SREBP-1 dimerization specificity maps to both the helix-loop-helix and leucine zipper domains: use of a dominant negative.

The mammalian SREBP family contains two genes that code for B-HLH-ZIP proteins that bind sequence-specific DNA to regulate the expression of genes involved in lipid metabolism. We have designed a dominant negative (DN), termed A-SREBP-1, that inhibits the DNA binding of either SREBP protein. A-SREBP-1 consists of the dimerization domain of B-SREBP-1 and a polyglutamic acid sequence that replaces the basic region. A-SREBP-1 heterodimerizes with either B-SREBP-1 or B-SREBP-2, and both heterodimers are more stable than B-SREBP-1 bound to DNA. Circular dichroism thermal denaturation studies show that the B-SREBP-1.A-SREBP-1 heterodimer is -9.8 kcal mol(-1) dimer(-1) more stable than the B-SREBP-1 homodimer. EMSA assays demonstrate that A-SREBP-1 can inhibit the DNA binding of either B-SREBP-1 or B-SREBP-2 in an equimolar competition but does not inhibit the DNA binding of the three B-HLH-ZIP proteins MAX, USF, or MITF, even at 100 molar eq. Chimeric proteins containing the HLH domain of SREBP-1 and the leucine zipper from either MAX, USF, or MITF indicate that both the HLH and leucine zipper regions of SREBP-1 contribute to its dimerization specificity. Transient co-transfection studies demonstrate that A-SREBP-1 can inhibit the transactivation of SREBP-1 and SREBP-2 but not USF. A-SREBP-1 may be useful in metabolic diseases where SREBP family members are overexpressed.

Eukaryotic Expression Vectors-The polylinker region of the pRc/ CMV eukaryotic expression vector (Invitrogen) was cut with HindIII and XbaI, and the following annealed oligonucleotide was inserted: 5Ј-AGCTCCACCATGGACTACAAGGACGACGATGACAAGCATATGT-GATGAAGCTT-3Ј (underlined bases are restriction sites for NcoI, NdeI, and HindIII) and 5Ј-CTAGAAGCTTCATCACATATGCTTGTCA-TCGTCGTCCTTGTAGTCCATGGTGG-3Ј (underlined bases are restriction site for XbaI). The HindIII restriction site was not regenerated and was reintroduced downstream of the FLAG sequence (DYKDDD-DK). We named this plasmid CMV500. Similarly, CMV566 with a hemagglutinin epitope (YPYDVPDYASL) and NdeI-HindIII cloning site was made using the following pair of complementary oligonucleotides: 5Ј-AGCTCCACCATGGCGTATCCCTACGACGTGCCCGATTATGCCC-ATATGTGATGAAGCTT-3Ј (underlined bases are restriction sites for NcoI, NdeI, and HindIII) and 5Ј-CTAGAAGCTTCATCACATATGGGC-ATAATCGGGCACGTCGTAGGGATACGCCATGGTGG-3Ј (underlined bases are restriction site for XbaI). For cloning A-HLH-ZIP inserts, CMV500 or CMV566 plasmid was partially digested with NdeI and HindIII. An NdeI-HindIII insert coding for the A-HLH-ZIP domain was cloned into a CMV500 or CMV566 NdeI-HindIII digest. The insert sequence was confirmed using the following sequencing primers: 5Ј-G-GCTAACTAGAGAACCCACTGC-3Ј (105 bases upstream) and 5Ј-TGG-CTGGCAACTAGAAGGC-3Ј (52 bases downstream).
Protein Expression and Purification-Plasmids containing DNA encoding the B-HLH-ZIP domain were transformed into Escherichia coli (BL21 LysE) (15). Transformed cells were grown in Super Broth containing 100 g/ml ampicillin at 37°C until optical density at 600 nm reached 0.6. Cells were induced to produce protein by the addition of 1 mM isopropyl-␤-D-thiogalactopyranoside and were harvested after 3 h of induction and processed as described earlier (16). After dialysis against low salt buffer (20 mM Tris-HCl (pH 8.0) with 50 mM KCl, 1 mM EDTA, and 0.2 mM phenylmethylsulfonyl fluoride) proteins containing B-HLH-ZIP motif were purified over heparin-Sepharose column (Amersham Biosciences). Proteins bound to the resin were eluted using potassium phosphate buffer (pH 7.4) containing increasing concentrations of KCl (150, 300, and 1000 mM). Proteins with acidic extensions were loaded on to a hydroxylapatite column (Bio-Rad) and eluted with 250 mM sodium phosphate buffer (pH 7.4). All protein samples were subsequently purified using a reverse phase analytical HPLC system (Rainin) equipped with C18 hydrophobic column (Varian). The eluents used were: eluent A, degassed deionized water with 0.1% (v/v) trifluoroacetic acid; eluent B, 100% (v/v) degassed acetonitrile (containing 0.1% (v/v) trifluoroacetic acid with a linear gradient of 0 -100% of B over 45 min with a flow rate of 1 ml/min. UV absorbance was monitored at 220 nm. HPLC purified and lyophilized protein samples were dissolved in 12.5 mM phosphate buffer, and concentrations were measured by taking absorbance at 230 nm. The molar absorption coefficient value for each protein sample was calculated as described earlier (16).
