INTRODUCTION

Top Previous Next References

Reb1p of Saccharomyces cerevisiae is an essential protein of 810 amino acids (28), with an unusual DNA binding region consisting of two Myb-like domains separated by nearly 150 amino acids (42). It binds to the 5' regions of numerous genes (8, 19, 26, 38, 46, 49, 54, 61). While the core consensus sequence for Reb1p binding has been identified as CGGGTAA (9, 37, 41, 61), substantial flexibility is permissible, and the adjacent nucleotides influence the binding affinity. (Reb1p has also been studied under the names Grf2p [9] and QBP [4].)

Reb1p plays a role in transcription by RNA polymerase II (Pol II). For instance, it is the major transcription factor for the ACT1 gene (38). Reb1p cooperates with other transcription factors, i.e., Rap1p and Abf1p, in regulating transcription of several glycolytic genes (46, 54). A Reb1p binding site is necessary for the threefold induction of the ILV1 gene in cells deprived of isoleucine and valine (49). In many other cases, Reb1p modestly influences the expression of genes by as-yet-undetermined mechanisms. In certain cases at least, Reb1p appears to play a role in chromatin organization. With a footprint of only 20 to 25 bp, the binding of Reb1p to sequences in the GAL1-10 upstream activation sequence (UAS) creates a 200-bp nucleosome-free region (17), although this is not the case for the UAS of the HSC82 gene (13).

Reb1p also plays one or more roles in the transcription of rRNA by RNA Pol I. It binds not only to the transcriptional enhancer, immediately downstream of the 3' end of the 35S rRNA transcript, but also just upstream of the promoter, about 200 nucleotides 5' of the start site of transcription (40, 41). In a construct placed within the ribosomal DNA (rDNA) locus, deletion of the two Reb1p binding sites causes a substantial reduction in transcription (31). However, deletion of the Reb1p binding sites has little effect on Pol I transcription of test genes on a variety of plasmid constructs (6, 32). These results imply that Reb1p might be involved in the three-dimensional arrangement of the rDNA within the nucleolus (31, 62). The binding of Reb1p has also been implicated in specifying the termination of the rRNA transcript by causing the polymerase molecule to pause. Release of the completed transcript soon follows (33-35). The mammalian counterpart to Reb1p is TTF1, identified by the Grummt laboratory as binding to the "Sal boxes" downstream of the human and mouse rRNA genes (3, 14, 15, 22). The binding of TTF1 not only is responsible for the termination of Pol I transcription (3, 14, 15) but also plays a role in blocking the DNA replication fork initiated from the intergenic region of rRNA genes (18). As in S. cerevisiae, the mouse rRNA genes have a TTF1 binding site just upstream of the Pol I promoter (21). The binding of TTF1 at that site can overcome chromatin-mediated repression of rRNA transcription (36), suggesting that TTF1 also shares with Reb1p the ability to exclude nucleosomes.

In this report on the regulation of transcription of the REB1 gene, we show that Reb1p binds to three sequences of the 5' region of its own gene. Two of the sites are upstream of the transcription initiation site. Each contributes to effecting transcription. Deletion of both abolishes transcription. The third is 30 bp downstream of the site of transcription initiation. Binding of Reb1p to the downstream site reduces transcription, probably by preventing the formation of a functional initiation complex. The concentration of REB1 mRNA is, on average, 1 molecule per cell (60, 63). The statistical fluctuation of the number of molecules in an individual cell could lead to substantial fluctuation in the concentration of Reb1p. We suggest that this arrangement of sites at the REB1 gene has evolved to minimize such fluctuations.

(This research is from a thesis submitted to the Sue Golding Division of the Albert Einstein College of Medicine in partial fulfillment of the requirements for the Ph.D. degree by Kevin L.-C. Wang.)

