Hsa_circ_0058124 promotes papillary thyroid cancer tumorigenesis and invasiveness through the NOTCH3/GATAD2A axis

This study was approved by the Ethics Committee of Nanjing Medical University and was conducted in accordance with the government policies and the Declaration of Helsinki. PTC and paired adjacent normal tissues were obtained from 92 patients (Nanjing Medical University Affiliated Cancer Hospital, Nanjing) with a definite pathological diagnosis of PTC during operation. The adjacent normal tissues were collected 3 cm away from the PTC edge and were confirmed no tumor cells by two specialized pathologists. All samples were immediately frozen in liquid nitrogen after collection. After histopathological vetting, a total of 12 samples, consisting of eight PTC samples (including four invasive tumor tissues with extrathyroidal extension and metastasis and four non-invasive tumor tissues with no extrathyroidal extension or metastasis) and four matching adjacent normal tissues, were ultimately selected for ceRNA microarray analysis.

Total RNA extraction and quantification, RNA purification, and cDNA synthesis were conducted as described previously [22]. For RNase R treatment, 2 μg total RNA was incubated for 15 min at 37 °C with or without 3 U/μg RNase R (Epicentre Technologies, WI, USA). To detect RNA expression, quantitative real-time PCR (qRT-PCR) was performed with PowerUp™ SYBR™ Green Master Mix (Thermo Fisher, MA, USA) and the Applied Biosystems StepOnePlus™ Real-time PCR Detection System (Life Technologies, CA, USA). The relative gene expression was calculated using the 2^-ΔCT method normalized to GAPDH, and the fold change of gene expression was calculated by the 2^-ΔΔCT method. The primer sequences are shown in Additional file 1: Table S1. Bulge-loop™ miRNA qRT-PCR Primer Sets (one RT primer and a pair of qPCR primers for each set) specific for miR-218-5p was designed by RiboBio (Guangzhou, China). The relative expression of miR-218-5p was normalized to human U6 snRNA.

TPC-1 and K1 cells were purchased from University of Colorado Cancer Center Cell Bank. These PTC cell lines were maintained in a humidified incubator (95% air/5% CO₂) at 37 °C in 90% RPMI 1640 (Invitrogen, CA, USA) supplemented with 10% fetal bovine serum (FBS; Invitrogen, CA, USA), 100 U/ml penicillin, and 100 μg/mL streptomycin. The cells were passaged at a 1:3 ratio in 10 cm dishes once 90% confluence was reached. All PTC cell lines used in this study were reported previously [21, 23]. Purchased siRNAs (RealGene, Nanjing, China) were used in transfection experiments. The TPC-1 and K1 cells were cultured to 50–60% confluence, and the transfection was carried out with Lipofectamine 3000 (Invitrogen, CA, USA). Sequences of the siRNAs are shown in Additional file 1: Table S2. To construct a plasmid expressing NUMB, the full-length human NUMB sequence (NCBI Reference Sequence: NM_001005745.1) was synthesized and cloned into the pCDNA3.1 vector (RealGene, Nanjing, China).

Cells were seeded at a density of 2 × 10³ cells per well (100 μL) in a 96-well plate. Cell viability was evaluated with the cell counting kit-8 (CCK-8; Bimake, Shanghai, China) after 0, 24, 48, 72, and 96 h of transfection. The absorbance was determined with a BioTek Synergy HTX multi-mode reader at 450 nm.

Crystal violet (KeyGen, China) was used to detect colony formation. About 200 cells were seeded in each well of 6-well plates. After transfection, the growth medium was changed every 3 days. After 9 days, the cells were fixed with 75% ethanol for 2 h, followed by staining with 0.2% crystal violet for 2 h, and colonies (> 50 cells/colony) were counted and photographed. Each experiment was performed three times independently.

