1 Institut de Pharmacologie Moléculaire et Cellulaire UMR 6097 Centre National de la Recherche Scientifique, Université de Nice-Sophia Antipolis, Valbonne
2 Institut National de la Santé et de la Recherche Médicale (INSERM) UMR-S 514, IFR 53, CHU Hôpital Maison Blanche, Reims
3 INSERM U576, Hôpital de lArchet, Nice
4 INSERM EMI 10014, Vandoeuvre-les-Nancy, France
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
transcriptome; microarray; inflammation; infection
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
We examined the host transcriptional response to the interaction between human airway epithelial cells, represented here by the human airway glandular cell line MM-39 (13, 34), and a reference S. aureus strain (8325-4 strain). To provide a molecular portrait of airways cells response to infection, a time course experiment corresponding to an interaction between airway epithelial cell and bacterial supernatant was first performed. It was followed by a comparison of the epithelial cells responses to live bacteria or to products secreted by bacteria after 3 h of contact. The aims of the present study were to analyze whether the cells in direct contact with live bacteria or only in presence with S. aureus soluble virulence factors induce different transcriptional responses in human airway cells.
![]() |
EXPERIMENTAL PROCEDURES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Bacterial Strain and Culture Conditions
S. aureus strain (8325-4), a reference laboratory strain was a gift from T. J. Foster (Department of Microbiology, Dublin, Ireland). Bacteria were grown at 37°C overnight in trypticase soy broth (TSB). Bacteria/cell interactions were analyzed under two distinct experimental conditions. In the first experimental condition, S. aureus bacteria were grown in TSB up to a concentration of 5 x 108 cfu/ml, then washed with phosphate-buffered saline (PBS 0.1 M, pH 7.2), resuspended in DMEM-F12, and added to the apical surface of confluent airway epithelial cells (MM-39). Interaction took place for 3 h at a ratio of 50 bacteria for 1 epithelial cell as previously described (5, 22). This experiment was named "live bacteria." In the second experimental condition, the initial suspension (5 x 108 cfu/ml) was centrifuged (1,500 g, 10 min) and the supernatant was diluted 10-fold in DMEM-F12 (9, 12). The airway epithelial cells were then incubated with the bacterial supernatant (soluble virulence factors) for 1, 3, 6, 9, and 24 h. Additional experiments were performed at 3 h, to compare with live bacteria. This experiment was named "bacterial supernatant." Each of these two experimental conditions included their own control corresponding to either airway cell culture medium for the live bacteria experiment or epithelial cell culture medium supplemented with 1/10 of bacterial growth medium for the bacterial supernatant.
RNA Extraction
At the end of the different incubations, cells were washed with PBS (0.1 M), and RNA was extracted with High Pure RNA Isolation Kit (Roche), according to the manufacturers instructions.
Microarray Technique
Array preparation.
Both oligonucleotide microarrays and cDNA microarrays were used for this study, as follows.
oligonucleotide microarrays.
The oligonucleotide microarrays contained 25,000 distinct oligonucleotide probes covering most of the known human transcripts. The list of the 25,279 probes (length
51 bp) is available online (http://www.microarray.fr:8080/mediante; also see Supplemental Table S1, available at the Physiological Genomics web site).1
Microarrays were printed with a ChipWriter Pro (Bio-Rad) on commercial UltraGAPS II slides (amino-silane-coated slides, Corning 40017) and processed according to the manufacturers instructions.
cdna microarrays.
The cDNA microarrays contained 4,200 distinct cDNA probes. Gene selection was based on relevance to inflammation, infection, differentiation, ion transport, cell signaling, cell migration, etc. A large fraction of the probes also corresponded to transcripts encoding membrane proteins. Such a microarray probed a significant portion of all known human transcripts and represented a powerful tool for studying molecular mechanisms of lung physiopathology. The list of the 4,200 probes is available at http://www.microarray.fr/IPMC/cDNA_microarray5k.html. The cDNA probes were PCR-amplified from cDNA derived from Universal Human Reference RNA (Stratagene) by reverse transcription. Probes 1) had a normalized length of 250 ± 19 bp; 2) had a normalized GC content of 52 ± 8%; 3) were specific for a unique human gene; and 4) were controlled by DNA sequencing. PCR products were purified by using QIAquick 96 PCR Purification Kit (Qiagen), resuspended in 3x SSC at a concentration of 200 ng/µl. Microarrays were printed with a SDDC-2 (Bio-Rad) on homemade aldehyde-coated glass microscope slides (11). Valid amplifications were characterized by the presence of a fragment at the correct molecular weight and by the identification of the correct sequence by direct sequencing. Data presented into the present manuscript only refer to sequence-verified probes. Microarray processing was then performed as described in Dayem et al. (11).
