Figures
Abstract
Background and Aims
In cystic fibrosis (CF), Pseudomonas aeruginosa is not eradicated from the lower respiratory tract and is associated with epithelial inflammation that eventually causes tissue damage. To identify the molecular determinants of an effective response to P. aeruginosa infection, we performed a transcriptomic analysis of primary human bronchial epithelial cells from healthy donors (CTRL) 2, 4, and 6 h after induced P. aeruginosa infection. Compared to noninfected cells, infected cells showed changes in gene activity, which were most marked 6 h postinfection and usually consisted in upregulation.
Results
By comparing for each time point of infection, the transcriptomic response of epithelial cells from CF patients and healthy donors, we identified 851, 638, 667, and 980 differentially expressed genes 0, 2, 4, and 6 h postinfection, respectively. Gene selection followed by bioinformatic analysis showed that most of the differentially expressed genes, either up- or downregulated, were in the protein-binding and catalytic gene-ontology categories. Finally, we established that the protein products of the genes exhibiting the greatest differential upregulation (CSF2, CCL2, TNF, CSF3, MMP1, and MMP10) between CF patients and CTRL were produced in higher amounts by infected cells from CF patients versus CTRL.
Citation: Balloy V, Varet H, Dillies M-A, Proux C, Jagla B, Coppée J-Y, et al. (2015) Normal and Cystic Fibrosis Human Bronchial Epithelial Cells Infected with Pseudomonas aeruginosa Exhibit Distinct Gene Activation Patterns. PLoS ONE 10(10): e0140979. https://doi.org/10.1371/journal.pone.0140979
Editor: Estelle Cormet-Boyaka, The Ohio State University, UNITED STATES
Received: June 29, 2015; Accepted: September 25, 2015; Published: October 20, 2015
Copyright: © 2015 Balloy et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: RNAseq raw datafiles are available in the European Nucleotide Archive (ENA) (primary accession number PRJEB9292). http://www.ebi.ac.uk/ena/data/view/PRJEB9292.
Funding: This work was supported by a grant from two French non-profit cystic-fibrosis organizations, Vaincre la mucoviscidose and Association Gregory Lemarchal. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Cystic fibrosis (CF) is caused by mutations of the gene encoding the cystic fibrosis transmembrane conductance regulator (CFTR), a membrane protein functioning as a chloride channel and expressed at the surface of many epithelia. Defective CFTR function is associated with ion-transport abnormalities that affect the function of a variety of organs. Pulmonary manifestations are usually at the forefront of the clinical picture. CFTR dysfunction in the respiratory epithelia is responsible for dehydration of the airway surface liquid, inspissation of secretions, and deficient mucociliary transport. These abnormalities impair the clearance of inhaled pathogens, thereby allowing opportunistic infections to develop [1]. The bacterial CF microbiota varies significantly with patient age [2, 3]. In young patients, Staphylococcus aureus and Haemophilus influenza are commonly found. Over time, the opportunistic pathogen Pseudomonas aeruginosa gains predominance, contributing over 80% of lung bacteria in adults with CF [4]. This chronic P. aeruginosa infection and the attendant prolonged inflammatory response cause tissue damage with a progressive decline in lung function that produces most of the morbidity and mortality [5] associated with CF.
The epithelium lining the conducting airways plays a central role in the innate immune response. It is the first barrier against pathogens and possesses several defense mechanisms against colonization by inhaled viruses, fungi, and bacteria [6]. Mechanical factors by which the airway epithelial cells clear pathogens include mucus secretion and the mucociliary escalator. In addition, these cells produce inflammatory mediators involved in the recruitment of cells such as neutrophils to the site of infection, and they are the main source of antimicrobial peptides. Absence of functional CFTR may impair the antimicrobial capabilities of epithelial cells by elevating NaCl concentrations and decreasing the pH of the airway surface liquid [7]. The airways of patients with CF contain high levels of proinflammatory mediators that continuously recruit neutrophils. Although links have been established between epithelial CFTR dysfunction, defective bacteria clearance, and intense inflammatory responses, the relationship between CF airway inflammation and infection remains unclear.
We hypothesized that P. aeruginosa infection triggers a maladaptive response in the CF epithelial cell mRNA profile. We assessed this hypothesis by comparing the transcriptomic response to P. aeruginosa infection in CF and normal epithelial airway cells. Our goal was to identify genes that were upregulated and downregulated in the CF cells, compared to the control (CTRL) cells, in response to P. aeruginosa infection, under the assumption that such changes in gene regulation might explain the strong inflammatory response and chronic P. aeruginosa infection in patients with CF.
Materials and Methods
Human bronchial epithelial cell culture
Human airway epithelial cells from bronchial biopsies (hAECBs) were purchased from Epithelix (Plan-les-Ouates, Switzerland), received at passage 1, and cultured in 75-cm2 culture flasks with serum-free Epithelix hAEC culture medium, which was changed every 3 days. One week later, the cells were subcultured in 6-well plates (105 cells/well). After the cells reached confluence, they were incubated overnight in DMEM containing 10% fetal calf serum, 10 mM Hepes, 1% penicillin, and 1% streptomycin before being infected.
Bacterial strain and growth conditions
The P. aeruginosa PAK strain used previously [8], expresses the full complement of virulence factors including pili; flagella; the type III secreted exoenzymes S, T, and Y; and a smooth lipopolysaccharide (LPS) belonging to serotype 6.
P. aeruginosa was grown for 12 h in Luria-Bertani medium, then diluted 4·10−6-fold, transferred to fresh medium, and grown overnight to the midlog phase. The culture was centrifuged and the bacterial pellet was washed twice with cold phosphate-buffered saline. The optical density measured at 600 nm was adjusted to give the desired concentration, as previously described [8]. The bacterial count was confirmed by plating serial dilutions on Luria-Bertani agar plates.
Infection protocol and flowchart of the experimental procedure
We used hAECBs from 4 patients with CF (3 females and 1 male; mean age: 26.5 ± 1.9 years) who were homozygous for the p.F508del mutation and from 4 healthy donors (2 females and 2 males; mean age: 64.5 ± 11.6 years) who had no known diseases or history of smoking (CTRL) (S1 Table). The CF and CTRL hAECBs were simultaneously infected with P. aeruginosa at 0.25 MOI for 0, 2, 4, 6, and 8 h in DMEM without antibiotics.
