The Bloom’s Syndrome Helicase Unwinds G4 DNA*

BLM, the gene that is defective in Bloom’s syndrome, encodes a protein homologous to RecQ subfamily helicases that functions as a 3′-5′ DNA helicase in vitro. We now report that the BLM helicase can unwind G4 DNA. The BLM G4 DNA unwinding activity is ATP-dependent and requires a short 3′ region of single-stranded DNA. Strikingly, G4 DNA is a preferred substrate of the BLM helicase, as measured both by efficiency of unwinding and by competition. These results suggest that G4 DNA may be a natural substrate of BLM in vivo and that the failure to unwind G4 DNA may cause the genomic instability and increased frequency of sister chromatid exchange characteristic of Bloom’s syndrome.

Bloom's syndrome is a rare genetic disorder characterized by genomic instability and a very high incidence of cancer. Affected individuals display a pleiotropic array of symptoms, including prenatal and postnatal growth retardation, sunlight sensitivity, impaired fertility, and immunodeficiency. Patients develop cancers of many different types, including solid tumors and leukemias. Cytogenetic studies have documented an extraordinary level of genomic instability in cells from Bloom's syndrome patients. Chromatid breaks and chromosome aberrations are common, including a characteristically increased level of sister chromatid exchanges. A cytogenetic signature of the disease is a greatly elevated frequency of symmetrical, quadriradial chromosomes, which appear to be homologous chromosomes caught in the act of recombination (reviewed in Refs. [1][2][3]. The gene that is defective in Bloom's syndrome, BLM, encodes a 1417-amino acid polypeptide containing seven signature motifs common to DNA and RNA helicases (4). Helicases play important roles in every aspect of DNA metabolism, including DNA replication, transcription, and recombination (reviewed in Refs. 5 and 6). Eukaryotic and prokaryotic cells contain a variety of helicases, and many viruses encode helicases that facilitate their replication or recombination. The distinctive functions of these different proteins are in many cases not completely understood.
The BLM protein is most highly homologous to helicases of the DExH-containing subfamily, of which the prototypical member is Escherichia coli RecQ (4,7). RecQ, which is only 607 amino acids long, is a 3Ј-5Ј helicase that functions to prevent illegitimate recombination and may also be involved in initiation of recombination (8 -10). Human RecQL (11,12) is a 659amino acid protein of no known function. The other members of this helicase subfamily thus far identified, Saccharomyces cerevisiae Sgs1p (13,14), Schizosaccharomyces pombe Rqh1p (15), and the human Werner's syndrome protein, WRN (16), are similar in length to BLM and share sequence homology not only within but also in a limited region outside of the helicase motifs. Sgs1p is the only structural homolog of BLM and WRN in the S. cerevisiae genome, and it appears also to be a functional homolog since yeast cells deficient in Sgs1p are characterized by mitotic hyperrecombination (14,17). The S. pombe rqh1 ϩ gene similarly functions to prevent recombination, particularly mitotic recombination resulting from DNA damage (15). Deficiency in the WRN protein results in many of the manifestations of premature aging, and cells from Werner's syndrome patients are characterized by an elevated frequency of gene deletions and chromosomal abnormalities but not by an elevated level of sister chromatid exchanges (16).
Purified recombinant BLM has recently been shown to unwind duplex DNA in vitro, functioning as an ATP-dependent helicase with 3Ј-5Ј directionality (18). However, the exact function of BLM in DNA metabolism remains unclear. It is particularly puzzling that, despite the presence of numerous helicase activities in mammalian cells, lack of a single helicase can have such a profound effect on genomic stability.
