AgfD, the checkpoint of multicellular and aggregative behaviour in Salmonella typhimurium regulates at least two independent pathways



The regulatory programme of multicellular behaviour in Salmonella typhimurium is determined by mutations in the agfD promoter. AgfD has already been identified to regulate the extracellular matrix associated with the multicellular morphotype composed of thin aggregative fimbriae (agf). To detect additional components contributing to the multicellular morphotype in S. typhimurium, we constructed a mutant in agfD, the positive transcriptional regulator of the agfBA(C) operon encoding for fimbrial subunit proteins. The agfD mutant lacked any form of multicellular behaviour as shown by analysis at the macroscopic and microscopic level. In contrast, the agfBA mutant unable to form thin aggregative fimbriae still maintained long-range intercellular adhesion. Promoter and expression analysis revealed that the genes downstream of agfD agfEFG most likely did not contribute to the remaining aggregative behaviour. Screening of transcriptional fusions for agfD dependency uncovered adrA, a homologue of yaiC in Escherichia coli. Environmental factors regulating adrA correspond to the regulation of thin aggregative fimbriae. AdrA is a putative transmembrane protein with a C-terminal GGDEF domain of unknown function although it is present in over 50 bacterial proteins. AdrA mutant cells, which still formed thin aggregative fimbriae with all binding characteristics, exhibited community behaviour but, unlike the wild type, lacked long-range intercellular adhesion. An agfBA adrA double mutant behaved like the agfD mutant. Therefore, it was concluded that agfD regulates at least two independent pathways contributing to the multicellular morphotype in S. typhimurium.


Multicellularity is a common behaviour of environmental and pathogenic bacteria that can be expressed in several forms such as fruiting body development, swarming, biofilm formation or cell clumping (Shapiro, 1998). Coordinated behaviour offers many advantages to a bacterial population in comparison with the state of single, planktonic cells. In the environment, cells in a community might be better protected against deleterious agents and amoebas grazing the ground (Costerton et al., 1987). They can also improve their metabolic state by trapping nutrients. In a host, biofilm-forming bacterial communities are not susceptible to the immune response of the host or to antibiotic treatment (Costerton et al., 1999). Considering these features of biofilm-forming cells it is not surprising that more than 50% of human bacterial infections were estimated to involve biofilms (News Focus, 1999) and that biofilms in industrial aquatic systems are a major problem (Costerton et al., 1987).

To fight adherent bacterial communities more effectively it is important to identify the pathways which are necessary for multicellular behaviour. Genetic dissection of the adherence process to abiotic surfaces (biofilm formation) has been started in several bacterial species (Heilmann et al., 1997; O′Toole and Kolter, 1998; Pratt and Kolter, 1998; Costerton et al., 1999). In Salmonella typhimurium a multicellular morphotype (rdar) was identified (Römling et al., 1998a). This morphotype which most likely occurs in all Salmonella spp. and Escherichia coli (Doran et al., 1993) had already been recognized at the beginning of the century (Lingelsheim, 1913). However, the morphotype was not examined closer until recently (Olsén et al., 1989; Collinson et al., 1991; 1993; Guard-Petter et al., 1996; Allen-Vercoe et al., 1997).

The rdar morphotype exhibited in the stationary phase of growth can be expressed by different regulatory programmes, a highly regulated (rdar28) and a constitutive (rdar28/37) programme. These programmes are manifested by single-point mutations in the promoter for the transcriptional regulator agfD (Römling et al., 1998a). Therefore, agfD seemed to be a major switch point for the regulation of multicellular behaviour in S. typhimurium. The regulated rdar morphotype was only expressed at 28°C on plates containing rich medium without salt and at 37°C on iron-depletion medium. Transcription at the agfD promoter was then completely dependent on the alternative sigma factor rpoS. In contrast, the constitutive rdar morphotype could be detected at 28°C and 37°C and under a variety of environmental conditions. No expression was seen in the logarithmic phase of growth nor at high osmolarity. Transcription from constitutive agfD promoters is rpoS independent (Römling et al., 1998a).

Multicellular behaviour by the rdar morphotype is manifested in different modes: expression of a wrinkled, dry and spreading colony which absorbs the dye Congo red in contrast to the common white and smooth colony (saw morphotype) on plates, adhesion to abiotic surfaces and human cell matrix components; and cell clumping in liquid culture are major characteristics (Olsén et al., 1989; Austin et al., 1998; Römling et al., 1998a).

So far only thin aggregative fimbriae had been shown to contribute to the rdar morphotype (Collinson et al., 1991; Römling et al., 1998b). In S. typhimurium as well as in E. coli (where the homologous fimbriae are called curli, csg) two divergently transcribed operons, agfDEFG and agfBA(C), are needed for the biogenesis of thin aggregative fimbriae (Hammar et al., 1995; Römling et al., 1998b), where agfD is the positive transcriptional regulator of the agfBA(C) operon (Hammar et al., 1995). However, in S. typhimurium there were indications that other genes might also contribute to the multicellular behaviour. Knocking-out the subunit for thin aggregative fimbriae, agfA, did not yield the saw morphotype, but a pink colony (pdar morphotype) coregulated with the agfDEFG operon where some aspects of intercellular interactions and adhesive behaviour remained (Austin et al., 1998; Römling et al., 1998a). On the other hand, rpoS mutants of the constitutive rdar morphotype still produced thin aggregative fimbriae, but nevertheless changed the colour of the colony from red to brown (bdar morphotype; Römling et al., 1998a), indicating that an unknown matrix component is regulated by rpoS.

Based on those data, two possibilities for other genes contributing to the rdar morphotype remained: the expression of agfEFG in the absence of agfBA or unknown genes somewhere in the chromosome regulated by agfD.

