Type IV pili-mediated secretion modulates Francisella virulence


*E-mail tguina@u.washington.edu; Tel. (+1) 206 616 3468; Fax (+1) 206 543 5383.


Francisella tularensis are the causative agent of the zoonotic disease, tularaemia. Among four F. tularensis subspecies, ssp. novicida (F. novicida) is pathogenic only for immunocompromised individuals, while all four subspecies are pathogenic for mice. This study utilized proteomic and bioinformatic approaches to identify seven F. novicida secreted proteins and the corresponding Type IV pilus (T4P) secretion system. The secreted proteins were predicted to encode two chitinases, a chitin binding protein, a protease (PepO), and a β-glucosidase (BglX). The transcription of F. novicida pepO and bglX was regulated by the virulence regulator MglA. Intradermal infection of mice with F. novicida mutants defective in T4P secretion system or PepO resulted in enhanced F. novicida spread to systemic sites. Infection with F. novicida pepO mutants also resulted in increased neutrophil infiltration into the mouse airways. PepO is a zinc protease that is homologous to mammalian endothelin-converting enzyme ECE-1. Therefore, secretion of PepO likely results in increased production of endothelin and increased vasoconstriction at the infection site in skin that limits the F. novicida spread. Francisella human pathogenic strains contain a mutation in pepO predicted to abolish its secretion. Loss of PepO function may have contributed to evolution of highly virulent Francisellae.


Francisella tularensis are highly infectious bacteria and the causative agent of the zoonotic disease tularaemia (Ellis et al., 2002). Tularaemia can be acquired by different routes including the bite of blood-feeding arthropods, such as ticks, biting flies, or mosquitoes, as well as by inhalation or ingestion (Ellis et al., 2002). Inhalational tularaemia is the most severe form of the disease and is associated with uptake of contaminated aerosols or dust (Dennis et al., 2001). Fewer than 10 inhaled bacteria can cause disease in humans (Ellis et al., 2002). Infection with the most virulent strains results in a rapid systemic spread, sepsis and significant morbidity in untreated individuals (Dennis et al., 2001).

Francisella tularensis have been divided into four subspecies that are highly similar in their genomic content (Johansson et al., 2004). F. tularensis ssp. tularensis (F. tularensis, also known as type A) is the most virulent form for humans. F. tularensis ssp. holarctica (F. holarctica, also known as type B) and ssp. mediasiatica are virulent for humans but less so than type A strains. F. tularensis ssp. novicida (F. novicida) is infectious only for immunocompromised humans (Ellis et al., 2002). Inbred mice are highly susceptible and succumb to infection by all Francisella subspecies including F. novicida (Fortier et al., 1991; Kieffer et al., 2003). Similarly to the symptoms of human tularaemia, mouse infection results in a rapid development of severe systemic disease, likely due to the ability of Francisella to evade the host innate immune system (Dreisbach et al., 2000; Telepnev et al., 2003; Bosio and Dow, 2005).

Very little is known about Francisella virulence determinants and the mechanism by which they promote evasion of host innate immunity. Francisella sole described virulence regulator MglA (Baron and Nano, 1998) activates the transcription of genes in a Francisella pathogenicity island (FPI) (Lauriano et al., 2004) that are essential for the intracellular replication of all F. tularensis subspecies (Nano et al., 2004 and A. Sjöstedt, pers. comm.). The FPI genes have no homology to genes of known function and little is known about how they promote virulence. MglA/FPI also modulate host signalling to inhibit proinflammatory responses by an unknown mechanism (Telepnev et al., 2003). A recent study has shown that Francisella strains lacking pilA, agene for a predicted Type IV pilin, were attenuated upon subcutaneous infection of mice (Forslund et al., 2006).

Many bacterial virulence factors are secreted proteins (Stathopoulos et al., 2000; Winstanley and Hart, 2001). Until now, no secreted proteins or secretion systems of Francisella have been described. This study utilized proteomic approaches to identify F. tularensis secreted proteins for the first time. Bioinformatics and genomics approaches were utilized to demonstrate that Francisella protein secretion is mediated by a system that utilizes components that are homologous to Type IV pili (T4P) in other organisms. Mutation in F. novicida T4P secretion system or one of the secreted proteins resulted in increased virulence of these strains in mouse model of infection. Our results suggested that loss of function is likely one of the mechanisms that contributed to evolution of highly virulent Francisella.


