Cdk8 attenuates lipogenesis by inhibiting SREBP-dependent transcription in Drosophila

Cellular lipids, such as fatty acids, triglycerides, phospholipids and sterols derived from isoprenoids, play diverse and critical roles during the normal development of multicellular organisms. For example, triglycerides serve as the major energy storage molecules; phospholipids and sterols form membrane structures that are essential for compartmentation of eukaryotic cells; and phosphatidylinositols, cholesterol and its derivatives can also function as signaling molecules or hormones in multicellular organisms (Fahy et al., 2009). Metazoans can obtain these lipids from the diet or through de novo lipogenesis and cholesterogenesis. The sterol regulatory element-binding protein (SREBP) family of transcription factors play critical and conserved roles in regulating the transcription of enzymes required for lipogenesis and cholesterogenesis (Brown and Goldstein, 1997; Jeon and Osborne, 2012; Osborne and Espenshade, 2009; Rawson, 2003; Shimano and Sato, 2017).

In mammals, there are three SREBP family members: SREBP-1A, SREBP-1C and SREBP-2. SREBP-1A and SREBP-1C are produced from alternative splicing of the sterol regulatory element binding transcription factor 1 (SREBF1) transcripts, with SREBP-1A possessing an additional 24 amino acid residues in its transactivation domain, thus displaying stronger transcriptional activity than SREBP-1C (Goldstein et al., 2002; Shimano and Sato, 2017). SREBP-1C is broadly expressed in most tissues and plays a major role in nutritional regulation of the expression of lipogenic enzymes, such as acetyl-CoA carboxylase (ACC; the rate-limiting enzyme for de novo lipogenesis; also known as ACAC in mammals), fatty acid synthase (FASN; the key enzyme for lipogenesis) and acetyl coenzyme A synthase (AcCoAS; also known as ACSS in mammals) (Amemiya-Kudo et al., 2002; Desvergne et al., 2006; Shimano et al., 1999; Shimomura et al., 1999). In contrast, SREBP-1A is predominantly expressed in certain tissues and most cultured cells and regulates the expression of genes encoding both lipogenic and cholesterogenic enzymes. SREBP-2, encoded by the SREBF2 gene, regulates the expression of factors involved in cholesterol metabolism (Goldstein et al., 2002; Shimano and Sato, 2017). SREBP transcription factors are unusual in that full-length SREBPs are localized on the membrane of the endoplasmic reticulum (ER). These full-length SREBPs are precursors consisting of a bHLH-Zip DNA-binding domain at the N-terminus, a transmembrane domain, and a regulatory domain at the C-terminus (Osborne and Espenshade, 2009). In the presence of sterols (or unsaturated fatty acids in the case of SREBP-1 processing), full-length SREBPs associate with two integral membrane proteins – SREBP cleavage-activating protein (SCAP) and insulin-induced gene (INSIG) – in the ER. When the intracellular level of sterols or fatty acids is low, INSIG is released and SREBP-SCAP is transported to the Golgi apparatus, where the full-length SREBP precursors are cleaved by site-1 protease (S1P; also known as MBTPS1 in mammals) and site-2 protease (S2P; also known as MBTPS2 in mammals), resulting in the release of the N-terminal fragment with the bHLH-Zip DNA-binding domain (Goldstein et al., 2002; Rawson, 2003; Wu et al., 2022). These N-terminal SREBP fragments enter the nucleus and stimulate the expression of SREBP target genes. Accordingly, these mature forms of SREBPs are also known as nuclear SREBPs (nSREBPs; unless otherwise specified, this work focuses on the mature or nSREBPs). This model provides an elegant explanation of how SREBPs maintain the homeostasis of intracellular sterols and fatty acids (Goldstein et al., 2002; Shimano and Sato, 2017), but the molecular mechanism that modulates the transcriptional activities of SREBPs in the nucleus is still not fully understood.

Consistent with the key role of SREBPs in regulating the transcription of lipogenic and cholesterogenic factors, dysregulation of SREBPs is found in metabolic disorders and various human cancers (Guo et al., 2014; Moslehi and Hamidi-Zad, 2018). Given that elevated fatty acid biosynthesis and increased expression of lipogenic enzymes such as FASN and ACC have been identified as a near-universal feature of human cancers, therapeutic approaches have been developed to target the key lipogenic enzymes, such as ACC, FASN and stearoyl-CoA desaturase 1 (Baenke et al., 2013; Cheng et al., 2018; Guo et al., 2014; Menendez and Lupu, 2007; Shimano and Sato, 2017). Understanding the function and regulation of SREBPs may provide mechanistic insights into these diseases, thereby aiding developing new approaches to treatment.

Although the components of the SREBP pathway are conserved in metazoans, the regulatory mechanisms are much simpler in invertebrates such as Drosophila and Caenorhabditis elegans. There is only one SREBP ortholog in Drosophila, in addition to single orthologs of other factors that control SREBP processing, including SCAP, S1P and S2P (Goldstein et al., 2002; Rawson, 2003). Instead of being inhibited by intracellular sterols, the proteolytic procession of SREBP is regulated by phosphatidylethanolamine (Dobrosotskaya et al., 2002; Seegmiller et al., 2002), consistent with the notion that arthropods and nematodes are auxotrophic for sterols due to the lack of critical cholesterogenic enzymes (Carvalho et al., 2010; Rawson et al., 1999; Rosenfeld and Osborne, 1998). Interestingly, unsaturated fatty acids such as oleate, linoleate and arachidodate inhibited the procession of SREBP-1A and SREBP-1C, but not that of SREBP-2, in cultured human embryonic kidney (HEK) 293 cells (Hannah et al., 2001). Together with the additional evidence summarized below, these discoveries suggest that Drosophila SREBP plays a similar role to mammalian SREBP-1C in regulating de novo lipogenesis.

De novo lipogenesis is under tight regulation of physiological conditions such as feeding and fasting. Extensive studies have established that the insulin/mechanistic target of rapamycin (mTOR) signaling pathway is activated by dietary nutrients, such as carbohydrates and amino acids (Edgar, 2006; Jewell et al., 2013; Sabatini, 2017). Interestingly, downregulation of CDK8-CycC by insulin signaling is dependent on (mTOR complex 1) mTORC1 in cultured mammalian cells (Feng et al., 2015), while depleting Cdk8 or CycC abolishes the effects of mTOR activation on autophagosome formation in Drosophila (Tang et al., 2018). Moreover, Cdk8 and CycC mutant larvae are hypersensitive to high-sugar diets and high levels of certain dietary amino acids that stimulate Tor in Drosophila (Gao et al., 2018). Collectively, these results suggest that mTORC1 functions upstream of Cdk8-CycC (Fig. 1A) (Li et al., 2019).

