Linc00673 and autophagy are up-regulated in cisplatin-resistant NSCLC cells

Since it has been discovered, the biological function of Linc00673 has been well investigated, including its role in tumor metastasis (Lu et al, 2017; Guan et al, 2019), proliferation (Huang et al, 2017; Schmidt et al, 2019), and senescence (Roth et al, 2018). Recent study has identified Linc00673-V4 as the prevailing isoform in lung adenocarcinoma, but the function of other isoforms remains largely unknown. To the best of our knowledge, there is no report about the participation of Linc00673 isoforms in therapeutic chemoresistance. In that case, we started our research by investigating whether Linc00673 was involved in NSCLC chemoresistance and which isoform contributed the most in NSCLC chemoresistance. First, we found that Linc00673 was up-regulated in cisplatin-resistant A549 cells (A549/CDDP) (Fig 1A). We wondered whether this phenomenon was repeatable in WT NSCLC cells, so we exposed the WT NSCLC cell lines A549, H1975 (Fig 1B) and H596 (Fig S1A) under cisplatin treatment in a concentration gradient way. We found that the expression of Linc00673 increased in a dose-dependent manner after treatment of cisplatin. We also observed that A549/CDDP exhibited enhanced LC3B-II transformation and decreased p62 protein level (Fig 1C). In WT NSCLC cell lines (Figs 1D and S1B), we found that the autophagy was induced in response to cisplatin treatment. To further validate this finding and monitor autophagic flux, stubRFP-sensGFP-LC3 vectors were transfected into cells. We observed significantly increased ratio of autophagolysosomes to autophagosomes in A549/CDDP (Figs 1E and S1C) and WT NSCLC cells under the treatment of cisplatin (Figs 1F and S1D–F).

So far, five different isoforms of Linc00673 have been reported in NSCLC, each may have distinctive biological function (Fig S1G) (Guan et al, 2019). We conducted further investigations to determine which isoform responded to cisplatin treatment and correlated with the heightened autophagy activity. The relative expression level of each isoform in WT A549 cells was in consistence with the previous study, as Linc00673-V3 to V5 had relatively high expression level (Guan et al, 2019) (Fig 1H). Intriguingly, Linc00673-V3 was significantly up-regulated in A549/CDDP cells compared with its parental cells, whereas the expression level of Linc00673-V4 and V5 remained unchanged (Fig 1G). Furthermore, similar phenomenon was also observed in WT A549 cells treated with cisplatin (Fig 1H). Considering the potential correlation between Linc00673 and autophagy, we treated WT A549 cells with rapamycin to induce autophagy and we found that Linc00673-V3 was significantly up-regulated by treatment of rapamycin (Fig 1I), which indicated that Linc00673-V3 could be induced upon autophagy induction and may be crucial for autophagy activation. Collectively, these data demonstrated that autophagy and Linc00673-V3 were up-regulated in response to cisplatin treatment in NSCLC cells.

Linc00673-V3 regulated autophagy contributes to chemoresistance of NSCLC cells in vitro

Having found that Linc00673-V3 displayed the most striking change among the five reported isoforms, we next detected its expression in various NSCLC cell lines. We found that Linc00673-V3 was highly expressed in A549 and H1975 cells, whereas its expression level in H1299 and H1650 were relatively low. In consistence with the results above, A549/CDDP cells had the highest expression level of Linc00673-V3 (Fig 2A). Next, we transfected A549 and H1975 cells that presented relatively high expression levels of Linc00673-V3 with two individual siRNAs against Linc00673-V3 that were validated by our previous works (Lu et al, 2017; Wu et al, 2020) (Fig S2A). Conversely, Linc00673-V3 overexpression assay was performed in H1299 cells which had relatively low Linc0067-V3 expression level, and the fold-change was 148 post overexpression (Fig S2B). As shown in Fig 2B and C, Linc00673-V3 overexpression significantly enhanced the resistance of H1299 and H596 cells towards the cytotoxicity of cisplatin, whereas knockdown Linc00673-V3 in A549/CDDP, A549, and H1975 cells significantly reduced the resistance (Fig 2D–F). Considering the essential role of autophagy in chemoresistance, we treated NSCLC cells with autophagy inhibitor chloroquine (CQ) and 3-methyladenine (3-MA). When autophagy was blocked, NSCLC cells were more sensitive to the cytotoxicity of cisplatin (Fig 2G–I). Taken together, these data suggested that Linc00673-V3 associated autophagy activity participated in the chemoresistance of NSCLC cells.

