Adherence of bacteria to eukaryotic cells is essential for the initiation of infection in many animal and human pathogens, e.g. Neisseria gonorrhoeae and Pseudomonas aeruginosa. Adhesion-mediating type IV pili, filamentous surface appendages formed by pilin subunits, are crucial virulence factors. Here, we report that type IV pilus-dependent adhesion is also involved in plant–bacteria and fungus–bacteria interactions. Nitrogen-fixing, endophytic bacteria, Azoarcus sp., can infect the roots of rice and spread systemically into the shoot without causing symptoms of plant disease. Formation of pili on solid media was dependent on the pilAB locus. PilA encodes an unusually short (6.4 kDa) putative pilin precursor showing 100% homology to the conserved N-terminus of the Pseudomonas aeruginosa type IV pilin. PilB encodes for a 14.2 kDa polypeptide showing similarity to FimF, a component of type I fimbriae of Escherichia coli. It was found to be extruded beyond the cell surface by immunofluorescence studies, and it may, therefore, be part of a pilus assembly complex or the pilus itself. Both genes are involved in the establishment of bacteria on the root surface of rice seedlings, as detected by fluorescence microscopy. Moreover, both genes are necessary for bacterial adhesion to the mycelium of an ascomycete, which was isolated from the same rhizosphere as the bacteria. In co-culture with the fungus, Azoarcus sp. forms complex intracytoplasmic membranes, diazosomes, which are related to efficient nitrogen fixation. Adhesion to the mycelium appears to be crucial for this process, as diazosomes were absent and nitrogen fixation rates were decreased in pilAB mutants in co-culture.
The first crucial event in the infectious process of many animal and human pathogenic bacteria is the attachment to epithelial cells of the host. Type IV pili play an essential role in mediating bacterial adherence to and colonization of host cell surfaces and are virulence factors in Neisseria gonorrhoeae, N. meningitidis, Pseudomonas aeruginosa and Moraxella bovis (Strom and Lory, 1993; Hahn, 1997). In addition, type IV pili are involved in the modulation of target cell specificity (Jonsson et al., 1994), bacteriophage adsorption (Bradley, 1974), twitching motility (Bradley, 1980; Darzins and Russell, 1997), transformation competence (Fusenegger et al., 1997) and social gliding in bacteria (Wu and Kaiser, 1995).
Type IV pili are filamentous cell appendages of 6 nm thickness assembled primarily from a single protein subunit, pilin, which consists of approximately 150 amino acid residues. A short, positively charged leader peptide, which is cleaved off the prepilin, and a highly conserved N-terminal region are characteristic of type IV pilins (Strom and Lory, 1993). The hydrophobic N-terminal region is postulated to form the centre of the pilus by hydrophobic interactions of 54-amino-acid α-helices of the pilin subunits, whereas the C-terminal part is exposed to the pilus surface (Parge et al., 1995; Forest and Tainer, 1997). This hypervariable region is the most highly exposed portion (Forest and Tainer, 1997) and the source of strain-specific antigenic variation and interaction with host surfaces (Nassif et al., 1993). Additional proteins located at the tip of the filament, adhesins, may also interact with host receptors in Neisseria (Rudel et al., 1995).
To date, the involvement of type IV pili in adhesion has been described for microbe–microbe (Wu and Kaiser, 1995) or animal/human–microbe interactions only. Here, we report for the first time evidence for a more general role of type IV pili, also mediating attachment of bacteria to plant roots and fungal mycelium.
