The missing nucleoside experiment: a new technique to study recognition of DNA by protein

We report a new technique for quickly determining which nucleosides in a DNA molecule are contacted by a sequence-specific DNA-binding protein. Our method is related to the recently reported "missing contact" experiment [Brunelle, A., & Schleif, R. F. (1987) Proc. Natl. Acad. Sci. U.S.A. 84, 6673-6679]. We treat the DNA molecule with the hydroxyl radical to randomly remove nucleosides. The ability of protein to bind to gapped DNA is assayed by gel mobility shift. Nucleosides important to protein binding are identified by sequencing gel electrophoresis. The missing nucleoside experiment can be used to scan a DNA molecule at single-nucleotide resolution in one experiment. The bacteriophage lambda repressor-OR1 and cro-OR1 complexes were analyzed to evaluate the method. For both proteins, the most important contacts are located in the protein monomer that binds to the consensus half of the operator. These contacts correspond well to those found by mutational studies, and in the cocrystal structure of the lambda repressor-operator. The missing nucleoside data show that the amino-terminal arms of lambda repressor make energetically important contacts with positions 7 and 8 and the central dyad base pair of the operator. The amino-terminal arm that makes the most extensive contacts to DNA appears to be the one that emanates from the repressor monomer that binds to the consensus half of the operator, in agreement with the cocrystal structure. The lambda cro protein does not have an amino-terminal arm, and the missing nucleoside experiment clearly shows a lack of contacts to DNA in the central region of the operator in this complex.     

Mutant λ Repressors with Increased Operator Affinities Reveal New, Specific Protein-DNA Contacts

The binding specificities of four mutant lambda cI repressor proteins with increased affinities for operator DNA were examined. Two mutant repressors (Glu34----Lys and Glu83----Lys) have the same specificity of binding as wild-type repressor, whereas two (Gly48----Ser and Gly48----Asn) have new binding specificities. The Gly48----Asn mutant repressor recognizes lambda operators with changes at base pair 3 with a different order of affinity than wild-type repressor, suggesting that the side chain of Asn48 makes additional specific DNA contacts at or near this base pair. When paired with a change that disrupts the specific interaction of the amino-terminal arm of lambda repressor with DNA (Lys4----Gln), one change that increases the affinity of repressor (Gly48----Ser) suppresses the binding defect of the Lys4----Gln repressor, resulting in a double mutant repressor with a new binding specificity different than that of both its parents and of wild type. These results lend strong support to the model of direct recognition of the lambda operator by lambda repressor proposed from the crystal structure of the repressor/operator complex.     

Crystal structure of an engrailed homeodomain-DNA complex at 2.8 A resolution: a framework for understanding homeodomain-DNA interactions

The crystal structure of a complex containing the engrailed homeodomain and a duplex DNA site has been determined at 2.8 A resolution and refined to a crystallographic R factor of 24.4%. In this complex, two separate regions of the 61 amino acid polypeptide contact a TAAT subsite. An N-terminal arm fits into the minor groove, and the side chains of Arg-3 and Arg-5 make contacts near the 5' end of this "core consensus" binding site. An alpha helix fits into the major groove, and the side chains of IIe-47 and Asn-51 contact base pairs near the 3' end of the TAAT site. This "recognition helix" is part of a structurally conserved helix-turn-helix unit, but these helices are longer than the corresponding helices in the lambda repressor, and the relationship between the helix-turn-helix unit and the DNA is significantly different.     

Molecular dynamics simulation reveals sequence-intrinsic and protein-induced geometrical features of the OL1 DNA operator

We have carried out molecular dynamics simulation of the lambda OL1 DNA operator on the free and the protein-bound forms. Our results lead us to conclude that the binding of the repressor actually makes the N-7 atom of Gua8' more solvent exposed, thereby enhancing its reactivity to chemical methylation. This increase in solvent accessibility surface area occurs simultaneously with the formation of hydrogen bonds between Lys-4 of the nonconsensus flexible N-terminal arm and Gua6' of the nonconsensus half-site operator DNA. Calculations of protein--DNA interaction energies reveal that among the residues of the arm, Lys-4 contributes the most favorably to the interaction energies. This result is consistent with mutagenesis studies that established that lysine at position 4 is absolutely required for tight binding. We find that the nonconsensus arm and the nonconsensus monomer interacts less favorably with DNA than do their respective counterparts of the consensus monomer. Moreover, the six-residue flexible arm accounts for at least half the total protein--DNA interactions energy. These results are in agreement with previous experimental studies. In accord with the diffuse electron density map observed in crystallographic studies of the nonconsensus flexible arm, we find that our model built for this region is more flexible and exhibits more conformations than its consensus counterpart. The simulation also reveals that DNA bending observed near the outer edge of the operator site is an intrinsic sequence-dependent property. By contrast, the DNA-bending features observed toward the center of the operator are induced by the protein. On the whole, stepwise protein-induced bending is more pronounced in the consensus half-site operator. We also find that the unusually large helical twist (49 degrees ) observed in the protein-bound form near the center of the operator results from the binding of the protein at a base step with some propensity for high twists.     

