Alveolar epithelial cell mesenchymal transitiondevelops in vivo during pulmonary fibrosis andis regulated by the extracellular matrixKevin K. Kim, Matthias C. Kugler, Paul J. Wolters, Liliane Robillard, Michael G. Galvez, Alexis N. Brumwell,Dean Sheppard, and Harold A. Chapman*

Department of Medicine and Cardiovascular Research Institute, University of California, San Francisco, CA 94143

Communicated by Shaun R. Coughlin, University of California, San Francisco, CA, July 6, 2006 (received for review May 5, 2006)

Mechanisms leading to fibroblast accumulation during pulmonaryfibrogenesis remain unclear. Although there is in vitro evidence oflung alveolar epithelial-to-mesenchymal transition (EMT), whetherEMT occurs within the lung is currently unknown. Biopsies fromfibrotic human lungs demonstrate epithelial cells with mesenchy-mal features, suggesting EMT. To more definitively test the capac-ity of alveolar epithelial cells for EMT, mice expressing �-galacto-sidase (�-gal) exclusively in lung epithelial cells were generated,and their fates were followed in an established model of pulmo-nary fibrosis, overexpression of active TGF-�1. �-gal-positive cellsexpressing mesenchymal markers accumulated within 3 weeks ofin vivo TGF-�1 expression. The increase in vimentin-positive cellswithin injured lungs was nearly all �-gal-positive, indicating epi-thelial cells as the main source of mesenchymal expansion in thismodel. Ex vivo, primary alveolar epithelial cells cultured on provi-sional matrix components, fibronectin or fibrin, undergo robustEMT via integrin-dependent activation of endogenous latent TGF-�1. In contrast, primary cells cultured on laminin�collagen mixturesdo not activate the TGF-�1 pathway and, if exposed to activeTGF-�1, undergo apoptosis rather than EMT. These data revealalveolar epithelial cells as progenitors for fibroblasts in vivo andimplicate the provisional extracellular matrix as a key regulator ofepithelial transdifferentiation during fibrogenesis.

apoptosis � fibronectin � integrin �v�6 � TGF-� �usual interstitial pneumonia

Idiopathic pulmonary fibrosis (IPF) is a disease characterizedby fibroblast accumulation, excessive collagen deposition, and

matrix remodeling, which lead to distortion of the alveolararchitecture, progressive decline in lung function, and ultimatelydeath (1). The pathologic hallmarks of IPF include heteroge-neous injury�fibrosis, scattered foci of fibroblast accumulationbeneath flattened alveolar epithelial cells (AECs), and scantinflammation (1). The pathways leading to excess collagen-producing fibroblasts in either IPF or experimental models oflung fibrosis are uncertain. Although expansion of residentfibroblasts likely occurs in response to injury, several lines ofevidence support the possibility that lung fibroblast developmentduring injury responses may also be derived from bone marrowand�or epithelial cells (2, 3). A greater understanding of thecellular origin of fibroblast accumulation in response to injury isimportant because the regulation of cellular transdifferentiationinto a fibroblast phenotype is likely very different from theregulation of fibroblast proliferation (3–5). Fibroblasts isolatedfrom IPF lungs have heterogenous phenotypes and propertiesdifferent from that of normal lung fibroblasts (6). One potentialexplanation for this heterogeneity is that fibroblasts are derivedfrom multiple cell types. There have been prior suggestions thatepithelial-to-mesenchymal transition (EMT) occurs in the lungduring fibrogenesis, but these suggestions derive largely fromstudies of transformed cells or primary AECs cultured on plastic,the in vivo significance of which is unclear (7–9). Reports using

primary AECs (8, 9) are more compelling, but current isolationmethods do not completely remove contaminating mesenchymalcells, thus making it uncertain whether factors thought to induceEMT may not actually favor the growth of contaminatingmesenchymal cells and�or promote the loss of epithelial cellsleading to increased measurements of mesenchymal markers.

To address these uncertainties, we used several approaches toexplore lung EMT during the pathogenesis of both experimentalpulmonary fibrosis and IPF. We found that IPF lung biopsiesdemonstrate epithelial cells which have acquired mesenchymalfeatures, raising the possibility of EMT during fibrogenesis. Wethen developed a triple transgenic mouse reporter system inwhich lung epithelial cells are genetically altered to permanentlyexpress �-gal, allowing us to map the fate of these cells anddefinitively test whether EMT occurs in vivo in an animal modelof pulmonary fibrosis. These studies provide the strongestevidence to date that EMT occurs during fibrogenesis andreveals EMT to be a surprisingly robust process. Furthermore,our data show that the provisional matrix drives this process byactivating the integrin �v�6 and thereby latent TGF-�1, reveal-ing a mechanism by which the provisional matrix initiates repair.

