M i d i o h g y (1998), 144,781-791
Printed in Great Britain
Rhizobium leguminosarum bv. viciae produces a novel cyclic trihydroxamate siderophore, vicibactin Michael J. Dilworth,’ Kerry C. Carson,’ Robin G. F. Giles,2 Lindsay T. Byrne3 and Andrew R. Glenn’ Author for correspondence: Michael J. Dilworth. Tel: +61 8 9360 2112. Fax: +61 8 9360 6486. e-mail :
[email protected]
1
Centre for Rhizobium Studies, School of Biological Sciences & Biotechnology, Division of Science, Murdoch University, Murdoch, Western Australia 6150
2
Chemistry Department, School of Mathematical & Physical Sciences, Division of Science, Murdoch University, Murdoch, Western Australia 6150
3
Department of Chemistry, The University of Western Australia, Nedlands, Western Australia 6009
Trihydroxamate siderophores were isolated from iron-def icient cultures of three strains of Rhilobiurn legurninorarum biovar wiciae, two from Japan (WSM709, WSM710) and one from the Mediterranean (WU235), and from a Tn5induced mutant of WSM710 (MNF7101). The first three all produced the same compound (vicibactin), which was uncharged and could be purified by solvent extraction into benzyl alcohol. The gallium and ferric complexes of vicibactin were extractable into benzyl alcohol a t pH 59, while metal-free vicibactin could be extracted with good yield a t pH 89. The trihydroxamate from MNF7101 (vicibactin 7101) could not be extracted into benzyl alcohol, but its cationic nature permitted purification by chromatography on Sephadex CM-25 (NHt form). Relative molecular masses and empirical formulae were obtained from fast-atom-bombardment MS. The structures were derived from one- and two-dimensional IH and I3C NMR spectroscopy, using DQF-COSY, NOESY, HMQC and HMBC techniques on the compounds dissolved in methanol-d, and DMSOd6.Vicibactin proves to be a cyclic molecule containing three residues each of (R)-2,5-diamino-N2-acetyl-N5-hydroxypentanoicacid (W-acetyl-M-hydroxy-Dornithine) and (R)-3-hydroxybutanoic acid, arranged alternately, with alternating ester and peptide bonds. Vicibactin 7101 differed only in lacking the acetyl substitution on the N2of the W-hydroxyornithine, resulting in net positive charge; it was still functional as a siderophore and promoted 55Fe uptake by iron-starved cells of WSM710 in the presence of an excess of phosphate. The rate of vicibactin biosynthesis by iron-def icient cells of WSM710 was essentially constant between pH 5.5 and 79, but much decreased at pH 59. When iron-starved cultures were supplemented with potential precursors for vicibactin, the rates of i t s synthesis were consistent with both fi hydroxybutyrate and ornithine being precursors. A t least three genes seem likely to be involved in synthesis of vicibactin from ornithine and fi hydroxybutyrate:a hydroxylase adding the -OH group to the N5of ornithine, an acetylase adding the acetyl group to the N2of ornithine, and a peptide synthetase system. Keywords: Rhizobium leguminosarum, siderophores, trihydroxamates, vicibactins
INTRODUCTION At neutral pH iron is virtually insoluble, the maximum concentration of Fe(II1) hydroxide being lo-’* M at Abbreviations: DQF-COSY, double quantum filtered correlation spectroscopy; FAB, fast atom bombardment; HMBC, heteronuclear multiple bond connectivity spectroscopy; HMQC, heteronuclear multiple quantum coherence spectroscopy; NOESY, nuclear Overhauser enhancement spectroscopy. 0002-2252 0 1998 SGM
physiological p H (Neilands, 1991). Nevertheless, iron is an essential macronutrient for all micro-organisms, with the exception of the lactobacilli (Pandey et al., 1994), and is associated with a number of important bacterial characteristics, such as virulence. It is also required as a structural component of a wide variety of haem and non-haem iron proteins. As a consequence microorganisms have evolved a variety of strategies for its acquisition which include the biosynthesis and excretion 781
M. J. D I L W O R T H a n d OTHERS
of Fe(II1)-chelating compounds (siderophores) and the membrane transport systems necessary for the translocation of the Fe-siderophore complex into the cell (Neilands, 1981, 1982).
bium leguminosarum biovar uiciae WSM710 (Carson et al., 1992), the iron complex of which has an E~~~ of 1510 M-' cm-' and a molecular mass of 828 (Carson et al., 1994). In this work vicibactin has been further characterized and its structure elucidated using 'H and 13C NMR spectroscopy. Vicibactin is shown to be a cyclic molecule comprised of three residues each of (R)-2,5-diamino-N2-acetyl-N5-hydr~xypentanoic acid (N2-acetyl-N5-hydroxy-~-ornithine) and three of (R)-3hydroxybutanoic acid, with alternating ester and peptide bonds. To our knowledge this is the first report of a bacterial trihydroxamate siderophore in which the hydroxamate ligands are formed by N5-hydroxyornithine and an adjacent carbonyl from P-hydroxybutyrate. It is also the first trihydroxamate siderophore from any root nodule bacterium for which the structure is known. A mutant strain of WSM710 (MNF7101) is shown to produce a functional siderophore which lacks the N2-acetylation found on ornithine in vicibactin.
