|Home | About | Journals | Submit | Contact Us | Français|
A comprehensive TnphoA mutant library was constructed in Yersinia pestis KIM6 to identify surface proteins involved in Y. pestis host cell invasion and bacterial virulence. Insertion site analysis of the library repeatedly identified a 9,042-bp chromosomal gene (YPO3944), intimin/invasin-like protein (Ilp), similar to the Gram-negative intimin/invasin family of surface proteins. Deletion mutants of ilp were generated in Y. pestis strains KIM5(pCD1+) Pgm− (pigmentation negative)/, KIM6(pCD1−) Pgm+, and CO92. Comparative analyses were done with the deletions and the parental wild type for bacterial adhesion to and internalization by HEp-2 cells in vitro, infectivity and maintenance in the flea vector, and lethality in murine models of systemic and pneumonic plague. Deletion of ilp had no effect on bacterial blockage of flea blood feeding or colonization. The Y. pestis KIM5 Δilp strain had reduced adhesion to and internalization by HEp-2 cells compared to the parental wild-type strain (P < 0.05). Following intravenous challenge with Y. pestis KIM5 Δilp, mice had a delayed time to death and reduced dissemination to the lungs, livers, and kidneys as monitored by in vivo imaging using a lux reporter system (in vivo imaging system [IVIS]) and bacterial counts. Intranasal challenge in mice with Y. pestis CO92 Δilp had a 55-fold increase in the 50% lethal dose ([LD50] 1.64 × 104 CFU) compared to the parental wild-type strain LD50 (2.98 × 102 CFU). These findings identified Ilp as a novel virulence factor of Y. pestis.
The genus Yersinia is comprised of 12 species, three of which, Yersinia enterocolitica, Yersinia pseudotuberculosis, and Yersinia pestis, are pathogenic for humans and rodents (5). Y. pestis, the causative agent of plague, is typically transmitted subcutaneously to humans by the bite of an infected flea, causing either bubonic or septicemic plague, but can also be transmitted by aerosols, causing pneumonic plague (25). Pathogenesis of Y. pestis is dependent on the presence of three plasmids: pCD1 encoding a type III secretion system, pPCP1 encoding plasminogen activator (Pla), and pMT1 encoding the capsular antigen fraction 1 (Caf1) (25). Y. pestis has been considered an extracellular pathogen because it contains mutations in two major invasin and adhesin homologues (Inv and YadA) found in invasive Y. pseudotuberculosis (24, 29). However, recent studies demonstrate that Y. pestis is able to invade eukaryotic cells mediated in part by Pla, capsular F1 antigen, OmpX (Ail), and Psa fimbriae (3, 18, 20, 21). None of these factors alone confers the full invasion phenotype.
To identify additional Y. pestis surface proteins that may contribute to invasion and virulence, a comprehensive TnphoA insertion library of avirulent Y. pestis KIM6(pCD1−) Pgm+ (pigmentation positive) was generated. This avirulent strain was used because it lacks the well-characterized type III secretion system and effector Yop genes on pCD1, and therefore the phoA fusions would be biased toward uncharacterized chromosomal loci. The DNA sequence of transposon insertion junctions was determined, and insertionally inactivated gene DNA sequences were compared to bacterial DNA sequence databases. In this screen, DNA sequence analysis repeatedly identified insertions in a large gene of 9,042 bp (YPO3944) which we designated ilp (intimin/invasin-like protein), with predicted similarity to the class of Gram-negative bacterial intimin/invasin/autotransporter proteins. Because computer analysis of Ilp predicted a three-dimensional structure similar to the C-type lectin-like domain found in the C-terminal region of Y. pseudotuberculosis invasin protein (11), we hypothesized that Ilp might contribute to the pathogenesis of Y. pestis by promoting the bacterial association with and/or internalization by arthropod or mammalian host cells. The purpose of this study was to test this hypothesis. Initial in vitro phenotypic analyses of ilp mutants were conducted in KIM6(pCD1−) Pgm+ (deleted type III secretion) and KIM5(pCD1+) Pgm− (partially attenuated strain), and initial in vivo virulence was analyzed in Y. pestis KIM5 because these strains can be used under biosafety level two (BSL2) conditions. The results of ilp contribution to virulence in Y. pestis KIM5 were confirmed by construction of the ilp deletion in the fully virulent Y. pestis CO92 strain, and in vivo virulence experiments were conducted under BSL3 containment.
Y. pestis and Escherichia coli strains used in this study are listed in Table 1. Y. pestis KIM5 strains were grown in brain heart infusion broth (BHIB) or on agar plates (BHIA) containing 2.5 mM CaCl2 in the presence of kanamycin (Km) and chloramphenicol (Cm) as appropriate. Y. pestis CO92 strains were grown in tryptone blood agar plates. E. coli strains were grown in Luria Bertani (LB) broth or on LB agar plates in the presence of Km (50 μg ml−1) as appropriate.
A luciferase reporter plasmid, pACYC177-pflhDC-lux was constructed to follow Y. pestis in vivo mouse infections. The lux operon from Vibrio harveyi, cloned into the shuttle vector RY111, was a gift of Mike Konkel (Washington State University). The lux operon (luxDCABE) contained on a 6-kb SacI-BamHI fragment was subcloned into plasmid pACYC177 (Kmr), and the promoter for the Yersinia master control gene for motility, pflhDC, contained within a 400-bp BamHI fragment was inserted upstream of the lux operon. This construct was transformed into Y. pestis strains used in this study, and bioluminescent colonies were detected on LB agar plates using an in vivo imaging system (IVIS). Cultures of Y. pestis were grown overnight in LB broth, and decimal dilutions were prepared for cell count determination and bioluminescence detection using a SprectraMax L plate reader to determine sensitivity levels. Additionally, this overnight culture was diluted 100-fold into fresh medium without antibiotic selection sequentially over 10 days, the normal period of our animal challenge experiments, to determine plasmid stability and maintenance of light emission.
