Elongin functions as a loading factor for Mediator at ATF6α-regulated ER stress response genes

Significance

Activation of gene expression by the transcription factor ATF6α is a critical downstream event in the endoplasmic reticulum (ER) stress response. The findings presented here shed light on the mechanism underlying ATF6α-dependent transcription activation. In particular, they identify a pathway for recruitment of Mediator and the RNA polymerase II transcription machinery to promoters, as well as an unexpected role for the transcription factor Elongin in this process.

Abstract

The bZIP transcription factor ATF6α is a master regulator of endoplasmic reticulum (ER) stress response genes. In this report, we identify the multifunctional RNA polymerase II transcription factor Elongin as a cofactor for ATF6α-dependent transcription activation. Biochemical studies reveal that Elongin functions at least in part by facilitating ATF6α-dependent loading of Mediator at the promoters and enhancers of ER stress response genes. Depletion of Elongin from cells leads to impaired transcription of ER stress response genes and to defects in the recruitment of Mediator and its CDK8 kinase subunit. Taken together, these findings bring to light a role for Elongin as a loading factor for Mediator during the ER stress response.

The Mediator complex is an evolutionarily conserved RNA polymerase II (Pol II) coactivator essential for optimal activation of transcription of most genes. The Mediator was first identified in yeast and later in higher eukaryotes (1, 2). Substantial evidence indicates that Mediator plays critical roles at multiple stages of transcription—beginning with assembly of Pol II preinitiation complexes to initiation, elongation, and termination (2, 3). Mediator functions as an integral component of Pol II preinitiation complexes, where it participates directly in recruitment of Pol II and general initiation factors, and subsequently of components of the Pol II elongation machinery (4–8).

A key step in activation of Pol II transcription is recruitment of Mediator to its target genes. The prevailing model posits that Mediator is recruited to target genes via direct, physical interactions with the activation domains of DNA binding transcription factors associated with enhancers and other regulatory elements. Indeed, the transcription activation domains of a growing number of transcription factors have been shown to interact specifically with surfaces on a number of the more than 25 distinct Mediator subunits (9, 10).

In studies of the mechanism underlying activation of Pol II transcription by ATF6α, a master regulator of endoplasmic reticulum (ER) stress response genes (11, 12), we observed that ATF6α is capable of recruiting Mediator and activating transcription in vitro in crude nuclear extracts (13, 14) but not in a purified, reconstituted enzyme system that supports GAL4-VP16– or GAL4-p53–dependent transcription at a model promoter. These findings raised the intriguing possibility that a dissociable Mediator “loading factor” was present in crude extracts and, further, that this factor might contribute to ATF6α-dependent transcription activation. In this report, we identify Elongin as such a factor.

Elongin is a multifunctional complex that regulates signal-dependent transcription by as yet poorly defined mechanisms (15–18). Elongin was originally identified as a Pol II transcription factor capable of stimulating the overall rate of transcript elongation in vitro (19, 20), and recent evidence from genomic studies indicates it travels with elongating Pol II and regulates CTD phosphorylation (21, 22). Elongin also assembles with Cullin family member CUL5 and RING finger proteins RBX1/2 to form an E3 ubiquitin ligase (23) that binds the Cockayne syndrome B protein (24, 25) and targets Pol II stalled at sites of DNA damage for ubiquitylation and degradation by the proteasome during the DNA damage response (26–28).

Here we identify an additional function for Elongin. We show that Elongin supports ATF6α-dependent activation of Pol II transcription from the HSPA5 promoter, where it functions at least in part through a direct interaction with Mediator to promote its efficient recruitment to the HSPA5 promoter. In addition, we present evidence that depletion of Elongin from cells impairs recruitment of Mediator to the HSPA5 and other ATF6α-dependent ER stress response genes following induction of ER stress. Notably, dynamic recruitment of the CDK8 kinase subunit of Mediator to ATFα-regulated genes is markedly altered, raising the possibility that Elongin plays a special role in ensuring timely recruitment of the form of Mediator containing the CDK8/Cyclin C kinase module.

Results

Identification in Nuclear Extracts of an Activity Required for ATF6α-Dependent Activation of Pol II Transcription In Vitro.

Activation of Pol II transcription from a variety of promoters can be recapitulated in vitro in a “minimal enzyme system” consisting of purified Pol II, general initiation factors TFIIB, D, E, F, H, and the multisubunit Mediator complex. Mechanistic studies exploiting the model GAL4-VP16 activator and model promoters containing multiple GAL4 binding sites and a viral core promoter have led to the model that activation of Pol II transcription begins with activator-dependent recruitment of Mediator and TFIID to promoters to form a stable, salt-resistant intermediate that serves as the site of assembly for the Pol II preinitiation complex (PIC) prior to initiation (29, 30).

