JOURNAL OF BACTERIOLOGY, Oct. 2004, p. 6430–6436
0021-9193/04/$08.00?0 DOI: 10.1128/JB.186.19.6430–6436.2004
Copyright © 2004, American Society for Microbiology. All Rights Reserved.
Vol. 186, No. 19
A Francisella tularensis Pathogenicity Island Required for
Francis E. Nano,1* Na Zhang,1Siobha ´n C. Cowley,2Karl E. Klose,3Karen K. M. Cheung,1
Michael J. Roberts,1Jagjit S. Ludu,1Gregg W. Letendre,1Anda I. Meierovics,2
Gwen Stephens,4and Karen L. Elkins2
Department of Biochemistry and Microbiology, University of Victoria, Victoria,1and British Columbia Centre for Disease Control,
Vancouver,4British Columbia, Canada; Center for Biologics Evaluation and Research, Food and Drug Administration,
Rockville, Maryland2; and Department of Microbiology and Immunology,
University of Texas Health Sciences Center, San Antonio, Texas3
Received 19 May 2004/Accepted 30 June 2004
Francisella tularensis is a gram-negative, facultative intracellular pathogen that causes the highly infectious
zoonotic disease tularemia. We have discovered a ca. 30-kb pathogenicity island of F. tularensis (FPI) that
includes four large open reading frames (ORFs) of 2.5 to 3.9 kb and 13 ORFs of 1.5 kb or smaller. Previously,
two small genes located near the center of the FPI were shown to be needed for intramacrophage growth. In
this work we show that two of the large ORFs, located toward the ends of the FPI, are needed for virulence.
Although most genes in the FPI encode proteins with amino acid sequences that are highly conserved between
high- and low-virulence strains, one of the FPI genes is present in highly virulent type A F. tularensis, absent
in moderately virulent type B F. tularensis, and altered in F. tularensis subsp. novicida, which is highly virulent
for mice but avirulent for humans. The G?C content of a 17.7-kb stretch of the FPI is 26.6%, which is 6.6%
below the average G?C content of the F. tularensis genome. This extremely low G?C content suggests that the
DNA was imported from a microbe with a very low G?C-containing chromosome.
Francisella tularensis is a highly infectious gram-negative coc-
cobacillus that causes the zoonotic disease tularemia (11). This
bacterial pathogen is known for its ability to cause a fulminat-
ing disease in humans after exposure to as few as 10 cells and
has raised considerable concern as a potential bioterrorist
agent (10). Because of its high infectivity and lethality, F.
tularensis is one of six types of microbes classified by the U.S.
Centers for Disease Control and Prevention as a category A
agent, one that poses the most serious threat as a vehicle of
There are a variety of subspecies and biotypes of F. tularen-
sis, but they all have greater than 95% DNA sequence identity.
Although the type A and type B biotype strains are highly
infectious, only type A strains, which are found exclusively in
North America, cause significant mortality in infected humans.
An attenuated variant of a type B biotype strain formed the
basis of a live vaccine strain (LVS) of F. tularensis. Understand-
ing the molecular basis of the differences in virulence levels of
F. tularensis strains may help in the development of a rationally
F. tularensis is a facultative intracellular pathogen. The cur-
rently available evidence suggests that F. tularensis resides in-
side a membrane-bound phagosome during its initial growth in
a macrophage and that it may be released into the cytoplasm
during a later phase of growth (3, 14). Very little is known
about the bacterial virulence factors needed for infection, al-
though it is clear that intracellular growth, especially in mac-
rophages, is essential to the virulence of F. tularensis. A bio-
chemical study of the LVS of F. tularensis showed that four
proteins are induced after F. tularensis entry into macrophages
(15). The gene encoding the most prominently induced pro-
tein, the 23-kDa IglC protein, has been molecularly cloned and
sequenced. Recently, Golovliov et al. (16) deleted this gene
and showed that the resulting mutant was unable to grow in
macrophages and unable to cause disease in mice.
Genetic approaches have also been used to discover other F.
tularensis genes needed for optimal intracellular growth. The
products of mglA and mglB, thought to be global regulators,
are both required for intramacrophage growth and virulence in
mice (5). Random insertional mutagenesis revealed that inac-
tivation of F. tularensis genes encoding homologues of
(purine biosynthesis), alanine racemase (peptidoglycan biosyn-
thesis), and the heat shock-inducible ClpB protease reduces
the ability of F. tularensis to grow in mouse macrophages (17).
Perhaps most significantly, transposon insertion into iglB and
iglC, which are part of the pathogenicity island described in this
work, profoundly affects intramacrophage growth (17).
The strategy of parasitizing host cells is a common theme
used by both bacterial and protozoan pathogens (2). In many
bacterial intracellular pathogens, a specific gene or set of genes
that promotes entry into host cells has been identified (12, 13,
20). Two general types of cell entry mechanisms have been
identified. One involves the tight binding of a bacterial surface
protein to a host cell receptor, followed by engulfment of the
bacterial cell by a zipper-like phagocytosis. A second type of
uptake uses type III secretion machinery to inject effector
proteins into host cells, inducing membrane ruffling and mac-
ropinocytosis. The genetic loci in Salmonella and Shigella spp.
* Corresponding author. Mailing address: Department of Biochem-
istry and Microbiology, P.O. Box 3055 STN CSC, University of Victo-
ria, Victoria, B.C. V8W 3P6, Canada. Phone: (250) 721-7074. Fax:
(250) 721-8855. E-mail: firstname.lastname@example.org.
that encode the products needed for entry into mammalian
cells are pathogenicity islands of common origin (18). The
horizontal movement of this cluster of genes has enabled dif-
ferent species of bacteria to gain the ability to invade cells.
