Rtg3p, a Basic Helix-Loop-Helix/Leucine Zipper Protein that Functions in Mitochondrial-induced Changes in Gene Expression, Contains Independent Activation Domains*

Rtg3p and Rtg1p are basic helix-loop-helix/leucine zipper protein transcription factors in yeast that interact and bind to sites in an upstream activation sequence element in the 5′-flanking region of CIT2, a gene encoding a peroxisomal isoform of citrate synthase. These factors are required both for basal expression of CIT2 and its elevated expression in cells with dysfunctional mitochondria, such as in respiratory-deficient petite cells lacking mitochondrial DNA (ρ°). This elevated expression of CIT2 is called the retrograde response. Here we show that fusion constructs between the Gal4p DNA binding domain and Rtg3p transactivate the expression of aLacZ reporter gene under the control of a GAL1promoter element. We have identified two activation domains in Rtg3p: a strong carboxyl-terminal domain from amino acids 375–486, and a weaker amino-terminal domain from amino acids 1–175; neither of these activation domains contain the bHLH/Zip motif. We have also identified a serine/threonine-rich domain of Rtg3p within amino acids 176–282 that is inhibitory to transactivation. In addition, the transcriptional activity of the Gal4-Rtg3p fusion proteins does not require either Rtg1p or Rtg2p; the latter is a protein containing an hsp70-like ATP binding domain that is also necessary for CIT2 expression. In contrast, transcriptional activation by Gal4-Rtg1p fusion proteins requires the Rtg1p basic helix-loop-helix/leucine zipper protein domain, as well as Rtg3p and Rtg2p. These data suggest that transcriptional activation by the Rtg1p-Rtg3p complex is largely the function of Rtg3p. Experiments are also presented suggesting that Rtg3p is limiting for gene expression in respiratory-competent (ρ+) cells.

Rtg3p and Rtg1p are basic helix-loop-helix/leucine zipper protein transcription factors in yeast that interact and bind to sites in an upstream activation sequence element in the 5-flanking region of CIT2, a gene encoding a peroxisomal isoform of citrate synthase. These factors are required both for basal expression of CIT2 and its elevated expression in cells with dysfunctional mitochondria, such as in respiratory-deficient petite cells lacking mitochondrial DNA (°). This elevated expression of CIT2 is called the retrograde response. Here we show that fusion constructs between the Gal4p DNA binding domain and Rtg3p transactivate the expression of a LacZ reporter gene under the control of a GAL1 promoter element. We have identified two activation domains in Rtg3p: a strong carboxyl-terminal domain from amino acids 375-486, and a weaker amino-terminal domain from amino acids 1-175; neither of these activation domains contain the bHLH/Zip motif. We have also identified a serine/threonine-rich domain of Rtg3p within amino acids 176 -282 that is inhibitory to transactivation. In addition, the transcriptional activity of the Gal4-Rtg3p fusion proteins does not require either Rtg1p or Rtg2p; the latter is a protein containing an hsp70-like ATP binding domain that is also necessary for CIT2 expression. In contrast, transcriptional activation by Gal4-Rtg1p fusion proteins requires the Rtg1p basic helix-loop-helix/leucine zipper protein domain, as well as Rtg3p and Rtg2p. These data suggest that transcriptional activation by the Rtg1p-Rtg3p complex is largely the function of Rtg3p. Experiments are also presented suggesting that Rtg3p is limiting for gene expression in respiratory-competent ( ؉ ) cells.