Francisella novicida secrete proteins during growth in rich medium

With a goal to identify putative secreted Francisella proteins, concentrates of F. novicida culture supernatants were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE). At least 10 Coomassie-stained bands were detected in F. novicida supernatants from exponential cultures (Fig. 1A). In contrast, culture supernatants of F. holarctica Live Vaccine Strain (LVS) did not contain any proteins (data not shown). Seven F. novicida culture supernatant proteins were identified using tandem mass spectrometry (Fig. 1A and Table 1). Sequence analysis showed that the open reading frames (ORFs) encoding F. novicida secreted proteins contain predicted N-terminal signal sequences essential for the efficient translocation of these proteins across the inner membrane via the Sec system (Oliver and Beckwith, 1982). Four F. novicida secreted proteins exhibited significant similarity to previously described enzymes that included three glycosylases: two predicted chitinases referred to as ChiA and ChiB (for chitinase A and chitinase B respectively; GenBank Accession DQ230365 and DQ230366), a β-glucosidase similar to the Escherichia coli BglX (BglX, GenBank Accession DQ230371) and a protease similar to bacterial PepO (PepO, GenBank Accession DQ230367). Three F. novicida secreted proteins including ChiA and ChiB contain at least one chitin binding domain (see Table 1). This was confirmed by binding of these proteins to the chitin beads (Fig. S1). The additional chitin binding protein is referred to as CbpA (chitinbindingprotein A). Two secreted proteins are not similar to proteins of known function; therefore they are referred to as Fsp58 and Fsp53 (for Francisellasecreted protein of 58 kDa and 53 kDa respectively).

Figure 1.

F. novicida culture supernatant proteins regulated by MglA.
A. Culture supernatant proteins were separated by 8% SDS-PAGE and visualized by Coomassie.
B. RT-PCR analysis of MglA-regulated genes. Total RNA was prepared from the F. novicida wild type (wt), mglA mutant and iglC mutant strains. PCR primers were specific to ORF regions of tul4, iglC, pepO and bglX.

Table 1. F. novicida culture supernatant proteins.
GenBank accessionkDaConserved domain search, database cdd.v2.03 http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi (E-value)NamePredicted function
  1. kDa, protein molecular weight based on the migration on the SDS-PAGE gel.

DQ23036593pfam00704, Glyco_hydro_18, Glycosyl hydrolases family 18 (1e-42)
cd00036, ChtBD3, Chitin binding domains (5e-07, 3e-07 and 2e-04)
cd00063, fibronectin binding domain (3e-06)
DQ23036675pfam00704, Glyco_hydro_18, Glycosyl hydrolases family 18 (1e-07)
COG3469, Chitinase (7e-06)
DQ23036773COG3590, PepO, Predicted metalloendopeptidase (4e-178)
pfam05649, Peptidase_M13_N, Peptidase family M13 (6e-79)
DQ23036860cd00036, ChtBD3, Chitin/cellulose binding domain (7e-04)CbpAChitin binding protein
DQ23036958No hitsFsp58Unknown
DQ23037053No hitsFsp53Unknown
DQ23037138pfam00933, Glyco_hydro_3, Glycosyl hydrolase family 3 N terminal domain (2e-29)
COG1472, BglX, Beta-glucosidase-related glycosidases (1e-50)

Francisella novicida pepO and bglX are positively regulated by MglA

  • MglA is the sole described Francisella virulence regulator (Baron and Nano, 1998). Analysis of the F. novicida culture supernatant protein showed that PepO and BglX were not secreted by the F. novicida mglA mutant (Fig. 1A). Results of semi-quantitative RT-PCR showed that transcription of both pepO and bglX were reduced in the F. novicida mglA mutant (Fig. 1B) when compared with the wild-type strain, or strain carrying a mutation in FPI non-regulatory gene iglC (Fig. 1B). In addition, neither the mglA nor iglC mutation affected transcription of the gene expressing F. novicida lipoprotein Tul4 as described previously (Lauriano et al., 2004) (Fig. 1B). These findings were confirmed by results of F. novicida whole cell protein proteomic analysis in which relative expression of F. novicida wild-type and mglA cellular protein was determined (T. Guina, unpublished). PepO abundance was increased 10-fold in the wild-type strain when compared with the isogenic mglA mutant (P < 0.01). BglX was detected in the wild-type F. novicida but not in the isogenic mglA strain. Positive regulation of PepO expression by MglA suggested that this protein could play a role during animal host infection.

