Multicellular and aggregative behaviour of Salmonella typhimurium strains is controlled by mutations in the agfD promoter


  • Ute Römling,

    1. Karolinska Institutet, Microbiology and Tumorbiology Center (MTC), Box 280, S-17177 Stockholm, Sweden.,
    Search for more papers by this author
    • *Present address: Gesellschaft für biotechnologische Forschung, Department of Cell Biology and Immunology, Mascheroder Weg 1, D-38124 Braunschweig, Germany.

  • Walter D. Sierralta,

    1. Max-Planck-Institut für experimentelle Endokrinologie, D-30603 Hannover, Germany.
    Search for more papers by this author
  • Kristina Eriksson,

    1. Karolinska Institutet, Microbiology and Tumorbiology Center (MTC), Box 280, S-17177 Stockholm, Sweden.,
    Search for more papers by this author
  • Staffan Normark

    1. Karolinska Institutet, Microbiology and Tumorbiology Center (MTC), Box 280, S-17177 Stockholm, Sweden.,
    Search for more papers by this author

Ute Römling. E-mail uto@gbf.dc; Tel. (531) 6181 318; Fax (531) 6181 444.


A colony morphology type is described in which cells of Salmonella typhimurium form a rigid multicellular network with expression of thin aggregative fimbriae that mediate tight intercellular bonds. Surface translocation of cells on plates and adherence to glass and polystyrene surfaces in biofilm assays are further characteristics of the morphotype. This morphotype (rdar) is normally expressed only at low temperature. However, in two unrelated S. typhimurium strains, spontaneous mutants were found forming rdar colonies independent of temperature. Allelic replacement proved a single point mutation in the promoter region of PagfD in each of the two mutants to be responsible for the constitutive phenotype of a multicellular behaviour. Transcription levels of the two divergently transcribed agf operons required for biogenesis of thin aggregative fimbriae by Northern blot analysis with gene probes for agfA and agfD as well as expression levels of AgfA by Western blotting were compared in the wild type, the constitutive mutants and their respective ompR and rpoS derivatives. In the wild type the rdar morphotype and expression of thin aggregative fimbriae are restricted to low temperature on plates containing rich medium of low osmolarity, but biogenesis of thin aggregative fimbriae occurs upon iron starvation at 37°C. In the upregulated mutants biogenesis of thin aggregative fimbriae is only abolished at high osmolarity at 37°C and in the exponential phase in broth culture. Control of expression of thin aggregative fimbriae and rdar morphology takes place at the transcriptional level at the agfD promoter. A functional ompR allele is required, however an rpoS mutation abolishes transcription only in the wild type, but has no influence on expression of thin aggregative fimbriae in the constitutive mutants.


Although bacteria are unicellular organisms, they are able to communicate with each other and therefore co-ordinately regulate behaviour that results in differentiation at the cellular and multicellular level (Roberts et al., 1996). Many bacteria enter a developmental programme in response to nutrient limitation, which can be considered as a collective defence against environmental stress (Costerton et al., 1987; Roberts et al., 1996). The co-operative behaviour ranges from a plain biofilm formation (Costerton et al., 1987) to a complex co-ordinated life cycle in bacteria such as Bacillus subtilis and Myxococcus xanthus (Dworkin, 1996). Two of the best-studied bacteria, Escherichia coli and Salmonella typhimurium, respond to starvation conditions with changes in cellular morphology and physiology (Hengge-Aronis, 1996; Roberts et al., 1996). However, self-organization of these organisms into multicellular systems is just beginning to be recognized (Shapiro and Hsu, 1989; Budrene and Berg, 1991; Harshey, 1994).

Habitats of bacteria are often limited in iron, an essential nutrient for growth (Guerinot, 1994). Besides the efforts of the cells to scavenge this ion (Wooldridge and Williams, 1993; Guerinot, 1994), iron starvation is also a signal to change cell morphology (Costerton et al., 1987; McCarter and Silverman, 1990). In some cases, sensing of iron starvation and surface properties are tightly linked (McCarter and Silverman, 1990; Zhang and Normark, 1996), which raised the question of the hierarchy in this signal cascade (McCarter and Silverman, 1990).

Thin, aggregative fimbriae and curli are homologous fibres expressed by Salmonella spp. and E. coli respectively (Doran et al., 1993). They were characterized in an enterotoxinogenic Salmonella enteritidis strain (Collinson et al., 1991; 1996), and in bacterial isolates causing bovine mastitis (Olsen et al., 1989) and acute salmonellosis in pigeons (Grund and Weber, 1988); however, a conclusive role for these fibres in pathogenesis remains to be elucidated. These appendages were shown to bind to a variety of serum proteins, e.g. soluble fibronectin (Olsen et al., 1989; Collinson et al., 1993; Sjöbring et al., 1994; Ben Nasr et al., 1996) and the dye Congo red (CR), which could be used as an convenient screening method for thin aggregative fimbriae (curli) expression (Collinson et al., 1993; Hammar et al., 1995).

In S. typhimurium and E. coli, the two divergently transcribed operons for the biogenesis of thin aggregative fibres (curli), agfDEFG and agfBA(C), which are called csgDEFG and csgBA(C) in E. coli, are separated by a 521 bp intergenic region (Hammar et al., 1995; Römling et al., 1998). Assembled by the extracellular nucleation–precipitation pathway (Hammar et al., 1996), the secreted fibre subunit CsgA is polymerized on the surface-exposed nucleator CsgB that, in addition, is present along the filament in minor amounts (Bian and Normark, 1997).

Not much is known about the regulation of curli and thin aggregative fimbriae expression on the molecular level. Transposon insertions in the csgD gene, which encodes for a transcriptional regulator belonging to the LuxR family, completely abolished transcription of the csgBA operon in E. coli (Hammar et al., 1995). Expressed on plates under conditions of low temperature, stationary phase and low osmolarity in S. typhimurium SR-11 and ATCC14028-1s (Römling et al., 1998) and E. coli YMel and MC4100 (Olsen et al., 1993; Hammar et al., 1995), thin aggregative fimbriae (curli) require the stationary sigma factor σS encoded by rpoS (Olsen et al., 1993; Römling et al., 1998) and transcriptional regulator ompR (Hultgren et al., 1996; Römling et al., 1998), at least for the transcription of the agfD/csgD operon. It was shown that mutations in hns encoding the histone-like protein H-NS relieve the necessity for rpoS in E. coli. However, curli expression remains dependent on the above-mentioned environmental conditions (Olsen et al., 1993).

