Fengrong Zuo*,
Naftali Kaminski*§, Elsie Eugui, John Allard, Zohar YakhiniΆ, Amir Ben-DorΆ,
Lance Lollini, David Morris‖, Yong Kim, Barbara
DeLustro, Dean Sheppard‖, Annie Pardo**, Moises Selman**,
Renu A. Heller.
From Roche
Bioscience, Palo Alto, California, Functional
Genomics, and Institute of Respiratory Medicine, Sheba Medical center, Tel
Hashomer, Israel. Ά Agilent technologies, Haifa, Israel, ‖ Lung Biology Center,
Cardiovascular Research Institute and the Department of Medicine University of
California, San Francisco, ** Facultad
de Ciencias, Universidad Nacional Autonoma de Mexico, and Instituto Nacional de
Enfermedades Respiratorias , Mexico.
§ To whom reprints and correspondences should be addressed at:
Functional Genomics, Institute of Respiratory
Medicine, Sheba Medical Center, Tel Hashomer, 52621, Israel
Phone: 972-3-5305772, Fax:
972-3-5302147, E-mail: kamins@sheba.health.gov.il
* Both authors contributed equally to the paper.
Abstract
Pulmonary
fibrosis is a progressive and largely untreatable group of disorders that
affects up to 100,000 people in the United States. To elucidate the molecular
mechanisms that lead to end-stage human pulmonary fibrosis we analyzed samples
from patients with histologically proven pulmonary fibrosis (Usual Interstitial
Pneumonia) using oligonucleotide 0microarrays. Gene expression patterns clearly
distinguished normal from fibrotic lungs.
Many of the genes that were significantly increased in fibrotic lungs
encoded proteins associated with extracellular matrix formation and
degradation, and proteins expressed in smooth muscle. Using a combined set of
scoring systems we determined that matrilysin (matrix metalloproteinase 7,
MMP-7), a metalloprotease not previously associated with pulmonary fibrosis,
was the most informative increased gene in our dataset. Immunohistochemisry demonstrated
increased expression of matrilysin protein in fibrotic lungs. Furthermore, matrilysin knockout mice
were dramatically protected from pulmonary fibrosis in response to
intratracheal bleomycin. Our
results identify matrilysin as a mediator of pulmonary fibrosis and a potential
novel therapeutic target. They also illustrate the power of global gene
expression analysis of human tissue samples to identify novel molecular
pathways involved in clinical disease.
Introduction
The molecular mechanisms
that lead to end-stage pulmonary fibrosis are poorly understood. Although a
multitude of conditions - including granulomatous disorders, exposure to
environmental dusts and toxins, autoimmune diseases, and drugs - have been
associated with pulmonary fibrosis, in many cases the causative agent remains
unknown. The most lethal and
refractory form of pulmonary fibrosis is Idiopathic Pulmonary Fibrosis. Idiopathic Pulmonary Fibrosis (IPF) is
defined histologically by the findings of Usual Interstitial Pneumonia (UIP),
characterized by the interposition of patchy foci of active fibroblast
proliferation associated with minimal inflammation, areas of advanced scarring
and fibrosis (honeycombing), and normal appearing lung (1, 2). The
etiology of this IPF/UIP has remained elusive, and the mean survival from the
time of diagnosis is only 3 years regardless of treatment (3).
Much
of our understanding of the mechanisms of pulmonary fibrosis comes from studies
of bleomycin-induced lung fibrosis in mice. The common paradigm postulates an
initial alveolar injury, which triggers an inflammatory response and,
subsequently, fibrosis (4). The mainstay of the current therapeutic approach to
UIP is suppression of inflammation. However, this approach is ineffective in
most patients (3). Recently, evidence has emerged that UIP may be the
result of disordered fibroblast regulation. Under this paradigm, the pathological process of fibrosis is
the result of persistent remodeling of the lung interstitium by fibroblasts and
smooth muscle cells rather than the result of persistent inflammation (5).
The
advent of microarray technology allows the simultaneous monitoring of the
transcriptional behavior of thousands of genes. This technology has been
repeatedly shown to be useful in the analysis of the response of a variety of
cellular systems to stimuli, in the classification of human cancer and in the
analysis of animal models of human disease (6-8). To characterize the transcriptional profile of UIP,
we analyzed gene expression patterns in lung tissue from patients with UIP
using oligonucleotide microarrays, and compared them with gene expression
patterns in normal lung tissue. Our results demonstrate a clear distinction
between gene expression patterns in the lungs of patients with UIP and control
lungs. In addition to the expected increase in extracellular matrix proteins,
we found a coordinated induction of genes that encode metalloproteases. One of
them, matrilysin (MMP-7), was the gene that was most distinctive between
fibrotic and normal lungs. To analyze the role of matrilysin in pulmonary
fibrosis we injected bleomycin into the tracheas of matrilysin knockout mice,
and found that matrilysin knockout mice are protected from bleomycin induced
pulmonary fibrosis. Our data suggest that matrilysin is a key regulator of
pulmonary fibrosis in mice and humans. Blocking matrilysin activity may thus
serve as a novel therapeutic intervention in this devastating disease.