Linc00673-V3 increases the transcription of LC3B

To further investigate the role of Linc00673-V3 in autophagy, we knocked down Linc00673-V3 in A549 and H1975 cells and found the protein level of both LC3B-I and LC3B-II significantly decreased (Figs 3A and S3A). Similar phenomenon was validated by immunofluorescence analysis, which showed decreased LC3B puncta formation after knockdown of Linc00673-V3 (Figs 3B and S3B). Moreover, rapamycin was unable to increase LC3B-II level in neither A549 nor H1975 cells knocked down by Linc00673-V3 (Fig 3A and B). There were two possibilities for this phenomenon. One was that knockdown of Linc00673-V3 may accelerate autophagic degradation, which led to the reduced protein level of LC3B. Another possibility was that knockdown of Linc00673-V3 could directly inhibit the initiation of autophagic process. We first treated cells with CQ to study the autophagic flux. Cells treated with CQ alone exhibited massive LC3B-II accumulation, whereas cells treated with both si-Linc00673-V3 and CQ did not show the same phenomenon (Figs 3C and S3H and I), suggesting the impairment of autophagosome biogenesis upon Linc00673-V3 knockdown. Conversely, when cells were overexpressed with Linc00673-V3, the protein level of LC3B was strikingly increased (Fig 3D). Moreover, LC3B lipidation and p62 degradation were markedly enhanced (Fig 3D) and immunofluorescence analysis suggested increased LC3B puncta formation after overexpressing Linc00673-V3 (Figs 3E and S3C). These findings together suggested that Linc00673-V3 was involved in autophagosome biogenesis and this may associate with the protein level of LC3B. To verify which isoform of Linc00673 was involved in autophagy, we knocked down overall Linc00673 and then overexpressed the five isoforms, respectively. We found that out of the five isoforms, only overexpression of Linc00673-V3 could significantly rescue the impairment of autophagy due to Linc00673 knockdown (Figs 3F and S3D).

Because it has been proved that knockdown of Linc00673-V3 could inhibit the initiation of autophagy, we further explored the potential mechanism underlying this phenomenon. We detected the expression of a series of ATG proteins required for autophagosome biogenesis and found that only LC3B was significantly reduced after knockdown of Linc00673-V3 (Figs 3G and S3J). We further detected the mRNA level of each gene and observed consistent result (Figs 3H and S3K). As we found the up-regulation of Linc00673-V3 under the treatment of cisplatin and the association between Linc00673-V3 and LC3B mRNA level (Fig S3E), we further explore the putative correlation between cisplatin treatment and LC3B transcription. The results showed that compared with the parental cells, A549/CDDP had higher LC3B transcript level (Fig S3F). When treated WT NSCLC cells with cisplatin, the mRNA level of LC3B were also elevated (Fig S3G). Taken together, these data suggested that Linc00673-V3 may participate in LC3B transcription regulation in NSCLC.

Linc00673-V3 directly interacts with Smad3 in NSCLC

To elucidate the underlying mechanism of which Linc00673-V3 up-regulated the mRNA level of LC3B in NSCLC, we performed RNA affinity pulldown (RAP) assay to identify the proteins that directly interacted with Linc00673-V3. The RNA-binding proteins (RBP) pulled down by biotin-labeled sense- or antisense-Linc00673-V3 were analyzed by silver staining and the bands around 55KD were different (as indicated by red arrows) (Fig 4A). The pulldown RBPs were then analyzed by mass spectrometry (Table S1). We overlapped the pulldown proteins with genes that were highly expressed in NSCLC tissues from TCGA and potential upstream transcriptional factors of LC3B predicted by AnimalTFDB, a comprehensive transcription factor database (Fig 4B) (Hu et al, 2019). It turned out that there were eight candidate proteins that met the above screening criteria (Fig 4C). We were particularly interested in Smad3 because its well-known function as a transcription factor and its implications in autophagy (Shim et al, 2021; Yang et al, 2021), whereas its exact functions and underlying mechanisms in autophagy remain unclear.