Sedimentation Equilibrium-Apparent molecular weights were determined by sedimentation equilibrium experiments using a Beckman XL-A Optima Analytical Ultracentrifuge equipped with absorbance optics and a Beckman An-60Ti rotor. Samples were dialyzed for 12 h against standard CD buffer (12.5 mM potassium phosphate buffer (pH 7.4) containing 150 mM KCl, 0.25 mM EDTA, and 1 mM dithiothreitol) and loaded at three concentrations, 20, 40, and 60 M, that correspond to 0.1, 0.2, and 0.3 absorbance at 276 nm, into a six-hole centerpiece and spun at 25,000 rpm for 24 h at 25°C. Proteins were scanned after every 4 h. Equilibrium was assumed to be reached (typically after 12-16 h) when two consecutive scans become indistinguishable. Partial specific volumes for all proteins were calculated from their primary sequences and amino acids values given by Zamyatnin (17).
Circular Dichroism-Samples for thermal denaturation studies were prepared in standard CD buffer. Prior to the experiment all samples were heated to 75°C for 15 min and then cooled at room temperature. The dimer protein concentration was 2 M in the homodimerizing system and was 4 M in mixtures. Thermal denaturation experiments were carried out in Jasco J-720 spectropolarimeter. Samples were heated in a water-jacketed cuvette holder attached to a programmable temperature controller accessory, interfaced with the computer that also controls the spectropolarimeter. Heat-induced denaturation curves were obtained by heating the protein sample from 6 to 85°C at a rate of 1°C/min, and measuring the changes in ellipticity () at 222 nm. About 850 data points ( 222 , T) were obtained for each transition curve. All thermal denaturations were fully reversible allowing thermodynamic parameters to be calculated.
Thermodynamic Calculations-At any given temperature protein sample is a mixture of dimer and monomer and total ellipticity is given by the sum of their fractions, where M and Di represent ellipticity values for the unfolded monomer and folded dimer at any temperature. For obtaining thermodynamic data from thermal transition curves following two assumptions were made. First, M is constant at higher temperature for all proteins studied and assumed to be temperature independent, i.e. M ϭ const. ϭ D (intercept at 0°C). Second, Di varies linearly with temperature and is given by, where N is the intercept of Di at 0°C, and T inter is the temperature at which linear dependences of dimeric and monomeric protein species intersect. For a reversible bimolecular reaction (2Monomer N Dimer) equilibrium constant is given by Equation 3.
The above relation is expressed in terms of fraction monomer (f M ϭ [Monomer]/[Total protein]) and total protein concentration (C), where where T is temperature in K, T m is the melting temperature (f M ϭ 1/2), ⌬H m is enthalpy change at T m , and R is the gas constant. Note that the ⌬C p term (constant pressure heat-capacity change upon dimer formation) is not included in the above equation. Because of the high degree of interdependence of ⌬H m and ⌬C p we were unable to simultaneously solve for these two variables. As a result, ⌬C p was initially assumed to be zero. Analysis of each transition curve gave the values of T m and ⌬H m . Fig. 3 shows the plot of ⌬H m as a function of T m . A linear least-squares fitting of all data (⌬H m versus T m ) gave a straight line where the slope corresponds to ⌬C p . The calculated value of ⌬C p (Ϫ1.74 Ϯ 0.13 kcal mol Ϫ1 dimer Ϫ1 K Ϫ1 ) is in good agreement with that reported earlier for other B-HLH-ZIP domains (12). ⌬H m at corresponding T m and ⌬C p were used to calculate ⌬G Di , the free energy change upon dimer formation using the following form of the Gibbs-Helmholtz equation, where T is in K, T o ϭ 310.15 K, R is the gas constant, and C is the total molar monomer protein concentration. All such values of ⌬G Di thus obtained are given in Tables II and III. The error in the values of ⌬G Di was calculated by considering the individual errors in values of ⌬H m and ⌬C p (18). DNA-HPLC-purified oligonucleotides (28-mer) containing the SRE-1 consensus site (5Ј-GTCAGTCAGACACCCCACTATCGGTCAG-3Ј) were purchased from Sigma-Genosys and were dissolved in water. The concentration of single-stranded DNA was determined by measuring the absorption at 260 nm. Double-stranded DNA was formed by mixing equimolar concentration of complimentary single-stranded oligonucleotides, heating the mixture to 80°C for 5 min, chilling on ice for 5 min, and then incubating at room temperature for 15 min. For CD experiments involving DNA (2 M) and protein (2 M) and their equimolar mixture (2 M of each), samples were prepared as described for protein thermal denaturation experiments using the same buffer system. CD thermal denaturations were recorded at 245 nm for DNA sample and 222 nm for protein and DNA-protein samples.
