Key points are not available for this paper at this time.
The yeast Saccharomyces cerevisiaeSgs1 protein is a member of a family of DNA helicases that include the Escherichia coli RecQ protein and the products of human Bloom's syndrome and Werner's syndrome genes. To study the enzymatic characteristics of the protein, a recombinant Sgs1 fragment (amino acids 400–1268 of the 1447-amino acid full-length protein) was overexpressed in yeast and purified to near homogeneity. The purified protein exhibits an ATPase activity in the presence of single- or double-stranded DNA. In the presence of ATP or dATP, unwinding of duplex DNA or a DNA-RNA heteroduplex by the recombinant Sgs1 fragment was readily observed. Similar to the E. coli RecQ helicase, displacement of the DNA strand occurs in the 3′ to 5′ direction with respect to the single-stranded DNA flanking the duplex. The efficiency of unwinding was found to correlate inversely with the length of the duplex region and was enhanced by the presence of E. colisingle-stranded DNA-binding protein. In addition, the recombinant Sgs1 fragment was found to bind more tightly to a forked DNA substrate than to either single- or double-stranded DNA. The yeast Saccharomyces cerevisiaeSgs1 protein is a member of a family of DNA helicases that include the Escherichia coli RecQ protein and the products of human Bloom's syndrome and Werner's syndrome genes. To study the enzymatic characteristics of the protein, a recombinant Sgs1 fragment (amino acids 400–1268 of the 1447-amino acid full-length protein) was overexpressed in yeast and purified to near homogeneity. The purified protein exhibits an ATPase activity in the presence of single- or double-stranded DNA. In the presence of ATP or dATP, unwinding of duplex DNA or a DNA-RNA heteroduplex by the recombinant Sgs1 fragment was readily observed. Similar to the E. coli RecQ helicase, displacement of the DNA strand occurs in the 3′ to 5′ direction with respect to the single-stranded DNA flanking the duplex. The efficiency of unwinding was found to correlate inversely with the length of the duplex region and was enhanced by the presence of E. colisingle-stranded DNA-binding protein. In addition, the recombinant Sgs1 fragment was found to bind more tightly to a forked DNA substrate than to either single- or double-stranded DNA. The SGS1 gene of Saccharomyces cerevisiaewas identified in a search for extragenic suppressors of the slow growth phenotype of cells deficient in DNA topoisomerase III, the product of the TOP3 gene (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar). Paradoxically, whereas null mutations in SGS1 suppress the growth defect oftop3 mutants, they significantly reduce the growth rate oftop1 mutants lacking a functional DNA topoisomerase I (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar, 2Lu J. Mullen J.R. Brill S.J. Kleff S. Romeo A.M. Sternglanz R. Science. 1996; 383: 678-679Google Scholar, 3West S.C. Cell. 1996; 86: 177-180Abstract Full Text Full Text PDF PubMed Scopus (119) Google Scholar). Based on the results of two-hybrid screens in yeast (4Fields S. Song O. Nature. 1989; 340: 245-246Crossref PubMed Scopus (4853) Google Scholar), it was suggested that Sgs1 protein interacted directly with DNA topoisomerase III (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar) and perhaps DNA topoisomerase II as well (5Watt P.M. Louis E.J. Borts R.H. Hickson I.D. Cell. 1995; 81: 253-260Abstract Full Text PDF PubMed Scopus (376) Google Scholar). Yeastsgs1 mutants show increased genome instability; the rate of mitotic recombination between homologous sequences is elevated (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar, 5Watt P.M. Louis E.J. Borts R.H. Hickson I.D. Cell. 1995; 81: 253-260Abstract Full Text PDF PubMed Scopus (376) Google Scholar,6Watt P.M. Hickson I.D. Borts R.H. Louis E.J. Genetics. 1996; 144: 935-945Crossref PubMed Google Scholar), and chromosome missegregation occurs more frequently during mitosis and meiosis (5Watt P.M. Louis E.J. Borts R.H. Hickson I.D. Cell. 1995; 81: 253-260Abstract Full Text PDF PubMed Scopus (376) Google Scholar, 6Watt P.M. Hickson I.D. Borts R.H. Louis E.J. Genetics. 