Francisella novicida PepO is a secreted metalloprotease

  • One of the F. novicida secreted proteins exhibited similarity to the M13 family of zinc metalloproteases that cleave vasoactive peptides. These include both mammalian and bacterial enzymes such as human endothelin-converting enzyme ECE-1 (Turner and Tanzawa, 1997), and Streptococcus parasanguis and Porphyromonas gingivalis PepO (Awano et al., 1999; Oetjen et al., 2001). ECE-1 processes pro-endothelin into its mature form endothelin, the most potent vasoconstrictor that is produced by various types of cells, but mostly by vascular endothelium and bronchiolar epithelium (Turner and Tanzawa, 1997). The carboxy-terminal region of F. novicida PepO contains all the conserved enzymatic active site residues in exact alignment to the corresponding residues of ECE-1 and bacterial PepO (not shown). These data suggested that F. novicida PepO is also a M13-family protease. ECE-1 and PepO from other bacterial species were previously shown to cleave the neuropeptide metenkephalin (YGGFM) in vitro (Awano et al., 1999; Oetjen et al., 2001). F. novicida strains expressing the chromosomal or plasmid-encoded PepO also cleaved YGGFM into YGG and FM (Fig. 2A). In contrast, a strain lacking pepO (a transposon at the pepO codon 353 of 687 that resulted in lack of expression of the C-terminal protease domain), and a strain expressing PepO point mutant with a change of histidine to glutamine (PepOHQ) in a predicted protease Zn-binding motif (H524EISH), lacked such activity (Fig. 2A). In trans-complemented pepO strains, PepO and PepOHQ were secreted by F. novicida (Fig. 2B). These data confirmed that F. novicida PepO is an ECE-1 orthologue.

Figure 2.

PepO is an M13 class metalloprotease.
A. F. novicida expressing PepO (wt and pepO + pKKpepO) cleave metenkephalin (YGGFM), while strains expressing mutant PepO (pepO and pepO + pKKpepOHQ) do not. Peptides were separated by thin layer chromatography. Ctrl, YGGFM and FM peptides.
B. Trans-complemented F. novicida strains secrete PepO and PepOHQ.

Francisella novicida secrete proteins by use of components that are homologous to T4P

With the goal to identify the F. novicida system involved in secretion of the above described proteins, we searched Francisella genomes for homologues of membrane-associated transporters and secretion systems components. Among 240 identified ORFs, 20 candidate genes encoding predicted transporter proteins of unknown function and transporter-associated nucleotide binding proteins were selected. The membrane-associated nucleotide binding proteins were selected because they are essential for secretion systems assembly or they provide energy for the secretion process (Pugsley, 1993). The culture supernatant protein profiles of F. novicida transposon mutants in 10 genes (see Experimental procedures) were examined as described above. Mutation in one of 10 selected genes resulted in a significant reduction of F. novicida protein secretion. In this mutant, the transposon was inserted in the first gene of a predicted bicistronic operon that encoded PilB and PilC (GenBank Accession DQ230373), orthologues of proteins that are essential for assembly of Gram-negative T4P (Peabody et al., 2003) (Fig. 3). Francisella predicted PilB and PilC are homologues of an inner membrane-associated ATPase and a polytopic inner membrane protein essential for T4P secretion and assembly. The transposon insertion in pilC also abolished F. novicida protein secretion, while trans-complementation of the pilC homologue mutant resulted in restoration of protein secretion (Fig. 3). Lack of secretion by F. novicida pilQ mutant was also observed (Fig. 3). pilQ is predicted to encode the only Francisella secretin. Secretins form outer membrane multimers that are essential for the functional assembly of Type II (T2S), Type III and T4P secretion systems (Sandkvist, 2001a; Peabody et al., 2003; Ghosh, 2004). In Gram-negative bacteria, T2S systems typically export chitinases and enzymes, some of which contribute to bacterial virulence in animals and plants, and also assist their survival in the environment (Sandkvist, 2001b). pilB, pilC and pilQ were the only genes in the Francisella spp. genomes that could participate in the assembly of either T2S or T4P systems, but homologues of other T2S-specific components such as gspC, gspM and gspL are missing (data not shown). Therefore, our results suggested that a Francisella secretion system utilizes homologues of a T4P system to accomplish protein secretion. Interestingly, a transposon insertion in a gene encoding predicted T4P ATPase PilT did not abolish F. novicida secretion (Fig. 3). PilT is essential for T4P retraction and is typically absent from the T2S systems (Burrows, 2005).

Figure 3.