Microbial siderophores may be classified by the nature of their Fe(II1)-chelating ligands as hydroxamates, catechols, carboxylates or mixed (Hofte, 1993).A number of fungal siderophores are di- and trihydroxamates based on ornithine, the 5-amino group of which has been hydroxylated and acylated to form the hydroxamate ligand (Hofte, 1993; Jalal & van der Helm, 1991). Bacterial siderophores in which the hydroxamate ligands are based on D- or L-ornithine include the ornibactins from Burkholderia (formerly Pseudomonas) cepacia, which contain one carboxylate and two hydroxamate ligands (Stephan et al., 1993) and the pyoverdines from Pseudomonas spp., which may have one or two hydroxamate groups (Hofte, 1993). The chelation of iron by exochelin MS from Mycobacterium smegmatis involves one N5-formyl-N5-hydroxyornithine and two N5-hydroxyornithine groups with METHODS adjacent carbonyl moieties (Sharman et al., 1995). Bacterial strains. Rhixobium leguminosarum bv. viciae (hereUnmodified D-ornithine can also be present in nonafter referred to as R . leguminosarum) WSM710 and R. hydroxamate siderophores such as staphyloferrin A leguminosarum WSM709 were isolated in Japan by John from Staphylococcus hyicus (Konetschny-Rapp et al., Howieson (Centre for Rhixobium Studies, Division of Science, 1990), where iron chelation involves two citrate Murdoch University) from a wild vetch (Vicia sp.) and a moieties. commercial pea crop (Pisurn sativum), respectively. R . The N,-fixing symbiosis between nodule bacteria and leguminosarum WU235 is of Mediterranean origin and was obtained from the Institute of Agriculture, The University of their respective host legumes is iron-dependent as iron is Western Australia. The three strains produce yellow haloes on a structural component of key proteins such as nitroCAS agar (Schwyn & Neilands, 1987), indicative of siderogenase, hydrogenase and leghaemoglobin. In symbiotic phore production. MNF7101 is a Tn5-induced mutant of legumes, iron deficiency has a variety of effects, preWSM710 which produces a larger halo than the wild-type. venting nodule initiation in lupins (Tang et al., 1991) or subsequent nodule development in peanuts (O'Hara et Glassware preparation. T o minimize iron contamination all al., 1988). The iron nutrition of root-nodule bacteria is glassware was cleaned in 16 o/' (w/v) aqueous HC1 and rinsed in deionized water from a reverse-osmosis desalinator therefore of agricultural importance; the role of sidero(Osmotron, Australia). This quality of deionized water was phore production in the success of nodulating strains used in all growth media and in the preparation of all reagents. has not yet been assessed. Siderophores produced by root nodule bacteria include carboxylates such as rhizobactin (Smith et al., 1985), citrate (Guerinot et al., 1990) and anthranilate (Rioux et al., 1986a, b), catechols (Modi et al., 1985; Nambiar & Sivaramakrishnan, 1987; Pate1 et al., 1988; Jadhav & Desai, 1994; Roy et al., 1994)and hydroxamates (Carson et al., 1992, 1994; Lesueur et al., 1993; Persmark et al., 1993). Rhizobactin 1021 from Rhixobium meliloti (recently renamed Sinorhizobium meliloti; de Lajudie et al., 1994) 1021, the only root nodule bacterial hydroxamate siderophore whose structure is known, is a dihydroxamate containing a central citrate moiety, substituted on one terminal carboxyl group by 1amino-3-(N-acetyJ-N-hydroxyamino)-propane and by 1-amino-3-(N-hydroxy -N-( E )-2-decenylamino)propane on the other (Persmark et al., 1993). The ability to synthesize siderophores appears to be restricted to a limited range of strains rather than widely distributed in root nodule bacteria (Carson et al., 1992).
Assay of hydroxamate concentrations. Siderophore concentrations were calculated from their absorbance at 450 nm in 3.3 mM Fe(ClO,), in 0.073 M HClO, (Carson et al., 1992), using an experimentally determined molar absorption coefficient of 1510 M-l cm-' (Carson et al., 1994) for vicibactin.
Vicibactin (previously called hydroxamate K) is a trihydroxamate-type siderophore produced by Rhizo-
To measure rates of siderophore synthesis, concentrations of
782
Growth of bacteria. Cultures were grown in the minimal salts medium (MSM) of Brown & Dilworth (1975) as described by Carson et al. (1994) with 20 mM mannitol as carbon source, but with NH,Cl at 10 mM. The concentration of added iron (in pM) is indicated by a number after MSM; thus MSM-20Fe indicates 20 pM added Fe. Cultures for siderophore isolation were grown in MSM-O-SFe.
All bacteria were grown in flasks containing 20% of their volume of medium on a rotary shaker (200-250r.p.m.) at 28 OC. Turbidity of rhizobial cultures was measured at 600 nm using a Hitachi G-1100 spectrophotometer ; if necessary cells were diluted with MSM-OFe so that the OD,,, fell within the range 0 1-43. vicibactin were measured over time in cultures growing in
Vicibactins from Rhizobium leguminosarum MSM-05Fe. In experiments in which the pH was varied, cells were grown with a mixture of 3 0 m M HEPES and 3 0 m M MES of the appropriate pH, centrifuged, washed once with fresh medium, and resuspended to an OD,,, of 0 1 in medium of the same pH. Growth and vicibactin production were followed for 4 h after siderophore was measurable in the supernatant; the pH was checked at the end of the experiment. In experiments with potential metabolic precursors of vicibactin, cells were grown in MSM-0-5Fe buffered with 60 mM HEPES to pH 6.8, centrifuged, washed and resuspended to an OD,,, of 0-1in the same medium containing added potential precursors at 10 mM. Optical density and vicibactin concentrations were measured as a time course extending into stationary phase. Vicibactin concentrations were then plotted against the accumulated value of JOD,,,.dt, approximated as the area under the optical density curve (de Hollaender & Stouthamer, 1979), the rate of synthesis being derived from the slope of the line. Rates are reported on a dry weight basis, using a value of 0.329 mg dry weight ml-l for a culture with an OD,,, of 1.0. Purification of hydroxamate siderophores from WSM709, WSM710 and WU235. Siderophores produced by these strains were purified by solvent extraction with benzyl alcohol, either at pH 5.0 for the ferrated complex (Carson et al., 1994) or at pH 8.0 for the unferrated form. T o form the galliumsiderophore complexes, ferrous sulphate was replaced with gallium perchlorate at the same molar ratio (2:l) to siderophore, and the complexes purified