1996; 144: 935-945Crossref PubMed Google Scholar). These characteristics are reminiscent of those of yeast top3 mutants lacking DNA topoisomerase III; in addition to a slow growth phenotype, top3 mutants show an elevated frequency of recombination between repetitive sequences and are defective in sporulation (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar, 7Wallis J.W. Chrebet G. Brodsky G. Rolfe M. Rothstein R. Cell. 1989; 58: 409-419Abstract Full Text PDF PubMed Scopus (453) Google Scholar). The hyper-rec phenotype oftop3 mutants is also suppressed by mutations inSGS1 (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar). The sequence of the Sgs1 protein suggests that it possesses a central region homologous to the Escherichia coli RecQ helicase (8Nakayama K. Irino N. Nakayama H. Mol. Gen. Genet. 1985; 200: 266-271Crossref PubMed Scopus (124) Google Scholar,9Umezu K. Nakayama K. Nakayama H. Proc. Natl. Acad. Sci. U. S. A. 1990; 87: 5363-5367Crossref PubMed Scopus (229) Google Scholar), and a helicase activity was detected in a rabbit reticulocyte coupled transcription/translation system expressing the Sgs1 protein (2Lu J. Mullen J.R. Brill S.J. Kleff S. Romeo A.M. Sternglanz R. Science. 1996; 383: 678-679Google Scholar). The plausible association between the yeast Sgs1 helicase and DNA topoisomerase III, a member of the type IA subfamily of DNA topoisomerases that also includes E. coli DNA topoisomerases I and III, is reminiscent of an enzyme termed “reverse gyrase,” which was previously found only in thermophilic organisms (for a review, see Ref. 10Duguet M. Eckstein F. Lilley D.M.J. Nucleic Acids and Molecular Biology. 9. Springer-Verlag, Berlin1995: 84-114Google Scholar). Reverse gyrase catalyzes positive supercoiling of DNA, and its sequence suggests that it possesses both a DNA helicase and a type IA DNA topoisomerase (Refs. 11Confalonieri F. Elie C. Nadal M. Bouthier de la Tour C. Forterre P. Duguet M. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 4753-4757Crossref PubMed Scopus (135) Google Scholar, 12Kozyavkin S.A. Krah R. Gellert M. Stetter K.O. Lake J.A. Slesarev A.I. J. Biol. Chem. 1994; 269: 11081-11089Abstract Full Text PDF PubMed Google Scholar, 13Nadal M. Couderc E. Duguet M. Jaxel C. J. Biol. Chem. 1994; 269: 5255-5263Abstract Full Text PDF PubMed Google Scholar; for the classification of DNA topoisomerases, see Ref. 14Wang J.C. Annu. Rev. Biochem. 1996; 65: 635-692Crossref PubMed Scopus (2070) Google Scholar). In the case of the Sulfolobus acidocaldarius enzyme, the putative helicase and the topoisomerase activity are present on the same polypeptide (11Confalonieri F. Elie C. Nadal M. Bouthier de la Tour C. Forterre P. Duguet M. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 4753-4757Crossref PubMed Scopus (135) Google Scholar). The association between these activities in a single enzyme led to the proposal that the enzyme acts by using its helicase activity to unwind DNA, generating both positive and negative supercoils; removal of the negative supercoils by the topoisomerase activity then results in a net accumulation of positive supercoils (11Confalonieri F. Elie C. Nadal M. Bouthier de la Tour C. Forterre P. Duguet M. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 4753-4757Crossref PubMed Scopus (135) Google Scholar). Recently, several additional homologues of the E. coli RecQ and yeast Sgs1 protein have been discovered. These include the products of the human Bloom's syndrome and Werner's syndrome genes (BLM and WRN, respectively) (15Ellis N.A. Groden J. Ye T.-Z. Straughen J. Lennon D.J. Ciocci S. Proytcheva M. German J. Cell. 1995; 83: 655-666Abstract Full Text PDF PubMed Scopus (1210) Google Scholar, 16Yu C.-E. Oshima J. Fu Y.-H. Wijsman E.M. Hisama F. Alisch R. Matthews S. Nakura J. Miki T. Ouais S. Martin G.M. Mulligan J. Schellenberg G.D. Science. 