We approached this question by constructing an in frame deletion mutant of agfD which was shown not to (negatively) influence the transcription and synthesis of the downstream genes agfEFG. A search for other genes regulated by agfD identified adrA (agfD regulated gene), a homologue to yaiC in E. coli. The multicellular morphotype mediated by adrA was of the bdar type; it still expressed thin aggregative fimbriae with all binding characteristics, but was missing an unknown extracellular substance. Characterization of the morphotype revealed that most of the community behaviour of the cells remained while long-range cell–cell interactions were weakened. Construction of an agfBA adrA double mutant yielded the saw morphotype of an agfD mutant completely lacking the multicellular behaviour. It was concluded that agfD regulates two independent pathways acting in an additive, but distinct, fashion to produce the multicellular behaviour.


Characterization of AgfD expression

AgfD is homologous to members of the UhpA (FixJ) family of transcriptional response regulators over the entire length of the protein showing the highest similarity (38%) to YhcY, a putative two-component response regulator in Bacillus subtilis. Proteins of the response regulator family consist of an N-terminal receiver domain and a C-terminal DNA-binding domain (Fig. 1; Parkinson and Kofoid, 1992). Two of five highly conserved amino acids (aa) in the receiver domain involved in the phosphorylation of an aspartate are conserved, including the putative phosphorylation site D59, otherwise three non-conservative substitutions occurred in positions D13, D14 and K109 which project into the active site (Fig. 1). The C-terminal helix-turn-helix motif is highly conserved (Baikalov et al., 1996).

Figure 1.

Comparison of the sequence of S. typhimurium AgfD to the most similar response regulator proteins in the UhpA/FixJ family. Alignments were performed using Pile Up (GCG package version 9, University of Wisconsin). Completely conserved aa are shown on a black background, 80% conservation is indicated by a dark-grey background and 60% conservation by a light-grey background. The extension of the receiver and DNA-binding domain, which is connected by a linker domain is shown. Five highly conserved aa involved in the catalysis of phosphorylation in the receiver domain are shown by asterisks (Volz, 1993); the asterisks in the DNA binding domain indicate residues involved in maintaining the tertiary structure of NarL (Baikalov et al., 1996). (sathy) S. typhimurium; (borpe) Bordetella pertussis; (bacsu) Bacillus subtilis; (bacbr) Bacillus brevis. Sequences were taken from SwissProt with the exception of YhcY (pir).

Expression of the rdar morphotype, as monitored by the expression of thin, aggregative fimbriae, only occurs in the stationary phase of growth in the regulated and constitutive rdar morphotype (Arnqvist et al., 1994; Römling et al., 1998a). However, traces of transcript for agfD can already be detected in the logarithmic phase of growth (Römling et al., 1998a). To elucidate AgfD accumulation during the growth phase of the constitutive rdar morphotype MAE52 quantitative AgfD immunoblot analyses of total cell lysates were carried out. AgfD could be detected as early as at an OD of 0.12 in the logarithmic growth phase and accumulated up to the stationary phase (Fig. 2). AgfD expressed from the arabinose regulated araBAD promoter in pBAD30 started to be produced at an OD of 0.6 while arabinose was in the medium. Therefore, factors other than AgfD might be needed for the expression of the rdar morphotype.

Figure 2.

Transcription of adrA requires AgfD whose synthesis is regulated by growth phase.

A. Growth (open symbols) and β-galactosidase activities (closed symbols) of strain ADR1a (adrA101::MudJ; triangles), strain MudJ1 (ΔagfD101 adrA101::MudJ pUMR15) grown in the presence (squares) or absence (circles) of 0.1% arabinose are shown. Cells were grown in LB medium without salt at 37°C at 210 r.p.m. Values are from one representative experiment, three independent experiments with duplicate measurements were done.

B. Immunoblot of AgfD protein using anti-AgfD antibody. Total cell lysates were prepared from cells at the corresponding time points. Protein was isolated from ADR1a (adrA101::MudJ) and MudJ1 (ΔagfD101 adrA101::MudJ pUMR15) medium supplemented with 0.1% arabinose. To control equality of the number of cells the blot was probed subsequently with antibody against OmpA.

Characterization of the agfD mutant morphotype and comparison with the agfBA morphotype

In order to characterize the morphotype mediated by agfD, the mutant MAE51 with an in frame deletion of 180 of 216 aa in agfD was constructed (Table 1). The description of phenotypic features examined is summarized in Table 2 and shown in Figs 3–5. With the deletion of agfD, the red colony of the rdar28/37 morphotype of MAE32 was converted into a white colony which was identical to the saw morphotype shown by the regulated rdar morphotype at 37°C on a Congo red (CR) plate (Fig. 3A). The rigid cell network of the rdar morphotype was completely lost in the agfD mutant (Fig. 3B). Electron microscopy analysis of the rdar morphotype revealed that the abundant extracellular matrix among cells which could be monitored as a tight fibrous network by scanning microscopy (Fig. 4A) or as Ruthenium red-positive material in thin sections (Fig. 5A) completely disappeared in the agfD mutant (Figs 4D and 5D). In the stationary phase of growth in liquid culture, the rdar morphotype formed huge, tight-cell clumps but the agfD mutant displayed only planctonic single cells (Fig. 3C). Cells expressing the rdar morphotype adhered to abiotic surfaces such as glass and plastic and to soluble fibronectin in contrast to the agfD mutant (Fig. 3F and Table 2). In standing culture, the rdar morphotype showed niche specificity together with multicellular behaviour by forming a pellicle with a tight mat of cells at the air–liquid interface (Fig. 3D). Individual cells settled at the bottom of the tube. Expression of thin aggregative fimbriae was detected by immunoblot analyses of formic acid depolymerized AgfA subunits in cells from the pellicle, but not from the pellet (data not shown). The agfD mutant, however, was equally distributed in the whole-liquid culture (Fig. 3D). In conclusion, the agfD mutant completely lost all forms of multicellular behaviour that were examined and is therefore absolutely necessary for the expression of the whole rdar morphotype.

