Pott's disease (spinal tuberculosis), a condition characterized by massive resorption of the spinal vertebrae, is one of the most striking pathologies resulting from local infection with Mycobacterium tuberculosis (Mt; Boachie-Adjei, O., and R.G. Squillante. 1996. Orthop. Clin. North Am. 27:95–103). The pathogenesis of Pott's disease is not established. Here we report for the first time that a protein, identified by a monoclonal antibody to be the Mt heat shock protein (Baird, P.N., L.M. Hall, and A.R.M. Coates. 1989. J. Gen. Microbiol. 135:931–939) chaperonin (cpn) 10, is responsible for the osteolytic activity of this bacterium. Recombinant Mt cpn10 is a potent stimulator of bone resorption in bone explant cultures and induces osteoclast recruitment, while inhibiting the proliferation of an osteoblast bone–forming cell line. Furthermore, we have found that synthetic peptides corresponding to sequences within the flexible loop and sequence 65–70 of Mt cpn10 may comprise a single conformational unit which encompasses its potent bone-resorbing activity. Our findings suggest that Mt cpn10 may be a valuable pharmacological target for the clinical therapy of vertebral tuberculosis and possibly other bone diseases.

Tuberculosis is epidemic, accounting for 7% of the annual worldwide death toll (1). Tuberculous infections of bone, particularly of the spinal vertebrae (Pott's disease), are still common in the third world (2). It is not known how Mycobacterium tuberculosis (Mt)1 infections of bone cause bone breakdown. Healthy bone is maintained by a dynamic equilibrium between the mesenchymal bone matrix–forming osteoblast cell lineage and the myeloid bone– resorbing osteoclast cell lineage (3). Mt infection of the spine obviously alters this dynamic equilibrium, resulting in the net loss of the extracellular matrix of vertebral bone and collapse of the vertebrae. Whether this loss of bone matrix is the result of the direct action of components of Mt, for example, the LPS-like cell surface molecule lipoarabinomannan (LAM) (4), on bone cells, or an indirect activation of inflammatory cells leading to bone cell activation, is not established. Evidence is appearing to suggest that molecular chaperones have biological actions in addition to their intracellular protein–folding activity (5). For example, chaperonin (cpn)10 has been found to be an essential growth and immunosuppressive factor in early pregnancy (6), and cpn60 induces cytokine synthesis (7) and resorption of bone (8).

In this study we have established that the bone resorbing activity of Mt is due to cpn10 which is as active as the most potent osteolytic cytokine, IL-1 (9, 10). Mt cpn10 also appears to induce the recruitment of osteoclasts in calvaria, and it is notable that calvarial bone resorption induced by this cpn can be completely blocked by the osteoclast-inhibiting hormone, calcitonin (11). Mt cpn10 was also found to inhibit the proliferation of cultured osteoblasts.

Using a series of NH2- and COOH-terminal truncated peptides, we have identified sites in Mt cpn10 responsible for the osteolytic activity of this molecular chaperone. We have identified the flexible loop of Mt cpn10 and the sequence 65–70 as regions most probably responsible for the bone-modulating bioactivity of this molecule.

Materials And Methods

Mycobacterial Sonicate.

The sonicate was prepared by sonicating a suspension of viable virulent Mt (strain H37Rv) at 4°C for 1 min intervals, followed by a 1 min rest period, for a total period of 1 h. The sonicated material was then centrifuged at 100,000 g for 1 h, and the supernatant was filtered through a 0.22-μm membrane filter.

mAbs.

Both the mAb to Mt cpn10 (SA12; 12) and the mAb to Mt cpn60 (TB78; reference 13) were obtained from murine ascites in a sufficiently high titer to bind to Mt cpn10 or cpn60 at the dilutions used in this study. SA12 is specific for mycobacterial cpn10 and TB78 is specific for mycobacterial cpn60. Neither of these mAbs are cross-reactive with any other Mt protein (14). The mAb to LAM (CS-35) of isotype IgG3 was obtained from concentrated tissue culture supernatant with a titer of 1:2,000 by Western blot analysis (Belisle, J.T., personal communication). CS-35 was raised against Mycobacterium leprae LAM and is cross-reactive with Mt LAM at a dilution of 1:1,000 by Western blot analysis (15). CS-35 was used at a 1:1,000 dilution in the bone resorption assay.