1996; 272: 258-262Crossref PubMed Scopus (1488) Google Scholar). Although the clinical features of the two diseases are rather different, patients of both syndromes exhibit chromosome instability and a predisposition to cancer (for reviews, see Refs. 17Ellis N.A. Curr. Opin. Genet. Dev. 1997; 7: 354-363Crossref PubMed Scopus (115) Google Scholar and 18Lombard D.B. Guarente L. Trends Genet. 1996; 12: 283-286Abstract Full Text PDF PubMed Scopus (18) Google Scholar). BLM and WRN proteins are similar to Sgs1 protein in size and share limited sequence homology outside of the central helicase domain, suggesting that these RecQ-type proteins might be functionally related. ASchizosaccharomyces pombe homologue of SGS1, termed rqh1 +, was also reported recently (19Stewart E. Chapman C.R. Al-Khodairy F. Carr A.M. Enoch T. EMBO J. 1997; 16: 2682-2692Crossref PubMed Scopus (328) Google Scholar). Similar to SGS1 of the budding yeast,rqh1 + has been shown to suppress recombination, especially during S phase arrest (19Stewart E. Chapman C.R. Al-Khodairy F. Carr A.M. Enoch T. EMBO J. 1997; 16: 2682-2692Crossref PubMed Scopus (328) Google Scholar). Interestingly, the E. coli RecQ protein has also been shown to suppress illegitimate recombination (20Hanada K. Ukita T. Kohno Y. Saito K. Kato J.-I. Ikeda H. Proc. Natl. Acad. Sci. U. S. A. 1997; 94: 3860-3865Crossref PubMed Scopus (239) Google Scholar). As a first step in studying the mechanistic and functional aspects of the Sgs1 protein, we have purified a truncated form of it containing the helicase domain. We report here the DNA-dependent ATPase and helicase activities of the purified protein. The regions of the SGS1 gene encoding the N- and C-terminal segments of the protein were amplified by the polymerase chain reaction (PCR) 1The abbreviations used are: PCR, polymerase chain reaction; ssDNA, single-stranded DNA; Ssb, ssDNA-binding protein; ATPγS, adenosine 5′-γ-thiotriphosphate; HA, a flu virus hemagglutinin epitope with the amino acid sequence YPYDVP. from pRS414-SGS1 (a gift from Dr. R. Sternglanz, State University of New York at Stony Brook). The N-terminal coding segment was amplified using the VENT polymerase (New England Biolabs) and a pair of oligodeoxynucleotides RB1 (5′-GAGTCTACGGGATCCGTAACCATGGTGACGAAGCCGTC-3′) and RB4 (5′-GGGGATCAAGCTTGCGGCAGTGTACGAGAGC-3′). The sequence of RB1 was designed to place a BamHI site GGATCC (underlined) of the of SGS1, that the be a previously for the of several proteins in The sequence of RB4 was designed to a site (underlined) to the of the N-terminal segment for the of The C-terminal coding segment of SGS1 was amplified using the and and was designed to the C-terminal to the sequence in an the site in both that a of be to the of Sgs1 protein. The and J. C. is a of a for the of yeast DNA topoisomerase II from the gene S.C. J. Biol. Chem. 1989; Full Text PDF PubMed Google Scholar). The N-terminal product was between the BamHI and of to The C-terminal product was then between the and of to The segment in between the site in the N-terminal coding region and the site in the C-terminal coding region of SGS1 was then by the fragment from DNA that the the coding The BamHI to segment of was between the same in for the of full-length Sgs1 protein with a at its expressing truncated SGS1 were from and for of the The used in the reported which a recombinant Sgs1 protein of amino acids 400–1268 of the protein, was from by the of the pair and In the 400–1268 of yeast SGS1 were between a and a site (underlined) in the The to segment was then The is to a recombinant protein the amino acids at its the first are from the of yeast DNA topoisomerase II and the the flu virus hemagglutinin epitope which was the recombinant protein to its by with a The at the was to the of the recombinant protein. was first a yeast P. J.C. Mol. Cell. Biol. 1994; 14: PubMed Scopus Google Scholar; was from Dr. R. was and in lacking and with was used to of lacking the was yeast with acid and and on a (New in a a of was the was by the addition