Y. pestis KIM6 TnphoA insertion mutants were constructed by mating an E. coli S17-1 nalidixic acid-sensitive (Nals) strain carrying suicide plasmid pUT (TnphoA) with strains of Y. pestis KIM6 that were nalidixic acid resistant and kanamycin sensitive (Nalr Kms). Conjugations were conducted at room temperature by spotting equal volumes of overnight cultures of donor and recipient cells on LB agar for 6- to 12-h incubations. Cells were harvested with a sterile toothpick and resuspended in phosphate-buffered saline (PBS), washed two times with PBS, diluted, and plated to yield 300 to 600 colonies per plate of LB Nal/Km medium supplemented with 40 μg ml−1 XP ([5-bromo-4-chloro-3-indolylphosphate] alkaline phosphatase indicator). Six independent conjugations were conducted to reduce sibling isolations of phoA fusions, and cells from each conjugation were plated and incubated at 25°C or 37°C. After 48 h, blue colonies were picked and purified by streak plating on selective indicator plates, and the PhoA phenotype was confirmed. The transposon insertion junctions were determined by DNA sequence analysis of PCR amplicons from individual mutants essentially as described by Jacobs et al. (17) using their series of phoA primer sequences and degenerate primers for flanking sequence identification.
Total RNA was extracted from Y. pestis KIM5 and KIM6+ cultures grown at 28°C or 37°C using TRIzol (Invitrogen) and Lysing Matrix B (Q-BIOgene) according to the manufacturers' instructions. The first strand of cDNA was synthesized from 1 μg of total RNA using random hexamer (Roche) and RNA Superscript II (Invitrogen). Primers for the quantitative reverse transcription-PCR (qRT-PCR) (Ilp1F/R and 16SF/R) were designed by Primer Express, version 2.0 (Applied Biosystems) (Table 2). The qRT-PCR was performed using the SYBR green I dye master mix (Applied Biosystems) and ABI Prism 7500 Real Time PCR System (Applied Biosystems) according to the manufacturer's instructions. The threshold cycle (CT) was calculated by Sequence Detector Systems software, version 1.2.2 (Applied Biosystems), when the cycle number at which the ΔRn (normalized reporter) crossed the baseline. The transcriptional level of each target gene was calculated using a following formula: −ΔCT = −[(CT of ilp) − (CT of the internal control)], where the internal control is the 16S rRNA gene.
To generate the Y. pestis KIM5 ilp deletion mutant, a 1,642-bp DNA fragment spanning a region 5′ to the ilp gene was amplified using Ilp7F/Ilp7R (Ilp7F/R) primers (Table 2) and digested with SmaI and XbaI. A 2,464-bp DNA fragment spanning a region 3′ to the ilp gene was amplified using Ilp8F/R primers (Table 2). Digested amplicons were cloned into pMS20 to generate pJW1, which was transformed into E. coli CC118 λpir. A 1,527-bp DNA fragment encoding a Km resistance cassette flanked with a flippase recognition target (FRT) was amplified from pKD4 (8) using NptF/R primers (Table 2), digested with PsiI and SfiI, and cloned into pJW1. This modified construct, designated pJW2 was used to transform E. coli S17-1 λpir. The E. coli S17-1(pJW2) λpir transformant was mated with Y. pestis KIM5 or KIM6 as described previously (33). The merodiploid strains were generated by single homologous recombination and served as complemented strains (ilp+/ilp::npt strains). The second homologous recombination was promoted by culturing the merodiploid strain on LB agar plates containing 5% sucrose. The deletion mutation was confirmed by PCR and qRT-PCR (data not shown).
Y. pestis CO92 Δilp was generated using λ-Red recombination with modification (8). Briefly, the Km resistance cassette was amplified from pKD4 using primers containing 36-nucleotide homology extensions 5′ and 3′ to the open reading frame of the ilp gene (Ilp9F/R) and transformed into Y. pestis CO92 carrying pJS06, a derivate of pKD46 containing the Red recombinase and the levansucrase gene (sacB) (23). Recombinants were selected on BHIA containing Km and verified by PCR. The pJS06 plasmid was cured from this recombinant by culturing on BHIA containing 5% sucrose and Km. The deletion mutation was confirmed by PCR and RT-PCR (data not shown).
Y. pestis KIM5 wild-type, Y. pestis KIM5 Δilp, and Y. pestis ilp+/ilp::npt strains and two Y. pestis KIM6 ilp::TnphoA fusions were used to determine if Ilp was localized to the bacterial outer membrane. Antibody against the N-terminal fragment of Ilp was used for immunization of rats. Animals were injected three times over the course of 6 weeks with 0.5 mg of purified fragment emulsified in Freund's incomplete adjuvant. Antiserum was prepared from harvested whole blood and stored at −80°C. Y. pestis KIM5 outer membrane preparations from the three strains above were prepared by growing cells to mid-exponential phase in 100 ml of LB broth, concentrating the cells by centrifugation, resuspending the cells in 10 mM Tris-EDTA buffer, and passing the cells through a French press. The membrane fraction from this lysis preparation was concentrated by centrifugation, and the pellet was solubilized with lysis buffer [7 M urea, 2 M thiourea, 4% CHAPS (3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate), 1% Triton, and 65 mM dithiothreitol (DTT)]. Protein samples (100 μg) from each membrane preparation were loaded in wells of a 10% SDS-PAGE gel. The proteins were immunoblotted, and Ilp was detected using a 1:1,000 dilution of anti-Ilp antibody followed by anti-rat-labeled alkaline phosphatase antibody.
