Advertisement

A Eukaryotic SWI2/SNF2 Domain, an Exquisite Detector of Double-stranded to Single-stranded DNA Transition Elements*

  • Rohini Muthuswami
    Footnotes
    Affiliations
    Department of Biochemistry and Molecular Genetics, School of Medicine, University of Virginia, Charlottesville, Virginia 22908
    Search for articles by this author
  • Patrick A. Truman
    Footnotes
    Affiliations
    Department of Biochemistry and Molecular Genetics, School of Medicine, University of Virginia, Charlottesville, Virginia 22908
    Search for articles by this author
  • Larry D. Mesner
    Affiliations
    Department of Biochemistry and Molecular Genetics, School of Medicine, University of Virginia, Charlottesville, Virginia 22908
    Search for articles by this author
  • Joel W. Hockensmith
    Correspondence
    To whom correspondence should be addressed: Dept. of Biochemistry and Molecular Genetics, P.O. Box 800733, School of Medicine, University of Virginia, Charlottesville, VA 22908. Tel.: 804-924-1230; Fax: 804-924-5069
    Affiliations
    Department of Biochemistry and Molecular Genetics, School of Medicine, University of Virginia, Charlottesville, Virginia 22908
    Search for articles by this author
  • Author Footnotes
    * This work was supported in part by United States Public Health Service Research Grant GM-43569 (to J. W. H.) and the University of Virginia School of Medicine. The Hewlett-Packard 8452A diode array spectrophotometer used in this work was purchased with funds provided by National Science Foundation Grant BIR-9216996.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.The nucleotide sequence(s) reported in this paper has been submitted to the GenBank™/EMBL Data Bank with accession number(s) AF173643.
    ‡ Present address: Dept. of Biochemistry and Molecular Genetics, Campus Box B-121, University of Colorado Health Sciences Center, 4200 E. Ninth Ave., Denver, CO 80262.
    § Predoctoral Trainee supported by United States Institutional Research Service Award GM-08136.
Open AccessPublished:March 17, 2000DOI:https://doi.org/10.1074/jbc.275.11.7648
      Many members of the SWI2/SNF2 family of adenosine triphosphatases participate in the assembly/disassembly of multiprotein complexes involved in the DNA metabolic processes of transcription, recombination, repair, and chromatin remodeling. The DNA molecule serves as an essential effector or catalyst for most of the members of this particular class of proteins, and the structure of the DNA may be more important than the nucleotide sequence. Inspection of the DNA structure at sites where multiprotein complexes are assembled/disassembled for these various DNA metabolic processes reveals the presence of a common element: a double-stranded to single-stranded transition region. We now show that this DNA element is crucial for the ATP hydrolytic function of an SWI2/SNF2 family member: DNA-dependent ATPase A. We further demonstrate that a domain containing the seven helicase-related motifs that are common to the SWI2/SNF2 family of proteins mediates the interaction with the DNA, yielding specific DNA structural recognition. This study forms a primary step toward understanding the physico-biochemical nature of the interaction between a particular class of DNA-dependent ATPase and their DNA effectors. Furthermore, this study provides a foundation for development of mechanisms to specifically target this class of DNA-dependent ATPases.
      ADAAD
      active DNA-dependent ATPase A domain
      bp
      base pair
      DNA recognition and the ensuing ATP hydrolysis by DNA-dependent ATPases are common denominators of a number of different macromolecular assembly/disassembly processes that are required for DNA metabolic events such as chromatin remodeling, replication, repair, recombination, and transcription. There are a variety of types of known nucleic acid-modifying enzymes that use the energy released by ATP hydrolysis to produce changes in a nucleic acid substrate (i.e. helicases, nucleases, topoisomerases, ligases, and recombinases). However, there are a few nucleic acid-dependent ATPases where the nucleic acid is not obviously modified and consequently does not appear to be a substrate for the enzyme. In these cases, the nucleic acid is generally regarded as an effector of the enzymatic activity. DNA-dependent ATPase A hydrolyzes ATP only in the presence of a DNA effector, without apparent modification of the effector. Originally isolated from calf thymus tissue, ATPase A is a 105-kDa protein that undergoes proteolysis to yield two smaller polypeptides with estimated molecular masses of 68 and 83 kDa (
      • Mesner L.D.
      • Truman P.A.
      • Hockensmith J.W.
      ). All three polypeptides possess ATPase activity that is manifested only in the presence of DNA. Initial studies with the 68-kDa polypeptide indicated that the protein is maximally active in the presence of DNA effectors possessing primer-template junctions (i.e. double-stranded to single-stranded transition). This effector preference is reminiscent of the gp44·62 and gp45 protein complex from T4 bacteriophage (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ,
      • Jarvis T.C.
      • Paul L.S.
      • Hockensmith J.W.
      • von Hippel P.H.
      ). The gp44·62 protein complex hydrolyzes ATP in the presence of primer-template DNA, with the energy release being coupled to the loading of a gp45 sliding clamp complex onto DNA, thereby increasing the processivity of DNA synthesis (
      • Jarvis T.C.
      • Paul L.S.
      • von Hippel P.H.
      ). The recognition of common epitopes by monoclonal antibodies (
      • Mesner L.D.
      • Sutherland W.M.
      • Hockensmith J.W.
      ) for gp44·62 and DNA-dependent ATPase A coupled with the use of similar DNA effectors led us to the suggestion that DNA-dependent ATPase A acts similarly to gp44·62,i.e. as a protein assembly factor.
      Assembly and the corresponding disassembly of multiprotein complexes are known to be essential for the DNA metabolic processes of transcription, recombination, repair, and chromatin remodeling, and the SWI2/SNF2 family of DNA-dependent ATPases is known to be involved in these processes (
      • Carlson M.
      • Laurent B.C.
      ). Although all members of this family contain seven helicase-related motifs, none of them have been reported to possess helicase activity, and not all members of this family require DNA for ATP hydrolysis (e.g. Mot1) (
      • Petukhova G.
      • Stratton S.
      • Sung P.
      ,
      • Auble D.T.
      • Hansen K.E.
      • Mueller C.G.
      • Lane W.S.
      • Thorner J.
      • Hahn S.
      ,
      • Pazin M.J.
      • Kadonaga J.T.
      ). Thus, the role of these helicase-related motifs, apart from being required for ATP hydrolysis, has not been established. Based on our observations with DNA-dependent ATPase A, we hypothesize that the “helicase” motifs function as DNA structure recognition elements (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ). Heretofore, data from studies of SWI2/SNF2 family members have not addressed two of the key components of this hypothesis: (i) the types of DNA secondary structures that effect ATP hydrolysis and/or function by these enzymes and (ii) the structural features of the proteins composing this family of ATPases that are responsible for binding to DNA in a structure-selective manner.
      We demonstrate here that DNA-dependent ATPase A belongs to the SWI2/SNF2 family of proteins and that a polypeptide domain encompassing the seven helicase-related motifs contains all the necessary regions for binding to DNA and hydrolysis of ATP. Furthermore, the domain is sufficient for recognition of specific DNA elements, and these DNA elements are necessary and sufficient for effecting ATP hydrolysis by DNA-dependent ATPase A. Our data support the proposal that the amino acid motifs common to the SWI2/SNF2 family of proteins are not helicase motifs, but are motifs that permit recognition of specific DNA elements. Similar to the structures generated by helicases, these elements are composed of double-stranded to single-stranded transition regions and as such are found in the DNA metabolic processes that include SWI2/SNF2 family members as participants. We suggest that one common theme for the SWI2/SNF2 family of proteins is specific recognition of DNA architecture.