Eukaryotic cells are able to monitor and respond to changes in mitochondrial function through changes in nuclear gene expression (1)(2)(3)(4)(5). We have termed this pathway retrograde regulation, and have suggested that it operates as a homeostatic mechanism for accommodating various cellular activities to alterations in mitochondrial function (6). One example of retrograde regulation in the yeast Saccharomyces cerevisiae is the elevated expression of the CIT2 gene in cells with dysfunctional mitochondria (7). CIT2 encodes a peroxisomal isoform of citrate synthase (8,9), which functions in the glyoxylate cycle in one of two steps that utilize acetyl coenzyme A. The glyoxylate cycle enables cells to carry out a net conversion of acetate, available either as a sole carbon source or produced via ␤-oxidation, to carbohydrate by the shuttling of glyoxylate cycle intermediates (e.g. succinate) from the peroxisomes to mitochondria (10). Under conditions where the activity of the tricarboxylic acid (TCA) 1 cycle is limiting, an increased activity of the glyoxylate cycle would allow for a more efficient use of two carbon compounds through anaplerotic pathways. We have found that CIT2 transcription is activated as much as 30-fold in cells that have blocks in the TCA cycle or are respiratorydeficient, for example, because of a complete loss of mitochondrial DNA ( o petites) (7,11,12).
We have identified three genes, RTG1, RTG2, and RTG3, that are required both for basal and retrograde-regulated expression of CIT2 (11,13). Surprisingly, these genes are also required for oleic acid-induced peroxisomal proliferation and the induction of enzymes of the ␤-oxidation pathway (12,14), both of which occur when oleic acid is present in the growth medium (15)(16)(17). Thus, the RTG genes appear to play a pivotal role in controlling metabolic interactions between mitochondria and peroxisomes.
RTG2 encodes a cytoplasmic protein containing an hsp70 type of ATP binding domain (18). Experiments to be presented elsewhere show that Rtg2p has properties of a heat shock protein that may function as a molecular chaperone in the control of gene expression. 2 RTG1 and RTG3 encode basic helix-loop-helix leucine zipper (bHLH/Zip) transcription factors (11,13). Rtg1p has some unusual features for a bHLH protein; in addition to an unusually long loop domain of some 39 amino acid residues, it has a very short basic DNA binding domain that lacks certain conserved amino acids found in the basic region of most other bHLH proteins. Those residues are known to make specific base contacts in the major groove of target site DNA (19 -21). Rtg3p, on the other hand, has the features of a typical bHLH/Zip protein, including the conserved amino acids in the basic region of the bHLH motif. Recently, we showed that Rtg1p and Rtg3p bind, most likely as a heterodimer, to sequences in a 76-bp UAS r element in the CIT2 promoter (13). The minimal DNA target site for binding of the Rtg1p-Rtg3p complex is not, however, an E box (22,23), CANNTG, which is the core binding site for the great majority of bHLH transcription factors, but a novel sequence, GGTCAC, which we have called an R box (13). The CIT2 UAS r element contains two such R boxes arranged as an inverted repeat separated by 28 bp of an AT-rich region. Both R box sites are required for full CIT2 expression and act synergistically in vivo (13).
In previous experiments, Rtg1p was tested for its ability to activate transcription by fusing to it the DNA binding domain of Gal4p (24). In wild-type cells, the Gal4-Rtg1p chimera was able to activate transcription of an integrated LacZ reporter gene under the control of a GAL1 promoter element (UAS G ). Moreover, the activity showed a typical retrograde response, i.e. expression was greater in °petites than in ϩ cells. Transcriptional activation by Gal4-Rtg1p was absolutely dependent upon the presence of Rtg2p and at least one other protein, which we suggested was a bHLH protein that interacted with Rtg1p. That protein was subsequently identified as Rtg3p (13). To characterize further the role of Rtg1p and Rtg3p in the control of CIT2 expression, we have made a variety of Gal4p fusions to Rtg3p. Here we show that Rtg3p contains at least two transcriptional activation domains that can function independently of Rtg1p and Rtg2p. We have also identified an apparent inhibitory domain of Rtg3p transactivation rich in serine and threonine residues that could be potential phosphorylation sites of the protein. We suggest that Rtg3p may be limiting for CIT2 expression in respiratory-competent ( ϩ ) cells and that Rtg2p could function to recruit Rtg3p for interaction with Rtg1p.
