|Home | About | Journals | Submit | Contact Us | Français|
Legionella pneumophila is able to survive inside phagocytic cells by an internalization route that bypasses fusion of the nascent phagosome with the endocytic pathway to allow formation of a replicative phagosome. The dot/icm genes, a major virulence system of L. pneumophila, encode a type IVB secretion system that is required for intracellular growth. One Dot protein, DotL, has sequence similarity to type IV secretion system coupling proteins (T4CPs). In other systems, coupling proteins are not required for viability of the organism. Here we report the first example of a strain, L. pneumophila Lp02, in which a putative T4CP is essential for viability of the organism on bacteriological media. This result is particularly surprising since the majority of the dot/icm genes in Lp02 are dispensable for growth outside of a host cell, a condition that does not require a functional Dot/Icm secretion complex. We were able to isolate suppressors of the ΔdotL lethality and found that many contained mutations in other components of the Dot/Icm secretion system. A systematic analysis of dot/icm deletion mutants revealed that the majority of them (20 of 26) suppressed the lethality phenotype, indicating a partially assembled secretion system may be the source of ΔdotL toxicity in the wild-type strain. These results are consistent with a model in which the DotL protein plays a role in regulating the activity of the L. pneumophila type IV secretion apparatus.
The gram-negative bacterium Legionella pneumophila is the causative agent of a potentially fatal form of pneumonia called Legionnaires' disease. L. pneumophila is found in freshwater environments, where it parasitizes many different species of protozoa (17). Humans become infected with L. pneumophila by inhaling aerosols generated from contaminated water sources. Upon entry into the human lung, L. pneumophila is internalized into bactericidal, alveolar macrophages. In contrast to phagosomes bearing most bacterial species, the compartment harboring L. pneumophila does not traffic into the lysosomal network and is not significantly acidified in the first few hours after uptake (26, 27). Instead, the phagosome interacts with early secretory vesicles at endoplasmic reticulum exit sites (29) and then undergoes a series of maturation events in which it sequentially associates with small vesicles, mitochondria, and eventually becomes surrounded by the rough endoplasmic reticulum (25, 60). Formation of this specialized compartment, called a “replicative phagosome,” allows the microorganism to grow intracellularly (25, 28). Later in the infective cycle, a majority of the replicative phagosomes fuse with acidified compartments containing late endocytic markers, and this is believed to play an important role in the replicative cycle of this pathogen prior to exit from its host cell (59).
The key to L. pneumophila's virulence is its ability to form a replicative phagosome, since mutants defective in this trait cannot replicate inside host cells and are thus unable to cause disease (24, 26). One large class of proteins that allow L. pneumophila to alter the endocytic pathway is encoded by the dot/icm genes (3, 5, 37). To date, over two dozen dot/icm genes have been identified and are clustered in two areas of the L. pneumophila chromosome (region I and region II) (63). Based on the similarity of the Dot/Icm proteins to proteins involved in conjugative DNA transfer, and the fact that the Dot/Icm system can transfer the mobilizeable plasmid RSF1010, it was proposed that the dot/icm genes of L. pneumophila encode a type IV secretion system (31, 50, 63).
Type IV secretion systems are able to export DNA and/or proteins out of the bacterial cell and include plasmid transfer systems (e.g., the tra and trb genes of the plasmid RP4), as well as systems involved in the delivery of virulence factors (10, 46, 66). The canonical type IV secretion system is encoded by the virB operon of the plant pathogen Agrobacterium tumefaciens (66). A number of other pathogens, including Bartonella tribocorum, Bordetella pertussis, Brucella abortus, Helicobacter pylori, and Rickettsia prowazekii, contain orthologues to the VirB proteins, and some of these systems have been shown to export proteins essential for virulence (10). In contrast to these type IV systems, the L. pneumophila Dot/Icm proteins have limited sequence similarity to the VirB proteins. Instead, the Dot/Icm proteins show high similarity to the transfer proteins from IncI plasmids (e.g., R64 and ColIb-P9) and compose a type IVB secretion system (31, 57).
As with most conjugative transfer systems, little is known about the specific function of many of the L. pneumophila Dot/Icm proteins. DotB was recently shown to possess ATPase activity and likely provides energy to the secretion apparatus (56). A second Dot protein, DotL, also contains a nucleotide binding motif and shows extensive sequence similarity to the conjugal transfer protein TrbC from IncI plasmids (19, 31). DotL also has detectable sequence similarity to a family of proposed ATPases known as TraG-like or type IV secretion system coupling proteins (T4CPs). The more notable members of the T4CP family include TraG (RP4 plasmid), TrwB (R388 plasmid), TraD (F plasmid), and the A. tumefaciens VirD4 protein (8, 18, 33).
The term “coupling protein” was proposed for this family because its members are believed to target, or couple, exported substrates to the secretion apparatus (8, 9, 15, 22, 23, 32, 61). This proposal was initially based on the phenotype of RP4 traG mutants, which were still able to process plasmid DNA into a secretion-competent intermediate and assemble a functional pilus but were unable to transfer the plasmid. This indicated that TraG plays a role in linking the two processes (9). Consistent with the idea of T4CPs linking substrates to the secretion apparatus, a number of T4CPs have been shown to interact with both exported substrates and with components of the secretion apparatus (2, 15, 20, 35, 61). Although T4CPs are absolutely required for export of substrates, their specific molecular function remains unknown (22).