EdU assays were performed using the Cell Light™ EdU kit (RiboBio, Guangzhou, China). Briefly, 6 × 10⁴ cells were seeded in well apiece of a 24-well plate. After transfection for 48 h, the culture medium was supplemented with 50 μM EdU for another 2 h. The cells were fixed with 4% formaldehyde for 30 min and incubated with glycine (2 mg/mL) for 10 min. The cells were then permeabilized with 0.5% Triton X-100 for 20 min, and then Hoechst 33342 were added. Cell proliferation (the percentage of EdU positive cells) was then determined by fluorescence microscopy (Olympus IX73, Japan). Assays were performed independently three times in triplicate.

After transfection for 48 h, the cells were collected for cell cycle and apoptosis assays. The apoptosis assay was performed with the Apoptosis Detection Kit (BD Pharmingen, CA, USA) by suspending the cells in 100 μL binding buffer containing 5 μL PE and 5 μL 7-ADD, incubating them in the dark for 15 min, and adding 400 μL binding buffer to resuspend the cells. The cells were divided into viable cells, dead cells, early apoptotic cells, and late apoptotic cells. The early and late apoptotic cells were measured. For cell cycle analysis, the cells were fixed in 75% ethanol overnight at − 20 °C, then stained with propidium iodide (PI) for 30 min, away from light. The cells were analyzed by flow cytometry (FACS Calibur Flow Cytometer, BD Biosciences, USA) to quantify the cell cycle or cell apoptosis. All experiments were performed independently in triplicate.

Transwell migration and invasion abilities were assessed using 8 μm pore transwell chambers either without Matrigel (for migration assays) or with Matrigel (for invasion assays). Each well of the upper chamber was filled with 300 μL of serum-free RPMI 1640 medium containing 1 × 10⁵/mL infected cells. The lower chambers contained 700 μL complete medium. The cells were fixed in 4% paraformaldehyde and stained with crystal violet after a 48-h incubation. The cells remaining in the top chamber were removed with cotton swabs, and imaged by light microscopy (Olympus, China). All assays were performed three times independently using triplicate wells.

Scratch wound assays were conducted by measuring the widths of the scratches at different times. Each well of 6-well plates was plated with 3 × 10⁵ cells. After 48 h incubation, when the well was completely covered with cells, a straight scratch was made with a 200 μl pipette tip. Images of the same fields were taken at different time points (0, 12, 24, 30, 36, and 40 h) after the scratch. Images at different time points were used to create an animation sequence in accordance with the timeline. The experiment was repeated at least three times independently.

We transiently transfected 3 × 10⁶ TPC-1 cells with 100 nM indicated siRNAs for 48 h, and the total RNA samples were collected using TRIzol reagent for whole-genome transcriptome profiling by RNA sequencing. The RNA sequencing and bioinformatics data analysis were performed by Shanghai Biotechnology Corporation. In brief, total RNA was extracted using the RNeasy Micro Kit (Qiagen, Hilden, Germany) and purified with the RNAClean XP Kit (Beckman Coulter, Inc., CA, USA). Paired-end libraries were then synthesized using the TruSeq® RNA Sample Preparation Kit (Illumina, USA) following the TruSeq® RNA Sample Preparation Guide. The purified libraries were sequenced on the Illumina HiSeq X-ten (Illumina, USA). For data processing, we used Hisat2 (version:2.0.4) [24] to map the cleaned reads to the human GRCh38 reference genome with two mismatches. After genome mapping, Stringtie (version:1.3.0) [25] was run with a reference annotation to generate FPKM (Fragments Per Kilobase of transcript per Million mapped reads) values for known gene models. The P-value and FDR were used to define the threshold of significance.

To gain insight into hsa_circ_0058124-mediated molecular pathways in PTC, we used the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis (https://​www.​kegg.​jp). GSEA was also performed using the Broad Institute GSEA version 3.0 software. The gene sets used for the enrichment analysis were downloaded from the Molecular Signatures Database (MsigDB, http://​software.​broadinstitute.​org/​gsea/​index.​jsp).