We have previously reported that CDK8 and its regulatory partner cyclin C (CycC; also known as CCNC in mammals) negatively regulate lipogenesis by inhibiting the transcriptional activity of the nSREBPs in both Drosophila and mammals (Zhao et al., 2012). CDK8 and CycC are two conserved subunits of the Mediator complex, a transcription co-factor complex required for RNA polymerase II-dependent transcription. CDK8 and CycC, together with MED12 and MED13, form the CDK8 kinase module (CKM). The CKM binds to the small Mediator complex, which is composed of ∼26 subunits conserved in eukaryotes, and forms the large Mediator complex (Allen and Taatjes, 2015; Dannappel et al., 2018; Soutourina, 2018; Yin and Wang, 2014). Specifically, reduction of CDK8 and CycC increases the transcription of lipogenic genes and fat accumulation in Drosophila, cultured mammalian cells and mouse livers (Zhao et al., 2012). Using in vitro kinase assays, we previously identified a conserved threonine residue, Thr402 of human SREBP-1C (or Thr390 of Drosophila SREBP), as the phosphorylation site by CDK8 (Zhao et al., 2012). Moreover, we showed that depleting CDK8 or CycC in cultured HEK293 cells reduces the levels of ubiquitinated nSREBP-1A, but increases the total protein level of nSREBP-1A, suggesting that phosphorylation of nSREBPs by CDK8 destabilizes nSREBP through ubiquitin-dependent protein degradation (Zhao et al., 2012). However, the exact mechanisms of how CDK8 interacts with SREBP and the biological consequences of SREBP phosphorylation by CDK8 in vivo are still not fully understood.

Here, we further investigated the role of phosphodeficient SREBP (nuclear form of Drosophila SREBP-Thr390Ala, designated as the SREBP^T390A or SREBP^TA mutant) in regulating its stability and transcriptional activities in Drosophila. In comparison to wild-type nSREBP, SREBP^T390A mutant proteins are more stable and more potent in stimulating the transcription of SREBP target genes and fat accumulation in vivo. Further mapping of the interactions between Cdk8 and SREBP led to the identification of six amino acids at the N-terminus of SREBP that are essential for the direct interaction between SREBP and Cdk8. Interestingly, these six amino acids, conserved from flies to humans, are also required for the direct interactions between SREBP and the MED15 subunit of the small Mediator complex. Collectively, these results support a key role of Cdk8-meditated SREBP phosphorylation in regulating lipogenesis in vivo, suggesting that the concerted interplay between Cdk8 and MED15 is important for the tight control of the transcriptional activities of nSREBP.

In previous work, we characterized the effects of Cdk8 and CycC mutations on gene expression using microarray analyses of Cdk8^K185 and CycC^Y5 null mutant larvae (Zhao et al., 2012). However, this method holds several drawbacks such as high background owing to cross-hybridization, limited dynamic range of detection due to saturated signals, and limited specificity and sensitivity, especially for low-abundance transcripts (Jaluria et al., 2007). To combat the disadvantages of microarray analysis, we carried out transcriptome profiling of the Cdk8^K185 homozygous larvae using RNA-sequencing (RNA-seq) analyses (see Materials and Methods for details). In comparison to the control group, 4034 genes were significantly upregulated, and 2867 genes were significantly downregulated, in Cdk8^K185 mutants. These genes were categorized using the Gene Set Enrichment Analysis of Gene Ontology (GseGO) function of the clusterProfiler package (Yu et al., 2012). The top 40 Gene Ontology categories are shown as a dot plot for Cdk8^K185 (Fig. 1B). The fatty acid biosynthesis process was identified as one of the most significantly upregulated gene categories among these top 40 Gene Ontology categories (Fig. 1B), consistent with our previous report of elevated levels of fatty acid biosynthesis in Cdk8^K185 mutant larvae (Zhao et al., 2012).

To display the changes in genes involved in the fatty acid biosynthesis process at high resolution, we used a heatmap to show all the genes within the category of biological triplicates of Cdk8^K185 and the control (w¹¹¹⁸). The triplicates of the same genotypes were clustered together according to the x-axis clustering (Fig. 1C), which validates correct sampling of these biological replicates. Importantly, transcription of several key lipogenic genes, such as FASN1, FASN2, ACC and AcCoAS, is significantly elevated in Cdk8^K185 mutant larvae. These lipogenic genes are known direct transcriptional targets of SREBP, consistent with our previous report that SREBP is negatively regulated by CDK8 (Zhao et al., 2012). The significantly changed genes were then mapped into the fatty acid synthesis pathway with the Pathview package (Luo and Brouwer, 2013). Although not all the genes related to lipid biosynthesis were increased, genes encoding key lipogenic enzymes, marked in red in Fig. S1, were significantly increased. Thus, these RNA-seq analyses provide further support for the inhibitory effects of Cdk8 on SREBP-dependent lipogenic gene expression.

Our previous in vitro biochemical analyses suggest that CDK8 may inhibit lipogenesis by directly phosphorylating SREBP at a highly conserved threonine residue, Thr390 in Drosophila SREBP or Thr402 in human SREBP-1C (Fig. 2A) (Zhao et al., 2012). To test whether SREBP phosphorylation by Cdk8 regulates the stability and the transcriptional activity of SREBP in vivo, we analyzed the effects of mutating the Thr390 residue of SREBP on those properties of SREBP in Drosophila. Specifically, we generated two transgenic lines to ectopically express the nuclear form [amino acids (AA)1-451] of either wild-type SREBP (UAS-SREBP^WT) or the phosphodeficient SREBP (UAS-SREBP^T390A). The T390A mutation was validated by sequencing as shown in Fig. 2B. To reduce potential variations caused by the chromatin environments of the insertion sites, we used the pVALIUM10-roe vector, which flanks the expression cassettes with gyspy insulators and supports site direct insertion with attB integrase target sequence (Ni et al., 2009; Perkins et al., 2015). The constructs were then inserted in the attP2 genomic landing site on the third chromosome.