We subsequently tested whether Smad3 was participated in Linc00673-V3 induced autophagy and chemoresistance in NSCLC. First, we validated the interaction between Linc00673-V3 and Smad3 in all the RBPs. It turned out that there was a significant enrichment in the samples pulled down by the Linc00673-V3 sense sequence but not the antisense sequence (Fig 4D). Consistently, the results of RNA immunoprecipitation (RIP) assay demonstrated greater enrichment of Linc00673-V3 captured by Smad3 specific antibody (Fig 4E). It was also noteworthy that there was slight enrichment of other isoforms. Given the similarity of the sequence and structure of those isoforms, such enrichment was also reasonable.

Based on the minimum free energy fold algorithm, we used the bioinformatic tool RNAfold to predict the secondary structure of Linc00673-V3 (Fig 4F). To identify the exact fragment that bound to Smad3, cross-linking RIP followed by RT-qPCR (CLIP-qPCR) was performed. As shown in Fig 4G, 23 RT-qPCR primers were designed for every 100-nt region of Linc00673-V3 (primers used in CLIP assay were listed in Table S2) and the CLIP-qPCR results showed that primer pair 1 and pair 23 were successfully amplified in both A549 and H1975 cells (Fig 4H), which were referred to as the first 1–100 and last 2,200–2,275 fragments accordingly. These two amplified regions were shown in the secondary structure of Linc00673-V3 (Fig 4F), and we found that these two regions were spatially close and formed stem-loop structures, which may partially explain the results of CLIP-qPCR assay. To further validate the exact region of Linc00673-V3 bound to Smad3, vectors carrying full-length or truncation mutations of Linc00673-V3 (V3-FL [1–2,275 nt], V3-T1 [100–2,175 nt], V3-T2 [1–2,175 nt], V3-T3 [100–2,275 nt], and V3-T4 [antisense sequence]) were constructed (Fig 4I). RAP assay was performed, only full-length Linc00673-V3 could bind to Smad3 (Fig 4I), suggesting that both 1–100-nt and 2,175–2,275-nt regions were essential for the interaction of Linc00673-V3 with Smad3.

Subcellular localization of LncRNA was thought to be correlated with its biological functions, thus RNA FISH and cellular fractionation assay were performed. Consistent with previous studies, Linc00673-V3 was mostly found in the cytoplasm of NSCLC cells (Fig S4A) and the results of RNA FISH also indicated the potential colocalization of Linc00673-V3 and Smad3 in cytoplasm (Fig 4J). Considering the transcriptional regulation capacity of Smad3, we further investigated whether Linc00673-V3 and Smad3 would regulate each other’s transcription reciprocally. We found that there was no significant difference in transcriptional level of Linc00673-V3 after Smad3 knockdown or overexpression (Fig S4B). Besides, Linc00673-V3 knockdown or overexpression also did not alter the transcriptional level of Smad3 (Fig S4C). Collectively, these results indicated the direct binding of Linc00673-V3 to Smad3 protein, and this interaction might be mediated by the stem-loop regions formed by the first 1–100-nt and the last 2,200–2,275-nt fragments of Linc00673-V3.

Linc00673-V3 elevates Smad3 protein level via inhibiting its ubiquitination in NSCLC

As Linc00673-V3 and Smad3 did not influence each other’s transcription, we further explore whether Linc00673-V3 could influence the Smad3 protein level. We observed the increased protein level of Smad3 after overexpressing Linc00673-V3 in H1299 cells (Fig 5A). Given that phospho-Smad3 (p-Smad3, Ser423/425) is its active form in terms of transcriptional regulation, we further detected the protein level of p-Smad3 in H1299. It turned out that overexpressing Linc00673-V3 could also increase the protein level of p-Smad3 (Fig 5A). Similar result was observed in HeLa cells (Fig 5B), which suggested that Linc00673-V3 could also influence the protein level of Smad3 in other cancer types. Consistently, Linc00673-V3 knockdown led to reduced protein levels of both Smad3 and p-Smad3 in A549 and H1975 cells (Fig 5C). These data thus indicated a positive regulatory role of Linc00673 in Smad3 protein level.