Ilp protein localization was also confirmed using two TnphoA fusions to Ilp. Outer membrane preparations were prepared from Y. pestis KIM6, the ilp deletion mutant, and two ilp::TnphoA insertion strains. Overnight cultures were grown at 37°C in 10 ml of LB broth, and cells were pelleted by centrifugation for 5 min at 5,000 × g. These cells were washed two times in physiological saline, then suspended in 1 ml of water, placed at −80°C for 1 h, thawed, and lysed by sonication for 12 s on setting 3 (Sonifier 150; Branson). Benzonase nuclease (1 μl; Novagen) was added to each tube and incubated with mixing at room temperature (RT) for 20 min. An equal volume of 2% (wt/vol) N-laurolylsarcosine (Sarkosyl; Sigma) in 20 mM HEPES, pH 7.4, was added, and samples were mixed at RT for 30 min. Outer membranes were pelleted at 14,000 × g for 3 h. Pellets were resuspended in water, protein concentrations were determined, and equal aliquots of each sample were separated by 10% SDS-PAGE on duplicate gels. One gel was stained with Coomassie R-250 (loading control), and the other gel was developed directly for alkaline phosphatase. Alkaline phosphatase in this second equivalent gel was detected directly by washing gels for 10 min in water, followed by exposure to 0.015% (wt/vol) 5-bromo-4-chloro-3-indolylphosphate/0.03% (wt/vol) nitroblue tetrazolium in Western developing buffer (0.1 M sodium bicarbonate, 1 mM magnesium chloride, pH 9.8) until blue bands developed. The gels were then stored in 5% acetic acid in the dark until scanned for documentation.
HEp-2 cells were grown in 12-well plates (Corning) using complete growth medium consisting of high-glucose Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum (HyClone) and 100 U ml−1 penicillin-streptomycin (Invitrogen) at 37°C under 5% CO2. HEp-2 cell monolayers (approximately 1 × 106 cells/well) were washed three times with invasion medium (complete growth medium except antibiotics) and coincubated with Y. pestis grown at 28°C or 37°C (5 × 106 CFU), washed, and resuspended in invasion medium for 1 h.
For adherence assays, cocultures were washed five times with PBS and detached using TrypLE express (Invitrogen) at 37°C for 5 min. Cells were lysed using 0.0025% (vol/vol) Triton X-100 solution (in deionized water). The number of bacteria associated with cells was determined by a serial dilution of cell lysates and plating on BHIA in triplicate. For invasion assays, cocultures were washed three times with PBS and treated with gentamicin (100 μg ml−1 in invasion medium) for 1 h. The number of bacteria internalized into HEp-2 cells was determined as described above.
For immunofluorescence assays, cocultures were incubated with polyclonal antibody specific to Y. pestis for 15 min, followed by coincubation with phycoerythrin-conjugated isotype-specific goat anti-IgG antibody for 15 min. Bacteria adhering to HEp-2 cells were observed under a fluorescence microscope (magnification, × 200; Nikon).
Xenopsylla cheopis fleas were infected by allowing them to feed on blood containing approximately 3.2 × 108 CFU ml−1 of Y. pestis KIM6 and Y. pestis KIM6 Δilp strains, using a previously described artificial feeding system (12, 13). Approximately equal numbers of male (n = 54) and female (n = 52) fleas that took a blood meal were maintained at 21°C and 75% relative humidity, fed twice weekly on uninfected mice for 4 weeks, and monitored for proventricular blockage as previously described (10, 12, 13). The infection rate was determined by counting the bacterial load in samples of 20 individual females collected immediately after the infectious blood meal and at 28 days postinfection (12, 13).
All animal experiments were approved by the University of Idaho Animal Use Committee (protocol number 2010-014). Specific-pathogen-free BALB/c mice (6 to 8 weeks old) were obtained from Simonsen Laboratories and housed in an air-filtered barrier unit (BioZone MiniRack). Y. pestis KIM5 strains were grown in BHIB at 28°C on a platform shaker at 250 rpm, washed, and resuspended in PBS. Mice were intravenously injected with 100 μl of Y. pestis bacteria in PBS. In some experiments, mice were anesthetized using isoflurane, and the progression of infection was monitored using an IVIS Lumina system (Califer). To monitor the bacterial load in specific organs during the course of infection, mice were humanely euthanized with an overdose of sodium pentobarbital (150 mg kg−1). The lungs, livers, spleens, and kidneys were surgically removed, weighed, and homogenized in 1 ml of PBS. The number of Y. pestis bacteria in the homogenates of each organ was determined using triplicate plate counts, and these counts were reported as CFU g−1 of tissue.
Y. pestis CO92 strains were grown on BHIA at 28°C, harvested using a sterile policeman, resuspended using PBS, filtered using nylon wool to remove clustered bacterial cells, aliquoted in 1-ml amounts, and stored at −80°C in the presence of glycerol (25%, vol/vol). Mice were anesthetized using isofluorine and intranasally instilled with 10 μl of serial dilutions of the Y. pestis frozen stock cultures. The 50% lethal dose (LD50) was determined by the method of Reed and Muench (27).