In contrast to the above-described regulation programme for biogenesis of thin, aggregative fimbriae (curli), one strain of S. enteritidis was reported to express the fibres in a temperature-independent manner on solid and in liquid medium (Collinson et al., 1991), suggesting that expression of thin aggregative fimbriae can be subject to a phase variation. Phase variation describes a dramatic change in the expression pattern of a gene because of physical changes on the DNA level leading to altered transcriptional or translational regulation. Phase variation is most intensively studied as the interplay between a pathogenic bacterium and its host environment (Robertson and Meyer, 1992), but also occurs in the developmental cycle of M. xanthus leading to the separation of the functions for spore production and formation (Dworkin, 1996). In general, fimbrial expression can be uniform or subject to phase variation, although both stages can respond to changes in environmental conditions (Low et al., 1996).

In this paper we propose a participation of thin aggregative fimbriae in the formation of multicellular complexes in S. typhimurium based on the phenotypic properties of the fibre-bearing cells. In the wild-type strains, expression is restricted to ambient temperature on plates and completely dependent on rpoS. The only detected environmental signal that induces biogenesis of thin aggregative fimbriae at higher temperatures is iron starvation. An almost constitutively multicellular behaviour was established by mutations that took place in the promoter region of agfD. Transcriptional and expression studies of the mutants and their respective rpoS derivatives showed that the main regulatory switch is on the level of the agfD promoter, whereby the promoter-up mutations abolished the need for rpoS.


Environmental cues affecting biogenesis of thin aggregative fimbriae in wild-type S. typhimurium

The two mouse virulent strains SR-11 and ATCC14028-1s display a temperature-dependent change in colony morphology on Congo red (CR) plates that is associated with the expression of thin aggregative fimbriae; the smooth and white (saw37) colonies at 37°C change to a rough, dry and red (rdar28) phenotype at 28°C (Römling et al., 1998).

To get further insights into the regulation pattern of thin aggregative fimbriae, the wild-type strain 14028-1s (always examined as the nalidixic acid-resistant mutant UMR1, the parent of all other strains) was analysed for the fibre subunit AgfA by Western blotting, choosing some of the growth conditions that S. typhimurium faces in the environment or in a host. A formic acid-dependent appearance of a signal was taken as the indication of intact fibres on the bacterial surface. However, absolute quantification of the AgfA signal is not possible due to the variable degree of depolymerization of AgfA by the formic acid treatment (Collinson et al., 1991).

UMR1 showed a highly regulated pattern of AgfA expression. Fibre expression was seen at 28°C on plates with LB medium without salt and under iron enrichment and starvation, but not on minimal medium under anaerobic or high salt conditions and in liquid culture (Fig. 1, Table 1[link] and data not shown). However, depletion of iron in the agar medium led to the expression of a significant amount of thin aggregative fimbriae at 37°C that was close to wild-type levels (Fig. 1, Table 1[link]).

Table 1. . Expression of AgfA and quantitative analysis of agfD and agfA transcripts under various environmental conditions. a. All transcriptional data were normalized setting the amount of transcript of the wild-type strain 14028-1s at 28°C to 100.b. Similar data resulted from the analysis of MAE52.c. Similar data resulted from the analysis of MAE42.T, traces of transcript; ND, not done.Thumbnail image of
Figure 1.

. Western blot analysis of the expression of thin aggregative fimbriae under various conditions monitored by the detection of the AgfA fibre subunit. A. Anaerobic conditions, cells were grown on LB without salt containing 0.2% KNO3 at 28°C. B. Overnight broth culture grown at 28°C. C. Cells grown at 42°C. D. Cells grown under iron starvation (D, 0.2 mM 2,2′-dipyridyl) at 37°C. E. Detection of the AgfA fibre subunit at 37°C. Minor signals for the fibre subunit that could vary in their intensities were regularly found in the slot (s) and in the gel corresponding to a dimer (d), whereas the major signal was in consistency with the running behaviour of a monomer (m). Slight cross-reactivity was seen with a band at ≈20 kDa, which was also present in the fibre subunit knockout mutant MAE18. Molecular masses of the size standard (Kaleidoscope polypeptide standard, Biorad) are indicated on the right. − and + indicate lanes containing cells without and with formic acid treatment. The positive control (p.c.) is MAE32 grown at 37°C.

Transcription analysis of the agfD and agfBA operon in wild-type S. typhimurium

To test whether expression of thin aggregative fimbriae is regulated at the transcriptional level at the conditions mentioned above, Northern blot analysis was carried out. RNA levels for the transcriptional regulator AgfD and the fibre subunit AgfA were determined by integration over all signals of distinct size detected by the respective probes (Fig. 2). This approach is feasible as the corresponding csgD gene in E. coli is the first gene in an operon that is needed for any expression of the csgBA message (Hammar et al., 1995).

Figure 2.

. Transcriptional analyses of the iron dependence of strains UMR1 and MAE40. Hybridizations with the agfD (A) and agfA (B) probes are shown. Arrows on the left mark the position and size of the transcripts (Römling et al., 1998). D, 0.2 mM 2,2′-dipyridyl; Fe, 100 μM FeCl3.

In the wild-type strain UMR1 significant signals for both probes were only found at 28°C on solid LB medium up to 0.35 M NaCl and under iron enrichment and depletion but not under any other conditions tested. A more than twofold rise in transcript levels was detected under conditions of iron depletion at 28°C and 37°C in contrast to LB medium without salt or supplemented with iron at 28°C (Fig. 2). The detection of agfD and agfA transcripts was concomitant with the expression of AgfA (Table 1), therefore regulation takes place at the agfD promoter.

Expression of thin aggregative fimbriae is dependent on a functional rpoS allele in wild-type S. typhimurium

Curli expression in E. coli YMel and MC4100 and expression of thin aggregative fimbriae in S. typhimurium 14028-1s is dependent on a functional rpoS allele at the level of transcription under standard conditions on plates (Olsen et al., 1993; Römling et al., 1998). The requirement for σS was extended by studying the transcription levels of agfA and agfD as well as the expression of AgfA by Western blotting in MAE40 (the rpoS mutant of UMR1) under all conditions mentioned above (UMR1 and MAE40 were always examined in parallel) (Fig. 1 and data not shown). Strain MAE40 gave no signals in Northern blot experiments under any growth conditions, even after severe overexposure of the blots.