RNA labeling and hybridization.
Two different RNA labelings were performed for this study: amplification protocol for oligonucleotide microarrays and direct RT labeling for cDNA microarrays.
amplification protocol (oligo array) We amplified 2 µg total RNA with the Amino Allyl MessageAmp aRNA kit (Ambion ref. 1752) according to the manufacturers instructions. Cy3- and Cy5-labeled aRNA were fragmented with the Ambion aRNA Fragmentation Reagents (ref. 8740), then added to 250 µl ChipHybe 80 (Ventana ref. 760-127) for hybridization in a Discovery station (Ventana, Illkirch, France). After a 10-min denaturation at 70°C, hybridization was performed 10 h at 48°C. Arrays were then washed twice in RiboWash solution (Ventana ref. 760-105) for 1 min, twice in 1x SSC solution (Invitrogen, 15557-044) for 1 min, rapidly dipped in water and ethanol, then spun dry.
rt protocol (cdna array). We reverse transcribed 10 µg of total RNA according to Dayem et al. (11), except that 500 µM dATP, 500 µM dCTP, 500 µM dGTP, 100 µM dTTP, and 400 µM amino-allyl-dUTP were used for the reaction. After 2.5 h of incubation at 42°C, RNA degradation, and removal of unincorporated nucleotides using the Nucleotide Removal Kit (Qiagen), cDNA were then labeled with Cy3/5 monofunctional reactive dyes (Amersham). Initial experiments were performed using a direct labeling method and provided similar results. Microarrays were then hybridized at 48°C, according to Dayem et al. (11).
Both oligonucleotide and cDNA microarrays were scanned with a ScanArray Express (version 2.0.19 microarray acquisition system; Packard BioScience, Rungis, France). The two red and green lasers operated at 633 nm and 543 nm to excite Cy5 and Cy3, respectively. The intensity was measured at 670 nm for Cy5 and 570 nm for Cy3. Laser power was set to 100%, and photomultiplier tube (PMT) power was set between 65% and 75% depending on the slides.
Bioinformatics Analysis
All results are available in GEO (http://www.ncbi.nlm.nih.gov/geo/), under the accession numbers GSE1853 (cDNA microarrays) and GSE1704 (oligonucleotide microarrays). For oligonucleotide microarrays, TIF images containing the data from each fluorescence channel were quantified with the GenePix Pro 5.0 program (Axon Instruments) using an "irregular features" quantification method. Lowess normalization was performed using the GeneSpring program (version 6.1). Genes characterized by a 75th percentile intensity of less than 1,000 for all measurements were discarded. For cDNA microarrays, TIF images containing the data from each fluorescence channel were quantified with the QuantArray program (version 3.0.0.0) (PerkinElmer, Rungis, France) using a fixed circle quantification method. For each spot, intensity and background values for Cy3 and Cy5 were obtained as average intensities. Negative controls ("neg") were spotted on each slide. These corresponded to nonmammalian mRNA sequences with no significant identity with any human sequences. Signals derived from these spots were subtracted from the specific signal so that a null gene expression was associated with a null fluorescence signal. The specific signal for spot i under condition j was thus defined as the total spot intensity (totij) minus local background (bgdij) minus the median of negj. Because incorporations of Cy3 and Cy5 dyes into cDNA can differ significantly, data was further normalized using a dye-swap method (8, 25, 55). This method required duplication of experiments, but improved the reproducibility of the quantification (Moreilhon and Barbry, unpublished data). For that purpose, a first microarray was hybridized with experimental and control samples labeled with Cy3 and Cy5, respectively. A second "swap" microarray was hybridized with experimental and control samples labeled with inverted dyes (experimental = Cy5; control = Cy3). The specific signal associated with spot i in sample j was defined by the geometrical average of the specific intensities in direct and swapped experiments:
![]() |
![]() |
Lists of genes significantly down- and upregulated under the different experimental conditions were established using two distinct approaches.