Nitrogen-fixing, Gram-negative bacteria of the genus Azoarcus occur in association with grasses (Reinhold-Hurek et al., 1993). They show similar colonization and infection patterns to their original host Kallar grass and rice seedlings in the laboratory (Hurek et al., 1994b; Reinhold-Hurek and Hurek, 1998a). Undifferentiated tissues above the root tips and the points of emergence of lateral roots are the sites of primary colonization of and entry into the plant (Reinhold-Hurek and Hurek, 1998b). Without causing symptoms of plant disease, these endophytes are capable of invading the central parts of the root (xylem) and of systemic spreading into the stele. In situ hybridization studies have recently demonstrated that Azoarcus nitrogenase genes can be expressed inside the roots of field-grown Kallar grass (Hurek et al., 1997a). Azoarcus also interacts with rhizosphere fungi (Hurek et al., 1995; 1997b). In co-culture with an asexual ascomycete originating from Kallar grass roots, Azoarcus sp. strain BH72 adheres tightly to the actively respiring mycelium. Under these conditions, the bacteria form novel intracytoplasmic membrane stacks, diazosomes (Hurek et al., 1995), that are related to a high efficiency of nitrogen fixation (Hurek et al., 1994b). Here, we present evidence that type IV pili are involved in the colonization of both plant and fungal host surfaces, indicating that there may be common traits in the initial steps of interactions of bacteria with eukaryotic organisms of their habitat.
A short type IV pilin is encoded in the pilAB locus of Azoarcus sp. BH72
Mutant BH1599, which had been obtained by transposon mutagenesis from wild-type Azoarcus sp. BH72, showed a pleiotropic phenotype: it was affected in attachment to plant roots and fungal mycelium (see below). Sequence analysis of the site of transposon insertion revealed two open reading frames (ORFs), pilA and pilB, the transposon disrupting pilA (Fig. 1A and B). Upstream of pilA, sequences homologous to σ54-dependent promoters and potential binding sites for the transcriptional activators NifA and NtrC were located (Fig. 1B). Northern blot analysis with a pilAB gene probe indicated that both genes were co-transcribed as a 1.1 kb transcript (Fig. 1C). The N-terminus of the deduced protein PilA showed high homology with type IV prepilins, being 100% identical to a Pseudomonas aeruginosa prepilin in amino acids 5–30 (Fig. 2A). The pilA ORF was unusually short (59 amino acids, 6.4 kDa) in comparison with prepilins of P. aeruginosa (149 amino acids) or N. gonorrhoeae (166 amino acids) (Fig. 2A). This result was confirmed by the sequencing of several different clones retrieved from a genomic library and of clones from a cDNA library constructed from mRNA of Azoarcus sp. BH72. Sequencing errors could also be excluded, as multiple stop codons were present in two reading frames. In the reading frame of pilA, stop codons were located at nucleotide positions 367, 370, 382, 433, 439, 502, 526 and 544. In frame +1, they were located at nucleotide positions 395, 449 and 563. In frame −1, another gene was identified (pilB ).
In order to screen for additional gene copies of pilA that might have the normal size of a pilin gene, we carried out Southern hybridization studies of genomic DNA with an oligonucleotide probe corresponding to the conserved region (nucleotides 207–233) of pilA. Only one hybridizing fragment could be detected at high stringency (72°C) (Fig. 3) and, thus, no other identical gene copy was present in the genome of Azoarcus sp. BH72. Hybridization at lower stringency (60°C) yielded an additional hybridizing fragment (1.0 kb in Pst I-digested DNA, lane 4 in Fig. 3), which might represent a gene with a homologous but not identical 5′ end. However, cloning and sequencing of this fragment showed that neither a pilin nor a pseudopilin gene was encoded by the oligonucleotide-binding region (not shown; J. Plessl and B. Reinhold-Hurek, unpublished results).
An additional small protein PilB (14.2 kDa) was encoded downstream of pilA. The presence of a second protein that was not fused to PilA was confirmed by two-dimensional gel electrophoresis. A protein that was missing in the protein extract of mutant BH1599 in comparison with the wild type (Fig. 4) was identified by its N-terminal amino acid sequence. The analysed sequence, VQLTTSGQVR, was identical to amino acids 20–29 of the deduced sequence of PilB (Fig. 1B). This would be expected to be the N-terminus of mature PilB after cleavage of a signal peptide typical of secreted proteins. This was in accordance with the observation that the first 19 amino acids of the deduced PilB sequence fulfilled the criteria proposed by von Heijne (1988) for secretory signal peptides.