We have been investigating mechanisms by which the master regulator of ER stress response genes, ATF6α, activates transcription of genes such as HSPA5, which encodes BiP, an Hsp70 family chaperone important for protein folding in the ER and for the ER stress response (11, 12). Optimal binding of ATF6α to ER stress response elements (ERSEs) found in many ER stress regulated genes requires CCAAT binding transcription factor NF-Y, which was previously shown to be required for stable, long-lived binding of ATF6α to its binding sites (31–33). Like GAL4-VP16 (29, 30), ATF6α is capable of recruiting Mediator, Pol II, and general initiation factors to a promoter and activating transcription in nuclear extracts (13, 14). As described below, unlike GAL4-VP16, ATF6α is not sufficient to support formation of functional preinitiation intermediates or ATF6α-dependent Pol II transcription in the minimal transcription system.

In these experiments, we immobilized on streptavidin beads two different DNA templates. The first, HSPA5-G219, contains HSPA5 promoter sequences from −269 to +1, which includes HSPA5’s core promoter and three ERSEs (34, 35), fused to a downstream cassette encoding a G-less RNA. The second, GAL4-E4-G210, contains five GAL4 binding sites upstream of the adenovirus E4 core promoter, also fused to a G-less cassette (Fig. 1A). As diagrammed in Fig. 1B, immobilized HSPA5-G219 or GAL4-E4-G210 were first incubated with ATF6α and NF-Y or GAL4-VP16. Reactions were then supplemented with Mediator, TFIID, or other protein fractions to allow formation of the initial intermediate. After a high salt wash to remove unbound proteins, Pol II and the remaining general transcription factors were added to complete PIC formation, and transcription was assayed in the presence of radioactive ribonucleoside triphosphates. In these reactions, the amount of transcript synthesized reflects the amount of stable, functional preinitiation intermediate that remains bound to the template after the high salt wash.

Fig. 1.

A cofactor for ATF6α/NF-Y–dependent Pol II transcription in vitro. (A) Diagram of pG5-E4T-G210 and HSPA5-G219 templates. Green, activator binding sites; blue, core promoters; orange, G-less cassettes. The gray oval represents the position of the streptavidin beads. (B) Diagram of assay used in this study to measure activator-dependent transcription. (C) The minimal transcription system is not sufficient for ATF6α/NF-Y–dependent transcription. Assays performed with immobilized pG5-E4T or HSPA5 templates contained GAL4-VP-16 or ATF6α/NF-Y, Mediator (MED), TFIID, or P1 fractions from HeLa cells (HeLa P1) or rat liver (rat P1) as indicated in the figure. Here and in subsequent figures, reactions were supplemented with Pol II, TFIIB, TFIIF, TFIIE, and TFIIH following the high salt wash. (D) Flowchart showing scheme used for fractionation of rat liver nuclear extracts and preparation of HMW and LMW fractions. (E) HMW and LMW fractions from G4000SW size exclusion support ATF6α/NF-Y–dependent transcription. Assays performed without added TFIID and Mediator contained HMW and/or LMW fractions as indicated in the figure. HMW and LMW fractions used in these assays were fractions 23 and 31 from G4000SW chromatography shown in F. (F) G4000SW fractions prepared as shown in D were subjected to immunoblotting with the indicated antibodies. M, markers. Numbers on the Left show the size of molecular weight markers in kilodaltons. (G) Purified TFIID and Mediator replace the HMW fraction in ATF6α/NF-Y–dependent transcription. Where indicated in the figure, assays contained LMW, Mediator, and TFIID. (H) Activity in the LMW fraction cofractionates with the peak of Elongin shown in F. Aliquots of G4000SW fractions were assayed for TFIID activity (red) or for the ability to support ATF6α/NF-Y–dependent transcription in the presence of purified TFIID and Mediator (blue) as described in SI Appendix, Supplementary Materials and Methods. Thyroglobulin (Thyro) and ovalbumin (Oval) size standards eluted from this column in peaks centered at fractions 29 and 38, respectively. See also SI Appendix, Fig. S1.

Under these conditions, we observed strong GAL4-VP16–activated transcription on the GAL4-E4-G210 template (Fig. 1C, lanes 1 and 2); however, there was no difference in the amount of transcript synthesized in the presence or absence of ATF6α and NF-Y (lanes 3 and 4), suggesting that transcription activation by ATF6α/NF-Y may depend on different or additional factors. Consistent with this possibility, we observed strong ATFα/NF-Y–dependent transcription when we replaced purified Mediator and TFIID with a 0.5 to 1 M salt step from phosphocellulose chromatography of either HeLa or rat liver nuclear extracts (Fig. 1C, lanes 5 through 8, and Fig. 1D); this crude “P1 fraction” contains Mediator, TFIID, and many additional proteins.

