American Journal ofPathology, Vol. 150, No. 3, March 1997Copyright X) American Societyfor Investigative Pathology

Transforming Growth Factors-f31, -132, and -f3Stimulate Fibroblast Procollagen Production inVitro but Are Differentially Expressed duringBleomycin-Induced Lung Fibrosis

Robina K. Coker,* Geoffrey J. Laurent,*Shahriar Shahzeidi,* Penny A. Lympany,tRoland M. du Bois,t Peter K. Jeffery,t andRobin J. McAnulty*From the Centrefor Cardiopulmonary Biochemistry andRespiratory Medicine,* University College London MedicalSchool, and the National Heart and Lung Institute,tImperial College, London, United Kingdom

Transforming growth factor (TGF)-f81 may po-tentiate wound healing andfibrosis by stimulat-ingfibroblast coUagen deposition. TGF-131 is im-plicated in the pathogenesis of pulmonaryf-brosis, but the role of TGF-182 and TGF-133 re-mains unclear. We examined their effects on lungfibroblast procoUagen metabolism in vitro andlocalized their gene expression during bleomy-cin-induced lungfibrosis using in situ hybridiza-tion with digoxigenin-labeled riboprobes. Althree isoforms stimulatedflbroblastprocollagenproduction. TGF-f33 was the mostpotent and alsoreduced procoUagen degradation. In normalmouse lung, TGF-f31 and TGF-183 mRNA tran-scripts were abundant in bronchiolar epithe-lium. After bleomycin, TGF-813 gene expressionwas maximaly enhanced at 10 days, with thesignal being predominant in macrophages. Sig-nal was also enhanced in mesenchymal, pulmo-nary endothelial, and mnesothelial cells. After 35days, the pattern of TGF-.81 gene expression re-turned to that of control lung. TGF-183 gene ex-pression remained unchanged throughout com-pared with controls. TGF-182 mRNA was notdetected with the antisense probe, but signal ob-tained with the sense probe suggests the pres-ence of a naturaly occurrng antisense. Thisstudy demonstrates that TGF-j81, -182, and -183 aUexert profibrotic effects in vitro. However,TGF-18 isoform gene expression is differentialy

controUed during experimentalpulmonaryfibro-sis with TGF-.81 the predominant isoform ex-pressed during pathogenesis. (Am J Pathol1997, 150:981-991)

The pathogenesis of pulmonary fibrosis remains in-completely understood. One current hypothesis isthat initial endothelial or epithelial cell injury triggersan influx of inflammatory cells from the circulation.Cytokines derived from these inflammatory cells aswell as from resident cells then stimulate fibroblaststo synthesize excessive amounts of extracellular ma-trix, including collagen.1One such group of cytokines is the transforming

growth factor (TGF)-f family. Three mammalianTGF-,B isoforms are now recognized, TGF-,B1, -12,and -13g. TGF-,1l is an extremely potent promoter ofextracellular matrix accumulation and acts via bothtranscriptional and post-transcriptional mecha-nisms.2 The effects of TGF-12 and TGF-f33 on fibro-blast collagen synthesis and degradation have notbeen studied.

There is now considerable evidence implicatingTGF-,B1 in the pathogenesis of pulmonary fibrosis.TGF-f1 and its mRNA levels increase during thedevelopment of experimentally induced lung fibro-SiS,3'4 and TGF-11 antibodies attenuate the fibroticresponse in the bleomycin mouse model5 and inimmune-induced lung fibrosis.6 TGF-j1 protein syn-thesis is increased in patients with idiopathic pulmo-

Supported by the Wellcome Trust (UK), the Arthritis and Rheuma-tism Council (UK), the British Lung Foundation, and the BritishCouncil.

Accepted for publication October 30, 1996.

Address reprint requests to Dr. Robina K. Coker, Centre forCardiopulmonary Biochemistry and Respiratory Medicine, TheRayne Institute, University College London Medical School, 5 Uni-versity Street, London WC1E 6JJ, UK.

981

982 Coker et alAJP Warcb 1997, Vol. 150, No. 3

nary fibrosis7'8 and with other interstitial lung diseas-es.9

Differential TGF-,B isoform expression during de-velopment,1 O-12 in normal adult tissues, 13-15 and indisease16,1 suggests that TGF-031 3 may each havedifferent functions in vivo. However, the role ofTGF-f32 and TGF-f33 in the pathogenesis of pulmo-nary fibrosis is at present controversial. Three stud-ies of bleomycin-induced pulmonary fibrosis sug-gest that TGF-P1 is the predominant isoforminvolved,18-20 whereas another proposes that allthree isoforms may be implicated.21

The aim of this study was to examine the role ofTGF-f32 and TGF-f33 in the pathogenesis of pulmo-nary fibrosis. We initially investigated whetherTGF-f32 and TGF-f33 shared the ability of TGF-f31 topromote extracellular matrix accumulation in vitro.Dose-response studies showed that TGF-f31l3 allstimulate fibroblast procollagen synthesis. TGF-f3was the most potent isoform in this assay and alsoreduced intracellular procollagen degradation. Hav-ing established that all three isoforms are potentiallyprofibrotic in vitro, we examined their gene expres-sion by in situ hybridization in a murine model of lungfibrosis. Our results show that gene expression of thethree isoforms is differentially regulated during thedevelopment of bleomycin-induced lung fibrosis.TGF-f31 but not TGF-132 or TGF-P3 gene expressionwas enhanced after bleomycin, suggesting thatTGF-f31 is the predominant isoform implicated in thepathogenesis of this disease.