ResultsAECs in IPF Lungs Demonstrate Features of Early EMT. Immunoflu-orescent staining of lung biopsies from patients with IPF revealedcells within remodeling matrices with features of both epithelial andmesenchymal cells, suggesting active EMT. In each of four diag-nostic biopsies proven to be IPF, a subset of cells stained weakly forpro-surfactant protein-C (pro-SPC), a type II AEC marker, andN-cadherin, a mesenchymal cadherin (Fig. 1). This appeared tofrequently develop near sites of alveolar collapse and AEC clus-tering. Epithelial cells lining the lumen of injured alveoli stainedmuch brighter for pro-SPC and had no discernible N-cadherinstaining (Fig. 1). The presence of pro-SPC-positive cells within theexpanded interstitium suggests AEC detachment from the base-ment membrane and migration within the extracellular matrix.Indeed, these epithelial cells have lost apical–basal polarity asdemonstrated by diffuse staining for the integrin �3�1 (Fig. 7,which is published as supporting information on the PNAS website), a laminin receptor normally expressed on the basal surface ofAECs. Cells that might have completed EMT would be expected tohave lost epithelial features, such as expression of pro-SPC. Becausesuch fibroblasts in the human tissues could not be distinguishedfrom resident fibroblasts, we turned to a mouse system in whichAECs are marked by genetic techniques to track their fate, allowing

Conflict of interest statement: No conflicts declared.

Freely available online through the PNAS open access option.

Abbreviations: IPF, idiopathic pulmonary fibrosis; AEC, alveolar epithelial cell; EMT, epi-thelial-to-mesenchymal transition; pro-SPC, pro-surfactant protein-C; Fn, fibronectin; Mg,matrigel.

*To whom correspondence should be addressed. E-mail: hal.chapman@ucsf.edu.

© 2006 by The National Academy of Sciences of the USA

13180–13185 � PNAS � August 29, 2006 � vol. 103 � no. 35 www.pnas.org�cgi�doi�10.1073�pnas.0605669103

Dow

nloa

ded

by g

uest

on

Janu

ary

29, 2

021

us to more definitively address the possibility of EMT duringfibrogenesis.

Generation and Verification of Triple Transgenic Reporter Mice. Tri-ple transgenic reporter mice containing at least one copy of thehuman SPC-rtTA, tetO-CMV-Cre (10, 11) and floxed ROSA26(12) transgenes (Fig. 2A) were generated by crossing, resultingin genetically tagged lung epithelial cells that permanentlyexpress �-gal. Mice were given doxycycline throughout gestation,and lung epithelial cell-specific expression of �-gal was con-firmed by several techniques. Frozen lung sections stained withX-gal revealed robust �-gal expression in AECs (Fig. 2C),whereas mesenchymal structures within the lung (smooth musclelayers within vessels and airways) were completely X-gal-negative. Sections from kidney, liver, and muscle also revealedabsence of X-gal staining. Protein lysates from these organs wereanalyzed by immunoblot demonstrating expression of �-gallimited to the lung (Fig. 2B). Primary AECs isolated from tripletransgenic mice were immediately fixed and stained with X-galfollowed by immunostaining for pro-SPC, revealing that �80–100% of the pro-SPC-positive cells are X-gal-positive, consistentwith prior reports of in vivo Cre-mediated recombination effi-ciency (13). AECs from triple transgenic mice were cultured onMatrigel (Mg), which is primarily composed of laminins andcollagen IV in media containing 10 ng�ml KGF (14) for 1 week,then stained with X-gal followed by immunostaining. AECscultured on Mg form tight clusters and maintain type II AECmarkers, such as pro-SPC, as has been reported. X-gal-positiveAECs all stain for pro-SPC and other epithelial markers, such asE-cadherin (Fig. 2 and Fig. 8, which is published as supportinginformation on the PNAS web site). Contaminating fibroblastsof the type II isolation are �-SMA-positive, E-cadherin-negative,and are all X-gal-negative. Thus, the triple transgenic reportermouse system results in lung epithelial-specific expression of�-gal. Littermate control mice lacking any of the three trans-genes revealed no expression of �-gal by these techniques.

AEC �-gal Reporter Mice Reveal EMT in Vivo. These reporter micewere next examined in a well described, cytokine-mediatedmodel of pulmonary fibrosis (15). Triple transgenic mice weretreated with intranasal instillation of AdTGF-�1 (2.5 � 108 pfuper mouse) in 30 �l of PBS or PBS alone. After 10 days, singlecell suspensions of AECs were prepared from injured andcontrol lungs. Cells were cultured on Mg-coated slides for 1

week, then stained with X-gal. AECs form tight clusters of cellsthat are X-gal-positive. Mesenchymal cells from control lungsare uniformly X-gal-negative (Fig. 2). In contrast, a subset ofX-gal-positive cells from injured lungs demonstrated a distinctelongated mesenchymal phenotype, forming stress fibers, filop-odia and lamellipodia (Fig. 9, which is published as supportinginformation on the PNAS web site).

