HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Wu, L. Articles by Hickson, I. D. Search for Related Content PubMed PubMed Citation Articles by Wu, L. Articles by Hickson, I. D. Social Bookmarking                      What's this?

J Biol Chem, Vol. 275, Issue 13, 9636-9644, March 31, 2000

The Bloom's Syndrome Gene Product Interacts with Topoisomerase III^*

Leonard Wu, Sally L. Davies, Phillip S. North, Hélène Goulaouic^§, Jean-François Riou^§, Helen Turley^¶, Kevin C. Gatter^¶, and Ian D. Hickson

From the  Imperial Cancer Research Fund Laboratories, Institute of Molecular Medicine, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DS, United Kingdom, ^§ Rhône-Poulenc Rorer, Centre de Recherche de Vitry-Alfortville, Vitry sur Seine Cedex, 94403 France, and the ^¶ Department of Cellular Science, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DU, United Kingdom

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Bloom's syndrome is a rare genetic disorder associated with loss of genomic integrity and a large increase in the incidence of many types of cancer at an early age. The Bloom's syndrome gene product, BLM, belongs to the RecQ family of DNA helicases, which also includes the human Werner's and Rothmund-Thomson syndrome gene products and the Sgs1 protein of Saccharomyces cerevisiae. This family shows strong evolutionary conservation of protein structure and function. Previous studies have shown that Sgs1p interacts both physically and genetically with topoisomerase III. Here, we have investigated whether this interaction has been conserved in human cells. We show that BLM and hTOPO III, one of two human topoisomerase III homologues, co-localize in the nucleus of human cells and can be co-immunoprecipitated from human cell extracts. Moreover, the purified BLM and hTOPO III proteins are able to bind specifically to each other in vitro, indicating that the interaction is direct. We have mapped two independent domains on BLM that are important for mediating the interaction with hTOPO III. Furthermore, through characterizing a genetic interaction between BLM and TOP3 in S. cerevisiae, we have identified a functional role for the hTOPO III interaction domains in BLM.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Bloom's syndrome is a genetic disorder characterized by retarded growth, sun sensitivity, immunodeficiency, and a predisposition to a wide variety of cancers (1). Cells from affected individuals show genomic instability, the hallmark feature being hyperrecombination between sister chromatids. The gene mutated in Bloom's syndrome, BLM, encodes a member of the RecQ family of DEXH box DNA helicases (2). This family of helicases has been conserved from bacteria to humans, and all members share a central core domain comprising seven highly conserved motifs found in all known DNA helicases (3). In addition to BLM, there are at least four other human RecQ homologs. Defects in two of these, WRN and RECQL4, are also associated with disease conditions (4, 5). WRN is the gene mutated in the premature aging disorder Werner's syndrome, and RECQL4 is mutated in Rothmund-Thomson syndrome, a rare condition associated with skin and skeletal abnormalities, as well as some features of premature aging (6). Like Bloom's syndrome, both of these diseases also give rise to an elevated incidence of cancer, although to a lesser extent than is seen in Bloom's syndrome.

In eukaryotes, the RecQ helicase family can be divided into two subfamilies: those that consist of essentially the helicase domain and those in which the helicase domain is flanked by large, poorly conserved N- and C-terminal domains (3). BLM falls into the latter class, along with WRN and RECQL4, as well as the Schizosaccharomyces pombe Rqh1 (7) and Saccharomyces cerevisiae Sgs1 proteins (8, 9). In addition to structural homology, there exists a certain degree of conservation of function among the various eukaryotic RecQ helicases. For example, Sgs1p, BLM and WRN display 3'-5' DNA helicase activity, suggesting that they share a similar mechanism of action (10-12). Mutations in BLM, WRN, SGS1, or rqh1⁺ result in genomic instability (3), and more specifically, BLM, sgs1, and rqh1 mutants display elevated levels of homologous recombination (7, 8, 13-15). Moreover, the hyperrecombination phenotype of sgs1 mutants can be partially suppressed by ectopic expression of either BLM or WRN (16).

The torsional stress introduced into DNA when the complementary strands are separated by helicases is relieved by a class of enzymes known as topoisomerases. These enzymes catalyze the passage of intact DNA strands across transient DNA breaks and fall into one of two classes, designated type I and type II, depending upon their mechanism of action. Type I topoisomerases generate single-stranded DNA nicks, whereas type II topoisomerases make double-stranded DNA breaks (17, 18). There is increasing evidence to suggest that the RecQ helicases act in concert with one member of the type I class, topoisomerase III. S. cerevisiae expresses a single topoisomerase III enzyme encoded by the TOP3 gene (19). top3 mutants grow very slowly, display hyperrecombination between repetitive sequences throughout the genome, and have defects in meiotic recombination (8, 20). Mutations in SGS1 partially suppress the pleiotropic effects of top3 mutations, suggesting that Sgs1p and Top3p act in the same pathway (8). Moreover, Sgs1p and Top3p interact physically and may, therefore, act in a coordinated manner as a heterodimeric or larger complex (8). What role such a complex plays in the cell is unclear. Harmon et al. (21) have shown that E. coli RecQ and topo III together can catalyze the linking and unlinking of covalently closed circular DNA molecules, and it has been suggested that such an activity may function in suppressing recombination.

In mammals, two topoisomerase III homologues, TOPO III and TOPO III, have been identified (22-24). Both of these enzymes can relax negatively supercoiled DNA, and TOPO III has been shown to be essential for embryonic development in mice (22, 24, 25). Human TOPO III (hTOPO III) is able to interact physically with Sgs1p when expressed in S. cerevisiae (26), suggesting that the interaction between RecQ helicases and topoisomerase III enzymes has been conserved between eukaryotic species.

