Placental circadian lincRNAs and spontaneous preterm birth

Abstract

Long non-coding RNAs (lncRNAs) have a much higher cell- and/or tissue-specificity compared to mRNAs in most cases, making them excellent candidates for therapeutic applications to reduce off-target effects. Placental long non-coding RNAs have been investigated in the pathogenesis of preeclampsia (often causing preterm birth (PTB)), but less is known about their role in preterm birth. Preterm birth occurs in 11% of pregnancies and is the most common cause of death among infants in the world. We recently identified that genes that drive circadian rhythms in cells, termed molecular clock genes, are deregulated in maternal blood of women with spontaneous PTB (sPTB) and in the placenta of women with preeclampsia. Next, we focused on circadian genes-correlated long intergenic non-coding RNAs (lincRNAs, making up most of the long non-coding RNAs), designated as circadian lincRNAs, associated with sPTB. We compared the co-altered circadian transcripts-correlated lincRNAs expressed in placentas of sPTB and term births using two published independent RNAseq datasets (GSE73712 and GSE174415). Nine core clock genes were up- or downregulated in sPTB versus term birth, where the RORA transcript was the only gene downregulated in sPTB across both independent datasets. We found that five circadian lincRNAs (LINC00893, LINC00265, LINC01089, LINC00482, and LINC00649) were decreased in sPTB vs term births across both datasets (p ≤ .0222, FDR≤.1973) and were negatively correlated with the dataset-specific clock genes-based risk scores (correlation coefficient r = -.65 ∼ -.43, p ≤ .0365, FDR≤.0601). Gene set variation analysis revealed that 65 pathways were significantly enriched by these same five differentially expressed lincRNAs, of which over 85% of the pathways could be linked to immune/inflammation/oxidative stress and cell cycle/apoptosis/autophagy/cellular senescence. These findings may improve our understanding of the pathogenesis of spontaneous preterm birth and provide novel insights into the development of potentially more effective and specific therapeutic targets against sPTB.

Keywords: GSVA analysis; lincRNA; molecular clock; preterm (birth); transcriptional co-alteration.

FIGURE 1

Analytic pipeline to identify placental circadian lincRNAs and pathways in sPTB vs term birth.

FIGURE 2

Violin plots of five differentially expressed core clock genes in sPTB vs Term in two datasets. Violin plots for (A) GSE73712, and (B) GSE174415. The red loosely dashed line in the center of each violin box represents the mean and the white dotted lines in the center represent the confidence limits for the population mean. All p-values on the top of the bars were generated by the moderated t-test in limma R package.

FIGURE 3

Violin plots of the expressions of five common circadian DE lincRNAs in sPTB vs Term in two datasets. Violin plots for (A) GSE73712, and (B) GSE174415. The red loosely dashed line in the center of each violin box represents the mean and the white dotted lines in the center represent the confidence limits for the population mean. All p-values on the top of the bars were generated by the moderated t-test in limma R package.

FIGURE 4

Visualization of the correlations between clock genes-based risk score and five common circadian DE lincRNAs in two datasets. Linear correlation between the clock genes-based risk score with each of the 5 DE lincRNA for (A) GSE73712 dataset, (B) GSE174415. The dots represent raw data points, and the lines represent linear regression lines for the relationships of clock genes-based risk score with each of five common circadian DE lincRNAs. The clock genes-based risk score = [(6-NPAS2) + (3-NR1D1) + (8-NR1D2) + (5-PER3) + (7-RORA)] for GSE73712 and [ARNTL + (6-CRY1) + (2-NPAS3) + (6-PER2) + (7-RORA)] for GSE174415, respectively.

FIGURE 5

Violin plots of top 5 decreased or increased pathway scores in sPTB vs Term in two datasets. Violin plots of the top 5 decreased pathways for (A) GSE73712, and (B) GSE174415, and the top five increased pathways for (D) GSE73712, and (E) GSE174415. The names of the predicted (C) down and (F) upregulated pathways identified in (A), (B) and (D), (E), respectively. The red loosely dashed line in the center of each violin box represents the mean and the white dotted lines in the center represent the confidence limits for the population mean. The pathway scores were generated by the GSVA R package. All p-values on the top of the bars were generated by the moderated t-test in limma R package (Tables S2 and S3).

FIGURE 6

Visualization of the correlations between five circadian lincRNAs-based risk score and top 5 decreased or increased pathway scores in two datasets. Linear correlation between the five circadian lincRNAs-based risk score and each of top 5 decreased pathway scores for (A) GSE73712 dataset, (B) GSE174415. Linear correlation between the five circadian lincRNAs-based risk score and each of top five increased pathways for (C) GSE73712 dataset, (D) GSE174415. The dots represent raw data points, and the lines represent linear regression lines for the relationships of five lincRNAs-based risk score with each of top five pathways. The five lincRNAs-based risk score = [(5-LINC01089) + (6-LINC00482) + (4-LINC00893) + (5-LINC00649) + (4-LINC00265)] for GSE73712, and [(4-LINC01089) + (5-LINC00482) + (5-LINC00893) + (5-LINC00649) + (5-LINC00265)] for GSE174415, respectively. The pathway scores were generated by the GSVA R package. The names of the identified up and downregulated pathways are indicated in Figures 5C, F, respectively.