PilB showed some similarity to FimF of E. coli (Fig. 2B), which is a minor component of type I fimbriae. For cellular localization of PilB, polyclonal antibodies were raised against a C-terminal peptide of PilB, the preimmmune serum showing no cross-reaction (not shown). The antibodies specifically detected a protein with a slightly different electrophoretic mobility (21 kDa) than the calculated molecular mass, which might result from tertiary structure. PilB was found in cell extracts of E. coli expressing pilAB and of wild-type Azoarcus sp. BH72 (Fig. 5A and B), but not in extracts of mutant BH1599 (Fig. 5A and B). When pilA or pilB were complemented in the transposon mutant (see below), PilB was not detected in extracts of BH1599pilA (Fig. 5A). In a complementation mutant harbouring intact pilB but an in frame deletion of pilA, the PilB protein was still present in protein extracts and of the same size as in wild-type cells (Fig. 5A, lanes 3 and 2). If a hypothetical PilAB fusion protein existed, the in frame deletion would result in a reduction of protein size of 5.5 kDa, which should be clearly visible at an electrophoretic mobility of 20 kDa. Besides the sequencing data, this confirms that pilA and pilB are distinct genes encoding different proteins in Azoarcus sp. BH72.
A higher molecular mass form of PilB (60 kDa), possibly indicating protein complexes, was present in a crude fraction of proteins of cell appendages of wild-type Azoarcus (Fig. 5B), but not in extracts heated with reducing agents (Fig. 5B) or in external fractions of mutant BH1599 (not shown). In the cell appendage fraction of wild-type cells, PilB appeared to be enriched in comparison with the fraction of total cellular proteins (Fig. 5B). Immunofluorescence studies on whole cells showed specific binding of PilB antiserum to microcolonies of strain BH72 (Fig. 5C–E). PilB appeared to be extruded beyond the cell surface. Typically, not the entire cells were stained, but localized regions adjacent to bacterial cells or between bacterial cells (Fig. 5F).
Similar experiments were not carried out with antiserum to PilA, because an attempt to raise polyclonal antibodies against a C-terminal peptide of PilA did not result in specific cross-reactions with cell extracts.
Pilus formation on solid media is dependent on pilA and pilB
Wild-type Azoarcus cells cultivated on agar surfaces were inspected for the presence of pili by transmission electron microscopy. In addition to bacterial flagella, thinner filaments of the typical diameter of pili (approximately 4.5–6 nm) were visible (Fig. 2C). However, strain BH72 was only sparsely piliated when grown on solid medium, mostly showing only one pilus per cell. Moreover, the pili were often broken and thus appeared to be very fragile. Pili were not detectable in colonies of mutant BH1599 when more than 200 cells were inspected. To investigate whether both genes (pilAB ) are involved in pilus formation, complementation mutants of strain BH1599 were constructed by stable integration of plasmids into the chromosome via homologous recombination. Either the intact operon under the control of its own promoter was integrated (strain BH1599pilAB ) or the integrated plasmid carried intact pilA and a disrupted pilB (strain BH1599pilA) or an in frame deletion of pilA and intact pilB (strain BH1599pilB ) (Fig. 1A, Table 1[link]). Electron microscopic inspection revealed that pilus formation was restored by complementation with both genes, pilAB. The wild-type piliation level (in 16% ± 10% of cells piliation visible) was achieved in mutant BH1599pilAB (20% ± 6%) (Fig. 2D). No pili were detected after complementation with pilA alone when 150 cells of BH1599pilA were inspected. Similarly, no cell-bound pili were detected after complementation with pilB (150 cells of BH1599pilB ); however, fragments of pili were rarely detected on the grid (Fig. 2E), demonstrating that pili were formed, albeit at a very low rate. Pilus fragments not associated with cells were not detected on grids of mutants BH1599 or BH1599pilA. Apparently, PilA being truncated by 7 terminal amino acids in mutant BH1599 was sufficient for an (inefficient) pilus formation, when PilB was present. This also confirms the contribution of two distinct genes, pilA and pilB, to pilus formation, as they were transcribed from two physically separated fragments on the chromosome in this experiment.