EMSA-Methods have been described previously (19). For assays involving SREBP-1 and SREBP-2 proteins, the SRE-1 DNA binding sequence was used (described in the above paragraph). For the other B-HLH-ZIP proteins, the E-box sequence (5Ј-GTCAGTCAGGCCACGT-GAGATCGGTCAG-3Ј) was used. Protein samples in gel shift reaction buffer (12.5 mM potassium phosphate buffer (pH 7.4) with 150 mM KCl, 0.25 mM EDTA, 2.5 mM dithiothreitol, 2 mg/ml bovine serum albumin, and 2% glycerol) were incubated with the appropriate radioactive probe for 45 min at 37°C and resolved in native 7.5% polyacrylamide gel in low ionic buffer 0.25ϫ TBE buffer (25 mM Tris base, 25 mM boric acid, and 0.5 mM EDTA) at room temperature. For time-dependent displacement experiments involving different mutants of A-SREBP-1, DNAprotein complex were incubated with DNs at 37°C for 5, 15, and 45 min and immediately gel electrophoresed.
Cell Culture-NIH-3T3 cells were grown in Dulbecco's modified Eagle's medium with 10% calf serum, 100 g/ml penicillin, 100 g/ml streptomycin (Invitrogen) and cultured in 10% CO 2 at 37°C. Cells were seeded in 12-well NUNC plates a day before performing transient transfections using a calcium phosphate procedure. Cells were transfected with 2 g of total DNA per well in 12-well plates. 8 g of total DNA was mixed with 140 l of CaCl 2 /HEPES (0.25 M CaCl 2 , 0.25 M HEPES, pH 7.1) and added dropwise to 140 l of NaCl/HEPES/ Na 2 HPO 4 (0.25 M NaCl, 25 mM HEPES, 1.5 mM Na 2 HPO 4 , pH 7.1).
After 22-min incubation at room temperature, a precipitate appeared and one-fourth of the DNA mixture was added slowly to a well. This was repeated two times for a triple estimation. After 5-to 6-h incubation the medium was replaced. After an additional 24 h, the cells were washed in phosphate-buffered saline, lysis solution (Tropix) was added, and after 15-20 min of incubation at room temperature on a shaking bench, lysates were transferred to an Eppendorf tube and snap frozen until performing luciferase assays.

Amino Acid Sequence of Proteins-
Molecular Weights of Proteins-We have determined the apparent molecular weight for three samples using sedimentation equilibrium experiments (Fig. 2). The three samples are the B-HLH-ZIP domain of SREBP-1 that we named B-SREBP-1, the dominant negative of SREBP-1 described in this report termed A-SREBP-1 that will be discussed shortly, and the mixture of these two proteins. Fig. 2A shows the optical scan of B-SREBP-1 protein after equilibrium is reached. Simulated curves for monomer and trimer are shown as solid lines. Actual experimental data were fitted to a dimer model. The lower panel shows the residuals of the fit showing no systematic error suggesting the model fits the experimental data. Goodness of the fit suggests that B-SREBP-1 exists as a dimer under experimental conditions. Panels B and C present the data for A-SREBP-1 and the B-SREBP-1⅐A-SREBP-1 mixture. The measured molecular weights of B-SREBP-1, A-SREBP-1, and their mixture are given in Table I. Because the dimer is the only oligomeric state at 25°C we have assumed that thermal  6), were obtained for each protein studied. We have used circular dichroism (CD) spectroscopy at 222 nm to monitor the thermal stability of B-SREBP-1, the B-HLH-ZIP domain of SREBP-1. The similarity of the CD spectra at 6°C before and after heating of SREBP-1 (Fig. 4A) shows that the thermal unfolding is reversible. The spectrum at 6°C for SREBP-1 shows the typical minima at 222 and 208 nm that is observed for helical proteins and the 222 / 208 ratio Ն 1 suggests the presence of a coiled-coil structure (20). Fig. 4B presents the CD monitored thermal denaturation and renaturation of SREBP-1. The denaturation is reversible and well fit assuming a two-state denaturation profile with a T m of 49.2°C and a calculated ⌬G Di of Ϫ10.0 kcal mol Ϫ1 dimer Ϫ1 (Table II). Retaining an additional 15 amino acids at the C terminus does not change the ellipticity or thermal stability indicating that the shorter version of SREBP-1 defines the entire B-HLH-ZIP domain. B-SREBP-1 is more stable than the five other B-HLH-ZIP domains combinations we have examined (Table II). We were unable to express sufficient quantities of B-SREBP-2 for CD characterization.
