Construction of circRNA-Mediated Immune-Related ceRNA Network and Identification of Circulating circRNAs as Diagnostic Biomarkers in Acute Ischemic Stroke

Figures & data

Table 1 Characteristics of the Expression Profiles in This Study

Dataset	Platform	Type	Sample Size		Author	Publication	Year
			Stroke	Control
GSE16561	GPL6883	Gene	39	24	Barr et al^Citation18	Neurology	2010
GSE110993	GPL15456	miRNA	20	20	Tiedt et al^Citation19	Circ Res	2017
GSE133768	GPL21825	circRNA	3	3	Zuo et al^Citation16	Stroke	2020

Figure 1 Flow diagram for the identification of immune-related circRNA biomarkers in AIS. (A) DECs, DEMs, DEIRGS were obtained from three different GEO datasets and two publicly accessible databases, and were analyzed by “limma” R package. The ceRNA pairs were predicted using Circbank and Mirwalk online databases, then constructed the ceRNA network. Identification the hub genes of the ceRNA network by centiscape algorithm and were further constructed the sub ceRNA network. The six highly expressed circRNAs were chosen as candidates for further investigation. (B) These six candidate circRNAs analyzed by qPCR and ROC curves in validation sample for functional validation. Finally, two circRNAs biomarkers were used to test the predictive accuracy in replication sample.

Figure 2 Screening and identification of DEGs, DEMs, DECs and DEIRGs in AIS patients and HCs. The volcano plot of DEGs (A), DEMs (C), DECs (E) in GSE16561, GSE110993, and GSE133768, respectively. Red dots represent upregulation while blue dots represent downregulation. The hierarchical cluster heatmaps of DEGs (B), DEMs (D), DECs (F) in GSE16561, GSE110993, and GSE133768, respectively. Orange red represents high expression while blue represents low expression. (G) Venn diagram of DEGs and IRGs. Purple and blue circles represent DEGs and IRGs, respectively, and their overlap region represents DEIRGs.

Figure 3 Functional enrichment analysis for DEIRGs. (A) Top 15 significantly enriched GO terms for biological process (BP), cellular component (CC), and molecular function (MF). The color scale from blue to red indicates the gradual increase of the level of enrichment. (B) KEGG pathway enrichment analyses of DEIRGs. The color scale from blue to red indicates the gradual increase of the level of enrichment. Each circle indicates a specific KEGG pathway terms and circle size represents the number of genes that have been enriched.

Figure 4 Construction of the dysregulated AISIRC network in AIS. Red and green triangles represent upregulated circRNAs and downregulated circRNAs, respectively. Orange and blue rectangles represent upregulated miRNAs and downregulated miRNAs, respectively. Yellow and purple ovals represent upregulated genes and downregulated genes, respectively. Gray lines represent the interactions between the parts of AISIRC network.

Table 2 Five Hub Genes Identified by Centiscape 2.2

Gene Name	Degree	Betweenness	Closeness	Gene Descriptions
FCGR3A	9	112.70	0.030	Low-affinity immunoglobulin gamma Fc region receptor III-A
FCGR3B	8	81.30	0.029	Low-affinity immunoglobulin gamma Fc region receptor III-B
IL7R	8	116.57	0.029	Interleukin-7 receptor subunit alpha
IL2RB	7	54.47	0.027	High-affinity IL-2 receptor subunit beta
MMP9	5	61.07	0.023	Matrix metalloproteinase-9

Figure 5 Construction of the PPI network and the circRNA–miRNA–hub gene subnetwork. (A) PPI network of DEIRGs in the AISIRC network. Node size is proportional to their degree in the PPI network, and edge thickness is proportional to the combined score between DEIRGs. The five important hub genes (dark orange nodes) were identified from the PPI network using the Centiscape algorithm. (B) Subnetworks of these five hub genes extracted from the AISIRC network. Red and green triangles represent upregulated circRNAs and downregulated circRNAs, respectively. Orange and blue rectangles represent upregulated miRNAs and downregulated miRNAs, respectively. Yellow and purple ovals represent upregulated genes and downregulated genes, respectively. Gray lines represent the interactions between the parts of AISIRC network.