Yeast Transformation, Cell Extracts, and ␤-Galactosidase Assays-Yeast were transformed using lithium acetate and heat shock (26). Transformants carrying the desired plasmid or plasmid pairs were selected for on YNBD supplemented with the necessary amino acids. Liquid precultures were inoculated with a pool of 15-20 independent transformants and grown in selective YNBG ( ϩ ) or YNBR (°). Selective YNBR cultures were inoculated at a low density (0.01 A 600 ), grown at 30°C and harvested at mid-log phase (ϳ0.9 A 600 ). Cell extracts and assays were carried out as described by Rose and Botstein (27). For each plasmid-strain combination, these extracts was assayed in triplicate.
Northern Blot Analysis-COP161 U7 ϩ and °strains were transformed with an empty control plasmid, or plasmid carrying RTG1 or RTG3 (either 2 or CEN). Selective YNBG ( ϩ ) or YNBR (°) precultures were used to inoculate selective YNBR cultures at a low density (0.01 A 600 ). Cells were harvested at mid-log phase and total RNA was isolated (28). RNA was fractionated on an agarose gel and probed for actin and CIT2 mRNA abundance as described previously (7). Signals were quantified using a Molecular Dynamics PhosphorImager.

Transactivation by Gal4-Rtg1p Is Dependent on Rtg2p and
Rtg3p-In previous experiments, we showed that a chimeric protein consisting of the DNA binding domain of Gal4p fused to Rtg1p was able to activate expression of an integrated LacZ reporter gene under the control of a Gal4p-responsive UAS G element (24). However, those studies revealed that Rtg1p by itself could not transactivate, but appeared to require, in addition to Rtg2p, one or more proteins for transcriptional activity. That conclusion was strengthened by the identification of a new bHLH/Zip protein encoded by the RTG3 gene, which is also essential for CIT2 expression, and the finding that Rtg3p binds with Rtg1p at two sites in the CIT2 UAS r (13). To determine whether Rtg3p is also required for transactivation mediated by the Gal4-Rtg1p chimeric protein, where Rtg1p is directed to the USA G element in the reporter gene solely by the Gal4p DNA binding domain, SFY526⌬rtg3 cells were transformed with p2H26 -177 encoding the Gal4p DNA binding domain fused directly to the HLH domain of Rtg1p (Fig. 1A). The ␤-galactosidase activity of this transformant was then compared with p2H26 -177 transformants of wild-type SFY526 cells and its ⌬rtg1 and ⌬rtg2 derivatives. Fig. 2 shows that there is a retrograde response in the transactivation by p2H26 -17 in SFY526 transformants, i.e. activity was greater in respiratory-deficient °petites than in ϩ cells of that strain. In addition, activity both in ϩ and in °c ells was comparably greater in SFY256⌬rtg1 cells, presumably because endogenous Rtg1p was not present to compete for limited interacting proteins required for transactivation of the Gal4-Rtg1p fusion protein. In contrast to these results, no transcriptional activation was observed in either SFY526⌬rtg2 or SFY256⌬rtg3 transformants. Altogether, regulation of expression of the UAS G -LacZ reporter gene by Gal4-Rtg1p resembles transcriptional control of the CIT2 gene, whereby expression is elevated in °cells, and is dependent on RTG1 and RTG3 in both ϩ and °cells (11,13). These results are consistent with the notion that transactivation occurs via an Rtg1p-Rtg3p heterodimeric complex, and that Rtg3p contains one or more transactivation domains.
Although Rtg2p is required for Gal4-Rtg1p-mediated transactivation of the UAS G -LacZ reporter gene, it is unlikely to be involved directly in transcriptional activation. First, we have fused full-length Rtg2p to the Gal4p DNA binding domain and found that this construct (pGBT-RTG2) by itself did not transactive UAS G -LacZ reporter gene expression in wild-type cells (Table I, part A). Second, in a standard two-hybrid experiment testing protein-protein interactions (29), when pGBT-RTG2, encoding the Gal4p DNA binding domain fused to Rtg2p, was co-transformed with either pGAD-RTG1 or pGAD-RTG3 encoding, respectively, Rtg1p or Rtg3p fused to the Gal4p activation domain, no increase in reporter gene expression was detected. This result indicates that there is no direct interaction between Rtg2p and either Rtg1p or Rtg3p.