Isolation and characterization of temperature-independent rdar-forming S. typhimurium mutants

Upon repeated subculturing, a temperature-independent rdar28/37 mutant derived from SR-11 was isolated by Lockman and Curtiss (strain SR-11b; Sukupolvi et al., 1997) resembling the basic phenotype of S. enteritidis described by Collinson et al. (1993). The same phenotype could be recovered directly from the original ATCC culture of strain 14028 at a frequency of about 1:1000 and one colony, ATCC14028-4r, was chosen for further studies. The clonality of the strain pairs was proven by pulsed-field gel electrophoresis (PFGE) (data not shown). To test the stability of the temperature-dependent phenotype of 14028-1s, cells were subcultured several times in liquid medium under shaking, thereby diluting the cells 1:100 every 24 h. Aliquots were spread on CR agar plates each day, but no change in phenotype could be detected, although more than 105 colonies were screened. The other three variants also proved to be stable under conventional handling in the laboratory; no change in phenotype could be detected during the course of this study. Therefore the wild type and the mutants seem to be phase locked under standard laboratory conditions. However, storage of rdar28/37 strains in stab cultures at 4°C led to revertants to rdar28 (data not shown).

The correlation between the rdar28/37 morphotype and the expression of thin aggregative fimbriae was established by detection of fibres by electron microscopy; the fibre subunit gene product, AgfA, by Western blotting; and by construction of deletions in agfA or agfBA, e.g. strain MAE18 derived from UMR2 (Figs 1E, 3B and C and data not shown).

Thin aggregative fimbriae producing rdar cells, irrespective of strain and temperature, had a larger diameter than saw colonies and formed irregular-shaped, flat colonies that successively developed a rugous, curled morphology. Extensive surface spreading of the cells occurred when freshly poured 1.5% agar plates were incubated in a humid atmosphere (Fig. 3A). An rdar colony could achieve a diameter of 6 cm after 48 h at 37°C, which was six times the diameter of a saw colony and colonized the whole plate after 72–98 h (data not shown). This multicellular behaviour is reminiscent of swarming (Harshey et al., 1994), but develops slower by taking place in the stationary phase of growth. Swimming (motility in low-concentration agar) and swarming [(motility on low-concentration agar (Harshey et al., 1994)] was assayed at 37°C for the wild type and the upregulated mutants. All isolates swam and swarmed in the same way, therefore these two phenotypes were independent of the rdar morphology type (data not shown). SR-11b and 14028-4r exhibited an even more pronounced colony morphology type, which developed earlier than rdar28 cells from the wild type (Fig. 3A). On plates, all rdar bacteria were connected in a dense, rigid network leaving no visible remnants of individual cells. The tangle could be peeled off from the agar plate like a skin without disruption (Fig. 3B) leaving an imprint of the bacterial lawn in the agar. The tight connection of the bacteria with the inanimate surface is also reflected by the fact that saw colonies grown overnight are completely washed away within 15 min from an agar plate surface that was overlaid with water and shaken, whereas rdar28/37 cells remain tightly clinging to the agar even after 5 h (data not shown). In overnight liquid cultures at 28°C and 37°C, the mutants formed large, macroscopically visible cell clusters (Fig. 3C) in contrast to the wild type, which showed no visible aggregation at either of the two temperatures. In addition, thin aggregative fimbriae-producing cells adhered to the wall of a glass tube or a polystyrene dish, when grown either in rich (LB without salt) or minimal medium with shaking, by forming a biofilm. The surface attachment was macroscopically visible and was confirmed by microscopic analysis and staining the bacteria with crystal violet (Fig. 3D and data not shown).

Figure 3.

. Phenotypes of bacterial isolates with a temperature-dependent and temperature-independent expression of thin aggregative fimbriae and their respective knock-outs. A. Phenotypic characterization of S. typhimurium strains 14028-1s and MAE32 representing the rdar28 and the rdar28/37 morphotype, respectively, grown on CR plates at different temperatures. 1, rdar28 (1s) and rdar28/37 (MAE32) morphotype; 2, agfA mutants MAE5 and MAE18 respectively; 3, rpoS mutants MAE40 and MAE41 respectively; 4, ompR mutants MAE46 and MAE49 respectively. The rdar28 morphotype of 1s is hardly expressed even at 28°C on these plates showing its high dependence on environmental conditions. B. Demonstration of the differences in intercellular connection between the rdar and the thin aggregative fimbriae deficient pdar morphotype. The bacterial lawn can be lifted from the plate without disruption in both cases. However, the rdar morphotype, here represented by MAE32 (left), has a skin-like structure, whereas the pdar morphotype (strain MAE18, right) has a rubber-like structure. Both strains were grown on LB-medium without salt at 37°C. C. Cell cluster formation in an overnight culture of S. typhimurium isolates. Cells were grown at 28°C in LB medium without salt with vigorous shaking. 14028-4r, MAE32 and MAE52 form large clusters, whereas MAE5 does not. However, smaller clusters are formed by MAE18, which sediment more slowly. D. Biofilm formation on a glass surface at 28°C (1) and 37°C (2) after overnight growth in LB medium without salt with shaking. The bacterial cells were stained with crystal violet.

A deletion in agfA or agfBA (e.g. MAE5 derived from UMR1 and MAE18 derived from UMR2 respectively) abolished the rdar colony morphology and resulted in a pink colony of slightly smaller diameter that developed a delayed roughness without a dry appearance (padr morphotype) on CR plates (Fig. 3A; Collinson et al., 1993; Römling et al., 1998). The padr morphotype expressed by a corresponding strain showed the same temperature dependence as the rdar morphotype. When peeled off from the plate the bacteria were still connected, but in a network of rubber-like consistency and did not stick to the surface of the agar. After overnight growth in liquid broth, MAE18 formed smaller but macroscopically visible cell clusters that sedimented more slowly than 14028-4r (Fig. 3C). No biofilm was formed by either deletion mutant (Fig. 3D).

Mapping of the mutations for the temperature-independent rdar morphotype

Preliminary transduction experiments with strains bearing a Km cassette about 3 kb downstream of agfC [INK1, INK7 (derived from SR-11 and SR-11b respectively), INK2 and INK3 (derived from UMR1 and UMR2 respectively) Table 2] indicated that the mutation for the temperature-independent rdar28/37 morphotype may lie close to the region for thin aggregative fimbriae. Successively, the more distant, but yet closest markers flanking the agf region, putA810 ::Tn10 and pyrC2688 ::MudJ, respectively, were transduced into the two strain pairs with conservation of their respective phenotypes. Series of phage crosses with the generalized transducing phages P22 and KB1 (data not shown) led to the conclusion that the mutation for the rdar28/37 morphotype most likely lay in the agf region.