We demonstrate here that a T4CP homologue, the DotL protein, is not only required for growth of L. pneumophila inside macrophages but is also essential for viability of certain strains on bacteriological media. The lethality caused by loss of dotL in those strains can be suppressed by mutations that inactivate the Dot/Icm complex, which is consistent with a DotL role in regulating the activity of this type IV secreton.
All L. pneumophila strains used in the present study are derived from Lp02 (hsdR rpsL thyA) or JR32 (hsdR rpsL), two separate isolates of L. pneumophila Philadelphia-1 (3, 7, 38) (Table (Table1).1). L. pneumophila strains were cultured on N-(2-acetamido)-2-aminoethanesulfonic acid (ACES)-buffered charcoal yeast extract agar (CYET) or ACES-buffered yeast extract broth (AYET) supplemented with thymidine (100 μg/ml). Salt sensitivity was assayed on CYET plates containing 0.65% sodium chloride (11, 45, 64). Antibiotics (kanamycin, 20 μg/ml; chloramphenicol, 5 μg/ml; streptomycin, 50 μg/ml; gentamicin, 5 μg/ml) and sucrose (5%) were added as needed. Escherichia coli strains were cultured on Luria-Bertani medium, and antibiotics (kanamycin, 20 μg/ml; ampicillin, 100 μg/ml; chloramphenicol, 17 μg/ml) were added as needed. Replication-competent plasmids were propagated in the E. coli strain XL1-Blue. In order to propagate suicide plasmids containing the R6K origin of replication, a strain expressing the R6K π protein, E. coli strain DH5α(λpir), was used (30, 67).
To make the ΔdotL suicide plasmid, pJB1001, two PCR-amplified fragments were cloned into the NotI/SalI sites of pSR47S (40). Fragment 1 was amplified by using the primers 5′-CCCAAACGGCCGCCAAACGAGTATTTACCATGC(JVP201 with the EagI site underlined) and 5′-CCCAAAGGATCCCGCATCATGGCTCTAATTCC(JVP202 with the BamHI site underlined). Fragment 2 was amplified by using the primers 5′-CCCAAAGGATCCGCTATTGGGCATGAAGAGAGC(JVP203 with the BamHI site underlined) and 5′-CCCAAAGTCGACCCTACTGATGCAACTTTAATCC(JVP204 with the SalI site underlined). Plasmid pJB1005 was constructed by inserting a gene encoding chloramphenicol acetyltransferase that was amplified from pKRP10 by using the primers 5′-CCCAAAGGATCCGAGGTTCCAACTTTCACC(JVP206 with the BamHI site underlined) and 5′-CCCAAAGGATCCCTGCCTTAAAAAAATTACGC(JVP207 with the BamHI site underlined) into the BamHI site of plasmid pJB1001.
To make the ΔdotN suicide plasmid, pJB3046, two PCR-amplified fragments were cloned into the NotI/SalI sites of pSR47S (40). Fragment 1 was amplified by using the primers 5′-CCCGCGGCCGCGGTGTATCGTTAGGTAAAATGG(JVP289 with the NotI site underlined) and 5′-CCCGGATCCCGCCATAGTTTGGTTCACATTCAGTC(JVP903 with the BamHI site underlined). Fragment 2 was amplified by using the primers 5′-CCCGGATCCGAGAAATGGGCTGCCAGTGC(JVP904 with the BamHI site underlined) and 5′-CCCGTCGACGCAGCTTTTAACTGATCGC(JVP286 with the SalI site underlined).
To make the ΔdotM suicide plasmid, pJB3050, two PCR-amplified fragments were cloned into the NotI/SalI sites of pSR47S (40). Fragment 1 was amplified by using the primers 5′-CCCGCGGCCGCGAAGCAATCTTCAGTCCTGG(JVP297 with the NotI site underlined) and 5′-CCCGGATCCCTGCTGTTGTTGTGCCATCTC(JVP901 with the BamHI site underlined). Fragment 2 was amplified by using the primers 5′-CCCGGATCCGATGAAGCGATTAGAGCTCTGG(JVP902 with the BamHI site underlined) and 5′-CCCGTCGACGCATACAGAGAGTTATCTCC(JVP294 with the SalI site underlined).
pJB1010, the His-tagged version of DotL, was constructed by amplifying the dotL open reading frame (ORF) using plasmid pJB359 and the primers 5′-GACATGCATGCGATGGGGTTGACTAATTAAGG (JVP217 with the SphI site underlined) and 5′-GACATGCATGCCCCGAAAGCAAAAGTTGCC(JVP218 with the SphI site underlined). The PCR product was digested with SphI and cloned into the SphI site of pQE-32 (Qiagen). The final construct can be used to express a fusion protein containing six histidines fused to amino acids 72 through 783 of DotL.