Table 1. Bacterial strains and plasmids used in this study.
Strain or plasmidGenotype or relevant phenotypeSource
S. typhimurium LT2
DA1705 hisD9953::MudJ his-9949::Mud1D. Andersson
metA22 metE551 ilv-452 leu-3121 trpC2 xyl-
404 galE856 hsdL6 hsdSA29 hsdSB121
rpsL120 H1-b H2-e,n,x fla-66 nml(–) Fel-2(–)
Bullas and Ryu (1983)
S. typhimurium ATCC14028
UMR1ATCC14028–1s Nalr; rdar28 Römling et al. (1998b)
UMR2ATCC14028-4r Nalr, PagfD1; rdar28/37 Römling et al. (1998a)
MAE14UMR1 ΔagfBA101::Kmr; pdar28This study
MAE18UMR2 ΔagfBA101::Kmr; pdar28/37 Römling et al. (1998a)
MAE28UMR1 ΔagfD101::Kmr; sawThis study
MAE32UMR1 PagfD2; rdar28/37 Römling et al. (1998a)
MAE41MAE32 rpoS::pRR10(ΔtrfA); bdar28/37 Römling et al. (1998a)
MAE51MAE32 ΔagfD101; sawThis study
MAE52UMR1 PagfD1; rdar28/37 Römling et al. (1998a)
MAE80MAE51, pBAD30This study
MAE81MAE51, pUMR15This study
MAE96MAE32 ΔagfBA102; pdar28/37This study
MAE97MAE52 ΔagfBA102; pdar28/37This study
MAE102MAE32 ΔagfBA102 adrA101::MudJ; sawThis study
MAE103MAE52 ΔagfBA102 adrA101::MudJ; sawThis study
MudJ1MAE81 adrA101::MudJThis study
MudJ6MAE81 adrA102::MudJThis study
ADR1aMAE52 adrA101::MudJ; bdar28/37This study
ADR1bMAE51 adrA101::MudJThis study
ADR6aMAE52 adrA102::MudJ; bdar28/37This study
ADR6bMAE51 adrA102::MudJThis study
ADR1cADR1a rpoS::pRR10(ΔtrfA); bdar28/37This study
ADR1fUMR1 adrA101::MudJ; bdar28This study
E. coli K-12
endA1 hdR17 supE44 thi-1 recA1 gyrA relA1
Δ (lacZYA-argF)U169 (m80lacZΔM15)
Laboratory collection
pBAD30Arabinose-regulated expression vector, Ampr Guzman et al. (1995)
pMAK700, pMAK705
Cmr, temperature-sensitive replicon derived from
Hamilton et al. (1989)
pSU19Cmr Bartolome et al. (1991)
pUC4KAmpr Kmr cassettePharmacia
pUMR3cpMAK700 ΔagfBA102This study
pUMR5pMAK700 ΔagfD101::KmThis study
pUMR8bpMAK705 ΔagfD101This study
pUMR15pBAD30::agfDThis study
Table 2. Phenotypic characterization of strains.
 Strain (genotype)

agfBA adrA)
  • a

    .+, Dye binding according to morphotype (red, pink or brown colour); −, white colony.

  • b

    . The relative level of dye binding was qualitatively assessed, using as reference strains MAE52 (binding designated ++) and UMR1 at 37°C (−).

  • c

    . The relative signal of formic acid-sensitive AgfA in Western blots was qualitatively assessed using as reference strains MAE52 (signal strength designated ++) and UMR1 at 37°C as negative control.

  • d

    . ND, not determined.

  • e

    .+, Fibronectin binding; −, no binding.

  • f . As shown in Fig. 3.

  • g . As shown in Figs 4 and 5.

Morphotyperdar28/37rdar28 saw37saw28/37pdar28/37bdar28/37saw28/37
binding e


in liquid
culturef (28°C)

Adherence to+++++++++
glassf (28°C)
Pellicle formationf
+, Rigid

+, Fragile

matrixg (37°C)
+, Fibrous

+, No structure
+, Fibrous

Figure 3.

Characterization of various forms of multicellular behaviour in agfD+ (MAE52/MAE32), agfBA (MAE18/MAE96/MAE97), adrA (ADR1a) and agfD (MAE51) mutants. The agfBA adrA mutants (MAE102/MAE103) behaved exactly alike the agfD mutant.

A. The morphotypes of MAE52, MAE97, ADR1a and MAE51 were grown on LB agar without salt plus CR at 37°C for 24 h.

B. Colony consistency of MAE32, MAE18, ADR1a and MAE51 grown for 24 h at 37°C on LB agar without salt.

C. Cell clumping in liquid cultures of MAE32, MAE97, ADR1a and MAE51 grown in LB medium without salt at 28°C for 24 h with shaking (210 r.p.m.). Magnification ∞ 630.

D. Adhesion patterns of MAE52, MAE96, ADR1a and MAE51 to glass (left panel). Cells were grown in minimal medium at 28°C for 24 h with shaking. Adherent cells were stained with Crystal violet as described in Experimental procedures. Pellicle formation of respective strains (right panel). Pellicle formation was judged after growth in standing LB medium without salt at 37°C after 24 h. Symbols are: (1) ring formation at the air–liquid interface; (2) pellicle formation; (3) turbid solution; (4) cells at the bottom.

Figure 4.

Scanning electron microscopy of plate grown cells. MAE52 (agfD+) exhibits the matrix structure of the rdar morphotype (A), whereas ADR1a (adrA) is characterized by a less dense and more fibrous structured matrix around individual cells (C). MAE18 (agfBA) shows a very densely packed unstructured matrix in direct contact to the bacterial surface and between cells (B). MAE51 (agfD) is devoid of matrix material between cells (D). All colonies were grown for 48 h on LB medium without salt at 37°C. Bars represent 1 µm.

