C-IGF1R encoded by cIGF1R acts as a molecular switch to restrict mitophagy of drug-tolerant persister tumour cells in non-small cell lung cancer

Abstract

The clinical efficacy of Epidermal Growth Factor Receptor Tyrosine Kinase Inhibitors (EGFR-TKIs) is limited by the emergence of drug resistance. We hypothesise that restoring dysregulated circular RNAs under initial treatment with EGFR-TKIs may enhance their effectiveness. Through high-throughput screening, we identify that combining circular RNA IGF1R (cIGF1R) with EGFR-TKIs significantly synergises to suppress tumour regrowth following drug withdrawal. Mechanistically, cIGF1R interacts with RNA helicase A (RHA) to depress insulin-like growth factor 1 receptor (IGF1R) mRNA splicing, negatively regulating the parent IGF1R signalling pathway. This regulation is similar to that of IGF1R inhibitor, which induces drug-tolerant persister (DTP) state with activated mitophagy. The cIGF1R also encodes a peptide C-IGF1R that reduces Parkin-mediated ubiquitination of voltage-dependent anion channel 1 (VDAC1) to restrict mitophagy, acting as a molecular switch that promotes the transition of DTP to apoptosis. Our study shows that combining cIGF1R with EGFR-TKIs efficiently reduces the emergence of DTP.

Introduction

The discovery of EGFR mutations as actionable oncogenic drivers has revolutionised cancer therapy for non-small-cell lung cancer (NSCLC) patients, especially for those in advanced stages, laying the foundation of precision cancer medicine [1, 2]. However, the emergence of resistance has limited the clinical success of EGFR tyrosine kinase inhibitors (EGFR-TKIs). The development of resistance to EGFR-TKIs is inevitable, which mediates the disease progression subsequently [2, 3]. In addition, the complexity of drug-resistant cells determines the ineffectiveness of current therapies [4, 5]. For further improved outcomes, EGFR-TKIs-based combination therapy has become a better option [6]. Clinical trials of biologically synergistic combinations building on the foundation treatment of EGFR-TKIs as first-line options to improve the outcomes are ongoing.

The NEJ009 study revealed that gefitinib (G) combined with chemotherapy improved progression-free survival (PFS) in untreated advanced patients with EGFR mutations with an acceptable toxicity profile compared with gefitinib alone [7]. The ARTEMIS study reported that the combination of erlotinib (Er) and bevacizumab, an antiangiogenic inhibitor, showed significant improvement in PFS for untreated patients with EGFR common mutation to those who received erlotinib alone [8]. However, the prolonged PFS and the acceptable side effects of the combination therapies have not translated to the overall survival (OS) benefit for advanced patients [9]. Another study also indicated that comparing Osimertinib (O) plus bevacizumab versus O alone, the combination arm failed to show prolongation of PFS in patients with advanced lung adenocarcinoma (LUAD) with EGFR-T790M mutation [10]. Additionally, the combination therapy of EGFR-TKIs and immune checkpoint inhibitors (ICIs) is controversial due to increased reporting of interstitial lung disease [6, 11–13]. Therefore, it is urgent to unravel the underlying vulnerabilities after the initial treatment of EGFR-TKIs to provide novel strategies for combination therapy.

Recent evidence has identified drug-tolerant persister (DTP) cells, which evade cell death under initial chemotherapy or EGFR-TKIs-based combination therapy, as a minor population of cancer cells that can constitute a reservoir from which drug-resistant tumours emerge [14–16]. However, targeting DTP cells also presents a therapeutic opportunity to impede tumour relapse [17]. Several mechanisms that contribute to the persistence of cancer cells have been reported in recent decades, they are mostly defined as diverse and reversible epigenetic changes without the selection for resistance mutations [14, 18]. Circular RNA (circRNA), as an endogenous RNA with a stable structure, can become a new generation of nucleic acid drugs [19, 20]. CircRNAs often act as epigenetic regulators, which have been reported to be associated with drug resistance [21–24]. However, whether circRNAs can be used as a potential enhancer of EGFR-TKIs to avoid DTP and improve drug efficacy remains unknown.

In this study, we identified that circRNA IGF1R (cIGF1R) significantly increased the efficacy of EGFR-TKIs by inhibiting the splicing of its parental gene insulin-like growth factor 1 receptor (IGF1R). The IGF1R signalling pathway, as a bypass of EGFR, mediates the resistance of EGFR-TKIs. However, combination therapy of EGFR-TKIs and IGF1R inhibitor (IGF1Ri) triggered tumour cells to step into a DTP state with activated selective mitophagy to evade cell death. In contrast, EGFR-TKIs plus cIGF1R significantly increased tumour cell apoptosis and inhibited regrowth. The cIGF1R could also translate peptide C-IGF1R, which acted as a molecular switch to inhibit Parkin-mediated ubiquitylation of VDAC1 to restrict activated mitophagy thereby promoting the transition of tumour cells from the sublethal DTP state with activated mitophagy to apoptosis.

Results

Restoring the dysregulation of cIGF1R expression induced by EGFR-TKIs treatment significantly improves the efficacy of EGFR-TKIs

The half-maximal inhibiting concentration (IC50) of several EGFR-TKIs (including Er, G, afatinib (Afa), and O) was measured in EGFR-mutant NSCLC cell lines. The results revealed that H1975 (EGFR L858R + T790M) was more sensitive to O, and PC9 (EGFR 19_Del) had a similar sensitivity to the EGFR-TKIs (Fig. S1a). To investigate the dysregulated circRNAs in living cells under the EGFR-TKIs therapy, H1975 and PC9 cells were treated with O and G, respectively, at the maximum effective concentration (Emax) for 3 days followed by circRNA-sequencing (circRNA-seq) (Fig. 1a). The results showed that at the onset of EGFR-TKI treatment, 141 and 470 circRNAs were up-regulated, and 308 and 564 circRNAs were down-regulated in H1975 and PC9 cells, respectively. The top 5 significantly up-regulated and down-regulated circRNAs in H1975 and PC9 cells were selected, which showed a consistent trend of their dysregulation in both cell lines (Fig. 1b, Fig. S1b). We subsequently confirmed the dysregulation of the 20 circRNAs in both cell lines using qRT-PCR analysis (Fig. S1c).

Fig. 1. cIGF1R increases the efficacy of EGFR-TKIs.

a H1975 cells were treated with DMSO or Osimertinib (O, 500 nM), and PC9 cells were treated with DMSO or Gefitinib (G, 100 nM) for 3 days followed by circRNA-sequencing (circRNA-seq) (n = 3). b Volcano plot displaying dysregulated circRNAs after EGFR-TKI treatment in H1975 and PC9 cells. Annotated probes include the 10 most significant circRNAs with decreased expression (blue) or increased expression (red). Gene expression analysis was conducted in n = 3 independent experiments. c Schematic diagram of barcode-labelled lentiviral plasmids. The OE pool contains 10 specific barcode-labelled overexpression vectors of the 10 most significantly down-regulated circRNAs, and the Sh pool contains 10 specific barcode-labelled knockdown vectors of the 10 most significantly up-regulated circRNAs. d Workflow for the construction of stably transfected OE pool or Sh pool in H1975 cells and their clones for variation screening in mouse subcutaneous xenografts. e qRT-PCR analysis of cIGF1R-barcode expression levels before inoculation (n = 3) and 30 days after inoculation (n = 10). f Real-time cell analysis (RTCA) experiments showing the proliferation of H1975 and PC9 cells under DMSO + empty vector (EV), EGFR-TKI, and EGFR-TKI + cIGF1R (O + cIGF1R; G + cIGF1R) treatment for 70 h (n = 3). g, h Proliferation (left) and weight (right) of mouse subcutaneous xenografts were measured under vehicle control, EGFR-TKI, and EGFR-TKI + cIGF1R (O + cIGF1R; G + cIGF1R) treatment for 35 days (n = 5). Data are shown as mean ± SD and were analysed using a two-tailed unpaired t-test. Source data are provided.

To investigate whether circRNAs could synergise with EGFR-TKIs, a lentiviral library embedding 20 individual barcodes was constructed corresponding to the candidate circRNAs, including the OE-pool overexpressing the 10 down-regulated candidates, and the Sh-pool knocking down the 10 up-regulated candidates (Fig. 1c). H1975 cells were infected with the OE pool and Sh pool, respectively, at an infection ratio of one barcode per cell, which was verified to detect individual barcode frequency through qRT-PCR (Fig. S1d). The transfected cells were subcutaneously transplanted into nude mice (10 mice per pool). After 30 days of continuous O treatment, the infected tumour cells (EGFP + ) were harvested from the subcutaneous tumours through fluorescence-activated cell sorting (FACS), and the individual barcode frequency was assessed (Fig. 1d). The results revealed that barcodes labelled cIGF1R (hsa_circ_0005035, circIGF1R (2)) were undetectable, suggesting that tumour cells overexpressing cIGF1R are barely able to sustain cell survival under O treatment (Fig. 1e, Fig. S1e). Importantly, cIGF1R abundantly expressed and was down-regulated in tumour tissues (Fig. S1f, g). In vitro and in vivo experiments confirmed that the treatment of EGFR-TKIs combined with cIGF1R significantly inhibited the growth of tumour cells (Fig. 1f–h). Moreover, cIGF1R alone was also found to inhibit tumour growth (Fig. S2a–d). To confirm whether cIGF1R specifically synergises with EGFR-TKIs, we compared the efficacy of cIGF1R plus EGFR-TKI with that of standard chemotherapy in advanced LUAD, which indicates that cIGF1R specifically synergises with EGFR-TKIs (Fig. S2e). Confirmation of the specific junction point sequence of cIGF1R, which originates from exon 2 of IGF1R, was accomplished via Sanger sequencing (Fig. S2f). The cIGF1R was exclusively amplified from cDNA, displaying augmented stability when compared to its linear counterparts, and distributed within both the cytoplasm and nucleus (Fig. S2g–j). These findings suggest that restoring the dysregulation of cIGF1R expression induced by EGFR-TKIs treatment significantly improves the efficacy of EGFR-TKIs.

