By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
In response to environmental signals in the host, bacterial pathogens express factors required during infection and repress those that interfere with specific stages of this process. Signalling pathways controlling virulence factors of the human respiratory pathogen, Haemophilus influenzae, are predominantly unknown. The lipooligosaccharide (LOS) outer core represents a prototypical virulence trait of H. influenzae that enhances virulence but also provides targets for innate and adaptive immunity. We report regulation of the display of the virulence-associated phosphorylcholine (PC) epitope on the LOS in response to environmental conditions. PC display is optimal under microaerobic conditions and markedly decreased under conditions of high culture aeration. Gene expression analysis using a DNA microarray was performed to begin to define the metabolic state of the cell under these conditions and to identify genes potentially involved in PC epitope modulation. Global gene expression profiling detected changes in redox responsive genes and in genes of carbohydrate metabolism. The effects on carbohydrate metabolism led us to examine the role of the putative H. influenzae homologue of csrA, a regulator of glycolysis and gluconeogenesis in Escherichia coli. A mutant containing an in-frame deletion of the H. influenzae csrA gene showed increased PC epitope levels under aerobic conditions. Furthermore, deletion of csrA elevated mRNA expression of galU, an essential virulence gene that is critical in generating sugar precursors needed for polysaccharide formation and LOS outer core synthesis. Growth conditions predicted to alter the redox state of the culture modulated the PC epitope and galU expression as well. The results are consistent with a multifactorial mechanism of control of LOS-PC epitope display involving csrA and environmental signals that coordinately regulate biosynthetic and metabolic genes controlling the LOS structure.
Haemophilus influenzae is a Gram-negative, facultative aerobe that colonizes the respiratory tract of humans, the only natural host known for this bacterium. It is a common cause of otitis media, upper and lower respiratory tract infections, septicemia and meningitis in children. Lipooligosaccharide (LOS), a principle component of the outer-membrane, mediates crucial interactions between the bacterium and the host's immune system contributing to its pathogenesis. In H. influenzae, the LOS lacks the long repetitive polysaccharide O-antigen side chain present in the lipopolysaccharide (LPS) of many Gram-negative bacteria (Moxon and Murphy, 2000). The LOS structure varies between strains including the acapsular or non-typeable strains that are important pathogens in otitis media and respiratory infections, yet several features are conserved. H. influenzae LOS consists of Lipid A, an inner core comprised of several sugars including a single 3-deoxy-D-manno-octulosonic acid (KDO) linked to three heptoses, and an outer core containing a heteropolymer of glucose and galactose generally not exceeding six residues and modified with sialic acid, N-acetylgalactosamine, and phosphorylcholine (PC) (Hood et al., 1999; Risberg et al., 1999). Expression of these surface-exposed substituents of the LOS has been shown to contribute to H. influenzae's ability to establish an infection in the host. Sialylated LOS is critical for colonization of the middle ear and nasopharynx in a chinchilla model of otitis media (Bouchet et al., 2003). Mutants deficient in the N-acetylgalactosamine structure exhibit reduced lethality in a mouse model of bacteremia (Hood et al., 1996a). PC modification of LOS (LOS-PC) in H. influenzae is important for colonization and persistence in the respiratory tract in an infant rat model of infection (Weiser et al., 1998). The presence of the PC epitope also enhances adherence and invasion of human epithelial cells by binding to the receptor for platelet-activating factor, leading to possible sequestration from host immune clearance (Swords et al., 2000).
The PC moiety is important for the infection process, yet expression of this epitope also increases susceptibility to serum killing mediated by complement and the acute-phase reactant, C-reactive protein (Weiser et al., 1998). In addition, the PC moiety in Haemophilus and in other bacteria is recognized by natural antibodies (Leon and Young, 1971; Shaw et al., 2000). Thus, rapid switching off of PC expression is likely to provide an important adaptation for surviving in different niches encountered by the bacterium during the colonization and infection process.
During the course of infection, H. influenzae is likely to encounter environmental niches of varying redox levels as it transits between the relatively high oxygen environment of the airway surface to sites lower in oxygen such as an interstitial location during traversal of the mucosal epithelium, entry into bloodstream, or spread to the middle ear. Previous work by others has identified genes likely to participate in redox responses by H. influenzae during pathogenesis. These genes include sodA, lctP and lpdA, whose respective homologues in Escherichia coli have been shown to be redox regulated (Compan and Touati, 1993; Lynch and Lin, 1996; Cunningham et al., 1998) and have been examined in infant rat models of H. influenzae infection. The sodA gene encoding superoxide dismutase was shown to be important for oxidative stress defence in vitro and for optimal nasopharyngeal colonization (D’Mello et al., 1997), lctP ( l-lactate permease) is required for survival in the bloodstream (Herbert et al., 2002), and lpdA (dihydrolipoamide dehydrogenase), a component of the pyruvate and α-ketoglutarate dehydrogenases, is needed for aerobic growth in vitro and for bacteremia (Herbert et al., 2003).
We postulate that modulation of gene expression in response to changes in redox conditions is required by H. influenzae in order to express the repertoire of genes needed for optimal survival as it transits between microenvironments within the host. In this report, we investigated the effect of varied aeration culture conditions on modulation of LOS-PC, an example of a classical virulence determinant of H. influenzae. We analysed the global expression profile of H. influenzae grown under the growth conditions that affect PC epitope levels in order to define H. influenzae's response to this condition on a genomic scale and to identify genes involved in modulating the PC phenotype. The results were integrated with a metabolic model to infer potential regulatory inputs into PC epitope display. Mutagenesis experiments validated this approach and detected a novel connection between LOS modification and genetic control of glycolytic pathways.
