Metformin Treatment Modulates Long Non-Coding RNA Isoforms Expression in Human Cells

Abstract

Long noncoding RNAs (lncRNAs) undergo splicing and have multiple transcribed isoforms. Nevertheless, for lncRNAs, as well as for mRNA, measurements of expression are routinely performed only at the gene level. Metformin is the first-line oral therapy for type 2 diabetes mellitus and other metabolic diseases. However, its mechanism of action remains not thoroughly explained. Transcriptomic analyses using metformin in different cell types reveal that only protein-coding genes are considered. We aimed to characterize lncRNA isoforms that were differentially affected by metformin treatment on multiple human cell types (three cancer, two non-cancer) and to provide insights into the lncRNA regulation by this drug. We selected six series to perform a differential expression (DE) isoform analysis. We also inferred the biological roles for lncRNA DE isoforms using in silico tools. We found the same isoform of an lncRNA (AC016831.6-205) highly expressed in all six metformin series, which has a second exon putatively coding for a peptide with relevance to the drug action. Moreover, the other two lncRNA isoforms (ZBED5-AS1-207 and AC125807.2-201) may also behave as cis-regulatory elements to the expression of transcripts in their vicinity. Our results strongly reinforce the importance of considering DE isoforms of lncRNA for understanding metformin mechanisms at the molecular level.

Keywords: RNA isoforms; lncRNA; metformin; transcriptome.

Figure 1

(A) Differential lncRNA isoforms counted per series using a cutoff of p-value < 0.05 and absolute log2foldchange > 0.5. Series are ordered according to descending metformin concentration. Number of upregulated lncRNA isoforms shown in dark red and number of downregulated lncRNA isoforms shown in dark blue. Upset plots of up- (B) and downregulated (C) lncRNA isoforms which intersect between series.

Figure 2

Heatmap of the 36 differentially expressed isoforms which are present in at least four of the six series. Heatmap was ordered according to Metformin concentration (lowest to highest, in green) and information on hours posttreatment (in purple) and cell type are present in the top. Color intensity represents log2FoldChange in the transcript in each series. Upregulated transcripts are in red, and downregulated transcripts are in blue. Panc-1: human pancreatic cell line isolated from pancreatic carcinoma. HeSC: human embryonic stem cells. 786-O: renal cell carcinoma. A549: human lung epithelial carcinoma cells. Primary hep: primary human hepatocytes.

Figure 3

(A) region around the AC016831.6-205 transcripts and its superposition with LINC-PINT-208. In A, region of some of the many annotated LINC-PINT and AC016831.6-205 transcripts; in B, approximation and superposition of the mentioned isoforms. In red, the AC016831.6 isoform encountered in our analysis; in green, other isoforms of the same gene; in orange, the LINC-PINT isoform associated with the one we found. (B) NEAT1 genomic region with its transcripts from Ensembl Genome Browser. In red, NEAT1-202, the isoform encountered in our data, which is also the canonic NEAT1 isoform. In green, all other NEAT1 isoforms. (C) NEAT1 isoforms differential expression per series. Bar length represents log2FoldChange in each isoform. Series are ordered according to metformin concentration, and NEAT isoforms are colored according to legend at the right side.

Figure 4

(A) total transcript counts in the 1 mb to each side region around the lncRNA for the 36 lncRNA selected for further analysis. (B) expression bar plots of the lncRNA isoform—transcript pairs which passed on the correlation cutoff (corr > 0.8 and p-value < 0.01) and on the genomic distance cutoff (1 mb to each side). On the top, AC125807.2-201 (ENST00000513358.3) and FOXM1-202 (ENST00000359843.8); on the bottom, ZBED5-AS1-207 (ENST00000664276.1) and EIF4G2-204 (ENST00000530211.6). Significance is computed as values inferior to 0.05. (C) Genomic region of around 300kilobases in chromosome 12. Tracks: GENCODE v37 annotation, ENCODE Hi-C, and Ensembl regulatory elements data. (D) genomic region of around 150kilobases in chromosome 11. Tracks: GENCODE v37 annotation, ENCODE Hi-C, Genehancer IM-PET, and Ensembl regulatory elements data.

Figure 5

(A) enrichment of AC016831.6-205 (ENST00000604514.1) targets using fgsea algorithm and the MsigDB C2 pathway dataset. NES was calculated using correlation values as the ranking; thus, positive NES indicates direct correlation between those target-transcripts and the pathway, and negative NES indicates inverse correlation between those target-transcripts and the pathway. Only protein-coding transcripts were used as target genes for this analysis. The use cutoff was p-adjusted inferior to 0.0001. x axis ordered according to series metformin concentration. (B) Network representation of target genes of AC016831.6-205 (ENST00000604514.1) are differentially expressed transcripts in at least 4 series and pass on the correlation cutoff (corr > 0.8 and p-value < 0.01). (C) enrichment of NEAT1-202 targets using fgsea algorithm and MsigDB C2 pathway dataset. Other information same as 5A.