The dotL complementing clone, pJB1014, was constructed by first amplifying the dotL ORF from Lp02 chromosomal DNA by using the primers 5′-GGGGTACCGGAATTAGAGCCATGATGCG(JVP227 with the KpnI site underlined) and 5′-GACATGCATGCGATGGGGTTGACTAATTAAGG(JVP217 with the SphI site underlined). The resulting product was digested with KpnI and SphI and ligated into KpnI/SphI-digested pJB908. pJB908, a derivative of the plasmid pKB5, has the following features: (i) an RSF1010 origin to permit replication in L. pneumophila, (ii) an ΔoriT mutation to prevent inhibition of growth in macrophages, and (iii) a tac promoter driving DotL expression (3). Constitutive expression from pJB1014 is able to rescue a dotL deletion strain for viability on plates and in macrophages and expresses similar levels of DotL compared to a wild-type strain.
pJB1242, the ΔlvhB suicide plasmid, was constructed by cloning two PCR-amplified fragments into the SalI and NotI sites of pSR47S. Fragment 1 was amplified by using the primers 5′-CCCGTCGACGTTTGGAGAAGTCAGTTTAAGG(JVP342 with the SalI site underlined) and 5′-CCCGGATCCTCATGGCGCCACCTTTTGC(JVP343 with the BamHI site underlined). Fragment 2 was amplified with the primers 5′-CCCGGATCCGAAGCACTCGAACTATAAACC(JVP344 with the BamHI site underlined) and 5′-CCCGCGGCCGCGTTTCGCCATTGTATCCC(JVP345 with the NotI site underlined).
pJB1304, containing the lvhB operon, was constructed by first amplifying the lvhB operon from JR32 chromosomal DNA by using the primers 5′-CCCGTCGACGTTTGGAGAAGTCAGTTTAAGG(JVP342 with the SalI site underlined) and 5′-CCCGCGGCCGCGTTTCGCCATTGTATCCC(JVP345 with the NotI site underlined). The resulting product was digested with SalI and NotI and ligated into SalI/NotI-digested pJB1300. pJB1300, a derivative of the plasmid pKB5 (3), has the HindIII site in the polylinker replaced with a unique NotI site.
pJB1010, a polyhistidine-tagged version of DotL in which the amino-terminal signal sequence of DotL was replaced with six histidines, was purified by using Ni-nitrilotriacetic acid chromatography (Qiagen). The purified His6-DotL fusion protein was injected into rabbits to raise polyclonal antibodies against DotL (Cocalico). The serum recognized a single protein from wild-type L. pneumophila extracts that was absent in extracts from an E. coli strain and a L. pneumophila strain lacking the dotL gene.
L. pneumophila was fractionated as previously described (55). Briefly, a culture of Lp02 was grown to mid-exponential phase, and the cells were pelleted and resuspended in 50 mM Tris-HCl (pH 8.0), 0.5 M sucrose, 5 mM EDTA, and 0.1 mg of lysozyme/ml. The cell suspension was incubated on ice for 1 h, MgSO4 was added to a final concentration of 20 mM, and spheroplasts were collected by centrifugation at 5,000 × g. The pellet was resuspended in 50 mM Tris-HCl (pH 8.0), sonicated, and then centrifuged at 5,000 × g to collect any unlysed cells. The supernatant was then centrifuged at 100,000 × g for 1 h at 4°C to obtain a total membrane fraction. The supernatant was removed, centrifuged at 100,000 × g, and saved as the cytoplasmic sample. The pellet was washed and resuspended in 50 mM Tris-HCl (pH 8.0). The inner membranes were solubilized by the addition of Triton X-100, and the outer membranes were collected by centrifugation at 100,000 × g. Fractions were resuspended in sodium dodecyl sulfate-polyacrylamide gel electrophoresis loading buffer, subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and either Coomassie blue stained for total protein or transferred to a membrane and probed with the anti-DotL serum (1:5,000).
The histiocytic cell line U937 (American Type Culture Collection) was cultured in RPMI 1640 media (BioWhittaker) containing 10% fetal bovine serum (BioWhittaker). Cells were differentiated with 12-O-tetradecanoylphorbol-13-acetate (TPA; Sigma) as described previously (3). Differentiated U937 cells were plated as a confluent monolayer in 24-well plates, with each well containing ca. 2 × 106 cells per well.
L. pneumophila chromosomal DNA was isolated by a combination of a high-salt precipitation to eliminate contaminating proteins, followed by isopropanol precipitation of the DNA. Chromosomal DNA was digested with 10 U of HaeII restriction enzyme overnight at 37°C. Southern blots were performed according to the ECL Southern hybridization kit (Amersham), with probes specific to regions flanking dotL (from pJB1001) or dotB (pJB921).
L. pneumophila was mutagenized by using the transposon delivery system encoded on pJK211-2 (13). pJK211-2 contains a temperature-sensitive origin that is not permissive for replication at 37°C, an altered sites transposase that increases the randomness of insertion, and a mini-Tn10 transposon containing a kanamycin cassette (KanR) and a conditional origin from plasmid R6K later used to recover the transposon insertions in E. coli strain DH5α(λpir). A pool of insertions was placed on sucrose chloramphenicol plates to select for recombinants (sucrose to select for recombinants and chloramphenicol to select for the ΔdotL::Cmr). Chloramphenicol-resistant (Cmr), kanamycin-sensitive, sucrose-resistant (Sucr) colonies were colony purified and scored for loss of the plasmid-encoded resistance cassette to ensure they had resolved the integrated plasmid. Insertions were recovered as previously described (13). The site of insertion was identified by sequencing by using the primers JVP348 (GGATCTGGTACCGGATCC) or JVP349 (TCAACAGGTTGAACTGCGGATC).