Methods
Lung specimens
Microarray analysis - Lung specimens from patients with
pulmonary fibrosis were either obtained from lungs removed for lung
transplantation (patients 4,5) or by diagnostic thoracoscopy (patients 6,7,8).
Only one of the patients had been treated with corticosteroids at the time of
biopsy (patient 8, prednisone 15mg/day). Control specimens were obtained from
the normal parts of lungs removed for lung cancer (patients 1,2,3). An
additional control was a purchased pool of total RNA from 5 lungs obtained from
Clontech (sample 0). Patients with pulmonary fibrosis (2 smokers, 3 nonsmokers)
showed a restrictive physiology with decreased diffusion capacity (DLCO),
ground glass attenuation and honeycombing on HRCT, and a typical pattern of UIP
on lung histology. Two suffered from an autoimmune disease (patient 7: Rheumatoid
arthritis, patient 8: Sjogrens Syndrome). The other 3 patients were diagnosed
as having IPF/UIP. The
experimental protocol was approved by the Institutional Internal review Board.
All surgical specimens for gene expression analysis were snap frozen in liquid
nitrogen.
For immunohistochemistry, lung samples were obtained from 7 other
patients with IPF/UIP (5
male and 2 female; 5 nonsmokers, 2 ex-smokers), aged 59 + 6 years by open lung
biopsy. None of the patients had been treated with corticosteroids or
immunosuppressive drugs. The diagnosis of IPF was based on clinical,
radiological and functional findings, and was corroborated by the
characteristic morphology of Usual Interstitial Pneumonia. Control lung tissue
samples were obtained from autopsies of patients who died from non-lung related
causes (2 men and 1 women, mean age 48 + 11 years).
Preparation of labeled cRNA
Probe preparation was performed as recommended by
the manufacturer of the microarrays (9). Briefly, total RNA was isolated by homogenization of
lung specimens in ice-cold Trizol.
Twice purified poly-A mRNA was isolated (Oligotex, Qiagen) and used as
template for double stranded cDNA synthesis with an oligo(dT)24
primer containing a T7 RNA polymerase promoter site added to the 3 end
(Genset). The cDNA was extracted with phenol/chloroform, ethanol precipitated
and used as a template for in vitro transcription (Ambion T7 Megascript system)
with biotin labeled nucleotides (Enzo Diagnostics). Labeled cRNA was fragmented
and a hybridization mix was generated as recommended (9)
Hybridization of microarrays
Aliquots of each sample (10mg cRNA in
200 ml hybridization mix) were hybridized to a Genechip Hugene FLâ array. After hybridization, each array was washed,
stained with streptavidin phycoerythrin (Molecular probes), washed again,
hybridized with biotin labeled anti streptavidin phycoerythrin antibodies,
restained with streptavidin phycoerythrin (Molecular probes) and scanned
(Hewlett Packard, GeneArrayTM scanner G2500A).
Analysis of Genechip data
Scanned output files were visually inspected for hybridization
artifacts. Arrays lacking significant artifacts were analyzed using Genechipâ 3.3 software (Affymetrix). Arrays were scaled to an
average intensity of 150 per gene and analyzed independently. The expression
value for each gene was determined by calculating the average of differences
(perfect match intensity minus mismatch intensity) of the probe pairs in use
for this gene. The expression analysis files created by Genechipâ 3.3 software were then transferred to a database
(Microsoft Access) linked to Internet genome databases (e.g. NHLBI, Swiss Prot,
and GeneCards) to update gene definitions. A value of 20 was assigned to all measurements lower than
20. For cluster analysis we used Cluster and Treeview programs described by
Michael Eisen (10). We did not include in the analysis genes that did
not have at least one average difference intensity value ³ 100 or one present call by Affymetrix criteria. Fold ratios were
calculated for each sample against the median of the controls. Average fold
ratios are the averages of the fold ratios of the UIP samples.
To determine the most informative
genes we used scoring methods previously described by Ben-Dor et al (10) and applied to the analysis of breast cancer and
melanoma using cDNA arrays (12). For detailed descriptions see reference (13). In
brief, a gene is designated as informative based on the degree to which its
tissue expression level is predictive of an independent classification of the
tissue sample as diseased or not diseased (11). The scores used in this study were:
TNoM (Total number of misclassifications) - a count of the number of
classification errors committed when using the best simple threshold to distinguish
between two classes (diseased or not diseased) based on the expression levels
of a specific gene.
Info an estimate of the uncertainty remaining about the sample
classification (diseased or not diseased) after the incorporation of predictions
based on expression of an individual gene is given (a lower Info score
indicates a higher predictive value for a given gene).
Gaussian The overlap between
distributions of expression levels for genes in two classes. The score is based on normality
assumptions.