as for the Fe(II1) compounds. Purification of hydroxamate siderophore from MNF7101. After centrifugation (10000 g, 15 min), the culture supernatant (225 1) was diluted with an equal volume of water, and ammonium sulfate (45 ml, 1M) added to a final concentration of 10 mM. The solution was loaded onto a column (19 cm x 2 1 cm diam.) of Sephadex CM-25(NHi) equilibrated with 10 mM ammonium sulfate. Elution was with a linear gradient of ammonium sulfate (50-600 mM). The unferrated siderophores eluted as two bands, the first at 210 mM ammonium sulfate and the major band at 430 mM. The major siderophore peak was collected, diluted with 9 vols water, and readsorbed onto a small column of Sephadex CM-25(NHi) equilibrated with 5 mM ammonium bicarbonate (pH 8.2). After washing with 25 mM ammonium bicarbonate (pH 8*2), the siderophore was eluted with 200 mM ammonium bicarbonate (pH 8-2), and freeze-dried for 3 d ; the yield relative to the original supernatant was 41 YO. FAB-MS of vicibactin and vicibactin 7101. High-resolution fast-atom-bombardment (FAB) mass spectra were recorded on a VG Micromass Autospec mass spectrometer at the University of Western Australia. Chirality of ornithine and phydroxybutyrate. Vicibactin (6 mg) was hydrolysed at 105 "C for 12 h in 0.6 m16 M HC1 in vials flushed with N,; HC1 was removed in a vacuum desiccator over NaOH. Methyl pentafluoropropionyl derivatives of D- and L-ornithine and of the vicibactin hydrolysate were prepared as described by Sharman et al. (1995) and separated by gas chromatography on a 0.5 mm x 50 m ChiraSil (Altech, Deerfield, IL, USA) column operated isothermally at 180 "C with helium as carrier. For /I-hydroxybutyrate, vicibactin samples (6 mg) were hydrolysed at 80 "C for 16 h in 3 M HCl(O.6 ml) in vials flushed with N,; acid was removed in a vacuum desiccator over NaOH. The dried residue was dissolved in 1.0 ml 50 mM Tris/HCl buffer, pH 8.2, and samples assayed for D-/3-hydroxybutyrate using D-/I-hydroxybutyrate dehydrogenase (EC1.1.1.30) and
NAD+. Other hydrolysates were converted to the N-trifluoroacetylpropan-2-yl esters (Sharman et al., 1995) and compared with authentic D- or L-/I--hydroxybutyrate derivatives by gas chromatography on the same ChiraSil column operated isothermally at 180 "C with helium as carrier. NMR spectroscopy. 'H and 13C NMR spectra of the hydroxamate siderophores [the Ga(II1) complexes of the trihydroxamates from WSM709, WSM710 and WU235, and the metal-free forms of vicibactin and vicibactin 71011 were recorded at 22 or 27 "C on a 500 MHz Bruker instrument. Chemical shifts are reported relative to tetramethylsilane (TMS) at 0 0 p.p.m. 55Fe uptake. Uptake of 55Fe complexes of vicibactin or vicibactin 7101 was measured as described by Carson et al. (1994). Spectroscopic studies of vicibactin and vicibactin 7101. UV and visible spectra were recorded with a Beckman DU-64 spectrophotometer . Acid hydrolysis. Purified, ferrated vicibactin (5 mg) was hydrolysed with 6 M HC1 at 100 "C for 15 h in a sealed ampoule; the acid was removed in a vacuum desiccator over NaOH and the hydrolysate redissolved in ethanol. Amino acids were separated on a Varian 5560 automatic HPLC with ninhydrin quantification. TLC was run on silica gel using 1butanol/acetone/acetic acid/water (70:70: 20 :40, by vol.) (Smith & Seakins, 1976).
RESULTS AND DISCUSSION UV and visible spectral properties Ferric-vicibactin has a A,, in aqueous solution at 445 nm and a molar absorption coefficient ( E ~ of ~ ~ 1510 M-l cm-l (Carson et al., 1994). It also absorbs strongly a t 244 nm (e,,, 2293 M-' cm-l). Ferric vicibactin 7101 absorbs maximally a t 435 nm (egS5 1082 M-l cm-l). The spectra of both ferric hydroxamates are unaffected by pH change within the range 1.5-7-0. Literature values for the molar absorption of trihydroxamate siderophores are generally about 25003000 M-' cm-' (Neilands, 1984;Crumbliss, 1991),reflecting hexadentate binding of the iron. On that basis, the value for vicibactin is unusually low despite its trihydroxamate-like lack of spectral response to decreases in pH. Since the acid-induced decrease in the molar absorption of ferrioxamine B is associated with an opening of the complex such that iron coordination by the ligands is decreased (Monzyk & Crumbliss, 1982), the low molar absorption of vicibactin may indicate that fewer than six ligands are attached to the iron. However, the chelation of iron by exochelin from M . smegmatis is hexadentate, involving three hydroxamate groups, but the molar absorption coefficient is also low, at 1044 M-' cm-' (Sharman et al., 1995).
Biological activity
Both vicibactin a n d vicibactin 7101 were functional
siderophores, since they were effective at mediating 55Fe uptake by cells of either WSM710 or MNF7101 grown o n low-iron medium (MSM-0-5Fe); the uptake rates with vicibactin were 0.19 a n d 0.21 nmol min-l (mg protein)-' for the two strains, a n d with vicibactin 7101 783
)
M. J. D I L W O R T H a n d O T H E R S
Table l a'H and 13C NMR data for vicibactin and vicibactin 7101 ~
Position (Figs 1 and 5)
Group*
Vicibactin (Ga complex) (methanol-d,) H
C
(4
[4J(Hz)l 1 2
-
446, dd, J 44,42 1.71, m 200, m 1.60, m 1-84,m 3.56, m 3-97, m
3 4 5
6 7
173.4t 53.7 28.8 243 51.8
1724t 262, dd, 37.8 J 145, 11.8 299, dd, J 145, 1.9 498, ddq, . 71.7 J 11.8, 6.1, 1.9 1-33,d, 21.0 J 6.1 -
8
9 10
-
11 12
1.98, s
Vicibactin (Ga complex) (DMSO-d,)
164.2 22.2
C
H
(4
[4J(Hz)l -
171.3
Vicibactin (metal-free) (DMSO-d,) H
[4J(Wl -
C
(4 171.6 520
4.30, m
51.5
405, m
1.49, m 1.83, m 1.51, m 1.62, m 3.38, m 3.94, m
23.1
1.46, m 1.52, m 1.52, m
23-0
3-23, m 3.56, m
46.5
-
2*47+ 2-94, br.d 1 13
27.3 500 169.4 35.9
447, ddq, J 11.5, 6.1, 1.6 1.22, d, J 6.1 8-19, d, J 7.7
-
1.89, s
703 206 -
161.8 223
28.2
-
169.37 2.50, dd, 37.9 J 16.0, 3-7 2.79, dd, J 16.0, 9.3 5.14, m 679 1.18, d, J 63 8.24, d, J 7.2 -
1-82,s
19.9
Vicibactin (metal-free) (methanol-d,) H
[4J(Wl -
429, dd, J 8.2, 4 8 1.60, m 1.79, m 1.68, m 3.55, m 3.64, m -
2.69, dd, J 15.3, 3.5 286, dd, J 15.3, 9.3 5-38, m 1.32, d,
J 6.2
C
(4 173.3153.8 29.6 243 48.4 172.3 39.3
70.1 205
-
-
-
169.45t 222
-
173.51225
1-99,s
_
_
_
_
_
Vicibactin 7101 (metal-free) (methanol-d,) H
14J(Hz)l -
3-43, m 1.58, m 1-70,m 1.70, m 1.70, m 3.62, m
C
(4 175.6t 549 322 23.7 48.5
172-3t 2.71, dd, 39-6 J 15.0, 4 5 2.86, dd, J 15.0, 8-4 5.33, ddq, 69.8 J 8.4, 63, 45 1.32, d, 204 J 63 -
* (0)represents ornithine ; (HB) represents hydroxybutyrate.
t Chemical shifts with this symbol may be transposed.