Competitive utilisation of RHA by cIGF1R hinders IGF1R splicing and stimulates cIGF1R biosynthesis

The activation of IGF1R, the parental gene of cIGF1R, is a critical bypass signalling pathway that mediates the resistance of EGFR-TKIs [5]. To investigate the mechanism by which cIGF1R increases the efficacy of EGFR-TKIs, we explored whether cIGF1R affected the expression of its parental gene. The results indicated that cIGF1R inhibited the protein levels of IGF1R and phosphorylated-IGF1R (pIGF1R), as well as the downstream signalling pathway, but had no effect on the levels of EGFR and pEGFR (Fig. 2a, Fig. S3a). Furthermore, cIGF1R inhibited the reactivation of EGFR downstream under EGFR-TKI treatment (Fig. 2b). Overexpression of cIGF1R reduced IGF1R mRNA levels, while knockdown of cIGF1R increased IGF1R mRNA levels (Fig. 2c, Fig. S3b, c). However, cIGF1R did not influence IGF1R pre-mRNA levels and the stability of IGF1R mRNA (Fig. 2d, Fig. S3d, e). In addition, cells treated with splicing inhibitors Madrasin [25] or (and) Isoginkgetin [26] damped the regulatory effect of cIGF1R on IGF1R (Fig. 2e, Fig. S3f), indicating that cIGF1R regulates the expression IGF1R upon splicing.

Fig. 2. cIGF1R interferes with RHA to regulate IGF1R and cIGF1R expression.

a Immunoblotting (IB) analysis of IGF1R and phosphorylated-IGF1R protein levels, as well as EGFR/IGF1R downstream signalling protein levels, in H1975 and PC9 cells (n = 3). b IB analysis of EGFR/IGF1R downstream signalling treated as indicated in H1975 and PC9 cells at 0, 2, 4, 6, and 8 days, respectively (n = 3). c, d IGF1R mRNA expression levels (c) and IGF1R pre-mRNA expression levels (d) treated as indicated (n = 3). e IGF1R mRNA expression levels were detected by qRT-PCR treated with Madrasin (30 μM), Isoginkgetin (10 μM), or Madrasin + Isoginkgetin for 24 h in H1975 cells with stable transfection of EV or cIGF1R (n = 3). f Biotinylated-cIGF1R-probe pull-down products were electrophoresed on SDS-page and stained with silver. g RHA and HNRPL protein were detected by IB in biotinylated-probe pull-down products from H1975 cells (n = 3). h cIGF1R levels of reciprocal RNA immunoprecipitation (RIP) products using anti-RHA antibodies were detected by qRT-PCR in H1975 cells (n = 3). i Fluorescence in situ hybridisation (FISH) using Cy3-labelled cIGF1R probe and immunofluorescence (IF) using anti-RHA antibody were performed in H1975 cells (n = 3). j IGF1R mRNA, cIGF1R and IGF1R pre-mRNA expression levels were detected by qRT-PCR in H1975 cells treated as indicated (n = 3). k Pearson’s correlation between IGF1R and RHA expression in LUAD from the TCGA database. l IGF1R mRNA levels were detected by qRT-PCR after RHA knocking down for 24 h in H1975 cells with stable transfection of EV or cIGF1R (n = 3). m Left upper, schematic showing the nearest inverted-repeat Alu elements flanking exon 2 of the IGF1R pre-mRNA. Left lower, the 3’UTR regions of luciferase reporter plasmids #1 and #3 contain forward-aligned reverse-complementary Alu elements, while the 3′UTR regions of reporter plasmids #2 and #4 contain reverse-aligned reverse-complementary Alu elements, the control plasmid used the 3′UTR region that least affected the active expression of the reporter gene as a positive control. Right, relative luciferase activity of the reporter plasmids containing different 3’UTR regions of IGF1R in H1975 cells (n = 12). n Integrative Genomics Viewer analysis of CLIP-seq shows the RHA peaks at Alu elements coverage of IGF1R (n = 2). o Relative luciferase activity of the reporter plasmids after overexpression of RHA or RHA + cIGF1R in H1975 cells (n = 12). p Left, schematics of minigene, details are provided in the “Methods”. Right, PCR products were analysed using primers specific to the minigene spliced mRNA in 293 T cells treated as indicated (n = 3). q IGF1R pre-mRNA levels of RIP products using anti-RHA antibodies were detected by qRT-PCR in H1975 cells treated as indicated (n = 3). r Left, schematic showing the positive control (Ctrl), Alu Sx-wild-type (WT) and Alu Sx-mutant (MUT) vectors. The “Front” and “Back” sequences in the control vector were the potent sequences that promote circularisation. Right, the expression levels of barcode 1 and barcode 2 were detected by qRT-PCR in H1975 cells transfected with Ctrl, WT, or MUT vectors (n = 4). s IGF1R, cIGF1R and IGF1R pre-mRNA levels were detected by qRT-PCR in H1975 cells treated as indicated (n = 3). t mechanism diagram of cIGF1R interferes with RHA to regulate IGF1R and cIGF1R expression. Data are shown as mean ± SD and were analysed by a two-tailed unpaired t-test, n.s. no significance, *p < 0.05, **p < 0.01, ***p < 0.001, ****p< 0.0001. Source data are provided.

Growing evidence suggests that circRNAs can modulate splicing of their parental genes by interacting with splicing factors [27, 28]. To identify the potential splicing factors that interacted with cIGF1R, the circRNA pull-down experiments were performed using two specific biotin-labelled cIGF1R probes followed by mass spectrometry (MS) analysis. The splicing factor, RNA helicase A (RHA, also known as DHX9) and Heterogeneous Nuclear Ribonucleoprotein L (HNRPL) were identified by intersecting with the RNA pull-down products of the two cIGF1R probes, the splicing factors and the RNA-binding proteins (RBPs) (Fig. 2f, Fig. S4a, b). However, the binding of cIGF1R and RHA, but not HNRPL was subsequently confirmed in both H1975 and PC9 cells through independent RNA pull-down assays (Fig. 2g, Fig. S4c). In addition, RNA Immunoprecipitation (RIP), Fluorescence in situ hybridisation (FISH), immunofluorescence (IF) experiments and computer molecular simulation assay verified the interaction between cIGF1R and RHA (Fig. 2h, i, Fig. S4d, e), and the binding of cIGF1R to RHA did not influence the expression of RHA (Fig. S4f). RHA was found to promote the expression of IGF1R mRNA, but not the IGF1R pre-mRNA and the positive correlation between the expression of IGF1R and RHA was observed in the TCGA-LUAD dataset and our cohort (Fig. 2j, k, Fig. S4g–j). Notably, the regulatory effect of cIGF1R on IGF1R was significantly weakened following RHA knockdown (Fig. 2l, Fig. S4k, l), indicating that RHA facilitates the splicing of IGF1R mRNA, and cIGF1R modulates the expression of IGF1R mRNA in an RHA-dependent manner.

We then examined the mechanism through which cIGF1R regulates the splicing of IGF1R pre-mRNA via RHA. RHA promotes the generation of linear transcripts by unwinding the double-stranded RNA formed by inverted-repeat Alu elements in introns [29, 30]. Screening the genome sequence of IGF1R, we focused on the longest introns of IGF1R, intron 1 and intron 2, containing many inverted-repeat Alu elements and flanking Exon 2 (genome.ucsc.edu) [31]. To identify the specific pair of Alu elements responsible for the alternative splicing of IGF1R pre-mRNA, we employed a luciferase reporter system with two different paired inverted-repeat Alus (Alu Y or Alu Jo) in intron 1, as well as Alu Sx in intron 2, all of which were inverted-repeat Alus and located closer to the exons (Fig. 2m) [29]. The results demonstrated a significant decrease in luciferase activity for constructs containing 3′ UTR inverted-repeat Alu elements (#2 and #4) compared to those with 3′ UTR homodromous-repeat Alu elements (#1 and #3) or 3’UTR Alu-free (Ctrl), and a more pronounced downregulation of luciferase activity was observed for constructs with Alu Jo and Alu Sx (#4) compared to Alu Y and Alu Sx (#2) (Fig. 2m). The RHA cross-linking immunoprecipitation sequencing data published in Nature in 2017 also revealed that RHA peaks on pre-mRNA of IGF1R were situated around the region of Alu Jo and Alu Sx, but not Alu Y (Fig. 2n). The luciferase activity of inverted-repeat Alu Jo and Alu Sx (#4), but not of homodromous-repeat Alu Jo and Alu Sx (#3), was enhanced by RHA and reduced by cIRF1R, while knockdown of RHA significantly decreased the luciferase activity of #4, indicating that the complementary pairing of Alu Jo and Alu Sx is crucial for the regulation of IGF1R pre-mRNA by RHA (Fig. 2o, Fig. S5a, b). Moreover, the minigene assay demonstrated that RHA facilitated the linear splicing of minigenes, but co-expression of cIGF1R hindered RHA-mediated linear splicing of minigenes (Fig. 2p). Knockdown of RHA also resulted in a significant decrease in the linear splicing of minigenes (Fig. S5c). Importantly, RIP assay indicated that RHA could interact with IGF1R pre-mRNA, but not the mRNA. The cIGF1R inhibited this interaction, and vice versa (Fig. 2q, Fig. S5d, e). These findings suggest that cIGF1R binds to RHA, hindering the interaction between RHA and IGF1R pre-mRNA, which impairs the unwinding of paired Alus by RHA and inhibits linear splicing.