Materials and Methods

Cell CultureProcollagen production was determined using pre-viously published methods.2 Human fetal lung fibro-blasts (HFL-1, American Type Culture Collection,Rockville, MD) were cultured in 12-well plates withDulbecco's modified Eagle's medium plus 5% new-born calf serum in a humidified atmosphere contain-ing 10% CO2 at 37°C until confluent. Dulbecco'smodified Eagle's medium was then removed andreplaced with 1 ml of preincubation medium contain-ing 4 mmol/L glutamine, 50 ,ug/ml ascorbic acid, 0.2mmol/L proline, and 2% newborn calf serum. After 24hours, the preincubation medium was removed andreplaced with 1 ml of fresh preincubation mediumcontaining one of the TGF-f3 isoforms at concentra-tions ranging from 0.05 to 5 ng/ml (2 to 200 pmol/L).Control cells had medium replaced without the ad-dition of TGF-j3. TGF-f31 and TGF-f32 were naturalporcine and TGF-P33 was recombinant chicken (R&D

Systems, Abingdon, UK). Cells were then incubatedfor an additional 24 hours before harvesting. Colla-gen production has previously been shown to belinear over this period of time.22The cell layer was scraped into the medium and

aspirated. Each well was washed with 1 ml of phos-phate-buffered saline (PBS) and the washings com-bined with the initial aspirate. Ethanol was added toa final concentration of 67% (v/v) and proteins pre-cipitated at 40C overnight. The samples were thenfiltered (0.45 ,um) as described previously.23 Afterhydrolysis in 2 ml of 6 mol/L hydrochloric acid over-night at 1100C and subsequent charcoal filtration(0.65 ,um), hydroxyproline was derivatized with7-chloro-4-nitrobenzo-2-oxa-1,3-diazole and sepa-rated using high pressure liquid chromatography.24Hydroxyproline content was then determined bycomparison with standards containing knownamounts of hydroxyproline and derivatized under thesame conditions. The linearity of the hydroxyprolineassay has previously been established between 5pmol and 20 nmol.24 Procollagen production wascalculated from the quantity of hydroxyprolinepresent in the ethanol-insoluble fraction. The propor-tion of newly synthesized procollagen degraded in-tracellularly (expressed as a percentage of total pro-collagen synthesis) was calculated from the quantityof hydroxyproline in the ethanol-soluble fractioncompared with that in both the ethanol-soluble andinsoluble fractions.2 Total procollagen synthesis isrepresented by the sum of hydroxyproline in theethanol-soluble and -insoluble fractions.

Animals

Adult mice (strain B6D2F1) aged 8 weeks and weigh-ing 24 to 26 g received a single dose of intratrachealsaline (0.14 mol/L) or saline containing bleomycinsulfate (6 mg/kg) in a volume of 0.05 ml and werekilled 3, 10, 21, or 35 days later by pentobarbitoneoverdose as previously described.25 Nine animalswere used in each group: six for hydroxyproline es-timations and three for in situ hybridization studies.Lungs were fixed by intratracheal instillation offreshly prepared 4% paraformaldehyde in PBS at apressure of 25 cm H20. The trachea was ligated justcaudal to the larynx and the thoracic contents re-moved together. After 4 hours of immersion in fixa-tive, lung tissue was transferred to 15% sucrose inPBS before dehydration and embedding in paraffinwax.

For total lung collagen estimation, tissue hy-droxyproline content was measured as previouslydescribed.26 Briefly, this was determined spectro-

TGF-,B Isoforms and Lung Fibrosis 983AJP March 1997, Vol. 150, No. 3

photometrically after oxidation with chloramine T andextraction of the toluene-miscible product. Collagencontent was calculated from hydroxyproline contentassuming that lung collagen contains 12.2% w/whydroxyproline.27

Tissue PreparationTwo sections were cut from each block and stainedfor collagen with Massons' trichrome. Prehybridiza-tion treatments were performed using techniquesdescribed previously.28'29 Sections (5 gm thick)were cut and placed on slides previously coated witha 2% v/v solution of 3-aminopropyltriethoxysilane inacetone. After dewaxing, sections were rehydratedthrough a series of ethanol washes of decreasingconcentration, followed by immersion in 0.14 mol/Lsodium chloride and PBS before refixing in freshlyprepared 4% paraformaldehyde. To maximize entryof the probe into cells, sections were treated withproteinase K (Life Technologies, Paisley, UK) at aconcentration of 20 ,ug/ml in 50 mmol/L Tris hydro-chloride, pH 7.5, 5 mmol/L EDTA for 10 minutesbefore refixing with paraformaldehyde. Sectionswere then acetylated by immersion in freshly pre-pared 0.1 mol/L triethanolamine containing 0.25%acetic anhydride and subsequently dehydratedthrough a series of increasing concentrations of eth-anol.

Probe Preparation

Riboprobes were synthesized from transcript-spe-cific murine TGF-3 cDNA constructs in pGEM vec-tors (TGF-f31 and TGF-f33) and in the SP72 vector(TGF-132). The constructs were obtained by deletingthe highly conserved regions of the three murinecDNAs, and the specificity of riboprobes synthe-sized from these templates is established.1730 TheTGF-f31 construct includes nucleotides 421 through1395 of the murine DNA and contains 764 bp of theamino-terminal glycopeptide (precursor) region and210 bp of the mature region. The TGF-f32 constructcontains 442 bp of the amino-terminal glycopeptideregion (1511 through 1953). The TGF-,33 constructcontains 609 bp of the amino-terminal glycopeptideregion (831 through 1440). Digoxigenin-labeled ribo-probes were synthesized by in vitro transcriptionfrom the linearized cDNA templates using the appro-priate RNA polymerase (SP6 or T7) according to themanufacturer's protocol (Boehringer Mannheim,Lewes, UK). For Northern analysis, cDNA probeswere labeled with 32p to a specific activity of 2 x 1 o6dpm/ng by random prime labeling (Amersham,

Slough, UK) according to the manufacturer's instruc-tions.

Northern Analysis

Northern analysis was performed to confirm probespecificity. Total RNA was extracted from murinelung using previously described methods.2829 PolyA' RNA was selected with a biotinylated oligo (dT)probe (Promega, Southampton, UK). Hybrids werethen captured and washed at high stringency usingstreptavidin coupled to paramagnetic particles anda magnetic separation stand (Promega). Poly A'RNA was subsequently eluted with water. The 1 0-,ugsamples of poly A' RNA were electrophoresedthrough an agarose-formaldehyde gel and blottedonto a Hybond N nylon membrane (Amersham). Hy-bridizations were performed overnight at 420C in50% formamide, 5X saline sodium phosphate, 5XDenhardt's reagent, 0.1% sodium dodecyl sulfate(SDS), and 100 jig/ml denatured salmon spermDNA. Post-hybridization washes were performed in2X standard saline citrate (SSC), 0.1% SDS for 10minutes at 42°C twice followed by 1 X SSC, 0. 1% SDSfor 15 minutes at 420C. Hybridized probe was de-tected by autoradiography followed by scanningwith laser densitometry.

