Pathway-specific repressor proteins are able to physically interact with pleiotropic corepressors Sin3, Cyc8 and Tup1

Transcriptional repressor Opi1 negatively regulates structural genes of yeast phospholipid biosynthesis by interaction with pleiotropic corepressors Sin3 and Cyc8 (Jäschke et al. 2011) both of which are able to counteract gene expression by recruitment of various histone deacetylases. This redundancy of corepressor interaction may be specific for Opi1 but may be also effective for additional repressor proteins. To find out whether functional redundancy is indeed a general mechanism of gene repression, we systematically investigated in vitro interaction of repressors affecting several unrelated regulatory pathways in yeast (Supporting Online Material, Table S1) with corepressors Sin3, Cyc8 and Tup1 by GST pull-down experiments. GST fusions of 18 full-length repressors (exception: Ume6, for which the minimal domain interacting with Sin3 was used) were synthesized in E. coli, bound to glutathione sepharose and subsequently incubated with protein extracts containing epitope-tagged corepressors HA-Sin3, HA-Cyc8 and HA-Tup1, respectively. Interaction studies were initially performed with corepressor-containing protein extracts from yeast (not shown) and then repeated with bacterial extracts. These in vitro studies summarized in Fig. 1 confirmed several previously described individual interactions but indeed revealed a complex pattern of interaction between corepressors and repressors some of which may bind three corepressors (5; Ume6, Cti6, Rox1, Yox1 and Dal80), two corepressors (9; Opi1, Yhp1, Mot3, Whi5, Fkh1, Rdr1, Gal80, Ure2 and Sko1) or a single corepressor (3; Xbp1, Mig1 and Nrg1). 13 out of 18 repressor proteins were able to interact with Cyc8, 12 with Sin3 and 11 with Tup1. Eight repressors were able to bind to both subunits of the Cyc8-Tup1 corepressor (Ume6, Cti6, Rox1, Yox1, Dal80, Gal80, Ure2 and Sko1), indicating that redundancy of interaction is effective even within the same complex. These studies were performed with proteins entirely produced by E. coli, indicating that direct interactions have been detected in these experiments. Among repressor proteins assayed for interaction, only Mth1 could not bind to a corepressor. Thus, Mth1 was not considered for subsequent investigations. In contrast to a previous publication reporting that Sko1 merely interacts with Tup1 (Pascual-Ahuir et al. 2001) we found that Sko1 could bind Cyc8 and Tup1.

Fig. 1

In vitro interaction of selected transcriptional repressor proteins with pleiotropic corepressors and gene repression by lexA fusions in vivo. For pull-down assays, GST fusions with expression cassettes encoding full-length proteins were used (exception: Ume6 aa 508–594, shown to able to bind PAH2 of Sin3; Kadosh and Struhl 1997). GST fusion proteins were immobilized on GSH sepharose and incubated with bacterial protein extracts, containing HA-Sin3 (pSW11; aa 1–1536, full-length), HA-Cyc8 (pFK77, aa 1–398, encoding TPR motifs) or HA-Tup1 (pFK76; full-length). Gene repression in vivo was measured in transformants of strain RTS-lexA (reporter gene lexAOp-CYC1-lacZ) containing plasmids which encode lexA fusions. Specific β-galactosidase activities are given in nmol oNPG hydrolyzed per min per mg of protein. Protein extracts prepared from at least 12 independent transformants were assayed. GST and lexA fusion plasmids are compiled in Supporting Online Material (Table S2). In vivo gene repression (RF, repression factor) was calculated as the ratio of specific β-galactosidase activities measured in transformants with empty vector pRT-lexA and individual lexA-repressor fusions, respectively. n. t. not tested, SD standard deviation,  +  in vitro interaction, - no interaction