To confirm and quantify the presence of X-gal-positive epi-thelial-derived mesenchymal cells, adult triple transgenic micewere given either AdTGF-�1 or vehicle alone by intranasalinstillation. After 21 days, lungs were analyzed for evidence offibrosis and EMT. Lung sections revealed moderate fibrosis andwhole lung lysate demonstrated increased expression of mesen-chymal proteins (Fig. 9) consistent with reports using this virus(15). Clusters of X-gal-positive cells are apparent within areas ofcollagen deposition in injured mice. To determine whether anyof these X-gal-positive cells were fibroblasts, lung sections werestained with X-gal followed by immunostaining (Fig. 3). Clustersof X-gal-positive cells are pro-SPC-negative and express themyofibroblast marker �-SMA (Fig. 3 A and B), demonstratingEMT in situ. Cellular costaining of X-gal and �-SMA wasconfirmed by confocal microscopy (not shown). To eliminatefurther the possibility that areas of costaining are due tooverlapping cells rather than cells expressing both �-gal and�-SMA, single cell suspensions were prepared from control andinjured lungs, then immediately fixed and stained with X-gal,followed by immunostaining for pro-SPC and �-SMA. Cells that

Fig. 1. Evidence of early EMT in IPF lung. IPF lung biopsy demonstrating cellsremoved from the alveolar basement membrane within the interstitiumcoexpressing pro-SPC and the mesenchymal protein, N-cadherin (arrows).

Fig. 2. Characterization of reporter mice expressing AEC �-gal. (A) Tripletransgenic mice express rtTA in lung epithelial cells using the SPC promoter. Inthe presence of doxycycline (dox), rtTA is an active transcriptional factorleading to expression of Cre recombinase and removal of the floxed poly(A)sequence in the ROSA allele. Thus, in cells expressing SPC, expression of lacZ isconstitutively and irreversibly activated. (B) Immunoblot of lysate from dif-ferent tissues demonstrating lung-specific expression of �-gal. (C) Expressionof �-gal in a lung section by X-gal staining. (D and E) Type II AECs isolated fromtriple transgenic mice, maintained on Mg for 1 week, then stained with X-gal(E) followed by immunostaining for E-cadherin and �-SMA (D). X-gal-positiveepithelial cells form clusters and stain for E-cadherin. Contaminating fibro-blasts are X-gal-negative and �-SMA-positive.

Kim et al. PNAS � August 29, 2006 � vol. 103 � no. 35 � 13181

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Janu

ary

29, 2

021

are clearly X-gal-positive, �-SMA-positive, and pro-SPC-negative (Fig. 3 C and D) are identified from mice treated withAdTGF-�1 representing in vivo EMT. Lung sections and singlecell suspensions from control mice do not show any X-gal-positive, �-SMA-positive cells. Because �-SMA expression islimited to a subset of fibroblasts (2), not all EMT-derivedfibroblasts may stain for �-SMA. Indeed, �5% of X-gal-positivecells from injured mice were �-SMA-positive.

To quantify the degree of EMT, single-cell suspensions frominjured and uninjured lungs were also stained with vimentin, amore general fibroblast marker (16). Staining of murine fibro-blasts and primary murine alveolar epithelial with this antibodyrevealed excellent specificity for fibroblasts (Fig. 10, which ispublished as supporting information on the PNAS web site).Many vimentin-positive cells are identified in injured lungs,approximately a third of which are also X-gal-positive (Fig. 3E–G). There is an overall 2-fold increase in vimentin-positivecells 21 days after AdTGF-�1 instillation. Surprisingly, X-gal-positive epithelial-derived cells account for nearly all of this invivo increase in vimentin-positive cells. Lung sections and singlecell suspensions from littermate control mice which were tetO-Cre-positive and ROSA26-positive but lacked the SPC-rtTA

transgene (n � 4) revealed no X-gal-positive cells 3 weeks afterAdTGF-�1, excluding the possibility that AdTGF-�1 couldinduce native �-gal activity or lead to nonspecific recombinationof the ROSA26 transgene.

Provisional Matrices Initiate EMT. Inspection of injured lungs 21days after exposure to active TGF-�1 indicated that clusters ofX-gal-positive, �-SMA-positive cells are frequently locatedwithin areas of dense fibronectin (Fn) deposition (Fig. 4 A andB). This physical association may result from Fn deposition byEMT-derived cells or from Fn-driven EMT. To test the hypoth-esis that matrix may regulate EMT, AECs were isolated fromuninjured triple transgenic mice and cultured on Mg or Fn. After1 week, cells were stained with X-gal, then immunostained. OnMg, no cells costaining for X-gal and �-SMA were found (Fig.2 D and E). On Fn, the AECs undergo dramatic phenotypicchanges with formation of stress fibers, lamellipodia and filop-odia; numerous �-SMA-positive cells are also X-gal-positive(Fig. 4 C and D) representing epithelial-derived myofibroblasts.Quantification of cells on these surfaces staining for X-gal and�-SMA (Fig. 4E) demonstrates that this process is robust; �40%of all �-SMA-positive cells were X-gal-positive. X-gal-positivecells maintained on Fn for 1 week are virtually all vimentin-positive (Fig. 10) and demonstrate very weak or absent stainingfor epithelial markers, pro-SPC and E-cadherin (Fig. 8). Immu-noblot analysis demonstrates that changes in protein expression