We have now asked whether BLM might be active on DNA substrates other than canonical Watson-Crick duplexes. One DNA structure that forms readily in vitro and is remarkably stable is G-DNA. In G-DNA, interactions between runs of guanines are stabilized by Hoogsteen bonding (19). G-DNAs can exist in different forms, distinguished by strand stoichiometry, strand orientation, and glycosidic conformation (reviewed in Ref. 20). One of the best characterized is G4 DNA, which contains four DNA strands (Refs. 19 and 21; see Fig. 1A). G-DNA structures have not been directly observed in vivo, but the ease with which G-DNA can form in vitro suggests that G-DNA may exist at least transiently in a cell. Runs of three or more Gs occur throughout the mammalian genome; and G-rich motifs are particularly abundant in the rDNA gene clusters, the immunoglobulin heavy chain gene switch regions (S regions), and telomeric repeats. As G-DNA is stable once formed, it could interfere with normal nuclear processes if there were no enzymes able to disrupt it efficiently.
In this report we demonstrate that the BLM helicase efficiently unwinds G4 DNA substrates to yield single strands. The G4 DNA unwinding activity of the BLM helicase has a 3Ј-5Ј directionality, requires a short single-stranded tail, and is ATP-dependent. These are also properties of the BLM duplex DNA-unwinding reaction (18). Most strikingly, by comparing the unwinding activity of BLM on G4 DNA and on synthetic duplex substrates, we have found that G4 DNA is a preferred substrate of the BLM helicase. G4 DNA (or other similar struc-tures) may therefore be a natural target of BLM helicase in vivo, and failure to displace interacting G-rich DNA sequences may explain the increase in genetic exchange characteristic of Bloom's syndrome.
Helicase Assays-Recombinant Bloom's syndrome protein (rBLM), expressed in S. cerevisiae and purified as described previously (18), was incubated with DNA in 20-l reactions in buffer containing 50 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 5 mM ATP, 50 mM NaCl, and 100 g/ml bovine serum albumin at 37°C. Figure legends specify the amount of DNA and protein used in each experiment; molar amounts of G4 DNA refer to moles of 4-stranded structures (1 mol of G4 ϭ 4 mol of single strands). For analysis of reaction kinetics, DNA was pre-equilibrated in the reaction mix for 10 min at 37°C, and reactions were initiated by addition of protein. Purified RecBCD (Ref. 24; a generous gift of Drs. A. F. Taylor and G. R. Smith, Hutchinson Cancer Research Institute, Seattle, WA) was preincubated in reactions containing DNA but no ATP, and reactions were initiated by addition of ATP. Reactions were terminated by addition of 5 l of a solution containing 2.5% SDS and 2.5 mg/ml proteinase K (Boehringer Mannheim), to achieve final concentrations of 0.5% SDS and 0.5 mg/ml proteinase K and then incubated at 50°C for 15 min to ensure complete proteolysis; or by addition of EDTA and SDS to final concentrations of 10 mM and 0.1%, respectively. Samples were analyzed on 10% nondenaturing polyacrylamide gels in 50 mM Tris borate, 1 mM EDTA, pH 8.3.

RESULTS
The BLM Helicase Unwinds G4 DNA-Synthetic oligonucleotides representing G-rich sequences from the S regions or telomeric repeats were used to form G4 DNA. The characteristic protection in G4 DNA of the N7 by Hoogsteen bonding (Fig. 1A) was verified by footprinting with dimethyl sulfate (Fig. 1B). To ask if G4 DNA is a substrate for the BLM helicase, we first examined the ability of purified recombinant BLM (rBLM) to unwind G4 DNA structures prepared from the TP oligonucleotide. Radiolabeled G4 DNA prepared from TP (referred to as G4-TP) 1 was incubated with purified rBLM at 37°C at a 2:1 molar ratio of rBLM:G4-TP. Reactions were terminated at indicated times, and the samples were deproteinized and analyzed by nondenaturing gel electrophoresis, which resolves G4 DNAs and their single-stranded counterparts. The fraction of G4 DNA unwound increased with time, and nearly all the starting material was unwound by about 10 min ( Fig. 2A). The ability of rBLM to unwind G4-TP was also dependent upon protein concentration (Fig. 2B). Within the range of 1 to 20 ng of rBLM tested, unwinding increased and reached a maximum at about 10 ng (67 fmol) of rBLM. This represents an approximately 2-fold molar excess of rBLM protein to G4 DNA.