In this study we provide evidence for a Francisella pathoge-
nicity island (FPI) that is required for intramacrophage growth
and virulence in mice. The presumed effector proteins encoded
by the FPI genes show no definitive similarity to known pro-
karyotic virulence proteins and thus represent novel factors
required for virulence and intramacrophage growth. The gene
encoding the PdpD protein appears to be absent in F. tularensis
type B strains, and this absence may play a role in the wide
difference in virulence of human infections between the type A
and type B strains.
MATERIALS AND METHODS
Strains and molecular techniques. The following F. tularensis strains were
used. F. tularensis B38 (ATCC 6223) is the type strain for the highly virulent F.
tularensis subsp. tularensis (type A biotype). However, the B38 strain has lost
virulence through laboratory passage. The F. tularensis LVS (ATCC 29684) is the
type strain for F. tularensis subsp. holarctica and represents the type B biotype. F.
tularensis subsp. novicida (type strain U112; ATCC 15482) was used for all gene
knockout and virulence work; the DNA sequence reported in this work is from
the U112 genome. Unless otherwise stated, the F. tularensis subsp. novicida
strains were grown in Trypticase soy broth supplemented with 0.1% cysteine
(TSBC). The generation time for F. tularensis subsp. novicida U112 and mutant
strains grown in TSBC was in the range of 80 to 90 min. F. tularensis Schu4 is a
fully virulent F. tularensis subsp. tularensis strain. Initial bioinformatics analysis
was performed using the DNA sequence of the genome project of F. tularensis
strain Schu4 (http://artedi.ebc.uu.se/Projects/Francisella/); however, all of the
analyses reported for the region covering pdpD through pdpA were done with the
F. tularensis subsp. novicida strain U112 sequence. Clinical isolates of F. tularen-
sis were from the following locations in British Columbia and were collected in
the years indicated: B1, Mission, 1993; B2, Vernon, 1997; B6, Vernon, 2002; B7,
Vandehoof, 2003. These clinical isolates and the Schu4 strain were manipulated
under biosafety level 3 (BSL3) containment, and all other strains were handled
under BSL2 conditions. Standard recombinant DNA and PCR techniques were
used to manipulate or analyze DNA (29). Transposon mutagenesis (8) and
transformation (4) of F. tularensis to create mutant 304-2 were performed as
previously described. To construct the pdpD mutant, JL12, regions of the chro-
mosome flanking pdpD were amplified by PCR and ligated to an erythromycin
resistance cassette (see Fig. 4B). The region to the left of pdpD was apparently
lethal in Escherichia coli in high-copy-number vectors. Therefore it was amplified
with the proofreading Pfx polymerase (Invitrogen) and ligated to the low-copy-
number vector pWSK29 (30), which had been digested with EcoRV, which
generates flush ends. The primers used for this amplification were pdpDL-F,
GGTACCTGGGTTATTTTTGCTGCTGA, and pdpDL-R, CTCGAGGATCC
ATACTTACTACTCTTACAAGTAAACC. The resulting amplicon was 1,864
bp. The right side of pdpD was amplified by standard PCR techniques and cloned
into pCR2.1 (Invitrogen) with primers pdpDR-F, CTCGAGCAATGATCTGG
GTTTAAATTTAGC, and pdpDR-R, GGTACCGCCATTTCTAAAGGGGT
TGG. The resulting amplicon was 1,315 bp. The two recombinant clones were
digested with XhoI and joined to an erythromycin resistance cassette (19) that was
engineered to contain flanking XhoI sites by PCR amplification using the primers
EmXhoF, CTCGAGTGAATCGTTAATAAGCAAAATTC, and EmXhoR, CTC
GAGTTAAGGGATGCAGTTTATGC. The replacement of pdpD by the erythro-
mycin cassette was verified by using PCR with three pdpD primer sets to show that
pdpD was absent and combining primers for the erythromycin cassette with primers
that hybridize to DNA flanking pdpD to generate amplicons.
DNA sequencing of the FPI region was performed with custom primers.
Sequence assembly and analysis were done with the LaserGene (DNAStar) suite
of programs. Comparison of amino acid sequences deduced from the FPI to
those in protein and nucleic acid databases was done using online BLASTP (1)
with the default settings. TBLASTN and BLASTN were also used. The locations
of transposon insertions were determined by amplifying the area of interest and
sequencing parts of the amplicon by initiating DNA sequencing reactions from
the TnMax2 transposon (19) with the primers AAACATGCAGGAATTG
ACGA and TTCCTGAGCCGATTTCAAAG. Mutant 304-2 was genetically
complemented by introducing DNA cloned into the kanamycin resistance,
broad-host-range plasmid pDSK519 (23). Several primers were used to produce
the amplicons shown in Fig. 2 and 3, and they are listed in Table 1; the relative
positions of some of these primers are shown in Fig. 4A.