Toward identifying the additional factor(s) needed to support ATF6α/NF-Y–dependent transcription, we subjected rat liver P1 to size-exclusion high-performance liquid chromatography. As shown in the complementation assays in Fig. 1E, we could reconstitute strong ATF6α/NF-Y–dependent transcription by combining a high molecular weight (HMW) fraction, which included Mediator and TFIID (Fig. 1F), with a second fraction that eluted in a smaller size range (low molecular weight [LMW] fraction). Importantly, the HMW fraction could be replaced with purified Mediator and TFIID (Fig. 1G).

Elongin Contributes to ATF6α/NF-Y–Dependent Transcription In Vitro.

Elongin was initially identified as a Pol II transcription factor that can stimulate the rate of transcript elongation in vitro (19). It is a three-subunit complex composed of the transcriptionally active Elongin A (ELOA) subunit and two small regulatory subunits, Elongin B and C, that bind to Elongin A and stimulate its elongation activity. We previously observed that Elongin, like the activity in the LMW fraction, elutes from phosphocellulose in the P1 fraction and from size-exclusion columns between the thyroglobulin and ovalbumin markers (19). As shown in Fig. 1 F and H and SI Appendix, Fig. S1B, the LMW activity cofractionates closely with Elongin through size-exclusion chromatography, raising the possibility that Elongin might contribute to ATF6α/NF-Y–dependent transcription.

We therefore asked whether recombinant Elongin, purified from Escherichia coli (Fig. 2A, wild type [WT]), could replace the LMW fraction in ATF6α/NF-Y–dependent transcription from the HSPA5 promoter. In the experiment shown in Fig. 2B, we observed strong stimulation of transcription by ATF6α and NF-Y when Elongin was included along with Mediator and TFIID prior to the high salt wash. Notably, Elongin did not increase the extent to which the model transactivators GAL4-VP16 and GAL4-p53 stimulated transcription from the GAL4-E4T promoter in our assays; however, Elongin addition led to a modest increase in both basal and activated transcription from GAL4-E4T (Fig. 2B).

Fig. 2.

Recombinant Elongin replaces the activity in the LMW fraction. (A) Coomassie blue-stained polyacrylamide gel showing wild-type and mutant Elongin complexes (ELO) used in transcription assays. Lane 1, molecular weight markers (M); lanes 2 through 4, Elongin composed of Elongin B and C and full-length Elongin A (WT), N-terminally deleted Elongin A (ΔN), or C-terminally deleted Elongin A (ΔC); and lane 5, Elongin BC complexes. (B) Elongin is needed for ATF6α/NF-Y–activated transcription but not for transcription activated by GAL4-VP16 or GAL4-p53. Transcription assays performed as diagrammed in Fig. 1B contained TFIID and Mediator and were supplemented or not with Elongin prior to the high salt wash. Reactions contained immobilized GAL4-E4T or HSPA5 templates and GAL4-VP16 or ATF6α/NF-Y as indicated in the figure. Here and in C, the numbers indicate the fold stimulation by activators in the presence or absence of Elongin. (C) Elongin supports ATF6α/NF-Y–dependent transcription in reactions containing Mediator purified through FLAG-MED29, FLAG-MED26, or FLAG-MED12. The indicated templates and activators were incubated with TFIID, Mediator, and Elongin prior to the high salt wash. (D) ATF6α/NF-Y– and Elongin-dependent Pol II transcription depends on TFIIB, TFIIE, TFIIF, and TFIIH. Reaction components were added before or after the high salt wash as indicated in the figure; those shown in red were omitted. (E) ATF6α/NF-Y– and Elongin-dependent Pol II transcription depend on Mediator and TFIID. (F) Intact Elongin A TFIIS-like N-terminal domain and C-terminal elongation stimulatory domain are needed for ATF6α-activated transcription. Transcription assays performed as diagrammed in Fig. 1B contained TFIID and Mediator and were supplemented or not with Elongin with wild-type or mutant Elongin A prior to the high salt wash. (G) Diagram showing Elongin A domain structure and deletion mutants. See also SI Appendix, Fig. S2.

Mediator exists in cells in at least two forms (36–38). One form, which includes Mediator subunit MED26, copurifies with Pol II. The other form lacks MED26 and Pol II and instead includes the Mediator kinase module, which is composed of the cyclin-dependent kinase CDK8/Cyclin C and additional subunits MED12 or MED12L and MED13 or MED13L. To determine whether Elongin function in ATF6α/NF-Y–dependent transcription exhibits specificity for a particular form of Mediator, we compared the effect of Elongin on ATF6α/NF-Y–dependent transcription in the presence of Mediator containing MED26 (purified through FLAG-tagged MED26), Mediator containing the CDK8 kinase module (purified through FLAG-tagged MED12), and Mediator purified through FLAG-tagged MED29, which is a mixture of the MED26- and kinase module–containing Mediator. As shown in Fig. 2C, the three different preparations of Mediator behaved similarly in our assays, arguing that Elongin does not discriminate between the various forms of Mediator.