In this study, we have examined the possibility that BLM acts in concert with one of the human TOPO III isozymes. Purified BLM and hTOPO III proteins were found to interact directly, and this association was mediated by two regions located in the N- and C-terminal domains of BLM that can independently bind hTOPO III. BLM was also found to genetically interact with yeast TOP3, and this interaction was shown to require the presence of either one of the two hTOPO III interaction domains.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Cell Lines-- The SV40-transformed normal human fibroblast cell line, WI-38/VA-13, which was obtained from the ATCC, was used as a representative of a human cell line from a normal individual. The GM08505 cell line is an SV40-transformed fibroblast cell line from a Bloom's syndrome patient (obtained from NIGMS, National Institutes of Health) and contains a BLM homozygous frameshift mutation resulting in premature truncation of the protein at position 739 (2). PSNB2 cells were derived from a clone of GM08505 cells stably transfected with pcDNA3/BLM, an expression construct containing the full-length BLM cDNA downstream of the CMV promoter. The derivation and characterization of these cells will be described in detail elsewhere.¹ All three cell lines were routinely cultured in -minimum Eagle's medium supplemented with 10% fetal bovine serum. HeLa S3 cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum.

Escherichia coli and S. cerevisiae Strains-- The E. coli BL21(DE3) strain was obtained from NEB. Gene disruptions were done in the S. cerevisiae YP-1 strain (his4-R leu2 MATa-URA3-MATa ura3-52 ade2-101 lys2). The SGS1 open reading frame was replaced with LYS2 as described previously (15). The TOP3 open reading frame was replaced with a kanMX cassette using the method of Wach et al. (27). Transformants were selected on plates containing G418, and the disruption of TOP3 was confirmed using polymerase chain reaction (PCR).²

Plasmids-- The entire BLM open reading frame was cloned between the BamHI and XhoI sites of pcDNA3 (Invitrogen) to generate pcDNA3/BLM. pMAL-C2 (New England Biolabs) and pGEX-4T-1 (Amersham Pharmacia Biotech) plasmids were used for the generation of maltose-binding protein (MBP) and glutathione S-transferase (GST) fusion peptides, respectively. Portions of the BLM cDNA that encoded various N- and C-terminal regions of BLM were amplified by PCR. Sense and antisense primers contained the seven terminal 5' sense and 3' antisense codons, respectively, of each desired fragment. The antisense primer had an additional in-frame antisense stop codon. EcoRI and XhoI sites were also engineered into the sense and antisense primers, respectively, to allow in-frame cloning of the PCR fragments into pMAL-C2 and pGEX-4T-1. pBLM-N, pBLM-C, and pBLM-NC contained residues 213-1417, 1-1265, and 213-1265 of BLM, respectively, and were generated by modification of pJK1, which comprises the entire BLM open reading frame with an additional 18 nucleotides encoding a C-terminal hexahistidine tag cloned into pYES (12). To generate the N-terminal deletion, PCR was used to amplify a 0.7-kb fragment from the BLM cDNA using the 5' primer BLMN6 (5'-GAGAGGTACCTAACCATGTCTGAAAGCGAGCAAATAGA-3') and the 3' primer BLMN7 (5'-TTCATAGAATTCCCTGTAGG-3'). The BLMN6 primer contains a unique KpnI site (underlined) and Kozak sequence in which the ATG is in-frame with Ser-213 of the BLM gene (boldface). The 0.7-kb PCR fragment contains the unique EcoRI site in pJK1 located at nucleotide position 1341 in the BLM cDNA. The BLMN6/7 PCR fragment was digested with KpnI and EcoRI and used to replace the single 1.4-kb KpnI/EcoRI fragment from pJK1 to generate pBLM-N.

To generate the C-terminal deletion, PCR was used to amplify a 0.5-kb fragment from the BLM cDNA using the 5' primer BLMC20 (5'-TGTAGGTCCTTCTGGAAGAT-3') and the 3' primer BLMC21 (5'-GAGATCTAGATCAGTGGTGGTGGTGGTGGTGACCATCAA TTTGAAGCAAAA-3'). The BLMC21 primer contains a unique XbaI site (underlined) and a stretch of 18 nucleotides (boldface) that encodes a hexahistidine tag in-frame to Gly-1265 of BLM. The 0.5-kb PCR fragment contains the unique SalI site in pJK1 located at nucleotide position 3343 of the BLM cDNA. The BLMC20/21 PCR fragments were each digested with SalI and XbaI and used to replace the single 1-kb SalI/XbaI fragment in pJK1 and pBLM-N, to generate pBLM-C and pBLM-NC, respectively.

Antibodies-- For the production of the IHIC33 anti-BLM antibody, rabbits were immunized with six 100-µg injections of an MBP fusion peptide containing residues 1-449 of BLM. The serum was then affinity-purified against the MBP/BLM (1-449) protein. Several mg of the MBP fusion peptide was subjected to SDS-PAGE and transferred to Hybond-ECL filters (Amersham Pharmacia Biotech). Filters were then incubated with 2 ml of serum overnight at 4 °C before being washed four times for 10 min each in TBS. Anti-BLM antibodies were eluted in 1 ml of 0.1 M glycine-HCl, pH 2.5, for 5 min at 20 °C and neutralized by the addition of 675 µl of 1 M Tris-HCl, pH 8.0.

A monoclonal antibody specific for BLM recombinant BLM was raised in Balb/C mice, following immunization with 10 µg of full-length recombinant BLM (28). Fusions were carried out using standard methods (29). Supernatants were screened by antibody capture on multiwell plates coated with recombinant BLM. An antiserum, designated BFL-103, was chosen for study and will be described in detail elsewhere.³

The D6 antibody specific for hTOPO III was raised in New Zealand White rabbits by subcutaneous injection of eight injections of 100 µg each of purified recombinant hTOPO III. The antiserum was used without further purification.

In Vitro Transcription/Translation-- pcDNA3 containing the entire open reading frame of either hTOPO III or hTOPO III was used in the TNT coupled reticulocyte system containing [³⁵S]methionine, following the manufacturer's recommendations (Promega). Half of each reaction was then boiled and separated by SDS-PAGE before being subjected to either autoradiography or Western blotting.

