TCF7L2 lncRNA: a link between bipolar disorder and body mass index through glucocorticoid signaling

Abstract

Bipolar disorder (BD) and obesity are highly comorbid. We previously performed a genome-wide association study (GWAS) for BD risk accounting for the effect of body mass index (BMI), which identified a genome-wide significant single-nucleotide polymorphism (SNP) in the gene encoding the transcription factor 7 like 2 (TCF7L2). However, the molecular function of TCF7L2 in the central nervous system (CNS) and its possible role in the BD and BMI interaction remained unclear. In the present study, we demonstrated by studying human induced pluripotent stem cell (hiPSC)-derived astrocytes, cells that highly express TCF7L2 in the CNS, that the BD-BMI GWAS risk SNP is associated with glucocorticoid-dependent repression of the expression of a previously uncharacterized TCF7L2 transcript variant. That transcript is a long non-coding RNA (lncRNA-TCF7L2) that is highly expressed in the CNS but not in peripheral tissues such as the liver and pancreas that are involved in metabolism. In astrocytes, knockdown of the lncRNA-TCF7L2 resulted in decreased expression of the parent gene, TCF7L2, as well as alterations in the expression of a series of genes involved in insulin signaling and diabetes. We also studied the function of TCF7L2 in hiPSC-derived astrocytes by integrating RNA sequencing data after TCF7L2 knockdown with TCF7L2 chromatin-immunoprecipitation sequencing (ChIP-seq) data. Those studies showed that TCF7L2 directly regulated a series of BD risk genes. In summary, these results support the existence of a CNS-based mechanism underlying BD-BMI genetic risk, a mechanism based on a glucocorticoid-dependent expression quantitative trait locus that regulates the expression of a novel TCF7L2 non-coding transcript.

Fig. 1. TCF7L2 transcription in the human brain.

a Clusters of human brain cells that were assigned cell types based on the expression of cell type-specific marker genes quantified by single-nucleus RNA-seq in 49,495 single nuclei from multiple cortical areas. Each dot represents a single cell. Cell types have been color-coded and labeled. OPCs: oligodendrocyte progenitor cells. b TCF7L2 expression in each of these brain cells was color-coded based on TCF7L2 expression levels. c Heatmap for the bulk expression level of TCF7L2 and marker genes in the clustered brain cell. Expression level in astrocytes is highlighted in the dashed box. a–c show data generated by the Allen Institute (https://portal.brain-map.org/atlases-and-data/rnaseq) [47]. d Percentages of TCF7L2-expressing cells in each sub-group of cell types are shown graphically. Numbers shown in parentheses are the numbers of TCF7L2-expressing cells/total number of sequenced cells. Single-cell RNA-seq data were generated from surgically removed human cerebral cortical tissues in two studies [48, 49]. The two data sets were combined and re-analyzed as reported in a previous study [55]. e TCF7L2 chromatin accessibility was quantified by single-cell ATAC-seq in 70,631 cells across human brain regions. ATAC-seq peaks across the TCF7L2 gene in each clustered cell type are shown separately from top to bottom. Physical positions on chromosome 10 (Chr.10) are labeled across the top. Based on its physical location on Chr.10 (GRCh38/p13), the TCF7L2 gene is depicted at the bottom with exons shown as bars and introns as lines. The scATAC-seq data were generated by Corces et al. [50] and visualized by using the WashU Epigenome Browser [81].

Fig. 2. TCF7L2 SNPs and glucocorticoid-regulated transcript variants.

a TCF7L2 gene structure. TCF7L2 exons are numbered from 1 to 14 based on the reference RNA transcript (ID: ENST00000369397). Alternative exons that map to introns of the reference transcript were named using the intron number, followed by a letter of the alphabet. Positions of the rs7903146 SNP and the rs12772424 SNP are indicated and they map 122.2 kb distant from each other. The rs12772424 SNP variant allele (A) created a half-palindrome glucocorticoid response element (GRE). b The rs12772424 SNP locus. Positions of rs12772424 and its “linked” SNPs (r² > 0.35, D’ >0.79 in European population) are labeled. TCF7L2 exons are depicted as boxes with exon numbers labeled above. Putative GREs are indicated below. Transcription factor (TF) binding sites for the glucocorticoid receptor (NR3C1), FOXA1, POLR2A, and CTCF are shown as boxes below the GREs. Darker colors represent stronger binding. c TCF7L2 transcript variants (T-1 to T-3) annotated by Ensembl are shown with their IDs shown in parentheses. d Immunofluorescent staining of astrocyte markers, GFAP and S100B, for hiPSC-derived astrocytes. Scale bars represent 50 µM. e Fold changes in RNA levels for TCF7L2 transcript variants in hiPSC-derived astrocytes after treatment with 100 nM DEX. Unique exon junctions (“J”) for specific transcript variants were quantified by qRT-PCR. GAPDH and FKBP5 expression was measured as internal and treatment controls, respectively. Data are mean ± s.d. (n = 3). ***P < 0.001 by Dunnett’s test. f DEX-dose-dependent induction of FKBP5 and repression of “T-2” and “T-3” in hiPSC-derived astrocytes. g Construction of reporter gene plasmid. DNA fragments that included the TCF7L2 rs12772424 SNP (373 bp) and exons 4b-4d (1812 bp) were cloned into the 5′-region of the LUC2 reporter gene. h Luciferase assays after cells were transfected with plasmids containing rs12772424 wild type (T) or variant (A) alleles and were treated with DEX or vehicle control. Data are presented as relative luciferase units (RLUs) after dexamethasone/vehicle (DEX/Veh) normalization. ***P < 0.001 by t test. i Relative RNA levels of TCF7L2 “T-2” and “T-3” compared with those of the housekeeping genes, VCP and C1orf43, in total RNA samples prepared from human tissue lysates. Data are mean ± s.d. (n = 3).