Table 1. . Bacterial strains, cloning vectors and recombinant plasmids used. a. Restriction site in polylinker of cloning vector.b. Restriction site of Tn5-gusA2.
A type IV pilin is involved in bacterial adhesion to the fungal mycelium and nitrogen fixation in co-culture with the rhizospheric asexual ascomycete
In co-culture with the rhizosphere fungus 2003 related to the Euascomycetes, most cells of Azoarcus sp. strain BH72 adhere to the fungal mycelium, where they form intracytoplasmic membrane stacks related to efficient nitrogen fixation (Hurek et al., 1995). During the first 2 days of incubation, the turbidity of the culture medium decreases because of adhesion of the bacterial inoculum to the mycelium, where bacterial microcolonies can be detected by using a fluorescent vitality stain. As fungal spores obscure turbidity measurements, bacterial numbers in the supernatant were determined by direct microscopic counts. Typical examples for the decrease in bacterial numbers are shown in 6Fig. 6A. The initial attachment within 2 days was significantly inhibited by disruption of either pilA or pilB or both, whereas the attachment of the transposon mutant was rescued by pilAB. Cell numbers present in the supernatant (in percentage of the initial inoculum as 100%) were 13.0 ± 5.4% (a) for wild-type strain BH72, 82.9 ± 8.9% (b) for transposon mutant BH1599, 10.1 ± 2.1% (a) for complementation mutant BH1599pilAB, 77.4 ± 4.9% (b) for complementation mutant BH1599pilA and 78.5 ± 24.2% (b) for complementation mutant BH1599pilB. Results are means ± standard deviations for four different cultures and are followed by different letters when they differ from each other significantly (P < 0.001).
The inactivation of the pilAB genes also affected nitrogen fixation and differentiation in association with the fungal mycelium. 6Figure 6B shows typical examples of the time course of nitrogen fixation in co-cultures. Nitrogen fixation measured by the acetylene reduction assay was significantly decreased in the pilAB mutant. The total amount of ethylene accumulated (μmol) was 17.3 ± 6.6 (a) for BH72, 6.6 ± 2.8 (b) for transposon mutant BH1599 and 16.8 ± 9.1 (a) for the complementation mutant BH1599pilAB. Values with standard deviations represent eight cultures each, and are followed by a different letter when significantly different (P < 0.05). Nitrogen fixation in pure cultures of mutant BH1599 was not affected in comparison with the wild-type Azoarcus. Generation times on N2 in an oxygen-controlled bioreactor (dissolved O2 concentrations 0.4–1.2 μM) on SM medium without combined nitrogen were 2.3 h for both strains.
In parallel, the ability of mutants to differentiate into diazosome-containing cells in association with the fungus was evaluated by transmission electron microscopy of ultrathin sections. Intracytoplasmic membranes were visible in wild-type strain BH72 and in the complementation mutant, but not in the pilAB mutant, where more than 120 cells were inspected.
Type IV pilin is involved in bacterial colonization of plant roots
Points of emergence of lateral roots and root tips are the preferred sites of grass root colonization and invasion for Azoarcus sp. BH72 (Hurek et al., 1994a; Reinhold-Hurek and Hurek, 1998a). Therefore, we determined the capacity of pil mutants to establish microcolonies on roots of young rice seedlings. Mutants differed significantly from wild-type Azoarcus in their colonization capacities. Viable cells of single or double pilAB mutants were rarely detected on the surface of main roots and were rarely found to colonize achsils of lateral roots (Fig. 7A–C). Plant cells of the root epidermis were rarely seen to be infected by mutants, in contrast to wild-type bacteria (Fig. 7D and E). The complementation of pilAB partially restored the wild-type phenotype: the percentage of lateral root achsils colonized (per main root) ± standard deviation was 29.6 ± 33.6 (a) for strain BH72, 1.3 ± 3.8 (b) for mutant BH1599, 13.7 ± 15.0 (c) for complementation mutant BH1599pilAB, and 1.1 ± 2.8 (b) and 1.8 ± 4.1 (b) for complementation mutants BH1599pilA and BH1599pilB respectively. For each strain, 40–55 main roots of five different plants obtained in two independent experiments were inspected, and numbers followed by different letters are significantly different from each other (P < 0.05).
