Mycoplasma pneumoniae Community Acquired Respiratory Distress Syndrome toxin expression reveals growth phase and infection-dependent regulation

Mycoplasma pneumoniae causes acute and chronic respiratory infections, including tracheobronchitis and community acquired pneumonia, and is linked to asthma and an array of extra-pulmonary disorders. Recently, we identified an ADP-ribosylating and vacuolating toxin of M. pneumoniae, designated Community Acquired Respiratory Distress Syndrome (CARDS) toxin. In this study we analysed CARDS toxin gene (annotated mpn372) transcription and identified its promoter. We also compared CARDS toxin mRNA and protein profiles in M. pneumoniae during distinct in vitro growth phases. CARDS toxin mRNA expression was maximal, but at low levels, during early exponential growth and declined sharply during mid-to-late log growth phases, which was in direct contrast to other mycoplasma genes examined. Between 7% and 10% of CARDS toxin was localized to the mycoplasma membrane at mid-exponential growth, which was reinforced by immunogold electron microscopy. No CARDS toxin was released into the medium. Upon M. pneumoniae infection of mammalian cells, increased expression of CARDS toxin mRNA was observed when compared with SP-4 broth-grown cultures. Further, confocal immunofluorescence microscopy revealed that M. pneumoniae readily expressed CARDS toxin during infection of differentiated normal human bronchial epithelial cells. Analysis of M. pneumoniae-infected mouse lung tissue revealed high expression of CARDS toxin per mycoplasma cell when compared with M. pneumoniae cells grown in SP-4 medium alone. Taken together, these studies indicate that CARDS toxin expression is carefully controlled by environmental cues that influence its transcription and translation. Further, the acceleration of CARDS toxin synthesis and accumulation in vivo is consistent with its role as a bona fide virulence determinant.


Introduction
Mycoplasma pneumoniae is a significant bacterial pathogen of the airways and accounts for 20-30% of all community acquired pneumonia. It is also implicated in other airway diseases including asthma, and in extrapulmonary manifestations, including neurological, gastrointestinal and dermatological disorders (Baseman and Tully, 1997;Waites and Talkington, 2004). M. pneumoniae colonizes the surfaces of epithelial cells and is also capable of invading host cells and establishing intracellular residence (Baseman et al., 1995;Dallo and Baseman, 2000). Recently, we identified a unique virulence factor designated Community Acquired Respiratory Distress Syndrome (CARDS) toxin, an ADP-ribosylating and vacuolating toxin that binds alveolar surfactant protein A (Kannan et al., 2005;Kannan and Baseman, 2006), likely contributing to additional colonization and pathogenic pathways. CARDS toxin ADP-ribosylates both similar and distinct human cell proteins when compared with the S1 catalytic subunit of pertussis toxin, leading to a cascade of events such as tissue disorganization, inflammation and airway dysfunction along with cell vacuolization (Kannan and Baseman, 2006). Hamster, murine and chimpanzee animal models and in vitro studies with tracheal organ cultures and human cell cultures have provided important insights in defining M. pneumoniae virulence potential (Hara et al., 1974;Hu et al., 1975;Barile et al., 1981;Baseman et al., 1995;Dallo and Baseman, 2000;Hardy et al., 2002;Kannan and Baseman, 2006). Recently, normal human bronchial epithelial (NHBE) cells were used as a model to study M. pneumoniae interactions (Krunkosky et al., 2007). NHBE cells maintained in the air-liquid interface culture system exhibit well differentiated heterogenous populations of ciliated and secretory cells remarkably similar to the lumen of the airway.
Clearly, M. pneumoniae must co-ordinate a wide range of virulence factors and circumvent host defenses in order to colonize, propagate, internalize, persist and be transmitted. Transcriptional and translational regulation in M. pneumoniae appears to be unique compared with other procaryotes, as this mycoplasma possesses only one authentic sigma factor and a limited number of genes encoding typical transcriptional and translational regulatory elements (Himmelreich et al., 1996;Dandekar et al., 2000), including helix-turn-helix (HTH) motifs (Samuelsson et al., 1997). Although there is no classical twocomponent system in M. pneumoniae (Himmelreich et al., 1996), there is evidence that M. pneumoniae is able to differentially regulate gene expression in response to environmental stimuli. For example, transcriptional regulation of mycoplasma heat shock genes has been observed in M. pneumoniae and other pathogenic Mycoplasma species (Weiner et al., 2003;Madsen et al., 2006;Musatovova et al., 2006;Chang et al., 2008;Kannan et al., 2008). Differential expression of lipoprotein genes in M. pneumoniae after acidic and oxidative stresses (Hallamaa et al., 2008) and regulation of ackA (acetate kinase) and ldh (lactate dehydrogenase) genes by glycerol (Halbedel et al., 2007) were recently reported. After contact with human lung epithelial cells, distinct patterns of M. pneumoniae lipoprotein gene expression were observed (Hallamaa et al., 2008). While upregulation of four heat shock genes (dnaJ, dnaK, lon and clpB) can be attributed to a functioning HrcA-CIRCE regulatory apparatus (Narberhaus, 1999;Weiner et al., 2003), in most cases the mechanisms of gene expression and regulation have not been identified. In addition, little is known regarding post-transcriptional and translational controls in M. pneumoniae, although evidence from two-dimensional gel electrophoresis, lipoprotein characterization, cleavage of signal peptides and phosphorylation analyses suggest that these processes exist (Razin et al., 1998;Regula et al., 2000;Ueberle et al., 2002;Jaffe et al., 2004).
Because only a paucity of data is available concerning regulation of virulence genes in M. pneumoniae, we focused on the identification of the putative promoter of the cards gene and monitored cards transcript levels during M. pneumoniae in vitro growth and after contact with host cells. We further demonstrated surface localization of CARDS toxin on intact mycoplasma cells with no evidence for release into the environment. Interestingly, we noted substantial increases in the synthesis of CARDS toxin protein per mycoplasma cell in infected mice. These data suggest that understanding how airway-associated environmental signals regulate CARDS toxin expression should provide important clues concerning M. pneumoniae virulence and associated pathologies.