We considered the possibility that Elongin might replace the function of one or more of the general transcription factors in these reactions. As shown in Fig. 2 D and E, omitting MED, TFIID, B, E, F, or H abrogated both basal and activated transcription. Thus, Elongin provides an additional function in ATF6α/NF-Y–dependent transcription.

The activity of Elongin in these assays depends on Elongin A; as shown in the last two lanes of Fig. 2F, the Elongin BC subcomplex cannot replace the complete Elongin complex in our assays. We next sought to determine whether the elongation stimulatory activity of Elongin A is sufficient to support activator-dependent transcription. As summarized in Fig. 2G, we previously observed that an Elongin A fragment that includes residues 400 to 772 (Elongin ΔN), but lacks a conserved domain also found at the N-termini of Pol II elongation factor TFIIS and Mediator subunit MED26, is sufficient to assemble with Elongin BC to form Elongin complexes that can interact with Pol II and stimulate its elongation activity. In contrast, Elongin complexes lacking C-terminal residues of Elongin A (Elongin ΔC) are unable to stimulate elongation (39). Neither Elongin ΔN nor Elongin ΔC can replace the intact Elongin complex in activation of ATF6α/NF-Y–dependent transcription (Fig. 2F). Thus, Elongin activity in ATF6α/NF-Y–dependent transcription requires not only an intact C-terminal domain, but also the N-terminal TFIIS-like domain of Elongin A. Taken together, these observations indicate that the function of Elongin in ATF6α-dependent transcription cannot be explained simply by its ability to stimulate elongation.

Elongin Promotes ATF6α/NF-Y–Dependent Recruitment of Mediator to the HSPA5 ERSEs.

We next sought to explore the role of Elongin in ATF6α/NF-Y–dependent activation of Pol II transcription. We therefore carried out order-of-addition experiments according to the protocol diagrammed in Fig. 3A to determine whether Elongin, Mediator, and TFIID are all needed before the wash to support formation of a functional preinitiation intermediate, or whether one or more of these factors can be added later, during PIC completion. We observed that ATF6α/NF-Y could still stimulate transcription when TFIID was added after the wash (SI Appendix, Fig. S2). As shown in Fig. 3B, however, ATF6α and NF-Y stimulated transcription in these assays only when both Elongin and Mediator were included in the first stage of the reaction (Fig. 3B), raising the possibility that Elongin might promote binding of Mediator to the promoter and/or place Mediator into an active conformation.

Fig. 3.

Elongin binds Mediator and promotes its activator-dependent recruitment to promoters in vitro. (A) Schematic diagram outlining protocol for reactions shown in B. (B) Elongin and Mediator support formation of a functional, ATF6α/NF-Y–dependent initial intermediate at the HSPA5 promoter. TFIID, with or without Mediator and Elongin, was incubated with immobilized HSPA5 promoter DNA prior to the high salt wash. Where indicated, Elongin or Mediator were added with Pol II, TFIIB, TFIIE, TFIIF, and TFIIH after the wash step. (C) Schematic diagram outlining protocol for binding reactions shown in D–F. (D) Elongin enhances ATF6α/NF-Y–dependent recruitment of Mediator containing MED26 or CDK8 to the HSPA5 promoter. MED26, CDK8, and MED6 were detected by immunoblotting with the indicated antibodies. (E and F) Cooperative binding of Mediator and Elongin to HSPA5 DNA depends on the ERSEs (E) and both ATF6α and NF-Y. Elongin A and MED6 were detected by Western blotting. (G) Schematic diagram outlining protocol used to measure Mediator–Elongin interaction. (H) Mediator binds Elongin containing wild-type but not mutant ELOA. Bound and unbound fractions from binding assays performed as diagrammed in G were analyzed by Western blotting with the indicated antibodies. (I) Elongin enhances GAL4-VP16–dependent recruitment of Mediator to immobilized promoter DNA containing GAL4 binding sites.

We then asked whether Elongin increases Mediator binding to ATF6α/NF-Y at the HSPA5 promoter. In these experiments, Mediator (purified through FLAG-MED29) and Elongin were incubated with bead-bound HSPA5 promoter DNA, with or without ATF6α and NF-Y, and washed, and bound proteins were analyzed by Western blotting against Mediator subunits (Fig. 3C). We observed strong, cooperative binding of Elongin and Mediator subunits MED6, MED26, and CDK8 to ATF6α/NF-Y–bound HSPA5 promoter DNA (Fig. 3 D and E). Hence, Elongin evidently enhances binding of both the MED26-containing and CDK8 kinase module-containing forms of Mediator. This observation is consistent with our evidence that Elongin supports ATF6α/NF-Y–dependent transcription in reactions performed with either MED26- or kinase module–containing Mediator. Mediator and Elongin binding to promoter DNA depended strongly on both ATF6α and NF-Y (Fig. 3F), and, as expected, detectable ATF6α/NF-Y–dependent binding of Mediator and Elongin to immobilized DNA was not observed with an HSPA5 template lacking the ERSEs (Fig. 3E).