The proteins in PTC cells were extracted using the Total Protein Extraction Kit (Thermo Fisher, MA, USA). The protein extracts (30–40 μg) were separated on 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis gels and then electrophoretically transferred to polyvinylidene difluoride membranes (Millipore, USA). After blocking in 5% non-fat milk for 2 h, the membranes were incubated overnight at 4 °C with primary antibodies recognizing N3ICD, HES1 (1:1000 dilution; Cell Signaling Technology, USA), NUMB (1:1000 dilution; Abcam, UK), GATAD2A (1:1000; Proteintech, USA), and GAPDH (1:1000 dilution; Proteintech, USA). After incubation with secondary antibodies (1:5000 dilution; Jackson ImmunoResearch, PA, USA), the protein bands were visualized by chemiluminescence using a GE Amersham Imager 600 (GE, USA).

To construct luciferase reporter vectors, the full length of hsa_circ_0058124 or the 3’UTR fragment of NUMB (NM_001005745.1; 3’UTR: 2104–3467) containing the predicted potential binding sites were cloned at the XhoI and NotI sites into the pmiR-RB-REPORT™ luciferase reporter vector (RiboBio., Guangzhou, China). The hsa_circ_0058124 mutant sequence or the 3’UTR mutant sequence of NUMB, in which seven nucleotides were mutated within the miR-218-5p binding sites, were also constructed into the vector.

For luciferase activity assay, each construct was cotransfected with indicated miRNAs (RiboBio) in 293 T cells using Lipofectamine 3000 (Invitrogen) for 48 h. Luciferase assays were performed with the Dual-Luciferase Reporter Assay System (Promega, WI, USA) according to the manufacturer’s instructions. Luminescent signal was quantified by BioTek Synergy HTX multi-mode reader, and luciferase activity was presented by relative hRluc/ hluc ratio.

All tissues were fixed overnight in a formalin solution, dehydrated in ethanol, embedded in paraffin, and sectioned at 5 μm. Then, the slides were treated with xylene and ethanol to remove the paraffin. The slides were blocked with 5% normal goat serum and incubated with anti-GATAD2A overnight at 4 °C. After washing with PBS, the slides were incubated with goat anti-rabbit horseradish peroxidase (Jackson ImmunoResearch, PA, USA) for 1 h at room temperature, and was detected the immunohistochemical reactions using a DAB kit (Vector Laboratories, CA, USA). The slides were examined under a phase contrast light microscope (Olympus IX73, Japan), and the average integral optical density of each positively stained slide was measured using an Image-Pro plus 6.0 image analysis system. Three areas were chosen randomly from each section for measurement.

For validation of differentially expressed genes, the IHC data of thyroid cancer tissues and normal thyroid tissues were downloaded from Human Protein Atlas [26] available from www.​proteinatlas.​org. Protein expression levels were scaled based on an overall antibody staining score ranging from not detected to high.

All animal experiments were performed in accordance with a protocol approved by the Institutional Animal Care and Use Committee of Nanjing Medical University. Nude mice (5 in each group) aged 5–6 weeks were injected subcutaneously with 0.1 ml of a cell suspension containing 3.0 × 10⁶ TPC-1 cells for the control group, the LV-sh-hsa_circ_0058124 group, the LV-sh-NOTCH3 group, the LV-sh-GATAD2A group, the LV-sh-hsa_circ_0058124 together with LV-sh-NOTCH3 group or the LV-sh-hsa_circ_0058124 together with LV-sh-GATAD2A group (Genechem, Shanghai, China) in the right flank. After the tumor was palpable, it was measured every 3 days. Its volume was calculated according to the formula (volume = length × width² × 0.5).

A total of twelve fresh PTC tissue samples, decribed in the previous paragraph, were selected for ceRNA microarray analysis. Detailed information of patient characteristics is provided in Additional file 1: Table S3.