In addition to in vitro binding studies we also investigated whether repressor proteins which can bind at least one corepressor are able to fulfil repressor function in the living cell, thus decreasing gene expression in vivo. An episomal expression plasmid (pRT-lexA) containing a MET25Prom-HA-lexADBD-NLS cassette with a versatile multi-cloning site was constructed and used to generate lexADBD-repressor fusions. To assay gene repression in vivo, derivatives of pRT-lexA encoding lexADBD-repressor fusions were transformed into a yeast strain containing an integrated reporter gene with a lexA operator sequence upstream of UAS1 and UAS2 of a CYC1-lacZ promoter fusion (Kadosh and Struhl 1997). Once recruited to the lexAOp upstream of the native CYC1 promoter, a functional repressor should significantly decrease expression of the reporter gene lacZ, leading to reduced activity of β-galactosidase. As is also shown in Fig. 1, lexADBD-repressor fusions were indeed able to downregulate expression of the reporter gene, although to varying degrees (2.6-fold repression mediated by Dal80 vs. 11.5-fold repression by Yox1). As is evident from the comparison of Yox1 and Dal80, which may both bind to Sin3, Cyc8 and Tup1, the strength of gene repression in vivo does not simply correlate with the number of corepressors a given repressor may contact.

Sin3-binding repressor proteins interact with PAH1 or PAH2

Although the four PAH domains of Sin3 were proposed as sites of protein–protein-interactions (Wang et al. 1990), previous studies with a limited number of proteins indicated that PAH1 and PAH2 are mainly responsible for repressor recruitment (Wagner et al. 2001; Washburn and Esposito 2001; Sahu et al. 2008). We thus systematically investigated which of its domains is contacted by 10 repressor proteins able to bind full-length Sin3 in vitro (not shown: Ume6 for which PAH2 was described as its target site, Washburn and Esposito 2001; Cti6, which binds to PAH1 and PAH2, Aref and Schüller 2020).

Epitope-tagged length variants of Sin3 were synthesized in S. cerevisiae and protein extracts subsequently used for interaction studies with immobilized GST fusions of repressor proteins. As is depicted in Fig. 2, several repressors are able to bind PAH1 and PAH2 in vitro (Opi1, Yox1, Rox1, Fkh1, Rdr1, Xbp1 and Whi5) while others show specificity for PAH1 (Dal80, Yhp1) or PAH2 (Mot3). As an exception, Whi5 could also interact with additional Sin3 length variants both of which contain the HDAC interaction domain 1, known to recruit the yeast major HDAC Rpd3 (Laherty et al. 1997; Grigat et al. 2012).

Fig. 2

Interaction of gene-specific repressor proteins with functional domains of Sin3. GST fusions with full-length repressor proteins (for expression plasmids cf. Supporting Online Table S2) were immobilized on GSH sepharose and incubated with protein extracts from yeast transformants, containing epitope-tagged Sin3 length variants 1–300 (PAH1; pCW83), 301–600 (PAH2; pYJ91), 601–950 (PAH3 + HID1; pYJ90), 801–1100 (HID1; pYJ89) and 1100–1536 (PAH4; pMP20), respectively. GST devoid of repressor fusion served as a negative control. HID HDAC interaction domain; PAH, paired amphipathic helix,  +  in vitro interaction, - no interaction

Although we have previously found that Opi1 contacts PAH1 of Sin3 (Wagner et al. 2001), these results show that PAH2 can be bound as well. To obtain evidence for the in vivo significance of Opi1-Sin3 interactions, we investigated regulated expression of a reporter gene dependent on the Opi1-controlled UAS element of phospholipid biosynthetic genes (ICRE-CYC1-lacZ, inositol/choline response element; Wagner et al. 2001) in the presence of several Sin3 deletion variants (Wang and Stillman 1993). As is shown in Table 1, our assays confirmed the severe deregulation of ICRE-dependent gene expression in a sin3 deletion strain (increased expression under repressing conditions, decreased expression under derepressing conditions). Removal of each single PAH from Sin3 still allowed a significant repression of the reporter gene while gene expression in the absence of both PAH1 and PAH2 was substantially increased. It should be mentioned that loss of PAH1 also prevents full derepression of the ICRE-dependent reporter gene. In summary, we conclude that PAH1 and PAH2 are indeed redundant for mediating Opi1-dependent gene repression. A similar functional redundancy may be also effective for other repressors shown to bind in vitro to both domains, PAH1 and PAH2.