Consistent with evidence that Mediator can bind directly to the ATF6α activation domain (14, 40), we detected a small amount of Mediator in association with the ATF6α/NF-Y–bound promoter in the absence of Elongin, while there was no detectable binding of Elongin when Mediator was omitted. Taken together, these observations suggest that Elongin interacts with Mediator and enhances the ability of Mediator to bind to ATF6α/NF-Y at the HSPA5 ERSEs.

We next asked whether Elongin can bind directly to Mediator in the absence of ATF6α/NF-Y and promoter DNA. To do so, we immobilized Mediator on agarose beads by incubating nuclear extract from HeLa S3 cells expressing a FLAG-epitope–tagged version of Mediator subunit MED26 (HeLa F-MED26) with anti-FLAG agarose. Control beads without Mediator were prepared by incubating anti-FLAG agarose with nuclear extracts from parental cells. After extensive washing to remove unbound proteins, beads bound to purified, immobilized Mediator or control beads were incubated with an excess of recombinant wild-type Elongin, washed again, and bound proteins were eluted with FLAG peptide and detected by Western blotting (Fig. 3G). As shown in Fig. 3H, Elongin bound to Mediator immobilized on FLAG-agarose, but not to the control beads, supporting the idea that Elongin binds directly to Mediator.

To ask whether there is a correlation between Elongin’s ability to bind Mediator and support activator-dependent transcription, we performed similar experiments using the Elongin mutants Elongin ΔN and Elongin ΔC, which do not support ATF6α/NF-Y–dependent transcription. Notably, we observed a striking reduction in the ability of Elongin ΔN and Elongin ΔC to bind Mediator compared to wild-type Elongin, consistent with the idea that direct interactions between Elongin and Mediator contribute to Elongin’s role in ATF6α/NF-Y–dependent transcription.

Finally, we asked whether the ability of Elongin to load Mediator onto promoter-bound transactivators is limited to ATF6α or whether it could reflect a more general Elongin function by performing binding experiments using immobilized GAL4-E4-G210 DNA and GAL4-VP16. As shown in Fig. 3I, the results of this experiment indicate that Elongin can also enhance binding of Mediator to promoter-bound GAL4-VP16, even though GAL4-VP16 can evidently recruit enough Mediator to support activated transcription in vitro without the help of Elongin.

Loss of Elongin Leads to Decreased Expression of ATF6α-Regulated Genes.

Thapsigargin (Tg) is a drug that induces ER stress by inhibiting the ER Ca²⁺-ATPase (41). To investigate further the role of ELOA in ATF6α-activated transcription of ER stress response genes, we first assessed the effect of siRNA-mediated depletion of ELOA on accumulation of Tg-induced transcripts (Fig. 4A). Reduction of ELOA mRNA in siELOA-treated cells was confirmed by qPCR (Fig. 4B and SI Appendix, Fig. S3A). Depletion of ELOA using either a SMARTpool or individual siRNAs led to a reduction in Tg-induced accumulation of HSPA5 mRNA (Fig. 4C and SI Appendix, Fig. S3B). As measured by qPCR, Tg-induced accumulation of new HSPA5 mRNA was reduced in cells transfected with ELOA siRNA compared to control. Similar results were observed with three additional well-characterized ATF6α targets, DDIT3, HERPUD1, and HYOU1 (SI Appendix, Fig. S3 C–E).

Fig. 4.

Effect of Elongin A depletion on expression of HSPA5 and other ATF6α target genes. (A) Outline of protocol for treatment of HCT116 cells with siRNA targeting Elongin A (siELOA) or control nontargeting siRNA (NonT), followed by induction of ER stress with 300 nM Tg. Cells were collected at various times after Tg treatment and processed for RT-qPCR, RNA-seq, or CUT&Tag, respectively. (B) Validation of Elongin A knockdown efficiency by RT-qPCR. Data points are the average of three independent experiments; error bars show SD. **P < 0.01,***P < 0.001, unpaired t test. (C) Elongin A depletion reduces ER stress–induced HSPA5 expression. Data points are the average of three independent experiments; error bars show SD. **P < 0.01, unpaired t test. (D) Gene Ontology analysis for the 162 ER stress–induced, ATF6α-bound genes. Shown are the five most significantly enriched biological processes in order of increasing P value. (E) Graph showing the effect of ELOA depletion on Tg-dependent activation of each of the 44 genes in group 2. Plotted is the ratio of Tg-dependent activation in cells treated with siELOA or nontargeting (NonT) siRNA. Each value represents the average of three biological replicates. See also SI Appendix, Fig. S3.