F. novicida secretes proteins via T4P system. Culture supernatant proteins were separated by 8% SDS-PAGE. wt, wild-type strain; pilB, pilC, pilQ, pilT, pilA2, pilA1 and pilA, mutants in T4P system homologues; pilC + pKKpilC, trans-complemented strain.

pilA is the first gene in a F. novicida three-gene operon (pilAA1A2) that corresponds to the pilAEV operon encoded by pathogenic Francisella (Forslund et al., 2006). These operons each encode three homologues of pilin genes that are essential for T4P assembly and fibre formation (Peabody et al., 2003). Examination of secreted protein profiles of F. novicida transposon mutants in each of the pilA, pilA1 and pilA2 genes showed that only pilA, the first gene in this operon, is essential for protein secretion during growth in rich medium (Fig. 3). Long pili-like structures have been demonstrated on the surface of the attenuated F. holarctica LVS (Gil et al., 2004). As LVS is a natural pilA mutant (Forslund et al., 2006), it is certain that the extended structures on the surface of LVS are not a result of a structure that contains PilA. In a recent study, Forslund et al. did not detect the long fibre structures on the surface of pathogenic F. holarctica (Forslund et al., 2006) though PilA formed oligomers expressed on the surface (Forslund et al., 2006). Our attempts to visualize the long pili-like fibres on F. novicida surface by using electron microscopy and previously described methods (Gil et al., 2004) were also not successful (data not shown). Therefore, it is likely that T4P homologous genes participate in the formation of a secretion system of Francisella spp. and may not form true T4P. Another possibility is that other predicted pilin homologues encoded by Francisella (data not shown) form LVS pili.

Mutation in the T4P secretion system does not affect F. novicida intracellular replication rates

To investigate the role of T4P secretion system in F. novicida intracellular survival, the wild-type strain and a pilC mutant were used to infect the mouse macrophage cell lines J774 and MH-S, and a human lung cancer epithelial cell line A549. Examination of bacterial numbers after 24 and 48 h of infection showed that intracellular growth rates of F. novicida pilC were similar to that of the wild-type strain in all three mammalian cell types (data not shown). In addition, the pilC mutation did not affect F. novicida uptake into the above described cell types as measured by the gentamicin protection assay (data not shown). Similar results were obtained for intracellular replication and uptake of a pepO mutant (data not shown). These results were consistent with those of Forslund et al. in which a pilA mutation did not affect the intracellular replication of F. holarctica (Forslund et al., 2006).

Mutations in the T4P secretion system and mutations in the gene expressing secreted protease PepO result in enhanced F. novicida spread to systemic sites

Francisella infection of mice and humans via intradermal (i.d.) route mimics the natural route of tularaemia transmission via arthropods. In a recent study, lack of pilA expression resulted in attenuation of a F. holarctica strain (Forslund et al., 2006). To determine whether T4P-mediated protein secretion contributed to F. novicida virulence, groups of BALB/cByJ mice were infected i.d. with various bacterial doses. Surprisingly, infection with the F. novicida pilC mutant resulted in enhanced mouse mortality when compared with infection with the wild-type strain. In one experiment, groups of 15 mice were infected i.d. with 100 × LD50 of either F. novicida wild type or F. novicida pilC. Forty-eight hours after infection, seven out of 15 mice infected with F. novicida pilC were dead and four were moribund. In contrast, only two out of 15 mice infected with the wild-type F. novicida were dead and one was moribund. Similar results were obtained upon mouse infection with F. novicida pilA mutant (data not shown). This difference in mortality was reflected by the difference in time needed for the bacteria to reach high numbers in the spleen. After 24 h of infection with two F. novicida LD50, there were 1.5–2 logs higher numbers of bacteria in spleens of mice infected with the F. novicida pilC when compared with mice infected with the isogenic wild-type strain (1.25e+5 cfu median versus 2.0e+3 cfu median, P < 0.01) (Fig. 4). Spleen burdens of mice infected with the trans-complemented pilC mutant were similar to those of mice infected with the wild-type F. novicida (8.2e+3 cfu median) (Fig. 4). Therefore, our data suggested that lack of T4P-mediated secretion resulted in increased virulence of F. novicida.

Figure 4.

Mouse infections with F. novicida wild-type (wt) and pilC mutant. Bacterial spleen burdens in mice (n = 5) infected with 2e+3 F. novicida by i.d. route. Spleen burdens of mice infected with wild-type strain (wt), pilC mutant (pilC) and a trans-complemented strain (pilC + pKKpilC) for 24 h. Results that are representative for more than three experiments are shown. Horizontal lines represent median values. **P < 0.01 for differences between two groups spanned by the lines.

Infection of mice with the F. novicida pepO transposon mutant (see above) also resulted in enhanced mouse morbidity and increased spleen burden at 24 h post infection when compared with mice infected with the wild-type strain (5.8e+5 cfu median versus 3.6e+4 cfu median, P < 0.01) (Fig. 5A). Similar to the results above, there was a significant increase in the number of bacteria recovered from the spleens of mice infected with trans-complemented F. novicida expressing and secreting PepOHQ, a point mutant in the protease Zn-binding site (see above), when compared with mice infected with the strain expressing and secreting wild-type PepO (3.7e+4 cfu median versus 3.0e+3 cfu median, P < 0.03) (Fig. 5B). These results suggested that the increased organ burden in mice infected with F. novicida that either lacked PepO, or secreted inactive PepO was likely due to lack of extracellular PepO protease activity. Due to the predicted role of PepO in conversion of pro-endothelin into a highly potent vasoconstrictor, it is possible that PepO secretion controls F. novicida spread from skin to the systemic sites.

