TFAM overexpression preferentially down-regulates gene expression in A549 cells.
We overexpressed human TFAM gene in human lung epithelial cancer cell line A549 to study TFAM-regulated transcriptome. A549 cells were transfected by TFAM overexpressing plasmid (TFAM-OE) and empty vector (control). The results we gained from the experiment have to be tested in order to prove their reliability. The success of TFAM overexpression was validated by western blot analysis using antibody against the FLAG tag (Fig. 1A). A total of six cDNA libraries were constructed and sequenced for the control and TFAM-OE in A549 cells, with three biological replicates for each group (Ctrl_1st, Ctrl_2nd, Ctrl_3rd, TFAM_1st, TFAM_2nd, TFAM_3rd). The raw reads from RNA-seq were cleaned, and.the cleaned reads were then mapped to the human genome. The amount of the cleaned reads that were mapped to the human genome was averagely 81 million per sample and the amount of uniquely mapped reads averagely 75 million (Table S2).
The uniquely mapped reads were used for the following analysis. The fragments per kilobase of transcript per million fragments mapped (FPKM) was used to present the level of gene expression. We obtained 24, 593 expressed genes when FPKM was set higher than 0 and 12,158 genes when FPKM > 1 in at least one sample (Table S2). On the base of the results of FPKM, a correlation of matrix was built up (Fig. 1B). The correlation matrix showed a high similarity among all six transcriptome, indicating the confidence of the experiments. Nevertheless,the three biological replicates in the same group of samples showed a high correlation than between two different groups (Fig. 1B), indicating that increased expression of TFAM imposed a consistent transcription difference.
To explore the gene expression differences between TFAM-OE and control, a volcano map was drawn on the base of edgeR. The volcano map showed the up-regulated genes and down-regulated genes. A standard of fold change ≥ 2 or ≤ 0.5 and a 5% false discovery rate (FDR) was used to seperate the significantly regulated genes from the other genes, resulting in 169 significantly up-regulated genes and 642 down-regulated genes (Fig. 1C). The details of the differentially expressed genes (DEGs) were presented in Table S4. Moreover, heatmap plot of the FPKM values of all DEGs showed a clear separation between the TFAM-OE and control group, and a high consistency among three replicates of the same groups (Fig. 1D). All the results above supported a conclusion that TFAM shows a significantly higher capacity of down-regulating than up-regulating gene expression.
TFAM-repressed genes are enriched in cytokine-mediated signaling, interferon response and signaling, viral and immune responses
To explore the potential biological functions of the TFAM-regulated genes, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were used to analyze the DEGs. There were 385 down-regulated and 102 up-regulated DEGs being annotated by GO, and 459 down-regulated and 120 up-regulated DEGs annotated by KEGG (Table S5). The GO biological process mostly enriched by the up-regulated genes was nervous development, which might be related to TFAM function in neurological diseases (Fig. 1E, upper panel, Table S4). The KEGG pathways enriched by TFAM-upregulated genes were mostly related to metabolisms (Fig. 1F, upper panel, Table S4).
Strikingly, the TFAM down-regulated genes were highly enriched in anti-viral GO biological processes, which involved 9 of the top ten enriched processes including cytokine-mediated signaling pathway, type I interferon-mediated signaling pathway, response/defense response to virus, interferon-gamma-mediated signaling pathway, immune response, response to interferon-beta/alpha, and innate immune response (Fig. 1E, lower panel). Consistently, Herpes simplex infection, influenza A, measles, cytokine-cytokine receptor interaction and Toll-like receptor signaling pathway were the top five enriched KEGG pathways by TFAM down-regulated genes, which included three viruses and two viral response pathways. We noted that most of the interferon-stimulated genes (ISGs) that have been previously shown an upregulated expression in Tfam deficiency mouse MEF cells [23] demonstrated an marked increase in their expression in the TFAM-overexpressing A549 cells (Fig. 1G). This consistency indicates that TFAM functions in repressing antiviral responses are similar in human and mice.
In other to verify the regulation of TFAM regulated gene expression, 2 upregulated and 5 down-regulated DEGs were randomly selected and tested by RT-qPCR experiment (Fig. 2). The cDNA generated from the control and TFAM-OE transfected cells were used as template. Among these genes, several were classical inflammation and antiviral response genes, CCL5, IL7, IRF7 and TLR3. TFAM-regulated expression of all these 7 genes were demonstrated a satisfied consistency by both RNA-seq and RT-qPCR experiments, supporting the reliability of the data of RNA-seq.
TFAM regulates thousands of alternative splicing events, particularly repressing exon skipping
The RNA binding capability of TFAM encourages us to propose that TFAM may affect some of the mRNA biogenesis and/or metabolic processes [24–26]. To pursue this possibility, we analyzed the changes in alternative splicing events between the transcriptomes of the control and TFAM-OE A549 cells. The alternative splicing events (ASEs) were classified into nine categories, which includes an intron retention (IR) event, and 8 non-IR (not intron-retention) events including mutually exclusive 3’ UTR (3pMXE), mutually exclusive 5’ UTR (5pMXE), alternative 3’ splice site (A3SS), A3SS&exon skipping (ES), alternative 5’ splice site (A5SS), A5SS&ES, cassette exon, exon skipping (ES).
