Extracellular matrix gene expression increases preferentially in rat lipocytes and sinusoidal endothelial cells during hepatic fibrosis in vivo

Extracellular matrix gene expression increases

preferentially in rat lipocytes and sinusoidal

endothelial cells during hepatic fibrosis in vivo.

J J Maher, R F McGuire

J Clin Invest.

1990;

86(5)

:1641-1648.

https://doi.org/10.1172/JCI114886

.

Whether parenchymal or nonparenchymal liver cells play a predominant role in the

pathophysiology of hepatic fibrosis has not been firmly established in vivo. We have

addressed this question by quantitating the relative abundance of specific mRNAs for

collagen types I, III, and IV, and laminin in purified populations of hepatocytes, sinusoidal

endothelial cells, and lipocytes from normal and fibrotic rat liver. In normal liver, type I

collagen gene expression was minimal in all cell types; mRNA for types III and IV collagen

were apparent in endothelial cells and lipocytes, but not in hepatocytes. Laminin mRNA

was present in all cell types. Induction of fibrogenesis by either bile duct ligation or carbon

tetrachloride administration was associated with a substantial increase in mRNA for types I

and III collagen in nonparenchymal cells. Lipocytes from fibrotic animals exhibited a greater

than 30-fold increase in type I collagen mRNA relative to normal lipocytes, and greater than

40-fold relative to hepatocytes. Type III collagen mRNA reached 5 times that in normal

lipocytes and greater than 120 times that in hepatocytes. Endothelial cells exhibited an

isolated increase in type I collagen mRNA, reaching five times that in normal liver. Type IV

collagen and laminin gene expression were not significantly increased in nonparenchymal

cells during fibrogenesis; in fact, mRNA for type IV collagen and laminin decreased by […]

Research Article

Find the latest version:

http://jci.me/114886/pdf

Extracellular

Matrix Gene

Expression

Increases

Preferentially

in Rat

Lipocytes

and

Sinusoidal Endothelial Cells

during Hepatic

Fibrosis In

Vivo

Jacquelyn J. Maher and Richard F. McGuire

Liver Core Center and Department ofMedicine, University of California, San Francisco, California 94110

Abstract

Whether parenchymal or nonparenchymal liver cells play a predominant role in the pathophysiology of hepatic fibrosis

has not beenfirmlyestablishedin vivo. We have addressed this question by quantitating the relative abundance of specific

mRNAsfor collagen typesI, III, and IV, and laminin in puri-fied populations of hepatocytes, sinusoidal endothelial cells, and lipocytes from normal and fibrotic rat liver. In normal liver, type I collagen geneexpression was minimal in all cell types; mRNA for types III and IVcollagenwereapparent in endothelial cells andlipocytes,butnotin hepatocytes. Laminin mRNA was present inall cell types.Induction offibrogenesis by eitherbile duct ligationorcarbontetrachloride administra-tion was associated witha substantial increase inmRNA for types I and III collagen in nonparenchymal cells. Lipocytes fromfibrotic animals exhibited a >30-fold increase in typeI collagen mRNA relative to normal lipocytes, and >40-fold relative to hepatocytes. Type III collagen mRNA reached 5 times that in normallipocytesand> 120 times that in hepato-cytes. Endothelial cells exhibitedanisolated increase in typeI collagen mRNA, reaching five times that in normal liver. Type IV collagen and laminin geneexpression were notsignificantly increased in nonparenchymal cells duringfibrogenesis; in fact, mRNAfor type IVcollagen and laminin decreased by up to 50% inendothelial cells. Despite the pronounced changes that occurred in matrix gene expression in nonparenchymal cells duringfibrogenesis, no change was noted in hepatocytes. We conclude that nonparenchymal liver cells, particularly lipo-cytes, are important effectors ofhepatic fibrosis in vivo. (J. Clin. Invest. 1990. 86:1641-1648.) Key words:

collagen-laminin*liver* cirrhosis-hepatocytes

Introduction

Liver fibrosis, regardless of its cause, is characterized by a

marked accumulation of extracellular matrix material within

the

perisinusoidal

spaceof Disse. Interstitial collagens (typesI andIII)comprisethe bulkofthenewlydepositedmatrix (1-3),

althoughbasement membrane (typeIV) collagenand

noncol-lagenousglycoproteins, suchaslaminin andfibronectin, also

Aportion of this work has beenpublishedin abstractform in 1989

(Hepatology[Baltimore] 10:629a).

Addressreprintrequests to Dr. J.Maher, Liver Center Laboratory, SanFrancisco General Hospital, Bldg. 40 Rm. 4102, 1001 Potrero Ave.,San Francisco,CA 941 10.