Mt cpn10 Peptides.

r-Mt cpn10 was expressed in Escherichia coli and purified by reversed-phase HPLC to >97% purity as previously described (16). The synthetic peptide fragments were prepared and purified by isoelectric focusing and by reversed phase HPLC to >95% purity as previously described (17). Before addition to the bone explants, r-Mt cpn10 was passed down a Polymyxin B–agarose column (Detoxigel column; Pierce, Rockford, IL) to remove any contaminating LPS. The composition of the peptides was confirmed by amino acid analysis and mass spectroscopy. All peptides were tested for LPS using the limulus amoebocyte lysate assay (Whittaker M.A. Bioproducts, Inc., Walkersville, MD). All peptides tested negative, indicating the presence of <0.03 endotoxin U LPS.

Calcium Release and Osteoclast Recruitment in Murine Calvaria.

The calvarial bone resorption assay was performed as described (18). In brief, calvaria were removed from 5-d-old MF1 mice, adherent connective tissue was dissected away, and the calvarial bone was halved, with each half being cultured separately on stainless steel grids. Calvaria were cultured in groups of 5 replicates in 30-mm dishes with 1.5 ml Biggers, Gwatkin, and Jenkins medium (ICN Biomedicals, Inc., Thame, UK) containing 5% heat-inactivated rabbit serum (GIBCO BRL, Paisley, UK) and 50 μg/ml ascorbic acid (Sigma Chemical Co., Poole, UK). After 24 h in culture, the media was replaced with media containing various concentrations of sonicated Mt, r-Mt cpn10 or Mt cpn10 peptides with or without a range of concentrations of mAb to Mt cpn10 (SA12; reference 12), mAb to Mt cpn60 (TB78; reference 13), mAb to Mt LAM at 1:1,000 dilution. Calvaria were cultured for a further 48 h and then the calcium released into the medium was measured by automated colorimetric analysis (19).

After removal of the media supporting the calvarial explants for calcium assay, the explants were then used for the measurement of osteoclast numbers by a modification of the method of Marshall et al. (20). The calvaria were fixed in 95% ethanol, 5% glacial acetic acid for 24 h, and were then washed in PBS containing 1 mg/ml BSA followed by reaction in the histochemical substrate mixture (obtained from Sigma Chemical Co., and used according to the manufacturer's instructions) for the localization of tartrate-resistant acid phosphatase (TRAP) activity, a marker enzyme for osteoclasts. Bones were then washed in PBS, decalcified, and fixed in 12.5% glutaraldehyde (BDH Chemicals, Ltd., Poole, UK) in 1 M hydrochloric acid (BDH Chemicals, Ltd.) for 5 min. Finally, the bone explants were washed and mounted in Aquamount (BDH Chemicals, Ltd.). Each calvarial explant was then scanned by transmitted light microscopy and TRAP-positive cells containing three or more nuclei were counted. The individual counting the cells was unaware of the treatment to which each explant had been exposed. Control cultures included unstimulated calvaria (to determine spontaneous release of calcium) and calvarial cultures stimulated with 1 μM prostaglandin (PG; to demonstrate that bone is responsive and to give a measure of the maximal response). r-Mt cpn10 and Mt cpn10 peptides were tested in a minimum of three experiments and gave reproducible results.

Osteoblast Proliferation.

The measurement of cell proliferation was as previously described (21). In brief, the human osteoblast-like cell line MG63 (CRL 1427; American Type Culture Collection, Rockville, MD) was cultured at a density of 15,000 cells/well in 96-well plates and incubated overnight at 37°C in DMEM (Gibco) plus 10% FCS (Sigma Chemical Co.) in 5% CO2/air. The media were then removed and cells were washed twice with sterile Hank's solution (Sigma Chemical Co.). To measure antiproliferative activity, various concentrations of r-Mt cpn10 or truncated peptides were added in DMEM containing 2% FCS, to the MG 63 cells. Cells were incubated for 24 h at 37°C. During the last 6 h of culture, 0.05 μCi of [3H]thymidine (Amersham International plc, Amersham, UK) was added to cells. The media were then removed and the cells fixed in 5% trichloroacetic acid. 100 μl of 0.5 M NaOH was used to lyse cells, this being neutralized by an equal volume of 0.5 M HCl. Radioactivity incorporated into nuclear DNA was measured by scintillation spectrometry. The cytotoxicity of the r-Mt cpn10 was determined by lactate dehydrogenase release, measured by the CytoTox 96 nonradioactive cytotoxicity assay (Promega, Heidelberg, Germany). Data has been generated from a minimum of three separate experiments.