cards gene organization and promoter mapping
In M. pneumoniae reference strain M129, the cards gene (mpn372, nucleotides 444341-446116) is flanked by mpn371 (nucleotides 444187-443552) and mpn373 (nucleotides 446741-446127) genes. Both mpn371 and mpn373 genes are transcribed from the complementary strand, in contrast to cards, and encode hypothetical proteins of unknown functions (Fig. 1A). cards is separated from upstream mpn371 by a 153-nucleotide long intergenic region (head-to-head orientation) and from downstream mpn373 by a 10-nucleotide short intergenic region (tail-to-tail orientation). Analysis of all three genes by reverse transcription PCR (RT-PCR) revealed three transcripts of expected polarity (Fig. 1B). Based on this gene organization, we predicted cards to have its own promoter. Primer extension (PE) analysis revealed a single transcriptional start point (TSP) at 10 nucleotides upstream of the cards translational start (Fig. 1C). Further examination of the sequence upstream of the TSP revealed additional consensus features of M. pneumoniae promoters, such as the presence of a -10 element (Pribnow box; TAAAAT; Fig. 1C) four nucleotides upstream from the identified TSP. The sequence immediately 5′ to the -10 element was AT-rich and contained polythymidine tracts (3 and 5 residues; Fig. 1C). While there was no strong consensus in the -35 region, the relatively conserved M. pneumoniae promoter-unique TTGA (Weiner et al., 2000) was found upstream of the -10 region. PE analyses did not identify individual transcriptional starts for mpn371 or mpn373. Although transcription of cards was readily confirmed by RT-PCR and low amounts of transcript were repeatedly demonstrated by slot blot analysis ( Fig. 2A), Northern blot analysis did not detect cards transcripts (data not shown) possibly because of low-level expression.

Is expression of cards temporal?
To determine the dynamics of cards transcript expression during in vitro mycoplasma growth in SP-4 broth, we performed slot blot analyses with total RNA from early-, mid-, late-exponential and stationary growth phases of M. pneumoniae cells. cards mRNA levels were compared with differential mRNA expression of selected M. pneumoniae metabolic genes (tuf, pdhB, pdhA, gap, eno and nox; Fig. 2A) and adherence-and heat shock-related genes (Fig. S1). Results indicated that cards expression occurred at low levels during early stages of growth, which decreased dramatically as growth progressed ( Fig. 2A). In contrast, other genes exhibited relatively stable or increased gene transcription throughout growth (eno, gap, nox, pdhA and pdhB) while tuf exhibited a similar pattern except at stationary phase where a noticeable decline in transcription occurred ( Fig. 2A). Adherence-and heat shock-related genes exhibited relatively stable or diminished transcription profiles during mid-to-late log growth phases but, in all cases, increased transcription was observed during later stages of growth in direct contrast to cards profiles (Fig. S1). To further confirm the specificity of cards expression during different mycoplasma growth intervals, we performed real-time quantitative reverse transcriptase polymerase chain reaction (qRT-PCR). As indicated in Fig. 2B, the expression of CARDS toxin transcription decreased dramatically (~130fold) between early-log and stationary phases.

Total M. pneumoniae cell protein profiles at different growth stages
In order to compare CARDS toxin protein levels with other mycoplasma proteins, we harvested mycoplasma cells during an extended (12-144 h) growth period in SP-4 broth and analysed protein profiles (Fig. 3A). Four distinct protein patterns were readily observed based upon SDS-polyacrylamide gel electrophoresis (PAGE) profiles and image analysis. One protein category exhibited increased intensity as mycoplasma growth proceeded from early to late stages (open arrow). Another category showed early intensity, which subsided or disappeared during mid-to-late log phase (closed small arrow). The third category exhibited increased intensity throughout logarithmic growth, which decreased during the stationary stage (large closed arrow). The last category represented proteins that remained mostly unchanged (no arrow). We selected specific proteins representing adherence-and stress-related functions (adhesin P1, adhesin P30, PDH-A, PDH-B, EF-Tu and ClpB) for comparison with CARDS toxin using immunoblotting (Fig. 3B). The intensities of immunoreactive bands demonstrated that CARDS toxin protein levels peaked between 24 and 48 h and declined to the lowest levels at 60 h, which continued over the remaining growth period (60-96 h; Fig. 3B). Other proteins demonstrated levels of detection consistent with their pattern of mRNA transcripts (Fig. 2). Growth stage-related dynamics of CARDS toxin mRNA and protein expression are depicted in Fig. 4 and are compared with the progression of culture growth (early-log through stationary phases). These quantitative data clearly reinforce the selective early synthesis of cards transcripts and CARDS toxin protein (Fig. 4). Note the relative stability of CARDS toxin following cards transcriptional decline (i.e. 60-144 h).