      EXPERIMENTAL PROCEDURES

      Chemicals

      Unless noted otherwise, chemicals were purchased from Fisher, Sigma, or Mallinckrodt Chemical Works. Restriction enzymes were purchased from Promega and New England Biolabs Inc. Automated Edman degradation peptide sequencing was conducted by the University of Virginia Biomolecular Research Facility, which also synthesized all DNA constructs using spectrophotometric and gas chromatographic analyses for quality control.

      Enzymes

      Calf thymus DNA-dependent ATPase A (105-kDa species) was purified as described (
      • Mesner L.D.
      • Truman P.A.
      • Hockensmith J.W.
      ). The activeDNA-dependent ATPase Adomain (ADAAD)1was purified from Escherichia coli cells following expression from the plasmid pRM102. Briefly, cDNA was generated from calf thymus mRNA using oligo(dT) primers. Oligonucleotide primers 502 (5′-GCTCGAATTCTTTATAGAGGAGAGGTAAAGCT-3′) and 503 (5′-TATACCATGGCAGGGACCCCGATGCACAGA-3′) were used in a polymerase chain reaction to amplify the region of the cDNA encoding amino acids 215–941 of ATPase A. The amplified polymerase chain reaction product was purified, restricted with EcoRI and NcoI, and ligated into the pET-24d(+) vector (Novagen) cut with these same two enzymes. The ligated product was transformed into JM109 cells, and transformants were selected in the presence of 51 μmkanamycin. A plasmid (pRM102) was isolated from this clone and transformed into BL21(DE3) cells (Novagen) for expression.
      E. coli BL21(DE3) cells containing the pRM102 plasmid were grown at 25 °C in the absence of kanamycin
      Muthuswami, R., Mesner, L. D., Wang, D., Hill, D. A., Imbalzano, A. N., and Hockensmith, J. W. (2000)Biochemistry, in press.
      to an absorbance of 1.0 at 600 nm and induced with 0.5 mmisopropyl-β-d-thiogalactopyranoside. The cells were grown at 25 °C for 2 h, harvested by centrifugation at 4225 ×g in a GS3 rotor for 15 min, weighed, and stored at −80 °C. For purification of ADAAD, 10 g of cells were thawed; resuspended in 200 ml of buffer containing 20 mm Tris-Cl (pH 7.5), 5% (v/v) glycerol, 5 mm EDTA, 5 mmEGTA, 50 mm NaCl, 5 mm β-mercaptoethanol, and 0.5 mm phenylmethylsulfonyl fluoride; homogenized with a Dounce homogenizer; and lysed by passing twice through a French press at 1500 p.s.i. The cell lysate was centrifuged at 12,000 ×g for 30 min at 4 °C, and 2 m NaCl was added to the supernatant. The resulting solution was centrifuged at 191,230 × g for 2 h at 4 °C, and the supernatant was loaded onto a P-60 gel filtration column (6.5 × 20 cm). The P-60 column was eluted with buffer containing 20 mm Tris-Cl (pH 7.5), 5 mm EDTA, 5 mm EGTA, 20% (v/v) glycerol, and 0.5 mmphenylmethylsulfonyl fluoride, and the resulting protein was loaded onto an immunoaffinity column, which was prepared and eluted according to a previously devised protocol (
      • Mesner L.D.
      • Truman P.A.
      • Hockensmith J.W.
      ). Enzymatic ATP hydrolysis was monitored by a colorimetric assay (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ). Protein content was estimated using the Bradford assay (
      • Bradford M.M.
      ).

      Calculation of the Apparent Binding Constants

      The binding of DNA to protein was monitored by the ability of the DNA molecule to effect ATP hydrolysis by the enzyme. To calculate the apparent binding constant for ATPase A/DNA or ADAAD/DNA interactions, we used a model that described a simple one-substrate, one-product reaction catalyzed by the enzyme. The interaction between the DNA-dependent ATPase and DNA leading to ATP hydrolysis was described as follows: enzyme·ATP + DNA → enzyme·ATP·DNA → enzyme·ADP + product + DNA, where the amount of phosphate released is a measure of the product formed. We described the rate of reaction (V) as follows:v i =V DNA[DNA]/(K DNA + [DNA]), where V DNA is the maximal velocity for the reaction described above. In this particular case, we have definedV max as V DNA andK DNA as the apparent dissociation constant since DNA is not a substrate, but is an effector.