+ Signal partially obscured by solvent.
they were 0.36 and 0.23 nmol min-l (mg protein)-', respectively. FAB mass spectroscopy
The FAB-MS spectrum for the Fe(II1) complex of vicibactin from WSM710 gave the molecular ion at 826.265332; the value calculated for the (M-H)+ ion C,,H,,N,O,,Fe is 826.268355. For the Ga(II1)complex of vicibactin from WSM709 the value found was 841.273680 and the calculated value for the (M H)' ion C3,H5,N,0,,Ga was 841,274647.
+
The corresponding values for the metal-free hydroxamate from MNF7101 were 649.343942 and 6490340847 calculated for the (M H)+ ion C2,H49N6012.
+
NMR spectroscopy Vicibactin. Ferric siderophores cannot be studied using
NMR spectroscopy since the paramagnetic iron causes fast relaxation and severe line broadening (Sharman et 784
al., 1995). Gallium was chosen as an alternative metal
since Ga3+is close to Fe3+in size but is not paramagnetic ; it therefore affords sharp spectra without greatly distorting the molecular conformation from that of the ferri-siderophore vicibactin. Clearly defined spectra were then obtained from Ga(II1) vicibactin and from both of the metal-free siderophores vicibactin and vicibactin 7101. Spectra were determined in methanol-d, and in dimethylsulfoxide-d, (DMSO) for vicibactin and Ga(II1) vicibactin and in methanol-d, for vicibactin 7101. The use of the two solvents proved to be complementary. While no solute signals were obscured by solvent peaks in methanol as they were in DMSO, signals for exchangeable protons disappeared in methanol-d, but were observable in DMSO-d,. In the latter solvent, this led to the unambiguous assignment of the regiochemistry of these siderophores through twodimensional double quantum filtered correlation spectroscopy (DQF-COSY) (Piantini et al., 1982)and nuclear Overhauser enhancement spectroscopy (NOESY) (Bodenhausen et al., 1984). DQF-COSY experiments indicate those protons in the NMR spectrum which
Vicibactins from Rhizobium leguminosarum
(a) M=Ga,,3
(b) M = H
Fig. 1. (a) Proposed structure of the Ga(lll) complex of vicibactin from R. leguminosarum WSM710. The numbering corresponds to the position assignments given in Table 1. The metal-binding site is derived from the N-OH of p-hydroxy-Dornithine and the carbonyl from /h-hydroxybutyrate. (b) Proposed structure of the metal-free vicibactin.
6 1-5. Thus, the doublet at 6 1.33 was coupled to a oneproton doublet of doublet of quartets ( J 11-8, 6.1 and 1.9 Hz) at 6 4-98, This latter signal was in turn coupled to each of two one-proton doublet of doublets at 6 2.62 (J,,, 14.5 and Jvic 11-8 Hz) and 6 2.99 (J,,, 145 and Jvic 1.9 Hz). The chemical shifts and coupling constants of this A,MXY system suggested 8-hydroxybutyrate as a partial structure, in which the two diastereotopic methylene protons resonated at significantly different chemical shifts. The chemical shift of the /?-methine proton (6 4.98) was deshielded considerably ( - 1 p.p.m.) from the value observed (6 406) for sodium /?hydroxybutyrate in methanol-d,, which suggested its esterification in vicibactin.
Seven of the remaining ten proton resonances were assigned to an ornithine unit. The proton on the asymmetrically substituted a-carbon of this amino acid residue appeared as a doublet of doublets ( J 4.4 and 4.2 Hz) at 6 4.46, and the DQF-COSY spectrum showed correlations of this a-proton with two one-proton multiplets, the 8-protons, at 6 1.71 and 6 2-00. These correlated, in turn, with the two y-protons at 6 1.60 and 6 1.84. The y-proton at 6 1.60 correlated strongly with the 6-proton at 6 3-56 and the alternative y-proton at 6 1.84 with the remaining &proton at 6 3.97. This spectroscopic assignment was confirmed by the fact that ornithine was identified as a major product when vicibactin was hydrolysed (6 M HCl, 15 h, 100 "C) and the hydrolysate separated on an automatic amino acid analyser.
10
8
6
4
2
0
8 (p.p.m.1 Fig. 2. 'H NMR spectrum for metal-free vicibactin in DMSO-d,. Peaks marked s are due to solvent and those marked i to an impurity.