In a first statistical approach, we used Significance Analysis of Microarrays (SAM, http://www-stat.stanford.edu/~tibs/SAM/index.html), a software developed by Tusher et al. (49) to compute a statistical value for each probe, which represented the strength of the relationship between gene expression and one of our qualitative response variables (bacteria, factor, or controls in our experiments). In the present study, 12 independent measurements for each condition [four log2 (ratio) measurements for each experiment, times three independent biological experiment] were tested for statistical significance. Only probes with less than five missing values were kept for analysis, since SAM performance was affected by an excess of missing values. A cutoff for significance, called the delta value, was chosen in order to minimize the false-positive rate, defined by the ratio between the number of called genes and the 90th percentile of the number of falsely called genes. An identical cutoff value was used for bacteria and for soluble factors. SAM was first run using a one-class function on normalized log2 (ratio) for each experimental conditions, to identify genes upregulated or downregulated by bacteria or by soluble factors. A two-class function was also selected to highlight genes differentially expressed between bacterial supernatant and live bacteria condition. Additional analyses were performed to compare log2 (ratio) between bacteria and soluble factors.
In a second statistical approach, we took into account the median signal intensity associated with each probe. With this second independent approach, signals derived from one probe were compared with a subgroup of spots having similar fluorescence intensity. Typically, for one experimental condition, a local Students t-test was run between the 12 spot ratio values associated with 1 probe and the 240 ratios associated with the 20 nearest neighbors (according to the median spot intensity). The analysis was performed using standard Microsoft Excel functions.
Ontologies attached to each gene were then used to classify altered genes according to main biological themes. For that, we used Expression Analysis Systematic Explorer (EASE) program, available at http://david.niaid.nih.gov/david/ease.htm (20), and MEDIANTE, a local database containing diverse information about our probe sets (http://www.microarray.fr/mediante/index?language=en). Additional statistical analyses, including K-means, principal component analysis (PCA), and unsupervised hierarchical classification, were performed using the XLSTAT program running on Microsoft Excel (Addinsoft, Paris, France) and MeV program (version 3) (44).
Real-Time Semiquantitative RT-PCR
cDNA was synthesized using the SuperScript system (Invitrogen) as previously described (51). Briefly, the reaction was carried out in 25 µl with 5 µg RNA, 100 ng oligo(dT), 100 ng random hexamers (Roche), and 200 U SuperScript II reverse transcriptase during 1 h at 45°C, followed by a 5-min incubation step at 95°C. The cDNA was then adjusted at 50 ng/µl. The real-time semiquantitative PCR (sqRT-PCR) was performed using the SYBR Green technology and a Applied Biosytems model 9600 apparatus. Primers (MWG Biotech, Courtaboeuf, France) were designed to span exon-intron junctions to prevent amplification of possible trace of genomic DNA and to result in amplicons between 100 and 150 bp. All pairs of primers were qualified by restriction enzyme digestion and electrophoresis. A 20-µl PCR reaction contained 50 mM Tris·HCl (pH 8.4, Invitrogen), 200 nM of each primer, 50 ng cDNA, 0.2 mM dNTP (Amersham), 2.5 mM MgCl2, 0.4 U Taq polymerase (Platinum, Invitrogen), and 1/10,000 dilution of SYBR Green (Molecular Probes) in optical PCR plates and caps (Applied Biosystems). Expression of target genes was measured after normalization of RNA with four different housekeeping genes, and values were expressed using the CT method, as fold increased expression above a theoretical negative control (user bulletin no. 2, Applied Biosystems, December 1997).
ELISA
The enzyme-linked immunoabsorbent assays (ELISA) for IL-6 (R&D systems, ELISA Quantikine ref. D6050), IL-8 (R&D Systems, ELISA quantikine ref. D8050), PGE2 (R&D Systems, high-sensitivity ELISA ref. DE2100), and TNF (R&D, Quantkine HS Human TNF Immunoassay systems ref. HSTA00C) were performed according to the manufacturers instructions. Data were expressed as picograms per 5 x 106 cells for IL-6 and IL-8 experiments and as picograms per milliliter for PGE2 and TNF experiments (sensitivity:
0.7 pg/ml for Il-6,
10 pg/ml for IL-8,
8.2 pg/ml for PGE2, and
0.12 pg/ml for TNF).