To test whether the stability of SREBP is affected in the phosphodeficient line, we performed a clonal analysis in the larval fat body using the cis-chromosomal recombination (FLP-out) system (Germani et al., 2018). Specifically, we constructed the larvae with the following genotypes: hs-flp/+; Actin>y⁺>Gal4, UAS-GFP/+; UAS-SREBP^WT/+ and hs-flp/+; Actin>y⁺>Gal4, UAS-GFP/+; UAS-SREBP^T390A/+. GFP-labeled adipocytes overexpress either wild-type SREBP (SREBP^WT) or SREBP^T390A mutant proteins, and the expression level of SREBP was detected by immunostaining using an anti-SREBP polyclonal antibody (Xie et al., 2015). Compared to the surrounding cells, ectopic expression of either form of SREBP in the GFP-positive adipocytes significantly increased the SREBP levels as expected (Fig. 2C,D). Importantly, a higher level of SREBP^T390A proteins compared to wild-type SREBP was observed in these clones (Fig. 2C′ versus D′), suggesting that phosphodeficient SREBP proteins are more stable than wild-type SREBP proteins in vivo.

To further test the effects of phosphodeficient SREBP on its stability, we ectopically expressed either wild-type or phosphodeficient SREBP proteins in the salary glands of third-instar larvae and then analyzed the levels of SREBP. Overexpression of either SREBP^WT or SREBP^T390A proteins in the salivary glands using Sgs3-Gal4 and C96-Gal4 causes lethality in the early larval stage (M.Z., X.L., and J.-Y.J., unpublished observations); thus, we used a weak salivary gland-specific Gal4 line, AbdB^RJ1-Gal4, which we generated in an unrelated project (R.-U.-S.H.-W. and J.-Y.J., unpublished data). This Gal4 line is weaker and more specific than other Gal4 lines that are active in the salivary glands, such as Sgs3-Gal4 and C96-Gal4. In addition, we generated a new set of transgenic lines by modifying the pVALIUM10-roe vector with 5xUAS, instead of the original 10xUAS (Ni et al., 2009; Perkins et al., 2015). To facilitate the detections, we also added a V5 epitope tag to the N-terminus of SREBP^WT or SREBP^T390A. The constructs were both targeted into the same attP2 loci to ensure the same genomic environment. These genetic tools allowed us to achieve low, but reliable, ectopic expression of SREBP proteins in salivary gland cells. As shown in Fig. 2F,F′, we detected weak SREBP protein expression in salivary glands cells, compared to the control, in which little anti-V5 signal was detected (Fig. 2E,E′). Compared to the salivary glands with ectopic expression of SREBP^WT-V5, we observed significantly stronger anti-V5 antibody staining in the nuclei of salivary glands with ectopic expression of SREBP^T390A-V5 (Fig. 2G-G″). These effects were quantified by measuring the nuclear-cytoplasmic ratio of the fluorescent intensities of the salivary gland cells (Fig. 2H). These observations indicate that phosphodeficient SREBP proteins are more stable than wild-type SREBP proteins.

We previously proposed a model to explain how CDK8-CycC inhibits SREBP-dependent gene expression, which posits that SREBP phosphorylation by CDK8 promotes the degradation of SREBP (Zhao et al., 2012). Consistent with this model, SREBP phosphorylation by Cdk8 increased the ubiquitination and degradation of SREBP (Zhao et al., 2012), and the E3 ligase SCF^Fbw7b is required for ubiquitination of SREBP (Sundqvist et al., 2005). Our new observations are consistent with the stabilization of the corresponding Thr402 to alanine mutation in human SREBP-1C (Zhao et al., 2012).

To test whether increased stability of SREBP^T390A correlates with more potent transcriptional activity at the cellular level, we generated a transgenic reporter line using a 4.4 kb promoter region of the Drosophila fatty acid synthase gene FASN1 to drive the expression of EGFP (Fig. 3A), designated as the FAS-EGFP reporter. We then genetically recombined this reporter with the Gal4/UAS flip-out system (Actin>y⁺>Gal4 UAS-RFP). Clonal depletion of SREBP in adipocytes, as marked with RFP, resulted in reduced EGFP levels compared with those of their neighboring cells (Fig. 3B,B′), suggesting that SREBP is required for the expression of this FAS-EGFP reporter. Conversely, overexpression of either SREBP^WT or SREBP^T390A significantly increased the levels of EGFP (Fig. 3C-D″), suggesting that SREBP is sufficient to drive expression of the FAS-EGFP reporter. Importantly, SREBP^T390A proteins are more potent than SREBP^WT in stimulating FAS-EGFP expression (compare Fig. 3D′ and C′), and the results from multiple clones are quantified and shown in Fig. 3E.

To further validate the effects of different forms of SREBP on its target gene transcription in a different tissue, we analyzed the expression of FANS1, AcCoAS and ACC in larval salivary glands. To directly visualize the transcripts of these genes, we used in situ hybridization chain reaction (HCR) RNA fluorescence in situ hybridization (RNA-FISH) imaging technology, a sensitive, quantitative and robust method that allows simultaneous imaging of specific mRNA transcripts at the cellular level (Choi et al., 2010, 2018; Dirks and Pierce, 2004). As expected, ectopic expression of SREBP^WT in salivary glands stimulated the expression of these SREBP target genes (compare Fig. 3G with the control shown in F). Compared to SREBP^WT, many more mRNA products of these genes were observed when SREBP^T390A was expressed (Fig. 3H; Fig. S2). Taken together, these observations show that the phosphodeficient SREBP is more potent than SREBP^WT in activating the transcription of SREBP target genes.

To investigate the function of phosphodeficient SREBP in promoting lipogenesis, we used a similar FLP-out system to generate somatic clones in the larval fat body, marked by GFP, to overexpress either SREBP^WT or SREBP^T390A, and then analyzed lipid accumulation by staining lipid droplets using Nile Red. As shown in Fig. 4A,A′, clonal expression of SREBP^WT in cells marked with GFP had no obvious effects on lipid droplets compared to their neighboring adipocytes, presumably because mature SREBP proteins are maximally activated in adipocytes at the wandering larval stage, or the ectopically expressed SREBP^WT proteins are actively degraded. In contrast, overexpressing phosphodeficient SREBP^T390A resulted in stronger Nile Red staining in adipocytes marked with GFP, compared to their neighboring adipocytes (Fig. 4B,B′).

The abundant lipid droplets accumulated in adipocytes of the late-third-instar larvae resulted in a high baseline level of lipid staining, making it challenging to make a quantitative comparison between the two forms of SREBP protein based on lipid accumulation. Thus, we tested these effects in salivary gland cells, which normally accumulate few or no detectable lipid droplets (Fig. 4C). Ectopic expression of SREBP^T390A in the salivary gland using AbdB^RJ1-Gal4 to overexpress SREBP^WT significantly increased the number of lipid droplets (Fig. 4D, quantification is shown in F). Compared to SREBP^WT, we observed significantly more lipid droplets per cell when SREBP^T390A was ectopically expressed using the same approach (Fig. 4E,F). Collectively, these results show that phosphodeficient SREBP^T390A proteins are more potent at stimulating lipogenic gene expression and lipid accumulation than the wild-type SREBP in vivo.