interact with others through coupling, the magnitude and pattern of which enable the formulation of structural entities within the molecule. NOESY experiments indicate the through-space proximity of protons, thereby further identifying structural components of the molecule. All the 'H NMR spectra were remarkably simple in spite of the relatively large number of protons indicated by the molecular formulae obtained from FAB-MS. Thus, when the spectrum of the Ga(II1) complex of vicibactin was determined in methanol-d,, a total of only sixteen protons was observed, consisting of two three-proton signals in the range 6 1-2 and ten one-proton signals in the range 6 1-5. The two three-proton methyl signals resonated as a doublet ( J 6.1 Hz) at 6 1.33 and a singlet at 6 1.98. A DQF-COSY spectrum identified the vicinal and geminal coupling between the protons in the region
The chemical shifts of the &protons at 6 3.56 and 6 3.97 and the a-proton at 6 4.46 were deshielded relative to the values of 6 2-83 and 6 3-37 for the 6- and a-protons, respectively, of ornithine itself, and these values suggested amidic character for both a- and 6-nitrogens. This was supported in the 13C NMR spectrum of the Ga(II1) complex of vicibactin by the presence of three carbonyl resonances in the range 6 16+174. This, together with the three-proton singlet at 6 1-98in the 'H NMR spectrum, indicated acetylation of one of the nitrogen atoms in each ornithine unit. The NMR data are summarized in Table 1. Thus, the methanol-d, spectra of the Ga(II1) complex of vicibactin indicated the presence of ornithine and /?hydroxybutyrate, giving rise to a repeating unit in which the alcohol group of the latter was esterified with the carboxyl of the former. However, three structural features remained to be identified. These were: (i) the nitrogen of ornithine to which the acetyl was attached, (ii) the nitrogen to which the /?-hydroxybutyrate carboxyl was linked, and (iii) the site of metal complexation. These were determined by acquiring spectra in DMSO-d, for both the same gallium-coordinated compound and for vicibactin. In the former 'H NMR spectrum, the data for which are summarized in Table 1, an additional strongly deshielded one-proton doublet ( J 7.7 Hz) was observed at 6 8.19, making a total of seventeen protons. The chemical shift supported the assignment of this signal to a hydrogen on an amide 785
M. J. D I L W O R T H and OTHERS
-i-
a
'
"
I,
S-
S-
. *
nitrogen, while the fact that it was a doublet suggested that it was coupled to the hydrogen on the asymmetric acarbon of the ornithine. The DQF-COSY spectrum confirmed this, since it showed a strong correlation between these two protons, and all the other correlations observed again confirmed the presence of ornithine and B-hydroxybutyrate.
A two-dimensional NOESY spectrum of vicibactin in DMSO-d, showed, in addition to the expected proximities within the respective ornithine and B-hydroxybutyrate residues, a significant correlation between the hydrogen on nitrogen and the acetyl group, indicating the attachment of the latter to the a-nitrogen of ornithine. This hydrogen on nitrogen also correlated strongly with the multiplet at 6 1.46-1.52 arising from the four p- and y-ornithine protons, thereby suppbrting the proximity of the N-H to these protons. \\\
The 'H NMR spectrum of vicibactin in DMSd-d, showed eleven signals which could be ascribed to only seventeen protons, while the 13Cspectrum showed only eleven carbon resonances. This information suggested threefold symmetry in vicibactin and a simplification of the FAB-MS-derived molecular formula to [ (C,,H,, N,O,),Ga]. The absence of a second low-field amidic proton in the 'H NMR spectrum, and the requirement in the mass spectrum for an additional oxygen for each repeating unit ( C,,H1,N,O,), suggested hydroxylation of the 6-N of ornithine to produce the corresponding hydroxamate through linking this &NOH to the car786
bony1 of p-hydroxybutyrate. Thus, conclusive evidence was obtained that the ornithine is linked to p-hydroxybutyrate through the &nitrogen. From the data presented, the structure shown in Fig. l ( a) was deduced for the Ga(II1) complex of vicibactin. The 'H NMR spectrum for desferri-vicibactin (Fig. l b ) in DMSO-d, is shown in Fig. 2. The chemical shift (6 9.88) of the hydroxyl hydrogen singlet is strongly deshielded, indicating that it is hydrogen-bonded. A two-dimensional DQF-COSY spectrum obtained in DMSO-d, (Fig. 3) once again showed a strong correlation between the amidic hydrogen doublet ( J 7.2 Hz) at 6 8.24 and the a-hydrogen of ornithine, which appeared as a multiplet at 6 4.05. All the other correlations were again entirely consistent. Two further two-dimensional NMR spectroscopic techniques were found to be invaluable in the structural identification of the vicibactins. These were heteronuclear multiple quantum coherence (HMQC) spectroscopy (Bax & Subramanian, 1986), which identifies the carbon atom to which each proton is directly attached, and heteronuclear multiple bond connectivity (HMBC) spectroscopy (Bax & Summers, 1986), which allows the identification of a hydrogen atom either two or three bonds away from a particular carbon atom. Fig. 4 shows the two-dimensional HMQC spectrum in the proton and carbon regions of 6, 0-6 and 6, 15-75, which allowed the unambiguous assignment of the eight
Vicibactins from Rhizobium leguminosarum
F 1 ~ ' " " ' ' 1 " " ' ' ' " " " " ' ' ' I " ' ' " ' ' ' 1 " ' ' ' ~
5
4
3
8 (p.p.m.)
2
non-carbonyl carbon resonances. The three carbonyl signals resonated at 6 169.37, 169.45 and 171.6. Individual assignments (Table 1) for each of these were achieved by two-dimensional HMBC spectroscopy, although the amide and hydroxamate carbonyl carbon resonances were almost isochronous and these assignments may be interchanged. For vicibactin, the 13C NMR spectrum in methanol-d, showed the three carbony1 resonances at 6 172.3, 173.3 and 173.5. It was possible to assign the first of these to the hydroxamate carbonyl carbon using HMBC spectroscopy, although in this solvent it was difficult to distinguish the amide and ester carbonyl carbon resonances, which were almost isochronous. The hydroxamate carbonyl carbon at 6 1723 was readily identified since it showed strong (two-bond) connectivity to each of the two a hydroxybutyrate protons, and three-bond connectivity to the B hydroxybutyrate proton as well as to each of the two 6ornithine protons. In establishing three-bond connectivity between the /I-hydroxybutyrate carbonyl carbon and each of the ornithine &protons, the last of these linkages provided additional confirmation of the structure shown in Fig. l(b), and excluded other possibilities. vicibactin 7101. The 'H NMR spectrum obtained for
metal-free vicibactin 7101 in methanol-d, showed a strong resemblance to that described above for vicibactin. Aside from chemical shift variations, the only major difference observed was the absence of the threeproton acetyl singlet. Individual proton and carbon
1
Fig. 4. Two-dimensional HMQC NMR spectrum for metal-free vicibactin in DMSOd6. The signals marked s are due to solvent and those marked i to an impurity.