Immunohistochemistry
Immunohistochemistry
was performed as previously described (14). Briefly, after deparaffinization and re-hydration
tissue sections were blocked with 3% H2O2 in methanol, followed
by antigen retrieval in citrate buffer (10 mM pH 6.0). Lung sections were
incubated with normal serum for 30 minutes, followed by overnight incubation
at 4oC with primary monoclonal anti-MMP-1 (20 mg/ml),
anti-MMP-9 (5 mg/ml) (Fuji Chemical Ind., Ltd., Toyama, Japan), or anti-MMP-7
(20 mg/ml) (Chemicon International, Inc. Temecula, CA). A secondary
biotinylated anti-immunoglobulin was applied followed by horseradish
peroxidase-conjugated streptavidin (BioGenex, San Ramon CA) according to manufacturer. 3-amino-9-ethyl-carbazole (AEC,
BioGenex) in acetate buffer containing 0.05% H2O2 was used as substrate.
The sections were counterstained with hematoxylin. The primary antibody was
replaced by non-immune serum for negative control slides.
Bleomycin Treatment
Age and sex matched 8 -16 weeks
old 129/Sv MMP-7+/+ (Jackson) and MMP-7-/-mice
(a gift from LM Matrisian), and 14-23 weeks old C57Bl/6 MMP-7+/+
(Charles River) and MMP-7-/- (LM Matrisian) mice were
maintained in a specific pathogen-free environment. Mice were anesthetized by
methoxyflurane and a 24-gauge needle was inserted into the trachea via the oral
cavity. 50 ml of bleomycin (0.05-0.08 units
in 0.9% saline, Sigma) or saline was slowly injected. Mice were sacrificed 14
or 21 days after bleomycin or saline injection, and lungs were collected for
either hydroxyproline determination or histology.
Hydroxyproline
assay
To estimate the total amount of collagen in the
lungs, hydroxyproline was measured as described (15). In brief, both lungs were removed and submitted to
acid hydrolysis in 5 ml 6N HCl for 21 hr at 110 oC. Precipitates
were removed by centrifugation. Supernatants were dried overnight and dissolved
in 100 ml of distilled water at room temperature. Each sample was tested in
duplicate. 50 ml of chloramine T (Sigma) in acetate-citrate buffer, pH 6.0, 50 ml of
perchloric acid, and 50 ml of Erlichs reagent were added sequentially to
each sample as previously described (15). Absorbance was measured at 570 nm and the amount of
hydroxyproline was calculated.
Histology
and Morphometry
Murine
lungs were fixed in 10% formalin by perfusion followed by submersion and
processed by routine histologic methods through paraffin into H&E stained sections
for evaluation by light microscopy. A single longitudinal section of each lung
lobe was examined from each mouse.
Scores were assigned for pulmonary fibrosis, increased numbers of
pulmonary alveolar macrophages, the presence leukocytes other than macrophages
within alveoli, and interstitial leukocytic infiltrates. Each parameter was marked as being
essentially normal or given a severity score from 1, minimal severity, to 5,
most severe. A score of 1
indicated a threshold level. A score of 2 indicated slight changes that did not
involve a significant amount of tissue. A score of 3 indicated the effect was
easily discernable in multiple sites and/or was present in larger foci in from
one to a few sites. A score of 4 indicated a relatively severe change that was
widespread within multiple lobes or which completely filled a single lobe. A
score of 5 indicated that a change was severe and involved all lobes.
Results
Gene expression patterns are distinct in fibrotic and non-fibrotic lungs - Global gene expression patterns clearly distinguished between
fibrotic and normal histology lungs (Fig 1a). Furthermore, when only the genes
that received the most significant score (TNoM = 0) are presented, the
differences in gene expression patterns are even clearer (Fig 1b). This
distinction was also evident when we applied hierarchical clustering to the
samples (data not shown).
A known issue in
analysis of microarray data is the small numbers of samples compared to the
large number of parameters measured a situation that can lead to spurious
statistical associations. To address this issue we subjected TNoM to rigorous
statistical benchmarking. We calculated the number (out of 7129) of genes
expected for every score (TNOM 0,1,2 etc), when a classification of the tissues
is uniformly drawn at random. In this case we expected 111 genes to have the
best score (TNoM = 0). We actually
observed 164 such highly informative genes. Such an overabundance of highly
informative genes has a p value < 10-6 when genes are independently
drawn and their TNoM is calculated with respect to a random partition. In
performing the calculations we treated the pool of 5 normal lungs as a single
sample, thus the figures presented here may even underestimate the information
content of this study. The complete dataset is provided on our website (http://FGUSheba.cs.huji.ac.il/)
Smooth muscle markers are increased in fibrotic lungs - One of the most intriguing phenomena that occurs
during the development of UIP is the formation of small aggregates of actively
proliferating myofibroblasts and fibroblasts called myofibroblast/fibroblast
foci (5). In addition, abnormal collections of smooth muscle
cells have been identified in UIP lungs (16). Interestingly, we observed a marked increase in the
expression of genes that encode for muscle proteins. These included markers of
smooth muscle differentiation such as vascular a smooth muscle actin, g smooth
muscle actin, calponin and integrin a7b1. Genes that encode for proteins
associated with cell contraction and actin filament organization, such as
myosin, SM22, tropomyosin and S100 A9 protein are also increased. The complete
list is provided in table 1.