Western Blot Analysis
MM-39 cells were washed twice in ice-cold PBS and lysed in lysis buffer [20 mM Tris-Cl (pH 7.5), 100 mM NaCl, 5 mM MgCl2, 0.5% Nonidet P-40, 1 mM sodium orthovanadate, 5 mM sodium fluoride] supplemented with protease inhibitors (Roche Applied Science). After 20 min at 4°C under continuous agitation, extracts were centrifuged at 12,000 g for 10 min at 4°C. We electrophoresed 50 µg of protein extracts on 9% SDS-PAGE, and these were electrotransferred onto a polyvinylidene fluoride membrane (Immobilon-P, Millipore). The membrane was incubated overnight at 4°C with a anti-HIF-1 monoclonal antibody at a 1,000th dilution (Novus Biologicals). Horseradish peroxidase-conjugated rabbit anti-mouse IgG antibody (1:5,000) was then applied for 30 min at room temperature. Immunoreactive bands were revealed by enhanced chemiluminescence (ECL, Amersham).
Quantification of the Cell-Bacteria Interaction
A fluorescence staining method using the LIVE/DEAD Bacterial Viability Kit (BacLight; Molecular Probes, Eugene, OR) was used to study the interaction between airway epithelial cells and S. aureus. BacLight is composed of two nucleic acid binding stains: SYTO 9, which stains cells (bacteria or host cells) with both intact and damaged membranes; and propidium iodide, which only stains damaged cells. The percentage of airway epithelial cells with attached and/or internalized bacteria after the 3-h incubation period was determined from the fluorescence images. Experiments were performed in triplicate.
Evaluation of Airway Epithelial Cell Death by Apoptosis or Necrosis
Three days after confluence, airway epithelial cells were incubated either with S. aureus at 5 x 108 CFU/ml or with bacterial supernatant, together with YO-PRO-1 (10 µg/ml, which stains apoptotic cells) and propidium iodide (1 µl/ml, which stains dead cells) probes. Fluorescent images were recorded every 30 min for 24 h as previously described (10). Variations of the YO-PRO-1 and of the propidium iodide fluorescence intensities were expressed as the ratio of the fluorescence intensity at a given time to the initial fluorescence intensity.
Experimental Design
Figure 1 describes the experimental design of the work presented in the manuscript.
|
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
To monitor the effects of live bacteria and bacterial supernatant on airway epithelial cells at 3 h, total RNA was isolated from the four experimental conditions and reverse transcribed into fluorescently labeled cDNAs. The entire microarray procedure (from infection to hybridization) was done independently three times for each experimental condition. Samples were hybridized to human cDNA microarrays comprising 4,200 distinct cDNA probes, then analyzed according to standard protocols (see EXPERIMENTAL PROCEDURES for microarray procedures) (Fig. 1).
Microarray Measurements
Significant changes in the expression profile were first identified with SAM. Table 1 shows that 36 SAM-positive genes were upregulated after a 3-h direct contact with S. aureus, and 7 SAM-positive genes were downregulated (live bacteria condition). Similarly, 72 SAM-positive genes were upregulated after contact with bacterial supernatant, and 42 SAM-positive genes were downregulated under the same experimental condition (bacterial supernatant condition). Thirteen genes were increased in both experimental conditions, whereas only one gene was found significantly decreased in both experimental conditions. A very similar qualitative picture was provided by two other independent analyses, based either on the use of a local Students t-test or on sqRT-PCR (sqRT-PCR results are described in the next paragraph). Students t-test revealed 58 altered genes in the live bacteria condition (43 repressed, 15 activated), whereas 359 genes were altered in the bacterial supernatant condition (170 repressed, 189 activated). Table 1 summarizes the results obtained with microarrays and sqRT-PCR. Whatever the quantification procedure, there are more altered genes in the bacterial supernatant condition than in the live bacteria condition. The poor overlap existing between the two conditions (only 10% of the genes were altered in both conditions) illustrated the existence of differential and specific responses of airway epithelial cells to a direct contact with live bacteria or to incubation with bacterial supernatant.
|
|
A differential expression superior to 5 was observed for 18 genes in the live bacteria condition (6 being downregulated and 12 being upregulated). On the other hand, 42 genes were regulated in the bacterial supernatant condition when using a 5-fold factor (16 being downregulated and 26 being upregulated). Among these genes, 7 genes were commonly upregulated in live bacteria and bacterial supernatant conditions with a ratio superior to 5. Only one gene was downregulated in the two experimental conditions, with a ratio superior to 5 (Table 1). These results confirmed that the transcriptional response to bacterial supernatant was prominent over the response to a 3-h direct contact with bacteria.