DNA Binding Stabilizes SREBP-1- Fig. 5A presents the CD spectra at 6°C of 1) B-SREBP-1 protein, 2) 28-bp doublestranded DNA containing a SRE-1 site, and 3) their equimolar mixture. The DNA has a minimum around 245 nm, and the protein has minima at 222 and 208 nm. The mixture of B-SREBP-1 and DNA shows a marked increase in negative ellipticity at 222 and 208 nm that we suggest represents the nonhelical basic region becoming ␣-helical upon DNA binding as has been observed for other B-HLH-ZIP proteins (8,12). We can monitor the stability of the protein in the presence of DNA   at 222 nm, because the CD signal at 222 nm of DNA does not change upon heating. Upon DNA binding, the thermal stability of B-SREBP-1 increased from 49.2°C to 56.3°C (Fig. 5B). Thus the DNA induced ␣-helical structure of SREBP-1 correlates with increased stability.
A-SREBP-1 Heterodimerizes with Both SREBP Family Members but Not Other B-HLH-ZIP Proteins-We have replaced the basic region of SREBP-1 and SREBP-2 with the polyglutamic acid sequence previously described (12) to generate A-SREBP-1 and A-SREBP-2, respectively (Fig. 1). Fig. 6A shows the CDmonitored heat-induced denaturation of B-SREBP-1, the dominant negative A-SREBP-1, and their equimolar mixture. All denaturation curves were cooperative with well defined preand post-transition baselines and well fit according to Equation 5 that assumes a two-state transition (Table III). The A-SREBP-1 protein has less negative ellipticity at 222 nm (Ϫ16 millidegrees) compared with B-SREBP-1 (Ϫ20 millidegrees). Replacing the basic region with the acidic region increases the stability of A-SREBP-1 (T m ϭ 73°C) by 23.8°C compared with B-SREBP-1 (49.2°C). This has been observed for other B-HLH-ZIP proteins (12).
The mixture of B-SREBP-1 and A-SREBP-1 is more stable than the calculated sum of the individual proteins indicating that a heterodimer is formed. The B-SREBP-1⅐A-SREBP-1 heterodimer is Ϫ9.8 kcal mol Ϫ1 dimer Ϫ1 and Ϫ1.8 kcal mol Ϫ1 dimer Ϫ1 more stable than the B-SREBP-1 and A-SREBP-1 homodimers, respectively. The B-SREBP-1⅐A-SREBP-1 heterodimer does not show an increase in ellipticity ( 222 at 6°C) compared with the sum of the individual proteins indicating that the stabilizing interaction between the acidic extension and the basic region does not induce ␣-helical structure. The thermal denaturation curve of the B-SREBP-1⅐A-SREBP-1 heterodimer shows a single transition indicating that the interaction between the basic region and the acidic extension denatures cooperatively with HLH-ZIP domain.
To determine if both members of the SREBP family could heterodimerize, we produced a dominant negative that contains the HLH-ZIP domain from SREBP-2. Fig. 6B shows the heat-induced denaturation curves of B-SREBP-1, A-SREBP-2, and their equimolar mixture. A-SREBP-2 interacts with B-SREBP-1 and melts with a single transition. The stability of SREBP-1 with either A-SREBP-1or A-SREBP-2 is similar (Table III) suggesting that the SREBP-1⅐SREBP-2 heterodimerize with similar stability to the SREBP-1or SREBP-2 homodimer.
