The Hfq Homolog in Legionella pneumophila Demonstrates Regulation by LetA and RpoS and Interacts with the Global Regulator CsrA

The Legionella homolog of Hfq lies in open reading frame (ORF) lpg0009 of the Legionella Genome Project (http://legionella.cu-genome.org/ ; last accessed November 2004). The Legionella Hfq protein is 64% identical to that in E. coli and 67% identical to that found in P. aeruginosa. Further analysis of the amino acid sequence revealed conservation of the amino acids at positions 8 (Gln), 42 (Tyr), 56 (Lys), and 57 (His), which have been proposed to be necessary for RNA binding (28, 32). However, the conserved chromosomal gene arrangement found in E. coli, V. cholerae, and other organisms is only partially found in L. pneumophila. In E. coli, the hfq gene is part of a multigene operon in which hfq is preceded by the gene miaA and followed by hflX, hflK, and hflC. The Legionella hfq gene is followed by an ORF with homology to the hflX gene, as seen in E. coli; however, these are the only two genes from the operon that are present in this region of the Legionella genome. Upstream of the L. pneumophila hfq gene is an ORF, running in the transverse direction, with homology to a pspA gene. Downstream of the hflX homolog ORF are two ORFs, again with transverse orientation, the first being a thioredoxin homolog and the further one being a putative phospholipase C protein. With the QIAGEN One-Step reverse transcription-PCR kit, experiments with JR32 wild-type exponential-phase cultures grown at 37 and 30°C determined that the hfq-hflX genes are cotranscribed (data not shown). In the hfq mutant background, no signal was obtained from the primers spanning the region of the hfq gene into the hflX gene (RT F2 and RT R2). Experiments with a primer 168 nucleotides upstream of the hfq gene (Hfq ORF F2) and a primer near the end of the hfq gene (RT R1) resulted in no product. Northern blot analysis, conducted as previously described (10), of hfq transcript expression in wild-type strain JR32, a serogroup 1 isolate, demonstrated strong expression during exponential-phase growth (optical density [OD] = 1.2 and 1.5) at 30 and 37°C. No transcripts were detected in stationary-phase bacteria (OD = 1.8) (data not shown). All primers used in this work are listed in Table 1.

An hfq mutant was constructed through homologous recombination in wild-type strain JR32 and verified by standard Southern blot analysis. Cultures for determination of growth and pigmentation kinetics were grown in either standard Legionella BCYE (buffered charcoal-yeast extract) medium or chemically defined medium (24). Experiments were conducted as previously described (10). The hfq mutant revealed defects in growth at both 37°C (Fig. 1a) and 30°C (Fig. 1b) in standard BYE growth medium, as well as in chemically defined minimal medium (Fig. 1c). Pigmentation, a virulence-associated trait, was reduced only when bacteria were grown at 30°C (Fig. 1d). Interestingly, the growth and pigmentation defects seen in the hfq mutant were similar to defects seen in the JR32 csrA mutant at 30°C (10). Under all conditions, the growth deficit appeared to be a delayed lag phase rather than a deficit in replication ability. Growth and pigmentation defects could be complemented by introduction of plasmid pBCKS carrying the hfq gene under the control of its own promoter region (pBC-hfq).

As hfq expression appeared to be growth phase dependent and mutants exhibited growth and pigmentation defects similar to those seen in the csrA mutant, we examined the relationship of hfq with the global regulatory protein CsrA. We examined csrA transcript expression in the hfq mutant in comparison with wild-type expression through standard Northern blot analysis. CsrA is normally expressed in replicative-phase cells during early and mid-exponential-phase growth and is necessary for replication in amoebae and macrophages (10, 22). In comparison with those in JR32 wild-type cells, csrA transcripts were present but reduced in the hfq mutant (Fig. 2a). Complementation of the mutant with pBC-hfq resulted in a wild-type level of csrA transcript expression. Experiments carried out at 30 and 37°C revealed the same expression pattern among the wild type, the mutant, and the complemented strain (hfq/C).

