CCT3- LINC00326 axis regulates hepatocarcinogenic lipid metabolism

Abstract

Objective: To better comprehend transcriptional phenotypes of cancer cells, we globally characterised RNA-binding proteins (RBPs) to identify altered RNAs, including long non-coding RNAs (lncRNAs).

Design: To unravel RBP-lncRNA interactions in cancer, we curated a list of ~2300 highly expressed RBPs in human cells, tested effects of RBPs and lncRNAs on patient survival in multiple cohorts, altered expression levels, integrated various sequencing, molecular and cell-based data.

Results: High expression of RBPs negatively affected patient survival in 21 cancer types, especially hepatocellular carcinoma (HCC). After knockdown of the top 10 upregulated RBPs and subsequent transcriptome analysis, we identified 88 differentially expressed lncRNAs, including 34 novel transcripts. CRISPRa-mediated overexpression of four lncRNAs had major effects on the HCC cell phenotype and transcriptome. Further investigation of four RBP-lncRNA pairs revealed involvement in distinct regulatory processes. The most noticeable RBP-lncRNA connection affected lipid metabolism, whereby the non-canonical RBP CCT3 regulated LINC00326 in a chaperonin-independent manner. Perturbation of the CCT3-LINC00326 regulatory network led to decreased lipid accumulation and increased lipid degradation in cellulo as well as diminished tumour growth in vivo.

Conclusions: We revealed that RBP gene expression is perturbed in HCC and identified that RBPs exerted additional functions beyond their tasks under normal physiological conditions, which can be stimulated or intensified via lncRNAs and affected tumour growth.

Keywords: hepatocellular carcinoma; lipid metabolism.

Figure 1

RBPs are deregulated in cancer and affect patient survival. (A, B) Boxplots of (A) gene expression level and (B) cox proportional hazard (coxph) coefficient of protein-coding genes grouped into RBP (green) and non-RBP (grey) genes. Each row represents a different cancer type defined by TCGA. Each cancer type consisted of 144 to1006 patients. Hinges correspond to the first and third quartiles, and whiskers correspond to the 1.5-times interquartile range. (C, D) Volcano plots demonstrate differentially expressed (DE) genes in the two HCC cohorts (C) TCGA-LIHC and (D) Australia HCC. Data points represent DE RBP genes (dark green), not significantly (ns)-DE RBP genes (FDR>0.05, light green) and all other genes (grey). (E, F) Four-way Venn diagrams intersect the number of (E) upregulated and (F) downregulated RBP genes in the TCGA and Australia HCC cohorts as well as in liver cancer cell lines Huh7 and HepG2. Intersections highlighted (bold) show the number of RBP genes commonly deregulated in HCC cohorts and cell lines. (G) Heatmap displays changes in expression levels for commonly upregulated RBP genes in HCC as highlighted in (E). Expression level is sorted by average tumour z-score from left to right. Black bar marks the top 10 highest expressed RBP genes, and purple asterisk marks RBPs with a canonical RBD. Colour gradient indicates z-score differences (green: high; grey: low). (H) Volcano plot displays comparison of the top and bottom tercile in RBP gene expression levels and hazard ratioswithin the TCGA-LIHC cohort (377 patients) (grey: downregulated and green: upregulated in tumour). The size of the circle represents the gene expression level of each RBP relative to each other (broad: high, narrow: low). (I) Kaplan-Meier plot shows the association of CCT3 gene expression level and 10-year survival within the top and bottom tercile within the TCGA-LIHC cohort (377 patients) (black: high and blue: low CCT3 gene expression levels). Statistics: log-rank (Mantel-Cox). BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; COAD, colon adenocarcinoma; DE, differentially expressed; ESCA, oesophageal carcinoma; GBM, glioblastoma multiforme; HCC, hepatocellular carcinoma; HNSC, head and neck squamous cell carcinoma; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LAML, acute myeloid leukaemia; LGG, brain lower grade glioma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; OV, ovarian serous cystadenocarcinoma; PAAD, pancreatic adenocarcinoma; RBP, RNA-binding proteins; READ, rectum adenocarcinoma; SARC, sarcoma; SKCM, skin cutaneous melanoma; TCGA, The Cancer Genome Atlas; UCEC, uterine corpus endometrial carcinoma.