In Situ HybridizationThis protocol was based on previously describedmethods.25 Hybridization buffer consisting of 50%deionized formamide, 300 mmol/L NaCI, 20 mmol/LTris/HCI (pH 7.4), 5 mmol/L EDTA, 10 mmol/Lmonosodium phosphate (pH 8.0), 10% dextran sul-fate, 1X Denhardt's solution, and 500 ,g/ml yeasttRNA was mixed with digoxigenin-labeled probe at aratio of 9:1 to give a final probe concentration of 20ng/ml. A 25-Al aliquot of hybridization solution wasapplied to each slide, and sections were incubatedovernight (16 hours) at 500C in a sealed chamberhumidified with a solution of 50% formamide in 2XSSC. After hybridization, all incubations were per-formed at room temperature. Slides were washed in4X SSC for 30 minutes and then in 0.2X SSC for 30minutes. Slides were then washed in Tris-bufferedsaline (TBS, consisting of 0. 1 mol/L Tris, pH 8.2, 0.15mol/L NaCI) for 5 minutes. They were then incubatedfor 30 minutes with antibody blocking solution con-sisting of 5% bovine serum albumin and 5% normalsheep serum diluted in TBS with 0.1% Tween 20(Sigma, Dorset, UK). After two additional 5-minutewashes in TBS, the slides were incubated with anti-body solution for 30 minutes. This consisted of a

984 Coker et alAJPMarch 1997, Vol. 150, No. 3

1:100 dilution of anti-digoxigenin-alkaline phospha-tase Fab fragments (Boehringer Mannheim) in 1%bovine serum albumin in TBS with Tween 20. Sec-tions were then washed twice in TBS for five minuteseach.

For detection of bound antibody, sections wereincubated with alkaline phosphatase substrate NewFuschin Red (Dako, High Wycombe, UK). This wasprepared according to the manufacturer's instruc-tions, and 1 mmol/L levamisole was added to inhibitendogenous alkaline phosphatase activity.31 A200-gl aliquot of freshly prepared reagent was ap-plied to each section for 20 minutes. The slides werethen rinsed in distilled water and counterstained withhematoxylin or methyl green. They were subse-quently mounted in glycerol without dehydration. TheNew Fuschin Red yields a red color at the site of thehybridized probe. Each in situ experiment was re-peated at least twice at each time point. Sectionswere examined and reported independently by threeof the authors (R. K. Coker, P. K. Jeffery, and R. J.McAnulty).

Statistical AnalysisData were analyzed with an unpaired t-test or one-way analysis of variance for multiple comparisonsusing the Newman-Keuls procedure. Mean valuesfor collagen production, degradation, and synthesiswere held to be significantly different when the theprobability of such differences arising, assuming thenull hypothesis to be true, were less than 5% (P <0.05). Where mean values are calculated, standarderrors of the mean (SEM) are also given.

Results

In Vitro StudiesDose-response curves obtained for each of the threeTGF-,3 isoforms are shown in Figure 1. Lung fibro-blast procollagen production is shown as a functionof TGF-,3 concentration in the culture medium duringthe 24-hour incubation period. Data are the means +SEM of four to six replicate cultures per condition.The results demonstrate that all three isoforms stim-ulated fibroblast procollagen production in a dose-dependent manner. TGF-033 was the most potentisoform, exerting a maximal effect at a concentrationof 4 pmol/L, whereas the other two peptides attaineda maximal stimulation only between 20 and 40pmol/L. Concentrations above 40 pmol/L did not fur-ther stimulate procollagen production. AlthoughTGF-133 achieved maximal stimulation at a lower con-

70

60

50z

c) 401-1U)c 00* 50a0-x

30o

0 30

Q._

-~ 0c)2

° 80

av-o

o 60

0-8c)

00~

40

0

A

nfl

*

1_

B * *

C **T1I

0 2 4 20 40

TGFf Concentration (pM)Figure 1. The ejfect of TGF-,/ isoforms on lung fibroblast procollagenproduction. HFL-1 procollagen production, assessed by measurementof hydroxyproline, is shown as a function of TGF-,/ isoform concen-tration. Each value represents the mean + SEM from four to sixreplicate cultures. A: TGF-(3j. B: TGF-(32* C: TGF-P33 *P < 0.01 comrpared with control.

centration than TGF-f31 or TGF-032, the maximal re-sponse was similar, with increases of approximately50% above control values for all three isoforms.

In this set of studies, the basal rate of procollagenproduction for experiments with TGF-f32 was lowerthan that for TGF-f31 and TGF-33. The basal ratevaried between experiments but was always be-tween 27 and 45 pmol/,lg DNA/hour.

Figure 2 shows the effects of each of the TGF-,Bisoforms on lung fibroblast procollagen production,procollagen synthesis, and intracellular degradationat a concentration of 40 pmol/L, which produced

III

I

TGF-f3 Isoforms and Lung Fibrosis 985A/PMarch 1997, Vol. 150, No. 3

TGFB1 TGFB2 TGFB3

Figure 2. The effect of TGF-l, -,B2, and -f3 on lungfibroblastprocol-lagen metabolism. Each value represents the mean percentage changefrom control + SEMfrom four separate experiments, each containingfour to six replicate cultures. Open bars, procollagen production;hatched bars, procollagen synthesis; solid bars, procollagen degrada-tion. Procollagen production is calculated from the quantity of by-droxyproline present in the ethanol-insoluble fraction. Procollagendegradation (expressed as a percentage) is calculated from theamount ofprocollagen present in the ethanol-solublefraction, whereasprocollagen synthesis represents the sum of procollagen present inethanol-insoluble and ethanol-soluble fractions.

maximal stimulation of procollagen production by allthree isoforms. Data are the means ± SEM of fourexperiments, each consisting of four to six replicatecultures. TGF-f31, TGF-f2, and TGF-933 stimulatedprocollagen production by approximately 46%, 46%,and 43%, respectively (P < 0.01 in all cases). Pro-collagen synthesis was significantly increased byeach of the three isoforms. Control synthesis rose byapproximately 38%, 37%, and 33% with TGF-f1,TGF-f32, and TGF-f33 respectively (P < 0.01 in allcases). There were no significant reproducible dif-ferences between the three isoforms in terms of theireffects on either procollagen production or synthesisat this concentration. There was a tendency for allthree isoforms to reduce intracellular procollagendegradation, but this was significant only in the caseof TGF-,33 (P < 0.05).