The back-splicing of circRNAs biogenesis and linear splicing compete against each other [28]. The Alu Jo and Alu Sx elements flanking exon 2 of the IGF1R pre-mRNA may facilitate the circularisation of cIGF1R. Our results have showed that RHA inhibited cIGF1R expression. To validate this, we developed barcode-tagged vectors to reflect the efficiency of Alu Jo and Alu Sx back-splicing. Barcode 2 was inserted reversely on both sides of exon 2 and could only be detected in circular products (Fig. 2r). RHA significantly suppressed the Alu Jo and Alu Sx-induced circularisation of exon 2, but this effect was abolished by cIGF1R. Knockdown of RHA slightly increased the circularisation efficiency of cIGF1R and eliminated the regulatory effect of cIGF1R (Fig. 2r, Fig. S5f). The ratio of IGF1R mRNA levels to IGF1R pre-mRNA levels could be increased by RHA but reversed by the addition of cIGF1R, and vice versa (Fig. 2s, Fig. S5g, h). Moreover, cIGF1R was also up-regulated in EGFR-mutant tissues (Fig. S5i). These findings suggest that cIGF1R binds RHA, which inhibits the splicing of IGF1R mRNA and promotes cIGF1R biosynthesis (Fig. 2t).

The combination of EGFR-TKIs and IGF1Ri treatment induces DTP, while EGFR-TKIs plus cIGF1R treatment promotes apoptosis

Linsitinib (L), also known as OSI-906, is a potent IGF1Ri that has been extensively utilised in both clinical and experimental settings [32]. In this study, we compared the efficacy of combining EGFR-TKIs with either cIGF1R or L. We found that the combination of O and L or O and cIGF1R significantly suppressed the proliferation of H1975 cells for the first three days (Fig. 3a). A similar depression was observed in PC9 cells treated with the combination of G and L or G combined with cIGF1R (Fig. 3a). Interestingly, after long-term treatment (8–20 days), viable cells were still detectable following treatment with EGFR-TKI plus IGF1Ri, whereas no viable cells were observed with EGFR-TKI plus cIGF1R treatment. Furthermore, cells treated with EGFR-TKI plus IGF1Ri regrew after drug washout (Fig. 3b, c, Fig. S6a, Supplementary Videos).

Fig. 3. Comparison of the efficacy of EGFR-TKIs plus cIGF1R and EGFR-TKIs plus IGF1R inhibitor (IGF1Ri).

a RTCA experiments were performed to evaluate the proliferation of H1975 cells treated with DMSO + EV, O + EV, O + cIGF1R, and O + Linsitinib (O + L), as well as PC9 cells treated with DMSO + EV, G + EV, G + cIGF1R (G + C), and G + Linsitinib (G + L), up to 70 h (n = 3). b Proliferation of H1975 and PC9 cells treated as described in a was assessed for 20 days followed by drug washout (n = 3). c H1975 cells were treated as described in a for 20 days followed by drug washout, and colony formation was evaluated. Three wells in each group (n = 33 wells/group) were randomly selected every 5 days and photographed after staining with crystal violet. d Left, H1975 cells were treated as described in a for 30 days in 3-D culture followed by drug washout, and proliferation was evaluated. Right, cellular organoids in the chamber were photographed every 10 days (n = 3). Scale bar, 200 μm. e, f Mice bearing H1975 cell xenograft tumours with EV or cIGF1R overexpression were treated with vehicle control, O, or O + L for 30 days followed by treatment cessation and follow-up (n = 5). g Tumour weight was measured for vehicle + EV group (at day 30), O + EV group (at day 30), O + cIGF1R group (at day 60) and O + L + EV group (at day 60) (n = 5). h Tumour weight was measured for vehicle + EV group (at day 30), O + EV group (at day 30), O + cIGF1R group (at day 60) and O + L + EV group (at day 60) (n = 5). Data are shown as mean ± SD and were analysed by a two-tailed unpaired t-test, **p < 0.01, ***p < 0.001. Source data are provided.

We observed that both EGFR-TKI plus IGF1Ri and EGFR-TKI plus cIGF1R similarly suppressed the EGFR signalling pathway during treatment, suggesting that the differences in regrowth after drug washout did not result from downstream reactivation (Fig. S6b). We confirmed these findings in 3-D culture conditions (Fig. 3d, Fig. S6c) and in vivo (Fig. 3e–g, Fig. S6d). These findings indicate that cells treated with EGFR-TKI plus IGF1Ri re-proliferate after drug withdrawal. DTP cancer cells are a subpopulation of cells that survive initial anti-cancer therapy, displaying a low proliferation rate and reversibility (relapse after drug withdrawal), similar to those observed with EGFR-TKI plus IGF1Ri treatment. Moreover, tumours treated with O combined with cIGF1R exhibited depressed Ki-67 expression and higher TUNEL positivity, suggesting greater apoptosis (Fig. 3h). These findings demonstrate that EGFR-TKI plus cIGF1R induces cell apoptosis, whereas EGFR-TKI plus IGF1Ri leads to tumour cells entering a DTP state but eventually reproliferating after drug withdrawal.

The DTP state induced by EGFR-TKIs and IGF1Ri combination therapy is dependent on autophagy

To gain insights into this phenomenon, we performed RNA sequencing (RNA-seq) on H1975 cells treated with O alone, O combined with cIGF1R, or O combined with L for 8 days (Fig. 4a). Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway enrichment analysis and Gene set enrichment analysis (GSEA) revealed that both O combined with cIGF1R and O combined with L treatments resulted in the arrest of growth and cell cycle gene programs, compared to the control (O alone). This was further confirmed by cell cycle experiments (Fig. 4b–d, Fig. S7a–c). However, we found that EGFR-TKI combined with cIGF1R treatment led to significantly increased cell apoptosis, as compared to EGFR-TKI combined with IGF1Ri treatment (Fig. 4e, f, Fig. S7d, e). Studies have shown that the DTP state induced by chemotherapy or targeted therapy is characterised by cell cycle arrest and senescence, which is consistent with the observed phenotypes in residual cells treated with a combination of EGFR-TKIs and IGF1Ri [15, 16, 18]. To further confirm this, senescence-associated beta-galactosidase (SA-β-Gal) activity assay was performed, which revealed that the majority of cells under EGFR-TKIs and IGF1Ri combination treatment stained positive for SA-β-Gal in both cell lines (Fig. 4g, Fig. S7f).

Fig. 4. Cells were apoptotic treated with O plus cIGF1R but entered a DTP state dependent on autophagy under the treatment of O plus IGF1Ri.

a RNA sequencing was performed on H1975 cells treated with O + EV, O + cIGF1R, or O + L + EV for 8 days (n = 3). b KEGG pathway enrichment analysis was performed on gene sets comparing OC-treated cells versus O + EV-treated cells and O + L + EV-treated cells versus O + EV-treated cells. c GSEA was performed on the cell cycle signature comparing OC-treated cells versus O + EV-treated cells or O + L + EV-treated cells versus O + EV-treated cells. d FACS analysis was used to determine the percentage of cells in G0&G1 phases in H1975 cells on day 0, day 2, day 4, day 6, and day 8 under O + EV, O + cIGF1R, or O + L + EV treatment (n = 3). e GSEA was performed on the apoptosis signature comparing O + cIGF1R-treated cells versus O + EV-treated cells. f FACS analysis was used to detect apoptosis levels of H1975 cells treated with O + EV, O + cIGF1R, or O + L + EV on day 0, day 2, day 4, day 6, and day 8 (n = 3). g Senescence-associated β-galactosidase (SA-β-Gal) staining was performed on H1975 cells treated as indicated for 8 days (n = 4). h Representative images of H1975 cells stably transfected with mCherry-EGFP-LC3B under the treatments indicated for 8 days (n = 3); scale bar, 20 μm. i Representative images of H1975 cells stably transfected with EGFP-LC3B under the treatments indicated, and bafilomycin (Baf) was added (200 nM) for 2 h to assess flux (n = 3); scale bar, 20 μm. j IB analysis of cell lysates was performed to determine the levels of LC3B and GAPDH in H1975 cells treated as indicated followed by 2 h of treatment with DMSO or Baf (200 nM) (n = 3). k FACS analysis was used to detect apoptosis levels of H1975 cells on day 0 and day 8 treated with DMSO, O, O + L, O + L + 3-Methyladenine (3MA, 5 μM), or O + L + Hydroxychloroquine (HCQ, 10 μM) for 24 h (n = 3). Data are shown as mean ± SD and were analysed by a two-tailed unpaired t-test, n.s. no significance, **p < 0.01, ***p < 0.001. Source data are provided.

Activated autophagy promotes the survival of tumour cells in a hostile environment [33], which was observed under EGFR-TKI and IGF1Ri combination treatment (Fig. S7g). Autophagic flux and LC3B conversion were utilised to monitor the autophagy activity, which revealed that EGFR-TKI combined with IGF1Ri treatment increased the autophagosome formation, lysosomal fusion, and autophagosome-associated LC3B-II, while the treatment of EGFR-TKI plus cIGF1R restricted the autophagy activity compared with the control (Fig. 4h–j, Fig. S7h–j). To explore whether autophagy is necessary for the survival of DTP cells induced by the treatment of EGFR-TKI combined with IGF1Ri, we utilised the autophagy inhibitor 3-Methyladenine (3-MA) or Hydroxychloroquine (HCQ) [34], which uncovered that autophagy inhibition significantly suppressed the survival of DTP tumour cells (Fig. 4k, Fig. S7k, l). These results suggest that EGFR-TKIs combined with IGF1Ri-treated cells exhibit DTP with activated autophagy, while EGFR-TKIs and cIGF1R combination treatment induce apoptosis and restrict autophagy.