Varied culture aeration conditions modulate phosphorylcholine (PC) epitope display and licA mRNA
We investigated whether environmental signals generated by a range of culture aeration levels during growth could affect the levels of the PC epitope displayed by H. influenzae Rd in which the PC moiety is attached to a glucose residue of the LOS, a feature it shares with the current model for the LOS structure of non-typeable strains (Risberg et al., 1999; Cox et al., 2001; Mansson et al., 2002; 2003). To attain different aeration levels, the volume of shaken cultures was varied as previously described (D’Mello et al., 1997). An unaerated culture was generated using a sealed tube filled to exclude air. The effect of these conditions on redox responses in H. influenzae was defined by global expression profiling of three replicate cultures grown under conditions predicted to promote aerobic (10 ml culture volume) versus microaerobic growth (200 ml culture volume) (Tables 1 and 2). We detected increased microaerobic expression of putative homologues of genes whose expression is increased under microaerobic and anaerobic conditions in E. coli, including genes encoding dimethyl sulphoxide reductase, dmsABC (Cotter and Gunsalus, 1989; Tseng et al., 1996) and nitrite reductase, nrfABCD (Choe and Reznikoff, 1993; Tyson et al., 1994). Moreover, the hxuCBA genes encoding the haem-haemopexin uptake system were more highly expressed in 10 ml cultures, consistent with H. influenzae's absolute requirement for exogenous haem sources for aerobic growth (White and Granick, 1963; Evans et al., 1974). Microarray results were independently confirmed by RT-qPCR (web supplement, Fig.S1). These results demonstrate differential expression of redox responsive genes under culture conditions we defined here as aerobic and microaerobic.
Table 1. Aerobically induced genes of Rd (P-values ≤ 0.01).
To evaluate PC epitope levels under varied redox conditions of growth, Western immunoblotting of whole-cell lysates from cultures grown with a range of aeration levels was performed with monoclonal antibody (mAb) TEPC 15, which is highly specific for the PC epitope (Fig. 1A). PC modification of the LOS in H. influenzae requires the lic1 locus consisting of licA (choline kinase), licB (choline transporter), licC (pyrophosphorylase) and licD (choline transferase) (Weiser et al., 1997). As expected, control lysates from a mutant containing a mariner transposon insertion in licA exhibited no reactivity with anti-PC mAb, confirming the antibody specificity (Fig. 1A, lanes 1 and 2). In wild-type Rd, the PC epitope was expressed poorly under aerobic conditions (10 ml volume), but was dramatically increased under low aeration conditions (Fig. 1A).
Because the licA gene is required for PC production, we investigated whether the modulation of PC epitope levels could be mediated through regulation of licA mRNA expression. We conducted primer extension analysis on RNA in wild-type Rd grown under conditions of varied culture aeration (Fig. 1B). Primer extension analysis mapped a licA transcriptional start site to a distance of approximately 46 bp upstream of putative licA ATG start codon γ. The level of abundance of licA transcripts containing a common 5′-transcriptional start site correlated with a corresponding increase in the culture volume, with the most abundant primer extension product seen in the 200 ml culture (Fig. 1B, lane 5).
In summary, these data indicate that varied culture aeration conditions result in modulation of both the display of the PC epitope on the LOS of H. influenzae and the abundance of licA-hybridizing transcripts in primer extensions in a similar manner. These conditions were also confirmed to modulate redox responsive genes indicating that the intracellular redox state may play a role in this regulation. Increased PC epitope levels were observed concomitant with increased abundance of licA primer extension products from RNA from microaerobic cultures compared to highly aerated cultures, suggesting that modulation of PC epitope levels may be mediated through regulation of licA mRNA expression. Based on these results, three growth conditions were used for subsequent experiments in this report to examine the mechanism of PC modulation. For simplicity, these conditions will be termed aerobic (‘+O2’, a 10 ml culture), microaerobic (‘M’, a 200 ml culture), and unaerated (‘–O2’, a sealed 50 ml tube filled to the rim).
Role of phase variation of licA in modulation of PC epitope display
The licA gene is subject to phase variation mediated by slipped-strand base mispairing within tandem tetrameric repeats of the sequence 5′-(CAAT)n-3′ within its N-terminal coding sequence (Weiser et al., 1989). To examine whether modulation of PC epitope levels in response to redox conditions is mediated by phase variation of licA, we created a strain, Δrep, which contains a deletion of the CAAT repeats and positions three potential licA ATG initiation codons in-frame (details in Experimental procedures) to allow full-length translation from all three initiation codons.
Western blot analysis was conducted to compare PC epitope levels in Δrep versus wild-type strains. As expected control lysates from a licA deletion strain, ΔlicA, did not react with anti-PC mAb (Fig. 2, lanes 3, 6 and 9). PC epitope levels of wild-type and Δrep were similar to each other under each aeration condition, indicating that the Δrep mutation does not affect redox modulation of this epitope (Fig. 2). In addition, fluorescent staining of a replicate gel for detection of the LOS carbohydrate structure revealed no appreciable differences between samples indicating that a change in total LOS production does not account for the modulation of the PC epitope (Fig. 2, bottom panel). In summary, phase variation of licA is not the exclusive means of regulating PC epitope levels, because deletion of the CAAT repeats did not abrogate modulation of this phenotype. Furthermore, this phase-locked Δrep strain that displays redox regulation of the PC epitope identical to wild-type Rd proved useful throughout these studies of PC modification of LOS by removing a mechanism that can spontaneously turn off PC production.