Fig. 3. Characterization of TCF7L2 Transcript “T-3”.

a TCF7L2 transcripts “T-1” and “T-3” and mapping sites of primers (P) that were used to amplify TCF7L2 cDNA. Primer mapping sites are labeled by arrows below the exons depicted as boxes in which exon numbers are labeled. Primers (P) are numbered from 5′ to 3′ of the gene. Forward primers are underlined. Primer sequences are listed in Supplementary Table S5. b The table lists eight polymerase chain reactions (PCR) used to amplify portions of the TCF7L2 cDNA. The combination of primers for each PCR reaction and the estimated base pair (bp) length of the PCR product are listed. NC = negative control reaction, which was not expected to yield a PCR product. c RT-PCR products using human brain total RNA or poly(A) + RNA as templates are shown after separation on a 1.2% agarose gel. PCR numbers (#) match the PCR assays listed in b. PCR assays that were designed to amplify “T-3” (from #1 to #4) yielded products of the expected sizes when total RNA was used as the PCR template but failed to amplify target sequences when poly(A) + RNA was the template. NC reactions (#5 and #6) and positive control reactions (#7 and #8) that amplify TCF7L2 mRNA transcript “T-1” produced the expected results. d Relative RNA levels of TCF7L2 “T-2” and “T-3” compared with GAPDH in hiPSC-derived astrocytes, hepatocytes, and pancreatic β-cells. Data are mean ± s.d. e RT-PCR products amplified TCF7L2 cDNA in hiPSC-derived astrocytes. f qRT-PCR determined “T-3” levels in hiPSC-derived astrocytes after transfection with antisense oligonucleotides (ASOs) targeting exon-4d. The Y axis represents “T-3” expression relative to control (non-targeting ASO). The combination of ASO 1 and 2 (ASOs 1+2) resulted in the best knockdown (KD) efficiency. g Volcano plot for RNA-seq identified differentially expressed genes (DEGs) in hiPSC-derived astrocytes. TCF7L2 mRNA expression was downregulated after “T-3” KD. h Western blot assays for TCF7L2 after knockdown of lncRNA-TCF7L2 “T-3” in hiPSC-derived astrocytes. i Top pathways enriched by DEGs of “T-3” KD (FC > 2.0; FDR < 0.05) in BioPlanet data set and j GWAS Catalog phenotypes.

Fig. 4. TCF7L2 protein function in hiPSC-derived astrocytes.

a TCF7L2 RNA levels in hiPSC-derived astrocytes after transfection with non-targeting siRNA control (siControl) or “pooled” TCF7L2-targeting siRNAs (siTCF7L2). b Western blot assay for TCF7L2. Vinculin was blotted as an internal control. c Expression levels as log₂ (reads per kilobase of transcript per million reads mapped [RPKM] +1) values for all detected genes in the RNA-seq libraries for siTCF7L2 (Y axis) compared with siControl (X axis). n = 2 independent replicates for both groups. d Heatmap of TCF7L2 ChIP-seq results, in which regions covering the called peaks ±2 kb were scaled to the same size relative to the center point of the peaks. The heatmap intensity represents fold enrichment of the TCF7L2 ChIP signal over input, with each row depicting each called peak, and with the X axis representing position relative to the center point of the peaks. The top panel is the average profile of the ChIP-seq peaks, with the Y axis representing fold enrichment of the TCF7L2 ChIP signal over input. e Distribution of 1858 identified TCF7L2 peaks (FDR < 0.01) across the genome. UTR untranslated region, TSS transcription start site, TTS transcription termination site. f The top three binding motifs and their binding transcription factors (TFs) were identified from HOMER de novo analysis based on p values. g Integration of RNA-seq and ChIP-seq data identified 186 genes that were both bound by TCF7L2 and were differentially expressed after TCF7L2 KD. Based on false discovery rate (FDR < 0.05), the “top” 500 up- and 500 downregulated DEGs and genes mapping ±100 kb from the TCF7L2 ChIP peaks (3541 genes) were used to identify “overlapping” TCF7L2 target genes. h Heatmap of expression for the 186 TCF7L2 target genes in control and TCF7L2 KD samples. Genes involved in the Wnt signaling pathways have been labeled. i Gene Ontology enrichment analysis for the 186 TCF7L2 target genes identified by integration of the RNA-seq and ChIP-seq data. Colors represent the number of genes enriched in each pathway. The X axis represents −log₁₀ of the FDR. Wnt signaling pathways have been highlighted.

Fig. 5. TCF7L2 directly regulates BD risk genes.

a Disease pathway enrichment of the 186 TCF7L2 target genes that showed bipolar disorder as the only significant enrichment in the Jensen’s Lab Disease database. No other pathway was significantly (FDR < 0.01) enriched in databases on the Enrichr website. b Expression levels of BD risk genes enriched in the disease pathway analysis were significantly changed after TCF7L2 KD in hiPSC-derived astrocytes. c Integrative Genomic Views of TCF7L2 ChIP peaks mapped to the BD risk genes shown in b. The orange panel represents the input control. The green panel represents TCF7L2 ChIP-seq reads. The purple panel represents TCF7L2 ChIP peaks that were called based on statistical significance (FDR < 0.01). Major ChIP peaks have been highlighted in dotted boxes. Genes mapping to the TCF7L2 ChIP peaks are shown in the blue panels at the bottom of each panel. The names of BD risk genes for which expression was changed after TCF7L2 KD are highlighted in red.