9
OI--H II
\
1 \
3
Fig. 5. Proposed structure for the altered vicibactin from a mutant strain (MNF7101) of R. leguminosarum WSM710. The numbering corresponds to the position assignments given in Table 1. This compound differs from vicibactin in lacking the acetylation at N'O.
assignments were confirmed by DQF-COSY and HMQC experiments. These assignments are presented in Table 1 and lead to the structure for vicibactin 7101 depicted in Fig. 5. Strong support for the attachment of the acetyl group to the a-nitrogen of ornithine in vicibactin (cf. Fig. l b and Fig. 5) arises through a comparison of the 'H NMR spectra determined in methanol-d, for vicibactin and vicibactin 7101. Significant shielding (0.87 p.p.m.) of the ornithine a-proton is observed for the latter compound, for which the a-nitrogen is no longer amidic. This is consistent with the basic character of vicibactin 7101 not observed for vicibactin itself. 787
M. J. DILWORTH a n d OTHERS
Table 2. Vicibactin biosynthesis as a function of pH Initial PH
Final pH
Mean generation time (h)
Rate of vicibactin production [nmol h-' (mg dry wt)-']
5.05
491 5.38
9 65 4.7 5 45
24.3 57.7 51.7 45-6 53-2
5-50 602 6.52 699
5.95
6.47 6.93
Chirality of ornithine and p- hydroxybutyrate
The methyl pentafluoropropionyl derivatives of D- and L-ornithine eluted from the ChiraSil column at 1 4 9 and 15.3 min, respectively. The HCl hydrolysate from vicibactin yielded a peak at 14.9 min, indicating that the ornithine was the D-isomer. There was no indication of a peak at 15.3 min, implying that all three ornithines are stereochemically identical. Hydrolysis samples to which D-ornithine was added gave only one peak, while samples with L-ornithine added yielded two distinct peaks. Reduction of NAD' catalysed by P-hydroxybutyrate dehydrogenase in the presence of acid hydrolysates of vicibactin confirms the presence of D-B-hydroxybutyrate in them. While the N-trifluoroacetylpropan-2-yl ester of L-/I-hydroxybutyrate could be identified by gas chromatography on ChiraSil, that of the D-8-hydroxybutyrate appeared to be too unstable to be detected. No peak for the ester of L-8-hydroxybutyrate was detected after treatment of the vicibactin hydrolysate. The positive identification of D-P-hydroxybutyrate using the stereospecific8-hydroxybutyrate dehydrogenase and the absence of the N-trifluoracetylpropan-2-yl ester of LP-hydroxybutyrate indicate that vicibactin contains only D-P-hydroxybutyrate.
Since the structural studies indicated that the building blocks for vicibactin are P-hydroxybutyrate and ornithine, the effects of adding these to cultures synthesizing vicibactin were studied. The addition of other amino acids (10 mM) was also investigated, since any effects of ornithine could be through the provision of an additional source of carbon and nitrogen (Table 3). L-Arginine had no effect on the rate of synthesis or final yield of vicibactin ; L-alanine, L-histidine and L-glutamate all increased the final yield (23-52%) and slightly (about 18 YO)increased the rate of synthesis. Addition of 10 mM L-ornithine increased the rate of vicibactin production by 40% and the total yield by 7 3 % ; by contrast, addition of 10 mM D-ornithine completely prevented growth. In similar experiments where only 20 mM HEPES was used, the p H fell significantly; amino acids like Lhistidine and L-glutamate, which are readily catabolized to produce ammonia, gave larger relative increases in rates of synthesis and final yield because the final pH was less acidic. Addition of 10 mM DL-8-hydroxybutyrate to cultures of WSM710 produced inhibition of growth with a low final yield (23%) of vicibactin; adding it with L-ornithine increased the rate of vicibactin synthesis by 66% but growth inhibition restricted the final yield of siderophore (Table 3). D-P-Hydroxybutyrate alone at 10 mM was similarly inhibitory to growth, while 10 m M L-p-hydroxybutyrate addition completely prevented growth. These data are consistent with both D-ornithine, racemized from L-ornithine, and D-8-hydroxybutyrate being precursors for biosynthesis of vicibactin. Route of biosynthesis
Factors affecting rates of vicibactin production
The biosynthesis of vicibactin from ornithine and Phydroxybutyrate involves the N-hydroxylation of ornithine, its acetylation and the formation of the three peptide and three ester bonds. The Tn5-induced mutation in MNF7101, which results in the unacetylated vicibactin 7101,clearly inactivates the specific acetylase required to acetylate the a-nitrogen of ornithine. By analogy with aerobactin synthesis (de Lorenzo & Neilands, 1986), where the order of reactions producing the hydroxamate group on lysine is N-hydroxylation and N-acetylation prior to incorporation of N-acetyl-Nhydroxylysine, this may well be the sequence leading to vicibactin in WSM710. Iron-starved cells of WSM710 take up iron complexed with vicibactin 7101, implying that the ferri-siderophore complex can be transported into the cells. However, such cells do not convert vicibactin 7101 to vicibactin when incubated with it, but synthesize the latter de novo (data not shown), implying that acetylation does not occur with the cyclized molecule.
The rate of vicibactin synthesis was essentially constant between pH 7.0 and 5.5, falling off significantly at p H 5-0 (Table 2).
It is not clear why these bacteria produce an acetylated siderophore (vicibactin) when the non-acetylated vicibactin 7101 is as functional in iron transport as vicibactin
Hydroxamates from WSM709 and WU235
The 'H and 13C NMR spectra of the Ga(II1) complexes of hydroxamates produced by iron-deficient cultures of WSM709 and WU235 were indistinguishable from those of Ga(II1) vicibactin. While we have no evidence of genetic difference between WSM709 and WSM710, there are clear physiological differences between these two strains and WU235, including the degree of iron repression of siderophore synthesis. That the siderophores produced by strains from regions as widely geographically separated as the Mediterranean and Japan are identical is remarkable.
788
Vicibactins from Rhizobium leguminosarum Table 3. Effect of adding potential precursors on rate of vicibactin production Addition (10 mM)
Maximal vicibactin concn
Rate of vicibactin synthesis [nmol h-' (mg dry wt)-']
417 582 422 635 511 98 111
75.7 88-4 702 88.8 91.1 666 125.8
721
106.1
(W)
None L-Alanine L- Arginine L-Histidine L-glut amate
DL-B-Hydroxybutyrate DL-B-Hydroxybutyrate L-ornithine L-Ornithine
+
ckz
/
\
/
I""
0
yH2
I
OH-[
O=
I
\
/CH-cH3
CH2
\
CHj-CH
\0
/'" \C
&-HC'
I NH
CH-
Lo
itself. Possible explanations include greater ease of excretion of an uncharged molecule (unlikely since vicibactin 7101 appears in the culture supernatant) and greater stability. The latter appears the more probable explanation, since vicibactin 7101 decomposes if culture supernatants are kept for more than 24 h, even at 6 "C. The rate of ornithine biosynthesis required to support the measured rates of vicibactin production by cells of R. leguminosarum WSM710 under iron-deficient conditions is very high compared to that required for arginine biosynthesis. The increase in rate is between 600- and 700-fold, raising some questions about the regulation of ornithine biosynthesis. One possible but speculative explanation is that there are two pathways for ornithine biosynthesis, one for arginine biosynthesis regulated by arginine, and the other repressed by high
..................................................................................................... Fig. 6. Structure of the cyclic vicibactin from R. leguminosarum bv. viciae WSM710. The three residues of N2-acetyl-A6hydroxy-oornithine and j?-D-hydroxybutyrate are arranged alternately and linked by alternating ester and peptide bonds.