At each postinfection time point, supernatants were collected and centrifuged at 3000 × g for 15 min to remove the bacteria. Before storage at -80°C, the supernatants were tested for viability of the infected and noninfected cells, by measuring lactate dehydrogenase (LDH) activity (CytoTox 96® NonRadioactive Cytotoxicity Assay, Promega, Madison, WI). Lysis of noninfected cells was used as the basal value to estimate lysis of infected cells. Interleukin (IL)-8 and IL-6 were measured using Duo-Set ELISA kits (R&D Systems, Abingdon, UK) in supernatants to assess epithelial-cell activation by P. aeruginosa.
Cell lysates were stored at -80°C until RNA isolation (see below). The transcriptomes of the 4 CF and 4 CTRL specimens were analyzed 0, 2, 4, and 6 h postinfection (32 analyses in all). Protein expression was assessed 8 h postinfection. We first evaluated gene regulation in CTRL cells infected with P. aeruginosa versus nonstimulated cells. Then, we compared the mRNA expression profiles of infected CF versus CTRL cells at each postinfection time point.
RNA and protein isolation
Small RNAs (< 200 nt), large RNAs (> 200 nt), and proteins were isolated in three separate fractions, using the NucleoSpin® miRNA kit (Macherey-Nagel, Düren, Germany).
RNA purification and library construction
We used 2 to 5 μg of large RNA to purify polyadenylated mRNAs and to build an RNA library, using TruSeq RNA Sample Prep Kit v2 (Illumina, #RS-122-2001 and #RS-122-2002, San Diego, CA) as recommended by the manufacturer. The nondirectional libraries thus obtained were CTRL led by Bioanalyzer DNA1000 Chips (Agilent Technologies, #5067–1504, Santa Clara, CA) and quantified using spectrofluorimetry (Quant-iT™ DNA High-Sensitivity Assay Kit, #Q33120, Invitrogen, Life Technologies, Carlsbad, CA).
Sequencing and Bioinformatic analysis
Sequencing of the 32 samples was performed on the HiSeq 2000 sequencer (Illumina) in single-end mode in order to have approximately 100 million reads of 50 bases per sample.
Reads were cleaned of adapter sequences and low-quality sequences using an in-house program (https://github.com/baj12/clean_ngs). Only sequences at least 25 nt in length were considered for further analysis. TopHat (version 1.4.1.1, default parameters, http://ccb.jhu.edu/software/tophat) was used for alignment on the reference genome (hg19). Genes were counted using HTseq-count (parameters: -m intersection-nonempty, -s yes, -t exon) [9, 10].
The PANTHER (Protein ANalysis THrough Evolutionary Relationships) (http://www.pantherdb.org/) Classification System was used to classify genes identified in our transcriptomics data according to their molecular function annotated using ontology terms [11, 12].
Multiplex detection immunoassays
To confirm that genes of interest showed significant differences in expression between CF and CTRL cells, protein levels were measured in stimulation supernatants, using the Human Magnetic Luminex Screening Assay (R&D Systems). We used an 11-plex panel to assay cytokines, chemokines, and growth factors; and a 3-plex panel to assay matrix metalloproteinase (MMP) levels.
Statistical analysis
Cytokines.
To determine whether differences between the CF and CTRL groups were statistically significant, we used Prism 6.00 software (GraphPad Software, La Jolla, CA), as indicated in the figure legends. P values <0.05 were considered significant. ANOVA was performed to compare quantitative variables across groups. To correct for multiple testing, we used either Bonferroni’s method (comparisons of all groups) or Dunnett’s method (comparisons of groups vs. a CTRL condition).
RNAseq data.
Counts were analyzed using R version 3.0.2 and the Bioconductor package DESeq2 version 1.2.9 (www.bioconductor.org) [13]. Data were normalized using DESeq2 and the default parameter. To estimate dispersion and test for differential expression, we used the default parameters (including outlier detection and independent filtering). The generalized linear model was set with time, condition (CF vs. CTRL), and patient as the main effects and with time x condition and condition x patient as the interaction terms, since the patient effect was nested within the condition effect and cells from each patient were studied at four different time points. To avoid model overfitting, patients were designated by the same label in each condition, and their nesting within the condition was included in the model through the condition x patient interaction term. Extracted contrasts included comparisons between the CF and CTRL groups at each time point and pairwise comparisons of time points in each condition. Raw P values were adjusted for multiple testing according to the Benjamini and Hochberg procedure, and genes with an adjusted P value <0.001 were considered differentially expressed [14].
Results
Activation of primary bronchial epithelial cells (hAEBCs) by P. aeruginosa
After infection, the kinetics of IL-8 and IL-6 production were comparable for CF and CTRL cells (Fig 1A and 1B, respectively). However, IL-6 and IL-8 concentrations differed significantly between CF and CTRL cells from 6 hours and at 8 hours postinfection, respectively. These data were obtained under experimental conditions in which P. aeruginosa infection did not alter the viability of CF or CTRL cells, as assessed based on LDH release (data not shown).
Kinetics of IL-8 (A) and IL-6 (B) production by hAEBCs infected with P. aeruginosa. At each time point, supernatants were collected for ELISAs of IL-8 and IL-6. The data are mean±SEM of IL-8 or IL-6 concentrations for four different patients. The statistical analysis consisted in ANOVA followed by Bonferroni’s multiple comparison test. *P<0.05 and **P<0.001. Each value in the cystic-fibrosis (CF) group was compared with the value at the same time point in the CTRL group.
Transcriptional response of CTRL primary bronchial epithelial cells (hAEBCs) to P. aeruginosa
When we analyzed the transcriptomic response of CTRL cells to P. aeruginosa infection, we found a significantly greater number of upregulated than of downregulated genes, compared to noninfected CTRL cells: 2, 4, and 6 h postinfection, the numbers of upregulated genes were 0, 13, and 186; whereas the numbers of downregulated genes were 0, 11, and 25; respectively (P<0.001).
Among the differentially expressed genes, we selected those whose fold change (FC) was ≥2 in the event of upregulation and ≤0.5 in the event of downregulation: 75 upregulated genes and 6 downregulated genes (S2 Table) were thus selected.
PANTHER (Protein ANalysis THrough Evolutionary Relationships) was used to classify genes according to gene-ontology molecular functions. Upregulated genes fell into four main functional categories: protein binding (37.1% of upregulated genes), catalytic activity (13.5%), nucleic-acid-binding transcription-factor activity (12.4%), and receptor activity (9%). The number of downregulated genes was too small for a meaningful classification.