Fig. 3. EMT develops in vivo after TGF-�1-mediated lung injury. (A and B)The same field (�60) of a lung 3 weeks after intranasal AdTGF-�1 stained byX-gal (B) then immunostained (A) for �-SMA and pro-SPC. Several X-gal-positive, pro-SPC-positive cells are demonstrated (arrowheads) as well as acluster of X-gal-positive, �-SMA-positive, and pro-SPC-negative cells (arrow).(C and D) The same field of a whole lung single-cell suspension from an injuredlung stained by X-gal (D) then immunostained (C) for �-SMA and pro-SPC. AnX-gal-positive, �-SMA-positive, pro-SPC-negative cell is demonstrated (ar-row). (E and F) The same field of a whole lung single cell suspension from atriple transgenic mouse three weeks after AdTGF-�1, stained by X-gal (F), andimmunostained (E) for vimentin. An X-gal- and vimentin-positive cell is indi-cated by arrow. (G) Quantification X-gal- and vimentin-positive single cellsfrom of AdTGF-�1-injured and control lungs (n � 3 per group).

Fig. 4. Fn drives AEC EMT ex vivo. (A and B) The same field (�60) of a lung3 weeks after intranasal AdTGF-�1 stained by X-gal (B), then immunostained(A) for �-SMA (green) and Fn (red). A cluster of X-gal- and �-SMA-positive cellswithin an area of Fn deposition is demonstrated (arrow). (C and D) The sameview of isolated AECs from an uninjured triple transgenic mouse cultured onFn-coated slides for 1 week, then stained with X-gal (D) then immunostained(C) for �-SMA. Several X-gal- and �-SMA-positive cells are shown (arrows) aswell as several X-gal-negative, �-SMA-positive cells (*). (E) Quantification ofprimary AECs from triple transgenic mice maintained on Fn- or Mg-coatedslides for 1 week and stained for X-gal and �-SMA. (F) Primary murine AECswere cultured on Fn- or Mg-coated plates for 2 days and then lysed andanalyzed by immunoblot.

13182 � www.pnas.org�cgi�doi�10.1073�pnas.0605669103 Kim et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

29, 2

021

begin within 2 days of culture; AECs maintained on Fn haveincreased expression of mesenchymal proteins, N-cadherin,�-SMA and PAI-1, and a marked decline in pro-SPC expressioncompared to the same pool of isolated AECs cultured on Mg(Fig. 4F). To determine whether other components of theprovisional matrix could drive EMT, primary AECs from tripletransgenic mice were cultured on surfaces coated with fibrin geland mixtures of fibrin and Fn. Primary AECs were found toundergo a similar EMT on both fibrin-coated and Fn�fibrin-coated matrices (not shown). Thus, culturing on either of the twomajor components of the provisional matrix is sufficient to driveAEC EMT.

What is the mechanism by which provisional matrix proteinspromote EMT ex vivo? Because activation of TGF-�1 is criticalto the development of tissue fibrosis in several settings, weconsidered the possibility that active TGF-�1 is preferentiallyproduced by cells contacting provisional matrix proteins. How-ever, AECs plated on Mg and Fn secreted equal levels of totalimmunoreactive TGF-�1 and no detectable active TGF-�1 (notshown). Fn-driven EMT also was not blocked by adding highconcentrations of a soluble TGF-�1 receptor (TGF-�RII-Fc) orantibodies against TGF-�1 (1D11), indicating that release ofactive TGF-�1 could not explain the process of Fn-driven EMT(not shown). However, primary AECs cultured on Fn for 2 daysreveal robust levels of p-Smad2 by immunoblot (Fig. 5 A and B)and nuclear localization of Smad2�3 by immunostaining (Fig. 11,which is published as supporting information on the PNAS website), consistent with activation of endogenous TGF-�1 signal-ing. In contrast, AECs cultured on Mg for 2 days are quiescent,revealing no detectable p-Smad2 and only cytoplasmic Smad2�3.Addition of a low molecular weight inhibitor of TGF-�1 receptorkinase (SB431542) largely blocked the morphological responsesof cells to Fn, inhibited expression of �-SMA, and allowed themaintenance of pro-SPC expression, indicating the persistenceof an epithelial phenotype on Fn (not shown). These resultsstrongly suggested that latent TGF-�1 was simultaneously beingactivated and sequestered, a mechanism previously ascribed tolatent TGF-�1 activation through the integrin �v�6 (17). Inte-grin �v�6 expression is limited to epithelial cells and known tobe required in animal models of lung fibrosis (18). To test thishypothesis, AECs were isolated from �6-null mice or wild-typecontrols and examined for their responses to culture on Fn.

