Previous Article | Next Article ![]()
Infection and Immunity, October 2004, p. 5582-5596, Vol. 72, No. 10
0019-9567/04/$08.00+0 DOI: 10.1128/IAI.72.10.5582-5596.2004
Copyright © 2004, American Society for Microbiology. All Rights Reserved.
Department of Infectious Diseases,1 Hartwell Center for Bioinformatics and Biotechnology St. Jude Children's Research Hospital,3 Department of Molecular Sciences, University of Tennessee Health Science Center Memphis, Tennessee2
Received 23 March 2004/ Returned for modification 19 May 2004/ Accepted 29 June 2004
|
|
|---|
|
|
|---|
Since Pasteur and Sternberg first described the pneumococcus, a number of virulence factors required for invasive disease have been identified. Major virulence determinants such as capsular polysaccharide, pneumolysin, and choline-binding proteins have clearly established roles in pathogenesis (18). More recently, large-scale identification of S. pneumoniae virulence determinants has been attempted. Studies, such as signature-tagged mutagenesis (STM), have used transposons and suicide vectors to pepper the chromosome of the bacteria with mutations and identify genes required for virulence (13, 25, 39). STM screens rely on negative selection of mutants through passage in animal models and, to date, STM has been used three times for the pneumococcus, each time in a different strain.
Ideally, characterization of pneumococcal virulence determinants should include analysis of gene expression during invasive disease. Confirmation of their expression in vivo would not only verify the contribution of these genes to pathogenesis but would also elucidate their contribution to discrete forms of disease (e.g., genes expressed during pneumonia versus those in the blood during bacteremia). Unfortunately, analysis of in vivo bacterial gene expression, much less at discrete sites, has been limited by the difficulties of isolating sufficient quantities of pure and intact bacterial RNA from infected host tissues. To circumvent the requirement for RNA, investigators have used differential fluorescence induction (DFI) to identify S. pneumoniae promoters that are induced during disease (29). In contrast to STM, DFI relies on the promoter activity of random fragments of DNA cloned upstream of an episomal, promoterless green fluorescent protein. Fluorescence is imparted by the promoter activity of the randomly integrated DNA fragment under the environmental conditions tested (in vivo). Fluorescent bacteria can then be sorted by fluorescence-activated cell sorting analysis, leaving only bacteria containing plasmids with active promoters. Using DFI, Marra et al. identified operons that are enhanced during bacterial growth in a chinchilla model of otitis media, lower respiratory tract infection in a mouse, and growth in an intraperitoneal chamber implant model (29).
Large-scale analysis of gene expression during invasive disease provides not only transcriptional data with regard to certain genes, such as virulence determinants but, after interpretation, also provides information as to the status of the bacteria during infection and the response of the bacteria to the host environment. In vivo gene expression therefore potentially identifies unknown or unappreciated targets for pharmaceutical or vaccine intervention since, presumably, the genes whose transcription is altered during invasive disease are those that are required by the bacteria for survival in the host (14).
We describe here a protocol for the collection of RNA from bacteria within infected blood and cerebrospinal fluid (CSF) in vivo and bacteria adherent to epithelial cells in vitro. Using genome-wide cDNA microarrays available from the Pathogen Functional Genomic Research Center (PFGRC) at The Institute for Genomic Research (TIGR; Rockville, Md.), we have examined pneumococcal gene expression during bacteremia, meningitis, and epithelial cell contact (ECC). Analysis of pneumococcal genes expressed in these models identified global patterns of expression unique to each condition tested. Patterns of expression were identified with regard to virulence factors, transporters, transcription factors, translation-associated proteins, metabolism, and genes with unknown function. These findings were then cross referenced to previous STM and DFI studies. Finally, we examined several candidate virulence genes with unknown function by insertion duplication mutagenesis and challenge of mice.
|
|
|---|
Bacterial strains and construction of mutants. D39 Xen7 (D39X), a stable bioluminescent isolate of S. pneumoniae, serotype 2, strain D39 (2), was created as previously described (11). TIGR4 Xen 35 (T4X) was created by transformation of TIGR4 with genomic DNA from D39X by using CSP-2 (35). Bioluminescent mutants deficient in genes identified as enhanced during invasive disease were created by insertion duplication mutagenesis (37). PCR primers with nested EcoRI and BamHI restriction sites were used to amplify 200- to 500-bp fragments from the N termini of the genes. Amplified PCR fragments were purified and digested with EcoRI and BamHI (New England BioLabs, Beverly, Mass.). Digested fragments were ligated into the vector pJDC9 and transformed into chemically competent XL1-Blue E. coli (Stratagene, La Jolla, Calif.). Single transformants containing the insert were identified and plasmid DNA from these clones was used to transform D39X by using CSP-1 or, alternatively, T4X with CSP-2 (7). Chromosomal integration of the vector at the correct locus was verified by PCR analysis and sequencing by using a primer contained within the integrated vector and a primer upstream of the gene of interest.