The circRNA expression in the PTC samples was profiled using a 4 × 180 K ceRNA microarray (circRNA, lncRNA, and mRNA). The differentially expressed circRNAs were identified by fold-change filtering (|fold change| > 2) and the Student’s t-test (P-value < 0.01), which revealed 607 circRNAs that were significantly differentially expressed in the invasive tumor vs. adjacent normal tissue set (Fig. 1a) and 49 circRNAs that were significantly differentially expressed in invasive vs. non-invasive tumor set (Fig. 1b). Hierarchical clustering was then performed to demonstrate the circRNA expression patterns among the sets (Fig. 1a, b).

To investigate the contribution of circRNAs to the tumorigenesis and invasiveness of PTC, we calculated the intersection of differential expressions of circRNA between the invasive vs. non-invasive tumor set and the invasive tumor vs. adjacent normal tissue set. This narrowed the number of candidate circRNAs down to 13. Figure 1c and d showed the Venn diagrams for this intersection and identified 11 significantly upregulated circRNAs and 2 downregulated circRNAs that overlapped the results of the previous comparisons (Additional file 1: Table S4). The mRNA expression profiles among the sets were also demonstrated using hierarchical clustering analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis (Fig. 2a, b). To investigate the potential target gene of the identified 13 circRNAs, a coexpression network analysis was used to screen these circRNAs in the eight PTC tumor samples (Fig. 2c). We also randomly selected 6 target genes to validate this circRNA-mRNA coexpression by qRT-PCR in 20 PTC tissues (Additional file 2: Figure S1). We then experimentally validated these 13 circRNA expression levels by qRT-PCR in larger-scale PTC tissue samples. The qRT-PCR results indicated hsa_circ_0058124 showed highest fold-change in the PTC tissues than in the control tissues, and was significantly upregulated in the invasive PTC tissues compared with non-invasive tumor tissues (Additional file 2: Figure S2). We therefore chose hsa_circ_0058124 for further analysis. The coexpression network for hsa_circ_0058124 was also shown in Fig. 2d.

Hsa_circ_0058124 is located on chromosome 2, is 864 base pairs (bp) in length, and consists of 5 exons (exons 15–19) from its host gene Fibronectin 1 (FN1) genome (Fig. 3a; Additional file 3). FN1, a fundamental component of the extracellular matrix, has been considered as a biomarker of PTC, and an important modulator of thyroid cancer aggressiveness [10, 27]. We avoided trans-splicing or genomic rearrangements, including head-to-tail splicing, by employing several universal circRNA detection methods [28]. We first designed convergent primers to amplify FN1 mRNA (a 184-bp fragment) and divergent primers to amplify hsa_circ_0058124 (a 83-bp fragment). We used cDNA and genomic DNA from two randomly selected PTC tissue and two PTC cell lines (TPC-1 and K1) as templates to amplify hsa_circ_0058124 from cDNA using only divergent primers (an expected 83-bp fragment), while no amplification product was observed from genomic DNA (Fig. 3b). We also used Sanger sequencing to confirm the head-to-tail splicing in the RT-PCR product of hsa_circ_0058124 with the expected size (Fig. 3c). RNase R is an exoribonuclease that can degrade RNA from its 3′ to 5′ end but does not act on circRNA [29]. In contrast to the linear FN1 mRNA, hsa_circ_0058124 was resistant to RNase R (Fig. 3d). The cell fraction assay and FISH analysis against hsa_circ_0058124 showed that hsa_circ_0058124 was primarily located in the nucleus and also existed in cytoplasm (Fig. 3e-f). Our results implied that hsa_circ_0058124 is a bona fide exonic circRNA that is abundant and stable in TPC-1 cells.