Figure 2

Reduced expression of RBP genes impact various cellular and molecular responses in HCC cell lines. (A) Bar graph displays RBP gene expression levels before (black bars) and after siRNA-mediated RBP-KD (coloured bar) relative to ACTB in HepG2 and Huh7 cells determined by qPCR (n=4 mean, ±SEM). Number above each bar shows the average KD efficiency in percent. Colour code: CCT3 (blue), DDX39A (magenta), HIST1H1C (turquoise), IGF2BP1 (red), KPNA2 (plum), NQO1 (grey), PEG10 (yellow), PKM (green), STMN1 (purple) and TOP2A (brown). Individual replicates are displayed by white circles. (B) Line graphs show the relative number of metabolically active cells (measured by optical density) over 7 days after siRNA-mediated RBP-KDs assayed by the MTT assay (n=5). Black line: non-targeting siRNA control (siNT-Ctrl), coloured line: RBP-specific siRNA KD. (C) Dot plot represents the percentage of dead cells after siRNA-mediated RBP-KDs (colour-coded) after 5 days (n=7, mean, ±SEM). (B, C) Statistics: paired two-tailed t-test, *p<0.05, **p<0.01, ***p<0.001. (D, E) Circle plots display number of genes per RNA biotype affected by RBP-KDs. The diameter of the circles corresponds to the number of genes in each category. Deregulated genes falling into different ncRNA subcategories are shown in figure 2E.(F) Volcano plots demonstrate DE genes after RBP-KDs. Data points represent significantly DE genes (colour-coded by RBP-KD, FDR<0.05) and not significantly DE genes (grey, FDR>0.05). Bolded numbers on the top of each graph indicate total number of DE genes (FDR<0.05). The wider circle in each plot highlights the downregulated RBP. Genes with an FDR value smaller than 1 × 10⁻¹⁰ were collapsed at 1 × 10⁻¹⁰. (G) Circle plot shows GO term and KEGG pathway enrichment analysis of deregulated genes after the CCT3-KD (FDR<0.05). The diameter of the circles corresponds to the number of genes in each GO or KEGG term and the colour code represents varying degrees of significance (white: high and blue: low p value). (H) Interaction network displays connections of the five most significant GO BP terms in figure 2G. GO term is bolded and gene names are highlighted. BP, biological process; CC, cellular compartment; DE, differentially expressed; KD, knockdown; MF, molecular function; OD, optical density; RQ, relative quantity.

Figure 3

RBP-KD affects lincRNA gene expression levels. (A, C) Bar graph show the number of (A) annotated and (C) novel DE lincRNA genes detectable after RBP-KDs. The frequency of lincRNA genes in one (black dot) or multiple (black dots connected by a line) RBP-KD experiments is shown. (B, D) Circle plots display the occurrence of (B) annotated and (D) novel lincRNA genes per RBP-KD. The diameter of the circles corresponds to varying degrees of significance (large: high, and narrow: low FDR value, black line: FDR<0.05). The colour code represents fold change (red: upregulated and blue: downregulated). Vertical bars specify the three most common clusters defining lincRNAs as either consistently downregulated (blue) or upregulated (red), or with varying pattern deregulation across the ten RBP-KD (purple). A star (*) marks lincRNAs used for further investigation. (E) The UCSC genome browser view demonstrates the genomic location of the novel lincRNA MSTRG.12891 in between genes encoding for PKM and PARP6. Arrows indicate direction of gene transcription. Gene expression patterns in Huh7 cells transfected with siNT-Ctrl (black) or siRNA-mediated KD of STMN1 (purple) are shown. The y-axis of each track specifies normalised RNA-seq read intensity.

Figure 4

Overexpression of lincRNA genes in HCC cell line causes molecular and cellular alterations. (A) Bar graphs exemplify the increase in relative gene expression of LINC00326 over 5 days after transfection of a CRISPRa vector with non-targeting gRNA-CRISPRa controls (CaNT-Ctrl, black gradient) or three lincRNA-specific gRNAs (LINC00326, blue gradient) determined by RT-qPCR. (B) Bar graph demonstrate the fold change in lincRNA gene expression 2 days after CRISPRa transfection of lincRNA-specific versus CaNT-Ctrl determined by RT-qPCR (n=2–4, mean, +SEM). The colour-code links the lincRNA to the respective RBP-KD experiment in which the lincRNA was identified (blue: CCT3, red: IGF2BP1, purple: STMN1). (C) Line graphs show relative increase in metabolically active HCC cell number over 7 days after CRISPRa transfection with CaNT-Ctrl (black) or lincRNA-specific gRNAs (coloured) determined by MTT assay. (D) Dot graph shows the percentage of early apoptotic cells 5 days after CRISPRa transfection with CaNT-Ctrl (black) and lincRNA-specific gRNAs (coloured) determined by FACS (n=2–8, mean, ±SEM). (A–D) Each biological replicate of Huh7 and HepG2 is displayed by circles. Graphs are coloured according to the colour-code selected for the RBP partner through which the lincRNA was identified (blue: CCT3, red: IGF2BP1, purple: STMN1). Statistics: paired two-tailed t-test, *p<0.05, **p<0.01. (E) Volcano plots demonstrate DE genes 2 days after CRISPRa transfection of lincRNA-specific versus CaNT-Ctrl determined by RNA-seq in Huh7. Data points represent significantly DE genes (coloured, FDR<0.01) and not significantly DE genes (grey, FDR>0.01). Bolded numbers on the top of each graph indicate total number of DE genes (FDR<0.01). Circle highlights lincRNA and lincRNA-interacting genes investigated. (F) Four-way Venn diagram intersects the number of DE genes after each lincRNA-OE experiment (FDR<0.01). (G) Circle plot displays the number of genes per RNA biotype affected by the lincRNA-OE. The diameter of the circles corresponds to the number of genes in each category.