Figure 5.

Mouse infections with F. novicida wild type (wt) and pepO mutants.
A. Bacterial spleen burdens in mice (n = 5) infected with 5e+4 F. novicida wild type (wt) and pepO transposon mutant (pepO) via i.d. route for 24 h. Results that are representative for more than three experiments are shown. Horizontal lines represent median values. **P < 0.01 for differences between two groups spanned by the lines.
B. Bacterial spleen burdens in mice (n = 4) infected with 1e+3 F. novicida pepO transposon mutant trans-complemented either with wild-type pepO (pepO + pKKpepO) or with pepO point mutant (pepO + pKKpepOHQ) via i.d. route for 24 h. Results that are representative for three experiments are shown. Horizontal lines represent means. *P < 0.03 for differences between two groups spanned by the lines.
C. Bronchoalveolar neutrophils in mice (n = 4) infected with 2e+3 of aerosolized F. novicida wild type (wt) and pepO transposon mutant (pepO) for 24 h. Results that are representative for more than three experiments are shown. *P < 0.02 for differences between two groups.

To investigate whether the effect of PepO on mouse infection was specific to the inoculation via skin, C57BL/6 mice were infected with aerosolized F. novicida wild type or F. novicida pepO. In contrast to the outcome of infection via the i.d. route, bacterial numbers in the spleen, liver and lungs of mice infected with aerosolized F. novicida pepO were similar to the organ burdens of mice infected the wild-type F. novicida (data not shown). However, the number of bronchoalveolar neutrophils was fourfold greater in mice infected with a F. novicida pepO mutant in comparison with mice infected with the wild-type strain (4.77e+5 ± 0.77e+5 versus 1.27e+5 ± 0.26e+5, mean ± SEM respectively; P < 0.02) at 24 h after infection (Fig. 5C). This result suggested that increased vascular permeability occurs in airways of mice infected with F. novicida pepO mutant. This difference was not associated with greater local induction of proinflammatory cytokines and chemokines that are active in recruiting neutrophils to the lungs; levels of TNF-α, MIP-2 and KC, in lung homogenates at 4 and 24 h after infection with pepO did not differ significantly from those in mice infected with wild-type F. novicida (data not shown). Due to the predicted role of PepO in conversion of pro-endothelin, it is likely that the increased number of airway neutrophils resulted from increased perfusion to the sites of infection in mice infected with the pepO mutant.

Based on its protease activity demonstrated by this study (see above), it is likely that F. novicida PepO secretion likely promotes increased pro-endothelin processing and increased vasoconstriction of the lymphatic and blood vasculature. As a result, PepO limits F. novicida spread from skin to the systemic sites during infection. Conversely, mutations that prevent PepO expression, secretion or enzymatic activity contribute to higher invasiveness of F. novicida. Overall, our results suggested that F. novicida PepO may function to moderate infection in the mammalian host.

PepO is mutated in human pathogenic F. tularensis strains

Genome analyses have shown that among Francisella spp., F. novicida is most similar to the Francisella progenitor strains, while human pathogenic F. tularensis and F. holarctica have evolved more recently (Johansson et al., 2004). Genome analysis of F. tularensis Schu S4 and F. holarctica OSU18, and attenuated F. holarctica LVS revealed that the insertion sequence (IS)-mediated genomic rearrangement has occurred at the 5′ end of pepO when compared with pepO region in F. novicida genome (Fig. 6A). As a result of this rearrangement, the first 43 codons that encode the N-terminal secretion signal of F. novicida PepO have been replaced by codons encoding other 12 amino acids (data not shown). This rearrangement likely resulted in an inability of the pathogenic Francisella strains to secrete PepO, and change in pepO expression levels when compared with F. novicida. Accordingly, the relative amounts of pepO mRNA were lower in F. holarctica LVS when compared with the amounts in F. novicida at the same culture density (Fig. 6B). In addition, F. holarctica genomes contain a point mutation that resulted in the removal of the PepO C-terminal 109 amino acids that resulted in inactivation of the protease domain. In conjunction with other results of this study that demonstrated increased invasiveness of F. novicida pepO mutants in mice, the genomic analysis suggested that loss of PepO secretion likely contributed to the evolution of highly virulent Francisella subspecies.

Figure 6.

