Biosynthesis of carbapenem antibiotic and prodigiosin pigment in Serratia is under quorum sensing control



Serratia sp. ATCC 39006 produces the carbapenem antibiotic, carbapen-2-em-3-carboxylic acid and the red pigment, prodigiosin. We have previously reported the characterization of a gene, carR, controlling production of carbapenem in this strain. We now describe further characterization of the carR locus to locate the genes encoding carbapenem biosynthetic and resistance functions. A novel family of diverse proteins showing sequence similarity to the C-terminal domain of CarF (required for carbapenem resistance) is described. We also report the isolation of the locus involved in the biosynthesis of the red pigment, prodigiosin. A cosmid containing ≈ 35 kb of the Serratia chromosome encodes synthesis of the pigment in the heterologous host, Erwinia carotovora, demonstrating, for the first time, that the complete prodigiosin biosynthetic gene cluster had been cloned and functionally expressed. We report the isolation of a third locus in Serratia, containing convergently transcribed genes, smaI and smaR, encoding LuxI and LuxR homologues respectively. SmaI directs the synthesis of N-acyl homoserine lactones involved in the quorum sensing process. We demonstrate that biosynthesis of the two secondary metabolites, carbapenem antibiotic and prodigiosin pigment, is under pheromone-mediated transcriptional regulation in this bacterium. Finally, we describe a new prodigiosin-based bioassay for detection of some N-acyl homoserine lactones.


Serratia sp. ATCC 39006 is able to produce the β-lactam antibiotic, 1-carbapen-2-em-3-carboxylic acid (carbapenem; Car) and the red pigment, prodigiosin (Williams and Quadri, 1980; Parker et al., 1982). We have investigated the genetics and biochemistry of carbapenem production in both Serratia and Erwinia carotovora and, in the latter species, have identified a cluster of nine genes involved in the production of the antibiotic (carRABCDEFGH;McGowan et al., 1996). With the exception of carH, we have been able to assign putative functions to each of the genes within the cluster. We have shown that these genes encode not only the enzymes of the carbapenem biosynthetic pathway, but also a novel β-lactam resistance mechanism and a positive regulator of carbapenem gene expression (McGowan et al., 1998).

The product of the first gene of the cluster, CarR, is a member of the LuxR family of transcriptional regulators (McGowan et al., 1995). CarR activates transcription of the remaining genes of the car cluster in response to N-(3-oxohexanoyl)-l-homoserine lactone (OHHL) synthesized by the product of the unlinked carI gene. This OHHL-dependent transcriptional activation allows the cells to coordinate expression of carbapenem with cell density and has been termed ‘quorum sensing’ (Bainton et al., 1992a; Fuqua et al., 1994).

We have previously shown that carbapenem and prodigiosin production in Serratia sp. ATCC 39006 is coordinately regulated by a novel regulatory protein, Rap (Thomson et al., 1997). The homologue of Rap in E. carotovora, Hor, also controls carbapenem production, while other Rap homologues regulate diverse physiological processes in an array of human and animal pathogens (Dehoux and Cossart, 1995; Thomson et al., 1997). By analogy with other closely related proteins, such as SlyA, PecS and MarR (Cohen et al., 1993; Reverchon et al., 1994; Oscarsson et al., 1996), Rap and Hor are thought to bind DNA directly to activate transcription of the genes within their regulon.

Although there are similarities in the control mechanisms of carbapenem production in Serratia and Erwinia, there are also several notable differences. Rather than the growth-phase dependent induction of the antibiotic in E. carotovora, for example (Bainton et al., 1992b), Serratia sp. ATCC 39006 reportedly produces carbapenem in parallel with growth (Bycroft et al., 1988). In addition, we have recently described the cloning and characterization of a homologue of the E. carotovora carR gene that controls carbapenem expression in Serratia (Cox et al., 1998). However, unlike its counterpart in E. carotovora, the product of this gene (which was also called carR) is reported as capable of functioning in a pheromone-independent fashion to activate carbapenem biosynthesis (Cox et al., 1998). In view of the observed differences in carbapenem production between Serratia sp. ATCC 39006 and E. carotovora, we set out to characterize the Serratia carbapenem biosynthetic operon, presumed to be located downstream of the carR gene.

In addition to the expression of carbapenem, Serratia sp. ATCC 39006 also produces a red pigment that is thought to be a member of the prodigiosins (Parker et al., 1982). Although perhaps one of the most well-known characteristics of the genus, the production of prodigiosin, or 2-methyl-3-pentyl-6-methoxyprodigiosin, is restricted to only three species: Serratia marcescens, Serratia rubidaea and Serratia plymuthica (Grimont and Grimont, 1984). Recent attention has focussed on the clinical utility of prodigiosins produced by species of Serratia and Streptomyces. Prodigiosins have been found to have potent immunosuppressive activities that selectively inhibit the proliferation of cytotoxic T-lymphocytes without affecting antibody formation (Tsuji et al., 1990; 1992). Research in murine models has shown that prodigiosins may have potential as supplementary therapeutic agents in the suppression of delayed type hypersensitivity and graft compared with host reactions (Nakamura et al., 1986; Kataoka et al., 1992; Ryohei et al., 1992; Magae et al., 1996). Prodigiosins also have other wide ranging biological properties, including antibacterial, antimalarial, antifungal and antiprotozoal activities (Williams and Quadri, 1980; Demain, 1995) and there remains considerable interest in the chemical synthesis of prodigiosin analogues which have potent immunosuppressive activities, but lack cytotoxicity (Gerber, 1975; Boger and Patel, 1987; Wasserman and Lombardo, 1989; D′Alessio and Rossie, 1996; D′Auria et al., 1999).

As well as the notion that prodigiosin plays an active role in the competitive survival of Serratia spp. (Williams and Quadri, 1980; Demain, 1995), there have been many other theories proposing a function for this pigment (Burger and Bennet, 1985; Paruchuri and Harshey, 1987). Curiously, the majority of clinical S. marcescens isolates are non-pigmented, whereas a far higher proportion of environmental isolates produce pigment (Grimont and Grimont, 1978; Ding and Williams, 1983; Aucken and Pitt, 1998). However, the explanation for this observation and the true physiological role of prodigiosin, if any, remains equivocal. Therefore, prodigiosin may conform to the classical description of a secondary metabolite, having no essential role in the growth or survival of the cell, but instead acting as an overflow for ‘waste’ products from primary metabolism (Bu'Lock, 1961). Supporting evidence for this idea comes from studies on the related pigment, undecylprodigiosin, from Streptomyces, which acts as a sink for excess proline (Hood et al., 1992).

The highly visible nature of prodigiosin has given it obvious appeal as a model system for the study of secondary metabolite production. Classical syntrophic cross-feeding studies have shown that prodigiosin is formed via a bifurcated pathway terminating in the enzymatic condensation of 2-methyl-3-amylpyrrole (MAP) and 4, methoxy-2.2′-bipyrrole-5-carboxyaldehyde (MBC; Williams et al., 1971; Williams and Quadri, 1980). Studies on the control of prodigiosin production have shown that the pigment is produced between trophophase and idiophase (Williams et al., 1971), consistent with the classical definition as a secondary metabolite (Bu'Lock, 1961). In addition, multiple extrinsic factors have been shown to affect prodigiosin production, such as media composition, pH and temperature (Williams et al., 1971; Williams and Quadri, 1980; Sole et al., 1994). Other studies have suggested that there are common regulatory links between the production of prodigiosin and several other phenotypes, such as the production of protease, haemagglutinating activity and flagellar variation (Paruchuri and Harshey, 1987; Hines et al., 1988; Goluszko et al., 1995). However, the genetic explanations for these observations remain unclear. Similarly, little is known about the enzymes or genes involved in the production of this pigment. In an attempt to clone the genes involved in this biosynthetic pathway, Dauenhauer et al. (1984) succeeded in isolating a S. marcescens genomic clone which, in Escherichia coli, could direct the condensation of MAP and MBC to form prodigiosin when they were both applied exogenously. However, no sequence data was reported for this clone and, despite the considerable attention paid to this secondary metabolite, the prodigiosin biosynthetic cluster has never been isolated as a functional unit from Serratia.