CARDS toxin location in mycoplasma membranes
To assess the location of CARDS toxin in M. pneumoniae cells, we performed immunoblot analyses on cellular fractions of M. pneumoniae clinical isolate S1. Whole cell lysates and cytoplasmic, membrane and culture supernatant preparations were quantified for the presence of CARDS toxin as described in Experimental procedures. In parallel, the localization of P1 (major surface-associated adhesin protein) and EF-G (intracellular cytosolic protein) was used as control. As shown in Fig. 5A, at 48 h of mycoplasma growth, CARDS toxin was detected in the density gradient-purified membrane fraction, which is free of intracellular protein contamination (note the absence of EF-G in lane 2). When compared with total CARDS toxin concentrations, 7-10% of CARDS toxin was associated with the mycoplasma membrane fraction. No full-length CARDS toxin was detected in culture supernatant (data not shown). Occasionally, we observed two minor CARDS toxin-related immunoreactive bands of low molecular weight in the supernatant of 72 and 96 h cultures, which may represent degraded or processed CARDS toxin (data not shown). Immunogold labelling and electron microscopic analysis revealed CARDS toxin distribution over the entire surface of M. pneumoniae cells, including A. Expression of cards along with other genes was analysed by DNA slot blot. M. pneumoniae gene-specific PCR products (Table S2) were blotted onto Zeta probe membranes. M. pneumoniae S1 cells were grown and harvested at early-log (24 h), mid-log (48 h), late-log (72 h) and stationary phases (120 h). 32 P-labelled cDNAs generated by reverse transcription of isolated total RNA were used as hybridization probes. Experiments were repeated three times. B. Expression of cards by relative real-time quantitative RT-PCR (qRT-PCR). M. pneumoniae S1 cells were grown and RNA isolated at specific growth phases. Real-time qRT-PCR was performed using SYBR green chemistry as detailed in Experimental procedures. Experiments were repeated two times. The average fold differences in expression levels of cards mRNA at various time points (when compared with the stationary phase) and standard deviations (SD) are presented. the tip organelle, a pattern similar to the major adhesin protein P1 (Fig. 5B).

Expression of CARDS toxin in M. pneumoniae during co-incubation with HeLa and differentiated NHBE cells
To evaluate the potential role of host cell factors in cards expression during infection, we incubated HeLa cells with M. pneumoniae and monitored the expression of cards after 15, 30 and 60 min. As appears in slot blot ( Fig. 6A), higher expression of cards mRNA was readily detectable when M. pneumoniae cells were co-incubated with HeLa cells in contrast to growth in broth culture alone. Real-time qRT-PCR data revealed that constant levels of cards transcription were maintained during the 60 min time frame in SP-4 broth whereas a fold increase in cards transcript levels was detected in M. pneumoniae cells after 1 h co-incubation with HeLa cells. To further reinforce the effect of co-incubation on CARDS toxin expression in M. pneumoniae cells, we used confocal immunofluorescence microscopy to detect CARDS toxin synthesis during mycoplasma infection of differentiated NHBE cells. Mycoplasmas were distributed over NHBE cell surfaces with concomitant CARDS toxin synthesis. Z-section analysis of infected NHBE cells revealed CARDS toxin both outside and inside NHBE cells (Fig. 7).

Expression of CARDS toxin during infection of mice
In M. pneumoniae-infected pneumonia patients, evidence for seroconversion against CARDS toxin is striking (Kannan and Baseman, 2006) indicating that CARDS toxin is expressed during human infection and is highly immunogenic. This contrasts with the low transient expression observed during M. pneumoniae growth in SP-4 broth (Figs 2 and 3). Since we observed increased cards transcript levels and readily measurable CARDS toxin production following co-incubation of mycoplasmas with host cells (Figs 6 and 7), we analysed the expression of CARDS toxin during in vivo infection. In intranasally inoculated mice, M. pneumoniae expressed CARDS toxin per mycoplasma genome in lung tissues at much higher levels (~8-40-fold at 24-48 h post infection, respectively) than during growth in SP-4 broth (Fig. 8).

