Attractive Interhelical Electrostatic Interactions in the Proline- and Acidic-rich Region (PAR) Leucine Zipper Subfamily Preclude Heterodimerization with Other Basic Leucine Zipper Subfamilies*

Basic region-leucine zipper (B-ZIP) proteins homo- or heterodimerize to bind sequence-specific double-stranded DNA. We present circular dichroism (CD) thermal denaturation data on vitellogenin promoter-binding protein (VBP), a member of the PAR subfamily of B-ZIP proteins that also includes thyroid embryonic factor, hepatocyte leukemia factor, and albumin site D-binding protein. VBP does not heterodimerize with B-ZIP domains from C/EBPα, JUND, or FOS. We describe a dominant negative protein, A-VBP, that contains the VBP leucine zipper and an acidic amphipathic protein sequence that replaces the basic region critical for DNA binding. The acidic extension forms a coiled coil structure with the VBP basic region in the VBP·A-VBP heterodimer. This new α-helical structure extends the leucine zipper N-terminally, stabilizing the complex by 2.0 kcal/mol. A-VBP abolishes DNA binding of VBP in an equimolar competition assay, but does not affect DNA binding even at 100-fold excess of CREB, C/EBPα, or FOS/JUND. Likewise, proteins containing the acidic extension appended to seven other leucine zippers do not inhibit VBP DNA binding. We show that conserved g ↔ e′ or i, i′ +5 salt bridges are sufficient to confer specificity to VBP by mutating the C/EBPα leucine zipper to contain the g ↔ e′ salt bridges that characterize VBP. A-VBP heterodimerizes with this mutant C/EBP, preventing it from binding to DNA. These conserved g ↔ e′ electrostatic interactions define the specificity of the PAR subfamily of B-ZIP proteins and preclude interaction with other B-ZIP subfamilies.

Basic region-leucine zipper (B-ZIP) 1 proteins are an abundant class of sequence-specific DNA-binding proteins found exclusively in eukaryotes (1,2). Crystal structures of the yeast GCN4 leucine zipper (3) and mammalian B-ZIP domains bound to DNA (4,5) indicate that the leucine zipper is a parallel dimeric coiled coil. The structural rules regulating dimerization specificity of leucine zippers have been examined using the GCN4 leucine zipper and a variety of mammalian leucine zip-pers (6 -8). Charged amino acids in the g and eЈ positions of the coiled coil produce interhelical g 7 eЈ or i, iЈ ϩ 5 interactions that either promote or inhibit dimerization depending on whether they are attractive or repulsive (9 -11).
We have previously designed a series of dominant negative proteins to several B-ZIPs in order to examine in vivo B-ZIP function. All dominant negatives, termed A-ZIPs, have the B-ZIP basic region replaced with an amphipathic acidic ␣-helical sequence that heterodimerizes with the basic region in the B-ZIP⅐A-ZIP heterodimer. The A-ZIP forms an N-terminal coiled coil extension of the leucine zipper that stabilizes the heterodimer by 2-5 kcal/mol (12)(13)(14) and prevents B-ZIP DNA binding at equimolar concentrations. A-ZIPs are being used as in vivo reagents to block activity of specific B-ZIP proteins (15).
We have used the A-ZIPs to examine dimerization specificity in vitro. In the present study, we have examined the VBP (vitellogenin promoter-binding protein) leucine zipper, the chicken ortholog (16) of the mammalian TEF (thyroid embryonic factor) protein (17). Two related mammalian B-ZIP proteins have also been identified: DBP (albumin site D-binding protein) (18) and HLF (hepatocyte leukemia factor) (19,20). This subclass of B-ZIP proteins is called PAR because it contains a proline-and acidic amino acid-rich region N-terminal to the B-ZIP domain. The PAR proteins share extensive sequence similarity (65-73%) in the B-ZIP domain, with four pairs of conserved attractive g 7 eЈ interhelical charge interactions containing glutamate and arginine. PAR proteins are expressed in the brain, liver, and thyroid and may be required for a functional circadian clock in Drosophila (21). A chimeric protein consisting of the B-ZIP domain of HLF and the E2A transactivation domain mediates a lethal form of childhood leukemia (19,20).