Figure 5.

Ultra-thin sections of Ruthenium-red embedded colonies. The matrix structure, which is already seen in scanning electron microscopy (Fig. 4), was confirmed by the ultrathin sections. (A) MAE52 (agfD+); (B) MAE18 (agfBA); (C) ADR1a (adrA); (D) MAE51 (agfD). The cells devoid of matrix (MAE51) look more damaged than all the other cells which have a matrix. All colonies were grown for 48 h on LB medium without salt at 37°C. Bars represent 0.5 µm.

Former studies had shown that knocking out the building of thin aggregative fimbriae by creation of a deletion in agfA or agfBA resulted in a pink mutant form, the pdar morphotype (Fig. 3A). The cells in the pdar colony are still connected in an elastic fashion (Fig. 3B) mediated by a matrix which appears to lack order (Figs 4B and 5B). In liquid culture, most of the cells are singular and only small clumps with loosely connected cells are seen (Fig. 3C). In the biofilm formed at the wall of the air–liquid interface of liquid cultures, the cells stick more to each other than to the abiotic surface. No pellicle is formed in standing culture, although a small ring of bacteria is maintained on the glass at the air–liquid interface (Fig. 3D). From these experiments, one can conclude that the absence of thin aggregative fimbriae abolished most of the multicellular behaviour of the rdar morphotype at the level of short-range cell–cell interactions and the community behaviour of the cells. However, extracellular factor(s) remained which mediated mainly a new type of long-range cell–cell interactions.

To establish that the rdar morphotype mediated by agfD is not merely a laboratory curiosity, virulence studies were carried out in BALB/c mice using the upregulated rdar morphotype MAE52, the regulated rdar morphotype UMR1 (the wild-type strain commonly used for infection studies) and its agfD and agfBA mutants, MAE28 and MAE14 respectively. As shown in Table 3, the constitutive rdar morphotype showed an ≈ 10-fold increase in LD50 while a complete lack of agfD expression decreased the LD50 threefold. An agfBA mutant remained as virulent as the wild type. Therefore, the absence or presence of the rdar morphotype influenced, although not dramatically, the virulence of S. typhimurium in the mouse model.

Table 3. LD50 values of strains expressing different rdar morphotype programmes.

UMR1 (rdar28)1.8 ∞ 104
MAE14 (pdar28)2.9 ∞ 104
MAE28 (saw28/37)<0.7 ∞ 104
MAE52 (rdar28/37)1.8 ∞ 105

Contribution of other genes to the multicellular morphotype

Comparison of the agfBA and agfD mutant morphotypes showed that genes other than agfBA, which was regulated by AgfD contributed to the multicellular behaviour (Römling et al., 1998a; this study). However, two further possibilities of regulatory patterns remained. AgfEFG could contribute to the pdar morphotype if AgfD would regulate its own promoter, or genes elsewhere in the S. typhimurium chromosome were responsible for the pdar morphotype. Therefore, autoregulation of the agfD promoter was tested by comparison of primer extension results from the wild-types UMR1 and MAE32 and their corrresponding in-frame agfD deletion mutants MAE50 and MAE51. Neither the constitutive nor the regulated agfD promoter was dependent on AgfD (Fig. 6A). Immunoblot analyses detecting agfG showed that the shortened transcripts were still translated into proteins (Fig. 6B). Therefore, expression of agfEFG most likely did not contribute to the long-range cell–cell interactions seen in the agfBA mutant and not in the agfD mutant.

Figure 6.

Transcription and translation at the agfDEFG operon is not dependent on AgfD.

A. Primer extension analysis of agfD+ and ΔagfD mutants of the regulated and constitutive rdar morphotypes at the agfD promoter. MAE50 is the ΔagfD mutant of UMR1 (rdar28), MAE51 the ΔagfD mutant of MAE32 (rdar28/37). Primer PEXD1 located 58 bp downstream of the agfDEFG start codon was used as described recently (Römling et al., 1998b). Cells were grown on LB agar without salt for 48 h at 28°C.

(B) Western blot analysis of AgfG in the agfD+ and ΔagfD mutants grown on LB agar without salt at 28°C and 37°C for 48 and 24 h respectively. Lanes: (1) UMR1 (rdar28); (2) MAE32 (rdar28/37); (3) MAE50 (rdar28, ΔagfD); (4) MAE51 (rdar28/37, ΔagfD).

To elucidate if other genes might be regulated by agfD we used random transcriptional fusion analysis. MAE51 carrying an in frame deletion in agfD was transformed with plasmid pUMR15 carrying the agfD gene under the control of the arabinose-inducible araBAD promoter in pBAD30 yielding MAE81. Random mutagenesis was then conducted in MAE81 using MudJ as previously described (Hughes and Roth, 1988) and mutants were plated on LB/EGTA plates. Sixteen thousand mutants were screened by patching the colonies onto Luria–Bertani (LB) plates without salt containing X-Gal in the presence and absence of arabinose. Mutants with altered β-galactosidase activities were isolated and differences in colour were confirmed. Next, each MudJ fusion was transduced into the constitutive rdar morphotype MAE52 and the agfD mutant MAE51 to confirm that expression of the fusions is agfD dependent (Fig. 7A). Two agfD-dependent fusions adrA101::MudJ and adrA102::MudJ, were finally found localized in the same gene at nucleotide position 222 and 635 respectively. The identified gene is homologous to yaiC in E. coli, the function of which is not yet characterized. We named this gene adrA.

Figure 7.

Dependency of adrA transcription on agfD, rpoS and environmental conditions.