EASE analysis
Time course analysis using ratios cutoff was performed. Hierarchical clustering was performed on modulated genes encoding proteins implicated in transcription, inflammation, and apoptosis using the MeV program (Fig. 2). Analysis of the 3-h interaction results showed that SAM-positive genes (see EXPERIMENTAL PROCEDURES) can be clustered in one of the following functional classes: 1) genes coding proteins involved in transcription (i.e., fos, jun, junB, EGR-1, etc.), 2) genes coding proteins related to inflammation (i.e., IL-8, IL-1 and IL-1ß, COX-2, etc.), and 3) genes coding proteins involved in regulation of cell apoptosis (i.e., Sgk, A20, etc.). These functional classes were also found after a direct analysis of the Gene Ontology terms associated with 405 genes selected with the Students t-test (not shown).
1) Regulators of transcription.
Analysis after a 3-h interaction revealed that several genes encoding proteins of the activator protein-1 (AP-1) complex were upregulated in the bacterial supernatant condition: fos, jun, junB. This was further confirmed by sqRT-PCR. In the live bacteria condition, fos, jun, and junB were also found upregulated and SAM positive, although the upregulation was less than in bacterial supernatant condition (Tables 2 and 3).
|
|
Although our experimental plan could not provide directly information about the initial steps leading to all these modifications, NF-B activation appears as one of the key players of the orchestrated gene regulation. It is already known that NF-
B can be activated at a protein level in many models of bacterial infections (15, 50). Transcriptional upregulation of NF-
BIA indeed indicates such activation. Moreover, Fig. 2 shows that NF-
B1 and NF-
BIE were significantly upregulated at a transcriptional level after 3 and 6 h of contact with bacterial supernatant. Rel was upregulated only at 3 h, whereas Rel A and Rel B were not regulated (Fig. 2). The activation of NF-
B target genes (39), especially by the bacterial supernatant condition, is also indicative of an early activation of the NF-
B protein complex. This was typically the case at 3 h for IL-8 and IL-6, Myc, NF-
B IA, junB, TNFAIP3, and NR4A2, which are known to be regulated by NF-
B. NR4A2 was also induced by live bacteria (Table 2). Incidentally, time course experiment revealed that TLR2, a potential activator of the NF-
B pathway, was upregulated when cells were in contact with bacterial supernatant (Fig. 2).
2) Inflammation.
jak/stat pathway.
Many genes encoding proteins known to induce the JAK/STAT pathway were upregulated during the time course interaction with bacterial supernatant: LIF, PDGFB, PDGFC, as well as interferon receptor proteins, IFNAR1, IFNGR1, and IFNGR2. Moreover, JAK1, Lyn, STAT1, STAT3, PIAS1, and SOCS-2, involved in the JAK/STAT pathway, were also upregulated (Fig. 2).
chemokines and interleukins.
The chemokines CXCL1, CXCL2, CXCL3 (also named GRO1, GRO2, and GRO3), and CCL20, as well as interleukins IL-8, IL-1 and IL-1ß, IL-20, and IL-24, and also IL-1RAP, were upregulated after interaction with bacterial supernatant. However, IL-20RA was downregulated, mostly after 6-h interaction with bacterial supernatant (Fig. 2). Analysis after a 3-h interaction showed that IL-1
, IL-1ß, IL-8, IL-6, and the leukemia inhibitory factor (LIF) were upregulated in the bacterial supernatant condition expressions but were marginally altered after a direct contact with live S. aureus. YARS, which can behave as an interleukin-8-like cytokine, was upregulated in both conditions (Tables 2 and 3).