In Vivo StudiesTable 1 shows the time course of changes in lungcollagen content during the development of bleomy-cin-induced pulmonary fibrosis. Total lung collagen

Figure 3. Northern analysis ofmurine lung with TGF-P,j3 probes. Tenmicrograms ofpoly A' RNA from murine lung was hybridized witheach ofthe three radiolabeled cDNA probes. Size markers (in kilobases)are indicated on the left.

increased with age and was slightly higher in controlanimals after 35 days. Total lung collagen in bleo-mycin-treated animals was unchanged at 3 days but10 days after treatment had risen by approximately40% (P < 0.05). It continued to rise and by 35 daysvalues were approximately 80% above those for con-trol animals (P < 0.01).

Staining of lung sections with Massons' trichromerevealed extensive fibrosis 35 days after bleomycin(data not shown). Taken together, the results of totallung collagen estimation and histological staining forlung collagen confirm that these mice developedlung fibrosis in response to bleomycin.

Figure 3 shows Northern analysis of poly A' RNAfrom murine lung. The following bands were seen:TGF-j31, 2.2 kb; TGF-12, four bands between 3.5 and6.3 kb; TGF-33, 3.3 kb. These transcripts are similarto those previously published12 and these resultstherefore confirm the specificity of the probes.

Figure 4 shows lung tissue from a representativecontrol animal hybridized with TGF-f31 (a, antisense;b, sense), TGF-,32 (c, antisense; d, sense), andTGF-33 (e, antisense; f, sense) probes. Both TGF-p1and TGF-f33 mRNA transcripts, stained red, were

Table 1. Time Course of Changes in Lung Collagen Content duing the Development of Bleomycin-Induced PulmonaryFIbrosis

Lung collagen content (mg)3 days 10 days 21 days 35 days

ControlBleomycin

1.24 + 0.021.20 + 0.04

Values are the mean (n = 6) + SEM.*P < 0.05 compared with controls.tp < 0.01 compared with controls.

60

zcW 40c00

E2 20

X -10

-20

H H Hr

TGF,1 TGF02 TGFB3

1.10 + 0.011.61 + 0.08*

1.23 ± 0.031.76 + 0.14t

1.60 + 0.022.89 ± 0.14t

Nm-i I - - _ __

7"5_

1.4 b

986 Coker et alAJP March 1997, Vol. 150, No. 3

e

......

Figure 4. TGF-f-3 gene expression in normal murine lung. a and b: Normal murine lung hybridized with TGF-f,3 antisense (a) and sense (b)riboprobes. c and d: Normal murine lung hybnidized with TGF-/32 antisense (c) and sense (d) riboprobes. Arrows show hybridization signal obtainedu'ith the sense probe in bronchial epithelium (d). e and f: Normal murine lung hybridized with TGF-f3 antisense (e) and sense (f) riboprobes.Hematoxjlin counterstain (a to d); methyl green (e and f); original magnification, X 200. Hybridization signal appears as a red color at the site ofmRNA detectiont.

present in normal mouse lung, and gene expressionfor both isoforms was predominant in bronchiolarepithelial cells. However, mRNA transcripts forTGF-f31 and TGF-f3 were also observed in the inter-stitium, where they were localized to alveolar macro-phages, mesenchymal cells, and cells lining the al-veolar wall thought to be alveolar type 11 cells. Signalfor both isoforms was also detectable in mesothelialcells at the pleural edge (not shown).The TGF-f32 probes generated hybridization signal

with the sense probe (arrowed in bronchial epithe-lium in Figure 4d) but little or no signal with theantisense probe (Figure 4c). Cellular localization ofsignal obtained with TGF-132 sense probe was similarto, but less widespread, than that obtained with theTGF-f31 antisense probe. These results were repro-ducible between experiments and between differentbatches of probe. Probe orientation was thereforeindependently confirmed using two methods. First,asymmetric restriction enzyme cuts were performedwith Clal and the resulting fragments examined byagarose gel electrophoresis. Second, dideoxy-medi-ated chain termination sequencing (United StatesBiochemical, Amersham, UK) yielded a sequencepattern that confirmed the greatest homology (87%)of the sense probe with mRNA for murineTGF-3232'33 when matched using the FASTA data-base searching program (Human Genome MappingProject CRC, Cambridge, UK). The possibility ofgreater digoxigenin labeling of the sense probe

compared with the antisense probe was excludedusing a chemiluminescence assay (BoehringerMannheim) according to the manufacturer's instruc-tions. The ratio of digoxigenin labeling of the anti-sense probe compared with the sense probe was1.2 to 1.

TGF-,31 gene expression was only slightly en-hanced 3 days after bleomycin (not shown) andappeared maximally enhanced 10 days after bleo-mycin administration. Figure 5 shows lung tissue atthis time hybridized with TGF-,B1 (a, sense; b-e,antisense), TGF-132 (f, sense; g, antisense), andTGF-j33 (h, sense; i, antisense) probes. After bleo-mycin, reduced TGF-f31 gene expression was appar-ent in bronchiolar epithelium (arrows in Figure 5b).Inflammation at this time was patchy and localizedmainly around airways and blood vessels and be-neath the pleura. TGF-f1 gene expression was pre-dominantly localized to inflammatory cells includingmacrophages (arrowheads in Figure 5c) but wasalso present in increased intensity and in greaternumbers of mesenchymal cells underlying bloodvessels (arrows in Figure 5c), luminal cells that maybe endothelial or subendothelial (Figure 5d), and inmesothelial and submesothelial cells adjacent to ar-eas of subpleural fibrosis (Figure 5e). In addition,there was intense signal throughout the interstitium,consistent with expression of TGF-f31 by capillaryendothelial cells, alveolar type epithelial cells, and

-i~ i, -N-L,

,.:7"i

I

..!X.i.... f

TGF-,B Isoforms and Lung Fibrosis 987AJP March 1997, Vol. 150, No. 3

Figure 5. TGF-f3_ geneexpression 10 days after bleornc in. a to e: TGF-I,' gene expression. Marl.ea uiyiuammarogsoy cuet uqu *,rau, em uj 3t,t.