      RESULTS

      Identification of a Novel SWI2/SNF2 Family Member

      Biochemical characterization of bovine 105-kDa DNA-dependent ATPase A (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ) coupled with amino acid analysis of the immunoaffinity-purified 83-kDa ATPase A polypeptide and its cyanogen bromide-generated cleavage products led to elucidation of the amino acid sequence for this protein (Fig. 1). The amino acid sequences of seven different peptides were obtained, and the sequences were used to generate oligonucleotide primers. The nucleotide sequence was then deduced using these primers in a polymerase chain reaction with template cDNA generated from two parallel systems: a bovine aortic endothelial cell cDNA library and a cDNA prepared from oligo(dT)-primed calf thymus mRNA. The amino acid sequence was generated by translation of the nucleotide sequence (Fig. 1).
      Figure thumbnail gr1
      Figure 1Amino acid sequence of calf thymus DNA-dependent ATPase A. The amino acids shown inboldface were determined by sequencing of peptide fragments generated by cyanogen bromide digestion of ATPase A or by N-terminal analysis of the 83-kDa polypeptide. Underlined amino acids represent the seven helicase-related motifs. Amino acid 215 constitutes the N terminus of ADAAD.
      Sequence comparison using BLASTP (NCBI) revealed that, despite having similar effector preference and common epitopes, ATPase A does not have any significant similarity to gp44·62. Instead, the sequence possesses seven helicase-related motifs in the carboxyl terminus (from amino acids 451 to 941), which are characteristic of the SWI2/SNF2 family of proteins (
      • Eisen J.A.
      • Sweder K.S.
      • Hanawalt P.C.
      ). This region of 490 amino acids shows up to 36% identity to members of the SWI2/SNF2 family of proteins (
      • Pearson W.R.
      • Lipman D.J.
      ). The sequence yielding the highest degree of similarity to ATPase A has been tentatively identified as encoding a helicase from Caenorhabditis elegans (NCBI accession number U41534; BLASTP value: 4 × 10−88). However, the C. elegans amino acid sequence showed homology only from amino acids 292 to 929 (33% identity) of ATPase A. Of the various proteins that have been at least partially purified by biochemical techniques, ATPase A showed 24% identity over 461 amino acids of hSNF2H (NCBI accession numberAB010882; BLASTP value: 6 × 10−27) and 30% identity over 289 amino acids of yeast Mot1 sequence (NCBI accession numberP32333; BLASTP value: 2 × 10−17) (
      • Davis J.L.
      • Kunisawa R.
      • Thorner J.
      ). Although BLASTP demonstrated significant similarity of ATPase A to otherSaccharomyces cerevisiae proteins such as the Snf2 and Sth1 proteins (
      • Laurent B.C.
      • Yang X.
      • Carlson M.
      ,
      • Laurent B.C.
      • Treich I.
      • Carlson M.
      ), no S. cerevisiae sequence showed significant similarity over the entire length of the protein. Despite having both the amino acid sequence and confirming nucleotide sequence, BLASTP analysis of the amino terminus of ATPase A (amino acids 1–450) did not show significant similarity to any protein sequence.
      Northern blot analysis of mRNA isolated from calf thymus tissue confirmed the presence of a 2.8-kilobase transcript corresponding to the 105-kDa protein (data not shown). Southern blot analysis of genomic DNA isolated from human, bovine, and murine tissues confirmed the presence of a gene encoding DNA-dependent ATPase A (data not shown). The Southern blot yielded the same results regardless of whether a probe was used for the 5′- or 3′-end of the gene and also supported the conclusion that multiple copies of this gene are present in these three organisms. Neither yeast RNA nor genomic DNA hybridized to any probe, confirming the absence of a sequence sharing full-length similarity in the yeast proteome data base and leading us to suggest that DNA-dependent ATPase A is unique to higher eukaryotes.

      Native DNA-dependent ATPase A Recognizes Double-stranded to Single-stranded Transition Regions