Because Linc00673-V3 has no effects on the transcription of Smad3 (Fig S4C), we thus asked whether it influenced the degradation of Smad3. To this end, we treated A549 and H1975 cells with cycloheximide to block de novo protein synthesis and found that knockdown of Linc00673-V3 significantly accelerated the degradation of both Smad3 and p-Smad3 (Fig 5D and E). Ubiquitin–proteasome system and autophagy are two major protein quality control pathways responsible for protein degradation and intracellular homeostasis (Silva-Fisher et al, 2020; Lu et al, 2021). We treated cells with the autophagy inhibitor CQ or the proteasome inhibitor MG132. Treatment of CQ did not attenuate the degradation of either Smad3 or p-Smad3 caused by Linc00673-V3 knockdown (Fig S5A and B). However, treatment of MG132 strikingly restored the protein levels of both Smad3 and p-Smad3 (Fig 5F), indicating Linc00673-V3 may stabilize Smad3 via preventing its degradation by the ubiquitin–proteasome system. As expected, the ubiquitination level of Smad3 dramatically increased after Linc00673 knockdown (Figs 5G and H and S5C and D). Taken together, these data demonstrated that the interaction between Linc00673-V3 and Smad3 could inhibit the ubiquitination of Smad3 and prevent Smad3 degradation through the ubiquitin–proteasome system.

Linc00673-V3 stabilizes Smad3 by counteracting the interaction between Smad3 and STUB1

Having established that Linc00673-V3 modulated Smad3 ubiquitination and degradation, we next aimed to study whether Linc00673-V3 disrupted the activity of certain E3 ligase that targeted Smad3. Several E3 ligases including Nedd4L (Gao et al, 2009; Louzada et al, 2021), STUB1 (Li et al, 2004; Shang et al, 2014; Shi et al, 2020), and VHL (Wang et al, 2016; Zhou et al, 2022) have been reported to target Smad3 for ubiquitination. To verify their effects on Smad3 protein level, siRNA has been designed to target each E3 ligase (Fig S6A and B). Knockdown of STUB1 increased Smad3 in A549 and H1975 cells (Fig 6A and B), whereas knockdown of Nedd4L or VHL had marginal impact on Smad3 protein level in A549 and H1975 cells (Fig S6C–F). Importantly, knockdown of STUB1 strongly increased the protein level of both LC3B-I and LC3B-II (Fig 6A and B). Moreover, the transcription level of LC3B also increased after STUB1 knockdown in A549 and H1975 cells (Fig 6C). Because Linc00673-V3 inhibited the ubiquitination of Smad3, we thus asked whether it disrupted the interaction of Smad3 with STUB1. To this end, we performed immunoprecipitation assay and confirmed the interaction between Smad3 and Nedd4L, STUB1, or VHL (Fig 6D and E). However, the interaction between Smad3 and STUB1 was markedly enhanced upon Linc00673-V3 knockdown in A549 and H1975 cells, whereas the interaction of Smad3 with the other two E3 ligases remained the same under the same conditions (Fig 6D and E). Furthermore, overexpression of STUB1 significantly increased the ubiquitination level of Smad3, whereas Linc00673-V3 overexpression attenuated the effect of STUB1 (Fig 6F and G). Taken together, these results suggested that Linc00673-V3 and STUB1 competitively bound to Smad3, and Linc00673-V3 stabilized Smad3 by preventing STUB1-mediated ubiquitination and proteasome degradation of Smad3.

LC3B is transcriptionally regulated by Smad3 and knockdown of Smad3 suppresses autophagy-associated chemoresistance in NSCLC

Using the bioinformatic tool JASPAR (Fornes et al, 2020), we found two putative Smad3-binding sites in the upstream promoter region of LC3B and they were referred to as region (a) and region (b) as shown in Fig 7A. We then cloned the LC3B promoter region and constructed the luciferase reporter plasmid pGL4-LC3B containing region (a) and region (b) (Fig 7A). The results of dual-luciferase reporter assay showed that overexpression of Smad3 significantly increased the luciferase activity, and this effect was much more striking when a constitutively activated form of Smad3 (Smad3-2SD) was overexpressed (Fig 7B). Consistently, the luciferase activity induced by Smad3 overexpression was strongly suppressed upon treatment of SIS3, a selective inhibitor of Smad3 (Fig 7B).

