Individual Matrix Metalloproteinases Control Distinct Transcriptional ...

2 downloads 178 Views 2MB Size Report
Jun 11, 2007 - ... Critical Care Medicine, Fred Hutchinson Cancer Research Center, ..... represented GO biological processes (using a Benjamini-Hochberg ...
INFECTION AND IMMUNITY, Dec. 2007, p. 5640–5650 0019-9567/07/$08.00⫹0 doi:10.1128/IAI.00799-07 Copyright © 2007, American Society for Microbiology. All Rights Reserved.

Vol. 75, No. 12

Individual Matrix Metalloproteinases Control Distinct Transcriptional Responses in Airway Epithelial Cells Infected with Pseudomonas aeruginosa䌤† Sean Y. Kassim,1 Sina A. Gharib,1,4 Brigham H. Mecham,2 Timothy P. Birkland,1 William C. Parks,1,4 and John K. McGuire1,3* Center for Lung Biology,1 Department of Genome Sciences,2 and Department of Pediatrics,3 University of Washington, Seattle, Washington 98109, and Pulmonary and Critical Care Medicine, Fred Hutchinson Cancer Research Center, Seattle, Washington 981094 Received 11 June 2007/Returned for modification 12 September 2007/Accepted 27 September 2007

Airway epithelium is the initial point of host-pathogen interaction in Pseudomonas aeruginosa infection, an important pathogen in cystic fibrosis and nosocomial pneumonia. We used global gene expression analysis to determine airway epithelial transcriptional responses dependent on matrilysin (matrix metalloproteinase 7 [MMP-7]) and stromelysin-2 (MMP-10), two MMPs induced by acute P. aeruginosa pulmonary infection. Extraction of differential gene expression (EDGE) analysis of gene expression changes in P. aeruginosa-infected organotypic tracheal epithelial cell cultures from wild-type, Mmp7ⴚ/ⴚ, and Mmp10ⴚ/ⴚ mice identified 2,091 matrilysin-dependent and 1,628 stromelysin-2-dependent genes that were differentially expressed. Key node network analysis showed that these MMPs controlled distinct gene expression programs involved in proliferation, cell death, immune responses, and signal transduction, among other host defense processes. Our results demonstrate discrete roles for these MMPs in regulating epithelial responses to Pseudomonas infection and show that a global genomics strategy can be used to assess MMP function. Pseudomonas aeruginosa is a common cause of nosocomial lung infection in hospitalized and immunocompromised patients (52), and it is the most significant respiratory pathogen in patients with cystic fibrosis (21). Consequently, a great deal of attention has been focused on understanding P. aeruginosa virulence, pathogenesis of P. aeruginosa pneumonia, and the host response to this pathogen. However, the ubiquity of this microorganism and the rapidity with which it develops resistance to antibiotics and evolves in vivo continue to make it problematic to treat (54). Hence, a greater understanding of the mechanisms regulating host-pathogen interactions in P. aeruginosa pulmonary infection may identify new strategies for this difficult clinical problem. The airway epithelium is the first point of host contact for many respiratory pathogens, including bacteria, and several innate airway epithelial mechanisms participate in the defense against bacterial colonization and infection in the airways. An intact epithelium maintains a barrier to the environment, the airway mucus layer confers physical protection from microbes and particles, and the mucociliary elevator is an important mechanism of mechanical clearance of pathogens (12). At the apical epithelial surface, the airway surface liquid includes many defense factors that prevent establishment of infection such as lysozyme, ␤-defensins, cathelicidin, and others (6). Epithelial cells also control inflammation as a secondary line of

defense and produce factors that attract and activate phagocytes and other immune cells to mount a larger, multitiered attack on invading microorganisms. Among the molecules produced by epithelial cells in response to infection are matrix metalloproteinases (MMPs). The MMPs are a family of zinc-containing enzymes with proteolytic activity against a wide range of extracellular proteins (13). MMPs are expressed in a variety of normal and disease processes, such as development, involution, repair, inflammation, and tumor growth. Although MMPs have historically been thought to mediate remodeling or destruction of structural components, studies with genetically modified mice have demonstrated predominant roles in controlling the activity of effector proteins, particularly those that function in immune processes (46). Thus, MMPs are viewed as key extracellular processing enzymes that regulate cell responses and signaling (19, 39). MMPs have been proposed to both protect against and contribute to pathology in infectious disease (20). For example, matrilysin (MMP-7), unlike many MMPs, is expressed in mucosal epithelium in most adult human and mouse tissues (50, 58). In the lung, matrilysin is constitutively expressed in tracheal glands and at low levels in tracheo-bronchial epithelium, and its expression is markedly increased in airway epithelium by injury (18). Additionally, a marked increase in matrilysin expression and secretion is an early epithelial marker of gramnegative bacterial infection, including infection with P. aeruginosa (34, 35). These observations, along with its reported roles in facilitating airway reepithelialization (18), processing antibacterial peptides (59), and regulating transepithelial migration of neutrophils in acute lung injury (33) position matrilysin, and perhaps other MMPs, as a key regulator of epithelial responses to early P. aeruginosa infection in the lung.

* Corresponding author. Mailing address: Center for Lung Biology, University of Washington, 815 Mercer Street, Seattle, WA 98109. Phone: (206) 897-1304. Fax: (206) 897-1546. E-mail: mcguirej@u .washington.edu. † Supplemental material for this article may be found at http://iai .asm.org/. 䌤 Published ahead of print on 8 October 2007. 5640