MATERIALS AND METHODS

Top Previous Next References

Strains, media, and general methods. All strains of S. cerevisiae used in this report are derived from W303a (MATa ade2-1 his3-11,15 leu2-3,112 trp1-1 ura3-1 ura3-1 can1-100) (59) and are listed in Table 1. Yeast cells were grown at 30°C in either YPD (1% yeast extract, 2% peptone, and 2% dextrose) or synthetic complete minimal medium with the necessary supplements. Strain DH5 was used for all DNA manipulations in Escherichia coli. Recombinant DNA methods and yeast manipulations are based on established protocols (50, 52).

Plasmids. Plasmids used in this study are listed in Table 2. Plasmid YIp357R (24) was used as the backbone for the integrating lacZ reporter gene constructs. A 1.2-kb DNA fragment of REB1, including 615 bp of the coding sequence, was subcloned to the BamHI and EcoRI sites of YIp357R, yielding a translational fusion of the REB1'-lacZ reporter gene in plasmid pLC6. This is the parental plasmid for all the other mutant constructs of the Reb1p binding site (Fig. 1A).

View larger version (33K): [in this window] [in a new window]   FIG. 1.   Elements controlling REB1 transcription. (A) Three potential Reb1p binding sites upstream of the REB1 ORF. The sequence from 200 to +158 is drawn to scale. The transcription initiation site (28) is designated as 1. A 1.2-kb DNA fragment of REB1, including 615 bp of the coding sequence, was fused to E. coli lacZ as a reporter gene. The orientation and relative distances to the transcription initiation site of the Reb1p binding sites A, B, and C are diagrammed in the REB1'-lacZ construct. (B) Sequence of the region upstream of the REB1 ORF. The A, B, and C sites are indicated. A consensus sequence of Reb1p binding sites is indicated in the lower box (37, 41). Mutations generated at individual sites are designated mtA, mtB, and mtC. The mutated nucleotides of each site are indicated with lowercase letters. Each mutated site was created to embed a restriction enzyme site for convenient verification: SpeI for mtA, NdeI for mtB, and SspI for mtC. For the C+75 construct, the original C site was mutated and a sequence identical over 15 bp with the original C site was constructed as shown at a position 105 bp downstream of the transcription initiation site.

Site-directed mutagenesis by PCR. Mutations were introduced at the sites indicated in Fig. 1B by the megaprimer method of PCR mutagenesis (1). In order to screen the mutated Reb1p binding sites easily, new restriction enzyme sites were created as follows: SpeI for the mutation at the A site, NdeI for the B site, and SspI for the C site. The resulting sequences, with mutated Reb1p binding sites, were used to replace the corresponding region of pLC6 (YIp357R-REB1'-lacZ) by an XbaI fragment (162 to +240), which includes about 80 bp downstream of ATG. All the mutations were confirmed by sequencing. All the PCR-based mutagenesis procedures in this study were designed to substitute nucleotides without altering the relative spacing of the Reb1p binding sites.

Construction of integration reporter genes and strains. For the reporter gene constructs with REB1'-lacZ in the YIp357R vector (listed in Table 1), the unique StuI site within the URA3 gene was used to linearize the REB1'-lacZ constructs, which were then transformed into WY119 for targeted integration to the URA3 locus (Table 1) (51). Ura⁺ cells were selected for Southern blot analysis to screen for single-copy integration of reporter genes. Two to three independent clones with a single integrated reporter gene were used for the -galactosidase activity assays (see Table 3). Triplicate data from every clone used for the -galactosidase assays were averaged to determine the final activity.

Southern and Northern blot analysis. Both methods followed protocols described previously (50, 52). Genomic DNA was isolated from mid-log-phase cells, digested with HindIII, separated by electrophoresis on a 1% agarose gel, and transferred to a nitrocellulose membrane. The blots were probed with a [³²P]dCTP-labeled fragment of the 1.1-kb URA3 gene. The results of Southern blot analysis were used to screen and verify the single-copy integration of the REB1'-lacZ reporter gene at the URA3 locus.