A. Expression of adrA is completely dependent on agfD. Strains used are ADR1a (adrA101::MudJ), ADR1b (ΔagfD adrA101::MudJ), ADR6a (adrA102::MudJ) and ADR6b (ΔagfD adrA102::MudJ).

B. Dependency of adrA transcription on environmental conditions. Strains: (1) ADR1a (adrA101::MudJ); (2) ADR1b (ΔagfD adrA101::MudJ).

C. Recovery of transcription activity of rdar28 at 37°C under iron-limiting conditions. ADR1f (rdar28adrA101::MudJ) was used.

D. Dependency of adrA transcription on rpoS. Strains used were ADR1a (adrA101::MudJ) and ADR1c (adrA101::MudJ rpoS). Cells were grown at 37°C on LB agar without salt or on the medium indicated for 24 h; growth under anaerobic conditions was for 6 days; growth in liquid culture was at 28°C for 24 h.

Phenotypic characterization of the adrA mutant

Mutation in the adrA gene changed the rdar morphotype from a red to a brown colony which looked identical to an rpoS mutant (Römling et al., 1998a) and was therefore classified as a bdar morphotype (Fig. 3A). The cells formed small clumps in the colony (Fig. 3B) which were connected by a less abundant and more fibrous extracellular matrix than in the rdar morphotype (Figs 4C and 5C). In liquid culture, besides single cells, clumps were visible, but they were smaller and often did not show such tight cell–cell interactions as in the rdar morphotype (Fig. 3C). The pellicle formed in standing culture was much more fragile than in the rdar morphotype. In addition, binding of Calcofluor was found to be a feature associated with agfD expression, but absent in the adrA mutant (Table 2). Therefore, Calcofluor binding was specific for the uncharacterized extracellular factor(s). In conclusion, most of the multicellular behaviour remained, in particular at the community level. However, the local cell–cell interactions turned out to be much more fragile.

We also examined whether the mutation in adrA had any effect on the expression or binding characteristics of thin aggregative fimbriae. Western blot analysis showed that formic acid-sensitive AgfA was present (Fig. 8), an indication for polymerized intact thin aggregative fimbriae on the surface of the bacteria (Hammar et al., 1996). The amount of fibronectin binding and binding to abiotic surfaces in the adrA mutant was comparable to the rdar morphotype (Table 2 and Fig. 3F). Thus, the binding characteristics assigned to the thin aggregative fimbriae were not changed.

Figure 8.

Biogenesis of thin aggregative fimbriae is not dependent on adrA expression. Strains used: (1) ADR1a (adrA101::MudJ); (2) MAE52 (rdar28/37adrA101::MudJ); (3) ADR1b (ΔagfD adrA101::MudJ); + with formic acid treatment; − without formic acid treatment. Cells were grown at 37°C for 24 h on LB agar without salt. Protein from an equal number of input cells was loaded in each lane and controlled by the detection of the OmpA protein.

Regulation of adrA by agfD, rpoS and environmental cues

Both transcriptional fusions, adrA101::MudJ and adrA102::MudJ, were completely regulated by agfD (Fig. 7A). AdrA102::MudJ showed an ≈ 3.6-fold lower activity than adrA101::MudJ indicating regulatory patterns at the level of RNA in adrA. The constitutive rdar morphotype is expressed under various environmental conditions, such as 28°C and 37°C, minimal medium, stationary phase in liquid medium, Fe-depletion and anaerobicity (Römling et al., 1998a). No expression of the rdar morphotype occurs at high osmolarity and in the exponential phase of growth. The regulated rdar morphotype, not expressed at 37°C, could be recovered under iron-limiting conditions (Römling et al., 1998a). The transcriptional fusions which were examined for adrA activity followed the described pattern of regulation (Figs 2 and 7). Transcriptional activity was completely dependent on agfD under all conditions examined.

To evaluate the dependence of adrA on rpoS, a double mutant ADR1c was constructed (Table 1). ADR1c has a brown colony morphology consistent with previous results showing that rpoS does not influence the expression of thin aggregative fimbriae in the constitutive rdar morphotype (Römling et al., 1998a). Because the adrA mutation alone also conferred a bdar morphotype we concluded that the adrA mediated pathway should be completely dependent on rpoS. However, analysis of the transcriptional activity of adrA showed that a significant signal of ≈ 50% remained in the rpoS mutant ADR1c (Fig. 7C) implying that the adrA pathway might contain other genes downstream of adrA.

Characterization of the adrA gene

The partial sequence derived from the identification of the insertion of the MudJ-fusion showed that adrA is homologous to yaiC in E. coli. The whole sequence of adrA was determined by sequencing a PCR product containing the adrA gene. Genomic DNA was amplified using primers for the closest adjacent genes sequenced in S. typhimurium, ddlA and proC, which were identified by comparison with the E. coli gene order. As concluded from the direction of the genes adjacent to adrA, the gene is the last in an operon or transcribed as a separate open reading frame (ORF). Therefore, no downstream effects from the MudJ-fusions are to be expected. The ORF from adrA is 1113 bp in size and encodes a protein of 41.5 kDa. The N-terminal domain from aa 1–190 has no homology to sequences in the database. Four transmembrane helices were identified in that region (Fig. 9). The C-terminal end is highly homologous to the GGDEF motif (Hecht and Newton, 1995) found in over 50 proteins of Gram-negative and Gram-positive bacteria. The function of the GGDEF motif is unknown. Sequence alignment revealed that AdrA is one of the simplest members of the GGDEF family, consisting of only two domains (Fig. 9). The similarity of AdrA to YaiC is 81.4% with 75.4% aa identity, whereby the C-terminal domain with 90.3% similarity is more conserved than the N-terminal domain with 71.2% similarity.

Figure 9.

