A novel DNA binding motif for yeast zinc cluster proteins: the Leu3p and Pdr3p transcriptional activators recognize everted repeats.

The Gal4, Put3, and Ppr1 yeast zinc cluster proteins bind as homodimers to DNA sequences composed of palindromic CGG triplets. Spacing between the triplets specifies the target site for a given zinc cluster protein. In addition, Hap1p, another zinc cluster protein, also recognizes CGG triplets but only when oriented as a direct repeat. Unexpectedly, our results show that Leu3p, another member of this family, also recognizes CGG triplets but oriented in opposite directions and spaced by 4 nucleotides (an everted repeat or inverted palindrome: CCG-N4-CGG). This constitutes a novel DNA motif for zinc cluster proteins. Moreover, the presence of this motif was shown to be essential for in vivo activation by Leu3p of a minimal reporter containing one copy of a target site for this activator. We also provide evidence that another member of this family, Pdr3p, binds to an everted repeat spaced by 0 nucleotides (CCGCGG). Thus, our results show that three CGG motifs are used by members of the zinc cluster family: palindromes, direct repeats, and everted repeats.

Repression of the alpha-fetoprotein gene promoter by progesterone and chimeric receptors in the presence of hormones and antihormones.

Using transient transfection assays, we showed that repression of the alpha-fetoprotein promoter by intact and deletion mutants of the progesterone receptor and by chimeric progesterone/glucocorticoid-estrogen receptors in the presence of their cognate hormones was closely correlated with their ability to bind to a progesterone/glucocorticoid-responsive element. This negative regulation was also observed in the presence of antihormones, providing evidence that receptor-antihormone complexes can bind to their responsive elements in vivo.

Fusions with histone H3 result in highly specific alteration of gene expression.

Hap1 is a yeast transcriptional activator which controls expression of genes such as CYC1 and CYC7. Our results show that Hap1 activity is dependent on a functional chromatin remodeling complex SWI/SNF. Using a modified two-hybrid screen with Hap1 as bait, we recovered expression vectors encoding the Gal4 activation domain fused to histone H3 [Gal4(AD)-H3]. Hap1 activity at CYC1 or CYC7 was increased by Gal4(AD)-H3 and the effect was dependent on the presence of the activation domain of Hap1 and a functional SWI complex. Importantly, overexpression of H3 alone had no effect on Hap1 activity. Analysis of Gal4(AD)-H3 revealed that the fusion is not incorporated into the nucleosome while a functional Gal4 activation domain is dispensable. Activity of many other transcriptional activators was unchanged or slightly affected in the presence of Gal4(AD)-H3. Thus, our results identify a new class of histone H3 variants that cause highly specific alteration of gene expression. Hap1 may interact directly with H3 favoring chromatin remodeling by the SWI/SNF complex.

Phenotypic analysis of genes encoding yeast zinc cluster proteins.

Zinc cluster proteins (or binuclear cluster proteins) possess zinc fingers of the Zn(II)2Cys6-type involved in DNA recognition as exemplified by the well-characterized protein Gal4p. These fungal proteins are transcriptional regulators of genes involved in a wide variety of cellular processes including metabolism of compounds such as amino acids and sugars, as well as control of meiosis, multi-drug resistance etc. The yeast (Saccharomyces cerevisiae) sequencing project has allowed the identification of additional zinc cluster proteins for a total of 54. However, the role of many of these putative zinc cluster proteins is unknown. We have performed phenotypic analysis of 33 genes encoding (putative) zinc cluster proteins. Only two members of the GAL4 family are essential genes. Our results show that deletion of eight different zinc cluster genes impairs growth on non-fermentable carbon sources. The same strains are also hypersensitive to the antifungal calcofluor white suggesting a role for these genes in cell wall integrity. In addition, one of these strains (YFL052W) is also heat sensitive on rich (but not minimal) plates. Thus, deletion of YFL052W results in sensitivity to a combination of low osmolarity and high temperature. In addition, six strains are hypersensitive to caffeine, an inhibitor of the MAP kinase pathway and phosphodiesterase of the cAMP pathway. In conclusion, our analysis assigns phenotypes to a number of genes and provides a basis to better understand the role of these transcriptional regulators.