In this article, we report the characterization of the complete carbapenem biosynthetic cluster and the cloning and heterologous expression of the prodigiosin gene cluster of Serratia sp. ATCC 39006. We define a novel family of proteins, found in both prokaryotes and eukaryotes, that exhibit homology to the carbapenem resistance protein, CarF. We also demonstrate that the production of both secondary metabolites, carbapenem and prodigiosin, is coordinately regulated via a quorum sensing system.


Subcloning the Serratia carbapenem gene cluster

The Serratia sp. ATCC 39006 genes, tentatively presumed to encode the biosynthesis of carbapenem, had been isolated previously by direct cosmid complementation of E. carotovora group 1 carbapenem mutants (strains defective in carR or carA–H, Table 1; Cox et al., 1998). Complementation by the two Serratia cosmids, pNRT1 and pNRT20, was identical to the complementation achieved with the E. carotovora cosmid cWU142 known to contain the cluster of nine car-related genes (McGowan et al., 1997; Cox et al., 1998). We therefore carried out reciprocal Southern blot hybridizations using cWU142 as the probe against restriction digests and fragments of pNRT1 and pNRT20 (data not shown). Non-vector homology was observed in the 1.0 kb and 5.9 kb BamHI fragments contiguous with the 2.8 kb BamHI fragment of pNRT1 previously shown to contain the Serratia carR gene (Cox et al., 1998). No hybridization to any of the remaining fragments of pNRT1 and pNRT20 was observed.

Table 1. Bacterial strains, plasmids and phages used in this study.
StrainGenotype or phenotypeaSource or reference
  • a

    . Ap, ampicillin; Sp, spectinomycin; Sm, streptomycin; Cm, chloramphenicol; Kn, kanamycin; Tc, tetracycline.

Escherichia coli
 DH1FsupE44 recA1 endA1 gyrA96 thi-1 hsdR17 (rk inline image) relA1 Hanahan (1983)
FendA1 hsdR17 (rk inline image) supE44 thi-1
recA1 gyrA (Nalr) relA1D(lacIZYA – argF) U169 deoR
Bethesda Research Laboratories
 ESSβ-lactam super sensitive Bainton et al. (1992a)
 ATCC 39006wild type (Car+, Pig+) Bycroft et al. (1987)
 CWM52ATCC 39006 (pMUT13), Pig, carR Cox et al. (1998)
 CWT12ATCC 39006 (pTROY9), Pig, carR Cox et al. (1998)
 CWT11, 13, 14,
 15, 18 and 31
ATCC 39006 (pTROY9), Pig, Car
Cox et al. (1998)
 Lac-A lacZ derivative of ATCC 39006This study
 LC13 smaI::mini-Tn5lacZ1, KnrThis study
 LIS smaI::mini-Tn5Sm/Sp, SprThis study
 MCA54 carA::mini-Tn5lacZ1, KnrThis study
MCCIA smaI::mini-Tn5Sm/Sp, carA::mini-Tn5lacZ1, Spr KnrThis study
Erwinia carotovora ssp. carotovora
 ATM101GS101, hor Thomson et al. (1997)
 GS101wild type Bainton et al. (1992a)
 PNP11 and 14GS101, carR McGowan et al. (1995)
 PNP21GS101, Car McGowan et al. (1995)
 PNP22GS101, carI McGowan et al. (1995)
 SCC3193wild typeSwedish Culture Collection,
 E. T. Palva
 SCRI102, 103,
 105, 106, 109,
 112–118, 120–127,
 130, 132, 135,
 139, 144, 149,
 152, 155, 166,
 169, 171, 172,
 174, 192, 193
 and 198
 wild type
Scottish Crop Research Institute
Plasmids and Cosmids
 pACYC184Cmr, Knr, low-copy cloning vector Chang and Cohen (1978)
 PbluescriptApr, lacZ, high-copy cloning vectorStratagene
 pMAC30pBluescript + ≈ 30kb Serratia strain LC13 chromosomal insertThis study
 pMAC555.5kb NotI subclone of pMAC30 carrying the partial ORF smaI′This study
 pMAC1282.3kb SfiI–SalI sunclone of pMAC30 carrying the partial ORFs'smaI and'smaRThis study
 pMUT13 lamB +, KnrE.T. Palva
 pNRT1 Serratia car genes in pSF6 Cox et al. (1998)
 pNRT20 Serratia car genes in pSF6 Cox et al. (1998)
 pNRT104pSF6 + mini-Tn5 and ≈ 25 kb of Serratia chromosomal insertThis study
 pNRT324 Serratia rap gene Thomson et al. (1997)
 pSF6Cloning vector; cos, mob, Spr Selvaraj et al. (1984)
 pT7–5T7φ10 promoter Tabor and Richardson (1985)
 pT7–6T7φ10 promoter Tabor and Richardson (1985)
 pT7–25≈ 3.4kb SspI–SalI fragment of pNRT1 in pT7–5; carA, carB, carC′This study
 pT7–35≈ 1.8kb BamHI–SalI fragment of pNRT1 in pT7–5; carB, carC′This study
≈ 4.6kb BamHI–EcoRI fragment of pNRT1 in pT7–6; carB, carC, carD, carE,
carF, carG′
This study
 pT7–56≈ 4.0kb SalI–BamHI fragment of pNRT1 in pT7–6; carD, carE, carF, carG, carHThis study
 pT7–65≈ 1.3kb EcoRI–BamHI fragment of pNRT1 in pT7–5; carHThis study
 pTON10210.1kb fragment of pNRT1 (BamHI partial) in pACYC184; Serratia carR-H genesThis study
 PTROY9pLAFRB malK::IS-3, LamB+ De Vries et al. (1984)
 pUTmini-Tn5Sm/Sp De Lorenzo et al. (1990)
 pUTmini-Tn5lacZ1  De Lorenzo et al. (1990)
 cWU142 E. carotovora car genes in pSF6 McGowan et al. (1995)
 λSerratia library Cox et al. (1998)
 φKP E. carotovora generalized tranducing phage Toth et al. (1993)
 φOT8 Serratia generalized tranducing phageUnpublished results

From a partial BamHI digest of pNRT1, a 10.1 kb fragment (comprising 2.84 kb, 1.04 kb, 5.94 kb and 0.4 kb BamHI fragments) was cloned in pACYC184 to give pTON102. In common with the parental cosmid, subclone pTON102 complemented all the carbapenem mutants of Serratia sp. ATCC 39006 and E. carotovora (data not shown). It was highly likely therefore, that the contiguous 2.8 kb, 1.0 kb, 5.9 kb and 0.4 kb BamHI fragments contained most, if not all, of the Serratia genes involved in carbapenem production. However, in contrast to the E. carotovora carbapenem biosynthetic cluster (McGowan et al., 1996) attempts at reconstitution of carbapenem production in E. coli using pTON102, or the parental cosmid, failed. The exogenous addition of OHHL or cotransformation with pNRT324, containing the Serratia carbapenem-regulatory gene (rap;Thomson et al., 1997) also failed to bring about any carbapenem production in E. coli.

Sequence analysis of the predicted Serratia carbapenem gene region

Analysis of the sequence data generated from the 2.8 kb, 1.0 kb and 5.9 kb BamHI fragments of pNRT1 revealed eight complete open reading frames (ORFs) in the region downstream of the carR gene (GenBank entry AF012907). Each of the eight ORFs exhibited a high degree of sequence similarity at the DNA level to the genes involved in carbapenem biosynthesis in E. carotovora (Fig. 1) and were therefore designated carA, carB, carC, carD, carE, carF, carG and carH. The similarity between the car gene clusters of both species was confined to the actual ORFs, with no significant sequence similarity apparent in the DNA outside the two regions. In order to confirm that carA–H were genuine ORFs, 35S-methionine labelling was used in the T7 expression system (Tabor and Richardson, 1985) to identify the gene products from a series of subclones carrying different regions of the car gene cluster (Table 1). Products corresponding to each of the eight putative genes in the Serratia carbapenem gene cluster were identified (data not shown).