Discussion
In this report we describe growth-regulated gene expression in M. pneumoniae, with special interest in the dynam- A. Equal amounts of S1 total cell proteins resolved on 4-12% Nu-PAGE gradient gel and stained with Coomassie Blue. Open arrows: increased protein intensity during early-to-late growth stages. Small closed arrows: decreased protein intensity during mid-to-late growth stages. Large closed arrows: increased protein intensity during stationary stage. B. Parallel gel transferred to nitrocellulose membrane for immunoblotting. Membrane strips were cut and treated with respective M. pneumoniae antibodies. Experiments were repeated two times.
ics of cards mRNA expression and CARDS toxin protein synthesis in vitro and in vivo. Growth conditions are invariably different in the test tube versus those encountered during the course of infection. Only recently have transcriptional studies been performed to examine M. pneumoniae responses to environmental cues (Weiner et al., 2003;Halbedel et al., 2004;Kannan et al., 2008). None have focused on monitoring bona fide virulence determinants, like CARDS toxin. Therefore, the information provided here represents an essential step in monitoring how M. pneumoniae may behave during the infectious process. With the availability of M. pneumoniae-purified recombinant proteins and specific antibody reagents, we were able to quantify and correlate transcriptional analyses with translational responses between CARDS toxin and a range of M. pneumoniae proteins.
Using M. pneumoniae clinical strain S1 (Kannan and Baseman, 2006;Musatovova et al., 2008) we demonstrated transcription of all three genes (cards and its adjacent genes: mpn371-mpn373) in their predicted orientations (Fig. 1). Previously, corresponding gene transcripts were analysed in M. pneumoniae strain M129 before and after temperature shift (Weiner et al., 2003). We also sequenced the corresponding regions in strain S1 and found very minimal differences in the sequences (see Fig. S2). As shown in Fig. 1C, the cards transcript contains a short 5′-untranslated region (transcription start point at T -10 ). Because cards is followed by a gene with opposite polarity (Fig. 1A), we expected cards to bẽ 1.8 kb single gene transcript. Recently, such monocistronic messages were shown for M. pneumoniae genes ldh and ptsH (Halbedel et al., 2004;. However, repeated Northern blots did not permit visualization of any clear band for cards. It is possible that the low amount of transcript we detected (Figs 2 and 4) is further degraded during isolation. Also possible and contrary to ldh and ptsH genes, a transcription termination signal is not predicted for cards (Ermolaeva et al., 2000), and transcripts of different lengths may be generated.
We investigated CARDS toxin synthesis by measuring changes in cards transcript levels and comparing these data with CARDS toxin protein concentrations. This is important because the relationship between mRNA levels and rates of protein synthesis can be non-linear (Vanbogelen et al., 1999). Moreover, protein analysis makes it possible to study critical post-translational control mechanisms, such as protein modification and stability, which may contribute greatly to the ultimate properties of a given protein. Although we observed a surprising demise of cards transcripts from early-to-mid log mycoplasma growth in SP-4 medium (Fig. 2B), we still detected CARDS toxin protein levels late into stationary phase (Figs 3 and 4). The detection of CARDS toxin many hours after the end of exponential growth can be explained either by toxin stability and/or by its continued low-level synthesis as a result of residual de novo transcription. In contrast to CARDS toxin, the toxins of Bortedella pertussis (Rambow-Larsen and Weiss, 2004) and Staphylococcus aureus (Payne, 2003) are maximally expressed at mid-to-late exponential growth phases (pertussis toxin) and late-to postexponential growth phases (a toxin). Similar high-level expression of toxins during late-log and stationary phase growth-related time points was observed in other bacteria (Thompson et al., 2003;Wagner et al., 2003;Snyder et al., 2004). Distinct from cards gene transcription, we observed high-level expression of adherence-and stress-related genes during late-log and stationary stage growth of M. pneumoniae, and similar growth dependent expression patterns of adherence-and stress-related genes have been observed in other bacteria (Strohmaier et al., 1995;Xu and Johnson, 1995;Nicholson et al., 2003;Thompson et al., 2003).
Earlier, we reported that CARDS toxin was a surfaceassociated, trypsin-sensitive surfactant protein A binding 68 kDa protein (Kannan et al., 2005). In the current study, we further reinforced its cell-associated distribution and lack of secretion into the medium during mycoplasma growth in SP-4 broth. Interestingly, only 7-10% of the toxin appears membrane-bound (Fig. 5A), and immunoelectron microscopy (Fig. 5B) further confirmed CARDS toxin surface localization. The majority of toxin is detected in the cytoplasm, which was confirmed by immunoblot comparisons between concentrated supernatant (ammonium sulphate precipitated and desalted), total mycoplasma lysate, membrane fractions and cytoplasmic fractions (Fig. 5A). Consistent with these results, we could not identify signal sequence specific for secreted proteins in CARDS toxin. Because higher expression of CARDS toxin is observed when mycoplasmas are co-incubated with host cells, it is possible that redistribution of toxin from the mycoplasma cytoplasm to mycoplasma mem- brane surface occurs, facilitating toxin-mediated events. For example, although most classical bacterial protein toxins are secreted into the surroundings, some are displayed on the bacterial surface or released upon contact with the host. The latter could be a possible scenario for CARDS toxin, as toxin colocalizes with mycoplasmas during colonization of host cells and can be detected inside NHBE cells based upon the merging of DAPI staining with anti-toxin immunostaining (Fig. 7). Therefore, mycoplasmas and toxin together appear to follow predictable patterns of other toxin-producing bacteria by evading host defenses, adhering to cells and tissue matrices, spreading within the host, eliciting inflammatory pathways and degrading host target cells and tissues for nutritional and survival benefits (Finlay and Falkow, 1997). Purified toxins of cholera, tetanus and diphtheria also serve critical functions in establishing infection and associated pathologies (Finlay and Falkow, 1997). This is consistent with the characteristic ciliostasis, cytoplasmic vacuolization, nuclear swelling, and extensive epithelial cell fragmentation and sloughing observed in hamster tracheal organ cultures infected with M. pneumoniae (Hu et al., 1976;Murphy et al., 1980). Importantly, this pathological pattern parallels observations when CARDS toxin alone is added to baboon tracheal organ cultures (Kannan and Baseman, 2006).
As with virulence factors in other pathogens, CARDS toxin probably relies on other mycoplasma components to fully potentiate its expression, delivery and impact on the host. Other possible microbial components include gene and protein regulatory systems, adherence factors for colonization and invasion and possible secretion systems A. Autoradiographic DNA slot blot analysis for cards and pdhA genes. M. pneumoniae-specific primers were used to generate 32 P-labelled cDNAs from total RNA isolated from HeLa cells alone, M. pneumoniae cells alone or M. pneumoniae cells co-incubated with HeLa cells for different time intervals. Experiments were repeated three times. B. Relative quantification of cards mRNA by real-time qRT-PCR. M. pneumoniae S1 cells were co-incubated with HeLa cells for 15, 30 and 60 min as described in Experimental procedures, and RNA was isolated. Real-time qRT-PCR was performed using SYBR green chemistry and pdhA as a normalizer. Experiments were repeated two times. The average fold differences in expression levels of cards mRNA at various time points (when compared with the cards mRNA level at time 0) and SD are presented.
for delivery of specific pathogenic products. Because CARDS toxin location is also associated with the tip organelle in addition to its membrane distribution (Fig. 5B), a likely scenario is that tip-mediated adherence brings M. pneumoniae into close contact with host target cells for delivery of toxin as observed with vibrio (Srivastava et al., 1980). In addition, CARDS toxin also can bind to mammalian cell surfaces through SP-A-like molecules and facilitate colonization, similar to B. pertussis pertussis toxin that appears to assist bacterial adher-