C-IGF1R encoded by cIGF1R restricts activated mitophagy to serve as a molecular switch from DTP to apoptosis

To investigate whether cIGF1R-induced apoptosis relied solely on IGF1R, we treated IGF1R-knockout H1975 cells with O or O plus cIGF1R. The treatment of O plus cIGF1R significantly suppressed tumour growth and increased cell apoptosis in vitro and in vivo, while O alone did not have this effect (Fig. 5a–c, Fig. S8a–e). These findings suggest that cIGF1R does not solely induce apoptosis by inhibiting IGF1R signalling. Evidence suggests that circRNAs can function as protein templates [35, 36]. Notably, ribosome sequencing databases detected the cIGF1R sequence, indicating that endogenous cIGF1R can be translated (Fig. S8f). The cIGF1R sequence contains an open reading frame (ORF) that potentially generates a 66 amino acid (a.a.) protein (C-IGF1R) that is completely non-overlapping with the IGF1R protein (Fig. 5d). The activity of the internal ribosome entry site (IRES) of cIGF1R was confirmed by a luciferase reporter assay (Fig. S8g). We used sucrose density gradient centrifugation-based polysome analysis to further investigate whether cIGF1R could be translated. H1975 cells were transfected with cIGF1R or cIGF1R ATG-deletion (ΔATG) vectors, followed by qRT-PCR in non-ribosome (N), monosome (M), and polysome (P) components (Fig. 5e). The cIGF1R, but not cIGF1R-ΔATG, was detected in large quantities in the P component (Fig. 5f, Fig. S8h). Linear IGF1R was detected mainly in the P component and was not affected by cIGF1R transfection, serving as a positive control (Fig. 5f). After overexpression of cIGF1R, a significantly different band was observed at around 10 kDa, and the specific sequence of C-IGF1R was subsequently verified by MS (Fig. 5g, Fig. S8i, j). Several vectors were generated to further validate cIGF1R translation. In addition to the previously used cIGF1R vector (C1), we also designed a Flag-tagged cIGF1R vector in which a Flag-tag was inserted at the end of the predicted C-IGF1R ORF (Flag-cIGF1R, C2), a negative control vector lacking ATG (Flag-cIGF1R-ΔATG, C3), and a Flag-tagged ORF linear sequence was used as a positive control (Flag-C-IGF1R, C4) (Fig. 5h). These vectors were transfected into H1975 cells, and the level of Flag-C-IGF1R was detected by IB. As expected, Flag-C-IGF1R was detected in the presence of Flag-cIGF1R, but not with Flag-cIGF1R-ΔATG (Fig. 5h). Finally, a specific monoclonal antibody against the antigenic determinant sequence of C-IGF1R was designed to confirm the endogenous expression of C-IGF1R, which showed that C-IGF1R was detectable in human tissues (Fig. S8k–m).

Fig. 5. cIGF1R-encoded protein C-IGF1R inhibited the mitophagy of DTP cells under EGFR-TKI and IGF1Ri combination therapy stress.

a–c Proliferation of H1975-IGF1R-knockout cells treated as indicated in 2-D culture (a) (n = 3), 3-D culture (b) (n = 3) or in vivo (c) (n = 5) followed by drug withdrawal. d Predicted open reading frame (ORF) in cIGF1R and the amino acid sequence of C-IGF1R. e H1975 cell lysates were collected by centrifugation on a 5–50% sucrose gradient to collect non-ribosomes (N), monomers (M), and polymers (P) transfected with cIGF1R or cIGF1R-ΔATG plasmids. f The proportion of cIGF1R and IGF1R mRNA in each fraction was measured by qRT-PCR. g Identification of C-IGF1R in H1975 cell lysate after overexpression of EV or cIGF1R by IB. h Left, cIGF1R vector, Flag-tagged cIGF1R vector, Flag-tagged cIGF1R-ΔATG vector, and linearised Flag-tagged ORF vector. Right, IB detection of H1975 cell lysates overexpressing EV or different vectors above using anti-Flag antibodies. i, j Representative images of H1975 cells stably transfected with mCherry-EGFP-LC3B or EGFP-LC3B under the treatments indicated (n = 3); scale bar, 20 μm. k IB of H1975 cell lysates treated as in (i, j) (n = 3). l Apoptosis levels of H1975 cells on day 0 and day 8 treated as in (i, j) (n = 3). m Proliferation of H1975 cells treated as in (i, j) for 20 days followed by drug washout (n = 3). n Mitochondria and Flag-C-IGF1R were stained in H1975 cells overexpressing Flag-cIGF1R using MitoTracker and anti-Flag antibodies, respectively. Scale bar, 10 μm. o H1975 cells transfected with EV or C-IGF1R were harvested after 8 days of O + L treatment and analysed by electron microscopy for mitochondrial morphology, red arrows mark intact autophagic vesicles surrounding mitochondria. Scale bar, 1 μm. p H1975 cells treated as indicated followed by staining using MitoTracker and LysoTracker. White arrows indicate the fusion of mitochondria and lysosomes. Scale bar, 5 μm. q Apoptosis levels of H1975 cells on day 0 and day 8 were detected by FACS treated as indicated. r Left, representative images of H1975 cells treated as in (i, j) for 0, 2, 4, 6 and 8 days followed by JC-1 staining (n = 3). Right, the value of the ratio of the red fluorescence intensity to the green fluorescence intensity (n = 3). s FACS analysis of H1975 cells treated as in (q) followed by mitoSOX staining (n = 3). t Apoptosis levels of H1975 cells on day 0 and day 8 were detected by FACS treated as indicated. Data are shown as mean ± SD and were analysed by a two-tailed unpaired t-test, n.s. no significance, *p < 0.05, **p < 0.01, ***p < 0.001. Source data are provided.

We investigated whether cIGF1R was dependent on C-IGF1R to induce apoptosis and restrict autophagy under EGFR-TKIs treatment. C-IGF1R did not induce apoptosis in cells treated with O, although no activated autophagy occurred, indicating that the inhibition of autophagy by C-IGF1R was not lethal to tumour cells treated with EGFR-TKI alone (Fig. S9a–e). However, it was found that C-IGF1R, rather than cIGF1R-ΔATG, effectively suppressed autophagy, stimulated apoptosis, and suppressed tumour cell regrowth under the EGFR-TKI and IGF1Ri combination treatment (Fig. 5i–m, Fig. S9f–j). These findings suggest that the functions of C-IGF1R were displayed only in the EGFR-TKI combined with IGF1Ri-induced sublethal DTP cells. It was observed that c-IGF1R harbours a mitochondrial targeting signal (MTS) sequence and was localised within the mitochondria (Fig. 5n, Fig. S9k, l). Mitophagy, a selective form of autophagy that targets damaged mitochondria, blocks mitochondrial outer membrane permeabilization (MOMP) and apoptosis [37, 38]. We hypothesised that C-IGF1R inhibited the activated mitophagy in the DTP cells to promote apoptosis. Electron microscopy analysis revealed that the O plus L-treated DTP cells increased the number of mitochondria enclosed in a double membrane but was reduced by Flag-C-IGF1R (Fig. 5o). Flag-C-IGF1R also impeded the colocalization of mitochondria and lysosomes in the DTP cells (Fig. 5p). These results demonstrate that C-IGF1R inhibits selective mitophagy of EGFR-TKI combined with IGF1Ri-induced DTP cells. Moreover, Mdivi-1 (mitophagy inhibitor), or knockdown of the mitophagy initiator protein PINK1 significantly increased cell apoptosis under the treatment with EGFR-TKI plus IGF1Ri, indicating that mitophagy was crucial for the survival of DTP cells (Fig. 5q, Fig. S9m). Drug stress, such as chemotherapy and immunotherapy, led to mitochondrial damage thereby elevating mitochondrial reactive oxygen species (mtROS) [16, 39]. It was found that mitochondrial membrane depolarisation and mtROS levels were significantly increased after O and IGF1Ri combination treatment but gradually recovered, indicating mitochondrial damage and activated mitophagy (Fig. 5r, s). However, the presence of Flag-C-IGF1R induced a sustained increase in mitochondrial membrane depolarisation and mtROS levels, suggesting that C-IGF1R reduces the clearance of damaged mitochondria under the stress of EGFR-TKI and IGF1Ri combination therapy (Fig. 5r, s). In addition, the combination of mitophagy inducer deferiprone (DFP) and the PINK1 activator, Kinetin, reduced C-IGF1R-induced apoptosis under the treatment of EGFR-TKI combined with IGF1Ri (Fig. 5t, Fig. S9n). These findings collectively demonstrate that cIGF1R-encoded C-IGF1R functions as a molecular switch, facilitating the transition of tumour cells from a DTP state with activated mitophagy to apoptosis under sublethal combination treatment with EGFR-TKI and IGF1Ri.

