Iron transport systems of Serratia marcescens

Serratia marcescens W225 expresses an unconventional iron(III) transport system. Uptake of Fe3+ occurs in the absence of an iron(III)-solubilizing siderophore, of an outer membrane receptor protein, and of the TonB and ExbBD proteins involved in outer membrane transport. The three SfuABC proteins found to catalyze iron(III) transport exhibit the typical features of periplasmic binding-protein-dependent systems for transport across the cytoplasmic membrane. In support of these conclusions, the periplasmic SfuA protein bound iron chloride and iron citrate but not ferrichrome, as shown by protection experiments against degradation by added V8 protease. The cloned sfuABC genes conferred upon an Escherichia coli aroB mutant unable to synthesize its own enterochelin siderophore the ability to grow under iron-limiting conditions (in the presence of 0.2 mM 2.2'-dipyridyl). Under extreme iron deficiency (0.4 mM 2.2'-dipyridyl), however, the entry rate of iron across the outer membrane was no longer sufficient for growth. Citrate had to be added in order for iron(III) to be translocated as an iron citrate complex in a FecA- and TonB-dependent manner through the outer membrane and via SfuABC across the cytoplasmic membrane. FecA- and TonB-dependent iron transport across the outer membrane could be clearly correlated with a very low concentration of iron in the medium. Expression of the sfuABC genes in E. coli was controlled by the Fur iron repressor gene. S. marcescens W225 was able to synthesize enterochelin and take up iron(III) enterochelin. It contained an iron(III) aerobactin transport system but lacked aerobactin synthesis. This strain was able to utilize the hydroxamate siderophores ferrichrome, coprogen, ferrioxamine B, rhodotorulic acid, and schizokinen as sole iron sources and grew on iron citrate as well. In contrast to E. coli K-12, S. marcescens could utilize heme. DNA fragments of the E. coli fhuA, iut, exbB, and fur genes hybridized with chromosomal S. marcescens DNA fragments, whereas no hybridization was obtained between S. marcescens chromosomal DNA and E. coli fecA, fhuE, and tonB gene fragments. The presence of multiple iron transport systems was also indicated by the increased synthesis of at least five outer membrane proteins (in the molecular weight range of 72,000 to 87,000) after growth in low-iron media. Serratia liquefaciens and Serratia ficaria produced aerobactin, showing that this siderophore also occurs in the genus Serratia.

Fe3" is insoluble at neutral pH, so iron supply in microbes occurs via siderophore compounds (12,39). The Fe3" siderophores are taken up in an energy-coupled process across the outer and cytoplasmic membranes of gram-negative bacteria (7,10). Transport across the outer membrane requires receptor proteins which specifically recognize the various Fe3" siderophores. These receptors are synthesized under iron-limiting growth conditions in amounts similar to those of the major outer membrane proteins, whereas they are barely detectable on Coomassie-stained polyacrylamide gels after growth at sufficient iron concentrations (8,15,33,41). Outer membrane translocation of the Fe3+ siderophores depends in addition on the TonB and ExbBD proteins, which have been localized in the cytoplasmic membrane. The TonB protein extends into the periplasmic space and is thought to be involved in the energy-dependent vectorial release of the Fe3+ siderophores from the receptors into the periplasm (19,20,23,27,29,43). Uptake across the cytoplasmic membrane seems to follow a periplasmic bindingprotein (PBP)-dependent transport mechanism, since the sequences and locations of the characterized proteins of the Escherichia coli Fe3+ hydroxamate (13,14) and Fe3+ citrate (44,49,53) transport systems are characteristic of PBP systems (1,2).
Recently, we described an Fe3+ transport system of Serratia marcescens which differed from the hitherto studied transport systems of members of the family Enterobac-* Corresponding author. teriaceae by the lack of a siderophore and an outer membrane receptor protein (3,52). Both compounds could have been missed if present in very low amounts, but since Fe3+ transport mediated by the cloned genes in genetically welldefined E. coli strains was independent of the TonB and ExbBD proteins, we suggested a novel type of Fe3+ uptake mechanism across the outer membrane. The sequences of the three transport proteins, derived from the nucleotide sequences of the sfiABC genes, and their localizations in the periplasm and cytoplasmic membrane were typical for PBP transport systems.
In this paper, we report physiological data on the Sfu transport system that support our previous conclusion of an unusual Fe3+ uptake across the outer membrane and a conventional PBP mechanism across the cytoplasmic membrane. Furthermore, we obtained evidence for additional iron supply systems in S. marcescens, indicating multiple ways for the difficult acquisition of this essential element.