Fig. 7. Localization of CARDS toxin in M. pneumoniae-infected NHBE cells. Fully differentiated NHBE cells grown in air-liquid interface were
infected with M. pneumoniae. Thirty-eight hours post infection, cultures were washed to remove unbound mycoplasmas, submerged in fixative, processed for laser scanning confocal microscopy and screened using mouse polyclonal anti-CARDS toxin sera. Luminal (A1 and B1) and cross-sectional (A2 and B2) views of uninfected (A1 and A2) and infected (B1 and B2) NHBE cells are presented. In the latter case, surface colonization by mycoplasmas and synthesis and distribution of CARDS toxin during infection are readily observed (red). Cross-sectional views of infected cells reveal intracellular localization of CARDS toxin (arrows). Also, DAPI staining indicates the abundance of surface associated M. pneumoniae DNA (small blue dots on the respiratory cell surface; B1 and B2). Merged images of the cross-sectional view indicate colocalization of CARDS toxin with M. pneumoniae at respiratory cell surfaces and also demonstrate that toxin is detectable inside target cells, possibly free of associated M. pneumoniae. Experiments were repeated three times. ence to mammalian cells (Tuomanen and Weiss, 1985).
While CARDS toxin expression is transient during M. pneumoniae growth in SP-4 medium, it was appreciably increased during contact with human cells (Figs 6 and 7) suggesting the influence of host cell factors. This was further supported by the many-fold increase in CARDS toxin protein molecules per M. pneumoniae cell ratio in mice versus broth culture (Fig. 8). The expression of bacterial toxins is generally tightly regulated and controlled at the transcriptional level by virtue of repression and/or transactivation via specific regulatory proteins (Gallegos et al., 1997;Rood, 1998;Hirst, 1999;Payne, 2003;Stibitz, 2003). The transient and decreased expression of CARDS toxin in M. pneumoniae during in vitro growth, cards transcriptional upregulation during co-incubation with host cells, detection of CARDS toxin in infected NHBE cells, CARDS toxin increased synthesis per mycoplasma cell in mice and the high-titre seroconversion to CARDS toxin in infected patients (Kannan and Baseman, 2006) suggest that the production of CARDS toxin is also controlled in vivo by environmental host signals. Uncovering the physiological cues that influence in vivo mycoplasma growth and CARDS toxin expression will be critical in clarifying the pathogen-host interaction and in devising strategies, including new antimicrobial agents, targeted immunotherapies and effective vaccines, to control and prevent M. pneumoniae-mediated disease progression.