A genome-wide random TnphoA insertion library was generated in Y. pestis KIM6 Pgm+ essentially as described for Pseudomonas aeruginosa (17). Insertion of mini-TnphoA confers kanamycin resistance and creates a blue colony on indicator plates when the insertion element is positioned in frame in genes encoding surface or exported proteins. Six separate conjugations were performed, and blue colonies were isolated from indicator plates incubated at either 25°C or 37°C to identify surface proteins regulated by temperature. DNA sequence analysis was performed on isolated blue colonies to identify insertion junctions and the disrupted gene (17). Over 1,200 individual phoA fusion constructs were isolated, of which 612 were sequenced. Of the isolates sequenced, 239 unique genes were identified, indicating approximately 2.6-fold genome coverage. In addition, the majority the surface-associated chromosomal virulence genes previously characterized were identified. A summary of the surface proteins identified is given in Table 3, and individual genes are listed in Table S1 in the supplemental material.
The most frequently identified single gene from both the 25°C and 37°C primary PhoA screens, identified 33 times in total, was gene YPO3944. This gene has a predicted 9,042-bp open reading frame encoding a protein of 3,014 amino acids with predicted sequence similarity to Gram-negative bacterial invasins (Fig. 1A). The Ilp protein showed structural similarity to invasin and intimin found in Y. pseudotuberculosis and Escherichia coli O157:H7, respectively. The predicted structure included a lysin motif (LysM) domain and repetitive bacterial immunoglobulin-like domains (BID) (Fig. 1A) (2, 24). Furthermore, the predicted three-dimensional structure of the C-terminal region of Ilp had similarity to the C-type lectin-like domain found in the C-terminal region of the Y. pseudotuberculosis invasin protein (InvA), showing interspersed α-helical and β-stranded regions (Fig. 1B and andC).C). Previous studies demonstrated that the C-terminal 479 amino acids of the Y. pseudotuberculosis invasin protein are exposed on the bacterial surface and interact with multiple β-chain integrin receptors on the host cell surface to promote cell association and internalization by host cells (11). For these reasons, we chose to study this gene in further detail.
Y. pestis cycles between its arthropod flea vector and mammalian hosts. Temperature (37°C) is a key environmental cue during this transition. Expression of many Y. pestis virulence factors, including but not limited to the type III secretion system and capsular F1 antigen, is regulated by temperature (30). To examine if ilp is temperature regulated, alkaline phosphatase assays were conducted with 15 independently isolated ilp-phoA fusions grown at 28°C and 37°C. We found significant variation in temperature-dependent levels of PhoA expression between different isolates; however, the general trend was increased levels of expression at 37°C, ranging from 2- to 9-fold, depending on which insertion mutant was examined (data not shown). To more precisely quantify transcription of the ilp gene, transcript levels from the two different culture conditions (28°C and 37°C) were measured using quantitative real-time RT-PCR (qRT-PCR) using RNA isolated from both Y. pestis KIM5 and KIM6 strains. The transcription of ilp at 37°C was 7.77 and 6.64 times higher than at 28°C in Y. pestis strains KIM5 Pgm− and KIM6 Pgm+, respectively (Fig. 2). Consistent with other identified Yersinia temperature-regulated genes, we noted that the promoter region of Ilp contains runs of As and Ts, and this region displayed intrinsic DNA bending by a two-dimensional gel assay (28; also data not shown).
Outer membrane preparations from Y. pestis CO92(pCD1−), Y. pestis CO92(pCD1−) Δilp::npt, and Y. pestis ilp+/ilp::npt were transferred to nitrocellulose membranes. These blots were developed with anti-Ilp rat antibody followed by anti-rat alkaline phosphatase antibody. A large protein (ca. 380,000 kDa) was observed in the two control strains but not the deletion mutant (Fig. 3). Additionally, outer membrane preparations from two independent Y. pestis KIM6 ilp::TnphoA fusion strains were compared to outer membrane preparations from Y. pestis KIM6 and the Y. pestis KIM6 Δilp strain. Outer membrane protein gels were stained directly for alkaline phosphatase activity, and truncated fusion proteins with enzyme activity were present only in the Ilp-PhoA fusion strains (data not shown). These data confirm that Ilp is a surface-associated protein.
Structural similarity to bacterial invasin and intimin suggested that Ilp might be involved in the interaction of Y. pestis with host cells. To test this, we compared Y. pestis KIM5 to an isogenic ilp deletion mutant (Δilp) and its complement for bacterial association with and internalization by HEp-2 cells. The association of Y. pestis KIM5 Δilp (grown at 28°C or 37°C) with HEp-2 cells was significantly lower (approximately 1.8 or 3.7 times at 28°C and 37°C, respectively; P < 0.01) than that of the parental wild-type or the Δilp-complemented strains (Fig. 4). These findings were further supported by immunofluorescence microscopy assays showing that a phycoerythrin signal, representing HEp-2-adhering bacteria, was noticeably weaker with Y. pestis KIM5 Δilp than that with the parental wild-type or the complemented strain (data not shown). Similar to these host cell association results, internalization of Y. pestis KIM5 Δilp grown at 28°C was also slightly but significantly lower (approximately 2.0 times; P < 0.01) than that of the parental wild type or the complemented strain (Fig. 4).
