Work overview

Section 03 of 04

Results

Transcriptomic landscape of human circular RNAs: unveiling molecular mechanisms in high-altitude adaptation

Buhe Bao, Mingxuan Yu, Wei Pang, Ruilin Wang, Wenbin Dong, Ying Bai, Rui Shi, and Renjie Wang · 2026

Contents

Section 03 of 04

  1. 01Introduction
  2. 02Materials and methods
  3. 03Results
  4. 04Discussion
Text size
Work overview

Section 3 of 4

Results

Buhe Bao, Mingxuan Yu, Wei Pang, Ruilin Wang, Wenbin Dong, Ying Bai, Rui Shi, and Renjie Wang · about 12 minutes

Original source ↗

Data source and research strategy

To investigate altitude-associated differential circRNA expression, we selected three population cohorts located at varying altitudes: Tianjin (sea level, 0 m), Ningxia Yinchuan (1,000 m), and Xizang Linzhi (3,000 m) (Figure 1A). The comprehensive study workflow is illustrated in Figure 1B and comprises four main stages: sample processing, expression profiling, bioinformatic analysis, and experimental validation. Sample processing involved rigorous preparation, total RNA extraction, and quality control (QC). Expression profiling included data acquisition, normalization, and identification of differentially expressed circRNAs. Bioinformatic analysis encompassed GO and KEGG pathway enrichment analyses and circRNA–miRNA regulatory network construction. Finally, specific candidate circRNAs were validated by qRT-PCR to further elucidate the circRNA–miRNA–mRNA (ceRNA) regulatory axis.

FIGURE 1: Panel A shows a map of China highlighting three cohorts at different altitudes: Xizang Linzhi cohort at three thousand meters, Ningxia Yinchuan cohort at one thousand meters, and Tianjin cohort at zero meters, illustrated with blue human figures and color-coded regions. Panel B presents a flowchart outlining four steps: sample processing, expression profile, bioinformatics analysis, and qRT-PCR verification, each listing associated tasks such as circRNA analysis and enrichment analysis.

FIGURE 1: Experimental design and the research strategies. (A) Three population groups from different altitudes were included in this retrospective study: Tianjin cohort (sea level, 0 m), Ningxia Yinchuan cohort (1,000 m), and Xizang Linzhi cohort (3,000 m). All samples were de-identified residual health examination specimens. (B) Research workflow of this study. The workflow comprised four main stages: sample processing, expression profiling, bioinformatic analysis, and qRT-PCR validation. Sample processing involved sample preparation, total RNA extraction, and quality control. Expression profiling included data acquisition, normalization, comparative analysis of common RNAs, and identification of differentially expressed circRNAs. Bioinformatic analysis encompassed GO enrichment, KEGG pathway enrichment, and circRNA–miRNA regulatory analysis. qRT-PCR validation involved the quantification of candidate circRNAs and the construction of the circRNA–miRNA–mRNA (ceRNA) regulatory network.

Copy number variation analysis

We first analyzed the copy number variations (CNVs) of the parent genes of the identified circRNAs (Figure 2). To examine CNVs in these regions, we calculated the frequencies of copy number loss and gain across the genome. The outer layer represents the chromosomal structure, while the inner layer displays the genome-wide distribution of CNVs. Gene expression profiles were normalized and compared against consistently expressed transcripts. Genes with identified CNVs are depicted in the innermost layer, where black and red dots represent copy number losses and gains, respectively. Large-scale copy number losses were observed in specific regions, such as chromosomes 5, 15, and 18, whereas a combination of both copy number losses and gains was present in other chromosomes. No CNVs were detected on the Y chromosome.

FIGURE 2: Circular genomic ideogram showing human chromosomes labeled one to twenty-two, X, and Y, arranged sequentially around a ring. Black and colored bands represent chromosomal regions, while gene names are annotated along inner arcs with connecting lines. Red, black, and gray marks indicate specific chromosomal loci, and outer colored segments separate chromosomes.

FIGURE 2: The Circos plot shows the CNVs of the human genome. The outer layer of the Circos plot represents the chromosomal structure, while the inner layer displays the genome-wide distribution of CNVs. Gene expression profiles were normalized and compared against consistently expressed transcripts. Genes harboring CNVs are illustrated in the innermost layer, where black and red dots represent copy number loss and gain, respectively.