MATERIALS AND METHODS
Bacterial strains, plasmids, and media. The S. marcescens and E. coli strains used are listed in Table 1. They were grown in tryptone-yeast extract (TY) medium and nutrient broth (NB) as described previously (52). To reduce the available iron in NB medium, 2.2'-dipyridyl (0.2 mM for NBD and 0.4 or 0.5 mM for NBD*) or ethylenediamine-di(ohydroxy)phenylacetic acid (EDDA; 0.1 mM) was added. Antibiotics were added to maintain the plasmids (ampicillin,  50 ,ug/ml; chloramphenicol, 40 ,g/ml; neomycin, 50 ,g/ml; tetracycline, 15 ,ug/ml). Mutants of S. marcescens W225 deficient in enterochelin synthesis were obtained by incubating log-phase cells with 0.2 mg of N-methyl-N'-nitro-N-nitrosoguanidine (MNNG) per ml for 55 min at 37°C, after which 99% of the cells had been killed. Cells were harvested by centrifugation, washed twice with M9 salt solution (38), and then incubated for 2 h in TY medium at 37°C. They were then transferred into M9 minimal medium (38) and incubated for 3 h and then overnight in the presence of 0.1 mg of cycloserine per ml. After being plated on M9 medium supplemented with tryptophan, tyrosine, and phenylalanine (0.1 mg/ml each) and p-amino benzoic acid and p-hydroxybenzoic acid (0.04 mM each), they were replica plated on M9 minimal medium. Strains DF1 to DF4 were able to grow only on M9 medium supplemented with all of the above-mentioned nutrients, indicating a biosynthetic block prior to chorismate. Overnight cultures (10 ,u) of strains DF1 to DF4 placed on filter paper disks supported growth of strain H1443 aroB seeded on M9 minimal medium agar, which had been supplemented with tryptophan, tyrosine, and phenylalanine at a concentration (0.004 mM) which was not sufficient to support growth of H1443. Strains DF1 to DF4 secreted a compound that fed H1443, placing the DF1 to DF4 mutations beyond the aroB mutation. Shikimate did not support growth, suggesting mutation sites in genes between shikimate and chorismate encoding biosynthesis functions (or inability to take up shikimate).
In a further experiment, S. marcescens was treated with MNNG, and selection for iron supply mutants was performed in NB medium in the presence of streptonigrin (5 ,ug/ml) as described previously (9). The resulting strain, NB22, grew weakly on NBD agar plates and showed a large yellow zone around colonies grown on chrome azurol S plates (see below).
The plasmids used are listed in Table 1. Their construction was described previously with the exception of the following plasmids. pAA2 was obtained by MluI digestion of pAAl followed by circularization. pAA100 contains the HindIII fragment of pSZ1 in pACYC184. pLA200 was constructed by cloning the larger EcoRI-EcoRV fragment of pSZ1 into pLG339 cut with EcoRI and SmaI. pSV100 contains the fecCDE genes of plasmid pST18, from which the smaller SalI fragment was excised (obtained from S. Veitinger). pAN302 contains the BamHI fragment of pEN2 carrying the iut gene (25) in pACYC184. Iron supply assays. Growth was determined on NBD or NBD* agar plates (17.5 ml) seeded with 0.2 or 0.1 ml of an overnight culture of the strain to be tested in 3 ml of overlay NBD or NBD* agar. Filter paper disks (8-mm diameter) were placed onto the NBD or NBD* agar, which contained 15 ,ul of the following 1 mM solutions: enterochelin, ferrichrome, coprogen, schizokinen, ferrioxamine B, and rhodotorulic acid. Several concentrations were used for assaying growth promotion on plates by dihydroxybenzoate (1 and 10 mM), sodium citrate (100, 50, 10, 5, and 1 mM), and iron citrate (5, 1, 0.5, and 0.05 mM iron). Growth in liquid cultures was measured in NBD medium and in NBD medium supplemented with 0.1 mM sodium citrate. Iron citrate solutions contained a 40-fold citrate surplus. Growth stimulation of the Sfu system with sodium citrate via FecA-TonB was also determined on NB plates containing 0.1 mM EDDA. Cells (2 x 107) were layered onto the NB-EDDA plates in 3 ml of NB soft agar (0.7%). To assay growth promotion by hemin, hemin (Sigma) was washed with 10 mM HCl and water and then dissolved (1 mM) in 10 mM NaOH. Sizes of growth zones around the disks were estimated by measuring their diameters (9,24,25). Production of siderophores was tested by placing 15 ,ul of an overnight culture of the producer strain in NB medium onto filter paper disks which were placed on NBD agar plates seeded with an indicator strain which could grow only when supplied with a siderophore that the indicator strain was able to use (24,25). Siderophore production was also tested on agar plates supplemented with chrome azurol S (48). Blue dye turned yellow around siderophore-producing colonies. For the quantitative determination of catecholate siderophores in culture supernatants of cells grown overnight, the method of Arnow (4) was employed.