VOL. 75, 2007

MMPs DICTATE EPITHELIAL TRANSCRIPTIONAL RESPONSE

Because matrilysin expression is induced by bacterial exposure and because several of the known and putative matrilysin substrates participate in signaling pathways modulate gene expression (heparin-binding epidermal growth factor, syndecan-1, E-cadherin, and insulin-like growth factor binding protein) (33, 40, 43, 61), we hypothesized that it controls distinct host cell responses to P. aeruginosa infection. Similarly, we report here that stromelysin-2 (MMP-10) is also rapidly induced by epithelial cells following P. aeruginosa exposure, yet mice with targeted deletion of matrilysin or stromelysin-2 have distinct inflammatory phenotypes in response to infection. To assess how these MMPs control host cell responses to infection, we used global oligonucleotide-based microarray expression analysis of P. aeruginosa-infected tracheal epithelial cells from wild-type and MMP-null mice grown as differentiated air-liquid interface (ALI) cultures. Our analyses demonstrate that matrilysin and stromelysin-2 influence distinct changes in gene expression that govern epithelial responses to acute P. aeruginosa infection. MATERIALS AND METHODS MMP-null mice. We designed a neomycin-containing construct targeting exons 3 to 5, which include the catalytic domain. Embryonic stem (ES) (129SvJ) transfections and selections were done at the Siteman Cancer Center ES Core, and the blastocyst injections were done by the Pulmonary Transgenic Mouse Core, both at Washington University in St. Louis, MO. ES clones positive for homologous recombination were injected into C57BL/6 blastocysts, and chimeric mice were bred to generate germ line heterozygotes, which were then bred to yield homozygous null mice (Mmp10⫺/⫺) and crossbred for 10 generations with C57BL/6 mice. Mmp7⫺/⫺ mice were originally obtained from L. Matrisian, Vanderbilt University, Nashville, TN. Wild-type C57BL/6 mice were bred as a congenic strain from MMP-deficient heterozygote mice and litter matched to MMP-deficient mice. In vivo infection model. The University of Washington Institutional Animal Care and Use Committee approved all animal use procedures. At 8 to 16 weeks of age, male mice were infected with live P. aeruginosa strain K by nebulization or by direct nasal inoculation with P. aeruginosa strain PA51673, a motile, nonmucoid, flagellated cystic fibrosis patient clinical isolate (34, 35). Bacteria were grown under standard laboratory conditions as overnight cultures in standard tryptic soy broth, centrifuged, washed in phosphate-buffered saline (PBS), and resuspended in PBS to an optical density at 600 nm of 0.2 and then diluted in PBS to working concentrations. For nebulization, mice were exposed to bacteria in a whole-animal chamber for 30 min with 107 CFU live bacteria delivered by aerosolization as described previously (24). For nasal inoculation, mice were anesthetized, and 25 ␮l of live bacterial suspension (4 ⫻ 108 CFU) was placed over the nares. At 4 or 24 h after infection, mice were sacrificed and lungs were processed for histology as described previously (33). Experiments were performed three times with a total of 8 to 10 mice per genotype at each time point. Mouse tracheal ALI cell culture. Primary mouse airway epithelial cells were cultured at an ALI as described previously (60). Culture medium supplements were from Sigma-Aldrich (St. Louis, MO) unless indicated. Mice were killed by CO2 euthanasia, and tracheas were sterilely resected from the larynx to the mainstem bronchi, cleaned of excess muscle and connective tissues, opened longitudinally, and incubated in Ham’s F-12 plus 100 U/ml penicillin and 100 ␮g/ml streptomycin (P/S) and 1.5 mg/ml Pronase (Roche Molecular Biochemicals, Indianapolis, IN) overnight at 4°C. The next day, tubes were inverted 10 to 12 times to dislodge cells from basement membranes. This process was repeated, and the contents of the three tubes were pooled and collected by centrifugation at 400 ⫻ g for 10 min at 4°C. Cells were resuspended in 200 ␮l/trachea of Ham’s F-12 plus P/S containing 0.5 mg/ml crude pancreatic DNase I (Sigma-Aldrich) and 10 mg/ml bovine serum albumin. The cells were then incubated on ice for 5 min, centrifuged at 400 g for 5 min at 4°C, and resuspended in MTEC basic media (1:1 [vol/vol] Dulbecco’s minimal essential medium–Ham’s F-12), 15 mM HEPES, 3.6 mM sodium bicarbonate, 4 mM L-glutamine, P/S, and 0.25 ␮g/ml amphotericin B [Fungizone]) with 10% fetal bovine serum. Twelve-well 1.1-cm2 Transwell clear polyester membrane supports (Corning-Costar, Corning, NY) were coated with filter-sterilized type I rat tail collagen (Becton-Dickinson) at 50

5641

␮g/ml in 0.02 N acetic acid using 1.0 ml/cm2 membrane overnight at 25°C. Membranes were seeded with 105 cells/cm2 in MTEC basic medium supplemented with 10 ␮g/ml insulin, 5 ␮g/ml transferrin, 0.1 ␮g/ml cholera toxin, 25 ng/ml epidermal growth factor (Becton-Dickinson, Bedford, MA), 26 ␮g/ml bovine pituitary extract (Cambrex, Charles City, IA), 5% fetal bovine serum, and 0.01 ␮M retinoic acid filling the upper and lower chambers and incubated in 5% CO2 at 37°C. Media were changed every 2 to 3 days until the transmembrane resistance was ⱖ1,000 m⍀/cm2, as measured by an epithelial Ohm-voltmeter (World Precision Instruments, Sarasota, FL). Media were then removed from the upper chamber to establish an ALI and were fed with MTEC basic medium supplemented with 2% NuSerum and 0.01 ␮M retinoic acid every 2 to 3 days added to the lower chamber only. Cells were infected with PA51673 at a multiplicity of infection (MOI), or ratio of bacteria to epithelial cells, of 1, 5, or 50 for 1 to 24 h at 14 to 21 days after initial plating (7 to 14 days after conversion to ALI). RNA isolation and microarray hybridization and processing. RNA was isolated from mouse lung tissues and ALI cultures with QIAGEN RNeasy mini kits. RNA quality was verified using the Agilent Bioanalyzer 2100, and all samples provided RNA integrity numbers ⱖ7. All subsequent RNA preparation and microarray chip hybridization and scanning procedures were performed by the Center for Array Technologies at the University of Washington (Seattle, WA) according to standard Affymetrix protocols. Briefly, the RNA samples were adjusted to a concentration of 2 ␮g/␮l and biotinylated. The samples were hybridized to GeneChip Mouse Genome Affymetrix 430 2.0 arrays, which contain 45,101 probe sets representing 20,973 unique genes. After hybridization overnight, chips were washed in the Affymetrix fluidics station, treated with phycoerythyrin-conjugated streptavidin, and scanned for hybridized targets with the GeneChip Scanner 3000. qRT-PCR. For quantitative real-time reverse transcription-PCR (qRT-PCR), 250 ng total RNA was reverse transcribed with a high-capacity cDNA archive kit (ABI-Applied Biosystems, Foster City, CA). Unlabeled PCR primers and TaqMan 6-carboxyfluorescein dye-labeled probes for target genes (Mmp7, Mmp10, Icam1, IL-6, Cxcl1, Rbm5, and Gapdh) were obtained from ABI. PCRs were performed according to the manufacturer’s protocol with an ABI HT7900 fast real-time PCR system using TaqMan reagents for 40 cycles. The threshold cycles (CT) for each probe set/cDNA were obtained from triplicate samples and averaged, and the ⌬CT was the calculated difference between the average CT for the target gene and the average CT for Gapdh as a control for total starting RNA quantity. The ⌬⌬CT was calculated from average ⌬CT at a given time point ⫺ the average ⌬CT of time zero samples. The data are expressed as relative quantification or change (fold), which is calculated as 2⫺⌬⌬CT and indicates the difference (fold) in gene expression relative to time zero. For time zero samples of Mmp7⫺/⫺ and Mmp10⫺/⫺ ALI cells, the difference (fold) in expression is as compared to wild-type expression at time zero. Data quality control and normalization. Data from the array were inspected visually and with the Affymetrix quality control metrics in the Gene Chip operating software. Two chips were removed (one for the wild type at 24 h and one for Mmp10⫺/⫺ at 24 h) from subsequent analysis due to high background and high scale factor, respectively. CEL files from the remaining 25 samples were extracted and normalized in Bioconductor in the R environment. To minimize contributions of batch effects, all samples were processed using the robust multiarray average (RMA) and mean centered by genotype using the RMA function of the Affymetrix package (11, 27) and the scale function of the base R environment (9). PCA and EDGE. Principal components analysis (PCA) was performed in R using the prcomp function of the stats package (9, 38, 56) on probe sets that changed ⱖ2-fold across all arrays. The normalized, filtered data were then analyzed for significant changes by the extraction of differential gene expression (EDGE) module running on top of the R environment. The time course methodology was applied to our data sets to identify differentially expressed genes of wild-type versus Mmp7⫺/⫺ and wild-type versus Mmp10⫺/⫺ comparisons, using length of exposure to P. aeruginosa as the time course covariate (32, 55). Biological process and pathway analysis. The Ingenuity Pathway Analysis package was used to map network interactions among significantly changed genes, and genes with 10 or more network interactions were defined as key nodes. Using the Database for Annotation, Visualization and Integrated Discovery (DAVID) (http://david.abcc.ncifcrf.gov/), the key node genes were analyzed for increased representation in the Gene Ontology (GO) Biological Processes database (17). Briefly, genes were uploaded using their ENTREZ gene identifications (IDs) and analyzed using DAVID’s Functional Annotation. Overrepresented GO biological processes (using a Benjamini-Hochberg score of ⬍0.001 to correct for multiple comparisons) that suggested functional differences were selected, and all of the genes in those processes were examined in the