The dimerization specificity of A-SREBP-1 was addressed by examining the thermal stability of mixtures with other B-HLH-ZIP proteins. Fig. 6C shows heat-induced denaturation curves    of B-MAX, A-SREBP-1, and their 1:1 mixture. The shape of the mixture curve is similar to that of the sum line curve suggesting these two proteins do not interact. Similar results were obtained when A-SREBP-1 was mixed with either B-USF of B-MITF (data not shown) indicating that the dimerization domains of these three B-HLH-ZIP proteins is sufficiently different from SREBP to prevent interactions even if stabilized by the interaction of the acidic region from A-SREBP-1 with the basic region of these proteins.
Derivatives of A-SREBP-1 with Different Stabilities-An ultimate goal in this work is to generate dominant negative to SREBPs that function in vivo. The high T m of A-SREBP-1 suggested it may not function in vivo, because it is so stable as a homodimer that it may not dissociate into monomers before heterodimerizing with a SREBP-1. Thus, we generated three additional versions of A-SREBP-1 with lower thermal stability, A-SREBP-1 T69 (T m ϭ 69.2°C), A-SREBP-1 T61 (T m ϭ 60.7°C), and A-SREBP-1 T51 (T m ϭ 50.8°C). The dominant negatives with T m of 69.2 and 60.7 were obtained by mutating either one or two leucines in the acidic extension to glutamic acid. Mutating the two leucines to glutamic acid and an isoleucine in helix 1 to alanine lowered the T m to 50.8°C. In CD thermal denaturation experiments, all three DNs preferentially heterodimerized with SREBP-1 (Table III).

Chimeric Proteins Map SREBP Dimerization Specificity to
Both the HLH and Leucine Zipper-Chimeric proteins were produced that contained the HLH domain from SREBP-1 and the leucine zipper domain of either MAX, USF, or MITF (SR-MAX, SR-USF, and SR-MITF). The amino acid sequences of the MAX, USF, and MITF B-HLH-ZIP domains are shown in Fig.  7. Two versions of these three chimeric dimerization domains were produced. One set contained the SREBP-1 basic region (B-SR-MAX, B-SR-MAX, and B-SR-USF), and one set contained the acidic extension (A-SR-MAX, A-SR-USF, and A-SR-MITF). The thermal denaturation of these six proteins is well fit by a two-state unfolding model suggesting that the HLH domain from SREBP-1 dimerizes cooperatively with the three leucine zipper domains to stabilize the HLH-ZIP domain (Table  IV). B-SR-MAX has a T m ϭ 50.0, which is similar to B-SREBP-1 (T m ϭ 49.2), indicating that the leucine zipper domain from SREBP-1 and MAX contribute similarly to stability. In contrast, the T m for B-SR-USF and B-SR-MITF is lower at 42.3°C and 40.9°C, respectively, indicating that these leucine zipper domains are less stable. The thermal stability of A-SR-MAX, B-SR-USF, and A-SR-MITF were all more stable than the corresponding B-HLH-ZIP version. Surprisingly, the increase in stability that occurred by replacing the basic region with the acidic region was different for these three chimeric proteins even though the only difference was the leucine zipper.
Heterodimers were identified that could occur by mixing one EMSA-We also used the electrophoretic mobility assay (EMSA) to examine the dimerization specificity of B-SREBP-1 and B-SREBP-2. We examined which of five A-HLH-ZIPs could inhibit DNA binding of B-SREBP-1 (Fig. 9A) or SREBP-2 ( Fig.  9B) bound to a radiolabeled DNA probe containing the SRE-1 site (5Ј-ACACCCCACT-3Ј). The inhibition of DNA binding we interpret represents the formation of a heterodimer between a B-SREBP family member and the dominant negative, which is unable to bind DNA. DNA binding of B-SREBP-1 is abolished in the presence of an equimolar concentration of either A-SREBP-1 or A-SREBP-2 as expected from the CD thermal denaturations that indicate that the B-SREBP-1⅐A-SREBP-1 heterodimer is more stable than SREBP-1 bound to DNA. The three SREBP-1 dominant negatives with lower homodimer stabilities also abolished B-SREBP-1 DNA binding (data not shown). The ability of A-SREBP-2 to inhibit the DNA binding of B-SREBP-1 demonstrates that the SREBP-1 and SREBP-2 HLH-ZIP domains can heterodimerize. 3 molar eq of A-MAX, A-USF, and A-MITF did not interfere with the binding of B-SREBP-1 (Fig. 9A) or SREBP-2 (Fig. 9B) to DNA, suggesting these dimerization domains do not interact.