of to of the recombinant protein was to for a at were by for at and and then in an of and a of and was to and the cells were by the addition of an of and in a In was by the S.C. J. Biol. Chem. 1989; Full Text PDF PubMed Google Scholar). The was from the and by and the were in A. The and were and by The was directly a in A. the with the same proteins were in The was then directly to a with proteins were with a of in in by a step containing the truncated Sgs1 protein were identified by were with to a to and was to The purified protein was then with to the of the protein to the The were then several in containing and a step with and the recombinant Sgs1 fragment was in containing and The protein was a The of the protein, by the of a protein with as was between and for several DNA were purified by The was from DNA were from and and were to form the pair double-stranded DNA substrate with a at the and and were to form the forked DNA substrate with a double-stranded segment with single-stranded at regions in and are In at its using and was also used as the single-stranded DNA DNA for helicase activity were by the of a DNA strand with single-stranded DNA. and in length were by and were at by with and or at the by with and The and were to at and and the sequence of the DNA was the same as in EMBO J. 1994; PubMed Scopus Google Scholar). were from the pair and the pair of the double-stranded form of These were at and purified on a The were then and with DNA. for DNA and of the products were previously S.C. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: PubMed Scopus Google Scholar). DNA and form DNA were from New England of DNA by virus topoisomerase was at in and The DNA was and were in ATPase and of for at reaction of and the of recombinant Sgs1 protein and DNA were by the addition of of to and of reaction was The were in acid and and the of ATP in the reaction was from the using a substrate DNA and were as for the ATPase ATP was were by the addition of of a by at for were either by in a containing or were on and were on The were using a DNA to that of the DNA in The E. coli single-stranded DNA protein used in of the helicase was from To the of the recombinant Sgs1 fragment to either single-stranded or form DNA, the of protein and DNA were in of DNA and were at for and the were by in in DNA and were by with To the of Sgs1 with or forked DNA, DNA and recombinant Sgs1 were at for in the DNA were by in a in were and the DNA was using a to full-length Sgs1 protein were of the of the protein overexpressed in S. To were for the of truncated proteins from an yeast gene was designed for expressing amino acid a for amino acids and a for amino acids the for the of an protein that with an of in a of the products that with of and with the of than that of of its the fragment was for In the for the of for acids 400–1268 of were of a of the yeast for two a epitope of flu virus and a of were to the and of the SGS1 to the and of the recombinant protein product for of a S. were with that the cells were and with a protein was overexpressed with an that the overexpressed protein was the product of the recombinant protein were then and the protein was purified to near The purified recombinant Sgs1 fragment was for its to ATP in the presence and of DNA. As shown in a DNA-dependent ATPase activity was readily activity with the of the recombinant Sgs1 protein from a of ATP was in the presence of several DNA either single- or double-stranded DNA In for the E. it was previously shown that single-stranded DNA was a more than double-stranded DNA K. Nakayama K. Nakayama H. Proc. Natl. Acad. Sci. U. S. A. 1990; 87: 5363-5367Crossref PubMed Scopus (229) Google Scholar). The of the of the shown in to a rate of of Sgs1 The DNA activity of the purified recombinant protein was using the The Sgs1 fragment was found to bind single-stranded DNA the presence of of the protein the of the DNA of the protein to to be and of the DNA was only at the of the recombinant protein used The of the Sgs1 fragment to