The above results showed that ilp expression demonstrated modest temperature regulation and that Ilp contributed to bacterial association with and internalization by mammalian host cells. Because Y. pestis cycles through an arthropod vector, we analyzed the Ilp effect on bacterial colonization of Xenopsylla cheopis (flea vector). Flea-feeding trials were conducted to compare bacterial flea gut blockage and long-term infection rates between Y. pestis KIM6 Pgm+Δilp and wild-type KIM6 Pgm+. The Y. pestis KIM6 Pgm+Δilp strain showed 35% blockage rate in fleas maintained at 21°C for 28 days. Similar blockage rates ranging from 21% to 38% were observed with the parental Y. pestis KIM6+ wild type. Longer-term Y. pestis infection and survival rates in unblocked fleas were determined by plate counts at 28 days after bacterial infection The Y. pestis KIM6 Pgm+Δilp strain showed approximately 94% of colonization rates, with an average bacterial load of 5.9 × 105 CFU/flea. Similar colonization rates (75% to 100%) and bacterial loads (4.8 × 105 to 5.8 × 105 CFU/flea) were observed with the Y. pestis KIM6+ parental wild-type strain. These results showed that loss of Ilp does not affect the ability of Y. pestis to cause flea gut blockage or long-term flea colonization.
Previous work from our laboratory and from others showed that the Y. enterocolitica flhDC operon is constitutively and highly expressed at 25°C and 37°C (15, 22). When the flhDC promoter was positioned in front of the lux operon carried on pACYC177 and transformed into Y. pestis KIM5 Pgm−, bioluminescent colonies were easily detected, as exemplified in Fig. 5B. One strongly bioluminescent colony was picked and purified. Sensitivity assays showed that 1 × 103 cells could be detected with a luminometer plate reader and that 1 × 108 cells produced ~1 million relative luminescent units (RLU), suggesting that light emission may be of sufficient intensity to detect cells in mice using the in vivo imaging system (IVIS) (Fig. 5C).
To determine plasmid stability and maintenance, Y. pestis pACYC177-lux was grown overnight at three temperatures (18°C, 25°C, and 37°C), and 100-fold dilutions were made into fresh LB medium daily without antibiotic selection for 10 days, the time equivalent for conducting a mouse challenge experiment. Colony count determinations showed no significant difference between counts made on LB and LB Km (40 μg ml−1) plates, nor did temperature affect plasmid stability. However, there was an approximate 15% loss in cellular bioluminescence over 4 days and approximately 2-fold loss over this 10-day period, as exemplified from a representative plate (Fig. 5B) from the final day of the 10-day experiment depicted in Fig. 5A. These results show that there is modest selection over time against lux expression with this reporter plasmid.
To determine if ilp is required for virulence, we used a murine infection model for morbidity and mortality assays. Y. pestis KIM5 Pgm− is attenuated for both intramuscular (i.m.) and intraperitoneal (i.p.) routes of infection but is lethal when introduced by intravenous (i.v.) injection (25). Groups of mice (n = 12) were intravenously challenged by tail vein injection of Y. pestis KIM5 Pgm−, the isogenic Δilp strain, and the Δilp-complemented strain. Each strain was transformed with pACYC177-lux to monitor infection progression daily with an IVIS Lumina system (Caliper Life Sciences, Hopkinton, MA). Challenge doses of 1 × 105 or 1 × 106 CFU were employed. Figure 6 depicts a representative subset of four animals infected with each strain on days 1 and 3 postinfection from this experiment. None of the mice challenged with 1 × 105 CFU Y. pestis KIM5 Δilp showed visible signals of infection at 24 h or 72 h (Fig. 6A, Δilp). In contrast, 6 out of 12 mice challenged with 1 × 105 CFU of the parental wild-type strain and 7 out of 12 mice challenged at this dose with the Δilp-complemented strain had clear signals of infection by day 3 (Fig. 6A, WT and Comp). All mice infected at this dose with all three strains succumbed to infection; however, there was a significant delay in time to death (24 to 36 h) among mice receiving the Y. pestis Δilp mutant (Fig. 6C). With the higher challenge dose of 1 × 106 CFU, 8 out of 12 mice receiving Y. pestis KIM5 Δilp had initial visible signs of infection that disappeared in 5 to 6 days postchallenge (Fig. 6B, Δilp, day 3; also data not shown). In contrast, all 12 mice injected with the parental wild-type strain and with complemented strain (1 × 106 CFU) had evidence of systemic infection by day 3, as shown by the representative subsample of this group (Fig. 6B, WT and Comp, day 3). Mice in all three groups succumbed to infection, but again there was a delayed time to death (1 to 2 day) in mice challenged with the Y. pestis KIM5 Δilp strain compared to the Y. pestis KIM5 wild-type strain (Fig. 6C). The ilp-complemented strain showed the equivalent time-to-death pattern observed in the parental wild-type strain (data not shown).
These experiments suggested that the progression of infection was altered by deletion of ilp. To assess the effect of Ilp on the systemic spread of Y. pestis KIM5, Y. pestis KIM5 wild-type, Y. pestis Δilp, and the Y. pestis Δilp-complemented strains were used. Mice were intravenously injected with 1 × 105 CFU. After 3 days, the numbers of CFU in various extracted and homogenized organs were determined using triplicate plate counts. The number of Y. pestis KIM5 Δilp bacteria recovered from spleen tissue was lower but not significantly different from that recovered from the parental wild-type or Δilp-complemented strain (Fig. 7D). However, the number of Y. pestis KIM5 Δilp bacteria colonizing lung, liver, and kidney tissue was significantly lower than that of the parental wild-type and complemented strains (P < 0.05) (Fig. 7A, ,B,B, And AndC).C). This suggested that Y. pestis was initially cleared from the blood by the spleen and that deletion of ilp delayed subsequent systemic bacterial spread to other organs.
