Interaction of ATPase Swi2 with transcriptional activation domains
Null mutant alleles of SWI2 encoding the ATPase subunit of the SWI/SNF complex cause severe loss of INO1 transcription and thus auxotrophy for inositol, among other pleiotropic phenotypes (Peterson et al. 1991; Peterson and Herskowitz 1992). Thus, Swi2 may be a candidate for interaction with transcriptional activation domains of Ino2 triggering stimulated expression of INO1 under inositol-limiting conditions. Swi2 has been previously identified as a target of activation domains from the ubiquitous activator VP16 and activator Gcn4, stimulating expression of amino acid biosynthetic genes (Neely et al. 2002). However, a functional domain within Swi2 mediating activator interaction has not been identified, possibly due to stability problems with truncated protein variants. To investigate whether Swi2 can also interact with activation domains of Ino2, we epitope-tagged the coding region of SWI2 and used yeast protein extracts containing HA-Swi2 for in vitro-interaction studies (GST pull-down experiments) with immobilized GST fusions of TAD1 (aa 1–35) and TAD2 of Ino2 (aa 101–135). As is shown in Fig. 1A, both activation domains are able to bind Swi2. For a precise mapping of the activator-binding domain within Swi2, various length variants of SWI2 were fused with HA and assayed for synthesis of stable HA-Swi2 proteins in yeast and/or E. coli. With the exception of variant aa 1-307 which could be obtained from both organisms, only bacterially produced length variants turned out to be stable and were subsequently used for pull-down studies (Fig. 1B). We focused on Swi2 truncations covering its N-terminus because several functional domains such as HSA, ATPase/translocase, SnAC, AT-hook and bromodomain have been already identified in the remaining part of the protein. Indeed, an N-terminal length variant (aa 1-307) synthesized in yeast and E. coli could interact with Ino2 TAD1. Further truncations finally enabled us to identify a 70 aa length variant of Swi2 (aa 238–307) as an activator-binding domain (ABD, Fig. 1B) which could not only interact with both TADs of Ino2 but also with TADs of unrelated activators Gal4, Gcn4, Rap1, Aro80 and Swi5 (Fig. 1C). Nevertheless, we cannot completely exclude that Swi2 contains an additional ABD within regions of the protein which we have covered less detailed by our deletion analysis.
As an in vivo assay for Swi2-dependent gene activation, we used Gal4DBD-Ino2 TAD fusions in combination with a GAL1-lacZ reporter gene and compared reporter gene expression in isogenic SWI2 and swi2 strains. As is shown in Fig. 2A, TAD1 and TAD2 mediated strong activation of the reporter gene when SWI2 is functional. Activation by TAD1 was substantially weakened in the swi2 null mutant (to 13.4%) while TAD2 was only partially affected (reduction to 72.5%). Although both TADs of Ino2 were able to interact with Swi2, TAD1 was significantly more dependent on Swi2 than TAD2.
To characterize the Swi2 ABD in more detail, we introduced missense mutations at selected positions into the ABD coding region and comparatively investigated whether the in vitro interaction of the Swi2 variants obtained with Ino2 TAD1 is compromised. We compared Swi2 sequences from various yeasts (Supplementary Fig. S1) and selected nine combinations of two, three or four conserved amino acids (F248 T249, E251 Q252 S253, L256 K257 L263 K264, I260 T261, L266 V267 N268, K270 P271, V279 I280 Q281, H286 P287 F290 K291 and R292 M293) which were replaced by alanine residues using site-directed mutagenesis. All variants of Swi2 ABD could be efficiently synthesized in E. coli but none of them was significantly impaired for in vitro interaction with TAD1 of Ino2 (Supplementary Fig. S2). We assume that a considerable degree of functional redundancy may exist among critical amino acids as we have shown previously for interaction of Ino2 with TFIID subunit Taf12 (Hintze et al. 2017).
To investigate whether the ABD of Swi2 is important for the function of the SWI/SNF complex we constructed two SWI2ΔABD deletion variants. Using the single-copy plasmid pECW24 (ARS CEN URA3 SWI2) and SWI2-specific primers, ΔABD variants lacking aa 240–298 and aa 160–298, respectively, were obtained by inverse PCR. The resulting plasmids pECW28 and pECW30 together with pECW24 as a positive control were transformed into a swi2Δ mutant and tested for complementation of mutant phenotypes (growth in the absence of inositol, utilization of carbon sources raffinose, ethanol and galactose, respectively). As is shown in Fig. 2B, both SWI2ΔABD variants could fully restore pleiotropic growth deficiencies of the swi2Δ mutant, indicating that the ABD of Swi2 is functionally dispensable, presumably because other subunits of the SWI/SNF complex are able to compensate for its loss.
