Article

• The EMBO Journal (1999) 18, 1621 - 1629
• doi:10.1093/emboj/18.6.1621

A trans-acting peptide activates the yeast a1 repressor by raising its DNA-binding affinity

Martha R. Stark¹, Dominik Escher² and Alexander D. Johnson^1,3

1. Department of Biochemistry and Biophysics, San Francisco, CA 94143-0414, USA
2. Institute for Molecular Biology, University of Zurich, CH-8057 Zurich, Switzerland
3. Department of Microbiology and Immunology, School of Medicine, University of California, San Francisco, CA 94143-0414, USA

Correspondence to:

Alexander D. Johnson, E-mail: ajohnson@socrates.ucsf.edu

Received 2 November 1998; Accepted 27 January 1999; Revised 27 January 1999

• Keywords:

  • homeodomain,
  • protein–DNA interactions,
  • protein–protein interactions,
  • VP16,
  • yeast cell-type determination

Introduction

In eukaryotes, proteins that regulate transcription typically act in specific combinations. A simple example of this principle is found in the specification of cell-types in the yeast Saccharomyces cerevisiae. In the a/ cell type, two homeodomain proteins, a1 and 2, form a heterodimer that binds with high affinity and specificity to a DNA sequence called the haploid-specific gene (hsg) operator (Goutte and Johnson, 1988, 1993, 1994; Dranginis, 1990). This operator is located upstream of many genes (collectively called the haploid-specific genes), and the binding of a1 and 2 to it recruits the SSN6–TUP1 complex, which represses transcription of each haploid-specific gene (Mukai et al., 1991; Keleher et al., 1992; Smith and Johnson, 1992; Komachi et al., 1994). These genes encode proteins required for the a and cells to mate, and for regulators of a/ cell-specific functions (for reviews see Herskowitz et al., 1992; Johnson, 1995). Because a1 and 2 are present together only in the a/ cell type, the haploid-specific genes are derepressed in the other two yeast cell types, a and cells.

The a1–2 heterodimer has been well studied genetically, biochemically and structurally (e.g. Goutte and Johnson, 1993; Phillips et al., 1994; Li et al., 1995; Vershon et al., 1995). Some of these studies have utilized the minimal fragments of a1 and 2 necessary to heterodimerize and to bind tightly to DNA. The 2 minimal fragment includes the homeodomain (60 amino acid residues which fold into three helices linked by two turns) and a short peptide tail (21 amino acid residues) which extends from the C-terminus of the homeodomain. The a1 minimal fragment includes only the 60-amino acid homeodomain. NMR studies have shown that the 2 tail is unstructured in the 2 monomer but folds into a short distorted -helix upon contact with the a1 homeodomain (Phillips et al., 1991, 1994). An X-ray crystallographic study of the a1 and 2 minimal fragment heterodimer bound to DNA (Li et al., 1995) has revealed that this helix rests on top of the a1 homeodomain (that is, on the surface opposite that which contacts DNA), making a series of hydrophobic contacts. In the crystal structure both homeodomains contact the DNA which is strongly bent. The only contact between the a1 and 2 minimal fragments occurs through the 2 tail (Stark and Johnson, 1994; Li et al., 1995), although in the intact protein additional protein–protein contacts are made (Goutte and Johnson, 1993; Ho et al., 1994). Many additional genetic and biochemical experiments support the biological relevance of the crystal structure (reviewed by Andrews and Donoviel, 1995).

Despite these studies, it has been difficult to account for the high DNA-binding specificity of the heterodimer in terms of the individual DNA-binding properties of its two constituents. Under experimental conditions in which the specificity of the heterodimer for the hsg operator over non-specific DNA was at least 3000-fold, the 2 monomer exhibited a DNA-binding specificity of 10-fold and the a1 monomer showed no reproducible sequence-specific DNA binding (Goutte and Johnson, 1993; Phillips et al., 1994).

In this paper, we address how the high DNA-binding specificity of the a1–2 heterodimer is generated. In particular, we distinguish between two hypotheses: (i) heterodimerization is simply a coupling of the two monomers; and (ii) heterodimerization involves the instruction of one monomer by the other. Since no significant structural changes occur in the 2 homeodomain upon heterodimerization (Wolberger et al., 1991; Phillips et al., 1994; Li et al., 1995), the second hypothesis can be reduced to the more specific proposal that, upon heterodimerization, the 2 tail instructs the a1 homeodomain to bind specifically to DNA. Previous work (Mak and Johnson, 1993; Stark and Johnson, 1994) established the importance of the 2 tail for heterodimer formation, but did not distinguish between these two models.

To test this idea, we constructed a series of chimeric molecules consisting of the homeodomain of a1 linked covalently to the tail of 2. These two elements were joined by linkers designed to be of sufficient length and flexibility to permit an intramolecular interaction between the a1 homeodomain and the 2 tail. A prediction of the instructional hypothesis (but not of the simple coupling model) is that such chimeric a1 molecules should be able to bind tightly and specifically to DNA as monomers; we show that this is indeed the case. In a second set of experiments, we show that an 2 tail peptide supplied in solution can induce the a1 homeodomain to bind to DNA, a result that provides additional support for the instructional model. Finally, we provide evidence that features of the interaction between a1 and 2 also apply to other combinations of gene regulatory proteins.

Construction of the a1::2 chimeric proteins

The design of the chimeric a1 homeodomains (in particular the length of the linkers) was based on inspection of the X-ray crystal structure of the a1–2 heterodimer bound to DNA (Li et al., 1995). They are shown schematically in Figure 1A. For two of the chimeric molecules, the 2 tail was attached via a linker to the C-terminus of the a1 homeodomain. In the X-ray structure the distance between the C-terminus of a1 and the N-terminus of the 2 tail is 32 Å. A linker of 11 amino acids (present in the a1::11::2 chimera) should, in principle, span this distance if it is assumed that the linker is fully extended. A chimera with a linker of 16 amino acids (a1::16::2) was also constructed to accommodate some degree of folding in the linker. The linkers were composed of glycine and serine to provide both flexibility and solubility. A third chimeric molecule was constructed in which the 2 tail was attached via a linker to the N-terminus of the a1 homeodomain. The distance between the C-terminus of 2 and the N-terminus of the a1 homeodomain in the heterodimer crystal structure is only 13 Å, and a glycine/serine linker of 6 amino acids was used to span this distance.

Figure 1.