Genomic organization and expression of pepO in F. tularensis spp.
A. Genomic organization of the pepO region in F. novicida (Fn) versus pathogenic F. tularensis Schu S4 and F. holarctica (Ft) strains.
B. RT-PCR analysis of gene expression in F. novicida strains wild type (wt), mglA and iglC; and F. holarctica LVS.


This study has identified, for the first time, Francisella secreted proteins and components necessary for their secretion. Mutations in T4P homologues pilB, pilC, pilQ and pilA abolished secretion in F. novicida. Though these proteins have similarity to T4P components, they may compose a secretion system rather than a pilus. Three secreted proteins were shown to bind chitin and one, PepO, was shown to be an M13 class zinc protease. Secreted chitinases are essential for the establishment of the arthropod infection by malaria parasite Plasmodium (Vinetz et al., 1999). The Plasmodium chitinase allows the parasite to penetrate the chitin-containing peritrophic matrix surrounding the blood meal in the mosquito midgut and establish the infection. Therefore, proteins secreted by Francisella T4P system described herein could play a similar role in arthropod infection and natural transmission of Francisella.

Two of the F. novicida T4P-secreted proteins were positively regulated by the virulence regulator MglA (Baron and Nano, 1998) that is essential for virulence of all Francisellae in mice, implicating the role of secretion system in alteration of the interaction with mammals. In a recent study, pilA mutation resulted in loss of virulence for mice of a F. holarctica strain inoculated by the subcutaneous route (Forslund et al., 2006). In addition, mutations in some of the genes encoding secreted proteins described in this study resulted in attenuation of human pathogenic Francisella virulence in mouse model of infection (T. Guina and A. Sjöstedt, in preparation). Therefore, protein secretion via T4P system described in this study contributes to virulence of pathogenic Francisella strains. There are other examples of protein secretion via T4P, some of which contribute to virulence. Myxococcus xantis pilQ mutation resulted in altered secreted protein profiles (Wall et al., 1999). Type IV fimbriae of Dichelobacter nodosus, a causative agent of footroot in sheep, are essential for virulence, secretion of a protease and natural competence (Kennan et al., 2001). Vibrio cholerae Type IV toxin coregulated pilus (TCP) mediates secretion of TcpF, a factor essential for the colonization of the mouse small intestine (Kirn et al., 2003).

Subunits essential for Gram-negative T2S and T4P assembly share significant similarity that suggests a common evolutionary origin (Burrows, 2005). It has been proposed that the T2S and T4P secretion could be promoted by pilus polymerization and retraction that could either push the proteins out through the outer membrane secretin pore, or open a gated secreted channel for protein translocation across the outer membrane (Sandkvist, 2001a). T4P pilus assembly and retraction are energized by action of the inner membrane-associated ATPases PilB and PilT respectively (Burrows, 2005). In this study, PilB, but not PilT, was essential for F. novicida secretion, suggesting against the role of PilT-mediated pilus retraction in secretion. A recent study suggested that the assembly and disassembly of the major B. subtilis competence pseudopilin occurs by exchange of intramolecular for intermolecular disulphide bonds between major pseudopilin subunits (Chen et al., 2006). Further studies should examine the possibility that the Francisella secretion is mediated by a similar process.

Surprisingly, this study showed that mutations in a component of the secretion system PilC, or secreted protease PepO, resulted in increased F. novicida systemic spread and enhanced replication in subcutaneously inoculated mice, consistent with a role of promotion of virulence. Expression of PepO also contributed to moderation of neutrophil influx in the airways of mice infected with aerosolized F. novicida. Bacterial PepO are homologues of mammalian ECE-1 that cleave precursors of the potent vasoconstrictor endothelin in vitro (Turner and Tanzawa, 1997; Awano et al., 1999; Oetjen et al., 2001). This activity is consistent with the hypothesis that F. novicida PepO can process endothelin and promote vasoconstriction, the result of which is limitation of bacterial spread upon infection via skin. Therefore, it is plausible that F. novicida virulence factors are moderated by the effects of PepO and other secreted proteins delivered through T4P. Results of this study and that by Forslund et al. (2006) suggest that the observed differences in T4P-mediated virulence among the Francisella spp. are possibly due to presence or absence of the specific genes encoding T4P substrates and/or strain-specific gene regulation.