Ligand-induced conformational changes in the crystal structures of Pneumocystis carinii dihydrofolate reductase complexes with folate and NADP+

Structural data from two independent crystal forms (P212121 and P21) of the folate (FA) binary complex and from the ternary complex with the oxidized coenzyme, NADP+, and recombinant Pneumocystis carinii dihydrofolate reductase (pcDHFR) refined to an average of 2.15 A resolution, show the first evidence of ligand-induced conformational changes in the structure of pcDHFR. These data are also compared with the crystal structure of the ternary complex of methotrexate (MTX) with NADPH and pcDHFR in the monoclinic lattice with data to 2.5 A resolution. Comparison of the data for the FA binary complex of pcDHFR with those for the ternary structures reveals significant differences, with a >7 A movement of the loop region near residue 23 that results in a new "flap-open" position for the binary complex, and a "closed" position in the ternary complexes, similar to that reported for Escherichia coli (ec) DHFR complexes. In the orthorhombic lattice for the binary FA pcDHFR complex, there is also an unwinding of a short helical region near residue 47 that places hydrophobic residues Phe-46 and Phe-49 toward the outer surface, a conformation that is stabilized by intermolecular packing contacts. The pyrophosphate moiety of NADP+ in the ternary folate pcDHFR complexes shows significant differences in conformation compared with that observed in the MTX-NADPH-pcDHFR ternary complex. Additionally, comparison of the conformations among these four pcDHFR structures reveals evidence for subdomain movement that correlates with cofactor binding states. The larger binding site access in the new "flap-open" loop 23 conformation of the binary FA complex is consistent with the rapid release of cofactor from the product complex during catalysis as well as the more rapid release of substrate product from the binary complex as a result of the weaker contacts of the closed loop 23 conformation, compared to ecDHFR.     

Crystal structure of the DNA-binding domain of BldD, a central regulator of aerial mycelium formation in Streptomyces coelicolor A3(2)

BldD is a central regulator of the developmental process in Streptomyces coelicolor. The 1.8 angstroms resolution structure of the DNA-binding domain of BldD (BldDN) reveals that BldDN forms a compact globular domain composed of four helices (alpha1-alpha4) containing a helix-turn-helix motif (alpha2-alpha3) resembling that of the DNA-binding domain of lambda repressor. The BldDN/DNA complex model led us to design a series of mutants, which revealed the important role of alpha3 and the 'turn' region between alpha2 and alpha3 for DNA recognition. Based on the fact that BldD occupies two operator sites of bldN and whiG and shows significant disparity in the affinity toward the two operator sites when they are disconnected, we propose a model of cooperative binding, which means that the binding of one BldD dimer to the high affinity site facilitates that of the second BldD dimer to the low affinity site. In addition, structural and mutational investigation reveals that the Tyr62Cys mutation, found in the first-identified bldD mutant, can destabilize BldD structure by disrupting the hydrophobic core.     

Template-directed interference footprinting of cytosine contacts in a protein-DNA complex: potent interference by 5-aza-2'-deoxycytidine.

In the template-directed interference (TDI) footprinting method (Hayashibara & Verdine, 1990), analogs of the naturally occurring DNA bases are incorporated into DNA enzymatically and assayed for interference of sequence-specific binding by a protein. Here we extend this method to include analysis of contacts of amino acid residues to the major groove surface of cytosine residues (TDI-C footprinting). The base analog 5-aza-2'-deoxycytidine, in which the hydrophobic 5-CH of cytosine is replaced by a hydrophilic aza nitrogen, was incorporated into DNA via the corresponding 5'-triphosphate. The analog was found to base pair with guanine during polymerization, resulting in substitution of 2'-deoxycytidine residues. TDI-C footprints of the lambda repressor-OL1 operator complex revealed apparent contacts to the cytosines at operator positions 7 and 8. Inspection of the high-resolution X-ray crystal structure of the lambda-OL1 complex (Clarke et al., 1992; Beamer & Pabo, 1992) revealed that C8 makes a hydrogen binding contact with the Lys3; C7, on the other hand, makes a previously unnoticed hydrophobic contact with the alkane side chain of Lys3. In only the consensus operator half-site was cytosine interference observed, suggesting that the nonconsensus arm binds DNA very differently if at all. The N-terminal arm represents the archetypal case of a sequence-specific peptide-DNA complex characterized at high resolution; thus, the present studies suggest strategies for design and screening of DNA binding peptides. The finding that 5-aza-2'-deoxycytidine inhibits sequence-specific DNA binding proteins may suggest an alternative rationale for the biological activities of this and related azapyrimidine nucleosides.     