Determination of iron transport rates. Cells were grown in 20 ml of NB medium or in NB medium supplemented with 0.1 mM sodium citrate to a density of 3 x 108 cells per ml. They were harvested by centrifugation and washed twice with 40 mM 3-(N-morpholino)propanesulfonic acid (MOPS) adjusted to pH 7.2 with KOH, 0.5 mM CaCl2, 50 mM NaCl, 100 mM MgSO4, 0.2% glucose, or M9 medium in the case of iron transport studies using sodium PP1. Cells were suspended in 5 ml of the wash solution supplemented with 0.1 mM nitrilotriacetate to a density of 5 x 108 cells per ml and incubated for 5 min prior to the addition of 50 ,ul of the radioactive 55Fe3" solution. The latter solution for the iron transport experiment in the presence of sodium PP1 consisted of 1. After incubation with V8 protease of Staphylococcus aureus (30 ,ug/ml) at 56°C for 0, 60, 90, and 120 min, samples (0.1 ml) were taken, and 50 ,ul of 30% trichloroacetic acid was added. The precipitates were washed with 0.1 ml of acetone and 0.1 ml of acetone-water (1:1) and then dissolved in the sample buffer for sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE).
Electrophoresis of outer membrane proteins. Outer membranes were prepared by lysis of cells with lysozyme followed by solubilization of the cytoplasmic membrane with Triton X-100 and differential centrifugation, as has been previously described in detail (19). The proteins were separated by electrophoresis on an 11% polyacrylamide gel in the presence of 0.1% SDS (19).
Recombinant DNA techniques. Isolation of chromosomal DNA and plasmids, use of restriction enzymes, ligation, agarose gel electrophoresis, recovery of fragments from agarose, transformation, and Southern DNA-DNA hybridization were done as previously described (49).

RESULTS
Sfu-mediated iron uptake in E. coli. Because of the lack of defined transport mutants of S. marcescens, the Sfu transport system was studied with E. coli K-12 mutants transformed with cloned sfi genes. The sfuABC genes confer upon E. coli aroB mutants unable to produce enterochelin the ability to grow on iron-limiting NBD. We have previously shown that E. coli single mutants lacking the outer membrane iron siderophore receptor proteins FhuA, FepA, FecA, FhuE, Cir, and Fiu were not impaired in Sfu-mediated iron transport (52). Since Fe3+ transport via dihydroxybenzoyl serine can alternatively utilize three receptors (Fiu, FepA, and Cir) (28) and catechol-substituted cephalosporins can utilize two receptors (Cir and Fiu) (16, 40), we measured growth of the triple-receptor mutant H1876 fiu cirfep transformed with plasmid pAA100 sfuABC on NBD agar plates. No growth inhibition was observed, which indicates that these receptors are not involved in Sfu-mediated iron transport.
In order to inhibit growth of E. coli sfuABC transformants on nutrient broth, the dipyridyl concentration had to be increased to 0.4 mM (NBD*). The Sfu system failed to enable E. coli K0280 fhuB, which lacks the cytoplasmic membrane protein FhuB, and K0295 fhuD, which lacks the periplasmic FhuD protein, to grow around filter paper disks containing ferrichrome or coprogen on NBD* plates. This means that iron carried as ferrichrome and coprogen across the outer membrane could not be transported via the Sfu system across the cytoplasmic membrane. The same results were obtained with dihydroxybenzoate, which did not support growth of the fepC mutant AN260 (26). This mutant cannot transport ferric dihydroxybenzoate across the cytoplasmic membrane and for this reason is also unable to synthesize enterochelin from dihydroxybenzoate.
In contrast to the above results, iron citrate was able to support Sfu-mediated iron transport. Strain ZI418 fecB:: Mud1, which is devoid of the periplasmic FecB protein required for citrate-mediated iron transport (44,49), showed after transformation with plasmid pAA100 on NBD* medium a distinct growth zone around filter paper disks containing citrate. The Mudl insertion exhibits a strong polar effect on the expression of the downstream fecCDE genes, as shown by complementation experiments with cloned fecBCDE genes (51a). Since the fecBCDE gene products are required for citrate-mediated iron transport across the cytoplasmic membrane, iron transport in this transformant was not catalyzed by the Fec system but rather by the Sfu system. However, uptake of iron dicitrate into the periplasm depends on the FecA outer membrane receptor protein and the TonB protein (33,44). Therefore, we tested whether the Sfucatalyzed, citrate-mediated iron uptake also required the FecA and TonB proteins. Indeed, no growth of strain Z1323 fecA(pAA100) or strain BR158 tonB(pSZ1) on NBD* plates was observed. No distinct growth zones appeared around filter paper disks with sodium citrate after 2 days of incubation at 27°C. Growth rates and the final cell density of strain ZI418 fecA+ tonB+(pAA100) in NBD liquid culture supplemented with 0.1 mM citrate were also twice as high as those of strains Z1323 fecA(pAA100) and BR158 tonB(pSZ1). J. BACTERIOL.
Apparently, iron uptake across the outer membrane under the extreme iron starvation conditions in NBD* medium requires the FecA receptor and the TonB protein (and presumably energy, which has not been examined).