C-IGF1R interacts with VDAC1 to reduce the Parkin-mediated ubiquitination of VDAC1

To dissect the mechanism by which C-IGF1R restricts mitophagy, co-immunoprecipitation (Co-IP) followed by MS to reveal that Flag-C-IGF1R could bind to the mitochondrial membrane (OMM) protein voltage-dependent anion channel 1 (VDAC1), which was subsequently confirmed by IB experiments in H1975 cells (Fig. 6a, b, Fig. S10a, b). To identify the region of VDAC1 that interacts with C-IGF1R, we designed four truncated plasmids of VDAC1, spanning from amino acids 1–90, 1–136, 1–194 and full-length, respectively (Fig. S10c). The protein truncation mapping assays revealed that Flag-C-IGF1R interacts with the region located at amino acids 91–136 of VDAC1 (Fig. 6c), which is consistent with the predictive interaction model generated by the ClusPro server (Fig. 6d). VDAC1 serves as an ion channel that regulates the outer mitochondrial membrane (OMM) potential, and its ubiquitination enables the recruitment of autophagic vesicles to the mitochondria [40]. The interaction between Flag-C-IGF1R and VDAC1 did not affect the activity of the mitochondrial permeability transition pore (MPTP) or the mitochondrial membrane depolarisation induced by CCCP, a well-known disruptor of the mitochondrial membrane potential and inducer of mitophagy [41] (Fig. 6e, Fig. S10d). However, Flag-C-IGF1R attenuated the polyubiquitylation of VDAC1 under CCCP treatment (Fig. 6f), which was subsequently confirmed by Co-IP experiments (Fig. 6g), indicating that Flag-C-IGF1R interacts with VDAC1 to depress the polyubiquitylation of VDAC1. To identify ubiquitinated lysine (K) residues in VDAC1 affected by Flag-C-IGF1R, we focused on K109 and K110, which are located in the domain interacting with Flag-C-IGF1R. The lysine residues were replaced by arginine (R) to respectively generate the VDAC1-K109R, VDAC1-K110R, and VDAC1-K109&K110R (KdualR) mutant proteins (Fig. S10e). The ubiquitination assays revealed that Flag-C-IGF1R significantly reduced the ubiquitination of VDAC1-wild-type and VDAC1-K109R, but not VDAC1-K110R and VDAC1-KdualR (Fig. 6h). Moreover, overexpression of wild-type or K109R VDAC1, but not K110R, was found to reverse C-IGF1R-induced apoptosis in DTP cells, implicating that K110 is an essential residue for VDAC1 ubiquitination regulated by Flag-C-IGF1R interaction (Fig. S10f).

Fig. 6. C-IGF1R inhibits Parkin-mediated ubiquitination of VDAC1.

a Identification of Co-Immunoprecipitation (Co-IP) products in H1975 cells overexpressing Flag-C-IGF1R using anti-Flag antibodies, as detected by immunoblotting (IB) and mass spectrometry (MS). b Immunoblotting (IB) of Co-IP products using anti-VDAC1, anti-TOMM20, and anti-GAPDH antibodies. c Co-IP of H1975 cells overexpressing Flag-C-IGF1R and Myc-tagged VDAC1 or truncated forms followed by IB. d Molecular docking analysis of the C-IGF1R and VDAC1 domains, with VDAC1 in yellow and C-IGF1R in cyan. e Fluorescence-activated cell sorting (FACS) analysis of H1975 cells treated as indicated and stained with Calcein AM, monitoring mitochondrial permeability transition pore (MPTP) (n = 3). f Time course experiment of carbonyl cyanide m-chlorophenylhydrazone (CCCP) treatment in H1975 cells stably transfected with empty vector (EV) or Flag-C-IGF1R, with total lysates detected by IB using specific anti-VDAC1, anti-HSP60, anti-TOMM20, and anti-GAPDH antibodies. All proteins tested were endogenously expressed (n = 3). g Co-IP in 293 T cells transfected with EV or Flag-C-IGF1R after overexpression of Myc-VDAC1 and 6×His-ubiquitin (His-UB), and the Co-IP products or whole cell lysates were detected by IB (n = 3). h Co-IP and IB experiments examining cell lysates from 293 T cells transfected with Myc-tagged wild-type VDAC1 (WT), Myc-tagged K109-mutated (K109R), Myc-tagged K110-mutated (K110R), or Myc-tagged K109 and K110 (KDualR) co-mutated plasmids for 24 h followed by 4 h CCCP treatment using anti-UB, anti-Myc, anti-Flag, and anti-GAPDH antibodies. i Co-IP in H1975 cells treated as indicated using anti-Myc antibodies, with IB examining the Co-IP products or cell lysates using anti-Parkin, anti-Myc, anti-Flag, and anti-GAPDH antibodies. j Co-IP in 293 T cells treated as indicated using anti-Myc antibodies, with IB examining the Co-IP products or cell lysates using anti-HA, anti-Myc, anti-Flag, anti-HA, and anti-GAPDH antibodies. k Staining of mitochondria and HA-Parkin in H1975 cells treated as indicated using MitoTracker and anti-HA antibodies, respectively. Scale bar, 10 μm.

VDAC1 is a substrate for Parkin, an E3 ubiquitin ligase involved in mitophagy [38]. We hypothesised that Flag-C-IGF1R interacted with VDAC1 to dampen the affinity for Parkin, which depressed the polyubiquitylation of VDAC1 (Fig. S10g). Co-IP assays revealed that overexpression of Flag-C-IGF1R disrupted the interaction between VDAC1 and Parkin (Fig. 6i, j). In addition, Parkin retained its cytosolic distribution and failed to translocate to mitochondria after overexpression of Flag-C-IGF1R in H1975 cells upon CCCP treatment (Fig. 6k). Knockdown of Parkin significantly increased the apoptosis of DTP cells, suggesting that Parkin-medicated VDAC1 ubiquitination is essential for the survival of DTP cells induced by EGFR-TKI plus IGF1Ri (Fig. S10h). Taken together, these data indicates that C-IGF1R interacts with VDAC1 to disrupt the interaction between VDAC1 and Parkin, which depresses the polyubiquitylation of VDAC1 to restrict the mitophagy.

Combination of O and cIGF1R improves survival in vivo

We have uncovered the binary function of cIGF1R, which hijacks RHA to negatively regulate the parent IGF1R signalling pathway, mimicking the IGF1R inhibitor to induce the DTP state with activated mitophagy, and also encoded C-IGF1R, acting as a molecular switch to promote the transition from DTP to apoptosis, two faces of which contribute synergistically to the enhancers of EGFR-TKIs. To further confirm the significance of cIGF1R in enhancing the efficacy of EGFR-TKIs, we used two different mouse models: a subcutaneous EGFR-mutant cell line-derived xenograft (CDX) model and an orthotopic CDX model. We designed a TET-ON-cIGF1R overexpression vector to test the therapeutic effect of cIGF1R after tumour formation (Fig. 7a). H1975 cells stably transfected with TET-ON-cIGF1R were expressed only when mice were fed with doxycycline (DOX) (Fig. S10i). H1975 cells stably transfected with TET-ON-cIGF1R were transplanted into BALB/c-nude mice subcutaneously. After mice developed palpable tumours, they were randomly assigned into 5 groups and treated with regimens as detailed in Fig. 7b (Fig. 7b). The time for tumours to reach 0.5 cm³ was significantly delayed in the O plus DOX and O plus L combined with HCQ groups compared to vehicle controls, O, and O plus L groups (Fig. 7c, d).

Fig. 7. The efficacy of cIGF1R combined with EGFR-TKIs was validated in mice.

a Schematic of the DOX-ON cIGF1R vector. b–d Mice bearing xenograft tumours of H1975 cells stably transfected with DOX-ON-cIGF1R were treated with vehicle, O, O + DOX, O + L, or O + L + hydroxychloroquine (HCQ), followed by treatment cessation and follow-up (n = 5). e–g Mouse lungs were orthotopically transplanted with Luc-H1975 cells stably transfected with DOX-ON-cIGF1R, and then mice were given vehicle, O, O + DOX, O + L, or O + L + HCQ followed by treatment cessation and follow-up (e), bioluminescence imaging of nude mice every 15 days (f), Kaplan–Meier survival analysis of nude mice (g). h Left, Hematoxylin-eosin (H&E, 1× and 40×) and immunohistochemistry (Ki-67 and TUNEL stain, 40×) analysis of mouse lungs from (e–g) on day 20 (Vehicle), day 45 (O), day 75 (O + L), day 100 (O + DOX, and O + L + HCQ). Scale bar, upper, 2 mm; lower, 50 μm. Right, percent Ki-67 positive cells and percent area necrosis are plotted. i Graphical summary of the key findings of the study. For d and g data were analysed by Log-rank (Mantel-Cox) test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Source data are provided.

We then implanted luciferase-labelled H1975 (Luc-H1975) cells stably transfected with TET-ON-cIGF1R into the lungs of nude mice in situ and confirmed the inoculation by bioluminescence imaging after surgery. These mice were randomly assigned into 5 groups and treated as indicated, and the tumour size was recorded using bioluminescence imaging every 15 days (Fig. 7e, f). The results revealed that mice treated with O combined with DOX had the longest OS, as well as those treated with O plus L combined with HCQ (Fig. 7g). Hematoxylin-eosin (H&E) staining of the mouse lungs showed minimal tumour volume and less Ki-67 expression, more TUNEL positivity in per cent of tumours under the O plus DOX, and O plus Lin combined with HCQ treatment, indicating decreased proliferation and increased apoptosis (Fig. 7h). However, there were no significant differences in Ki-67 expression or TUNEL positivity between vehicle controls (at day 20), O (at day 45), and O plus L regrowth tumours (at day 100) (Fig. 7h). These findings confirmed that combining cIGF1R could enhance the efficacy of EGFR-TKIs in vivo.

Discussion

In this study, we first detected the deregulated circRNAs after EGFR-TKIs therapy by circRNA-seq. By employing high-throughput deregulated circRNA screens in vivo, we identified that cIGF1R significantly synergised with EGFR-TKIs. We subsequently demonstrated that cIGF1R hijacked the splicing factor RHA to negatively regulate the expression of IGF1R, which inhibited the reactivation of EGFR downstream under EGFR-TKIs treatment. Moreover, we showed that EGFR-TKIs plus cIGF1R significantly induced apoptosis while the combination of EGFR-TKIs and IGF1Ri led to DTP. We identified that the DTP cells were characterised by selective mitophagy activation and were highly dependent on mitophagy. In addition, we found that cIGF1R could also encode the peptide C-IGF1R, which restricted Parkin-mediated polyubiquitination of VDAC1, thus inhibiting the activated mitophagy to induce apoptosis. Finally, we confirmed that EGFR-TKIs combined with cIGF1R could significantly improve the survival of mice.