Protein purification. To construct plasmids for overexpressing Reb1p in the bacterial strain BL21(DE3), the 2.4-kb coding sequence of REB1 was generated by PCR by using Pfu DNA polymerase (Stratagene) with NdeI and BamHI sites for subcloning to pET11a (58). His₆-Reb1p was generated in a similar way by subcloning the coding sequence of REB1 to pET-His (10) at the XhoI and BamHI sites, leading to a fusion protein with six histidine residues fused to the N terminus of Reb1p. Partially purified Reb1p was prepared by 75% ammonium sulfate precipitation as previously described (42). His₆-Reb1p was further purified by using nickel-nitriloacetic acid resin according to the procedures of the manufacturer (Qiagen). The purity of the fusion protein was estimated to be 90% by sodium dodecyl sulfate-polyacrylamide gel electrophoresis followed by staining with Coomassie blue.

-Galactosidase activity assay. Yeast cells were inoculated from fresh overnight cultures into 10 ml of medium with either 2% dextrose or 2% galactose. For dextrose medium, cells were collected when the OD₆₀₀ reached between 1 and 2. Overexpression of Reb1p in the cells slows growth, with the doubling time estimated to be about 6 h. Cells were collected from galactose medium at 12 h after the inoculation (final OD₆₀₀ between 1 and 2). The cells were washed once with Z buffer (60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 10 mM MgSO₄, 50 mM -mercaptoethanol; adjusted to pH 7.0) and resuspended in 5 ml of Z buffer. Triplicate (1 ml each) samples from each strain were assayed. -Galactosidase activity was calculated as follows: OD₄₂₀/(OD₆₀₀ of assayed culture × volume assayed × time) (50).

RESULTS

Top Previous Next References

Three Reb1p binding sites are located upstream of the REB1 open reading frame (ORF). Examination of the upstream non-coding sequence of REB1 reveals three sites that can potentially bind Reb1p (Fig. 1): sites A and B, 110 and 80 bp upstream of the transcription initiation site (28), respectively, and site C, 30 bp downstream of it. To verify the authenticity of these sites, we generated several DNA fragments in which certain of the sites had been mutated (Fig. 2A). The mutations altered only the sequences of the sites without changing the distance between them (Fig. 1B). The binding of purified recombinant Reb1p to fragments containing the three individual sites is shown in the three left panels of Fig. 2B. It is clear that Reb1p binds to each of the putative sites, with relative affinities as follows: A  C  B. The dissociation constants (K_d) of the Reb1p binding sites were estimated to be about 25 and 70 nM for the A and C sites, respectively (7). The K_d value for the B site is roughly 10-fold higher.

View larger version (46K): [in this window] [in a new window]   FIG. 2.   A, B, and C are authentic Reb1p binding sites. (A) The indicated probes were generated by PCR with an oligonucleotide end labeled (*) with [-32P]ATP as one of the primers (JW202 [203 to 185] and JW243 [+68 to +87]). Thus, the specific activities of the probes should be identical. Plasmids containing mutations in the Reb1p binding sites (Fig. 1B) were used as templates in the PCR. (B) Each of the four probes was incubated with increasing amounts of purified His6-Reb1p (i.e., 0, 4.5, 9, 13.5, 18, and 22.5 nM as indicated) and subjected to EMSA. Positions of single-bound (A, B, and C) and double-bound (AC) complexes are indicated. Free probes are denoted by arrowheads. The dashed box exhibits a shorter exposure of lanes 19 to 24.

View larger version (48K): [in this window] [in a new window]   FIG. 3.   Binding of Reb1p to the A and B sites. (A) The probe was generated by PCR with an end-labeled (*) oligonucleotide (JW821 [162 to 143]; Fig. 1B) as one of the primers. (B) A fixed amount of probe was incubated with partially purified Reb1p (0, 0.6, 1.2, 1.8, 2.4, 3.0, 4.5, 6.0, and 7.5 µg in lanes 1 to 9, respectively) and subjected to EMSA. Positions of bound probes (single and double) are indicated. Free probe is denoted by an arrowhead. The faster-migrating complexes indicated by the brace are due to degradation of the partially purified Reb1p (42).