Since dipyridyl is primarily an Fe2" chelator, additional experiments have been done with the Fe3+ chelator EDDA. The experimental conditions had to be adapted to this very strong iron chelator (stability constant, 33 for EDDA compared with 17 for dipyridyl). In fact, we could not observe any improved growth of E. coli H1443 transformed with plasmid pAA100 compared with growth of the untransformed strain on NB agar containing 0.1 mM EDDA. However, when EDDA was given only to the solid NB nutrient agar (1.5%) and the number of cells seeded in 3 ml of NB soft agar (0.7%) over the solid agar was reduced from 1 x 108 to 2 x 107, a clear growth promotion zone around filter paper disks with 100 mM sodium citrate could be discerned. Growth promotion was dependent on a functional FecA-TonB outer membrane transport system, since growth of fecA and tonB mutants was not supported by citrate.
Since iron dicitrate served as an iron donor for the Sfu system, we attempted to ascertain whether the SfuBC proteins could be replaced by the functionally analogous Fec-CDE proteins. The SfuB protein, like the FecCD proteins, is very hydrophobic and is located in the cytoplasmic membrane. The SfuC protein and the FecE protein are hydrophilic but associated with the cytoplasmic membrane and contain nucleotide-binding domains, so that both could drive the Sfu-mediated iron transport through ATP hydrolysis, as has been shown for histidine (5) and maltose transport (17).
Strain H1443 transformed with the two plasmids pAA2 sfuA and pSV100 fecCDE did not grow on NBD, excluding transfer of iron from the periplasmic SfuA protein to the FecCDE proteins in the cytoplasmic membrane.
In order to obtain quantitative data, transport rates with 5sFe3" citrate were measured. Cells were grown in NB medium supplemented with citrate. The available iron in this medium is low but sufficient for normal growth in contrast to the extremely iron-deficient conditions in NBD*. All fec mutants, including fecA mutants, which were transformed with pAA100 sfiABC, were able to transport iron at high rates ( Fig. 1, curves 1, 2, and 3). Under conditions of low iron supply, the FecA receptor is not required, which also holds true for the TonB protein. No transport into untransformed cells was seen (Fig. 1, curves 5, 6, and 7). Iron transport via the Sfu system was faster than that via the Fec system of strain AB2847 fec+ (Fig. 1, curve 4). This may be caused by the multiple sfuABC gene copies in the transformants in contrast to the single fecABCDE copy in AB2847. However, transformation of ZI418 fecB (polar on fecCDE expression) with plasmid pLZ30, which carries a 5'-truncated 'fecA fecBCDE, did not increase the citrate-mediated iron transport rate but instead slightly decreased it (data not shown). Previously, it has been shown by growth and transport assays that sodium PP, serves as an iron donor for the Sfu system at low iron concentrations (52). At the extremely iron-deficient conditions in NBD*, sodium PP, up to 30 mM did not support growth. At 40 mM sodium PP1, growth was observed; this growth, however, did not depend on the Sfu system, since it also occurred to the same extent in the untransformed H1443 aroB strain and the untransformed BR158 aroB tonB strain. The sodium PP1-mediated iron transport rates (measured at 0.25 mM sodium PPj) of cells grown in NB were similar to those of citrate-mediated transport via the Sfu system, indicating that more than one fecB(pAA100) (curve 2), ZI325 fecA fecB(pAA100) (curve 3), and the untransformed strains Z1342 (curve 5), ZI418 (curve 6), and ZI325 (curve 7) lacking plasmid pAA100. For comparison, citratemediated iron transport into E. coli AB2847 via the Fec system has been included (curve 4). ligand could solubilize Fe3+ and donate it to the Sfu system (Fig. 2, curves 1 and 2, transformants carrying pAA100). The same strains transformed with plasmid pLZ30 'fecABCDE were unable to transport Fe3+ supplied as sodium PP1 salt (curves 3 and 4).
Iron regulation of sfu gene expression. A potential consensus sequence for the Fur iron repressor protein (11,18) was found at nucleotides 54 to 72 upstream of the sjuiA gene (3), suggesting iron-regulated transcription of the sfiiABC genes. To examine this notion, plasmid pLA200 containing the sfuABC genes on the low-copy-number vector pLG339 was transferred into E. coli SC1l fir. The iron transport rates obtained in the presence of sodium PPi were three times higher than those obtained for the parent strain H1717 fiur+(pLA200) (data not shown).