An antibody was raised against IglA in a rabbit by injecting recombinant
protein. The primers CTCGAGGGCGTTGTTAAGGTAACTTGC and CTCG
AGCAACTTCTGTAGATCCCCCAAA were used to amplify the iglA gene as
an XhoI-XhoI fragment. This amplicon was cloned into the SalI site of pTZ18U
and transformed into BL21?DE3(pLysS) (Novagen). The hyperexpressed IglA
was separated on sodium dodecyl sulfate–12% polyacrylamide gel electrophore-
sis gel and purified with a Bio-Rad model 491 Prep cell. The acetone-precipitated
protein was emulsified in TiterMax adjuvant (Sigma) and injected subcutane-
ously three times at intervals of 3 weeks into a New Zealand White rabbit, in
accordance with Canadian Council on Animal Care protocols. Immune serum
was used in immunoblots at a dilution of 1:2,500. The antibody reactivity was
visualized by reacting the blots with IRDye800-conjugated goat anti-rabbit im-
munoglobulin G (Rockland Immunochemicals, Gilbertsville, Pa.) and exposing
the filters to excitation light in a LiCor Odyssey imaging system.
Infection of murine bone marrow-derived macrophages with bacteria. Bone
marrow macrophages were infected with F. tularensis subsp. novicida strain U112
or mutants as previously described (7). Briefly, bone marrow cells obtained from
femurs of healthy BALB/cByJ male mice were plated at 2 ? 106/well in 24-well
plates (Costar, Corning, N.Y.) for 1 week in complete tissue culture medium
containing L929 supernatants (cDMEM). Macrophages were then infected with
F. tularensis strains at a multiplicity of infection of 1:20 (bacterium-to-macroph-
age ratio), and monolayers were incubated for 2 h in cDMEM, washed, and
incubated at 37°C in 5% CO2for the remainder of the experiment. The common
practice of adding gentamicin to kill extracellular bacteria cannot be used with
the F. tularensis subsp. novicida strain in this assay as it is exquisitely sensitive to
this treatment; it has been demonstrated that macrophages do not support F.
tularensis extracellular growth in cDMEM (3). To determine bacterial uptake
and replication, infected macrophages were lysed at the time points indicated in
Fig. 5 with sterile distilled water for 3 min, diluted immediately in phosphate-
buffered saline, and plated on Mueller-Hinton (MH) agar plates containing the
appropriate antibiotics. The plates were incubated for 1 to 2 days at 37°C in 5%
CO2, and colonies were counted.
Bacterial stocks. For macrophage and animal experiments, isolated colonies of
bacteria were inoculated into modified MH broth (Difco Laboratories, Detroit,
Mich.) supplemented with ferric pyrophosphate and IsoVitaleX (Becton Dick-
inson, Cockeysville, Md.) as previously described (7). Broth cultures were grown
to mid-log phase as previously described (7), and 1-ml aliquots of bacteria were
frozen in broth alone at ?70°C. These were periodically thawed for use, and
viable bacteria were quantified by plating serial dilutions on MH agar plates. The
number of CFU after thawing varied less than 10% over a 12-month period.
Animals and mouse infections. Six- to eight-week-old male specific-pathogen-
free BALB/cByJ mice were purchased from the Jackson Laboratory (Bar Har-
bor, Maine). Animals were housed in sterile microisolator cages in a barrier
environment at the Center for Biologics Evaluation and Research. Mice were fed
autoclaved food and water ad libitum, and all experiments were performed under
Institutional Animal Care and Use Committee guidelines. Mice were given 0.1
ml of appropriately diluted bacteria intradermally at the base of the tail; actual
doses of inoculated bacteria were simultaneously determined by plate count. All
materials used in animals, including bacteria, were diluted in phosphate-buffered
saline (BioWhittaker, Walkersville, Md.) containing ?0.01 ng of endotoxin/ml.
The sequences of the pdpD regions of two type B strains were assigned
GenBank accession numbers AY626806 and AY626807. The DNA sequence of F.
tularensis subsp. novicida strain U112 has been assigned GenBank accession no.
The identification of an FPI. The sequencing of the F. tula-
rensis genome (22, 28) and the development of simple genetic
tools with F. tularensis subsp. novicida have facilitated analysis
of virulence factors of Francisella. We previously isolated two
mutations in the linked genes iglB and iglC that reduce the
ability of F. tularensis to grow in macrophages (17). We per-
formed bioinformatics analysis of the DNA in the region of the
F. tularensis genome surrounding the location of these inser-
tions and discovered an apparent FPI of approximately 30 kb
(Fig. 1). In the left half of the FPI are eight open reading
VOL. 186, 2004FRANCISELLA PATHOGENICITY ISLAND6431
frames (ORFs), four of which, iglABCD, appear to be orga-
nized into an operon. The deduced products of iglA and iglB
have about 30% identity to hypothetical proteins found in
several bacterial species, most of which are animal or plant
pathogens or plant symbionts. One set of similar genes, impBC
of Rhizobium leguminosarum, encode proteins thought to be
involved in protein secretion (6). Bladergroen and colleagues
(6) noted that presumed homologues of impBC are found
organized in an identical fashion in operons in a number of
gram-negative bacteria; this organization is maintained in the
F. tularensis iglABCD operon. The product of iglC has previ-
ously been shown (15) to be very highly induced after entry of
F. tularensis into mouse macrophages and has recently been
shown by us and others to be needed for growth in macro-
phages (16, 17). However, the deduced proteins IglC and IglD
show no significant similarity to other known proteins. In the
right half of the FPI are three large ORFs, named pdpABC (for
pathogenicity determinant protein). The region between pdpB
and pdpC has eight relatively short ORFs, seven of which are
below 800 bp and one of which is 1,431 bp. None of the amino
acid sequences deduced from pdpABC or the smaller ORFs
show substantial similarity to those of known proteins.