Whereas wild-type cells underwent the expected mesenchymaltransition on Fn, �6-null AECs did not translocate Smad2�3 tothe nucleus, developed little �-SMA, and maintained a pro-SPC-positive epithelial phenotype (Fig. 5). Thus, Fn itself leads to�v�6-dependent activation of latent TGF-�1 with a resultantrobust EMT. This finding correlates with the apparent distribu-tion of epithelial-derived cells devoid of pro-SPC in foci of Fnaccumulation in vivo during matrix remodeling and indicates thatat least one component of the provisional lung matrix alone caninitiate EMT.

Matrigel Promotes AEC Apoptosis in Response to TGF-�1. To inves-tigate the response of AECs on Mg to exogenous TGF-�1,primary AECs from triple transgenic reporter mice were isolatedand cultured on Mg or Fn. After 1 day, TGF-�1 (4 ng�ml) wasadded to some wells. After an additional 2 days, cells wereanalyzed for apoptosis by TUNEL assay. Cells on Mg treatedwith TGF-�1 revealed a marked increase in TUNEL-positivecells compared to cells on Mg without TGF-�1 or cells of Fn withor without TGF-�1 (Fig. 6B). The contaminating fibroblastsfrom the AEC isolation were resistant to TGF-�1-inducedapoptosis (Fig. 6A). One week after TGF-�1 treatment, cellswere stained with X-gal followed by immunostaining for �-SMA.On Fn, exogenous TGF-�1 had no effect on the number ofX-gal-positive or �-SMA-positive cells. In contrast, there was adramatic loss of X-gal-positive AECs on Mg in the presence ofexogenous TGF-�1. No X-gal-positive, �-SMA-positive cellswere identified within Mg-coated wells treated with TGF-�1(Fig. 6 C and D).

DiscussionType II AECs have long been considered the precursor ofalveolar type I cells, and are known to possess proliferativepotential (19). In this report, we are able to address several majorpoints of uncertainty in identifying AECs as precursors tofibroblast-like cells and a mechanism regulating this transition.By using genetically modified mice in which AEC fate can betracked, we have been able to identify epithelial-derived mes-enchymal cells, in vivo and ex vivo, that have lost epithelialfeatures and would have otherwise been indistinguishable fromother mesenchymal cells. We verified histological costaining

Fig. 5. AEC EMT requires �v�6-dependent latent TGF-�1 activation. (A)Primary murine AECs were cultured for 2 days on Mg�Col or Fn, then lysed andanalyzed by immunoblot for p-Smad2 and total Smad2�3. (B) Primary AECsfrom wild-type and �6-null mice were cultured on Fn � SB431542 (10 �M) for2 days, then lysed and immunoblotted for p-Smad2 and total Smad2�3. (C andD) AECs from wild-type (C) and �6-null (D) mice were cultured on Fn for 4 days,then stained for �-SMA and pro-SPC.

Fig. 6. TGF-�1-induced AEC apoptosis on Mg but not Fn. (A) Overlay of phasewith TUNEL (green) and Dapi (blue) staining of AECs plated on Mg and treatedwith TGF-�1 (4 ng�ml) for 2 days. (B) Quantification of TUNEL-positive cells. (Cand D) AECs from triple transgenic mice cultured on Mg then stimulated withTGF-�1 for 1 week, stained with X-gal (C) followed by immunostaining for�-SMA (D). A remaining X-gal-positive cell is shown (arrow) as well as severalapoptotic remnants (arrow heads).

Kim et al. PNAS � August 29, 2006 � vol. 103 � no. 35 � 13183

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Janu

ary

29, 2

021

with staining from single-cell suspensions to definitively elimi-nate the possibility of overlapping cells leading to apparentcostaining. Our observations that these cells undergo EMT bothex vivo and in vivo provides direct evidence that AECs, and typeII cells in particular, are progenitors for mesenchymal cells,which can contribute significantly to the pool of expandedfibroblasts after lung injury. AECs appear to be positioned in thealveolus to respond to injurious events by both proliferating toreplace alveolar type I cell loss and by transdifferentiating topromote repair.