To figure out the exact binding site of p-Smad3 in the promoter region of LC3B, chromatin immunoprecipitation (ChIP-qPCR) assay was performed. In the absence of additional stimulations, we found the binding of p-Smad3 to region (a) or (b) had no significant difference (Figs 7C and S7A). However, under the treatment of cisplatin (Figs 7D and S7B) or rapamycin (Figs 7E and S7C), the region (b) showed a greater binding capacity with p-Smad3 than region (a). In addition, a reference region (c) within the promoter region of LC3B was chosen to verify the specificity of the interaction of p-Smad3 with region (b). As shown in Fig S7D and E, there was no significant binding in region (c). Based on these findings, we then cloned the (b) region and constructed a luciferase reporter plasmid, pGL4-LC3B(b), and the results of dual-luciferase assay also confirmed the interaction between region (b) and Smad3 (Fig 7F).

Apart from dual-luciferase assay and ChIP-qPCR, we directly detected the LC3B transcription level in A549 and H1975 cells. After overexpressing Smad3-2SD, the transcriptional level of LC3B was elevated, and it could partially attenuate the down-regulation of LC3B mRNA level caused by Linc00673-V3 knockdown (Fig 7G and H). The protein level of LC3B was also detected, and we found that knockdown of Smad3 led to down-regulation of both LC3B-I and LC3B-II protein level, whereas overexpression of Smad3-2SD caused the accumulation of LC3B-I, which were in consistence with the up-regulation of LC3B transcriptional level (Fig 7I and J). We then detected the protein level of p-Smad3, and it appeared that p-Smad3 increased under the treatment of cisplatin or rapamycin (Fig 7K and L). These results suggested that Smad3-mediated transcription of LC3B was dependent on Linc00673-V3 upon autophagy induction.

Because we had established that Smad3 could regulate LC3B gene expression and thus affected autophagy, we further explored its role in NSCLC chemoresistance. The results showed that knockdown of Smad3 increased the sensitivity of NSCLC cells to cisplatin (Fig 7M–O). Taken together, these data suggested that Linc00673-V3–mediated Smad3 activation could directly up-regulate LC3B gene expression and participated in autophagy-associated chemoresistance of NSCLC.

Linc00673-V3 correlates with poor prognosis of NSCLC patients and Linc00673-V3–targeted antisense oligonucleotide suppresses autophagy-associated chemoresistance in vivo

Data from KM plotter demonstrated that higher Linc00673, Smad3, and MAP1LC3B (LC3B) expression levels were correlated with worse overall survival outcomes (Fig 8A–C), and prognosis data from GSE37745 also validated this finding (Fig S8A and B). To investigate the effects of Linc00673-V3 on NSCLC chemoresistance in vivo, tumor xenograft nude mouse model has been established. Moreover, specific antisense oligonucleotide (ASO) of Linc00673-V3 (ASO-Linc00673-V3) has been designed to explore the potential role of Linc00673-V3 as a therapeutic target in NSCLC. In vitro validation confirmed the efficiency of ASO as it decreased Linc00673-V3 and LC3B expression level (Fig 8D and E). LC3B lipidation was also decreased after treatment of ASO-Linc00673-V3, which was accompanied by accumulation of p62 (Fig 8F). As expected, treatment with ASO-Linc00673-V3 enhanced the sensitivity of A549 cells towards cisplatin in vitro (Fig 8G). In vivo validation demonstrated that both tumor volume and weight were significantly reduced in cisplatin-treated mice (Fig 8H and J). When combined cisplatin with ASO-Linc00673-V3, the tumor volume and weight further reduced compared with cisplatin treatment alone (Fig 8H and J), which indicated that Linc00673-V3 knockdown increased the sensitivity of tumor cell towards cisplatin in vivo. IHC staining revealed enhanced Linc00673-V3, Smad3, and LC3B expression in tumor tissues when treated with cisplatin, which is largely attenuated by ASO-Linc00673-V3 (Fig 8I). RT-qPCR of tumor tissues further confirmed the reduced expression of Linc00673-V3 and LC3B when treated with ASO-Linc00673-V3 (Fig 8K and L). Taken together, these results validated the role of Linc00673-V3 in chemoresistance in vivo via up-regulation of autophagy and suggested that ASO-Linc00673-V3 could be a promising antitumor therapeutic strategy especially in patients with chemoresistance of cisplatin.