5642

KASSIM ET AL.

INFECT. IMMUN.

complete MMP-dependent data sets. A gene-gene interaction network was created from the temporal transcriptional profiling experiments using Ingenuity Systems’ software and database (14). This database has been manually curated from ⬎200,000 full-text, peer-reviewed scientific articles and consists of published protein relationships from ⬃10,000 mammalian gene orthologues. An algorithm utilizing the connectivity of the differentially expressed genes served as the basis for creating gene-gene interaction networks, and a visual depiction of these network interactions was drawn based on the connectivity matrix of its members using Pajek’s visualization software (7). Microarray data accession number. Minimal Information About Microarray Experiments (MIAME)-compliant microarray data have been submitted to the NIH-GEO (accession no. GSE7957).

RESULTS P. aeruginosa infection induces expression of matrilysin and stromelysin-2 in vivo and in organotypic airway epithelium cultures. C57BL/6 wild-type mice and litter-matched mice with targeted deletion of Mmp7 or Mmp10 were infected with P. aeruginosa by nebulization or direct nasal inoculation. Mmp7⫺/⫺ mice have been characterized in previous studies (57), but we recently generated stromelysin-2-null (Mmp10⫺/⫺) mice. Similar to Mmp7⫺/⫺ mice, Mmp10⫺/⫺ mice are healthy, with no overt defects in fertility, litter size, gross appearance, organ structure, or tissue histology. In fact, we breed the homozygote null mice of both genotypes. In addition, the airways of Mmp7⫺/⫺ and Mmp10⫺/⫺ mice are morphologically normal with the expected proportion of ciliated and Clara cells. The lack of a phenotype in the unchallenged Mmp10⫺/⫺ mice is not unexpected. Stromelysin-2 is not expressed in developing tissues, and other than in the small intestine and some osteoclasts, it is also not expressed in mature organs (2, 37). Both matrilysin and stromelysin-2 mRNAs were expressed at low levels in normal mouse lungs (Fig. 1). However, after P. aeruginosa infection, matrilysin expression in whole lung increased 5- and 10-fold at 4 and 24 h postinfection in wild-type mice. Similarly, stromelysin-2 expression increased fourfold above baseline at 4 h postinfection and remained elevated at 24 h (Fig. 1). The expression of matrilysin in Mmp10⫺/⫺ mice and of stromelysin-2 in Mmp7⫺/⫺ mice did not differ from the levels measured in wild-type lung (data not shown). Thus, induction of matrilysin and stromelysin-2 is part of the early epithelial response to acute P. aeruginosa infection. Although both matrilysin and stromelysin-2 are induced by P. aeruginosa infection, a comparison of tissue histology revealed distinct inflammatory phenotypes in null mice (Fig. 2). Mmp7⫺/⫺ mice had less severe pneumonia at 4 h compared to wild-type mice, but by 24 h postinfection, the differences between genotypes were less pronounced. These findings indicate that the absence of matrilysin delays the host response to infection. In contrast, at 4 h after infection, pulmonary inflammation, particularly around airways, was increased in Mmp10⫺/⫺ mice compared to wild-type mice, and at 24 h, Mmp10⫺/⫺ mice had much more severe pneumonia than did wild-type mice, with extensive hemorrhage and diffuse pulmonary consolidation (Fig. 2). Similar results were obtained with both infection protocols and bacterial strains. (A more thorough characterization of these inflammatory phenotypes is the focus of a separate report.) These findings indicate that matrilysin and stromelysin-2 regulate distinct host pulmonary responses in acute P. aeruginosa lung infection. Overall, matrily-

FIG. 1. MMP-7 and MMP-10 mRNA expression is induced in lung and ALI epithelium after Pseudomonas aeruginosa infection. Mice and mouse tracheal epithelial cells cultured at an ALI were infected with P. aeruginosa, and total RNA was isolated at 0 to 24 h after infection. qRT-PCR was performed for Mmp7 and Mmp10, and expression levels are shown as change (fold) compared to expression level at time zero. Lung data are pooled from three animals per time point and run in duplicate. ALI data are pooled from three experiments with two to four samples per time point run in duplicate. Values are means ⫾ standard error. ALI Mmp7 and Mmp10 changes are significant over the time course. †, P ⬍ 0.0001 by analysis of variance; *, P ⬍ 0.01 by Dunnett’s multiple-comparison posttest for individual time points as compared to t ⫽ 0 h.

sin functions to promote inflammation, whereas stromelysin-2 acts to restrain inflammation. P. aeruginosa first encounters the respiratory tract via the airway epithelium, and thus, we used a model of differentiated organotypic airway epithelial cell cultures grown at an ALI to study how matrilysin and stromelysin-2 function in regulating the initial epithelial responses to this pathogen. These cultures reflect the cellular makeup of airway epithelium in vivo, replete with fully differentiated ciliated and mucus-producing cells, and provide a more relevant model than standard monolayer epithelial cultures (16, 60). In addition, use of differentiated ALI cultures allows us to assess the intrinsic epithelial response to infection without the influence of other cell types. Tracheal epithelial cells from Mmp7⫺/⫺ and Mmp10⫺/⫺ mice established ALI cultures that were largely indistinguishable from wild-type cultures, with no differences in time to differentiation, cellular morphology, or transepithelial resistance (data not shown). To model the natural route of airway epithelial infection seen in vivo, cultures from wild-type mice were infected on the apical surface with P. aeruginosa at an MOI, or ratio of bacteria to epithelial cells, of 1, 5, or 50 for 1 to 24 h. The epithelial response to P. aeruginosa was gauged by tracking ICAM-1 expression (1), which was robustly increased with an MOI of 5 (see the supplemental material). Thus, this dose, which also did not produce the extensive cell death seen with an MOI of 50 (data not shown), was used for subsequent studies. As in