Reb1p binds preferentially to the A site when both the A and C sites are on the same molecule. Careful examination of Fig. 2 reveals that although all probes are the same size, the Reb1p-C complex migrates somewhat faster than either the A or the B complex. This is presumably due to bending of DNA at a site relatively near the end of the molecule (64). In any case, the difference in migration provides a means to distinguish binding at the A site from that at the C site. At low concentrations of Reb1p, a probe carrying both A and C sites shows only complexes corresponding to binding at the A site. At higher concentrations there are complexes due to binding at both sites (Fig. 2B, lanes 19 to 24). In no case is there a band corresponding to binding at the C site alone, as is clear in the brief exposure (above lanes 19 to 24). This is a surprising result since the two sites differ in K_d by only a factor of three. To determine whether these observations depend on the A and C sites being on the same molecule, we incubated Reb1p with two separate fragments, each containing one of the sites. As shown in Fig. 4, lanes 3 and 4, Reb1p binds to both the A and the C sites when they are on different molecules. Thus, the apparent loss of C site binding occurs only when the two sites are on the same molecule of DNA. In contrast, when the B and C sites are on the same molecule, the C site is filled at a normal rate (data not shown).

View larger version (66K): [in this window] [in a new window]   FIG. 4.   Binding of Reb1p to the A and C sites on separate probes. The end-labeled probes A and C (see Fig. 2A) were incubated with purified His6-Reb1p as indicated (lanes 2 and 5). For lane 3 (A+C), individual A and C probes were mixed and then incubated with His6-Reb1p. For lane 4 (A/C), the A and C probes were incubated separately with Reb1p and were then mixed just before electrophoresis. All samples were subjected to EMSA. Positions of bound complexes are indicated. Free probes are indicated by the arrowhead.

View larger version (32K): [in this window] [in a new window]   FIG. 5.   The off-rate of Reb1p bound at the A and the C sites. Purified His6-Reb1p was added to the master tubes containing either probe A or probe C (see Fig. 2A). After 20 min of incubation, specific competitor was added, and samples were then loaded onto a running gel at the indicated times (in minutes) (see Materials and Methods for details). Results for the A and C sites are shown in panels A and B, respectively. PhosphorImager quantitation is shown in panel C, where the lines represent the best fit to a single decay rate. The estimated half-lives for the A and C complexes are 43 and 12 min, respectively.

Transcriptional activity of Reb1p binding sites. To study the role of Reb1p in the transcription of its own gene, we constructed a chimeric reporter gene in which 1.2 kb of REB1, including 615 bp of the coding region, was fused upstream of the -galactosidase ORF (see Fig. 1A). Several constructs that included the mutations of the Reb1p binding sites used for the binding studies (see Fig. 1B) were integrated as a single copy at the URA3 locus as described in Materials and Methods. The wild-type REB1 sequence (ABC) supports the expression of the lacZ reporter gene (Table 3). Mutation of either the B site (AC) or the A site (BC) reduces expression partially. Mutation of both the A and the B sites (C) abolishes expression almost completely. Thus, these two sites are essential for REB1 transcription. The presence of either the A site or the B site leads to significant expression; they are approximately additive (Table 3, compare AC + BC with ABC). It is interesting that the B site alone, with less than one-tenth the affinity for Reb1p binding in vitro, has nearly half the transcription activity of the A site, suggesting that the intrinsic activity of a Reb1p molecule is greater when bound at that position. These results are reminiscent of the situation of the PDR3 gene, in which two Pdr3p binding sites not far upstream from the transcription initiation site are essential for transcription and provide additive effects (12).