CircRNA expression profiles

To identify key circRNAs, we analyzed differential expression profiles obtained from the circRNA expression microarray. Scatter plots were employed to evaluate the overall expression variations in each pairwise comparison (Figure 3A), and differentially expressed circRNAs (DE-circRNAs) were identified using volcano plots (Figure 3B). A fold change (FC) > 2 and a p-value <0.05 were adopted as the thresholds for defining significant differential expression. The expression profiles of these DE-circRNAs were visualized through hierarchical clustering analysis (heatmap), where each row and column represents a probe set and a sample, respectively (Figure 3C). In the comparison between the 3,000 m and 1,000 m groups, 135 DE-circRNAs were identified (36 upregulated and 99 downregulated). Between the 1,000 m and sea level (0 m) groups, 20 DE-circRNAs were detected (18 upregulated and 2 downregulated). Finally, the comparison between the 3,000 m and sea level (0 m) groups yielded 41 DE-circRNAs (26 upregulated and 15 downregulated).

FIGURE 3: Panel A shows three MA plots comparing gene expression at different altitudes, with each plot displaying dots in gray, red, and blue to represent expression changes; x- and y-axes are labeled accordingly. Panel B presents three volcano plots for the same comparisons, with points colored to indicate significant changes, and axes indicating fold change and significance. Panel C displays three clustered heatmaps illustrating gene expression patterns for each altitude comparison, using a blue to red color scale to represent expression levels across multiple gene identifiers and sample groups.

FIGURE 3: CircRNA analysis identified differentially expressed circRNAs among the three population groups. (A) Scatter plots showing circRNA expression differences in the 3,000 m vs. 1,000 m; 1,000 m vs. sea level (0 m); and 3,000 m vs. sea level (0 m) comparisons. Red and blue points indicate significantly dysregulated circRNAs in each comparison. (B) Volcano plots showing circRNA expression differences in the 3,000 m vs. 1,000 m; 1,000 m vs. sea level (0 m); and 3,000 m vs. sea level (0 m) comparisons. CircRNAs with fold change (FC) > 2 and p < 0.05 are colored red (upregulated) or blue (downregulated). (C) Heatmap of circRNA expression profiles based on hierarchical clustering analysis. Red indicates higher expression, and blue indicates lower expression.

Potential biological functions and pathways

To explore the potential biological functions of the differentially expressed circRNAs, we performed GO and KEGG pathway enrichment analyses. Figure 4 displays the top 30 enriched GO terms for the parent genes of circRNAs identified in each pairwise population comparison. In the 3,000 m vs. 1,000 m comparison (Figure 4A), parent genes were primarily enriched in BPs related to translational initiation, signal recognition particle (SRP)-dependent cotranslational protein targeting to membrane, and mRNA catabolic processes. CC terms were mainly associated with ribosomes and cytosolic large ribosomal subunits, while MF terms were mainly associated with structural constituents of the ribosome and rRNA binding. For the 1,000 m vs. sea level (0 m) comparison (Figure 4B), enriched BP terms included translation, peptide metabolic processes, and peptide biosynthetic processes. CC and MF terms showed similar trends to the 3,000 m vs. 1,000 m comparison, with primary enrichment in ribosomal components and structural constituents of the ribosome. In the 3,000 m vs. sea level (0 m) comparison (Figure 4C), enriched BP terms shifted toward cellular protein catabolic processes and carbohydrate derivative metabolic processes. CC terms were largely enriched in extracellular matrix components, while MF terms included structural molecule activity, identical protein binding, and calcium ion binding.

FIGURE 4: Panel A, B, and C display bubble plots of the top thirty Gene Ontology enrichment terms for biological process, cellular component, and molecular function categories in different altitude comparisons, with enrich factor, p-value color scale, and gene count size shown. Panels D, E, and F present bar charts of Gene Ontology classifications for each comparison, showing the number of genes identified per classification, color-coded by GO domain.