Gal4-Rtg3p Can Transactivate UAS G -LacZ Reporter Gene Independent of Rtg1p or Rtg2p-The results of the previous experiments suggest that, for transactivation of the UAS G -LacZ reporter gene, either there is a necessary interaction between Rtg1p and Rtg3p, or Rtg3p by itself is sufficient for transcriptional activation. To distinguish between these possibilities, we constructed a series of chimeric proteins consisting of the Gal4p DNA binding domain fused to various domains of Rtg3p, including the full-length protein (Fig. 1B). These con-structs were then tested for their ability to transactivate expression of the UAS G -LacZ reporter gene in wild-type and mutant SFY526 cells. Fig. 3 shows that the full-length Gal4-Rtg3p chimera, p3GB1-486 (Fig. 1B), could transactivate in wild-type cells. Unlike Gal4-Rtg1p, however, transactivation did not show a typical retrograde response of higher activity in °cells. Moreover, Gal4-Rtg3p could transactivate in all three ⌬rtg mutant backgrounds as well as, or better than, in wildtype cells. Except for expression in ⌬rtg1 cells, activity was lower in the °petite than in ϩ cells. These results are in sharp contrast to Gal4-Rtg1p transactivation, which requires both Rtg2p and Rtg3p and which consistently shows a greater activity in °cells. From these experiments, we conclude that Rtg3p itself contains one or more domains responsible for transcriptional activation, but these are not by themselves responsive to retrograde control.
Rtg3p Contains Independent Amino-and Carboxyl-terminal Activation Domains-To define the activation domain(s) of Rtg3p, we constructed a series of Gal4-Rtg3p fusions linking different portions of the protein to the DNA binding domain of Gal4p (Fig. 1B). These constructs were then tested for activity in ϩ and °derivatives of strain SFY526. The results of these experiments (Fig. 4) show, first, that there are at least two independent activation domains in Rtg3p, a strong COOHterminal domain (amino acids 375-486) and a weaker NH 2 - terminal domain (amino acids 1-175). Second, neither of these domains contain the HLH/Zip protein dimerization motif, indicating that transactivation does not require an HLH/Zip-mediated protein-protein interaction. Although there was somewhat greater activity in construct pGB300 -486 containing the HLH/ Zip domain, these results, nevertheless, contrast sharply with those obtained with the Gal4-Rtg1p fusions, where the HLH/ Zip domain was found to be essential for activity in UAS G -LacZ reporter gene expression (24).
A third significant outcome of these experiments is that the region of Rtg3p from amino acids 176 to 282 was not only inactive in supporting reporter gene expression but also appeared to be inhibitory to the activity of the NH 2 -and COOHterminal activation domains. For example, construct p3GB1-282 was less active than construct p3GB1-175, and construct p3GB176 -286 was considerably less active than p3GB300 -486 or p3GB375-486 (Figs. 1B and 4). It is noteworthy that the inactive, and apparently inhibitory, domain of Rtg3p from amino acids 176 -282 is very rich in serine and threonine residues, and thus may be potentially important as phosphorylation sites in the protein. Finally, as is the case with the fulllength fusion construct, all of the active, truncated Gal4-Rtg3p fusions were also active in ⌬rtg1, ⌬rtg2, and ⌬rtg3 cells (data not shown). From these data, we conclude that Rtg3p contains NH 2 -and COOH-terminal transactivation domains, which, when directed to the Gal4p DNA binding site, do not require the bHLH/Zip dimerization domain or the products of the two other known RTG genes for activity.