G4 DNA Unwinding Is ATP-dependent-Helicases generate the driving force to catalyze DNA unwinding by hydrolysis of nucleoside triphosphates (reviewed in Ref. 5). The helicase activity of rBLM on duplex substrates has been shown to de- 1 The abbreviations used are: G4-TP, G4 DNA prepared from TP; nt, nucleotide(s); ccc, covalently closed circle. pend upon the presence of ATP in the reaction (18). We tested unwinding of G4 DNA by rBLM in the presence and absence of ATP and in the presence of the nonhydrolyzable analogue, ATP␥S. Unwinding occurred only when ATP was present and was not supported by ATP␥S (Fig. 2C).
Polarity and Substrate Requirements for G4 DNA Unwinding-The direction of translocation of a DNA helicase is defined operationally as the polarity of flanking single-stranded DNA that is necessary to initiate unwinding in vitro. rBLM was previously shown to unwind most efficiently duplex DNA molecules with 3Ј single-stranded tails, and it is, therefore, a 3Ј-5Ј helicase (18). We compared unwinding of G4 DNA formed from G4-TP, which has a 7-nt 3Ј single-stranded tail; G4-TP-S, which has a 4-nt single-stranded 3Ј tail; and G4-TP-B, which has no 3Ј single-stranded region. Unwinding of G4-TP-S was much less efficient than unwinding of G4-TP; and there was essentially no unwinding of G4-TP-B (Fig. 2D). A similar dependence of unwinding on a 3Ј single-stranded tail was also evident in assays using G4 DNA formed from the OX-1 and OX-1T oligonucleotides, which carry the Oxytricha telomeric repeat d(TTTTGGGG). G4-OX-1T has a 3Ј single-stranded tail 7 nt in length, similar to G4-TP; and G4-OX-1 has no 3Ј singlestranded region. G4-OX-1T was fully unwound by rBLM, whereas G4-OX-1 was not (Fig. 2D).
The directionality of unwinding G4 DNA by BLM was therefore 3Ј-5Ј, and unwinding required the presence of a 3Ј singlestranded end. Unwinding was efficient if the 3Ј tail was 7 nt long but was significantly diminished on substrates with a 4-nt tail. These aspects of the unwinding activity of BLM on G4 DNA are consistent with previous studies using Watson-Crick duplex substrates (18 Fig. 2). Comparison of the unwinding activity of rBLM on G4 DNA and the partial duplex substrates showed that significantly more rBLM was required to unwind duplex DNA than G4 DNA. In reactions containing approximately 12 fmol of each DNA substrate, approximately 5 ng (33 fmol) of rBLM was sufficient to unwind all the G4 DNA substrate (Fig. 3A). Relatively little unwinding of the partially duplex ccc substrate was evident with this amount of rBLM, but when 100 ng of protein was included, the reaction neared completion (Fig. 3A). Essentially none of the TP partial duplex was unwound in reactions containing as much as 50 ng of rBLM (Fig. 3B). The H1/K1 duplex fork was a better substrate than the TP partial duplex, but the amount of enzyme required to unwind all the H1/K1 fork substrate was about 10-fold higher than that required to unwind a comparable amount of G4 DNA (Fig. 3B). Since the T m of the H1/K1 duplex is 68°C while the T m of G4-TP is over 90°C (21), this is unlikely to reflect relative thermodynamic stabilities of the two substrates. We conclude that the BLM helicase is more active on G4 DNA substrates than on duplex DNA. The preference of the BLM helicase for G4 DNA substrates was further borne out by competition experiments. rBLM was incubated with labeled G4-TP in the presence of cold G4-TP or duplex H1/K1 fork substrates, reaction products resolved by gel electrophoresis, and the percentage of DNA unwound was quantified by phosphoimaging (Fig. 3C). In parallel experiments, the ability of cold G4-TP or duplex H1/K1 fork substrates to compete with unwinding of labeled H1/K1 fork substrates was analyzed (Fig. 3D). A molar equivalent of G4-TP partially competed unwinding of labeled G4-TP, and competi-tion was nearly complete at 5-fold excess; but H1/K1 could not compete for G4-TP unwinding, even at 20-fold molar excess (Fig. 3C). Conversely G4-TP competed for about 80 and nearly 100% of H1/K1 unwinding at 3-and 5-fold molar excess, respectively (Fig. 3D). We note that although it may appear that H1/K1 fork substrates cannot compete for unwinding of labeled H1/K1 with 100% efficiency, this probably reflects the fact that high concentrations of competitor