Comparison of the domain structures of proteins with GGDEF motif for which a phenotype or function is proposed. PDEA1 (phosphodiesterase A from Acetobacter xylinum; AF052517, Tal et al., 1998); DGC1 (diguanylate cyclase from A. xylinum; AF052517; Tal et al., 1998); PleD from Caulobacter crescentus (Q46020; Hecht and Newton, 1995); AdrA from S. typhimurium. OS1, oxygen-sensing domain similar to FixL; OS2, oxygen-sensing domain similar to NifL; GGDEF, EAL domains of unknown function; D1/2, receiver domain similar to the CheY phosphorylation motif (Volz, 1993). Shaded boxes indicate putative transmembrane domains. A consensus sequence for the GGDEF domain which is divided into four regions (Hecht and Newton, 1995) is shown below the AdrA sequence.

AgfD regulates two independent pathways

The experiments described above showed that the biogenesis and function of thin aggregative fimbriae is not influenced by the expression of adrA (Fig. 8 and Table 2). Also, adrA transcription is not influenced by the expression of agfBA (data not shown). In addition, an agfBA mutant showed a pink colour, bound Calcofluor and produced extracellular factor(s) mediating mainly long-range cell–cell interactions (Figs 3A, 4 and 5; Table 2). However, from the experiments conducted so far it was not clear whether adrA was involved in the biogenesis of these extracellular factor(s). It was also not clear whether adrA and the thin aggregative fimbriae are the only pathways regulated by agfD. In order to answer these two questions the adrA101::MudJ mutation was transduced into the agfBA mutants MAE96 and MAE97. The resulting adrA agfBA double mutants, MAE102 and MAE103, respectively, showed the behaviour of agfD mutants (Figs 3–5 and Table 2), the multicellular morphotype was abolished. These experiments showed that adrA is involved in the production of the extracellular factor(s) mediating long-range cell–cell interactions. In addition, the complete lack of multicellular behaviour of the agfBA adrA double mutant implies that agfD regulates only two independent pathways. Figure 10 shows a scheme of the possible regulatory network mediated by agfD.

Figure 10.

Model illustrating the role of agfD in the regulation pathway leading to the rdar morphotype. The thin aggregative fimbriae, the unknown extracellular substance(s) and flagella, which are not regulated by AgfD (Römling and Rohde, 1999), are all required for a fully expressed rdar morphotype.


In this report, we have identified agfD as playing a key role in the expression of the multicellular behaviour in S. typhimurium. By sequence comparison, we present evidence that AgfD is a DNA-binding response regulator. AgfD has homology with the UhpA (FixJ) family over the entire length of the protein with the putative phosphorylation site D59 conserved. However, AgfD might harbour a new subtype of a receiver domain, because invariant residues required for phosphorylation and activation are not conserved (Volz, 1993). AgfD accumulates through the logarithmic growth phase, being expressed predominantly in the stationary phase. Because the rdar morphotype is only expressed during the stationary phase, we postulate that phosphorylation of AgfD is needed to turn on the rdar morphotype. However, the corresponding kinase is not known. Experiments with AgfD expressed from a plasmid showed that under inducing conditions AgfD expression starts later than in the wild type and overshooting AgfD expression is downregulated. Therefore, AgfD expression is highly regulated for growth phase-dependent expression and quantity either on the RNA or protein level.

The deregulation (rdar28/37) or absence (ΔagfD) of the rdar morphotype did not dramatically influence the virulence of S. typhimurium in the mouse model of typhoid fever. This virulence study using strains with defined mutations confirmed and extended earlier experiments with an uncharacterized upregulated rdar morphotype of strain SR-11 (Sukupolvi et al., 1997). Lack of the extracellular matrix decreased the LD50 dose of the organism. No influence on the virulence of a mutant only missing thin aggregative fimbriae was also shown by other investigators (van der Velden et al., 1998) and for Salmonella enteritidis in a chicken model (Allen-Vercoe et al., 1999).

The ΔagfD mutant of S. typhimurium completely lacked the multicellular behaviour (Fig. 3) in contrast to the ΔagfBA mutants where long-range cell–cell interactions were still present (Römling et al., 1998a; this study). Electron microscopy revealed a gum-like matrix which appears to lack order among the cells in the ΔagfBA mutants which specifically bound the dye Calcofluor (Figs 4B and 5B and Table 2). However, the nature of the matrix, which could consist of proteins or carbohydrates or both is presently unknown.

In order to answer the question of alternative pathways regulated by agfD we identified two transcriptional fusions located in the adrA gene. AdrA is homologous to yaiC in E. coli whose function has not been identified (Blattner et al., 1997). AdrA contains two domains, a transmembrane domain and the GGDEF motif, the function of which is unknown (Fig. 9). Despite the GGDEF motif being present in over 50 proteins from Gram-positive and Gram-negative bacteria, a function or a phenotype has been unravelled for only a few of these genes (Hecht and Newton, 1995; Alksne and Rasmussen, 1997; Tal et al., 1998). Because the GGDEF motif was found in combination with the receiver domain of response regulators (Parkinson and Kofoid, 1992) a regulatory function of the GGDEF domain either by protein–protein interactions or DNA binding was proposed (Aldridge and Jenal, 1999). Therefore, adrA is most likely not the only gene in the cascade leading to the unknown extracellular substance(s) (Fig. 10), but may have transport and/or signalling functions. Other genes involved in the biosynthesis of the extracellular substance(s) remain to be determined.