Cell culture and reagents

A549, H1975, and cisplatin-resistant A549 (A549/CDDP) cell lines were cultured in RPMI-1640 medium (KEYI), supplemented with 10% FBS (SERRANA), 1% GlutaMAX (Gibco), and 1% sodium pyrurate (Gibco), final concentration of 100 U/ml penicillin (Gibco) and 100 U/ml streptomycin (Gibco). The A549/CDDP cells were purchased from the Cell Resource Center, Peking Union Medical College (PCRC), China. During culturing, the A549/CDDP cells were continuously exposed to 20 μM concentration of CDDP to maintain their resistance to cisplatin. H1299, H596, H1650, H520, H293T, and HeLa cell lines were cultured in DMEM medium (Wisent), supplemented with 10% FBS and final concentration of 100 U/ml penicillin (Gibco) and 100 U/ml streptomycin (Gibco). All cell lines were cultured at 37°C in an atmosphere containing 5% CO₂.

Reagents used in this study were as follows: autophagy inhibitors chloroquine (CQ) and 3-methyladenine (3-MA) (Selleck), autophagy activator rapamycin (Selleck), protein synthesis inhibitor cycloheximide, proteasome inhibitor MG132 (Selleck), and Smad3-specific inhibitor SIS3 (Selleck).

Cell transfection

The siRNAs of Linc00673 and Smad3, purchased from GenePharma, were transfected into cells using PowerFect siRNA Transfection Reagent (SignaGen Laboratories). The nucleotide sequences of si-Linc00673-#3, si-Linc00673-#5, si-Smad3, si-Nedd4L, si-STUB1, and si-VHL were listed in Table S3. The sequences of Linc00673-V1, Linc00673-V2, Linc00673-V3, Linc00673-V4, Linc00673-V5, Smad3, and STUB1 were cloned into pcDNA3.1 vectors. The stubRFP-sensGFP-LC3 plasmid was kindly given by Jianbing Zhang from Zhejiang Provincial People’s Hospital. In general, the stubRFP-sensGFP-LC3 plasmid codes the fusion protein components include the red fluorescent protein stubRFP, the GFP sensGFP, and the autophagy marker protein LC3. SensGFP is an acid-sensitive protein, whereas stubRFP is a stable fluorescent protein unaffected by pH changes. When the autophagosomes fuse with lysosomes to form autolysosomes, the sensGFP fluorescence is quenched and only red fluorescence punctate aggregates can be detected. The ratio of sensGFP to stubRFP bright spots was used to evaluate the progression of autophagic flux. All the plasmids were transfected with Lipofectamine 3000 (Thermo Fisher Scientific).

Western blot analysis

Cells were lysed to extract proteins with lysis buffer (NCM biotech) containing protease and phosphatase inhibitors cocktail (NCM biotech). The proteins were separated by 8–13.5% SDS–PAGE gels and transferred to nitrocellulose filter membranes, and these membranes were incubated with primary antibodies overnight at 4°C, followed by incubation with anti-mouse or anti-rabbit HRP-conjugated secondary antibodies (1:5,000; Promega). Finally, the protein–antibody complex was detected with the ECL substrate (Cyanagen) by ChemiScope 3300 Mini (Clinx). Primary antibodies were as follows (detail information listed in Table S4): GAPDH (1:5,000; Proteintech), LC3B (1:1,000; CST), p62 (1:1,000; CST), Smad3 (1:1,000; Abcam), p-Smad3 (1:1,000; CST), Ubiquitin (1:1,000; CST), HA (1:1,000; CST), ULK1 (1:1,000; CST), Beclin1 (1:1,000; CST), ATG3 (1:1,000; CST), ATG4 (1:1,000; CST), ATG5 (1:1,000; CST), ATG7 (1:1,000; CST), ATG12 (1:1,000; CST), Nedd4L (1:1,000; Abclonal), STUB1 (1:1,000; Abclonal), and VHL (1:1,000; Abclonal).