iron, but unaffected by arginine. An alternative is a single pathway with a dual control involving iron and arginine and/or ornithine. An obvious parallel to the synthesis of the cyclic vicibactin (Fig. 6 ) is that of enniatin B biosynthesis (Kleinkauf & von Dohren, 1990; Pieper et al., 1995; Stachelhaus & Marahiel, 1995), which also involves the production of a cyclic molecule containing three units each of a hydroxyacid (D-2-hydroxy-3-methylbutanoic acid) and an amino acid (N-methyl-L-valine), linked with alternating peptide and ester bonds. Synthesis of enniatin B is carried out by enniatin synthetase, the 347 kDa product of the 9-5 kb esnl gene of the fungus Fusarium oxysporurn (Haese et al., 1993). Enniatin synthetase is a two-domain multifunctional enzyme, one domain responsible for activation and thioester for789
M. J. D I L W O R T H and OTHERS
Carson, K. C., Holliday, S., Glenn, A. R. & Dilworth, M. J. (1992).
mation for 2-hydroxy-3-methylbutanoicacid, and the other for activation, N-methylation and thioester formation for L-valine (Billich & Zocher, 1987). Since the mutation in MNF7101 which results in non-acetylation does not prevent the other reactions of vicibactin 7101 production, acetylation is not a function of a putative vicibactin synthetase. By analogy with the biosynthesis of some other siderophores, hydroxylation at N5 of ornithine is also unlikely to be a function of such a synthetase. That vicibactin contains D-ornithine rather than the Lisomer was surprising in view of the responses of irondeficient cells to addition of these to the medium. Cells of WSM710 appear to be sensitive to D-ornithine, possibly because it interferes with arginine biosynthesis. Under normal biosynthetic conditions, where L-ornithine is synthesized as an intermediate in arginine biosynthesis, a racemase activity will therefore be required to produce the D-ornithine for vicibactin. Whether this racemase activity is a discrete enzyme, or another function of a peptide synthetase, remains to be determined, but several peptide synthetases catalysing peptide antibiotic production do carry racemase activities for amino acids (Kratzschmar et al., 1989; Fuma et al., 1993; Stachelhaus & Marahiel, 1995) and Dornithine is found in such antibiotics, e.g. bacitracin (Kleinkauf & von Dohren, 1990). Cellular synthesis of the appropriate isomer of /?-hydroxybutyrate presents no such problem as poly-/?-hydroxybutyrate contains the D-form. It is therefore possible that a vicibactin synthetase could racemize L-ornithine to D-ornithine as well as forming the ester and peptide linkages necessary for eventual cyclization to vicibactin; in that case there would still only be three genes specifically associated with vicibactin synthesis.
Siderophore and organic acid production in root nodule bacteria. Arch Microbiol 157,264-271. Carson, K. C., Glenn, A. R. & Dilworth, M. 1. (1994). Specificity of siderophore mediated transport of iron in rhizobia. Arch Microbiol 161, 333-339. Crumbliss, A. L. (1991). Aqueous solution equilibrium and kinetic studies of iron siderophore and model siderophore complexes. In Handbook of Microbial Zron Chelates, pp. 177-234. Edited by G. Winkelmann. Boca Raton : CRC Press. Fuma, S., Fujishima, Y., Corbell, N., D’Souza, C., Nakano, M. M., Zuber, P. & Yamane, K. (1993). Nucleotide sequence of 5’ portion
of srfA that contains the region required for competence establishment in Bacillus subtilis. Nucleic Acids Res 21, 93-97. Guerinot, M. L., Meidl, E. J. & Plessner, 0. (1990). Citrate as a siderophore in Bradyrhizobium japonicum. J Bacteriol 172, 329 8-3303. Haese, A,, Schubert, M. & Herrimann, M. (1993). Molecular characterization of the enniatin synthetase gene encoding a multifunctional enzyme catalysing N-methyldepsipeptide formation in Fusarium scripi. Mol Microbiol7,905-914. HBfte, M. (1993). Classes of microbial siderophores. In Zron Chelation in Plants and Soil Microorganisms, pp. 3-26. Edited by L. L. Barton & B. C. Hemmings. San Diego: VCH Press. de Hollaender, J. A. & Stouthamer, A. H. (1979). Multicarbonsubstrate growth of Rhizobium trifolii. FEMS Microbiol Lett 6, 57-59. Jadhav, R. S. & Desai, A. (1994). Role of siderophore in iron uptake in cowpea rhizobium gnl (peanut isolate) - possible involvement of iron repressible outer membrane proteins. FEMS Microbiol Lett 115, 185-189. Jalal, M. A. F. & van der Helm, D. (1991). Isolation and spectroscopic identification of fungal siderophores. In Handbook o f Microbial Zron Chelates, pp. 235-269. Edited by G. Winkelmann. Boca Raton: CRC Press. Kleinkauf, H. & von DUhren, H. (1990). Nonribosomal biosynthesis of peptide antibiotics. Eur J Biochem 192, 1-15.
ACKNOWLEDGEMENTS
Konetschny-Rapp, S., Jung, G., Meiwes, J. & Zahner, H. (1990).
Staphyloferrin A : a structurally new siderophore from staphlylococci. Eur J Biochem 191,65-74. Krltzschmar, J., Krause, M. & Marahiel, M. A. (1989). Gramicidin S biosynthesis operon containing the structural genes grsA and grsB has an open reading frame encoding a protein homologous to fatty acid thioesterases. J Bacteriol 171, 5422-5429.