Table 3 Baseline Characteristics of the Validation and Replication Sample

Baseline Characteristics	Validation Sample			Replication Sample
	AIS	HCs	P-value^a	AIS	TIA	HCs	P-value^a	P-value^b	P-value^c
Demographics
Total, n	36	36	-	200	66	100	-	-	-
Age, mean (SD), years	61.8(12.6)	65.8(11.9)	0.177*	64.8(10.7)	65.0(11.0)	64.9(11.3)	0.999^§	0.989^§	0.991^§
Female, n (%)	16(44.4)	15(41.7)	0.812^†	67(33.5)	27(40.9)	43(43)	0.107^†	0.789^†	0.275^†
Vascular risk factors, n(%)
Hypertension	21(58.3)	16(44.4)	0.238^†	134(67)	40(60.6)	43(43)	<0.001^†	0.026^†	0.344^†
Diabetes mellitus	9(25)	6(16.7)	0.384^†	58(29)	17(25.8)	15(15)	0.008^†	0.086^†	0.612^†
Dyslipidemia	11(30.6)	8(22.2)	0.422^†	47(23.5)	24(36.4)	24(24)	0.923^†	0.086^†	0.041^†
Atrial fibrillation	4(11.1)	1(2.8)	0.164^†	34(17)	1(1.5)	1(1)	<0.001^†	0.766^†	0.001^†
Coronary artery disease	3(8.3)	4(11.1)	0.691^†	32(16)	14(21.2)	11(11)	0.244^†	0.072^†	0.332^†
Smoking history	9(25)	5(13.9)	0.234^†	40(20)	14(21.2)	13(13)	0.134^†	0.161^†	0.832^†
Alcohol	2(5.6)	1(2.8)	0.555^†	13(6.5)	4(6.1)	3(3)	0.203^†	0.337^†	0.899^†
Biochemical indexes on admission, mean (SD)/median (interquartile range)
Serum glucose, mmol/L	5.82(2.49)	5.23(1.15)	0.024^‡	5.96(2.13)	5.27(1.39)	5.36(1.17)	<0.001^§	0.980^§	<0.001^§
Triglycerides, mmol/L	1.27(0.71)	1.27(0.78)	0.434^‡	1.26(0.85)	1.29(0.93)	1.18(0.74)	0.210^§	0.335^§	0.559^§
Total cholesterol, mmol/L	4.32(1.10)	4.40(0.93)	0.736*	4.33(1.04)	4.11(0.97)	4.13(0.98)	0.099^∇	0.938^∇	0.131^∇
HDL, mmol/L	1.04(0.38)	1.29(0.64)	0.161^‡	1.04(0.30)	1.04(0.37)	1.05(0.43)	0.210^§	0.335^§	0.559^§
LDL, mmol/L	2.76(0.74)	2.58(0.79)	0.329*	2.66(1.17)	2.54(1.16)	2.38(1.07)	0.008^§	0.502^§	0.365^§

Notes: ^aAIS patients vs HCs; ^bTIA patients vs HCs; ^cAIS patients vs TIA patients; *Student’s t-test; ^†Chi-squared test; ^‡Mann–Whitney U-test; ^§Kruskal–Wallis ANOVA test; ^∇one-way ANOVA test.

Abbreviations: AIS, acute ischemic stroke; HCs, healthy control subjects; HDL, high-density lipoprotein; LDL, low-density lipoprotein; SD, standard deviation; TIA, transient ischemic attack.

Figure 6 Relative levels of the five candidate circRNAs via qRT-PCR in the validation sample (n=36/36, healthy control [HC]/acute ischemic stroke [AIS]). Two circRNAs were significantly differentially expressed: (A) hsa_circ_0007364 and (E) hsa_circ_0045128. Three circRNAs were not significantly differentially expressed: (B) hsa_circ_0000479, (C) hsa_circ_0047460, and (D) hsa_circ_0017252. Median ± interquartile range, Mann–Whitney U-test.

Figure 7 Relative levels of (A) circPTP4A2 and (B) circTLK2 via qRT-PCR in the replication sample (n=100/66/200, healthy control [HC]/transient ischemic attack [TIA]/acute ischemic stroke [AIS]). Median ± interquartile range, Kruskal–Wallis ANOVA test followed by Bonferroni correction pairwise comparisons. *P < 0.05; **P < 0.01; ***P < 0.001.