CircRNAs are unusually stable RNAs produced by the circularisation of exons and have been described as generally down-regulated transcripts in cancer [20, 42–44]. It has been argued that circRNA expression levels cannot generally be explained by simple correlation with the expression of their linear isoforms [27]. In contrast, circRNAs are typically generated at the expense of canonical mRNA isoforms [28]. Here, we identified that the expression of cIGF1R was a negative association with IGF1R (linear transcript of its parental gene) after EGFR-TKIs treatment. The splicing factor RHA medicated the competitive splicing of IGF1R and circulation of cIGF1R. Interestingly, cIGF1R could hijack RHA to reduce the splicing of IGF1R and concurrently promote cIGF1R biosynthesis as positive feedback. A previous study also indicated that total circRNA abundance showed a consistent negative correlation of a panel of cell cycle progression genes in prostate carcinoma [43], indicating a potentially widespread layer of previously unknown gene regulation. Our results suggested that circRNA biogenesis per se might be an important regulator of mRNA production.

Our results showed that the efficacy of IGF1R inhibition using IGF1Ri or cIGF1R combined with EGFR-TKI was completely different, IGF1Ri caused DTP, while cIGF1R led to apoptosis. A recent study also indicated that inhibition of EGFR and its downstream ERK1/2 reactivation makes tumour cells dormant (also known as DTP). We confirmed that the DTP state induced by EGFR-TKI combined with IGF1Ri using 2-D, 3-D cultures and animal models. Studies revealed that DTP cells are seen following the use of anti-cancer agents and exhibit a reversible and slow-cycling status that is not driven by any genetic variations [14, 16, 45]. We identified that the DTP cells were characterised by similar senescence, cycle arrest, and relapse after drug withdrawal, consistent with the results of previous studies. Our study also revealed the activation of mitophagy in DTP cells and its dependence on autophagy, especially mitophagy for survival. Combination therapy increased mitochondrial damage and mtROS, and cells maintain survival by clearing damaged mitochondria through mitophagy. Combined with autophagy inhibitors or blocking mitophagy in cells can significantly increase the apoptosis of DTP cells. Previous studies have shown that chemotherapy-induced DTPs in colorectal cancer cells were dependent on autophagy for survival, and fusobacterium nucleatum could promote chemoresistance to colorectal cancer by modulating autophagy [16, 46]. However, the therapeutic targeting of autophagy in cancer is sometimes viewed as controversial [34]. Recent studies indicate that a combination of ERK and autophagy inhibition could be an effective treatment approach for KRAS-mutant pancreatic cancer [47, 48]. This suggests the prospect of combining targeted therapy with autophagy inhibition in cancer.

Interestingly, the combination therapy of EGFR-TKIs and cIGF1R-induced apoptosis restricting mitophagy. We also confirmed the different effects between O + cIGF1R and O in IGF1R-knockout cells, suggesting that cIGF1R-induced apoptosis was not merely dependent on the regulation of its parent gene IGF1R. The cIGF1R also translates the peptide C-IGF1R, which interacts with VDAC1 to disrupt the Parkin-mediated polyubiquitylation of VDAC1 to restrict mitophagy, leading to EGFR-TKI combined with IGF1Ri-treated sublethal DTP cells apoptosis (Fig. 7i). Accumulating evidence suggests that circRNAs can be translated into peptides to exert similar or completely different functions from their linear counterparts [35]. According to a recent study, circRNAs only encoded peptides that were much shorter than their linear cognates, with a peak length distribution of less than 100 a.a. [36]. In our study, we identified that the C-IGF1R, 66 a.a. in length, has a completely different sequence from that of IGF1R, suggesting that C-IGF1R and IGF1R have distinct functions. We observed that C-IGF1R binds to OMM protein VDCA1, which is crucial for selective mitophagy through polyubiquitin modification. We also found that O combined with C-IGF1R did not increase apoptosis due to the low activity of mitophagy under O single-drug therapy, which indicated the specificity of C-IGF1R. Additionally, the combination of O + L and cIGF1R-ΔATG could not induce apoptosis. On the contrary, O + L and cIGF1R-ΔATG combination treatment led to DTP with activated autophagy, which indicated that cIGF1R increases cell apoptosis under EGFR-TKIs treatment through binary mechanisms.

In conclusion, our findings suggest that the combination of cIGF1R can significantly improve the efficacy of EGFR-TKIs in vitro and in vivo. These data support the exploration of EGFR-TKIs in combination with cIGF1R as a strategy to improve initial response rates and expand benefits in NSCLC patients with EGFR mutations.

Methods

Tissue samples

The clinical characteristics of the patients are detailed in Supplementary Table 1.

Cell lines

The HEK-293T cell line and human non-small cell lung cancer (NSCLC) cell lines (H1975, PC9) were procured from the China Centre for Type Culture Collection. The NSCLC cell lines (H1975, PC9) were cultured in 1640 with 10% foetal bovine serum (FBS, Corning), while the HEK-293T cells were cultured in DMEM with 10% FBS (Corning). Prior to the experiment, the cells were screened for mycoplasma contamination, cross-contamination between species, and authenticity. The cell lines used in the experiments were cultured for a maximum of 20 passages.

Animal models

For animal research, four-week-old female BALB/c nude mice were purchased from GemPharmatech. All animals were housed in a pathogen-free environment at Nanjing Medical University. All animals were subjected to a 12 h light-dark cycle. The room temperature was kept at 22 °C, and the humidity level was kept between 55%–70%.

For the subcutaneous xenograft mouse model in Fig. 1, 20 mice were randomly assigned to two groups (n = 10 mice per group). H1975 cells (5 × 10⁶, stably transfected with OE pool or Sh pool) with 50% Matrigel (Corning) were inoculated subcutaneously to the mice. The mice were treated with O (10 mg/kg once daily) for 30 days and the subcutaneous tumours of the mice were taken out to prepare a single-cell suspension, the EGFP-positive tumour cells were sorted by flow cytometry. The DNA of tumour cells was extracted and detected by qRT-PCR. For the subcutaneous xenograft mouse model in Fig. 1, Fig. 3, Fig. 5 and Fig. S2, mice (n = 5 mice per group) were subcutaneously inoculated with H1975 cells (5 × 10⁶, stably transfected with cIGF1R or EV) in 50% Matrigel (Corning) and treated as described in the figure legend. Tumour volumes were monitored every 5 days. The nude mice were killed 35 or 60 days after inoculation, and tumours were removed, weighed, and photographed. A portion of the tumour tissue was fixed in 4% paraformaldehyde for HE and IHC, while the remainder was frozen in liquid nitrogen for further analysis.

For the subcutaneous xenograft model in Fig. 7, each nude mouse was injected subcutaneously with 5 × 10⁶ H1975 cells (stably transfected with TET-ON-cIGF1R) in 50% Matrigel (Corning). For efficacy studies, when the xenografts reached approximately 100 mm³, the mice were randomised into 5 groups (n = 5 mice per group): (1) Vehicle control, (2) O (10 mg/kg once daily), (3) O (10 mg/kg once daily) + Dox (100 mg/kg once daily), (4). O (10 mg/kg once daily) + L (10 mg/kg once daily), and (5) O (10 mg/kg once daily) + HCQ (100 mg/kg once daily). All the mice were treated as described in the figure legend and sacrificed after the treatment, then the tumours were removed for further study.

For the orthotopic xenograft mouse model in Fig. 7, Luc-H1975 cells (stably transfected with TET-ON-cIGF1R) were Intrapulmonary injected into the mice. After injection for 5 days, lung orthotopic xenografts were assessed by bioluminescence imaging (PerkinElmer). Mice were randomly divided into 5 groups (n = 5 mice per group) according to the intensity of the region of interest (ROI): (1) Vehicle control, (2) O (10 mg/kg once daily), (3) O (10 mg/kg once daily) + Dox (100 mg/kg once daily), (4). O (10 mg/kg once daily) + L (10 mg/kg once daily), and (5) O (10 mg/kg once daily) + HCQ (100 mg/kg once daily). 40 days later, all treatment was withdrawn. Lung tumours were recorded every 15 days by bioluminescence imaging. After 100 days, the mice were sacrificed, and their lungs were extracted, weighed, and photographed. For HE and IHC investigations, the lung tissue was fixed in 4% paraformaldehyde.

Compounds

All compounds used in this study were listed in Supplementary Table 2.

Lentivirus, plasmids and siRNAs

Barcode-tagged lentivirus vectors expressing cATG2B, cCARD6, cKPNB1, cCENPI, cCTBP1, cIGF1R, cDLG1, cNAV3, cMTHFD1L, cSLC15A4 and shRNA targeting cBRWD1–1, cHMGCR, cSPG11, cTTBK2, cPARP12, cMANBA, cBRWD1-2, cMKLN1, cTBC1D4, cTRPM7 were co-transfected into HEK-293T cells using Lipofectamine 3000 (Thermo) for lentivirus packaging. 48 h after transfection, supernatants containing lentivirus particles were collected. H1975 cells were infected with lentiviruses (OE pool or Sh pool) in the presence of 2 μg/ml polybrene (MedChemExpress). Lentivirus vector expressing cIGF1R was created by cloning the IGF1R exon 2 sequence using pCDH-CMV-MCS-EF1-Puro, and an additional splicing sequence (Front: AAAGTGCTGAGATTACAGGCGTGAGCCACCACCCCCGGCCCACTTTTTGTAAAGGTACGTACTAATGACTTTTTTTTTATACTTCAG, Back: GTAAGAAGCAAGGAAAAGAATTAGGCTCGGCACGGTAGCTCACACCTGTAATCCCAGCA) was added. The TET-ON-cIGF1R expression plasmid was created by cloning the IGF1R exon 2 sequence using pHBLV-TetOn-SV40-Puro-TRE3GS-MCS, and an additional splicing sequence (same to the above) was added. Lentivirus vectors expressing mCherry-EGFP-LC3B or EGFP-LC3B were purchased from Hanbio. Subsequently, H1975 cells stably overexpressing TET-ON-cIGF1R, mCherry-EGFP-LC3B, or EGFP-LC3B were constructed using the same method as above. Plasmids overexpressing RHA, cIGF1R, cIGF1R-ΔATG, Flag-cIGF1R, Flag-cIGF1R-ΔATG, Flag-C-IGF1R, Myc-VDAC1-1, Myc-VDAC1-2, Myc-VDAC1-3, Myc-VDAC1-4, Myc-VDAC1-K109R, Myc-VDAC1-K110R, Myc-VDAC1-109&110R, and HA-Parkin were purchased from Bioworld. All plasmids were verified by DNA sequencing. Lipofectamine 3000 was used to transfect plasmids according to the manufacturer’s instructions. RiboBio provided the siRNAs targeting cIGF1R, RHA, and PINK1. Lipofectamine RNAiMAX (Thermo) was used for siRNA transfections.