Among upregulated genes in the protein-binding category, those having the greatest fold change values encoded the transcription factors Kruppel-like factor 2 (KLF2) and Nuclear receptor subfamily 1, group D, member 1 (NR1D1); the cytokine IL-17C; the chemokine (C-C motif) ligand 20 (CCL20); the hormonally active protein adrenomedullin 2 (ADM2); and the apoptosis inhibitor baculoviral IAP repeat-containing protein 3 (BIRC3) (Table 1).
Among the downregulated genes, we identified three genes having known molecular functions, namely, the two transcriptional regulators early growth response protein 1 (EGR1) and SMAD family member 6 (SMAD6); and the nucleosome assembly protein (NAP) family member NAP 1-like 3 (NAP1L3).
Differential gene activation in cystic fibrosis (CF) versus CTRL cells in response to P. aeruginosa infection
We compared genes that were differentially expressed in CF and CTRL cells at each postinfection time point. The heat-map of the mean gene-expression level under each condition revealed two distinct clusters that separated CF and CTRL cells, one of upregulated genes and the other of downregulated genes (Fig 2A).
A- The heat-map of the mean expression levels of all the genes in each condition revealed two distinct clusters that separated cystic-fibrosis (CF) cells from CTRL cells. B and C- Venn diagram of differentially expressed genes between cystic-fibrosis and CTRL cells upon P. aeruginosa infection. Circles: Differentially expressed genes upregulated (B) or downregulated (C) at a single postinfection time point or at two or three consecutive postinfection time points (thus, 0h was not considered). We selected the genes whose fold-change in expression level was ≥2 in the event of upregulation and ≤0.5 in the event of downregulation. Squares: Differentially expressed genes upregulated (B) or downregulated (C) at two, three, or four consecutive time points (thus, 0h was considered). We selected the genes for which the ratio of FC in infected cells over FC in noninfected cells was ≥1.5 in the event of upregulation or ≤0.6 in the event of downregulation.
We then identified the genes whose level of expression differed between CF and CTRL cells with P<0.001. The numbers of these genes were 851, 638, 667, and 980 at the four postinfection time points (0, 2, 4 and 6 h), respectively (data not shown).
An MA plot (transcripts data have been transformed onto the M (log ratios) and A (mean average) scale) was created to visualize differences in gene expressions between CF and CTRL cells at each postinfection time point. Among the differentially expressed genes 541, 343, 415, 532 were upregulated and 310, 295, 252, and 448 were downregulated at 0, 2, 4, and 6 h postinfection, respectively (data not shown).
We then created Venn diagrams representing upregulated (Fig 2B) and downregulated genes (Fig 2C). We identified genes that were upregulated or downregulated (134 and 57, respectively) only at baseline (0h). These genes were not considered further, since our goal was to study gene expression changes induced by P. aeruginosa infection. We then separated the genes differentially expressed at a single postinfection time point and those differentially expressed at two or three consecutive time points (i.e., 0, 2, 4, 6; or 0, 2, 4; or 0, 2; or 2, 4, 6; or 4, 6; or 2, 4 h). To exclude genes whose differences in expression between CF and CTRL cells were similar at baseline (0 h) and after infection, and therefore probably related to the CFTR genotype and not to P. aeruginosa infection, for each gene, we computed the ratio of the fold change value at each postinfection time point over the fold change value at 0h. We then selected the genes for which this ratio was ≥1.5 (arbitrary choice) in the event of upregulation and ≤0.6 (arbitrary choice) in the event of downregulation, at one or more postinfection time points. For genes differentially expressed at a single time point or at two or three consecutive time points (2 and 4 h; or 2, 4 and 6 h; or 4 and 6 h), we defined a significant change in gene expression as fold change ≥2 in the event of upregulation and fold change ≤0.5 in the event of downregulation. These criteria identified 171 upregulated genes and 236 downregulated genes in CF versus CTRL cells, which are listed in S2 Table.
Genes differentially upregulated in cystic fibrosis (CF) versus CTRL epithelial cells infected with P. aeruginosa
PANTHER was then used to identify key functions differentially affected by the response to infection of CF and CTRL cells. In all, seven gene-ontology molecular-function categories were overexpressed in CF compared to CTRL cells after P. aeruginosa infection (Table 2).
The two main categories were protein binding and catalytic activity, with 34.6% and 25.5% gene hits vs. total genes, respectively. The five genes the most differentially expressed in the protein-binding and catalytic-activity categories are reported in Table 3A and 3B, respectively.
We then focused on genes for which theoretical considerations suggested a role in the intense inflammatory response to infection seen in patients with CF. Some of these genes belonging to protein-binding family were involved in the inflammatory response, such as tumor necrosis factor (TNF), IL-17C, colony-stimulating factor (CSF) 2 (also called GM-CSF), CSF3 (also called G-CSF), and CCL2. Others code for proteolytic enzymes classified in the catalytic category, namely MMP1, MMP10, and MMP13.
IL-17C and CSF2 mRNA expressions were differentially upregulated in CF cells 4 and 6 h postinfection, respectively. TNF, CSF3, and CCL2 mRNAs were already differentially upregulated in CF cells at baseline (0 h) and the difference was increased 4 h postinfection. Similarly, MMP1 and MMP13 mRNAs showed significant differential upregulation in CF cells at 0h, with increases in the differences 4 h postinfection. By contrast, MMP10 was differentially upregulated only at the 4 h postinfection time point.
Among proteins (Fig 3), IL-17C and MMP13 were not detected, even 8 h postinfection. TNF and CSF3 protein levels differed significantly between CF and CTRL cells 8 h postinfection. After infection, particularly at the 6 h and 8 h time points, CSF2, CCL2, MMP1, and MMP10 proteins were expressed at higher levels in CF than CTRL cells, although the differences were not statistically significant.
Protein levels were measured in supernatants of infected CF and CTRL cells 0, 2, 4, 6, and 8 hours postinfection, using the Human Magnetic Luminex Screening Assay. Data are mean±SEM of cytokine concentrations for four different patients. The statistical analysis consisted in ANOVA followed by Bonferroni’s multiple comparison test. *P<0.05 and **P<0.001. Each value in the cystic-fibrosis (CF) group was compared with the value at the same time point in the CTRL group.