Table 2. . Bacterial strains used in this study (Bullas and Ryu (1983); Fang et al. (1992); Torreblanca and Casadesus (1996); Römling et al. (1998)).Thumbnail image of

Characterization of the mutation responsible for the rdar28/37 morphotype

As the mutation for rdar28/37 and concomitant upregulation of expression of thin aggregative fimbriae at 37°C mapped within the agf region, the transcriptional regulator agfD and the intergenic region between the two agf operons were prime target candidates. Examination of the sequences of the two strain pairs, SR-11 and SR-11b, 14028-1s and 14028-4r, revealed one individual single base pair mutation in the promoter region of agfD in each of the temperature deregulated strains, whereas the sequences of SR-11 and 14028-1s were identical. Evaluation of the transcriptional start sites for SR-11b and 14028-4r by primer extension at 28°C and 37°C showed identity with the wild-type site (data not shown). Therefore, the mutations could be classified as an insertion of a single T after position −17 (between the −10 and −35 box) in strain 14028-4r and a transversion from G to T at position −44 in strain SR-11b (Fig. 4). These mutations did not change the characteristics of the agfD promoter, which appears to be a promoter transcribed by σD: the −35 and −10 box spaced by 16 bp or 17 bp contain the most conserved bases from the consensus sequences. The −10 box and the transcriptional start site that originates at a G (the second most common base at this position) are 7 bp apart (Hertz and Stormo, 1996).

Figure 4.

. Nucleotide sequence of the region of the agfD promoter in strains SR-11 and 14028-1s . The −10 and −35 box sequences are indicated by lines over the text, consensus bp to the σD promoter are underlined. The transcriptional start site is in bold and the spacer distances between the transcriptional start site and the −10 region and between the −10 and −35 region in 14028-1s and SR-11 are indicated. Arrows mark the insertion after position −17 in 14028-4r and the transversion from G→T in SR-11b. The region of the consensus sequence for OmpR is boxed, consensus bases have a dark background, the mismatch base a light one.

To prove unambiguously that the mutations found in the two strains are solely responsible for the rdar28/37 morphotype and the upregulated expression of thin aggregative fimbriae at 37°C, phage transduction and allele replacement experiments were performed. Creating a restriction site for Tsp509I the mutation found in SR-11b could be easily followed and correlated with the temperature-independent rdar28/37 morphotype by phage transduction. Screening of 17 independent clones after the introduction of the flanking markers, putA810 ::Tn10 and pyrC2688 :: MudJ, from a saw37 background into INK7 confirmed the link between the point mutation and the rdar28/37 morphotype (data not shown). Allelic replacement experiments were performed using PCR3 fragments of strains SR-11b and 14028-4r covering the whole intergenic region. These were cloned into the pMAK705 vector that harbours a temperature-sensitive replicon yielding pUMR11-1 and pUMR11-10 respectively (Table 3). Allelic replacement on the SR-11 and 14028-1s chromosomes resulted in strains MAE31 and MAE52 respectively. To show the independence of the mutation of the strain background, the mutation from strain SR-11b was also introduced into 14028-1s creating strain MAE32. The link of the mutations with the rdar28/37 morphotype was confirmed by Tsp509I digest of PCR4 fragments or sequencing; strain identity was validated by PFGE (Fig. 5).

Table 3. . Plasmids used and constructed in this study (Hamilton et al. (1989); Römling et al. (1998)).Thumbnail image of
Figure 5.

. Confirmation of strains identity and allelic exchange in strains SR-11 and 14028-1s. A. Pulsed-field gel electrophoresis of XbaI-digested chromosomes of wild-type strains and their respective allelic replacement derivatives. Lanes: s, λ-oligomer size standard in 48.5 kb increments; 1, SR-11; 2, MAE31; 3, MAE32; 4, 14028-1s. B. Tsp509I-digested PCR4 products showing the presence (343 bp) or absence (460 bp) of the PagfD2 mutation. s, PstI-digested λ-DNA with sizes 516 bp, 467 bp, 448 bp and 339 bp seen from the top. Other lanes as in (A).

Strains MAE31, MAE32 and MAE52 exhibited no phenotypic differences with the originally isolated mutants SR-11b and 14028-4r when the following features were compared: colony morphology on CR plates; biofilm formation; behaviour in liquid medium; and the pattern for expression of thin aggregative fimbriae under different conditions. Being derivatives of the well characterized LT2 strain, MAE32 and MAE52 were used for further studies and only data referring to these strains are reported.

Effect of the rdar28/37 mutations on expression of thin aggregative fimbriae under different conditions

MAE32 and MAE52 showed a high-level expression of thin aggregative fimbriae under a series of different growth conditions at 28°C and 37°C (Table 1 and data not shown). Signals of AgfA on Western blots were detected on plates with rich medium, on minimal medium with glucose, under conditions of iron depletion or enrichment, anaerobicity, high osmolarity at 28°C and late stationary phase in liquid culture. (Fig. 1, Table 1[link] and data not shown). As S. typhimurium is also an important pathogen in birds that have a body temperature of 40°C and higher, the expression of thin aggregative fimbriae at 42°C was checked. Only a weak signal for AgfA was detectable on the blot (Fig. 1C).

In general, the mutations in the agfD promoter region led to an expression phenotype of thin aggregative fimbriae that differed significantly from the pattern seen in the wild type by making it independent of several but not all environmental cues; expression of thin aggregative fimbriae is still dependent on the very late stationary phase of growth (Arnqvist et al., 1994) and moderate osmolarity conditions on plates at 37°C. Gross differences between the two mutant strains were not found under any conditions examined.

Transcription analysis of the agfD and agfBA operon in the temperature-independent rdar28/37 mutants

Transcriptional analysis of the agfD and agfA messages were carried out at the growth conditions used to monitor thin aggregative fimbriae expression (see above). MAE32 and MAE52 exhibited an approximately twofold rise of signal intensity for both messages compared with the wild-type strains UMR1 at 28°C on LB plates without NaCl. Unlike the wild type, the transcript levels did not change significantly upon temperature shift to 37°C (Fig. 6, Table 1[link]).