Our results suggest that the role of type IV pili in adhesion and infection is not restricted to animal pathogens, but is also a common trait in plant–microbe and fungus–microbe interactions. In the diazotroph Azoarcus sp. BH72, an unusually short pilin, PilA, is essential for pilus formation and affects the colonization of plant root and fungal surfaces.
With 59 amino acids, PilA is extremely short compared with other pilin precursors of approximately 160 amino acids (Strom and Lory, 1993). Truncated pilins may also occur in N. gonorrhoeae. However, these strains are non-piliated unless they possess an additional, complete gene (Bergström et al., 1986). Therefore, it is highly unusual that such a short pilin may form a functional pilus. As (i) Southern hybridization revealed no additional copies of identical genes in the Azoarcus genome and (ii) pilA mutants lack pili, this ‘truncated’ pilin appears to be essential for pilus formation in Azoarcus. It has been recognized recently that pilins may also play a role in protein excretion as a component of the secretion apparatus for exoenzymes, as proposed for P. aeruginosa (Lu et al., 1997). This may be the case also for PilA, as we observed a decreased cellulase excretion in the Azoarcus pilA mutant (unpublished data).
Pilus formation can also be influenced by mutations in pilus assembly proteins. A complex machinery depending on at least 30 genes is required for the biogenesis and function of type IV pili (Alm and Mattick, 1997). Pilus biogenesis and extracellular protein secretion appear to be evolutionarily related processes. This is reflected in the sequence similarities of some proteins involved in pilus assembly (Russell and Darzins, 1994; Alm and Mattick, 1996; Alm et al., 1996) and in the type II secretion pathway (Bally et al., 1992; Bleves et al., 1998). Several of these proteins share homology with the highly conserved N-terminal region of the type IV prepilin, including the short leader sequence, which is processed by the signal peptidase PilD. The protein encoded by pilA of Azoarcus sp. BH72 contains the highly conserved G-1F + 1/E + 5 motif that is required for cleavage of the 6- to 7-amino-acid leader sequence and the subsequent methylation of the mature pilin (Strom et al., 1993). In a stretch of 25 amino acids, pilA showed 100% identity to the conserved region of prepilin of P. aeruginosa and only limited homology to pilus assembly proteins (PilE, PilW, PilV, FimT and FimU) or secretory apparatus components (XcpT, XcpU, XcpV and XcpW). Also, most of these proteins are in the range of 15–18 kDa (Alm et al., 1996) and therefore not similar to PilA in size. This suggests that PilA is indeed a prepilin and not primarily a part of a pilus assembly or secretion apparatus. Moreover, mutations in proteins of the general secretion pathway are not likely to result in a non-piliated phenotype, as pilA does. If PilA was an unusual type IV pilus assembly protein, this would still indicate that type IV pili are involved in the phenotype we observed.