Figure 1.

The relative size, organization and homology of the nine genes, and their predicted products, from the carbapenem gene clusters of Serratia sp. ATCC 39006 and E. carotovora.

Comparative sequence analysis of the car clusters of both species also revealed a number of other features within the sequence that have been conserved. The carR–carA intergenic regions of each cluster contain a direct repeat overlapping a region of hyphenated dyad symmetry which in each case would allow the formation of a stable stem–loop structure (Welch et al., 2000, and unpublished data). Immediately downstream of this intergenic region, both species initiate translation of carA at a GTG codon rather than an ATG codon.

We have previously reported evidence that the CarF and CarG proteins of E. carotovora encode a novel resistance function, allowing self-resistance to carbapenem (McGowan et al., 1997). Both proteins were originally reported as having no homologues in the database and the mechanism of resistance is not yet known. (McGowan et al., 1996). However, a large number of protein sequences have since been reported that each exhibit homology with the C-terminal end of CarF from both Serratia sp. ATCC 39006 and E. carotovora (Fig. 2).

Figure 2.

Alignment (using Clustalw analysis) of the C-terminal amino acid sequences of the CarF proteins from Serratia sp. ATCC 39006 and E. carotovora, and the C-termini of various predicted proteins from the database. The positions of the three motifs along a linear representation of Serratia CarF have been indicated. The database homologues are from a variety of sources: Serratia sp. ATCC 39006 (CarF – this paper); E. carotovora (CarF, GenBank U17224); A. thaliana (GenBank AQ249564); B. stearothermophilus (XaiF, GenBank U15985); B. subtilis (YfmG, GenBank D86417); B. pertussis (Sanger); C. crescentus (TIGR); C. pneumoniae (Cpn0148, GenBank AE001601); C. trachomatis (Pkn1, GenBank AE001288); D. melanogaster (GenBank AC006169); H. sapiens (expressed sequence tag − THC132493, TIGR); M. avium (TIGR); M. tuberculosis 1 (MTCY210.31, GenBank Z84395), M. tuberculosis 2 (MTV025.051c, GenBank AL022121); N. meningitidis (Sanger); N. gonorrhoeae (GenBank M97926 and Sanger); P. gingivalis (TIGR); P. aeruginosa 1 (ORF2, GenBank U07359); P. aeruginosa 2 (PGP); P. fluorescens (GenBank U69893); Pseudomonas sp. G-179 (NirV, GenBank AF040987); Synechocystis sp. strain PCC6803 (slr1303, GenBank D90901); V. cholerae (TIGR). Where applicable genome sequence data was retrieved from: Sanger (The Sanger Centre, UK at, TIGR (The Institute for Genomic Research, USA at and PGP (The Pseudomonas genome project at Dark and light shading has been used to indicate positions at which at least 11 of the 23 residues are either identical or similar respectively.

Isolation of the Serratia prodigiosin biosynthetic cluster

When the Serratia sp. ATCC 39006 chromosomal library (Cox et al., 1998) was used to complement the E. carotovora group 1 carbapenem mutants, some transductants exhibiting a red pigmented phenotype (Pig+) were observed (data not shown). Approximately 1% of the Spr colonies, recovered after transduction with the Serratia chromosomal library, produced this bright-red pigment, presumed to be prodigiosin. However, interestingly, in an analogous experiment to isolate additional regulatory elements common to both E. carotovora and Serratia, no Pig+ transformants were observed when the Serratia sp. ATCC 39006 chromosomal library was introduced into an E. carotovora carI mutant (strain PNP22) deficient for the production of OHHL. Restriction analysis of one of the cosmids, pNRT104, recovered from a pigmented CarI+E. carotovora strain GS101 transductant, revealed that pNRT104 carried ≈ 35 kb of chromosomally derived insert (data not shown).

Prodigiosin displays characteristic absorbance maxima in isopropanol of 466 nm under alkali conditions (appearing orange in colour) and 540 nm under acid conditions (appearing red–pink in colour) (Williams and Quadri, 1980). To confirm that the red pigment encoded by cosmid pNRT104 was prodigiosin, an overnight culture of E. carotovora strain GS101 harbouring pNRT104 was extracted with isopropanol. Isopropanol-extracted pigment was acidified or made basic, by the addition of one drop of either HCL (10 M) or NaOH (10 M), respectively, and the absorbance was measured between the wavelengths of 350 nm and 600 nm. The characteristic absorbance peaks at 466 nm and 540 nm were observed under acid and alkali conditions, confirming that the red pigment produced by E. carotovora strain GS101 (pNRT104) was indeed a prodigiosin (data not shown).

Prodigiosin production in heterologous hosts and colonial patterning

In the light of the previous, unproductive, attempts to reconstitute prodigiosin production in E. coli (Dauenhauer et al., 1984), we wanted to discover whether the ability to heterologously express prodigiosin from cosmid pNRT104, was peculiar to E. carotovora strain GS101. Various members of the Enterobacteriaceae (Table 1) including E. coli, Proteus mirabilis and Enterobacter agglomerans, as well as additional strains of E. carotovora, were transformed with pNRT104. Transformants were selected on media supplemented with spectinomycin. In all cases successful transfer and integrity of cosmid pNRT104 was confirmed by reextraction of cosmid pNRT104 and subsequent restriction analysis of the recovered cosmid DNA.

Of the Enterobacteriaceae tested, only certain strains of E. carotovora carrying pNRT104 exhibited a Pig+ phenotype (data not shown). Eleven of the 36 E. carotovora ssp. carotovora strains tested exhibited a Pig phenotype when transformed with pNRT104. The remaining 25 strains exhibited markedly different patterns of pigment production (Fig. 3A), highlighting some interesting aspects of differential gene expression within the bacterial colony. For example, colonies of strain SCRI114 display concentric rings of pigmentation, whereas strain SCRI144 only produces pigment at the centre of each colony. Colonies of strains SCRI205 and SCRI174 are pigmented throughout but show rings of more intense pigmentation at different positions within the colonies. The molecular biology and physiology underlying this differential gene expression inside a colony will be investigated further. Erwinia carotovora subspecies atroseptica was also able to express pigment encoded by pNRT104, and differences in the expression patterns between strains were also observed in this genetic background (data not shown).

Figure 3.

A. Colonial patterning of prodigiosin gene expression in five strains of E. carotovora carrying pNRT104. The strains of E. carotovora used were: (i) SCRI114; (ii) SCRI144; (iii) SCRI174; (iv) SCRI109; and (v) SCRI205.

B. Expression of prodigiosin following exposure of E. carotovora to OHHL. Paper discs soaked with either (i) water, or (ii) OHHL (5 µl of 5 mM) were placed on a lawn of E. carotovora strain PNP22 (carI) carrying pNRT104.

The regulation of prodigiosin production in a heterologous genetic background

As previously described, when the E. carotovora carI mutant strain PNP22 was transduced using the Serratia sp. ATCC 39006 chromosomal library, no red pigmented colonies were observed. This observation led us to believe that the heterologous expression of prodigiosin, in E. carotovora, may require the bacterial pheromone, OHHL. To test this hypothesis, E. carotovora strain PNP22 was transformed with cosmid pNRT104. None of the recovered Spr transformants, harbouring pNRT104, exhibited a red (Pig+) phenotype. However, pigment production could be restored by the exogenous addition of OHHL, or by cross-feeding with the wild-type E. carotovora strain GS101 (Fig. 3B).

Exogenous addition of OHHL to the other members of the Enterobacteriaceae, harbouring pNRT104, had no effect on their ability to express the prodigiosin encoded by pNRT104 (data not shown). In addition, because all of the E. carotovora ssp. carotovora strains that were transformed with pNRT104 (this study) were known to produce OHHL (Holden et al., 1998), the requirement for OHHL did not explain the inability of 11 of the 36 E. carotovora ssp. carotovora strains to express pigment. Moreover, this indicated that, although OHHL is essential for prodigiosin expression in E. carotovora, an additional factor(s) is also required.