Table 1 Influence of Sin3 deletion variants on regulated expression of an ICRE-dependent reporter geneTup1-binding repressor proteins interact with WD40 repeats

It has been shown that repressor interactions with Cyc8 are mediated by TPR domains clustered at its N-terminus (Schultz et al. 1990; Tzamarias and Struhl 1994, 1995; Jäschke et al. 2011). Although WD40 repeats of Tup1 are generally presumed to mediate interaction with repressor proteins, experimental evidence for this assumption has been reported only in details for alpha2 (α2; Komachi et al. 1994). To investigate whether Tup1 recruitment by other repressors is exclusively dependent on its WD40 domain, we generated Tup1 length variants separating the N-terminus responsible for binding of Cyc8, mediator and histones H3/H4 (aa 1–339) from the C-terminus containing seven WD40 repeats (aa 340–713). We also constructed a Tup1 truncation encoding only two complete WD40 repeats (aa 599–713; WD40-6 and -7; Fig. 3a). Using alpha2 as a reference, we could indeed show that Yox1, Gal80, Ure2, Cti6 and Dal80 are able to bind to WD40 repeats but not to the N-terminus of Tup1 (Fig. 3b). Repressor-Tup1 interaction does not require the entire set of seven WD40 repeats but is also effective with only two WD40 repeats, confirming a previous study which showed that alpha2 can bind to WD40-2 (Komachi et al. 1994). We conclude that WD40 repeats exhibit only a limited degree of specificity but are in fact substantially redundant for repressor interaction.

Fig. 3

Interaction of selected gene-specific repressor proteins with N- and C-terminal length variants of Tup1. a Structural features of Tup1 and position of HA-tagged length variants used for interaction studies. b GST fusions with repressor proteins alpha2, Yox1, Gal80, Ure2, Cti6 and Dal80 (for expression plasmids cf. Supporting Online Table S2) were immobilized on GSH sepharose and incubated with bacterial protein extracts, containing epitope-tagged Tup1 length variants 1–339 (no WD40 repeats; pVS1), 340-713 (WD40 repeats 1–7; pVS2) and 599–713 (WD40 repeats 6 and 7; pVS5), respectively

Mapping and mutational analysis of corepressor binding domains

We could previously show that a single domain within repressor Opi1 mediates contact to both corepressors Sin3 and Cyc8 (OSID, aa 1–106) and that Opi1 OSID missense variants identically affect interaction with Sin3 and Cyc8 (Jäschke et al. 2011). To identify minimal domains within repressors involved in repressor-corepressor interaction and possibly to derive common sequence patterns we constructed GST fusions encoding length variants of repressors Yox1, Dal80, Xbp1, Ure2, Rox1, Mot3, Rdr1 and Sko1, which were used for binding studies with epitope-tagged corepressors. Similar studies have been previously described for Mig1 (Östling et al. 1996), Ume6 (Kadosh and Struhl 1997), Cti6 (Aref and Schüller 2020), Fkh1 (Aref et al. 2021) and Gal80 (Lettow et al. 2022). No mapping studies were performed with Nrg1, Yhp1 and Whi5.