Indirect Immunofluorescence Analysis-- Cells were grown on coverslips and fixed with 4% paraformaldehyde in 250 mM HEPES, pH 7.4, 0.1% Triton X-100 at 4 °C for 20 min. After 3-5 washes over 20 min in PBSA, cells were permeabalized in 0.5% Triton X-100 for 20 min and then washed as before. Blocking in 5% fetal bovine serum in PBSA was followed by 16 h incubation with the primary antibodies BFL-103, IHIC33, or D6, which were used at a 1:50, 1:200, and 1:200 dilution, respectively. 3-5 PBSA washes over 20 min were followed by incubating for a further 30 min with anti-mouse Cy3 (Sigma) or anti-rabbit fluorescein isothiocyanate (Dako) secondary antibody diluted to 1/800 and 1/200 respectively. Cells were washed five times in PBSA, and the DNA was stained using Hoechst 33258 at 50 ng/ml. Stained slides were mounted in 90% glycerol, 20 mM Tris-HCl, pH 8.0, and 50 µg/ml paraphenylenediamine. Slides were viewed at × 100 magnification on a Zeiss Axioskop microscope. Image acquisition and analysis were performed using Kromascan (Kinetic Imaging), and the images were pseudocolored and merged in Adobe Photoshop.

Whole Cell and Nuclear Extracts-- Nuclear extracts were prepared from exponentially growing HeLa S3 cells. Approximately 10⁸ cells were washed in PBSA and the pellet lysed in 5 ml of lysis buffer (10 mM Tris-HCl, pH 7.5, 1.5 mM MgCl₂, 10 mM NaCl, 1% Nonidet P-40 and 1 mM PMSF) on ice for 45 min. The nuclei were then harvested at 5000 × g for 5 min, and the supernatant was discarded. The pellet was then suspended in 0.15 ml of TKM buffer (50 mM Tris-HCl, pH 7.5, 5 mM MgCl₂, 25 mM KCl, and 1 mM PMSF) to which 15 µl of 0.2 M EDTA, pH 8.0, and 2 volumes of Buffer D (80 mM Tris, pH 7.5, 2 mM EDTA, 0.53 M NaCl, 20% glycerol supplemented with 1 mM DTT, 1 mM PMSF, and Complete protease inhibitor mixture (Roche Molecular Biochemicals), used at the manufacturer's recommended concentration), were added, and the mixture was incubated on ice for 30 min. The nuclear extract was then cleared by centrifugation at 14,000 rpm in a microcentrifuge at 4 °C, and the supernatant stored at 70 °C. Whole cell extracts were typically prepared from 2 liters of exponentially growing HeLa S3 cultures. After rinsing of cells in cold PBSA, the cell pellet was resuspended in 3 ml of lysis buffer (10 mM Pipes, pH 6.8, 100 mM NaCl, 300 mM sucrose, 3 mM MgCl₂, 1 mM EGTA supplemented with 0.5% Triton X-100, 1 mM DTT, 1 mM NaF, 1 mM -glycerophosphate, 1 mM sodium orthovanadate, 5 mM sodium pyrophosphate, 1 mM glucose 1-phosphate, 10 nM microcystin, 0.1 mM para-nitrophenylphosphate, 1 mM PMSF, and Complete protease inhibitor mixture (Roche Molecular Biochemicals), used at the manufacturer's recommended concentration), for 5 min at 4 °C. The lysate was then cleared by centrifugation at 5000 × g at 4 °C for 5 min and used on the day of preparation.

Immunoprecipitations and Western Blotting-- Typically, whole cell extracts prepared from 300 ml of exponentially growing HeLa S3 cells were used for each immunoprecipitation. D6 or IHIC33 antibody was added to the extract to give a 1:50 dilution and incubated for 30 min at 4 °C. Antibody was captured by adding to each sample 100 µl of a 50% slurry of protein A-Sepharose (Sigma) in whole cell extract lysis buffer. Samples were then rotated end-over-end for 30 min at 4 °C. Immunoprecipitates were pelleted in a benchtop microcentrifuge for 5 s at 14,000 rpm and washed in 1 ml of cold whole cell extract lysis buffer. The pellet was washed a further four times before being boiled in sample loading buffer (50 mM Tris-HCl, pH 6.8, 100 mM DTT, 2% SDS, 0.2% bromphenol blue, 20% glycerol) for 5 min. Samples were separated on 7.5% SDS-polyacrylamide gels and transferred to Hybond-ECL filters (Amersham Pharmacia Biotech) using a TE 70 semi-dry transfer unit (Amersham Pharmacia Biotech). BLM and hTOPO III were detected by conventional Western analysis using 1:500 dilutions of IHIC33 or D6 as primary antibody, respectively. Anti-rabbit IgG/horse radish peroxidase conjugates (Sigma) were used as secondary antibody at a 1:10,000 dilution and detected using ECL (Amersham Pharmacia Biotech) following the manufacturer's instructions.

Expression and Purification of Recombinant Proteins-- Recombinant BLM and hTOPOIII were expressed and purified as described previously (12, 28, 30). MBP and GST fusion peptides were expressed and purified from BL21 (DE3) cells (New England Biolabs) transformed with the pMAL-C2 or pGEX-4T-1 expression plasmids containing various portions of the BLM cDNA. Overnight cultures of transformants inoculated into 400 ml of LB at a 1:100 dilution were grown at 37 °C to an A₆₀₀ of 0.5-0.6. Isopropyl-1-thio--D-galactopyranoside was then added to a final concentration of 0.4 mM, and the cultures were allowed to grow for a further 2-3 h before being chilled on ice for 30 min. After centrifugation at 10,000 rpm in a Beckman JA10 rotor, the cell pellet was resupended in 10 ml of PBSA supplemented with 1 mM PMSF and Complete protease inhibitor mixture (Roche Molecular Biochemicals) at the manufacturer's recommended concentration. Cells were then lysed by sonication, and the lysate was cleared by centrifugation at 42,000 rpm in a Beckman 70 TI rotor for 30 min at 4 °C. 1 ml of 50% slurries of amylose-agarose (NEB) or GSH-agarose resin (Sigma) were then added to the lysates to capture MBP or GST fusion proteins, respectively, and rotated end-over-end for 30 min at 4 °C. The agarose resin was then pelleted in a benchtop microcentrifuge at 14,000 rpm for 5 s and washed with 1 ml of cold PBSA. The resin was washed a further four times before being boiled in sample loading buffer (50 mM Tris-HCl, pH 6.8, 100 mM DTT, 2% SDS, 0.2% bromphenol blue, 20% glycerol) for 5 min.