Figure 6.

. Transcriptional analyses of strain MAE32, and its respective rpoS and ompR mutant at 28°C and 37°C. 1, MAE32; 2, MAE41; 3, MAE49. Hybridizations with the agfD (A) and agfA (B) gene probes are shown. Arrows on the left mark the position and size of the transcripts.

In general, the expression of thin aggregative fimbriae is determined over the regulation at the agfD promoter as the two messages, agfD and agfA, changed in parallel and were concomitant with the detection of the fibre subunit AgfA (Table 1). Exceptions were anaerobic conditions and the logarithmic-phase growth in liquid culture in which the agfA transcript levels were more dramatically reduced than the ones of agfD, implying that additional factors act beyond the initiation of the agfD transcript.

Influence of dam, ompR and rpoS on the rdar morphotype and expression of thin aggregative fimbriae in the temperature-independent rdar28/37 mutants

As a few fibres in E. coli are regulated by the difference in methylation pattern (Low et al., 1996), the potential influence of a dam locus on expression of thin aggregative fimbriae was analysed (Torreblanca and Casadesus, 1996). As no change in the rdar morphotype was seen in UMR1, MAE32 and MAE52, which is in accordance with the lack of GATC sites in the intergenic region (data not shown), no further analysis was performed.

An ompR locus abolished the rdar morphotype in MAE32 and MAE52 and restored the saw morphotype typical of the wild type grown at 37°C (Fig. 3A). No agfD transcript was detected and subsequently neither an agfA transcript nor an AgfA signal on Western blots was seen (Fig. 6; data not shown). Therefore, expression of thin aggregative fimbriae is completely dependent, directly or indirectly, on a functional ompR locus as in the wild type (Römling et al., 1998).

Introduction of an rpoS mutation in MAE32 and MAE52 (MAE41 and MAE42, respectively, are the rpoS mutants) created colonies that remained dry and rough, but their colour changed to brownish (bdar morphotype) (Fig. 3A). The cells remained aggregative, however the network structure was lost. Production of thin aggregative fimbriae was not affected by the rpoS mutation under standard conditions as judged by the detection of AgfA by Western blotting and RNA transcript analysis of agfA and agfD (Fig. 6 and data not shown). Therefore, a segregation of the development of the rdar morphotype and the expression of thin aggregative fimbriae had taken place. The fact that two promoter mutations in different regions made these promoters independent of σS suggested that another sigma factor may now recognize the agfD promoter, as the transcriptional start site of the agfD operon was not subjected to change in the rpoS mutants MAE41 and MAE42 (checked at 28°C; data not shown).

Effect of an rpoS background on agfD transcription and biogenesis of thin aggregative fimbriae in the temperature-independent rdar28/37 morphotype

The transcription levels of agfA and agfD and the expression of the fibre subunit by MAE41 and MAE42 remained unchanged after growth on LB plates without salt at 28°C and 37°C (Fig. 6 and data not shown) compared with MAE32 and MAE52 respectively. Growth under anaerobicity and iron depletion or enrichment of the medium did not alter the transcription and expression levels (Fig. 1, Table 1[link] and data not shown). However, in the rpoS mutants agf transcription was more sensitive to high osmolarity and exhibited a lower production of thin aggregative fimbriae in liquid broth medium after overnight growth and on minimal medium (Fig. 1, Table 1[link] and data not shown). An opposite effect of the rpoS mutation was seen at 42°C on LB plates without salt. Whereas MAE32 and MAE52 did not produce thin aggregative fimbriae at 42°C, expression and RNA levels in the rpoS mutants were as high as in MAE32 and MAE52 at 37°C (Fig. 1, Table 1[link] and data not shown). When transcript levels of agfD and agfA were subject to change in the rpoS mutants, they altered correspondingly, suggesting that no gross rpoS-dependent regulation of transcripts or gene products in the pathway leading to the biogenesis of thin aggregative fimbriae is involved.


In the present paper we describe the rdar morphotype as being a developmental programme in S. typhimurium triggered by starvation, which leads to the formation of a multicellular network with thin aggregative fimbriae as a component of the intercellular matrix. Differential response of the multicellular behaviour to environmental conditions is controlled by mutations in the promoter for the agfD operon. Thereby, the expression of thin aggregative fimbriae becomes almost constitutive in contrast to the wild type in which biogenesis of thin aggregative fimbriae occurs only on plates with a rich medium of low osmolarity at ambient temperature to be overcome only by iron starvation. However, production of thin aggregative fimbriae remains completely dependent on a functional ompR gene encoding for a transcriptional regulator in all strains, but becomes independent of rpoS.

Detection and characterization of the rdar28/37 morphotype

The formation of a distinct colony morphology, the rdar morphotype, is tightly linked to thin aggregative fimbriae (curli fibres), the genes of which are present and expressed in strains of virtually all Salmonella spp. and E. coli (Doran et al., 1993; Olsen et al., 1993; Bäumler et al., 1997). Therefore, we expect the multicellular bacterial behaviour that results in the rdar morphotype to be expressed and play an important role in the lifestyle of Salmonella and E. coli. We have also found that the rdar morphotype is exhibited by several S. typhimurium strains from the collection of Lilleengen type (LT) strains (our unpublished data). However, the rdar morphotype and thin aggregative fimbriae (curli fibres) have escaped wide attention until recently as they are not produced by many E. coli and S. typhimurium laboratory strains because of rpoS, ompR and mutations within the thin aggregative fimbriae (curli) gene cluster (Zambrano et al., 1993; Lee et al., 1995; our unpublished data). In addition, as cells are usually streaked on plates for single colonies first, rdar28/37 cells remain tightly clinging together at the beginning of a streak and never reach the state of an individual colony.

S. typhimurium rdar cells adhere to plastic surfaces and glass in an assay used to analyse biofilm formation. A biofilm represents the developmental state of sessile bacterial populations in aquatic environments (Costerton et al., 1987). As Salmonella spp. are more prevalent in sediments than in free-floating water (Murray, 1991) this organism might prefer to interact with inanimate surfaces rather than exhibit a planktonic life form when present in the environment. Bacterial cells resulting in a rdar colony morphology can tightly adhere to inanimate surfaces and could therefore effectively colonize particulate matter in the environment. Nutrient trapping and protection of the population by, e.g., providing an extracellular surface layer protecting against mechanical stress are characteristics of biofilms (Costerton et al., 1987). Such characteristics are also reflected by thin aggregative fimbriae-producing cells binding a variety of proteins (Sjöbring et al., 1994; Ben Nasr et al., 1996) and establishing rigid intercellular bonds (Fig. 3B).