Rtg1p and Rtg3p Can Interact in the Absence of Rtg2p-The findings presented thus far suggest that Rtg3p is the functional transactivator of the Rtg1p-Rtg3p complex. Hence, reporter gene expression mediated by Gal4-Rtg1p would require recruitment of Rtg3p and interaction of these proteins via their bHLH/ Zip motifs. Gal4-Rtg3p, however, which contains independent transactivation domains, would not require Rtg1p for transactivation of reporter gene expression from a UAS G element. What then is the function of Rtg2p in transactivation? Rtg2p has no known DNA binding motifs, but it does contain an hsp70-like ATP binding domain (18) and has some properties of a heat shock protein and molecular chaperone, 2 such as a peptide-stimulated ATPase activity (30). Thus, Rtg2p might function to promote a direct interaction between Rtg1p and Rtg3p. To examine whether Rtg1p and Rtg3p can interact in the absence of Rtg2p when both proteins are targeted to the nucleus via plasmid-encoded heterologous nuclear localization signals, we used the construct, p2H26 -177, which fuses the Gal4p DNA binding domain to Rtg1p, in a two-hybrid assay together with a second plasmid, pGAD-RTG3, encoding fulllength Rtg3p linked to the Gal4p activation domain. The two constructs, p2H26 -177 and pGAD-RTG3, were examined singly (along with the control plasmids) and in combination for their ability to transactivate the UAS G -LacZ reporter gene in ϩ and °derivatives of wild-type, ⌬rtg3, and ⌬rtg2 SFY526 cells. Table I, part B, shows that, as observed previously (24), the construct encoding the Gal4p DNA binding domain fused to Rtg1p (p2H26 -177) was able to transactivate the reporter gene in wild-type SFY526 cells in a retrograde-responsive manner. By contrast, no transactivation was observed in cells transformed with pGAD-RTG3, which encodes the Gal4p activation domain fused to Rtg3p. However, in wild-type cells transformed with both p2H26 -177 and pGAD-RTG3, a high level of transactivation was observed that was 2-3.6-fold greater than with p2H26 -177 alone. In ⌬rtg3 cells, neither p2H26 -177 or pGAD-RTG3 alone showed any activity, but activity was observed in cells transformed with both plasmids, indicating an interaction between Rtg1p and Rtg3p. In ⌬rtg2 cells, some transactivation by p2H-177 was observed but the activity was only a fraction (10% or less) of that seen in wild-type cells. Importantly, strong transactivation was observed in ⌬rtg2 cells containing both p2H26 -177 and pGAD-RTG3, indicating that Rtg2p is not required for the direct interaction between Rtg1p and Rtg3p. These findings are consistent with previous observations that an Rtg1p-Rtg3p dependent complex is formed in electrophoretic mobility shift assays of the 76-bp CIT2 UAS r using whole cell extracts from a ⌬rtg2 strain (11) and that recombinant Rtg1p and Rtg3p can interact and bind to their target DNA sites in vitro (13). Finally, it is interesting to note that the response of the p2H26 -177/pGAD-RTG3 pair in all cases showed an elevated expression in °cells, suggesting that some component of the retrograde response is post-transcriptional, since both proteins were over expressed from constitutive promoters.
Overexpression of RTG3-Previous experiments showed that the level of Rtg1p detected by Western blot analysis does not differ significantly between ϩ and °cells (24), suggesting that the CIT2 retrograde response is unlikely to be controlled by a change in the level of Rtg1p. We have not been able to determine whether the CIT2 retrograde response might be con-

FIG. 4. Rtg3p has NH 2 -and COOH-terminal transactivating domains.
Wild-type SFY526 ϩ and °cells were transformed with the various Gal4-Rtg3p fusion constructs indicated in Fig. 1. The transformants were assayed for ␤-galactosidase activity as described in Fig. 2.   FIG. 3. Activation of LacZ reporter gene expression by a Gal4-Rtg3p fusion protein is independent of Rtg1p and Rtg2p. Wildtype SFY526 ϩ and °cells and their ⌬rtg1, ⌬rtg2, and ⌬rtg3 derivatives were transformed with the fusion plasmid, p3GB1-486, encoding the DNA binding domain of Gal4p fused to the full-length Rtg3p. ␤-Galactosidase activity was assayed as described in the legend to Fig. 2. trolled by changes in the level of Rtg3p because it is a very low abundance protein both in ϩ and °cells (data not shown). We have been able to detect the Rtg3p mRNA by RNase protection experiments, however, and observed that °cells contain about twice as much Rtg3p mRNA as ϩ cells (13).