Genes that encode immunoglobulins, complement and some chemokines were
upregulated in fibrotic lungs
Surprisingly, gene expression patterns in patients with pulmonary fibrosis did
not suggest increased acute inflammatory activity. We did not observe any activation
of an IL-1 or TNFa related
pathway, nor did we observe any increase in apoptosis related genes. However,
we did observe an increase in genes that encode for immunoglobulins,
potentially reflecting chronic B cell infiltration, and an increase in some
complement factors. The genes that encode the chemokines MCP 4, GRO 1 and
eotaxin were increased (Table 1).
Genes
that encode proteins that are involved in extracellular matrix formation,
degradation and signaling are increased in fibrotic lungs - As expected expression of the genes that encode for collagens I, and
III were increased in fibrotic lungs, as was expression of the genes that
encode for collagen VI, tenascin C, osteopontin and fibronectin (Table 1).
Several other genes that encode for proteins involved in extracellular matrix
formation such as BIGH3, filamin and fibrillin were increased, as were genes
associated with ECM related signaling such as cell adhesion kinase b and the collagen receptor tyrosine kinase DDR (Table 1). Another interesting subset was of genes
that encoded proteases. The
expression levels of Matrix metalloproteinase 1 (MMP-1), Matrix
metalloproteinase 2 (MMP-2), MMP-7, Matrix metalloproteinase 9 (MMP9) (table 1)
and cathepsin E were all relatively increased in fibrotic lungs. MMP7
(matrilysin) was the gene that scored as the most informative (TNoM = 0, Info =
0, Gaussian = 0.00625, complete dataset, http://FGUSheba.cs.huji.ac.il/) among
the genes that were upregulated (Figure 2A).
Increased expression of MMP-7 is confirmed by
immunohistochemistry - To
verify the gene expression data at the level of protein expression, and to
identify the cellular source of MMP-7 we performed immunohistochemiistry on
lung tissue from patients with UIP. Immunoreactive MMP-7 was abundantly
localized in the epithelium including in alveolar and bronchiolar epithelial
cells, (Figures 2B and C, and Supplemental figure 4A-C at http://www.pnas.org/). MMP-7 was also detected in the extracellular space
as shown by staining in thickened alveolar septa and in the extracellular
matrix (Figure 2B). MMP-7 was
minimally detected in normal lung parenchyma (Figure 2D). We also stained for
MMP-1 and MMP-9. MMP-1 was primarily present on hyperplastic type 2 pneumocytes
and bronchiolar cells lining honeycomb cysts, as well as in reactive
alveolar epithelial cells. (Supplemental Figure 5A and B, http://www.pnas.org/). Alveolar macrophages also intensely stained for
MMP-1 in tissue sections from patients with fibrotic lung disease (1A,
Supplemental figure 5 insert A, http://www.pnas.org/). MMP-9 was detected in neutrophils, both inside
vessels and in the interstitium (Supplemental Figure 6A and insert, http://www.pnas.org/). MMP-9 was also localized in the
extracellular matrix and in some areas of dense scars (Figure 3B, supplemental
data). Control lungs showed scattered staining for MMP-1 in (Supplemental
figure 5C, http://www.pnas.org/), and minimal MMP-9 staining in neutrophils
(Supplemental figure 6D, http://www.pnas.org/). Control samples incubated with non-immune sera
were negative as shown in Figure 4D supplemental data (http://www.pnas.org/).
Matrylisin knockout mice do not develop lung
fibrosis following treatment with bleomycin - To determine whether MMP-7 was actually involved
in the fibrotic process, we administered bleomycin to lungs of MMP-7-/-mice,
and to age and sex matched wild type controls from two different backgrounds:
129Sv and C57Bl/6. Two to three weeks after intratracheal bleomycin,
hydroxyproline level increased by 96 - 119% in wild type 129 Sv mice compared
to 55 - 59% in matrilysin knockout mice, in three separate experiments using 5
- 9 mice/group (Figure 3A). The differences were statistically significant with
p values from 0.0002 to 0.02 in all three experiments. This finding was not
dependent on the genetic background of the mice, since comparable results were
obtained in three separate experiments with mice of the C57Bl/6 background (Fig
3A). These results were supported by histologic evaluation that revealed milder
changes in the lungs of matrilysin knockout mice (Fig 3D). The mean severity
for all histopathologic parameters assessed was significantly less in the
bleomycin-challenged MMP-7-/- mice than in the corresponding WT
controls (Fig 3B).
Discussion
The results of this study support
the view of UIP as a disease of persistent matrix deposition and destruction
that is associated with relatively modest, chronic inflammation. Interestingly,
despite the prevailing view of honeycomb lung as an irreversibly scarred and
quiescent tissue, our gene expression data show a tissue that is quite actively
remodeling with a high level of gene expression of both matrix degrading
proteins and proteins involved in matrix deposition.