Quantitative real-time PCR analysis was conducted to detect hsa_circ_0058124 expression patterns in 92 paired PTC tissue samples. The hsa_circ_0058124 expression was significantly upregulated in most tumor tissues (63/92, 68.48%) when compared with the adjacent normal counterparts (Fig. 4a, b). We also investigated the clinical significance of hsa_circ_0058124 by analyzing the correlations between its expression level and clinicopathological characteristics. The median value of the relative expression of hsa_circ_0058124 in PTC tissues was 1.514, and this was used as a cut-off value to divide the patients into two groups: a high hsa_circ_0058124 relative expression (≥1.514; n = 46) and a low hsa_circ_0058124 relative expression (< 1.514; n = 46). The clinicopathological characteristics of the 92 PTC patients are shown in Table 1 and Fig. 4. The analysis results showed that hsa_circ_0058124 expression level was significantly higher in patients with large tumor size, advanced stage, extrathyroidal extension, lymph node metastasis, or distant metastasis (Table 1, Fig. 4c). No significant correlation was noted between hsa_circ_0058124 expression and age, gender, or multifocality (p > 0.05; Table 1). These results suggest that hsa_circ_0058124 was upregulated in PTC and correlated with unfavorable prognosis. Hsa_circ_0058124 might therefore be considered as a novel prognostic biomarker.

We further studied the functional role of hsa_circ_0058124 in PTC cell lines by designing two siRNA oligonucleotides that would target the unique backsplice junction. These backsplice junction-specific siRNAs successfully decreased hsa_circ_0058124 expression but did not affect the linear FN1 mRNA level in TPC-1 and K1 cells (Fig. 5a). We then observed that hsa_circ_0058124 knockdown significantly decreased the cell viability determined by CCK-8 assays (Fig. 5b). Similar results were obtained with the EdU assays, as the percentage of EdU positive cells was significantly decreased when compared with the control group (Fig. 5c). Similarly, colony formation assays revealed that hsa_circ_0058124 knockdown greatly attenuated the numbers of visible colonies (Fig. 5d). Transwell migration and invasion assays demonstrated that hsa_circ_0058124 knockdown impaired the migration and invasion ability (Fig. 6a), and the wound-healing assays showed that hsa_circ_0058124 silencing significantly suppressed the migration (Fig. 6b). Further flow cytometry investigations of the effects of hsa_circ_0058124 on cell cycle and apoptosis revealed the percentages of apoptotic cells significantly increased after hsa_circ_0058124 knockdown in TPC-1 cells (Fig. 6c), while the proportions of cells in each cell cycle phase were not significantly different between the si-hsa_circ_0058124 group and the control group (Fig. 6d).

We investigated the potential functional signaling pathways related to hsa_circ_0058124 in PTC using unbiased transcriptome profiling by RNA sequencing in TPC-1 cells transfected with si-hsa_circ_0058124. The depletion of hsa_circ_0058124 affected the expression levels of 457 genes: 239 genes were upregulated and 218 genes were downregulated (q-value < 0.05, |fold change| > 2, Fig. 7a). The KEGG pathway enrichment analysis showed that hsa_circ_0058124 affected several downstream signaling pathways involved in cell proliferation and metastasis, including the Notch, TNF, and FoxO signaling pathways (Fig. 7b). Since the Notch signaling pathway ranked first above all signaling pathways and plays an essential role in cell migration and invasion [30], we investigated this signaling pathway as a potential functional downstream pathway for hsa_circ_0058124. GSEA analysis showed that the gene set of KEGG_NOTCH_SIGNALING_PATHWAY was significantly enriched in si-hsa_circ_0058124 group compared to si-NC group, indicating that this gene set could be significantly associated with hsa_circ_0058124 expression in PTC (Fig. 7c-d). We further validated the related genes in the NOTCH signaling pathways and found that the NOTCH3/HES1 axis showed the most significant differential expression in the pathway [NOTCH4 and MAML3 were excluded because of too low expression (CT value>30)] (Fig. 7e). We also verified the NOTCH3 and HES1 protein levels by western blotting and confirmed that NOTCH3 pathway dramatically negatively related to hsa_circ_0058124 (Fig. 7f). Taken together, these findings demonstrated that hsa_circ_0058124 is an oncogenic driver that function by down-regulating the NOTCH3 signaling pathway in PTC.