Effects of licA mRNA overexpression on PC epitope levels
Phosphorylcholine modification of lipooligosaccharide epitope levels and transcript levels of the lic1 locus essential for its production are similarly influenced by redox conditions of growth. Therefore, a simple hypothesis to account for the observed modulation of the PC epitope level is that regulation of lic1 mRNA mediates this control. We postulated that if the lower abundance of lic1 mRNA in aerobic cells accounts for the aerobic decrease in PC epitope levels, then generating elevated amounts of lic1 mRNA aerobically will cause elevated aerobic PC epitope levels. We engineered a strain Rhel-licA, in which the native licA gene is under the transcriptional control of a strong promoter from the hel gene (The Institute for Genomic Research, TIGR locus HI0693) encoding an outer-membrane lipoprotein involved in NAD and NMN uptake (Green et al., 1991; Reilly et al., 1999; Kemmer et al., 2001), which we had found by Northern analysis to be highly expressed (data not shown). Rhel-licA contains the hel promoter immediately 5′ of three in-frame potential licA initiation codons and an in-frame deletion of the CAAT repeat units (Experimental procedures).
Total RNA from Rhel-licA grown aerobically or microaerobically was examined by reverse transcriptase quantitative real-time PCR (RT-qPCR) to assess the level of abundance of licA-specific transcripts (Fig. 3A). Control reactions using primers specific for the hel gene verified that equivalent amounts of RNA were analysed for each strain (Fig. 3A, legend). Results of the RT-qPCR with primers specific for licA showed an ∼7.5-fold and ∼13-fold increase in the level of licA cDNA in Rhel-licA compared to parent strain Δrep in cultures grown aerobically (Fig. 3A, columns 1 and 2) and microaerobically (Fig. 3A, columns 3 and 4) respectively. The ∼7.5-fold aerobic increase corresponding to the level of licA mRNA in Rhel-licA compared to Δrep did not result in an increase of PC epitope levels in Rhel-licA (Fig. 3B, lanes 1 and 2). Similarly, the microaerobic increase in licA mRNA levels in Rhel-licA also did not increase the PC epitope levels under this growth condition (Fig. 3B, lanes 3 and 4). Thus, overexpression of lic1 mRNA from a strong promoter did not alleviate the low aerobic PC epitope levels. Although the lic1 locus is essential for PC epitope production, increasing lic1 expression is not sufficient to alter the regulation of PC epitope levels.
Differential expression of genes of carbohydrate metabolism in response to PC modulating conditions
The results described above indicate that lic1 regulation alone does not account for modulation of this epitope by redox conditions of growth and other mechanisms are likely to be involved. Building of the complete LOS structure requires numerous biosynthetic steps to produce the final configuration for PC attachment. These steps are potentially subject to regulation. The observed changes in lic1 expression in response to culture aeration levels coupled with the inability of lic1 overexpression to induce an aerobic increase in PC levels suggest a model in which coordinate regulation of lic1 and additional genes involved in LOS biosynthesis is responsible for modulation of PC epitope display. To identify new and previously characterized genes/pathways that are involved in modulation of the LOS-PC epitope in H. influenzae, we examined the microarray expression data. Several genes detected by expression profiling to be more highly expressed aerobically or microaerobically are components of metabolic pathways related to LOS biosynthesis. Figure 4 illustrates a model of known and predicted pathways of central carbohydrate metabolism of H. influenzae as it relates to LOS biosynthesis and the gene expression patterns we detected in H. influenzae with expression profiling as discussed below. Additional interpretation of the microarray data not directly discussed here is available as supplemental discussion in the web supplement.
Aerobically induced genes
Under the aerobic condition, microarray analysis detected higher expression levels of genes involved in conversion of sugars to energy (see Fig. 4 and Table 1 for reference to genes discussed below). The pgk gene (HI0525) encoding the glycolytic enzyme, phosphoglycerate kinase was induced ∼twofold. Thiamine pyrophosphate (TPP), the biologically active form of thiamine, is a cofactor required by several metabolic enzymes critical for conversion of carbohydrates to energy. TPP is required by pyruvate dehydrogenase for oxidative decarboxylation of pyruvate, transketolase for metabolism of pentose sugars and α-ketoglutarate dehydrogenase for oxidation of α-ketoglutarate (Begley et al., 1999). Higher aerobic expression levels were detected for thiM (HI0415) (5.6-fold), thiE (HI0417) 2.7-fold and thiD (HI0416) (5.5-fold, web supplement, TableS1), encoding putative homologues of E. coli enzymes known to generate the thiazole and pyrimidine moieties that form thiamine pyrophosphate in that organism (Vander Horn et al., 1993). We also observed higher aerobic expression of HI0357 (4.5-fold), a putative thiamine biosynthesis gene located in a probable operon with HI0354, HI0355 and HI0358. These three genes have conserved domains that show sequence similarity to proteins that function in either thiamine transport or biosynthesis and were also highly expressed aerobically (∼six- to 11-fold). Therefore, the expression pattern suggests increased cofactor production for pyruvate dehydrogenase and elevated phosphoglycerate kinase expression, though post-transcriptional regulation also likely contributes to the activity of these enzymes. Because these enzymes mediate entry of glycolytic products into the citric acid cycle for energy generation, their increased expression could lead to a relative depletion under the aerobic condition of sugar precursors needed for synthesis of the LOS outer core to which the PC epitope is attached.
Microaerobically induced genes
Several genes with known or potential functions in LOS synthesis distinct from PC epitope display were more highly expressed microaerobically (see Table 2 for reference to genes discussed below). We observed higher microaerobic expression of lgtD (HI1578) twofold, which encodes an N-acetylgalactosaminyltransferase for addition of an N-acetylgalactosamine LOS extension in H. influenzae (Shao et al., 2002). This specific modification occurs at heptose III of the LOS inner core, distinct from heptose I that is linked to the glucose residue substituted with PC in Rd (Risberg et al., 1999). Expression of a cluster of putative glycosyltransferase genes (HI1695, HI1697, HI1698) and a probable sialyltransferase gene (HI1699) with proposed functions in lipopolysaccharide biosynthesis was also increased ∼2.5-fold. Modulation of these genes in response to redox conditions suggests that, in addition to PC epitope display, other components of the LOS structure could potentially respond to this condition.