A dominant negative specific to the PAR family would therefore be a valuable reagent both for biochemical studies and as a clinical reagent. We describe A-VBP, a dominant negative protein that inhibits VBP DNA binding in an equimolar competition assay, but does not inhibit the DNA binding of other B-ZIP domains. To gain insight into dimerization specificity in the PAR family, we used a C/EBP␣ mutant with the same g 7 eЈ interhelical salt bridges as VBP. This mutant protein interacts with A-VBP, suggesting that the g 7 eЈ salt bridges are important structural determinants in regulating VBP dimerization specificity with other leucine zippers.

EXPERIMENTAL PROCEDURES
Proteins-The sequence of the 96-amino acid VBP protein is ASMTGGQQMGRDP-LEE-KVFVPDEQKDEKYWTRRKKNNVAA-KRSRDARRLKENQI 1 TIRAAFL 2 EKENTAL 3 RTEVAEL 4 RKEVGRC-5 KNIVSKY 6 ETRYGPL. The d positions are in bold type. The first 13 amino acids are from 10, the next three amino acids are a cloning linker, and the remaining 80 amino acids are the C terminus of VBP. All * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. B-ZIP proteins described in this paper contain the 10 leader and 37 amino acids N-terminal of the first d leucine position of the leucine zipper ( Fig. 1). All A-ZIP proteins, except for FOS and ATF2, which are truncated, contain the leucine zipper region starting at the first leucine position (Fig. 1) and extend until the natural C terminus. All A-ZIPs, except for A-CREB, contain the acidic extension LEQRAEELA-RENEELLEKEAEELEQENAELE (12). A-CREB contains a leucine instead of an asparagine (in bold) in the acidic extension (14). The LE in bold is the XhoI site that is the border between the first d position in the leucine zipper and the N-terminal acidic extension.
Proteins were expressed in Escherichia coli using the T7 isopropyl-1-thio-␤-D-galactopyranoside-inducible system (22). B-ZIP domains were purified over a heparin column and subsequently purified on a Rainin high performance liquid chromatography system using a C18 column chromatographed from 0% to 100% acetonitrile in 0.1% trifluoroacetic acid (12). The A-ZIP and O-VBP (the VBP leucine zipper without an N-terminal extension) constructs were purified over a hydroxylapatite column, eluting with 200 mM phosphate and subsequently purified as for the B-ZIPs.
Circular Dichroism-Circular dichroism (CD) studies were performed using a Jasco J-720 spectropolarimeter. All protein stock solutions were in 12.5 mM potassium phosphate (pH 7.4), 150 mM KCl, and 0.25 mM EDTA. For the assay, 1 mM DTT and 4 M protein sample in 1.5 ml of stock buffer (described above) was heated to 65°C for 20 min to reduce the cysteines, cooled to room temperature for 5 min, and added to a 5-mm rectangular CD cell.
Thermodynamic Calculations-Melting temperature (T m ) and enthalpy (⌬H) values were determined from denaturation curves assuming a two-state equilibrium dissociation of ␣-helical dimers into unfolded monomers using ⌬C p of Ϫ1.2 kcal mol Ϫ1 (12). ⌬G values are reported at 37°C.
DNA Binding Assay-Purified B-ZIP domains (5 ϫ 10 Ϫ6 M dimer) were heated for 10 min at 65°C in the presence of 1 mM DTT. These proteins were then diluted to a final concentration of 20 -100 nM in 20 l of the gel-shift reaction buffer (12.5 mM KPO 4 , pH 7.4, 150 mM KCl, 0.25 mM EDTA, 2.5 mM DTT, 2 mg/ml bovine serum albumin, 2% glycerol). This solution was incubated for 10 min at 65°C, then mixed with 0.1 pmol of the probe ( 32 P-labeled double-stranded oligonucleotide). A-ZIP proteins were similarly heated with DTT and added at the indicated molar equivalence before the second heating step. The binding complexes were resolved on a 7.5% polyacrylamide gel in 0.25ϫ TBE buffer at room temperature. The sequence of the four 28-mer DNA probes is as follows (reported consensus binding site is underlined): AP1, GTCAGTCAGAATGACTCATATCGGTCAG; CREB, GTCAGTCA-GATGACGTCATATCGGTCAG; VBP, GTCAGTCAGATTACGTAATA-TCGGTCAG; C/EBP, GTCAGTCAGATTGCGCAATATCGGTCAG.