Attempts to identify the substrate for the Sfu system. The only transport protein which can readily be isolated under mild conditions and therefore in an active state is the periplasmic SfuA protein. It was released by transferring cells from a medium of high osmolarity into a medium of low osmolarity (osmotic shock treatment [35]). The E. coli cells contained the sfuABC genes on plasmid pAAl downstream of the phage T7 gene 10 promoter, which was transcribed by the T7 RNA polymerase encoded on plasmid pGPl-2 under the control of a temperature-sensitive repressor (51). The cellular RNA polymerase was inhibited by rifampin. Under these conditions, only the SfuA protein of the periplasmic fraction was labeled with [35S]methionine, as revealed by SDS-PAGE (Fig. 3). We then tried to degrade SfuA by using proteases to see whether binding of the substrate would alter the peptide degradation pattern. This approach has been successful in the case of the and coprogen) protected the periplasmic protein from being degraded (35). The SfuA protein was largely resistant to degradation by trypsin and proteinase K. Only the staphylococcal V8 protease at 56°C degraded part of SfuA (Fig. 3, lanes 1 to 4). Degradation could be partially prevented by the addition of 0.1 mM iron chloride (lanes 5 to 8) and to a lesser extent but reproducibly by 0.1 mM iron citrate (ratio of Fe3+ to citrate, 1:20) (lanes 9 to 12). When the citrate concentration was increased (ratio of Fe3+ to citrate, 1:200) to further reduce the free-iron concentration, no protection was observed (lanes 13 to 16). There was also no inhibition of SfuA degradation by ferrichrome (lanes 17 to 20). These data suggest that Fe3+ may bind directly to the SfuA protein and that citrate may function as an iron carrier from which iron dissociates to bind to the SfuA protein. Iron chloride did not inhibit the V8 protease, since the FhuD protein was degraded as fast in its presence as in its absence (data not shown). Siderophore production of S. marcescens W225. Studies to see whether the unusual Sfu system is the only iron transport system of S. marcescens have been undertaken. A yellow zone appeared around colonies of S. marcescens grown on chrome azurol S nutrient agar plates, indicating that the cells secreted a compound that withdrew iron from the blue dye and turned it yellow. To identify the presumed siderophore(s), growth promotion of E. coli indicator strains on NBD medium was tested around filter paper disks containing samples of the S. marcescens culture. The E. coli indicator bacteria produced no siderophore and were either impaired in the uptake of Fe3+ enterochelin (IR20fepA, IB1691fepB) and its degradation product dihydroxybenzoyl serine and its precursor dihydroxybenzoate (H1876 fepA fiu cir), other catecholates (H1875fepA cir, H1877fepA fiu, H1728fiu cir), Fe3+ aerobactin (H1443 iut), ferrichrome (K0280 fhuB), coprogen (MS172 fluE), and Fe3`dicitrate (ZI314 fecB) or were deficient in all iron uptake routes (BR158 tonB, W3110-6 exbBD). Stimulation of growth by secretion products of S. marcescens W225 was observed only with E. coli cells which took up enterochelin. The fep mutants showed no growth zone (fepB) or a reduced growth zone (fepA). The weak growth of the latter strain was not observed with strain H1876, indicating utilization of enterochelin degradation products or other catecholate compounds as iron sources. Only the triple mutant H1876 showed no growth, while growth of the E. coli double mutants carryingfepA cir, fepA fiu, andfiu cir mutations was supported. E. coli mutants with mutations in tonB and exbBD did not respond to S. marcescens secretion products. Apparently, enterochelin was the only secretion product of S. marcescens which could serve as a siderophore for E. coli.
Determination of catecholates in the culture supernatant by using the Arnow procedure resulted in A564/A578 ratios of 1.92 for S. marcescens W225 and 0.14 for E. coli W3110. There was no enterochelin-specific absorbance for AB2847 aroB. These data show that S. marcescens secretes much larger amounts of enterochelin into the culture supernatant than E. coli W3110. Secretion of aerobactin was tested with E. coli H1876(pAN302) iut+, which takes up but does not synthesize aerobactin. No growth stimulation was observed with S. marcescens W225, excluding aerobactin synthesis by this strain. The additional S. marcescens strains tested (Table 1) were all aerobactin negative. In contrast, Serratia liquefaciens and Serratia ficania stimulated growth of E. coli H1876(pAN302), indicating synthesis and secretion of aerobactin.