Homodimer
Heterodimer with B-SREBP-1(S) a Letters S and L in parentheses denote short and long version of the same protein (see text for details). b Same as in Table II.   In addition, we used EMSA to analyze the dimerization properties of the chimeric proteins containing the SREBP-1 HLH domain and the leucine zipper from MAX, USF, and MITF. Fig. 11 examines the ability of A-SR-MAX at 1, 10, or 100 molar eq to inhibit the DNA binding of B-SR-MAX, B-SREBP-1, and B-MAX. At one molar equivalent A-SR-MAX abolishes the DNA binding of B-SR-MAX. However, even at 100 molar eq, A-SR-MAX does not affect the DNA binding of B-SREBP-1 or B-MAX indicating that these proteins do not dimerize. Fig. 12A shows that A-SR-USF inhibits the DNA binding of B-SR-USF but not B-SREBP-1 or B-USF. Fig. 12B shows that A-SR-MITF inhibits the DNA binding of B-SR-MITF but not B-SREBP-1 or B-MITF. Collectively, these results demonstrate that both the HLH and leucine zipper domains mediate dimerization specificity.
Different SREBP-1 Dominant Negatives Have Different Kinetics of Displacing SREBP-1 from DNA-We studied the timedependent displacement of homodimer B-SREBP-1 bound to DNA by the four SREBP-1 dominant negatives. Unlike in other gel shift experiments, the mixtures of B-SREBP-1 and A-SREBP-1 were not heated to 75°C, instead B-SREBP-1 was allowed to bind to the radioactive probe at 37°C and then A-SREBP-1 was added after 30 min. These reaction mixtures were incubated for 5 min (Fig. 13 (panel A), 15 min (panel B), and 45 min (panel C) and ran on 7.5% native polyacrylamide gel for 90 min. After 5 min DNA binding of B-SREBP-1 is perturbed by 1 M of A-SREBP-1 T51 (the mutant with the lowest thermal stability). After 15 min even 100 nM of A-SREBP-1 T51 is effective in displacing the double-stranded B-SREBP-1 from its DNA binding site. Furthermore, after 15 min of incubation mutant A-SREBP-1 T61 also interacts with B-SREBP-1, although less effectively (panel B). After 45 min of incubation time DNA binding of B-SREBP-1 is abrogated by 10 molar eq of A-SREBP-1 and all its mutants.
A-SREBP-1 Activity in Transient Transfections-To investigate the ability of A-SREBP to inhibit the transactivation of either SREBP-1 or SREBP-2 in a cellular system, we used a transient transfection assay. For that purpose the N-terminal parts (i.e. the constitutive active forms) of the human SREBP-1a and SREBP-2 were cloned into the pcDNA3.1 expression vector and used for transfection of the NIH3T3 cell line, which express very low endogenous levels of SREBPs. SREBP dimers often need auxiliary transcription factors such as Sp1 and NF-Y to bind efficiently to their response elements. Thus, all natural SREBP-responsive promoters contain a core SRE where SREBP binds and adjacent binding sites for one or more of these auxiliary factors. To determine the efficiency of the A-SREBPs to antagonize SREBP function it is therefore important to use a reporter construct that reflects the conditions at natural promoters. We have recently identified a functional SRE and adjacent NF-Y-and Sp-1-responsive elements in the proximal promoter of the rat acyl-CoA-binding protein (ACBP) gene. 2 These three DNA elements mediate the synergistic action of SREBPs, NFY, and Sp1, respectively, and are contained in the rACBP(Ϫ182/ϩ1)-luc reporter construct.
The activity of the rACBP(Ϫ182/ϩ1)-luc reporter construct is low in NIH-3T3 cells in the absence of SREBP co-transfection. However, co-transfection with either SREBP-1a or SREBP-2 efficiently activates the promoter (Figs. 14 and 15). This activation is significantly decreased in a dose-dependent manner by co-transfection with A-SREBP-1 and A-SREBP-2. However, co-transfection with A-USF, a dominant negative that inhibits USF action (14) exerted only a slight and non-dose-dependent inhibition of SREBP transactivation.
To further investigate the specificity of the A-SREBPs, we co-transfected NIH-3T3 cells with a USF expression plasmid, a 2 M. Boysen and S. Mandrup, unpublished data. ZIP transcription factors. Thus, our A-SREBP constructs can be used to efficiently and specifically abolish SREBP activity in vivo. DISCUSSION We have designed a dominant negative (DN), termed A-SREBP-1, that inhibits the function of SREBP-1 or SREBP-2, two related B-HLH-ZIP transcription factors. A-SREBP-1 inhibits the in vitro DNA binding of both SREBP family members without interfering with the activity of three B-HLH-ZIP pro-teins MAX, USF, and MITF. Transient co-transfection studies demonstrate that A-SREBP-1 can inhibit the transactivation of SREBP-1 and SREBP-2 but not USF. A-SREBP-1 consists of the dimerization domain of B-SREBP-1 and a polyglutamic acid sequence that replaces the basic region. Chimeric proteins were produced containing the SREBP-1 HLH domain and the leucine zipper from MAX, USF, or MITF. Dimerization data obtained for these chimeric proteins indicate that both the HLH and leucine zipper domains are critical for the dimerization specificity of SREBP-1.