or To the that to the single-stranded DNA might be by Sgs1 to a duplex DNA with single-stranded at forked DNA was The forked DNA was by the of two in was found that forked DNA with the Sgs1 fragment in the of a as detected by the same only to either a single-stranded or duplex DNA was observed. To for DNA helicase activity of the recombinant protein, we it a from single-stranded DNA. The Sgs1 fragment was found to the of the from the single-stranded DNA in the presence of ATP and and and activity at a of ATP to of activity was ATP was by the ATP in the presence of the addition of to the reaction was found to significantly the unwinding activity ATP be by and DNA unwinding by the Sgs1 fragment These are similar to those in the unwinding of DNA K. Nakayama K. Nakayama H. Proc. Natl. Acad. Sci. U. S. A. 1990; 87: 5363-5367Crossref PubMed Scopus (229) Google Scholar). Similar to the case of E. coli the addition of or significantly the unwinding reaction K. Nakayama K. Nakayama H. Proc. Natl. Acad. Sci. U. S. A. 1990; 87: 5363-5367Crossref PubMed Scopus (229) Google Scholar). of the strand displacement reaction is shown in of the fragment was and the reaction was to The DNA unwinding activity was to the or in the reaction was a in the unwinding activity at a between and at the helicase activity was than of that in the The recombinant Sgs1 fragment was also for its to unwind a DNA-RNA or DNA was to DNA and with the recombinant Sgs1 fragment in the presence of and of the by the protein with an efficiency similar to that of the the helicase unwinding of the DNA-RNA was by the of ATP or the To the of the helicase the DNA shown in were These were by first a or in with The duplex region of was then by to a with either a or a to of its The recombinant Sgs1 fragment was found to the In unwinding and of the by the protein was The was to unwinding of its the was with an efficiency to that of the These results that the Sgs1 protein has a 3′ to 5′ of unwinding with respect to the flanking the duplex as was with E. coli RecQ helicase K. Nakayama K. Nakayama H. Proc. Natl. Acad. Sci. U. S. A. 1990; 87: 5363-5367Crossref PubMed Scopus (229) Google Scholar). To the of the efficiency of DNA unwinding on the length of the duplex several DNA were using of or As in substrate was with the of the recombinant Sgs1 and the products of the reaction were by was found that the efficiency of strand displacement was by its with more Sgs1 protein to the DNA the recombinant Sgs1 fragment of the fragment at a of and only of the fragment and of the fragment was at a enzyme of and and The of E. coli on the helicase activity of the recombinant Sgs1 protein was also a DNA substrate a it was that the Sgs1 helicase activity was by the presence of of the of displacement of the DNA strand was enhanced by more than to of Sgs1 helicase activity by was also using a strand to its single-stranded We have shown that a purified recombinant Sgs1 which amino acid 400–1268 of the full-length protein, possesses a helicase activity with duplex DNA The DNA unwinding reaction is on the presence of ATP or of the to be as for ATP and the helicase activity in the presence of the E. coli RecQ helicase, which is by double-stranded DNA, the of ATP by the recombinant Sgs1 protein is by both of DNA. The of unwinding is 3′ to 5′ with respect to the single-stranded DNA. The characteristics of the unwinding activity are similar to those of the E. coli RecQ protein K. Nakayama K. Nakayama H. Proc. Natl. Acad. Sci. U. S. A. 1990; 87: 5363-5367Crossref PubMed Scopus (229) Google Scholar), which homology with Sgs1 the helicase domain. the of the RecQ family of DNA helicases the human protein has previously been purified and shown to have DNA unwinding M. H. S. J. T. Nucleic Acids 1994; PubMed Scopus Google Scholar, J. Biol. Chem. 1994; 269: Full Text PDF PubMed Google Scholar). Recently, the Werner's syndrome and Bloom's syndrome proteins were also overexpressed and shown to a DNA helicase activity N. A. O. J. S. M. Y. Nucleic Acids 1997; PubMed Scopus Google Scholar, A. Martin G.M. Oshima J. Genet. 