To further validate the effects of different forms of SREBP on FASN1 transcription in a biological context, we carried out a similar experiment in the adult intestine and analyzed the effects of starvation. Starvation is expected to inhibit mTOR, which correlates with elevated Cdk8 protein levels in Drosophila larvae (Xie et al., 2015). This model predicts that elevated Cdk8 phosphorylates and stimulates the degradation of nSREBP^WT, but not phosphodeficient SREBP^T390A. If so, only SREBP^T390A is expected to stimulate the expression of the FAS-EGFP reporter under starvation conditions (Fig. 1A). To test this prediction and to avoid potential positional effects in the midgut for our comparison, we focused our analyses on a posterior region of the midgut, near the midgut and hindgut boundary, in adult female flies. The expression of FAS-EGFP in cells within this region was low (Fig. 4G). As expected, overexpression of either SREBP^WT or SREBP^T390A in cells marked with RFP (Fig. 4G′,H′) stimulated the expression of the FAS-EGFP reporter under a normal feeding condition (Fig. 4G,H). However, under starvation conditions, ectopic expression of SREBP^WT in cells marked with RFP (Fig. 4I′) failed to stimulate FAS-EGFP expression; its level was similar to that in the neighboring cells (Fig. 4I). In contrast, ectopically expressing the phosphodeficient SREBP^T390A (the cell marked with RFP, Fig. 4J′) can still stimulate the expression of the FAS-EGFP reporter compared to that in neighboring cells (Fig. 4J). These observations suggest that both SREBP^WT and phosphodeficient SREBP^T390A can stimulate lipogenic gene transcription in vivo, but only SREBP^T390A is resistant to starvation in this process. These results further support the key function of Cdk8-dependent phosphorylation of SREBP in the context of nutritional regulation of de novo lipogenesis (Fig. 1A).

The N-terminus of human SREBP-1A (AA30-40) fused with GST can pull down several Mediator subunits, including CDK8, MED1, MED6 and MED15, from the HeLa cell nuclear extract (Yang et al., 2006). However, it is unclear whether Cdk8 directly interacts with SREBP, or indirectly through the small Mediator complex, and whether the physical interactions between SREBP and MED15 are conserved in Drosophila. To test whether Cdk8 directly interacts with SREBP and to further map the specific regions mediating this interaction, we performed three rounds of GST pull-down analyses.

Specifically, we generated polyhistidine (His)-tagged Cdk8 (AA1-262), containing the ATP-binding site, A-loop, CycC interface and substrate-binding domain (Schneider et al., 2011; Xu et al., 2014). In addition, we generated GST-fusion proteins by dividing SREBP into three partially overlapping fragments (Fig. 5A), designated as GST-SREBP-1 (AA1-250), GST-SREBP-2 (AA201-350, containing the bHLH-Zip domain) and GST-SREBP-3 (AA301-451), respectively. As shown in Fig. 5B, His-Cdk8-N (AA1-262) interacted with GST-SREBP-1, but not GST-SREBP-2 or GST-SREBP-3, suggesting that AA1-200 of SREBP interacts with Cdk8 in vitro. Next, we generated three smaller and partially overlapping fragments of GST-SREBP-1 and carried out a similar GST pull-down assay. Only GST-SREBP-1-A (AA1-100; Fig. 5A), but not GST-SREBP-1-B (AA51-200) or GST-SREBP-1-C (AA151-250), interacted with His-Cdk8-N (Fig. 5C), suggesting that AA1-50 of SREBP directly binds to Cdk8. In our third round of mapping, we observed that GST-SREBP-1-A-2 (AA1-50) and GST-SREBP-1-A-3 (AA1-75), but not GST-SREBP-1-A-1 (AA1-25), interacted with His-Cdk8-N (Fig. 5A,D). These observations show that the AA25-50 at the N-terminus of SREBP directly interacts with Cdk8 in vitro, which is consistent with the notion that SREBP protein is a direct target of Cdk8.

The nSREBP contains a structured bHLH-Zip DNA-binding domain (AA284-334), but other parts of the nSREBP are largely intrinsically disordered (Fig. 6A). This is perhaps a reason why only the structure of the DNA-binding domain of human nSREBP-1A has been resolved using X-ray crystallography (Parraga et al., 1998). Given that the AA30-40 region of human SREBP-1A can directly interact with the KIX domain of MED15 (Yang et al., 2006), we tested whether the interaction is conserved in Drosophila. As shown in Fig. 6B, GST-SREBP-1-A-3 (AA1-75) can also pull down His-tagged full-length Drosophila MED15 protein, suggesting that the interaction between the N-terminus of SREBP and MED15 is evolutionarily conserved. Using the yeast two-hybrid assay, we validated the interactions between SREBP AA1-50 and full-length Cdk8 or MED15 (Fig. 6C). Notably, these analyses suggest that the same region of SREBP (AA25-50) can directly interact with Cdk8 or MED15 (Fig. 5D, Fig. 6B,C). These interactions may have important implications in our understanding of the dynamic processes that involve the small Mediator complex and the CKM in SREBP-dependent gene expression (see Discussion).

To identify the specific amino acids within the N-terminus of SREBP that are required for direct interaction with CDK8 or MED15, we analyzed the intrinsically disordered regions of SREBP from different species. We used the ANCHOR2 program, which can robustly predict protein disorder based on an energy estimation approach, and can also detect regions that probably gain energy by interacting with globular proteins (Dosztanyi et al., 2009). As shown in Fig. 6D, we used a heatmap to visualize the ANCHOR2 scores (ASs) of the N-terminus of SREBP proteins from Drosophila melanogaster, zebrafish (Danio rerio; SREBP1), frog (Xenopus laevis; SREBP1), chicken (Gallus gallus; SREBP1), mouse (Mus musculus; SREBP-1A) and human (Homo sapiens; SREBP-1A). Interestingly, these analyses revealed a hot spot overlapping from AA35-44 across the species (Fig. 6D), suggesting that this region has higher potential to interact with other proteins. Within this region, six amino acids (DMLDII, AA36-41), are highly conserved during evolution (Fig. 6E).