FIG. 14.
A-SREBPs antagonize SREBP-1 activity in transient transfections. NIH-3T3 cells were transiently transfected with 50 ng of ACBP(Ϫ182/ ϩ1), 50 ng of pcDNA-SREBP-1 and where indicated increasing molar eq of either A-SREBP-1, A-SREBP-2, or A-USF (equivalent molar amounts, 2, 4, and 8 molar excess). pCMV-␤-galactosidase was co-transfected as control. All transfections were performed as triplicate standard calcium phosphate transfections in 12-well plates with a total of 2 g of DNA/plate. Empty expression vectors were added to compensate for promoter load. Luciferase values have been normalized to ␤-galactosidase values. Standard deviations are indicated. The results are representative of three independent experiments.

FIG. 15.
A-SREBPs antagonize SREBP-2 activity in transient transfections. NIH-3T3 cells were transiently transfected with 50 ng of ACBP(Ϫ182/ ϩ1), 50 ng of pcDNA-SREBP-2 and where indicated increasing molar eq of either A-SREBP-1, A-SREBP-2, or A-USF (equivalent molar amounts, 2, 4, and 8 molar excess). pCMV-␤-galactosidase was co-transfected as control. All transfections were performed as triplicate standard calcium phosphate transfections in 12-well plates with a total of 2 g of DNA/plate. Empty expression vectors were added to compensate for promoter load. Luciferase values have been normalized to ␤-galactosidase values. Standard deviations are indicated. The results are representative of three independent experiments.
Over 125 B-HLH proteins have been identified in the human genome and have been grouped into 44 orthologous families (21). At least seven of these families contain a leucine zipper domain immediately C terminus of the HLH dimerization domain (AP4, MYC, MAX, MAD, USF, MITF, and SREBP). Both domains denature cooperatively indicating that they form an extended dimerization interface. Both B-HLH and B-HLH-ZIP proteins are known to homodimerize and/or heterodimerize (22). The amino acid determinants that regulate dimerization specificity of the leucine zipper have begun to be unraveled (23)(24)(25)(26)(27). The contribution of leucine zippers found in B-HLH-ZIP proteins has been studied. For example, the normally heterodimerizing B-HLH-ZIP Myc protein can be made to homodimerize by changing two repulsive arginines in the e and a position of the leucine zipper to glutamine and asparagines, respectively, indicating that dimerization specificity of this transcription factor lies in the leucine zipper region (28). In contrast, amino acids that contribute to the dimerization specificity of the HLH domain has not been as extensively examined (29).
We have used a dominant negative strategy to produce a protein (A-SREBP-1) that preferentially heterodimerizes with the B-HLH-ZIP domain of SREBP-1, being 9.8 kcal mol Ϫ1 dimer Ϫ1 more stable than the B-SREBP-1 homodimers. This preferential heterodimerization between B-HLH-ZIP and A-HLH-ZIP proteins has allowed us to use both circular dichroism spectroscopy and EMSA to examine the dimerization specificity of the HLH-ZIP domain. Mixtures of four B-HLH-ZIP and five A-HLH-ZIP proteins indicate that the SREBP-1 dimerizes with both SREBP proteins but not with the three other B-HLH-ZIP proteins. To further map dimerization specificity, we produced chimeric B-HLH-ZIP and A-HLH-ZIP proteins containing the SREBP-1 HLH domain and the leucine zipper of MAX, USF, or MITF. These proteins denatured cooperatively indicating that these heterologous dimerization domains could cooperate to produce a more stable dimerization interface. Both the CD-monitored thermal denaturations and EMSA indicate that mixtures of B-HLH-ZIP and A-HLH-ZIP proteins only heterodimerized if the HLH domains and the leucine zipper domains were identical. These two domains could be derived from different B-HLH-ZIP proteins. When heterodimerization does not occur, it indicates that repulsion between the HLH-ZIP domains can not override the 9.8 kcal mol Ϫ1 dimer Ϫ1 attraction between the acidic extension and the basic region.
