Clinical characteristics of participants
The participants were assigned to two groups according to their hemoglobin and hematological characteristics. The “high-HbF-β-thalassemia carriers” group (F) included 20 participants (HbF > 5.0%, HbA2 < 3.5%) while the control group (C) included 20 healthy individuals. The mean HbF and HbA2 values are listed in Table 1. These differed significantly between the F and C groups.
CircRNA profiles determined by microarray
To identify differentially expressed circRNAs, 10958 circRNAs were detected in three pairs of healthy subjects and high-HbF-β-thalassemia carriers by circRNA microarray. The fold changes (FC) in circRNA expression were determined, showing the presence of 2183 circRNAs with fold changes ≥ 1.5 and significantly different expression (P < 0.05) between the F and C groups. Of these, 1209 circRNAs were up-regulated and 974 were down-regulated (Figure 1A-C). GO and KEGG analyses were performed (Supplementary Figure S1).
Identification of HbF-related circRNAs in β-thalassemia
Pearson’s correlation analysis was used to evaluate the association between circRNA expression and hematological parameters. Positive circRNAs were defined as those whose expression showed significant association with hematological parameters, showing significant association HbF, HbA2, Hb, MCV, MCH, MCHC and RDWCV (|Pearson R| > 0.8 and P < 0.05, data not shown). A circRNA whose expression value correlated with HbF (|Pearson R |> 0.9 and P < 0.01) was defined as a significant HbF-related circRNA. The top 10 positive HbF-correlated and negative HbF-correlated circRNAs are shown in the heatmap (Figure 2). The down-regulated circRNA, has-circRNA-100466, showed a strong negative correlation with both HbF and HbA2 (P < 0.001) and was thus analyzed further.
Construction of the circRNA-associated ceRNA network
To explore the role of has-circRNA-100466 in β-thalassemia, we established a ceRNA regulatory network centered on has-circRNA-100466. This was based on the premise that circRNAs sponge miRNAs to modulate mRNA activity.
The ceRNA network identified 7 miRNA nodes and 28 mRNA nodes (Figure 3A). We further analyzed the hub genes and miRNAs using the “Cytohubba” plugin. This led to the final identification of five hub miRNAs (miR-19b-3p, miR-19a-3p, miR-130b-3p, miR-618, and miR-30c-1-3p) and four hub genes were identified, including SOX6, PARVA, SPIRE1, and MED13L. As the sub-ceRNA network showed, has-circRNA-100466 modulates the expression of the hub genes through binding to (miR-19b-3p, miR-19a-3p, miR-130b-3p, miR-618, and miR-30c-1-3p (Figure 3B). These predicted interactions offer insight into HbF induction.
Verification of the ceRNA network
Firstly, we used dataset GSE93973 from our previous microarray studies in which aberrantly expressed miRNAs were identified in β-thalassemia patients with high HbF compared to healthy controls. The five hub miRNAs observed in the present study were also detected in GSE93973, but only miR-19b-3p met the differential expression criteria of P < 0.05 and was identified as a differentially expressed miRNA. Meanwhile, the hub gene SOX6 encodes a transcription factor that plays an important role in erythropoiesis. Therefore, the has-circRNA-100466▁miR-19b-3p▁SOX6 pathway was selected for further verification.
Next, qRT-PCR was used to verify the differentially expressed has-circRNA-100466, miR-19b-3p, and SOX6 in the F and C groups. The results were consistent with the microarray data, showing down-regulation of has-circRNA-100466 and SOX6 and up-regulation of miR-19b-3p in the F group (Figure 4).
Numerous investigations of the interactions between miRNAs and mRNAs have suggested a pivotal role for Argonaute2 (AGO2), which has been shown to form complexes with miRNAs and thereby modulate their actions12. To investigate this, RIP on K562 cell extracts was performed using an anti-Ago2 antibody. This showed that both has-circRNA-100466 and miR-19b-3p were specifically associated with AGO2, with no enrichment seen with the control IgG, indicating that miR-19b-3p targets has-circRNA-100466 (Figure 5A-B). Similarly, co-precipitation of SOX6 mRNA and miR-19b-3p with AGO2 was also seen (Figure 5C), suggesting that miR-19b-3p interacts and may sponge SOX6.
Association between has-circRNA-100466/miR-19b-3p/SOX6 and HbF levels
Spearman correlations were used to examine the relationships between the qRT-PCR-determined levels of has-circRNA-100466, miR-19b-3p, and SOX6 and HbF. While significantly positive correlations were seen between miR-19b-3p and HbF (r = 0.783, P < 0.001), significantly negative correlations were seen between has-circRNA-100466 and HbF (r = 0.709, P < 0.001), and SOX6 and HbF (r =0.710, P < 0.001) (Figure 6A-6C).