oligonucleotides drive the newly denatured single strands back into duplexes. The competition analysis therefore supports the notion that G4 DNA is RecBCD Helicase Cannot Unwind G4 DNA-To determine whether the ability to unwind G4 DNA is shared by other DNA helicases, we asked if RecBCD helicase could unwind G4 DNA. RecBCD is known to be involved in DNA repair and recombination in E. coli (reviewed in Ref. 25). RecBCD possesses ATPdependent exonuclease, endonuclease, and helicase activities (24, 26 -28). RecBCD was assayed for helicase activity on duplex and G4 DNA substrates under conditions designed to minimize interference by its associated nuclease activity (24). RecBCD was preincubated with DNA substrates in reactions containing Mg 2ϩ but lacking ATP, and reactions were initiated by adding ATP. When rBLM was assayed under these conditions, it unwound G4 DNA; but RecBCD displayed no G4 DNA unwinding activity even at very high enzyme concentrations (Fig. 4A). Similar results were obtained with longer incubation times (data not shown). Although RecBCD was inactive on G4 DNA, this same preparation of enzyme actively unwound both the H1/K1 fork and TP partial duplex DNAs under these reaction conditions (Fig. 4B). These results show that the ability to unwind G4 DNA is not a common feature of all DNA helicases.

DISCUSSION
The experiments reported here demonstrate that the BLM helicase can unwind G4 DNA. Like duplex DNA unwinding by rBLM (18), G4 DNA unwinding requires a 3Ј single-stranded region and is ATP-dependent. Most strikingly, G4 DNA is a better substrate for rBLM than is duplex DNA. Two lines of evidence support this conclusion. First, the amount of rBLM needed to unwind a given amount of G4 DNA is at least 10-fold lower than that needed to unwind a comparable number of moles of a duplex DNA substrate. Second, G4 DNA outcompetes duplex DNA in rBLM unwinding assays performed with either G4 or duplex substrates. These results argue that G4 DNA (or related G-DNA structures) may be a natural target for the BLM helicase in vivo.
G4 DNA Forms Readily and Is Very Stable in Vitro-In G4 DNA, four parallel DNA strands are stabilized by Hoogsteen Replication to produce two daughter chromosomes (left) or recombination between sister chromatids (right) may expose single-stranded regions of nucleic acids containing multiple G runs. This could create a high local concentration of G runs that drives formation of G quartets which stabilize G-DNA junctions between strands. In normal cells, the BLM helicase could efficiently remove a G quartet-stabilized junction. In the absence of BLM function, this junction might persist, resulting in elevated recombination levels. Left, top, replication fork (proceeding right to left); center, G-DNA junction forms; bottom, replication (left to right) proceeds nearly to completion but fork remains immobilized by G-DNA junction. Right, top, sister chromatids; center, initiation of recombination by an invading 3Ј-end; bottom, G-junction immobilizes further recombination. Open and closed circles denote guanines on sister chromatids; two of the four guanines in each G quartet are shown as deriving from each sister chromatid, although other combinations are possible. G-DNA is shown as involving antiparallel strands, to simplify the diagram. Formation of G-DNA structures in vivo could involve DNA spanning hundreds of nucleotides or more, in either parallel or antiparallel configuration. bonding between guanines (Fig. 1A). Although G4 DNA structures have not been directly identified in vivo, G-rich DNAs spontaneously form G4 DNA in vitro in a reaction that shows a quadratic dependence upon concentration (21). In fact, G4 DNA forms so readily that laboratory stocks of G-rich synthetic oligonucleotides typically contain a small fraction of G4 DNA. Once formed, G4 DNA is exceedingly stable: the T m of G4 DNA derived from the TP synthetic 49 mer used in our experiments is above 90°C (21), whereas the T m of the same sequence as duplex DNA is 76°C. One can imagine that a structure this stable would be deleterious in vivo if it persisted once formed, and it is reasonable that cells would possess mechanisms to disrupt G4 DNA structures. The results presented here provide an enzymatic candidate for this role.