Bacterial adhesion to surfaces also plays an important role in various infectious disease processes ranging from bacterial colonization of foreign body implants to the adhesion of mucosal membranes in the gastrointestinal tract and elsewhere (Costerton et al., 1987). The few epidemiological data that are available suggest a temperature-regulated curli expression in non-invasive E. coli strains (Olsen et al., 1993; Ben Nasr et al., 1996). As the suppression of biogenesis of thin aggregative fimbriae at 37°C is relieved by iron deprivation in the rdar28 morphotype, a participation of thin aggregative fimbriae (curli) in the colonization of the vertebrate host by Salmonella spp. and E. coli is possible, but remains to be elucidated.

Adhesion to surfaces does not seem to be the only purpose for bacteria yielding the rdar morphotype. The larger diameter of rdar colonies in contrast to either saw or pdar colonies (see Fig. 3A) suggests that gaining of territory after surface adhesion is another feature of bacteria participating in the rdar morphotype.

The contribution of curli to the rdar phenotype could be to mediate surface adhesion by penetration of the agar surface to establish the rigid intercellular bonds between cells, thereby forming a skin-like surface layer and to contribute to the surface motility.

Why and when does the switch in rdar morphotype expression occur?

The ability to adapt to distinctive habitats ensures the survival of bacterial isolates in changing environments. Thus, the ability to alter the expression pattern of thin aggregative fimbriae (expression at 28°C, at 28°C and 37°C, solely at 37°C or no expression at all) would ensure the tuning from a multicellular and/or adherent lifestyle to the single-cell state. In aquatic environments bacteria must accomplish the change between a sessile and planktonic phenotype in order to colonize favourable niches. The marine organism Caulobacter crescentus has solved this problem by making both phenotypes an integral part of its life cycle (Roberts et al., 1996). A potential role of the phase switch of the biogenesis of thin aggregative fimbriae in a vertebrate host needs to be investigated. Phase variation is thoroughly studied in Neisseria spp. that change a variety of surface components because of the microenvironment in the human host (de Vries et al., 1996).

Under standard laboratory conditions the rdar28 and the rdar28/37 morphotype remained stable. However, as the original ATCC culture displayed rdar28/37 cells at a high frequency, we suspect the storage procedure or drying by lyophilization (Rhode et al., 1975) as selective pressures for mutations leading to a permanent rdar morphotype.

The rdar morphotype and expression of thin aggregative fimbriae are not always linked

In other bacteria, such as Vibrio parahaemolyticus, Proteus mirabilis, M. xanthus, C. crescentus, Pseudomonas aeruginosa and Staphylococcus epidermidis, colonization of and moving on surfaces is a well-studied phenomenon (Costerton et al., 1987; McCarter and Silverman, 1990; Harshey, 1994; Dworkin, 1996; Heilmann et al., 1996; Roberts et al., 1996). Recently, it was also shown that S. typhimurium and E. coli are capable of swarming on agar surfaces by differentiation into multinucleated, hyperflagellated cells (Harshey, 1994). Altered behaviour on inanimate or animate surfaces probably involves bacterial matrix components such as pili, exopolysaccharides, lipopolysaccharides and exolipids (Costerton et al., 1987; Matsuyama, 1995; Dworkin, 1996). These are suspected to change in composition and amount as a bacterial strategy to adhere selectively and adapt to the particular surface to be inhabited (Matsuyama, 1995).

The agfD operon seems to play a key role in the response of S. typhimurium to surfaces as a single point mutation in PagfD confers the whole rdar37 morphotype. Consequently, a polar mutation in agfD led to a saw morphotype (data not shown). Most likely, the transcriptional regulator AgfD influences the expression of other genes than agfBA or AgfE, AgfF and AgfG are produced in the absence of agfBA as the pdar morphotype manifested by bacteria lacking thin aggregative fimbriae was clearly distinct from saw. The sensitivity of co-ordinated gene expression on morphotypes was also reflected by the rpoS mutant derivatives of rdar28/37 S. typhimurium. Even although absence of σS did not significantly affect expression of thin aggregative fimbriae, the resulting bdar morphotype was clearly distinct from rdar as well as from pdar and saw (Fig. 3B). The bdar morphotype of these colonies indicates differential expression of other (cell surface) components than thin aggregative fimbriae necessary for the rdar morphotype that require a functional rpoS gene.

In this context, we have to state that the phenotype of MC4100, which produces curli at 28°C, is also distinct from rdar as it is missing the extensive surface spreading and the tight network structure, although the cells are autoaggregative and form rough and dry colonies (our unpublished observations). Therefore, thin aggregative fimbriae might be part of a differentially regulated extracellular network.

Regulation of agfD transcription and expression of thin aggregative fimbriae on the molecular level in wild-type S. typhimurium

Besides environmental growth conditions and the general participation of agfD, ompR, rpoS and hns, nothing is known about the direct interactions that result in a changed transcriptional status at the agfD and agfB promoters. Our data suggest that most of the regulation takes place at the agfD promoter as agfA and agfD transcript levels usually change in parallel. If AgfD, a member of the LuxR family of transcriptional regulators, is transcribed and successively expressed, it may directly activate transcription at the agfB promoter.

Based on the environmental signals that act on gene expression of thin aggregative fimbriae (Fig. 7), such as iron, reduced O2 pressure and osmolarity, several known regulatory factors can be proposed. However, only for OmpR has a convincing binding site been found that has one mismatch to the recently proposed consensus sequence for independent binding (Huang and Igo, 1996) and is centred at position −50.5 in S. typhimurium (Fig. 4) relative to the transcriptional start site of the agfDEFG operon (Römling et al., 1998). The mutation found in SR-11b, which is located at −44, does not weaken this consensus sequence (see Fig. 4). Besides, its essential requirement in initiating transcription at the agfD promoter OmpR could participate in regulation of transcription to osmolarity. This regulation pattern would reflect the role of OmpR at the ompF promoter in E. coli (Pratt and Silhavy, 1995).

Figure 7.

. Diagram showing the influence of various environmental conditions and genes on transcription from the agfD and agfB promoter (A) in the wild-type strains; (B) in the temperature upregulated mutants. l. m., liquid medium; m. m., minimal medium; temp., temperature; exp., exponential; stat., stationary.