with sense probe (a). Hybridization signal was reduced in bronchiolar epithelium (arrows in b) but enhanced in alveolar macrophages and

inflammator cells (arrowheads in c) and mesenchymal cells underlying blood vessels (arrows in c). Signal was also inicreased in luminal cell;s,which may be pulmnonary endothelial or subetndothelial cells (arrows in d), and in mesothelial cells adjacent to areas ofsubpleuralfibrosis (arrows

in e). f and g: TGF-132 gene expression. Positive signal was obtained with senseprobe (arrowheads in f), but little signal was observed with antisense

probe (g). h and i: TGF- 33 gene expression, showing absence ofsignal ivith senseprobe (h). Very little signal was observed in bronchiolar epitheliuin,macrophages, or inflfammatoryfoci Hematoxylin counterstain; original magnification, X 200 (a, b, and f to i), x 400 (c and e), and X1000(d).

fibroblasts. Cells lining alveolar walls, probably alve-olar type II cells, also expressed TGF-f1 (not shown).

Hybridization with the TGF-132 riboprobes againyielded positive signal with the sense probe (arrow-heads in Figure 5f) but little or no signal with theantisense probe. In contrast to TGF-l31, TGF-,33 geneexpression was not enhanced after bleomycin, andvery little signal was observed either in bronchiolarepithelium or in macrophages and inflammatory foci.

Figure 6 shows lung tissue examined 21 and 35days after bleomycin hybridized with TGF-f31 sense

(Figure 6, a and c) and antisense (Figure 6, b and d)probes, respectively, and lung tissue 21 days afterbleomycin hybridized with the TGF-,32 sense (Figure6e) and antisense (Figure 6f) probes. Patchy fibrosiswas evident at 21 days with TGF-,31 expression pre-

dominant in alveolar macrophages (arrowheads inFigure 6b). At 35 days after bleomycin, there was a

return toward the control pattern of gene expressionfor TGF-,31, with mRNA transcripts predominant inbronchiolar epithelium (arrowheads in Figure 6d).

Similarly, at 35 days, the pattern of signal localizationfor TGF-,33 was identical to that seen in control ani-mals (data not shown). Hybridization with the TGF-f2probes again yielded positive signal with the sense

probe (arrowheads in Figure 6e) but little or no signalwith the antisense probe.

DiscussionIn this study we have examined the effects of TGF-I1-3 on fibroblast procollagen synthesis and degra-dation. We have shown that TGF-132 and TGF-f3share the ability of TGF-f31 to promote collagen dep-osition by stimulating fibroblast procollagen synthe-sis. This is consistent with their high degree of ho-mology and previous assays demonstrating similarbiological effects.34 However, TGF-f33 was 10 timesmore potent than the other two isoforms in stimulat-ing procollagen production and also reduced intra-cellular procollagen degradation. Our findings are

I.

988 Coker et alA/P March 1997, Vol. 150, No. 3

p

*

Figure 6. TGF- 3, and TGF-,12 gene expression 21 and 35 days after bleornycin. a and b: TGF-)3, genle expression in lung tissue 21 days afterbleomycin. a: Sense. b: Antisense. Patchy fibrosis uas evident at 21 days, with signalfor TGF-,1 promitnentt in macrophages (arrowheads in b). cand d: TGF-,B1 gene expression in luntg tissue 35 days after bleomrycin. c: Sense. d: Antisense. TGF- 3, gene expression returned toward the patternseen in control animals, with hybridization signalpredominant in bronchiolar epithelium (arrowheads in d). e and f: TGF-,12 gene expression inlung tissue 21 days after bleomycin. e: Sense. f: Antisense. Hbridization signal was obtained uvith the sense probe (arrowheads in e), but little otno signal uas obtained with the antisense probe. Methyl green counterstain; original magnification, X 400.

consistent with the observation that TGF-,B3 is morepotent than TGF-p1 in stimulating collagen produc-tion by fetal rat bone cells35 and may reflect differingreceptor affinities for TGF-31-3.

Using digoxigenin-labeled riboprobes, we local-ized TGF-,3 isoform gene expression in normal mu-rine lung and during bleomycin-induced lung fibro-sis. In normal lung, TGF-f31 and TGF-f33 mRNAtranscripts were identified in a wide variety of cellsincluding bronchiolar epithelium, alveolar macro-phages, alveolar type 11 cells, and mesenchymal andmesothelial cells. The widespread distribution ofTGF-,1l and TGF-f33 mRNA transcripts in adult mu-rine lung adds significant new information to previ-ous data using radiolabeled probes30 that showedgene expression for TGF-j31-3to be limited to smoothmuscle cells and fibroblasts in the bronchioles.These findings suggest that digoxigenin-labeled ri-boprobes are a more sensitive tool for cytokine de-tection than radiolabeled ones. They also suggestthat TGF-f31 and TGF-f33 play important roles in nor-mal lung homeostasis consistent with the recognizedregulatory effects of TGF-,31 on epithelial cell prolif-eration and differentiation,36 immunomodulation,37and matrix protein turnover.22A striking finding 10 days after bleomycin was the

shift from predominant signal in airway epithelium ofnormal animals to predominant signal in alveolarwalls. This may reflect both epithelial cell damageand inflammatory cell influx. It was not seen in controlanimals, indicating that it was specific to bleomycin

injury. After bleomycin, TGF-f31 gene expression wasmainly localized to inflammatory cells including mac-rophages. Macrophages are recognized as an im-portant source of TGF-f31 in pulmonary fibrosis.48However, we have shown that signal in mesenchy-mal, alveolar type 11, and mesothelial cells was alsoenhanced after bleomycin. Furthermore, the gener-alized increase in alveolar wall signal suggests thatalveolar epithelial, microvascular endothelial, andsubendothelial cells were also expressing TGF-f1.

Mesenchymal cells expressing TGF-f31 may befibroblasts or smooth muscle cells. Lung fibroblastsproduce TGF-,B in vitro38 and auto-induction may oc-cur.39 Fibroblast and endothelial cell TGF-,31 geneexpression increases after in vitro bleomycin expo-sure,40.41 and TGF-P13 augments endothelial matrixprotein synthesis.42 In the light of this evidence, ourdata suggest that increased TGF-P13 gene expres-sion by mesenchymal cells and pulmonary and cap-illary endothelial cells after bleomycin administrationcontributes to interstitial matrix accumulation.