View this table: [in this window] [in a new window]   TABLE 3.   -Galactosidase activity assay of REB1'-lacZ reporter genesa

Reb1p blocks transcription at the initiation step. We sought to determine whether Reb1p binding at the C site prevents initiation of transcription or acts as an attenuator of ongoing transcription. To do so, we moved the C site. Starting with a reporter gene that had a mutant C site (AB in Table 3), we generated a 15-bp sequence identical to the C site 105 bp downstream of the transcription initiation site (C₊₇₅ in Fig. 1B). EMSA was used to verify the binding of Reb1p to the C₊₇₅ site (data not shown). If the C site is an attenuator (30), it should block the elongation of transcription even at a downstream site. As shown in Table 3, moving the C site downstream had nearly the same effect on transcription as did mutation of the C site (compare ABC, AB, and ABC₊₇₅ in both the DEX and the GAL columns). This result suggests that the repressive effect of Reb1p occupying the C site depends on its proximity to the transcription initiation site. Thus, we conclude that Reb1p binding to the C site blocks transcription at a step prior to elongation, perhaps by preventing the proper alignment of a competent transcription initiation complex.

A Reb1p binding site can block transcription dependent on a heterologous UAS. Does the inhibitory effect of an occupied C site depend on the presence of the A and/or B site upstream? In short, is it specific for the REB1 gene? To answer this question, we introduced a C site 35 bp downstream of the transcription initiation site of the RPL32 gene, which encodes ribosomal protein L32 (11). RPL32 has no apparent Reb1p binding sites. In this reporter construct, the ORF of GFP was fused to the second exon of RPL32 (Fig. 6A).

View larger version (29K): [in this window] [in a new window]   FIG. 6.   Reb1p bound at the C site can block transcription of a heterologous promoter. (A) Diagram and partial sequence of the reporter genes. An 860-bp fragment of RPL32 of S. cerevisiae fused to GFP was used as the backbone for the reporter gene, pLC151 (pRS314-RPL32'-GFP). The transcription initiation site is indicated as 1 (11). The nucleotides designating the 5' (GT) and 3' (AG) sites of the RPL32 intron are underlined. By substitution of only seven nucleotides it was possible to generate a sequence identical to the original C site over 14 bp at positions +32 to +45 of the RPL32 transcription unit, as shown in the box. The resulting reporter gene was designated pLC151C. (B) Decreased transcription of RPL32'-GFP by Reb1p binding to the C site. Total RNA prepared from WY189 (with pLC151, lanes 1 and 2) and WY190 (with pLC151C, lanes 3 and 4) were subjected to Northern blot analysis and probed with an end-labeled oligonucleotide (JW969 [+31 to +10]; panel A) that will detect transcripts of both RPL32 and RPL32'-GFP (arrows). Both strains contained pLC30 (pGAL1-REB1, see Tables 1 and 2), in which overexpression of Reb1p can be induced by using galactose as the carbon source (lanes 2 and 4). The relative amounts of the two mRNAs, as determined by PhosphorImager analysis, are indicated. Three independent clones from each strain were used for this set of experiments. One representative example of each strain is shown.

The presence within a small region of DNA of three sites that bind Reb1p, all with different properties, provides an opportunity to consider the relationship among adjacent sites from a new perspective. On the one hand it is interesting that two sites within 30 bp of each other can bind with neither interference nor cooperativity. The 30 bp of extended DNA extends over about 100 Å, although we expect it to be far more tightly coiled in vivo. With 810 amino acids, Reb1p would fill a sphere with a diameter of ca. 60 Å, although it may, of course, be elongated. In vitro, however, we find that the presence of a filled A site has little effect on the binding of Reb1p to the considerably weaker B site (Fig. 3).

The second point relates to the relationship between the A and the C sites. Reb1p has similar affinities for these two sites when they are on different molecules, yet when they are on the same molecule, the C site is not filled in the presence of the A site under the nonequilibrium conditions of an EMSA (Fig. 2B). We interpret this result as implying that the C site can donate its Reb1p ligand to the A site. That is, Reb1p binds to the C site but over time, because the off-rate is greater at the C site, the Reb1p molecule migrates to the A site. This migration could represent either a true release followed by rebinding or a partial release followed by a sliding from the C to the A site. While this migration has been observed only in vitro, it seems likely that the environment in vivo, with its high concentration of proteins and nucleic acids, would partially mimic the "cage" effect of a polyacrylamide gel lattice. We consider below the biological implications of this arrangement.