The response of epithelial cells to S. aureus infection was also quantified at the protein level by an ELISA test for IL-6 and IL-8. Figure 4, A and B, showed an increase in IL-8 and IL-6 in the cell culture supernatant after a 3-h incubation with bacterial supernatant. Smaller variations were noticed after a direct contact with the bacteria.
|
ALOX5 and LTA4H, two genes related to leukotrienes synthesis were downregulated by bacterial supernatant in a similar time-dependent manner (Fig. 2).
3) Apoptosis.
Time course experiment on bacterial supernatant condition showed that genes encoding proteins that inhibit apoptosis such as BIRC3, SGK, PIM, and BCL2A1 were upregulated, mainly after 3 h. SCYE1, described at sites of apoptosis, and pro-apoptotic genes encoding CASP1 and DAPK1 were downregulated rapidly after interaction with bacterial supernatant. This major anti-apoptotic response was, however, concomitant with a pro-apoptotic response as interleukins IL-1, IL-1ß, IL-24, STK17B, INHBA, and TNFRSF10B were upregulated rapidly after interaction with bacterial supernatant (Fig. 2). Analysis on cDNA microarrays after a 3-h interaction rather confirmed an anti-apoptotic transcriptional response of airway cells after contact with live bacteria or bacterial supernatant. Indeed, several genes that encoded products with anti-apoptotic properties such as PIM and TNFAIP3 were upregulated in both conditions after 3-h interaction. At 3 h, most airway cells incubated with live bacteria or bacterial supernatant exhibited a necrotic phenotype. Apoptosis was only transiently observed 810 h after interaction with bacterial supernatant and live bacteria, respectively (Fig. 5).
|
4) Genes regulated by hypoxia or by oxidative stress.
Some genes mainly regulated in the live bacteria condition (bhlhb2, MIG-6, TFRC, and EGR1) have been reported in the literature to be induced by hypoxia, through a HIF-1-dependent cascade. HIF-1
was not regulated at the transcriptional level (neither after 3 h of interaction with live bacteria nor after interaction with bacterial supernatant at 1, 3, 6, 9, or 24 h), suggesting a stabilization of HIF-1
in presence of hypoxia and oxidative stress (23, 24). In line with this hypothesis, expression of the HIF-1
protein was found increased in MM-39 cells after 3-h interaction with live bacteria and with bacterial supernatant (Fig. 6).
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A first experiment analyzed the interaction of bacterial supernatant with epithelial cells during a time course and was performed using a pangenomic microarray (Fig. 2). This first analysis clearly showed an early alteration (i.e., around 3 h) of many genes encoding proteins implicated in transcription, inflammation, and apoptosis.
To compare the response of epithelial cells with bacterial supernatant and the response of cells with live bacteria, another more specific analysis was performed after 3 h of interaction with either live bacteria and bacterial supernatant. Three reasons made us choose this time point. We first observed that transcriptional response to bacterial supernatant was prominent in the early response to interaction (1, 3, and 6 h) compared with late response (9 and 24 h) (Fig. 2). In contrast with the experiments of Boldrick et al. (6), S. aureus bacteria were not heat killed, leading to a potential effect of in vitro cytotoxicity and differential bacterial growth rate. We indeed observed that prolonged interaction between live bacteria and airway epithelial cells led to irreversible airway cell damage. Moreover, PBS-washed bacteria were in log-phase growth after a 3-h incubation with MM-39 cells (not shown). During log-phase growth, bacteria synthesize predominantly cell surface virulence factors such as protein A. On the other hand, genes encoding secreted virulence factors, such as -toxin, are transcribed at low levels during log-phase growth and are upregulated during the stationary phase of growth (31a, 38).
We anticipated that a 3-h incubation with washed live bacteria would differ significantly from a 3-h incubation with bacterial supernatant, which was largely enriched in soluble virulence factors. We therefore designed our study to discriminate the responses of host cells after a 3-h contact with live bacteria, or with soluble secreted virulence factors without live bacteria. To analyze the different responses between live bacteria and bacterial supernatant, more specific microarray consisting of 4,200 sequence-verified probes, representing
1/6th of the protein coding capacity of the human genome was selected. Although the fraction covered by this microarray may appear small, available probes were associated with 2,487 terms of the Gene Ontology database, representing 49% of all annotations for all known human transcripts. Moreover, this microarray was enriched with genes coding proteins implicated in inflammation, infection, differentiation, ion transport, cell signaling, and cell migration.
sqRT-PCR was also performed on a set on 537 pairs of validated primers. Larger stimulation factors were found with sqRT-PCR than with microarray, in general agreement with previous reports by Gavin et al. (18).