One of the most impressive
features of our results is the clear distinction in gene expression patterns,
and therefore the fundamental tissue biology, of UIP lungs compared to
non-fibrotic lungs. A usual concern with microarray analysis of diseases, in
which tissue availability is a challenge, is the difficulty in obtaining data
that is both biologically meaningful and statistically significant. The use of
traditional statistics is complicated in microarray experiments because of the
troubling asymmetry between the numbers of samples and the number of parameters
measured, which may result in spurious statistical associations. One of the
advantages of the analytic methods that we used here is their amenability to
rigorous statistical benchmarking (11). We can calculate the number of informative genes per score expected in a
random classification and then compare this estimated number of high scoring
(or informative) genes to the actual number of informative genes (per score)
measured in our dataset (13). In our study, despite the small sample size and the
genetic heterogeneity of our patients we observed overabundance of informative
genes suggesting that the differences in gene expression that we observed
between the two groups were biologically meaningful. Indeed, gene expression
patterns were highly robust and also evident when we applied other statistical
tools (t test, Wilcoxon), clustering methods and self-organizing maps (data not
shown). The results of our analysis are an example of overcoming some of the
limitations of a small sample size using computational tools specifically
designed for microarray experiments.
Close inspection of the
informative genes in our data set sheds light on several important aspects of
fibrotic lung disease. One
impressive feature is the coordinated upregulation of matrix
metalloproteinases, predominantly MMP-1 and MMP-7, in the lungs of
patients with UIP. It has been
suggested that the tissue phenotype of fibrosis (primarily the exaggerated extracellular matrix deposition)
results from dysregulation of the synthesis and degradation of extracellular
matrix proteins - a process that could involve several members of the MMP
family. Thus, MMP-1
(collagenase-1 that degrades fibrillar collagens) as well as MMP-2 and MMP-9
(gelatinases A and B which have a broad range of substrates including type IV
collagens from basement membranes) have previously been reported to be
upregulated in human pulmonary fibrosis and in animal models of pulmonary
fibrosis (17-19). In our
dataset, MMP-1, 2, and 9 were significantly higher in UIP lungs. These
observations were verified at the protein expression level for MMP-1 and 9
(Supplemental figures 5,6). We did not observe any significant increase in gene
expression levels of tissue inhibitors of metalloproteinases in UIP lungs.
Surprisingly we detected a decrease in TIMP3 gene expression (Complete data set
table, http://FGUSheba.cs.huji.ac.il/), an observation that has been seen in
other microarray experiments (Pardo A. Unpublished observation).
MMP-7, a metalloproteinase not previously associated with pulmonary
fibrosis, was the most informative gene among the increased genes in UIP lungs.
This was confirmed by immunohistochemical staining on lung tissue from a
separate group of patients with UIP. The intense staining confirmed increased
MMP-7 protein expression by epithelial cells as well as an increase in MMP-7
bound to the extracellular matrix. The protection of MMP-7-/-
knockouts from bleomycin-induced fibrosis shows that this metalloproteinase
plays an important role in the development of pulmonary fibrosis. MMP-7 is
expressed in epithelial tissues that are in contact with the environment. Although it is not expressed by mice
living in germ free environment (20), its expression is rapidly induced by infection and
air pollution (21, 22). MMP-7 is required for activation of defensins in the
small bowel and MMP-7-/- mice cannot kill enteric
pathogens (23). Recently, MMP-7 has been shown to play a role in two
pathways that are considered important in the development of bleomycin-induced
pulmonary fibrosis, the TNFa, and Fas
pathways (24, 25). Like MMP-1, MMP-7 was primarily expressed by
alveolar epithelial cells, and both may have a role in cell migration as has
been proposed for keratinocytes in skin wound healing (26, 27). Similarly, MMP-1 and MMP-7 may participate in
alveolar and bronchiolar cell migration over different matrices during UIP lung
remodeling (28). In addition, MMP-7 may enhance a procoagulant
microenvironment in the alveolar spaces since it is able to rapidly cleave
tissue factor pathway inhibitor, an inhibitor of tissue factor-induced
coagulation (29). Increased local procoagulant activity has been found
in fibrotic lung disorders, including idiopathic pulmonary fibrosis, and
alveolar epithelial cells seem to contribute to the increased procoagulant and
antifibrinolytic activities in this disorder (5,
30). Based
on our data and its known functions, matrilysin is an excellent candidate to be
a key player in sustained tissue injury and fibrosis. On one side it is induced
by epithelial injury, a common pathological process leading to pulmonary
fibrosis, and on the other hand it may be an activator of cell death,
chronic inflammation, and procoagulant activity. The focus on MMP-7 in our work
is a good example of large-scale gene expression experiments as hypothesis generating
tools. Based on previous knowledge we had little reason to suspect that MMP-7
played a role in pulmonary fibrosis.
However, no matter what type of analysis was used, MMP-7 was the most
significant gene that was induced in UIP samples, thus almost demanding its
analysis.