The sigma subunit responsible for promoter recognition and transcription initiation at the wild-type agfD promoter seems to be σS. Arguments for this scenario are the absolute dependence of the wild-type promoter on a functional rpoS allele under all conditions examined and the lack of evidence for additional activator loci besides those influencing the levels of RpoS in the cell (M. Hammar, unpublished data). Thin aggregative fimbriae (curli) expression in wild-type bacteria coincides with high expression of RpoS in the late exponential growth phase at low temperature and carbon starvation (Hengge-Aronis, 1996; this paper; M. Hammar, unpublished), suggesting that at least part of the vast regulation of expression of thin aggregative fimbriae (curli) is performed through variations in RpoS levels.

Why do the promoter mutations in agfD cause a different transcription pattern?

Two mutations in different regions of the agfD promoter, the upstream element (UP) and the spacer between the −10 and −35 boxes (Busby and Ebright, 1994; Hertz and Stormo, 1996), respectively, were found to confer a similar phenotype of promoter upregulation using a temperature-independent transcription at an approximately twofold higher level than in the wild type at 28°C. Otherwise, the promoter region is not a hot-spot for mutations as it is highly conserved between S. typhimurium 14024-1s and E. coli MC4100 (4 bp difference in the 60 bp stretch upstream of the transcriptional start site). However, the promoter region of S. enteritidis 27655-3b, which has a temperature-independent expression pattern of thin aggregative fimbriae (Collinson et al., 1996), differs from 14028-1s by 3 bp. Interestingly, one of the base pair exchanges corresponds to the mutation found in SR-11b and might therefore confer the temperature-independent regulation in this strain.

Surprisingly, the mutations found in SR-11b and 14028-4r make the transcription from the agfD promoter completely independent of RpoS. Doing so they might enhance the specificity of RNA polymerase (RNAP) loaded with σD for the mutant agfD promoter and therefore abolish the need for an activator(s) encoded by an rpoS-dependent pathway. Another possibility is a sigma factor switch in which the two promoter mutations alter the specificity to RNAP holoenzyme loaded with different sigma factors. The wild-type promoter would require σS for transcription. To date, the candidate sigma factor for transcription of the mutant promoter can only be σD under conditions of stationary phase of growth (Ozaki et al., 1992). A high sequence homology between σD and σS is reflected by overlapping promoter recognition properties (Tanaka et al., 1993; Altuvia et al., 1994). Consequently, only 1 or 2 bp changes in a promoter site can convert the transcription accessibility from σS to σD and vice versa (Utsumi et al., 1995; Wise et al., 1996). A sigma factor switch would also help to explain the different effect of environmental conditions on the transcription from the mutant agfD promoters compared with wild type as rpoD and rpoS are differently regulated (Jishage et al., 1996). In any case, the mutant agfD operon adds to the increasing number of stationary induced genes that are not regulated by rpoS under most environmental conditions (Hengge-Aronis, 1996).

By affecting transcription from the agfD promoter, the selected mutations might not only affect transcription of the subunit gene for thin aggregative fimbriae, but also the other genes required for its biogenesis allowing a balanced expression of the various proteins that result in the formation of the highly adhesive surface organelles. In line with this argument, it was recently shown that overexpression of the lipoprotein CsgG encoded by the csgDEFG operon enhances the amount of curli produced (Loferer et al., 1997). Hence, up-promoter mutations affecting only the agfBA(C) operon would not establish the appropriate balance of gene products.

One might wonder why two independent and spontaneous events leading to the rdar28/37 morphotype were due to mutations affecting the property of the agfD promoter. Besides strong selection, it suggests that no other more frequent genetic events in the two agf operons or elsewhere in the chromosome can cause this behavioural change in Salmonella wild-type strains.

Experimental procedures

Bacterial strains, growth conditions and phages

All bacterial strains used and constructed in this study are listed in Table 2. For monitoring expression of thin aggregative fimbriae by various techniques, Salmonella typhimurium strains were grown on Luria–Bertani (LB) agar plates without NaCl at 28°C and 37°C, unless otherwise stated (Ausubel et al., 1994). LB agar without salt supplemented with Congo red (40 μg ml−1)/Coomassie brilliant blue (20 μg ml−1) was used to judge colony morphology and colour. If required, LB agar without salt was supplemented with NaCl (0.15, 0.3 and 0.5 M), 100 μM FeCl3 or 0.2 mM 2, 2′-dipyridyl. To establish anaerobic conditions, LB agar plates without salt, supplemented with 0.2% KNO3 were incubated in a GasPak anaerobic jar (BBL Microbiology Systems, Becton Dickinson). Other media to support cell growth on agar plates were minimal M9 supplemented with 0.2% glucose (Ausubel et al., 1994) and Yesca (1% Bacto-casamino acids, 0.1% Bacto-yeast extract). Growth in liquid medium was performed by filling one-tenth of the flask to ensure proper aeration by vigorous shaking. Phages P22 HT105/1 int-201 (Schmieger, 1972) and KB1 int (McIntire, 1974) were used to carry out transductions among LT2 and SR-11 strains, respectively, according to recommended protocols (McIntire, 1974; Maloy et al., 1994). For cloning purposes all cells were grown on LB agar supplemented with antibiotics used in the following concentrations: ampicillin, 100 μg ml−1; chloramphenicol, 20 μg ml−1; kanamycin, 30 μg ml−1; nalidixic acid, 50 μg ml−1.

DNA techniques

Isolation of plasmid, cosmid and chromosomal DNA was performed according to standard procedures (Ausubel et al., 1994). Enzymatic manipulations such as restriction digestion, ligation and PCR was carried out according to standard protocols using enzymes from Boehringer Mannheim or New England Biolabs. DNA fragments were recovered from agarose gels using the Geneclean II kit (Bio101) and PCR fragments were purified using the Quiaquick PCR purification kit (Quiagen). Southern transfer, hybridization, probe labelling, non-radioactive detection and all manipulations concerning PFGE technology have been described previously (Römling et al., 1994). Plasmid DNA used for sequencing was purified by PEG precipitation (Applied Biosystems, user bulletin 18). Sequencing was carried out using the Cycle Sequencing Ready Reaction Kit (Perkin-Elmer). All sequence analysis was performed with the help of the GCG version 8 (Wisconsin Package).