FIGURE 4: GO analysis of the parent genes of differentially expressed circRNAs. (A) Top 30 enriched GO terms of parent genes of differentially expressed circRNAs in the 3,000 m vs. 1,000 m comparison. (B) Top 30 enriched GO terms of parent genes of differentially expressed circRNAs in the 1,000 m vs. sea level (0 m) comparison. (C) Top 30 enriched GO terms of parent genes of differentially expressed circRNAs in the 3,000 m vs. sea level (0 m) comparison. (D) GO classification of parent genes of differentially expressed circRNAs in the 3,000 m vs. 1,000 m comparison. (E) GO classification of parent genes of differentially expressed circRNAs in the 1,000 m vs. sea level (0 m) comparison. (F) GO classification of parent genes of differentially expressed circRNAs in the 3,000 m vs. sea level (0 m) comparison. GO analysis was categorized into three categories: biological processes, molecular functions, and cellular components. Dot size represents the number of enriched genes, and dot color represents the statistical significance of the corresponding GO category.

Furthermore, we performed GO functional classification analysis on the parent genes of the differentially expressed circRNAs for each pairwise population comparison (Figures 4D–F). In all three comparison groups (3,000 m vs. 1,000 m; 1,000 m vs. sea level; and 3,000 m vs. sea level), the parent genes exhibited consistent distribution patterns across GO categories. In the BP domain, the majority of genes were categorized under “cellular process,” “metabolic process,” and “regulation of biological process.” For CC terms, the genes were primarily associated with “cell,” “cell part,” and “organelle.” Within the MF domain, “binding” was the most common term across all comparisons, supplemented by “catalytic activity” and “structural molecule activity” in specific groups.

To further elucidate the functional roles of the identified circRNAs, we performed KEGG pathway enrichment analysis on their parent genes. The top 30 enriched pathways for each pairwise population comparison are summarized in Figure 5. In the 3,000 m vs. 1,000 m comparison (Figure 5A), the parent genes were primarily associated with the ribosome, other glycan degradation, and ECM–receptor interaction pathways. For the 1,000 m vs. sea level (0 m) comparison (Figure 5B), the ribosomal pathway was the most prominent enrichment. Finally, in the 3,000 m vs. sea level (0 m) comparison (Figure 5C), the parent genes were mainly enriched in pathways related to protein processing in the endoplasmic reticulum and ECM–receptor interaction.

FIGURE 5: Panel A shows a dot plot of the top thirty pathway enrichment results for three thousand meters versus one thousand meters, with gene count indicated by dot size and p-value by color gradient. Panel B displays a similar pathway enrichment dot plot for one thousand meters versus zero meters. Panel C presents the same analysis for three thousand meters versus zero meters. Panel D is a horizontal bar chart of KEGG classifications for three thousand meters versus one thousand meters, showing gene distribution among pathway classes. Panel E gives KEGG classification for one thousand meters versus zero meters in a bar chart, and panel F shows KEGG classification for three thousand meters versus zero meters in a similar format.

FIGURE 5: KEGG analysis of the parent genes of differentially expressed circRNAs. (A) Top 30 enriched KEGG pathways in the 3,000 m vs. 1,000 m comparison. (B) Top 30 enriched KEGG pathways in the 1,000 m vs. sea level (0 m) comparison. (C) Top 30 enriched KEGG pathways in the 3,000 m vs. sea level (0 m) comparison. (D) KEGG classification of parent genes of differentially expressed circRNAs in the 3,000 m vs. 1,000 m comparison. (E) KEGG classification of parent genes of differentially expressed circRNAs in the 1,000 m vs. sea level (0 m) comparison. (F) KEGG classification of parent genes of differentially expressed circRNAs in the 3,000 m vs. sea level (0 m) comparison. KEGG analysis was categorized into six categories: cellular processing, environmental information processing, genetic information processing, human diseases, metabolism, and organismal systems. Dot size represents the number of enriched genes, and dot color represents the statistical significance of the corresponding pathway.