Given that circRNAs have been reported to function as miRNA sponges, we speculated that hsa_circ_0058124 might regulate the NOTCH3 pathway by sponging miRNAs. Western blottings have indicated that the NOTCH3 pathway is significantly downregulated by hsa_circ_0058124, and NUMB was reported to be a strong suppressor of the NOTCH pathway (Fig. 8a) [32, 33], we hypothesized that hsa_circ_0058124 may upregulate NUMB by sponging miRNA, which resulted in the NOTCH3 pathway suppression. Two powerful web tools, CircInteractome [34] and starBase [35], were employed to predict the potential miRNA binding partners of hsa_circ_0058124 and NUMB. As shown in Fig. 8b, we further performed Venn diagram to intersect the predicted target miRNAs and narrowed the number of miRNAs down to 7 (miR-665, miR-513a-5p, miR-133a-3p, miR-3064-5p, miR-218-5p, miR-4784, miR-133b). Dual luciferase assays were conducted to validate whether these miRNAs could directly target hsa_circ_0058124 and NUMB. Luciferase reporters were cotransfected with each miRNA mimic into 293 T cells. We found that 2 miRNAs (miR-665, miR-218-5p) significantly inhibited the luciferase reporter activity (Fig. 8c), suggesting that hsa_circ_0058124 could bind to these miRNAs. The pmiR-RB-Report vector containing 3′-UTR regions of NUMB cotransfected with miR-218-5p mimic resulted in a remarkable decrease in the Renilla/firefly ratio compared with the control vector, but not in the NUMB 3′-UTR vector cotrasfected with miR-665 (Fig. 8d), suggesting that miR-218-5p could directly bind to NUMB. Concordantly, the protein level of NUMB was significantly reduced in cells transfected with miR-218-5p mimic when compared with control cells (Fig. 8e). The results above indicated that miR-218-5p could bind both hsa_circ_0058124 and NUMB directly, and modulated NUMB expression. To further verify the miR-218-5p binding sites in hsa_circ_0058124 and NUMB, mutant (mut) and wild-type (wt) luciferase reporter constructs were transfected into TPC-1 cells together with miR-218-5p mimics or controls (miR-NC). miR-218-5p mimics significantly inhibited luciferase activity in cells transfected with wild-type constructs, but luciferase activity was not affected in cells transfected with mutant constructs (Fig. 8f, g), indicating that miR-218-5p directly targets hsa_circ_0058124 and NUMB. Then, we determined the expression of miR-218-5p and NUMB by qRT-PCR in 20 pairs of PTC and matched normal tissues. We observed that miR-218-5p expression was downregulated in PTC tissues compared with normal tissues, while the expression of NUMB was significantly upregulated in PTC tissues (Additional file 2: Figure S3). There was also a negative correlation of hsa_circ_0058124 and miR-218-5p, similar to that of miR-218-5p and NUMB (Additional file 2: Figure S4). Moreover, miR-218-5p expression was significantly upregulated with hsa_circ_0058124 knockdown (Fig. 8h), and the expression of NUMB at both the mRNA (Fig. 8i) and protein level (Fig. 8j) were significantly decreased after hsa_circ_0058124 silencing in TPC-1 and K1 cells. As shown in Fig. 8k, we demonstrated that the NOTCH3 pathway was regulated by hsa_circ_0058124/miR-218-5p/NUMB axis. NOTCH3 pathway was significantly activated by hsa_circ_0058124 suppression, while si-hsa_circ_0058124 induced NOTCH3 pathway activation was attenuated with miR-218-5p inhibition or NUMB overexpression. These results indicated that hsa_circ_0058124 suppress NOTCH3 pathway via miR-218-5p/NUMB. Furthermore, miR-218-5p inhibition or NUMB overexpression significantly attenuated the si-hsa_circ_0058124-induced inhibition of growth (Fig. 8l), migration and invasion (Fig. 8m). Taken together, these results suggest that hsa_circ_0058124 promotes PTC progression through miR-218-5p/NUMB axis to suppress NOTCH3 pathway.