A total of 87437 ASEs were identified in this study, involving 23810 known ASEs and 61663 novel ASEs (Table S6). To identify the TFAM-regulated alternative splicing events (RASEs), the changes of the AS ratio between TFAM-OE and control transcriptomes were compared. By setting a cut-off of P-value 0.05, 2563 RASEs were identified, containing 331 IR and 2232 non-IR (NIR) RASEs (Fig. 3A, Table S7). Several examples of the quantification results of the RASEs using RNA-seq junction reads were shown in Fig. 3B. These RASEs were primarily ES, CassetteExon, A3SS and A5SS (Fig. 3A). We noticed that when compared to the cassette exon events, TFAM overexpression resulted in more ASEs that demonstrated an decreased level of exon skipping. This suggested that TFAM may specifically represses exon skipping in an unknown mechanism.
TFAM regulated alternative splicing events are enriched in DNA repair, metabolic pathways and transcriptional regulation
We assessed the potential function of TFAM-regulated alternative splicing by performing GO and KEGG analyses on RASGs. GO analysis showed that TFAM-regulated alternative splicing were mostly enriched in DNA repair pathway, followed by protein localization and nerve growth factor receptor signaling pathway (Fig. 3C). Metabolic pathways were most enriched by TFAM-regulated RASGs in the KEGG analysis results (Fig. 3D). Interestingly, we noticed that regulation and negative regulation of transcription from DNA polymerase II promoter were among the top 10 GO terms enriched by TFAM-regulated RASGs. These results suggested that TFAM may exert some of its biological functions via alternative splicing regulation.
When the 1796 regulated alternative splicing genes (RASG) were overlapped with 811 DEGs, and there were only 39 shared genes, showing no statistic significance (Fig. 3E). We therefore speculated that most of these genes might be regulated by TFAM independently at the transcription and alternative splicing levels. Interestingly, these 39 shared genes were strongly enriched in positive regulation of transcription from RNA polymerase II promoter (GO-BP), which included DDIT3, IRF7, CSF3, IFI16, HELZ2, NLRC5, SMARCA4 (Fig. 3F, Table S8). This indicated that transcription regulation genes may be coordinately regulated by TFAM at the transcription and alternative splicing levels.
To assess the confidence of the TFAM-regulated ASEs, six RASEs in DNA repair and cell apoptosis were randomly selected for performing RT-qPCR experiments. To enable a more clear comparison, we showed the RNA-seq quantification results, the RNA-seq reads distribution map with the number of the splicing junction reads labeled, and also the RT-qPCR results side by side (Fig. 4 and Fig. S1). The exon skipping event in Among these tested RASEs, four were statistically consistent with the RNA-seq quantification results, which included CD44, HNRNPH1, BRCA1 and IKBKP, while those in HNRNPF and HNRNPR were not (Fig. 4 and Fig. S1). We noticed that the inconsistency RASEs showed smaller change in the alternative splicing ratios upon the TRAM overexpression.
TFAM-regulated alternative splicing is strongly linked to its repressed gene expression
Given that the genes containing TFAM-regulated ASEs were significantly enriched in regulation of transcription from DNA polymerase II promoter, we proposed that if TFAM may regulate target gene transcription of via affecting the alternative splicing of transcription factors. To test this hypothesis, we subjected all the RASE-containing genes to TFBSTools to identify transcription factors (TFs) with known binding motifs, which resulted in 45 such TFs (Supp Table S9A). A total of 683 transcription factors with their corresponding motif sequences were collected in TFBSTools. We then identified DNA motifs enriched in the promoter regions (1K-3K adjacent to the transcription start sites) of the TFAM-repressed genes and TFAM-promoted genes, demonstrating more consistently predicted motifs in the down-regulated genes than the upregulated genes (Fig. 5A, Table S9B). We than ran TFBSTools to correspond the DNA binding motifs of TFs whose splicing was regulated by TFAM to the motifs enriched by TFAM-repressed and promoted genes separately. It was shown that the DNA binding motifs 26 of the 45 TFAM-regulated TFs (RASGs) were overlapped with 36 of the 147 motifs present in the promoters of TFAM-repressed genes (Fig. 5B); the overlap was remarkably significant (P-value 6.45e-9). One overlapped motif was usually present in many TFAM-repressed genes (Fig. 5C) ; 36 overlapped motifs were present in 638 of the 642 TFAM-repressed genes. As such, TFAM regulation of alternative splicing events in transcription factors has a potential to regulate the transcription of almost all of the TFAM-repressed genes.
We further analyzed the Pearson correlation between the TFAM-regulated alternative splicing and TFAM-regulated gene expression. Setting the cutoffs as correlation coefficient ≥ 0.95 and ≥ P-value 0.005, we found that the expression change of 602 TFAM-repressed genes were correlated with the alternative splicing ratio change of TFAM-regulated TFs. Among these 602 TFAM-repressed genes, we selected those involved in cytokine-mediated signaling (one GO-BP term), interferon signaling and response (4 GO-BP terms), response/defense response to virus (2 GO-BP terms), immune/innate immune response (2 GO-BP terms) (Fig. 1E, Fig. S2C). These genes were integrated to three groups, many of which were shared (Fig. 5D). The correlation network between these selected TFAM-repressed genes and the corresponding TFAM-regulated TFs was drawn (Fig. 5E). The results further supported that TFAM-regulated alternative splicing may generally contribute to the TFAM-repressed expression of anti-viral response genes.
The connections between TFAM-regulated alternative splicing of transcription factors and TFAM promotion of gene expression were analyzed in parallel (Table S9), showing a strong overlap between the DNA binding motifs of TFAM regulated TFs and motifs enriched by TFAM-upregulated genes (Fig. S2A), and most of the TFAM-upregulated genes (transcription) containing DNA binding motifs of TFAM-regulated TFs (alternative splicing) (Fig. S2B).