The model system described here not only reveals the intrinsicpotential of AECs for mesenchymal transdifferentiation, as hadbeen suspected (2, 8), but also allows quantification of howrobust this process may be and provides insight in the mecha-nisms involved. In the context of AdTGF-�1-mediated lunginjury, most of the expansion of vimentin- and �-SMA-positivecells appears to derive from epithelial cells at 21 days after theinitiation of injury. Although more subjective, inspection of thehistopathology of these injured lungs indicates that one commonsite of epithelial-derived fibroblasts in vivo is at sites of Fnaccumulation (Fig. 4). This observation could result from Fn-deposition by EMT-derived cells, or more interestingly, suggestthat the provisional matrix could activate AEC transdifferentia-tion. The ability of provisional matrix proteins to drive EMT wasdemonstrated in our ex vivo system. AECs maintained on Fn orfibrin, major components of the provisional matrix, surprisinglyundergo spontaneous EMT without the need for exogenousTGF-�1, unlike prior reports of other in vitro models of EMT (8,9). Addition of TGF-�1 to cells on Fn had virtually no effectbecause of robust �v�6 integrin-dependent activation of endog-enous TGF-�1 signaling. This observation alone invites inter-pretation of a large body of prior work on cultured AEC.Although AECs cultured on Fn have been known to spread andchange their phenotype for many years (20), our data reveal thatthis is actually EMT and is driven by a specific signaling pathwayinitiated by engagement of fibrin or Fn. In contrast, AECsmaintained on Mg, composed primarily of alveolar basementmembrane components, do not activate TGF-�1 and maintainan AEC phenotype.

TGF-�1 signaling has been implicated in both AEC apoptosis(21) and EMT (8, 9), making the in vivo response of AECs toTGF-�1 during fibrogenesis unclear. Indeed, epithelial celldeath is a prominent feature of fibrotic lung diseases, and ourcurrent data support a prominent role for AEC EMT, suggestingthat both responses occur during fibrogenesis and that thedeterminants of the AEC response to TGF-�1 may be critical tothe progression of fibrogenesis. Prior obscurity on the AECresponse to TGF-�1 stems from the limitation of previoustechniques. Immortalized AEC-like cells may have a blunted orabsent apoptotic response, potentially blocking an importantphysiologic response to TGF-�1. Culture of primary AECs witheven a small amount of fibroblast contamination may ultimatelyresult in a pure fibroblast culture through either EMT orpreferential epithelial cell death. By using a system in whichprimary AECs are permanently tagged, it is clear that, at leastex vivo, the content of the matrix contacting AECs determinesthe response to TGF-�1. Treatment of AECs on Mg withexogenous TGF-�1 leads to AEC apoptosis, whereas AECs onFn undergo EMT through activation of endogenous TGF-�1.Collectively, these findings suggest a model of fibrosis in whichan initiating stimulus leads to AEC apoptosis and enhancedalveolar permeability, then later, AECs within a Fn�fibrin-richprovisional matrix, undergo EMT and directly contribute tofibrosis. This finding is consistent with previous reports thatdemonstrate that alveolar accumulation of Fn precedes fibrosis,implicating Fn in the accumulation of fibroblasts in IPF and inanimal models of pulmonary fibrogenesis (22). The combined invivo and ex vivo findings in this report provide a clear link

between formation of a provisional matrix in response to injury,focal activation of the repair process, and accumulation ofEMT-derived fibroblasts.

To what extent does EMT contribute to the progression offibrotic lung disease in humans? Elevated levels of activeTGF-�1 and deposition of a provisional matrix rich in Fn andfibrin are prominent components of fibrotic lung disease, sup-porting the potential for EMT by the mechanism described inthis report. Immunostaining of IPF lung biopsies also supportsthe possibility of EMT (ref. 8 and Figs. 1 and 7) and isolatedprimary human type II cells undergo an apparently identicalconversion on Fn (not shown). However, detection of putativetransitional cells in vivo by immunostaining requires cells toexpress detectable levels of epithelial markers. Our data (Fig. 4)show that AECs rapidly lose these markers on Fn. If EMT-derived fibroblasts are found to have a unique expression profile,they might be distinguished from other fibroblasts, allowingfurther validation and quantification of EMT in humans.

The high prevalence of EMT during experimental lung injurysuggests that this is a previously unrecognized component of anormal remodeling process. If so, what normally limits andreverses this process? Preliminary observations indicate thatEMT-derived fibroblasts are more prone to cell death thanresident lung fibroblasts in culture, suggesting a mechanism forlimiting physiologic repair after injury and inviting the specula-tion that, in pathologic conditions, an inherent tendency ofEMT-derived fibroblasts to die may be suppressed leading todysregulated and progressive fibrosis. Elucidation of additionaldeterminants of AEC EMT and their subsequent cell survivaland function could provide new therapeutic insights for fibroticdisorders.

