure S1G). Kaplan-Meier survival analysis of 672 LUAD patients showed that high expression of APCDD1L-AS1 was significantly associated with worse overall survival (OS) (log rank test, p = 0.0048; Fig. 2h). Additionally, flow cytometry and western blot analyses confirmed that APCDD1L-AS1KD promoted icotinib-induced apoptosis in icotinib-resistant cells (Fig. 2i-j; Additional file 1: Figure S1H and S1I). Collectively, these results supported a role for APCD D1L-AS1 in EGFR upregulation and icotinitib resistance.
Results of qRT-PCR showed that the levels of all three miRNAs were lower in icotinib-resistant cells compared to parental cells ( Fig. 3e; Additional file 1: Figure S2A). In contrast, the levels of all these miRNAs were significantly increased in APCDD1L-AS1-KD cells, while decreased in APCDD1L-AS1-OE cells (Additional file 1: Figure S2B and S2C; Figure S3A and S3B). On the other hand, the mimics of miR-1322, miR-1972 and miR-324-3p decreased, whereas their inhibitors increased the level of APCDD1L-AS1 (Additional file 1: Figure S3C). All these results indicated that APCDD1L-AS1 could sponge with miR-1322, miR-1972 and miR-324-3p in a reciprocally suppressive manner.

SIRT5 contributed to icotinib resistance in LUAD cells
To verify the role of SIRT5 in icotinib resistance, SIRT5 was knocked down in icotinib-resistant LUAD cells (Additional file 1: Figure S6A). The MTT assays showed that SIRT5-KD significantly reduced the IC50 values for icotinib in icotinib-resistant cells (Fig. 5a). Meanwhile, both protein and phosphorylation levels of EGFR were significantly decreased in SIRT5-KD icotinib-resistant cells ( Fig. 5b; Additional file 1: Figure S6B). In addition, flow cytometry and western blot analyses indicated that SIRT5 KD significantly increased apoptosis in icotinibresistant cells (Fig. 5c-d). These results indicated that SIRT5 contributed to EGFR upregulation following icotinib resistance.

SIRT5 up-regulated EGFR by inhibiting autophagic degradation
The next question is how SIRT5 up-regulated EGFR expression. No changes of EGFR mRNA levels was observed in SIRT5-KD cells (Additional file 1: Figure  S7A) indicating that SIRT5 might increase EGFR protein synthesis and/or inhibiting EGFR degradation. Then, after treated with the protein synthesis inhibitor cyclohexane (CHX), the protein level of EGFR in SIRT5-KD icotinib-resistant cells was analyzed by western blot. The results showed that SIRT5-KD dramatically shortened the half-life of EGFR ( Fig. 6a; Additional file 1: Figure S7B), suggesting that SIRT5 might upregulate EGFR by inhibiting its degradation. However, the proteasome inhibitor MG-132 could not elevate EGFR levels in SIRT5-KD cells, excluding the possible involvement of proteasome pathway in the process ( Fig. 6b; Additional file 1: Figure S7C). Autophagy is another well known pathway for the degradation of damaged proteins. Therefore, we determined to test whether EGFR undergone the autophagic degradation in the SIRT5-KD cells. Our results showed the decreased level of p62, and increased level of LC3B-II in SIRT5-KD cells (Fig. 6c, Additional file 1: Figure S7D), indicating that SIRT5 KD could promote autophagic flux in icotinibresistant LUAD cells. Furthermore, two autophagy inhibitors, CQ and 3-MA, partially rescued SIRT5-KD-induced phenotypes including EGFR downregulation, decreased EGFR activation and increased PARP cleavage (Fig. 6d-e; Additional file 1: Figure S7E and S8A). On the contrary, the combination of autophagy initiator rapamycin with icotinib partially reversed the icotinib resistance of LUAD cells (Fig. 6f). More importantly, the images of confocal microscopy showed that EGFR partially colocalized with the LC3B autophagic vesicle marker under SIRT5 knockdown in HCC827/IcoRH cells (Additional file 1: Figure S8B). Taken together, these results suggested that SIRT5 promoted icotinib resistance by inhibiting EGFR autophagic degradation.