The inner layer of N. gonorrhoeae pili is a parallel coiled-coil made of α-helices, which are formed by the 54 N-terminal amino acids, hydrophobic interactions giving stability and flexibility to the fibre (Parge et al., 1995; Forest and Tainer, 1997). For the corresponding 53 amino acids of Azoarcus PilA, an α-helix was predicted by different secondary structure prediction algorithms, and they might therefore form a similar core of a fibre. In this case, the C-terminal amino acids might be exposed to the entire surface of the pilus, in contrast to the corresponding amino acids 49–56, which appear to be exposed only at the end of the N. gonorrhoeae pilus fibre (Forest and Tainer, 1997). However, it is not clear whether, in Azoarcus, additional proteins are linked to PilA as a coat of the pilus, especially as the pilus appears to be only slightly smaller in diameter than usual type IV pili. Also, it is not clear how attachment to both plant and fungal surfaces can be mediated, as the Azoarcus pilin is lacking domains that are antigenic determinants and involved in receptor recognition or antigenic variation in pathogens (Seifert, 1996). It can be speculated that, as proposed in Neisseria (Rudel et al., 1995), additional proteins might serve as adhesins in Azoarcus, which could mediate the specificity of binding to different plant or fungal surface components. In this context, it is tempting to speculate on the role of pilB, which is co-transcribed with pilA and essential for pilus formation. PilB and FimF of E. coli both have an N-terminal signal sequence typical of secreted proteins and share some, albeit weak, amino acid similarity throughout the sequence (Fig. 2A). FimF is likely to be a minor component of the mature type I fimbriae and also of the assembly apparatus in a FimC–FimF pilus–chaperone complex (Jones et al., 1995). As PilB appears to be extruded from intact cells, according to immunological analyses, it may be speculated that it might have a similar role in biogenesis or function of pili in Azoarcus sp. BH72, albeit for type IV and not type I pili. As there is no high homology to proteins known to participate in type IV pilus biogenesis and PilB was found to be essential for piliation in Azoarcus sp. BH72, PilB appears to be a novel protein involved in this process. It appears to be sufficient to support pilus production, albeit inefficiently, even when PilA is truncated: in mutant BH1599, in which PilA is shortened by 7 terminal amino acids and pilB is disrupted as a result of transposon insertion, complementation with PilB leads to the production of fragments of pili at a very low rate.
It has been recognized recently that surface appendages play a role in plant–pathogen interactions. Plant tumour induction by Agrobacterium tumefaciens may depend on a pilus structure (Fullner et al., 1996), and a new type of pilus (Hrp pilus) of Pseudomonas syringae is involved in eliciting the hypersensitive response of plants (Roine et al., 1997). However, this is the first report that type IV pili are involved in plant–microbe and fungus–bacteria interactions. In the plant pathogen Xanthomonas campestris, a type IV pilin gene has been detected but has not been observed to have an influence on virulence (Ojanen-Reuhs et al., 1997). In contrast, the Azoarcus sp. BH72 type IV pilin appears to be crucial for the establishment of microcolonies on grass roots. As colonization of roots by bacteria is a common but poorly understood phenomenon, the detection of genes involved in attachment to roots is an important step towards elucidating this process. As these genes also mediated bacterial adhesion to the mycelium of a rhizosphere fungus, type IV pili appear to have a more general role in the interaction of Azoarcus with eukaryotes of its habitat. Non-piliated mutants failed to form diazosomes in co-culture with the fungus, suggesting that spatial proximity to the mycelium is a prerequisite for the formation of these membranes related to efficient nitrogen fixation. One possible reason for this might be the lower oxygen concentrations likely to prevail in niches of the actively respiring mycelium.
Our findings suggest that there are common traits in animal–pathogen and other eukaryote–microbe interactions. Comparative studies of both processes might give important insights into the evolution of pilus-mediated interactions.
Rice seeds (Oryza sativa IR36) were surface sterilized and inoculated as described previously (Hurek et al., 1994a). However, plants were grown in a 16 h day/8 h night cycle in a greenhouse, and the following plant medium was used (per litre): DL-malic acid, 0.2 g, titrated with KOH to pH 6.8; L-proline, 0.58 g; KH2PO4, 1.8 g; KH2PO4, 1.2 g; ferric citrate, 13 mg; NaCl, 0.1 g; CaCl2, 26.4 mg; MgSO4 × 7H2O, 0.2 g; Na2MoO4 × 2H2O, 4 mg; H3BO3, 3 mg; MnSO4 × H2O, 12 mg, ZnSO4 × 7H2O, 0.2 mg; CuSO4 × 5H2O, 0.1 mg; agar, 4 g. After 3 weeks, rice roots were rinsed with water and stained overnight at 4°C with the LIVE/DEAD BacLight bacterial viability kit (Molecular Probes), and colonization was inspected by fluorescence microscopy (see below).