The heterologous expression of prodigiosin in an E. carotovora hor mutant

One of the few known positive regulators of prodigiosin expression in Serratia is encoded by the rap gene (Thomson et al., 1997). E. carotovora possesses a closely related homologue of Rap, Hor. To determine whether Hor was also required for the heterologous expression of prodigiosin, pNRT104 was introduced into E. carotovora, strain ATM101 (hor) by conjugation. Transconjugants were selected on minimal salts media supplemented with spectinomycin, but the resulting transconjugants failed to express pigment (data not shown). The ability to express prodigiosin encoded by pNRT104 could be restored to E. carotovora strain ATM101 by supplying hor in trans using plasmid pNRT324 (data not shown).

Clearly, in E. carotovora, the expression of prodigiosin biosynthetic genes on pNRT104 is subject to both Hor- and OHHL-mediated control. Based on the high level of sequence similarity between Rap and Hor it was perhaps not surprising that Hor also modulated the heterologous expression of prodigiosin in E. carotovora. However, as previously mentioned, there was no evidence for OHHL production in this strain of Serratia, despite the fact that it possesses a gene, carR, related to the transcriptional activators involved in quorum sensing (Cox et al., 1998). Consequently, the dependence of heterologous pigment expression in Erwinia on the production of OHHL led us to reexamine the possibility that Serratia sp. ATCC 39006 did in fact produce an N-acyl homoserine lactone(s) (AHLs).

Detection of bacterial pheromones in Serratia sp. ATCC 39006

To re-evaluate the possibility that Serratia sp. ATCC 39006 produced an N-acyl homoserine lactone(s) we tested culture supernatants with both the Chromobacterium violaceum strain CV026 bioassay (McClean et al., 1997), and the lux bioassay, using E. coli strain JM101 transformed with the lux biosensor plasmid pSB401 (Winson et al., 1998). The former assay differs from the lux bioassay in that it is able to detect a broader range of AHLs via the induction and interference bioassays (McClean et al., 1997). Although the lux biosensor failed to detect AHL production in this strain of Serratia[Ahl, Cox et al. (1998) and data not shown], a defined blue halo was observed on the C. violaceum assay (Fig. 4A). This result indicated that Serratia sp. ATCC 39006 does indeed produce an AHL(s), although not OHHL. We therefore decided to pursue the nature and biological function(s) of the AHL(s) eliciting a response in the CV026 bioassay.

Figure 4.

A. Detection of AHL(s) produced by Serratia sp. ATCC 39006. A 3 µl aliquot of the Serratia strain was spotted onto a lawn of CVO26.

B. Activation of prodigiosin and carbapenem biosynthesis in Serratia sp. strain LIS (smaI). Pigment and antibiotic production is restored in strain LIS, left spot, by cross-feeding from the red pigmented AHL producing strain MCA54, right spot. Then, 3 µl of each strain was inoculated onto a lawn of E. coli strain ESS. The carA strain was used because the antimicrobial action of the wild-type in the E. coli lawn would obscure the zone of clearing caused by the production of carbapenem by strain LIS.

C. Identification of AHLs produced by Serratia sp. ATCC 39006. AHL extracts from overnight cultures of Serratia sp. ATCC 39006 (wt) and strain LIS (smaI) were resolved on TLC plates and developed using the LIS (left panel) and CV026 (right panel) biosensor strains. Samples were resolved alongside synthetic BHL, HHL and OHL (middle panel shows the superimposed standards lanes in the LIS and CV026 bioassays).

Random transposon mutagenesis and isolation of AHLSerratia mutants

Random mini-Tn5lacZ1 transposon mutants of strain Lac-A were isolated as described (see Experimental procedures). Of the 11 000 putative Serratia mini-Tn5lacZ1 transposon insertion mutants tested, 0.3% failed to produce a blue halo when screened on the C. violaceum strain CV026 bioassay and were therefore considered to be defective for the production of the putative bacterial pheromone (Ahl, data not shown). One of these strains was isolated and named LC13. In an identical manner, another transposon-induced Ahl mutant of Serratia (strain LIS) was generated using mini-Tn5Sm/Sp. Interestingly, all of the Serratia Ahl mutants isolated were also phenotypically Pig and, when tested on the E. coli strain ESS bioassay, were also unable to produce the β-lactam antibiotic, carbapenem (Car) (data not shown). Both pigment and antibiotic production could be restored to these strains by cross-feeding from the wild-type Serratia sp. ATCC 39006 (Fig. 4B) as well as by the exogenous addition of either N-butanoyl-l-homoserine lactone (BHL) or N-hexanoyl-l-homoserine lactone (HHL) (Fig. 5A). Attempts to cross-feed the AhlSerratia mutants with E. carotovora strain GS101, or the exogenous addition of OHHL to an Ahl-deficient strain, failed to restore carbapenem production or pigment biosynthesis (data not shown) unless physiologically unrealistic and excessive concentrations were provided (e.g. see Fig. 5A).

Figure 5.

Comparison between AHL-detection bioassays. The two test strains (A) Serratia strain LC13 and (B) Chromobacterium violaceum strain CV026 were plated out as lawns and exposed to 1 ml of each of the various AHLs (at ≈ 2.5 mM). The AHLs used were: BHL (N-butanoyl-l-homoserine lactone), OBHL [N-(3-oxobutanoyl)-l-homoserine lactone], HHL (N-hexanoyl-l-homoserine lactone), OHHL [N-(3-oxohexanoyl)-l-homoserine lactone], OHL (N-octanoyl-l-homoserine lactone), OOHL [N-(3-oxooctanoyl)-l-homoserine lactone], DHL (N-decanoyl-l-homoserine lactone), ODHL [N-(3-oxodecanoyl)-l-homoserine lactone], dDHL (N-dodecanoyl-l-homoserine lactone), OdDHL [N-(3-oxododecanoyl)-l-homoserine lactone].Next, 1 ml Acetonitrile was also spotted onto the lawn as a control.

All of the above results suggested that the Serratia pheromone(s) was BHL and/or HHL. When Serratia supernatant extracts were tested in the C. violaceum TLC bioassay, species comigrating with BHL and HHL were detected − although relatively weakly (Fig. 4C). In contrast, no activity was detected in SmaI mutant supernatants. The CV026 bioassay strain is more sensitive to HHL than BHL (Fig. 5B). Therefore, the more intense violacein production due to the putative BHL from Serratia suggests that, although HHL is made by this strain of Serratia, BHL is made in considerably higher concentration.

Furthermore, when the SmaI strain (LIS) was used as a biosensor in place of the CV026 mutant, prodigiosin production was strongly restored by supernatant extracts from wild-type Serratia but not the SmaI mutant. The active molecule comigrated with BHL. No activity comigrating with the HHL control was detected. However, this LIS biosensor assay strain responds less well to HHL than to BHL, even at high concentrations (Fig. 5A). Once again, these observations are consistent with the notion that Serratia sp. strain ATCC 39006 makes considerably more BHL than HHL.

Finally, to confirm absolutely that the mutants with an Ahl phenotype were the result of a single transposon insertion in each case, the novel generalized transducing phage, φOT8, was first used to transduce the transposon insertions into a clean genetic background. All of the resultant antibiotic-resistant transductants exhibited the same pleiotropic Pig, Car, Ahl phenotype as the parental strain (data not shown). We then proceeded to use a transposon-tagged Ahl mutant to clone the gene responsible for pheromone production in the Serratia strain.