To test whether mutating these six amino acids (DMLDII) can alter the protein-binding potential in this disordered region of the SREBP, we replaced them with six alanines from different species in silico and then analyzed the ASs of these mutant SREBPs using ANCHOR2, a web-based application developed for predicting disordered protein-binding regions (Dosztanyi et al., 2009). As shown in Fig. 6D, the AS hotspot is diminished within this region, indicating that these six conserved amino acids are important for the SREBP, Cdk8 and MED15 interactions. To verify this prediction, we mutated these six amino acids to six alanines in the GST-SREBP-1-A-3 (AA1-75) fragment, and then performed GST pull-down assays. As shown in Fig. 6B (lane 3), the mutation of these conserved six amino acids in GST-SREBP-1-A-3 abolished its interaction with both MED15 and Cdk8. These observations suggest that the conserved region within the N-terminus of SREBP, particularly the highly conserved but disordered region DMLDII (AA 36-41), is crucial in mediating the interactions between SREBP and Cdk8, or between SREBP and MED15.

Previous studies have provided an elegant model to explain how SREBPs control intracellular fatty acid and cholesterol homeostasis. These studies mainly focus on understanding the mechanisms of how the full-length SREBP precursors are processed in the cytoplasm in response to intracellular levels of sterols and fatty acids (Brown and Goldstein, 1997; Jeon and Osborne, 2012; Osborne and Espenshade, 2009; Rawson, 2003; Shimano and Sato, 2017). However, exactly how the nSREBP proteins activate the transcription of lipogenic and cholesterogenic genes, and how this process is turned off in the nucleus are still not fully understood. In this study, we analyzed the in vivo effects of Cdk8 phosphorylation of SREBP on lipogenic gene transcription and de novo lipogenesis, and then further mapped the specific regions of SREBP that directly interact with Cdk8 and MED15 using in vitro and in silico approaches.

Post-translational modifications of SREBPs, such as acetylation and phosphorylation, play critical roles in fine-tuning SREBP activities (Raghow et al., 2019; Zhao and Yang, 2012). Using in vitro kinase assays, we previously identified one specific phosphorylation site of SREBP by CDK8, i.e. Thr402 of human SREBP-1C, or Thr390 of Drosophila SREBP (Zhao et al., 2012). Further biochemical analyses using cultured mammalian cells suggest that Thr402 phosphorylation negatively regulates human SREBP-1C (Zhao et al., 2012). In this study, we investigated the biological consequences of this phosphorylation in Drosophila, particularly its effects on the stability of SREBP, as well as the activity of SREBP in stimulating target gene expression and lipid accumulation. Compared to the wild-type SREBP, SREBP^T390A phosphodeficient proteins were more stable, and were more potent in stimulating SREBP target gene expression and lipid accumulation in vivo. These results suggest that Cdk8 is a key kinase that restrains the transcriptional activity of SREBP in the nucleus (Fig. 1A).

Besides CDK8, previous studies using different experimental systems have shown that SREBPs can be phosphorylated at multiple threonine and serine residues by a number of protein kinases, including AMPK, GSK-3β, ribosomal protein S6 kinase (p70S6K), p38ERK, PKA and c-JUN N-terminal protein kinase (JNK) (Dong et al., 2015, 2014, 2020; Knebel et al., 2014; Li et al., 2011; Lu and Shyy, 2006; Punga et al., 2006). As discussed below, p70S6K phosphorylation of SREBP-1C positively regulates the processing of full-length SREBP-1C and stimulates nSREBP-1C activity (Dong et al., 2020), while phosphorylation by other kinases inhibits SREBP-1 activity (Fig. S3A). To consider the evolution implications of these mechanisms, we aligned the amino acid sequences of SREBP-1A and SREBP-1C homologs in a few representative species, including Drosophila melanogaster, zebrafish, Xenopus, chicken (Gallus gallus), mouse, rat and human. Of all the reported phosphorylation sites, the threonine residue phosphorylated by CDK8 (Fig. S3B), as well as several threonine/serine residues phosphorylated by GSK-3β and p70S6K (Fig. S3B,C), appear to be highly conserved from flies to humans. GSK-3β can phosphorylate human SREBP-1A at Thr426 (equivalent to Thr402 in human SREBP-1C) and Ser430 (Fig. S3B), thereby promoting ubiquitination and degradation of SREBP-1A (Punga et al., 2006; Sundqvist et al., 2005).

Similarly, rat SREBP-1C can be phosphorylated at Ser418, Ser419 and Ser422 residues upon insulin treatment (Fig. S3C), and p70S6K, a kinase downstream of insulin signaling, was shown to enhance SREBP-1C precursor maturation and protect SREBP-1C from proteasomal degradation (Dong et al., 2020). These phosphorylation sites or serine/threonine-rich regions seem to be conserved, indicating mechanistic conservation of SREBP phosphorylation by p70S6K in Drosophila and humans (Fig. S3C), which remains to be experimentally validated.

In contrast to Cdk8, GSK-3β and p70S6K, the phosphorylation sites of AMPK, p38ERK, PKA and JNK are not conserved in Drosophila. For example, it was reported that activation of AMPK by polyphenols and metformin can inhibit lipogenesis by phosphorylating SREBP-1C at Ser372, thereby suppressing SREBP-1C cleavage and nuclear translocation in mice and cultured human HepG2 hepatocytes (Li et al., 2011). The Ser372 residue of human SREBP-1C is conserved in lancelets and vertebrates, but the sequence in that region is not conserved in Drosophila (Fig. S3D). Similarly, Ser73 of rat SREBP-1C can be phosphorylated by GSK-3β, resulting in the dissociation of the SREBP-1C-SCAP complex and ubiquitination-dependent proteasomal degradation of SREBP-1C (Dong et al., 2015). However, the Ser73 residue appears to be only conserved in SREBP proteins of certain vertebrate species (Fig. S3E). In addition, Ser117 of human SREBP-1A can be phosphorylated by ERK and JNK, and mutating this phosphorylation site protects mice under normocaloric conditions from developing enlarged fatty livers (Kotzka et al., 2012), yet the Ser117 residue is also only conserved in mammals (Fig. S3F). Moreover, p38MAPK, ERK and JNK can also phosphorylate Ser39 and Thr402 of human SREBP-1C (Knebel et al., 2014). Taken together, these discoveries have revealed potential variations in the detailed molecular mechanisms of how insulin signaling stimulates SREBP-dependent lipogenic gene expression and promotes de novo lipogenesis during the evolution of metazoans. To our knowledge, the role of GSK-3β, NJK, p38ERK and p70S6K in regulating SREBP activity has not been tested in Drosophila and other invertebrates. Based on amino acid sequence conservation of SREBPs, phosphorylation of SREBP by CDK8, GSK-3β and p70S6K might be a more ancient regulatory mechanism than phosphorylation of SREBP by AMPK and MAPKs. Compared to in invertebrates, such as insects, additional kinases that regulate SREBP and lipogenesis may offer adaptive advantages in coping with more complex physiological needs that require fine-tuned control of the activities of the SREBP family of transcription factors in vertebrates or mammals.

