The TRIM-NHL RNA-binding protein Brain Tumor coordinately regulates expression of the glycolytic pathway and vacuolar ATPase complex

Abstract

The essential Drosophila RNA-binding protein Brain Tumor (Brat) represses specific genes to control embryogenesis and differentiation of stem cells. In the brain, Brat functions as a tumor suppressor that diminishes neural stem cell proliferation while promoting differentiation. Though important Brat-regulated target mRNAs have been identified in these contexts, the full impact of Brat on gene expression remains to be discovered. Here, we identify the network of Brat-regulated mRNAs by performing RNA sequencing (RNA-seq) following depletion of Brat from cultured cells. We identify 158 mRNAs, with high confidence, that are repressed by Brat. De novo motif analysis identified a functionally enriched RNA motif in the 3' untranslated regions (UTRs) of Brat-repressed mRNAs that matches the biochemically defined Brat binding site. Integrative data analysis revealed a high-confidence list of Brat-repressed and Brat-bound mRNAs containing 3'UTR Brat binding motifs. Our RNA-seq and reporter assays show that multiple 3'UTR motifs promote the strength of Brat repression, whereas motifs in the 5'UTR are not functional. Strikingly, we find that Brat regulates expression of glycolytic enzymes and the vacuolar ATPase complex, providing new insight into its role as a tumor suppressor and the coordination of metabolism and intracellular pH.

Graphical Abstract

Figure 1.

Identification of Brat-regulated mRNAs by RNA-seq. (A) Diagram of the domain architecture that defines TRIM-NHL proteins. (B) Diagram for Brat protein, following the naming convention used by Loedige et al. (12). (C) Western blot showing efficient depletion of V5-tagged Brat in DL1 cells by RNA interference (RNAi), relative to non-targeting control (NTC) RNAi. RNAi used two independent double-stranded RNAs (dsRNAs #1 or #2), which targeted non-overlapping regions of the brat mRNA. Western blot of tubulin served as a control for equivalent loading of the samples. Size of protein bands in kDa. Depletion of Brat for RNA sequencing (RNA-seq) was conducted in quadruple replicates. (D) Volcano plot showing differential gene expression following depletion of Brat. Genes are represented by log₂ fold change (lfc), relative to NTC RNAi and adjusted P-value (P^adj). Genes are additionally color-coded by significance-weighted fold change (z-score). (E) Venn diagram showing overlap between genes upregulated (lfc > 0, P^adj < 0.05, baseMean ≥ 50) in both RNAi conditions. Significance of overlap determined via Fisher’s exact test (***P< 0.001). (F) Violin plot, with inset boxplot, showing distribution of fold changes for genes upregulated in both RNAi conditions. (G) RNA levels of upregulated genes, relative to 7SK level, measured with reverse transcription-quantitative polymerase chain reaction (RT-qPCR). Mean ± standard deviation values are plotted. Significant differences determined via ANOVA with post-hoc Tukey–Kramer tests (*P< 0.05, **P< 0.01, ***P< 0.001, n = 4). Abbreviations: Dm,Drosophila melanogaster; RING, Really Interesting New Gene; B1 and B2, B-Box zinc fingers 1 and 2; CC, coiled-coil domain; S, serine-rich domain; Q, glutamine-rich domain; H, histidine-rich domain.

Figure 2.

A significant number of Brat-repressed genes are bound by Brat. (A) Venn diagram of genes upregulated by Brat RNAi and those identified in published Brat RNA-immunoprecipitations studies (RIPs). In all cases, Brat-associated genes were limited to those expressed in DL1 cells. (B) Fractions of upregulated genes that are bound by Brat in embryo extracts reported in RIP experiments by Loedige et al. (11) or Laver et al. (3). Genes were additionally considered ‘bound’ by Brat if identified in either RIP, and a smaller subset was identified in both RIPs. Significance of the overlaps between upregulated genes and these bound subsets were determined via Fisher’s exact tests (***P< 0.001). The number of genes per subset is indicated in each column. Kernel density estimate showing distribution of significance-weighted log₂ fold changes from (C) NTC versus dsRNA #1 and (D) NTC versus dsRNA #2 for genes either bound in Brat RIPs or that were not bound. Median significance-weighted log₂ fold changes for bound (solid) and unbound (dashed) genes displayed as vertical lines. Significance of differences between distributions was determined via a Wilcoxon rank sum test (***P< 0.001). Number of genes per subset in key.