Isolation of bacterial RNA from infected mouse blood. Female BALB/cJ mice (4 to 5 weeks old; The Jackson Laboratory, Bar Harbor, Maine) were maintained in a biosafety level 2 facility at St. Jude Children's Research Hospital. All experimental procedures were done with mice anesthetized with either inhaled isoflurane (Baxter Healthcare Corp., Deerfield, Ill.) at 2.5% or intraperitoneal MKX (1 ml of ketamine [Fort Dodge Laboratories, Fort Dodge, Iowa] at 100 mg/ml; 5 ml of xylazine [Miles Laboratories, Shawnee Mission, KA] at 100 mg/ml; 21 ml of phosphate-buffered saline [PBS]) at 5 µl/g of body weight. Bacteria in the blood were collected by using a modified version of the protocol described by Ogunniyi et al. (33). Mice were infected intratracheally with 105 CFU of D39X and imaged by using a charge-coupled device camera at 48 h postchallenge. Mice with severe sepsis, as determined by bioluminescent imaging (>50,000 relative light units/mouse) (11), were exsanguinated, and the blood was immediately transferred to a tube containing RNAprotect (Qiagen, Valencia, Calif.) at a ratio of 35:1 (vol/vol; RNAprotect/expected total mouse blood collected) and vortexed for 5 s. Bacteria were harvested by centrifugation of the suspension at 825 x g for 10 min to remove debris, followed by centrifugation at 15,000 x g for 15 min to pellet the bacteria. RNA isolation was performed by using a Qiagen RNeasy minikit (Qiagen) with the following modifications. Bacteria were lysed in the presence of 400 mg of 0.1-mm zirconia-silica beads (BioSpec Products, Inc., Bartlesville, Okla.) by using the Mini-Beadbeater 3110BX (BioSpec Products, Inc.) and then incubated at 70° C for 10 min. Bacterial lysate was spun through a Qiashredder column (Qiagen) to remove the beads, and the remaining solution was subsequently processed according to the manufacturer's protocol with an on-column DNase digestion step. Quantitation of RNA was performed by using a UV spectrophotometer (UV-1601; Shimadzu Corp., Kyoto, Japan) at an optical density at 260 nm (OD260), whereas degradation was assessed by visualization of the RNA on a 1% agarose Tris-borate-EDTA (TBE) gel.
Isolation of bacterial RNA from infected rabbit CSF. Male New Zealand White rabbits (5 kg; Myrtles, Thompson Station, Tenn.) were anesthetized with 35 ml of 25% urethane administered subcutaneously and pentobarbital sodium (15 mg/kg) given intravenously. Anesthetized rabbits were immobilized on a stereotaxic frame and challenged by direct intracisternal injection with 108 CFU of T4X by using a 25-gauge spinal needle (27). At 4 h after challenge, CSF was collected and transferred immediately to a sterile tube containing RNAprotect at a ratio of 3:1 (RNAprotect/rabbit CSF). RNA was then collected as described above.