Using RNA-sequencing (RNA-seq), we explored the effect of ELOA knockdown on the expression of a wider set of genes that are regulated by ER stress and targeted by ATF6α. We identified a set of ∼162 genes that are 1) induced by Tg and 2) exhibit increased ATF6α occupancy near their transcription sites by cleavage under targets and tagmentation (CUT&Tag; see below) (42, 43) (SI Appendix, Table S1). Not surprisingly, these genes were strongly enriched for Gene Ontology (GO) terms related to ER stress and the unfolded protein response (Fig. 4D). As shown in the violin plots in SI Appendix, Fig. S3F , siRNA-mediated ELOA depletion led to a small but significant reduction in activation of this set of genes (group 1) after 4 h of Tg treatment, and this reduction was enhanced when the gene set was further limited to the 44 genes (group 2) from group 1 annotated with the GO term “response to endoplasmic reticulum stress” (SI Appendix, Fig. S3F and Table S1). We also generated plots showing the effect of ELOA depletion on Tg-dependent activation of individual genes in both groups 1 and 2. As shown in Fig. 4E and SI Appendix, Fig. S3G, the majority of these genes exhibited reduced Tg-dependent activation.

Loss of Elongin Leads to Decreased Mediator Occupancy at HSPA5 during the ER Stress Response.

We next assessed the effect of ELOA depletion on recruitment of Mediator and other components of the Pol II transcription machinery to the HSPA5 gene by performing CUT&Tag on cells that were not treated with Tg and cells that were treated with Tg for 30, 60, or 120 min. As shown in Fig. 5 A, Upper, ATF6α accumulates similarly in both ELOA-depleted and control cells in response to Tg at the HSPA5 promoter region, which includes the ERSEs. ATF6α appeared rapidly within 30 min of Tg treatment and persisted throughout the time course of the experiment, although its occupancy decreased somewhat at the 120-min time point. A smaller, intergenic ATF6α peak was also observed ∼1.5 kb upstream of the major ATF6α peak at the HSPA5 promoter proximal region. Inspection of ENCODE datasets from unstimulated HCT116 cells indicates the upstream ATF6α peak falls in a region that is rich in H3K27Ac and H3K4me1 (44), which often flank enhancer elements (45). Elongin appears with similar kinetics at both the HSPA5 promoter region and upstream ATF6α peak. Interestingly, the ratio of Elongin to ATF6α was greatest at the upstream ATF6α site. Elongin also appears throughout the body of the HSPA5 gene (SI Appendix, Fig. S4). These results are consistent with two recent reports from Dynlacht and coworkers (21) and Kingston and coworkers (22), who observed that Elongin accumulates not only at the promoters of genes, but also throughout the bodies of genes and at many intergenic loci with properties of enhancers. Confirming that our siELOA treatment effectively reduced levels of ELOA, we observed substantially reduced ELOA CUT&Tag signals in siELOA-treated cells (Fig. 5A and SI Appendix, Fig. S4).

Fig. 5.

Effect of ELOA depletion on recruitment of Mediator subunits and other transcription components. (A) IGV browser shots showing ATF6α, ELOA, MED15, MED1, and CDK8 occupancy at a region surrounding the HSPA5 promoter, detected by CUT&Tag. Tracks from HCT116 cells treated with nontargeting siRNA (solid colors) or siRNA against ELOA (black lines) are overlaid. Gray boxes indicate the positions ATF6α-occupied loci overlapping the ERSEs and at a region ∼1.5 kb upstream. (B) Average gene plots showing ATF6α, ELOA, MED15, MED1, and CDK8 occupancies at ER stress–induced, ATF6α-bound genes that show the greatest decrease in MED15 and MED1 occupancy after Elongin A knockdown (n = 77). (C) Graphs show background subtracted areas under the metaplot curves for each time point shown in B. Arb units, arbitrary units. See also SI Appendix, Fig. S4.

To investigate the contribution of ELOA to ATF6α-dependent recruitment of Mediator to the HSPA5 gene, we measured the occupancies of Mediator middle module subunit MED1, tail module subunit MED15, and kinase module subunit CDK8 before and after Tg treatment in control and ELOA-depleted cells. As shown in Fig. 5A, ELOA depletion led to reductions of all three Mediator subunits. Notably, there were differences in both the location and kinetics of accumulation of core Mediator subunits MED1 and MED15 and of kinase module subunit CDK8. Whereas MED1 and MED15 accumulated in parallel with ELOA at both the HSPA5 promoter and upstream regions, there was little CDK8 at the upstream region. The accumulation of CDK8 at the promoter was delayed relative to ATF6α, Elongin, and the other Mediator subunits, reaching maximum accumulation 60 min after Tg treatment and decreasing to a level near that observed in untreated cells after an additional 60 min (Fig. 5A).