Among the genes that where
significantly upregulated, we did not find many inflammatory related genes. We
observed evidence of a chronic inflammatory infiltration probably associated
with B-cell infiltration (immunoglobulins etc), but we did not observe any
evidence for T cell activation, or for any significant activation of apoptotic
pathways. Although in animal models of pulmonary fibrosis the fibrotic response
may be dependent on intact TNFa or Fas signaling pathways (31-35), activation of these pathways was not apparent in
this study.
One limitation of using whole
lung homogenates in our study is our relative inability to assess the exact
cells that are overexpressing genes that were higher in UIP lungs. For some
typical cellular markers it is very hard to even guess whether the changes
result from changes in cellular content or real gene expression patterns. For
instance we observed changes in three cytokeratins (5, 8, 15). This may
represent a change in epithelial gene expression, a change in epithelial
cellular subset (epithelial migration), or a phenotypic change in another cell
type. So far such analysis is cumbersome, but the creation of databases that
will contain gene expression signatures of specific cell types and analytic
approaches to dissect them out computationally could ameliorate this hardship.
Despite this limitation one of the impressive clusters of genes that that were
significantly increased in UIP lungs was that of muscle related genes. An important feature in the
histological description of UIP is the presence of fibroblastic foci (5). Fibroblast foci possibly represent microscopic areas
in which fibroblasts migrate, proliferate, and maintain the abnormal
architecture of the lung. A major subpopulation of the fibroblasts in these
foci expresses markers of smooth muscle differentiation, such as a-smooth muscle actin (aSMA) (16,
17). In
addition, clusters of smooth-muscle cells are usually seen in UIP lungs, probably
contributing to the decreased lung compliance (16).
We observed an increase in muscle-related genes that ranged from 2.5 to
25 fold (table 1). These coordinated increases in smooth muscle markers, and in
cellular contractile machinery, reflect our ability to detect a significant
phenotypic change in even in a relatively small cell population in the lung.
In this study we used
oligonucelotide microarrays to analyze gene expression patterns in UIP. We
distinguished specific subsets of genes that differentiated UIP from
non-fibrotic lungs. These genes included genes that encoded proteins associated
with B cells and with chronic inflammation, muscle related genes and genes that
encoded proteins involved in extracellular matrix production, degradation and
signaling. Using advanced computational scoring methods we determined that
MMP-7, a multifunctional matrix metalloprotease not previously associated with
pulmonary fibrosis, was the most informative increased gene in our dataset. The hypothesis that MMP-7 is indeed
involved in the pathogenesis of pulmonary fibrosis was supported by our
observation that MMP-7 knockout mice were protected from bleomycin induced
fibrosis. Our results support the view of UIP as a chronic destructive process,
in which an increase in myofibroblasts and cells with smooth muscle-like
phenotype, active abnormal reepithelialization, and aberrant repair are
interwoven.
The authors thank Gady Cojocaro and Dr Hal van Wart for their support
and, Dr. Tom Raffin and Ms. Susan Jacob for their
support to obtain fresh human lung tissue samples from Stanford Medical Center,
Stanford University. Dr.
Isasshar Ben-Dov provided helpful comments on the manuscript. Dr Kaminski's
work was supported in part by a generous grant from the Tel-Aviv chapter of the
Israeli Lung Association.
References
1. Costabel,
U. & King, T. E. (2001) Eur Respir J 17, 163-7.
2. Ryu, J. H.,
Colby, T. V. & Hartman, T. E. (1998) Mayo Clin Proc 73,
1085-101.
3. King, T.
E., Jr. (1998) Ann Intern Med 129, 806-12.
4. Coker, R.
K. & Laurent, G. J. (1998) Eur Respir J 11, 1218-21.
5. Selman, M.,
King, T. E. & Pardo, A. (2001) Ann Intern Med 134, 136-51.
6. Cojocaru,
G. S., Rechavi, G. & Kaminski, N. (2001) Isr Med Assoc J 3,
292-6.
7. van Berkum,
N. L. & Holstege, F. C. (2001) Curr Opin Biotechnol 12,
48-52.
8. Kaminski,
N., Allard, J. D., Pittet, J. F., Zuo, F., Griffiths, M. J., Morris, D., Huang,
X., Sheppard, D. & Heller, R. A. (2000) Proc Natl Acad Sci U S A 97,
1778-83.
9. (1998) in Affymetrix,
Inc. Eukaryotic expression analysis target preparation, product enclosure.
10. Eisen, M. B., Spellman,
P. T., Brown, P. O. & Botstein, D. (1998) Proc Natl Acad Sci U S A 95,
14863-8.
11. Ben-Dor, A., Bruhn, L.,
Friedman, N., Nachman, I., Schummer, M. & Yakhini, Z. (2000) J Comput
Biol 7, 559-83.
12. Hedenfalk, I., Duggan,
D., Chen, Y., Radmacher, M., Bittner, M., Simon, R., Meltzer, P., Gusterson,
B., Esteller, M., Kallioniemi, O. P., Wilfond, B., Borg, A. & Trent, J.
(2001) N Engl J Med 344, 539-48.
13. Ben-Dor, A., Friedman,
N. & Yakhini, Z. (2000), (Agilent Laboratories Technical report
AGL-2000-13: http://www.labs.agilent.com/resources/techreports.html.