Homology Modeling of Mt cpn10.

The Mt cpn10 model was generated by homology modeling of the Mt cpn10 sequence onto the atomic coordinates of the GroES structure of the monomer with the flexible loop assigned. The model was energy minimized with QUANTA/CHARMm (Molecular Simulations, Inc., San Diego, CA) using a nonbonded cutoff of 14 Å and a dielectric constant distance dependence until the root mean squared deviations were <0.001 Kcal/Å. The side chains were minimized first, keeping the backbone fixed. This was followed by minimization of the whole monomer. The heptamer was generated from the monomer by the symmetry operations relating the GroES subunits. The same energy minimization procedure was repeated for the final Mt cpn10 heptamer model, which was displayed with the SYBYL molecular modeling package (Tripos UK, Milton Keynes, UK).

Results And Discussion

Sonicates of viable Mt added to explants of murine calvarial bone produced a dose-dependent stimulation of bone resorption, measured as calcium release into the tissue culture medium. Osteoclast numbers in calvarial explants were counted and showed a parallel increase (Fig. 1,A). The Mt sonicate–induced stimulation of bone resorption was dose dependently and completely inhibited by a neutralizing mAb to Mt cpn10 (SA12; reference 12), but not by a subclass-matched neutralizing mAb to Mt cpn60 (TB78; reference 13). Likewise, the mAb SA12 caused a dose-dependent decrease in the numbers of osteoclasts present in the calvarial explants (Fig. 1,B). In contrast, SA12 had no effect on the stimulation of bone resorption induced by PG (Fig. 1 C). Purified Mt LAM, added at a concentration of 1 μg/ ml, had no osteolytic activity, and neutralizing mAb to LAM did not inhibit the bone resorption induced by the Mt sonicate (results not shown). Addition of polymyxin B had no effect on the bone resorbing activity of the Mt sonicate.

Purified r-Mt cpn10 caused a dose-dependent stimulation of calcium release from cultured calvaria with osteolytic activity being noted at a concentration of 1 ng/ml (equivalent to 100 pmol) that was reproducible and statistically significant (P < 0.01; Fig. 2,A). The bone resorbing activity of Mt cpn10 was dose dependently and completely inhibited by mAb SA12 (Fig. 2 B). The osteoclast-inactivating hormone, calcitonin, at a concentration of 10 ng/ml, also blocked r-Mt cpn10-induced bone resorption (results not shown). Addition of polymyxin B had no effect on the bone resorbing activity of r-Mt cpn10.

Addition of r-Mt cpn10 to subconfluent cultures of the human osteoblast-like cell line MG63 caused significant inhibition of cell proliferation at concentrations ⩾1 nM (Fig. 3). Inhibition of proliferation was not due to cytotoxicity of the r-Mt cpn10.

A panel of 11 NH2- and COOH-terminal truncated peptides and short peptides (16) were used to define the specific structural features of r-Mt cpn10 responsible for its osteolytic and osteoblast antiproliferative activities. These peptides corresponded to residues 1–25, 1–58, 26–99, 46– 99, 51–99, 54–99, 59–99, 65–99, 71–86, 75–99, and 91–99. Graded concentrations of each peptide were tested separately in each assay in three separate experiments. 2 of these 11 peptides, 26–99 and 65–99, exhibited reproducible osteolytic activity (Fig. 4). Polymyxin B had no inhibitory effects on the activity of these peptides. Peptide 26–99 contains sequences that are within the flexible loop region of Mt cpn10 (residues 16–35). To determine if this flexible loop contributed to the osteolytic activity, two short peptides within the flexible loop in Mt cpn10 (21–35: TTTASGLVIPDTAKE) and in the E. coli cpn10 (GroES residues 23–33: GGIVLTGSAAA) were synthesized and were also found to have osteolytic activity in the calvarial assay (Fig. 4). Mt, unlike E. coli, is able to secrete extracellular cpn10 (22), which has important implications for the pathogenic effects of Mt cpn10 in vivo.

All 12 Mt cpn10 peptides were repeatedly tested for antiproliferative activity but even at very high concentrations, none showed any ability to inhibit osteoblast proliferation.