VOL. 75, 2007

MMPs DICTATE EPITHELIAL TRANSCRIPTIONAL RESPONSE

5643

FIG. 2. Differential response between Mmp7⫺/⫺ and Mmp10⫺/⫺ mice to bacterial infection. Mouse lung tissues were isolated at 4 and 24 h after infection with P. aeruginosa and formalin fixed, and sections were stained with hematoxylin and eosin. Bars, 200 ␮m. Samples are representative of 8 to 10 mice of each genotype at each time point.

whole lung, basal expression of matrilysin and stromelysin 2 was low in established ALI cultures and infection with P. aeruginosa induced the expression of both Mmp7 and Mmp10 as early as 1 h, and Mmp7 and Mmp10 mRNA levels peaked at 4 h and remained elevated above baseline levels at 24 h (Fig. 1). At 1 h after infection, transmembrane resistance was similar to that prior to infection; however, in response to persistent infection, resistance dropped ⬎80% by 24 h after infection (data not shown), an observation consistent with published reports (5, 10, 28, 62). Thus, gene expression changes measured at 1 h after infection reflect initial contact with the bacteria, while the epithelium is still intact, and later changes reflect progression of acute infection. Duration of bacterial exposure is the primary determinant of gene expression changes in P. aeruginosa-infected mouse airway epithelium. We used a global gene expression analysis approach to assess the consequences of matrilysin and stromelysin-2 activity in the airway epithelial responses to P. aeruginosa infection. A 1-h infection time point was used to identify gene expression changes induced by early epithelial contact

with bacteria, and a 24-h time point was selected to identify later changes. After infection, RNA was isolated, converted to cDNA, labeled, and hybridized to Affymetrix mouse 430 2.0 array chips. After scanning, data were normalized by RMA and mean centering. A global filter was then applied across the entire data set to identify probe sets that significantly changed at least twofold across all arrays. This cutoff identified 14,858 probe sets representing 9,593 genes. PCA was applied to the data set of significantly changing genes and indicated that the primary source of variability in the data was the length of exposure to bacteria. Three distinct clusters corresponding to baseline and 1 h and 24 h of exposure to P. aeruginosa were identified in the principal component space (Fig. 3). Principal component 1, accounting for 55% of gene expression variability, grouped samples collected at 24 h of exposure, implying that this time point captures the majority of the data variance. We did not observe a significant segregation of data points across genotypes. Because direct contact with bacteria is such a potent stimulus for epithelial cells, this observation indicates that the majority of the gene expression

5644

INFECT. IMMUN.

KASSIM ET AL.

FIG. 3. PCA of microarrays. PCA was performed in the R environment. Each microarray chip is represented individually by shape and color: circles, wild type (WT); triangles, Mmp7⫺/⫺; squares, Mmp10⫺/⫺; blue, 0 h; green, 1 h; red, 24 h.

changes were due to duration of bacterial exposure, independent of MMP genotype. Importantly, baseline differences among genotypes at time zero prior to infection had minimal contribution to gene expression variability compared to the changes induced by infection. We used qRT-PCR to validate gene expression changes seen on microarrays for selected genes. For example, Icam1 and Cxcl1 mRNA levels increased at 1 and 24 h after infection, and the changes in expression determined by microarray analysis matched closely the level of expression determined by qRTPCR (see the supplemental material). Similar validation was performed for several other genes, including Il6, Rbm5, and Bcl2, with generally a strong correlation between the microarray and qRT-PCR data. EDGE analysis shows distinct MMP-dependent gene expression changes. We used EDGE analysis within the Bioconductor package of the R environment to define gene expression changes dependent on either matrilysin or stromelysin-2 for the 9,593 genes with a ⬎2-fold change in expression. To identify significant gene expression changes dependent on individual MMPs, we made comparisons between wild-type and Mmp7⫺/⫺ arrays across all time points and performed an identical comparison between wild-type and Mmp10⫺/⫺ arrays. EDGE analysis provides advantages in revealing changes over time that may not be identified by a conventional two-way comparison (32, 55). Using a global false discovery rate of 0.05, EDGE identified 2,432 probe sets representing 2,091 genes with unique ENTREZ gene IDs as significantly changed in the Mmp7⫺/⫺ arrays compared to the wild type and 1,840 probe sets representing 1,628 genes in the Mmp10⫺/⫺ arrays compared to the wild type (Fig. 4). A global interpretation of the heat maps suggests that stromelysin-2 functions to activate changes in gene expression for large clusters of genes, particularly at 24 h after bacterial exposure. In the Mmp10⫺/⫺ cells, epithelial gene expression was largely unchanged across the time course for many genes (black in the heat maps). In contrast, in the matrilysin data set, the effect was large clusters of genes changing in similar directions but matrilysin-dependent differences were revealed by the degree of change (intensity in the heat maps). Comparing these lists, we identified 608 genes that were present in both MMP-dependent gene sets. Because

duration of bacterial exposure was the major determinant of changes in gene expression in this system, effects of individual MMP activities would be predicted to regulate some of the same genes, although not necessarily in the same direction or via the same mechanisms. Key node network analysis identifies distinct MMP-dependent regulation of biological processes in epithelial host response to P. aeruginosa. Using Ingenuity Pathway Analysis, we mapped the significantly changed genes in the matrilysin-dependent and stromelysin-2-dependent data sets to gene interaction networks based on known biological interactions. Significant genes with 10 or greater network connections were defined as key nodes and were selected for further analysis because they were the most likely to illustrate the biological processes and pathways regulated by the MMP-dependent gene expression changes (Fig. 5). There is increasing evidence that the functional robustness of biological networks is highly dependent on these densely connected hubs (36). The key node genes were then subjected to GO analysis to gain insights into which aspects of the epithelial response to P. aeruginosa were regulated by the individual MMPs. Both matrilysin- and stromelysin-2-regulated key nodes were enriched in genes that are predicted to be activated in epithelial responses to P. aeruginosa, with significant enrichment in GO biological process modules such as immune response, cell proliferation, cell death, and signal transduction, and the matrilysin-regulated key nodes were also significantly enriched in the DNA-dependent transcription and regulation of metabolism modules (Fig. 5). A comparison of the heat map for matrilysin-dependent key nodes in the cell proliferation module to that of all of the matrilysin-dependent genes in that module showed that the patterns of expression in the key node genes between wild-type and Mmp7⫺/⫺ cells are representative of the larger gene set for that module (Fig. 6). Similarly, the patterns of change for stromelysin-2-dependent key nodes in the immune response module reflected those of all stromelysin-2-dependent genes in that module. A comparison of the matrilysin- and stromelysin-2-dependent genes for each biological process module shows that although some of the highly networked genes that are regulated by these individual MMPs were the same, the majority are distinct. For example, the cell death module shows that most of the matrilysin- and stromelysin-2-dependent genes are uniquely present in either one or the other data set but not both (Fig. 7). Focusing on one key node present in both MMPdependent data sets in the cell death module, Bcl2, for example, shows that the significantly changing genes that this gene product may interact with were quite distinct between the two genotypes. Thus, it is highly likely that the gene expression differences in important epithelial host response processes such as apoptosis and immune response reflect the differences in acute pneumonia phenotypes observed between genotypes in the in vivo infection models. DISCUSSION In this study, we used an unbiased global microarray-based genomics approach to assess the function of matrilysin and stromelysin-2 in epithelial responses to acute P. aeruginosa infection. Collectively, our data show that global transcrip-