Prophage lambda induces terminal recombination in Escherichia coli by inhibiting chromosome dimer resolution. An orientation-dependent cis-effect lending support to bipolarization of the terminus

A prophage lambda inserted by homologous recombination near dif, the chromosome dimer resolution site of Escherichia coli, is excised at a frequency that depends on its orientation with respect to dif. In wild-type cells, terminal hyper- (TH) recombination is prophage specific and undetectable by a test involving deletion of chromosomal segments between repeats identical to those used for prophage insertion. TH recombination is, however, detected in both excision and deletion assays when Deltadif, xerC, or ftsK mutations inhibit dimer resolution: lack of specialized resolution apparently results in recombinogenic lesions near dif. We also observed that the presence near dif of the prophage, in the orientation causing TH recombination, inhibits dif resolution activity. By its recombinogenic effect, this inhibition explains the enhanced prophage excision in wild-type cells. The primary effect of the prophage is probably an alteration of the dimer resolution regional control, which requires that dif is flanked by suitably oriented (polarized) stretches of DNA. Our model postulates that the prophage inserted near dif in the deleterious orientation disturbs chromosome polarization on the side of the site where it is integrated, because lambda DNA, like the chromosome, is polarized by sequence elements. Candidate sequences are oligomers that display skewed distributions on each oriC-dif chromosome arm and on lambda DNA.     

Regulation of arginine biosynthesis in the psychropiezophilic bacterium Moritella profunda: in vivo repressibility and in vitro repressor-operator contact probing

We report the cloning of the arginine repressor gene from the psychropiezophilic Gram-negative bacterium Moritella profunda, the purification of its product (ArgR(Mp)), the identification of the operator in the bipolar argECBFGH(A) operon, in vivo repressibility studies, and an in vitro analysis of the repressor-operator interaction, including binding to mutant and heterologous arginine operators. The ArgR(Mp) subunit shows about 70% amino acid sequence identity with Escherichia coli ArgR (ArgR(Ec)). Binding of purified hexameric ArgR(Mp) to the control region of the divergent operon proved to be arginine-dependent, sequence-specific, and significantly more sensitive to heat than complex formation with ArgR(Ec). ArgR(Mp) binds E.coli arginine operators very efficiently, but hardly recognizes the operator from Bacillus stearothermophilus or Thermotoga maritima. ArgR(Mp) binds to a single site overlapping the -35 element of argC(P), but not argE(P). Therefore, the arrangement of promoter and operator sites in the bipolar argECBFGH(A) operon of M.profunda is very different from the organization of control elements in the bipolar argECBH operon of E.coli, where both promoters overlap the common operator and are equally repressible. We demonstrate that M.profunda argC(P) is about 44-fold repressible, whereas argE(P) is fully constitutive. A high-resolution contact map of the ArgR(Mp)-operator interaction was established by enzymatic and chemical footprinting, missing contact and base-specific premodification binding interference studies. The results indicate that the argC operator consists of two ARG box-like sequences (18bp imperfect palindromes) separated by 3bp. ArgR(Mp) binds to one face of the DNA helix and establishes contacts with two major groove segments and the intervening minor groove of each ARG box, whereas the minor groove segment facing the repressor at the center of the operator remains largely uncontacted. This pattern is reminiscent of complex formation with the repressors of E.coli and B.stearothermophilus, and suggests that each ARG box is contacted by two ArgR subunits belonging to opposite trimers. Moreover, the premodification interference patterns and mutant studies clearly indicate that the inner, center proximal halves of each ARG box in the M.profunda argC operator are more important for complex formation and repression than the outermost halves. A close inspection of sequence conservation and of single base-pair O(c)-type mutations indicate that the same conclusion can be generalized to E.coli operators.     

Comparison of the structures of operator DNA free and in complex with lambda repressor

The structures of operator DNA unbound and in complex with lambda repressor protein are compared. The conformation of the left 10 base pairs of a lambda right regulatory operator DNA sequence has been previously determined in solution using nuclear magnetic resonance techniques and the structure of a homologous left regulatory operator DNA bound to lambda repressor N-terminal domain had been previously solved using X-ray crystallography. The DNA adopts an overall linear B-form DNA both in the absence and presence of lambda repressor. Superimpositioning of the DNA structures reveals small differences between them that are due to the binding of protein and not to the different techniques used for their determination.     