Peptide 1–58 was inactive in the bone resorption assay, although it contains the predicted flexible loop. The most likely explanation is that the structure of 1–58 differs from that of whole protein, because peptide 1–58 is a dimer (16), the aggregation of which is unusual as it occurs via the NH2-terminal region in contrast to the whole protein in which contact between two neighboring protomers involves the COOH-terminal tail of one protomer and the NH2-terminal region of the other (23).

Peptides 46–99, 51–99, 54–99, and 59–99 were also inactive, although they contain the active 65–99 sequence, and again, structural differences are the likely explanation for this discrepancy. For example, the structure of the inactive peptide 59–99 has been assigned to that of four antiparallel β strands (24), but circular dichroism spectroscopy data with the active peptide 65–99 (data not shown) indicate that the latter is mainly composed of the random coil conformation. Peptide 26–99 is active in the bone resorption assay since most of the mobile loop is part of its NH2-terminal tail and is probably accessible to solvent (and hence a receptor) as often happens to the NH2-terminal and COOH-terminal regions of polypeptides and proteins. For the same reason, amino acids 65–70 would be considered the active sequence in peptide 65–99.

A molecular model of heptameric Mt cpn10 was derived from the E. coli cpn10 crystal structure (reference 23; Fig. 5). The sequences derived from the peptide data which contribute to the osteolytic activity are colored red and correspond to the flexible loop (21–35) at the bottom outer edge of the heptamer which is in close proximity to the sequence 65–70. Although the flexible loop is exposed in the heptameric model, the sequence 65–70 is inaccessible at the subunit interface. Furthermore, based on studies with GroES, which dissociates to monomers <1 μM (25), Mt cpn10 would be expected to dissociate at the concentrations used in all these biological assays. This suggests that an alternative oligomeric form of Mt cpn10 may be required for osteolytic activity. A tetrameric form of Mt cpn10 has been reported (16), and this may be the osteolytically active form. Alternatively, Mt cpn10 might assemble as a heptamer only at the putative cell receptor. The flexible loop also binds to cpn60 in the cpn60–cpn10 protein-folding complex (26), suggesting that the putative cell receptor for Mt cpn10 has some structural homology with cpn60. If this structural homology is significant, it would require the Mt cpn10 to assemble as a heptamer on the receptor. Considering the gross structural rearrangements that occur in the cpn10 subunits of the heptamer when it binds to cpn60 (27), it may be possible for the two active sequences in Mt cpn10 to make contact with the receptor. None of the peptide fragments containing the active sequences appear to be involved in the interaction of Mt cpn10 with the human osteoblast-like cell line MG63, possibly due to their inability to assemble as a heptamer. In this regard, it may be important to note that mitochondrial cpn60 is expressed on the surface of human cells (28). These receptors are likely to be of therapeutic importance in the treatment of bone tuberculosis and possibly in other bone diseases.

Based upon the peptide activity data, the Mt cpn10 structure model gives an approximate guide to the location of the osteolytically-active sequences on the Mt cpn10 structure. The precise molecular structure accounting for the bioactivity of Mt cpn10 will be obtained by ongoing work on solving the Mt cpn10 structure by x-ray crystallography (Roberts, M.M., A. Coker, G. Fossati, P. Mascagni, A.R.M. Coates, and S.P. Wood, unpublished data) and the use of site-directed mutagenesis.

It is not known which Mt strains are associated with Pott's disease. In this study we have tested the Mt strain H37Rv, which is a virulent strain commonly used in research into tuberculosis. We have shown that the obligate protein, Mt cpn10, is the osteolytically-active component produced by this organism. All strains of Mt must contain this protein and therefore have the potential to induce bone disease. There may be additional factors to consider in the propensity of Mt to cause Pott's disease, and further studies into this area are clearly necessary.

Acknowledgments

We are grateful to the Medical Research Council for funding P.A. White and K. Heron, to the Arthritis and Rheumatism Council for funding S.P. Nair, B. Henderson, and A.R.M. Coates, to Italfarmaco for support to A.R.M. Coates, and to the Sir Jules Thorne Charitable Trust for supporting K. Reddi. The Mt LAM and neutralizing mAb to LAM were kind gifts of Professor John T. Belisle, Department of Microbiology, Colorado State University, Fort Collins, CO, and the mAb to Mt cpn10 (SA12) was a kind gift from Dr. Percy Minden, Department of Medicine, Colorado State University, Fort Collins, CO.