Bacterial strains, human tissue cultures, growth conditions and collection of samples
For growth curve studies, M. pneumoniae S1 cells  were grown in 400 ml of SP-4 medium (4 ¥ 100 ml in 150 cm 2 tissue flasks) and incubated at 37°C for 72 h. Surface-attached mycoplasmas were harvested, concentrated, pooled and divided into forty 150 cm 2 tissue culture flasks, which were incubated at 37°C for 12,24,36,48,60,72,84,96 and 120 h. Three flasks were collected at each time interval [one for RNA isolation and two for protein characterization, colony counts, optical density (OD600) determinations and protein measurements]. For OD600 reading, cells were washed in phosphate-buffered saline (PBS, pH 7.4) and scraped and passed through syringes with 25-gauge needles. Additional flasks were used at appropriate time intervals for membrane isolation and other analyses described below. Each experiment was repeated at least three times. All samples were stored at -80°C until use. For evidence of secreted CARDS toxin during M. pneumoniae growth in SP-4 broth, culture-grown supernatant samples were filter-sterilized using 0.2 mm filters to remove mycoplasmas and then concentrated using ammonium sulphate (30-80% saturation). Supernatants were dialysed with PBS and analysed for the presence of CARDS toxin by immunoblot. Recombinant Escherichia coli cells were grown in Luria-Bertani (LB) broth or LB agar containing ampicillin (100 mg ml -1 ) at 37°C.
Normal human bronchial epithelial cells were grown in air-liquid interface as previously described (Krunkosky et al., 2007). Briefly, NHBE cells (Cambrex, San Diego, CA) were seeded onto Transwell-clear culture inserts (24.5, 0.45 mm pore size; Costar, Cambridge, MA) thin-coated with rat tail collagen type I (Collaborative Res., Bedford, MA). The apical surfaces of NHBE cells were exposed to a humidified 95% air/5% CO2 environment and medium changed daily during 21-28 days of air-liquid culture. To study expression and localization of CARDS toxin during M. pneumoniae infection of NHBE cells, stocks of clinical strain S1 were quantified by dilution and plating. Prior to infection, M. pneumoniae inocula were passed 10¥ through 25-gauge needles and added apically to each culture well (50 mycoplasmas: 1 NHBE cell) for specific times. Total epithelial cells exposed to apical luminal surfaces in 1 cm 2 culture inserts were estimated to be approximately 450 000 by examination of confocal microscopic images from multiple fields of fully differentiated NHBE cells. Each experiment was repeated in triplicate.
To further examine experimental conditions that could influence the degree of expression of cards transcripts in M. pneumoniae cells, we incubated mycoplasma cell monolayers alone or with HeLa cells (ATCC), the latter had been resuspended in SP-4 broth supernatant (0.2 mm filtered) from 48 h depleted mycoplasma growth medium in order to eliminate the effect of fresh SP-4 components on cards expression. Ten millilitres of depleted mycoplasma growth medium suspended HeLa cells was overlaid on mycoplasma monolayers for 15, 30 and 60 min. Test samples were collected at specific intervals, RNA isolated each time and DNA blot slot blots and real-time qRT-PCR performed. Experiments were repeated at least two times for each experiment.

DNA manipulations, primer design and DNA slot blot construction
Mycoplasma pneumoniae chromosomal DNA was isolated using Easy-DNA kit (Invitrogen, Carlsbad, CA), quantified by OD260, and 100 ng were used in amplification reactions. Sequences of selected genes were obtained from NCBI databases, and primers for PE analysis, RT-PCR and amplification of genomic regions of selected genes were designed (Table S1).
To obtain sequencing templates in PE analysis, corresponding regions of the M. pneumoniae chromosome were amplified using specific primers and high-fidelity AccuTaq LA DNA polymerase (Sigma-Aldrich, St. Louis, MO). Generated amplicons were TA-cloned into pCR2.1 and transformed into TOP10 cells following manufacturer's instruction (Invitrogen). Recombinant plasmids were purified from E. coli cultures using the QIAprep spin kit (Qiagen, Valencia, CA).
For DNA slot blot construction, specific regions of selected M. pneumoniae genes were amplified using corresponding primers (Table S2) and high-fidelity AccuTaq LA DNA polymerase (Sigma-Aldrich). All generated amplicons were purified by gel extraction, and 200 ng of each specific PCR product was blotted in triplicate on Zeta probe membranes by using Bio-Dot SF microfiltration apparatus as suggested by the manufacturer (Bio-Rad Laboratories, Hercules, CA).