Cultured mesothelial cells express TGF-f31mRNA,4344 but to our knowledge, ours is the firststudy to show this in vivo. Mesothelial cells have nothitherto been shown to express TGF-P33 mRNA. Wehave previously demonstrated procollagen gene ex-pression in the subpleural region of fibrotic mouselung.2529 Mesothelial cells may thus participate inthe pathogenesis of pulmonary fibrosis by producingTGF-31, which stimulates subpleural fibroblast colla-gen synthesis. TGF-f31 also stimulates mesothelial

IIIII

I9oh

A

I.,I

TGF-f Isoforms and Lung Fibrosis 989AJP March 1997, Vol. 150, No. 3

cell proliferation in vitro;45'46 TGF-31 produced bythese cells may therefore also act in autocrine fash-ion to stimulate their replication after injury. Our find-ings strengthen previous evidence suggesting thatmesothelial cells participate in repair after pleuralinjury.47-49

Cells lining alveolar walls expressing TGF-,31 maybe alveolar type 11 cells or adherent macrophages.Bleomycin induces proliferation and metaplasia oftype II cells.50 TGF-31-3 inhibit alveolar type 11 cellreplication, and TGF-,3 secretion by these cells afterbleomycin, as measured by bioassay, is inverselyrelated to their proliferation.20 TGF-t31 production bymetaplastic type 11 cells may therefore act in anautocrine manner to regulate their proliferation anddifferentiation after injury.The time course for TGF-131 gene expression ob-

served in this study using in situ hybridization issimilar to that observed in previous studies thatquantified TGF-P31 mRNA by Northern analysis.3'51'52TGF-j31 gene expression increased to a maximumaround 10 days and then declined, with the patternof expression returning to that seen in control ani-mals by 35 days.

In contrast to TGF-/1, TGF-f33 gene expressionwas not enhanced after bleomycin nor associatedwith macrophage influx. Little signal was detected inbronchiolar epithelium or inflammatory cells afterbleomycin. Gene expression of TGF-f31 and TGF-J3is therefore differentially regulated, suggesting thatTGF-f1 but not TGF-f33 is implicated in the patho-genesis of bleomycin-induced pulmonary fibrosis.Our findings are consistent with in vitro studies thathave shown that TGF-,B1 but not TGF-f32 or TGF-f33secretion by alveolar macrophages and alveolar ep-ithelial type 11 cells increases during the evolution ofthis disease.1820 Furthermore, increased mRNA forTGF-,31 but not TGF-P32 has been demonstrated inbleomycin-treated mice.19 In contrast, one study ofbleomycin-induced pulmonary fibrosis has pro-posed that TGF-31 are all implicated in the patho-genesis of pulmonary fibrosis.21

The results obtained with the TGF-42 probes wereunexpected. There are several potential explana-tions. First, the sense probe could be inserted in theopposite direction to that predicted within the vector.This possibility was excluded by restriction mappingwith C/al and dideoxynucleotide sequencing. Sec-ond, the sense probe could be hybridizing with anunrelated species. This is relatively unlikely given thesequencing results. Third, the sense probe could bemore heavily labeled with digoxigenin than the anti-sense probe. This was excluded using a chemilumi-nescence assay to assess labeling. Finally, a natu-

rally occurring antisense molecule may be present inlung tissue. A precedent is the demonstration of anendogenous TGF-f3 antisense molecule in chickembryo heart during cardiac valve formation.53 Itstemporally controlled appearance suggests it mayregulate TGF-f33 production during development.Pelton and colleagues30 were able to demonstrateTGF-f32 mRNA using the same probes as ours, sug-gesting that the appearance of a natural antisensemolecule in mice may be strain or age specific.Additional studies using Northern analysis of lungwith the TGF-,32 riboprobes are required to establishwhether such an antisense molecule exists. If con-firmed, the implications of an antisense RNA mole-cule regulating TGF-f32 gene expression will be im-portant not only for our understanding of regulationof its gene expression and its role in the pathogen-esis of fibrosis but also for the design of future ther-apies directed at modifying TGF-,3 function in a va-riety of fibrotic disorders.

In summary, we have shown that TGF-f31, TGF-,32,and TGF-f33 all stimulate fibroblast procollagen syn-thesis in vitro, TGF-,33 being the most potent in thisassay. TGF-f33 also reduced intracellular procolla-gen degradation. TGF-f31 and TGF-f33 gene expres-sion was localized to a wide diversity of cell types innormal murine lung. After bleomycin, TGF-f31 geneexpression was maximally enhanced at 10 days andpredominantly localized to macrophages. TGF-133gene expression was not enhanced after bleomycin.TGF-f32 mRNA was not detectable at any stage usingthe antisense probe, but signal was obtained usingthe sense probe, suggesting the presence of a nat-urally occurring antisense molecule. Differentialgene regulation of the isoforms during the course ofbleomycin-induced pulmonary fibrosis is consistentwith data emerging from other models of lung injury.Our results suggest that, whereas TGF-f31 is impli-cated in the pathogenesis of bleomycin-induced pul-monary fibrosis, TGF-f33 may not be. The role ofTGF-f32 remains unclear.

Our findings have important implications for thedevelopment of anti-TGF-,3 strategies in the treat-ment of fibrotic lung disorders. A number of suchstrategies have been proposed,'154 including theuse of antisense molecules, soluble receptor antag-onists, and antibodies. If our results are confirmed inpatients, TGF-31 will become the key target for suchtherapies.

AcknowledgmentsWe are grateful for the gift of cDNA probes for TGF-1-3 from Professor H. L. Moses (Nashville, TN).