We also used CUT&Tag to assess the effect of ELOA depletion on TBP and Pol II. As shown in SI Appendix, Fig. S4, TBP accumulated at the HSPA5 promoter but not the upstream region in response to Tg treatment; however, neither the magnitude nor the kinetics of TBP accumulation were altered by depletion of ELOA. In contrast, marked reductions of both the CTD serine 2- and serine 5-phosphorylated forms of Pol II were observed following depletion of ELOA (SI Appendix, Fig. S4). Before induction of ER stress, the amount and distribution of S5-phosphorylated and S2-phosphorylated Pol II was very similar in both control and Elongin A–depleted cells. Following induction of ER stress, the amount of S5P and S2P Pol II increased dramatically in both control and knockdown cells. However, loss of Elongin was associated with a gradual accumulation of S5P-phosophorylated Pol II downstream of the transcription start site, within the first few hundred bases of the transcribed region, and a substantial reduction in the amount of both S5- and S2-phosphorylated Pol II at the 3′ end of the transcription unit. These results are reminiscent of recent findings from the Dynlacht and coworkers, who observed that depletion of ELOA led to reduced levels of both S5- and S2-phosphorylated Pol II over the bodies of many genes during steady-state transcription in human DLD1 cells (21).

Loss of Elongin Leads to Decreased Mediator Occupancy at Other ATF6α-Regulated Genes.

In addition to the HSPA5 gene (our model for biochemical studies), loss of ELOA led to reduced accumulation of Mediator subunits at many, but not all, ATF6α target genes. Fig. 5B shows metaplots depicting the average occupancy of ATF6α, Elongin, MED15, MED1, and CDK8 at a subset of the ATF6α target genes (n = 77); the graphs in Fig. 5C show background subtracted areas under the metaplot curves for each time point. As at HSPA5, MED15 and MED1 accumulate more rapidly than CDK8 in response to Tg treatment and remain associated with the promoter regions of these genes throughout the time course of the experiment, while CDK8 occupancy increases dramatically 1 h after Tg treatment and then returns to the level seen in unstimulated cells. In general, the effect of Elongin depletion on MED15 and MED1 was observed 30 min after Tg treatment but not at later times, suggesting that Elongin depletion delays, but does not prevent, accumulation of these subunits. The most striking and widespread effect, however, was misregulation of CDK8 recruitment. As shown in Fig. 5 B and C, depletion of ELOA led to a dramatic reduction in the transient accumulation of CDK8 at these ATF6α target genes, raising the possibility that ELOA has a special role in regulating the subform of Mediator possessing the kinase module.

Discussion

The findings described in this report expand the repertoire of functions for the RNA polymerase II transcription factor Elongin. Elongin was originally identified as a regulator of Pol II elongation that can control polymerase activity in at least two distinct ways: 1) by interacting directly with transcribing Pol II and stimulating the overall rate of transcript synthesis and 2) by assembling with Cullin family member CUL5 and RING finger protein RBX1/2 to form an E3 ubiquitin ligase that targets stalled Pol II for removal from genes during the DNA damage response. Here, we uncover an additional function of Elongin in transcription initiation. In particular, it supports ATF6α/NF-Y–dependent formation of functional preinitiation complexes at the HSPA5 promoter in vitro. In addition, we show that it binds directly to Mediator and can enhance Mediator loading onto promoters both in vitro and in cells. In the future, it will be of interest to determine whether the same Elongin molecules that promote Mediator loading then transition from the preinitiation complex into the early elongation complex, or whether distinct Elongin molecules participate in Mediator loading and elongation.

Our analysis of the contribution of Elongin to ER stress–regulated transcription in cells revealed that Elongin depletion leads to altered pattens of Mediator occupancy at the promoters of a subset of ER stress–induced, ATF6α target genes. Notably, Elongin depletion differentially affects both the extent and dynamics of recruitment of Mediator core subunits (MED1 and MED15) and the Mediator kinase module subunit CDK8. MED1 and MED15 are rapidly recruited to HSPA5 and other genes upon induction of ER stress, coincident with the appearance of ATF6α and Elongin. Decreased MED1 and MED15 occupancy following Elongin depletion was most evident shortly after Tg treatment, suggesting Elongin may not be essential for recruitment of the Mediator core, but rather may help bring it rapidly to promoters following ER stress. In contrast, under normal conditions, CDK8 is recruited only transiently, well after MED1 and MED15, and this recruitment is largely blocked by Elongin depletion. Thus, as noted earlier, Elongin could have a particularly important role in recruiting and/or regulating forms of Mediator associated with the kinase module.