SWI/SNF subunits Swi1, Snf5 and Snf6 also interact with Ino2
It was previously shown that the Gcn4 TAD not only interacts with Swi2 but also with SWI/SNF subunits Swi1 and Snf5 (Neely et al. 2002). Swi1 contains a mostly α-helical ARID (AT-rich interaction domain, aa 405–506) being able to bind DNA non-specifically (Wang et al. 2012). The ARID is contained within an internal sequence of Swi1 (aa 329–657) for which Prochasson et al. (2003) demonstrated interaction with activation domains of Gcn4 and VP16. We thus examined whether this region can also bind to activation domains of Ino2. Using Gcn4 TAD as a positive control we could show that both TADs of Ino2 indeed interact with epitope-tagged Swi1 in vitro (aa 329–657; Fig. 3A; lanes 3–5). For a more precise mapping of the ABD, truncated length variants of Swi1 were constructed and assayed for binding to Ino2 TAD1. As is shown in Fig. 3B, Swi1 variant aa 428–606 was still able to interact with TAD1 efficiently while variant aa 329–531 (comprising the entire ARID) failed to bind. Assuming that the DNA-binding ARID and the Swi1 ABD do not functionally overlap, aa 507–606 may represent the essential core of the ABD. Mapping studies with Ino2 TAD2 gave identical results (not shown).
Snf5 contains an internal sequence similar to the human INI1 protein (integrase interactor 1; aa 453–672, depicted in Fig. 4B) which was initially identified as a binding partner of HIV integrase and later shown to function as a tumor suppressor (Kalpana et al. 1994). The same domain of Snf5 is also similar to the Sfh1 subunit of chromatin remodelling complex RSC (“Snf Five Homolog”). The N-terminal sequence of Snf5 (aa 1-334) was previously shown to interact with VP16 and Gcn4 activation domains (Prochasson et al. 2003). This sequence contains a glutamine-rich region of unknown function (aa 221–270: 94% Gln; Qn in Fig. 4B). As is apparent from Fig. 4A, aa 1-334 of Snf5 can be also bound by both TADs of Ino2 (lanes 3 and 4; TAD of Gcn4 as a positive control, lane 5). The same result was obtained with bacterially produced Snf5 aa 1-334 (Fig. 4B). We next synthesized shorter length variants of Snf5 and assayed for interaction with TAD1 of Ino2. The results of these studies argue for the existence of two activator binding domains within Snf5, ABD1 (aa 130–275) and ABD2 (aa 265–334; Fig. 4B). Although these domains have a short overlap (11 amino acids, 7 of which are glutamine), this region is certainly too small to form the core of an ABD, arguing for individual interaction of ABD1 and ABD2 with TADs.
Snf6 is an additional subunit of SWI/SNF which has been described as an interaction partner of activators Pho4 and Swi5 (Neely et al. 2002) but is devoid of sequences similar to known functional domains. We could demonstrate that TAD1 and TAD2 of Ino2 are also able to recruit Snf6 synthesized in yeast and in E. coli (Fig. 4C, lanes 3 and 4; using Pho4 as a positive control, lane 5), arguing for a direct interaction. This interaction may be mediated by Snf 6 aa 1-225 which similarly bind to Ino2 TAD1 and TAD2. However, we cannot rule out the existence of an additional ABD since aa 1-225 was the only truncated Snf6 variant investigated.
SWI/SNF subunit Swi3 has been also demonstrated to bind selected activator proteins such as Swi5 (Neely et al. 2002). We thus finally investigated whether epitope-tagged Swi3 (full-length protein) can interact with TADs of Ino2, using Swi5 as a positive control. However, these experiments gave no evidence for Swi3 as an interaction partner of Ino2 (not shown).
Mutational analysis of Sth1 interaction with transcriptional activation domains
Molecular anatomy of ATPases Swi2 and Sth1 of chromatin remodelling complexes SWI/SNF and RSC, respectively, is highly conserved. We thus reasoned that Sth1 may possibly also interact with TADs of Ino2, mediated by a region at a similar position as demonstrated above for Swi2. A length variant of Sth1 (aa 1-300) could be stably synthesized in S. cerevisiae and indeed interacted with both TADs of Ino2 (Fig. 5A). Using various Sth1 truncations of its N-terminus, we finally identified a length variant of 73 amino acids (aa 160–232) as activator-binding domain which could be synthesized in E. coli and in yeast and was able to bind Ino2 TAD1, arguing for a direct interaction which is not dependent on other proteins from S. cerevisiae (Fig. 5B). The ABD of Sth1 is contained within the so-called “scaffold II”-domain (aa 154–318) which was identified by cryo-EM analysis of the RSC complex (Patel et al. 2019; scaffold I and II together are part of the “body module”; Wagner et al. 2020). As described above for Swi2, we also studied whether additional unrelated TADs can interact with Sth1. As shown in Fig. 5C, TADs of activators Gal4, Rap1, Leu3 and Aro80 were indeed able to bind the ABD of Sth1. It should be mentioned that we cannot exclude the existence of an additional ABD within the remaining sequence of Sth1 which was not investigated in this work.