Interaction of the bacteriophage P22 Arc repressor with operator DNA

Are repressor binds to a single, partially symmetric, 21 base-pair operator site that is centered between the -10 and -35 regions of the Pant promoter. Protection and interference experiments show that Arc makes contacts with the operator on one side of the DNA helix. Although Arc is a small protein (53 residues/subunit), it makes contacts that are farther from the center of the operator than those made by many larger repressors. These extended contacts include the phosphate groups at the ends of the 21 base-pair site. Under standard conditions (pH 7.5, 100 mM-KCl, 3 mM-MgCl2, 22 degrees C) half-maximal operator binding is observed at an Arc concentration of 2.5 X 10(-9) M and the protein-DNA complex is very stable (t1/2 approximately equal to 80 min).     

The Mu repressor-DNA complex contains an immobilized 'wing' within the minor groove

We have determined the solution structure of the complex between the 'winged-helix' enhancer binding domain of the Mu repressor protein and its cognate DNA site. The structure reveals an unusual use for the 'wing' which becomes immobilized upon DNA binding where it makes intermolecular hydrogen bond contacts deep within the minor groove. Although the wing is mobile in the absence of DNA, it partially negates the large entropic penalty associated with its burial by maintaining a small degree of structural order in the DNA-free state. Extensive contacts are also formed between the recognition helix and the DNA, which reads the major groove of a highly conserved region of the binding site through a single base-specific hydrogen bond and van der Waals contacts.     

Ribosome interactions of aminoacyl-tRNA and elongation factor Tu in the codon-recognition complex

The mRNA codon in the ribosomal A-site is recognized by aminoacyl-tRNA (aa-tRNA) in a ternary complex with elongation factor Tu (EF-Tu) and GTP. Here we report the 13 A resolution three-dimensional reconstruction determined by cryo-electron microscopy of the kirromycin-stalled codon-recognition complex. The structure of the ternary complex is distorted by binding of the tRNA anticodon arm in the decoding center. The aa-tRNA interacts with 16S rRNA, helix 69 of 23S rRNA and proteins S12 and L11, while the sarcin-ricin loop of 23S rRNA contacts domain 1 of EF-Tu near the nucleotide-binding pocket. These results provide a detailed snapshot view of an important functional state of the ribosome and suggest mechanisms of decoding and GTPase activation.     

Contacts between gamma delta resolvase and the gamma delta res site.

We have investigated the interaction between resolvase and the res site of the transposon gamma delta by methylation and ethylation interference experiments. We have examined the effect of these DNA modifications both on binding and resolution in vitro. Major groove methylations within a 9 bp sequence that borders each site inhibit binding of resolvase to that site. Ethylation of certain phosphates within, and adjacent to, this border sequence inhibits binding. Together, these interference points define a contact region, present at all three res sites. In vitro resolution is inhibited only by modifications within site I. Inhibition of resolution by methylation of adenines at the center of site I suggests that minor groove contacts near the crossover may be required for resolution activity.     

Crystallization of a complex of cro repressor with a 17 base-pair operator

Crystals of the lambda cro repressor complexed to a 17 base-pair synthetic binding site related to the OR3 operator have been obtained. The complex crystallizes in the hexagonal space group P6(2) (or P6(4)) with unit cell dimensions a = b = 154.8 A, c = 85.6 A. Preliminary photography reveals that the crystals are stable to X-rays and display measurable reflections to a resolution of about 3.7 A. The diffraction patterns suggest that the cro-DNA complexes are arranged in an open hexagonal network with the DNA fragments stacked end-to-end. The DNA is in the B-form but appears to be bent or curved into an approximate superhelix.     

Structure of the human NF-kappaB p52 homodimer-DNA complex at 2.1 A resolution.

The crystal structure of human NF-kappaB p52 in its specific complex with the natural kappaB DNA binding site MHC H-2 has been solved at 2.1 A resolution. Whereas the overall structure resembles that of the NF-kappaB p50-DNA complex, pronounced differences are observed within the 'insert region'. This sequence segment differs in length between different Rel proteins. Compared with NF-kappaB p50, the compact alpha-helical insert region element is rotated away from the core of the N-terminal domain, opening up a mainly polar cleft. The insert region presents potential interaction surfaces to other proteins. The high resolution of the structure reveals many water molecules which mediate interactions in the protein-DNA interface. Additional complexity in Rel protein-DNA interaction comes from an extended interfacial water cavity that connects residues at the edge of the dimer interface to the central DNA bases. The observed water network might acount for differences in binding specificity between NF-kappaB p52 and NF-kappaB p50 homodimers.