Can the Persistence of G-DNA Explain the Cellular Phenotype of Bloom's Syndrome?-Genetic instability because of greatly increased levels of recombination characterizes Bloom's syndrome. Most of the recombination appears to reflect mitotic exchanges between sister chromatids or homologous chromosomes. Mitotic recombination is also elevated in yeast mutants in the BLM homologs, S. cerevisiae SGS1 (14,17) and S. pombe rqh1 ϩ (15). Replication or recombination that precedes mitosis could potentiate G-DNA formation in vivo by generating singlestranded regions, which interact to form a junction stabilized by G quartets rather than standard Watson-Crick pairing. Fig. 5 shows how a G-DNA junction might form in vivo. Recombination or replication would expose a single-stranded G-rich region, allowing G-DNA to form. In normal cells, the BLM helicase could efficiently remove a G-DNA junction, but in the absence of BLM function, this junction might persist, prolonging the opportunity for strand invasion by the freed 3Ј-end. Formation of the structures diagrammed in Fig. 5 could account for the very striking karyotypic abnormalities seen in Bloom's syndrome cells: the high frequency of sister chromatid exchanges and symmetric quadriradial chromosomes that appear to represent unresolved recombination intermediates.
Not All Helicases Can Unwind G4 DNA-Specific helicases are associated with replication and recombination, but these helicases could mitigate against formation of G-DNA only if they can unwind this substrate efficiently. Our results show that the ability to unwind G4 DNA is not a universal property of DNA helicases. At least one very potent recombinationassociated helicase, E. coli RecBCD, is essentially inactive on G4 DNA. One viral helicase has been shown to unwind G4 DNA, the SV40 T antigen helicase (29), but the activity of BLM on G4 DNA substrates may be much greater than that of SV40 T antigen. BLM helicase reactions reach completion in 20 min at a 2:1 molar ratio of polypeptide:G4 DNA (Fig. 2B), whereas SV40 T antigen reactions have thus far been shown to require a 360:1 molar ratio of polypeptide:G4 DNA and 45 min of incubation (29).
BLM is a member of a helicase subfamily that includes S. cerevisiae Sgs1p, S. pombe Rqh1p, and the human WRN helicase. BLM (18), WRN (30), and Sgs1p (31,32) can all unwind duplex DNA substrates in vitro, with 3Ј-5Ј directionality. The observation that the BLM helicase is more active on G4 DNA than on duplex substrates raises the possibility that at least some of the helicases related to BLM may similarly be active on G-DNA. Ribosomal DNA is G-rich and has considerable potential to form G-DNA, and the possibility that Sgs1p may unwind G-DNA is particularly interesting in light of the alterations in rDNA maintenance and nucleolar structure that characterize S. cerevisiae Sgs1 mutants (33,34).
Molecular genetic analysis of Bloom's syndrome patients predicts that deficiency in the helicase alone should be sufficient to account for the pathology of the disease (4). If at least one specialized function of BLM in vivo is G4 DNA unwinding, then absence of this helicase could explain the elevated recombination levels in Bloom's syndrome cells, particularly if other helicases cannot unwind G4 DNA efficiently.