Our observation that CDK8 is transiently recruited to many ER stress response genes raises the intriguing possibility that different forms of Mediator are recruited to ER stress response genes at different times during their transcription to perform different functions and/or that Mediator composition may change during the ER stress response. Indeed, evidence suggests that Mediator is often recruited to genes in association with the kinase module, which is then released as transcription activation proceeds (46–50). The kinase module has both activating and repressive functions in different contexts (51), and whether it contributes to Mediator-dependent activation of ER stress response genes, dampens their expression, or both, remains to be determined. Future exploration of the causes and consequences of the Mediator kinase module’s transient enrichment at ER stress response genes will be needed to understand its contribution to regulation of ER stress–responsive transcription.

The finding that Elongin increases the amount of Mediator recruited to promoters bound by both ATF6α/NF-Y and GAL4-VP16 in vitro, even though Elongin is not needed for formation of active GAL4-VP16–dependent preinitiation complexes, was unexpected. This suggests that, under our assay conditions, GAL4-VP16 is able to recruit enough Mediator in the correct orientation or conformation to support transcription, while ATF6α/NF-Y cannot. While we do not understand why ATF6α/NF-Y and GAL4-VP16 behave differently, it is known that both ATF6α and GAL4-VP16 bind to Mediator through a portion of its MED25 subunit referred to as its activator interaction domain (AcID) (14, 52–54). The ATF6α transactivation domain binds to a surface on AcID referred to as H2, while the larger VP16 transactivation domain binds to H2 as well as a surface on the opposite face of AcID referred to as H1 (40). Of note, the ATF6α and VP16 activation domains bind the MED25 AcID with similar affinities, but induce distinctly different AcID conformations, with potentially different consequences for Mediator transcription activity. These findings, together with evidence that AcID conformation can be altered allosterically through interaction with an additional binding partner(s), makes it tempting to speculate that, in addition to facilitating Mediator recruitment by ATF6α, Elongin binding to Mediator could alter the structure of ATF6α-bound MED25 in a way that more effectively sends a transcription activating signal.

Our findings add Elongin to a growing list of Pol II elongation factors that function with Mediator. In yeast, the genes encoding Pol II elongation factor TFIIS and several Mediator subunits exhibit synthetic lethality associated with transcription defects; these defects can, at least in some cases, be rescued by TFIIS mutants that are inactive in Pol II elongation (55, 56). In addition, chromatin immunoprecipitation experiments have shown that 1) yeast TFIIS is recruited to the GAL1 UAS in a step that depends on GAL4, SAGA, and Mediator (57) and, further, that 2) yeast lacking wild-type TFIIS exhibit reduced occupancy of Mediator and/or Pol II at several promoters, suggesting that TFIIS could contribute to Mediator recruitment (57). This idea could be consistent with biochemical experiments suggesting that yeast TFIIS promotes formation of Pol II preinitiation complexes in yeast extracts and purified enzyme systems (56, 58). Although it is presently not known whether, like Elongin, TFIIS binds directly to purified Mediator and is able to facilitate its recruitment to promoters by DNA-bound transcription factors, it is noteworthy that both Elongin A and TFIIS share a conserved N-terminal domain that, in the case of Elongin, facilitates Mediator recruitment. Therefore, it will be of interest to test TFIIS directly in similar biochemical experiments to determine whether it also functions as a Mediator loading factor.

Finally, it is worth noting that our biochemical experiments indicate that Elongin not only promotes binding of Mediator to promoters but can itself be recruited to promoters through its interaction with Mediator. This finding is reminiscent of earlier findings implicating Mediator subunits MED26 and MED23 and the Mediator kinase module in recruitment and/or regulation of Pol II elongation factors ELL/EAF, P-TEFb, and associated proteins that make up the Super Elongation complex (SEC), as well as of the ELL/EAF-containing Little Elongation complex (LEC) (59–62). In the future, it will be of interest to explore the possibility that Mediator acts as a general platform for recruitment of these and perhaps other elongation factors to promoters, where they are poised for transfer to Pol II after initiation occurs.

Materials and Methods

Pol II and TFIIH were purified from rat liver, and Mediator and TFIID were purified from HeLa cells. All other transcription factors were expressed in and purified from E. coli. Detailed protocols for cell culture and cell lines, preparation of Pol II and transcription factors, assays for transcription and transcription factor recruitment, siRNA transfection and thapsigargen treatment, RNA extraction, cDNA synthesis and RT-qPCR, CUT&Tag assay, data analysis, and antibodies are described in SI Appendix, Supplementary Materials and Methods. Next-generation sequencing (NGS) datasets are available at Gene Expression Omnibus (GEO): GSE172134. All other original data underlying this manuscript can be accessed from the Stowers Original Data Repository at https://www.stowers.org/research/publications/LIBPB-1620.

Data Availability

RNA-seq and CUT&Tag (NGS) data have been deposited in GEO (GSE172134) (63). All other original data underlying this manuscript can be accessed from the Stowers Original Data Repository at https://www.stowers.org/research/publications/LIBPB-1620. Previously published data were used for this work (GEO: GSE96299 and GSE95958, ref. 44).