A scheme of the regulatory pathway leading to the rdar morphotype is shown in Fig. 10. Transcription of agfD is completely dependent on ompR, a global response regulator. Both agfD and ompR mutants show a white colony, the saw morphotype (this study; Römling et al., 1998a). Further down, AgfD regulates the expression of at least two pathways, thin aggregative fimbriae and the production of unknown extracellular substance(s). Those two pathways can be discriminated by the differential binding of the dye CR. Deletion of thin aggregative fimbriae lead to a pink colony (pdar), whereas deletion of genes involved in the production of the unknown extracellular substance, adrA and rpoS, created a brown colony (bdar). On the molecular level, AgfD most likely initiates transcription directly at the agfBA promoter. In addition, AgfD regulates adrA on the transcriptional level; whether this regulation is direct or indirect is not known. RpoS downregulates the extracellular substance(s) 100% as judged by Calcofluor binding (data not shown). However, in the upstream pathway adrA transcription is only diminished by 50% in an rpoS mutant (Fig. 7D). Therefore, there might be an additional regulation point below adrA in the cascade leading to the unknown extracellular substance(s).

AgfD plays a key role in the activation of the two different pathways. So far, it seems that both pathways leading to thin aggregative fimbriae and to the unknown extracellular substance(s) are jointly switched on in the wild-type strains. Finding conditions or strains where only one of the extracellular substances is expressed would help in the identification of the responsible regulatory elements.

With the help of the mutations in adrA and agfBA it was possible to separate the roles of the two pathways leading to multicellular behaviour; the expression of thin aggregative fimbriae regulates the community behaviour and short-range cell–cell interactions, whereas adrA is involved in long-range cell–cell interactions. However, an extracellular matrix also serves frequently as a protective shield against harmful substances such as oxidizing agents and antibiotics (Costerton et al., 1999; Yildiz and Schoolnik, 1999). Subsequent functional assays will show whether the two substances play distinct protective roles.

Using several phenotypic (e.g. binding of different dyes) and microscopic assays only two extracellular substances could be differentiated so far. However, the possibility remains that additional pathways contribute to the rdar morphotype (Guard-Petter et al., 1996; Allen-Vercoe et al., 1997; Austin et al., 1998). In the pink morphotype agfA colonies of S. enteritidis, uncharacterized novel fimbrial structures were detected (Allen-Vercoe et al., 1997). In another study, type I fimbriae seem to reduce the ability of S. enteritidis cells to form aggregates when adhering to an abiotic surface, biofilm formation (Austin et al., 1998). Whether these two fimbrial types are regulated by agfD remains to be elucidated. It is possible that those fibres contribute only to certain aspects of multicellular behaviour that were investigated in those studies, colony morphology or biofilm formation respectively. For example, flagella which are not regulated by agfD contribute to the community behaviour of the cells but not to colony morphology, matrix formation or quantity of adherence (Römling and Rohde, 1999; Fig. 10) The regulation of the physiological requirements and the metabolism of cells in aggregates are certainly different from cells in the planctonic status. Oxygen limitation, diffusion barriers for various substances and tight cell–cell interactions occur in the cell community. Elucidation of these physiological pathways may unravel new targets for the prevention of bacterial adherence.

Experimental procedures

Bacterial strains and growth conditions

All E. coli and S. typhimurium strains used in this study are listed in Table 1. For maximal expression of the multicellular morphotype S. typhimurium strains were grown on Luria–Bertani (LB) agar plates without NaCl at 28°C and 37°C, unless otherwise stated. Calcofluor white (fluorescent brightener 28; 50 µg ml−1) plates and Congo red (CR)–LB agar without salt (Römling et al., 1998a) were used to judge colony morphology and colour. LB without salt agar was supplemented with 0.5 M NaCl, 2,2′-dipyridyl or 0.2% KNO3 and minimal M9 medium with 0.2% glucose (Ausubel et al., 1994) was also used. Aerobic conditions in liquid cultures were created by filling one-tenth of the flask while shaking at 210 r.p.m. Antibiotics were used at the following concentrations: ampicillin (100 µg ml−1), chloramphenicol (20 µg ml−1), kanamycin (30 µg ml−1) and nalidixic acid (50 µg ml−1).

Construction of plasmids and strains

Plasmids used in this study are listed in Table 1. The agfBA deletion in MAE14 was constructed as in MAE18 (Römling et al., 1998a) but with UMR1 as parent strain. An in frame deletion in agfBA was constructed as follows: plasmid pUMR3a (Römling et al., 1998a) carrying an out of frame deletion for agfBA was cut with PstI and successively treated with T4-DNA-Polymerase to create blunt ends. Religation yielded plasmid pUMR3c with an in frame deletion. To generate a deletion in agfD, an allelic exchange plasmid was constructed by cloning DNA fragments amplified with primers AGFF (CGTAAGCTTATCCAACGCTGAGG, HindIII-site underlined)-AGFD2 (TGGATGCATACCCAGGCAGTTTCATGG, NsiI-site underlined) and AGFD1 (TGGATGCAT CGTAGCTTGCAGAGATGG, NsiI-site underlined)-AGFB3 (GGAGGATCCAGATCATAATTTGTCG, BamHI-site underlined) in pMAK700 and inserting a Kmr cassette into the NsiI site (pUMR5). To generate an in frame deletion in agfD without any resistance marker, the AGFF–AGFD2 fragment cloned in pUC8 and the AGFD1–AGFB3 fragment cloned in pBluescript were cut out with HindIII/SalI and SalI/BamHI, respectively, and cloned in HindIII/BamHI cut pMAK705 to yield pUMR8b. With the allelic exchange procedure, 180 of 216 aa of agfD were removed. The primer Sp31 (CTCGCATGCTTACCGCCTGAGATTATCG, SphI-site underlined) and AGFD3 (GAGTCTAGAGTGGAGGTTCATCATGTTTA, XbaI-site underlined, mismatch for optimal Shine-Dalgarno sequence italic) were used to amplify the whole promoterless agfD gene, which was subsequently cloned into XbaI/SphI-cut pBAD30 to yield pUMR15. Arabinose-dependent expression was confirmed by the ability to complement the ΔagfD101 mutation achieving a full rdar morphotype with 0.1% arabinose at 28°C. Allelic exchange with the pMAK-system was carried out as described earlier (Römling et al., 1998b; Hamilton et al., 1989). Phage P22 HT105/1 int-201 was used for transduction of S. typhimurium strains and transductants were checked for pseudolysogenicity and lytic phages according to recommended protocols (Maloy et al., 1996). Random MudJ insertions in strain MAE81 were carried out as described (Hughes and Roth, 1988). All constructed strains were checked by Southern hybridization and/or PCR. To isolate genes containing MudJ-fusions, SalI-digested genomic DNA was cloned into pSU19 and Cm/Km-resistant colonies were selected. A primer complementary to the left end of MudJ and a primer in the plasmid were used to determine the chromosomal DNA sequence. Analyses of sequences were performed with the Genetics Computer Group package, version 9 (GCG, University of Wisconsin).