Isolation of bacterial RNA from pneumococci attached to Detroit cells. A T-175 flask of confluent Detroit pharyngeal epithelial cells (38) was activated with tumor necrosis factor alpha (Sigma) at 10 ng/ml for 2 h. Cells were challenged with 25 ml of a 4 x 106-CFU/ml suspension of T4R (12), an unencapsulated derivative of TIGR4, and incubated at 37°C for 3 h to allow for adherence. After the incubation, cells were washed three times with PBS (BioWhittaker, Walkersville, Md.) to remove nonadherent bacteria and covered with 10 ml of RNAprotect. Cells were scraped off the surface of the flask, and the suspension was transferred to a new T-175 flask containing 3-mm glass beads. Cells were sonicated for 5 min by using a tabletop ultrasonic cleaner (FS20; Fisher Scientific, Pittsburgh, Pa.) to lyse the epithelial cells but leave the bacteria intact. The suspension was then centrifuged at 800 x g for 5 min to remove cellular debris and then centrifuged at 4,600 x g for 10 min to pellet the bacteria. RNA was collected from the bacterial/eukaryotic pellet by using a Qiagen RNeasy minikit. To remove the remaining eukaryotic RNA, bacterial RNA isolated from the bacterial/eukaryotic pellet was then enriched by using MICROBEnrich (Ambion, Austin, Tex.). Quantitation of RNA was performed by using a UV spectrophotometer at OD260, whereas degradation was assessed by visualization of the RNA on a 1% agarose TBE gel. As a control, RNA was collected from bacteria grown in tissue culture medium in parallel and was processed by using MICROBEnrich.
Microarray analysis of bacterial RNA.
Microarray experiments were performed by using whole-genome S. pneumoniae cDNA microarrays obtained from the PFGRC at TIGR (http://pfgrc.tigr.org.). The S. pneumoniae genome microarray consisted of PCR products representing segments of 2,131 open reading frames from S. pneumoniae strain TIGR4 (44) and 118 unique open reading frames from strains R6 (17) and G54 (39). Microarray experiments, including RNA quality control, Cy3 and Cy5 dye labeling, hybridization, washing, and scanning, were performed at the Functional Genomics lab, Hartwell Center for Bioinformatics and Biotechnology, St. Jude Children's Research Hospital, by using protocols from the PFGRC (http://pfgrc.tigr.org/protocols.shtml). The hybridization probe was constituted by mixture of differentially labeled cDNA derived from (i) total RNA isolated from S. pneumoniae obtained from either infected mouse blood or rabbit CSF or (ii) RNA from bacteria grown in C+Y. Alternatively, probe for ECC analysis was constituted with (i) RNA isolated from bacteria adhered to Detroit cells or (ii) RNA from parallel cultures grown in tissue culture media. RNA samples from both conditions were labeled with monofunctional Cy3 and Cy5 dyes by using an indirect amino-allyl labeling method, combined, and hybridized overnight to the printed slides. Slides were washed and scanned by using an Axon 4000B dual channel scanner (Axon Corp., Union City, Calif.) to generate a multi-TIFF image of each slide. Images were analyzed by using Axon GenePix 4.1 image analysis software, and the resulting text-data files were imported into Spotfire DecisionSite for Functional Genomics (version 7.2; Spotfire, Sommerville, Mass.). A series of filtration algorithms were applied to remove spots that consistently generated bad data (based on the frequency with which a particular spot failed to reach a minimum required signal-to-noise ratio [SNR] and the frequency with which a particular spot was flagged bad by the image analysis software GenePix Pro 4.1). Genes that were flagged
66% of the time were removed from analysis. Similarly, spots that failed to meet the SNR criteria 75% of the time (9 of 12 times for blood and ECC microarrays and 6 of 8 for CSF microarrays) were removed from consideration. Intensity-based global normalization was then performed to remove dye-specific bias, and background correction was performed by subtracting the normalized median pixel intensity of the background value from the normalized median pixel intensity of the spot itself. Cy5/Cy3 ratios (fold changes) were then calculated for every spot. Since each gene was spotted four times per glass microarray, only genes whose corresponding spots were not flagged at least 75% of the times were considered.
Statistical analysis of microarray data. Microarray analysis examining RNA from bacteria adhered to Detroit cells and blood experiments was performed on total RNA isolated from three independent biological replicate experiments. Microarray analysis examining RNA from infected rabbit CSF was performed on total RNA isolated from two independent biological replicate experiments. Accuracy and statistical significance of the gene expression differential over the course of the replicate experiments was calculated by using a Student t test-analysis of variance algorithm available in Spotfire DecisionSite (19). Genes with high levels of significance (P > 0.001) and a minimum fold change of 2.0 were considered up- or downregulated.