Far Western Analysis-- Typically, 0.2-1.0 µg of each polypeptide was subjected to SDS-PAGE and transferred to Hybond-ECL filters (Amersham Pharmacia Biotech) using a TE 70 semidry transfer unit (Amersham Pharmacia Biotech), after which all subsequent steps were performed at 4 °C. Filters were immersed twice in denaturation buffer (6 M guanidine-HCl in PBSA) for 10 min and then incubated six times for 10 min each in serial dilutions (1:1) of denaturation buffer supplemented with 1 mM DTT. Filters were blocked in TBS containing 10% powdered milk, 0.3% Tween 20 for 30 min before being incubated in BLM or hTOPO III (0.5 µg/ml) in TBS supplemented with 0.25% powdered milk, 0.3% Tween 20, 1 mM DTT, and 1 mM PMSF for 60 min. Filters were washed four times for 10 min each in TBS containing 0.3% Tween 20, 0.25% powdered milk. The second wash contained 0.0001% glutaraldehyde. Conventional Western analysis was then performed to detect the presence of BLM or hTOPO III using BFL-103 or D6, respectively, as primary antibody. Anti-rabbit IgG/horse radish peroxidase conjugate (Sigma) was used as secondary antibody at a 1/10,000 dilution and detected using ECL (Amersham Pharmacia Biotech) following the manufacturer's instructions.

Expression of Wild-type and Mutant BLM cDNAs in S. cerevisiae-- Plasmid transformation of yeast strains was carried out by the modified protocol of Gietz et al. (31). Transformants were selected for uracil prototrophy and maintained under selection thereafter. Exponentially growing cultures in 2% glucose were diluted to an A₆₀₀ of 0.1, and 10 µl of each culture was streaked onto agar plates containing 2% glucose or 1% raffinose/0.5% galactose to activate expression from the GAL1 promoter. Plates were then incubated at 30 °C for 2-4 days to allow colony formation. For the preparation of whole cell extracts, 50-ml cultures of transformants were grown in 1% raffinose to an A₆₀₀ of 0.4, at which point 0.5% galactose was added. Cells were then cultured for a further 20 h and pelleted by centrifugation. After rinsing in 50 ml of lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM KCl), the cell pellet was resuspended in 1 volume of lysis buffer, and 1 volume of glass beads was added. Cells were lysed by vortexing for 10 min at 4 °C, after which, 10× SDS loading buffer was added, and the sample was boiled for 5 min. Samples were separated by SDS-PAGE, transferred to Hybond-ECL filters (Amersham Pharmacia Biotech), and then subjected to conventional Western analysis using mouse monoclonal anti-histidine antibody (Dianova) as the primary antibody.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Generation of Anti-BLM and Anti-hTOPO III Antibodies-- Anti-BLM antibodies were raised in rabbits against a chimeric protein consisting of MBP fused to residues 1-449 of BLM. The antiserum chosen for study, designated IHIC33, was affinity-purified using the MBP-BLM fusion protein, as described under "Experimental Procedures." Using Western blotting, IHIC33 recognized purified recombinant BLM (12, 28), and a single major protein in HeLa cell nuclear extracts with a molecular mass of approximately 180 kDa that co-migrated with the recombinant BLM protein (Fig. 1A). No such protein was detected in cell extracts made from the fibroblast cell line, GM08505, which was derived from a patient with Bloom's syndrome and has been shown previously to lack expression of BLM (32) (Fig. 1A). However, extracts from PSNB2 cells (GM08505 cells transfected with pcDNA3/BLM, an expression construct carrying the BLM cDNA) did contain an immunoreactive species of the predicted molecular mass (Fig. 1A), confirming that the IHIC33 antiserum is specific for the BLM protein.

View larger version (30K): [in this window] [in a new window]   Fig. 1.   Characterization of anti-BLM. A, left panel, Western blot using the IHIC33 antibodies against either purified recombinant BLM (rBLM) or a HeLa nuclear extract, as indicated above the lanes. The sizes of molecular mass standards (in kDa) run in parallel are shown on the left. The position of BLM is indicated on the right. Right panel, Western blot using the IHIC33 antibodies against a whole cell extract of GM08505 or PSNB2 cells, as indicated above the lanes. B, Western blot with the BFL-103 antibody against recombinant BLM and purified recombinant proteins comprising MBP alone or MBP fused to the N-terminal (residues 1-447) or C-terminal (residues 966-1417) domains of BLM, as indicated above the lanes.

A monoclonal antibody specific for BLM, designated BFL-103, was raised in mice and detected recombinant BLM by Western blotting (Fig. 1B). Western blotting of different recombinant fragments of BLM revealed that the epitope recognized by BFL-103 resides in the nonconserved, N-terminal domain of BLM (Fig. 1B). Additional validation of this antiserum is described below, in Fig. 4.

Antibodies were also raised using purified recombinant hTOPO III as antigen. Serum from immunized rabbits, designated D6, contained antibodies that recognized the purified recombinant hTOPO III (30) (Fig. 2A). Two degradation products of the hTOPO III protein present in the preparation were also recognized by this antiserum. In HeLa cell extracts, D6 antibodies detected two proteins of approximately 97 and 95 kDa (Fig. 2A). The same two proteins were also detectable in extracts from WI-38/VA-13 cells, a normal human fibroblast cell line, and from GM08505 cells (data not shown and see Fig. 4D, below). Because hTOPO III and hTOPO III share 36% homology and the D6 antiserum was generated using full-length recombinant hTOPO III, it was possible that the reason for the presence of two immunoreactive bands on the Western blot was that the D6 antiserum can cross-react with hTOPO III. To eliminate this possibility, expression vectors containing the hTOPO III and III cDNAs were used to synthesize the two human isozymes in an in vitro transcription/translation system. Using this system, both proteins were produced at comparable levels, as judged by [³⁵S]methionine incorporation, but only hTOPO III was recognized by D6 antiserum (Fig. 2B), confirming that the D6 antiserum is specific for hTOPO III. It is possible, therefore, that the two proteins recognized by the D6 antiserum in cell extracts represent either the full-length and a proteolytically cleaved version of hTOPO III or different posttranslationally modified forms of the protein that exist in human cells.