Figure 5

The CCT3-LINC00326 interactome regulates lipid metabolism. (A) Bar graph shows enrichment of LINC00326 compared with GAPDH and HULC (negative controls) over input control after RNA immunoprecipitation with a TCP1 (CCT1)-(purple), CCT2- (white) or CCT3-specific (blue) antibody versus the beads-only (no antibody) control (orange) followed by RT-qPCR with gene- and strand-specific primers (n=3, mean, +SEM). Each biological replicate is displayed by circles. Statistics: ANOVA with a Bonferroni’s multiple comparison test, ***p<0.001, ns: non-significant. (B) Microscopic image of single-molecule RNA FISH using exonic probes for LINC00326 (white dots and arrows) in LINC00326-OE Huh7 HCC cells. DAPI (blue) marks the nucleus. Pie chart represents the fraction of signals in the nucleus (blue) or cytoplasm (pink) in cells, or cells without any signal (grey). Scale bar: 5 µm. (C) Violin plots show quantification of LINC00326 RNA FISH signal localisation in Huh7 HCC cells (n=40–53, statistics: paired two-tailed t-test, ns: non-significant). (D) Line graphs shows the relative number of metabolically active cells (measured by optical density, OD) over 7 days after the siRNA-mediated CCT3-KD and/or LINC00326-KD assayed by MTT assay (n=3, mean, +SEM, statistics: ANOVA with a Bonferroni’s multiple comparison test, *p<0.05, **p<0.01). (E) Bar charts show siRNA-KD-efficiencies of targeted RBP genes (figure 5D) (n=3, mean, +SEM). (F) Two-way Venn diagram intersects the number of deregulated genes after CCT3-KD and LINC00326-OE. (G) Circle plot shows GO term and KEGG pathway enrichment analysis of the 70 commonly deregulated genes after CCT3-KD and LINC00326-OE. The diameter of the circles corresponds to the number of genes in each GO or KEGG term and the colour code represents varying degrees of significance (white: high and blue: low p value). (H) Interaction network displays connections of the five most significant GO BP terms shown in figure 5G. GO term is bolded and gene names are highlighted. (I) Heatmap (unsupervised clustering) displays the fold change in expression levels for lipid metabolic process genes (figure 5H) when comparing HCC cohorts and cell lines, CCT3-KD and LINC00326-OE over non-cancerous or NT controls, respectively. Colour gradient indicates log₂FC differences (red: high; blue: low). (J) Circle plot demonstrates enrichment of TF-binding motifs in the promoter regions of the 70 commonly deregulated genes of the CCT3-KD and LINC00326-OE. The diameter of the circles corresponds to the fold change over background controls and the colour code represents varying degrees of significance (white: high and blue: low p value). Identified motifs for each TF are shown in online supplemental figure S9f. (K) StringDB interaction network shows the links of the TFs identified in figure 5F and known interaction partners (direct: green, indirect: grey). Two direct connections to lipid metabolism genes are highlighted (blue). (L) The UCSC genome browser view demonstrates genomic location of three lipid metabolism-associated genes and the LINC00326 gene. Arrows indicate direction of gene transcription. Horizontal bars indicate ChIP-seq signals (black: strong; grey: weak) for available TF-binding events in HepG2 or liver cells (ENCODE).

Figure 6

The CCT3-LINC00326 network affects lipid metabolism and tumour growth in vitro and in vivo. (A–C) Bar graphs show comparison of (A) malondialdehyde (MDA) production (lipid degradation), (B) Oil Red O staining (lipid accumulation) and (C) ROS production of CCT3-KD or LINC00326-OE (blue) to the respective NT controls (black) 48 hours after transfection (n=4–6, mean, +SEM). Each biological replicate is displayed by circles. Statistics: paired two-tailed t-test, (D) Dot graph displays library size-normalised CCT3 mRNA expression level in adult human individuals with normal and obese weight, MAFLD and MASH (n=2–8, mean). LINC00326 was not assayed. Statistics: one-way ANOVA. (E) Boxplot of normalised CCT3-expression in the TCGA-LIHC cohort divided by main pathological cancer stage. Statistics: one-way ANOVA with Tukey Honest Significant Differences test. (F–H) Microscopy images of TUBULIN-GFP expressing Huh7 cells (F) in vitro (scale bar: 100µM) and (G–H) in vivo in zebrafish xenografts (scale bar: 500 µM).(I, J) Box plots show changes in tumour area in zebrafish xenografts after (I) CCT3-KD (n=20–21, mean, ±SEM) or (J) LINC00326-OE (n=21–28, mean, ±SEM). Individual zebrafish were followed over 5 days and tumour area is given relative fold change to day 1 (D1) after injection. Statistics: one-way ANOVA, *p<0.05, ***p<0.001. (K) Schematic model for CCT3-LINC00326 regulation of lipid metabolism. Reducing CCT3 or increasing LINC00326 gene expression in liver cancer cells inhibits lipid accumulation and promotes lipid degradation (peroxidation). Due to the strong dependency of cancer cells towards high lipogenesis, this in turn slows down cancer growth and promotes cell death. *p<0.05, **p<0.01, ***p<0.001, MAFLD, metabolic-associated fatty liver disease; MASH, metabolic-associated steatohepatitis; NS, non-significant; ROS, reactive oxygen species.