To gain some insight into whether Rtg3p might be limiting in the retrograde response, we determined the effect of overexpression of Rtg3p on UAS G -LacZ reporter gene expression in cells containing the Gal4-Rtg1p fusion protein. Both ϩ and °d erivatives of wild-type and ⌬rtg2 SFY526 cells were transformed with p2H26 -177, and either p181-RTG3, a 2-based multicopy plasmid containing the full-length RTG3 gene under control of its own promoter or, as a control, the empty vector, yplac181. Table II shows that in the presence of the control plasmid, pGBT9, 181-RTG3 does not transactivate in either ϩ or °wild-type SFY526 cells or in the SFY526⌬rtg2 derivative. As expected, the Gal4-Rtg1p fusion protein encoded by p2H26 -177 transactivated in the presence of the control plasmid, yplac181, in wild-type SFY526 cells but not in the ⌬rtg2 derivative. When Rtg3p was overexpressed in the transformants containing 181-RTG3, the Gal4-Rtg1p-mediated transactivation was increased 2.8-fold in ϩ cells and 1.9-fold in °cells of the wild-type SFY526 strain. These data suggest that Rtg3p is limiting both in ϩ and °cells for Gal4-Rtg1p-mediated transactivation of reporter gene expression. Table II also shows that cells overexpressing Rtg3p (p181-RTG3) can overcome the failure of Gal4-Rtg1p to mediate transactivation in ⌬rtg2 cells, suggesting that Rtg2p is not required per se for Rtg3p function. These findings raise the possibility that Rtg2p is required to make Rtg3p available for interaction with Rtg1p.
Overexpression of RTG1 and RTG3 Can Increase Transcription from the CIT2 Promoter-The preceding experiments demonstrate that increasing expression of RTG3 can increase the activity of a Gal4-Rtg1p/Rtg3p complex directed to the Gal4p DNA binding sites in the UAS G . They do not, however, address whether increasing levels of Rtg3p can increase transcription from the CIT2 UAS r containing the natural DNA target sites for the Rtg1p-Rtg3p complex. To examine that point, as well as to determine whether overexpression of RTG1 also affects transcription from the CIT2 promoter, we transformed wild-type COP161U7 cells with the RTG1 or RTG3 genes on either a centromere-based (pRS426) or 2-based multicopy (yplac195) plasmids. In all cases, expression of RTG1 and RTG3 was under control of their own promoters. Total cellular RNA was isolated from log-phase cultures grown on YNBR ϩ casamino acids medium and analyzed by Northern blotting to determine CIT2 expression relative to an internal control for actin mRNA (Fig. 5). The data were quantified by PhosphorImager analysis, and the relative amount of CIT2 mRNA, normalized to the level of actin mRNA, is present below each lane.
In this series of experiments, the retrograde response for CIT2 expression (°/ ϩ ) was 5.6 ( Fig. 5, lanes 1 and 2). In ϩ cells, overexpression of Rtg1p from either the centromere or 2-based plasmid resulted in a less than 3-fold increase in CIT2 mRNA abundance and little or no increase in °cells (lanes 3-6). There was a somewhat larger increase in CIT2 mRNA when RTG3 was overexpressed in ϩ cells from the centromere plasmid (lane 7), and nearly a 10-fold increase in when RTG3 was overexpressed from the 2-based plasmid (lane 9). In °cells, overexpressing RTG1 or RTG3 from either vector led only to a relatively small increase in CIT2 expression above the control cells (compare lane 2 with lanes 4, 6, 8, and  10). These data suggest, first, that the retrograde response is effectively limited by Rtg3p and less so by Rtg1p, and second, that CIT2 transcription in °cells is close to the limit of expression that is regulated by the Rtg1p-Rtg3p heterodimeric complex.