Expression profiling also detected a microaerobic increase in levels of the licABCD genes required for LOS-PC display (Table 2, TableS1 in web supplement) and this result was confirmed by primer extension (Fig. 1B) and Northern analysis (data not shown). Another LOS related gene, galU (HI0812), appeared to be moderately increased under this condition (1.5-fold) (web supplement, TableS1). In E. coli, galU encodes a UDP-glucose pyrophosphorylase that catalyses the interconversion of glucose-1-phosphate to uridine diphosphate glucose (UDP-glucose) (Weissborn et al., 1994), the activated form of the sugar used in the biosynthesis of various carbohydrates, including LPS (Sundararajan et al., 1962). E. coli galU mutants produce a truncated LPS core that lacks glucose and galactose (Sundararajan et al., 1962). H. influenzae GalU is 72% identical to E. coli GalU at the predicted amino acid level and galU mutants in H. influenzae also produce truncated LOS molecules (Hood et al., 1996b). Although the galU induction level was low as detected by the microarray hybridization experiment, the importance of this gene in linking metabolism to LOS biosynthesis led us to evaluate this result by Northern analysis and subsequent confirmation of its increased expression under the microaerobic condition is described below.
Overview of expression profiling data and potential pathway for LOS-PC modulation
The microarray data indicates a global expression pattern consistent with an increase in the breakdown of carbohydrates for energy generation aerobically, which would predict a decrease in LOS precursor abundance for LOS biosynthesis (Fig. 4). Depletion of precursors such as UDP-glucose could lead to a decrease in modification of the LOS outer core, and contribute to the aerobic decrease in LOS-PC epitope levels. In microaerobiosis, several genes with known and proposed roles in LOS/LPS modification or biosynthesis were upregulated including licABCD, needed for generating the PC epitope and attaching it to LOS, in addition to galU which is required for production of UDP-glucose, the precursor of the glucose residue of the LOS to which PC is added. Together these changes suggested that redox modulation of LOS-PC display could be mediated by differential abundance and incorporation of glucose residues into the LOS combined with modulation of the licABCD locus itself.
Global regulator csrA modulates PC epitope levels
The microarray results indicated that the aerobic and microaerobic culture conditions used here are likely to influence central carbon flux. In E. coli, central carbohydrate metabolic pathways are subject to regulation by a conserved, pleotropic post-transcriptional regulator, CsrA (Romeo, 1998). CsrA negatively regulates glycogen biosynthesis (Romeo et al., 1993; Sabnis et al., 1995), glycogen degradation (Yang et al., 1996) and gluconeogenesis (Sabnis et al., 1995), and positively regulates glycolysis (Sabnis et al., 1995). We hypothesized that the putative csrA (HI0813) homologue in H. influenzae (67% amino acid identity to E. coli CsrA) could potentially regulate aspects of similar pathways, thereby affecting the cellular flux of sugar precursors needed for LOS modification. This model predicts that a deletion of csrA will alter the level of LOS-PC epitope display. To investigate this possibility, we took advantage of an ordered mutant bank we previously generated by in vitro mutagenesis of each 10 kb segment of the Rd genome with a mariner derived minitransposon (Akerley et al., 2002). This bank provides us with access to a transposon mutation in any gene of interest that is nonessential for growth, and we used it to obtain the mutant containing an insertion in csrA. We examined the LOS-PC phenotype of the csrA transposon insertion mutant and found elevated levels of PC epitope under aerobic and microaerobic conditions (data not shown). To rule out effects that may be due to transcriptional polarity because galU is directly downstream of csrA in the H. influenzae genome, we generated an in-frame csrA deletion mutation and evaluated the effects of CsrA on modulation of PC epitope levels. Our Western blot results showed an increase in PC epitope levels in the non-polar csrA mutant compared to the wild-type under the aerobic culture condition suggesting that CsrA is required for the decrease in epitope production observed under this condition (Fig. 5). CsrA may negatively affect the level of the sugar precursors needed for LOS extension, similar to the repressive effects of CsrA on gluconeogenesis and glycogen biosynthesis in E. coli (Romeo, 1998).
Regulation of galU mRNA expression by redox conditions of growth and csrA
The sugar residues of the H. influenzae LOS outer core are the target substrates for PC addition (Risberg et al., 1999; Schweda et al., 2000). Therefore, CsrA-mediated regulation of galU could contribute to csrA's effect on PC epitope levels by controlling glycosyl residue addition to the LOS outer core. We investigated both redox regulation of galU and the potential role of CsrA in the regulation of galU in H. influenzae by Northern analysis in the csrA mutant versus wild-type under aerobic and microaerobic conditions (Fig. 6A). Our csrA deletion mutation is a precise in-frame replacement of the csrA coding region with the coding region of the kanamycin resistance gene, aphI (Fig. 6B). Northern analysis showed that the level of abundance of the ∼0.9 kb galU mRNA was increased in the csrA deletion mutant under both conditions as compared to wild-type (Fig. 6A). The changes in culture aeration did not affect the level of abundance of the 0.9 kb galU-specific transcript, which was a minor species in wild-type Rd and was markedly more abundant in the csrA mutant (Fig. 6A). Longer transcripts containing both csrA and galU were also detected. A multicistronic transcript of ∼1.5 kb hybridizing to both galU- and csrA-specific probes in the wild-type was more abundant microaerobically (Fig. 6A, lanes 1, 3, 5 and 7). Consistent with the insertion of the kanamycin resistance gene and deletion of csrA, the ∼1.5 kb csrA/galU transcript was absent in the csrA mutant and replaced by a ∼2.4 kb transcript which hybridized to galU but not csrA. Similar to the csrA/galU transcript of the wild-type strain the ∼2.4 galU hybridizing transcript in the csrA mutant was more abundant under microaerobic conditions.