A csrA deletion mutant is characterized by premature expression of transmission and stationary-phase traits. fliA and flaA transcript expression at mid-exponential phase, as well as production of the flagellar protein FlaA, is seen in the csrA deletion mutant (10, 22). Analysis of these traits revealed no defects in the hfq mutant compared to the wild type (data not shown) at either 30 or 37°C. Western blot analysis with a Legionella-specific FlaA antibody (15) provided by K. Heuner (Würzburg University) also revealed no differences from wild-type expression. Because of this flagellar phenotypic difference in the hfq and csrA mutants, CsrA protein levels were examined in the hfq mutant. Western blot assays with an E. coli CsrA-specific antibody (provided by T. Romeo, Emory University) revealed no significant difference in CsrA protein levels between the wild type and the hfq mutant (Fig. 2b). In both strains, as well as an rpoS mutant strain, CsrA protein was present during exponential-phase growth and at the beginning of the stationary phase. For comparison, our regulatory csrA mutant, which still contained minute amounts of csrA mRNA, did not produce detectable amounts of the corresponding protein. Therefore, although csrA mRNA levels were reduced in the hfq mutant, the amount present was capable of being translated to produce levels of CsrA protein necessary for repression of stationary-phase traits. We therefore tested if we could complement the growth and pigmentation defects of the hfq mutant by expressing csrA from a plasmid. With the expression of csrA from plasmid pMMB2002 (27), we were able to restore normal growth and pigment production to the hfq mutant at 30°C, demonstrating that these defects are at least in part due to the interaction of Hfq with csrA RNA (Fig. 1b).

In the process of identifying genes whose expression is affected by loss of the Hfq protein, we also examined the transcription kinetics of the fur gene and the translation efficiency of the fur promoter region in an hfq mutant background. The interaction of the Hfq protein with the ferric uptake regulator has been described in E. coli, functioning through the rhyB sRNA (35). Although not as prominent as the reduction of csrA, the amount of fur transcript was reduced in the L. pneumophila hfq mutant and clearly complemented to more than wild-type levels with the introduction of the hfq gene in trans on multicopy plasmid pBCKS (Fig. 3a). Further analysis at 30 and 37°C with a luciferase reporter gene fusion to the fur promoter region revealed high activity of the fur promoter in wild-type cells during exponential-phase growth, as expected. Luciferase assays were conducted as previously described (17). In comparison, the same pattern of expression was present but clearly reduced in the hfq mutant background. Interestingly, this same reduction in expression was also seen in the csrA and rpoS mutants (Fig. 3b).

In order to determine if the defect in Fur expression was a cause for the growth defect, we examined the growth kinetics of the wild type, the mutant, and the complemented strain in BYE medium with twice the normal amount of iron supplement (120 μg) added, as well as in CDM without iron. Experiments were conducted at 30°C, as this is the temperature at which the greatest variation between the wild type and mutant exists. Interestingly, in medium without an iron source the JR32 and hfq/C strains were able to reach OD levels seen in complete CDM but took longer to reach this level. The hfq mutant, however, grew poorly in this medium (Fig. 3c). In medium containing excess iron, the hfq mutant was able to grow at wild-type rates; however, pigment production remained lower than that of the wild type (data not shown). These data point to a role for Hfq in the iron uptake and storage system of L. pneumophila.