Virulence assessment of mutants. To assess the virulence potential of mutants deficient in genes with altered expression in vivo, mice were challenged intratracheally and monitored for 4 days or challenged intranasally and monitored for 7 days. Exponential cultures (OD620 = 0.5) of D39X, T4X, or their derivatives were centrifuged, and the bacteria were washed with and suspended in PBS. Mice were anesthetized with isoflurane and challenged intranasally with 107 CFU in 25 µl or intratracheally with 105 CFU in 100 µl of PBS. At 2 days postchallenge, all mice were sampled by intranasal lavage, blood collection from the tail vein, and bioluminescent imaging with a Xenogen IVIS camera. Bioluminescent imaging provided a noninvasive method for assessing the bacterial burden in the lungs of all of the mice (data not shown). After imaging, 6 to 10 mice were randomly selected from each cohort and sacrificed, and bacterial titers in lungs were determined. The number of CFU present per gram of homogenized lungs was used to assess bacterial burden in the lungs. Remaining mice (6 to 10 per group) were then observed to determine the percent survival over time. In all instances after challenge, the infectious dose was confirmed by serial dilution and plating of the bacterial suspension on blood agar.
|
|
|---|
Collection of RNA samples from infected blood and CSF.
Collection of bacterial RNA from infected tissues has been limited by the inherent difficulties of separating large quantities of host cells and debris away from bacteria in a manner sufficiently timely to prevent degradation of RNA or to prevent novel gene expression in response to conditions presented during the separation process (14, 36). To avoid these problems, we harvested infected mouse blood and rabbit CSF directly into RNAprotect (Qiagen). RNAprotect served to stabilize bacterial RNA, allowed purification of the RNA without degradation (Fig. 1), and lysed contaminating host cells. Preliminary experiments determined that high titers of bacteria (>108 CFU/ml) in body fluids were required to obtain sufficient RNA (data not shown). Although strain T4X was used for meningitis, preliminary experiments determined its yield in blood was too low (data not shown). Therefore, for bacteremia we used strain D39, which attains titers in blood of 109 CFU/ml by 48 h after challenge (35). Bioluminescence of T4X and D39X permitted visualization of the bacteria in living mice and allowed us to identify animals with sufficiently high bacterial titers for RNA isolation. Bioluminescent imaging was necessary, since only half of the mice infected intratracheally had sufficient bacteria in the bloodstream 48 h postchallenge. The remaining mice eventually progressed to similar levels of bacteremia (
109 CFU/ml of blood within 12 to 24 h); however, this occurred in a manner too staggered for satisfactory bacteria collection. Exsanguination of the mice collected at 48 h yielded
0.75 ml of blood from each infected animal, which in turn allowed collection of
25 µg of pure bacterial RNA from the pooled blood of three to four mice.
![]() View larger version (44K): [in a new window] |
FIG. 1. RNA collected from S. pneumoniae in vivo and in vitro. After isolation of pneumococcal RNA from infected blood, CSF, and bacteria adherent to the Detroit pharyngeal epithelial cell line, RNA was visualized on a 1% TBE gel to confirm purity and assess degradation. Bacterial RNA (open arrowheads) and eukaryotic RNA (shaded arrowheads) are indicated. Bacterial RNA isolated from the ECC model was enriched after collection of the initial pellet (raw) to remove contaminating eukaryotic RNA.
|
1.0 ml of infected CSF was collected and yielded
5 µg of pure bacterial RNA from each rabbit (Fig. 1).
Collection of RNA samples from pneumococci attached to epithelial cells.
To examine bacterial gene expression in response to ECC, we infected confluent monolayers of Detroit cells (38), a pharyngeal epithelial cell line, with pneumococcus T4R. T4R is an unencapsulated derivative of T4 (12). As a control, RNA was collected from bacteria grown in the same tissue culture media without cells. These analyses therefore reflect only changes in gene expression that are a result of host cell exposure. Stimuli resulting in alterations in gene expression should therefore be limited to (i) intimate contact with the host cells, (ii) soluble factors released by the host cell during normal processes, and (iii) soluble factors released by the host cell in response to the bacterial infection. Collection of RNA from bacteria attached to Detroit cells resulted in the collection of
25 µg of mixed eukaryotic and prokaryotic RNA from the bacteria and host cell pellet. Using MICROBencrich, we were able to purify the prokaryotic RNA (Fig. 1), with a final yield of
10 µg per T-175 flask.
Microarray analysis of gene expression relies on the comparison of two RNA species. Levels of expression reported in the test condition are reported in relation to levels of transcription observed in the control condition. A direct comparison of test versus control can be set up in vitro with tissue culture (medium to medium plus Detroit cells). With regard to in vivo gene expression, an appropriate in vitro control condition does not exist. However, reference to the same arbitrary control media (blood versus C+Y or CSF versus C+Y) allows for indirect comparisons between the different test conditions (blood versus CSF). As a baseline, >99% of genes showed similar expression levels in D39 and T4 grown in C+Y (data not shown).