DISCUSSION
Components of the CIT2 Transcription Apparatus-Transcription of the CIT2 gene requires the bHLH/Zip proteins, Rtg1p and Rtg3p, which bind as heterodimers to two identical, non-E box target sites (R boxes, GGTCAC) in the UAS r element of the CIT2 promoter (13); neither protein alone is able to bind to the R box in vitro or to activate CIT2 transcription in vivo. CIT2 transcription also requires Rtg2p, a protein with no obvious DNA binding motifs, that is related to the hsp70 family of heat shock proteins (18). 2 Based on two-hybrid analysis carried out in the current study, we conclude that Rtg2p neither binds to Rtg1p or Rtg3p nor is required for the direct interaction of these transcription factors with each other.
Transactivation by Rtg3p-To understand better how the Rtg proteins function both in basal and retrograde regulated expression of the CIT2 gene, we have extended our previous analysis of the transcriptional activity of fusion proteins between the Gal4p DNA binding domain and Rtg1p (24) to include a similar analysis of Gal4-Rtg3p chimeras; these were analyzed in wild-type ϩ and °cells and in different rtg mutant backgrounds. Such fusions allow an assessment of the transcriptional activity of Rtg3p and its various domains without requiring DNA binding to endogenous target sites. Our experiments show that Gal4-Rtg1p fusion proteins require Rtg3p and Rtg2p to activate transcription of a LacZ reporter gene. By contrast, Gal4-Rtg3p fusion proteins can transactivate reporter gene expression in the absence of either Rtg1p or   (31)(32)(33). From the analysis of a series of constructs of fusions of the Gal4p DNA binding domain to regions of Rtg3p, we identified two transactivation domains, one within the NH 2 -terminal region from amino acids 1 to 175 and the other within the COOHterminal region from amino acids 375 to 486. This organization is similar to that of the skeletal muscle specific bHLH transcription factor, myogenin, which also has NH 2 -and COOHterminal transactivation domains (34). It is important to note that neither of the Rtg3p activation domains contain the bHLH/Zip protein dimerization motif. This finding is consistent with the observation that Gal4-Rtg3p can activate transcription in the absence of Rtg1p. Conversely, transactivation by Gal4-Rtg1p requires both its bHLH/Zip domain (24) and Rtg3p, suggesting that in this one-hybrid assay the Rtg1p bHLH/Zip domain in the Gal4-Rtg1p fusions directs endogenous Rtg3p to the Gal4p DNA binding site for activation of reporter gene expression.
A Potential Regulatory Domain of Rtg3p-A central domain of Rtg3p from amino acids 176 to 282 that was tested for transactivation was not only inactive but also inhibitory to activation by other domains; constructs lacking that central domain had significantly more activity than the constructs containing it (for example, p3GB1-282 versus p3GB1-175 shown in Fig. 4). This apparently inhibitory domain of Rtg3p is very rich in serine and threonine residues and thus could contain potential phosphorylation sites that might play a role in regulating the activity of the Rtg1p-Rtg3p complex. That phosphorylation can play a role in regulating the activity of bHLH proteins is suggested from studies showing that there is a dimerization-dependent phosphorylation of the muscle-specific bHLH transcription factors myogenin (35) and MyoD (36). The potential for a direct regulatory role of this phosphorylation is suggested by the finding that mutations of the dimerization-dependent phosphorylation sites enhance the transcriptional activity of myogenin heterodimeric complexes (35). It will be of interest to learn whether the serine/threoninerich region of Rtg3p is a target for phosphorylation of the protein that might affect the transcriptional activity of the Rtg1p-Rtg3p complex.
The dioxin receptor, also known as the aryl hydrocarbon receptor, is another example of a bHLH protein with a regulatory domain that may bear on the activity of the Rtg1p-Rtg3p complex. The dioxin receptor is a ligand-responsive member of the bHLH/PAS family, which forms a heterodimer with the bHLH/PAS protein, Arnt, and binds to specific xenobiotic-responsive elements regulating transcription of genes involved in drug metabolism (37)(38)(39)(40). The dioxin receptor has been shown to have a strong COOH-terminal transactivation domain, which is attenuated by a central 82-amino acid ligand binding domain of the protein (41). It is interesting that this inhibitory domain of the dioxin receptor is a site where a regulator molecule, an hsp90, interacts to prevent dimerization with Arnt (42,43).