990 Coker et alAJPMarch 1997, Vol. 150, No. 3

References

1. McAnulty RJ, Laurent GJ: Pathogenesis of lung fibrosisand potential new therapeutic strategies. Exp Nephrol1995, 3:96-107

2. McAnulty RJ, Campa JS, Cambrey AD, Laurent GJ:The effect of transforming growth factor-p on rates ofprocollagen synthesis and degradation in vitro. BiochimBiophys Acta 1991, 1091 :231-235

3. Hoyt DG, Lazo JS: Alterations in pulmonary mRNAencoding procollagens, fibronectin, and transforminggrowth factor-f3 precede bleomycin-induced pulmo-nary fibrosis in mice. J Pharmacol Exp Ther 1988, 246:765-771

4. Khalil N, Bereznay 0, Sporn MB, Greenberg AH: Mac-rophage production of transforming growth factor-/3and fibroblast collagen synthesis in chronic pulmonaryinflammation. J Exp Med 1989, 170:727-737

5. Giri SN, Hyde DM, Hollinger MA: Effect of antibody totransforming growth factor-,B on bleomycin-inducedaccumulation of lung collagen in mice. Thorax 1993,48:959-966

6. Denis M: Neutralization of transforming growth fac-tor-f31 in a mouse model of immune-induced lung fibro-sis. Immunology 1994, 82:584-590

7. Khalil N, O'Connor RN, Unruh HW, Warren PW,Flanders KC, Kemp A, Bereznay OH, Greenberg AH:Increased production and immunohistochemical local-ization of transforming growth factor-,B in idiopathicpulmonary fibrosis. Am J Respir Cell Mol Biol 1991,5:155-162

8. Broekelmann TJ, Limper AH, Colby TV, McDonald JA:Transforming growth factor-f31 is present at sites ofextracellular matrix gene expression in human pulmo-nary fibrosis. Proc Natl Acad Sci USA 1991, 88:6642-6646

9. Corrin B, Butcher D, McAnulty RJ, du Bois RM, BlackCM, Laurent GJ, Harrison NK: Immunohistochemicallocalization of transforming growth factor-,81 in thelungs of patients with systemic sclerosis, cryptogenicfibrosing alveolitis, and other lung diseases. Histopa-thology 1994, 24:145-150

10. Gatherer D, ten Dijke P, Baird DT, Akhurst RJ: Expres-sion of TGF-,B isoforms during first trimester humanembryogenesis. Development 1990, 110:445-460

11. Millan FA, Denhez F, Kondaiah P, Akhurst RJ: Embry-onic gene expression patterns of TGF-/31, ,2, and ,3suggest different developmental functions in vivo. De-velopment 1991, 111:131-144

12. Schmid P, Cox D, Bilbe G, Maier R, McMaster GK:Differential expression of TGF J31, f2, and ,33 genesduring mouse embryogenesis. Development 1991,111:117-130

13. Unsicker K, Flanders KC, Cissel DS, Lafyatis R, SpornMB: Transforming growth factor-, isoforms in the adultrat central and peripheral nervous system. Neuro-science 1991, 44:613-625

14. Yamanaka Y, Friess H, BOchler M, Beger HG, Gold LI,

Korc M: Synthesis and expression of transforminggrowth factor ,B-1, /3-2, and /3-3 in the endocrine andexocrine pancreas. Diabetes 1993, 42:746-756

15. Schmid P, Cox D, Bilbe G, McMaster G, Morrison C,Stahelin H, Luscher N, Seiler W: TGF-f3s and TGF-/3type 11 receptor in human epidermis: differential ex-pression in acute and chronic skin wounds. J Pathol1993, 171:191-197

16. Milani S, Herbst H, Schuppan D, Stein H, Surrenti C:Transforming growth factors /1 and ,B2 are differen-tially expressed in fibrotic liver disease. Am J Pathol1991, 139:1221-1229

17. Perkett EA, Pelton RW, Meyrick B, Gold LI, Miller DA:Expression of transforming growth factor-,8 mRNAsand Proteins in pulmonary vascular remodeling in thesheep air embolization model of pulmonary hyperten-sion. Am J Respir Cell Mol Biol 1994, 11:16-24

18. Khalil N, Whitman C, Zuo L, Danielpour D, GreenbergA: Regulation of alveolar macrophage transformingfactor-,8 secretion by corticosteroids in bleomycin-in-duced pulmonary inflammation in the rat. J Clin Invest1993, 92:1812-1818

19. Baecher-Allan CM, Barth RK: PCR analysis of cytokineinduction profiles associated with mouse strain varia-tion in susceptibility to pulmonary fibrosis. Reg Immu-nol 1993, 5:207-217

20. Khalil N, O'Connor RN, Flanders KC, Shing W, WhitmanC: Regulation of type 11 alveolar epithelial cell prolifer-ation by TGF-,B during bleomycin-induced injury in rats.Am J Physiol 1994, 267:L498-L507

21. Santana A, Saxena B, Noble NA, Gold LI, Marshall BC:Increased expression of transforming growth factor /3isoforms (/31, /32, /33) in bleomycin-induced pulmonaryfibrosis. Am J Respir Cell Mol Biol 1995, 13:34-44

22. McAnulty RJ, Chambers RC, Laurent GJ: Regulation offibroblast procollagen production: transforming growthfactor-,81 induces prostaglandin E2 but not procolla-gen synthesis via a pertussis toxin-sensitive G-protein.Biochem J 1995, 307:63-68

23. Chambers RC, McAnulty RJ, Shock A, Campa JS,Newman Taylor AJ, Laurent GJ: Cadmium selectivelyinhibits fibroblast procollagen production and prolifer-ation. Am J Physiol 1994, 267:L300-L308

24. Campa JS, McAnulty RJ, Laurent GJ: Application ofhigh-pressure liquid chromatography to studies of col-lagen production by isolated cells in culture. Anal Bio-chem 1990, 186:257-263

25. Shahzeidi S, Jeffery PK, Laurent GJ, McAnulty RJ: In-creased type procollagen mRNA transcripts in thelungs of mice during the development of bleomycin-induced fibrosis. Eur Respir J 1994, 7:1938-1943

26. Laurent GJ, McAnulty RJ, Corrin B, Cockerill P: Bio-chemical and histological changes in pulmonary fibro-sis induced in rabbits with intratracheal bleomycin. EurJ Clin Invest 1981, 11:441-448

27. Laurent GJ, Cockerill P, McAnulty RJ, Hastings JRB: Asimplified method for quantitation of the relative

TGF-, Isoforms and Lung Fibrosis 9914/P March 1997, Vol. 150, No. 3

amounts of type and type IlIl collagen in small tissuesamples. Anal Biochem 1981, 113:301-312