View larger version (35K): [in this window] [in a new window]   Fig. 2.   Characterization of anti-hTOPO III antibodies. A, Western blot using the D6 antibodies against purified recombinant hTOPO III (left panel) or a HeLa whole cell extract (right panel). The position of hTOPO III is indicated on the right. B, in vitro transcription/translation reactions containing no DNA (lane 1), pcDNA3-hTOPO III (lane 2), or pcDNA3-hTOPO III (lane 3). Left panel, autoradiograph of reaction products. Right panel, Western blot using the D6 antiserum. The positions of the hTOPO III and hTOPO III bands are indicated on the right.

BLM and hTOPO III Exist as a Complex in Human Cells-- Given that Sgs1p and Top3p interact physically in yeast (8), we tested the possibility that BLM may form a complex with topoisomerase III in human cells. Initially we sought evidence that BLM and hTOPO III could be co-immunoprecipitated from HeLa cell extracts. For this, immunoprecipitates were prepared using either IHIC33 (anti-BLM) or D6 (anti-hTOPO III) antibodies. These precipitates were then separated by SDS-PAGE and Western blotted for the presence of hTOPO III and BLM. In D6 immunoprecipitates, two proteins of 97 and 95 kDa that reacted with the D6 antiserum could be detected (Fig. 3, upper panel). The slower migrating form could also be detected in IHIC33 immunoprecipitates, indicating that BLM and at least one of the hTOPO III forms is tightly associated in HeLa cells. This interaction appeared to be specific because the faster migrating D6-immunoreactive species was not present in IHIC33 immunoprecipitates. As expected, Western blotting with IHIC33 revealed BLM to be present in the IHIC33 immunoprecipitate (Fig. 3, lower panel). However, BLM was also detected in the D6 immunoprecipitate (Fig. 3, lower panel), indicating that the complex of BLM and hTOPO III can be captured using either antibody.

View larger version (56K): [in this window] [in a new window]   Fig. 3.   BLM and hTOPO III can be co-immunoprecipitated from HeLa cell extracts. Immunoprecipitates were prepared using either anti-BLM (IHIC33) or anti-hTOPO III (D6) antibodies as indicated above the lanes. After separation by SDS-PAGE and transfer to nitrocellulose filters, immunoprecipitates were probed for the presence of either hTOPO III, using the D6 antibody (upper panel), or BLM, using the IHIC33 antibody (lower panel).

The co-immunoprecipitation of BLM and hTOPO III from human cell extracts is consistent with these proteins forming a complex. To provide additional evidence for this, we analyzed whether BLM and hTOPO III co-localize in the nucleus of intact human cells. In order for co-staining studies to be undertaken without the requirement to directly conjugate antibodies to fluorochromes, the BFL-103 mouse monoclonal antibody to BLM and the D6 rabbit polyclonal antibody to hTOPO III were used for this part of the study. As a first step, however, we wanted to confirm that the staining patterns obtained with the IHIC33 and BFL-103 anti-BLM antibodies were the same. Indirect immunofluorescence of exponentially growing WI-38/VA-13 cells using either IHIC33 or BFL-103 revealed BLM to be localized to prominent nuclear foci in many of the cells (Fig. 4A). This was in accordance with the results of Neff et al. (32). Merging the fluorescent signals for BLM obtained with the BFL-103 and IHIC33 antisera revealed a very strong concordance (Fig. 4A).

View larger version (20K): [in this window] [in a new window]   Fig. 4.   A, BLM localizes to nuclear foci in WI-38/VA-13 cells. The foci detected with the anti-BLM BFL-103 monoclonal (red) and IHIC33 polyclonal (green) antisera co-localize, as revealed in the merged image. B, co-localization of BLM (red) and hTOPO III (green) in the nucleus of WI-38/VA-13 cells. C, lack of staining for BLM and a reduced number of hTOPO III foci in the nucleus of GM08505 cells (upper panel) compared with PSNB2 cells (lower panel). Nuclear DNA was revealed by staining with Hoechst 33258. D, upper panel, Western blot using the D6 antibody against whole cell extracts of GM08505 and PSNB2 cells, as indicated above the lanes. The position of the hTOPO III protein is indicated on the right. The lower panel shows the same blot probed with anti--tubulin antibodies to ensure uniform loading.

Analysis of the subcellular localization of hTOPO III also revealed a punctate pattern of nuclear staining (Fig. 4B). Merging the signals for BLM and hTOPO III proteins also indicated a strong concordance, indicating that BLM and hTOPO III co-localize in human cell nuclei. As expected from the Western blotting data described in Fig. 1, BLM staining using the BFL-103 antiserum was absent from Bloom's cell line, GM08505 (Fig. 4C), providing additional evidence that BFL-103 is specific for BLM. In contrast, staining of PSNB2 cells (GM08505 containing the BLM cDNA) revealed that BLM resides in nuclear foci that co-localized with hTOPO III. BLM was also found distributed in the nucleoli of PSNB2 cells, a staining pattern that was also seen in nontransfected cells, albeit to a quantitatively lesser extent than in PSNB2 cells. This striking nucleolar staining in PSNB2 cells was presumably due to the overexpression of BLM from the CMV promoter in this line. Interestingly, although hTOPO III staining was still evident in the Bloom's syndrome mutant cell line, the number of nuclear foci detected using the D6 antibody was consistently and substantially reduced compared with that seen in PSNB2 cells (Fig. 4C). This could have been due to one of two reasons: either hTOPO III is mislocalized in GM08505 cells, or hTOPO III is stabilized in PSNB2 cells, perhaps through its association with BLM. This latter possibility was excluded, however, because Western analysis using the D6 antiserum on whole cell extracts from GM08505 and PSNB2 cells revealed similar levels of hTOPO III (Fig. 4D). This suggests that the correct localization of hTOPO III to nuclear foci is at least partially dependent upon the presence of BLM. Together, these data strongly suggest that BLM and hTOPO III exist as a complex in vivo.