Materials and MethodsGenetically Modified Mice. The conditional ROSA26 reportermouse (12), was a generous gift from Brian Black (University ofCalifornia, San Francisco, CA). Briefly, this reporter mousecontains a transgene (ROSA) consisting of a lacZ gene and anupstream floxed triple polyadenylation [poly(A)] sequence pre-venting transcription of the lacZ gene. The poly(A) sequence isf lanked by loxP sites subjecting it to removal by Cre recombinase,resulting in transcription and constitutive expression of lacZ incells expressing Cre. Lung epithelial specific recombinationbetween the loxP sites was achieved by using the tetO-CMV-Cretransgene and the surfactant protein-C promoter-reverse tetra-cycline transactivator transgene (SPC-rtTA) (10). Compoundmice containing at least one copy of all three transgenes wereobtained by breeding and were genotyped by PCR (10–12).Pregnant females were maintained on doxycycline containingwater (1 mg�ml) throughout gestation. Freshly frozen lungsamples for staining were obtained from mice anesthetized witha mixture of ketamine and xylazine. Mice were exsanguinatedand then perfused with PBS. Lungs were lavaged twice with PBSand then frozen in OCT for sectioning and staining. Lysate fromlung, kidney, and liver were routinely screened by immunoblotto ensure lung-specific expression of �-gal in all mice used forexperiments. Integrin �6-null mice and matched controls weregenerated as described (17).

X-gal Staining. Expression of lacZ in 5- to 7-�m-thick freshlyfrozen tissue sections or isolated primary AECs was determinedby X-gal staining (23). In some samples, X-gal staining wasfollowed by immunofluorescent staining (23) or picosirius redcollagen staining performed by the University of CaliforniaResearch Morphology Core Facility (San Francisco, CA). Forcellular quantification, 300 cells per sample stained for �-gal andfor mesenchymal proteins were visualized and quantified inde-pendently by two investigators in blinded fashion. For evaluationof group differences, the two-tailed Student’s t test was used

13184 � www.pnas.org�cgi�doi�10.1073�pnas.0605669103 Kim et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

29, 2

021

assuming equal variance. A P value of �0.05 was accepted assignificant.

Type II Cell Isolation and Culture. Murine alveolar epithelial type IIcells were isolated from adult mice (6–12 weeks old) followingthe method of Corti et al. (20) with minor modifications and werecultured on tissue culture plates and chamber slides coated withMg, supplemented with 5% collagen type I and 25% SABM(Cambrex, East Rutherford, NJ) or coated with 100 �g�ml Fnin PBS. Cells were maintained in SAGM without hydrocortisonecontaining 5% charcoal�dextran-treated FBS and 10 ng�mlKGF in a 37°C, 5% CO2 incubator as described (14). For furtherdetails, see Supporting Text, which is published as supportinginformation on the PNAS web site.

Whole Lung Single-Cell Suspensions. Single-cell suspension wasobtained from murine lungs as described (24) with minormodifications. Cells were either cultured or immediately fixedand stained with X-gal in suspension. Cells stained with X-gal insuspension were then centrifuged onto superfrost-coated slides(Fisher) and immunostained. For further details, see SupportingText.

Intranasal Adeno-TGF-�1. A replication-deficient adenovirus en-coding a mutated TGF-�1 (AdTGF-�1) as described (15) was agenerous gift from Jack Gauldie (McMaster University, Ham-ilton, ON, Canada). This adenovirus encodes porcine TGF-�1with two point mutations (C223S and C225S) that preventTGF-�1 binding to latency-associated protein and thus producesa constitutively active TGF-�1. Adult (6–12 weeks old) triple

transgenic mice were lightly anesthetized with isoflurane, theninstilled with 2.5 � 108 pfu of AdTGF-�1 in 30 �l of PBS throughintranasal delivery. This dose was empirically derived based onpreliminary experiments and prior reports. Littermate controltriple transgenic mice received PBS alone. Mice were killed andanalyzed 10–21 days after AdTGF-�1 instillation, as indicated.

Immunohistochemistry. Cryosections (5–7 �m) of tissue and cul-tured cells were stained by immunofluorescence as described(17, 25). Stained sections were visualized on a Nikon fluorescentmicroscope, and images were captured with a SPOT 2.3.1 camera(Diagnostic Instruments, Sterling Heights, MI) and analyzed byusing SPOT 4.0.9 software (Diagnostic Instruments).

Immunoblot. Tissue and cells were lysed in RIPA buffer (150 mMNaCl�50 mM Tris, pH 8.0�1% Triton X-100�0.5% sodium deoxy-cholate�0.1% SDS, supplemented with protease and phosphataseinhibitors) and analyzed by immunoblot as described (25).

TUNEL Assay. Assessment of AEC apoptosis was performed byTUNEL assay (Roche, Basel, Switzerland) according to packageinsert. Percentages of TUNEL-positive cells were quantifiedfrom five low-power fields per sample.

We thank Jack Gauldie and Jeffrey Whitsett (Cincinatti Children’sHospital Medical Center, Cincinatti, OH) and Paul Weinreb (BiogenIdec, Cambridge, MA) for contributing valuable reagents for thesestudies. This work was supported in part by National Institutes of HealthGrants GM-075419 (to K.K.K.), HL-04055 (to P.J.W.), HL-53949 (toD.S.), and HL-44712 (to H.A.C.).