      For DNA-dependent ATPase A, generation of free phosphate via ATP hydrolysis is manifested only in the presence of DNA. In particular, single-stranded and double-stranded heteropolymeric DNAs, either closed circular or linear, act as effectors of ATP hydrolysis with no indication of base sequence specificity (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ). Optimal effectors have the capability of possessing secondary structures such as double-stranded to single-stranded transitions. The analysis of the interaction between SWI2/SNF2 family members and DNA effectors has been limited in scope, and although sequence specificity for binding has been suggested, no DNA consensus sequence has been reported (
      • Quinn J.
      • Fyrberg A.M.
      • Ganster R.W.
      • Schmidt M.C.
      • Peterson C.L.
      ). We hypothesized that DNA-dependent ATPase A recognizes specific DNA structures and although the formation of these structures is sequence-dependent, there is no a prioriassumption of specific base recognition. To delineate the specific structures required to effect ATP hydrolysis, we chose a reductionist approach and employed a set of synthetic oligonucleotides to define a minimal effector (Table I).
      Table IOligonucleotide effectors and predicted structures
      OligonucleotideSequence (5′ to 3′)Possible structurePredicted structureΔG
      502GCTCGAATTCTTATAGAGGAGAGGTAAAGCTSingle-stranded0
      1027AGCTTTACCTCTCCTCTATAAGAATTCGAGCSingle-stranded0
      1026GCTCGAATTCTTATTATAGAGAGGTAAAGCTSingle-stranded0
      650GCGCAATTGCGCDouble-strandedGCGCAATTGCGC23.3
      CGCGTTAACGCG
      1009GACTCGAGTCGACTTTTTTTTTTGGGGGGGGGG5′-Recessed termini5′-GACTCGAGTCGACTTTTTTTTTTGGGGGGGGGG10.8
      ———-CAGCTGAGCTCAG-5′
      995CGCGCGTACGCACGTACGACGCGCACACGTGTGTG5′-Recessed termini5′-CGCGCGTACGCACGTACGACGCGCACACGTGTGTG12
      —ACGCATGCGCGC-5′
      993CGCGCGTACGCATACGTTGTATGCACACGTGTGTG5′-Recessed termini5′-CGCGCGTACGCATACGTTGTATGCACACGTGTGTG12
      ——–CATGCGCGC-5′
      1008CCCCCCCCCCCCCCCTCGATGTCGACTCGAGTC3′-Recessed terminiCCCCCCCCCCCCCCCTCGATGTCGACTCGAGTC-3′10.8
      3′-CTGAGCTCAGCTGTA——–
      994CACACACGTGTGCATACAACGTATGCGTACGCGCG3′-Recessed terminiCACACACGTGTGCATACAACGTATGCGTACGCGCG-3′12
      3′-GCGCGCATGC——-
      1027 + 502AT-rich DNAAGCTTTACCTCTCCTCTATAAGAATTCGAGC46.7
      TCGAAATGGAGAGGAGATATTCTTAAGCTCG
      1008 + 100910-bp mismatchCCCCCCCCCCCCCCCTCGATGTCGACTCGAGTC41.9
      GGGGGGGGGGTTTTTTTTTTCAGCTGAGCTCAG
      1027 + 10264-bp mismatchAGCTTTACCTCTCCTCTATAAGAATTCGAGC41.4
      TCGAAATGGAGAGATATTATTCTTAAGCTCG
      950GCGCAATTGCGCTCGACGATTTTTTAGCGCAATTGCGCStem-loopCGACG18.2
      5′-GCGCAATTGCGCT     A
      3′-CGCGTTAACGCGA     T
      TTTTT
      The structures were predicted using the Mfold, Plotfold, and FoldRNA programs contained within GCG software. These programs also calculate the free energy for the formation of the helix. As these programs predict the structure of RNA and not DNA, we also tried to predict the possible structures manually and calculated the free energy of destabilization by the method of Breslauer et al. (
      • Breslauer K.J.
      • Frank R.
      • Blocker H.
      • Marky L.A.
      ). The final structure we predicted for each sequence was the one that possessed the highest free energy of destabilization. The ΔG of destabilization of the duplex (kcal/mol of interaction) was calculated according to the method of Breslaueret al. (
      • Breslauer K.J.
      • Frank R.
      • Blocker H.
      • Marky L.A.
      ) using ΔG total = −(Δg i + Δg sym) + ΣxΔg x. The calculation assumes that the free energy of the double helix is a sum of its nearest neighbor interactions (ΣxΔg x). Δg i represents the helix initiation energy: duplexes containing G · C base pairs are assigned 5 kcal, whereas helices composed exclusively of A · T base pairs are assigned a value of 6 kcal. Δg sym represents the free energy involved in forming a duplex from a self-complementary sequence: a value of 0.4 kcal is assigned in such cases, whereas a value of zero is assigned when duplexes are formed from two complementary sequences.
      The ability of oligonucleotides to interact with DNA-dependent ATPase A was measured as a function of ATP hydrolysis. We initially found that oligonucleotides <40 bases in length varied widely in their ability to effect ATP hydrolysis by DNA-dependent ATPase A, varying from no induced hydrolysis up to hydrolytic levels greater than we had previously seen with any effector. Possible secondary structures of the oligonucleotide effectors were predicted using the FoldRNA, Mfold, and Plotfold programs contained within the GCG software package (
      • Zuker M.
      ,
      • Zuker M.
      ). As these programs are designed to predict the secondary structure for an RNA molecule, we also predicted possible structures for each of our DNA constructs by analyzing the sequence manually. Subsequently, we calculated the free energy of destabilization for possible duplexes using the method of Breslauer et al. (
      • Breslauer K.J.
      • Frank R.
      • Blocker H.
      • Marky L.A.
      ). The most probable structure for a given DNA construct was selected from the possible distribution of structures as the one that was the most energetically favorable (Table I).
      Neither a single-stranded DNA molecule (oligonucleotides 502, 1027, and 1026) nor a double-stranded, blunt-ended DNA (oligonucleotide 650) was capable of effecting ATP hydrolysis by DNA-dependent ATPase A (Table II). However, DNA molecules with both single-stranded and duplex character were effectors of ATP hydrolysis by DNA-dependent ATPase A. In particular, a stem-loop DNA molecule (oligonucleotide 950) was the optimal effector of DNA-dependent ATPase A. We also found that duplex DNA molecules that possess the character necessary to yield a double-stranded to single-stranded transition region, such as mismatches of 4 and 10 base pairs or AT-rich sequence, were effectors of DNA-dependent ATPase A.
      Table IIEffector utilization and kinetics
      DescriptionOligonucleotideATPase A ActivityADAAD KDNAADAAD V DNA
      nmol Pireleased/hnmnmol Pi released/min
      Single-stranded1026None NM
      NM, not measurable; ND, not determined.
      NM
      502NoneNMNM
      1027NoneNMNM
      Blunt-ended duplex650NoneNMNM
      5′-Recessed ends, no stem loops1009NoneNMNM
      995NoneNMNM
      993NoneNMNM
      3′-Recessed ends, no stem loops1008174.160 ± 171.7 ± 0.1
      994240.6NDND
      AT-rich duplex1027 + 502210.349 ± 114.5 ± 0.5
      10-Base pair mismatch duplex1008 + 1009280.215.8 ± 4.03.9 ± 0.5
      4-Base pair mismatch duplex1027 + 1026284.34.1 ± 1.53.8 ± 0.5
      Stem-loop950320.61.9 ± 0.42.4 ± 0.1
      DNA-dependent ATPase A hydrolytic activity was measured in the presence of 10 nm effector at 37 °C for 1 h. Duplex DNA molecules were generated by mixing and heating equimolar amounts of two single-stranded DNAs to 70 °C for 3 min and slow cooling the mixture to room temperature prior to addition to the enzymatic reaction. The amount of phosphate released was measured using a colorimetric assay (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ). For K DNA andV DNA calculations, ADAAD hydrolytic activity was measured as a function of DNA concentration for 1 h at 37 °C.K DNA and V DNA were calculated as described under “Experimental Procedures.”
      a NM, not measurable; ND, not determined.
      To establish the affinity of DNA-dependent ATPase A for these double-stranded to single-stranded transition regions, we used a kinetic phosphate release assay to estimate the dissociation constant for the interaction between the enzyme and its optimal effector, stem-loop DNA. We assumed that the protein/DNA interaction is in rapid equilibrium and that the formation of this complex is not the rate-limiting step. Under these conditions, the dissociation constant was calculated to be 0.5 nm, indicating that DNA-dependent ATPase A has a high affinity for this type of DNA structure. The yeast SWI/SNF multiprotein complex has previously been reported to have a high affinity for some promoter sequences and DNA-binding properties that permit recognition of structured DNA (four-way junctions), but the protein components responsible for binding the DNA have not been identified (
      • Quinn J.
      • Fyrberg A.M.
      • Ganster R.W.
      • Schmidt M.C.
      • Peterson C.L.
      ). Our results lead to the proposal that the SNF2 component of the SWI/SNF complex may be solely responsible for the high affinity DNA binding.
      With the apperception that DNA-dependent ATPase A recognizes DNA in a highly structure-dependent fashion that is not dependent on other protein cofactors and with a desire to elucidate the kinetic mechanism of DNA-dependent ATPase A, we sought to establish whether the helicase-related motifs constitute a functional domain. To characterize the protein/DNA interaction, we overexpressed and purified the molecular motor from ATPase A using a heterologous expression system.