VOL. 75, 2007

MMPs DICTATE EPITHELIAL TRANSCRIPTIONAL RESPONSE

5645

FIG. 4. Matrilysin and stromelysin-2 control different transcriptional responses. The Venn diagram represents significantly changed genes identified by EDGE analysis (false discovery rate, ⬍0.05), from a twofold-change filtered data set, of wild-type (WT) versus specific MMP gene-deleted mice. Unique probe sets are presented in row-normalized heat maps. All numbers are totals of unique ENTREZ gene IDs represented by the Affymetrix probe sets.

tional responses dependent on either matrilysin or stromelysin-2 in the airway epithelium were largely distinct, suggesting discrete roles for these MMPs in airway epithelial cell regulation of immune response, cell proliferation, cell death, signal transduction, and other processes central to the initial host defense against invading bacteria. The relevance of these gene expression differences to in vivo host responses was supported by our observations of histological differences in the different genotypes in the first 24 h after P. aeruginosa infection. One advantage of this pathway-based approach to interrogating MMP function is that it can guide future studies directed at identifying specific substrates and targets for intervention. Several reports describe increased epithelial expression in a variety of MMPs in response to bacterial infection. For example, Porphyromonas gingivalis induces gelatinase A (MMP-2) and gelatinase B (MMP-9) in human gingival epithelial cells (3) and Helicobacter pylori induces expression of collagenase 1 (MMP-1) and matrilysin in gastric epithelium (8, 31). P. aeruginosa induces gelatinase B in corneal epithelium, and work

from our laboratory demonstrated that matrilysin is induced in airway epithelium and human lung epithelial cell lines by exposure to gram-negative bacteria, including P. aeruginosa (35). Matrilysin is also highly expressed in the lungs of patients with cystic fibrosis that are chronically infected with P. aeruginosa (18, 35). However, the function of matrilysin or other MMPs in bacterial infection in these models has not been previously defined. One problem that has confounded efforts to ascribe specific mechanisms of action to individual MMP proteolytic targets is that MMPs may have multiple potential substrates in the pericellular environment, and therefore, sorting out how an individual MMP substrate cleavage might contribute to epithelial host defense is difficult. Moreover, multiple MMPs can be expressed in the same tissues, and most single MMPdeficient mice lack developmental phenotypes, with the notable exception of the MMP-14 (MT1-MMP)-deficient mouse, which has severe musculoskeletal abnormalities (25). One interpretation of lack of developmental phenotypes among MMP-null mice is functional compensation by other MMPs

5646

KASSIM ET AL.

INFECT. IMMUN.

FIG. 5. Key nodes reveal differences in GO biological processes influenced by matrilysin and stromelysin-2. The results of Ingenuity Pathway Analysis of MMP-dependent data sets were filtered to identify genes that had 10 or more network interactions, and these interactions were drawn based on the connectivity matrix of its members using Pajek’s visualization software. Unique matrilysin key nodes are cyan, unique stromelysin-2 key nodes are magenta, and key nodes in both data sets are yellow. The table indicates the number of genes in each biological process from key node and all significant gene data sets identified by DAVID analysis as significantly overrepresented GO biological processes (Benjamini corrected P value of ⬍0.001). NS, not significant.

VOL. 75, 2007

MMPs DICTATE EPITHELIAL TRANSCRIPTIONAL RESPONSE

5647

FIG. 6. Key nodes reflect changes in entire MMP-dependent data sets. Shown are heat maps for key nodes and all significantly changing genes in the matrilysin-dependent cell proliferation module and stromelysin-2-dependent immune response GO biological process modules.

(49). However, the revelation of discrete phenotypes in MMPdeficient mice upon experimental challenge indicates that in specific circumstances, individual MMPs have unique functions that are not compensated for by different MMPs or other proteinases. Furthermore, many of the MMPs thus far knocked out in mice are not widely, if at all, expressed in developing tissues (44). We hypothesized that by understanding the global gene expression changes of MMP activity that are downstream from proteolytic events, we could define potential MMP functions independent of characterizing specific substrates. Because the biological processes these MMPs regulate may vary over time in our infection model, we chose to analyze the data for gene expression changes in wild-type versus MMP-deficient mice across a time course encompassing both early and late responses to infection. EDGE analysis allowed us to perform an initial significance testing across the entire 25-chip database, dramatically strengthening the statistical power over a static analysis design that would include only a few chips in each comparison. Thus, with this approach, we markedly lessen the probability of missing significant gene changes that may not be identified in strict pairwise comparison between samples at individual times. Another strength of our strategy is that it accounts for the dramatic time-dependent changes in gene expression seen in wild-type cells over the time course that could otherwise overwhelm genotype-dependent differences. Indeed, nearly 80% of the significantly changing genes identified by EDGE were not dependent on either MMP but rather reflected the potent effects of P. aeruginosa infection in this system, a point illustrated by the PCA. Furthermore, we used the key node network analysis to assess the functional consequences of the induction of matrilysin or stromelysin-2 expression on important biological processes, thus avoiding biases introduced by focusing on which genes increase or decrease the most. Performing gene ontology analysis based on number of genes in a given process assigns equal weight to every significantly chang-

ing gene present in the data set, whereas the key node approach focuses the analysis on the most networked genes (key nodes), about which more is known, and thus emphasizes the genes most likely to be informative in ascribing biological functions. That our data revealed matrilysin to be involved in immune responses, cell proliferation, cell death, and signal transduction responses to P. aeruginosa airway infection is consistent with known functions of matrilysin reported for other systems. For example, matrilysin regulates acute alveolar neutrophil influx by shedding syndecan-1/KC complexes (33), facilitates airway reepithelialization by cleaving E-cadherin (40), promotes apoptosis by generating soluble Fas ligand (47), and regulates growth factor signaling by epidermal growth factor and insulinlike growth factor (43, 61). The high representation of matrilysin-dependent key node genes in signal transduction, cell proliferation, DNA-dependent transcription, and the regulation of metabolism modules points to growth factors and their receptors and adhesion molecules as potential matrilysin substrates for controlling these responses. Thus, a correlation between predicted consequences of the processing of known matrilysin substrates and gene expression changes in the biological modules that these candidate substrates may regulate supports the use of a global genomics approach to interrogate MMP function. Indeed, we are now assessing whether these proteins are matrilysin targets in airway P. aeruginosa infection. In contrast to matrilysin, little is known as to how stromelysin-2 functions in infection or injury. Several reports described marked expression of stromelysin-2 in a variety of carcinomas, in wounded epidermis, and in atherosclerosis (4, 22, 37, 41, 45, 48, 51), indicating that this MMP is a component of the tissue response to challenge. In vitro, stromelysin-2 can degrade several extracellular matrix proteins and can activate other MMPs (29, 42, 53); however, data on an in vivo role are scant. Overexpression of stromelysin-2 by keratinocytes potentially affects cell-matrix interactions, and other studies have suggested a role in vascular homeostasis (41, 53). However, evidence for

5648

KASSIM ET AL.