Discussion
In the present study, we found the upregulation of a novel lncRNA APCDD1L-AS1 in icotinib-resistant LUAD cells, and proposed a novel working model wherein APCDD1L-AS1 induced icotinib resistance by inhibiting autophagic degradation of EGFR and Aberrant expressions of numerous lncRNAs have been found in various cancers, including lung cancer. The ceRNA network composed of lncRNAs-miRNA-mRNA is one of the main mechanisms promoting cancer progression. A single lncRNA can interact with single or multiple miRNAs, and a single miRNA can also interact with single or multiple lncRNAs or mRNAs. Therefore, the ceRNA network is highly complex [27]. Studies concerning EGFR-TKI resistance in NSCLC reported that lncRNA LOC554202 upregulating miR-31 could reduce sensitivity of NSCLC cells to gefitinib [28]. The lncRNA RHPN1-AS1 downregulation promoted gefitinib resistance by targeting the miR-299-3p/TNFSF12 pathway in NSCLC [29]. However, these studies just focused on individual ceRNA interactions. In this study, based on the prediction of target miRNAs and mRNAs and the confirmation of luciferase reporter assay and RIP, we found that APCDD1L-AS1 could simultaneously sponge with three miRNAs (miR-1972, miR-1322 and miR-324-3p), and the three miRNAs targeted the common gene, SIRT5, establishing a stable and powerful ceRNA network. As a result, APCDD1L-AS1 exerted strong drug resistance-promoting function through this ceRNA network. We found that SIRT5 could promote EGFR-TKI resistance. SIRT5, a member of Sirtuin family, is known to play important roles in autophagy, apoptosis and drug resistance [30,31], A study based on online analysis indicated that SIRT5 was positively correlated with complete response to neoadjuvant chemotherapy in triple negative breast cancer, suggesting that SIRT5 might play an inhibitory role in drug resistance [32]. However, it was also reported that SIRT5-positive colorectal cancer cells with wild-type KRAS were resistant to either chemotherapeutic agents or cetuximab [33]; SIRT5 promoted cisplatin resistance in ovarian cancer. In this study, we found the upregulation of SIRT5 in icotinib-resistant LUAD cells and gefitinib-resistant tumors from GSE80344 dataset, supporting that the resistance promoting role of SIRT5. Also, the findings that SIRT5-KD could promote autophagic degradation of EGFR and enhance icotinib sensitivity provided further evidence that SIRT5 functioned as a resistant factor by inhibiting autophagy. SIRT5 has been reported to regulate autophagy in a context-dependent manner. SIRT5 could promote autophagy in CRC [30], and suppress autophagy in human breast cancer cells [34]. Our results showed that SIRT5 functioned as an autophagy inhibitor, and were consistent with a previous study showing that SIRT5 inhibited autophagy by decreasing ammonia production [34]. Therefore, SIRT5 functioned as an autophagy suppressor in the context of EGFR-TKI resistance.
Autophagy is a conserved catabolic process resulting in self-digestion and the removal of dysfunctional proteins and organelles [35]. Accumulating evidence supported that autophagy, as a double-edged sword, could either induce cell death or cell protection, and might therefore be associated with drug resistance of cancer cells [36]. It was reported that the knockdown of Rab5a or CCL2 could stimulate autophagy to reverse cisplatin resistance in gastric cancer [37]. Our results that the autophagy inhibitors, CQ and 3-MA, rescued SIRT5-KD-induced EGFR autophagic degradation and promoted icotinib resistance, strongly suggested the effect of SIRT5 on icotinib resistance is mediated by its autophagy inhibitory function. Targeting the WEE1 kinase was reported to strengthen the antitumor activity of imatinib via promoting KIT autophagic degradation in gastrointestinal stromal tumors [38]. In our study, we found that knockdown of SIRT5 could decrease EGFR expression and increased the colocalization of EGFR with autophagic vesicles. Also, SIRT5 knockdown promoted autophagic degradation of EGFR, suggesting that the LncRNA-miR-1322/miR-1972/miR-324-3p-SIRT5 axis inhibited the autophagic degradation of EGFR. As the key target of EGFR-TKI in NSCLC, EGFR was reported to be up-regulated at both protein and phosphorylation levels in EGFR-TKIs resistant cells [39], which was similar to our results. Therefore, accelerating autophagic degradation of EGFR may be the potential strategy for overcoming EGFR-TKI resistance in LUAD. In the current study, mTOR inhibitor rapamycin, which can also induce cell death via promoting autophagy [40], actually did increase icotinib sensitivity in icotinibresistant LUAD cells. Other mTOR inhibitors, such as torin2 and BIBW2992) [8,9] were also reported to be able to induce apoptosis and inhibit cell proliferation in Fig. 9 The schematic diagram of APCDD1L-AS1 contributing to icotinib resistance in lung adenocarcinoma cells. In this progress, up-regulated APCDD1L-AS1 as a miRNA sponge to decoy miR-1322, miR-1972 and miR-324-3p, promote the expression of SIRT5, inhibit autophagic degradation of EGFR, increase EGFR phosphorylation, inhibit apoptosis and induce icotinib resistance EGFR-TKI-resistant NSCLC cells by negative feedback regulation of Akt/mTOR signaling and inducing autophagy, suggesting promising therapeutic strategy in NSCL C with EGFR-TKI resistant phenotype. In addition, a recent study reported that the combination of bisdemethoxy curcumin and icotinib could enhance the sensitivity of primary EGFR-TKI resistant NSCLC cell lines to icotinib via autophagy induction [41], which was similar to ours. Therefore, our study provided strong evidence that manipulating the activity of autophagy might be a useful therapeutic strategy to enhance the drug sensitivity in cancer [42,43]. Certainly, due to the complexity of autophagy, the roles of autophagy in icotinib resistance need to be further clarified. Also, it needs to be further addressed whether rapamycin could reverse EGFR-TKI resistance by modulating lncRNA-miR-1322/ miR-1972/ miR-324-3p-SIRT5 axis-mediated autophagy in clinical application in the future.

Conclusions
In summary, this study demonstrated that APCDD1L-AS1 could induce icotinib resistance by inhibiting autophagic degradation of EGFR via a ceRNA network of simultaneously decoying miR-1322, miR-1972 and miR-324-3p to upregulate SIRT5 in LUAD. These findings revealed a novel mechanism of EGFR-TKI resistance, and provided a series of new targets and potential strategy for overcoming EGFR-TKI resistance in LUAD.