1997; PubMed Scopus Google Scholar, Hickson I.D. J. Biol. Chem. 1997; 272: PubMed Scopus Google Scholar). that of the of the RecQ family share a helicase The efficiency of strand displacement by the recombinant Sgs1 fragment with length of the strand to be The unwinding used only displacement and be is plausible that with the of the DNA the protein reduce the efficiency of the recombinant Sgs1 fragment might exhibit a in its reaction and might the DNA the displacement of the The increased efficiency of unwinding of the fragment in the presence of E. coli protein is with either of these We have also that Sgs1 has the to an strand to a DNA with the same efficiency and as DNA from DNA. The product of the Werner's syndrome gene was also found to unwind to DNA N. A. O. J. S. M. Y. Nucleic Acids 1997; PubMed Scopus Google Scholar). The to unwind is a of In of E. coli DNA helicases Proc. Natl. Acad. Sci. U. S. A. 1989; 86: PubMed Scopus Google Scholar), only protein) unwinding of is the unwinding of heteroduplex by the recombinant Sgs1 fragment is in of the of Sgs1 protein. an on the in the of a DNA by protein, the recombinant Sgs1 protein fragment was shown to have a for a forked DNA substrate than either single- or double-stranded DNA. have for the of DNA unwinding by it has been suggested that of the enzyme to a might be for helicase Annu. Rev. Biochem. 1996; 65: PubMed Scopus Google Scholar). In addition, the for to a DNA be the protein is to as a in with a DNA topoisomerase The the of the RecQ family of of the functional and association between yeast Sgs1 protein and DNA topoisomerase III, it was suggested that a between the two as a gyrase to positive regions in DNA (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar). a is in of a are to a positive supercoiling activity of the recombinant Sgs1 fragment in the presence of yeast DNA topoisomerase III have been and J. C. The of the N- and C-terminal regions of the full-length Sgs1 protein in the recombinant protein its of a with the topoisomerase or its with DNA. for the of Sgs1 protein in is that it might by or with DNA topoisomerase III in the of DNA a activity might be near the of DNA to the DNA (1Gangloff S. McDonald J.P. Bendixen C. Arthur L. Rothstein R. Mol. Cell. Biol. 1994; 14: 8391-8398Crossref PubMed Scopus (617) Google Scholar, 5Watt P.M. Louis E.J. Borts R.H. Hickson I.D. Cell. 1995; 81: 253-260Abstract Full Text PDF PubMed Scopus (376) Google Scholar, J.C. J. Biol. Chem. Full Text PDF PubMed Google Scholar, J.C. J. Biol. Chem. Full Text PDF PubMed Google Scholar). to to DNA strand and an in DNA recombination and chromosome The Sgs1 topoisomerase III pair might also directly on recombination to DNA and reduce the mitotic recombination frequency J.C. J. Biol. Chem. Full Text PDF PubMed Google Scholar, J.C. J. Biol. Chem. Full Text PDF PubMed Google Scholar). In the case of E. coli DNA topoisomerase III, which yeast DNA topoisomerase III in its J.C. J. Biol. Chem. Full Text PDF PubMed Google Scholar, J. Biol. Chem. Full Text PDF PubMed Google Scholar), the enzyme has been shown to be in the of in an in DNA system J. Biol. Chem. Full Text PDF PubMed Google Scholar, H. J. Biol. Chem. 1994; 269: Full Text PDF PubMed Google Scholar). The of the Bloom's and the Werner's syndrome as homologues of the RecQ-type proteins in E. coli and has in of The functional and between the yeast Sgs1 protein and yeast DNA topoisomerase III that homologues might also Interestingly, results that are two of DNA topoisomerase III by genes to chromosome R. J.C. Proc. Natl. Acad. Sci. U. S. A. 1996; PubMed Scopus Google Scholar, K. S. E. K. A. J. N. 1997; 7: PubMed Scopus Google Scholar) and chromosome K. S. E. K. A. J. N. 1997; 7: PubMed Scopus Google Scholar). The plausible between these DNA topoisomerases and and a of helicases the WRN, and in mechanistic and functional We for on substrate Sternglanz for and and for and of the
Bennett et al. (Wed,) studied this question.