Plasmids pJB1001 and pJB1005 were transferred into L. pneumophila strains by using an RP4 conjugation system encoded on pRK600 (14). L. pneumophila strains containing the integrated plasmid were selected by plating on CYET containing kanamycin and streptomycin. Resulting merodiploid strains that had a second crossover event were selected by plating on CYET plates containing 5% sucrose. Resolution of the integrated plasmid was confirmed by loss of kanamycin resistance. In the case of strains containing the ΔdotL::Cmr cassette, sucrose-resistant colonies were streaked onto chloramphenicol to screen for the wild-type or mutant dotL alleles.
L. pneumophila strains were resuspended in phosphate-buffered saline to an optical density at 600 nm of 1. The bacterial suspensions were then diluted 1:1,000 in RPMI 1640 containing 10% fetal bovine serum, and 2 mM glutamine. A monolayer of TPA-treated U937 cells were infected with various L. pneumophila strains at a multiplicity of infection of one for 1 h. The monolayers were washed with fresh RPMI and then incubated in RPMI 1640 containing 10% fetal bovine serum and 2 mM glutamine at 37°C and 5% CO2. Thymidine was added when appropriate. At 1, 24, 48, and 72 h postinfection, cells were lysed in sterile ddH2O and dilutions were plated on CYET. Plates were incubated for 4 days at 37°C, and viable counts were determined.
GenBank accession numbers for submitted sequences are as follows: DotU is AF533658 and DotV is AF533657.
The L. pneumophila DotL protein has extensive similarity to several proteins found in GenBank (Fig. (Fig.1A),1A), with the highest degree of similarity (56% identity) to an uncharacterized ORF found in Coxiella burnetii that has been proposed to be part of a type IV secretion system (57). DotL also has similarity to TrbC (27% identity), a protein required for the transfer of the IncI plasmids R64 and ColIb-P9 (Fig. (Fig.1,1, top) (19, 31), and to an ORF on a plasmid found in Pseudomonas syringae strains that may be part of the conjugative transfer apparatus for this plasmid (57). DotL also has sequence similarity, extending primarily over the Walker A box, to members of the type IV coupling protein family, most notably TraD, TraG, TrwB, and VirD4 (Fig. (Fig.1A).1A). In addition, DotL shares a number of characteristics with members of the T4CP family. These include a predicted size of 86 kDa, the presence of a potential nucleotide binding motif (a Walker A box), and an amino-terminal hydrophobic sequence that would likely target the protein to the bacterial inner membrane (52, 65). These characteristics, combined with its homology, suggest DotL may be a T4CP.
To confirm the subcellular localization of DotL, L. pneumophila extracts were prepared, fractionated, and the protein was detected by Western analysis with a DotL specific antibody. The DotL protein was primarily localized to the membrane fraction (Fig. (Fig.1B,1B, lane 3). Moreover, the majority of the protein was Triton X-100 soluble, indicating it was likely to be in the inner membrane of the bacterial cell (Fig. (Fig.1B,1B, lane 4) (48). In addition, a smaller cross-reacting species of ca. 75 kDa could be detected that localized completely to the cytoplasmic fraction (Fig. (Fig.1B,1B, lane 2) and is consistent with a DotL breakdown product lacking the hydrophobic amino-terminal transmembrane domains.
To investigate the function of the DotL protein, we attempted to delete the dotL gene from the chromosome of Lp02, a strain of L. pneumophila with an intact dot/icm system (3, 63). Previous attempts to delete dot/icm genes have been uniformly successful, indicating that the Dot/Icm complex is not required for viability on bacteriological media (1, 3, 63). To construct an in-frame deletion of the dotL gene, ca. 500 bp of DNA upstream and downstream adjacent to the dotL gene was cloned into the suicide vector pSR47S, generating plasmid pJB1001 (Fig. (Fig.2).2). The dotL deletion plasmid was electroporated into strain Lp02 and introduced onto the chromosome by selecting for a single crossover event generating a dotL/ΔdotL merodiploid strain. Merodiploids that had resolved were selected by plating on sucrose, a toxic compound for gram-negative organisms containing the counterselectable marker sacB (4). Resolution of the merodiploid should result in an equal proportion of strains containing either the wild-type copy of dotL or ΔdotL on the chromosome (Fig. (Fig.22).
Examination of 14 independent sucrose resistant recombinants derived from the dotL/ΔdotL merodiploid strain revealed no strains that lacked the wild-type copy of dotL (Fig. (Fig.3,3, top panel). In contrast, sucrose resistant recombinants derived from a similarly constructed dotB/ΔdotB merodiploid (55) re sulted in ten strains containing wild-type dotB and four strains containing ΔdotB (Fig. (Fig.3,3, bottom panel). To ensure that the recombination event in the dotL/ΔdotL merodiploid strain was not theoretically impossible, recombinants were selected in a merodiploid containing the dotL+ plasmid pJB1014. In this situation, chromosomal dotL deletions were recovered, indicating that a ΔdotL could be obtained in a strain exogenously expressing DotL (data shown below). Finally, the ΔdotL::Cmr reporter plasmid pJB1005 could not be introduced directly onto the chromosome of the L. pneumophila strain Lp02 by using natural transformation (56), confirming the difficulty of constructing a dotL deletion. These results indicated a strong bias against deleting the wild-type version of dotL and suggested that loss of dotL may result in lethality of L. pneumophila on bacteriological media.