RNA extraction and RT-qPCR analysis

Total RNA was extracted from cultured cells using Trizol reagent (Sangon Biotech) and RNA was reverse transcribed into cDNA using PrimeScrip RT reagent Kit (Takara), according to the manufacturer’s recommendations. RT-qPCR reactions were performed by Takara’s SYBR Premix Ex Taq II (Tli RNaseH Plus) in Applied Biosystems 7500 Fast Real-Time PCR System, with gene-specific primers which were shown in Table S5. Transcripts were normalized to GAPDH mRNA.

Dual-luciferase reporter assay

H293T cells cultured in 24-well plates were transfected with experimental or control firefly reporter vectors and Renilla co-reporter vector. These cells were lysed with passive lysis buffer at 24 h after transfection, and the luciferase activity was detected according to the protocol of Dual-Luciferase Reporter Assay System (Promega). Constitutively activated Smad3-2SD plasmid was generated by using PCR-based mutagenesis to substitute Ser⁴²³ and Ser⁴²⁵ to aspartic acid in Smad3 and cloned into pcDNA3.1 vector.

Chemosensitivity and CCK-8 assay

Transfected or CDDP-resistant or parental lung cancer cells were incubated into 96-well plates and treated with different concentration of CDDP, with or without CQ or 3-MA. After 48 h, cell viability was assessed by the absorbance (450 nm) using Cell Counting Kit-8 (CCK-8; NCM biotech) according to the manufacturer’s protocol. The cell growth curve was drawn and the 50% inhibition of growth (IC₅₀) value of each cell line was calculated using GraphPad Prism7.0.

Subcellular RNA fractionation

To determine the cellular localization of Linc00673-V3, the cytoplasm and nuclear RNA fractions were isolated and purified using PARIS system (Thermo Fisher Scientific), according to the manufacturer’s protocol. Then the levels of Linc00673-V3, GAPDH, and U6 in the extracted RNAs were analyzed by RT-qPCR.

FISH

FISH was performed according to the manufacturer’s instructions (RNA FISH Kit; GenePharma). In brief, A549 and H1975 cells grown on the slides were washed with PBS twice and fixed in 4% paraformaldehyde. Then slides were incubated with prehybridization buffer at 37°C for 30 min. Biotin-labeled probes were premixed with streptavidin-labeled Cy3 to form the working solution. After prehybridization, slides were incubated with probe working solution at 37°C overnight. Images were captured using an Olympus FV1000 confocal microscope (Olympus) the next day.

RAP assay

Affinity pulldown of biotinylated RNA for detection of protein–RNA complexes was performed using Ribo RNAmax-T7 biotin-labeled transcription kit (RiboBio), according to the manufacturer’s protocol. In vitro transcription of the biotinylated RNA and two 10-cm dishes of ∼70% confluent cells were prepared for the pulldown reaction, incubated at RT for 30 min. Washed beads were added to the RNA–protein mixture and incubated for another 30 min. With the supernatant discarded, 1x Laemmli sample buffer supplemented with β-mercaptoethanol was added to the magnetic beads for Western blotting.

RIP and cross-linking RIP (CLIP) assay

RIP assay was performed using BersinBio RIP Kit, according to the manufacturer’s protocol. The cultured cells were lysed by RIP lysis buffer containing protease inhibitor and RNase inhibitor, and the lysate was incubated with specific antibodies. The mixture was incubated with magnetic beads, and these beads were washed three times to obtain the co-immunoprecipitated RNAs, which were extracted by Trizol reagent later. The RNAs were subjected to RT-qPCR analysis, with IgG used as control.

CLIP assay was performed using BersinBio CLIP-qPCR Kit (BersinBio), according to the manufacturer’s protocol. Briefly, cells were cultured in medium containing 100 μM 4-thiourdine overnight. The next day, ultraviolet irradiation was applied to cause covalent binding between RNA fragments and proteins. After harvesting the cells with NP-40 lysis buffer, RNase was applied to digest unbound RNA fragments and the targeted protein–RNA complex was captured using anti-Smad3 monoclonal antibody. The RNA fragments obtained by immunoprecipitation were subjected to reverse transcription and poly(A) tailing reaction at 3′OH ends of each RNA fragment. Then, RT-qPCR was conducted, primers were designed every 200 nt with 100 nt overlapping intervals to cover the full length of Linc00673-V3 as depicted in Fig 4G and detail sequences of each primer were listed in Table S2.