Isolation of the Serratia smaI and smaR genes

Chromosomal DNA from strain LC13 was digested with various restriction enzymes that would not cut within the transposon (De Lorenzo et al., 1990) and hybridized with a mini-Tn5lacZ1 DNA probe in a Southern blot. Southern blot analysis revealed that the transposon insertion was located on a 20–30 kb pstI fragment. Serratia chromosomal DNA was redigested with pstI and DNA fragments of > 20 kb were recovered. The size fractionated DNA was ligated into pBluescript (Stratagene) and used to transform E. coli strain DH5α. Plasmid pMAC30 was isolated from one of the transformants and found to carry the mini-Tn5lacZ1 transposon and ≈ 25kb of flanking Serratia chromosomal DNA.

Restriction analysis of pMAC30 (data not shown) showed that the mini-Tn5lacZ1 transposon could be isolated, along with flanking Serratia chromosomal DNA, on a 5.5 kb NotI fragment (containing the transposon lacZ gene 5′ region and ≈ 2.1 kb of Serratia chromosomal DNA) and on a 2.8 kb SfiI–SalI fragment (containing the transposon kanamycin resistance gene 3′ region and ≈ 0.8 kb of Serratia chromosomal DNA). The 5.5 kb NotI fragment was ligated into the compatible EagI site of pACYC184 (Chang and Cohen, 1978), forming pMAC55. The 2.8 kb SfiI–SalI fragment was end-filled and ligated into the SmaI site of pACYC184, forming pMAC128.

Partial sequence analysis of pMAC128 revealed the presence of the 3′ end of two partial ORFs, denoted ′smaI and ′smaR (secondary metabolite activator). ′smaI and ′smaR were convergently transcribed and overlapped by 84 bp at the 3′ ends. Preliminary database searches predicted that the partial protein product of ′smaI was homologous to members of the LuxI family of autoinducer synthases and that the ′smaR protein product was homologous to members of the LuxR family of transcriptional activator proteins. Oligonucleotide primer DRPR, designed to the ′smaR sequence, was used to complete the sequence of smaR. The complete smaR gene was predicted to encode a protein of 247 amino acids with a predicted molecular weight of 29 kDa. Database searches revealed that the putative SmaR protein shared a significant level of amino acid identity with EchR (66% similarity, 55% identity) and ExpR (64% similarity, 55% identity) from E. chrysanthemi and E. carotovora ssp. carotovora respectively.

Plasmid pMAC55 was also partially sequenced and was found to carry an incomplete ORF, predicted to encode the N-terminus of a protein with sequence similarity to autoinducer synthases and therefore likely to be the missing 5′ end of ′smaI (assuming no chromosomal rearrangements had occurred during the transposon insertion). To confirm the sequence of the smaIsmaR region, primers PR18A and URPR, designed to sequences 5′ of the convergently transcribed smaI and smaR genes (respectively), were used to amplify this region from wild-type Serratia sp. ATCC 39006 chromosomal DNA. A product of ≈ 1.8kb was amplified and sequenced by ‘primer walking’ (see Experimental procedures). The complete smaI ORF was predicted to encode a protein of 234 amino acids with a molecular weight of 27 kDa. Database searches revealed that SmaI shared significant sequence similarity with SwrI (76% similarity, 66% identity) and EcbI (67% similarity, 53% identity) from Serratia liquefaciens and E. carotovora ssp. betavasculorum respectively. To determine the nature of any genes contiguous with smaI and smaR we extended the nucleotide sequence upstream of both genes, using the progenitor plasmid pMAC30 as template DNA (see Experimental procedures). We identified one complete ORF, a partial ORF and a gene remnant upstream of smaI. The protein product of the complete ORF, denoted fecE, displayed significant sequence identity with the E. coli proteins FecE (41%) and FepC (40%), both of which are iron transport proteins (GenBank accession numbers: M26397and X57471respectively; Staudenmaier et al., 1989; Shea and McIntosh, 1991). Immediately upstream of fecE, the partial ORF (designated ′fecD) was predicted to encoded a protein that shared a significant level of similarity with FecD (47% identity, 56% similarity; GenBank accession number M26397) from E. coli. This protein is also thought to be involved in the active uptake of iron (Ochsner and Vasil, 1996) and, like the Serratia homologue, is also found upstream of fecE (Staudenmaier et al., 1989).

In addition to the complete and partial ORFs fecE andfecD, we identified a sequence of 190 bp (within the smaIfecE intergenic region) that displayed significant nucleotide sequence identity (81%) to the 3′ end of the insB gene (required for the active transposition of insertion element IS1;Ohtsubo et al., 1981). Analysis of the sequence upstream of smaR identified a hypothetical ORF with no significant protein homologues. However, we also found a 24 bp sequence that shared a high level of sequence identity with a portion of the Tn903 insertion sequence. The complete 4.7 kb quorum sensing locus of Serratia sp. ATCC 39006 has been deposited in the GenBank database under accession number EMBL AJ275980.

AHL-mediated regulation of carbapenem production in Serratia sp. ATCC 39006

In order to confirm the quorum sensing control of transcription of the Serratia car genes, we decided to examine the transcriptional expression of one of these genes, carA. We isolated strain MCA54 (carA::mini-Tn5lacZ1) following mutagenesis of strain Lac-A using the mini-Tn5lacZ1 transposon. PCR amplification of chromosomal DNA was used to confirm the generation of a transcriptional fusion between the promoterless lacZ gene of the transposon, and the carbapenem biosynthetic gene, carA. A smaI derivative of Serratia sp. ATCC 39006 was isolated following mutagenesis using mini-Tn5Sm/Sp and was named LIS. A φOT8 lysate made from strain LIS was used to transduce strain MCA54, selecting streptomycin resistant transductants, thereby generating strain MCCIA (carA::mini-Tn5lacZ1, smaI::mini-Tn5Sm/Sp). The β-galactosidase activity of the carA::lacZ fusion strains MCA54 and MCCIA was then determined by the method of Miller (1972)(Fig. 6).

Figure 6.

Levels of β-galactosidase activity throughout growth resulting from a carA::lacZ fusion in strain MCCIA (smaI::mini-Tn5Sm/Sp) in the absence of exogenously added AHL (–▪–), or in the presence of either 0.5 µM BHL (–▴–) or 0.5 µM HHL (–●–) added to the culture at the start of growth. The level of β-galactosidase may be compared with that of the same fusion in a wild-type background (strain MCA54; –◆–). Growth (absorbance at 600 nm) is represented by appropriate open symbols.

These data indicated that the transcription of the carA::mini-Tn5lacZ1 fusion in strain MCA54 was induced after ≈ 5 h, reaching a maximum β-galactosidase activity after 10 h. The highest β-galactosidase activity generated by strain MCCIA was ≈ 2.5% of the equivalent activity recorded for the progenitor strain MCA54. The exogenous addition of either BHL or HHL, to the growth medium of strain MCCIA (at time point zero), restored the peak β-galactosidase activity to 85% and 56%, respectively, of the peak activity of strain MCA54. Clearly, the transcription of the car biosynthetic genes is controlled by the quorum sensing molecules.



We set out to analyse the Serratia sp. ATCC 39006 carbapenem biosynthetic cluster in order to compare the mechanisms by which Serratia and Erwinia produce carbapenem. Our interest focused on both the evolutionary relatedness of the biosynthetic clusters and on the differences in regulation of carbapenem production between these bacteria.

We had previously shown that Serratia sp. ATCC 39006 possessed a carR gene that was similar to the first gene in the E. carotovora car cluster (McGowan et al., 1996; Cox et al., 1998). Analysis of the DNA downstream of carR revealed that all of the genes in the Serratia car cluster shared a high level of sequence identity with those found in E. carotovora. Interestingly, the level of conservation between the corresponding proteins of the two clusters was noticeably higher in those previously shown to be absolutely required for carbapenem biosynthesis (McGowan et al., 1997) (CarABC; 75.0–91.6%), when compared with the proteins encoded by the remainder of the cluster (59.3–71.7%), although the evolutionary and physiological reasons driving such conservation are not clear. As in E. carotovora, the unidirectional expression and organization in relation to carR, together with the potential for translational coupling between some of the genes, suggested possible operonic expression.