Although it appeared not to be feasible to isolate a strain lacking dotL, it was possible that an insufficient number of events were examined in order to identify such a strain. To screen a larger number of recombination events, the deletion strategy was repeated with a chloramphenicol-marked version of the dotL deletion. A dotL/ΔdotL::Cmr merodiploid was subjected to selection on sucrose, in the absence of chloramphenicol, and the presence of the ΔdotL::Cmr cassette was subsequently screened by plating sucrose resistant recombinants on medium containing chloramphenicol. Examination of a larger number of sucrose-resistant strains still failed to detect a recombinant that contained just the ΔdotL::Cmr allele (0 of 753 events scored). Based on these results, we conclude that dotL is required for the viability of the L. pneumophila strain Lp02 on bacteriological media.
In order to determine whether it was possible to suppress the lethality caused by loss of dotL, we plated an even greater number of the dotL/ΔdotL::Cmr merodiploid on plates containing sucrose and chloramphenicol, thereby directly selecting for loss of dotL. Rare sucrose-resistant, chloramphenicol-resistant recombinants were isolated at a rate of ~10−6. This was consistent with dotL being an essential gene, with the chloramphenicol-resistant colonies that arose being pseudorevertants due to spontaneous mutations in other genes. To identify the nature of the pseudorevertants, random transposon insertions were generated in the dotL/ΔdotL::Cmr merodiploid strain background by using a mini-Tn10 transposon, and the insertion pool was plated on sucrose and chloramphenicol to select for strains that could tolerate loss of dotL. Thirty-three such insertions were isolated from independent pools. These strains were first analyzed by Southern blot to ensure that they had only one insertion. To confirm that the phenotype was linked to the transposon insertion, the strains were recreated by transforming the transposon and flanking chromosomal DNA into the original, unmutagenized merodiploid strain by using natural transformation (56). Examination of the 33 strains by this assay demonstrated that, in each case, the phenotype was linked to the transposon insertion. Finally, the transposons and flanking DNA were recovered on a plasmid, and the sites of the transposon insertions on the L. pneumophila chromosome were identified by sequencing off the end of each transposon.
Surprisingly, approximately one-half of the insertions (16 of 33) were in other dot/icm genes. This included four insertions in dotA, two in dotG, one in dotI, five in dotO, three in icmF, and one in icmX (Fig. (Fig.4).4). In most cases, the phenotype appeared to be due to inactivation of the gene the transposon was inserted in, because the insertions were in terminal genes of proposed operons (e.g., dotA, dotO, icmF, and icmX). Among the insertions that were not in known dot/icm genes, three mutants (JV1308, JV1343, and JV1499) were defective for intracellular growth of L. pneumophila when the insertions were separated from the ΔdotL (data not shown). The three mutants each contained an insertion in a different site of the same gene, which is located ca. 20 kb from region II (Fig. (Fig.4)4) (63). This gene codes for a small protein of 180 amino acids that has extensive homology to DotE (40% amino acid identity over 171 amino acids). We have designated this gene dotV because it is required for proper targeting of the L. pneumophila phagosome and for intracellular growth (unpublished results) (accession no. AF533657). Finally, the remaining insertions were not in known dot/icm genes or homologous genes and, when separated from the dotL deletion, caused the corresponding strains to exhibit various degrees of growth inhibition inside host cells (data not shown).
To confirm that loss of a specific dot gene could suppress a ΔdotL, we attempted to delete dotL in a strain containing an in-frame deletion of the dotA gene. In contrast to the previous attempt to delete dotL (Fig. (Fig.3),3), both dotL and ΔdotL loopouts were obtained from the ΔdotA dotL/ΔdotL merodiploid, demonstrating that loss of a single dot gene could allow the isolation of the ΔdotL mutation (Fig. (Fig.5).5). Because the ΔdotL suppressor hunt identified only a subset of dot/icm genes, we investigated whether they were the only dot/icm genes that, when inactivated, could suppress the ΔdotL lethality. In-frame deletions were constructed in 23 of the 26 dot/icm genes, and the ΔdotL suicide plasmid was integrated into each strain to assay for the ability to tolerate loss of dotL. Remarkably, dotL could be deleted in almost all of the strains containing different dot/icm mutations (Fig. (Fig.6).6). This suppression was specific in that dotL could not be deleted in a strain lacking a housekeeping gene found in region II, citA, which is not required for intracellular growth (Fig. (Fig.6)6) (42).