28. Maniatis T, Fritsch EF, Sambrook J: Molecular Cloning:A Laboratory Manual. Cold Spring Harbor, NY, ColdSpring Harbor Laboratory Press, 1989

29. Shahzeidi S, Mulier B, de Crombrugghe B, Jeffery PK,McAnulty RJ, Laurent GJ: Enhanced type IlIl collagengene expression during bleomycin induced lung fibro-sis. Thorax 1993, 48:622-628

30. Pelton RW, Johnson MD, Perkett EA, Gold LI, MosesHL: Expression of transforming growth factor-/3l, -,B2,and -f33 mRNA and protein in the murine lung. Am JRespir Cell Mol Biol 1991, 5:522-530

31. Ponder BA, Wilkinson MM: Inhibition of endogenoustissue alkaline phosphatase with the use of alkalinephosphatase conjugates in immunohistochemistry.J Histochem Cytochem 1981, 29:981-984

32. Madisen L, Webb NR, Rose TM, Marquardt H, Ikeda T,Twardzik D, Seyedin S, Purchio AF: Transforminggrowth factor-,82: cDNA cloning and sequence analy-sis. DNA 1988, 7:1-8

33. Webb NR, Madisen L, Rose TM, Purchio AF: Structuraland sequence analysis of TGF-,32 cDNA clones pre-dicts two different precursor proteins produced by al-ternative mRNA splicing. DNA 1988, 7:493-497

34. Roberts AB, Sporn MB: The transforming growth factor,Bs. Peptide Growth Factors and Their Receptors I.Edited by MB Sporn, AB Roberts. London, Springer-Verlag, 1990, pp 419-472

35. ten Dijke P, Iwata KK, Goddard C, Pieler C, Canalis E,McCarthy TL, Centrella M: Recombinant transforminggrowth factor type 13: biological activities and recep-tor-binding properties in isolated bone cells. Mol CellBiol 1990, 10:4473-4479

36. Masui T, Wakefield LM, Lechner JF, La Veck MA, SpornMB, Harris CC: Type 13 transforming growth factor is theprimary differentiation-inducing serum factor for normalhuman bronchial epithelial cells. Proc Natl Acad SciUSA 1986, 83:2438-2442

37. Shull MM, Ormsby I, Kier AB, Pawlowski S, Diebold RJ,Yin M, Allen R, Sidman C, Proetzel G, Calvin D, Annun-ziata N, Doetschman T: Targeted disruption of themouse transforming growth factor-,11 gene results inmultifocal inflammatory disease. Nature 1992, 359:693-699

38. Kelley J, Fabisiak JP, Hawes K, Absher M: Cytokinesignaling in lung: transforming growth factor-,B secre-tion by lung fibroblasts. Am J Physiol 1991, 260:L123-L128

39. Kelley J, Shull S, Walsh JJ, Cutroneo KR, Absher M:Auto-induction of transforming growth factor 13 in hu-man lung fibroblasts. Am J Respir Cell Mol Biol 1993,8:417-424

40. Phan SH, Gharee-Kermani M, Wolber F, Ryan US:Bleomycin stimulates production of transforminggrowth factor-,8 by rat pulmonary artery endothelialcells. Chest 1991, 99:66S

41. Breen E, Shull S, Burne S, Absher M, Kelley J, Phan S,Cutroneo KR: Bleomycin regulation of transforminggrowth factor-,8 mRNA in rat lung fibroblasts. Am JRespir Cell Mol Biol 1992, 6:146-152

42. Muller G, Behrens J, Nussbaumer U, BohIen P, Birch-meier W: Inhibitory action of transforming growth factor,B on endothelial cells. Proc Natl Acad Sci USA 1987,84:5600-5604

43. Bermudez E, Everitt J, Walker C: Expression of growthfactor and growth factor receptor RNA in rat pleural me-sothelial cells in culture. Exp Cell Res 1990, 190:91-98

44. Gerwin BI, Lechner JF, Reddel RR, Roberts AB, Rob-bins KC, Gabrielson EW, Harris CC: Comparison ofproduction of transforming growth factor-,B and plate-let-derived growth factor by normal human mesothelialcells and mesothelioma cell lines. Cancer Res 1987,47:6180-6184

45. Franzen L, Dahlquist C: The effect of transforminggrowth factor-,B on fibroblast cell proliferation in intactconnective tissue in vitro. In Vitro Cell Dev Biol 1994,30A:460-463

46. Gabrielson EW, Gerwin BI, Harris CC, Roberts AB,Sporn MB, Lechner JF: Stimulation of DNA synthesis incultured primary human mesothelial cells by specificgrowth factors. FASEB J 1988, 2:2717-2721

47. Owens MW, Grimes SR: Pleural mesothelial cell re-sponse to inflammation: tumor necrosis factor-inducedmitogenesis and collagen synthesis. Am J Physiol1993, 265:L382-L388

48. Adamson IYR, Bakowska J, Bowden DH: Mesothelialcell proliferation: a nonspecific response to lung injuryassociated with fibrosis. Am J Respir Cell Mol Biol1994, 10:253-258

49. Davila RM, Crouch EC: Role of mesothelial and sub-mesothelial stromal cells in matrix remodeling followingpleural injury. Am J Pathol 1993, 142:547-555

50. Adamson IYR, Bowden DH: Bleomycin-induced injuryand metaplasia of alveolar type 2 cells. Am J Pathol1979, 96:531-544

51. Raghow R, Irish P, Kang AH: Coordinate regulation oftransforming growth factor 3 gene expression and cellproliferation in hamster lungs undergoing bleomycin-induced pulmonary fibrosis. J Clin Invest 1989, 84:1836-1842

52. Phan SH, Kunkel SL: Lung cytokine production in bleo-mycin-induced pulmonary fibrosis. Exp Lung Res1992, 18:29-43

53. Potts JD, Vincent EB, Runyan RB, Weeks DL: Senseand antisense TGF-,B3 mRNA levels correlate with car-diac valve induction. Dev Dynam 1992, 193:340-345

54. Coker RK, Laurent GJ: Pathogenesis of pulmonaryfibrosis: implications for pharmacological intervention.Immunology and Management of Interstitial Lung Dis-eases. Edited by EH Walters, RM du Bois. London,Chapman and Hall Medical, 1995, pp 19-36