We detected significant differential production of IL-8 and IL-6 proteins between CF and CTRL cells (see Fig 1). As expected, the transcriptional analysis shows significant induction of IL-8 or IL-6 mRNA expression (P<0.05).
To explain the excessive inflammation triggered by P. aeruginosa in patients with CF, we assessed the mRNA expression level of Pattern Recognition Receptors (PRRs) known to be involved in P. aeruginosa recognition. P. aeruginosa expressed numerous pathogen-associated molecular patterns (PAMPs) such as LPS recognized by toll-like receptors (TLRs) 2, 4 and flagellin recognized by TLR5. None of the PPRs met the stringent P or fold change criteria used in our study. Nevertheless, some PRRs may be involved in the upregulation of proinflammatory genes. Thus the mRNAs for TLR2, 4, and 6 were significantly (P<0.05) upregulated in CF cells both at 0 h and after P. aeruginosa infection (S1 Fig). Other PRRs showed nonsignificant differences in expression: although TLR9 was upregulated and TLR5 downregulated, in CF cells compared to CTRL cells. Nucleotide-binding oligomerization domain containing 2 (NOD2) was similarly induced by infection in CF and CTRL cells
Genes differentially downregulated in cystic fibrosis (CF) versus CTRL epithelial cells infected with P. aeruginosa
PANTHER identified a total of 10 gene-ontology molecular-function categories that were expressed at lower levels in CF cells than in CTRL cells (Table 4).
Catalytic activity and protein binding were the two main underexpressed categories, with 33% and 27.5% of genes, respectively. The five genes in the catalytic-activity and protein-binding categories exhibiting the largest differences in expression between CF and CTRL cells are listed in Table 5A and 5B, respectively.
Genes in the catalytic-activity category included kallikreins (KLK12 and KLK13) and ubiquitin-specific peptidase 9, Y-linked (USP9Y), which are enzymes exhibiting peptidase activity. Genes in the protein-binding category had a variety of functions, such as hormonal (ADM2), transcriptional (E74-like factor 5 [ELF5]), translational (eukaryotic translation initiation factor 1A, Y-linked [EIF1AY]), and calcium-ion binding (troponin C type 1 [TNNC1]) activities. For most of these genes, the difference in downregulation between CF and CTRL cells was significantly larger 6 h postinfection than at other time points; the exceptions were KLK12 and EIF1AY, for which the differences were largest 2 h postinfection.
Discussion
The respiratory tract epithelium is a critical environmental interface that participates to the regulation of the innate immune response of the host to infection. Host-pathogen interactions determine the intensity of the innate inflammatory response and, therefore, the risk of tissue injury. Our primary objective was to characterize the transcriptomic response of healthy primary bronchial epithelial cells to P. aeruginosa infection, in order to identify effectors of the innate immune response involved in combatting bacterial infections. We also planned to compare this transcriptomic response with that of infected epithelial cells from patients with CF. We reasoned that differences between normal and CF cells might point to abnormalities in innate immune function responsible for failure of CF epithelial cells to eradicate P. aeruginosa, as well as for the generation of an exaggerated inflammatory response.
We first investigated primary bronchial epithelial cells from healthy CTRL to assess their transcriptional response to alive P. aeruginosa infection. We performed tests infection of CTRL cells at 2, 4, 6 h to allow sufficient time for expression of P. aeruginosa virulence factors and of host genes and then compared gene expressions of infected to non-infected cells (0 h). The response was strongest 6 h postinfection; in contrast, after 2 h the transcriptome of P. aeruginosa-infected epithelial cells was unchanged. After 4 h, only two upregulated and three downregulated genes showed an at least 2-fold increase or decrease in their expression level versus uninfected cells. After 6 h, the same criterion was met by 75 upregulated and 6 downregulated genes.
The upregulated genes included genes encoding innate immunity effectors, such as IL-17C, CCL20, and TNF, which contribute to combat bacterial infections. The gene for the transcription factor KLF2 was also upregulated. A previous study established that KLF2 overexpression was induced by ExoS, a toxin delivered through the P. aeruginosa type III secretion system and associated with epithelial-cell death [15].
Among downregulated genes, we found NAP1L3, SMAD6, cell division cycle-associated protein 7 (CDCA7), EGR1, and zinc finger protein 367 (ZNF367). SMAD6 promotes the antiinflammatory response by playing an important role in transforming growth factor-beta 1 (TGF-ß1)-mediated negative regulation of proinflammatory signaling [16]. SMAD6 downregulation during Gram-negative bacterial infection has been reported previously: the bacterial endotoxin LPS can inhibit the TGF-ß1 signaling pathway leading to SMAD6 expression [17]. EGR1 is a nuclear protein that functions as a transcriptional regulator. The EGR1 downregulation observed in our infected CF cells may be ascribable to acyl-homoserine lactones, which are intercellular signaling molecules used in quorum sensing by P. aeruginosa and Burkholderia cepacia [18].
Previous studies have identified genes whose expression in respiratory epithelial cells changed after P. aeruginosa infection. High-density DNA microarrays have been used to identify regulated genes in A549 cells infected with live PAK P. aeruginosa then studied for 3 h [19]. An at least 2-fold change in the expression level was found for only 22 upregulated genes and 2 downregulated genes. Some of these genes were also upregulated in our study such as Ras homolog family member B (RHOB), E74-like factor 3 (ELF-3 or ESE-1), nuclear factor of kappa light polypeptide gene enhancer in B-cells inhibitor, alpha (NFKBIA or MAD3), and macrophage cationic peptide 1 (MCP-1)). In contrast, the two downregulated genes were not found. Serial analysis of gene expression (SAGE) has been used for a transcriptomic analysis of primary bronchial epithelial cells infected with heat-inactivated P. aeruginosa (PAO1 strain) for 6 h [20]. The results indicated regulation of four categories of genes: keratins, proteinase inhibitors, IL-1 family members, and S100 calcium-binding proteins. None of these genes were regulated in our study. These discrepancies across studies may be ascribable to differences in respiratory epithelial cell types (e.g., bronchial, alveolar, immortalized, or primary cells), bacteria (strains, inactivated vs. alive), MOI, time of infection, or even expression-profiling techniques.