Microarray analysis of bacterial physiology in different body sites.
Filtration algorithms and SNR analysis of the microarrays by GenePix 4.1 removed genes whose corresponding spots on the microarray were not valid
75% of the time. As such, the results reported in the present study correspond to 69% of the genome during growth in the bloodstream, 53% of the genome during meningitis, and 68% of the genome after ECC. A list of the genes that did not meet the above criteria and therefore were not analyzed for alternate gene expression in vivo can be found at http://www.stjuderesearch.org/vivogene. Stringent requirements imposed by the filtration algorithms ensured that any alterations in gene expression that are reported here are rigorously supported by the data collected from the microarrays.
For genes that met the established criteria for transcription analysis, it was determined that the majority (92% in the blood, 85% in CSF, and 90% after ECC) were expressed in a fashion similar to growth in C+Y. Table 1 lists the genes with altered transcription sorted based on their putative cellular role and their location on the bacterial chromosome. Inclusion criteria were a >2-fold change in levels of expression and a P value of <0.01. Of the genes with altered expression in vivo, 55% had enhanced expression in blood, while 81% were enhanced during ECC. In contrast, only 35% were enhanced during growth in CSF. Overall, the patterns of gene expression were determined to be distinct for bacteria in each anatomic site. Figure 2 is a Venn diagram illustrating the differences and similarities between the expression profiles of the pneumococci in blood, in CSF, and during ECC. Highlighting the disparity between these conditions, only eight genes had similar alterations in gene expression during bacterial growth in blood, in CSF, or during ECC: two that encode the virulence determinants PspA (46) and PrtA (4), three that encode genes in the psa operon (manganese acquisition and transport) (30), two that are involved in energy metabolism (an enolase [eno] and a glycosyl hydrolase [SP0265]), and a transporter (SP1587). Their common expression warrants further investigation as potential targets for intervention, since they may reflect a core set of genes required for virulence.
|
View this table: [in a new window] |
TABLE 1. Differential expression of S. pneumoniae genes as determined by microarray analysis
|
![]() View larger version (25K): [in a new window] |
FIG. 2. Venn diagrams highlighting disparity between infectious models for genes with altered expression. Numbers indicate the amount of genes determined to have altered expression that are either shared or exclusive to growth of the pneumococci in blood, CSF, or ECC.
|
In the CSF, the pneumococcus decreased transcription of at least 20 genes involved with competence (15), including comX1, comA, comB, comE, and comD. Likewise, 11 genes involved in the biosynthesis of fatty acids were also markedly reduced. Decreased expression of the fab operon (31) may indicate that sufficient fatty acids are available in the CSF. Other differences observed in CSF cultures included enhanced expression of the lic operon (50), which indicates that the bacteria are decorating their cell wall with phosphorylcholine. SP1804, a gene encoding a general stress protein, is enhanced in the CSF, perhaps indicating that the bacteria are under stress in adapting from the rich medium of blood to the poor medium of CSF.
Pneumococci adherent to epithelial cells also had unique alterations in their expression profiles. We observed enhanced expression of the lic operon (50) and the dlt operon (49) during ECC. These changes indicate enhanced addition of phosphorylcholine and D-alanine to teichoic acids. Enhanced expression of VncS and CiaR/H, genes encoding two-component systems (42), indicate global changes in the synthesis of cell wall polymers, peptide pheromones, bacteriocins, and htrA expression. Manganese acquisition also appears to be particularly important since the psa operon showed enhanced expression in all three of the conditions tested and has recently been shown to undergo enhanced transcription in vivo (30).
With regard to energy metabolism, during growth in blood the pneumococcus reduces the expression of the genes encoding 1-phosphofructokinase and fructose-bisphosphate aldolase. Phosphorylation of 1-phosphofructose by 1-phophofructokinase is the rate-limiting step of glycolysis and targets the molecule for glycolysis (22). Reducing expression levels of 1-phosphofructokinase suggests that the pneumococcus has a readily available carbon source in the blood. At the same time, concurrent enhanced expression of SP0265 and SP2021, glycosyl hydrolases, and enhanced expression of scrR and malR, the repressors for sucrose and maltose operons, suggest that the pneumococcus is also utilizing alternative sources of energy.