The GATA family is known to interact with the NOTCH signaling pathway [36‐38]. Our mRNA-sequence data also showed GATAD2A was significantly dysregulated in si-hsa_circ_0058124 group. Therefore, we questioned whether hsa_circ_0058124 may promote PTC progression through down-regulating NOTCH3/GATAD2A axis. Consistently, hsa_circ_0058124 silencing significantly increased the expression of Notch3 and GATAD2A at the protein level. In addition, blocking the NOTCH3 pathway with si-NOTCH3 also downregulated GATAD2A, even with suppressed hsa_circ_0058124, while suppression of GATAD2A did not significantly affect the NOTCH3 expression in TPC-1 cells (Fig. 9a). These findings indicated that the suppression of GATAD2A by hsa_circ_0058124 depends on NOTCH3.

We confirmed whether the oncogenic effects of hsa_circ_0058124 knockdown were mediated by NOTCH3/GATAD2A axis in PTC cells by cotransfecting si-hsa_circ_0058124 and si-NOTCH3 or si-GATAD2A into two PTC cell lines. Subsequent functional experiments showed that the suppression of NOTCH3 or GATAD2A could partly impair the si-hsa_circ_0058124-induced inhibition of growth (Fig. 9b-c), migration, and invasion (Fig. 9d). Thus, suppression of NOTCH3 or GATAD2A could restore the malignant behavior of hsa_circ_0058124 knockdown. However, the results also indicated that suppression of NOTCH3 or GATAD2A only partly impaired the si-hsa_circ_0058124-induced inhibition of the malignant behavior in PTC cells, which suggested that NOTCH3/GATAD2A axis might just be one of the major downstream target genes regulated by hsa_circ_0058124.

To investigate the GATAD2A expression levels in PTC tissues, we collected 5 PTC tissues with paired adjacent normal tissues, and measured the expression levels of GATAD2A by IHC assays. The results showed that GATAD2A expression was much lower in the PTC tissues compared with adjacent normal tissues (Fig. 9e). We also validated the differential expression of GATAD2A in PTC by examining IHC staining data of thyroid tumor tissues and normal thyroid tissues from the Human Protein Atlas. The fraction of samples with high, medium, low, or not detected levels of protein expression were provided by the blue-scale color-coding. As shown in Fig. 9f, all 4 thyroid tumor samples showed medium expression of GATAD2A, while all 3 normal samples showed high expression. Collectively, these data indicated that the underlying mechanism by which hsa_circ_0058124 promotes PTC progression is associated with a suppression of the NOTCH3/GATAD2A axis.

To further explore the effects of hsa_circ_0058124/NOTCH3/GATAD2A axis on PTC cells in vivo, we injected nude mice (aged 5–6 weeks, 5 in each group) with TPC-1 cells transfected with indicated shRNAs (3 × 10⁶ cells) subcutaneously for the control group, the hsa_circ_0058124 stably knocked down group, the NOTCH3 stably knocked down group, the GATAD2A stably silencing group, the stable knock down hsa_circ_0058124 with NOTCH3 inhibition group or the stable knock down hsa_circ_0058124 with suppressing GATAD2A group. Increases in tumor sizes were examined and recorded in each group every 3 days. At 27 days after TPC-1 cell injection, the mice were humanely euthanized, and the tumor volume and weight of the dissected tumors were measured. As demonstrated in Fig. 10a-d, the in vivo tumor formation assay suggested that hsa_circ_0058124 knockdown dramatically inhibited tumor growth when compared to the negative control group, whereas suppression of NOTCH3 or GATAD2A partly abolished this reduction of tumor growth by hsa_circ_0058124 knockdown. Taken together, these findings demonstrated that hsa_circ_0058124 is an oncogenic driver that functions by suppressing the NOTCH3/GATAD2A axis in PTC in vivo.