The transcriptomic response of CTRL cells to P. aeruginosa may constitute the signature of the ideal immune response of respiratory epithelial cells to infection. To identify gene expression differences that might explain the inadequate innate immune response of CF cells to infection, we compared the transcriptomes of CF and CTRL primary bronchial epithelial cells infected with P. aeruginosa. The two groups of cells were exposed to identical experimental conditions, to avoid introducing confounding factors. An important difference with our data concerning the response of infected CTRL cells, for which a small number of gene expressions were modulated, is the large number of genes whose expression levels differed between CF and CTRL cells. After P. aeruginosa infection, many genes whose expression was not altered in CTRL cells showed upregulation or downregulation in CF cells.
Interestingly, the proinflammatory genes that were upregulated in CF cells compared to CTRL cells included the genes encoding IL-17C, TNF, CSF2, CSF3, and CCL2, all of which are major mediators of inflammation. However, protein expression was not evidenced for all of these genes. Thus, IL-17C was not detected in supernatants of CF or CTRL cells, even 8 h postinfection. After 6 h, among mRNA levels, those for TNF were higher than those for CCL2 and CSF2, and similar to those for CSF3. These differences were not reflected in the protein concentrations in the supernatants: TNF protein concentrations were very low (7 and 13 pg/mL after 6 h and 12 and 28 pg/mL after 8 h in CTRL and CF cell supernatants, respectively); whereas CSF2, CCL2, and CSF3 protein levels were high. Nonetheless, 8 h postinfection, TNF and CSF3 protein concentrations were significantly higher in CF than in CTRL cells, in keeping with the results of the transcriptomic analysis. CSF2 and CCL2 protein levels were higher in CF cells than in CTRL cells and increased further with time after infection, although the differences were not statistically significant. In general, for the genes with the greatest differences in expression, we also found differences in protein levels. CSF2 and CSF3 are involved in the host response to microbial infections [21, 22]. CSF2 secretion promotes neutrophil maturation and activation [23]. Thus, CSF2 activates the Jak/STAT pathway, thereby delaying spontaneous neutrophil apoptosis [24]. CSF3 may contribute to lung inflammation, as it increases neutrophil chemotactic activity. Whereas the production of CSF2 is well documented, less is known about the regulation of CSF3 in airway epithelial cells. In one study, lung epithelial cells stimulated by LPS from P. aeruginosa released CSF3 [25]. These findings are consistent with the presence in CF of airway inflammation characterized by a profuse influx of neutrophils into the lungs.
The main family of mediators in the catalytic category is the MMP family. MMPs are zinc-dependent enzymes with proteolytic activity against a wide range of extracellular proteins. MMPs are therefore involved in extracellular matrix turnover, tissue degradation, and tissue repair [26]. MMPs are anchored to the cell surface or secreted as proenzymes that are activated by proteolytic cleavage [27]. Their potential substrates include membrane proteins and proteins in the extracellular compartment. MMP activity is controlled by specific tissue inhibitors of metalloproteases (TIMPs) [28]. MMPs are expressed in a variety of normal and disease processes, such as development, involution, repair, inflammation, and tumor growth. They play a major role in regulating the host immune response to infections. By degrading extracellular matrix components and by modulating cytokine and chemokine activity, MMPs allow the migration of inflammatory cells from the bloodstream and drive their recruitment [29–31]. In our model, MMP1, MMP10, and MMP13 mRNAs, which were upregulated, showed the greatest differences between CF and CTRL cells. MMP1 and MMP10 proteins were detected in the supernatants of infected cells; in contrast, MMP13, although described as a secreted MMP, was not detected. As with CCL2 and CSF2, differences in MMP1 and MMP10 protein levels between CF and CTRL cells were not significant 8 h postinfection; nevertheless, in the CF cells these levels were higher and increased over time. In an in vivo model, MMP10 seemed essential in activating protective host responses initiated by the epithelium after P. aeruginosa infection [32]. However, it has been demonstrated that MMP10 shares with the neutrophil serine proteases an ability to activate MMP1 [33]. Although nothing is known about the role for MMP1 or MMP10 in CF, our data suggest that these two enzymes may be involved in tissue damage related to excessive inflammation and may contribute to infection-related CF pathology. We also detected significant differences in IL-8 and IL-6 production after infection in CF and CTRL cells. In lung samples from patients with CF, IL-8 and IL-6 levels were elevated and correlated with airway obstruction [34–37].
The increases in TLR2 and TLR4 expression may contribute to the proinflammatory response of CF cells. In one study, LPS recognition was sufficient to activate TLR2- and TLR4-dependent signaling in epithelial cells [38]. Similarly, increased TLR9 mRNA expression in infected CF cells may be clinically relevant, as TLR9 has been found to play a detrimental role in the pulmonary response to P. aeruginosa [39]. However, investigations of TLR protein expression would be of interest, as one study showed TLR6 mRNA expression without TLR6 protein expression in epithelial cell lines [40].
In contrast to the upregulated genes encoding cytokines and MMPs, the genes that were downregulated in CF cells in our study have not been previously identified as relevant in CF. Nevertheless, biological considerations suggest that downregulation of these genes may explain phenotypic features of CF. Adrenomedullin 2 (ADM2) protects against tissue injury in various organs including the lung, central nervous system, cardiovascular system, and kidney. In a rat model of testis inflammation, ADM2 treatment protected against LPS-induced damage by decreasing the levels of reactive oxygen species, TNF-α, IL-6, and IL-1 ß [41]. ADM2 attenuated the severity of ventilator-induced lung injury in mice [42] and, when used in low doses, protected mouse lungs from early ischemia/reperfusion injury by preserving the integrity of the blood-air barrier and by strongly decreasing the influx of leukocytes into the alveolar spaces [43]. ELF5 (also called ESE-2), an epithelial-specific member of the ETS transcription factor family, is expressed at high levels in the salivary glands, mammary glands, and trachea and at far lower levels in the lung [44]. ELF5 regulates the epithelium-specific gene K18 and may contribute to epithelial regeneration [45].
Other groups have performed studies to evaluate changes in the transcriptional response of CF compared to CTRL epithelial cells after P. aeruginosa infection. Differences in the experimental protocols preclude detailed comparisons of the results. However, the findings fall into two main categories. Thus, in one study, infected CF cells (IB3 cell line) overexpressed proinflammatory mediators such as cytokines, chemokines, and growth factors but underexpressed protease inhibitors, compared to CTRL cells (S9 cell line) [46]. Although the genes described do not exactly match those identified in our study, they belong to the same biological families. In contrast, another study produced contradictory results to ours, with higher levels of IL-8, IL-6, and ICAM-1 expression in CTRL cells (CFT1-LCFSN) compared to CF cells (CFT1-ΔF508) after infection [47].