Pathogenicity islands are often recognized by the aberrant
G?C content in their DNA, which differs from that for the rest
of the resident genome. The F. tularensis genome has an over-
all G?C content of 33.2% (28). The FPI has different regions
that have variable G?C content. The region corresponding to
pdpD through iglD has a G?C content of 31% (Fig. 1). The
region from 1 kb to the left of pdpC to 204 bp to the right of
the start codon of pdpA has a G?C content of 26.6% (bp 7969
to 25635 in the sequence with GenBank accession no.
AY293579). Immediately to the right of the presumed pro-
moter region of pdpA lies a 5,050-bp region that is 51% G?C
and that encodes rRNA. The very different G?C content of
TABLE 1. Primers used to amplify FPI genes
Gene or region
Primer Sequence (5?-)Amplicon size(s) (bp)
TGC TTT TAG TGG GTC ATG GA
CTA CGG CAT AAA TGG CTG GT
pdpB 3,282 pdpB-F
TTA GGT ACC GCC TTG CTC TTC TAA GTT GA
ATA CTC GAG GCT AGC AAT GTC ATC AAA GG
TGC CTG AGT CAT TGC TTG AT
TTG CCT CAA CAA CTG CTT TG
CAA GTG CTT GGT GGT GGT AA
TGA TGT TTG ACC TGA ATT AGT GG
TGG GTT ATT CAA TGG CTC AG
TCT TGC ACA GCT CCA AGA GT
TCC TGG CTT TGA TTT TGA GC
AAA TCT TGT TCA TCA AAC GCA AT
CTC GAG CAA TGA TCT GGG TTT AAA TTT AGC
GGT ACC GCC ATT TCT AAA GGG GTT GG
pdpD–iglD NALA-F TAA AAT TGC ACA GCA GAT AAG AGC9,932 (U112),
LA-RCGT ATA GCT GAT GGC TGG GCC
CTC GAG GGC GTT GTT AAG GTA ACT TGC
CTC GAG CAA CTT CTG TAG ATC CCC CAA A
CTC GAG CTC TTG TGA TGC TGC TGA GTC T
CTC GAG TCG CCA CTT GTT ACC TGT TG
CTC GAG TTT GAA GGA ATG AAT ACT ACA ATG A
CTC GAG CCA TCT TCC CAA TAA ATC CTT
iglD 1,197 iglD-F
CTC GAG GCG CAG CTA GCA CAG ATA AA
CTC GAG GCT GGG CTA TCC CTC ATT AT
TTG CGC AGC TAG CAC AGA TA
TCT GCG AAC TTC AAT TCT CTT TC
Region to left of pdpD NA pdpDL-F
GGT ACC TGG GTT ATT TTT GCT GCT GA
CTC GAG GAT CCA TAC TTA CTA CTC TTA CAA GTA AAC C
Region ca. 4 kb to
left of pdpD
NA O5-F AGT GTA ATG GAG CCC AAC CA420
O5-R GGT TTG CCA AAG CAG ATG AT
aNA, not applicable.
6432NANO ET AL. J. BACTERIOL.
this region is consistent with the need for conservation of the
rRNA sequence, which permits changes in the G?C content
primarily to adapt to the optimum growth temperature of the
The recently released raw sequence data for the genome of
F. tularensis LVS show that there are two copies of the FPI in
this genome. (These sequence data were produced by the
BBRP Sequencing Group at Lawrence Livermore National
Laboratory and can be obtained from ftp://bbrp.llnl.gov/pub
/cbnp/F-tularensis/F.tularensis.html.) One of the LVS forms of
the FPI is essentially identical to the Schu4 form (excluding the
pdpD deletion; see below) from one set of inverted repeats on
the left end to the other set on the right end. A second copy of
the FPI in the LVS strain is identical from the inverted repeats
on the left end through the rRNA genes. Thus, the presence of
two copies of this region in LVS suggests that the FPI region
was capable of movement at one time and may still have the
capacity to be mobile.
The pdpD gene is absent from type B strains of F. tularensis.
Upstream of iglABCD is a large ORF that we named pdpD.
The deduced amino acid sequence shows no significant simi-
larity to that of any known protein. The F. tularensis subsp.
novicida form of the PdpD protein is composed of 1,245 amino
acids and is predicted to be 141 kDa with a pI of 6.84. Com-
parison of the deduced amino acid sequence of PdpD found by
us in F. tularensis subsp. novicida strain U112 to the PdpD
found in F. tularensis subsp. tularensis strain Schu4 showed that
the F. tularensis subsp. novicida form has 50 additional amino
acids. Forty-eight of these amino acids were found in a con-
tinuous stretch between Q494 and Q543. We examined the
nature of the corresponding region of the pdpD gene by am-
plifying this region by PCR using DNA from the three widely
available type strains of F. tularensis (B38, LVS, and F. tula-
rensis subsp. novicida U112); Schu4 DNA, which represents the
sequenced genome, was also included (Fig. 2A). Several con-
trol PCRs, involving two other regions of the pdpD gene, as
well as regions of iglC-D, pdpC, pdpB, and pdpA (see below),
were performed. As expected, the PCR product that corre-
sponds to codons 468 to 512 (for Schu4; codons 468 to 560 for
U112) generated a product that was 136 bp when B38 or Schu4
DNA was used as the template and 280 bp when U112 DNA
FIG. 1. Gene organization and G?C content of the FPI. Top
graph, fractional G?C content of the 300-kb region of the F. tularensis
subsp. tularensis (strain Schu4) chromosome that encompasses the FPI
(http://artedi.ebc.uu.se/Projects/Francisella/); bottom graph, G?C
content of the FPI. The ORFs (arrows) from pdpD through pdpA are
derived from the DNA sequence of F. tularensis subsp. novicida strain
U112, determined in this work (GenBank accession no. AY293579)
and the remaining ORFs and sequence data are derived from the F.
tularensis subsp. tularensis Schu4 sequence. At the left end of the FPI
are ORFs tnpAB, which show exact identity to genes encoding pre-
sumed transposases previously found in F. tularensis (GenBank acces-
sion no. AAL06399 and AAL064100); at the right end is tnpA only.