INFECT. IMMUN.

FIG. 7. Matrilysin and stromelysin-2 differentially control apoptotic transcriptional response. Shown are the results of Ingenuity Pathway Analysis of significantly changed genes from the MMP-7 and MMP-10 data sets in the apoptosis pathway. Unique matrilysin-dependent genes are cyan, unique stromelysin-2-dependent genes are magenta, and genes present in both data sets are yellow. Gray lines indicate network connections with the connections for the Bcl2 key node indicated in black to demonstrate distinct Bcl2 interactions between data sets. EC, extracellular; MEM, membrane; CYT, cytoplasmic; NUC, nuclear.

specific function of stromelysin-2 in vivo is lacking. Our observations of increased inflammation and pulmonary consolidation in infected Mmp10⫺/⫺ lung provide evidence for a role in restraining the inflammatory response. These data—in conjunction with the striking lack of transcriptional change in many of the stromelysin-2-dependent genes in the Mmp10⫺/⫺ cells—suggest that stromelysin-2 may be essential in activating protective host responses initiated by the epithelium early after P. aeruginosa infection. Apoptosis is one process that the microarray data indicate may be differentially regulated by both matrilysin and stromelysin-2 in airway epithelium responses to P. aeruginosa, which is interesting as the potential role of apoptosis in P. aeruginosa in

airway epithelial infection is controversial. It is been proposed that P. aeruginosa-induced apoptosis in airway epithelium is a mechanism of pathogen clearance (15, 23) and that this is a CD95 (Fas)-dependent process (23). However, the general applicability of these findings has been questioned (26). Our data show marked differences in the expression of genes related to apoptosis in both MMP-dependent data sets, including CD95/Fas in the stromelysin-2-dependent data set. The observations of increased keratinocyte apoptosis in skin wounds of mice overexpressing a constitutively active mutant stromelysin-2 under the control of a keratinocyte-specific promoter (30) and the aforementioned report of matrilysin generation of soluble Fas ligand support further investigation of the poten-

VOL. 75, 2007

MMPs DICTATE EPITHELIAL TRANSCRIPTIONAL RESPONSE

tial roles for matrilysin and stromelysin-2 in controlling apoptosis in our in vivo and ALI P. aeruginosa infection models. Overall, our data suggest that a global microarray approach can be used to assess the consequences of MMP function in a defined biological system. We have identified candidate biological processes regulated by two different MMPs both expressed in a similar time course and in the same cell type after acute P. aeruginosa infection. Whereas our long-term goal is to characterize the substrates that matrilysin and stromelysin cleave to initiate these gene expression changes, our genomics analysis identified activated pathways that can be investigated independent of knowing these substrates. Since it will be important to make phenotypic correlations between gene expression data and their in vivo physiologic and pathological consequences, our approach provides a framework to define the in vivo phenotypes in the mouse pneumonia model by focusing on specific processes such as immune response and apoptosis. ACKNOWLEDGMENTS We thank Shawn Skerrett, University of Washington, for assistance with bacterial nebulization; Robert Ernst, E. Peter Greenberg, and Pradeep Singh, University of Washington, for helpful discussions regarding bacterial strains; Ron McCarthy and Dale Kobayashi, Washington University, for help with generating Mmp10⫺/⫺ mice; Teresa Tolley and Darlene Stuart, Morphology Core, Washington University, for excellent histology services; Thomas Mariani, Harvard University, for helpful discussions on microarray analysis; Benjamin Jacobson for help with ALI cultures; and Ying Wang for maintaining our mouse colony. This work was supported by grants HL068780, HL029594, HL077555, and HL074223 from the National Institutes of Health and by the University of Washington NHLBI Microarray Center (HL072730) headed by Dan Sabath and Roger Bumgarner. S.Y.K. is supported by training grant T32-HL007312. REFERENCES 1. Aldallal, N., E. E. McNaughton, L. J. Manzel, A. M. Richards, J. Zabner, T. W. Ferkol, and D. C. Look. 2002. Inflammatory response in airway epithelial cells isolated from patients with cystic fibrosis. Am. J. Respir. Crit. Care Med. 166:1248–1256. 2. Andersen, T. L., M. del Carmen Ovejero, T. Kirkegaard, T. Lenhard, N. T. Foged, and J. M. Delaisse. 2004. A scrutiny of matrix metalloproteinases in osteoclasts: evidence for heterogeneity and for the presence of MMPs synthesized by other cells. Bone 35:1107–1119. 3. Andrian, E., Y. Mostoeaoui, R. Mahmoud, and D. Grenier. 2007. Regulation of matrix metalloproteinases and tissue inhibitors of matrix metalloproteinases by Porphyromonas gingivalis in an engineered human oral mucosa model. J. Cell. Physiol. 211:56–62. 4. Aung, P. P., N. Oue, Y. Mitani, H. Nakayama, K. Yoshida, T. Noguchi, A. K. Bosserhoff, and W. Yasui. 2006. Systematic search for gastric cancer-specific genes based on SAGE data: melanoma inhibitory activity and matrix metalloproteinase-10 are novel prognostic factors in patients with gastric cancer. Oncogene 25:2546–2557. 5. Azghani, A. O. 1996. Pseudomonas aeruginosa and epithelial permeability: role of virulence factors elastase and exotoxin A. Am. J. Respir. Cell Mol. Biol. 15:132–140. 6. Bals, R., and P. S. Hiemstra. 2004. Innate immunity in the lung: how epithelial cells fight against respiratory pathogens. Eur. Respir. J. 23:327– 333. 7. Batagelj, V., and A. Mrvr. 2004. Pajek analysis and visualization of large networks. Springer, New York. 8. Bebb, J. R., D. P. Letley, R. J. Thomas, F. Aviles, H. M. Collins, S. A. Watson, N. M. Hand, A. Zaitoun, and J. C. Atherton. 2003. Helicobacter pylori upregulates matrilysin (MMP-7) in epithelial cells in vivo and in vitro in a Cag dependent manner. Gut 52:1408–1413. 9. Becker, R. A., J. M. Chambers, and A. R. Wilks. 1988. The new S language: a programming environment for data analysis and graphics. Wadsworth & Brooks/Cole Advanced Books & Software, Pacific Grove, CA. 10. Beisswenger, C., C. B. Coyne, M. Shchepetov, and J. N. Weiser. 2007. Role of p38 MAP kinase and TGF-␤ signaling in transepithelial migration of invasive bacterial pathogens. J. Biol. Chem. 282:28700–28708. 11. Bolstad, B. M., R. A. Irizarry, M. Astrand, and T. P. Speed. 2003. A com-

12. 13. 14.