Throughout this analysis we have made two assumptions, first that the proteins denature as a two-state process, and second that the different structural elements contribute to stability independently of each other. It is tempting to suggest that the absence of an interaction between a B-HLH-ZIP and A-HLH-ZIP protein indicates that repulsion between the dimerization domains is greater than the attraction between the basic region and acidic extension. However, leucine zipper folding is not truly two-state (30) and may require nucleation (31). Disruption of nucleation might prevent dimerization between the basic region and acidic extension. This prevents us from concluding that the repulsion between dimerization domains that do not form in greater than 9.8 kcal mol Ϫ1 dimer Ϫ1 .
We examined the amino acid sequence of the HLH and leucine zipper dimerization domains in an attempt to rationalize these dimerization specificity observed in this report. There are 18-amino acid differences between SREBP-1 and SREBP-2 HLH-ZIP domains (see Fig. 7). Because these two domains have similar homodimer and heterodimer stabilities, we conclude these amino acids are not critical for either stability of dimerization specificity. The leucine zippers of SREBP-1 and MAX have many similarities but do not interact. In both cases, the leucine zipper is in register with helix 2 of the HLH (32). Both contain an attractive g-eЈ salt bridge in the third heptad that would facilitate both homodimerization and heterodimerization. The hydrophobic interfaces are also similar with a notable exception. The leucine zipper of B-SREBP-1 has an N in the second "a" position, whereas B-MAX has an I. Using a heterodimerizing B-ZIP protein system, the interaction between N and I was found to be energetically unfavorable with a calculated coupling energy of ϩ4.3 kcal mol Ϫ1 dimer Ϫ1 (27). This single amino acid difference may partially explain why SREBP-1 and MAX do not heterodimerize (27,33).
Heterodimerization between the SREBP-1 and the USF or MITF leucine zippers is also not observed. However, unlike the MAX leucine zippers, these two leucine zippers are out of phase with helix 2. To preserve the heptad repeat of leucines, we need to postulate a three-amino acid stutter in the coiled-coil structure (Fig. 7). Having the SREBP-1 leucine zipper out of register with either the USF or MITF leucine zipper prevents dimerization.
The parallel, left-handed, four-helix bundle structure of the HLH dimer is more complex than the leucine zipper, and our knowledge of the dimerization specificity is much more limited. Three putative electrostatic interactions between charged amino acids in helix 1 and helix 2 on the surface of the HLH dimer have been implicated in mediating dimerization specificity between the B-HLH proteins MyoD and E12 (29). The SREBP-1 and MITF proteins have the same potential electrostatic interactions, and the x-ray structure of the B-HLH-ZIP domain of SREBP-1 bound to DNA identifies one postulated electrostatic interaction between an aspartic acid in helix 1 and an arginine in helix 2 that are 3.18 Å apart (Fig. 7) (10). USF and MAX do not have any of these electrostatic interactions indicating these interactions may be important for the observed dimerization specificity. However, MITF also has these potential salt bridges observed in SREBP-1 indicating these interactions can not be used to rationalize the inhibition of dimerization between these two proteins. Thus, although we find that HLH domains possess structural determinants that regulate dimerization specificity, the rules that govern dimerization specificity are inadequate to explain our results.
Previously, two dominant negatives to SREBP have been described. The DNA binding specificity of SREBP-1c was altered by replacing in the DNA binding region a tyrosine with an arginine that is more typically found in B-HLH-ZIP proteins (6). This mutant SREBP-1 homodimer no longer bound to the SRE DNA sequence but instead bound to carbohydrate response element E-Box. A heterodimer with endogenous SREBP-1 bound to both E-box and SRE site. In other studies, a dominant negative was produced that heterodimerized with SREBPs but did not bind DNA. This was accomplished by replacing the tyrosine already mentioned, with alanine (34,35). The dominant negative described in this report also acts by preventing SREBPs from binding to DNA but is more potent due to the extra stability imparted to the heterodimer by the interacting basic region and acidic extension.
A-SREBP-1 in vivo can inhibit both SREBP-1 and SREBP-2, related proteins with potentially overlapping function. This potential for redundant function has complicated interpretation of many genetic experiments in mammals. However, combining A-SREBP-1, which inhibits both SREBP family members, with small interference RNA technology, which inhibits only individual members of the family, should allow an evaluation of any redundant function in this family of transcriptional regulators. and Asha Acharya for comments on the manuscript. Plasmids pPac-hSREBP-1a and pPac-hSREBP-2 were kindly provided by T. Osborne.