In contrast, inactivation of three dot/icm genes, dotK, icmS, and icmW, did not suppress loss of dotL (Fig. (Fig.6).6). dotK encodes an outer membrane protein with homology to OmpA and is only partially required for growth in amoebae (53). icmS and icmW encode two small, acidic, cytoplasmic proteins that have been proposed to function as secretion chaperones (12). Similar to dotK, icmS and icmW are not absolutely required for the growth of L. pneumophila in permissive cell lines such as U937s and HL60s (12, 53). Therefore, it was possible that inactivation of these genes failed to suppress loss of dotL simply because they are not absolutely required for intracellular growth. However, inactivation of three other dot/icm genes—icmF, dotU, and icmR—did suppress loss of dotL (Fig. (Fig.6),6), even though loss of these genes caused only a partial inhibition of growth in permissive hosts (12, 53, 54, 62, 68). These results indicate that, although inactivation of the majority of the dot/icm genes can suppress the lethality caused by loss of dotL, there is specificity to the suppression.
dotL, also known as icmO, is essential for viability in the Lp02 background. However, it has been previously published that loss of dotL in another L. pneumophila strain, JR32, is not a lethal event (52). Lp02 and JR32 were independently derived from L. pneumophila Philadelphia-1, an organism isolated from the original Legionnaires' disease outbreak in 1976 (3, 38). Each strain was individually selected to be streptomycin resistant and to lack host restriction, and Lp02 was then also selected to be a thymidine auxotroph. Due to the relatedness of these two strains, it was surprising that dotL/icmO was essential for viability in Lp02 but was dispensable in JR32. To confirm that dotL was not an essential gene in JR32, a clean ΔdotL was constructed in the JR32 background and was indeed found to be viable on buffered CYE plates (data not shown). To examine intracellular growth, monolayers of U937 macrophages were challenged with wild-type L. pneumophila strains Lp02 and JR32. Both strains were able to multiply >1,000-fold in 3 days (Fig. (Fig.7).7). In contrast, an Lp02 strain lacking a functional dotA gene, Lp03, was unable to replicate inside macrophages. As previously shown, deletion of dotL in JR32 prevented the strain from replicating in U937 macrophages, and this defect could be complemented by the addition of a plasmid containing dotL+ (Fig. (Fig.7)7) (52).
Although the JR32 ΔdotL strain was viable, closer examination revealed that it displayed a key difference from other dot/icm mutants (Table (Table2).2). Wild-type L. pneumophila stains, such as Lp02 and JR32, exhibit a significantly decreased plating efficiency on buffered CYE plates containing a low amount of sodium chloride (0.65%) compared to growth on plates lacking sodium chloride (11, 45, 64). Most dot/icm deletions (e.g., ΔdotA) exhibit an increased plating efficiency on plates supplemented with salt (Table (Table2).2). However, the JR32 ΔdotL strain was even more sensitive to sodium chloride than JR32 (Table (Table2),2), suggesting that the physiology of the JR32 ΔdotL is perturbed. These data suggest that loss of dotL in either the Lp02 strain or the JR32 strain is detrimental to the cell.
Since the sequences of the dot/icm genes are identical between Lp02 and JR32, it is likely that there is an additional genetic difference between the two strains responsible for the more severe effect of deleting dotL in Lp02. For example, a gene may have been inactivated or lost during the derivation of JR32 that allows the JR32 ΔdotL strain to survive. Alternatively, a gene may be absent in Lp02 that is normally able to suppress the lethality caused by loss of dotL. In fact, a number of differences have been reported between various L. pneumophila serogroup I isolates including Lp01, the progenitor strain of Lp02, and JR32 (Table (Table1)1) (6, 36, 47). One potential candidate is lvhD4, which is present in JR32 but not in Lp01 (47). lvhD4 is encoded in the lvhB1-11/lvhD operon and is a component of a second type IV secretion system found in L. pneumophila strains such as JR32 (51). lvhD4 encodes a protein with similarity to T4CPs, most specifically to the A. tumefaciens VirD4 (51), and could in theory functionally substitute for DotL.
To confirm that the JR32 and Lp02 isolates we were working with contained and lacked lvhD4, respectively, we performed Southern analysis with a probe specific to lvhD4 (Fig. (Fig.8,8, top). Consistent with previous reports, Lp01 and Lp02 lacked lvhD4, whereas JR32 and Philadelphia-1, the progenitor strain for both Lp02 and JR32, both contained it (Fig. (Fig.8A).8A). Lp01 and Lp02 may have lost the lvhB-lvhD region during their derivation to become restriction minus, since a number of restriction or modification genes are located adjacent to the lvhB/lvhD4 system (47). To determine whether there was a connection between the presence of lvhD4 and ΔdotL lethality, we deleted lvhD4 from JR32 or added it back to Lp02 and then assayed the consequence of deleting dotL. dotL could still be deleted in a JR32 strain lacking the lvhB-lvhD4 region, indicating that lvhD4 was not responsible for the viability of the JR32 ΔdotL strain (Fig. (Fig.8B).8B). Likewise, the addition of the lvhB-lvhD4 region from JR32 to the Lp02 strain did not suppress loss of dotL. Therefore, lvhD4 does not appear to be responsible for the altered requirements of dotL in these two L. pneumophila strains.
In addition to the dot/icm and the lvhB systems, certain L. pneumophila strains contain an additional type IV secretion system (6). This system is encoded in an ca. 65-kb locus, LpPI-1, that bears the hallmarks of a pathogenicity island. It contains homologues to a type IV secretion system that resembles the F plasmid, including a T4CP that resembles the F plasmid TraD protein, mobile genetic elements, and several putative virulence factors (6). In contrast to lvhD4, LpPI-1 is present in Lp02 but is absent from JR32. However, deletion of the TraD-like protein from Lp02 did not suppress the lethality caused by loss of dotL (C. Vincent and J. P. Vogel, unpublished data). Thus, some additional, as-yet-uncharacterized, mutation must exist in one of these strains that is responsible for the differential requirement of dotL for viability.