The small opposing arrows indicate the approximate positions of 16-bp
inverted repeats that have previously been shown to be associated with
tnpA (21). The ORF labeled pmcA shows 44% identity to conserved
domains found in putative molecular chaperones, the most closely
related of which is the vdcC gene of Bacillus anthracis (NP_656320).
The short nature of seven of the eight ORFs between pdpB and pdpC
suggests that the area has many sequencing errors or is composed of
nonfunctional vestigial genes. However, the DNA sequence is highly
conserved between F. tularensis subsp. tularensis and novicida, suggest-
ing that these ORFs represent functioning genes. Arrows labeled 62
and 116, locations of the transposon insertions that originally indicated
a cluster of virulence-associated genes; arrow labeled 304-2, location of
the insertion that gave rise to the mutant described in this work, codon
401 of pdpA (821 codons). Hatched line, region of the allelic replace-
ment in the JL12 mutant; solid thick lines, extent of the F. tularensis
DNA insert in the complementing plasmid pVIC314 and the non-
complementing plasmid pVIC311. rrsH and rrlH, 16S and 23S rRNA
genes, respectively; fdn, A subunit of formate dehydrogenase. The
smaller ORFs are not drawn to scale; the numbers of ORFs cited in the
text are based on the F. tularensis subsp. novicida DNA sequence.
FIG. 2. PCR amplification of FPI segments. (A) Results when
primers were used to amplify internal fragments of pdpA, pdpB, pdpC,
pdpD, and the junction between iglC and iglD. The resulting PCR
products are shown; they provide evidence that the amplified regions
in pdpA, pdpB, pdpC, and iglC-D in the F. tularensis strains that were
tested are very similar. The three primer pairs used to amplify the
region encompassing codons 123 to 356 (pdpD-1), codons 468 to 560
(in U112; codons 468 to 512 in Schu4; pdpD-2), and codons 906 to
1131 for pdpD-3 demonstrate that the pdpD gene is missing or is
substantially different in sequence from the form found in strains
U112, Schu4, and B38. The different sizes of the PCR products with
the pdpD-2 set of primers with U112 DNA as the template relative to
the sizes for the other template products confirm the difference in the
DNA sequences. (B) Long-range PCR encompassing pdpD to iglD.
Primers were used to amplify a region of approximately 10 kb in strain
U112. The product in the LVS reaction is approximately 5.5 kb. Lane
MW, molecular weight markers. (C) Left to right, PCR products ob-
tained by using primers for iglA, iglB, iglC, and iglD. Molecular weight
markers are as in panel B. Lane water, PCR done with primers for iglA
but with no template. Similar reactions were done with the three other
sets of primers with identical results (data not shown).
VOL. 186, 2004FRANCISELLA PATHOGENICITY ISLAND 6433
was used as the template. Surprisingly, DNA from the LVS
showed no PCR amplicon for three different PCR primer pairs
for pdpD, while showing the expected amplicons for pdpA,
pdpB, pdpC, and iglC-D. To corroborate that a substantial
portion of pdpD was missing from the LVS strain, we per-
formed long-range PCR with primers that surrounded the
pdpD gene (Fig. 2B). With Schu4 DNA as the template a
9.9-kp product was generated; however, the LVS template
generated a 5.5-kb product. To test if the full iglABCD operon
was present in LVS, we performed further PCRs. LVS tem-
plate DNA generated products corresponding to the full
lengths of iglA (555 bp), iglB (1,545 bp), iglC (636 bp), and iglD
(1,197 bp) (Fig. 2C). DNA sequencing of this region in the
LVS strain showed that a 4,249-bp region is deleted from this
region. The deletion extends from 107 bp upstream of pmcA to
the first base pair of codon 980 of pdpD (Schu4 form; codon
1030 of the U112 form of pdpD; Fig. 1; see Fig. 4A). The
genome sequence of the LVS strain confirms this deletion and
shows that it occurs in both copies of the FPI in the LVS strain.
Since the LVS strain is an attenuated variant of a type B
strain, the question arose as to whether the absence of pdpD
represents a feature specific to the LVS or is a feature of type
B strains in general. To address this question, we examined the
DNA of four type B clinical isolates for the presence of pdpD.
As shown in Fig. 3, PCR amplification of three regions of pdpD
indicates that this gene is missing from the clinical isolates or
is significantly different in its nucleotide sequence. PCR am-
plification of other regions of the chromosome close to pdpD
suggests that the clinical isolate forms are very similar to those
found in the type strains, B38, LVS, and U112, of F. tularensis.
Disruption of pdpD results in a F. tularensis mutant defective
for intramacrophage growth and virulence in mice. The region
corresponding to the pdpD gene was replaced with an eryth-
romycin cassette by a recently described technique (25) (Fig.