Figure 3.

Brat motifs are enriched in the 3′UTRs of upregulated genes. (A) Fraction of genes bound in either Loedige et al. (11) or Laver et al. (3) Brat RIP experiments with at least one putative Brat binding site (BBS) motif in the 5′UTR, coding sequence (CDS) or 3′UTR. (B) The fraction of upregulated genes with at least one putative BBS motif in each mRNA feature. In panels (A) and (B), significance of the overlap between genes containing 5′UTR, CDS or 3′UTR motifs and upregulated genes was determined via Fisher’s exact test (***P< 0.001). The number of genes per subset is indicated in the columns. (C) Upregulated genes bound by Brat in published RIPs were analyzed for BBS content and location. The percentage of these Brat bound subsets that contain at least one putative Brat motif in each mRNA feature is displayed. Over-representation was determined via Fisher’s exact test (***P< 0.001). The number of genes per subset is reported in Supplementary Figure S3A. Kernel density estimate showing distribution of significance-weighted log₂ fold changes from Brat dsRNA #1 (D) and Brat dsRNA #2 (E) for genes that contain 3′UTR BBS motifs versus those that do not. Median values for genes with motifs (solid) and without (dashed) displayed as vertical lines. Significance of differences between the distributions were tested via Wilcoxon rank sum test (***P< 0.001). (F) The 5′-UUGUUD motif was significantly enriched in the 3′UTRs of genes that were upregulated by Brat RNAi (E-value 6.1 × 10⁻²¹), identified using the MEME algorithm. (G) The fraction of upregulated genes with at least one putative 5′-UUGUUD motif in each mRNA feature. (H) The fraction of upregulated genes, grouped by binding in published RIP datasets, that contain at least one 5′-UUGUUD motif in the 3′UTR. In panels (A)–(C), (G) and (H), significance of the overlap between genes containing 5′UTR, CDS or 3′UTR motifs and upregulated genes was determined via Fisher’s exact test (***P< 0.001). The number of genes per subset is reported in each column.

Figure 4.

The number of 3′UTR BBS motifs sensitizes mRNAs to repression by Brat. (A, B) Genes were divided by whether they were bound by Brat in either RIP dataset or unbound. The percentage of genes with the indicated number of 3′UTR BBS motifs are displayed as histograms. Both the minimum (A) and maximum (B) number of BBS motifs per gene were considered for genes with multiple mRNA isoforms. Similarly, histograms display the number of 3′UTR motifs among genes upregulated following Brat RNAi and unregulated, when the minimum (C) and maximum (D) number of BBS motifs per gene are considered. In panels (A)–(D), the number of genes per subset are indicated in the key. The number of genes in each category is presented in Supplementary File S3. (E) Renilla luciferase reporters (Rluc) with various numbers of 5′-UUGUUG BBS motifs in the 3′UTR were co-transfected with a control firefly luciferase, and either empty vector (EV) or V5-tagged Brat cDNA. (F) The log₂ fold change in reporter expression caused by Brat for each reporter, relative to EV, is plotted for three independent experiments, each with three biological replicates. Biological replicates within the same experimental replicate are denoted (squares, circles and triangles). Mean ± standard deviation values are plotted. Statistically significant differences were determined via ANOVA and post-hoc Tukey–Kramer test, using a mixed linear model (n = 9, ***P < 0.001, n.s.: non-significant).

Figure 5.

Brat does not repress through the 5′UTR. (A) Four 5′-UUGUUG BBS motifs, or mutated (5′-UCC UUG) versions, were placed in the 5′UTR of Renilla luciferase reporters. (B) Luciferase assays testing the activity of Brat, or a mutant that does not bind RNA in vitro (N933A) on these reporters. For all reporters, this activity was compared to an empty vector (EV) negative control. At the bottom, western blot detection of V5-tagged wild type and mutant Brat proteins. Tubulin served as a loading control. Sizes of bands are in kDa. Mean ± standard deviation values are plotted, with replicate measurements (n = 3) within the same experimental replicate (n = 3) denoted (squares, circles and triangles). Significance of differences between conditions was determined via ANOVA with post-hoc Tukey–Kramer Test (***P < 0.001, n.s.: non-significant).

Figure 6.