IC50 analysis

Single-cell suspensions were seeded into 96-well plates at a density of 5000 cells per well. Viable cells were fixed with 4% formaldehyde and stained with 1% crystal violet after 3 days of drug incubation.

Cell growth and viability assays

Cell growth was monitored using the real-time cell analysis (RTCA) system (ACEA Biosciences) and cell proliferation plates. RPMI-1640 or DMEM containing 10% FBS was added to the chamber, followed by plating 10,000 cells onto each well of the E-plate. Readings were taken every hour until the experiment ended (up to 70 h). Viability of NSCLC cell lines was also assessed using a 3-D cell culture. In 96-well plates, 500 cells were plated with 10 μl Matrigel (Corning) droplets per well. After 24 h, the organoids were treated with a drug-containing solution (drug renewed every 2 days) for 10, 20, 30, 40, 50, or 60 days, and the cell ability was measured using a Glomax Luminometer (Promega). For the colony formation assay, 200 cells/well were seeded into 12-well plates and treated as shown in the figures. The colonies were gently rinsed with PBS, fixed in 4% paraformaldehyde for 20 min at room temperature, and then stained with 0.1% crystal violet. Colonies with more than 50 cells were counted under the microscope.

Cell cycle analysis

After the designated treatments, the cells were harvested using Trypsin (Gibco), washed with PBS, and fixed overnight at −20 °C in 75% ice-cold ethanol. Subsequently, the fixed cells were washed with PBS and resuspended in RNase PBS with 0.1% Triton X-100. The cells were then stained with propidium iodide (PI) for 30 min at 37 °C. The DNA content of the cells was measured using a flow cytometer (BD Biosciences).

PI/Annexin V apoptosis assay

After the specified treatments, both floating and adherent cells were trypsinized and washed with PBS. Apoptotic cells were detected using the Annexin V-FITC Apoptosis Detection Kit I (BD Biosciences), which stains cells with Annexin V-FITC and PI. The manufacturer’s instructions were followed for staining. A flow cytometer (BD Biosciences) was used to analyse the cells for apoptosis.

Cell senescence β-galactosidase staining

After the indicated treatments, H1975 and PC9 cells were subjected to β-galactosidase staining to detect cell senescence using the β-galactosidase staining kit (Beyotime). The staining procedure was performed according to the manufacturer’s instructions.

TUNEL

Apoptotic cell death in mouse xenografts and patient tissue was determined using a TUNEL-Detection Kit (Beyotime) according to the manufacturer’s instructions. The slides were then examined under a fluorescence microscope (Carl Zeiss).

Dual-luciferase reporter assay

The dual-luciferase reporter plasmid was transfected into H1975 cells using Lipofectamine 3000 (Thermo) for the RHA luciferase reporter experiment. Luciferase activity in cell lysates was tested 48 h after transfection using a Dual-Luciferase Reporter Assay System (Promega). The activity of firefly luciferase was adjusted to that of Renilla luciferase. The experiment was repeated a total of twelve times.

Autophagic flux assay

H1975 and PC9 cells stably transfected with pmCherry-C1-EGFP-LC3B or pEGFP-C3-MAP1LC3B were plated in confocal dishes and treated as described in the figure. Fluorescence confocal microscopy (Carl Zeiss) was used to investigate LC3 puncta.

Mitochondrial status analysis

For mtROS and MPTP analysis, H1975 cells were incubated for 10 min at 37 °C in a PBS medium with 5 μM MitoSOX Red dye to examine mtROS species formation. Cells were rinsed three times with PBS before being stained with DAPI for 5 min. After H1975 cells were treated according to the experimental design, an adequate volume of Calcein AM staining solution was added and incubated in the dark at 37 °C for 30 min. It was replaced with a fresh pre-warmed medium at 37 °C and incubated in the dark for 30 min after the incubation. Then, MitoSOX and MPTP were examined using a flow cytometer (BD Biosciences). For OMM polarisation analysis, H1975 cells were incubated for 20 min at 37 °C in serum-free 1640 media containing 2.5 μg/ml JC-1 dye. Cells were rinsed three times with full media before being viewed under a fluorescence microscope (Carl Zeiss). The ratio of red-JC-1 staining (polarised mitochondria) / green-JC-1 staining (depolarised mitochondria) was calculated using Image J software.

RNA extraction and qRT-PCR

Patient tissue and cell RNA was extracted using the TRIzol reagent (Thermo) followed by reverse transcription using the RNA-to-cDNA kit (Takara). To quantitatively express the contents of RNA, qRT-PCR analysis was performed using SYBR Green Premix (Vazyme) in the QuantStudio 6 Flex system (ABI) or ViiA 7 Dx system (ABI). ACTB, GAPDH, or U6 served as the internal control for circRNA and mRNA, and qRT-PCR primer sequences are listed in Supplementary Table 3.

Minigene assay

A minigene splicing reporter system was designed to evaluate the effects of Alu elements, RHA, and cIGF1R on IGF1R linear splicing. The Exon 2 and flanking intron sequence of IGF1R (99244001-99255000 hg19) together with exon 1 and exon 3 of IGF1R (11423 bp) were synthesised using GenBrick technology (GenScript, China) and subsequently inserted into the pcDNA3.1 + C-eGFP vector. The forward primer was designed in exon 2 of IGF1R (TGAGTACAACTACCGCTGCT), and the reverse primer was located in EGFP (GAACTTCAGGGTCAGCTTGC), which specifically amplified the minigene-derived spliced mRNA. The minigene plasmids together with RHA overexpression plasmids or RHA siRNAs were transfected into 293 T cells. RNA was extracted 24 h later, and the expression of minigene-derived IGF1R linear splicing products was analysed by qRT-PCR.

Polysome profiling analysis

H1975 cells were grown in 15 cm dishes and transfected with cIGF1R or cIGF1R-ATG plasmids. After 48 h, the cells were rinsed with ice-cold PBS and treated for 5 min with 100 ng/ml Cycloheximide (MedChemExpress). Cells were then lysed on ice for 15 min in 500 μl polysome lysis solution, followed by centrifugation at 4 °C for 7 min at 16000 g to pellet nuclei and mitochondria. The supernatant was then put onto a sucrose density gradient ranging from 5 to 50% and ultracentrifuged at 20,000 g for 2 h at 4 °C in a Beckman SW41 Ti rotor. An absorbance detector (Thermo) linked to the fraction collector was used to measure absorbance at 254 nm. TRIzol solution was used to extract RNA from fractions.

CRISPR-cas9

Using the CRISPR Design Tool (crispr.mit.edu), single guide RNAs (sgRNAs) targeting IGF1R exons were created. The IGF1R-targeting sgRNAs (sgRNA-IGF1R-1: CCACGACGGCGAGTGCATGC, sgRNA-IGF1R-2: TCAGTACGCCGTTTACGTCA) were cloned into eSpCas9-2A-Puro (PX459) V2.0 plasmid (Purchased from GenScript). The IGF1R-targeting sgRNA plasmids were transiently transfected into H1975 cells for 36 h. Cells were exposed to puromycin (2 μg/ml) selection for one week after transfection (for 72 h). Puromycin-resistant cells were resuspended and planted into 96-well plates as single cells. Finally, IGF1R-knockout clones were tested by IB 4 weeks later.

CircRNA-seq

H1975 or PC9 cells were treated with O (500 nM) or G (100 nM) for 3 days, after which apoptotic cells were washed away with ice-cold PBS, and the total RNA of the remaining cells together with the cells before treatment was extracted using TRIzol reagent. Subsequently, the concentration and integrity of the extracted total RNA were determined using the Qubit 3.0 Fluorometer (Invitrogen) and the Agilent 2100 Bioanalyzer (ABI). Total RNA was treated with 10 units of RNase R (Genesee) for 30 min at 37 °C. The quality and concentration of the library were determined using a DNA 1000 chip on an Agilent 2100 Bioanalyzer. The KAPA Biosystems Library Quantification kit (Kapa Biosystems) was used to assess the accuracy of quantification for sequencing applications. Before clustering, each library was diluted to a final concentration of 10 nM and pooled equimolarly. All samples were subjected to Paired-End (Nova-seq 6000) sequencing. Reads were first mapped to the most recent using Bowtie2 version 2.1.0 [49]. The UCSC transcript collection STAR [50] and DCC [51] STAR were used to map the data to the mapped genome for circRNA expression investigation. The edgeR tool was used to identify differentially expressed genes. The figure was created using R. Differentially expressed circRNAs are listed in Supplementary Table 4.