A number of other potential regulatory features within the car clusters of both species have also been conserved. For example, the carR–carA intergenic region (Welch et al., 2000) contains alternative inverted repeat and direct repeat sequences, in both Erwinia and Serratia. In Serratia the potential stem–loop is followed by a run of Ts, and it could function as a terminator to punctuate transcription between carR and the putative carA-H operon. However, more importantly, we have shown recently that physical binding of the CarR transcriptional regulator in the intergenic region is crucially influenced by the repeat sequences (Welch et al., 2000). Furthermore, as we have shown for E. carotovora (McGowan et al., 1996) the translational initiation codon of the Serratia carA gene is GTG, in place of the more common ATG codon. It is intriguing that this relatively unusual initiation codon has been retained in both species. The choice of initiation codon has been shown to not only affect the efficiency of translational initiation, but has also been shown to be important for other aspects of post-transcriptional auto-regulation (Parsons et al., 1988; Ringquist et al., 1992; Blattner et al., 1997). In tandem with a common and relatively poor ribosome binding site, the retention of the GTG initiation codon for the carA genes of both species could be important in maintaining the expression of CarA at levels below those of the other Car proteins encoded by the transcript. Further work will be required to understand the significance of the use of the rare GTG codon for carA.

Comparative sequence analysis also revealed that the carH gene is highly conserved between both Serratia sp. ATCC 39006 and E. carotovora (Fig. 1). In Erwinia there is no evidence, to date, of the involvement of CarH in either carbapenem biosynthesis or resistance (McGowan et al., 1997) and it displays no significant homology to any protein currently in the databases. Consequently, although we do not yet know the function of this gene, its conservation in both species argues strongly that it is involved in some aspect of either process.

Other proteins in the Erwinia and Serratia car clusters for which there were previously no protein homologues in the database include CarF and CarG (McGowan et al., 1997). We have shown that, in Erwinia, carF and carG encode a novel carbapenem self-resistance mechanism. More recent database searches, with both the E. carotovora and Serratia CarF proteins, have shown that many homologues of these proteins have now been reported. Homologues of CarF come from a diverse range of organisms, and include proteins from eukaryotes (Homo sapiens, Arabidopsis thaliana and Drosophila melanogaster) as well as prokaryotes (Fig. 2). The N-termini of these protein homologues exhibit no sequence relatedness either to CarF or, in most cases, to each other.

Like the resistance function of E. carotovora, where we have shown CarF to be exported across the cytoplasmic membrane (McGowan et al., 1997), some of these homologues appear to be associated with extracytoplasmic phenotypes. The xaiF(xynR) gene from Bacillus stearothermophilus for example, is adjacent to xynA encoding the secreted enzyme, xylanase and has been reported to be involved in its production (Cho and Choi, 1998). A partially sequenced gene from three different species of Pseudomonas is in each case adjacent to genes involved in the synthesis or uptake of the secreted siderophores, pseudobactin/pyoverdine. In Bacillus subtilis, the homologue YfmG is located immediately downstream of an operon of four genes (ymfCDEF) encoding an iron uptake system belonging to the ABC transporter family. It is also clear from genome sequence data that the homologues from Neisseria meningitidis and Neisseria gonorrhoeae are located immediately downstream of the ainA gene, encoding the major anaerobically induced, outer membrane protein, Pan1 − a copper-containing nitrite reductase (Mellies et al., 1997). In addition, like CarF, at least seven of the protein homologues were predicted to encode a N-terminal signal peptide to allow export of the protein across the cytoplasmic membrane.

Within this new family of proteins, the homology at the C-terminus is apparent as three distinct motifs (Fig. 2) at fixed relative positions and perhaps these motifs together might represent a single and novel protein domain. Work currently in progress to precisely locate the CarFGH proteins and to characterize the mechanism of resistance, may help to clarify the role of this putative novel domain.

Finally, it is interesting that the same regulatory proteins (CarR) from Serratia and Erwinia respond to different molecules (BHL and OHHL, respectively) in their cognate genetic backgrounds, yet activate similar target genes. Sequence analysis of the two proteins shows multiple differences and, interestingly, particularly in the putative AHL-binding domain. However, it is not yet possible to define precisely which residues are the key determinants of AHL binding specificity, i.e. representing a ‘recognition signature’. The molecular specificity of the Erwinia CarR protein has been defined recently (Welch et al., 2000), although the contact residues are not known. It will therefore be a project of considerable interest to determine precisely how the Erwinia and Serratia proteins differentiate between OHHL and BHL – and this will be pursued in future.

It is also curious that in Serratia, the SmaR protein shows only weak sequence identity (36.7%) with CarR from the same strain. To date, smaR mutants have no obvious phenotype (data not presented) and so its function is not transparent. Nevertheless, in keeping with the SwrR nomenclature of Serratia liquefaciens (Givskov et al., 1997) and the ExpR/RexR nomenclature of Erwinia carotovora (Pirhonen et al., 1993; Salmond et al., 1995) where neither proteins have a proven function, we have provisionally retained the smaR nomenclature for the gene encoding the LuxR homologue, because it is adjacent to smaI.


As a fortuitous consequence of using Erwinia as a surrogate host to clone the Serratia carbapenem biosynthetic and regulatory genes, we observed that several colonies harbouring cosmids from the Serratia sp. ATCC 39006 chromosomal library produced a red pigment. Solvent extraction of this pigment from overnight cultures of E. carotovora carrying the prodigiosin biosynthetic cosmid pNRT104, confirmed that the pigment encoded by this cosmid was in fact a prodigiosin. This represents the first reported cloning of the Serratia prodigiosin biosynthetic genes as a functional unit.

Prior to this study the biosynthetic cluster of the related pigment undecylprodigiosin (Red), produced by Streptomyces coelicolor A3(2), had also been cloned (Tsao et al., 1985). Because S. coelicolor A3(2) Red mutants were able to cross-feed defined S. marcescens prodigiosin mutants (Feitelson and Hopwood, 1983), undecylprodigiosin is thought to be produced along a similar biosynthetic route to that proposed for the biosynthesis of prodigiosin by S. marcescens. The red biosynthetic cluster of S. coelicolor A3(2) was cloned from a cosmid-based chromosomal library on a 35.7 kb chromosomal insert (Malpartida et al., 1990) and is thought to comprise at least 18 genes (Narva and Feitelson, 1990; Coco et al., 1991). Preliminary restriction analysis of the Serratia cosmid pNRT104 revealed that it includes ≈ 35kb of chromosomally derived insert (data not shown). Therefore, by analogy with the red cluster, it is likely that pNRT104 contains the entire prodigiosin biosynthetic (pig) cluster and a detailed analysis of this cosmid clone is now in progress.

Although the prodigiosin biosynthetic cosmid was introduced into an array of bacterial species, only members of the erwinias could support the heterologous production of this pigment (under the conditions tested). Our inability to reconstitute pigment production in E. coli is consistent with previous studies (Dauenhauer et al., 1984) and highlights the fact that the Erwinia spp. must contain one (or more) regulatory factor(s) required for pigment gene activation that is not present in E. coli. Moreover, the highly ordered patterning of prodigiosin expression observed in the heterologous E. carotovora colonies harbouring pNRT104 (Fig. 3A), suggests that the expression of the prodigiosin biosynthetic cluster fell under very precise and strain-specific host gene regulation systems. It is unlikely that the strain variation in pigment expression is a consequence of quorum sensing control of the prodigiosin biosynthetic genes. All of the E. carotovora strains tested were known to produce AHLs and, given the diffusibility of these molecules, it is unlikely that they alone are responsible for the patterning of colony pigment expression. In addition, when OHHL was spotted onto a lawn of E. carotovora strain PNP22 harbouring pNRT104, no discrete patterns of pigment expression were observed.

This is not the first time differential gene expression within a bacterial colony has been so readily observed (reviewed by Shapiro, 1998). However, the strong visual assay for gene expression that these recombinant E. carotovora colonies afford, should prove useful in the study of patterning of gene activation within the complex community of a bacterial colony, and this will be pursued in a separate study.