To investigate whether a variant of the essential STH1 gene lacking its ABD can functionally complement a sth1 null mutation, we used the plasmid shuffle strategy, introducing a centromeric URA3 STH1 rescue plasmid into a wild-type strain with subsequent deletion of the genomic STH1 copy to give strain KSY1. Using a centromeric LEU2 STH1 plasmid as a template, we next constructed two deletion variants by inverse PCR, lacking the core ABD (aa 160–232) and the more extended scaffold II-region (aa 154–318), respectively. Following transformation of these plasmids into strain KSY1, transformants were cultivated in the presence of FOA to select for loss of the URA3 STH1 rescue plasmid. As is shown in Fig. 6A, neither STH1(Δ160–232) nor STH1(Δ154–318) could functionally replace STH1, indicating that the deleted domains are indispensable for the function of the protein. The failure to complement a sth1 null mutation is not caused by instability of the resulting truncated Sth1 variants as epitope-tagged proteins could be detected in cellular extracts (Fig. 6B).
Although we hypothesize that the essential regions deleted in STH1(Δ160–232) and STH1(Δ154–318) mediate recruitment of Sth1 to DNA-bound activators, it cannot be excluded that other functions such as formation of the RSC complex are also affected. We thus investigated activator binding of Sth1 more precisely by introduction of missense mutations at selected positions. Since evolutionary conservation of amino acids may indicate functional importance, we aligned sequences from various yeasts similar to the ABD of S. cerevisiae Sth1 (aa 160–232; Supplementary Fig. S3). As previously shown for TAF subunits of TFIID, combinations of basic and hydrophobic residues may be important for coactivator recruitment (Hintze et al. 2017). We thus replaced such amino acids by alanine (Sth1 R198A I199A, R202A I203A, N212A L213A and N212A L213A G214A T215A Y216A S217A L218A, respectively), synthesized mutational Sth1 variants in E. coli and used them for in vitro interaction studies with TAD1 of Ino2. It should be emphasized that the heptapeptide motif aa 212–218 is almost completely conserved among the yeast species compared (Fig. S3). While binding of variant R202A I203A was unaffected, variants R198A I199A and N212A L213A could no longer interact with Ino2 TAD1 in vitro (Fig. 7A). This result differs from what we have shown above for the ABD of Swi2 which had turned out as resistant to mutational modification. To study the influence of these mutations in vivo, we constructed corresponding variants of full-length STH1 which were tested for functional complementation of a sth1 null mutation by plasmid shuffling (Fig. 7B). To exclude severe alterations of Sth1 conformation by alanine replacements, we compared structural predictions for wild-type and mutant variants using AlphaFold (Jumper et al. 2021). No significant difference is predicted for variants R198A I199A, R202A I203A and N212A L213A while the septuple variant affecting aa 212–218 may exhibit an extended α-helix, possibly as a result of conversion of a conserved glycine into alanine (Supplementary Fig. S4). Although double variants R198A I199A and N212A L213A were functional in vivo, a combination of both of them (giving the quadruple variant R198A I199A N212A L213A) could no longer replace wild-type STH1. This was also true for the septuple variant N212A L213A G214A T215A Y216A S217A L218A which had failed to interact with Ino2 in vitro (Fig. 7A). In summary we conclude that the Sth1 sequence aa 198–218 represents the core of the protein being required for interaction with transcriptional activators.
Identification of a multifunctional activator binding domain in ATPase Ino80
We have previously identified the pleiotropic ino80 mutation leading to activation defects of genes involved in phospholipid biosynthesis and various unrelated metabolic pathways (Ebbert et al. 1999). The corresponding INO80 gene encodes an ATPase (1489 aa) distantly related to Swi2 and Sth1 but similar to Swr1 (Bao and Shen 2007). The phenotype of ino80 mutations indicates that Ino80 (or other subunits of the corresponding complex) may also interact with Ino2. Using an epitope-tagged variant of full-length Ino80, we were indeed able to demonstrate its binding to both TADs of Ino2 (Fig. 8A). For a more precise mapping of the domain mediating activator binding, we again used length variants of Ino80 and finally identified aa 455–620 as its ABD which was functional when synthesized either in yeast or in E. coli (Fig. 8B). Importantly, this sequence co-localizes with the previously identified DBINO (aa 504–601; Bakshi et al. 2004; Shen et al. 2003) and the HSA domain interacting with actin and actin-related proteins (aa 462–598; Szerlong et al. 2008). Not only TADs of Ino2 but also activators Rap1, Leu3, Gcn4 and Swi5 were able to interact with the ABD of Ino80 (Fig. 8C).
Although an INO80 gene deletion variant lacking sequences which encode amino acids 356–682 was unable to complement an ino80 null mutation (Shen et al. 2003), we constructed an INO80 variant devoid of its ABD as defined in this work (aa 455–620). The resulting protein could be stably synthesized in yeast (Fig. 9B) but failed to complement an ino80Δ mutation (Fig. 9A), confirming the functional importance of Ino80 aa 455–620. Because of the versatile function of this domain, we cannot conclude that the observed deficiency of Ino80Δ455-620 is a result of missing activator recruitment. Similar to what we found for various subunits of SWI/SNF, other subunits of INO80 (such as the poorly characterized proteins Ies1-Ies6) may be also contact partners of activator proteins.