Previous studies and our new observations suggest a physiological context for the role of Cdk8 in inhibiting SREBP and lipogenesis (Fig. 1A). Insulin signaling and amino acids can activate mTORC1, which can downregulate CDK8 or CycC (Feng et al., 2015; Xie et al., 2015). For simplicity, the following amino acids are numbered based on the equivalent sites in Drosophila SREBP. On the one hand, mTORC1 activates p70S6K, which can phosphorylate the SREBP precursors at Ser423, Ser424 and Ser429, thereby protecting SREBP from proteasome degradation and promoting its processing into the nuclear form of SREBP (Dong et al., 2020) (Fig. 1A; Fig. S3C). On the other hand, activated mTORC1 can downregulate CDK8 (Feng et al., 2015; Xie et al., 2015), which allows mature SREBP to stimulate lipogenic gene expression (Zhao et al., 2012) (Fig. 1A). According to this model, under feeding conditions, Cdk8 can no longer inhibit SREBP, thus the gain of either wild-type SREBP or the SREBP^T390A mutant can equally stimulate the expression of SREBP target genes (Fig. 4G,H).

Conversely, we have observed that the level of Cdk8 proteins is significantly elevated upon starvation in Drosophila larvae, with a concurrent reduction of the nSREBP proteins (Xie et al., 2015). This increase in CDK8 proteins is likely due to the inactivation of mTORC1 (Feng et al., 2015). In addition, it has been reported that starvation or overexpression of TSC1/TSC2 inhibits the activity of mTORC1 and p70S6K (Blagosklonny, 2011; Garami et al., 2003). Low p70S6K activity may lead to hypophosphorylation of Ser423, Ser424 and Ser429 residues of SREBP, thereby reducing the efficiency of the procession of full-length SREBP (Dong et al., 2020) (Fig. 1A). On the other hand, increased CDK8 can phosphorylate SREBP at Thr390 (Zhao et al., 2012), which may be subsequently phosphorylated by GSK-3β at Ser394 (Punga et al., 2006), thereby promoting SREBP degradation (Punga et al., 2006; Zhao et al., 2012) (Fig. 1A). Thus, overexpressed wild-type SREBP can be quickly eliminated during starvation, resulting in unchanged FAS-EGFP expression (Fig. 4I). However, the effect of starvation on SREBP activity is abolished by mutating the Thr390 residue, as the phosphodeficient form of SREBP can still activate FAS-EGFP transcription under starvation conditions (Fig. 4J). These observations suggest that Cdk8 plays a critical role in inhibiting SREBP-dependent lipogenesis during starvation, and this mechanism is repressed by downregulating Cdk8 protein through mTORC1 under feeding conditions (Fig. 1A). Given that both Cdk8 and GSK-3β are shown to phosphorylate Thr390 of SREBP, it is necessary to further investigate how these different kinases coordinate with each other in phosphorylating SREBP and lipogenesis in different biological contexts in vivo.

Our biochemical and yeast two-hybrid analyses suggest that either Cdk8 or MED15 can directly interact with a small fragment within the N-terminus of Drosophila SREBP. Mutation of six amino acids (DMLDII, AA36-41) in Drosophila SREBP abolished its interaction with either Cdk8 or MED15, suggesting that these six amino acids are essential for mediating these physical interactions. These results further extend the previous report that SREBP in C. elegans can interact with MED15 in both C. elegans and mammalian cells, and that MED15 is required for SREBP-activated gene expression (Yang et al., 2006).

It is unclear how Cdk8 interacts with the N-terminus of SREBP in Drosophila, yet the phosphorylation site (Thr390) is close to the C-terminus of mature SREBP. Except the bHLH DNA-binding domain, most regions of nSREBP, including both the N-terminus and C-terminus, seem to be intrinsically disordered (Fig. 6A). We speculate that the interaction between Cdk8 and the N-terminus of SREBP causes conformational change and brings the kinase domain of Cdk8 close to the C-terminus of SREBP, thereby enabling Cdk8 to phosphorylate the Thr390 residue. Despite the biochemical and in silico analyses presented in this study and previous works (Yang et al., 2006; Zhao et al., 2012), evidence showing direct interactions among these proteins in vivo is still lacking. Perhaps, solving the structure of the SREBP-CDK8 complex using cryogenic electron microscopy, or three-dimensional structural and simulation analyses, may provide structural insights into this process.

There are several possible scenarios to explain the dynamic interplay among SREBP, Cdk8 and MED15. Given the current understanding of the role of Mediator complexes and how the CKM regulates the transcription of a few transactivators, the most parsimonious model is that the mature SREBP homodimer binds to the sterol regulatory elements (SREs) of its target genes, and the N-terminus of SREBP directly interacts with the MED15 subunit of the small Mediator complex, which interacts with RNA polymerase II and other general transcription factors, allowing for transcription initiation (Fig. 7). Prior studies suggest that the CKM is recruited to the transcription start sites shortly after the transcription initiation, thereby limiting the subsequent rounds of transcription re-initiation (Fryer et al., 2004; Hahn, 2004). Exactly how the CKM is subsequently recruited to the promoter remains unclear. Interestingly, recent analyses of the budding yeast CKM structure using cryo-electron microscopy have revealed a PIWI domain in MED13 (Li et al., 2021). PIWI proteins bind to RNAs and are known for their roles in regulating RNA interference and transposon silencing (Ross et al., 2014). Thus, we speculate that nascent mRNAs interact with MED13 via its PIWI domain, which facilitates CKM recruitment to the promoter region, where the CKM binds to the small Mediator complex, while CDK8 interacts with specific transactivators such as SREBP. This could explain how the CKM is recruited to promoters after transcription elongation, thereby limiting transcription re-initiation. It is also unclear whether CDK8 directly competes with MED15 in binding to the same N-terminus of SREBP or to two different N-termini of the SREBP homodimer. Subsequently, CDK8 phosphorylates the Thr402 of human SREBP-1C or Thr390 of Drosophila SREBP, thereby reducing SREBP binding to the SREs, favoring the export of phospho-SREBPs to the cytoplasm for degradation. In this model, MED15 is necessary for SREBP-activated expression of lipogenic genes, whereas Cdk8 is required for turning this process off (Fig. 7). This model can explain the biological consequences that we observed in Cdk8 and CycC mutants, and our analyses of SREBP^T390A mutant proteins provide further support for the dynamic model described above.