Plasmid and strain construction

Vectors and selected plasmid constructs are listed in Table 3. Primers for PCR were: AGFD1: dTGGATGCATCGTAGCTTGCAGAGATGG (NsiI, restriction site underlined) AGFBD: dACGAAAGCTTGCACTGCTGTGGGTTG (HindIII); AGFB1: dGCTCTGCAGCCAGATC-ATAATTTGTCG (PstI); AGFB3: dGGAGGATCCAGATCATAATTTGTCG (BamHI); AGFA1: dCGTCTGCAGGATTGCTGCGAATGCTGC (PstI); AGFA2: dCGTCTGCAGTGG-AACGCTAAAAACTC (PstI); AGFC: dCGAGGATCCGGCCATTGTTGTGATAAA (BamHI); Sp22: dGAGATATCTTCCAGAGAACG used to create the fragments PCR1 (AGFA2 ↔ AGFC), PCR2 (AGFBD ↔ AGFB1), PCR3 (AGFD1 ↔ AGFB3), PCR4 (Sp22 ↔ AGFA1). To create a plasmid that could generate a deletion in agfBA, PCR1 and PCR2 were directly cloned into pMAK700 cut with HindIII and BamHI. This plasmid was named pUMR3a, a derivative harbouring a Km cassette in the single PstI site pUMR3b-1. PCR3 fragments were first cloned into pBS cut with PstI and HindIII (pUMR10) for sequencing and then transferred into HindIII/XbaI cut pMAK705 (pUMR11). Construction of pUMR7b-2 has been described previously (Römling et al., 1998). DNA translocation into bacteria was performed using competent cells (E. coli ; Inoue et al., 1990) or electrocompetent cells (S. typhimurium; Gene pulser electroprotocol, Biorad). pMAK derivatives constructed in E. coli were passed through the restriction deficient strains S. typhimurium LB5010 before electroporation into other Salmonella strains. Sequential temperature shifts were performed to isolate S. typhimurium strains that underwent gene replacement events (Hamilton et al., 1989). The resistance marker or the phenotype on CR plates were used to isolate the strains and the genotype was verified by Southern hybridization and/or PCR.

Northern blotting

Total RNA was prepared from 10 mg of S. typhimurium cells (grown for 17 h at 37°C or for 24 h at 28°C) using the hot phenol method, and the RNA concentration was determined spectroscopically. An aliquot (10 μg) of RNA was loaded on a 1.2% MOPS–formaldehyde gel (Ausubel et al., 1994), which was run for 4 h a 4 V cm−1 with the 0.24–9.5 kb RNA ladder (Life Technologies) as standard. After soaking the gel in H2O (twice for 20 min), the RNA was transferred to a Amersham Hybond-N membrane overnight by capillary blotting using 20× SSC (Ausubel et al., 1994). Single-stranded probes complementary to the RNA template on the blot were constructed using an asymmetric PCR on symmetric PCR templates covering the region of the gene of interest and labelled with RadPrime Labelling System (LifeTechnologies) using 30 μCi [α-32P]-dCTP (3000 Ci nmol−1; Amersham). Hybridization (using less than 6 ng ml−1 of probe) and washing of blots was carried out according to standard procedures (Ausubel et al., 1994). The quality of transfer to the membrane was checked by probing with part of the 16S RNA sequence from plasmid pKK3535 (Brosius et al., 1981) cut by HindIII. Signals were detected by a radioisotope imagine system (Phosphoimager 445SI, Molecular Dynamics). Quantification of the message was performed by integrating over all bands detected by a single probe.

Protein techniques

Whole-cell preparations used as samples for SDS–PAGE and immunoblot analysis were performed as follows: 5 mg of cells (grown for 24 h at 37°C or for 48–60 h at 28°C) were resuspended in 200 μl of SDS sample buffer (Ausubel et al., 1994) and heated at 95°C for 5 min. To detected polymerized fibre-derived AgfA subunits, 5 mg of cells (or an equivalent of 3 OD600 in 1 ml of liquid culture), resuspended in 100 μl of 99% formic acid and treated as described previously (Collinson et al., 1991). The proteins were transferred to PVDF membranes (Immobilon P, Millipore) after electrophoresis, incubated with primary and secondary antibody (horseradish peroxidase-linked anti-rabbit IgG; 1:2000; Boehringer Mannheim) and detected using chemiluminescence (Boehringer Mannheim) on Hyperfilm-MP (Amersham). Polyclonal antisera against the E. coli proteins CsgA (1:2000), OmpA (1:90000) and DsbA (1:20000) were used as primary antibodies. To check the protein content and transfer to the membrane, all blots were probed with OmpA or DsbA antibodies after AgfA detection.

Biofilm assay on polystyrene and glass

The adherence assay to solid surfaces was carried out as described recently by Heilmann et al. (1996). Either 1% safranin or 0.4% crystal violet was used to stain surface-attached bacteria. Glass tubes were inspected visually, the absorbance to polystyrene was measured at 492 and 540 nm, respectively, in order to judge biofilm formation.

Nucleotide sequence accession number

The nucleotide sequence of two operons, agfDEFG and agfBAC, required for the biogenesis of thin aggregative fimbriae is in the EMBL data library under accession no. AJ002301.


  1. *Present address: Gesellschaft für biotechnologische Forschung, Department of Cell Biology and Immunology, Mascheroder Weg 1, D-38124 Braunschweig, Germany.


We would like to thank the following people who kindly and promptly provided plasmids, strains and antibodies: Z. Bian for the anti-CsgA antibody; S. Casabadeus for strains SV3000 and SV1244; D. G. Guiney for SF1005; S. R. Kushner for pWKS30; E. Morfeldt and S. Arvidson for Anti-OmpA antibody; M. Rhen for LB5010; K. Sanderson for the KB1 phage and several bacterial strains and M. Wunderlich and R. Gockshuber for the DsbA-antiserum. A. Eklund and M. Wahlgren made their pulsed-field gel electrophoresis apparatus generously available for use. We are grateful to H. Loferer, M. Hammar and E. Morfeldt for technical advice. We appreciate the critical reading of the manuscript by R. Hurme and M. A. Ellmann. U.R. is the recipient of a fellowship from the programme ‘Infektionsbiologie’ from the Bundesministerium für Forschung und Technologie (BMFT). This work was supported by grants from Medicinska Forskningsrådet (B96-16X-10843-03A) and an unrestricted grant for infectious disease research from Bristol-Myers Squibb to S. Normark.