LetA has been well characterized as a activator of stationary-phase phenotypes through relief of CsrA repression (3, 14). We therefore examined the relationship of hfq and letA. hfq expression levels of letA mutant cultures grown at 30°C to the early exponential, mid-exponential, and early stationary phases were compared to wild-type expression levels. At mid-exponential phase, when hfq transcripts were at their highest in wild-type cells, practically no transcripts were found in the letA mutant (Fig. 4a). This defect could be complemented when the letA gene was introduced into the mutant on plasmid pMMB-letA (17). Interestingly, at stationary phase, when hfq transcripts are no longer detectable in wild-type cells, we could detect more hfq transcripts in the letA mutant than seen during the exponential phase (Fig. 4b). It has been previously shown that RpoS is reduced in a letA mutant background (17). Therefore, we were interested in determining if the reduction in hfq transcripts in the letA mutant was due solely to the loss of LetA or perhaps to the reduction of RpoS. Investigations of transcript expression in an rpoS mutant (generously provided by H. Shuman, Columbia University) showed a very low or no signal of hfq transcripts at both 30 and 37°C (Fig. 4c). The increase in hfq transcripts seen in the letA mutant at stationary phase was not seen in the rpoS mutant at that time point. Therefore, hfq transcription appears to be negatively regulated by the LetA protein in the stationary phase but positively regulated in the exponential phase by the stationary-phase sigma factor RpoS.

Since Hfq has been shown to be necessary for RpoS translation in various other bacterial species (5, 29, 38), we wanted to determine if that was the case for L. pneumophila. We examined rpoS RNA and protein expression in the hfq mutant background in comparison to that in wild-type cells at 30 and 37°C. rpoS mRNA in L. pneumophila strain LP02 has been reported to be expressed during exponential-phase growth but not during stationary phase (2). We confirmed this expression pattern to also be the case in L. pneumophila JR32 (Fig. 4d). Additionally, comparison of rpoS transcripts of the hfq mutant to those of our wild type showed no significant differences in the expression pattern. Further analysis of the translation of RpoS with luciferase assays (17) also demonstrated no differences between the wild type and the hfq mutant (data not shown). This was not caused by a necessary secondary structure of the rpoS mRNA for Hfq function, as a reporter gene fusion vector that included the first 20 nucleotides of the coding region also led to no difference in luciferase expression levels between the wild type and the hfq mutant. In L. pneumophila, hfq is apparently regulated by RpoS and is not, under these growth conditions, necessary for RpoS expression. Thus, the regulation is similar to the regulation of hfq found in L. monocytogenes, where hfq expression is regulated by the RpoS functional homolog σ^B (6). In gram-negative bacteria such as E. coli and P. aeruginosa, an inverse direction of regulation has been observed (29, 30). Along with this difference in Hfq function in comparison with the E. coli hfq gene, the Legionella hfq gene is also not able to complement an E. coli mutant for Qβ phage replication (30), demonstrating further functional differences in this protein for the two species (data not shown).

Hfq has been shown to be capable of stabilizing mRNA transcripts by protecting the transcripts from RNase E degradation (21). Most likely, the interaction of Hfq with csrA and fur is necessary for stability during the replicative cycle. Analyses of both the csrA and fur gene sequences demonstrate the presence of AU-rich areas that are possible targets for both RNase E and Hfq binding. From the growth curves we could see that the hfq mutant demonstrated more of a prolonged lag phase but, once replication began, was capable of replication at a similar rate and reached the same levels as the wild type. Perhaps this prolonged lag phase is due to degradation of csrA transcripts where at the beginning of the replication cycle the first transcripts are immediately degraded. However, over time a threshold level of csrA transcripts can be reached where more transcripts are produced than can be simultaneously degraded. The excess of transcripts would then allow translation of the CsrA protein, which is then sufficient for replication and repression of stationary-phase traits. The growth and pigmentation defects found in the hfq mutant could then be due at least in part to the reduction in csrA or by other CsrA-independent functions. This theory is supported by the ability of plasmid-expressed csrA to cure the growth and pigment defects of the hfq mutant, as well as by the growth kinetics of hfq and hfq/C in medium with excess iron or without iron.