      The Helicase-related Domain of DNA-dependent ATPase A Is Sufficient for ATPase Activity

      During our biochemical characterization of ATPase A, we identified shorter proteolytic forms (Fig. 1), which retain DNA-dependent ATP hydrolysis comparable to the full-length parent polypeptide; thus, the amino terminus of the enzyme is dispensable for ATP hydrolytic function. The cDNA sequence encoding one of these proteolytic products (82 kDa) was cloned into a pET-24d(+) vector. Since the 82-kDa polypeptide comprised both the ATP- and DNA-binding domains of ATPase A as well as the seven helicase-related motifs, we termed it the activeDNA-dependent ATPase Adomain (ADAAD) (Fig. 1, amino acids 215–941). Western blot analysis using monoclonal antibodies raised against ATPase A (
      • Mesner L.D.
      • Sutherland W.M.
      • Hockensmith J.W.
      ) showed the presence of the ADAAD protein following induction of expression with isopropyl-β-d-thiogalactopyranoside.
      The specific activity of ADAAD (16 μmol of ATP hydrolyzed per min/mg), overexpressed and purified in the absence of kanamycin,2 was nearly identical to that of the 83-kDa ATPase A purified from calf thymus tissue (
      • Mesner L.D.
      • Sutherland W.M.
      • Hockensmith J.W.
      ). Also similar to the proteolytic products, ADAAD exhibited a stability of >80% remaining activity after 20 days at room temperature. We conclude that ADAAD constitutes a correctly folded domain composed of the helicase-related motifs and that this domain is sufficient for effector recognition and subsequent ATP hydrolysis. This region of DNA-dependent ATPase A can thus be considered as the “motor” domain of the SWI2/SNF2 family (
      • Pazin M.J.
      • Kadonaga J.T.
      ). Using ATPase assays, the relative effector recognition properties of the expressed ADAAD were established as being essentially the same as those of the native protein; thus, we employed ADAAD to expand our observations regarding the essential components of DNA effector structure. Furthermore, the relative abundance of ADAAD (∼1 mg of purified ADAAD/6 g of cells) has allowed us to establish dissociation constants in addition to relative levels of activity (Table II).

      DNA-dependent ATPase A and ADAAD Discriminate between 3′- and 5′-Recessed Termini

      DNA-dependent ATPase A and ADAAD are able to recognize DNA constructs possessing double-stranded to single-stranded transition elements, including primer-template junctions, stem-loops, and mismatch regions. From earlier experiments using micrococcal nuclease- and pancreatic nuclease-treated DNAs, we hypothesized that these enzymes may be able to differentiate between a 5′-hydroxyl and a 3′-hydroxyl terminus (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ).
      To test this hypothesis, we prepared DNA partial duplexes, each containing three elements: 1) a duplex region, 2) a single-stranded region, and 3) recessed termini. One set of molecules contained recessed 5′-hydroxyl termini (oligonucleotides 1009, 995, and 993), and the other set of DNA molecules possessed recessed 3′-hydroxyl termini (oligonucleotides 1008 and 994). We found that the DNA constructs containing recessed 5′-hydroxyl termini were not effectors of ATP hydrolysis, but those containing recessed 3′-hydroxyl termini were effectors of ATP hydrolysis (Table II). We conclude that DNA-dependent ATP hydrolysis responds differentially to 3′- and 5′-recessed termini. ATP hydrolysis is preferentially effected by DNA molecules with a double-stranded to single-stranded transition and a 3′-hydroxyl terminus in close proximity to the duplex region.

      DNA-dependent ATPase A and ADAAD Hydrolyze ATP in the Presence of AT-rich DNA

      Anticipating that DNA-dependent ATPase A might be able to transiently distort DNA structure, we prepared double-stranded, blunt-ended oligonucleotides that possessed a region with high AT content (oligonucleotides 1027 and 502). Although neither of the single-stranded oligonucleotides was able to effect ATP hydrolysis by itself, we found that DNA-dependent ATPase A was able to recognize the duplex DNA effector with resultant ATP hydrolysis (TableII). This result leads us to suggest that DNA-dependent ATPase A can melt into the AT-rich regions of a DNA molecule or at least trap local denaturation of AT-rich regions. We suggest that the resulting protein·DNA complex results in a stabilized double-stranded to single-stranded transition element that permits proper enzymatic conformation for effecting ATP hydrolysis.

      Correlation of the Free Energy of Destabilization of a DNA Molecule with Its Effector Function

      Hybridization of single-stranded DNAs to form a duplex region or destabilization of double-stranded DNA to form a single-stranded region may both be modeled by the use of thermodynamic parameters. To understand the interaction between an effector DNA molecule and DNA-dependent ATPase A, we attempted to correlate the ability of an oligonucleotide to function as an effector with its free energy of duplex destabilization. From a thermodynamic perspective, a DNA duplex can be considered as the sum of its nearest-neighbor interactions (
      • Breslauer K.J.
      • Frank R.
      • Blocker H.
      • Marky L.A.
      ). In any Watson-Crick DNA molecule, 10 different nearest-neighbor interactions are possible: AA/TT, AT/TA, TA/AT, CA/GT, GT/CA, CT/GA, GA/CT, CG/GC, GC/CG, and GG/CC. Thus, the stability of a DNA duplex can be predicted from its primary sequence if the relative stability and the temperature-dependent behavior of each DNA nearest-neighbor interaction are known. Breslauer et al. (
      • Breslauer K.J.
      • Frank R.
      • Blocker H.
      • Marky L.A.
      ) have characterized each of the possible nearest-neighbor interactions thermodynamically, and we have used these published values to calculate the free energy of destabilization of the predicted duplex structures for our DNA constructs (Table I).
      The DNA molecules can be divided into two groups based on their structure and their ability to yield a double-stranded to single-stranded transition in the DNA. The first group consists of molecules that anneal to form primer-template junctions. The second group consists of duplex double-stranded, blunt-ended DNA molecules containing internal regions either single-stranded due to mismatches or easily denatured. Although the number of oligonucleotides presented in Table II is small, all the DNA molecules support the principle that a stable double-stranded to single-stranded transition element present in the effector DNA results in better effector function.
      For the first group of DNA constructs, we found that as the free energy of destabilization increased, the ability of the molecule to function as an effector increased (Table II, compare oligonucleotides 1008 and 994). We ascribe this behavior to the fact that as the free energy of destabilization increases, the DNA primer-template junction becomes more stable, leading to an increase in double-stranded to single-stranded transition character. This leads, in turn, to an increase in the ability of the DNA molecule to function as an effector of DNA-dependent ATPase A. Thus, we hypothesize that for single-stranded DNA molecules forming primer-template junctions, the amount of ATP hydrolyzed by DNA-dependent ATPase A increases as the free energy of destabilization for the duplex region increases.
      In the second group of molecules, the DNA molecule is primarily a double-stranded, blunt-ended duplex. Since DNA-dependent ATPase A recognizes only the double-stranded to single-stranded transition elements, a fully duplexed DNA molecule can become an optimal effector only if destabilized. Destabilization can be introduced into a double-stranded, blunt-ended DNA molecule in two ways: (i) by introducing mismatches (oligonucleotides 1008 and 1009 or oligonucleotides 1027 and 1026) and (ii) by increasing the AT content of DNA (oligonucleotides 502 and 1027). An increase in destabilization implies that the lower free energy will lead to DNA molecules that have larger proportions of double-stranded to single-stranded transitions and that can then function as effectors of DNA-dependent ATPase A. The results show that as the free energy of destabilization of a double-stranded, blunt-ended DNA molecule decreases, the amount of ATP hydrolyzed by DNA-dependent ATPase A increases (TablesI and II). Double-stranded to single-stranded transition elements may exist in the form of stem-loops, internal bubbles due to mismatches, or AT-rich regions, and we conclude that DNA-dependent ATPase A and ADAAD can recognize all these types of structures as effectors of ATP hydrolysis.