Proteins (A), peptides (B) and operators (C) used in the experiments described in this paper. (A) The top two diagrams depict the minimal fragments of wild-type a1 and 2 proteins sufficient for heterodimer formation and tight DNA binding. The additional proteins are chimeras in which the tail of 2 has been covalently linked to the homeodomain of a1. The linkers between the tail and the homeo- domain are composed of alternating (glycine)₂ and (serine)₂, and are depicted as a string of filled black circles. (B) 2 tail peptides were synthesized by California Peptide Research, Inc. All three peptides are identical except for position 196; they are 19 amino acids in length, beginning at residue 189 of 2 and ending at residue 207. The intact 2 protein ends with residue 210 and residues 208–210 are not required for a1–2 heterodimer formation (Mak and Johnson, 1993). (C) The a1–a1 operator has the same spacing and binding site orientation as a naturally occurring hsg operator (which contains an a1 and an 2 binding site), except the 2 binding site of the hsg operator has been replaced by a second a1 binding site (Goutte and Johnson, 1994). The two a1 binding sites are separated by six base pairs and for the experiments of Figures 2, 5 and 6 are contained within an 80-bp DNA fragment.

View full figure (54 KB)

a1::2 chimeric proteins bind DNA as monomers

The four chimeric proteins depicted in Figure 1A were expressed in Escherichia coli, purified to >90% homogeneity, and tested for their binding to a synthetic operator composed of two a1 half-sites (a1–a1 in Figure 1C). In contrast to the a1 homeodomain, which failed to bind DNA (Figure 2, lanes 5–7), all three chimeras exhibited efficient DNA binding in the 30–100 nM range (Figure 2, lanes 11–13, 17–19, 23–25). The chimeras a1::11::2 (lanes 11–13) and a1::16::2 (lanes 17–19) each formed two distinct protein–DNA complexes, whereas 2::6::a1 (lanes 23–25) produced only a single species. Based on a comparison with previous results (Smith and Johnson, 1992; Goutte and Johnson, 1993) and with the migration of the a1–2 fragment heterodimer bound to DNA (Figure 2, lanes 2–4), we conclude that the 2::6::a1–DNA species and the faster migrating of the two a1::11::2–DNA and a1::16::2–DNA species are monomers of the a1 chimera bound specifically to DNA. Consistent with this assignment, DNase I footprinting of the chimeric proteins on an hsg operator (which contains an a1 half-site and an 2 half-site; see Goutte and Johnson, 1994) demonstrated that all three of the chimeric proteins bind only to the a1 half of the operator (Figure 3). In contrast, the a1–2 fragment heterodimer binds to both the a1 and 2 sites of the operator (Figure 3).

Figure 2.

Binding of the a1::2 chimeras to the a1–a1 operator. The ³²P-labeled DNA fragment (80 nucleotide pairs) containing the a1–a1 operator was incubated with the purified protein indicated for 30 min at room temperature and electrophoresed through a 5% native Tris-borate–EDTA polyacrylamide gel (as described in Stark and Johnson, 1994). Lane 1 contains labeled DNA alone. Lanes 2–4 contain 3 nM 2 hd + tail (2_128-210) in addition to a1 hd. The a1 hd alone (lanes 5–7) and the a1::2 chimera homodimer (lanes 8–10) were included along with the a1–2 fragment heterodimer in order to demonstrate the different mobility shifts expected for monomers and dimers bound to the DNA. a1::11::2 (lanes 11–13) and a1::16::2 (lanes 17–19) both give two shifts, consistent with monomeric and dimeric DNA binding, whereas 2::6::a1 (lanes 23–25) gives only one shift consistent with monomeric DNA binding. All three chimeras containing the 2 tail mutation have reduced DNA binding (lanes: 14–16, a1::11::2_L196S; 20–22, a1::16::2_L196S; 26–28, 2_L196S::6::a1). The a1 homeodomain and chimera concentrations for each set of three reactions are 30, 100 and 300 nM.

View full figure (44 KB)

Figure 3.

DNA binding of a1::2 chimeras to an a1–2 operator. DNase I protection of an 80 nucleotide pair ³²P-labeled fragment containing the hsg operator. An operator with only one a1 binding site was chosen for this experiment so that monomeric binding of proteins would be revealed. Lane 1, 3M a1 hd plus 100 nM 2_128-210; lane 2, 50 M a1::2; lane 3, no protein; lane 4, 50 M 2::6::a1; lane 5, 100 M a1::16::2; lane 6, 50 M a1::11::2. The a1–2 fragment heterodimer and the a1::2 chimeric protein described previously as binding as a dimer (see text) were included as controls, to show protection of both half-sites in the operator. In contrast, 2::6::a1, a1::16::2 and a1::11::2 show protection only over the a1 half of the operator.

View full figure (40 KB)

In addition to complexes of a monomer bound to DNA, a1::11::2 and a1::16::2 each form a more slowly migrating species, which is very likely to be two monomers bound to DNA. As pointed out above, the binding of 2::6::a1 appears solely monomeric. One explanation for this difference is that the placement of the linker on the N-terminus of a1 in 2::6::a1 sterically blocks the binding of a second molecule to the second a1 half-site, while the C-terminal placements permit the binding of a second monomer. Another possibility is that the dimeric species observed in this experiment arise, at least in part, from favorable contact between two monomers, and that this contact is suboptimal in a2::6::a1. In any case, the important point is that all three a1::2 chimeras are capable of binding as monomers to specific DNA sites, whereas the a1 homeodomain itself shows no DNA binding under similar conditions.

We showed previously that a different a1::2 chimera, one that contains the tail of 2 linked covalently to the C-terminal end of the a1 homeodomain but that lacks a linker, binds DNA only as a dimer (Stark and Johnson, 1994). For comparison, the behavior of this protein is shown in Figure 2 (lanes 8–10). According to the a1–2–DNA crystal structure, this chimera (due to the absence of a linker) should not be capable of undergoing an intramolecular interaction to bring the 2 tail in contact with the proper surface of the a1 homeodomain. Thus, the only way for this chimera to efficiently bind the operator is through the interaction of the tail of one molecule with the homeodomain of a second molecule. We believe this is the explanation for the dimer requirement of this chimera. In contrast, the chimeras described in this paper are all capable of binding DNA as monomers. The predicted structures of the chimeras, based on the crystal structure of the a1–2 fragment heterodimer, are shown in Figure 4A, B and C.

Figure 4.

Cartoons depicting interaction of the 2 tail with the a1 homeodomain. a1 is shown in red and 2 is shown in blue. The green ovals refer to the glycine/serine linkers (see Figure 1). (A) The a1–2 heterodimer fragment crystal structure (Li et al., 1995). (B) Inferred structure of the two chimeras (a1::11::2 and a1::16::2) in which the 2 tail and linker were fused to the C-terminus of a1. (C) Inferred structure of the 2::6::a1 chimera, in which the 2 tail and linker were fused to the N-terminus of the a1 homeodomain. (D) Inferred structure of the a1 homeodomain bound by wild-type 2 tail peptide. For each of these cases, we have provided evidence that the interaction of the 2 tail with the homeodomain of a1 induces a conformational change, which stimulates the binding of the a1 homeodomain to hsg operator DNA. The contacts between a1 and DNA occur through helix 3 and the loop between helices 1 and 2 (Li et al., 1995). As described in the text, it has been proposed that the conformational change in a1 occurs in this loop.

View full figure (30 KB)

Specific point mutations in the tail reduce DNA binding by the a1::2 chimeras

To rule out the possibility that the enhanced DNA-binding of the chimeric a1 proteins was due to non-specific contributions of the linker or of the tail, a set of additional chimeras was constructed, each of which contains a specific point mutation in the tail. Leu196 in the 2 tail is critical for a1–2 function in vivo and in vitro (Strathern et al., 1988; Stark and Johnson, 1994). In the X-ray crystal structure of the a1–2 heterodimer, this leucine makes an important contact with the a1 homeodomain (Li et al., 1995). In all three chimeras, this leucine was changed to serine, and in all three cases (Figure 2, lanes 14–16, a1::11::2_L196S; lanes 20–22, a1::16::2_L196S; and lanes 26–28, 2_Ll96S::6::a1) DNA binding was significantly reduced compared with chimeras that carried the wild-type tail. This result demonstrates that efficient DNA binding by the a1 homeodomain chimeras specifically requires a functional 2 tail.

The 2::6::a1 chimeric protein shows sequence-specific DNA binding

The DNase protection experiments of Figure 3 showed that the three a1 chimeras specifically recognized the a1 half-site. We verified the sequence-specific DNA-binding of one chimera, 2::6::a1, using an additional technique. Using a gel mobility assay, we directly compared its affinity for a DNA fragment that contained the synthetic a1–a1 binding site with its affinity for the same fragment lacking these sites (Figure 5). Both 2::6::a1 (lanes 9–11) and the a1–2 fragment heterodimer, used as a control (lanes 3–5), bind specifically to the a1–a1 operator fragment, but not to the operator that lacks a1 sites (lanes 6–8 and 12–14). These results verify the footprinting results, showing that the 2::6::a1 chimera has a marked preference for a known a1 binding site over other DNA.

Figure 5.

DNA-binding specificity of the 2::6::a1 chimera. For this electrophoretic mobility shift experiment, an 80 nucleotide pair ³²P-labeled fragment containing the a1–a1 operator and the same labeled fragment with the operator deleted (consisting of 51 nucleotide pairs) were utilized. The two operators have different mobilities in the gel due to their size difference. Lanes 1, 3–5 and 9–11 contain the a1–a1 fragment while 2, 6–8 and 12–14 contain the fragment that lacks specific a1-binding sites. Lanes 1 and 2 lack protein and the other lanes have the indicated protein or proteins added. The binding conditions and protein concentrations were the same as those used in Figure 2.

View full figure (35 KB)

An 2 tail peptide is sufficient to induce the a1 homeodomain to bind DNA

It seemed plausible, based on the results described above, that the isolated tail of 2 could induce the a1 homeodomain to bind specifically to DNA even if the two polypeptides were not covalently joined. To test this idea, a wild-type and two mutant 2 tail peptides were synthesized. All three peptides were 19 amino acids in length and end at residue 207 of 2. The final 3 amino acid residues in the 2 tail (positions 208–210) are not required for a1–2 function (Mak and Johnson, 1993). One mutant contains a serine at position 196 and the other an alanine, changes known to disrupt the interaction of the tail with a1 (Strathern et al., 1988; Li et al., 1995). The results illustrated in Figure 6 show that the amount of DNA bound by the homeodomain increases as the concentration of wild-type peptide is raised (lanes 2–6), with half-maximal stimulation reached at a peptide concentration of 0.3 mM. This value is in excellent agreement with the K_D of 0.2–0.3 mM seen for the interaction of the 2 and a1 fragments as measured by NMR spectroscopy (Baxter et al., 1994; Phillips et al., 1994). The two mutant peptides also stimulate a1 binding (Figure 6, lanes 7–11 and 12–16), but to a significantly lesser extent. In other experiments, the difference between the wild-type and mutant peptides was less pronounced, suggesting that the a1–peptide interaction is highly sensitive to the conditions employed. The predicted structure of the a1 homeodomain–peptide complex, based on the crystal structure of the a1–2 heterodimer fragment, is shown in Figure 4D.

Figure 6.

A 19 amino acid 2 tail peptide, supplied in trans, induces a1 to bind the a1–a1 operator. Each lane contains 30 nM a1 hd, which does not bind DNA on its own (lane 1). In addition to a1 hd, lanes 2–6 contain successive 2-fold increases of the wild-type 2 tail peptide beginning with a concentration of 0.15 mM in lane 2 and ending with 2.5 mM in lane 6. Both mutant peptides, present in the same concentrations as the wild-type peptide, are reduced in their ability to induce a1 DNA binding (L196S, lanes 7–11; L196A, lanes 12–16). The binding conditions are the same as those of Figure 2, except that the incubations of DNA and protein were carried out at 4°C, as was the electrophoresis.

View full figure (32 KB)

Can heterologous 'tails' also mediate cooperative binding of a1 and 2 to DNA?

We have presented evidence that the 2 tail, in addition to joining a1 and 2 together to form the heterodimer, acts as a ligand to convert the a1 homeodomain to a high-affinity DNA-binding form. In this section, we test the generality of the tail–homeodomain interaction by testing whether a heterologous tail can substitute for the 2 tail in bringing about the cooperative binding of a1 and 2. We tested two heterologous tails for this activity. The first was from the 2 protein of Kluyveromyces lactis, a yeast closely related to S.cerevisiae. As judged by sequence comparison, the K.lactis 2 tail is slightly shorter than that of S.cerevisiae (Figure 7), but the last several residues of the S.cerevisiae 2 tail are dispensable for its function (see above). The four hydrophobic residues that form the basis of the interaction between the S.cerevisiae 2 tail and the a1 homeodomain are also hydrophobic in the K.lactis 2 tail, but only one of the 12 tail positions (Leu200 in 2) contains the same amino acid in both proteins.

Figure 7.

Sequence alignment of the 'tails' experimentally attached to the 2 homeodomain. Amino acid sequence of the S.cerevisiae (S.c.) and K.lactis (K.l.) 2 C-terminal tails correspond to the C-terminal 21 and 12 residues, respectively, of the two proteins. The VP16 sequence numbers correspond to residues 371–389 in the full-length VP16 protein (1–490). The residues of the S.cerevisiae 2 tail shown in bold are those that interact with the surface of the a1 hd. The corresponding residues in the other two tails are also shown in bold for comparison.