Quorum sensing

In E. carotovora strain GS101, the heterologous expression of prodigiosin was found to require OHHL and the hor gene product. We had anticipated that Hor, a closely related homologue of the Serratia Rap protein (a positive regulator of prodigiosin production in Serratia sp. ATCC 39006; Thomson et al., 1997) may affect the expression of prodigiosin in Erwinia. However, there was no evidence for a bacterial pheromone in Serratia sp. ATCC 39006 (Cox et al., 1998) and so it was difficult to see how the expression of prodigiosin could be regulated by quorum sensing in this heterologous host. To clarify this issue we used the lux based biosensor (Swift et al., 1993; Winson et al., 1998) and the recently developed Chromobacterium violaceum bioassay (McClean et al., 1997) in order to screen for any AHL production by this strain of Serratia. The data obtained from plate assays suggested that Serratia sp. ATCC 39006 produced AHLs. Production of BHL and HHL was confirmed by the use of two TLC-based bioassays exploiting the CV026 violacein reporter and the novel LIS prodigiosin reporter. Furthermore, the latter two assays suggested that BHL was the more abundant molecule in the Serratia strain.

Using a random transposon mutagenesis approach, we isolated two Ahl mutant trains, LC13 and LIS. In each case, the chromosomal DNA flanking the transposon was cloned and sequenced revealing the presence of two genes, smaI and smaR (secondary metabolite activator). Both of these genes encode homologues of proteins known to be involved in quorum sensing. The smaI gene product shared the highest level of sequence similarity with SwrI, the autoinducer synthase from S. liquefaciens (Eberl et al., 1996; Eberl et al., 1999). SwrI directs the synthesis of two bacterial pheromones, BHL and HHL (in an ≈ 10:1 ratio) that are involved in the regulation of swarming motility and the production of extracellular proteases. These data are consistent with the results obtained using the lux and Chromobacterium biosensors, plus TLC assays, which indicated that Serratia sp. ATCC 39006, like S. liquefaciens, produces some HHL, but considerably more BHL. In addition, the smaI mutants also exhibit a pleiotropic phenotype, being defective for both prodigiosin and carbapenem production. Cross-feeding by wild-type Serratia, or the exogenous addition of either BHL or HHL restored the smaI mutant to Car+, Pig+ (Fig. 4B and C, Fig. 5A and data not shown). Therefore, we have shown that production of both carbapenem and prodigiosin is regulated by quorum sensing in Serratia sp. ATCC 39006.

The production of an AHL(s) by Serratia sp. ATCC 39006 was initially surprising but, in fact, is not inconsistent with the findings of our previous report (Cox et al., 1998). The Serratia CarR protein is a homologue of the OHHL-dependent CarR protein of E. carotovora. We showed previously that the Serratia CarR protein is functionally interchangeable with the Erwinia homologue yet did not require the cognate Erwinia pheromone, OHHL, to activate the E. carotovora carbapenem gene cluster (Cox et al., 1998). In this study, we have shown, using the lux bioassay, that this strain of Serratia does not make OHHL. We have confirmed therefore the OHHL-independent nature of the Serratia CarR activity. We conclude that a reinterpretation of the previously reported activation of the car gene cluster in the absence of AHL, is that it may have been due to a copy number effect involving the plasmid-encoded carR gene in this heterologous genetic background. Despite the use of low-copy vectors in the latter experiments, these observations echo the need to conduct quorum sensing experiments using strains with single chromosomal genes (or, at least, in very low copy) if physiologically meaningful data are to be obtained.

It is interesting to note that the addition of high concentrations of OHHL to smaI mutant strain LC13 resulted in only trace amounts of prodigiosin being synthesized (Fig. 5A). In contrast, we have shown that, in E. carotovora, OHHL is essential for prodigiosin expression via PNRT104, and induces a high level of pigment production in this genetic background. We have recently shown that expression of prodigiosin production in recombinant E. carotovora strains is independent of the LuxR-type regulators, EccR/RexR/ExpR or CarR (data not presented). Therefore, we presume that novel DNA-binding transcriptional regulators that activate prodigiosin production in E. carotovora and Serratia remain to be discovered. It is tempting to speculate that these regulators must respond to different AHLs, while binding to the same DNA target sequences. Consequently, we are currently attempting to isolate these new, predicted activator proteins in order to compare the DNA-binding and inducer-binding domains of each.

β-Galactosidase assays of the lacZ transcriptional gene fusion to carA revealed that an intact autoinducer synthase, SmaI, is essential for the full transcription of this carbapenem biosynthetic gene. This was consistent with phenotypic studies that showed that the smaI mutant failed to produce any detectable carbapenem (this study). Surprisingly however, the apparent induction of β-galactosidase was delayed until around 5 h following inoculation of strain MCA54. At first sight, this apparent growth phase-dependent induction of carA transcription seems inconsistent with the report of Bycroft et al. (1988), that carbapenem was produced in parallel with growth. However, reexamination of the published data reveals that they are not inconsistent with our finding, and indicate that, like E. carotovora, the production of carbapenem by Serratia sp. ATCC 39006 is induced in an AHL-dependent manner in late-log to early stationary phase of the bacterial growth cycle.

The exogenous addition of BHL almost fully restored the β-galactosidase activity of the carA::mini-Tn5lacZ1 fusion to the levels of the progenitor strain, whereas the addition of HHL only partially induced the carA gene. These observations, combined with the failure to detect a C6 AHL using the lux assay (and the relatedness of SmaI and SwrI) strongly suggested that the cognate AHL produced by Serratia sp. ATCC 39006 is BHL and the production of BHL (and HHL) by this species was confirmed via TLC-based bioassays.

Interestingly, we have shown that the smaIsmaR locus has several features that are commonly found on pathogenicity islands in other bacteria (Shea and McIntosh, 1991; Hacker et al., 1997; Vokes et al., 1999). These include two putative ORFs, ′fecD and fecE (homologues of which are involved in the uptake of iron), several insertion sequence ‘gene remnants’ (which share significant levels of sequence identity with portions of insertion sequence IS1 and transposon Tn903) and the smaI and smaR genes themselves, homologues of which are associated with the transcriptional control of multiple virulence factors in other bacterial species. It is formally possible therefore that the smaI/R‘quorum sensing locus’ of Serratia sp. ATCC 39006 may have been part of a mobile genetic element at one time.

Finally, it has not escaped our attention that, like the C. violaceum strain CV026 bioassay, Serratia strains LIS or LC13 (smaI) also represent an aesthetically attractive, facile screen for AHL production. The Serratia-based bioassay is activated by AHLs with acyl side-chains of between C4 and C8, being most sensitive to BHL and HHL. Moreover, unlike the C. violaceum bioassay, the Serratia assay can also be used to detect OBHL (Fig. 5). Thus, by using our Serratia bioassay in conjunction with that of C. violaceum we have extended the range of AHLs that can be easily and accurately detected.

Experimental procedures

Bacterial strains and media

The bacterial strains and plasmids used during this study are listed in Table 1. A derivative (Lac-A) of Serratia sp. ATCC 39006 was identified as being LacZ following mutagenesis using EMS. Serratia and E. carotovora strains were routinely grown at 30°C and E. coli strains at 37°C in Luria broth. Minimal salts media was as described in Sambrook et al. (1989). The E. coli strain ESS bioassay for detection of carbapenem was carried out at 25°C (McGowan et al., 1997).

The Chromobacterium violaceum strain CV026 bioassay was performed as previously described by McClean et al. (1997). The Serratia strain LC13 bioassay was performed essentially as described for the C. violaceum strain CV026 bioassay with the exception of using a modified PG agar [peptone (0.5%), glycerol (1%) and 1.5% agar] (Morrison, 1969). When required, LB medium was supplemented with BHL and HHL at a final concentration of 0.5 mM. AHLs were provided by S. R. Chhabra.