Differential regulation of normal and tumor alpha 1-fetoprotein genes in fetal hepatocyte x hepatoma hybrids.

Fetal rat hepatocytes and mouse hepatoma cells actively expressing alpha 1-fetoprotein (AFP) and albumin genes were fused with the use of Sendai virus, and the expression of normal (rat) and tumor (mouse) AFP and albumin genes was analyzed in hybrid clones. The tumor AFP gene and both albumin genes were active in 103 hybrids. Expression of the normal fetal rat AFP gene, however, was maintained in only 3 hybrids, and it was frequently lost or decreased selectively upon subcloning. Furthermore, the normal AFP gene, when expressed, was more reactive than the tumor AFP gene to repression by a glucocorticosteroid hormone. These results suggest constitutive differences in the manner an oncofetal gene is activated and regulated in normal and neoplastic states. AFP gene expression in normal hepatocytes appears to be subordinated to a differentiation program degenerated and bypassed in hepatoma cells.

The acidic transcriptional activation domains of herpes virus VP16 and yeast HAP4 have different co-factor requirements.

The acidic transcriptional activation domain of the herpes virus activator VP16 requires an accessory protein complex for function, termed an adaptor. Although the activation domain of the yeast activator HAP4 is also highly negatively charged, its function is independent of at least one component of the adaptor complex, ADA2. In this study, we have used an in vitro inhibition assay to determine whether the activation domains of VP16 and HAP4 use a similar mechanism to potentiate transcription. Both domains had potent activation ability, indicating a similar strength of action. However, the capacity of each domain to inhibit activation of a heterologous test promoter (dA/dT) was sharply dissimilar. VP16 selectively inhibited activated transcription of dA/dT, without affecting basal transcription, implying that VP16 and the activator protein of the dA/dT promoter share a mechanism for activation. In contrast, HAP4 was totally unable to inhibit activated transcription of the dA/dT template. In the second part of the study, a genetic selection was used to obtain mutations in putative cofactor genes for HAP4. The spectrum of phenotypes caused by these mutations was strikingly different than mutations in the adaptor for the VP16 activation domain. These results strongly suggest that HAP4 and VP16 have distinct cofactor requirements, although they are both acidic activators.

Progestin receptors: isoforms and antihormone action.

We present evidence that the two isoforms of A and B of the chicken (cPR) and human progesterone receptor (hPR) originate from two different mRNA populations. One of these encodes the isoforms A which originate by initiation of translation at an in-frame AUG found 127 (cPR) and 165 (hPR) codons downstream of the AUG which gives rise to the isoforms B. Two estrogen-inducible hPR promoters were identified which are responsible for the generation of these two classes of transcripts. Characterization of the cPR promoter suggested the possible existence of cell-type and isoform-specific auto-regulation of cPR transcription and provided evidence that estrogen-induction of cPR expression occurs at a post-transcriptional level. Finally, we demonstrate promoter-specific transcriptional activation by the hPR isoforms A and B, and we discuss the mechanism of action of the anti-progestin RU486.

HAP1 positive control mutants specific for one of two binding sites.

The expression of the yeast CYC1 and CYC7 genes is controlled by the HAP1 activator. A GAL4-like zinc finger (residues 1-148) specifies binding to the dissimilar sites UAS1 (of CYC1) and CYC7, and an acidic domain (residues 1307-1483) is essential for activation of transcription. To analyze how HAP1 binds to UAS1 and CYC7, we performed saturation mutagenesis of the DNA-binding domain and recovered mutants with altered activity. Class 1 mutants had a reduced activity at both UAS1 and CYC7, and class 2 mutants selectively eliminated activity at CYC7. Surprisingly, several mutants of both classes exhibited wild-type DNA binding, indicating that they were specifically defective in activation. These positive control (PC) mutants alter residues that bracket the zinc finger. We explain these mutants in a model involving cofactor proteins that bind UAS1 and CYC7 along with HAP1. The existence of PC mutants that only affect activity at CYC7 raises the possibility that different cofactors may exist for UAS1 and CYC7.

Zinc cluster proteins Leu3p and Uga3p recognize highly related but distinct DNA targets.

Members of the family of fungal zinc cluster DNA-binding proteins possess 6 highly conserved cysteines that bind to two zinc atoms forming a structure (Zn2Cys6) that is required for recognition of specific DNA sequences. Many zinc cluster proteins have been shown to bind as homodimers to a pair of CGG triplets oriented either as direct (CGG NX CGG), inverted (CGG NX CCG), or everted repeats (CCG NX CGG), where N indicates nucleotides. Variation in the spacing between the CGG triplets also contributes to the diversity of sites recognized. For example, Leu3p binds to the everted sequence CCG N4 CGG with a strict requirement for a 4-base pair spacing. Here, we show that another member of the family, Uga3p, recognizes the same DNA motif as Leu3p. However, these transcription factors have distinct DNA targets. We demonstrate that additional specificity of binding is provided by nucleotides located between the two everted CGG triplets. Altering the 4 nucleotides between to the two everted CGG triplets switches the specificity from a Uga3p site to a Leu3p site in both in vitro and in vivo assays. Thus, our results identify a new mechanism that expands the repertoire of DNA targets of the family of zinc cluster proteins. These experiments provide a model for discrimination between targets of zinc cluster proteins.

Mutations in target DNA elements of yeast HAP1 modulate its transcriptional activity without affecting DNA binding.

The yeast zinc cluster protein HAP1, a member of the GAL4 family, is a transcriptional activator that binds as a homodimer to target DNA sequences. These targets include the upstream activating sequences of the CYC1 and CYC7 genes, which have no obvious sequence similarity. Even though both sites have the same affinity for HAP1, activation differs at these two sites, even when the sequences are placed in an identical promoter context. In addition, mutants of HAP1 that can bind to both sites but are specifically transcriptionally inactive at CYC7 have been previously isolated. In order to identify nucleotides that are responsible for this differential activity, we have performed random and site-directed mutagenesis of these target sites and assayed their binding to HAP1 in vitro and their activity in vivo in reporter plasmids. Our results show that HAP1 binding sites are degenerate forms of the direct repeat CGG N3 TA N CGG N3 TA. Moreover, we show that activity of HAP1 mutants defective for activation of the CYC7gene is restored by specific mutations in the CYC7 binding site. Conversely, other mutations of the target sites prevent activation by HAP1, without interfering with DNA binding. The results suggest that the sequence of the target sites influences the conformation and, hence, the activity of DNA-bound HAP1.

Decreased expression of specific genes in yeast cells lacking histone H1.

Chromatin plays an important role in regulating eukaryotic gene expression. Chromatin is composed of DNA wrapped around a nucleosome core (consisting of two copies of the well conserved histones H2A, H2B, H3, and H4) and a more variable linker histone H1. Various in vitro and in vivo studies have implicated histone H1 as a repressor of gene expression or as an activator, but its exact role is still unclear. Sequencing of the yeast genome has led to the identification of a putative histone H1 gene. Biochemical studies demonstrated that yeast does indeed possess a bona fide histone H1. However, deletion of the unique yeast H1 gene is not associated with any phenotypes, and it was questioned whether it plays any role. To address this issue, we performed whole-genome microarray analysis to identify genes that are affected by H1 removal. Surprisingly, deletion of the gene encoding histone H1 does not result in increased gene expression but rather in a modest reduction. Northern blot analysis of selected genes confirmed the results obtained with the microarray analysis. A similar effect was observed with an integrated lacZ reporter. Thus, our data demonstrate that removal of yeast histone H1 only results in decreased gene expression.

Antibody-promoted dimerization bypasses the regulation of DNA binding by the heme domain of the yeast transcriptional activator HAP1.

The yeast transcriptional activator HAP1 contains a DNA-binding domain homologous to the zinc finger of GAL4 and an adjacent regulatory domain that blocks DNA binding in the absence of the inducer heme. We show that short HAP1 fragments containing the zinc finger are unable to bind to DNA but can be rescued by antibody to the HAP1 zinc finger. These fragments are missing a coiled-coil sequence similar to that within the dimerization domain of GAL4 and dimerization domains of myosin heavy chain. We surmise that the antibody promotes DNA binding by bringing together two monomers. Interestingly, the antibody will also promote DNA binding of a larger HAP1 fragment containing the DNA-binding and the heme-regulatory domains. This suggests that the regulatory domain acts by preventing dimerization of HAP1 in the absence of heme. Consistent with this view is an in vivo assay that also reveals that heme promotes HAP1 dimerization in yeast cells.

Rat alpha 1-fetoprotein messenger RNA: 5'-end sequence and glucocorticoid-suppressed liver transcription in an improved nuclear run-off assay.

Cloned cDNA fragments spanning nearly the entire coding regions of rat AFP and albumin genes were used in liver nuclear run-off assays. Under standard assay conditions, transcription signals detected with 5' probes were systematically stronger than with 3' probes. Heparin eliminated this phenomenon, which suggests that nuclear run-off assays are subject to in vitro reinitiation occurring preferentially in promoter gene regions. Transcription in the presence of heparin indicates that very few polymerases are engaged on the AFP gene in adult rat liver. Dexamethasone treatment of developing rat liver results in the loss of transcribing polymerases from all regions of the AFP gene. Albumin gene transcription is unaffected. Inhibition of liver protein synthesis with cycloheximide does not modify the AFP gene suppressive action of dexamethasone. Glucocorticoid hormone receptors may thus directly interact with the AFP locus, blocking polymerase initiation. We also report the sequence analysis of rat AFP mRNA, which reveals the existence of two potential initiation codons on this molecule.

A linker region of the yeast zinc cluster protein leu3p specifies binding to everted repeat DNA.

Yeast zinc cluster proteins form a major class of yeast transcriptional regulators. They usually bind as homodimers to target DNA sequences, with each monomer recognizing a CGG triplet. Orientation and spacing between the CGG triplet specifies the recognition sequence for a given zinc cluster protein. For instance, Gal4p binds to inverted CGG triplets spaced by 11 base pairs whereas Ppr1p recognizes a similar motif but with a spacing of 6 base pairs. Hap1p, another member of this family, binds to a direct repeat consisting of two CGG triplets. Other members of this family, such as Leu3p, also recognize CGG triplets but when oriented in opposite directions, an everted repeat. This implies that the two zinc clusters of Leu3p bound to an everted repeat must be oriented in opposite directions to those of Gal4p or Ppr1p bound to inverted repeats. In order to map the domain responsible for proper orientation of the zinc clusters of Leu3p, we constructed chimeric proteins between Leu3p and Ppr1p and tested their binding to a Leu3p and a Ppr1p site. Our results show that the linker region, which bridges the zinc cluster to the dimerization domain, specifies binding of Leu3p to an everted repeat. We propose that the Leu3p linker projects the two zinc clusters of a Leu3p homodimer in opposite directions allowing binding to everted repeats.

Control of transcription of the chicken progesterone receptor gene. In vitro and in vivo studies.

To study the promoter of the chicken progesterone receptor (cPR) gene and the relevance of several progestin-responsive elements therein, chimeric genes were constructed which contained the 5'-flanking region of the cPR gene linked to promoterless globin or chloramphenicol acetyltransferase sequences. Cell-specific initiation of transcription was observed in transiently transfected chicken embryo fibroblasts when using 876 base pairs of the cPR gene upstream region. Transcription from these reporter genes could be induced by progestins in the presence of cPR form A but not of form B. In keeping with these data, three in vitro progesterone receptor (PR)-binding sites were identified in the cPR promoter region by DNase I protection assays. However, in vivo, nuclear run-on transcription demonstrated that neither primary stimulation with progestins, nor treatment of secondarily estrogen-stimulated chicks with progestins, glucocorticoids, or androgens resulted in any significant change of cPR gene transcription in the oviduct, thus suggesting a cell- and/or development-specific role for these progestin-responsive elements. Although estrogen is known to increase PR levels in the chick oviduct, this effect does not involve stimulation of PR gene transcription, as demonstrated here by nuclear run-on experiments, the analysis of DNase I hypersensitive sites, and transient cotransfection studies. Since acute withdrawal from estrogen-stimulation markedly decreased the level of cPR mRNAs in chick oviduct when analyzed by Northern blotting, we conclude that estrogen-dependent stimulation of PR levels in the oviduct is a post-transcriptional process.

DNase I hypersensitivity and methylation of the 5'-flanking region of the alpha 1-fetoprotein gene during developmental and glucocorticoid-induced repression of its activity in rat liver.

Three major regions of DNase I hypersensitivity (DH) were found in alpha 1-fetoprotein (AFP) chromatin of rat liver. DH site I is located at the transcription initiation site and associated with ongoing AFP transcription. DH site II is located 2.5 kb upstream from the cap site: it is developmental stage-dependent but dissociable from ongoing AFP transcription. DH site III, 3.7 kb upstream from the cap site, behaves as hepatocyte-constitutive. DH sites are present in similar regions of liver albumin chromatin. Dexamethasone-induced AFP gene repression is accompanied by the selective loss of AFP DH site I, a likely result of glucocorticoid receptors binding to a DNA recognition sequence located 5'-adjacent to DH site I. Sl nuclease-hypersensitive sites were found on naked superhelical AFP and albumin DNA, but do not appear to contribute DH sites in liver chromatin. The extent of hypomethylation of HpaII sites at the 5'-end of the AFP gene correlates positively with the level of potential and actual expression of the gene. We conclude that developmental and hormonal regulation of the AFP gene is confined within congruent to 4 kb of 5'-flanking DNA, and we discuss possible hierarchical interactions among DH sites, in relation to DNA methylation and replication.

The N-terminal region of the chicken progesterone receptor specifies target gene activation.

Steroid hormone receptors belong to a family of nuclear receptors that trigger transcriptional activation of target genes by specific binding to DNA recognition sequences, usually located in the 5'-flanking region of the target gene. Nuclear receptors appear to be segmented proteins and extensive structure-function analyses have attempted to elucidate the functional significance of individual segments. Two of these regions have been defined as the domains responsible for recognition of responsive elements of target genes (region C) and hormone binding (region E) (refs 2-7). But the functional significance of the N-terminal region (A/B), which diverges extensively even for a given receptor between different species, has remained obscure. We have previously cloned, expressed and analysed the chicken progesterone receptor (cPR) (ref. 8). This receptor and its human homologue from T47D breast cancer cells are unique among the steroid hormone receptors in that two forms, A and B, are present in equal amounts in cytosolic extracts, the latter having the higher molecular weight. For the chicken progesterone receptor, we have presented evidence suggesting that the cPR form A corresponds to an N-terminally truncated form of B (ref. 8). Here we report on the functional difference between the forms A and B in the transcriptional activation of two target genes.

Steroid hormone receptors compete for factors that mediate their enhancer function.

Stimulation of transcription of reporter genes by the progesterone receptor (PR) was inhibited in transfected HeLa cells by co-expressing the estrogen receptor (ER) in an ER-dose- and estrogen-dependent manner. Both the N-terminal A/B region and the hormone binding domain of ER were involved in this inhibition, which was antagonized by antiestrogens and did not appear to involve direct interaction between ER and either reporter gene or PR. ER expression also inhibited activation by the glucocorticoid receptor (GR), and both PR and GR expression inhibited activation by ER, albeit to a lower extent. Similar transcriptional interference was observed between the endogenous PR and ER present in T47D and MCF-7 breast cancer cells transfected with an ER reporter gene. Moreover, transcription of the resident estrogen-induced pS2 gene was partially inhibited by exposing MCF-7 cells to progestins or glucocorticoids. We propose that these observations reflect competition for a functionally limiting transcription factor(s).

Two distinct estrogen-regulated promoters generate transcripts encoding the two functionally different human progesterone receptor forms A and B.

The human progesterone receptor (hPR) cDNA, synthesized from T47D breast cancer cells, and the hPR gene 5'-flanking region were cloned and sequenced. Comparison of the cDNA-deduced amino acid sequence with other PR homologues demonstrated the modular structure characteristic of nuclear receptors. As in the case of the chicken homologue, there are two hPR forms, A and B, which originate from translational initiation at AUG2 (codon 165) and AUG1, respectively. Northern blot analysis of T47D mRNA using various cDNA derived probes identified two classes of hPR mRNAs, one of which could code for hPR form B, while the other one lacked the 5' region upstream of AUG1. S1 nuclease mapping and primer extension analyses confirmed that the second class of hPR transcripts are initiated between +737 and +842 and thus encode hPR form A, but not form B. By using the hPR gene 5'-flanking sequences as promoter region in chimeric genes, we show that a functional promoter (located between -711 and +31) directs initiation of hPR mRNAs from the authentic start sites located at +1 and +15. Most importantly, initiation of transcription from chimeric genes demonstrated the existence of a second promoter located between +464 and +1105. Transient co-transfection experiments with vectors expressing the human estrogen receptor showed that both promoters were estrogen inducible, although no classical estrogen responsive element was detected in the corresponding sequences. When transiently expressed, the two hPR forms similarly activated transcription from reporter genes containing a single palindromic progestin responsive element (PRE), while form B was more efficient at activating the PRE of the mouse mammary tumor virus long terminal repeat. Transcription from the ovalbumin promoter, however, was induced by hPR form A, but not by form B.

Transient expression of human and chicken progesterone receptors does not support alternative translational initiation from a single mRNA as the mechanism generating two receptor isoforms.

Two isoforms (A and B) of the human (hPR) and chicken (cPR) progesterone receptors originate from a single PR gene. cPR form A results from initiation of translation at a downstream ATG codon (ATG2) which in the cPR cDNA-deduced open reading frame is found 128 amino acids C-terminal to and in-frame with the first ATG codon (ATG1) that gives rise to form B. Our recent observation of an abundant cPR mRNA which encodes only form A suggested to us that the two isoforms are translated from different transcripts (Jeltsch, J. M., Turcotte, B., Garnier, J. M., Lerouge, T., Krozowski, Z., Gronemeyer, H., and Chambon, P. (1990) J. Biol. Chem. 265, 3961-3974). This view is, however, at variance with data obtained by transient transfection with expression vectors containing most of the cPR cDNA (downstream of nucleotide +53) since both isoforms were generated in transiently transfected COS cells (Conneely, O. M., Kettelberger, D. M., Tsai, J. J., Schrader, W. T., and O'Malley, B. W. (1989) J. Biol. Chem. 264, 14062-14064). To further support our above conclusion, vectors containing either hPR or cPR cDNAs were introduced into HeLa and COS-1 cells. Only hPR form B originated from a vector containing the entire cDNA (containing nucleotides 1 to approximately 4400), and form A was produced only from a vector expressing hPR transcripts (nucleotides 814 to approximately 4400) lacking ATG1. Vectors expressing the 5'-untranslated and coding region of the cPR mRNA (nucleotides 29-2921) generated only traces of form A in the two cell lines. Similar traces of form A were observed in COS-1 cells transfected with a vector lacking the 5'-untranslated region. Collectively, these results do not support the hypothesis that similar amounts of the two PR isoforms are generated by alternative initiation of translation on a single PR transcript. We discuss data indicating that for hPR and cPR, isoforms A and B are in fact translated from different mRNAs.