4B). The resulting pdpD mutant, named JL12, still produced
IglA, albeit at a lower level than the wild type (Fig. 4C). This
mutant failed to grow in mouse bone marrow-derived macro-
phages (Fig. 4D). In two independent experiments, a total of
11 of 11 BALB/cByJ mice died within 7 days following intra-
FIG. 3. PCR analysis of clinical isolates of F. tularensis. The chro-
mosomes of four type B isolates were amplified with primers corre-
sponding to three regions of the pdpD gene as well as surrounding
areas of the chromosome. The orf5 locus is 4 kb to the left of pdpD as
shown in Fig. 1. These results show that pdpD is missing or substan-
tially different in type B clinical isolates compared to the form found in
type A strains or the F. tularensis subsp. novicida biotype. Lane water
is as defined for Fig. 2.
FIG. 4. Properties of the pdpD genomic region and phenotypes of pdpD mutant. (A) Relative sizes of the pdpD genes in F. tularensis strain
Schu4, F. tularensis subsp. novicida, and the LVS. (B) Recombinant constructs used to make the pdpD::Emrallelic replacement mutant JL12, as
described in the Materials and Methods. (C) Immunoblot of E. coli-produced, isolated recombinant IglA (lane 1) and IglA expressed in the pdpD
mutant, JL12 (lane 2), and wild-type U112 (lane 3). When normalized to the amount of total protein loaded per lane, the intensity of the IglA band
is sixfold higher in the U112 lane than in the JL12 lane. (D) Relative growth of wild-type F. tularensis subsp. novicida U112 and JL12 in bone
marrow-derived macrophages after 72 h. Representative data from one of three repetitions are shown.
6434 NANO ET AL. J. BACTERIOL.
dermal infections with 105cells of wild-type F. novicida U112.
In contrast, 12 of 12 mice survived intradermal infection with
105cells of JL12. Multiple attempts were made to complement
the pdpD mutation. However, we were unable to recover re-
combinant plasmids carrying pdpD in E. coli that were not
lethal in F. tularensis subsp. novicida.
Disruption of the pdpA gene renders F. tularensis unable to
grow in macrophages and avirulent in mice. We wished to test
whether one or more of the genes associated with the most
extreme low-G?C-content region of the FPI, pdpA through
pdpC, were needed for virulence. Because of the dramatic
difference between the G?C content of the rRNA genes and
that of the adjacent pdpA gene, a putative promoter region for
pdpA could be surmised. Hence, we inactivated pdpA of F.
tularensis subsp. novicida by transposon insertion and tested
the resulting mutant for virulence properties. Several inser-
tions in pdpA were found to diminish intramacrophage growth,
and one mutant, 304-2, was studied in detail. As seen in Fig. 5,
the F. tularensis parent strain (U112) and a random-insertion
mutant (R4) replicated exponentially in primary murine mac-
rophages. In contrast, the mutant 304-2 failed to grow in mac-
rophages. Efforts to complement this mutant were successful.
When the pdpA gene on plasmid pVIC314 was reintroduced to
F. tularensis strain 304-2, the wild-type intramacrophage
growth phenotype was restored. The same pattern was seen
with intradermal infections of mice. Doses of 2 ? 105CFU of
F. tularensis strains U112 and R4 (about 50 times the 50%
lethal dose) (24) were lethal for BALB/cByJ male mice; in
contrast, mice survived doses of 2 ? 107CFU of strain 304-2,
over 100-fold more bacteria (Fig. 6), demonstrating that the
pdpA mutant is extremely attenuated. When pdpA on plasmid
pVIC314 was reintroduced into F. tularensis strain 304-2, mice
succumbed to infection with a dose of 2 ? 107CFU. Experi-
ments in which 2 ? 105CFU of the pdpA mutant or the
complement strain were used for infection gave essentially
identical results as when 2 ? 107bacteria were used for infec-
tion, indicating that complementation was essentially com-
plete. Thus, pdpA is required for growth in macrophages and
virulence in mice.
There are multiple lines of evidence that the genomic region
described in this work can be classified as a pathogenicity
island. First, this work and others demonstrate that this region
contains a cluster of genes encoding virulence factors (16, 17).
Second, much of the region has DNA with a G?C content that
differs significantly from that of the rest of the F. tularensis
chromosome. Third, this area is surrounded by transposable
elements. Although transposable elements are common in F.
tularensis (21), there is good evidence that this region is actu-
ally mobile. Golovliov et al. (16) demonstrated that iglC is
duplicated in the LVS of F. tularensis, and the genome se-
quence data on the LVS strain show that the entire FPI region
is duplicated in that strain. This, in turn, suggests that this
genomic area can move to and from replicons other than the F.
tularensis genome. Very recent work shows that iglA, iglC,
pdpA, and pdpD are all regulated by MglA, suggesting that
these FPI genes are coordinately regulated to produce a viru-
lence phenotype in F. tularensis (26). Together, these features
of this genomic area justify applying the term pathogenicity
FIG. 5. Growth of F. tularensis pdpA mutant and control strains in
mouse macrophages. Bone marrow-derived macrophages were in-
fected, and samples were lysed every 24 h to determine the number of
viable bacteria (CFU) on agar medium. Squares, growth of parent
strain F. tularensis (U112); inverted triangles, control strain R4 with a
random transposon insertion not affecting intracellular growth; trian-
gles, mutant 304-2; diamonds, mutant 304-2 with complementing plas-
mid pVIC314; circles, mutant 304-2 with plasmid pVIC311, which
should not complement the genetic defect. Representative data from
one of three repetitions are shown.