Therefore, similar to the modulation pattern of the PC epitope, regulation of galU mRNA abundance in H. influenzae is mediated through both csrA and in response to culture aeration levels during growth. Consistent with the genomic organization of these genes in a potential operon, we detect both a monocistronic galU transcript and multicistronic transcripts containing csrA and galU. Irrespective of the growth condition, deletion of csrA leads to increased abundance of the monocistronic galU mRNA in H. influenzae, while the multicistronic transcript containing galU is more abundant in microaerobiosis versus aerobiosis in both wild-type and the csrA mutant. An increase in the galU-encoded UDP-glucose pyrophosphorylase is likely to increase levels of the cellular UDP-glucose pool, potentially contributing to increased incorporation of glycosyl residues in the LOS for subsequent PC attachment.
Modulation of PC phenotype
In this report, we demonstrate that environmental conditions influence lic1 transcript abundance and the level of mAb reactivity specific to the LOS-PC epitope of H. influenzae, an important surface antigen contributing to the pathogenicity of this organism. The levels of steady-state lic1 transcripts as well as PC epitope were highest under microaerobiosis and repressed under high aeration. This modulation is not mediated through phase variation of lic1. Overexpression of lic1 mRNA could not relieve aerobic repression of PC epitope levels. However, we have observed that decreasing lic1 expression leads to decreased PC epitope levels as expected (data not shown) because the genes of the lic1 locus are essential for PC epitope addition to the LOS. While lic1 regulation is likely to contribute to modulation of PC epitope levels, our results indicate a role for additional or alternative mechanisms for redox modulation of this epitope.
Global gene expression patterns and pathways of carbon metabolism
Genomic expression profiling performed on cells grown under the conditions that modulate PC epitope levels indicated gene expression patterns that are consistent with aerobic depletion of precursor sugars of the LOS outer core, microaerobic induction of LOS synthesis and PC epitope production (Tables 1 and 2, Fig. 4). We observed an aerobic increase in expression of genes with known and proposed functions in thiamine biosynthesis and of a putative glycolytic enzyme gene, pgk (phosphoglycerate kinase), a key enzyme of glycolysis. Several thiamine-dependent enzymes that play important roles in carbohydrate metabolism require TPP as a cofactor, including pyruvate dehydrogenase (PDH). Increased levels of PDH activity could signify a faster rate of sugar utilization for energy generation, thereby reducing the levels of precursors available for LOS modification.
Conversely, microaerobiosis led to increased expression of several genes with known or proposed roles in LOS biosynthesis or modification including both licABCD and galU, the latter a glucose pyrophosphorylase required for addition of glucosyl residues to the LOS outer core on which the PC epitope is displayed. In general, the expression profile suggested that these LOS-PC modulating conditions could exert changes in the generation and utilization of LOS precursors by central sugar metabolic pathways. If this model is correct, then disruption of the ratio of glycolysis to carbohydrate synthesis should influence the levels of LOS-PC. Therefore, these results led us to examine the role of a putative regulator of these metabolic pathways, csrA, which could influence the balance of sugar precursors needed for LOS modification.
csrA and galU
Disruption of csrA in E. coli results in increased gluconeogenesis by increasing expression of genes or enzymes of central carbohydrate metabolism (pckA, phosphoenolpyruvate carboxykinase; Fbp, fructose-1,6-bisphosphatase; Pps, phosphoenolpyruvate synthase; and Pgm, phosphoglucomutase), and decreased glycolysis by decreasing levels of glycolytic enzymes (Pgi, glucose-6-phosphate isomerase; PfkA, 6-phosphofructokinase; Tpi, triosephosphate isomerase; and Eno, enolase) (Romeo, 1998). Homologues of csrA are widespread among plant and human pathogenic bacteria and act mainly as negative regulators of virulence properties, such as the production of extracellular proteases, quorum sensing molecules, secondary metabolites (e.g. hydrogen cyanide, pyocyanin) (Chatterjee et al., 1995; Cui et al., 1995; Pessi et al., 2001) and regulation of invasion genes (Lawhon et al., 2003). Furthermore, in the phytopathogen, Erwinia amylovora, the csrA homologue, rsmA represses the production of capsular exopolysaccharide, which is required for pathogenicity in this organism (Ma et al., 2001).
The non-polar csrA mutant in H. influenzae showed an increase in PC epitope levels under aerobic culture conditions compared to wild type. The abundance of the monocistronic galU transcript was dramatically increased in the H. influenzae csrA mutant as well. This change is predicted to lead to increased formation of UDP-glucose, a donor of glycosyl groups needed for LOS extension and substrate for PC addition, providing a potential mechanism for csrA-mediated PC regulation.
To our knowledge, csrA has not been reported to regulate galU in any other bacteria. Mutants deficient in galU have defective LOS/LPS structures and are attenuated for colonization in animal models of H. influenzae infection and galU is a major virulence factor in several other bacterial pathogens as well (Hood et al., 1996b; Rioux et al., 1999; Marra and Brigham, 2001; Nesper et al., 2001). Given the role of csrA in regulation of galU, a gene that is critical for H. influenzae pathogenesis, it will be important to examine whether the H. influenzae csrA regulates additional virulence genes and whether it controls gluconeogenic and glycolytic genes as does its probable homologue in E. coli. Such studies can provide insight into the relative contributions of csrA and its regulation of galU in H. influenzae pathogenesis.
Culture aeration conditions influenced galU mRNA production by a mechanism that appears to be independent of csrA-mediated regulation. Levels of the multicistronic csrA/galU transcript, the predominant form of galU mRNA detected in wild-type cells, increased in microaerobiosis compared to aerobiosis and this relative increase under the microaerobic condition in the csrA mutant was retained, indicating that csrA is not required for control of galU transcript levels in response to these conditions.