Transcriptional analysis
Total RNA from M. pneumoniae was isolated with RNase-free reagents and plastic ware. Surface-adherent mycoplasmas alone and M. pneumoniae cells co-incubated with HeLa cells were washed twice with 25 ml of ice-cold PBS. Sixteen millilitres of TriReagent (Sigma-Aldrich) was added to each T150 flask, and RNA was extracted as suggested by the manufacturer. Pellets were dissolved in RNase-free water and quantified at 260 nm. RNA was subjected to DNase I treatment (Gibco-BRL) prior to further use.
To determine TSP we performed PE with primers that were end-labelled with [g-32 P]-ATP using polynucleotide kinase (United States Biochemical, Cleveland, OH). Twenty-five micrograms of total mycoplasma RNA and 15-20 ng of radiolabelled primer were heated to 70°C for 10 min and cooled slowly. After annealing, 5 ml of 10¥ reverse transcription buffer (200 mM Tris-HCl, pH 8.4, 500 mM KCl), 5 ml of 25 mM MgCl2, 5 ml of 100 mM dithiothreitol, 2.5 ml of 5 mM dNTPs and 1 ml of Superscript II RT (Gibco-BRL) were added, and PE was performed at 42°C. PE reactions were terminated by incubation at 70°C for 15 min, E. coli RNase H (2 units per sample) added, and incubation continued at 37°C for 20 min. Individual PE products were phenol-chloroform extracted, precipitated with ethanol and dissolved in 6 ml of H2O and 3 ml of loading buffer. Five microlitres of each sample was analysed in 6% sequencing gels alongside sequencing reactions (dsDNA Cycle Sequencing System, Gibco-BRL) generated with the same end-labelled oligonucleotide and recombinant plasmid containing corresponding sequence.
Transcription of mpn371, cards and mpn373 genes was demonstrated by RT-PCR using isolated RNA. Five micrograms of total RNA was subjected to DNase I treatment (Gibco-BRL), and individual reactions were performed with corresponding antisense primers (Table S1) and SuperScript first-strand synthesis system as suggested by the manufacturer (Gibco-BRL). Generated complementary DNAs were used in PCR with appropriate primers (Table S1) by standard procedures. PCR products were analysed in 1% agarose. Chromosomal DNA was used with each set of primers to demonstrate amplification of a single specific region in each reaction.
Relative quantification of cards transcription was done by the comparative CT (DDCT) method in two-step real-time qRT-PCR using SYBR Green chemistry. Prior to RNA analysis, cards-and pdhA-specific primers (Table S1, MPN372P/M and MPN393P/M, respectively; the later served as a normalizer) were examined for specificity. Melting curves of amplicons confirmed generation of a single specific product in each reaction. Using primers with HeLa chromosomal DNA yielded no product. Validation of the assay was done using primer sets with serial dilutions of S1 chromosomal DNA. For analysis, 2 mg of total RNA was treated with DNase I (Invitrogen) and first-strand cDNA was synthesized using reverse cards-and pdhA-specific primers and SuperScript reverse transcriptase as suggested by manufacturer (Invitrogen). PCR reactions (50 ml) contained 25 ml of SYBR Green PCR Master Mix (Applied Biosystems), 2.0 ml of generated cDNAs (diluted 1:10) and corresponding primer set (1 mM each) and amplification was performed under default conditions using ABI PRISM 7900HT SDS (Applied Biosystems). All reactions were run in triplicate and plates included cDNA synthesis reactions without added reverse transcriptase (controls for DNA contamination). When analysing the mycoplasma response to co-incubation with HeLa cells, cDNA synthesis reactions performed on RNA isolated from uninfected HeLa cells were also included. Analysis was done using RQ Manager 1.2 (Applied Biosystems). cards transcript amounts are represented as a fold change with corresponding standard deviation.
Hybridization probes for slot blot analysis were also generated using the SuperScript first-strand synthesis system (Gibco-BRL). Twenty-five micrograms of RNA plus a mixture containing 0.2 mM of each gene-specific antisense oligonucleotide (Table S2) were heated to 70°C for 10 min and cooled to room temperature. Reverse transcription was performed with both [a-32 P]dCTP and [a-32 P]-TTP (NEN) replacing corresponding nucleotides in the reaction buffer. Reactions were heat inactivated at 70°C for 15 min, and E. coli RNase H (2 units per sample) was added prior to a 20 min incubation at 37°C. After unincorporated nucleotides were removed by gel filtration using G-25 Sephadex column chromatography (Roche), an equal volume of formamide was added. Denatured (boiling for 5 min) probes were added to hybridization buffer.
Individual PhosphorImager screens were analysed using PhosphorImager 400E (Molecular Dynamics) with a pixel size of 176 mm, and resulting image files were assessed by determining pixel density for each band using ImageQuant (version 5.0) (Molecular Dynamics). Numeric files were exported into a Microsoft Excel spreadsheet for analysis. To analyse transcript amounts for different experimental conditions, we expressed values as percentage of pdhA transcript amount after background subtraction.

Expression and purification of recombinant M. pneumoniae proteins and immunoblot analysis
Recombinant N-terminal His-tagged M. pneumoniae CARDS toxin, ClpB , DnaK, PDH-A, PDH-B, EF-Tu and carboxyl P1 truncation (unpublished data) were overexpressed in E. coli BL21(DE3) cells and purified using Ni-NTA (nickel-nitrilotriacetic acid) column chromatography (Kannan and Baseman, 2006). Verification of fusion proteins was performed by SDS-PAGE, Coomassie blue staining and immunoblotting with rabbit-or mouse-specific antisera. Concentrations of all proteins were determined by BCA protein assay (Pierce, Rockford, IL). For measurement of each M. pneumoniae protein, immunoblot analysis was employed using polyclonal rabbit or mouse antisera generated against M. pneumoniae CARDS toxin, ClpB, P1, P30, EF-Tu, PDH-B and Mycoplasma genitalium PDH-A and DnaK. To quantify protein expression by immunoblotting, a dilution series of each purified recombinant protein was loaded onto individual gels, probed with corresponding antibodies and signal strengths determined by densitometry. Quantification of band intensities was performed by scanning immunostained bands and analysing images with KODAK Image software. Using Microsoft Excel linear regression, we estimated protein concentrations in test samples, which were compared with standard curves. As necessary, samples were diluted so that the amount of loaded protein fell within the linear range of standards. Relative levels of protein expression were determined by comparing test unknowns to the 24 h sample as an internal standard. Transcript or protein amounts are represented in means with corresponding standard deviation.