FIG. 6. Infection of mice with F. tularensis pdpA mutant and con-
trol strains. Mice were infected intradermally with doses ranging from
2 ? 105to 2 ? 107CFU of the indicated bacteria; actual infection
doses were confirmed by retrospective plate count. Squares, survival of
mice following infection with 2 ? 105CFU of the parent strain of F.
tularensis (U112); inverted triangles, infection with 2 ? 105CFU of the
control strain R4, which has a random transposon insertion not affect-
ing intracellular growth; triangles, infection with 2 ? 107CFU of
mutant 304-2; diamonds, infection with 2 ? 107CFU of mutant 304-2
with complementing plasmid pVIC314; circles, infection with 2 ? 107
CFU of mutant 304-2 with plasmid pVIC311, which should not com-
plement the genetic defect. Although lines in this graph are offset
slightly for clarity, all mice survived (100% survival) infection with
mutant 304-2 and mutant 304-2 with the noncomplementing plasmid
pVIC311, and similarly all mice died (0% survival) following infection
with the wild type and mutant 304-2 with complementing plasmid
pVIC314. Representative data from one of four repetitions are shown.
VOL. 186, 2004FRANCISELLA PATHOGENICITY ISLAND6435
island. Thus, this work represents the first description of a Download full-text
cluster of virulence genes, and the first description of a patho-
genicity island, in F. tularensis.
There are numerous examples of large strain-to-strain dif-
ferences in virulence levels among microbial pathogens. For F.
tularensis, type A strains are thought to produce approximately
10% mortality in untreated human cases, and infections with
the type B strains are rarely fatal (10). Until now, there was no
known potential virulence factor that is present in the type A
strains that is not present in type B strains. The presence or
absence of pdpD may account, in part, for the strain-to-strain
difference in virulence. Strangely, the pdpD mutant F. tularen-
sis described here has a lower virulence in mice than the LVS
strain, which also lacks pdpD. Conceivably, the duplication of
the FPI in the LVS strain and the consequent increased dosage
of the FPI-encoded genes may partially compensate for the
absence of pdpD. Alternatively, the apparently polar effect on
transcription of the erythromycin resistance cassette insertion
may account for the lowered virulence phenotype in the pdpD
mutant, JL12. Interpreting the virulence role of the altered
form of pdpD in F. tularensis subsp. novicida is complicated by
the different endotoxin and O antigen that this strain possesses
(9). The endotoxin in particular clearly diminishes the ability of
F. tularensis subsp. novicida to grow in the macrophages of
some animals (9).
At present the most significant clue as to the function of IglA
and IglB comes from the study of putative homologues in
Rhizobium leguminosarum (6). In this species the IglAB homo-
logues are thought to be needed for secretion of certain pro-
teins. Hence, one possible hypothesis for the role of IglAB in
F. tularensis is that they have a role in secretion of PdpABCD
and perhaps other proteins.
This work was supported in part by a grant from the Natural Sci-
ences and Engineering Council of Canada to F.E.N.
We are grateful to the members of the F. tularensis genome sequenc-
ing consortium for granting access to unpublished DNA sequence
data. We thank A. Kari for use of facilities; P. Keeling, D. Burns, T.
Pearson, and E. S. Stibitz for critical reading of the manuscript; and the
C. Upton group and R. Roper for assistance with bioinformatics anal-
1. Altschul, S. F., T. L. Madden, A. A. Schaffer, J. Zhang, Z. Zhang, W. Miller,
and D. J. Lipman. 1997. Gapped BLAST and PSI-BLAST: a new generation
of protein database search programs. Nucleic Acids Res. 25:3389–3402.
2. Amer, A. O., and M. S. Swanson. 2002. A phagosome of one’s own: a
microbial guide to life in the macrophage. Curr. Opin. Microbiol. 5:56–61.
3. Anthony, L. D., R. D. Burke, and F. E. Nano. 1991. Growth of Francisella spp.
in rodent macrophages. Infect. Immun. 59:3291–3296.
4. Anthony, L. S. D., M. Z. Gu, S. C. Cowley, W. W. Leung, and F. E. Nano.
1991. Transformation and allelic replacement in Francisella spp. J. Gen.
5. Baron, G. S., and F. E. Nano. 1998. MglA and MglB are required for the
intramacrophage growth of Francisella novicida. Mol. Microbiol. 29:247–259.
6. Bladergroen, M. R., K. Badelt, and H. P. Spaink. 2003. Infection-blocking
genes of a symbiotic Rhizobium leguminosarum strain that are involved in
temperature-dependent protein secretion. Mol. Plant-Microbe Interact. 16:
7. Bosio, C. M., and K. L. Elkins. 2001. Susceptibility to secondary Francisella
tularensis live vaccine strain infection in B-cell-deficient mice is associated
with neutrophilia but not with defects in specific T-cell-mediated immunity.
Infect. Immun. 69:194–203.
8. Cowley, S. C., C. J. Gray, and F. E. Nano. 2000. Isolation and characteriza-
tion of Francisella novicida mutants defective in lipopolysaccharide biosyn-
thesis. FEMS Microbiol. Lett. 182:63–67.
9. Cowley, S. C., S. V. Myltseva, and F. E. Nano. 1996. Phase variation in
Francisella tularensis affecting intracellular growth, lipopolysaccharide anti-
genicity and nitric oxide production. Mol. Microbiol. 20:867–874.