References

References
1
Murray
CJ
,
Styblo
K
,
Rouillon
A
Tuberculosis in developing countries: burden, intervention and cost
Bull Int Union Tuberc Lung Dis
1990
65
6
24
[PubMed]
2
Boachie-Adjei
O
,
Squillante
RG
Tuberculosis of the spine
Orthop Clin North Am
1996
27
95
103
[PubMed]
3
Nair
SP
,
Meghji
S
,
Wilson
M
,
Reddi
K
,
White
P
,
Henderson
B
Bacterially induced bone destruction: mechanisms and misconceptions
Infect Immun
1996
64
2371
2380
[PubMed]
4
Fenton
MJ
,
Vermeulen
MW
Immunopathology of tuberculosis: roles of macrophages and monocytes
Infect Immun
1996
64
683
690
[PubMed]
5
Coates, A.R.M. 1996. Immunology of chaperonins. In The Chaperonins. R.J. Ellis, editor. Academic Press, London. 267–296.
6
Cavanagh
AC
,
Morton
H
The purification of early-pregnancy factor to homogeneity from human platelets and identification as chaperonin 10
Eur J Biochem
1994
222
551
560
[PubMed]
7
Beagley
KW
,
Fujihashi
K
,
Black
CA
,
Lagoo
AS
,
Yamamoto
M
,
McGhee
JR
,
Kiyono
H
The Mycobacterium tuberculosis71-kDa heat-shock protein induces proliferation and cytokine secretion by murine gut intraepithelial lymphocytes
Eur J Immunol
1993
23
2049
2052
[PubMed]
8
Kirby
AC
,
Meghji
S
,
Nair
SP
,
White
P
,
Reddi
K
,
Nishihara
T
,
Nakashima
K
,
Willis
AC
,
Sim
R
,
Wilson
M
,
Henderson
B
The potent bone-resorbing mediator of Actinobacillus actinomycetemcomitansis homologous to the molecular chaperone GroEL
J Clin Invest
1995
96
1185
1194
[PubMed]
9
Baird
PN
,
Hall
LM
,
Coates
ARM
Cloning and sequence analysis of the 10K antigen gene of Mycobacterium tuberculosis.
J Gen Microbiol
1989
135
931
939
[PubMed]
10
Seckinger
P
,
Klein-Nulend
J
,
Alander
C
,
Thompson
RC
,
Dayer
JM
,
Raisz
LG
Natural and recombinant human IL-1 receptor antagonists block the effects of IL-1 on bone resorption and prostaglandin production
J Immunol
1990
145
4181
4184
[PubMed]
11
Chambers
TJ
,
McSheehy
PM
,
Thomson
BM
,
Fuller
K
The effect of calcium-regulating hormones and prostaglandins on bone resorption by osteoclasts disaggregated from neonatal rabbit bones
Endocrinology
1985
116
234
239
[PubMed]
12
Minden
P
,
Kelleher
PJ
,
Freed
JH
,
Nielsen
LD
,
Brennan
PJ
,
McPheron
L
,
McClatchy
JK
Immunological evaluation of a component isolated from Mycobacterium bovis BCG with a monoclonal antibody to M. bovisBCG
Infect Immun
1984
46
519
525
[PubMed]
13
Coates
ARM
,
Hewitt
J
,
Allen
BW
,
Ivanyi
J
,
Mitchison
DA
Antigenic diversity of Mycobacterium tuberculosis and Mycobacterium bovisdetected by means of monoclonal antibodies
Lancet
1981
2
167
169
[PubMed]
14
Coates
ARM
,
Nicolai
H
,
Pallen
MJ
,
Guy
A
,
Chaparas
SD
,
Mitchison
DA
The 45 kilodalton molecule of Mycobacterium tuberculosisidentified by immunoblotting and monoclonal antibodies as antigenic in patients with tuberculosis
Br J Exp Pathol
1989
70
215
225
[PubMed]
15
Chatterjee
D
,
Lowell
K
,
Rivoire
B
,
McNeil
MR
,
Brennan
PJ
Lipoarabinomannan of Mycobacterium tuberculosis: capping with mannosyl residues in some strains
J Biol Chem
1992
267
6234
6239
[PubMed]
16
Fossati
G
,
Lucietto
P
,
Giuliani
P
,
Coates
ARM
,
Harding
S
,
Colfen
H
,
Legname
G
,
Chan
E
,
Zaliani
A
,
Mascagni
P