The limitation in this study is the overall difference between the mean ages of the patients and CTRL, which may influences their response against P. aeruginosa. Influence of these parameters was not tested in our analysis since mutation status is shared with age (aged CTRL vs. young CF). Also, we choose to grow hAECBs in 2D monolayers for two main reasons. Firstly, CF cells grown at air-liquid interface produce more mucus than control cells that may prevent bacteria to physically interact with the cells and thus the subsequent cell host response. Also, introducing Pseudomonas aeruginosa, at the apical side of the cells, even in a reduce liquid volume would have resulted in the lost of the air-liquid interface.
In summary, we performed a comparative transcriptomic analysis to obtain a complete picture of the genes differentially regulated in CF versus CTRL respiratory epithelial cells in response to alive P. aeruginosa infection. Our results show that the CF respiratory epithelium responds to the presence of P. aeruginosa by producing, depending on the time of infection, an array of proteins that magnify the inflammatory response: thus, our protein analysis revealed that the infection induced overproduction of TNF, CSF2, CSF3, CCL2, MMP1, and MMP10 in CF cells compared to CTRL cells. Most of these proteins were produced by CTRL cells, but in lower levels. In contrast, after P. aeruginosa infection, some genes were activated in CTRL epithelial cells but not in CF cells. Interestingly, most of these genes are beneficial for tissue integrity. Our data on downregulated genes in infected CF cells may help to understand the mechanisms that underlie the persistence of P. aeruginosa in the CF lung. Furthermore, our identification of upregulated genes in infected CF cells may shed light on the mechanisms of the inflammatory response. We will now investigate the consequences of manipulating the expression of the genes that are differentially regulated in CF and CTRL cells, using markers for inflammation and antimicrobial activity as the endpoints. The expected results may suggest strategies for combatting P. aeruginosa infection, thereby improving both survival and quality of life in patients with CF.
Supporting Information
S2 Table. Lists of genes upregulated or downregulated in CTRL cells upon P. aeruginosa infection.
We defined significant regulation as an at least 2-fold increase (tab: up-regulated genes CTRL cells) or decrease (tab: down-regulated genes CTRL cells) in expression after infection versus baseline (0 h), with P<0.001. ID, unique gene identifier; 0 h, nonstimulated; 4 h and 6 h, 4 and 6 hours postinfection; FC, fold change; adjpBH, adjusted P value to which the α cutoff was applied. Lists of the 171 selected upregulated or 236 selected downregulated genes differentially expressed between CF and CTRL cells in response to P. aeruginosa infection: For genes differentially expressed between CF and CTRL cells at baseline (0 h) and after infection, we computed the ratio of the FC value at each postinfection time point over the FC value at 0 h and then selected ratio ≥1.5 (upregulation: tab: upreg CF vs CTRL) and ≤0.6 (downregulation; tab: down reg CF vs CTRL), at one or more postinfection time points. For genes differentially expressed at a single time point or at two or three consecutive time points (2 and 4 h; or 2, 4 and 6 h; or 4 and 6 h), we selected FC≥2 (upregulation; tab: upreg CF vs CTRL) and FC ≤0.5 (downregulation; tab: down reg CF vs CTRL).
https://doi.org/10.1371/journal.pone.0140979.s002
(XLSX)
S1 Fig. Transcript abundance of innate immune receptors in CF and CTRL cells upon P. aeruginosa infection kinetic profiled by RNA-seq data.
Normalized read count of Toll-Like Receptor (TLR), CD14 and Nod Receptors in CF and CTRL cells upon P. aeruginosa. Asterisks indicate statistically significant differences (*p<0.05; **p<0.01; § = p<0.001, #p<10e-8).
https://doi.org/10.1371/journal.pone.0140979.s003
(TIF)
Author Contributions
Conceived and designed the experiments: LG VB MC JYC HC. Performed the experiments: LG VB CP. Analyzed the data: LG VB BJ MAD OT. Wrote the paper: VB LG MC HV.
References
- 1. Stuart B, Lin JH, Mogayzel PJ Jr. Early eradication of Pseudomonas aeruginosa in patients with cystic fibrosis. Paediatric respiratory reviews. 2010;11(3):177–84. Epub 2010/08/10. pmid:20692633; PubMed Central PMCID: PMCPmc4001925.
- 2. Cox MJ, Allgaier M, Taylor B, Baek MS, Huang YJ, Daly RA, et al. Airway microbiota and pathogen abundance in age-stratified cystic fibrosis patients. PloS one. 2010;5(6):e11044. Epub 2010/06/30. pmid:20585638; PubMed Central PMCID: PMCPmc2890402.
- 3. Klepac-Ceraj V, Lemon KP, Martin TR, Allgaier M, Kembel SW, Knapp AA, et al. Relationship between cystic fibrosis respiratory tract bacterial communities and age, genotype, antibiotics and Pseudomonas aeruginosa. Environmental microbiology. 2010;12(5):1293–303. Epub 2010/03/03. pmid:20192960.
- 4. Pressler T, Bohmova C, Conway S, Dumcius S, Hjelte L, Hoiby N, et al. Chronic Pseudomonas aeruginosa infection definition: EuroCareCF Working Group report. Journal of cystic fibrosis: official journal of the European Cystic Fibrosis Society. 2011;10 Suppl 2:S75–8. Epub 2011/06/17. pmid:21658646.
- 5. Goss CH, Ratjen F. Update in cystic fibrosis 2012. American journal of respiratory and critical care medicine. 2013;187(9):915–9. Epub 2013/05/03. pmid:23634859.
- 6. Boucher RC. New concepts of the pathogenesis of cystic fibrosis lung disease. The European respiratory journal. 2004;23(1):146–58. Epub 2004/01/24. pmid:14738247.
- 7. Pezzulo AA, Tang XX, Hoegger MJ, Alaiwa MH, Ramachandran S, Moninger TO, et al. Reduced airway surface pH impairs bacterial killing in the porcine cystic fibrosis lung. Nature. 2012;487(7405):109–13. Epub 2012/07/06. pmid:22763554; PubMed Central PMCID: PMCPmc3390761.