The reduced spread to the various organs and delayed time to death of Y. pestis KIM5Δilp suggested that Ilp affected the virulence of Y. pestis. To test this, the bacterial LD50 in BALB/c mice was determined for intranasal challenge with the fully virulent Y. pestis CO92 strain and a Y. pestis CO92 Δilp mutant. Mice received increasing decimal doses (1.2 × 102 through 1.2 × 105 CFU) of the fully virulent Y. pestis CO92 strain or an isogenic Y. pestis CO92 Δilp mutant. The inocula for these experiments were harvested from tryptone blood agar plates (32), and the number of CFU ml−1 was determined before cryogenic storage. The number of CFU ml−1 was also repeated if inocula were prepared from frozen stocks. Mice (n = 8 per group) were monitored twice daily for the development of disease symptoms and survival. After 2 days, all mice receiving the two highest doses (1.2 × 104 or 1.2 × 105 CFU) of Y. pestis CO92 wild type showed visible signs of infection, including lethargy, hunched back, labored breathing, rough fur, and huddling. All mice in these two dosage groups succumbed to infection by day 4 postchallenge (Fig. 8A). Five out of eight (62.5%) mice exposed to 1.2 × 103 CFU of Y. pestis CO92 wild type succumbed to the infection after 6 days. Interestingly, one out of eight mice infected with 1.2 × 102 CFU of Y. pestis CO92 wild type succumbed to the infection without any visible disease symptoms at 5 days postchallenge. In contrast, none of the mice challenged with 5 × 102 or 5 × 103 CFU of Y. pestis CO92 Δilp had visible disease symptoms, and all mice in these challenge groups survived (Fig. 8B). Six of eight mice exposed to 5 × 104 CFU of Y. pestis CO92 Δilp died by day 5 postchallenge. Mice challenged with 5 × 105 CFU of Y. pestis CO92 Δilp succumbed to infection at 7 days postchallenge. The LD50s of Y. pestis CO92 and Y. pestis CO92 Δilp were calculated as 2.98 × 102 and 1.64 × 104 CFU, respectively. Thus, the deletion of ilp resulted in a 55-fold LD50 increase.
Consistent with previous reports (19), bacterial growth in the lungs and spleens gradually increased during infection with the CO92 wild-type strain (Fig. 9). However, bacterial growth rates in the lungs and spleens of mice infected with CO92 Δilp were significantly different. At 24 h after bacterial challenge there was no significant difference in the numbers of Y. pestis in the lungs among animals receiving the CO92 wild type or the Δilp mutant (Fig. 9A). However, the number of bacteria in the lungs of mice infected with the CO92 wild type proceeded to increase rapidly after 48 h (Fig. 9A), and similar numbers of bacteria were observed in the spleens (Fig. 9B). In contrast, the number of bacteria in the lungs and spleens of mice infected with the CO92 Δilp strain decreased to below the limit of detection (with one exception) after 48 h (Fig. 9A and andB)B) and then increased to approximately 1 × 103 and 1 × 104 CFU g−1 in the lungs and spleens, respectively, after 72 h (Fig. 9A and andB).B). We concluded that Ilp contributed to Y. pestis systemic spread during infection and to overall virulence.
The most important finding in this study was the identification of a novel Y. pestis surface-associated virulence protein. The gene for this protein, Ilp, was discovered by screening a comprehensive TnphoA expression library. Ilp was shown to be temperature regulated, involved in host cell adhesion, nonessential in flea colonization, and required for systemic spread and virulence in a murine disease model, as evidenced by a 55-fold increase in LD50 for the ilp deletion mutant. Mouse trials were facilitated, in part, by the use of a novel lux expression plasmid (pACYC177-lux) that allowed tracking the systemic spread of Y. pestis in vivo.
Data from comprehensive mutational analysis for bacterial surface/secreted proteins is limited, but based on results from Pseudomonas aeruginosa (17), it seemed reasonable to assume that between 5 and 10% of the 4,327 putative Y. pestis genes were potential TnphoA targets. Therefore, we initially planned on collecting and conducting sequence analysis on 1,200 independent phoA fusions to obtain an approximate 3-fold coverage of the chromosome. This level of target saturation (239 unique genes) was nearly reached after sequencing 612 phoA isolates. Based on this result, we conservatively estimate that 6% of the Y. pestis genome codes for surface-directed proteins that are not essential for viability. This may be an artificially low estimate, given the selection conditions (temperature, aerobic environment, and nutrient rich medium), the use of Y. pestis KIM6(pCD1−), and the fact that Y. pestis has undergone significant genomic reduction compared to its progenitor Y. pseudotuberculosis (4). Nevertheless, we were encouraged by the fact that the majority of previously characterized surface-associated chromosomal virulence genes were identified by this screen. It is also of interest that a large number of unique genes identified in this screen (52/239, or 22%) are hypothetical genes with no known functions. Conspicuously missing from this screen are the majority of 22 histidine kinases associated with two-component systems, all of which we previously screened for roles in virulence (23).