Therefore, CsrA and the signals generated by varying culture aeration levels are not necessarily components of the same regulatory pathway controlling galU mRNA abundance levels. Additional studies are required to determine whether CsrA-mediated control of the PC epitope can be dissociated from the response of PC epitope levels to these conditions that modulate expression patterns of redox responsive genes.
CsrA-mediated control suggests an additional signal may participate in PC epitope regulation, although signals that could regulate activity of CsrA in H. influenzae are not known. The activity of CsrA homologues in other species is modulated by small untranslated regulatory RNAs (sRNAs) first identified in E. coli (Romeo, 1998). sRNAs are in turn regulated in a complex manner involving transcriptional control by a two-component signal transduction system and quorum sensing (Heeb et al., 2002; Valverde et al., 2003; Weilbacher et al., 2003). No cognate sRNA-controlling csrA in H. influenzae has been identified. It is possible that H. influenzae modulates the PC epitope under several distinct environmental conditions, those that trigger CsrA activity and those that activate redox signalling systems. Alternatively, H. influenzae may encounter a condition that activates both regulatory pathways simultaneously.
Our data implicate a multifactorial mechanism for modulation of PC epitope display in H. influenzae. First, our microarray and Northern data indicate that multiple genes contributing to LOS biosynthesis are regulated by redox conditions of growth and it is likely that this coordinated regulation is required for modulation of LOS-PC. Combined regulation of the lic1 operon and galU together is likely to account, at least in part, for the modulation of PC epitope levels under these conditions, with the possibility that regulation of genes and enzymes of central carbohydrate metabolism also plays a role. Second, our Northern data indicate that csrA mediates negative control of galU. Given the role of csrA in other organisms, its putative homologue in H. influenzae is likely to coordinately regulate additional factors such as genes of central carbohydrate metabolism that could participate in modulating PC epitope levels. Together these results provide insight into the mechanism of PC epitope modulation and indicate that additional virulence-associated LOS modifications are likely to be influenced by csrA and redox conditions of growth.
These studies are consistent with a model in which multiple systems regulate genes of carbohydrate metabolism and LOS synthesis in H. influenzae and this control provides a likely mechanism for modulation of PC epitope levels in response to culture aeration levels. The ability to control levels of PC epitope displayed on the bacterial surface in response to signals could play a role in rapid adaptation to environmental and physiological conditions encountered during infection. This regulation would provide a mechanism complementary to phase variation for inactivating epitope production under unfavourable conditions. For example, initial attachment to the mucosal surface likely represents an aerobic condition in which very low numbers of bacteria must avoid host innate immune defences such as PC-reactive antibodies, a major component of the innate B-cell repertoire that is well represented at the mucosal surface (Conley and Briles, 1984). Subsequently, H. influenzae is thought to grow on the mucosal surface and form microcolonies or biofilms (Ehrlich et al., 2002). Notably, comparisons between plate-grown cells and cells grown in a biofilm model system in vitro indicate numerous antigenic changes including modulation of sialic acid residues and an epitope that is part of the LOS core and includes KDO (Campagnari et al., 1990; Murphy and Kirkham, 2002; Greiner et al., 2004). Microenvironments generated by such multicellular structures during bacterial colonization are likely to include low oxygen or reducing conditions, which our data indicate can lead to increased PC levels. Because PC has been implicated in the invasion of epithelial mucosal barrier (Swords et al., 2000), an increase in PC can facilitate spread to a submucosal niche or invasion into the bloodstream. The LOS structure mediates diverse interactions with the host and it will be important to determine whether the PC epitope and other LOS modifications are controlled by environmental signals during infection and pathogenesis.
Haemophilus influenzae growth conditions
The non-encapsulated Rd derivative of H. influenzae type d (BA042) (Akerley et al., 2002) was grown at 35°C in Brain Heart Infusion agar or broth supplemented with 10 µg ml−1 nicotinamide adenine dinucleotide and 10 µg ml−1 haemin (sBHI). DNA was transformed into naturally competent H. influenzae prepared as previously described (Barcak et al., 1991). Kanamycin (Km) and Tetracycline (Tet) were added to sBHI at 20 µg ml−1 and 8 µg ml−1 respectively (Table 3).
Table 3. Bacterial strains and plasmids used in this study.
Strains and plasmids
Source or reference
. Deletion of 16 of 17 CAAT repeats with in-frame positioning of three potential licA ATG initiation codons.
Non-encapsulated H. influenzae Rd; referred to as Rd or wild-type in this study
pLic1 containing a deletion of lacZα from the pCR-BluntII-TOPO vector pLic1 containing deletion of 16 tandem CAAT repeats and translational in-frame positioning of three potential licA ATG initiation codons
This study This study
pΔCAAT containing deletion of lacZα from the pCR-BluntII-TOPO vector
pLic1ΔZα containing E. coli lacZ expressed from the putative licA initiation codons
pΔCAAT2 containing replacement of the licA coding region with E. coli lacZ
pΔCAAT2 containing Phel immediately upstream of the licA ORF
pCR-BluntII-TOPO vector carrying 3.8 kb PCR product containing wild-type csrA and flanking genomic regions
Deletion construct derived from p814-812 containing aphI KmR cassette from Tn903 in place of the csrA ORF
Plasmid and H. influenzae strain construction
Standard molecular biology methods were used for plasmid construction, primer extension, Northern and Western blot analysis (Ausubel et al., 1995). Strain RlicA41 was isolated from an ordered mutant strain collection representing the genome of H. influenzae Rd mutagenized with magellan1, a derivative of the mariner-family transposon Himar1 as described previously (Akerley et al., 2002). All primer sequences are provided as a web supplement (TableS2).