15.

16.

17.

18.

19. 20.

21.

22.

23.

24.

25.

26.

27.

28.

29. 30.

31.

32.

33.

34.

35.

36.

5649

parison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19:185–193. Boucher, R. C. 2004. New concepts of the pathogenesis of cystic fibrosis lung disease. Eur. Respir. J. 23:146–158. Brinckerhoff, C. E., and L. M. Matrisian. 2002. Matrix metalloproteinases: a tail of a frog that became a prince. Nat. Rev. Mol. Cell Biol. 3:207–214. Calvano, S. E., W. Xiao, D. R. Richards, R. M. Felciano, H. V. Baker, R. J. Cho, R. O. Chen, B. H. Brownstein, J. P. Cobb, S. K. Tschoeke, C. MillerGraziano, L. L. Moldawer, M. N. Mindrinos, R. W. Davis, R. G. Tompkins, and S. F. Lowry. 2005. A network-based analysis of systemic inflammation in humans. Nature 437:1032–1037. Cannon, C. L., M. P. Kowalski, K. S. Stopak, and G. B. Pier. 2003. Pseudomonas aeruginosa-induced apoptosis is defective in respiratory epithelial cells expressing mutant cystic fibrosis transmembrane conductance regulator. Am. J. Respir. Cell Mol. Biol. 29:188–197. Davidson, D., F. M. Kilanowski, S. H. Randell, D. N. Sheppard, and J. R. Dorin. 2000. A primary culture model of differentiated murine tracheal epithelium. Am. J. Physiol. Lung Cell. Mol. Physiol. 279:L766–L778. Dennis, G., Jr., B. T. Sherman, D. A. Hosack, J. Yang, W. Gao, H. C. Lane, and R. A. Lempicki. 2003. DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol. 4:P3. Dunsmore, S. E., U. K. Saarialho-Kere, J. D. Roby, C. L. Wilson, L. M. Matrisian, H. G. Welgus, and W. C. Parks. 1998. Matrilysin expression and function in airway epithelium. J. Clin. Investig. 102:1321–1331. Egeblad, M., and Z. Werb. 2002. New functions for the matrix metalloproteinases in cancer progression. Nat Rev. Cancer 2:161–174. Elkington, P. T., C. M. O’Kane, and J. S. Friedland. 2005. The paradox of matrix metalloproteinases in infectious disease. Clin. Exp. Immunol. 142: 12–20. Gibson, R. L., J. L. Burns, and B. W. Ramsey. 2003. Pathophysiology and management of pulmonary infections in cystic fibrosis. Am. J. Respir. Crit. Care Med. 168:918–951. Gill, J. H., I. G. Kirwan, J. M. Seargent, S. W. Martin, S. Tijani, V. A. Anikin, A. J. Mearns, M. C. Bibby, A. Anthoney, and P. M. Loadman. 2004. MMP-10 is overexpressed, proteolytically active, and a potential target for therapeutic intervention in human lung carcinomas. Neoplasia 6:777–785. Grassme, H., S. Kirschnek, J. Reithmuller, A. Riehle, G. von Kurthy, F. Lang, M. Wellwe, and E. Gulbins. 2000. CD95/CD95 ligand interactions on epithelial cells in host defense to Pseudomonas aeruginosa. Science 290:527– 530. Hawn, T. R., K. D. Smith, A. Aderem, and S. J. Skerrett. 2006. Myeloid differentiation primary response gene (88)- and toll-like receptor 2-deficient mice are susceptible to infection with aerosolized Legionella pneumophila. J. Infect. Dis. 193:1693–1702. Holmbeck, K., P. Bianco, J. Caterina, S. Yamada, M. Kromer, S. A. Kuznetsov, M. Mankani, P. G. Robey, A. R. Poole, I. Pidoux, J. M. Ward, and H. Birkedal-Hansen. 1999. MT1-MMP-deficient mice develop dwarfism, osteopenia, arthritis, and connective tissue disease due to inadequate collagen turnover. Cell 99:81–92. Hotchkiss, R. S., W. M. Dunne, P. E. Swanson, C. G. Davis, K. W. Tinsley, K. C. Change, T. G. Buchman, and I. E. Karl. 2001. Role of apoptosis in Pseudomonas aeruginosa pneumonia. Science 294:1783. Irizarry, R. A., B. Hobbs, F. Collin, Y. D. Beazer-Barclay, K. J. Antonellis, U. Scherf, and T. P. Speed. 2003. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4:249–264. Kim, J. Y., U. S. Sajjan, G. P. Krasan, and J. J. LiPuma. 2005. Disruption of tight junctions during traversal of the respiratory epithelium by Burkholderia cenocepacia. Infect. Immun. 73:7107–7112. Knauper, V., G. Murphy, and H. Tschesche. 1996. Activation of human neutrophil procollagenase by stromelysin 2. Eur. J. Biochem. 235:187–191. Krampert, M., W. Bloch, T. Sasaki, P. Bugnon, T. Rulicke, E. Wolf, M. Aumailley, W. C. Parks, and S. Werner. 2004. Activities of the matrix metalloproteinase stromelysin-2 (MMP-10) in matrix degradation and keratinocyte organization in wounded skin. Mol. Biol. Cell 15:5242–5254. Krueger, S., T. Hundertmark, T. Kalinski, U. Peitz, T. Wex, P. Malfertheiner, M. Naumann, and A. Roessner. 2006. Helicobacter pylori encoding the pathogenicity island activates matrix metalloproteinase 1 in gastric epithelial cells via JNK and ERK. J. Biol. Chem. 281:2868–2875. Leek, J. T., E. Monsen, A. R. Dabney, and J. D. Storey. 2006. EDGE: extraction and analysis of differential gene expression. Bioinformatics 22: 507–508. Li, Q. L., P. Y. Park, C. L. Wilson, and W. C. Parks. 2002. Matrilysin shedding of syndecan-1 regulates chemokine mobilization and transepithelial efflux of neutrophils in acute lung injury. Cell 111:635–646. Lopez-Boado, Y. S., C. L. Wilson, L. V. Hooper, J. I. Gordon, S. J. Hultgren, and W. C. Parks. 2000. Bacterial exposure induces and activates matrilysin in mucosal epithelial cells. J. Cell Biol. 148:1305–1315. Lopez-Boado, Y. S., C. L. Wilson, and W. C. Parks. 2001. Regulation of matrilysin expression in airway epithelial cells by Pseudomonas aeruginosa flagellin. J. Biol. Chem. 276:41417–41423. Luscombe, N. M., M. M. Babu, H. Yu, M. Snyder, S. A. Teichmann, and M.

5650

37.

38. 39. 40. 41.

42. 43.

44. 45.

46. 47.

48. 49.

KASSIM ET AL.