Phase variation. S. pneumoniae undergoes spontaneous phase variation between a transparent phenotype and an opaque phenotype (21); the transparent phenotype has an enhanced capacity to adhere and colonize the nasopharynx, whereas the more phagocytosis-resistant opaque phenotype predominates in blood. An increased capacity to adhere by the transparent phenotype corresponds to higher levels of CbpA, phosphorylcholine, teichoic acid, and autolysin than do opaque variants (21, 41, 48). In contrast, the opaque phenotype produces more capsular polysaccharide and hydrogen peroxide and requires up to 30-fold more human immune serum to achieve 50% opsonophagocytic killing than related transparent strains (20, 43).
Analysis of the genes enhanced during ECC confirmed a selection for the transparent phenotype. We observed enhanced expression of cbpA and the lic and dlt operons. Similar results for bacteria in CSF taken together with downregulation of spxB in CSF indicate that the transparent phenotype may reappear after bacteria leave the bloodstream. Interestingly, we did not observe phenotype-enhanced expression of spxB or of the genes involved in capsule production in the blood.
Microarray analysis of known virulence determinants. Microarray analysis of 20 known pneumococcal virulence determinants indicated site-specific expression for 17 of them (Table 1). Only two virulence determinants with altered gene expression, pspA and prtA, were expressed in a similar fashion in blood, CSF, or ECC (see above). pspA, encodes a choline-binding protein that inhibits complement deposition on the surface of the bacteria and is upregulated in blood by Northern analysis (46). prtA encodes a conserved serine protease (4).
Expression profiles were consistent with much of the current understanding of pneumococcal pathogenesis. During attachment to pharyngeal epithelial cells, pneumococci enhanced expression of the genes encoding choline-binding protein A (CbpA) and HtrA, two proteins shown to contribute to nasopharyngeal colonization (41, 42). Expression of ply was decreased, suggesting attenuation of the ability to injure host cells with the toxin pneumolysin. We observed enhanced expression in the blood of the operon encoding choline-binding protein G (CbpG) and choline-binding protein F (CbpF) (12). CbpG has been demonstrated in our laboratory to be required for the development of high-grade bacteremia (preliminary data). Lastly, we observed decreased expression of the genes encoding pneumolysin (ply) (8), autolysin (lytA) (45), and pyruvate oxidase (spxB) (43) in the CSF. Pyruvate oxidase, SpxB, is responsible for the production of hydrogen peroxide by the pneumococcus. This trio was striking since pneumolysin, hydrogen peroxide, and inflammatory cell wall components released by autolysin are the principle agents by which the pneumococcus induces neuronal damage (6, 47). Decreased expression of these genes in the CSF suggests that the pneumococcus adapts so as to decrease damage to neurons.
We also observed alterations in the gene expression of less-well-characterized virulence determinants. In the blood, we observed enhanced expression of lmb, a gene homologous to a laminin-binding protein in Streptococcus anginosus (1). lmb transcription was subsequently reduced in the CSF. Likewise, we observed a decrease in the expression of genes involved in bacteriocin synthesis and immunity (10). These expression profiles suggest a potentially underappreciated role for these genes during pathogenesis.
Cross-reference to STM studies and DFI analyses. Investigators have used STM and DFI to identify pneumococcal genes required for invasive disease (13, 25, 29, 39). In Table 1, we highlight genes with altered expression that have been identified by STM or DFI. Comparison of the three STM studies determined that the majority of loci identified were hit by only one study. Although this may reflect differences in methodology, it is likely that the use of three different strains of S. pneumoniaeG54, a serotype 19F (39); strain 0100993, a serotype 3 (25); and T4, a serotype 4 (13)contributed to the discrepancy observed between the three STM studies. This disparity suggests that strain-dependent variations may occur with regard to virulence. As such, genes identified by more than one method are therefore particularly interesting for the current analysis since they may represent a set of core virulence elements required by all pneumococci for the disease process. Moreover, core genes need not have altered transcription in vivo. The genes encoding immunoglobulin A1 protease (40), ZmpB (a metalloprotease that elicits inflammation in the lower respiratory tract) (5), and PavA (a fibronectin-binding protein) (16) all maintained unchanged levels of expression in vivo and yet were all identified by more than one STM study (Table 2). Table 2 is a comprehensive list of genes with unaltered expression that have been identified by more than one STM study. This approach is not comprehensive, however, since none of the major virulence determinants, such as pneumolysin, capsular polysaccharide, or autolysin, was identified by more than one STM study. This may indicate a more site-specific contribution to pathogenesis (34). Finally, the majority of genes hit by more than one STM analysis encoded proteins whose function is unknown (Table 1). Those whose expression was altered in vivo and that are characterized include the pur operon (32) and rib operon (23) for purine and riboflavin biosynthesis, respectively, as well as radA, a DNA repair enzyme (3).