Use of Fe3" siderophores by S. marcescens W225. To test whether S. marcescens takes up Fe3" siderophores, which it does not synthesize itself, aro mutants were isolated to avoid competition with the genuine enterochelin. Four such strains were obtained after MNNG mutagenesis and cycloserine selection. They showed no yellow zones around colonies on chrome azurol S plates, and the Arnow test also revealed no catecholate compound in the culture supernatant. However, strains DF1 and DF4 growing on filter paper disks supported somewhat reduced growth of E. coli indicator strains able to take up enterochelin. The diameters of the growth zones were 11 mm; those of the parent strain were 16 mm (diameter of the filter paper disk, 8 mm). Since the diameter of the  (578 nm) the NBD* medium. When iron citrate instead of sodium citrate was used, growth of DF2R was stimulated down to a concentration of 0.05 mM. Growth of the tonB mutant was also supported but only down to a concentration of 0.5 mM. At all iron concentrations used (5, 1, 0.5, and 0.05 mM), the 2 growth zones of the tonB mutant were smaller than those of the tonB+ parent strain (3.4 and 2.5, 2.5 and 1.5, 2.0 and 1.1, 0.9 and 0 cm, respectively). Of the two control strains (E. coli H1443 and BR158 tonB) employed, only growth of H1443 tonB+ was supported by sodium and iron citrate, 3 whereas the tonB mutant did not grow. Growth of H1443 was supported down to a sodium citrate concentration of S mM and down to an iron citrate concentration of 0.5 mM. p */ 4 An additional mutant (NB22) obtained after MNNG mutagenesis and streptonigrin selection formed a larger yellow zone on chrome azurol S agar plates than the aro+ parent 5ĩ /strain. However, NB22 was unable to grow on NBD* plates g * > 5 (0.5 mM dipyridyl), and growth was not enhanced by the addition of enterochelin. Apparently, strain NB22 synthesized and secreted enterochelin but was unable to take up Hybridization of DNA fragments from E. coli iron transport genes with chromosomal DNA fragments of S. marcescens. To corroborate the data on S. marcescens iron transport sysvth zone is logarithmically related to the concentration of tems and to obtain an indication of how closely those compound to be tested, the difference between wild type systems are structurally related to the E. coli transport mutants is substantial. Mutants DF2 and DF3 did not systems, we hybridized fragments excised from E. coli )ort growth of any of the E. coli indicator strains, so they transport genes with fragments of the chromosome of S. enterochelin negative according to this very sensitive marcescens W225, which was cleaved with the same restricy. However, upon prolonged incubation, low growth tion enzymes as the E. coli genes. The plasmids used for ulation was also obtained with DF2 and DF3. They isolating the DNA fragment are listed in Table 1. The DraI Lrently secreted an enterochelin precursor which was fragment offlhuA composed about half of the gene, the KpnI n up and converted to enterochelin by the E. coli aroB fragment of fluE was 90% of the gene, the PvuII-EcoRV cator strains (H1875, H1877, and H1728). Growth of the fragment was 95% of iut, the HindII fragment was the entire e mutant H1876 and of BR158 tonB, which do not take tonB gene plus 800 bp of flanking E. coli DNA, the EcoRI atecholate compounds, was not supported. fragment was the entire exbB gene, the HindIII-BglI fragcubation in NBD liquid cultures also demonstrated the ment was half of the fur gene plus 0.2 kb of the vector, and rent growth rates of strains DF1 and DF2 (Fig. 4). It also the SalI-BamHI fragment was the entire fecA gene plus a ved the growth-enhancing role of the sfu genes when E.
portion of the fecB gene. The 4.8-kb HindIII fragment of sfu H1443 was transformed with pSZ1 sfu. served as a positive control. Hybridization with flhuA, iut, :rains DF1 to DF4 were used to determine growth exbB, and fur was obtained (data not shown). The sizes of ulation on NBD* agar plates (containing 0.5 mM dipythe fragments of S. marcescens and E. coli were the same for 1), onto which the siderophores enterochelin, aerobactin, fluA (1.5 kb) and iut (2 kb) and differed for e-xbB (3.7 and 2 chrome, schizokinen, coprogen, rhodotorulic acid, ferkb, respectively) and fur (1 and 2 kb, respectively). No amine B, and hemin at concentrations of 1 mM (15 ,ul) on hybridization between a 4.8-kb HindIII DNA fragment of the r paper disks were placed. All siderophores supported S. marcescens sfu region and HindIII-digested chromosomal vth of the S. marcescens aro mutants to a similar extent. DNA of E. coli H1443 was observed.
oli sfiwABC transformants were unable to use heme iron, Iron-regulated outer membrane proteins. The large variety he heme uptake system in S. marcescens remains un-of Fe3+ siderophores that supported growth of S. marcesvn.
cens W225 prompted an examination of iron-regulated outer itrate-mediated iron uptake into S. marcescens DF2R membrane proteins, which in E. coli and many other gramved a complex behavior. On NBD* plates (containing negative bacteria serve as receptors for Fe3+ siderophores. nM dipyridyl) seeded with DF2R, growth zones of 3, 2, We used the rough derivative DF2R of W225 to avoid wavy 1.6 cm appeared around filter paper disks soaked with 15 protein bands after SDS-PAGE caused by the 0 antigens. f 50, 10, and 5 mM sodium citrate, respectively. No Synthesis of S. marcescens proteins in the molecular weight vth zone was observed with 1 mM sodium citrate. A range of 70,000 to 85,000 was low in TY medium, which I mutant of DF2R exhibited no growth, suggesting a contains sufficient amounts of iron (Fig. 5, lane 1). Reduction ptor-and TonB-dependent uptake of iron across the of the available iron by addition of dipyridyl (1 mM) resulted r membrane. Under these conditions, citrate had to in a large increase in the amounts of these proteins (Fig. 5, pete for iron with the high concentration of dipyridyl in lane 2). Growth on nutrient broth medium, which contains , which agrees with the superior iron supply of the latter via the Sfu transport system. Addition of citrate suppressed synthesis of these proteins except for the FecA protein (lane 10). The electrophoretic mobilities of the iron-regulated proteins and the major outer membrane proteins differed in Serratia spp. and E. ccli.