General molecular biology methods

Plasmid DNA was purified using the method recommended by Applied Biosystems. Isolation of chromosomal DNA, purification of PCR products, plasmid transformation and enzymatic manipulations (restriction digestion, ligation, phosphorylation, PCR and cycle sequencing) were carried out using standard protocols (Ausubel et al., 1994; Sambrook et al., 1989). AdrA was amplified by long-range PCR (Boehringer) using primers DDLA: CAGCAAATCCTGATGACTTTCG and PROC: GAAGCGATGACGAAGTGTATGG according to the recommended protocols. Isolation of RNA by the hot-phenol method and primer extension were carried out as described recently (Römling et al., 1998b).

Protein techniques

Bacterial lysates for the detection of AgfA in Western blots and gel separation of proteins were performed as reported recently (Römling et al., 1998b) Otherwise, standard methods were applied for bacterial lysates, protein gels, Western blots and detection of antigens (Ausubel et al., 1994). The following primary antibodies were used: anti-AgfA (1:4000), anti-AgfD (1:5000), anti-AgfG (1:4000; Loferer et al., 1997), anti-OmpA (1:90 000; a gift from E. Morfeldt) and anti-DnaK (1:2000; Biomol). Peptide antibodies against AgfD were created using the sequences (C)EADKKLIHYWQDNLSRKNN (aa 62–80) and (K)TPDDYPYREIENWP (aa 89–102) located in the less conserved N-terminal receiver module of the molecule. Affinity purification of antiserum was carried out with Hi-Trap columns coupled with the two peptides.

β-Galactosidase levels were measured as described by Miller (1972). β-Galactosidase activity is calculated using the formula: units = 1000 × {[OD420 − (1.75 × OD550)]/(t × V × OD660)} with t = reaction time in min; V= volume of cell suspension in assay in ml; OD420 and OD550 of reaction solution; OD660 of original cell suspension.

Fibronectin-binding assay

The fibronectin-binding assay was carried out as described by Olsén et al. (1989).

Evaluation assays for multicellular behaviour

The rdar morphotype was judged visually on CR-plates. Adherence assays to polystyrene and glass were carried out as described elsewhere (Römling and Rohde, 1999). The morphology of cell clumps was inspected visually and by light microscopy under phase contrast. Pellicles from standing cultures were examined visually.

Electron microscopy

Transmission electron microscopy. Plate grown colonies were fixed in a solution containing one part 0.2 M cacodylate, pH 7.0, one part 3% glutaraldehyde and one part 0.15% ruthenium red for 1 h on ice. After washing with a solution containing one part 0.2 M cacodylate, pH 7.0 and one part 0.15% ruthenium red, a further fixation step was carried out with a solution consisting of one part 0.2 M cacodylate, pH 7.0, one part 3% glutaraldehyde and one part 1% aqueous osmium tetroxide for 3 h at room temperature. Subsequently, colonies were washed with 0.2 M cacodylate, pH 7.0, containing 0.15% ruthenium red and embedded in aqueous 1.5% water agar. The embedding was carried out according to the Spurr protocol (Spurr, 1969). Ultra-thin sections were cut with a diamond knife and counterstained with uranyl acetate and lead citrate. Samples were examined in a Zeiss TEM910 at an acceleration voltage of 80 kV.

Scanning electron microscopy. Colonies were fixed in a fixation solution containing 3% glutaraldehyde and 5% formaldehyde in phosphate buffered saline (PBS) for 1 h on ice. After washing with PBS, colonies were dehydrated with a graded series of acetone, dried at the critical point of liquid CO2, mounted onto aluminium stubs and sputter-coated with an ≈ 10-nm-thick gold film. Some samples were fractured with tweezers after being critical point dried. Samples were examined in a Zeiss field emission scanning electron microscope DSM 982 Gemini using the Everhart-Thornley Se-detector and the in-lens SE-detector in a 50:50 ratio.

Virulence assays

Six- to eight-week-old female BALB/c mice were deprived of food and water for 6 h prior to per oral (p.o.) inoculation of 30 µl bacterial suspension. Virulence of strains was established by the 50% lethal dose (LD50) assay by applying serial 10-fold dilutions of bacterial suspensions ranging from 104 to 108 CFU to groups of four mice. Viability of the animals was assayed 30 days post-infection; the LD50 dose was determined by the method of Reed and Muench (1938).

Note added in proof

Sequence data were submitted to the EMBL data library under the accession number AJ271071.


This work was in part supported by a DFG grant (RO 2023/3–1) to U.R. We would like to thank G. Maaß and J. Wehland for continuous support and interest. We gratefully appreciate the introduction to mouse experiments by Per Falk. H. Loferer and E. Morfeldt are acknowledged for their generous gifts of antibodies. We appreciate the skillful technical assistance of E. Müller in the electron microscopy work. Thanks to P. Hagendorf and R. Munder for running the sequencing gels. U.R. is a recipient of a fellowship from the program ‘Infektionsbiologie’ from the Bundesministerium für Forschung und Technologie (BMFT).