We suggest that Rtg3p, which is a low abundance protein both in ϩ and °cells, is recruited to the R box through its interaction with the more abundant and constitutively expressed Rtg1p. Although the details of how the complex contacts the R box sites remain to be established, it is likely that Rtg3p interacts with the CA dinucleotide of the R box because the basic region of the Rtg3p bHLH/Zip motif contains the critical conserved amino acid residues, such as an Arg, His, and Glu, with characteristic spacing that have been shown for other bHLH proteins to make specific contacts with the CA dinucleotide of the CANNTG E box half-site in the major groove of DNA (19,21,44). Those key amino acid residues are not present as such in the truncated basic domain of Rtg1p. It is also significant that the Rtg1p-Rtg3p complex does not bind in vitro to common E box sites that were tested under conditions where robust binding was observed to the R box (13). Thus, the unique structural characteristics of Rtg1p may be responsible for determining the unusual DNA binding specificity of this bHLH/ Zip complex.
What is the role of Rtg2p in the transcriptional process mediated by Rtg1p-Rtg3? Since overexpression of Rtg3p can elevate target gene expression in ϩ cells, it is conceivable that Rtg2p could function to modulate the expression of RTG3. This is unlikely, however, for a number of reasons. First, we have not detected any effect of an rtg2 null allele on the level of expression of RTG3 mRNA (data not shown). Second, the abundance of RTG3 mRNA does not vary by more than 2-fold between ϩ and °cells (13). And finally, electrophoretic mobility shift assays using whole cell extracts from a ⌬rtg2 mutant strain, in which there is no CIT2 expression, yield the same Rtg1p-Rtg3p-dependent complex with a CIT2 UAS r probe as do extracts from wild-type cells (11,13). An alternative possibly suggested by our data is that in ϩ cells, Rtg3p is sequestered in an inactive form by one or more proteins, and that Rtg2p functions to regulate the availability of Rtg3p for transcriptional activation. This would be analogous to the sequestration of transcription factors such as NF-B by I-B (reviewed by Verma et al. (45)), the E protein family of bHLH proteins by Id (46) and the bHLH/PAS dioxin-responsive and related Sim family of transcription factors by hsp90 (43, 47, 48). FIG. 5. The influence of increased copy number of RTG1 and RTG3 on CIT2 mRNA levels in ؉ and o cells. COP161U7 ϩ and °cells were transformed with a low copy centromere-based vector (pRS416) or a high copy 2-based vector (yplac195), both with a URA3 marker, and an insert containing either the RTG1 or RTG3 gene. The ϩ and °control cells were transformed with the empty pRS416 plasmid. Total RNA was isolated from mid-log phase cultures grown on selective YNBR ϩ casamino acids as described under "Materials and Methods." The blots were hybridized with probes specific for CIT2 and ACT1 as described previously (11). The hybridization signals were quantified using a Molecular Dynamics PhosphorImager and the ratio of the relative signals for CIT2 and ACT1 are indicated below the figure.
Overexpressing Rtg3p in ϩ cells, in which the majority of the endogenous protein might be sequestered in an inactive form, could result in an increase in gene expression by effectively titrating out the protein(s) responsible for the sequestration. The finding that overexpression of Rtg1p also resulted in some increase in CIT2 expression in ϩ cells, but to a lesser extent than was observed with Rtg3p, could be accounted for by a competition for the free and sequestered forms of Rtg3p. The presence of an hsp70-like ATP binding domain in Rtg2p (18) and observations that Rtg2p has the properties of a heat shock protein 2 raise the possibility that Rtg2p could function as a molecular chaperone to promote the formation of the Rtg1p-Rtg3p heterodimer for transcriptional activation.