Plasmid pLic327 contains 5′ sequences upstream of licA amplified with primers licA5′1 and licAorfout2. The BamHI-digested PCR product was cloned into a derivative of pXT10 (Wong and Akerley, 2003) containing E. coli lacZ (encoding β-galactosidase) and the aphI kanamycin resistance gene. Strain Δrep which contains a deletion of 16 tandem CAAT repeats in licA and an in-frame positioning of three potential licA ATG initiation codons was created as follows: a 5.44 kb product was amplified from Rd using primers HI1534-5′ and HI1540-3′, and cloned into the pCR-Blunt II-TOPO vector (Invitrogen) to create pLic1. pLic1 was used as template in PCR with primers, licAMout2 and licAΔrep to generate a ∼8.9 kb product that was digested with StuI and recircularized to generate plasmid, pΔCAAT. pΔCAAT was transformed into a Rd derivative, strain ZlicA containing E. coli lacZ translationally fused to licA immediately 3′ of the licA ATG initiation codon γ and screened for white colonies on 5-brom-4-chloro-3-indolyl-beta-D-galactopyranoside (Xgal) plates to create strain Δrep. Strain ZlicA was created as follows: the E. coli lacZ gene was amplified with primers 5′Z-ATG2 and 3′Z-TAA, digested with AscI and StuI and ligated to a ∼8.5 kb PCR product of a pLic1 derivative, pLic1ΔZα (contains a deletion of lacZα) amplified with primers licAMout2 and licAΔrep and digested with MluI and StuI to create pZlicA. pZlicA was transformed into Rd to create ZlicA.
Strain Rhel-licA, which contains the native licA gene under the transcriptional control of the putative hel (HI0693) promoter was created as follows: pΔCAAT was digested with BsaI and NotI to remove lacZα, followed by klenow end-filling and recircularization to generate pΔCAAT2. pΔCAAT2 was used as template in PCR with primers licATG1out and licATG1in to generate a ∼8.5 kb PCR product that was digested with SalI and ligated to an ∼294 bp PCR product containing the putative hel promoter sequence amplified from Rd with primers hel5′ATGout and 692–5′ATGout and digested with SalI to create plasmid pHel-licA. pHel-licA was transformed into strain ZlicA and screened for white colonies on X-gal plates to create Rhel-licA. Strain ΔlicA containing a licA deletion was created as follows: the E. coli lacZ gene was amplified as described above and ligated to a ∼7.6 kb PCR product amplified from template pΔCAAT2 with primers licAMout2 and licA3′ORF and digested with MluI and StuI to create pZΔlicA. Rd was transformed with pZΔlicA to create strain ΔlicA.
Strain Δ8kan which contains a non-polar deletion of csrA was created as follows: A 3.8 kb product was amplified from Rd using primers HI0814-5′ and HI0812-3′, and cloned into the pCR-Blunt II-TOPO vector (Invitrogen) to create p814-812. p814-812 was used as template in PCR with primers, 8ATGOUT and 8TAGOUT to amplify an ∼7 kb PCR product that was digested with StuI and SalI. This fragment was ligated to a StuI- and SalI-digested PCR product containing the kanamycin resistance gene, aphI from Tn903 amplified with primers MER5kanSDATG and MER3kanTAA (Wong and Akerley, 2003) to create pΔ8kan. Rd was transformed with pΔ8kan to create Δ8kan. Strains generated in this report were confirmed to contain the appropriate mutations by sequencing or PCR amplification across recombinant junctions of the respective mutations with the H. influenzae chromosome.
Western blot analysis
Haemophilus influenzae Rd was grown in sBHI at 35°C to an OD600 = 0.2–0.3 under culture aeration conditions ranging from 10 ml to 300 ml in 500 ml Erlenmeyer flasks shaken at 250 r.p.m. Unaerated cultures filled to capacity in sealed 50 ml tubes were grown similarly. 0.25–0.5 OD600 units of cells were pelleted and resuspended in solution 21, a component of the M-IV competence inducing medium (Barcak et al., 1991) followed by addition of loading buffer containing 2-mercaptoethanol and boiled at 100°C for 5 min. Boiled whole cell lysates (0.1–0.25 OD600 units) were separated by SDS-PAGE with 18% polyacrylamide gels and electrotransferred onto Immobilon-P (Millipore Corporation). Equivalent numbers of cells in whole-cell lysates were loaded in each lane. Equal loading was verified either by staining replicate gels run in parallel with Commassie Blue or by staining the upper one-eighth of the transferred blot with Ponceau S. Immunoblotting was performed using a 1:10 000 dilution of anti-PC IgA mAb TEPC 15 (Sigma-Aldrich) and bands visualized using either anti-mouse immunoglobulin A conjugated to peroxidase or alkaline phosphatase (Rockland Immunochemicals). To visualize the LOS, whole-cell lysates from 0.02 OD600 units of cells were resolved by SDS-18% PAGE, stained with the fluorescent dye, Pro-Q Emerald 300 of the Pro-Q Emerald 300 Lipopolysaccharide Gel Stain Kit (Molecular Probes), and visualized with UV transillumination for photography.