The distinct profiles observed between the live bacteria and bacterial supernatant conditions (only 10% of all SAM-positive genes were common to both) was indicative of different responses of airway epithelial cells in contact with either soluble S. aureus factors present in the supernatant of stationary phase bacteria or in contact with log-phase growing bacteria.
To identify relevant metabolic pathways participating to the host response at 3 h, several gene lists were established. Two independent statistical approaches were used. The first one used the classic SAM method, developed by Tusher et al. (49). Since this method hardly takes into account the intensity of the signals, we analyzed our results with a second approach, where probes were sorted according to their intensity level. Probes were selected when their ratio measurements differed significantly from the ratio measurements of the 20 nearest probes (as measured by the intensity), using a t-test. Although this test probably captured a few false-positive probes, the corresponding gene list included a majority of genes already identified by the SAM method (64%). The gene list obtained with this method defined a clear-cut distinction between live bacteria and bacterial supernatant (Fig. 3). Overall, the magnitude of alteration of these genes was stronger in the bacterial supernatant than in the live bacteria condition, as well for the number of altered genes as for the amplitude of the stimulation factors.
A second important observation was the upregulation of proinflammatory molecules such as cytokines and chemokines, as previously reported by Boldrick et al. (6).
Analysis of a time course experiment showed that genes associated with the JAK/STAT pathway were upregulated shortly after incubation with bacterial supernatant (Fig. 2). The JAK kinase activity is critical for normal transmission of cytokine and growth factor signals in host cells (42). The chemokines CXCL2, CXCL3, and CCL20 were upregulated, in agreement with a previous report by Kobayashi et al. (26) about the contact of S. aureus with human polymorphoneutrophils. The release of cytokines IL-1, IL-1ß, and IL-6, as well as the chemokine IL-8, was significantly upregulated in the bacterial supernatant condition but not or only slightly in the live bacteria condition. IL-6 and IL-8 interleukins have been extensively identified in response to bacterial infection and during the ensuing inflammation (4, 32, 45, 56). Most of the regulated genes were increased both at mRNA and protein levels and can be under a tight control by the NF-
B complex. Indeed, NF-
B-responsive elements are present on IL-1, IL-6, and IL-8 promoters (28, 30, 36, 39).
Parallel with the downregulation of the leukotriene pathway, there was an increased expression of PGE2, correlated with an increased expression of cyclooxygenase (COX-2) in the bacterial supernatant condition. This COX-2 upregulation was in agreement with results found by Rose et al. (43) on A549, and by Lin et al. (31) on human pulmonary epithelial cells. Xie et al. (54) have shown that jun, which is upregulated in the bacterial supernatant condition, can mediate v-src-induced COX-2 gene expression, through its participation to the COX-2 promoter CRE site. This suggests that the activation of jun in the bacterial supernatant condition may contribute to the observed stimulation of COX-2. Additional factors may play a role in the regulation of COX genes. For instance, COX-2 expression is known to be induced by several cytokines (35), and its promoter contains NF-B binding sites. As for IL-1
, IL-1 ß, IL-6, and IL-8, increased transcription of COX-2 was mainly found in the bacterial supernatant condition, consistent with a possible direct regulation by NF-
B.
NF-B effects are indeed not restricted to S. aureus, and many reports indicate similar effects after interaction with gram-negative bacteria such as Escherichia coli (1, 27) or Shigella flexneri (40). These other reports attribute a key role to Toll-like receptors (TLR) to initiate this mechanism, and it was interesting to notice an increased transcription of TLR2 in bacterial supernatant condition (Fig. 2).