As described above, the binding of Reb1p has been experimentally implicated in a variety of functions involving both RNA Pol I and II. The results presented here demonstrate two additional functions of Reb1p, first as an essential transcription factor for its own gene (occupying sites A and B) and also as an obstruction to transcription initiation (occupying site C). Considering first the contribution of the individual sites to the transcription of REB1, we will then attempt to integrate the role of all three sites. The A site has the highest affinity and a very low off-rate. Loss of the A site leads to loss of about 60% of transcription of the gene. Does Reb1p bound at the A site activate transcription by providing an activation domain? Indeed, the N-terminal region of Reb1p can serve as an activator (unpublished data) and contains a glutamine-rich region between residues 60 and 200. Alternatively, Reb1p binding at the A site could recruit another transcription factor that itself activates transcription. Finally, Reb1p binding at the A site could alter the chromatin structure of the region sufficiently to permit the binding of TATA binding protein and the associated factors necessary to initiate transcription, no true activator being necessary for such an infrequent transcription event (see below).

The B site remains puzzling. Although Reb1p binds to it very weakly in vitro, it provides a major element of transcription activation in vivo. At normal levels of Reb1p it has half the activity of the A site; at elevated levels of Reb1p it has activity equal to that of the A site. One possible explanation is that binding of Reb1p to the B site leads to an intrinsically higher activation of transcription. Alternatively, another protein may enhance binding of Reb1p at the B site in vivo, as the SWI-SNF complex appears to do for a weak Gal4p binding site (5).

Mutation of both the A and the B sites leads to almost complete inhibition of transcription of REB1. Thus, Reb1p is an essential factor for the transcription of its own gene. Yet these two sites are not sufficient to replace the UAS of the CYC1 gene (data not shown). Indeed, it should be noted that the location of the A and B sites is much closer to the site of initiation than most transcription activators, which usually bind 150 to 500 bp upstream. In the case of the REB1 gene, perhaps the presence of at least one Reb1p molecule prevents the chromatin structure from becoming so condensed that no activation can occur.

Although one could raise the hypothesis that Reb1p bound to the A or the B site is somehow necessary to overcome the repression of transcription due to the occupancy of the C site, the data in Table 3 show that such is not the case. The transcription from a construct missing all three sites is no greater than that from a construct missing just the A and B sites.

As a negative element just downstream of transcription initiation, the C site plays an unusual role in the transcription of REB1. While transcription can be inhibited by the binding of a factor to a site artificially placed downstream of the transcription initiation site (25, 39, 47, 55), REB1 is unusual in having such an arrangement in its natural environment. The binding of Reb1p to an element of the rRNA gene plays a role in the termination of Pol I transcription by leading to a transcriptional pause (33). Yet, that seems not to be the mechanism at work here because moving the C site downstream, away from the initiation site, abolishes its effect. Such a result suggests that occupancy of the C site prevents the polymerase complex from aligning at the proper site to initiate transcription. This conclusion is consistent with a recent report based on intermolecular cross-linking that the Pol II preinitiation complex extends to at least position +20 (29). The effect of a bound Reb1p molecule seems to be general, for transcription is also reduced when the Reb1p binding site is inserted at position +30 in a different gene (Fig. 6). This result demonstrates two points. Occupancy of the C site represses transcription independent of the presence of other Reb1p molecules acting to activate transcription. Furthermore, occupancy of the C site can inhibit transcription from a very active gene to the same degree as from a modestly active gene.

In considering the biological role of the positive and negative regulation of REB1 transcription by Reb1p, it is important to bear in mind the dynamic state of both REB1 and its product within the cell. We have estimated that a cell has 2,000 to 3,000 molecules of Reb1p (41), which is abundant for a transcription factor but only 1 to 2% of the level of ribosomal proteins, for example. Recent measurements of the abundance of mRNAs in S. cerevisiae (60) found only a single REB1 mRNA among 60,000 mRNAs tabulated (our analysis of the complete data set, kindly provided by V. E. Velculescu). This result suggests that on average there is less than one mRNA per cell. An alternative approach led to a value of about one REB1 mRNA per cell (37a, 63). Although one might initially think that these mRNA values seem somewhat low, they are consistent with the relative values obtained for ribosomal protein mRNAs. These values have two important implications.