Abbreviation: ns, not significant.

Table 4 Logistic Regression Analysis for Predictive Power of circPTP4A2 and circTLK2

Variable	Univariate Analysis^a		Multivariate Analysis^b
	OR (95% CI)	P-value	OR (95% CI)	P-value
AIS^c
circPTP4A2	3.204 (2.183–4.705)	<0.001	2.788(1.876–4.145)	<0.001
circTLK2	2.576 (1.768–3.754)	<0.001	2.480(1.676–3.670)	<0.001
TIA^c
circPTP4A2	1.782 (1.142–2.780)	0.011	1.628(1.036–2.560)	0.035
circTLK2	1.877 (1.230–2.863)	0.003	1.736(1.141–2.639)	0.010
AIS^d
circPTP4A2	1.799 (1.291–2.505)	0.001	1.713(1.220–2.405)	0.002
circTLK2	1.373 (1.045–1.804)	0.023	1.429(1.078–1.894)	0.013

Notes: ^aOnly one of the candidate circRNAs was included in the model. ^bThe age, sex, vascular risk factors and biochemical indexes were adjusted in the model. ^cThe reference category was the HCs group. ^dThe reference category was the TIA group.

Abbreviations: AIS, acute ischemic stroke; CI, confidence interval; HCs, healthy control subjects; OR, odds ratio; TIA, transient ischemic attack.

Table 5 Diagnostic Property of Individual circRNAs and the Combined circRNAs

Variables	AUC (95% CI)	P-value	Sensitivity	Specificity
AIS patients vs HCs
circPTP4A2	0.762 (0.703–0.820)	<0.001	0.695	0.820
circTLK2	0.734 (0.675–0.793)	<0.001	0.665	0.730
The combined circRNAs	0.805 (0.751–0.860)	<0.001	0.900	0.630
TIA patients vs HCs
circPTP4A2	0.708 (0.628–0.787)	<0.001	0.742	0.680
circTLK2	0.564 (0.470–0.658)	0.164	0.409	0.810
The combined circRNAs	0.664 (0.580–0.748)	<0.001	0.580	0.748
AIS patients vs TIA patients
circPTP4A2	0.704 (0.637–0.770)	<0.001	0.560	0.879
circTLK2	0.632 (0.546–0.719)	0.001	0.905	0.379
The combined circRNAs	0.707 (0.640–0.773)	<0.001	0.570	0.879

Abbreviations: AIS, acute ischemic stroke; AUC, area under the curve; CI, confidence interval; HCs, healthy control subjects; TIA, transient ischemic attack.

Figure 8 Diagnostic property of individual circRNAs and the combined circRNAs. Receiver operating characteristics (ROC) curves were calculated using the baseline levels of circRNAs based on the replication sample and their corresponding AUC values for discriminating AIS patients from HCs (A), for discriminating AIS patients from TIA patients (B), and for discriminating TIA patients from HCs (C).

Figure 9 Temporal expression profile of validated circRNAs after AIS. The results for 18 patients with available data for five time points after AIS: circPTP4A2 (A) and circTLK2 (B). Median ± interquartile range, Kruskal–Wallis ANOVA test followed by Bonferroni correction pairwise comparisons. *P<0.05 compared with HC; ^#P<0.05 compared with day 1.

Barr TL, Conley Y, Ding J, et al. Genomic biomarkers and cellular pathways of ischemic stroke by RNA gene expression profiling. Neurology. 2010;75(11):1009–1014. doi:10.1212/WNL.0b013e3181f2b37f

 PubMed Web of Science ®Google Scholar

Tiedt S, Prestel M, Malik R, et al. RNA-Seq identifies circulating miR-125a-5p, miR-125b-5p, and miR-143-3p as potential biomarkers for acute ischemic stroke. Circ Res. 2017;121(8):970–980. doi:10.1161/CIRCRESAHA.117.311572

 PubMed Web of Science ®Google Scholar

Zuo L, Zhang L, Zu J, et al. Circulating circular RNAs as biomarkers for the diagnosis and prediction of outcomes in acute ischemic stroke. Stroke. 2020;51(1):319–323. doi:10.1161/STROKEAHA.119.027348

 PubMed Web of Science ®Google Scholar