Brat represses genes in the glycolysis pathway. (A) Log₂ fold change of expression of the indicated glycolytic enzymes significantly (P^adj > 0.05, baseMean ≥ 50) upregulated following RNAi of Brat, as measured by RNA-seq (n = 4). (B) Conversion of glucose to pyruvate via glycolysis (blue arrows) and conversion of oxaloacetate to glucose via gluconeogenesis (red arrows) in Drosophila. Additional metabolites such as trehalose, starch and ɑ-d-glucose also enter this pathway (black arrows). Finally, pyruvate is converted to acetyl-CoA and enters the TCA cycle (dashed arrow). Enzymes that catalyze each step of the pathway, that are expressed in DL1 cells, are represented by gene symbols. Genes significantly upregulated following Brat RNAi (discorectangle) and genes expressed but not significantly upregulated (rectangle). Note enzymes required for two steps of gluconeogenesis are not expressed in DL1 cells (dashed border). Created in BioRender. R. Connacher, 2024, BioRender.com/j19h070. Adapted from ‘Glycolysis and Glycolytic Enzymes’, by BioRender.com (2023). Retrieved from https://app.biorender.com/biorender-templates. Fly-specific genes from KEGG pathway dme00010. (C) Western blot verifying the depletion of endogenous, tagged V5::Brat from DL1 cells grown in normal media (+GlutaMax) and glutamine-starved media (−GlutaMax). Glucose-6-phosphate abundance in cells grown in normal media (D) or glutamine-starved media (E). Replicate measurements (n = 3) within the same experimental replicate (n = 4) are denoted (squares, circles, upward triangles and downward triangles). (F) Renilla luciferase reporters containing 3′UTRs of glycolytic enzymes. (G) These reporters were co-transfected with exogenous Brat under the control of an inducible promoter, and compared to corresponding empty vector (EV) samples. Mean ± standard deviation values are plotted, with replicate measurements (n = 3) within the same experimental replicate (n = 3) denoted (squares, circles, upward triangles). In panels (E)–(G), statistically significant differences were determined via ANOVA and Tukey–Kramer tests for post-hoc comparisons, utilized a mixed linear model (***P< 0.001, n.s.: non-significant).

Figure 7.

Brat represses V-ATPase complex subunits. (A) Structure of the human V-ATPase complex, PDB #ID 6WM2. The ATPase hexamer of subunits A and B and proton-transporting subunits c and c″ are surrounded by a scaffold of subunits G, E, C, H, a and e. Rotation is transferred between the ATPase and proton-transporting components by the central stalk of subunits D and d and F. Additional subunits include S1 and M8.9. (B) Log₂ fold change (± standard error) of V-ATPase subunits significantly (P^adj > 0.05, baseMean ≥ 50) upregulated following RNAi of Brat, as measured by RNA-seq (n = 4). (C) RNA levels of V-ATPase subunits and the metabolic gene Treh, relative to 7SK level, in heterozygous and brat mutant larvae determined via RT-qPCR. (D) Similarly, RNA levels of V-ATPase subunits were determined in larvae in which ubiquitous expression of Brat was induced (Da>Brat), compared to larvae only containing the Da-Gal4 driver (Da>). For all RT-qPCR, mean ± standard deviation values are plotted. Significant differences determined via Welch’s t-test (*P< 0.05, **P< 0.01, ***P< 0.001 and n = 4). (E) Endogenous Brat was depleted via RNAi, and the level of tagged Vha100-2 detected via Western blot. (F) Quantitation of band intensity via densitometry. Mean ± standard deviation values are plotted. Significance of differences were determined via ANOVA with a post-hoc Tukey–Kramer test, using a linear model (**P< 0.001 and n = 3). In all blots, sizes of bands are displayed in kDa.

Figure 8.

The 3′UTRs of V-ATPase subunits confer sensitivity to Brat (A) and (B) Renilla luciferase reporters containing 3′UTRs of V-ATPase subunits VhaAC39-1 and VhaM8.9 (A) or Vha100-2 and VhaPPA1-1 (B). These reporters were co-transfected with exogenous Brat under the control of an inducible promoter, and compared to corresponding empty vector samples. Mean ± standard deviation values are plotted, with replicate measurements (n = 3) within the same experimental replicate (n = 4) denoted (squares, circles and triangles). Statistically significant differences were determined via ANOVA and Tukey–Kramer tests for post-hoc comparisons, utilized a mixed linear model (***P< 0.001, n.s.: non-significant).