RESULTS
The VBP Subfamily of B-ZIP Proteins-The VBP B-ZIP domain is 80 amino acids long, containing a 37-amino acid basic region and a 43-amino acid leucine zipper. The B-ZIP domain is sufficient for sequence-specific DNA binding (16,23). Native VBP, including the transactivation domain, is 313 amino acids long. The VBP leucine zipper homodimer is schematically presented in Fig. 1. The lower panel of Fig. 1 presents a sequence alignment showing the conserved amino acids in the PAR family of leucine zippers. Four pairs of attractive g 7 eЈ interhelical salt bridges in VBP are highlighted by bold arrows. The first pair is R 7 E, and the next three are all E 7 R pairs. Biochemical studies indicate that PAR proteins can readily form heterodimers within the family but not with C/EBP (16,19,24). However, the ability of VBP to heterodimerize with a range of leucine zippers has never been tested. We mixed VBP with C/EBP␣ (Fig. 2) and saw no increase in thermal stability of the mixtures by CD spectroscopy. Similar results were ob- tained for mixtures of VBP with CREB, JUND, or FOS B-ZIP domains (data not shown). This indicates that VBP does not preferentially heterodimerize with B-ZIPs from other subfamilies. From these data, it is clear that homodimers are thermo-dynamically significantly favored over heterodimers, but does not reveal whether heterodimers are thermodynamically unfavored relative to monomers.
A-VBP Interacts with VBP and Can Prevent VBP DNA Bind-   Fig. 3A) or A-VBP (Table I; Fig. 3B). The CD trace for the VBP B-ZIP domain thermal denaturation has an inclined low temperature base line that probably represents the non-cooperative denaturation of the basic region. Deleting the basic region from VBP to produce O-VBP reduced the ellipticity () from Ϫ42 mdeg to Ϫ35 mdeg at 6°C and increased the T m , suggesting that the basic regions are destabilizing by 1.1 kcal/mol/dimer. The low temperature base line of O-VBP is nearly level, suggesting that the VBP leucine zipper domains denatures more cooperatively than the VBP B-ZIP domain. A-VBP is 0.4 kcal/mol less stable than O-VBP, indicating that the acidic extension is also slightly repulsive in the A-VBP homodimer. Unlike the basic region that unfolds non-cooperatively at low temperatures, the acidic extension displays a cooperative unfolding at around 20°C that is independent of the VBP leucine zipper cooperative unfolding at 50°C.
The mixture of A-VBP and VBP is 2.0 kcal/mol more stable than the VBP homodimer. The 43% increase in ellipticity at low temperatures implies that the basic region and acidic extension form an ␣-helical coiled coil extension of the leucine zipper. The absence of a low temperature transition for the B-VBP⅐A-VBP heterodimer suggests that the basic and acidic regions form an extended coiled-coil that denatures cooperatively with the VBP zipper in the heterodimer. Analytical ultracentrifugation of the B-VBP⅐A-VBP mixture indicates the sample is a dimer (data not shown). Fig. 3C shows the results of a gel shift assay after adding varying concentrations of either O-VBP or A-VBP to a DNA binding reaction containing 20-nm VBP and its cognate DNA site (ATTACGTAAT) in a 28-base pair double-stranded oligonucleotide. O-VBP inhibits VBP DNA binding at between 10 and 30 molar eq. A-VBP, in contrast, almost completely abolishes DNA binding at equimolar concentrations. Fig. 3D presents a schematic of the heterodimer between A-VBP and the VBP B-ZIP domain. (Fig. 4). Analytical ultracentifugation of the mixture at 6°C indicates a greater than dimer molecular weight (data not shown). We interpret this to be a A-VBP homodimer and a C/EBP␣ homodimer whose acidic and basic regions are interacting, resulting in both an increase in ellipticity and molecular weight. The 2.4 kcal/mol interaction between the acidic extension and the basic region of C/EBP␣ (12) did not overcome the repulsion between their leucine zipper regions. Similarly, A-VBP did not heterodimerize with CREB, FOS, or JUND (data not shown).