A potential caveat of this study is that the experimental evidence for the direct phosphorylation of Drosophila SREBP on Thr390 by Cdk8 is lacking; this assumption is inferred based on an in vitro CDK8 kinase assay using human SREBP-1C as the substrate and the conservation of the threonine residue in difference species (Zhao et al., 2012). Owing to the lack of phosphospecific antibodies for the Drosophila SREBP Thr390 residue, it is also unknown whether SREBP is phosphorylated at the Thr390 residue in vivo, and whether manipulation of Cdk8 activity in vivo alters the phosphorylation status of this site. Therefore, we still cannot establish beyond doubt that SREBP is a bona fide substrate of Cdk8. However, the functional analyses of the phosphodeficient SREBP summarized in this work further improve confidence for the inhibitory effects of Cdk8 on SREBP and de novo lipogenesis, which we proposed previously based on genetic and biochemical analyses in Drosophila and mammalian systems (Zhao et al., 2012).

Taken together, this work advances our understanding of how mature SREBP activates its target gene transcription and how this process is turned off. It serves as yet another example of how transcription activation is tightly coupled with its inactivation and explains why transactivators generally have a short half-life (Tansey, 2001). Perhaps, further work using fluorescent-tagged subunits of the Mediator complex and live imaging will help to define the concerted and dynamic actions of the small Mediator complex and the CKM in the tight control of SREBP activities in the nucleus and SREBP-dependent lipogenesis.

Flies were raised at 25°C on a standard cornmeal, molasses and yeast medium. To construct the UAS-SREBP transgenic lines, we first amplified the DNA sequence of the mature SREBP using high-fidelity PrimeSTAR Max DNA polymerase (Takara, R045A) and primers SREBP-5.10 and SREBP-3.10 (Table S1). After purification, the PCR fragment was inserted into the pENTR/D-TOPO vector (Thermo Fisher Scientific, K240020), and then amplified in Escherichia coli strain DH5α. The T390A mutation was generated with a QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies, 200523) with primers SREBP-TA-5.1 and SREBP-TA-3.1 (Table S1). After validation by sequencing, the two pENTR constructs were recombined into the pVALIUM10-roe vector (https://fgr.hms.harvard.edu/trip-plasmid-vector-sets), which contains the attB sequence for site-specific insertion into the Drosophila genome (Perkins et al., 2015) using Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific, 11791100), following the manufacturer's instructions. The amplified constructs were micro-injected into the integrase-expressing fly strain with attP2 site (BL-25710: P{y[+t7.7]=nos-φC31^NLS}X, y¹ sc¹ v¹ sev²¹; P{y[+t7.7]=CaryP}attP2) using the service provided by Rainbow Transgenic Flies (Camarillo, CA, USA). To achieve potent expression of genes of interest, the original pVALIUM10-roe vector contains ten Gal4-binding sites (10xUAS) with a 5xUAS sequence flanked by two loxP sites (Perkins et al., 2015). We observed that transgenic lines using 10xUAS vector to ectopically express SREBP were too strong and caused early larval lethality when driven with salivary-specific Gal4 lines. Thus, we removed this 5xUAS cassette using the bacteriophage Cre recombinase (NEB, M0298S) in vitro and generated 5xUAS-SREBP^WT and UAS-SREBP^T390A transgenic lines with these constructs inserted at the attP2 landing site. To facilitate detection of these SREBP proteins, we added a V5 tag to the N-terminus of these proteins; primer sequences are included in Table S1.

The first-instar larvae were heat shocked at 37°C for 2 min, and then raised at 25°C until the late-third-instar wandering stage. The fat body was dissected in phosphate-buffered saline (PBS) and fixed in 4% formaldehyde at room temperature for 10 min. For FAS-EGFP imaging, the fat body was incubated with 1 μM 4′,6-diamidino-2-phenylindole (DAPI) in PBS at room temperature for 10 min, washed three times with PBS, and then mounted in VECTASHIELD antifade mounting medium (Vector Laboratories, H-1000). For Nile Red staining, the fat body was washed three times with PBS-Tween 20 for 10 min each, and then incubated with 100 ng/ml Nile Red at room temperature for 30 min, followed by DAPI staining and mounting as described above for FAS-EGFP. We used ImageJ to quantify the levels of FAS-EGFP, and P-values were calculated using one-tailed unpaired t-tests.

To generate clones in intestines, we first starved 3-day-old adult flies of specific genotypes (as described in the Results section) by keeping them in vials containing 1.5% agar for 24 h, and then performed heat-shock treatment at 37°C for 40 min. These animals were then kept in agar vials for a further 6 h before dissection at 25°C. The control group was kept in vials with normal food but heat shocked and dissected at the same time point. Intestines of these animals were dissected, fixed, stained with DAPI and mounted as described above. Confocal images were taken using a Nikon Ti Eclipse or a Zeiss LSM900 confocal microscope system and then processed by Adobe Photoshop 2021 software.

Salivary glands from third-instar larvae at the wandering stage were dissected in PBS and fixed in 4% formaldehyde for 10 min at room temperature. After rinsing, the salivary glands were then incubated with BODIPY 493/503 (1 µg/ml, Invitrogen, D3922; diluted in PBS) for 2 h at room temperature. After rising three times in PBS (10 min each), the tissues were mounted in VECTASHIELD medium.

For immunostaining, we used an anti-SREBP antiserum (Xie et al., 2015). Anti-SREBP antibody was purified using Dynabeads Protein A (Thermo Fisher Scientific, 10001D). The fixed fat body was permeabilized with PBS-0.1% Triton X-100 three times, 10 min each, and blocked with PBS+0.1% Triton X-100+5% normal goat serum+0.2% bovine serum albumin (PBS-Triton X-100-NGS-BSA) at room temperature for 1 h. The fat body was then incubated with anti-SREBP (1:100 diluted in PBS-Triton X-100-NGS-BSA) at 4°C overnight. After rinsing with PBS-Triton X-100 three times, the tissues were incubated with a secondary antibody (Alexa Fluor 488 AffiniPure donkey anti-rabbit, Jackson Immunological Laboratories, 711-545-152; 1:1000 diluted in PBS-Triton X-100-NGS-BSA) at room temperature for 1 h, followed by the same DAPI staining, rinsing and mounting steps as described above. Anti-V5 (Invitrogen, R960-25; 1:500) was used for V5-tagged SREBP staining, with secondary antibody (Alexa Fluor 488 AffiniPure goat anti-mouse, Jackson Immunological Laboratories, 115-545-003; 1:1000). ImageJ was used for quantifying the fluorescent intensity of single-plane confocal images.