Gerstein. 2004. Genomic analysis of regulatory network dynamics reveals large topological changes. Nature 431:308–312. Madlener, M., C. Mauch, W. Conca, M. Brauchle, W. C. Parks, and S. Werner. 1996. Regulation of the expression of stromelysin-2 by growth factors in keratinocytes: implications for normal and impaired wound healing. Biochem. J. 320:659–664. Mardia, K. V., J. T. Kent, and J. M. Bibby. 1979. Multivariate analysis. Academic Press, London, United Kingdom. McCawley, L. J., and L. M. Matrisian. 2001. Matrix metalloproteinases: they’re not just for matrix anymore! Curr. Opin. Cell Biol. 13:534–540. McGuire, J. K., Q. L. Li, and W. C. Parks. 2003. Matrilysin mediates Ecadherin ectodomain shedding in injured lung epithelium. Am. J. Pathol. 162:1831–1843. Montero, I., J. Orbe, N. Varo, O. Beloqui, J. I. Monreal, J. A. Rodriguez, J. Diez, P. Libby, and J. A. Paramo. 2006. C-reactive protein induces matrix metalloproteinase-1 and -10 in human endothelial cells: implications for clinical and subclinical atherosclerosis. J. Am. Coll. Cardiol. 47:1369–1378. Nakamura, H., Y. Fujii, E. Ohuchi, E. Yamamoto, and Y. Okada. 1998. Activation of the precursor of human stromelysin 2 and its interactions with other matrix metalloproteinases. Eur. J. Biochem. 253:67–75. Nakamura, M., S. Miyamoto, H. Maeda, G. Ishii, T. Hasebe, T. Chiba, M. Asaka, and A. Ochiai. 2005. Matrix metalloproteinase-7 degrades all insulinlike growth factor binding proteins and facilitates insulin-like growth factor bioavailability. Biochem. Biophys. Res. Commun. 333:1011–1016. Nuttall, R. K., C. L. Sampieri, C. J. Pennington, S. E. Gill, G. A. Schultz, and D. R. Edwards. 2004. Expression analysis of the entire MMP and TIMP gene families during mouse tissue development. FEBS Lett. 563:129–134. Orbe, J., I. Montero, J. A. Rodriguez, O. Beloqui, C. Roncal, and J. A. Paramo. 2007. Independent association of matrix metalloproteinase-10, cardiovascular risk factors and subclinical atherosclerosis. J. Thromb. Haemostasis 5:91–97. Parks, W. C., C. L. Wilson, and Y. S. Lopez-Boado. 2004. Matrix metalloproteinases as modulators of inflammation and innate immunity. Nat. Rev. Immunol. 4:617–629. Powell, W. C., B. Fingleton, C. L. Wilson, M. Boothby, and L. M. Matrisian. 1999. The metalloproteinase matrilysin proteolytically generates active soluble Fas ligand and potentiates epithelial cell apoptosis. Curr. Biol. 9:1441– 1447. Ramos, M. C., H. Steinbrenner, D. Stuhlmann, H. Sies, and P. Brenneisen. 2004. Induction of MMP-10 and MMP-1 in a squamous cell carcinoma cell line by ultraviolet radiation. Biol. Chem. 385:75–86. Rudolph-Owen, L. A., D. L. Hulboy, C. L. Wilson, J. S. Mudgett, and L. M. Matrisian. 1997. Coordinate expression of matrix metalloproteinase family members in the uterus of normal, matrilysin-deficient, and stromelysin-1 deficient mice. Endocrinology 138:4902–4911.

Editor: A. Camilli

INFECT. IMMUN. 50. Saarialho-Kere, U. K., E. C. Crouch, and W. C. Parks. 1995. Matrix metalloproteinase matrilysin is preferentially expressed in human exocrine epithelium. J. Investig. Dermatol. 105:190–196. 51. Saarialho-Kere, U. K., S. O. Kovacs, A. P. Pentland, W. C. Parks, and H. G. Welgus. 1994. Distinct populations of keratinocytes express stromelysin-1 and -2 in chronic wounds. J. Clin. Investig. 94:79–88. 52. Sadikot, R. T., T. S. Blackwell, J. W. Christman, and A. S. Prince. 2005. Pathogen-host interactions in Pseudomonas aeruginosa pneumonia. Am. J. Respir. Crit. Care Med. 171:1209–1223. 53. Saunders, W. B., K. J. Bayless, and G. E. Davis. 2005. MMP-1 activation by serine proteases and MMP-10 induces human capillary tubular network collapse and regression in 3D collagen matrices. J. Cell Sci. 118:2325–2340. 54. Smith, E. E., D. G. Buckley, Z. Wu, C. Saenphimmachak, L. R. Hoffman, D. A. D’Argenio, S. I. Miller, B. W. Ramsey, D. P. Speert, S. M. Moskowitz, J. L. Burns, R. Kaul, and M. V. Olson. 2006. Genetic adaptation by Pseudomonas aeruginosa to the airways of cystic fibrosis patients. Proc. Natl. Acad. Sci. USA 103:8487–8492. 55. Storey, J. D., W. Xiao, J. T. Leek, R. G. Tomkin, and R. W. Davis. 2005. Significance analysis of time course microarray experiments. Proc. Natl. Acad. Sci. USA 102:12837–12842. 56. Venables, W. N., and B. D. Ripley. 2002. Modern applied statistics with S, 4th ed. Springer, New York, NY. 57. Wilson, C. L., K. J. Heppner, P. A. Labosky, B. L. M. Hogan, and L. M. Matrisian. 1997. Intestinal tumorigenesis is suppressed in mice lacking the metalloproteinase matrilysin. Proc. Natl. Acad. Sci. USA 94:1402–1407. 58. Wilson, C. L., K. J. Heppner, L. A. Rudolph, and L. M. Matrisian. 1995. The metalloproteinase matrilysin is preferentially expressed by epithelial cells in a tissue-restricted pattern in the mouse. Mol. Biol. Cell 6:851–869. 59. Wilson, C. L., A. J. Ouellette, D. P. Satchell, T. L. Ayabe, Y. S. Lopez-Boado, J. L. Stratman, S. J. Hultgren, L. M. Matrisian, and W. C. Parks. 1999. Regulation of intestinal ␣-defensin activation by the metalloproteinase matrilysin in innate host defense. Science 286:113–117. 60. You, Y., E. J. Richer, T. Huang, and S. L. Brody. 2002. Growth and differentiation of mouse tracheal epithelial cells: selection of a proliferative population. Am. J. Physiol. Lung Cell. Mol. Physiol. 283:L1315–L1321. 61. Yu, W. H., J. F. Woessner, J. D. McNeish, and I. Stamenkovic. 2002. CD44 anchors the assembly of matrilysin/MMP-7 with heparin-binding epidermal growth factor precursor and ErbB4 and regulates female reproductive organ remodeling. Genes Dev. 16:307–323. 62. Zulianello, L., C. Canard, T. Kohler, D. Caille, J. S. Lacroix, and P. Meda. 2006. Rhamnolipids are virulence factors that promote early infiltration of primary human airway epithelia by Pseudomonas aeruginosa. Infect. Immun. 74:3134–3147.