Separation of mycoplasma membrane, cytoplasmic and secreted proteins
In order to localize CARDS toxin in specific mycoplasma cell fractions, M. pneumoniae S1 cells were grown in SP-4 broth as described above. Cell cultures were harvested at 24, 48, 72 and 96 h by scraping and centrifugation, washed two times in 10 ml of 4°C Tris-NaCl [20 mM Tris-HCl (pH 7.4), 200 mM NaCl, 1 mM EDTA] and suspended in 1 ml of Tris-NaCl containing 20 ml of protease inhibitor cocktail (Sigma-Aldrich). Mycoplasma cells were disrupted by osmotic shock and sonication as detailed by us earlier (Dallo et al., 2002). Then, samples were centrifuged at 100 000 g for 1 h to separate membrane and cytosolic cell fractions. Cytosolic fractions were precipitated with trichloroacetic acid at 20% saturation for 30 min on ice, suspended in 1 ml of Tris-NaCl and divided into 100 ml of aliquots or concentrated using Amicon centrifugal filter (Millipore; 3 kDa cutoff). Membrane fractions were further purified by sucrose gradient centrifugation to separate from unbroken cells. Then, enriched membrane fractions were suspended in 1 ml of Tris-NaCl, sonicated for 1 min, divided into 100 ml of aliquots, and membrane yield calculated relative to total cell lysate (Razin, 1983). To analyse secreted proteins, broth supernatants (500 ml) from mycoplasma cultures at each time point (24, 48, 72 and 96 h) were precipitated by ammonium sulphate in a three stepwise concentration (30%, 50% and 85%) (Englard and Seifter, 1990). Pellets from each ammonium sulphate fraction were resuspended in PBS containing protease inhibitors, dialysed against the same buffer, concentrated by Amicon centrifugal filter (Millipore; 3 kDa cutoff) and divided into 100 ml of aliquots. Membrane and supernatant fractions were stored at -80°C, and equal amounts of proteins were examined by immunoblotting for the presence of CARDS toxin (anti-rabbit antisera 1:2000), cytoplasmic EF-G (anti-rabbit sera 1:1000, gift from Dr Richard Herrmann) and membrane adhesin P1 (anti-mouse monoclonal antibody 1:2000). Based on immunoblot band intensities in each fraction and by membrane yield (constitutes~9-10% of the total protein), the percentage of membrane-associated CARDS toxin was determined.
Immunogold electron microscopy of CARDS toxin location in M. pneumoniae Immunogold labelling of M. pneumoniae cells was performed as reported previously (Dallo et al., 2002;Balasubramanian et al., 2008). Forty-eight hour cultures of SP-4-grown M. pneumoniae cells were washed with 100 mM Tris-HCl buffer (pH 7.5) containing 1% BSA and 1% heat-inactivated goat serum and incubated individually with anti-CARDS toxin, anti-P1 or anti-EF-G rabbit antisera at 1:100 dilution for 120 min at room temperature. After incubation for 60 min at room temperature with goat anti-rabbit IgG-gold (20 nm particles) complex diluted 1:40 in PBS, pH 7.4, with 1% BSA, M. pneumoniae cells were mounted on Formvar-coated nickel grids and fixed with 1% glutaraldehyde/4% formaldehyde for 20 min at room temperature. Individual grids were examined with a JEOL 1230 transmission electron microscope at 80 kV accelerating voltage after staining with 7% uranyl acetate followed by Reynolds lead citrate.

Immunostaining and confocal microscopy to detect CARDS toxin in NHBE cells
For immunostaining of infected and control differentiated NHBE cells, airway cultures were gently washed with PBS and fixed with 4% paraformaldehyde (v/v) diluted in PBS containing 0.05% Tween (PBST). Fixed cells were permeabilized for 15 min with PBST containing 0.1% Triton X-100 (v/v), washed in PBST and incubated for 1 h at room temperature in blocking solution consisting of 10% normal goat serum diluted in PBST (v/v). Cells were then probed overnight at 4°C with mouse anti-CARDS toxin serum diluted 1:400 in blocking solution. The following day cells were washed with PBST and incubated for 1.5 h with Cy-5conjugated anti-mouse IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) diluted 1:75 in blocking solution. Cells were next washed with PBST followed by nuclei and actin counterstaining for 20 min using 1 mg ml -1 DAPI and 2 U ml -1 of Alexa Fluor 488-labelled phalloidin (Molecular Probes, Eugene, OR) respectively. After washing cells with PBST, membranes were excised from Transwell supports and mounted on slides containing Citifluor antifade (Ted Pella, Redding, CA). Optical sections of DAPI, Alexa Fluor 488 and Cy-5 fluorescence were imaged and reconstructed using a Leica TCS SP2 spectral confocal microscope (Leica Microsystems, Exton, PA).

Animals and inoculation
Two-month-old mycoplasma-and murine virus-free female BALB/c mice (Charles River and Harlan) were housed in filter-top cages and allowed to acclimate to their new environment for 1 week. As detailed earlier (Hardy et al., 2001), mice were intranasally inoculated once (day 0) with 2.5-7 ¥ 10 7 genomes of M. pneumoniae in 50 ml of SP-4 broth. Control mice were inoculated with sterile SP-4 broth. Animal guidelines were followed in accordance with the Institutional Animal Care and Research Advisory Committee at The University of Texas Southwestern Medical Center at Dallas.