Mycobacterium tuberculosischaperonin 10 forms stable tetrameric and heptameric structures: implications for its diverse biological activities
J Biol Chem
1995
270
26159
26167
[PubMed]
17
Lucietto
P
,
Fossati
G
,
Ball
HL
,
Giuliani
P
,
Mascagni
P
M. tuberculosischaperonin 10 and N-truncated fragments. Their synthesis and purification by the isoelectric focusing technique carried out in solution
J Pept Res
1997
49
308
323
[PubMed]
18
Nair
S
,
Song
Y
,
Meghji
S
,
Reddi
K
,
Harris
M
,
Ross
A
,
Poole
S
,
Wilson
M
,
Henderson
B
Surface-associated proteins from Staphylococcus aureusdemonstrate potent bone resorbing activity
J Bone Miner Res
1995
10
726
734
[PubMed]
19
Gitelman
HJ
An approved automated procedure for the determination of calcium in biological specimens
Anal Biochem
1967
18
521
531
20
Marshall
MJ
,
Holt
I
,
Davie
MW
The number of tartrate-resistant acid phosphatase-positive osteoclasts on neonatal mouse parietal bones is decreased when prostaglandin synthesis is inhibited and increased in response to prostaglandin E2, parathyroid hormone, and 1,25 dihydroxyvitamin D3
Calcif Tissue Int
1995
56
240
245
[PubMed]
21
White
PA
,
Wilson
M
,
Nair
SP
,
Kirby
AC
,
Reddi
K
,
Henderson
B
Characterization of an antiproliferative surface-associated protein from Actinobacillus actinomycetemcomitanswhich can be neutralized by sera from a proportion of patients with localized juvenile periodontitis
Infect Immun
1995
63
2612
2618
[PubMed]
22
Orme
IM
,
Miller
ES
,
Roberts
AD
,
Furney
SK
,
Griffin
JP
,
Dobos
KM
,
Chi
D
,
Rivoire
B
,
Brennan
PJ
T lymphocytes mediating protection and cellular cytolysis during the course of Mycobacterium tuberculosisinfection
J Immunol
1992
148
189
196
[PubMed]
23
Hunt
JF
,
Weaver
AJ
,
Landry
SJ
,
Gierasch
L
,
Deisenhofer
J
The crystal structure of the GroES co-chaperonin at 2.8Å resolution
Nature (Lond)
1996
379
37
45
[PubMed]
24
Chan
E
,
Fossati
G
,
Giuliani
P
,
Lucietto
P
,
Zaliani
A
,
Coates
ARM
,
Mascagni
P
Sequence and structural homologies between M. tuberculosischaperonin 10 and the MHC class I/II peptide binding cleft
Biochem Biophys Res Commun
1995
211
14
20
[PubMed]
25
Zondlo
J
,
Fisher
KE
,
Liu
Z
,
Ducote
KR
,
Eisenstein
E
Monomer-heptamer equilibrium of the Escherichia colichaperonin GroES
Biochemistry
1995
34
10334
10339
[PubMed]
26
Landry
SJ
,
Zeilstra-Ryalls
J
,
Fayet
O
,
Georgopoulos
C
,
Gierasch
LM
Characterization of a functionally important mobile domain of GroES
Nature (Lond)
1993
364
255
258
[PubMed]
27
Saibil
H
The lid that shapes the pot: structure and function of the chaperonin GroES
Structure (Lond)
1996
4
1
4
[PubMed]
28
Soltys
BJ
,
Gupta
RS
Immunoelectron microscopic localization of the 60-kDa heat shock chaperonin protein (Hsp60) in mammalian cells
Exp Cell Res
1996
222
16
27
[PubMed]
1

Abbreviations used in this paper: cpn, chaperonin; LAM, lipoarabinomannan; Mt, Mycobacterium tuberculosis.

Author notes

Address correspondence to Dr. Michael Mark Roberts, RM 1.241A, Jenner Wing, Level 1, Division of Molecular Microbiology, St. George's Hospital Medical School, Cranmer Terrace, London SW17 0RE, United Kingdom. Phone: 44-181-725-5722; FAX: 44-181-672-0234; E-mail: m.roberts@sghms.ac.uk