Northern and primer extension analysis
Total RNA from H. influenzae Rd was obtained from cultures grown in sBHI to OD600 = 0.3–0.4 under varied culture aeration conditions ranging from 10 ml to 200 ml and unaerated as described above. RNA was isolated using TRIzol Reagent (Invitrogen), treated with DNase I (Ambion) and phenol extracted. For Northern blotting, 15 µg of total RNA was electrophoresed on a 1.5% agarose gel containing 1.1% formaldehyde and transferred to a Nytran nylon membrane (Amersham Pharmacia Biotech). Probes were generated by amplification from Rd using 5′ and 3′ primer pairs for licA (HI1537), licB (HI1538), licC (HI1539), licD (HI1540), galU (HI0812) and csrA (HI0813). PCR products were labelled with the Gene Images AlkPhos Direct Labeling Kit and signals visualized with CDP-Star chemiluminescent detection system (Amersham Pharmacia Biotech). Washing and hybridization conditions were according to the manufacturer's instructions. Primer extension analysis was performed on 20 µg total RNA from Rd using a [γ-32P]-ATP (Amersham Pharmacia Biotech) labelled primer, licAorfout2 which is located 99 bp 3′ of the putative licA ATG start codon γ (Fig. 1B). Products were analysed by electrophoresis in a 7 M urea, 6% polyacrylamide gel. Sequence ladder was generated using the same [γ-32P]-ATP-labelled primer with plasmid pLic327 as template using the Sequenase 2.0 DNA Sequencing Kit (USB Corporation).
Total RNA from triplicate cultures of H. influenzae Rd grown aerobically or microaerobically (10 ml and 200 ml in 500 ml Erlenmeyer flasks respectively) to OD600 = 0.3–0.4 was obtained and treated with DNase I as described above. A total of 8 µg of RNA from each triplicate culture was used as template for generation of cDNAs using random primers (New England Biolabs) followed by coupling to the fluorescent dye, Cy3 (Amersham Biosciences) according to the protocol for the BD Atlas PowerScript Fluorescent Labeling Kit (BD Biosciences Clontech). Fluorescently labelled cDNAs were used in hybridization on Corning UltraGAPS slides printed with the H. influenzae Genome Oligo set (Qiagen Operon) using a GMS 417 Arrayer (Affymetrix). The genome set contains 1714 optimized 70mer probes representing 1714 H. influenzae Rd open reading frames (ORFs) and 12 unique negative control probes. Controls for cDNA synthesis and fluorescent label coupling were also printed onto each slide to ascertain the efficiency of these reactions using the reagents supplied with the BD Atlas PowerScript Fluorescent Labeling Kit. Oligos were diluted in 50% dimethyl sulphoxide to a concentration of 40 micromolar and heated at 96°C for 2 min prior to printing in triplicates onto each slide at a constant temperature of 22°C and 40–50% relative humidity. Slides were hybridized at 42°C in hybridization chambers (Corning) for 16–24 h, washed as described (http://brownlab.stanford.edu) and scanned using a GMS 418 Array Scanner (Affymetrix). Images were processed and hybridization signals quantified with ImaGene and GeneSight (BioDiscovery). The total signal intensity for every gene represented on the array was corrected by subtracting the local background, merging identical spots by obtaining the average signal intensity value from triplicate spots on the same slide, and normalized by dividing by the mean of the values for all of the genes represented on the array. The corrected signal intensity for each gene represents the mean of triplicate samples from three independent hybridization experiments to Cy3-labelled cDNAs derived from independent cultures grown aerobically and microaerobically. Expression ratio data were generated by comparing the corrected mean signal intensity values from arrays hybridized to cDNA generated from aerobically versus microaerobically grown cultures. Statistical analysis of the expression data was performed with the Cyber-T Bayesian statistics program from the Institute for Genomics and Bioinformatics at the University of California, Irvine (http://visitor.ics.uci.edu). Genes expressed more highly under one condition were considered to be induced by that condition relative to the other during the experiment. Genes whose expression ratios had Bayesian P-values based on the regularized t-test ≤0.01 were considered to be significant in their fold induction.
Reverse transcriptase-quantitative PCR (RT-qPCR)
Quantification of licA and hel mRNA expression in strains Rhel-licA and Δrep grown aerobically (10 ml cultures) and microaerobically (200 ml cultures) was performed using the DyNAmo SYBR green qPCR kit (MJ Research) in quantitative real-time PCR measured with the DNA Engine Opticon II System (MJ Research). Briefly, 0.5 µg of DNase I-treated total RNA from strains Rhel-licA and Δrep, grown aerobically or microaerobically, was used as template in cDNA synthesis using SuperScript II reverse transcriptase (Invitrogen). licA cDNA was generated using primer licAmid3′. hel cDNA was generated using primer 693int3. A total of 1/20 of the reverse transcriptase reactions were used as template in qPCR for amplification of licA using primers, licArep and licAmid3, and hel using primers, 693-5 and 693int3. Real-time cycler conditions were as follows: 95°C for 2 min, followed by 39 cycles of 96°C for 30 s, 58°C for 30 s, and 72°C for 1 min, followed by one cycle of 72°C for 7 min. Fluorescence was read at 78°C. Control reactions were performed in parallel with mock cDNA reactions generated without reverse transcriptase. Samples were electrophoresed on agarose gels to confirm product sizes. A standard curve (r2 ≥ 0.98) was generated from a dilution series of wild-type Rd genomic DNA as template using each primer pair in a parallel set of reactions in qPCR. Quantification of mRNA expression of the hxuAC, dmsAC, nrfABCD genes from Rd wild-type was measured in real-time PCR assays essentially as described above, except random primers (New England Biolabs) were used to generate the cDNA templates used in qPCR. The 5′ and 3′ primer pairs used in qPCR are specific for hxuA (HI0264), huxC (HI0262), dmsA (HI1047), dmsC (HI1045), nrfA (HI1069), nrfB (HI1068), nrfC (HI1067), nrfD (HI1066).
We thank Colin Cox and Jennifer Wyrzkowski-Lapierre for technical assistance and TIGR for finished and unfinished microbial genome sequences. We thank Drs John Leong and Kishore Alugupalli for helpful comments and advice on the manuscript. This work was supported in part by grants from the National Institutes of Health (AI49437) and the American Heart Association to B.J.A.; a postdoctoral fellowship from the American Cancer Society and in part, a grant from the Cystic Fibrosis Foundation to S.M.W.