ChIP assay

ChIP was performed using ChIP Pierce Magnetic ChIP Kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. Briefly, cells (4 × 10⁶ cells) were fixed and cross-linked, and the cross-linked cells were lysed by buffer containing protease/phosphatase inhibitors and digested by MNase. Supernatant containing the digested chromatin was incubated with anti–p-Smad3 antibody overnight at 4°C. The magnetic beads were added to bind the immunoprecipitation complex. Beads were washed and purified to obtain the DNA for RT-qPCR analysis. The anti-RNA polymerase II antibody was used as a positive control and the IgG was used as a negative control.

Co-immunoprecipitation assay

The cultured cells were lysed by NP40 lysis buffer (Beyotime) containing protease inhibitor cocktail at 4°C for 45 min. Rabbit or mouse IgG and Protein A/G PLUS-Agarose (Santa Cruz Biotechnology) were added for pre-cleaning at 4°C for 30 min, and the cell lysate was incubated with primary antibody at 4°C overnight. The complex was added with Protein A/G PLUS-Agarose for co-immunoprecipitation, which was collected with sample buffer for Western blotting analysis.

Immunofluorescent staining

Transfected or control cells were incubated on sterilized coverslips in six-well plates, with or without autophagy activator rapamycin. After 24 h, cells were fixed using 4% paraformaldehyde for 5 min at RT, and fixed cells were permeabilized with 1% Triton X-100 for 10 min. After blocked with 5% BSA in PBST for 30 min, cells were incubated in primary antibody against LC3B (1:200; Proteintech) at 4°C overnight and in fluorochrome-labeled anti-rabbit secondary antibody (1:500; Beyotime) for 1.5 h at RT in the dark. Afterwards, the coverslips were stained with DAPI (1:10,000; Beyotime) and sealed with anti-fluorescence quenching sealing liquid (Beyotime). Images were taken from Olympus FV1000 confocal microscope (Olympus).

Animal experiment

BALB/c nude mice (6 wk old) were purchased from Shanghai SLAC Laboratory Animal Co., Ltd and sustained in the pathogen-free environment. The animal experiment protocol was approved by the Animal Care and Use Committee of the Laboratory Animal Center Zhejiang University.

A549 cells (2 × 10⁶) were resuspended in 0.9% saline with 0.1 ml Matrigel (Corning). Then, the cells were injected subcutaneously into the right flank of 6-wk-old female BALB/c nude mice. Mice bearing tumors of 150–200 mm³ were randomly assigned into four groups: vehicle control group (saline), ASO-Linc00673-V3 group (5 nmol intratumoral injection), cisplatin group (3 mg/kg), and cisplatin (3 mg/kg) together with 5 nmol ASO-Linc00673-V3 group. The treated mice were intraperitoneally injected twice per week for 4 wk. Mice were euthanized for tumor dissection, immunohistochemistry staining, and RT-qPCR analysis. Tumor weight was measured at the end point of the experiment. Data were analyzed using t test.

All tumor grafts were subjected to immunohistochemical staining according to the manufacturer’s instructions (SP Rabbit &Mouse HRP Kit; CWBIO). In brief, paraffin-embedded tissue slices were dewaxed and rehydrated. After that, antigens were retrieved by boiling in 0.01 M citrate buffer for 30 min. Tissues were incubated with 3% hydrogen peroxide followed with goat serum. Antibodies used in immunohistochemical staining were LC3B (1:300; Proteintech) and Smad3 (1:300; Proteintech). DAB was applied to immunodetection after incubating tissue slices with primary antibodies at 4°C overnight.

Bioinformatics and statistical analysis

The secondary structure of Linc00673-V3 was predicted by bioinformatic tool RNAfold (http://rna.tbi.univie.ac.at). The Smad3-binding sites in LC3B promotor were putative using the bioinformatic tool JASPAR (Fornes et al, 2020) (https://jaspar.genereg.net). The correlation between expression levels of Linc00673-V3, Smad3, and LC3B (also known as MAP1LC3B) and overall survival outcomes was obtained from KM plotter (https://kmplot.com/analysis).

Statistical analyses were performed using Graphad Prism7.0. All experiment data were expressed as mean ± SD of at least three independent experiments, and the P-value was calculated by two-tailed t test or analysis of variance. Unless specified, P < 0.05 was considered significant.