|
View this table: [in a new window] |
TABLE 2. Loci or genes within the same operon identified by multiple STM analyses with constitutive gene expression as determined by in vivo microarray analysis
|
|
View this table: [in a new window] |
TABLE 3. Virulence assessment of mutants deficient in genes with enhanced expression during ECC or growth in blood
|
Considerations. Although the present study is the first to directly measure pneumococcal gene expression during invasive disease, it must be acknowledged that to obtain sufficient quantities of bacterial RNA from the disease models, certain aspects of the experimental design were not optimal. Foremost, it was necessary to use a different strain for each disease model (D39X for blood, T4X for CSF, and T4R for ECC). Although preliminary data from in vitro experiments (data not shown) demonstrates that D39 and T4 respond in a similar manner to growth in vitro (>99% of the genes examined when grown in C+Y), how each strain responds to different anatomical sites remains unknown. It was also necessary to use a different host species for each disease model (mice for bloodstream infection, rabbits for CSF infection, and humans for ECC), as well as a different RNA extraction protocol for ECC. The relationship between different host species and the pneumococcus may also result in alterations of the pneumococcal response. As a result of these caveats, care must be taken when gene expression levels for different anatomical sites are compared.
A second consideration is the requirement for very high titers of pneumococci in the biological samples and the various durations of infection. For blood, bacterial titers were ca. 109 CFU/ml and were collected 2 days after intratracheal challenge. As such, the expression profile observed does not represent the initial phases of bloodstream infection. Likewise, rabbits were infected intracisternally with 108 CFU, and the bacteria were collected 4 h later. Since the bacteria were collected after only 4 h, our model represents bacterial gene expression during early meningitis. During a natural infection, only a few animals would have a bacterial titer as high as 108 CFU/ml in the CSF, and leukocytes and other host factors would be present in the CSF.
A final consideration is that not all of the genes in the genome were examined. Although our analysis of in vivo gene expression is the most comprehensive to date, we were unable to report on the expression of 31% of the genome during growth in blood, 47% in the CSF, and 32% after ECC. As such, the present study is not comprehensive and further studies, including confirmation by Northern analysis of specific genes, are warranted.
Conclusions.
Microarray analysis of pneumococcal gene expression during invasive disease sheds light on the complex mechanisms of pneumococcal pathogenesis. Despite limitations imposed by technical challenges associated with in vivo RNA collection, we observed dramatic changes in a variety of genes, thus providing important information as to the physiology of the pneumococcus during invasive disease. Most genes (
90%) do not change expression from in vitro growth medium to in vivo. However, strikingly, the pneumococcus seems to adapt to the blood, CSF, and ECC in a site-specific manner. These changes may be a result of the pneumococcus adapting to nutrient availability and the stresses placed on the bacteria. In blood, cell wall and membrane synthesis and competence were unchanged, whereas the bacteria strongly decreased ribosomal transcription, reduced glycolysis, and increased CbpG and CbpF. In CSF, pneumococci drastically curtailed gene expression for competence, autolysis, and production of toxins, while increasing amino acid biosynthesis and choline on the cell surface.
Cross-referencing of microarray analyses with that of STM and DFI and the virulence assessment of our own mutants strongly implies that pneumococcal virulence is dependent on a network of genes, the majority of which are not fully understood. Although an understanding of this network may seem daunting, it seems that the continued use of global analytical techniques will shed light on the core genes that are required for virulence. Such an attempt will require synthesis of data from multiple strains to derive a general picture of virulence above the noise of strain variation.
This study was supported by NIH grant R01 AI27913 and The American Lebanese Syrian Associated Charities.
|
|
|---|
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Copyright © 2009 by the American Society for Microbiology. For an alternate route to Journals.ASM.org, visit: http://intl-journals.asm.org | More Info»