- 8. Ramphal R, Balloy V, Huerre M, Si-Tahar M, Chignard M. TLRs 2 and 4 are not involved in hypersusceptibility to acute Pseudomonas aeruginosa lung infections. Journal of immunology (Baltimore, Md: 1950). 2005;175(6):3927–34. Epub 2005/09/09. pmid:16148139.
- 9. Trapnell C, Pachter L, Salzberg SL. TopHat: discovering splice junctions with RNA-Seq. Bioinformatics (Oxford, England). 2009;25(9):1105–11. Epub 2009/03/18. pmid:19289445; PubMed Central PMCID: PMCPmc2672628.
- 10. Anders S, Huber W. Differential expression analysis for sequence count data. Genome biology. 2010;11(10):R106. Epub 2010/10/29. pmid:20979621; PubMed Central PMCID: PMCPmc3218662.
- 11. Mi H, Muruganujan A, Thomas PD. PANTHER in 2013: modeling the evolution of gene function, and other gene attributes, in the context of phylogenetic trees. Nucleic acids research. 2013;41(Database issue):D377–86. Epub 2012/11/30. pmid:23193289; PubMed Central PMCID: PMCPmc3531194.
- 12. Mi H, Muruganujan A, Casagrande JT, Thomas PD. Large-scale gene function analysis with the PANTHER classification system. Nature protocols. 2013;8(8):1551–66. Epub 2013/07/23. pmid:23868073.
- 13.
R: A Language and Environment for Statistical Computing [Internet]. 2013. Available from: http://www.R-project.org/.
- 14. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Statist Soc B. 1995;57:289–300.
- 15. O'Grady EP, Mulcahy H, O'Callaghan J, Adams C, O'Gara F. Pseudomonas aeruginosa infection of airway epithelial cells modulates expression of Kruppel-like factors 2 and 6 via RsmA-mediated regulation of type III exoenzymes S and Y. Infection and immunity. 2006;74(10):5893–902. Epub 2006/09/22. pmid:16988269; PubMed Central PMCID: PMCPmc1594899.
- 16. Choi KC, Lee YS, Lim S, Choi HK, Lee CH, Lee EK, et al. Smad6 negatively regulates interleukin 1-receptor-Toll-like receptor signaling through direct interaction with the adaptor Pellino-1. Nature immunology. 2006;7(10):1057–65. Epub 2006/09/05. pmid:16951688.
- 17. Kim EY, Kim BC. Lipopolysaccharide inhibits transforming growth factor-beta1-stimulated Smad6 expression by inducing phosphorylation of the linker region of Smad3 through a TLR4-IRAK1-ERK1/2 pathway. FEBS letters. 2011;585(5):779–85. Epub 2011/02/08. pmid:21295571.
- 18. Sbarbati A, Tizzano M, Merigo F, Benati D, Nicolato E, Boschi F, et al. Acyl homoserine lactones induce early response in the airway. Anatomical record (Hoboken, NJ: 2007). 2009;292(3):439–48. Epub 2009/02/28. pmid:19248172.
- 19. Ichikawa JK, Norris A, Bangera MG, Geiss GK, van 't Wout AB, Bumgarner RE, et al. Interaction of pseudomonas aeruginosa with epithelial cells: identification of differentially regulated genes by expression microarray analysis of human cDNAs. Proceedings of the National Academy of Sciences of the United States of America. 2000;97(17):9659–64. pmid:10931941; PubMed Central PMCID: PMC16921.
- 20. Vos JB, van Sterkenburg MA, Rabe KF, Schalkwijk J, Hiemstra PS, Datson NA. Transcriptional response of bronchial epithelial cells to Pseudomonas aeruginosa: identification of early mediators of host defense. Physiological genomics. 2005;21(3):324–36. Epub 2005/02/11. pmid:15701729.
- 21. Saba S, Soong G, Greenberg S, Prince A. Bacterial stimulation of epithelial G-CSF and GM-CSF expression promotes PMN survival in CF airways. American journal of respiratory cell and molecular biology. 2002;27(5):561–7. Epub 2002/10/25. pmid:12397015.
- 22. Ballinger MN, Paine R 3rd, Serezani CH, Aronoff DM, Choi ES, Standiford TJ, et al. Role of granulocyte macrophage colony-stimulating factor during gram-negative lung infection with Pseudomonas aeruginosa. American journal of respiratory cell and molecular biology. 2006;34(6):766–74. Epub 2006/02/14. pmid:16474098; PubMed Central PMCID: PMCPmc2644237.
- 23. Vliagoftis H, Befus AD, Hollenberg MD, Moqbel R. Airway epithelial cells release eosinophil survival-promoting factors (GM-CSF) after stimulation of proteinase-activated receptor 2. The Journal of allergy and clinical immunology. 2001;107(4):679–85. Epub 2001/04/11. pmid:11295658.
- 24. Epling-Burnette PK, Zhong B, Bai F, Jiang K, Bailey RD, Garcia R, et al. Cooperative regulation of Mcl-1 by Janus kinase/stat and phosphatidylinositol 3-kinase contribute to granulocyte-macrophage colony-stimulating factor-delayed apoptosis in human neutrophils. Journal of immunology (Baltimore, Md: 1950). 2001;166(12):7486–95. Epub 2001/06/08. pmid:11390502.
- 25. Koyama S, Sato E, Nomura H, Kubo K, Miura M, Yamashita T, et al. The potential of various lipopolysaccharides to release IL-8 and G-CSF. American journal of physiology Lung cellular and molecular physiology. 2000;278(4):L658–66. Epub 2000/04/06. pmid:10749742.
- 26. Nagase H, Woessner JF Jr. Matrix metalloproteinases. The Journal of biological chemistry. 1999;274(31):21491–4. Epub 1999/07/27. pmid:10419448.
- 27. Van Wart HE, Birkedal-Hansen H. The cysteine switch: a principle of regulation of metalloproteinase activity with potential applicability to the entire matrix metalloproteinase gene family. Proceedings of the National Academy of Sciences of the United States of America. 1990;87(14):5578–82. Epub 1990/07/01. pmid:2164689; PubMed Central PMCID: PMCPmc54368.
- 28. Brew K, Dinakarpandian D, Nagase H. Tissue inhibitors of metalloproteinases: evolution, structure and function. Biochimica et biophysica acta. 2000;1477(1–2):267–83. Epub 2000/03/10. pmid:10708863.