The total RNA from ten third-instar larvae of w¹¹¹⁸ (as the wild-type control) and Cdk8^K185 homozygotes at the wandering stage (triplicate for each genotype) was extracted using 1.0 ml TRIzol Regent (Invitrogen), and then purified using an RNeasy Mini Kit (Qiagen, 74104) following the manufacturer's instructions. The preparation and sequencing of the RNA libraries, as well as quality control and adapter trimming, were performed by the Texas A&M University (TAMU) Genomics and Bioinformatics Service using an Illumina HiSeq 2500 following the standard protocols. The processed files were uploaded to and analyzed on the TAMU High Performance Research Computing cluster Terra. RNA-seq sequences were aligned to the Drosophila melanogaster genome FlyBase release 6.30 using STAR open source software (Dobin et al., 2013). The gene counts were generated by the featureCounts function from the Subread package (Liao et al., 2014). Differential expression analysis was performed using the DESeq2 R package (Love et al., 2014), with a Benjamini–Hochberg method adjusted P-value of 0.05 used to identify statistically significant differentially expressed genes. The heatmaps were generated by the pheatmap R package (https://cran.r-project.org/web/packages/pheatmap/pheatmap.pdf), with normalized gene counts provided by DESeq2. The clusterProfiler R package was used to perform Gene Ontology analysis of significantly altered genes associated with FlyBase gene ID (Yu et al., 2012). The Benjamini–Hochberg multiple testing correction method was used, and the output dot plot was generated by the enrichplot R package (https://bioconductor.org/packages/release/bioc/html/enrichplot.html).

Five pairs of salivary glands from the third-instar larvae were dissected in PBS and then fixed in 4% formaldehyde for 20 min at room temperature. Fixed salivary glands were then rinsed three times in 1× PBS with 0.1% Tween 20 (PBST), 5 min each. After rehydrating the salivary gland in 100%/75%/50%/25% methanol in PBST, the samples were washed three times with PBST and then incubated in 10 μg/ml proteinase K for 5 min. Following three washes in PBST, the samples were fixed again in 4% formaldehyde for 20 min. For the rest of the steps, we followed the multiplexed HCR RNA-FISH protocol provided by Molecular Instruments (https://files.molecularinstruments.com/MI-Protocol-RNAFISH-GenericSolution-Rev7.pdf) with the following modifications: we added DAPI for the first 30 min of 5× sodium chloride sodium citrate with 0.1% Tween 20 washing, then mounted the samples in Antifade Mounting Medium (VECTASHIELD) for microscopy. The probe set for FANS1 (lot number PRQ415) was used with B1-Alexa Fluor 488 amplifiers, the probe set for AcCoAS (lot number PRQ416) was used with B2-Alexa Fluor 594 amplifiers, and the probe set for ACC (lot number PRQ417) was used with B3-Alexa Fluor 647 amplifiers (all purchased from Molecular Instruments) in multiplexed in situ HCR. Confocal images were taken using a Zeiss LSM900 confocal microscope system and then processed by Adobe Photoshop CS6 software.

Cdk8 (AA1-262), MED15 (full length) and SREBP fragments were amplified with PrimeSTAR Max premix (Takara, R045A), using the primers listed in Table S1. These fragments were inserted into pENTR/D-TOPO vector (Thermo Fisher Scientific, K240020). pENTR-SREBP-1-A-3 was mutated by a QuikChange II Site-Directed Mutagenesis Kit (Agilent, 200523) using primers SREBP-mut-5.1 and SREBP-mut-3.1 (Table S1). The Cdk8 and MED15 fragments in pENTR vectors were recombined into the pDEST17 vector (N-terminal 6XHis tag) with Gateway LR Clonase II Enzyme mix (Thermo Fisher Scientific, 11791020). Similarly, the SREBP fragments, including mutated pENTR-SREBP-1-A-3, were recombined into the pDEST15 vector (N-terminal GST tag). The constructs were transformed into E. coli strain Rosetta for protein expression using standard protocols. Purification of GST-tagged proteins and the GST pull-down assays were performed using the same protocol as described previously (Li et al., 2020).

The yeast two-hybrid assay was performed as described before using AH109 yeast strain (Li et al., 2020). pENTR-SREBP-1-A-2 (SREBP1-50) was recombined into pGBKT7-GW vector with Gateway LR Clonase II Enzyme mix, and pENTR-Med15 (full length) was recombined with the pGADT7-GW vector. The pGADT7-CDK8 plasmid was described previously (Li et al., 2020). The SD/−Leu/−Trp/−His medium (MP Biomedicals) was supplemented with 2.5, 5 or 10 mM 3-amino-1,2,4-triazole (3-AT; Sigma-Aldrich, A8056).

The amino acid sequences of SREBP proteins from different species were downloaded from the National Center for Biotechnology Information (NCBI) Reference Sequence (RefSeq) database and aligned using Clustal Omega with default parameters (https://www.ebi.ac.uk/Tools/msa/clustalo/) (O'Leary et al., 2016; Sievers et al., 2011). The SREBP sequences of different species in FASTA format were individually uploaded to https://iupred2a.elte.hu/ and performed ANCHOR2 and IUPred2A predictions with default parameters (Dosztanyi et al., 2009; Mészáros et al., 2018). The heatmaps were drawn by pheatmap R package (https://cran.r-project.org/web/packages/pheatmap/pheatmap.pdf), based on the predicted scores.

P-values were calculated using one-tailed Type 3 t-tests, and error bars indicate s.d. The numbers of samples analyzed are indicated in Fig. 2H and Fig. 3E.

We thank Jin Sun and Jian-Quan Ni for helping us generate the FAS-EGFP reporter, Liya Pi for the AH109 yeast cell stock, the Bloomington Drosophila Stock Center (NIH P40OD018537) for fly strains, and the Texas A&M University Genomics and Bioinformatics Service and the Texas A&M High Performance Research Computing for RNA-seq analyses. We also thank Sarah Bondos for advice on analyses of intrinsically disorder proteins, and Fajun Yang, Robert Schofner and Jasmine Sun for critical comments on the manuscript.