DISCUSSION
The Sfu iron transport system of S. marcescens accepts Fe 3 1solubilized with oxaloacetate, sodium PP1. and citrate (52). This excludes a ligand specificity for uptake across the outer membrane and stands in sharp contrast to all the siderophore-mediated iron transport systems hitherto characterized, which exhibit a high ligand specificity that is even narrower for iron transport across the outer membrane than for transport across the cytoplasmic membrane (7,8). We therefore conclude that the Sfu system transports iron without a specific ligand. However, not all Fe"~siderophores served in iron uptake via Sfu. Ferrichrome, coprogen, and dihydroxybenzoate were inactive in experiments with E. ccli sfuBC transformants that were lacking components required to transport these siderophores across the cytoplasmic membrane (flhull and fepC). This also held true under extremely iron-deficient conditions (0.4 mM dipyridyl or 0.1 mM EDDA), under which citrate was an active iron donor for the Sfu system. Interestingly, this extreme iron deficiency required the FecA iron citrate receptor and the TonB protein for transport across the outer membrane, since E. coli sfuABC transformants mutated in fecA and tonB could not grow with sodium citrate. E. coli sfuABC transformants mutated in the fecBCDE genes, which are required for citrate-mediated iron transport across the cytoplasmic membrane, took up iron, showing that citrate carried iron across the outer membrane and the Sfu system carried it across the cytoplasmic membrane. Apparently, at high dipyridyl concentrations and in the presence of the strong iron chelator EDDA, entry of iron through the outer membrane is too low to support growth. For this reason, transport catalyzed by FecA and TonB is necessary. The iron citratemediated transport for the Sfu system across the outer membrane clearly demonstrates receptor and TonB dependence under conditions of strong iron deficiency and no such requirement under a lower iron limitation. We take this as evidence that under the latter conditions, the iron diffusion rate is high enough to fulfill the iron requirement. At very low iron concentrations, the energy-coupled receptor and TonB-dependent uptake across the outer membrane are necessary.
The reason for the inability of ferrichrome, coprogen, and dihydroxybenzoate to serve as iron donors for the Sfu system, despite their efficient uptake across the outer membrane, may be sought in the very high stability constants of these iron complexes (ferrichrome, 29.1; coprogen, 30.2; dihydroxybenzene, 43.7 for dihydroxybenzoate) (30), which presumably exceed that of iron citrate by several orders of magnitude (numbers at pH 7 are for various reasons not available [22,37]). The iron citrate complex has just the right properties in that it is strong enough that an outer membrane transport system has evolved for transport of the complex and sufficiently weak that it donates iron to the Sfu system. The other siderophores tested are apparently inactive in donating iron to the Sfu system in the periplasm. Also, EDDA is less suitable than dipyridyl for characterizing the Sfu iron transport system because of its very high iron complex formation constant, so that iron is not transferred from EDDA to Sfu.
The primary acceptor for iron in the Sfu system is probably the periplasmic SfuA protein. Indeed, binding of iron supplied as iron chloride and iron citrate could be demonstrated by showing an altered V8 protease degradation in their presence compared with the pattern obtained in their absence. In contrast, the transport-inactive ferrichrome exerted no effect on protease degradation of SfuA. These results are similar to the results obtained with the FhuD protein, where only the Fhu transport-active iron siderophores of the hydroxamate type (aerobactin, ferrichrome, and coprogen) prevented degradation by proteinase K and trypsin, whereas the transport-inactive ferrichrome A, iron citrate, and iron sulfate were without effect (35). We found no metal-binding motif in the SfuA protein.
The SfuABC-mediated iron transport rate in E. coli sfuABC transformants is higher than the iron citrate transport rate via the Fec system. This indicates a high efficiency of transport across the cytoplasmic membrane which is sufficient to support growth despite the presumably low entry rate of iron through the outer membrane.