BLM and hTOPO III Interact Directly in Vitro-- We next wanted to determine whether the interaction between BLM and hTOPO III was a direct one, because in vivo the interaction may be mediated via an accessory protein. Far Western analysis was therefore performed to determine whether purified recombinant BLM and hTOPO III could interact directly in vitro. Essentially, this procedure involves the immobilization of one of the proteins of interest to a nitrocellulose filter, which is then incubated in buffer containing the second protein. The filter is then washed to remove any unbound material, and the presence of the second protein is detected by conventional Western analysis. When BLM was used to probe hTOPO III bound to nitrocellulose, a BFL-103-immunoreactive band could be detected at the position where hTOPO III would be expected to migrate (Fig. 5). This was due to BLM binding to hTOPO III, as opposed to cross-reactivity of the BFL-103 antibody with hTOPO III, because this band was absent on an identical, control blot that had been incubated in TBS alone but had otherwise been treated in the same way (Fig. 5). The interaction between BLM and hTOPO III appeared to be specific because BLM did not bind to either of two control proteins, MBP and bovine serum albumin (BSA), which were run in parallel on the same blot. To eliminate the possibility that recombinant hTOPO III was inherently "sticky," the reciprocal experiment was carried out, wherein hTOPO III was used to probe nitrocellulose-bound BLM. A D6-immunoreactive band, which was absent on the control blot, was found migrating at the position of BLM, confirming that BLM and hTOPO III interact directly (Fig. 5). Again, the interaction was specific because no signal was detected with either of two control proteins, MBP and GST (Figs. 5 and 6B). A second D6-immunoreactive band found migrating at the position of BSA was not due to interaction between BSA and hTOPO III (Fig. 5). Rather, this latter band was due to low level cross-reactivity of the D6 antibodies with BSA because the same band with equal intensity was also present on the control blot that had been incubated with TBS alone and hence had not been exposed to hTOPO III (Fig. 5). We conclude that purified recombinant BLM and hTOPO III interact directly in vitro.

View larger version (19K): [in this window] [in a new window]   Fig. 5.   Purified recombinant BLM and hTOPO III interact directly. Upper panel, purified BLM, hTOPO III, MBP, and BSA were subjected to SDS-PAGE and stained with Coomassie Blue. Lower panel, the proteins were transferred to nitrocellulose filters and incubated with either purified BLM or hTOPO III, as indicted above the four images. Conventional Western analysis was then used to detect the presence of either BLM, using the BFL-103 antibody, or hTOPO III, using the D6 antibody, as indicated above the images.

View larger version (18K): [in this window] [in a new window]   Fig. 6.   Mapping of the hTOPO III interaction domains on BLM. A, schematic representation of the 1417-amino acid BLM protein indicating the position of the conserved helicase domain. Below this is depicted a series of fusion proteins comprising either MBP or GST fused to the indicated portions of BLM. The numbers represent the amino acid positions in BLM that form the C-terminal residue (for the MBP fusions) or either the C-terminal or the N-terminal residue (for the GST fusions) of each BLM fragment. MBP is depicted as a black box, and GST is depicted as a stippled box. The right column indicates the ability of each peptide to bind to hTOPO III as detected by far Western analysis. B, far Western analysis showing the shortest BLM fusion proteins (with both MBP and GST) that bind hTOPO III (lanes 1, 5, and 6), the longest MBP fusion peptide that is negative for hTOPO III binding (lane 2), and the MBP and GST tags alone as controls (lanes 3 and 4). Lanes 2 and 6 also contain some degradation products of the respective fusion proteins. The left panels show Coomassie Blue-stained gels of the various proteins, and the right panels show the same proteins transferred to nitrocellulose and subjected to far Western analysis using hTOPO III as a probe. Bound hTOPO III was detected using the D6 antibody.

BLM Contains Two hTOPO III Interaction Domains-- We next investigated which region of the BLM protein was responsible for mediating interactions with hTOPO III. The BLM protein essentially consists of three domains: a central conserved helicase domain found in all RecQ helicases, flanked by divergent N- and C-terminal regions (2, 3). In yeast, Top3p has been shown to interact with the N-terminal domain of Sgs1p (8). We therefore analyzed whether the N-terminal domain of BLM was required for interaction with hTOPO III. Recombinant fusion peptides were generated that consisted of portions of the N-terminal domain of BLM fused to the C terminus of MBP to aid affinity purification. These fusion peptides were then transferred to nitrocellulose and tested for their ability to bind hTOPO III using far Western analysis. hTOPO III was found to bind to a peptide containing residues 1-447 of BLM (Fig. 6A). Analysis of further truncation derivatives of BLM revealed that residues 1-212 were sufficient to direct binding of hTOPO III (Fig. 6, A and B, upper panel). To eliminate the possibility of an artifact caused by the binding of hTOPO III to the junction where MBP is fused to the BLM N-terminal domain, a second peptide was generated consisting of GST fused to residues 1-212 of BLM. Switching the tag to GST in this way had no effect on the extent of hTOPO III binding, confirming that the hTOPO III interaction domain resides solely in those residues derived from the BLM protein (Fig. 6B, lower panel). An MBP fusion peptide that contained residues 1-142 of BLM did not bind to hTOPO III, indicating that residues 143-212 of BLM are important for the formation of the hTOPO III binding site (Fig. 6B, upper panel).

To examine the possibility of additional sites of interaction between BLM and hTOPO III, far Western analysis was performed using the C-terminal domain of BLM. A GST fusion peptide containing residues 966-1417 of BLM was found to interact with hTOPO III (Fig. 6A). The location of this second site of interaction was further mapped by analysis of various truncated derivatives of the C-terminal domain and was found to reside between residues 1266-1417 (Figs. 6, A and B, lower panel). Hence, two independent interaction domains for hTOPO III are present in the BLM protein.