An IS-mediated rearrangement occurred at the 5′ end of pepO in human pathogenic Francisella resulting in the expression of a non-secreted form of PepO. F. holarctica genomes contain an even shorter pepO that lacks the functional protease domain. As infection with F. novicida strains that do not secrete PepO enhanced mouse mortality, it is possible that the higher invasiveness of the human pathogenic Francisella strains, when compared with F. novicida, could be in part explained by loss of PepO secretion. Genes for several other T4P-secreted proteins have been mutated in pathogenic Francisella, suggesting these mutations could also contribute to evolution of pathogenic Francisella and their adaptation to a specific host or environmental niche. Increased bacterial virulence and adaptation to the specific hosts that were result of gene loss and inactivation have been observed in other pathogens such as Yersinia, Salmonella and Bordetella (Parkhill et al., 2003; Chain et al. 2004; McClelland et al., 2004). Our study suggests that alterations in Francisella secreted proteins have likely occurred during adaptation to ecological environments and hosts that support strains that have increased pathogenesis for humans. Studies that address the role of other T4P-secreted proteins in Francisella virulence are currently under way.

Experimental procedures

Bacterial strains and growth conditions

Francisella novicida U112 wild type and mglA (Baron and Nano, 1998) were obtained from Francis Nano (University of Victoria, Victoria, Canada). F. novicida iglC was obtained from Karl Klose (Lauriano et al., 2004). F. holarctica LVS was obtained from Karen Elkins (U.S. Food and Drug Administration, Rockville, MD). Bacteria were grown in Tryptic soy broth (Gibco BRL, Grand Island, NY) supplemented by 0.1% cysteine (TSB-C). F. novicida transposon insertion mutants were selected from a defined genomic transposon mutant library (L. Gallagher et al., unpublished). This mutant library was generated according to previously described methods (Kawula et al., 2004) using the commercially available <KAN-2> ‘transposome’ (Epicentre, Madison, WI). Insertion locations were determined by semi-degenerate PCR amplification of the transposon–genome junctions followed by DNA sequencing. The mutants were colony purified and phenotypes confirmed by analysis of F. novicida culture supernatant protein profiles (see below).

Secreted protein preparation and protein identification

For preparation of secreted proteins, bacterial cultures were grown to mid-logarithmic phase and proteins were concentrated upon precipitation in the presence of 10% (v/v) trichloroacetic acid as previously described (Kimbrough and Miller, 2000). Proteins were separated by SDS-PAGE and visualized by Coomassie blue. Protein identification was performed using tandem mass spectrometry and by matching the resulting peptide fingerprint spectra (Guina et al., 2003) with the Francisella genome databases [F. tularensis Schu S4, GenBank Accession No. AJ749949; and the preliminary F. novicida and F. holarctica LVS and OSU18 genome annotations that were generated using the microbial gene finder algorithm Glimmer (Salzberg et al., 1999) respectively]. Concentrated F. novicida culture supernatants used for identification of secreted proteins did not contain cytoplasmic proteins, as confirmed by mass spectrometry/peptide fingerprinting, and by immunoblotting using the polyclonal rabbit antiserum raised against 21 kDa cytoplasmic protein IglA (gift of Francis Nano, University of Victoria, Victoria, Canada). Protein sequence alignments and analysis were performed by using the following algorithms: blast (http://www.ncbi.nlm.nih.gov/BLAST), the Conserved Domain Database search (http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi), psort (http://psort.nibb.ac.jp/form.html) and signalp (http://www.cbs.dtu.dk/services/SignalP).

Isolation and identification of secreted F. novicida chitin binding proteins

For protein binding to chitin beads, filtered culture supernatant was concentrated at the room temperature using Amicon Stirred Cells (Millipore, Billerica, MA) and Ultracell Amicon YM30 cellulose filter (30 000 Da NMWL; Millipore, Billerica, MA) under nitrogen gas pressure. When the supernatant reached 10% of the original volume, sample was washed by addition of two starting bacterial culture volumes of 20 mM Tris-Cl pH 8.3. Protein was finally concentrated to the 0.3% starting bacterial culture volume. Chitin beads (NEB, Ipswich, MA) were washed twice with PTE buffer (phosphate-buffered saline pH 7.2, 0.1% Triton-X100 and 0.1 mM EDTA) and mixed with the increasing amounts of concentrated supernatant proteins. Mixture was rotated at room temperature for 2–3 h, after which the unbound protein fraction was saved. Chitin beads were washed thrice in PTE buffer and suspended in the Laemmli buffer. Proteins were separated by SDS-PAGE and visualized by Coomassie blue.

Proteins were identified by using mass spectrometry and bioinformatics tools as described elsewhere in the text.