While constructing a collection of dot/icm deletions, we were able to generate in-frame deletions in 23 of the 26 known dot/icm genes. However, similar to the dotL deletion, we could not construct a deletion in dotM, the gene upstream of dotL (Table (Table3).3). dotM, also known as icmP, codes for a predicted inner membrane protein with similarity to TrbA of the IncI plasmids R64 and ColIb-P9 (24% amino acid identity). Since dotL and dotM are likely to be cotranscribed in a two gene operon, dotML, it was possible that the lethality of the dotM deletion was due solely to polarity on the downstream dotL gene (Fig. (Fig.4).4). However, the ΔdotM mutation could not be obtained in the presence of a complementing clone containing a wild-type version of dotL, suggesting that the dotM lethality was not due to polarity but reflected the essentiality of dotM (Table (Table3).3). Moreover, insertions in dotM were obtained in a screen for genes that resemble dotL, i.e., genes that are essential in the presence of a functional Dot/Icm complex (13).
A third gene, dotN, also proved difficult to delete from the L. pneumophila chromosome. dotN, also known as icmJ, is located ca. 12 kb downstream of the proposed dotML operon and is the first gene of another predicted operon, dotNO (Fig. (Fig.4).4). dotN codes for a small protein of 208 amino acids that contains a high proportion of cysteines (4.3%). The lethality of the ΔdotN could not be due to simple polarity on the downstream gene dotO because deletions could easily be made in the dotO gene. Similar to dotL, both dotM and dotN could each be deleted in strains lacking a functional dot complex (Table (Table3).3). These results indicate that three dot genes, dotL, dotM, and dotN are each essential for viability on bacteriological media in the Lp02 background and in each case, the lethality can be suppressed by inactivation of the Dot/Icm complex.
The dot/icm genes are required for the intracellular replication of L. pneumophila and encode a type IVB secretion system that appears to have evolved from the conjugation apparatus of an IncI plasmid. We have demonstrated here that three dot genes, dotL, dotM, and dotN, are essential for growth of L. pneumophila strain Lp02 on bacteriological media. This is in direct contrast to the established paradigm that the dot/icm genes are dispensable under the laboratory conditions of growth on plates (3, 37). In addition, we were able to isolate a large collection of suppressors of the ΔdotL lethality and have shown that the majority of these map to other dot/icm genes. However, inactivation of several dot/icm genes (dotK, icmS, and icmW) did not suppress loss of dotL, indicating specificity to the suppression.
DotL has limited homology to the T4CP family of proteins. T4CPs have been proposed to play a central role in type IV secretion systems (34). They have been shown to bind substrates synthesized in the cytoplasm and target them to the secretion apparatus in the inner membrane (2, 15, 61). T4CPs have also been shown to interact with other components of the secretion apparatus, namely, the VirB10-family of proteins (20, 35). Finally, T4CPs are absolutely required for export of substrates (22). Based on their homology to Escherichia coli FtsK and Bacillus subtilis SpoIIIE, and their ability to bind DNA, T4CPs have been proposed to function as molecular pumps, driving export of substrates via hydrolysis of ATP (22). In consideration of these traits, T4CPs would appear to be likely candidates to function as regulators of the type IV secretion complexes.
Based on the similarity of DotL to T4CPs, it is surprising that inactivation of the dotL gene in strain Lp02 is lethal. No other known T4CP is essential for viability. Moreover, the only proteins associated with conjugative transfer that that are required for bacterial viability are inhibitors of plasmid toxin segregation factors (43). DotL, however, shows no sequence similarity to such factors. In addition, if DotL functioned as an inhibitor of a plasmid segregation toxin, then the ΔdotL lethality suppressors would be predicted to map to the toxin. In contrast, many of the ΔdotL suppressors are components of the Dot/Icm machinery, and the non-dot/icm suppressors do not have homology to any known toxin inhibitors.
To explain these overall observations regarding toxicity induced by loss of dotL, we propose that loss of the DotL protein results in the accumulation of a toxic structure consisting of a portion of the Dot/Icm complex (Fig. (Fig.9).9). This partial Dot/Icm complex could be deleterious for a number of different reasons. First, a partial Dot/Icm complex could misassemble or misfold in the absence of DotL, disrupting the membrane in some fashion. Alternatively, loss of dotL could be toxic because the type IV secretion system forms an unregulated pore in the membrane in the absence of DotL (Fig. (Fig.9).9). In this model, DotL would play the role of a regulator of the complex, controlling the opening and closing of the pore.
We favor the unregulated pore model for the following reasons. First, if a misfolded subcomplex were the cause of the lethality one would not anticipate that inactivation of the majority of dot/icm genes (20 of 23) would suppress the loss of dotL. Second, the JR32 ΔdotL phenotype, increased sensitivity to sodium relative to a wild-type strain, is much more consistent with an unregulated pore. Although the sodium sensitivity of wild-type L. pneumophila strains is not well understood, it is believed to result from leakage of sodium ions through the Dot/Icm secretion apparatus (11, 64). This model is supported by the observation that strains resistant to sodium chloride often contain mutations in dot/icm genes (63). Taken in this context, loss of a regulator of the secretion pore is predicted to enhance the effect of exogenous sodium and is consistent with the hypersensitivity of the JR32 ΔdotL. Finally, there is precedence in the literature of an example in which loss of a protein resulted in an unregulated pore that can be lethal under certain circumstances. Inactivation of Yersinia pestis lcrG results in an unregulated type III secretion pore under certain conditions and has led to the model where LcrG forms a plug at the base of the apparatus (39, 58).