Boldrick et al. (6) found that most of genes encoding NF-B/Rel proteins (i.e., NF-
B1, NF-
B2, relB, cRel, I-
B
) can be regulated after infections by diverse bacteria. We identified an effect on the NF-
B/Rel complex transcription (NF-
B1/p105, Rel, and NF-
BIE) in bacterial supernatant condition during the time-course experiment (Fig. 2). Unfortunately, this effect was not detected at 3 h, neither by cDNA microarray (Table 2) nor by sqRT-PCR (Table 3). Since Bottero et al. (7) have shown that activation of the NF-
B protein complex closely correlates with the transcriptional level of NF-
B IA, the measurement of its mRNA level appears as a rapid, sensitive, and powerful method to quantify the transcriptional power of NF-
B. The increased expression of NF-
B IA observed in our experiments therefore signed the activation of the NF-
B complex in the bacterial supernatant condition and to a lesser extent in the live bacteria condition. Bacterial activation of NF-
B is in agreement with results of DiMango et al. (14), who showed that NF-
B was activated after interaction between Pseudomonas aeruginosa and respiratory epithelial cells. We also noticed the activation of many NF-
B-responsive genes as a strong indication of the activation of this cascade.
Expression of several "anti-apoptotic" genes was observed in the bacterial supernatant condition, and, to a lesser extent, in the live bacteria condition. The development of an anti-apoptotic response may appear surprising, since many reports rather suggested pro-apoptotic mechanisms consecutive to infection. This apoptotic response was shown in bovine mammary epithelial cells (MAC-T) (3, 53) and in endothelial cells, via adhesion and internalization of S. aureus (33). However, Weinrauch et al. (52) showed that low doses of -toxin induce DNA fragmentation and cell death, whereas high doses result in massive necrosis without DNA fragmentation. In agreement with Da Silva et al. (10), who have worked on the same MM-39 cell line, we observed a prominent necrosis phenotype of the MM-39-cells, with a transient apoptotic process 9 h after interaction (Fig. 5).
TNF expression was not altered at a RNA level (microarray and sqRT-PCR), but TNF
secretion was slightly increased in the bacterial supernatant condition. Under these experimental conditions, TNF
concentration reached 15 pg/ml of MM-39 cell culture medium, whereas it reached almost 1,000 pg/ml in the experiments of Haddad and coworkers (19), when epithelial cells were in contact with E. coli LPS. Our results were in agreement with those found by Stout et al. (47), who showed that addition of staphylococcal glycocalyx induced production of PGE2 and IL-1 by murine peritoneal macrophage, but not of TNF
. The modest increase in TNF
release in our experiment might very well be related to the inhibitory effect of PGE2 on TNF
secretion in macrophages, which does not affect IL-1 release (16, 29, 46). Bacterial supernatant also induced an upregulation of the gene encoding BIRC3 (AIP1), which can act as a repressor of TNF signaling by binding to the tumor necrosis factor receptor-associated factors TRAF1 and TRAF2.
In conclusion, our results show that the transcriptional response of airway epithelial cells to S. aureus differ markedly when cells interact with live bacteria in log-phase growth or with soluble factors from stationary growth phase. The activation of the JAK/STAT pathway, as well as NF-B and AP-1 represent major activating pathways after stimulation by soluble virulence factors. It accounts for the induction of proinflammatory cytokines such as IL-8 as well as inflammatory mediators such as PGE2. These "stereotypic" properties are consistent with many other studies performed with other bacteria and other cells. These results therefore appear a posteriori as a good validation of our experimental approach. Besides, our study shows that additional transcription factors, such as HIF-1
, may play a role in the orchestrated response of airway cells to S. aureus infection. S. aureus infection also lead cells to a necrotic, rather than an apoptotic cell death. One important aspect about S. aureus infection concerns its capacity to favor subsequent infections by gram-negative bacterial infections such as P. aeruginosa in cystic fibrosis patients. We anticipate that upregulation of inflammatory mediators and/or any other modifications observed in the present study may provide interesting new pharmacological targets to improve the clinical status of S. aureus infected patients.
![]() |
GRANTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
ACKNOWLEDGMENTS |
---|
![]() |
FOOTNOTES |
---|
Address for reprint requests and other correspondence: P. Barbry, Institut de Pharmacologie Moléculaire et Cellulaire UMR 6097 CNRS, Sophia-Antipolis, 06560 Valbonne, France (E-mail: barbry{at}ipmc.cnrs.fr).
10.1152/physiolgenomics.00135.2004.
1 The Supplementary Material for this article (Supplemental Table S1) is available online at http://physiolgenomics.physiology.org/cgi/content/full/00135.2004/DC1.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|