A proposed scheme of hfq regulation is that in exponential phase RpoS induces the expression of Hfq, which contributes to the stability of exponential-phase RNA, allowing the bacteria to quickly adapt to and efficiently use the replicative environment. Then, when cells respond to signals of nutrient deprivation and must transcend to an infectious, virulent form, LetA, in its role as an inducer of stationary-phase traits (2, 3), either directly or indirectly turns off hfq transcription. This turning off of transcription occurs through an RpoS-independent pathway, as hfq transcripts are not found in rpoS mutants at stationary phase.

Although much about the function of the RNA binding Hfq protein has been published, only few studies have examined its regulation. Our studies open a new area for Hfq research in examining the regulation by the LetA transcriptional activator, as well as the stationary-phase sigma factor RpoS. Additionally, Hfq is well known to be essential for several regulatory mechanisms through binding with sRNAs. To date, no sRNAs have been identified in legionellae. BLAST searches for homology to the known sRNAs from E. coli, salmonellae, or yersiniae have produced no hits in the Legionella genome (data not shown). Through the use of Hfq, it may now be possible to identify these molecules and determine what role they may play in regulating the biphasic life cycle of legionellae. Further studies should also concentrate on determining the mechanism behind the proposed stabilization of csrA and fur mRNA by Hfq.

Because of the pleiotropic effects of a deletion of hfq, we were interested to see if this also translates into defects in multiplication in living cells. Therefore, Acanthamoeba castellanii host cells were infected with the JR32, hfq, and hfq/C strains at a multiplicity of infection of 10 in accordance with previously published protocols (17). Although no significant differences were found between the wild type and the hfq mutant, the trend demonstrated that the hfq mutant invaded with slightly less efficiency than the wild type (Fig. 5a) at both 30 and 37°C. Replication in the amoebae was not affected at 37°C, but at 30°C it demonstrated a lag in replication reminiscent of that seen in the in vitro growth curves (Fig. 5b). The host cells responsible for supporting human infection are alveolar macrophages. The MH-S mouse alveolar macrophage cell line has been demonstrated to be a competent model for Legionella infection of the lung (19, 37). Infection assays were conducted by standard protocols. Briefly, MH-S cells (10⁶) in six-well plates were infected in duplicate at a multiplicity of infection of 10 with either the JR32, hfq, or hfq/C strain. Bacterial suspensions were plated on BCYE to determine accurate bacterial counts. Plates were centrifuged briefly in order to ensure Legionella contact with the macrophages and then incubated for 2 h. Cells were then washed twice with phosphate-buffered saline and incubated for a further 60 min in culture medium with gentamicin (200 μg/ml). Cells were then again washed twice in phosphate-buffered saline and incubated until the appropriate time point for analysis. To determine invasion ability, the first set of cells was immediately lysed and bacteria were plated on BCYE agar to determine CFU. The percentage of bacteria capable of invading was calculated as the number of CFU at 2 h divided by the number of CFU in the initial bacterial suspension. Replication ability was calculated as the log of the number of CFU at each time point divided by the number of CFU at 2 h. Each experiment was done with duplicate wells for each time point, and the experiment was conducted three times. The macrophage infection assays again showed no significant differences in invasion between the wild type and the mutant (Fig. 5a) but revealed the same trend seen in the amoeba infections. (Fig. 5b). After 72 h of intracellular growth in macrophages, the hfq mutant had a significant decrease in number compared to the JR32 and hfq/C strains.

We therefore conclude that hfq plays a crucial role in the regulatory network of legionellae that is different from that in other gram-negative bacteria. Despite this, hfq deletions only slightly affect the ability of legionellae to invade cells and multiply intracellularly. It remains to be examined if the defect results in alterations in virulence in an in vivo animal model.

We thank K. Mueller and A. Flossdorf for excellent technical assistance, B. Yang for performance of the fur luciferase assays, A. Flieger for advice and critical review of the manuscript, T. Romeo and K. Heuner for providing antibodies, N. Cianciotto for the pMMB2002 vector, and I. Moll for the Qβ phage, bacterial strains, and technical advice.

This work was supported by the BMBF Community Acquired Pneumonia Network (CAPNETZ) program.