      The Apparent Dissociation Constant for the Interaction between DNA and the Protein Is Dependent upon the Structure of DNA

      Fig.2 shows that theV DNA and K DNA for ADAAD are both dependent upon the structure of the DNA effector. Similar to the parent DNA-dependent ATPase A effector specificity, the apparent dissociation constant for the interaction between ADAAD and the DNA effector possessing a stem-loop structure is much lower than that for the interaction between ADAAD and a primer-template junction (3′-recessed end) (Table II). We propose that the difference in these dissociation constants can be explained on the basis of the free energy of destabilization. The free energy of destabilization for the stem-loop molecule is greater than that for the primer-template junction; thus, the double-stranded to single-stranded transition region present in the stem-loop DNA is more stable, leading it to function as a much more efficient effector than a primer-template junction.
      Figure thumbnail gr2
      Figure 2Effect of different DNA structures on the activity of ADAAD. The ATPase activity of ADAAD was measured in the presence of stem-loop DNA (●), 4-bp mismatch DNA (○), 10-bp mismatch DNA (▪), AT-rich DNA (■), or primer-template junction DNA (▴) under the reaction conditions described under “Experimental Procedures.”
      Surprisingly, the apparent dissociation constant for the interaction between ADAAD and a DNA molecule possessing a 4-bp mismatch was 4-fold lower than that for the interaction between ADAAD and a DNA molecule possessing a 10-bp mismatch. Examination of the sequence of the 4-bp mismatch DNA molecule revealed that the DNA contained AT-rich DNA. Based on AT-rich DNA acting as an effector of ADAAD, we hypothesized that ADAAD can melt into the DNA molecule containing a 4-bp mismatch and thus generate a double-stranded to single-stranded transition element with the appropriate amount of single-stranded character to effect ATP hydrolysis. Furthermore, we hypothesized that the distance of the 3′-hydroxyl from the double-stranded to single-stranded transition plays an important role in effecting ATP hydrolysis and that this is an additional variable for the 4- and 10-bp mismatch DNAs.
      R. Muthuswami and J. W. Hockensmith, manuscript in preparation.

      ADAAD Melts into AT-rich Regions

      The hypothesis that ADAAD melts into AT-rich regions present in a DNA effector was tested by comparing the effector properties of these DNA effectors as a function of the NaCl concentration. We presumed that there is a minimal length of single-stranded DNA required in the double-stranded to single-stranded transition element in order to effect ATP hydrolysis. This minimal length could be either available as single-stranded DNA in the model effectors that we have tested or generated by sacrificing binding free energy of the DNA-dependent ATPase to provide some denaturation of the DNA duplex. These two extremes can be differentiated by the addition of salt, which should result in a more stable duplex and hence a poorer effector when denaturation is required to provide the optimal effector structure (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ).
      We found that the addition of 50 mm NaCl decreased ADAAD ATP hydrolytic activity by 50% when AT-rich duplex DNA was used as effector (Fig. 3 A), presumably by blocking melting of the DNA duplex by the enzyme. In contrast, the activity of ADAAD decreased by only 20% when primer-template junction or stem-loop DNA molecules having the requisite double-stranded to single-stranded transition element were used as effectors. Similarly, at 100 mm NaCl, the enzyme lost 80% of its activity when AT-rich DNA was used as effector, but only 50% of its activity in the presence of primer-template junction DNA. In contrast, at 100 mm NaCl, the stem-loop DNA lost a disproportionally larger share of its activity (80%) and behaved more like the AT-rich DNA than the primer-template junction. We note that the stem-loop DNA is closed by an AT base pair and that the enzyme could presumably melt this base pair. This ability to melt into this base pair may become too energetically costly as the base pair is stabilized at higher NaCl concentrations. At salt concentrations higher than 200 mm, all DNA-dependent ATPase activity was lost presumably as a result of a global effect such as disrupted protein/ATP or protein/DNA interactions.
      Figure thumbnail gr3
      Figure 3A, effect of salt on the effector properties of the stem-loop DNA, primer-template junction DNA, and AT-rich DNA. The effect of NaCl on the ATPase activity of ADAAD was measured in the presence of stem-loop DNA (●), AT-rich DNA (○), and 3′ recessed termini (■). 10 nm DNA was used in these experiments, and the reaction conditions described under “Experimental Procedures” were used. B, effect of salt on the effector properties of 4- and 10-bp mismatch DNAs. The effect of NaCl on the ATPase activity of ADAAD was measured in the presence of 4-bp mismatch DNA (●) and 10-bp mismatch DNA (○). 10 nmDNA was used in these experiments, and the reaction conditions described under “Experimental Procedures” were used.
      Fig. 3 B provides an additional salt titration that compares the effect of salt on the effector characteristics of a DNA molecule containing a 4-bp mismatch (oligonucleotides 1027 and 1026) and one containing a 10-bp (oligonucleotides 1008 and 1009) mismatch. Again, we observed that a DNA molecule requiring more denaturation (4-bp mismatch) to generate a double-stranded to single-stranded transition element loses its ability to function as an effector much more rapidly than a DNA molecule containing a larger single-stranded region (10-bp mismatch). Thus, at 50 mm NaCl, ATP hydrolysis decreased by 50% in the presence of the 4-bp mismatch, whereas ∼90% activity was retained in the presence of a DNA molecule containing a 10-bp mismatch. These results support our hypothesis that ADAAD can melt into AT-rich regions present in a DNA molecule, bind to the resulting double-stranded to single-stranded transition elements, and effect ATP hydrolysis.

      DISCUSSION

      Sequence analysis shows that DNA-dependent ATPase A belongs to the SWI2/SNF2 family of proteins, which possess seven helicase-related motifs at the carboxyl terminus. The members of the SWI2/SNF2 family have proposed roles in many DNA metabolic processes (i.e. transcription, repair, recombination, etc.) that may have a common fundamental structural element. The DNA at the site where these processes are occurring undergoes a transition from double-stranded to single-stranded character. Experimental results suggest that a yeast SWI/SNF family complex and DNA-dependent ATPase A recognize this type of structure specifically (
      • Mesner L.D.
      • Truman P.A.
      • Hockensmith J.W.
      ,
      • Quinn J.
      • Fyrberg A.M.
      • Ganster R.W.
      • Schmidt M.C.
      • Peterson C.L.
      ), as does gp44·62 (
      • Jarvis T.C.
      • Paul L.S.
      • Hockensmith J.W.
      • von Hippel P.H.
      ). We hypothesize that the recognition of double-stranded to single-stranded transition elements in DNA is a specific property of some molecular motor domains and propose that one common theme for the SWI2/SNF2 family of proteins is specific recognition of DNA architecture. We have demonstrated that the presence of a double-stranded to single-stranded transition region in a DNA molecule is a necessary and sufficient element for effecting ATP hydrolysis by DNA-dependent ATPase A. The double-stranded to single-stranded transition elements utilized in our studies differed significantly in their sequences, but not necessarily in their ability to effect ATP hydrolysis, thereby leading us to suggest that the interaction between DNA and DNA-dependent ATPase A is DNA sequence-independent, but structure-dependent.
      Since members of the SWI2/SNF2 family vary widely in size, we felt that development of a biochemical characterization of the conserved molecular motor domain would allow insight into the molecular mechanisms of action common to the entire family. Here we have narrowed the functional DNA-dependent ATPase activity to a polypeptide containing the helicase-related motifs. Thus, we have demonstrated that the DNA and ATP recognition elements reside within these helicase-related motifs, which constitute the core motor domain of the SWI2/SNF2 family of DNA-dependent ATPases. Whereas neither ADAAD nor ATPase A has helicase activity, both are capable of recognizing double-stranded to single-stranded transitions in DNA molecules. As none of the proteins belonging to the SWI2/SNF2 family have so far been shown to possess any helicase activity, it raises the important question of the role of these helicase-related motifs in these proteins and whether we should refer to these sequences as helicase motifs. We propose that the helicase-related domains of this molecular motor are essential for recognition of specific DNA structures and that these structures drive conformational changes and alter ATP hydrolysis.
      The interaction between DNA and DNA-dependent ATPase A leading to ATP hydrolysis is confined to a 727-amino acid region that possesses the seven helicase-related motifs. This region, also known as ADAAD, shows extensive amino acid sequence similarity to the SWI2/SNF2 family of proteins and thus can be considered as the core or motor domain responsible for the interaction between DNA and ATP. Regions outside this domain are not necessary for the structure-specific recognition of DNA elements, although we have no evidence to exclude them from additional interactions with DNA or other proteins. ADAAD was expressed based on natural cleavage sites without the aid of tags or fusions, yielding a protein with high specific activity (16 μmol of ATP hydrolyzed per min/mg). The expression and specific activity of other SWI2/SNF2 family members have been previously reported for (i) a fusion product of the C-terminal portion of yeast SNF2(∼0.02 μmol/min/mg) (
      • Laurent B.C.
      • Treich I.
      • Carlson M.
      ), (ii) a fusion product of the C-terminal portion of yeast MOT1 (∼0.33 μmol/min/mg) (
      • Auble D.T.
      • Hansen K.E.
      • Mueller C.G.
      • Lane W.S.
      • Thorner J.
      • Hahn S.
      ), and (iii) yeast Rad54 protein (∼13 μmol/min/mg) (
      • Petukhova G.
      • Stratton S.
      • Sung P.
      ). Although these latter SWI2/SNF2 family members were prepared as fusion proteins, native SWI/SNF complexes have reported specific activities that vary over a similarly broad range (
      • Cote J.
      • Quinn J.
      • Workman J.L.
      • Peterson C.L.
      ,
      • Cairns B.R.
      • Lorch Y.
      • Li Y.
      • Zhang M.
      • Lacomis L.
      • Erdjument-Bromage H.
      • Tempst P.
      • Du J.
      • Laurent B.
      • Kornberg R.D.
      ). The high specific activity coupled with the high yields has permitted us to initiate the pursuit of a kinetic mechanism for the recognition of DNA and the effecting of ATP hydrolysis.
      Furthermore, the domain known as ADAAD is sufficient for recognition of the DNA structural element, and additional protein components are not necessary for this recognition. As a supporting example, it has recently been reported that BAF57, a component of the mammalian SWI/SNF complex, possesses a high mobility group motif and therefore may be responsible for the binding of the SWI/SNF complex to four-way DNA junctions (
      • Wang W.
      • Chi T.
      • Xue Y.
      • Zhou S.
      • Kuo A.
      • Crabtree G.R.
      ). Mutations in the high mobility group domain of BAF57 did not prevent the SWI/SNF complex from binding to DNA (
      • Wang W.
      • Chi T.
      • Xue Y.
      • Zhou S.
      • Kuo A.
      • Crabtree G.R.
      ), which is consistent with our evidence that ADAAD and, by inference, SWI2/SNF2 family members of the SWI·SNF complex have the ability to recognize specific DNA structural elements. Thus, it appears that the different proteins constituting SWI/SNF multiprotein complexes might possess independent abilities to recognize, bind, and thereby distinguish between similar DNA-binding motifs, i.e. the SWI2/SNF2 family protein component of the SWI/SNF complex might be responsible for binding to double-stranded to single-stranded transition elements in preparation for various DNA metabolic activities ranging from DNA repair to transcription, whereas proteins containing the high mobility group box add a degree of specificity necessary for the recognition of specific subsets of double-stranded to single-stranded transitions such as a four-way junctions.
      Our studies can also be used to provide an explanation for earlier reports showing the ability of the SWI2/SNF2 family members to effect ATP hydrolysis in the presence of double-stranded or single-stranded phage DNA (
      • Petukhova G.
      • Stratton S.
      • Sung P.
      ,
      • Cairns B.R.
      • Kim Y.J.
      • Sayre M.H.
      • Laurent B.C.
      • Kornberg R.D.
      ). Both of these types of DNA are large and heterogeneous in base sequence, and we have previously demonstrated that such DNAs can result in domains where double-stranded to single-stranded transition regions occur (
      • Hockensmith J.W.
      • Wahl A.F.
      • Kowalski S.
      • Bambara R.A.
      ). Thus, because secondary structures are difficult to eliminate from large heteropolymers and because the DNA-binding domain of the protein may be able to melt into duplex structures, such DNAs should be expected to be capable of effecting ATP hydrolysis by proteins belonging to the SWI2/SNF2 family.
      Similar to the effecting of DNA-dependent ATP hydrolysis, we have found that the apparent dissociation constant for the interaction between ADAAD and DNA was also dependent upon the structure of the DNA molecule. The examples provided under “Results” demonstrate that the free energy of helix destabilization can be directly linked to the apparent dissociation constant for ADAAD/DNA interaction. Thus, DNA-dependent ATP hydrolysis by ADAAD is a direct measure of the presence of a double-stranded to single-stranded transition element and also of the stability of that element.
      Our studies show that ADAAD, an SWI2/SNF2 family member, belongs to the class of DNA-binding proteins that interact with DNA in a DNA structure-specific fashion. Other proteins of this class include proteins containing the high mobility group box, DNA-dependent protein kinase, and the gp44·62 complex from T4 bacteriophage, all of which recognize DNA in a structure-specific fashion (
      • Jarvis T.C.
      • Paul L.S.
      • Hockensmith J.W.
      • von Hippel P.H.
      ,
      • Pohler J.R.
      • Norman D.G.
      • Bramham J.
      • Bianchi M.E.
      • Lilley D.M.
      ,
      • Grosschedl R.
      • Giese K.
      • Pagel J.
      ,
      • Morozov V.E.
      • Falzon M.
      • Anderson C.W.
      • Kuff E.L.
      ). Unlike these other proteins, ADAAD is a highly active DNA-dependent ATPase whose activity can be measured in a colorimetric assay. The ability of DNA-dependent ATPase A to recognize DNA in a structure-specific fashion leading to ATP hydrolysis is easily adapted for identifying the presence of secondary structures in short DNA oligonucleotides. Although not all double-stranded to single-stranded DNA elements are equally effective catalysts, the high level of ADAAD activity yields a broad range of discrimination for detection of these elements. Through empirical observations, we have found that measurement of ADAAD activity yields an exquisite detection of secondary structure in DNA molecules.
      The identification of the specific DNA structure necessary and sufficient for ATP hydrolysis by DNA-dependent ATPase A has set the stage for exploring a possible model for the interaction between the DNA effector and the protein. Furthermore, the DNA structure-specific recognition and concomitant ATP hydrolysis by DNA-dependent ATPase A provide an exquisite opportunity for discovery and synthesis of specific inhibitors of the SWI2/SNF2 family of ATPases. Using the findings reported here, we have explored this avenue and have discovered a family of inhibitors that specifically inhibit members of the SWI2/SNF2 family of proteins.2

      Acknowledgments

      We thank Dennis Rinehart and Christian Anderton for technical assistance. We are also grateful to Jamie Kennedy and LeeAnn Swanegan for numerous comments and contributions.

      REFERENCES

        • Mesner L.D.
        • Truman P.A.
        • Hockensmith J.W.
        Biochemistry. 1993; 32: 7772-7778
        • Hockensmith J.W.
        • Wahl A.F.
        • Kowalski S.
        • Bambara R.A.
        Biochemistry. 1986; 25: 7812-7821
        • Jarvis T.C.
        • Paul L.S.
        • Hockensmith J.W.
        • von Hippel P.H.
        J. Biol. Chem. 1989; 264: 12717-12729
        • Jarvis T.C.
        • Paul L.S.
        • von Hippel P.H.
        J. Biol. Chem. 1989; 264: 12709-12716
        • Mesner L.D.
        • Sutherland W.M.
        • Hockensmith J.W.
        Biochemistry. 1991; 30: 11490-11494
        • Carlson M.
        • Laurent B.C.
        Curr. Opin. Cell Biol. 1994; 6: 396-402
        • Petukhova G.
        • Stratton S.
        • Sung P.
        Nature. 1998; 393: 91-94
        • Auble D.T.
        • Hansen K.E.
        • Mueller C.G.
        • Lane W.S.
        • Thorner J.
        • Hahn S.
        Genes Dev. 1994; 8: 1920-1934
        • Pazin M.J.
        • Kadonaga J.T.
        Cell. 1997; 88: 737-740
        • Bradford M.M.
        Anal. Biochem. 1976; 72: 248-254
        • Eisen J.A.
        • Sweder K.S.
        • Hanawalt P.C.
        Nucleic Acids Res. 1995; 23: 2715-2723
        • Pearson W.R.
        • Lipman D.J.
        Proc. Natl. Acad. Sci. U. S. A. 1988; 85: 2444-2448
        • Davis J.L.
        • Kunisawa R.
        • Thorner J.
        Mol. Cell. Biol. 1992; 12: 1879-1892
        • Laurent B.C.
        • Yang X.
        • Carlson M.
        Mol. Cell. Biol. 1992; 12: 1893-1902
        • Laurent B.C.
        • Treich I.
        • Carlson M.
        Genes Dev. 1993; 7: 583-591
        • Quinn J.
        • Fyrberg A.M.
        • Ganster R.W.
        • Schmidt M.C.
        • Peterson C.L.
        Nature. 1996; 379: 844-847
        • Zuker M.
        Methods Enzymol. 1989; 180: 262-288
        • Zuker M.
        Science. 1989; 244: 48-52
        • Breslauer K.J.
        • Frank R.
        • Blocker H.
        • Marky L.A.
        Proc. Natl. Acad. Sci. U. S. A. 1986; 83: 3746-3750
        • Cote J.
        • Quinn J.
        • Workman J.L.
        • Peterson C.L.
        Science. 1994; 265: 53-60
        • Cairns B.R.
        • Lorch Y.
        • Li Y.
        • Zhang M.
        • Lacomis L.
        • Erdjument-Bromage H.
        • Tempst P.
        • Du J.
        • Laurent B.
        • Kornberg R.D.
        Cell. 1996; 87: 1249-1260
        • Wang W.
        • Chi T.
        • Xue Y.
        • Zhou S.
        • Kuo A.
        • Crabtree G.R.
        Proc. Natl. Acad. Sci. U. S. A. 1998; 95: 492-498
        • Cairns B.R.
        • Kim Y.J.
        • Sayre M.H.
        • Laurent B.C.
        • Kornberg R.D.
        Proc. Natl. Acad. Sci. U. S. A. 1994; 91: 1950-1954
        • Pohler J.R.
        • Norman D.G.
        • Bramham J.
        • Bianchi M.E.
        • Lilley D.M.
        EMBO J. 1998; 17: 817-826
        • Grosschedl R.
        • Giese K.
        • Pagel J.
        Trends Genet. 1994; 10: 94-100
        • Morozov V.E.
        • Falzon M.
        • Anderson C.W.
        • Kuff E.L.
        J. Biol. Chem. 1994; 269: 16684-16688

      Linked Article