Alignment studies of Ilp identify it as a member of the intimin/invasin/autotransporter superfamily of surface proteins. This family, for which E. coli O157 intimin and Y. pseudotuberculosis invasin are archetypal, is now recognized in a number of Gram-negative organisms (31). The Ilp protein conforms to the general characteristics of this family, evidenced by an N-terminal signal sequence and N-terminal LysM domain, a highly conserved β-domain predicted to conform to a transmembrane β-barrel structure, and a C-terminal passenger domain localized on the cell surface. The conserved LysM domain at the N terminus is predicted to interact with peptidoglycan and thus may serve a periplasmic anchoring function for this large protein. The β-barrel domains are speculated to be pore-forming domains perhaps required for export of the C-terminal passenger domain across the outer membrane. Empirical data for the latter function are still subject to interpretation, being primarily based on similarity to other better described autotransporter proteins. The C-terminal passenger region contains repeated bacterial immunoglobulin-like domains (BID) which are highly repeated in Ilp and may form the context for surface adhesion. The majority of TnphoA insertions isolated in ilp mapped in this highly repetitive passenger domain, suggesting that this region is surface exposed. The large size of Ilp and its repetitive BID motifs undoubtedly accounted for the high representation of ilp-phoA fusions isolated in our transposon screen.
Our studies examined several regulatory parameters of ilp expression. Transcription of ilp was not under the control of the pCD1 virulence plasmid, as the plasmid's presence or absence did not influence ilp expression. Comparison of transcript levels between 25°C and 37°C showed an approximate 7-fold level of induction relative to rRNA controls. Temperature-induced conformational changes of intrinsic DNA bends can contribute to temperature regulation (28). Sequence analysis of the promoter region of the ilp gene showed spaced tracts of A and T residues typically associated with intrinsic DNA bending. Two-dimensional gel electrophoresis showed that the migration of PCR amplicons of the ilp promoter region was retarded relative to that of “unbent” DNA ladder fragments (data not shown). These preliminary results suggest that DNA topology of the ilp promoter region is subject to temperature, and such conformational changes may explain, in part, enhanced expression at 37°C.
Although pneumonic plague is the least prevalent form of plague, it is the most virulent form of the disease, showing rapid progression to lethality from low infectious doses. Aerosolization of Y. pestis is also the most likely form of deployment of this organism as a bioweapon (16), so understanding the systemic spread of the organism is crucial. However, it is not clearly understood how Y. pestis invades the pulmonary lymphoid tissues to result in subsequent systemic disease. The ability of bacterial pathogens to adhere to and invade host cells such as macrophages and epithelial cells not only provides an advantage in evading host defense mechanisms but may also be required for efficient systemic spread of infection. Several lines of evidence have demonstrated that Y. pestis is able to adhere to and/or invade host cells, and this ability is partially mediated by bacterial cell surface molecules such as Pla, capsular F1 antigen, OmpX, and Psa fimbriae, despite its lack of the major adhesins and invasins found in other enteropathogenic Yersinia species (3, 18, 21, 24). Identification of Y. pestis Ilp as yet another potential adhesion/invasin warranted further investigation to assess its contribution to pathogenicity.
HEp-2 cell association assays with Y. pestis KIM5, Y. pestis KIM5 Δilp, and the Y. pestis KIM5 Δilp-complemented strain indicated that Ilp acted as an adhesin, showing a reduction of cell association by Y. pestis KIM5Δilp compared to that of the parental wild-type and Δilp-complemented strains. Internalization assays also showed a slight reduction of internalization by Y. pestis KIM5 Δilp compared to the parental wild-type strain, suggesting that Ilp may also be involved in the internalization process. Interestingly, internalization by Y. pestis KIM5 grown at 37°C was approximately 860 times lower than that by Y. pestis KIM5 grown at 28°C. This finding is consistent with other reports that Y. pestis KIM5 grown at 37°C expresses F1 capsular antigen, which might prevent efficient interaction of Y. pestis surface proteins with the host cell and result in reduced bacterial internalization (6, 21). Importantly, Y. pestis KIM5 Δilp grown at 28°C showed a significant, albeit slight, decrease in HEp-2 cell internalization compared to the parental wild-type strain, yet a substantial number of mutant cells were still internalized. Further study is necessary to determine how other adhesins and invasins such as Pla, OmpX, and Psa are coordinately involved in this process. A specific mechanism for Ilp in invasion is still lacking and is the subject of current studies. We do not know if Ilp acts directly as an adhesion molecule or stabilizes the outer membrane and thus facilitates the activity of Pla, OmpX, and Psa.
Previous studies have identified three virulence factors involved in the transmission from fleas to warm-blooded hosts. The yersinia murine toxin (ymt) gene, encoding a phospholipase D, is required for survival in the flea midgut (14). The hmsHFRS and gmhS genes, encoding an extracellular polysaccharide and a modification of lipopolysaccharide core, respectively, are required for the formation of biofilm and blockage in the flea (7, 13). Because several members of the intimin/invasin family of proteins have been implicated in biofilm formation (31) and because Ilp is expressed at lower temperatures, we investigated the potential role of Ilp on blockage and infectivity in the flea. Results showed that there were no differences in the blockage and infection rates by Y. pestis KIM6 Δilp compared to those of the parental wild-type strain, suggesting that Ilp has a limited role, if any, in the transmission of Y. pestis from a flea to a mammalian host.