A-VBP Does Not Heterodimerize with Other B-ZIP Domains-CD thermal denaturation experiments of the mixture of A-VBP and the C/EBP␣ B-ZIP domain show increased ellipticity at low temperatures but no increase in thermal stability
We also used gel-shift assays to measure inhibition of DNA binding by A-VBP. Fig. 5C shows that A-VBP did not inhibit the DNA binding of C/EBP, CREB, or JUND/FOS bound to their cognate sites, even at 100 molar excess.
In a corollary experiment (Fig. 5B), we examined the ability of seven A-ZIP leucine zippers to inhibit the DNA binding of VBP. The leucine zipper protein sequences used in these experiments are shown in Fig. 5A. The number of putative attractive or repulsive g 7 eЈ interactions in homodimer or heterodimer with VBP is also calculated. The control A-VBP inhibited binding at equimolar concentration. However, none of the other zippers inhibited VBP DNA binding, even at 100-fold molar excess. All of these A-ZIPs did inhibit DNA binding of their normal dimerization partners.
A C/EBP␣ g 7 eЈ Mutant Leucine Zipper Heterodimerizes with VBP-We have designed a C/EBP␣ leucine zipper mutant protein, called C/EBP␣-L, that has the same g 7 eЈ salt bridges that are found in VBP (25). This required changing four amino acids in the g and e position of C/EBP␣ leucine zipper (Fig. 6). We examined the interaction of C/EBP␣-L with A-VBP to investigate the importance of these g 7 eЈ salt bridges to VBP dimerization specificity. The thermal stabilities of C/EBP␣ and C/EBP␣-L homodimers are shown in Table II. Fig. 7 (A and B) presents the CD thermal denaturation curves for C/EBP␣-L alone or mixed with either O-VBP or A-VBP. O-VBP did not preferentially heterodimerize with C/EBP␣-L (Fig. 7A). However, A-VBP did heterodimerize with C/EBP␣-L as shown by a 10°C increase in the melt line relative to the sum line at ϳ50°C (Fig. 7B). Previous work has shown that the acidic extension heterodimerizes with the C/EBP basic region contributing 2.4 kcal/mol to stability (12). Thus, heterodimerization between C/EBP␣-L and A-VBP implies that repulsion between the two zippers is less than 2.4 kcal/mol. In contrast, heterodimerization is not observed between A-VBP and another mutant C/EBP␣ protein called C/EBP␣-J. C/EBP␣-J has the g and e residues in the third heptad reversed (compared with C/EBP␣-L) so that there are six attractive and two repulsive g 7 eЈ interactions with the VBP leucine zipper (Fig. 6B). The lack of interaction between C/EBP␣-J and A-VBP indicates that there are repulsive interactions between the zippers that destabilize the heterodimer by more than 2.4 kcal/ mol (Fig. 7, C and D). Fig. 8 is a gel shift that shows that the DNA binding of C/EBP␣-L is abolished by A-VBP of between 1 and 10 molar eq. This is consistent with the CD results. In contrast the DNA binding of C/EBP␣-J is not inhibited by 100 molar eq of A-VBP. These data indicate that the attractive g 7 eЈ salt bridges play an important role in regulating the dimerization specificity of the VBP leucine zipper.  5. A, sequence alignment of leucine zippers used in experiments with VBP. Except for FOS and ATF2 that are truncated as denoted by the double dashes, all protein sequences for the B-ZIP leucine zippers start in the basic region at an invariant arginine, continue through the leucine zipper and terminate at the natural C terminus of these proteins. The proline (bold P) in the C terminus of the leucine zipper is likely to be the helix-breaking residue at the natural C-terminal boundary of the ␣-helical leucine zipper. Below the protein sequences is a consensus sequence for the B-ZIP motif, where represents any hydrophobic amino acid. The leucine zipper sequence is broken into heptads (g, a, b, c, d, e, f). Above the sequences is a bar connecting amino acids in the g position with those in the following e position, indicating amino acids that may interact interhelically in either homodimers or heterodimers. Only four heptads are shown because of the presence of the proline and lack of charged residues at the C terminus of the these molecules. The number of attractive and repulsive g 7 eЈ salt bridges formed in a hypothetical heterodimer between VBP and other B-ZIP proteins are presented. B, VBP DNA binding is inhibited by A-VBP but not other A-ZIPs. 30 nm VBP was bound to its cis element and competed with 1, 10, or 100 molar eq of the indicated A-ZIP. Lane 1 contains the free probe. Lane 2 contains VBP bound to its consensus DNA sequence. 