We have measured the half-life of the REB1 mRNA by using the rpb1-1 mutant allele of RNA Pol II (44) and obtained a value of 15 to 20 min (data not shown), a reasonable value for an mRNA of S. cerevisiae (23). If there is only one mRNA per cell, then transcription of the REB1 gene occurs, on average, only once every 30 to 40 min. While we do not know the rate at which association and dissociation of transcription factors and DNA occur in the complex milieu of the nucleus, it is clear that this is within the time frame of the dissociation we observed in vitro. Thus, consideration of the influence of the association and dissociation of the Reb1p molecules at the REB1 gene should incorporate the fact that the gene is idle most of the time. The apparently anomalous activity of Reb1p bound to the B site could be a manifestation of such a nonequilibrium situation.

A second implication of the very small number of REB1 mRNAs per cell is that there will necessarily be substantial fluctuations in the number of mRNAs in an individual cell, leading to fluctuations, albeit damped, in the level of Reb1p itself.

Numerous instances of transcription factors controlling their own synthesis have been reported. The proto-oncogenes myb and jun positively regulate their own expression (2, 43); how that expression is extinguished is unclear. In contrast, fos and myc negatively regulate their own transcription, and loss of this control through mutation or genomic rearrangement leads to oncogenesis (16, 45). An interesting example in S. cerevisiae is Rap1p, which can partially repress the transcription of its own gene (20). The amplification implicit in positive autoregulation is used in Drosophila spp. to establish programmed cell types during development (reviewed in reference 56). Similarly, it is used to mount a vigorous response to stress, e.g., in the case of the AMT1 gene of Candida glabrata in responding to heavy metals (65) and perhaps in the case of the drug resistance gene PDR3 of S. cerevisiae (12).

On the other hand, examples of a transcription factor regulating its own gene in both a positive and negative way are rare. One that does is the cI gene, encoding the  repressor, which also uses three sites to maintain a constant, tightly controlled level of repressor. Cooperative binding of repressor to the two distal sites stimulates transcription of the cI gene. As the level of repressor rises, it binds to the third site, very close to transcription initiation, where it blocks the entry of polymerase (reviewed in reference 48). With the exception of the cooperative binding at the upstream sites, REB1 shares these elements.

Let us now consider the REB1 gene as a biological entity. Reb1p can bind to at least several hundred sites within the genome, participating in a number of important transcriptional events. We suggest that the regulatory elements of REB1 have evolved so as to maintain a stable production of Reb1p, since either overproduction or underproduction of Reb1p is deleterious for growth (unpublished data). At first glance it would seem that having the gene product as an essential component for its transcription would lead to an unstable situation, one that is autocatalytic in either direction. However, the presence of the C site changes that scenario entirely, as we see from Table 3, where overproduction of Reb1p leads to severe repression of REB1 transcription. Under normal growing conditions, it appears that the A site is more or less fully occupied, since in strain WY102 there is little difference in lacZ activity whether or not Reb1p is abundant. On the other hand, it appears that the C site is occupied far less, since the repression due to the C site is much greater when Reb1p is overproduced. We suggest that the arrangement of sites has evolved to buffer the fluctuation of Reb1p concentration that must occur due to the statistical nature of transcription, given the small average number of REB1 mRNA molecules per cell. As the level of Reb1p falls, the C site can serve as a recruiter for Reb1p, transferring it to the A site and even perhaps to the B site, where it can activate transcription. As the level of Reb1p rises, the C site becomes more fully occupied because the occupied A site is no longer available as a sink, and transcription is repressed. The more rapid on- and off-rates of the C site will render it more responsive to changing levels of Reb1p.