14. Pardo, A., Barrios, R.,
Gaxiola, M., Segura-Valdez, L., Carrillo, G., Estrada, A., Mejia, M. &
Selman, M. (2000) Am J Respir Crit Care Med 161, 1698-704.
15. Woessner, J. F. (1961) Arch.
Biochem. Biophys. 93, 440-447.
16. Ohta, K., Mortenson, R.
L., Clark, R. A., Hirose, N. & King, T. E., Jr. (1995) Am J Respir Crit
Care Med 152, 1659-65.
17. Selman, M., Ruiz, V.,
Cabrera, S., Segura, L., Ramirez, R., Barrios, R. & Pardo, A. (2000) Am
J Physiol Lung Cell Mol Physiol 279, L562-74.
18. Lemjabbar, H., Gosset,
P., Lechapt-Zalcman, E., Franco-Montoya, M. L., Wallaert, B., Harf, A. &
Lafuma, C. (1999) Am J Respir Cell Mol Biol 20, 903-13.
19. Swiderski, R. E.,
Dencoff, J. E., Floerchinger, C. S., Shapiro, S. D. & Hunninghake, G. W.
(1998) Am J Pathol 152, 821-8.
20. Parks, W. C.,
Lopez-Boado, Y. S. & Wilson, C. L. (2001) Chest 120, S36-41.
21. Su, W. Y., Jaskot, R.
H., Richards, J., Abramson, S. R., Woessner, J. F., Jr., Yu, W. H. &
Dreher, K. L. (2000) Am J Physiol Lung Cell Mol Physiol 279,
L152-60.
22. Lopez-Boado, Y. S.,
Wilson, C. L., Hooper, L. V., Gordon, J. I., Hultgren, S. J. & Parks, W. C.
(2000) J Cell Biol 148, 1305-15.
23. Wilson, C. L.,
Ouellette, A. J., Satchell, D. P., Ayabe, T., Lopez-Boado, Y. S., Stratman, J.
L., Hultgren, S. J., Matrisian, L. M. & Parks, W. C. (1999) Science 286,
113-7.
24. Mitsiades, N., Yu, W.
H., Poulaki, V., Tsokos, M. & Stamenkovic, I. (2001) Cancer Res 61,
577-81.
25. Haro, H., Crawford, H.
C., Fingleton, B., Shinomiya, K., Spengler, D. M. & Matrisian, L. M. (2000)
J Clin Invest 105, 143-50.
26. Dunsmore, S. E.,
Saarialho-Kere, U. K., Roby, J. D., Wilson, C. L., Matrisian, L. M., Welgus, H.
G. & Parks, W. C. (1998) J Clin Invest 102, 1321-31.
27. Planus, E., Galiacy,
S., Matthay, M., Laurent, V., Gavrilovic, J., Murphy, G., Clerici, C., Isabey,
D., Lafuma, C. & d'Ortho, M. P. (1999) J Cell Sci 112,
243-52.
28. Parks, W. C., Sudbeck,
B. D., G.R. Doyle &
Saariahlo-Kere, U. K. (1998) in Matrix Metalloproteinases, eds. Parks,
W. C. & Mecham, R. P. (Academic Press, San Diego), pp. 263-297.
29. Belaaouaj, A. A., Li,
A., Wun, T. C., Welgus, H. G. & Shapiro, S. D. (2000) J Biol Chem 275,
27123-8.
30. Loskutoff, D. J. &
Quigley, J. P. (2000) J Clin Invest 106, 1441-3.
31. Ortiz, L. A., Lasky,
J., Hamilton, R. F., Jr., Holian, A., Hoyle, G. W., Banks, W., Peschon, J. J.,
Brody, A. R., Lungarella, G. & Friedman, M. (1998) Exp Lung Res 24,
721-43.
32. Zhang, K.,
Gharaee-Kermani, M., McGarry, B., Remick, D. & Phan, S. (1997) J Immunol
158, 954-9.
33. Aoshiba, K., Yasui, S.,
Tamaoki, J. & Nagai, A. (2000) Am J Respir Crit Care Med 162,
695-700.
34. Kuwano, K., Hagimoto,
N., Tanaka, T., Kawasaki, M., Kunitake, R., Miyazaki, H., Kaneko, Y., Matsuba,
T., Maeyama, T. & Hara, N. (2000) J Pathol 190, 221-9.
35. Kuwano, K., Hagimoto, N., Kawasaki, M.,
Yatomi, T., Nakamura, N., Nagata, S., Suda, T., Kunitake, R., Maeyama, T.,
Miyazaki, H. & Hara, N. (1999) J Clin Invest 104, 13-9.
Figure Legends
Figure 1: Gene
expression infogram for all 7129 genes (A) and for all the 164 most informative
genes (TNoM = 0) (B). To eliminate outlier effect, genes were normalized to a
range of [0,1], meaning that the maximum value for every gene was set to be 1,
the minimum value to be 0, and the rest of the values were linearly fitted to
this range. Yellow is maximal expression and blue is minimal.