We thank Dr A. Reeder (University of Western Australia) for running the FAB mass spectra and the Australian Research Council for financial support. Additionally, we greatly appreciate the constructive criticism and help given by Professor Colin Ratledge, University of Hull, UK, without whose comments we might have got the structure wrong.
de Lajudie, P., Willems, A., Pot, B., Dewettinck, D., Maestrojuan, G., Neyra, M., Collins, M. D., Dreyfus, B., Kersters, K. & Gillis, M. (1994). Polyphasic taxonomy of rhizobia : emendation of the
REFERENCES
genus Sinorhizobium and description of Sinorhizobium meliloti comb. nov., Sinorhizobium sahlei sp. nov., and Sinorhizobirrm teranga sp. nov. Znt J Syst Bacteriol44, 715-733. Lesueur, D., Diem, H. G. & Meyer, J. M. (1993). Iron requirement and siderophore production in Bradyrhizobium strains isolated from Acacia mangium. J Appl Bacteriol74, 675-682. de Lorenzo, V. & Neilands, J. B. (1986). Characterization of iucA and iucC genes of the aerobactin system of plasmid ColV-K30 in Escherichia coli. J Bacteriol 167, 350-355. Modi, M., Shah, K. 5. & Modi, V. V. (1985). Isolation and characterisation of catechol-type siderophore from cowpea Rhizobium RA-1. Arch Microbiol 141, 156-158. Monzyk, B. & Crumbliss, A. L. (1982). Kinetics and mechanism of the stepwise dissociation of iron(II1) from ferrioxamine B in aqueous acid. J A m Chem Soc 104,49214929,
Bax, A. & Subramanian, 5. (1986). Sensitivity-enhanced two-
dimensional heteronuclear shift correlation NMR spectroscopy. J Magn Res 67,565-569. Bax, A. & Summers, M. F. (1986). ‘H and 13C assignments from sensitivity-enhanced detection of heteronuclear multiple-bond connectivity by 2D multiple quantum NMR. J Am Chem SOC 108, 2093-2094. Billich, A. & Zocher, R. (1987). N-Methyltransferase function of the multifunctional enzyme enniatin synthetase. Biochemistry 26, 8417-8423. Bodenhausen, G., Kogler, H. & Ernst, R. R. (1984). Selection of coherence - transfer pathways in NMR pulse experiments. J Magn Res 58,370-388. Brown, C. M. & Dilworth, M. J. (1975). Ammonia assimilation by Rhizobium cultures and bacteroids. J Gen Microbiol86,39-48. -
790
Vici bactins from R hizobium leguminosarum Nambiar, P. T. C. & Sivaramakrishnan, 5. (1987).Detection and assay of siderophores in cowpea rhizobia (Bradyrhizobium) using radioactive Fe (59Fe).Appl Microbiol Lett 4, 37-40. Neilands, J. B. (1981).Microbial iron compounds. Annu Rev Biochem 50,715-731. Neilands, J. B. (1 982). Microbial envelope proteins related to iron. Annu Rev Biochem 36,285-309. Neilands, J. B. (1984).Methodology of siderophores. Struct Bond 58, 1-24. Neilands, J. B. (1991).A brief history of iron metabolism. Biol Metals 4, 1-6. O'Hara, G. W., Dilworth, M. J.,
Boonkerd, N. & Parkpian, P. (1988). Iron deficiency specifically limits nodule development in
peanut inoculated with Bradyrhizobium sp. New Phytol 108, 51-57. Pandey, A., Bringel, F. & Meyer, J. M. (1994).Iron requirement and search for siderophores in lactic acid bacteria. Appl Microbiol Biotech 40,735-739. Patel, H. N., Chakraborty, R. N. & Desai, 5. B. (1988).Isolation and partial characterization of phenolate siderophore from Rhizobium leguminosarum IARI102. FEMS Microbiol Lett 56,131-134. Persmark, M., Pittman, P., Buyer, J. S., Schwyn, B., Gill, P. R. & Neilands, J. B. (1993).Isolation and structure of rhizobactin 1021,
a siderophore from alfalfa symbiont Rhizobium meliloti 1021.
J Am Chem SOC 115,3950-3956.
Piantini, U., Ssrenson, 0. W. & Ernst, R. R. (1982). Multiple
quantum filters for elucidating NMR coupling networks. J Am Chem SOC 104, 6800-6802. Pieper, R., Haese, A., SchrWer, W. & Zocher, R. (1995).Arrangement of catalytic sites in the multifunctional enzyme enniatin synthetase. Eur J Biochem 230, 119-126. Rioux, C. R., Jordan, D. C. & Rattray, 1. B. M. (1986a). Iron requirement of Rhizobium leguminosarum and secretion of
anthranilic acid during growth on an iron-deficient medium. Arch Biochem 248, 175-182. Rioux, C. R., Jordan, D. C. & Rattray, J. B. M. (1986b). Anthranilate-promoted iron uptake in Rhizobium leguminosarum. Arch Biochem 248, 183-189. Roy, N., Bhattacharyya, P. & Chakrabartty, P. K. (1994).Iron acquisition during growth in an iron deficient medium by Rhizobium sp. isolated from Cicer arietinum. Microbiology 140, 2811-2820. Schwyn, B. & Neilands, 1. B. (1987).Universal chemical assay for the detection and determination of siderophores. Anal Biochem 160,47-56. Sharman, G. J., Williams, D. H., Ewing, D. F. & Ratledge, C. (1995). Isolation, purification and structure of exochelin MS, the extracellular siderophore from Mycobacterium smegmatis. Biochem J 305, 187-196. Smith, 1. & Seakins, J. W. T. (1976)Chromatographic and Electrophoretic Techniques vol. 1, 4th edn. Bath: William Heinemann Medical Books. Smith, M. J., Shoolery, J. N., Schwyn, B., Holden, 1. & Neilands, J. B. (1985).Rhizobactin, a structurally novel siderophore from
Rhizobium meliloti. J Am Chem SOC 107, 1739-1743. Stachelhaus, T. & Marahiel, M. A. (1995).Modular structure of genes encoding multifunctional peptide synthetases required for non-ribosomal peptide synthesis. FEMS Microbiol Lett 125,3-14. Stephan, H., Freund, S., Meyer, J. M., Winkelmann, G. & Jung, G.
(1993). Structure elucidation of the gallium ornibactin complex by 2D NMR spectroscopy. Liebigs Ann Chem 43-48. Tang, C., Robson, A. D. & Dilworth, M. 1. (1991).Which stage of nodule initiation in Lupinus angustifolius L. is sensitive to iron deficiency ? New Phytol 117, 243-250. Received 27 October 1997; accepted 5 November 1997.
791