1, 10, and 100 molar eq of each A-ZIP are competed against the VBP-DNA interaction as shown. At the 100 molar eq, some of the A-ZIPs show minimal VBP binding, notably A-C/EBP and A-CREB. C, A-VBP does not inhibit DNA binding of other B-ZIP proteins. VBP (30 nM), C/EBP (30 nM), CREB (20 nM), and JUND/FOS (30 nM) were bound to their respective probes and challenged with 1, 10, or 100 molar eq of A-VBP. As previously observed, A-VBP selectively inhibits VBP DNA binding at 1 molar eq. binding to its cognate site at equimolar concentrations in EMSA, but does not interact with other B-ZIPs, even at 100fold excess. Using eight A-ZIP dominant negatives, we show by EMSA and CD thermal melts that the PAR B-ZIPs do not interact with the other four known subfamilies of mammalian B-ZIP proteins: FOS, C/EBP, JUN, and CREB. These data strongly suggest that A-VBP may be a potent and specific reagent for inhibiting both the activation and repressive properties of the VBP family of B-ZIP proteins in biological systems. To facilitate further design of A-ZIP dominant negatives, we need to have a detailed understanding of the interactions that govern dimerization specificity.
We have used A-VBP to investigate the structural determinants regulating the preferential dimerization within the PAR subfamily. We find that conserved attractive interhelical g 7 eЈ (i, iЈϩ5) salt bridges play a major role. C/EBP␣-L protein, containing the VBP g 7 eЈ salt bridges in a C/EBP␣ background, interacts with A-VBP while C/EBP␣ does not. This suggests that these g 7 eЈ salt bridges are critical for regulating homodimerization of the PAR subfamily. Their importance is reinforced by the observation that a reversal of a single E 7 R salt bridge to R 7 E in the mutant C/EBP␣-J prevented heterodimerization with A-VBP.
The expected difference in stability between the A-VBP⅐C/ EBP␣-L and A-VBP⅐C/EBP␣-J heterodimers is 2.9 kcal/mol ( Fig. 9) (10, 23). Of this, 2.5 kcal/mol is due to coupling energy. 40% of the coupling energy is from attractive g 7 eЈ interactions (Ϫ1.0 kcal/mol), and 60% is from repulsive interactions (ϩ1.5 kcal/mol). The C/EBP␣-L⅐A-VBP heterodimer would be stabilized by Ϫ1.3 kcal/mol for each of the two E 7 R salt bridges relative to A 7 A in the third heptad, resulting in a stabilization of Ϫ2.6 kcal/mol and a coupling energy of Ϫ1.0 kcal/mol. In contrast, the C/EBP␣-J⅐A-VBP heterodimer would be destabilized by 0.3 kcal/mol because repulsive E 7 E and the R 7 R interactions result in a destabilizing coupling energy of ϩ1.5 kcal/mol. These attractive g 7 eЈ interactions are not observed in the GCN4 leucine zipper (26).
A general strategy to measure repulsion between protein surfaces is to entropically constrain them. The strategy we present in this report is more limited, the heterodimer between the basic region and the acidic extension must denature cooperatively with the leucine zippers. Additionally, to measure the repulsion between the leucine zippers of interest, it must be less than the basic region and acidic region interaction.
PAR proteins are expressed in the brain (27), liver, and thyroid and have a circadian rhythm in their expression pattern (28 -30). Recent work has demonstrated that a Drosophila B-ZIP protein with homology to PAR B-ZIP domains is required for a functional circadian clock (21). A chimeric protein between the B-ZIP domain of HLF and the transactivation domain E2A mediates a lethal form of childhood leukemia (24). Previous work using a TEF dominant negative containing mutations in the basic region had dominant negative properties (31). Expression of this dominant negative in UOC-B1 cells, a human leukemia cell line that contains a chimeric protein containing the TEF B-ZIP domain, lead to apoptosis. The VBP/ TEF dominant negative described in this manuscript should be more effective because the acidic extension confers higher affinity to the target B-ZIP domain. When gene therapy procedures are in place, A-VBP could be an effective therapy for this particular leukemia.