A highly conserved region has been identified in the cytoplasmic membrane components of a number of PBPdependent uptake systems, including the family of iron uptake proteins (33b). If a matrix reflecting this pattern is generated and used for a search in a data base (PC/GENE program MATSCAN, SwissProt Release 17), it is possible to distinguish between members of the iron and noniron uptake systems. Interestingly, the corresponding region of SfuB (amino acids 419 to 450) is detected only with the noniron matrix. The possibility of distinguishing between iron and noniron uptake systems with such a matrix is in agreement with the fact that all the iron uptake systems possess an outer membrane TonB-dependent receptor, whereas the other uptake systems do not need one. Furthermore, significant homologies between all the cytoplasmic membrane components of the iron uptake systems except SfuB (33a) have been detected (PC/GENE program PCOMPARE). These findings support the unique properties of the Sfu iron transport system. Enterobacteriaceae usually express more than one iron transport system (7,8,15,41). E. coli K-12 has at least seven Fe3+ and one Fe2+ transport systems. To obtain a more complete view of the iron transport systems of S. marcescens, of which we had studied only the Sfu system because of its unusual properties, we performed a survey study, without going into detail, when the systems seem to be similar to those of E. coli. The results obtained indicate the existence of Fe3+ enterochelin, Fe3+ aerobactin, ferrichrome, Fe3+ coprogen, ferrioxamine B, Fe3+ citrate, and hemin transport systems. Enterochelin was the only siderophore synthesized by the S. marcescens strains tested. Aerobactin was not formed by S. marcescens but was formed by S. liquefaciens and S. ficania. However, S. marcescens contained an Fe3+ aerobactin transport system. S. marcescens and S. liquefaciens are the two clinically relevant species of the genus Serratia. Apparently, S. marcescens contains fep, fes, ent, fhu, fec, iut, and fiu (cir) genes, which are equivalent to those characterized in E. coli. S. marcescens also expresses a number of iron-regulated outer membrane proteins. Five protein bands could be identified on gels, indicating at least five iron receptors. Unfortunately, it was not possible to correlate the proteins with the various iron transport systems by isolating drugresistant mutants. Growth inhibition by albomycin and the rifamycin derivative CGP4832 (45) (both in E. coli, indicative for the FhuA receptor) and by catecholate-substituted cephalosporins (Cir and Fiu) (16,40) occurred at higher concentrations than in E. coli, and no inhibition was observed with the colicins (M for FhuA, and B and D for FepA), so that resistant, frequently receptor deletion mutants were not obtained with these agents. In these assays, a rough derivative (DF2R) has been used to avoid partial resistance due to the long lipopolysaccharide 0 antigens, which may inhibit access to the receptors. Therefore, we have to conclude that the transport components of S. marcescens differ from those of E. coli so that no or only weak cross-sensitivity to these agents exists. On the other hand, hybridization with chromosomal DNA fragments of S. marcescens and a number of DNA fragments of E. coli iron transport genes (fhuA, iut, exbB, andfur) was obtained, despite the difference in the GC content of the two organisms (S. marcescens, 58%; E. coli, 50%). Although no hybridization with the tonB gene was found, sequencing of the S. marcescens gene revealed a high homology to the E. coli protein but too low a DNA homology to be detected at the stringent conditions used in the hybridization experiments (23a). Therefore, the failure to find additional DNA hybridizations does not necessarily demonstrate large structural and functional differences from the E. coli transport systems. Growth on Fe3+ coprogen also suggests an FhuE-equivalent coprogen receptor in S. marcescens, despite the lack of hybridization with the E. coli fluE gene. A wide distribution of FhuE equivalent receptor proteins among gram-negative bacteria is suggested by the occurrence of such a protein in the plant growth-promoting Pseudomonas putida strain WCS358 (6). The protein, designated PupA, shows the greatest homology (31.4%) to FhuE among all the iron receptor proteins. It serves as a receptor for pseudobactin 358, which contains both a hydroxamate and a catecholate group.
Two systems are probably involved in citrate-mediated iron uptake in S. marcescens. The experiments with E. coli sfuABC transformants clearly indicated that citrate served as an iron donor for the Sfu system. The same mechanism was probably operating in the TonB-independent uptake of added iron citrate in S. marcescens DF2R tonB. On the other hand, when iron had to be withdrawn by sodium citrate from NBD* medium, growth of DF2R depended on TonB activity. The low amounts of iron citrate formed under these conditions required a receptor and TonB-dependent uptake through the outer membrane. However, we were unable to induce an outer membrane iron citrate receptor protein by adding citrate, as has been done in E. coli (44). Therefore, other than the physiological data, we have no protein analysis or genetic indications of a citrate-mediated iron transport system in S. marcescens.
Threefold enhancement of the Sfu-mediated iron transport rate in an E. coli fur mutant compared with uptake in the fur+ parent strain suggests a Fur-regulated expression of the sfubABC genes. In addition, synthesis of the iron-related outer membrane proteins of S. marcescens was strongly increased after cells were grown in iron-limiting dipyridylcontaining medium. Interestingly, the S. marcescens hemolysin genes (shlA and shlB) are also derepressed by a factor of 10 when cloned into an E. coli fur mutant (42). Sequences typical of Fur-binding regions do occur in the -35 region upstream of the sftA (3) and shlB (42) genes, and they are apparently recognized by the E. coli Fur repressor. It seems that S. marcescens shares several iron transport systems as well as their regulation with E. coli and other gram-negative bacteria and that only the Sfu system is unique for S. marcescens.