BLM and Yeast TOP3 Genetically Interact-- Mutation of the SGS1 gene, encoding the sole S. cerevisiae RecQ homologue, suppresses the slow growth and hyperrecombination phenotype of top3 mutants (8). This genetic interaction indicates that Sgs1p and Top3p act in the same biochemical pathway and suggests that Sgs1p generates a DNA structure that if not resolved by Top3p leads to slow growth (8). We therefore analyzed whether BLM could also genetically interact in this way with yeast TOP3. In an sgs1 top3 double mutant, complementation of sgs1 by plasmid-encoded Sgs1p should result in the appearance of a slow growth phenotype, reminiscent of a top3 mutant. This effect has indeed been demonstrated previously by others following ectopic expression of not only Sgs1p, but also BLM (16, 32, 33). In our study, we also found that expression of the BLM protein from a galactose-inducible promoter induced slow growth in sgs1 top3 cells (Fig. 7A). However, expression of BLM in an sgs1 mutant (Fig. 6A) or in wild-type cells (Fig. 7B) had little effect on growth, indicating that this BLM-induced effect was dependent on the absence of functional Top3p.

View larger version (35K): [in this window] [in a new window]   Fig. 7.   Genetic interaction between BLM and yeast TOP3. A, galactose-inducible expression plasmids containing either BLM or various BLM mutant cDNAs (see text for details) were transformed into an sgs1 top3 strain (upper square panels) or an sgs1 strain (lower square panels). As a control, both strains were also transformed with the empty vector. Shown is the growth of each transformant on plates containing either 2% glucose, which represses plasmid expression (left panels), or 0.5% galactose, which induces GAL1 promoter expression (right panels). B, wild-type yeast were transformed with galactose-inducible expression plasmids containing either the BLM cDNA, the mutant BLM-NC cDNA, or an empty vector. Composition of plates was as in A. C, analysis of expression levels for the BLM and BLM-NC proteins in an sgs1 strain. The left panel shows protein extracts stained with Coomassie Blue to indicate that the protein loadings were equivalent. The right hand panel shows a Western blot of the same extracts using an anti-His-tag antibody, because both cDNAs encode a C-terminal hexahistidine tag. Lane 1, strain harboring the pYES2 vector alone; lane 2, strain harboring the BLM cDNA; lane 3, strain harboring the BLM-NC cDNA.

The hTOPO III Interaction Domains Are Required for the Genetic Interaction between BLM and Yeast TOP3-- We next tested the possibility that the genetic interaction between BLM and yeast TOP3 requires the hTOPO III interaction domains on BLM. Galactose-inducible expression constructs were generated containing cDNAs encoding mutant BLM proteins that lack the N-terminal (BLM-N), C-terminal (BLM-C), or both (BLM-NC) hTOPO III interaction domains. Because these modifications to BLM involved truncations that could have global effects on protein folding, it was important to confirm that the mutant proteins were still functionally active. It has been shown previously that expression of BLM cDNAs containing missense, helicase-inactivating mutations are not functional when expressed in sgs1 top3 cells (32). Similarly, in our study, the ability to induce slow growth in sgs1 top3 cells was used as an indicator of BLM function. As with wild-type BLM, we found that all three of the truncated BLM proteins induced slow growth when expressed in sgs1 top3 cells (Fig. 7A), demonstrating that loss of either or both of the hTOPO III interaction domains on BLM does not adversely affect BLM activity in yeast. However, although expression of wild-type BLM was tolerated in wild-type and sgs1 strains, expression of the BLM-NC mutant protein was still able to induce slow growth in sgs1 cells, albeit to a slightly lesser extent than it was in sgs1 top3 cells (Fig. 7A). Hence, the deleterious effect induced by BLM-NC is largely independent of TOP3 status, indicating that BLM-NC and TOP3 do not genetically interact. In contrast, expression of the mutant BLM-N and BLM-C proteins, which each lack only one of the hTOPO III interaction domains, had little effect in sgs1 cells (Fig. 7A). Therefore, as with wild-type BLM, BLM-N and BLM-C induce slow growth that is dependent upon the absence of Top3p, indicating that either interaction domain is sufficient for establishing a genetic interaction between BLM and yeast TOP3.

Taken together, these data suggest that the hTOPO III interaction domains target yeast Top3p to a BLM-induced DNA structure (which, if not resolved, is detrimental to the cell), probably via a direct interaction with BLM itself. However, there are two alternative explanations for the results described above. First, the BLM-NC-induced slow growth in sgs1 cells is not due to uncoupling of the activities of BLM and Top3p; rather, the mutant BLM-NC protein has an additional, dominant-negative effect that acts via a mechanism independent of the pathway in which Sgs1p and Top3p act. Second, loss of both hTOPO III interaction domains has a stabilizing effect on the protein, resulting in the accumulation of higher levels of active BLM helicase and hence an increase in the number of BLM-generated deleterious DNA structures. These two alternative explanations were excluded, however, by the following observations: (i) expression of BLM-NC had little effect in wild-type cells (Fig. 7B), indicating that the BLM-NC protein only induces slow growth in the absence of Sgs1p and is unlikely, therefore, to be acting through a process independent of the pathway in which Sgs1p and Top3p act; and (ii) BLM-NC was expressed at a comparable level to that of full-length wild-type BLM (and possibly at an even lower level, given that approximately 50% of the wild-type protein in the extracts was found to be partially degraded) (Fig. 7C).

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

We have shown that the Bloom's syndrome gene product, BLM, forms a direct physical association with the human topoisomerase III isozyme, and we have identified two independent sites of interaction on BLM. Deletion of both sites was found to severely compromise the ability of BLM to functionally interact with topoisomerase III in S. cerevisiae. Increasing evidence suggests that the biochemical functions of RecQ helicases in different species are intimately associated with those of topoisomerase III enzymes (3), because the genetic interaction between Sgs1p and Top3p, first revealed in S. cerevisiae (8), has been shown to be conserved in the distantly related fission yeast S. pombe. We have described elsewhere that mutations in the gene encoding the only known fission yeast RecQ homologue, rqh1⁺, can rescue the lethality caused by deletion of top3⁺ (34). Here, we have shown a similar genetic interaction exists between one of the human RecQ homologs, BLM, and budding yeast TOP3. In common with Sgs1p and Top3p, which form a complex in budding yeast (8), BLM also directly interacts with hTOPO III, and a complex containing both proteins can be detected in vivo.