Based on the phenotype of a strain lacking dotL, mutations that cause lowered viability in the presence of an intact Dot/Icm apparatus were previously isolated (13). lidA was shown to encode a protein exported by the Dot/Icm system that may interact directly with DotL (13). Other lid genes may encode proteins necessary for proper assembly of the Dot/Icm complex, particularly a subcomplex consisting of DotL, DotM, and DotN. For example, three Lid proteins are involved in disulfide bond metabolism and, since the DotN protein is rich in cysteine residues, it may be that mutations affecting the formation of disulfide bonds could disrupt folding of DotN (13).
The ΔdotL lethality phenotype in Lp02 has proven to be useful for several additional reasons. First, it has provided a convenient plate selection for additional dot/icm mutants. This is noteworthy because many of the dot/icm genes were identified by labor-intensive screens that have never been performed to saturation (1, 3, 45). The only selection for dot/icm mutants previously available was based on the phenomenon that sodium-resistant L. pneumophila strains were often avirulent, although this phenomenon is poorly understood and may be mutagenic (11, 64). The benefit of our new selection is amply demonstrated since we have already identified an additional dot/icm gene, dotV, by this procedure.
The ΔdotL lethality phenotype also provides information about existing Dot/Icm proteins. A number of Dot/Icm proteins that appear to be primarily cytoplasmic and not membrane associated were still able to suppress the loss of DotL when their genes were inactivated. For example, IcmQ and IcmR have been shown to be soluble proteins in the cytoplasm of L. pneumophila where IcmR appears to function as a chaperone for IcmQ (16). Although the specific function of IcmQ remains unknown, the fact that ΔicmQ and ΔicmR were able to suppress the lethality caused by the ΔdotL suggests that they are directly required for the assembly or activity of the Dot/Icm complex. Another example is the DotB ATPase (55). Although DotB does not appear to be an integral component of the Dot/Icm membrane complex, it is required for expression of the ΔdotL lethality trait, thus indicating that the protein plays a role in the assembly and/or function of the apparatus.
In contrast, inactivation of icmS or icmW did not suppress loss of dotL. Since icmS and icmW are predicted to encode cytoplasmic proteins and have been proposed to function as chaperones for secreted substrates (12), their failure to suppress is consistent with our model. Moreover, inactivation of a secreted substrate ralF (41) also failed to suppress loss of dotL (unpublished results). One additional Dot/Icm protein, the putative lipoprotein DotK, was also not required for ΔdotL lethality. Combined with the observation that a ΔdotK strain shows only mild defects for intracellular growth (53), this suggests that DotK is not essential for the formation of the Dot/Icm complex. Further examination of how various dot/icm mutants are able to suppress loss of dotL may reveal information on which components are key to formation of the secretion pore.
A third interesting observation that resulted from our analysis of the ΔdotL lethality involved dotM and dotN. Similar to dotL, we discovered that dotM and dotN are also essential for viability in the Lp02 background and are not essential for the viability of JR32 on bacteriological media but are required for growth of JR32 inside macrophages (50). Since all three proteins appear to code for inner membrane components of the secretion apparatus, it is possible that DotM and DotN interact with DotL and regulate its activity, perhaps by modulating its proposed nucleotide hydrolysis capability. In fact, we have recently shown that DotM can be coimmunoprecipitated by using DotL specific antibodies (Vincent and Vogel, unpublished).
It is interesting that deleting dotL in two very closely related strains results in very different phenotypes: death versus life. This is likely to be due to a genetic difference between the two strains acquired during their derivation. The JR32 strain may have acquired a suppressor mutation or Lp02 may have lost a gene that prevents ΔdotL lethality. One difference between these strains is that Lp02 lacks the second type IV secretion system encoded by the lvhB operon (47). However, deletion of the lvhB operon in JR32 did not cause the dotL deletion to be lethal, and therefore the identity of the suppressor(s) remains to be discovered. Nevertheless, the difference between these two strains may not be as profound as it initially appeared, since the JR32 ΔdotL strain is less fit than a wild-type strain, as demonstrated by its hyper-NaCl sensitivity. It is possible that the difference in phenotypes between the two strains is more a matter of degrees of sensitivity to loss of dotL rather than JR32 being impervious to its loss.
The ΔdotL phenotype described here is consistent with the proposal that T4CPs function as inner membrane gates for exported substrates (49). Further characterization of this interesting phenomenon should shed light not only on the function of DotL and other T4CPs but also on the L. pneumophila Dot/Icm complex and other type IV secretion systems.
We thank James Kirby for the generous gift of pJK211-2. We also thank Jessica Sexton and Carr Vincent for suggestions and critical reading of the manuscript.
G.M.C. was supported by training program J32AI1007422. This study was funded by the Whitaker Foundation (J.P.V.), the American Lung Association (J.P.V.), and NIH grant AI48052-01A2 (J.P.V.) and by funding from the Howard Hughes Medical Institute to R.R.I.