1. American Thoratic Society & European Respiratory Society (2002) Am. J.Respir. Crit. Care Med. 165, 277–304.

2. Kalluri, R. & Neilson, E. G. (2003) J. Clin. Invest. 112, 1776–1784.3. Phillips, R. J., Burdick, M. D., Hong, K., Lutz, M. A., Murray, L. A., Xue, Y. Y.,

Belperio, J. A., Keane, M. P. & Strieter, R. M. (2004) J. Clin. Invest. 114,438–446.

4. Moore, B. B., Kolodsick, J. E., Thannickal, V. J., Cooke, K., Moore, T. A.,Hogaboam, C., Wilke, C. A. & Toews, G. B. (2005) Am. J. Pathol. 166, 675–684.

5. Zeisberg, M., Hanai, J., Sugimoto, H., Mammoto, T., Charytan, D., Strutz, F.& Kalluri, R. (2003) Nat. Med. 9, 964–968.

6. Thannickal, V. J., Toews, G. B., White, E. S., Lynch, J. P., III, & Martinez, F. J.(2004) Annu. Rev. Med. 55, 395–417.

7. Kasai, H., Allen, J. T., Mason, R. M., Kamimura, T. & Zhang, Z. (2005) Respir.Res. 6, 56.

8. Willis, B. C., Liebler, J. M., Luby-Phelps, K., Nicholson, A. G., Crandall, E. D.,du Bois, R. M. & Borok, Z. (2005) Am. J. Pathol. 166, 1321–1332.

9. Yao, H. W., Xie, Q. M., Chen, J. Q., Deng, Y. M. & Tang, H. F. (2004) LifeSci. 76, 29–37.

10. Mucenski, M. L., Wert, S. E., Nation, J. M., Loudy, D. E., Huelsken, J.,Birchmeier, W., Morrisey, E. E. & Whitsett, J. A. (2003) J. Biol. Chem. 278,40231–40238.

11. Tichelaar, J. W., Lu, W. & Whitsett, J. A. (2000) J. Biol. Chem. 275,11858–11864.

12. Soriano, P. (1999) Nat. Genet. 21, 70–71.13. Lewandoski, M. (2001) Nat. Rev. Genet. 2, 743–755.

14. Rice, W. R., Conkright, J. J., Na, C. L., Ikegami, M., Shannon, J. M. & Weaver,T. E. (2002) Am. J. Physiol. 283, L256–L264.

15. Bonniaud, P., Kolb, M., Galt, T., Robertson, J., Robbins, C., Stampfli, M.,Lavery, C., Margetts, P. J., Roberts, A. B. & Gauldie, J. (2004) J. Immunol. 173,2099–2108.

16. Radisky, D. C., Levy, D. D., Littlepage, L. E., Liu, H., Nelson, C. M., Fata, J. E.,Leake, D., Godden, E. L., Albertson, D. G., Nieto, M. A., et al. (2005) Nature436, 123–127.

17. Munger, J. S., Huang, X., Kawakatsu, H., Griffiths, M. J., Dalton, S. L., Wu,J., Pittet, J. F., Kaminski, N., Garat, C., Matthay, M. A., et al. (1999) Cell 96,319–328.

18. Breuss, J. M., Gallo, J., DeLisser, H. M., Klimanskaya, I. V., Folkesson, H. G.,Pittet, J. F., Nishimura, S. L., Aldape, K., Landers, D. V., Carpenter, W., et al.(1995) J. Cell Sci. 108, 2241–2251.

19. Adamson, I. Y. & Bowden, D. H. (1974) Lab. Invest. 30, 35–42.20. Corti, M., Brody, A. R. & Harrison, J. H. (1996) Am. J. Respir. Cell Mol. Biol.

14, 309–315.21. Solovyan, V. T. & Keski-Oja, J. (2006) J. Cell Physiol. 207, 445–453.22. Limper, A. H. & Roman, J. (1992) Chest 101, 1663–1673.23. Snyder, E. Y., Deitcher, D. L., Walsh, C., Arnold-Aldea, S., Hartwieg, E. A.

& Cepko, C. L. (1992) Cell 68, 33–51.24. Kim, C. F., Jackson, E. L., Woolfenden, A. E., Lawrence, S., Babar, I., Vogel,

S., Crowley, D., Bronson, R. T. & Jacks, T. (2005) Cell 121, 823–835.25. Zhang, F., Tom, C. C., Kugler, M. C., Ching, T. T., Kreidberg, J. A., Wei, Y.

& Chapman, H. A. (2003) J. Cell Biol. 163, 177–188.

Kim et al. PNAS � August 29, 2006 � vol. 103 � no. 35 � 13185

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Janu

ary

29, 2

021