- 29. Leppert D, Waubant E, Galardy R, Bunnett NW, Hauser SL. T cell gelatinases mediate basement membrane transmigration in vitro. Journal of immunology (Baltimore, Md: 1950). 1995;154(9):4379–89. Epub 1995/05/01. pmid:7722295.
- 30. Faveeuw C, Preece G, Ager A. Transendothelial migration of lymphocytes across high endothelial venules into lymph nodes is affected by metalloproteinases. Blood. 2001;98(3):688–95. Epub 2001/07/27. pmid:11468168.
- 31. Warner RL, Beltran L, Younkin EM, Lewis CS, Weiss SJ, Varani J, et al. Role of stromelysin 1 and gelatinase B in experimental acute lung injury. American journal of respiratory cell and molecular biology. 2001;24(5):537–44. Epub 2001/05/15. pmid:11350822.
- 32. Kassim SY, Gharib SA, Mecham BH, Birkland TP, Parks WC, McGuire JK. Individual matrix metalloproteinases control distinct transcriptional responses in airway epithelial cells infected with Pseudomonas aeruginosa. Infection and immunity. 2007;75(12):5640–50. Epub 2007/10/10. pmid:17923522; PubMed Central PMCID: PMCPmc2168342.
- 33. Saunders WB, Bayless KJ, Davis GE. MMP-1 activation by serine proteases and MMP-10 induces human capillary tubular network collapse and regression in 3D collagen matrices. Journal of cell science. 2005;118(Pt 10):2325–40. Epub 2005/05/05. pmid:15870107.
- 34. Armstrong DS, Hook SM, Jamsen KM, Nixon GM, Carzino R, Carlin JB, et al. Lower airway inflammation in infants with cystic fibrosis detected by newborn screening. Pediatric pulmonology. 2005;40(6):500–10. Epub 2005/10/07. pmid:16208679.
- 35. Dean TP, Dai Y, Shute JK, Church MK, Warner JO. Interleukin-8 concentrations are elevated in bronchoalveolar lavage, sputum, and sera of children with cystic fibrosis. Pediatric research. 1993;34(2):159–61. Epub 1993/08/01. pmid:8233718.
- 36. Sagel SD, Chmiel JF, Konstan MW. Sputum biomarkers of inflammation in cystic fibrosis lung disease. Proceedings of the American Thoracic Society. 2007;4(4):406–17. Epub 2007/07/27. pmid:17652508; PubMed Central PMCID: PMCPmc2647605.
- 37. Sagel SD, Kapsner R, Osberg I, Sontag MK, Accurso FJ. Airway inflammation in children with cystic fibrosis and healthy children assessed by sputum induction. American journal of respiratory and critical care medicine. 2001;164(8 Pt 1):1425–31. Epub 2001/11/13. pmid:11704590.
- 38. Raoust E, Balloy V, Garcia-Verdugo I, Touqui L, Ramphal R, Chignard M. Pseudomonas aeruginosa LPS or flagellin are sufficient to activate TLR-dependent signaling in murine alveolar macrophages and airway epithelial cells. PloS one. 2009;4(10):e7259. Epub 2009/10/07. pmid:19806220; PubMed Central PMCID: PMCPmc2752798.
- 39. Benmohamed F, Medina M, Wu YZ, Maschalidi S, Jouvion G, Guillemot L, et al. Toll-like receptor 9 deficiency protects mice against Pseudomonas aeruginosa lung infection. PloS one. 2014;9(3):e90466. Epub 2014/03/07. pmid:24595157; PubMed Central PMCID: PMCPmc3942450.
- 40. Greene CM, Carroll TP, Smith SG, Taggart CC, Devaney J, Griffin S, et al. TLR-induced inflammation in cystic fibrosis and non-cystic fibrosis airway epithelial cells. Journal of immunology (Baltimore, Md: 1950). 2005;174(3):1638–46. Epub 2005/01/22. pmid:15661927.
- 41. Li L, Ma P, Liu Y, Huang C, O WS, Tang F, et al. Intermedin attenuates LPS-induced inflammation in the rat testis. PloS one. 2013;8(6):e65278. Epub 2013/06/12. pmid:23750251; PubMed Central PMCID: PMCPmc3672160.
- 42. Muller-Redetzky HC, Kummer W, Pfeil U, Hellwig K, Will D, Paddenberg R, et al. Intermedin stabilized endothelial barrier function and attenuated ventilator-induced lung injury in mice. PloS one. 2012;7(5):e35832. Epub 2012/05/09. pmid:22563471; PubMed Central PMCID: PMCPmc3341380.
- 43. Korner C, Kuchenbuch T, Pfeil U, Jung K, Padberg W, Kummer W, et al. Low-dose adrenomedullin-2/intermedin reduces pulmonary ischemia/reperfusion injury. Peptides. 2014. Epub 2014/10/08. pmid:25290159.
- 44. Oettgen P, Kas K, Dube A, Gu X, Grall F, Thamrongsak U, et al. Characterization of ESE-2, a novel ESE-1-related Ets transcription factor that is restricted to glandular epithelium and differentiated keratinocytes. The Journal of biological chemistry. 1999;274(41):29439–52. Epub 1999/10/03. pmid:10506207.
- 45. Yaniw D, Hu J. Epithelium-specific ets transcription factor 2 upregulates cytokeratin 18 expression in pulmonary epithelial cells through an interaction with cytokeratin 18 intron 1. Cell research. 2005;15(6):423–9. Epub 2005/07/01. pmid:15987600.
- 46. Virella-Lowell I, Herlihy JD, Liu B, Lopez C, Cruz P, Muller C, et al. Effects of CFTR, interleukin-10, and Pseudomonas aeruginosa on gene expression profiles in a CF bronchial epithelial cell Line. Molecular therapy: the journal of the American Society of Gene Therapy. 2004;10(3):562–73. Epub 2004/09/01. pmid:15336656.
- 47. Reiniger N, Ichikawa JK, Pier GB. Influence of cystic fibrosis transmembrane conductance regulator on gene expression in response to Pseudomonas aeruginosa infection of human bronchial epithelial cells. Infection and immunity. 2005;73(10):6822–30. Epub 2005/09/24. pmid:16177360; PubMed Central PMCID: PMCPmc1230967.