Yox1

The homeodomain-containing repressor Yox1 is phosphorylated by the cyclin-dependent kinase Cdc28/Cdk1 and counteracts gene activation by the MADS box transcription factor Mcm1, which is responsible for expression of genes with an ECB promoter motif (early cell cycle box; Pramila et al. 2002). Since Yox1 mediated strong gene repression in vivo and interacted with corepressors Sin3, Cyc8 and Tup1 (Fig. 1), we performed a detailed analysis of its molecular functions. Length variants of Yox1 were fused with GST and incubated with bacterial protein extracts containing Sin3, Cyc8 and Tup1, respectively. As is shown in Fig. 4, a short region comprising aa 220–280 of Yox1 following its homeodomain was able to interact with Sin3, Cyc8 and Tup1 in vitro and could also repress expression of a lexAOp-dependent reporter gene in vivo when fused with LexA. Corepressor interaction was also observed with an even shorter Yox1 domain (aa 235–280). For a more precise analysis, we introduced missense mutations into Yox1 length variant aa 220–280 at selected positions, focusing on several hydrophobic amino acids at positions reminiscent of a heptad-like sequence pattern. Gene repression was strongly weakened when residues V257, L262, V266 and I270 L271 were replaced with alanine (Fig. 4). As is evident from a double mutation replacing R272 D273, not only hydrophobic residues are important for functional repression, indicating that at least Yox1 aa 257–273 define the functional core of its repression domain.

Fig. 4

Molecular analysis of Yox1 interaction with corepressors and gene repression by lexA fusions. Yox1 length variants were fused with GST, immobilized on GSH sepharose and incubated with bacterial protein extracts, containing HA-Sin3 (pJL34; aa 1-480, comprising PAH1 and PAH2), HA-Cyc8 (pFK77, aa 1-398, encoding TPR motifs) or HA-Tup1 (pFK76; full-length). GST expression plasmids and lexA fusion plasmids encoding length variants or missense mutations of Yox1 are compiled in Supporting Online Material (Table S2). Gene repression in vivo was measured in transformants of strain RTS-lexA (reporter gene lexAOp-CYC1-lacZ). Empty vector pRT-lexA devoid of effector domains was used as a negative control for maximal reporter gene expression. Specific β-galactosidase activities are given in nmol oNPG hydrolyzed per min per mg of protein. Protein extracts prepared from at least 12 independent transformants were assayed. HD homeodomain, n. t. not tested, RF repression factor, SD standard deviation,  +  in vitro interaction, - no interaction

Dal80

The zinc finger repressor Dal80 (degradation of allantoin; Cunningham and Cooper 1993) negatively regulates genes required for acquisition of nitrogen from poor sources by binding to URSGATA sequence motifs but its mechanism of repression has not been described previously. As demonstrated above (Fig. 1), Dal80 can interact with corepressors Sin3, Cyc8 and Tup1.

GST fusions of Dal80 length variants showed that its N-terminal part comprising the zinc finger domain was unable to interact with corepressors (Fig. 5a). In contrast, truncations representing the C-terminus and internal sequences could bind to Sin3 and Cyc8. Indeed, aa 151–200, which are common to these truncations were able to interact with three corepressors and could also mediate strong gene repression in vivo (repression factor 7.3), which is substantially more effective than observed with entire Dal80. Possibly, full-length Dal80 contains sequences counteracting maximal gene repression executed by the minimal core domain aa 151–200.

Fig. 5

Molecular analysis of corepressor interaction in vitro and gene repression in vivo by lexA fusions of repressors Dal80 (a), Xbp1 (b) and Ure2 (c). Length variants of repressor proteins were fused with GST, immobilized on GSH sepharose and incubated with bacterial protein extracts, containing HA-Sin3 (pJL34; aa 1-480, comprising PAH1 and PAH2), HA-Cyc8 (pFK77, aa 1-398, encoding TPR motifs) or HA-Tup1 (pFK76; full-length). GST expression plasmids and lexA fusion plasmids encoding length variants of Dal80, Xbp1 and Ure2 are compiled in Supporting Online Material (Table S2). Gene repression in vivo was measured in transformants of strain RTS-lexA (reporter gene lexAOp-CYC1-lacZ) which contain plasmids encoding lexA fusions. Empty vector pRT-lexA was used as a negative control for maximal reporter gene expression. Specific β-galactosidase activities are given in nmol oNPG hydrolyzed per min per mg of protein. Protein extracts prepared from at least 12 independent transformants were assayed. ZF zinc finger, MR Mbp1-related DNA-binding domain, n. t. not tested, RF repression factor, SD standard deviation,  +  in vitro interaction, - no interaction