Based on reduced internalization of avirulent Y. pestis KIM6 Δilp by HEp-2 cells, we constructed a Δilp mutation in the partially attenuated Y. pestis KIM5(pCD1+) Pgm− strain to test in a mouse infection model. Surprisingly, mice i.v. infected with the Y. pestis KIM5 Δilp strain lacked the overt infection symptoms observed with the parental Y. pestis KIM5 strain and had a consistent delay in the time to death. To further examine this pattern, we employed an IVIS using Y. pestis KIM5(pACYC177-lux). Mice infected with Y. pestis KIM5(pACYC177-lux) showed systemic spread of infection early on (by day 3), whereas the KIM5 Δilp strain did not spread systemically until just before death. This was verified by determining the number of Y. pestis CFU g−1 of spleen, kidney, lung, and liver tissues. This analysis showed decreased systemic spread of Y. pestis Δilp from the spleen. These experiments also demonstrated the utility of tracking plague infection in vivo using bioluminescence. Such studies were significantly facilitated by our pACYC177-lux reporter construct, which has a high level of constitutive expression conferred by the flhDC promoter, plasmid stability in vitro and in vivo without antibiotic selection, the Tn5 kanamycin resistance gene that is allowed for use with this select agent, and the expression of the entire lux operon that negated the need to infuse luciferin prior to light detection (a requirement when a luciferase reporter [luxAB] is used alone). Additionally, we opted to use the pACYC177 vector over the construction of a chromosomal lux operon insertion for several reasons. This plasmid is temperature stable without antibiotic selection maintenance and has a low copy number (20 copies per cell). Therefore, we reasoned this would maintain an amplified reporter signal (20 copies of lux per cell versus a single chromosomal copy) without increased metabolic maintenance burdens on the organism. The latter point seems to be justified since there was no difference in virulence between wild-type and pACYC177-containing cells. Finally, because Y. pestis is easily transformable with plasmid DNA, introducing a plasmid into test strains is technically easier than constructing a chromosomal lux-bearing strain for each Y. pestis variant under study if a general reporter signal is only required.
Previous reports examining pulmonary infection of fully virulent Y. pestis showed that the dissemination of Y. pestis from the lung to other tissues and organs and recirculation back into lung tissue induce pulmonary edema and collapse (20). However, it is not fully understood whether the lethality of pneumonic plague is primarily attributed to the bacterial outgrowth in the lung and/or other locations. In this study, the number of bacteria in the lungs infected with the CO92 Δilp strain was significantly lower than that in lungs infected with the CO92 wild type, and the lethality of Y. pestis CO92 Δilp showed a 55-fold reduction. Interestingly, previous studies demonstrate that intranasal murine challenge with Y. pestis CO92 lacking Pla resulted in 100% survival even though a substantial number of bacteria can be recovered from the spleen (approximately 1 ×105 CFU), i.e., 1,000 more than the number recovered from the lungs after 3 days of infection (20). Additionally, challenging mice intranasally with Y. pestis CO92 lacking pPCP1 and the Braun lipoprotein resulted in 100% survival even though there was no significant difference in the number of bacteria in the spleens compared to that in wild-type infection (1). These results suggest that the lethality of pneumonic plague is related to the number of Y. pestis bacteria in the lung tissue rather than the number in the spleen, and this might explain the reduced virulence of CO92 Δilp in pneumonic plague. Consistent with this hypothesis was our initial observation that mice challenged with KIM5 Δilp intravenously remained asymptomatic until just before death, unlike mice infected with the wild-type strain that displayed early symptoms. This observation suggested that sepsis was delayed, and this was substantiated using in vivo bioluminescence assays. Thus, similar to the pneumonic infections, mice harboring wild-type bacteria had systemic infection early on, whereas Δilp bacteria were not detected in the infected mice until just prior to death.
After this paper was submitted for publication, Pisano et al. (26) described two related Y. pseudotuberculosis autotransporters, designated Ifp and InvC, and their respective roles in virulence of this Yersinia enteropathogen. Y. pseudotuberculosis InvC is highly similar to that of Y. pestis Ilp but contains multiple repeats of the immunoglobulin-like binding domain, yielding a protein of 5,337 amino acids compared to the 3,014 amino acids of Ilp. InvC, like Ilp, shows modest temperature regulation and is localized to the outer membrane, and its deletion results in delayed time to death in a mouse infection model. In contrast to the deletion of Y. pestis ilp that resulted in reduced adhesion to host cells, deletion of invC does not significantly alter host cell binding. However, expression of Y. pseudotuberculosis InvC in nonadherent E. coli K-12 enhances adherence but not host cell entry. This suggests that other adhesions can compensate for loss of InvC in Y. pseudotuberculosis, and together these data suggest that InvC and, by inference, Ilp function as adhesion molecules. The most significant contrast between Y. pestis Ilp and the similar Y. pseudotuberculosis InvC protein is that InvC is not required for virulence. Although Y. pseudotuberculosis ifp mutants show detectable attenuation of virulence, Y. pseudotuberculosis invC mutations did not display significant attenuation. This suggests that multiple adhesions of Y. pseudotuberculosis (Ail, InvA, YadA, Ifp, and InvC) may compensate for the loss of a single factor while the natural mutations of adhesions in Y. pestis may have elevated the importance of Ilp for Y. pestis virulence.
In summary, our findings indicate that the Ilp plays an important role in Y. pestis pathogenesis by promoting host cell association and invasion and colonization in the lung. Furthermore, its critical role in bacterial outgrowth in the lung suggests that Ilp might be a useful vaccine candidate. Importantly, these studies have identified a new virulence determinant of Y. pestis.
This study was supported by National Institutes of Health grants P20 RR15587, P20 RR16454, and P20 GM103408 and the Idaho Agricultural Experimental Station.
We thank Mike Konkel for the lux plasmid, Megan Steele for technical assistance, and Colin Manoil and Larry Gallagher for technical advice and protocol assistance for the transposon experiments.
Published ahead of print 30 July 2012
Supplemental material for this article may be found at http://iai.asm.org/.