DNA manipulations and genetic constructions

Generalized transduction experiments using φKP were carreid out as described in Toth et al. (1993). Transduction with the new Serratia bacteriophage φOT8 (unpublished; to be described elsewhere) used a similar protocol for lysate preparation and transduction as that used for φKP (Toth et al., 1993). A chromosomal DNA library of Serratia sp. ATCC 39006 was prepared and packaged into phage l as previously described in McGowan et al. (1995). Conjugal mobilization of pNRT104 was achieved in a biparental patch mating with E. coli strain DH1, containing the mobilizing plasmid pRK2013 (Ditta et al., 1980) and cosmid pNRT104, as the donor strain and various E. carotovora ssp. as the recipients. Transconjugants were selected on minimal salts media supplemented with spectinomycin and with sucrose (0.2% final concentration) as sole carbon source (to counter select the donor E. coli strain).

Random transposon mutagenesis of Serratia strain Lac-A (lacZ) was performed in a patch mating with E. coli strain S17–1 carrying plasmid pUTmini-Tn5lacZ1 or plasmid pUTmini-Tn5Sm/Sp (De Lorenzo et al., 1990) on complex media. Transconjugants were selected on minimal salts media with sucrose as sole carbon source and supplemented with kanamycin or spectinomycin respectively. Transconjugants in which the insertion of the mini-Tn5lacZ1 transposon had generated a transcriptional fusion were identified by the addition of Xgal to the selection media (30 mg ml−1). These transconjugants were further screened on the C. violaceum strain CV026 bioassay, for AHL production, and on the E. coli strain ESS bioassay for carbapenem production. Colony PCR, using primers CAR3 (AGCGCTAAGTTGTGATGG; designed to sequence 5′ of the Serratia carA gene) and LAC1 (ATCGGTAATCATGGTCAT; designed to sequence 5′ of the lacZ gene carried on mini-Tn5lacZ1), was used to identify Serratia carA::mini-Tn5lacZ1 mutant strain MCA54. The smaIsmaR locus was identified from Serratia strain LC13 using a mini-Tn5lacZ1-specific gene probe made by excising the ≈ 1.5kb NotI fragment, carrying the kanamycin resistance gene, from plasmid pUTmini-Tn5lacZ1 (De Lorenzo et al., 1990). The gene probe was prepared for the Southern blot using a ‘Geneclean kit’ (BIO101). Southern blots were performed using a ‘non-radioactive DIG-11-dUTP kit (Boerhinger Mannheim).

Mapping and subcloning of the Serratia carbapenem gene cluster

Single-, partial- and double-restriction digests using BamHI and SalI were used to construct restriction maps of the Serratia cosmids pNRT1 and pNRT20. Fragments of between 10 kb and 12 kb resulting from a partial BamHI restriction digest of cosmid pNRT1 were cloned in pACYC184 digested with BamHI. The resulting subclones were transferred into various Car mutants of E. carotovora and Serratia by electroporation, and assayed for carbapenem activity using E. coli strain ESS (McGowan et al., 1996).

Sequence analysis

Restriction fragments of pNRT1 after digestion using BamHI, or BamHI and SalI, were subcloned by ligation into M13mp18/19. Random subclones of these fragments generated by sonication and ligation into M13mp18, were sequenced by the dideoxynucleotide chain-termination method using a ‘Sequenase’ kit (USB). Sequence data was also derived from across the restriction sites to confirm the contiguous nature of the sequence. Partial sequence of pMAC55 was generated using a primer, LAC1. Nucleotide sequence was generated for plasmid pMAC123, using the pBluescript T3 primer (Stratagene) and DRPR (GCTTTCAGAGAAGCATCA). Using the incomplete nucleotide sequence generated from pMAC55 and pMAC123, synthetic oligonucleotide primers PR18A (CGATCGATTCAGTTGCATG) and URPR (TTAGTTAGGCAGAATCGTG) were designed to sequences at positions 5′ of the convergently transcribed smaI and smaR genes respectively. Amplification conditions for PCR (using Taq DNA polymerase, Promega) using primers PR18A and URPR were: 92°C for 5min; 48°C, 0.5 min; 72°C, 1.5 min followed by 35 cycles of 92°C for 0.5min; 48°C, 0.5min; and 72°C, 1.5min. Sequence of the smaIsmaR PCR product was generated on an Applied Biosystems 373 automated sequencing machine using Taq FS (Perkin-Elmer) and primers PR18A and URPR. Additional primers were designed to complete the sequence as it progressed: PR18B (TCGCCTTAACTGGAGAGT), MRU (TCATCGCAGGATTTGTCC), MRD (CGGCATCATGCTGTTTT), DIPR (GAAATCAATGACACCCATCAG), LSIU (AGTGCTCTAATGCGCTCT), I0 (GCGTTTGTTGAGCAGC), I1 (ATCCTCACTGAGTCAC), I2 (TTAACTGGAGAGTGAG), I3 (CCCAATTATCTCCTTG), I4 (GGCCATCATTGTTTGAATG), R0 (GGGGTATAACTGGTTC), DR3 (GTACAACCTAAGCATGGC), R2 (ATCATGCTGTTTTACCC), R3 (ATGTGGGTAATGCTGTC), R4 (CTAATGCGCTCTTTATC), R5 (AGAGCCTATGTGACAAG), R6 (AGCATCATATCGCCGC), FEP1 (TCTTGATGGTATTGGCCG) and FEP2 (GATGACTTACCACAGCCG). Nucleotide sequence data was analysed using GCG (Genetics Computer Group, University of Wisconsin) and blast, searching the GenBank/EMBL and SWISSPROT databases. All protein transmembrane domain predictions were performed using psort from Osaka University, Japan (

β-Galactosidase assays

β-Galactosidase assays were performed in triplicate using ONPG (o-nitrophenol-β-galactoside). The average of these results are expressed as ‘Miller’ units and are proportional to the increase in the absorbance of free o-nitrophenol per minute per constant cell density.

Gene product identification

A series of DNA fragments carrying different parts of the Serratia car gene cluster was cloned, in the correct orientation for expression of the genes, in the T7 expression vectors pT7–5 or pT7–6. The resulting plasmids were used to express the genes in the presence of [35S]-methionine by the methods previously described (Tabor and Richardson, 1985; McGowan et al., 1996) and the protein products were visualized after autoradiography, following their separation by SDS–PAGE.

Isopropanol extraction of prodigiosin

Salt (NaCl)-saturated isopropanol was mixed with an overnight culture of either Serratia or E. carotovora harbouring pNRT104. To prevent the isopropanol becoming miscible with aqueous solution additional NaCl was added to the extraction mix. After vigorous shaking the solvent and aqueous phases were allowed to resolve. Prodigiosin partitions to the solvent layer and this was removed and measured on a Philips PU 8720 spectrophotometer over the range of 350 nm to 700 nm.

Extraction of AHLs from overnight cultures of Serratia sp.

AHLs were extracted from 10 ml overnight cultures of Serratia sp. using ethyl acetate as described by Shaw et al. (1997).

Identification of AHL species using thin-layer chromatography

Next, 5 µl aliquots of AHL extract were applied to C18 reversed phase TLC plates (200 µm layer; Whatman) and developed with methanol–water [60:40 (v/v); Shaw et al., 1997}. TLC plates were air dried, then placed in 24 cm × 24 cm bioassay dishes and overlaid with LB agar (1.5% w/v) seeded with CV026 (McClean et al., 1997) or modified PG agar seeded with strain LIS (or LC13). AHL extracts were resolved alongside known synthetic AHL ‘standards’. After incubation for 48 h at 25°C, the position of AHLs on the TLC plates was indicated by purple or red spots on a light background in the CV026 and LIS bioassays respectively.


This work was carried out with support from the BBSRC, UK. We would particularly like to thank S. R. Chhabra and colleagues, B. W. Bycroft, P. Williams and the late G. S. A. B. Stewart, for the generously given samples of AHLs and for the Chromobacterium violaceum CV026 bioassay strain. M.A.C. was funded by a BBSRC studentship awarded to G.P.C.S.