The interaction between Top3p and Sgs1p is mediated by the N-terminal domain of Sgs1p (8). Similarly, hTOPO III was also found to interact with the N-terminal domain of BLM, suggesting that the functional domain organization of these two helicases may be similar, despite the lack of extensive primary sequence conservation between their respective N-terminal domains. A second hTOPO III interaction domain was also identified in the extreme C-terminal portion of BLM. Interestingly, although both hTOPO III interaction domains map to the nonconserved regions of BLM, either one was sufficient for BLM to genetically interact with S. cerevisiae TOP3. There may, therefore, be conserved features of the secondary and tertiary structure of the topoisomerase III interaction domains of BLM and Sgs1p that are not evident by analysis of their amino acid sequences. It is not known whether the C-terminal domain of Sgs1p is also involved in binding to Top3p.

The ability of mutations in SGS1 to suppress the slow growth phenotype of top3 mutants has been interpreted previously to suggest that the Sgs1 helicase generates a DNA structure that must be acted upon by Top3p in order to prevent accumulation of an as yet unidentified, toxic intermediate (8). The requirement for either of the hTOPO III interaction domains to be present on BLM in order for the genetic interaction between BLM and yeast TOP3 to be maintained suggests that Top3p is recruited to its sites of action via a direct interaction with the BLM protein itself. The fact that hTOPO III appeared to be mislocalized in GM08505 cells is also consistent with a role for BLM in targeting hTOPO III to its correct site of action.

The presence of two independent hTOPO III interaction domains on BLM raises the possibility that BLM and hTOPO III interact with a 1:2 stoichiometry, which has mechanistic implications for how topoisomerase III might resolve a RecQ helicase-generated DNA structure. Topoisomerase III is a type I topoisomerase and therefore only makes single-stranded DNA nicks. However, the recruitment of two topoisomerase III molecules to a DNA structure by a RecQ helicase could, in principle, be used to generate a double-stranded DNA break, with each topoisomerase III molecule making a nick in close proximity on opposite strands of the duplex. The possibility that topoisomerase III can, under certain circumstances, act coordinately to cleave both strands of a DNA duplex is consistent with the recent observations of Harmon et al. (21). These authors showed that together, the E. coli RecQ and Top3 proteins can catalyze the passage of double-stranded DNA through a break in a second, covalently closed double-stranded DNA molecule. It has been suggested that such an activity could act either to decatenate newly replicated daughter DNA molecules prior to cell division or to disrupt early recombination intermediates between inappropriately paired DNA molecules (21, 35).

Bloom's, Werner's, and Rothmund-Thomson syndromes are clinically distinct entities indicating that the human RecQ homologues perform at least some nonoverlapping functions within the cell. sgs1 mutants are to a certain extent a phenocopy of both Bloom's and Werner's syndromes. BLM mutants display hyperrecombination and have DNA replication abnormalities (1, 3). Similarly, sgs1 mutants also display hyperrecombination (8, 15) and are sensitive to the ribonucleotide-reductase inhibitor hydroxyurea (16), suggesting that they have some defect in replication. In common with the phenotype of WRN mutants, sgs1 mutants have a reduced replicative life span and display, prematurely, several markers of aging, such as sterility and redistribution of Sir proteins from telomeres to the nucleolus (36). The Sgs1p and WRN proteins also both localize to the nucleolus (36-38), suggesting a conservation of function in rDNA metabolism. Furthermore, expression of either BLM or WRN can partially suppress the hyperrecombination phenotype of sgs1 mutants (16). It is possible, therefore, that the Sgs1/Top3 interaction has been conserved not only between BLM and hTOPO III but also between other RecQ family helicases in humans and one of the topoisomerase III isozymes. It is likely that the evolutionarily divergent N- and C-terminal domains play a key role in functionally distinguishing the mammalian RecQ helicases through either directing additional enzymatic functions or mediating specific protein-protein interactions. For example, a DNA exonuclease function resides in the N-terminal domain of WRN (39, 40), an activity that is apparently absent from the other human RecQ homologs. The fact that the two hTOPO III interaction domains map to nonconserved regions in BLM might suggest that the interaction between BLM and hTOPO III is specific for this pair of proteins. However, hTOPO III has been shown to interact with Sgs1p when expressed in yeast (26), suggesting that in human cells hTOPO III is also likely to interact with one or more of the RecQ helicases. Therefore, it is possible that although the ancestral recQ gene has undergone several duplication and functional diversification events during evolution, the RecQ helicases may all share a common mechanism of action in that they act in concert with a topoisomerase III partner. Hence, the phenotype of RecQ helicase mutants may be at least partially the consequence of a functional impairment of topoisomerase III. The fact that mutations in at least three of the known human RecQ helicases give rise to cancer-prone disorders therefore raises the possibility that the gene encoding hTOPO III may also be a tumor suppressor gene. Ongoing work is aimed at addressing this possibility.

	ACKNOWLEDGEMENTS

We thank Julia Karow for the BLM protein, Guy Debousker for hTOPO III protein, Adele Goodwin and Jonathan Kearsey for yeast strains, Connie Wilson for preparing the manuscript, and members of the ICRF Genome Integrity Group for helpful discussions.

	Note Added in Proof

Johnson et al. (Johnson, F. B., Lombard, D. B., Neff, N., Mastrangelo, M.-A., Dewolf, W., Ellis, N. A., Marciniak, R. A., Yin, Y., Jaenisch, R., and Guarente, L. (2000) Cancer Res., in press) have also identified an interaction between BLM and topoisomerase III protein.

	FOOTNOTES

^* This work was supported by the Imperial Cancer Research Fund and Rhone-Poulenc Rorer.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.

To whom correspondence should be addressed. Tel.: 44-1865-222417; Fax: 44-1865-222431; E-mail: hickson@icrf.icnet.uk.

¹ P. S. North et al., manuscript in preparation.

³ H. Turley et al., manuscript in preparation.

	ABBREVIATIONS

The abbreviations used are: PCR, polymerase chain reaction; BSA, bovine serum albumin; DTT, dithiothreitol; kb, kilobase(s); GST, glutathione S-transferase; MBP, maltose-binding protein; PBSA, phosphate buffered saline; PAGE, polyacrylamide gel electrophoresis; Pipes, 1,4-piperazinediethanesulfonic acid; PMSF, phenylmethylsulfonyl fluoride; TBS, Tris-buffered saline.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES