Screening of translated and abnormally expressed circRNAs in GBM

To explore differentially expressed circRNAs with coding potential in tissues from patients with primary GBM and recurrent GBM, we performed circRNAs microarray analysis of 4 primary GBM samples and 4 recurrent GBM samples after standardized therapy. Through circRNAs microarrayanalysis, a total of 7367 circRNAs were found to be upregulated in recurrent GBM compared with primary GBM, and 6250 circRNAs were downregulated in recurrent GBM compared with primary GBM. According to the filter conditions |FC|≥ 1.3 and P < 0.05, 412 significantly upregulated circRNAs and 173 significantly downregulated circRNAs were identified via circRNAs microarrayin recurrent GBM compared with primary GBM (Fig. 1A, B). Figure 1C shows a schematic of the screening of differentially expressed circRNAs. Furthermore, the top ten downregulated circRNAs and the top ten upregulated circRNAs were selected according to their FC values, and a total of 20 circRNAs were used for subsequent analysis. The CircRNADb, circBank and TransCirc databases were used to explore the coding potential of the top 20 circRNAs identified above. Owing to the fact that the differences between the circRNADb and circBank scores were small, the ranking was mainly based on the TransCirc coding potential score. The results demonstrated the top 11 circRNAs with coding potential, including hsa_circ_0000745, hsa_circ_0002538, hsa_circ_0000348, hsa_circ_0007940, hsa_circ_0008950, hsa_circ_0008732, hsa_circ_0113446, hsa_circ_0134742, hsa_circ_0003275, hsa_circ_0006341 and hsa_circ_0008291 (Fig. 1D). Two downregulated circRNAs (hsa_circ_0000745 and hsa_circ_0002538), and one upregulated circRNA (hsa_circ_0007940) with the highest coding potential scores were selected for further coding potential analysis.

Fig. 1

Screening of translated and abnormally expressed circRNAs in GBM. A Heatmap of the expression of circRNAs in recurrent and primary GBM samples. B Volcano plot of dysregulated circRNAs in recurrent and primary GBM samples. C Screening of primary and recurrent GBM samples for differentially expressed circRNAs (|FC|≥ 1.3 and P < 0.05). D CircRNADb, circbank and TransCirc databases were applied to forecast the coding potential of the top ten upregulated and top ten downregulated circRNAs. E Upper: schematic representation of the interaction of 18S rRNA on the ribosomal 40S subunit with the complementary sequence of 18S on the IRES. Lower: secondary structure analysis of three candidate circRNAs with coding potential and their complementary sequences of 18S rRNA and SuRE structural elements. F Upper: schematic representation of p-circRNA-Flag plasmids construction for verifying of coding abilities of three candidate circRNAs. Lower: the coding abilities were verified by western blot experiment after transfection of p-circRNA-Flag plasmids in 293 T cells. G The circ_0000745 expression levels in non-tumor brain tissues (NTT) and glioblastoma (GBM) were analyzed via GSE86202, GSE109569 and GSE92322 datasets. H The circ_0007940 expression levels in NTT and GBM or high-grade glioma (HGG) were analyzed via GSE86202, GSE109569 and GSE165926 datasets. I Analysis of circ_0000745 and circ_0007940 expression in GBM cell lines (A172, U87, U251, B19, T98G, and LN229) and normal astrocytes (UC2)

Recent studies have demonstrated that the complementary sequence of 18S rRNA and the 40–60 nt SuRE on circRNA IRESs are important for promoting the translational initiation of endogenous circRNAs [23]. The SuRE may cause a pause in RNA unwinding, thereby increasing the chance that complementary 18S rRNA sequences on the IRES interact with 18S rRNA on the ribosome and consequently facilitating circRNA translation in a cap-independent manner (Fig. 1E upper). Therefore, the characteristics of the structural elements (18S rRNA complementary sequence and SuRE) in the IRES sequences of the abovementioned three circRNAs were analyzed via RNAfold WebServer. The results showed that there were 18S rRNA complementary sequences and SuRE elements in the circ_0000745, circ_0002538 and circ_0007940 IRESs, and all of the 18S rRNA complementary sequences occurred before the SuRE element. These results indicated that the IRESs of circ_0000745, circ_0002538 and circ_0007940 may encode proteins (Fig. 1E, lower). To further explore whether circ_0000745, circ_0002538, and circ_0007940 encode proteins or peptides, p-circRNA-Flag plasmids with 3 × Flag tag sequences were constructed (Fig. 1F, upper panel). The splicing site was located between the 3 × Flag tag sequences. Only when translation across the splice site was met, the plasmid could translate and express the complete Flag protein. The results showed that there were no Flag protein bands in the vector and p-circ_0002538-Flag groups, whereas p-circ_0000745-Flag and p-circ_0007940-Flag had a Flag protein band, and the protein size was consistent with the predicted size of the online website, thus suggesting that circ_0000745 and circ_0007940 both had the ability to encode proteins (Fig. 1F lower).

Compared with that in primary GBM samples, the circ_0000745 expression level was downregulated, and the circ_0007940 expression level was upregulated, in recurrent GBM samples. Afterwards, the expression levels of circ_0000745 and circ_0007940 in nontumor brain tissues (NTTs) and GBM or high-grade glioma (HGG) tissues were analyzed through four GEO datasets, including GSE86202, GSE109569, GSE92322, and GSE165926. The results showed that circ_0000745 expression was downregulated in patients with GBM compared with in patients with NTT in the GSE109569, GSE92322, and GSE92322 datasets (Fig. 1G). In contrast, circ_0007940 expression did not significantly differ between NTT and GBM or HGG in the GSE86202, GSE109569, and GSE165926 datasets (Fig. 1H). The results of the expression level analysis of various cell lines also demonstrated that circ_0000745 was highly expressed in UC2 cells and that it was expressed at considerably lower levels in GBM cell lines (A172, U87, U251, SNB19, T98G, and LN229), which is consistent with the abovementioned three GEO database analyses. However, the expression level of circ_0007940 had no obvious trend in UC2 or GBM cell lines, which was consistent with the results of the GEO database analysis (Fig. 1I). Therefore, circ_0000745, which has both coding potential and abnormal expression, was selected as the research object.

CircSPECC1 characteristics in GBM

According to the circBase database, hsa_circ_0000745 is located on the short arm of chromosome 17. It is formed via the splicing of exon 4 of the 15 exons of the NM_001243439 transcript that is transcribed by the parental gene sperm antigen with calponin homology and coiled-coil domains 1 (SPECC1). With a size of 1580 nt, it was named circSPECC1 (Fig. 2A). We used Sanger sequencing to confirm the specific back-splicing junction site sequence to confirm that circSPECC1 was circular (Fig. 2A, left panel). Agarose gel electrophoresis was used to determine the specificity and product size of the divergent circSPECC1 primer that was used to amplify the back-splicing junction site (Fig. 2A, right panel). The results showed that circSPECC1 could be amplified by this divergent primer with great specificity, and the product size was as expected. SPECC1 mRNA was amplified by both random primers and oligo (dT) primers; however, circSPECC1 could not be amplified by oligo (dT) primers compared with random primers (Fig. 2B). Furthermore, circSPECC1 demonstrated greater resistance to RNase R than SPECC1 mRNA, which suggests that circSPECC1 is nonlinear (Fig. 2C).

Fig. 2

Validation of the circular structure and subcellular localization of circSPECC1. A The location of circSPECC1 in the human genome (upper panel). Schematic diagram of circSPECC1 structure and Sanger sequencing to verify the splice site of circSPECC1 (left panel). The specificity of the divergent primers and product size were checked by agarose gel electrophoresis following quantitative reverse transcription polymerase chain reaction (RT-qPCR) (right panel). B RT-qPCR analysis of circSPECC1 expression after amplification with random primer or oligo (dT) primer. C RT-qPCR analysis of circSPECC1 expression after treated with RNase R. D CircSPECC1 expression detected by RT-qPCR after Act D treatment. E The presence of circSPECC1 in cDNA and gDNA samples from 293T cells was detected using divergent and convergent primer. F The distribution of circSPECC1 examined by cytoplasmic and nuclear RNA isolation. GAPDH and U6 in the cytoplasm and nucleus were used as positive controls, respectively. G CircSPECC1 distribution in the cytoplasm and nucleus was demonstrated by RNA fluorescence in situ hybridization. Every experiment was carried out a minimum of three times, and the mean ± standard deviation (SD) of the results was given. ns, P > 0.05; ****, P < 0.0001

Following actinomycin D (Act D) treatment, circSPECC1 mRNA remained more stable than SPECC1 mRNA (Fig. 2D). The divergent primers could specifically amplify circSPECC1, whereas the convergent primers could amplify SPECC1 mRNA. Agarose gel electrophoresis further confirmed that circSPECC1 was resistant to RNase R degradation and was highly stable (Fig. 2E). Afterward, we performed cell component extraction and RNA FISH analysis in GBM cell lines (LN229 and U251) and corresponding TMZ-resistant GBM cell lines (LN229R and U251R), which showed that circSPECC1 was distributed in both the cytoplasm and nucleus in these cells, with a predominance in the cytoplasm (Fig. 2F, G). Overall, these findings demonstrated that circSPECC1 is a circular and stable transcript in GBM cell lines.

CircSPECC1 suppressed the proliferation, migration, invasion, and colony formation abilities of GBM cells and restored GBM sensitivity to TMZ

To investigate circSPECC1 expression levels in clinical samples, we analyzed circSPECC1 expression in nine samples from recurrent GBM patients after receiving standardized therapy and 32 samples from patients with primary GBM. The results showed that circSPECC1 expression in recurrent GBM tissues was lower than that in primary GBM tissues (Fig. 3A). Furthermore, circSPECC1 expression in LN229R cells was lower than that in LN229 cells. Similarly, the circSPECC1 expression level in U251R cells was lower than that in U251 cells (Fig. 3B). These findings demonstrated that circSPECC1 may function as a tumor suppressor gene in GBM and is associated with TMZ sensitivity.

Fig. 3

CircSPECC1 inhibited GBM and TMZ-resistant GBM cell lines proliferation, migration, and invasion abilities in vitro. A Analysis of circSPECC1 expression levels in primary (n = 32) and recurrent (n = 9) GBM samples. B Analysis of circSPECC1 expression in GBM cell lines (LN229 and U251) and corresponding TMZ-resistant GBM cell lines (LN229R and U251R). C Cell viability assay of LN229/LN229R cells treated with various concentrations of TMZ for 48 h. D Cell viability assay of U251/U251R cells treated with various concentrations of TMZ for 48 h. E CCK-8 assays for LN229 and U251 cells proliferation with circSPECC1 knockdown after dimethylsulfoxide (DMSO) or temozolomide (TMZ) co-treatment. F CCK-8 assays for LN229R and U251R cells proliferation with circSPECC1 overexpression after DMSO or TMZ co-treatment. G Wound healing assays for LN229 and U251 cells migration with circSPECC1 knockdown after DMSO or TMZ co-treatment. H Wound healing assays for LN229R and U251R cells migration with circSPECC1 overexpression after DMSO or TMZ co-treatment. I Transwell assays for LN229 and U251 cells invasion with circSPECC1 knockdown after DMSO or TMZ co-treatment. J Transwell assays for LN229R and U251R cells invasion with circSPECC1 overexpression after DMSO or TMZ co-treatment. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

To explore the biological functions of circSPECC1, we used two siRNAs targeting the junction site of circSPECC1 to knock down circSPECC1 expression in GBM cell lines (LN229 and U251), as well as a circSPECC1 overexpression plasmid to overexpress circSPECC1 in TMZ-resistant GBM cell lines (LN229R and U251R). The knockdown efficiency of circSPECC1 siRNAs and the overexpression efficiency of the circSPECC1 overexpression plasmid were verified via RT‒qPCR (Supplementary Fig. 1A‒B).

Through CCK-8 assays, we obtained the half-maximal inhibitory concentration (IC50) values of TMZ in TMZ-resistant GBM cell lines (LN229R and U251R) and GBM cell lines (LN229 and U251). The results demonstrated that the IC50 values of GBM cell lines (LN229 and U251) were significantly lower than those of TMZ-resistant GBM cell lines (LN229R and U251R), thus indicating that TMZ-resistant GBM cell lines developed resistance to TMZ (Fig. 3C, D). Furthermore, the CCK-8 assay results demonstrated that knockdown of circSPECC1 increased the proliferation of U251 and LN229 cells (Fig. 3E), and circSPECC1 overexpression inhibited the proliferation of LN229R and U251R cells (Fig. 3F), regardless of whether DMSO or TMZ was used for co-treatment. A wound healing assay showed that knockdown of circSPECC1 expression resulted in increased migration of both the LN229 and U251 cell lines (Fig. 3G), and overexpression of circSPECC1 inhibited the migration of the LN229R and U251R cell lines (Fig. 3H) after treatment with either DMSO or TMZ. Transwell experiments demonstrated that knockdown of circSPECC1 expression increased the invasion ability of the LN229 and U251 cell lines (Fig. 3I), and the overexpression of circSPECC1 inhibited the invasion ability of the LN229R and U251R cell lines (Fig. 3J), regardless of DMSO or TMZ treatment. Knockdown of circSPECC1 expression also led to an increase in the colony formation ability of the LN229 and U251 cell lines (Fig. 4A). The overexpression of circSPECC1 inhibited the colony formation ability of LN229R and U251R cells treated with either DMSO or TMZ (Fig. 4B).

Fig. 4

Colony formation and comet assays with circSPECC1 knockdown or overexpression in vitro. A Colony formation assays for LN229 and U251 cells colony formation ability with circSPECC1 knockdown after DMSO or TMZ co-treatment. B Colony formation assays for LN229R and U251R cells colony formation ability with circSPECC1 overexpression after DMSO or TMZ co-treatment. C Comet assays for LN229 and U251 cells DNA damage degree with circSPECC1 knockdown after DMSO or TMZ co-treatment. D Comet assays for LN229R and U251R cells DNA damage degree with circSPECC1 overexpression after DMSO or TMZ co-treatment. ns, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001

Due to the difference in circSPECC1 expression between GBM cell lines and TMZ-resistant GBM cell lines, the impact of circSPECC1 expression on the TMZ sensitivity of GBM cells was verified by using an IC50 assay, a γH2AX immunofluorescence assay and a comet assay. The IC50 significantly increased when circSPECC1 was knocked down in GBM cell lines (LN229 and U251) and significantly decreased when circSPECC1 was overexpressed in resistant cell lines (LN229R and U251R) (Supplementary Fig. 2A, B). The results of the γH2AX immunofluorescence assay indicated that there was no discernible variation in the intensity of γH2AX fluorescence in LN229 or U251 cells after knocking down circSPECC1 expression under DMSO treatment, whereas knocking down circSPECC1 expression under TMZ treatment increased the DNA damage repair ability of LN229 and U251 cells and decreased their sensitivity to TMZ (Supplementary Fig. 2C, D). The γH2AX fluorescence intensity in LN229R and U251R cells did not significantly differ from one another after overexpression of circSPECC1 under DMSO treatment, whereas the overexpression of circSPECC1 decreased the DNA damage repair ability of LN229R and U251R cells under TMZ treatment. In addition, the sensitivity of LN229R and U251R cells to TMZ was restored (Supplementary Fig. 3A, B). Moreover, the comet assay results demonstrated that there was no significant difference in the degree of DNA tailing in LN229 or U251 cells after knocking down circSPECC1 expression under DMSO treatment, whereas the knockdown of circSPECC1 resulted in a significant reduction in the degree of DNA tailing in LN229 or U251 cells after TMZ treatment. The findings demonstrated that the sensitivity of LN229 and U251 cells to TMZ was markedly decreased after circSPECC1 knockdown (Fig. 4C). In LN229R and U251R cells, there was no significant difference in the degree of DNA tailing after overexpressing circSPECC1 under DMSO treatment, whereas overexpression of circSPECC1 caused a significant increase in the degree of DNA tailing in LN229R and U251R cells after TMZ treatment (Fig. 4D). These findings demonstrated that circSPECC1 overexpression markedly increased the sensitivity of LN229R and U251R cells to TMZ.

These results showed that circSPECC1 functions as a tumor suppressor gene and inhibits the proliferation, migration, invasion and colony formation ability of GBM cells; additionally, the sensitivity of TMZ-resistant GBM cells to TMZ can be restored by circSPECC1.

CircSPECC1 encoded the protein SPECC1-415aa

The abovementioned studies confirmed that circSPECC1 has an IRES sequence and that it has an active structure that drives translation (Fig. 1E). Western blot experiments also showed that circSPECC1 could encode proteins (Fig. 1F). Subsequently, we further explored whether circSPECC1 can encode a protein in GBM. Based on the prediction information from circBase and TransCirc online, the open reading frame (ORF) of circSPECC1 translated a new protein known as SPECC1-415aa, with 415 amino acids and a molecular weight of approximately 47 kDa, as well as 18 specific amino acids at the end (GPLQQLNGQAFQPHGNFQ). The amino acid sequence 1–397 of SPECC1-415aa completely overlapped with the amino acid sequence 225–621 of the protein encoded by the parent SPECC1 mRNA, and the amino acid sequence 398–415 was specific for SPECC1-415aa (Fig. 5A). This observation suggests that this protein has a unique biological function.

Fig. 5

Identification of circSPECC1 coding capacity. A Diagram of circSPECC1 structure and explanation of amino acid alignment between the encoded new protein by circSPECC1 and the parent SPECC1 protein. B Full-length IRES sequences of circSPECC1 or IRES sequences of its different truncated mutants were cloned between Rluc/Luc reporters with independent start (ATG) and stop (TGA) codons, including full-length IRES (165-286nt), IRES-1 (165-225nt) and IRES-2 (225-286nt). C The relative luciferase activity of Luc/Rluc in empty vector, full-length IRES, IRES-1 and IRES-2 vector detected by dual luciferase assay. D Description of vectors used to detect whether circSPECC1 can encode the protein. Control vector: the flanking sequences were deleted and circSPECC1 could not be expressed. SPECC1-415aa-3 × Flag vector: a linearized ORF of circSPECC1 with a 3 × Flag tag was cloned into a linear plasmid and used as a positive control. CircSPECC1-3 × Flag vector: this vector had the flanking sequences, splice acceptor (SA) and splice donor (SD) and could be able to express circSPECC1 and SPECC1-415aa with Flag tag protein. CircSPECC1-3 × Flag-Del vector: the start codon (ATG) was censored, which could express circSPECC1 but not translate SPECC1-415aa. All vectors had CMV promoters. E The expression of SPECC1-415aa in 293 T cells transfected with the above plasmids by western blot assay. F IP and western blot experiments verified that the circSPECC1-3 × Flag plasmid expressed the Flag tag protein in 293 T cells and was specifically enriched by Flag antibody. G Schematic diagram of the new protein SPECC1-415aa translated by circSPECC1 and the secondary spectrum of the specific peptide of SPECC1-415aa-3 × Flag protein after IP-MS. **, P < 0.01; ****, P < 0.0001

Studies have shown that the IRES of circRNA contains two key regulatory elements, including the complementary sequence of 18S rRNA and the SuRE element (40–60 nt), which can promote the translation of circRNA in a cap-independent manner[23]. The ability of the IRES of circSPECC1 to drive its encoded protein and to encode the new protein SPECC1-415aa was verified. First, the IRES sequence activity of circSPECC1 was detected. Dual luciferase plasmids containing the full-length (165–286 nt) and truncated IRES mutant 1 (IRES-1, 165–225 nt) IRESs, as well as the truncated IRES mutant 2 (IRES-2, 226–286 nt) IRESs, were constructed to detect IRES-driven translation (Fig. 5B). The dual luciferase assay results showed that the full-length IRES (nt 165–286) sequence of circSPECC1 drove translation, and the IRES activity was mainly localized in the IRES-2 (nt 226–286) region, which represents the second half of the IRES (Fig. 5C). To further verify that the ORF of circSPECC1 could be translated into the SPECC1-415aa protein driven by the IRES sequence, we constructed a linear cDNA plasmid encoding SPECC1-415aa with a 3 × Flag tag sequence (p-SPECC1-415aa-Flag), a circSPECC1 overexpression plasmid (p-circSPECC1-Flag), and a circSPECC1 plasmid with an ATG deletion at the start codon (p-circSPECC1-Del-Flag) (Fig. 5D). Western blot results demonstrated that both the p-SPECC1-415aa-Flag and p-circSPECC1-Flag plasmids could express the predicted size protein (SPECC1-415aa). In contrast, the p-circSPECC1-Del-Flag plasmid without ATG at the start codon failed to express the protein of the predicted size (Fig. 5E). This indicated that the ORF of circSPECC1 was able to be translated into the protein SPECC1-415aa in the presence of the ATG initiation codon. These results indicated that the IRES sequence of circSPECC1 was active in driving translation and encoded a protein (SPECC1-415aa) with a molecular weight of approximately 47 kDa (approximately 50 kDa when the 3 × Flag tag was included).

Furthermore, the abovementioned assays confirmed that circSPECC1 could encode a protein and further confirmed that the protein encoded by circSPECC1 was consistent with the predicted protein amino acid sequence. 293 T cells were transfected with the constructed p-circSPECC1-Flag plasmid, after which immunoprecipitation and mass spectrometry amino acid sequence analysis (IP-MS) were conducted. Immunoprecipitation (IP) demonstrated that the p-circSPECC1-Flag plasmid was translated and expressed a protein with a molecular weight of approximately 50 kDa in 293 T cells, thus indicating that the SPECC1-415aa protein could be successfully expressed in 293 T cells and was highly enriched by the Flag tag antibody (Fig. 5F). The IP products enriched by the Flag tag antibody were analyzed by using mass spectrometry (MS). MS demonstrated a specific peptide segment unique to SPECC1-415aa (the red marked part represents the specific peptide sequence of SPECC1-415aa, and the purple marked part represents the 3 × Flag sequence) (Fig. 5G). The translation of the protein SPECC1-415aa was confirmed in circSPECC1.

CircSPECC1 regulated GBM biological function and sensitivity to TMZ by encoding the protein SPECC1-415aa

In contrast to primary GBM samples, recurrent GBM samples had aberrant expression of circSPECC1, thus regulating GBM cell sensitivity to TMZ. We hypothesized that circSPECC1 regulates GBM sensitivity to TMZ through its encoded protein SPECC1-415aa. First, CCK-8 assays demonstrated that the proliferation of LN229R and U251R cells transfected with p-circSPECC1 and p-SPECC1-415aa plasmids (both of which encode the SPECC1-415aa protein) was significantly inhibited by treatment with DMSO or TMZ. Cell proliferation was more inhibited in the TMZ cotreatment group than in the DMSO group. However, transfection of the p-circSPECC1-Del overexpression plasmid (which does not encode the SPECC1-415aa protein) did not significantly inhibit the proliferation of LN229R and U251R cells in the TMZ cotreatment group (Fig. 6A, B).

Fig. 6

CircSPECC1 regulated the sensitivity of GBM to TMZ through SPECC1-415aa protein. A The effects of p-circSPECC1, p-SPECC1-415aa or p-circSPECC1-Del plasmids transfection on the proliferation ability of LN229R cells in the company of DMSO or TMZ. B The effects of p-circSPECC1, p-SPECC1-415aa or p-circSPECC1-Del plasmids transfection on the proliferation ability of U251R cells in the company of DMSO or TMZ. C The effects of p-circSPECC1, p-SPECC1-415aa or p-circSPECC1-Del plasmids transfection on the degree of DNA damage of LN229R and U251R cells by immunofluorescence assay in the presence of DMSO or TMZ. D In the presence of DMSO or TMZ, the effects of p-circSPECC1, p-SPECC1-415aa or p-circSPECC1-Del plasmids transfection on the degree of DNA damage in LN229R and U251R cells by comet assay. E The expression level of SPECC1-415aa protein in 9 samples of recurrent GBM patients and 32 samples of primary GBM patients via western blot experiment. ns, P > 0.05; ***, P < 0.001; ****, P < 0.0001

In addition, γH2AX immunofluorescence assays demonstrated that the extent of DNA damage did not significantly differ between LN229R and U251R cells transfected with p-circSPECC1, p-SPECC1-415aa or p-circSPECC1-Del plasmids under DMSO treatment. However, when cotreated with TMZ, the degree of DNA damage in LN229R and U251R cells transfected with p-circSPECC1 or p-SPECC1-415aa was significantly increased. Transfection of the p-circSPECC1-Del plasmid had no significant effect on the degree of DNA damage in LN229R and U251R cells treated with TMZ (Fig. 6C). Comet assays also demonstrated that DNA damage did not significantly differ between LN229R and U251R cells transfected with p-circSPECC1, p-SPECC1-415aa or p-circSPECC1-Del plasmids after DMSO treatment. However, compared with control cells, LN229R and U251R cells transfected with p-circSPECC1 and p-SPECC1-415aa plasmids showed a significant increase in comet tailing after TMZ treatment, indicating a significant increase in DNA damage. The overexpression of the p-circSPECC1-Del plasmid had no significant effect on DNA damage in LN229R or U251R cells (Fig. 6D). These results suggested that circSPECC1 regulated the DNA damage repair capacity of GBM to affect GBM sensitivity to TMZ by encoding the SPECC1-415aa protein.

The abovementioned results verified that the biological function of circSPECC1 in GBM was mainly achieved through its encoded protein SPECC1-415aa. We further analyzed the clinical characteristics of SPECC1-415aa. The mRNA expression level of circSPECC1 in primary and recurrent GBM tissues was consistent with the previous results, which demonstrated that the expression level of SPECC1-415aa in recurrent GBM tissues was significantly lower than that in primary GBM tissues (Fig. 6E).

M6A modification promoted circSPECC1 expression

Our study demonstrated that circSPECC1 was expressed at low levels in recurrent GBM samples, and TransCirc online database prediction indicated the presence of m6A modification sites on circSPECC1. Therefore, we hypothesized that m6A modification may occur on circSPECC1 and affect its expression level in GBM. First, the predictive analysis of the SRAMP database demonstrated multiple m6A modification sites on circSPEEC1, including four sites with very high confidence (score > 0.8) (Supplementary Fig. 4). MeRIP assays demonstrated m6A modification on circSPECC1, and the level of m6A modification on circSPECC1 in a TMZ-resistant GBM cell line (LN229R/U251R) was lower than that in a GBM cell line (LN229/U251). These results indicated that the level of m6A modification on circSPECC1 was lower in TMZ-resistant GBM cell lines than in GBM cell lines (Fig. 7A).

Fig. 7

M6A modification on circSPECC1. A Verifying the m6A modification on circSPECC1 in TMZ-resistant GBM cell lines and GBM cell lines by MeRIP assay. B Verifying the knockdown efficiency of IGF2BP1/2/3 siRNA in LN229R and U251R cell lines. C RT-qPCR assay verifying the effect of IGF2BP1/2/3 knockdown on the expression level of circSPECC1 in LN229R and U251R cell lines. D RIP assay used to verify the binding of IGF2BP1 and circSPECC1 in LN229R and U251R cell lines. E Immunofluorescence assay used to verify the colocalization of IGF2BP1 and circSPECC1 in LN229, LN229R, U251 and U251R cells. F Effect of IGF2BP1 knockdown on the stability of circSPECC1 through Act D assay. G Western blot assay used to verify the expression level of IGF2BP1 between TMZ-resistant GBM cells and corresponding GBM cells. H Correlation analysis of IGF2BP1 mRNA and circSPECC1 expression in primary (n = 32) and recurrent (n = 9) GBM samples. ns, P > 0.05; *, P < 0.05; **, P < 0.01. ***, P < 0.001; ****, P < 0.0001

Recent studies have shown that m6A reader proteins, including YTHDC1/2, YTHDF1/2/3, and IGF2BP1/2/3, can increase or reduce RNA stability [24,25,26,27]. Studies have shown that YTHDF and YTHDC family members mainly play biological roles in promoting circRNA degradation [28, 29]. However, IGF2BP family members mainly play biological roles in promoting circRNA stabilization [30, 31]. For example, circARHGAP12 was shown to contribute to the development of cervical cancer through the IGF2BP2/FOXM1 pathway, which is dependent on m6A. Moreover, RNA immunoprecipitation (RIP), fluorescence in-situ hybridization (FISH), and Act D experiments confirmed the binding of IGF2BP2 to circARHGAP12 and the ability of IGF2BP2 to increase the stability of circARHGAP12 [26]. Owing to the fact that the IGF2BP family can promote the stability of circRNAs, which is consistent with our meRIP experimental findings, IGF2BP1/2/3 were selected for further assessments.

To investigate the impact of IGF2BP1/2/3 on the stability of circSPECC1, siRNAs were used to knockdown the expression of IGF2BP1, IGF2BP2, and IGF2BP3. RT‒qPCR was used to assess the knockdown efficiency of si-IGF2BP1, si-IGF2BP2, and si-IGF2BP3 siRNAs in LN229R and U251R TMZ-resistant GBM cell lines. The results showed that the si-IGF2BP1, si-IGF2BP2, and si-IGF2BP3 siRNA sequences were effective (Fig. 7B). Afterwards, RT‒qPCR was used to further determine the effect of IGF2BP family members on circSPECC1 expression. The findings demonstrated that only IGF2BP1 affected the expression level of circSPECC1 after IGF2BP1/2/3 knockdown in the LN229R and U251R cell lines (Fig. 7C). CircSPECC1 expression was consistent with the trend of IGF2BP1 expression. Therefore, it was speculated that IGF2BP1 may be a m6A reader protein that binds to circSPECC1. Further RIP assays showed that IGF2BP1 could bind to circSPECC1 in the LN229R and U251R cell lines (Fig. 7D). Immunofluorescence experiments also demonstrated the colocalization of IGF2BP1 and circSPECC1 in LN229, LN229R, U251, and U251R cells, thus indicating that IGF2BP1 binds to circSPECC1 (Fig. 7E). Moreover, the Act D assay demonstrated that knockdown of IGF2BP1 expression in the LN229R and U251R cell lines resulted in a shortened half-life and decreased stability of circSPECC1 (Fig. 7F).

The level of m6A modification of circSPECC1 was lower in TMZ-resistant GBM cells than in corresponding GBM cells (Fig. 7A) and the m6A reader IGF2BP1 affected the stability of circSPECC1 (Fig. 7F). On the other hand, the protein expression level of IGF2BP1 was not significantly different between TMZ-resistant GBM cells and corresponding GBM cells (Fig. 7G). The above results suggested that the reduction of circSPECC1 expression in TMZ-resistant GBM cells compared with corresponding GBM cells was associated with the reduction of m6A modification occurring on circSPECC1. These findings indicated that the m6A reader protein IGF2BP1 bound to circSPECC1 and promoted its stability, which could also explain why circSPECC1 expression was reduced in TMZ-resistant GBM cells.

To validate the abovementioned findings in clinical samples, we further analyzed the correlation between circSPECC1 and IGF2BP1 mRNA in 9 recurrent GBM samples and 32 primary GBM samples. The results of the correlation analysis demonstrated a positive correlation between the levels of IGF2BP1 mRNA and circSPECC1 expression (Fig. 7H).

SPECC1-415aa competitively bound ANXA2 with EGFR

Previous studies have verified that circSPECC1 functions as a tumor suppressor gene in GBM by encoding the protein SPECC1-415aa, and SPECC1-415aa can affect the sensitivity of GBM cells to TMZ. Afterward, the molecular mechanism by which SPECC1-415aa regulates TMZ sensitivity in GBM cells was explored. First, transcriptome sequencing was performed in the LN229 GBM cell line after knocking down circSPECC1. The cluster heatmap and volcano map showed the differentially expressed mRNAs after knocking down circSPECC1, and a total of 335 significantly downregulated mRNAs and 196 significantly upregulated mRNAs were identified (Supplementary Fig. 5A, B). Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed on the 531 differentially expressed mRNAs. KEGG enrichment analysis demonstrated that circSPECC1 mainly affected the PI3K-AKT signaling pathway, NF-kappa B signaling pathway, NOD-like receptor signaling pathway, IL-17 signaling pathway and ECM-receptor interaction (Supplementary Fig. 5C). In the GO analysis, biological process (BP) enrichment analysis showed that circSPECC1 mainly affected the negative regulation of cellular processes, response to stress, the cell surface receptor signaling pathway, regulation of apoptotic processes, regulation of cell death, and response to biotic stimuli (Supplementary Fig. 5D). Cellular component (CC) analysis demonstrated that the functional sites of circSPECC1-regulated genes were mainly localized in the cytosol, extracellular region, and extracellular matrix (Supplementary Fig. 5E). In addition, molecular function (MF) analysis demonstrated that the main molecular functions impacted by circSPECC1 included receptor regulation activity, signaling receptor binding, G protein-coupled receptor binding and receptor ligand activity (Supplementary Fig. 5F). These results suggested that circSPECC1 mainly affects the PI3K-AKT and NF-kappa B signaling pathways and mainly regulates biological processes, such as cell apoptosis, receptor ligand activity, signal receptor binding, and receptor regulatory activity, by regulating related mRNAs in the cytoplasm and extracellular matrix.

To investigate the molecular mechanism by which SPECC1-415aa regulates the sensitivity of GBM cells to TMZ, immunoprecipitation-mass spectrometry (IP-MS) experiments were performed in 293 T cells using a linear SPECC1-415aa overexpression plasmid with a 3 × Flag tag (p-SPECC1-415aa-Flag). After screening for the number of unique peptides (more than 2), the results showed that IgG and Flag had 77 common binding proteins, with 61 specific binding proteins in the IgG group and 38 specific binding proteins in the Flag group (Fig. 8A, left panel). Among these 38 proteins, we further screened proteins with more than five unique peptides, and a total of six proteins were found, including HSPA1A, MYH6, MYL3, CASP14, ANXA2, and ACTC1. Through comprehensive analysis of its subcellular localization and biological functions in tumors, as well as via enrichment analyses via transcriptome sequencing, annexin A2 (ANXA2) was identified for further study. Figure 8A (right panel) shows the secondary spectra of the specific peptide of ANXA2 that can bind to SPECC1-415aa in the IP‒MS assay. Further IP experiments showed that exogenous SPECC1-415aa-Flag was able to bind to the endogenous ANXA2 protein in the LN229R and U251R cell lines (Fig. 8B, left panel). The IP experiments further showed that endogenous SPECC1-415aa could bind to the endogenous ANXA2 protein in LN229 and U251 cell lines (Fig. 8B, right panel). Immunofluorescence assays also verified that ANXA2 colocalized with SPECC1-415aa in LN229, LN229R, U251, and U251R cells and mainly colocalized to the cytoplasm (Fig. 8C). These results indicated that SPECC1-415aa can bind to the ANXA2 protein in GBM cell lines and corresponding TMZ-resistant GBM cell lines.

Fig. 8

Validation of SPECC1-415aa binding to ANXA2 to regulate the EGFR-AKT pathway. A The binding proteins (Unique peptides > 2) of IgG group and Flag group and the secondary spectrum of ANXA2 specifically bound to the Flag protein in 293 T cells by IP-MS. B Verifying the bindings between endogenous ANXA2 protein and exogenous SPECC1-415aa-Flag (Left) in LN229R and U251R cells via IP assay. Verifying the bindings between endogenous ANXA2 protein and endogenous SPECC1-415aa (Right) in LN229 and U251 cellsby IP experiment. C Immunofluorescence assay verifying the interaction between ANXA2 protein and SPECC1-415aa protein in LN229, LN229R, U251 and U251R cell lines. D IP experiment verifying the binding level of ANXA2 and EGFR after the overexpression of circSPECC1 in U251R and LN229R cells. E IP experiment verifying the binding level of ANXA2 and EGFR after the circSPECC1 knockdown in U251 and LN229 cells. F Impact of circSPECC1 overexpression on activation of EGFR-AKT pathway in LN229R and U251R cells. G Effect of circSPECC1 knockdown on activation of EGFR-AKT pathway in LN229 and U251 cells. H Effect of circSPECC1 knockdown or circSPECC1 knockdown combined with EGFR blocker Osimertinib on the activation of EGFR-AKT pathway in LN229 and U251. I Osimertinib could restore the sensitivity of LN229 and U251 cells to TMZ after knocking down circSPECC1 via EdU experiment. J Osimertinib could restore the sensitivity of LN229 and U251 cells to TMZ after knocking down circSPECC1 via TUNEL experiment. ns, P > 0.05; ***, P < 0.001; ****, P < 0.0001

Current studies have shown that proteins encoded by circRNAs can not only promote the binding of target proteins with their interacting proteins but also competitively inhibit the binding of target proteins with their interacting proteins to perform biological functions in tumors [32, 33]. For example, Wang et al. reported that circMAPK14-encoded MAPK1