RNA-seq

H1975 cells were treated with O (n = 3), O plus cIGF1R (n = 3), or O + L + EV (n = 3) for 8 days followed by extracting total RNA. RNA sequencing was performed to analyse RNA samples according to the manufacturer’s guidelines. In brief, raw data were generated by sequencing RNA samples on the Illumina HiSeq 4000 platform (Illumina). The R software tool was used to identify genes with substantial differential expression between O and O plus cIGF1R or O and O + L + EV groups based on fold changes of 2.0 and p < 0.05. DESeq2 was used for differential expression analysis, and differentially expressed genes (FDR < 0.05) were submitted to KEGG analysis and GSEA. The differentially expressed genes are listed in Supplementary Table 5.

RNase R treatment

Total RNA was isolated from H1975 and PC9 cells and separated into two groups. After pre-treatment with RNase R (Genesee) at 37 °C for 30 min according to the manufacturer’s recommendations, qRT-PCR was performed to determine the expression of cIGF1R, IGF1R, and GAPDH with or without RNase R. GelRed was used to visualise the products after they were separated on a 1.5% agarose gel.

Nuclear and cytoplasmic extraction assay

According to the manufacturer’s procedure, a nucleocytoplasmic separation kit (Thermo) was used to extract nucleocytoplasmic protein or RNA. The nucleus and cytoplasmic RNAs were used to detect the expression of target genes by qRT-PCR. GAPDH as a cell plasma localisation reference and U6 as a nuclear localisation reference was applied to normalise the target gene expression and to calculate the ratio of the nucleus to the cytoplasm.

RNA fluorescence in situ hybridisation (FISH)

A cy3-labelled oligonucleotide probe complementary to the cIGF1R junction region was designed using the Clone Manager Kit analysis tool. H1975 and PC9 cells were seeded in covered glass-bottom confocal dishes and cultured overnight. In situ staining was performed using the RNA FISH kit (GenePharma) following the manufacturer’s instructions. Nuclei were stained with DAPI. Images were acquired on an LSM 880 laser confocal microscope (Carl Zeiss). The sequence of the cIGF1R probe is ATGCCTGGCCCGCAGATTTTCTGG.

Circ-RNA pull-down

RNA pull-down experiments were carried out according to the manufacturer’s instructions using a Pierce Magnetic RNA-Protein Pull-Down kit (Thermo). A biotinylated cIGF1R probe (0.2 nmol, Riobio) was pre-treated with magnetic beads before being incubated with H1975 cell lysates overnight at 4 °C. The biotinylated cIGF1R probe sequence is CGCAGAUUUUCUGGCAGCGGUUUGU (Bio-cIGF1R-probe 1), and AUGCCUGGCCCGCAGAUUUUCUGG (Bio-cIGF1R-probe 2), and the sequences of the negative controls for the biotinylated cIGF1R probes were their antisense sequences, respectively. Detection of cIGF1R-interacting RNAs was performed by RT-qPCR as described in our previous study [52]. IB and MS were used to examine the isolated proteins, and the proteins detected by MS are listed in Supplementary Table 6.

RIP

The RIP assay was carried out according to the manufacturer’s guidelines using a Magna RIP RNA-Binding Protein Immunoprecipitation kit (Millipore). In brief, 5 μg of control IgG antibody (1:20; Thermo), anti-RHA (Abcam, ab26271) was coupled to magnetic beads and treated overnight at 4 °C with H1975 cell lysates. The protein was then digested with proteinase K, and the RNA was extracted using phenol-chloroform. Finally, reverse transcription and qRT-PCR were used to measure cIGF1R expression.

Co-IP and MS

Cells were lysed in immunoprecipitation lysis buffer (Thermo) by spinning the cell lysate on a rotor wheel for 10 min at 4 °C. All supernatants were pooled after centrifugation of the samples at 15,000 g. The immunoprecipitation input was removed, and the samples were incubated with appropriate magnetic beads (Thermo) for 3 h at 4 °C with rotation. Afterwards, the beads were washed three times with immunoprecipitation wash buffer and eluted with Laemmli buffer. The purified proteins were analysed by IB and MS. MS-identified proteins in H1975 cells overexpression of EV or cIGF1R are listed in Supplementary Table 7, and MS-identified proteins in Co-IP products using anti-Flag antibodies in H1975 cells are listed in Supplementary Table 8.

Immunoblotting and antibodies

After washing the cells 3 times with ice-cold PBS, the cells were harvested by centrifugation at 1000 rpm for 5 min and then lysed using RIPA buffer (Thermo) containing phosphatase and protease inhibitors (Thermo) for 30 min at 4 °C. Collect the supernatant by centrifugation at 15,000 g for 15 min at 4 °C. Protein concentration was determined by BCA Protein Assay (Beyotime). Denatured lysates were separated on 4–20% SDS-PAGE gels and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore). Membranes were blocked in 5% non-fat dry milk for 2 h and then immunoblotted with the primary antibody overnight. Standard immunoblotting procedures were then followed using Alexa 680 and 800-conjugated species-specific secondary antibodies (Rockland). Western blots were visualised with an infra-red scanner (LI-COR). The following primary antibodies were used: from Cell Signalling Technology (all at 1:1000 dilution), IGF1R (9750 T), phospho-IGF1R Tyr1316 (28897 S), EGFR (4267 S), phospho-EGFR Tyr1068(3777 S), AKT (9272 S), phospho-AKT Ser473 (4060 S), ERK1/2 (4695 S), phospho-ERK1/2 Thr202/Tyr204 (4370 S), S6 (2217 S), phospho-S6 Ser235/236 (4858 S), Ubiquitin P4D1 (3936 S), DYKDDDDK Tag (14793 S); from Abcam, RHA (ab26271), TOMM20 (ab186735), Hsp60 (ab190828), USP30 (ab219969), Parkin (ab77924); from Proteintech, VDAC1 (55259-1-AP), LC3B (18725-1-AP), SLIRP (26006-1-AP), IMMT (26006-1-AP), ECH1 (11385-1-AP), MRPL12 (14795-1-AP), Beta-Tubulin (10094-1-AP), Hsc70 (66442-1-Ig), Flag-Tag (66008-4-Ig), MYC-Tag (16286-1-AP, 60003-2-Ig), 6×His-Tag (10001-0-AP, 66005-1-Ig), HA-Tag (51064-2-AP, 66006-2-Ig) and GAPDH (60004-1-Ig). The antibodies were employed at the manufacturer’s recommended dilutions. The full-length western blots can be found in the supplementary file titled “Uncropped blots and gels”.

IF

Immunofluorescence was performed using a specific antibody against Flag-Tag (Cell Signalling Technology, 8146 T) in Fig. 6n. H1975 cells were fixed with 4% formaldehyde for 15 min before being blocked for 60 min at room temperature with 5% normal goat serum. Alexa Fluor 488-labelled, or Alexa Fluor 555-labelled secondary antibodies (Thermo) were used for immunostaining. Nuclei were counterstained with DAPI.

IHC

IHC was performed on the formalin-fixed, paraffin-embedded mouse or human tumour tissue sections using an anti-Ki-67 (Proteintech, 27309-1-AP) antibody. IHC was carried out using an automated protocol developed for the BenchMark XT automated slide-staining system (Ventana Medical Systems) and was detected using an ultraView Universal DAB detection kit (Ventana Medical Systems). Hematoxylin II (Ventana-Roche) was used as a counterstain.

Molecular simulation assay

The crystal structure of VDAC1 was retrieved from the PDB bank (PDB ID: 6G6U, https://www.rcsb.org/) and was prepared using the Protein Preparation Wizard in Maestro with default options. The C-IGFR1 was built and refined by the “Build Peptide from Sequence” module of Maestro as well. Subsequently, the algorithm SiteMap was used to predict the potential binding site of VDAC1 [53]. The binding mode of VDAC1-C-IGFR1 was predicted by the “Protein-Protein Docking” module of Schrödinger [54].

Statistics and reproducibility

All data were obtained from independent experiments with independent biological replicates. No statistical methods were used to predetermine the sample size. Most results were qualitatively replicated in two different cell lines and provided consistent results using independent techniques. Investigators were not blinded to the assignment. No samples or experiments were excluded during the experiment and outcome assessment. Experimental data are expressed as mean and individual points, mean and standard deviation, or mean and range, as indicated in the figure legend. Immunofluorescence experiments were performed in at least three independent biological replicates, and for each replicate, several hundred cells were scored in each condition. The significance was tested using an unpaired two-tailed Student’s t-test for multiple comparisons with a single control. Tests were performed under the assumption that the values follow a normal distribution and have similar variances. Statistical analysis was performed using Prism v.9.4.

Author contributions

FJ and GD designed experiments and wrote the manuscript. XS performed statistical analysis with the R language. HW, YL, TZ, YC, LX, and XY completed the basic experiment part. BC, WX and QM were responsible for clinical sample collection and subsequent sample delivery. GD and FJ helped to revise the manuscript. All authors read and approved the final manuscript.

Data availability

CircRNA-seq and RNA-seq data have been deposited in the Gene Expression Omnibus (ncbi.nlm.nih.gov/geo) under accession number GSE211854. The TCGA Research Network (cancergenome.nih.gov) provided the RNA expression data of EGFR-wild-type and EGFR-mutant human LUAD. This document includes source data. All further data supporting the conclusions of this study are accessible upon reasonable request from the corresponding author.

Competing interests

The authors declare no competing interests.

Ethics

The research conducted in this manuscript adhered to all applicable ethical regulations. Human tissue samples were obtained from the Department of Thoracic Surgery at Jiangsu Cancer Hospital. Ethical approval was granted by Nanjing Medical University. All mouse research was authorised by the Institutional Animal Care and Use Committee (IACUC) of Nanjing Medical University (IACUC-2209003).

Supplementary information

The online version contains supplementary material available at 10.1038/s41418-023-01222-0.