RNA isolation and RT-PCR

Total RNA was isolated from mid-logarithmic cultures (OD600 = 0.3–0.4) by TRI REAGENT (Molecular Research Center, Cincinnati, OH) per manufacturer's recommendations. Total RNA was treated with DNase I (Invitrogen, Carlsbad, CA), precipitated using ice-cold ethanol, and RNA amounts were then normalized to each other. RT-PCR reactions were performed as described (Lauriano et al., 2004). Control experiments demonstrated a lack of contaminating genomic DNA in the purified total RNA. Gene-specific PCR primers were designed based upon corresponding 300–400 nt internal sequences within tul4, pepO, bglX and iglC.

Mutant complementation and generation of a pepO site-directed mutant

For trans-complementation of a F. novicida pilC mutant, the pilC ORF with the upstream Shine–Dalgarno sequence was cloned downstream of the F. tularensis LVS groESL promoter into the PstI and EcoRI sites of pKK214-GFP (Golovliov et al., 2003), resulting in pKKpilC. The pKKpilC plasmid was then introduced into F. novicida pilC by electroporation (Maier et al., 2004). Return of the secretion function by trans-complementation with pKKpilC was verified by analysis of the F. novicida secreted protein profiles. The pepO mutant was complemented in a similar fashion, except that a complementing pKKpepO plasmid contained the F. novicida pepO ORF and 255 nucleotides of upstream DNA. Return of PepO secretion by trans-complementation was verified by analysis of the F. novicida secreted protein profiles. A F. novicida pepO point mutant in which PepO histidine 254 in the coding sequence (His524) was changed to Gln (pepOHQ; pepO nucleotide 1571 was changed from T to A) was generated by using QuikChange Site-Directed Mutagenesis kit (Stratagene, La Jolla, CA).

Francisella novicida uptake and intracellular replication in macrophages and epithelial cells

All cell lines were obtained from American Type Culture Collection (ATCC, Manassas, VA). The murine macrophage cell line J774 was cultured in DMEM medium (GibcoBRL, Grand Island, NY) supplemented with 10% fetal bovine serum (FBS) at 37°C in 5% CO2 atmosphere. Alveolar macrophage cell line MH-S and human lung epithelial cell line A549 were cultured in RPMI 1640 supplemented with 25 mM HEPES, 0.3 g l−1l-glutamine and 10% FBS. Macrophage and epithelial cell infections were performed as described earlier (Baron and Nano, 1998) at moi of 1, 10 and 100 bacteria per mammalian cell.

Mouse infections

Six- to eight-week-old specific pathogen-free (SPF) BALB/cByJ and C57BL/6 mice were purchased from the Jackson Laboratory (Bar Harbor, ME). Animals were housed in sterile cages in a barrier environment at the University of Washington SPF facility. Mice were fed autoclaved food and water, and all experiments were performed under the University of Washington IACUC guidelines. Male BALB/cByJ mice were given 50 μl of appropriately diluted bacteria i.d. at the shaved mid-belly. Doses of inoculated bacteria were simultaneously determined by plate count. All materials used in animals, including bacteria, were diluted in endotoxin-free phosphate-buffered saline (Gibco BRL, Grand Island, NY). Moribund animals that exhibited hunched posture, piloerection, laboured breathing, and lack of mobility were humanely euthanized as determined by IACUC guidelines. For determination of the infected organ burden, groups of five mice were euthanized after 24 h of infection; their spleens were collected, homogenized and incubated for 5 min with PBS-buffered 0.1% (w/vol) saponin. Organ burden was determined by plating serial dilutions of the homogenized organs onto TSB-C agar. Male and female C57BL/6 mice were exposed to aerosolized bacteria in a nose-only exposure chamber (InTox Products, Moriarty, NM). Bacterial suspensions were aerosolized using Uni-Heart nebulizers (Westmed, Lakewood, CO) driven at 38 psi to a flow rate of 3 l min−1. Airflow through the chamber was maintained at 15 l min−1. After a 10 min exposure, bacterial deposition in homogenized lung tissue of three mice was determined. At subsequent time points, quantitative organ cultures, bronchoalveolar lavage (BAL) cell counts and protein amounts were determined as described previously (Skerrett et al., 1999; 2004). All experiments were repeated at least three times. Statistical data analysis was performed using Prism software (GraphPad Software, San Diego, CA). As data were not normally distributed, comparisons between groups were performed using the Mann–Whitney test. A P-value of < 0.05 was considered statistically significant.


We would like to thank the staff of the University of Washington Genome Center for access to the unpublished F. novicida genomic sequence; Beth Ramage and Colin Manoil for help in construction of F. novicida transposon library; Jimmie Lara for help with electron microscopy; Denise Monack and Anders Sjöstedt for critical reading of the manuscript. This study was funded by the NIAID award for the WWAMI Research Center of Excellence for Biodefense and Emerging Infectious Diseases #5 U54 AI057141-03 to S.I.M., T.G., M.B. and S.J.S.