Xbp1

The DNA-binding repressor Xbp1 prevents expression of cyclin genes of the G1 phase and thus causes delay of cell cycle progression (Mai and Breeden 2000). Xbp1 contains an Mbp1-related DNA-binding domain and was shown to interact with Sin3 but not with Cyc8 and Tup1 (Fig. 1). In contrast to other repressor proteins studied here, Xbp1 contains at least two non-overlapping domains mediating corepressor interaction and gene repression in vivo (aa 1–430 and aa 430–647; Fig. 5b). Using additional Xbp1 truncations, we could finally map a repression domain of 74 aa at the ultimate C-terminus of the protein. No further mapping was done for the length variant aa 1–430.

Ure2

This negative regulator of nitrogen catabolite repression is devoid of an obvious DNA-binding domain but contains two domains related to glutathione-S-transferases (GST) of unknown significance. In the presence of a favorable nitrogen source, Ure2 prevents gene activation by GATA factors Gln3 and Gat1 (Cooper 2002). Ure2 can convert into the conformational variant [URE3], forming polymeric amyloid prion-like filaments being able to self-propagate but unable to mediate gene repression (Baxa et al. 2002). Interestingly, the prion-forming domain of Ure2 (aa 1–80; Baxa et al. 2002) is part of its Cyc8 and Tup1 interaction domain, which also mediates efficient gene repression in vivo (5.2-fold; Fig. 5c). In contrast, GST-related domains of Ure2 (aa 106–228 and aa 201–354) were unable to bind corepressors.

Rox1

Oxygen limitation (hypoxia) requires activation of certain respiratory enzymes such as cytochrome and cytochrome oxidase which, however, are repressed by Rox1 under aerobic conditions. Rox1 contains an N-terminal HMG box sequence motif (high mobility group) mediating its binding to promoter sites upstream of hypoxic genes such as ANB1 and CYC7 (Deckert et al. 1999). Rox1 is able to interact with corepressors Sin3, Cyc8 and Tup1 in vitro, utilizing a repression domain comprising amino acids 124–200 (Supporting Online Fig. 1a).

Mot3

Similar to Rox1, the zinc finger protein Mot3 is required for repression of certain hypoxic genes among which are genes of ergosterol biosynthesis (Hongay et al. 2002). As shown above (Fig. 1), Mot3 binds to Cyc8 and Tup1 in vitro, mediated by two non-overlapping domains before and behind its zinc finger motif. However, efficient repression in vivo can be observed only with aa 231–347, which are thus considered as a genuine repression domain (Supporting Online Fig. 1b).

Rdr1

PDR5 encoding an ABC transporter of the plasma membrane responsible for efflux of various drugs is negatively regulated by the Zinc cluster repressor Rdr1 (Hellauer et al. 2002). Our mapping results show that an internal domain comprising aa 364–455 of Rdr1 mediate interaction with Sin3 (Supporting Online Fig. 1c).

Sko1

Initially, Sko1 was identified as a negative regulator of the invertase gene SUC2 with a CREB-related leucine zipper (Nehlin et al. 1992). However, its major function is regulation of genes required to counteract hyperosmotic stress (such as the sodium exporter ENA1). Importantly, Sko1 can be converted into an activator by Hog1-dependent phosphorylation, which is now able to recruit complexes SAGA and SWI/SNF to its target promoters (Proft and Struhl 2002). In this work we demonstrate that Sko1 interaction with Cyc8 requires a central domain (aa 201–427), which also mediates gene repression in vivo (Supporting Online Fig. 1d).

Using experimental results from this work and mapping data previously described, we now can compile sequences responsible for corepressor recruitment from 16 repressor proteins of S. cerevisiae (Table 2).

Table 2 Compilation of corepressor-recruiting and gene repression-mediating domains within 16 transcriptional repressor proteins of S. cerevisiae