Construction of plasmids for complementation assays and mutant strains
Complementation with pilAB and with pilA or pilB separately was achieved by transformation of the transposon mutant strain BH1599 with the plasmids pJBLP1, pJBLP11 or pJBLP13, respectively, by electroporation and subsequent integration of the plasmids into the chromosome by homologous recombination. Plasmids were constructed as follows. The insert of pJBLP1 (Fig. 1A) containing pilAB originates from pJBL1, which was obtained from a λ-ZAP-Express library from partial SauIIIAI digests of chromosomal DNA of strain BH72. The pilB gene of pJBLP1 was inactivated by cloning a Km-resistance cartridge (Pharmacia) into the MfeI restriction site of pJBLP1, resulting in pJBLP11. In frame deletion of pilA was achieved by polymerase chain reaction (PCR) using a primer spanning only the first and the last 12 bp of pilA (nucleotides 161–201 and 355–402; Fig. 1B). The PCR product containing the Eco 47III restriction site downstream of the pilA deletion was cloned, the correct deletion controlled by sequencing, and the part of pJBLP1 located downstream of the Eco 47III site was cloned behind the pilA deletion, resulting in pJBLP13 (Fig. 1A).
Rabbit polyclonal antibodies to PilA or PilB were raised against custom-synthesized peptides (amino acids 29–54 of PilA or amino acids 128–142 of PilB in Fig. 1B) (Eurogentec). Proteins were separated by SDS–PAGE, electroeluted on nitrocellulose and detected on Western blots as outlined previously (Hurek et al., 1994a) or by using a chemiluminescence reaction (ECL Western blotting detection reagents; Amersham). To obtain proteins of cell appendages, cells were passed five times through a hypodermic needle, and proteins of the supernatant were precipitated with trichloroacetic acid.
For fluorescence microscopy, bacteria grown on agar plates were transferred to glass slides silanized with 0.185% Bind-Silane (Pharmacia) and 0.185% acetic acid in ethanol. Bacteria were incubated with antibody solutions diluted as follows: anti-PilB, 1:500; swine anti-rabbit fluorescein isothiocyanate (FITC)-conjugated antibody (Dako), 1:30. Before and after antibody incubations, bacteria were washed with PBS.
To visualize fluorescence, bacteria and plant samples were inspected as described previously (Egener et al., 1998). Micrographs were recorded by a 3 Chip RGB colour camera (Hamamatsu Photonic Systems). To reduce haze, digital confocal imaging was achieved using Improvision Openlab software (Egener et al., 1998).
For transmission electron microscopy, bacteria were grown on agar plates, resuspended in PBS, transferred to charged, Formvar-/carbon-coated copper grids, washed with PBS, fixed with 0.5% glutaraldehyde and negatively stained with 2% neutralized phosphotungstic acid. To visualize intracytoplasmic membranes, cells were embedded in resin, sectioned and stained according to Hurek et al. (1994a).
The GraphPad InStat software package was used for statistical analysis by ANOVA followed by the Tukey–Kramer test.
We thank E. Mörschel and B. Kastner for the use of the electron microscope, I. B. Zhulin for advice in computational analysis, C. Staubitz (MPI) for help with photographic reproductions, T. Karg (MPI) for two-dimensional gel electrophoresis, and J. Plessl (MPI) for cloning and sequence analysis of a putative pseudopilin gene. This work was supported by grants from the Deutsche Forschungsgemeinschaft (DFG) (Re 756/5-1 and SFB395) to B.R.-H.