View full figure (31 KB)

The second tail we tested derives from the herpes simplex virus transcriptional regulator, VP16. Although not itself a homeodomain protein, VP16 interacts with the mammalian Oct-1 homeodomain to maximally activate transcription of some of its target genes. Baxter et al. (1994) and Li et al. (1995) have pointed out that the region of VP16 shown to interact with Oct-1 (Werstuck and Capone, 1989a, b; Stern and Herr, 1991; Walker et al., 1994; Shaw et al., 1995; Lai and Herr, 1997) bears rough amino acid similarity to that of the 2 tail (Figure 7). Moreover, the results of mutagenesis experiments (Lai et al., 1992; Pomerantz et al., 1992) have suggested that the surface of Oct-1 contacted by VP16 is approximately equivalent to the surface of a1 contacted by 2.

Analysis of the K.lactis 2 tail

We substituted the C-terminal tail of the 2 protein from S.cerevisiae with that of the K.lactis 2 to give an 2 hd::K.l. tail chimera. We expressed this chimeric protein in E.coli and showed that it was capable of binding DNA cooperatively with the S.cerevisiae a1 homeodomain protein (Figure 8A). In this DNA-binding experiment, the a1 homeodomain on its own shows no detectable DNA binding, but in the presence of the 2 hd::K.1. tail chimeric protein, significant DNA binding is observed. The migration of the complex is consistent with a heterodimer bound to DNA. Approximately 10-fold more 2 hd::K.1. tail protein than S.cerevisiae 2 homeodomain protein is required to bind an equivalent amount of DNA in the presence of the same concentration of a1 (data not shown), indicating that the interaction between the K.lactis 2 tail and the S.cerevisiae a1 homeodomain is less favorable than that between the two S.cerevisiae proteins.

Figure 8.

The 2 chimeric proteins containing heterologous tails bind to an a1–2 operator cooperatively with the a1 homeodomain protein. A ³²P-labeled DNA fragment containing an a1–2 operator was incubated with the indicated proteins for 45 min on ice before being electrophoresed through a 5% native Tris-borate–EDTA polyacryl- amide gel. The a1 homeodomain protein was purified, and the 2 chimeric proteins were all present in bacterial extracts. (A) Lane 1 contains no protein. All lanes marked with the a1 hd contain 3-fold increases of protein, beginning with 30 nM and ending with 1 M. The 2 hd::K.l. tail and the 2 hd::K.l. tail_I218S proteins are present in the lanes indicated at a concentration of 2 nM. Note that the extracts containing the wild-type and mutant 2 hd::K.l. tail proteins both show weak binding to the DNA in the absence of a1 hd protein. (B) The lanes marked with the a1 hd protein contain 3-fold increases of protein, beginning with 30 nM and ending with 300 nM. The 2 hd::VP16 and 2 hd::VP16_I377S proteins are present in the lanes indicated at a concentration of 80 nM. The extracts containing the wild-type and mutant 2 hd::VP16 proteins both show weak binding to the DNA in the absence of the a1 hd protein.

View full figure (58 KB)

To determine whether the K.lactis 2 tail was interacting with the a1 homeodomain in a manner analogous to that of the S.cerevisiae 2 tail, we changed one of the hydrophobic residues in the K.lactis tail, isoleucine 218, to serine. In S.cerevisiae, the equivalent mutation (L196S) disrupts the interaction between a1 and 2 (see above). The chimera 2 hd::K.l. tail_I218S fails to bind cooperatively with the a1 homeodomain (Figure 8A). From these results, we conclude that the C-terminal tail of K.lactis 2 can at least partially substitute for the S.cerevisiae 2 tail. Thus, the interaction with the a1 homeodomain is maintained even though the two tails are identical at only a single position.

Analysis of the VP16 'tail'

To test further the generality of the a1 homeodomain–2 tail interaction model, we grafted a region of the herpes virus activator protein, VP16, onto the 2 homeodomain, in place of 2's own tail (2 hd::VP16). This region of VP16 is predicted to form an amphipathic helix and to interact with the exposed surface of the Oct-1 homeodomain protein (Hayes and O'Hare, 1993; Lai and Herr, 1997). In the VP16 tail, four hydrophobic residues correspond to the hydrophobic residues in the 2 tail that form the surface with which it interacts with the a1 homeodomain; only two of these four residues are identical between the two proteins. Overall, this region of VP16 is identical to the 2 tail at these two positions out of a total length of 19.

The DNA-binding experiment of Figure 8B shows that this 2 hd::VP16 chimera is also capable of binding to DNA cooperatively with the a1 homeodomain. However, the binding is 50-fold weaker than that observed for the wild-type 2 protein. As discussed above, the VP16 and 2 tails differ considerably in amino acid sequence, and this difference may account for the poorer interaction seen with this chimera when compared with the 2 hd::K.l. tail.

When the residue in VP16 that corresponds to Leu196 in the 2 tail is mutated to serine (I377S), the interaction of the mutant chimeric protein with the a1 homeodomain is reduced by more than a factor of 10 (Figure 8B), indicating that this residue plays a crucial role in the interaction with a1 and suggesting that the VP16 tail interacts with the a1 homeodomain in a manner similar to that of the 2 tail.

a1 and 2 are homeodomain proteins that regulate cell identity in the budding yeast S.cerevisiae. Like other proteins that bind DNA through homeodomains, a1 and 2 each lack the specificity and affinity to select target DNA efficiently on their own. However, they form a heterodimer in solution which can then bind tightly and specifically to the hsg operators, thereby turning off transcription of the haploid-specific genes (for review see Johnson, 1995). In the DNA-bound state, the homeodomains of a1 and 2 both make extensive contacts with the hsg operator (Li et al., 1995).

The work presented in this paper shows that heterodimer formation is not simply the joining together of the a1 and 2 homeodomains. Heterodimerization also induces changes in each monomer, and the change in a1 is crucial for the DNA binding specificity of the heterodimer. A key contact in the heterodimer is formed by a flexible tail that extends from the 2 homeodomain and that becomes ordered upon contact with an exposed surface of a1 (Phillips et al., 1994; Li et al., 1995). This conformational change in the 2 tail has been well studied by NMR and X-ray crystallographic methods. In the 2 monomer, this tail is unstructured, but it assumes a distorted -helix upon contact with a1 (Phillips et al., 1991, 1994; Wolberger et al., 1991; Li et al., 1995). However, this change in 2 has no apparent consequence on the DNA-binding properties of 2 (Mak and Johnson, 1993), and the overall structure of the 2 homeodomain (excluding the tail) is the same whether or not it is complexed with a1.

In this paper, we have provided evidence that contact with the 2 tail induces a change in the a1 homeodomain which renders it competent to bind tightly and specifically to DNA; without this instructional change, the a1 homeodomain binds DNA only weakly, if at all. We propose that the change in the DNA-binding properties of a1 induced by contact with 2 results from a conformational change in a1. What might this conformational change be? Baxter et al. (1994) showed by applying isotope-edited NMR spectroscopy to ¹⁵N-labeled a1 homeodomain and ¹⁴N-labeled 2 homeodomain plus tail that the resonances of many positions in the a1 homeodomain changed upon addition of 2. Some of these changes can be accounted for by direct contact by the 2 tail; however, others lie in positions more distant from the sites of direct contact. A cluster of 2-induced changes lies in the loop between helix 1 and helix 2 of the a1 homeodomain. As seen in the heterodimer crystal structure, this loop makes contact with both the DNA (via a water molecule) and with the 2 tail, and it is plausible that contact with the 2 tail repositions this loop to maximize the affinity of a1 for DNA (Li et al., 1995). A network of protein–DNA contacts involving the backbone of this loop is clearly observed in the heterodimer–DNA crystal structure, and even a subtle change in the conformation of this loop (predicted to occur upon dissociation of the 2 tail) could disrupt this network of contacts and significantly weaken the affinity of a1 for DNA.

We have also provided evidence that the principles underlying the a1–2 interaction apply to the association of other homeodomain proteins. In particular, we have shown that two other tails can substitute for that of 2 in promoting heterodimer formation and DNA binding of a1 and 2. All three tails used in this study (from S.cerevisiae 2, from K.lactis 2 and from herpes virus VP16) are predicted (or in the case of S.cerevisiae 2, known) to form amphipathic helices. The VP16 region contains only two residue positions identical to those in S.cerevisiae 2 and the K.lactis tail only a single identical residue. However, the overall conservation of residue type across the entire tail region is higher in the K.lactis tail than in VP16, and this difference may explain why, of the two, the K.lactis tail is more efficient at cooperating with a1. Despite its lower efficiency, we believe that this short region of VP16 is able to interact with the a1 homeodomain in a manner similar to that of the 2 tail. This idea is based in part on the observation that a specific point mutation in the VP16 tail, located at a position corresponding to one crucial for the a1–2 interaction, destroys the cooperative binding with a1. Two possibilities, not mutually exclusive, could account for the observation that the heterologous tails do not function as well in mediating a1–2 cooperative DNA binding as does the S.cerevisiae 2 tail. First, the affinity between the heterologous tail and the surface of the a1 homeodomain may differ, with the S.cerevisiae 2 tail having the highest affinity. Secondly, the appropriate instructional change in a1 may be induced only by the 2 tail; the two other tails may effectively link a1 and 2 together, but may not fully induce the correct change in a1.

The fact that the VP16 tail can interact with the a1 homeodomain strongly supports the view that this region of VP16 interacts with the Oct-l homeodomain in a similar manner. A number of additional observations also support this view. Mutations in this region of VP16 disrupt cooperative binding to the DNA with Oct-1 without affecting the ability of VP16 to interact with another factor, HCF (Werstuck and Capone, l989a, b; Stern and Herr, 1991; Walker et al., 1994; Shaw et al., 1995; Lai and Herr, 1997). Peptides corresponding to this region of VP16 can affect the DNA-binding specificity of the Oct-l homeodomain, as does the intact VP16 protein (Stern and Herr, 1991). Similar peptides can inhibit formation of the Oct-1–VP16 complex, presumably by binding to the same surface of Oct-l as is normally bound by the full-length VP16 protein (Haigh et al., 1990; Hayes and O'Hare, 1993; Wu et al., 1994). Moreover, the surface of the Oct-1 homeodomain that contacts VP16 has been mapped through mutagenesis studies, and it corresponds to the same region of the a1 homeodomain that is contacted by 2 (Figure 9; Lai et al., 1992; Pomerantz et al., 1992).

Figure 9.

Comparison of the exposed surface of the a1 homeodomain and the Oct-1 homeodomain. Helices 1 and 2 of the a1 homeodomain are shown in red (left), and helices 1 and 2 of the Oct-1 homeodomain are shown in yellow (right). This surface is exposed when the homeodomain is bound to DNA. Residue numbering follows the standard convention established in previous studies of homeodomains (Qian et al., 1989). The labeled residues on a1 have been shown to interact with the 2 tail (Li et al., 1995). These residues form a hydrophobic depression, flanked on one end by F15 and on the other by a salt bridge (indicated by green dots), in which the hydrophobic face of the 2 tail helix rests. The surface of Oct-1 shows similarities to that of a1. Residues shown in blue, when mutated, affect the ability of Oct-1 to interact with VP16 (Lai et al., 1992; Pomerantz et al., 1992). Of these, S19 and E30 correspond to important interaction residues on the surface of a1. Residues shown in yellow have not been mutagenized, so their roles in the interaction with VP16 are unknown. However, these residues (in addition to S19 and E30) could form a hydrophobic patch and salt bridge, providing an interaction interface for the hydrophobic face of the proposed VP16 helix.

View full figure (51 KB)

The observations presented in this paper support the hypothesis that the principles underlying the a1–2 interaction are conserved among other homeodomain proteins. There are now numerous examples of cooperative interactions involving homeodomain proteins, some of which might also be mediated by these same types of interactions, specifically by an amphipathic helix binding to the surface of its partner homeodomain. Only a limited number of solvent-exposed residues are available on a homeodomain bound to DNA, so it is plausible that the interaction between a1 and 2 (an -helix resting on helices 1 and 2) may have appeared early during the evolution and diversification of homeodomain proteins.

There are also indications that additional pairs of homeodomain proteins undergo conformational changes upon heterodimerization. Chan and co-workers (1996) have proposed that the YPWM hexapeptide of the fly labial protein, a homolog of the mouse Hoxb1 protein, inhibits its DNA binding and that interaction with Exd removes this inhibition. A second example is found in the mammalian Pbx1 protein. The affinity of the Pbx1 protein is enhanced by YPWM-containing peptides derived from several of the partners of Pbx1, and one of several models that could explain this result is a peptide-induced conformational change in the Pbx1 protein (Knoepfler and Kamps, 1995; Peltenburg and Murre, 1996; Sanchez et al., 1997). Conformational changes induced by the heterodimerization of Hoxb8 and Pbx1 have been determined by circular dichroism spectroscopy (Sanchez et al., 1997), although the exact nature of these changes and their consequences for DNA binding remain to be determined.

Finally, the idea that 2 carries a ligand that increases the affinity of a1 for DNA is similar in principle to the many cases of small molecules that directly activate the binding of proteins to DNA. Examples include such ligands as cAMP for E.coli CAP (Beckwith, 1987; Reznikoff, 1992; Ebright, 1993) and tryptophan for the E.coli trp repressor (Yanofsky and Crawford, 1987; Somerville, 1992). In the case of a1, the small molecule is, in a sense, carried by the partner protein 2. A ligand-induced change seems an efficient way of ensuring that a1 is inactive in cells that lack 2 (a cells), but becomes activated only when 2 is also present, the condition that determines the a/ cell-type.

Construction of expression plasmids

The DNA encoding the glycine/serine linkers was synthesized as complementary oligonucleotides, annealed, and ligated between DNA encoding the a1 hd and DNA encoding the tail of 2 (pMS5; see Stark and Johnson, 1994). The linkers are composed of alternating (glycine)₂ and (serine)₂. The oligonucleotides that made up the linkers in the a1::2 chimeras are as follows: a1::11::2–GATCTAAAGGTGGTTC
TTCTGGCGGCTCCTCCG; a1::16::2–GATCTAAAGGTGGTTCTTCTGGCGGC
TCCTCCGGTGGCTCTTCCGGCG; 2::6::a1–GGTGGTTCTTCTGGT. The first two oligonucleotides have an overhanging GATC at the 5' end of each oligonucleotide of the pair, and were cloned into the BglII site at the junction of the a1 homeodomain and the 2 tail, yielding pMS74 and pMS87. The third linker oligonucleotide was part of a larger oligonucleotide which contains the 2 tail sequence (aa 189–210) immediately upstream of the linker. This oligonucleotide pair has an overhanging TA at each 5' end and was cloned into the NdeI site at the 5' end of the a1 homeodomain (pMS4; see Philips et al., 1994), resulting in pMS76.

The L196S mutants of the a1::11::2 and a1::16::2 chimeras were made by inserting the original oligos into the BglII site at the junction of the a1 homeodomain and the 2_L196S tail (pMS18), resulting in pMS91 and pMS92. The 2_L196S::6::a1 was generated by incorporating the codon change into the oligonucleotide pair and cloning into pMS4 (pMS77).

All 2 chimeric constructs were made by replacing the wild-type 2 tail with an oligo duplex consisting of either the K.lactis 2 tail (nucleotides corresponding to residues 212–223, MRS124 5'-TCGACGAAAAGAAAAAACAACTGCCGTTT
CGTCAGATATAAGAAACATTCTTAATTAAG, MRS125 5'-GATCCTTAATTAAGAATGTTTCTTATATCT
GACGAAACGGCAGTTGTTTTTTCTTTTCG), a mutant K.lactis 2 tail, I218S (MRS126 5'-TCGACGAAAAGAAAAAACAACTGCCGTTTCGTCA
GATTCGTCAGATTCTAGAAACATTCTTAATTAAG, MRS 127 5'-GATCCTTAATTAAGAATGTTTCTAGAATCTGACGAAACGGCAGTTGTTT
TTTCTTTTCG), the VP16 'tail' (nucleotides corresponding to residues 371–389, MRS100 5'-TCGACGAAAAGAAAAAACAAACAATTACGGGTCTACCATCGAGGGCCTGCT
CGAGGGCCTGCTCGATCTCCCGGACGACGACGCCCCCTAAG, MRS101 5'-GATCCTTAGGGGGCGTCGTCGT
CCGGGAGATCGAGCAGGCCCTCGA
GCAGGCCCTCGATGGTAGACCCGTAATTGTTTGTTTTTTCTTTTCG), or a mutant VP16 'tail', I377S (MRS118 5'-TCGACGAAAAGAAAAAACAAACAATTACGGGTCTACCTCTGAGGGCCTG
CTCGAGGGCCTGCTCGATCTCCCG
GACGACGACGCCCCCTAAG, MRS119 5'-GATCCTTAGGGGGCGTCGTCGTCCGGGAGATCGAGC
AGGCCCTCGAGCAGGCCCTCAGAGGTAGACCCG
TAATTGTTTGTTTTTTCTTTTCG). The tail oligo duplexes contain a SalI overhang on the 5' end and a BamHI overhang on the 3' end, with the first 19 nucleotides corresponding to 2 sequence upstream of the tail.

A SalI site was introduced into 2 19 bp upstream of the tail by site-directed mutagenesis (pMS20). The SalI site was removed from the yeast CEN ARS vector pAV115 containing the MAT locus (pMS21). pMS21 was cut with BglII and BamHI to remove the majority of the 2 gene. This was replaced with the corresponding BglII–BamHI fragment from pMS20 that contains 2 with the SalI site upstream of the 2 tail (pMS22). The SalI–BamHI fragment containing the 2 tail was removed from pMS22 and replaced with either the wild-type or mutant K.lactis 2 tail or VP16 'tail' (pMS104, pMS105, pMS85, pMS103, respectively). The 2 hd versions of these chimeras were generated by PCR using pMS104, 105, 85 and 103 as templates. The 5' primer for all of these PCR reactions introduces an NdeI site at the beginning of the 2 hd (o14039 5'-GATAAACAAACATATGAAACCTTACAGAG). The 3' primers contain a BamHI site and are specific for each tail: K.lactis 2 tail–MRS128 5'-GCCGGATCCTTAATTAAGAATGTTTC; VP16 'tail'–MRS106 5'-GCCGGATCCTTAGGGGGCGTC. The resulting PCR fragments were cut with NdeI and BamHI and cloned into these sites in the bacterial expression vector, pHB40P, under the control of the T7 promoter (Studier and Moffatt, 1986), resulting in pMS109–2 hd::K.l. tail, pMS110–2 hd::K.l. tail_I65S, pMS89–2 hd::VP16, pMS108–2 hd::VP16_I377S.

Peptides

2 tail peptides were synthesized by California Peptide Research, Inc. All three peptides are identical except for position 196; they are 19 amino acids in length, beginning at residue 189 of 2 and ending at residue 207. The amino acid sequence of the wild-type 2 tail peptide is TITIAPELADLLSGEPLAK. Residue 196 is shown in bold. Peptides were HPLC-purified, resuspended in H₂O, and concentrations were determined by the quantitative Ninhydrin assay (Sarin et al., 1981).

Operators

The a1–a1 operator has the same spacing and binding site orientation as a naturally occurring hsg operator (which contains an a1 and 2 binding site), except the 2 binding site of the hsg operator has been replaced by a second a1 binding site (Goutte and Johnson, 1994). The two a1 binding sites are separated by six base pairs and, for the experiments shown in Figures 2, 5 and 6, are contained within an 80-bp DNA fragment. A second DNA fragment, identical in sequence to the a1–a1 operator-containing fragment except that it contains no specific a1-binding sites, was used in the experiment of Figure 5. The removal of the a1-binding sites results in a 51 nucleotide pair DNA fragment.

Protein purification

The a1 hd protein and all a1::2 chimeric proteins were overexpressed in E.coli strain BL21(DE3)pLysS. Protein purification from cell lysates was by adhesion to a cation-exchange resin (Sephadex SP-C50, Pharmacia) followed by elution with a NaCl gradient (Phillips et al., 1994). The 2 hd+ tail fragment was a gift from A.Vershon.

All 2 hd chimeric proteins were present in bacterial extracts. These extracts were made from the protease-deficient E.coli strain CAG597D overexpressing plasmids pMS109, pMS110, pMS89 and pMS108. Cells were grown overnight to saturation at 30°C, harvested, resuspended in 7 ml/g lysis buffer {100 mM Tris–HCl, pH 8.0, 1 mM EDTA, 10 mM -mercaptoethanol, 500 mM NaCl and 0.1 mM [4-(2-aminoethyl)benzenesulfonyl fluoride] (AEBSF) (Calbiochem)}, sonicated to lyse the cells, and centrifuged at 30 000 g for 40 min. The supernatant was used in the gel shift assays after quantitating the amount of 2 hd chimeric protein in each extract by SDS gel followed by Coomassie Blue staining.

DNA-binding assays

For the electrophoretic mobility shifts containing only purified proteins (a1::2 chimeric proteins), proteins were incubated with a ³²P-labeled DNA fragment for 30 min at room temperature and electrophoresed through a 5% native Tris-borate–EDTA polyacrylamide gel as described previously (Stark and Johnson, 1994). For the electrophoretic mobility shifts containing overexpressed 2 hd chimeric proteins in bacterial extracts, the extracts were incubated, either in the presence or absence of purified a1 hd protein, with labeled DNA for 45 min on ice before electrophoresis. The binding conditions for the 2 tail peptides and labeled a1–a1 operator were the same as in the other a1::2 experiments except that the incubations of DNA and protein were carried out at 4°C, as was the electrophoresis.

The DNase I protection experiment was carried out under the same conditions as the mobility shift assays using purified proteins, except that 10–50 times more DNA was used (a ³²P-labeled DNA fragment containing an hsg operator), and the binding buffer contained no glycerol or E.coli genomic DNA, but was supplemented with 10 mM CaCl₂ and 2.5 mg/ml calf thymus DNA. Reactions were cleaved for 10 min at room temperature with 1.5 mg DNase I (Worthington) and then stopped and precipitated with 1.6 M ammonium acetate. Samples were electrophoresed through a 10% denaturing TBE gel.

We thank T.Li and C.Wolberger for providing us with the coordinates for the a1–2–DNA crystal structure prior to publication, S.Rader, S.Triezenberg and members of the Johnson laboratory for helpful discussions and comments on the manuscript, S.Rader for help with the structure representations and W.Schaffner for his support of D.E. This work was funded by a NIH Grant RO1 GM37049 to A.D.J.

Andrews BJ and Donoviel MS (1995) A heterodimeric transcriptional repressor becomes crystal clear. Science, 270, 251–253. | PubMed | ISI | ChemPort |

Baxter SM, Gontrum DM, Phillips CL, Roth AF and Dahlquist FW (1994) Heterodimerization of the yeast homeodomain transcriptional regulators 2 and a1: secondary structure determination of the a1 homeodomain and changes produced by 2 interactions. Biochemistry, 33, 15309–15320. | PubMed | ISI | ChemPort |

Beckwith J (1987) The lactose operon. In Neidhardt,F.C. (ed.), Escherichia coli and Salmonella Typhimurium: Cellular and Molecular Biology. American Society for Microbiology, Washington, DC, pp. 1444–1452.

Chan SK, Pöpperl H, Krumlauf R and Mann RS (1996) An extradenticle-induced conformational change in a Hox protein overcomes an inhibitory function of the conserved hexapeptide motif. EMBO J, 15, 2476–2487. | PubMed | ISI | ChemPort |

Dranginis AM (1990) Binding of yeast a1 and 2 as a heterodimer to the operator DNA of a haploid-specific gene. Nature, 347, 682–685. | Article | PubMed | ISI | ChemPort |

Ebright RH (1993) Transcription activation at class I CAP-dependent promoters. Mol Microbiol, 8, 797–802. | Article | PubMed | ISI | ChemPort |

Goutte C and Johnson AD (1988) a1 protein alters the DNA binding specificity of 2 repressor. Cell, 52, 875–882. | Article | PubMed | ISI | ChemPort |

Goutte C and Johnson AD (1993) Yeast a1 and 2 homeodomain proteins form a DNA-binding activity with properties distinct from those of either protein. J Mol Biol, 233, 359–371. | Article | PubMed | ISI | ChemPort |

Goutte C and Johnson AD (1994) Recognition of a DNA operator by a dimer composed of two different homeodomain proteins. EMBO J, 13, 1434–1442. | PubMed | ISI | ChemPort |

Haigh A, Greaves R and O'Hare P (1990) Interference with the assembly of a virus-host transcription complex by peptide competition. Nature, 344, 257–259. | Article | PubMed | ISI | ChemPort |

Hayes S and O'Hare P (1993) Mapping of a major surface-exposed site in herpes simplex virus protein Vmw65 to a region of direct interaction in a transcription complex assembly. J Virol, 67, 852–862. | PubMed | ISI | ChemPort |

Herskowitz I, Rine J and Strathern J (1992) Mating-type determination and mating-type interconversion in Saccharomyces cerevisiae. In Jones,E.W., Pringle,J.R. and Broach,J.R. (eds), The Molecular and Cellular Biology of the Yeast Saccharomyces. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp. 583–656.

Ho CY, Adamson JG, Hodges RS and Smith M (1994) Heterodimerization of the yeast MATa1 and MAT2 proteins is mediated by two leucine zipper-like coiled-coil motifs. EMBO J, 13, 1403–1413. | PubMed | ISI | ChemPort |

Johnson AD (1995) Molecular mechanisms of cell-type determination in budding yeast. Curr Biol, 5, 552–558. | Article | PubMed | ISI | ChemPort |

Keleher CA, Redd MJ, Schultz J, Carlson M and Johnson AD (1992) Ssn6-Tup1 is a general repressor of transcription in yeast. Cell, 68, 709–719. | Article | PubMed | ISI | ChemPort |

Knoepfler PS and Kamps MP (1995) The pentapeptide motif of Hox proteins is required for cooperative DNA binding with Pbx1, physically contacts Pbx1 and enhances DNA binding by Pbx1. Mol Cell Biol, 15, 5811–5819. | PubMed | ISI | ChemPort |

Komachi K, Redd MJ and Johnson AD (1994) The WD repeats of Tup1 interact with the homeo domain protein 2. Genes Dev, 8, 2857–2867. | PubMed | ISI | ChemPort |

Lai JS and Herr W (1997) Interdigitated residues within a small region of VP16 interact with Oct-1, HCF and DNA. Mol Cell Biol, 17, 3937–3946. | PubMed | ISI | ChemPort |

Lai JS, Cleary MA and Herr W (1992) A single amino acid exchange transfers VP16-induced positive control from the Oct-1 to the Oct-2 homeo domain [published erratum appears in Genes Dev. 1992, 6, 2663]. Genes Dev, 6, 2058–2065. | PubMed | ISI | ChemPort |

Li T, Stark MR, Johnson AD and Wolberger C (1995) Crystal structure of the MATa1/MAT2 homeodomain heterodimer bound to DNA. Science, 270, 262–269. | Article | PubMed | ISI | ChemPort |

Mak A and Johnson AD (1993) The carboxy-terminal tail of the homeo domain protein 2 is required for function with a second homeo domain protein. Genes Dev, 7, 1862–1870. | PubMed | ISI | ChemPort |

Mukai Y, Harashima S and Oshima Y (1991) AAR1/TUP1 protein, with a structure similar to that of the subunit of G proteins, is required for a1–2 and 2 repression in cell type control of Saccharomyces cerevisiae. Mol Cell Biol, 11, 3773–3779. | PubMed | ISI | ChemPort |

Peltenburg LT and Murre C (1996) Engrailed and Hox homeodomain proteins contain a related Pbx interaction motif that recognizes a common structure present in Pbx. EMBO J, 15, 3385–3393. | PubMed | ISI | ChemPort |

Phillips CL, Vershon AK, Johnson AD and Dahlquist FW (1991) Secondary structure of the homeo domain of yeast 2 repressor determined by NMR spectroscopy. Genes Dev, 5, 764–772. | PubMed | ISI | ChemPort |

Phillips CL, Stark MR, Johnson AD and Dahlquist FW (1994) Heterodimerization of the yeast homeodomain transcriptional regulators 2 and a1 induces an interfacial helix in 2. Biochemistry, 33, 9294–9302. | Article | PubMed | ISI | ChemPort |

Pomerantz JL, Kristie TM and Sharp PA (1992) Recognition of the surface of a homeo domain protein. Genes Dev, 6, 2047–2057. | PubMed | ISI | ChemPort |

Qian YQ, Billeter M, Otting G, Muller M, Gehring WJ and Wuthrich K (1989) The structure of the Antennapedia homeodomain determined by NMR spectroscopy in solution: comparison with prokaryotic repressors. [published erratum appears in Cell 1990, 61, 548] Cell, 59, 573–580. | Article | PubMed | ISI | ChemPort |

Reznikoff WS (1992) Catabolite gene activator protein activation of lac transcription. J Bacteriol, 174, 655–658. | PubMed | ISI | ChemPort |

Sanchez M, Jennings PA and Murre C (1997) Conformational changes induced in Hoxb-8/Pbx-1 heterodimers in solution and upon interaction with specific DNA. Mol Cell Biol, 17, 5369–5376. | PubMed | ISI | ChemPort |

Sarin VK, Kent SB, Tam JP and Merrifield RB (1981) Quantitative monitoring of solid-phase peptide synthesis by the ninhydrin reaction. Anal Biochem, 117, 147–157. | Article | PubMed | ISI | ChemPort |

Shaw P, Knez J and Capone JP (1995) Amino acid substitutions in the herpes simplex virus transactivator VP16 uncouple direct protein–protein interaction and DNA binding from complex assembly and transactivation. J Biol Chem, 270, 29030–29037. | Article | PubMed | ISI | ChemPort |

Smith DL and Johnson AD (1992) A molecular mechanism for combinatorial control in yeast: MCM1 protein sets the spacing and orientation of the homeodomains of an 2 dimer. Cell, 68, 133–142. | Article | PubMed | ISI | ChemPort |

Somerville R (1992) The Trp repressor, a ligand-activated regulatory protein. Prog Nucleic Acid Res Mol Biol, 42, 1–38. | PubMed | ISI | ChemPort |

Stark MR and Johnson AD (1994) Interaction between two homeodomain proteins is specified by a short C-terminal tail. Nature, 371, 429–432. | Article | PubMed | ISI | ChemPort |

Stern S and Herr W (1991) The herpes simplex virus trans-activator VP16 recognizes the Oct-1 homeo domain: evidence for a homeo domain recognition subdomain. Genes Dev, 5, 2555–2566. | PubMed | ISI | ChemPort |

Strathern J, Shafer B, Hicks J and McGill C (1988) a/-specific repression by MAT2. Genetics, 120, 75–81. | PubMed | ISI | ChemPort |

Studier FW and Moffatt BA (1986) Use of bacteriophage T7 RNA polymerase to direct selective high-level expression of cloned genes. J Mol Biol, 189, 113–130. | Article | PubMed | ISI | ChemPort |

Vershon AK, Jin Y and Johnson AD (1995) A homeo domain protein lacking specific side chains of helix 3 can still bind DNA and direct transcriptional repression. Genes Dev, 2, 182–192.

Walker S, Hayes S and O'Hare P (1994) Site-specific conformational alteration of the Oct-1 POU domain-DNA complex as the basis for differential recognition by Vmw65 (VP16). Cell, 79, 841–852. | Article | PubMed | ISI | ChemPort |

Werstuck G and Capone JP (1989a) Identification of a domain of the herpes simplex virus trans-activator Vmw65 required for protein–DNA complex formation through the use of protein A fusion proteins. J Virol, 63, 5509–5513. | ISI | ChemPort |

Werstuck G and Capone JP (1989b) Mutational analysis of the herpes simplex virus trans-inducing factor Vmw65. Gene, 75, 213–224. | Article | PubMed | ISI | ChemPort |

Wolberger C, Vershon AK, Liu B, Johnson AD and Pabo CO (1991) Crystal structure of a MAT2 homeodomain–operator complex suggests a general model for homeodomain–DNA interactions. Cell, 67, 517–528. | Article | PubMed | ISI | ChemPort |

Wu TJ, Monokian G, Mark DF and Wobbe CR (1994) Transcriptional activation by herpes simplex virus type 1 VP16 in vitro and its inhibition by oligopeptides. Mol Cell Biol, 14, 3484–3493. | PubMed | ISI | ChemPort |

Yanofsky C and Crawford IP (1987) The tryptophan operon. In Neidhardt,F.C. (ed.), Escherichia coli and Salmonella Typhimurium: Cellular and Molecular Biology. American Society for Microbiology, Washington, DC, pp. 1453–1472.

Main navigation

• Search EMBO Journal

  Search

• Advanced search

• The EMBO Journal

  • Home
  • Aims and scope
  • Current issue
  • Advance online publication
  • Web focuses
  • Archive:
  • Browse by issue
  • Browse by subject
  • Browse by category
  • Free online sample issue
  • Press releases

• Authors and Referees

  • Editorial process
  • Guide for authors
  • Submit an article
  • Guide for referees
  • Editors and Editorial Board
  • Contact editorial office

• Customer Service

  • Subscribe
  • Order sample copy
  • Purchase articles
  • Reprints and permissions
  • Contact NPG
  • Advertising