To gain further insight into the functional categories of the parent genes, we performed KEGG pathway classification analysis for each pairwise population comparison. Figures 5D–F illustrate the distribution of these genes across various KEGG functional categories. In the 3,000 m vs. 1,000 m comparison (Figure 5D) and the 1,000 m vs. sea level (0 m) comparison (Figure 5E), the parent genes were primarily associated with the “translation” category. In the 3,000 m vs. sea level (0 m) comparison (Figure 5F), the parent genes were mainly classified under “translation,” “signaling molecules and interaction,” and “signal transduction” pathways.

circRNA analysis

We identified 135 differentially expressed circRNAs between the 3,000 m and 1,000 m groups, 130 of which were unique to this comparison. In the 1,000 m vs. sea level (0 m) comparison, 20 circRNAs were differentially expressed, with 15 being unique to this group. A total of five common circRNAs—hsa_circ_0044526, hsa_circ_0022498, hsa_circ_0044534, hsa_circ_0044520, and hsa_circ_0026102—were identified across both comparisons (Figure 6A). Further analysis provided detailed characteristics and predicted RNA-binding proteins (RBPs) for these five candidates (Table 1). Notably, AGO2 and EIF4A3 were identified as shared RBPs for all five circRNAs, suggesting a potential role for these candidates in post-transcriptional regulation.

FIGURE 6: Venn diagram illustrating differentially expressed genes for two elevation comparisons, with 130 genes unique to three thousand meters versus one thousand meters, 15 unique to one thousand meters versus zero meters, and 5 genes overlapping. Overlapping genes are listed as hsa_circ_0044526, hsa_circ_0022498, hsa_circ_0044534, hsa_circ_0044520, and hsa_circ_0026102.

FIGURE 6: Venn diagram of differentially expressed circRNAs among the three altitude groups. A total of 135 differentially expressed circRNAs were identified between the 3,000 m and 1,000 m groups, and 20 differentially expressed circRNAs were identified between the 1,000 m and sea level (0 m) groups. Five circRNAs were shared in both comparisons: hsa_circ_0044526, hsa_circ_0022498, hsa_circ_0044534, hsa_circ_0044520, and hsa_circ_0026102.

circRNA ID | Host gene | Genomic location | Spliced sequence length | Rbp (site count) | Database | Prediction threshold
hsa_circ_0044526 | COL1A1 | chr17:48,265,890–48,273,337 | 2,205 | AGO2 (1)EIF4A3 (20) | CircBank | Score ≥0.5
hsa_circ_0044534 | COL1A1 | chr17:48,266,528–48,273,337 | 1,935 | AGO2 (2)EF4A3 (22) | CircBank | Score ≥0.5
hsa_circ_0044520 | COL1A1 | chr17:48,264,375–48,273,337 | 2,529 | AGO2 (3)C170RF85 (1)EIF4A (14)IGF2BP3 (1) | CircBank | Score ≥0.5
hsa_circ_0022498 | UBXN1 | chr11:62,443,971–62,446,527 | 1,289 | AGO2 (1)EIF4A3 (2)HNRNPC (1) | CircBank | Score ≥0.5
hsa_circ_0026102 | MLL2 | chr12:49,435,871–49,448,534 | 5,933 | AGO1 (1)AGO2 (7)DGCR8 (2)EIF4A3 (1) | CircBank | Score ≥0.5

qRT-PCR validation

To further validate the differential expression of these five circRNAs, qRT-PCR was performed. As shown in Figure 7A, the relative expression levels of these five circRNAs were higher in the 1,000 m population group and highest in the 3,000 m population group. To further confirm the results in vitro, the relative expression levels of these circRNAs were quantified in human endothelial cells after transfection with the HIF-1α overexpression vector. As shown in Figure 7B, HIF-1α overexpression enhanced the relative expression levels of these five circRNAs in human endothelial cells. This result reveals that the five specific circRNAs contribute to HIF-1α-mediated high-altitude adaptation.

FIGURE 7: Panel A presents a scatter plot with error bars showing relative circRNA expression of five circRNAs at altitudes of zero, one thousand, and three thousand meters, with significant increases at higher altitudes indicated by asterisks. Panel B displays a bar graph comparing relative circRNA expression in human endothelial cells between a control group and HIF-1a treated group, showing HIF-1a significantly increases expression for all five circRNAs.

FIGURE 7: qRT-PCR verification of the microarray analysis results. (A) Relative expression levels of the five circRNAs were significantly upregulated with increasing altitude. Expression levels in the 1,000 m group were higher than those in the sea level (0 m) group, and expression levels in the 3,000 m group were higher than those in both the 0 m and 1,000 m groups. (B) Relative expression levels of the five circRNAs in HUVECs under control and HIF-1α overexpression conditions. HIF-1α overexpression significantly increased the expression levels of the five circRNAs.