Figure 2: MMP-7 gene
and protein expression levels are significantly increased in fibrotic lungs.
The Gaussian distribution for MMP-7 reveals minimal overlap (Score of 0.00625)
between normal lungs (green line, A) and fibrotic lungs (red line, A). All
actual expression levels of MMP-7 in fibrotic lungs (red X marks, A) were
higher than in normal lungs (green X marks, a) and no samples were
misclassified (TNoM and Info = 0). The y-axis in the gaussian figure (A) is the
estimated density of the expression level. This density is computed based on
the normality assumption of expression level (for each class). MMP-7 staining
verified the gene expression predictions and demonstrated both cell-associated
and matrix MMP-7 in distal airspaces (B) and proximal airway (C) from fibrotic
lungs. No MMP-7 immunoreactivity was seen in normal lungs (D). (Original
magnification B, C, and D:40X).
Figure 3. MMP7-/-
mice are protected against bleomycin-induced pulmonary fibrosis. Increase in
lung hydroxyproline content in lungs of wild type (MMP-7+/+) mice 14
days after bleomycin injection is significantly higher than in MMP-7-/- mice
on both C57BL6 and 129SvJ background. Increase values were calculated by
subtracting mean hydroxyproline content (mg hydroxyproline/lung)
of lungs of saline injected animals from the hydroxyproline content of lungs 14
days after bleomycin (A). Comparison of semi quantitative morphometric analysis
of lung histology 14 days after bleomycin (B) and representative histology of
MMP-7+/+ (C) and MMP-7-/- mice (D). Data (means ± SEM)
are representative of at least two comparable experiments with 6 mice per
group; * P < 0.01 relative to bleomycin treated MMP+/+
mice.
Tables
Table 1. Functional
groups of selected genes substantially upregulated in fibrotic lungs. (Numbers
in parentheses are average fold changes)
Table 1.
Functional groups of selected genes substantially upregulated in
fibrotic lungs (fold induction)
Muscle markers
[D17408]
calponin (9.9)
[S81419] dystrophin (2.9) [D00654] enteric smooth muscle g-actin (4.2) [M12125]
fibroblast tropomyosin muscle-type (26) [X74295] Integrin a7b1 (1.7) [U48959] myosin light chain kinase MLCK (6.7) [U14391] myosin-IC (4.2) [AF001548]
myosin heavy chain (2.9) [M63603] PLN (2.5) [L10678] profilin II (3.2) [M95787] SM22 (2.3) [M26311] S100 calcium-binding protein A9 (2.7) [Z24727] tropomyosin isoform (2.1) [X13839] vascular smooth muscle a-actin
(2.6) ECM, growth factors and
proteases
[M22489]
BMP2a (3.9) [M77349]
BIGH3 (2.2) [X65784] CAR (1.7) [J05036] cathepsin E (9.8) [U43522] cell adhesion kinase b (4.5) [M55998]
Collagen I-a1 (2.4) [Z74616] Collagen I-a2 (5.2) [X06700] Collagen III-a1 (4.7) [X52022] Collagen VI-a3 (2.5) [U48705] DDR (11.3) [L13923] fibrillin (2.0) [D83920] ficolin-1 (4.5) [X53416] filamin (4.3) [S37730] IGFBP2 (5.8) [L27560] IGFBP5 (2.4) [M62403]
IGFBP4 (2.0) [M21389]
keratin 5 (20.3) [X74929] keratin 8 (2.9) [X07696] keratin 15 (14.1) |
Cytokines, chemokines
and antioxidants
[U25182] antioxidant enzyme AOE37 (25.6) [U72511] BAP (2.3) [L33930] CD24 (3.4) [D49372] eotaxin (4.0) [X54489] GRO1 (2.9) [X65727] GSTa (7.0) [L76191] IL-1 r-associated kinase (6.1) [U00672] IL-10 r (2.8) [S74221] IK (4.5) [U46767] MCP-4 (8.6) [M37766]
MEM-102 glycoprotein (6.8) [L19686] MIF (8.9) [M94250]
Retinoic acid inducible factor MK (16.5) [X65965] SOD-2 (4.6) [D50663] TCTEL1 (3.6) [X16662]
vascular anticoagulant-b (2.7) Complement,
Immunoglobulins and Amyloid
[U50939]
amyloid protein-binding protein 1 (3.2) [M84526]
adipsin/complement factor D (2.2) [L15702]
complement factor B (2.7) [K02765]
complement C3, aand b (2.0) [M65292]
factor H homologue (2.7) [U28488]
G protein-coupled r AZ3B (2.5) [M63438]
Ig rearranged g chain,
V-J-C region (4.8) [S71043]
Ig a2 (4.3) [M12759]
Ig J chain (3.6) [L02326]
l -17
(4.7) [M34516]
o light
chain protein (2.4) [U22178]
microseminoprotein b (8.4) [X57809]
rearranged immunoglobulin l light
chain (4.2) [X51441]
SAA protein (10.9) [M63379]
TRPM-2 (1.8) [V00563] V00563 (3.9) |