10. Dennis, D. T., T. V. Inglesby, D. A. Henderson, J. G. Bartlett, M. S. Ascher,
E. Eitzen, A. D. Fine, A. M. Friedlander, J. Hauer, M. Layton, S. R. Lil-
libridge, J. E. McDade, M. T. Osterholm, T. O’Toole, G. Parker, T. M. Perl,
P. K. Russell, and K. Tonat. 2001. Tularemia as a biological weapon: medical
and public health management. JAMA 285:2763–2773.
11. Ellis, J., P. C. Oyston, M. Green, and R. W. Titball. 2002. Tularemia. Clin.
Microbiol. Rev. 15:631–646.
12. Gaillard, J. L., P. Berche, C. Frehel, E. Gouin, and P. Cossart. 1991. Entry
of L. monocytogenes into cells is mediated by internalin, a repeat protein
reminiscent of surface antigens from gram-positive cocci. Cell 65:1127–1141.
13. Galan, J. E. 1996. Molecular genetic bases of Salmonella entry into host cells.
Mol. Microbiol. 20:263–271.
14. Golovliov, I., V. Baranov, Z. Krocova, H. Kovarova, and A. Sjostedt. 2003.
An attenuated strain of the facultative intracellular bacterium Francisella
tularensis can escape the phagosome of monocytic cells. Infect. Immun.
15. Golovliov, I., M. Ericsson, G. Sandstrom, A. Tarnvik, and A. Sjostedt. 1997.
Identification of proteins of Francisella tularensis induced during growth in
macrophages and cloning of the gene encoding a prominently induced 23-
kilodalton protein. Infect. Immun. 65:2183–2189.
16. Golovliov, I., A. Sjostedt, A. Mokrievich, and V. Pavlov. 2003. A method for
allelic replacement in Francisella tularensis. FEMS Microbiol. Lett. 222:273–
17. Gray, C. G., S. C. Cowley, K. K. Cheung, and F. E. Nano. 2002. The
identification of five genetic loci of Francisella novicida associated with in-
tracellular growth. FEMS Microbiol. Lett. 215:53–56.
18. Groisman, E. A., and H. Ochman. 1993. Cognate gene clusters govern
invasion of host epithelial cells by Salmonella typhimurium and Shigella
flexneri. EMBO J. 12:3779–3787.
19. Haas, R., A. F. Kahrs, D. Facius, H. Allmeier, R. Schmitt, and T. F. Meyer.
1993. TnMax-a versatile mini-transposon for the analysis of cloned genes and
shuttle mutagenesis. Gene 130:23–31.
20. Isberg, R. R., D. L. Voorhis, and S. Falkow. 1987. Identification of invasin: a
protein that allows enteric bacteria to penetrate cultured mammalian cells.
21. Johansson, A., I. Goransson, P. Larsson, and A. Sjostedt. 2001. Extensive
allelic variation among Francisella tularensis strains in a short-sequence tan-
dem repeat region. J. Clin. Microbiol. 39:3140–3146.
22. Karlsson, J., R. G. Prior, K. Williams, L. Lindler, K. A. Brown, N. Chatwell,
K. Hjalmarsson, N. Loman, K. A. Mack, M. Pallen, M. Popek, G. Sand-
strom, A. Sjostedt, T. Svensson, I. Tamas, S. G. Andersson, B. W. Wren,
P. C. Oyston, and R. W. Titball. 2000. Sequencing of the Francisella tularensis
strain Schu 4 genome reveals the shikimate and purine metabolic pathways,
targets for the construction of a rationally attenuated auxotrophic vaccine.
Microb. Comp. Genomics 5:25–39.
23. Keen, N. T., S. Tamaki, D. Kobayashi, and D. Trollinger. 1988. Improved
broad-host-range plasmids for DNA cloning in gram-negative bacteria. Gene
24. Kieffer, T. L., S. C. Cowley, F. E. Nano, and K. L. Elkins. 2003. Francisella
novicida LPS has greater immunobiological activity in mice than F. tularensis
LPS, and contributes to F. novicida murine pathogenesis. Microbes Infect.
25. Lauriano, C. M., J. R. Barker, F. E. Nano, B. P. Arulanandam, and K. E.
Klose. 2003. Allelic exchange in Francisella tularensis using PCR products.
FEMS Microbiol. Lett. 229:195–202.
26. Lauriano, C. M., J. R. Barker, S.-S. Yoon, F. E. Nano, B. P. Arulanandam,
D. J. Hassett, and K. E. Klose. 2004. MglA regulates transcription of viru-
lence factors necessary for Francisella tularensis intraamoebae and intram-
acrophage survival. Proc. Natl. Acad. Sci. USA 101:4246–4249.
27. Nakashima, H., S. Fukuchi, and K. Nishikawa. 2003. Compositional changes
in RNA, DNA and proteins for bacterial adaptation to higher and lower
temperatures. J. Biochem. (Tokyo) 133:507–513.
28. Prior, R. G., L. Klasson, P. Larsson, K. Williams, L. Lindler, A. Sjostedt, T.
Svensson, I. Tamas, B. W. Wren, P. C. Oyston, S. G. Andersson, and R. W.
Titball. 2001. Preliminary analysis and annotation of the partial genome
sequence of Francisella tularensis strain Schu 4. J. Appl. Microbiol. 91:614–
29. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning, 2nd
ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.
30. Wang, R. F., and S. R. Kushner. 1991. Construction of versatile low-copy-
number vectors for cloning, sequencing and gene expression in Escherichia
coli. Gene 100:195–199.
6436NANO ET AL.J. BACTERIOL.