Receivedfor publication 13 April 1990 and in revisedform 28 June 1990.

contribute (4). Numerousefforts have been made toidentify the cellular sources of extracellular matrix proteins in intact liver. Immunohistochemical studies suggest that in normal liver, extracellular matrix proteins are presentinboth

paren-chymal and nonparenchymal liver cells (5-7). After carbon tetrachlorideadministration, typeIVcollagen and lamininare

localizedinsinusoidal endothelialcells andlipocytes (Ito cells, fat-storing cells), whereas types I and III collagen reportedly are present in hepatocytes (8, 9). These observations suggest thatparenchymal liver cells playacentral role in the

produc-tion of interstitial collagenthataccompanies hepatic fibrosis.

Invivo labeling studiesalso suggest that the majority of colla-gen extracted from liver after carbon tetrachloride

administra-tion derives from hepatocytes (10). Otherexperiments, how-ever, have ledto adifferentconclusion, thathepatic extracel-lular matrix synthesis is largely a function of mesenchymal liver cells.Experimentsexamining fibrotic animal and human liver by in situ hybridization demonstrate that messenger RNAsforcollagen types I, III, and IV, and lamininare con-centrated inperisinusoidalcells (1l1-13). Basedonthese

con-flictingresults, debatestill exists regarding whichcell popula-tion is themajorcontributortohepaticfibrosis.

Althoughimmunohistochemistryand insituhybridization

are useful forexaminingtheorigin ofextracellular matrixin

intacttissue,theyprovideonlyqualitativeinformation,andat the level oflight microscopycannotpinpointtheexact source

ofextracellular matrix protein or mRNA. These limitations

can be circumvented in part by adopting a cell culture ap-proach, which permits quantitative measurementsof matrix synthesisinpurified populations of livercells. Studiesin cul-turefavor nonparenchymal cellsastheprimarysourceof ex-tracellularmatrixinliver, demonstratingthatlipocytesfar ex-ceed hepatocytes in collagen synthesis on a per-cell basis (14-16).Eventhesefindingsmustbeinterpreted with caution,

however, inviewof evidencethat cells in cultureare

suscepti-ble to changes in collagen phenotype (17-19). An ideal

methodfordefiningthe cellular source(s)ofextracellular

ma-trix inliverin vivomustincorporatetheadvantages of both approaches above, providing a means toquantitate matrix

withinindividual liver cell populations without imposing the

influence of a foreign (culture) environment. Clayton and Darnell(20)have shown that the mRNAprofile of livercells

immediately following isolation fromtheintact organ

recapit-ulates that invivo. Inthecurrentstudy, this method has been

appliedto parenchymalandnonparenchymal liver cellsfrom

normaland fibroticliver,to assessexpression ofvarious extra-cellularmatrix genes invivo.Wedemonstrate intwo

mecha-nistically distinct models ofhepatic fibrosis in the rat, that extracellular matrix gene expression increases exclusively in

mesenchymal livercells. Furthermore,expression ofthe gene for typeIcollagen isenhancedtothe greatest extent inhepatic lipocytes.AstypeIcollagen is thepredominant constituent of

thefibrotic

hepatic

matrix,theseresults suggest thatlipocytes

J.Clin.Invest.

© TheAmericanSocietyforClinicalInvestigation,Inc.

0021-9738/90/1 1/1641/08 $2.00

figure prominently in the pathophysiology of hepatic fibrosis invivo.

Methods

Materials

[Alpha-32P]cytidine-5'-triphosphate ([a-32P]CTP) (>800Ci/mmol) (SP6grade) waspurchasedfrom Amersham Corp., Arlington Heights,

IL. Pronase and collagenaseB were purchased from Boehringer MannheimBiochemicals,Indianapolis,IN; DNaseI wasfrom Calbio-chem-Behring Corp.,LaJolla, CA.DMEand Ham'sF-12mediawere

obtained from Flow Laboratories, Inc., McLean, VA; Eagle'sMEM

withoutcalcium (MEM-E)waspreparedin the laboratory using amino acids fromSigmaChemicalCo.,St.Louis,MO. Larex* (arabinogalac-tan)wasobtained from Larex International, Tacoma,WA. Formalin-fixed S.aureus waspurchased from Zymed Laboratories,Burlingame, CA. Humanacetoacetylatedlow-densitylipoprotein, conjugatedwith 3,3'-dioctadecyl indocarbocyanine (Dil), wasgenerouslysuppliedby

Dr. R. Pitas, GladstoneFoundation, SanFrancisco, CA. Guanidine thiocyanate was purchased from Fluka ChemicalCorp., Ronkon-koma,NY.Cesiumchloride (molecular biology grade) and formamide

werefromInternationalBiotechnologies,Inc., NewHaven, CT. Poly-acrylamideand agarose(molecularbiology grade)wereobtained from Bio-Rad Laboratories, Richmond,CA; ultrapure urea wasfrom

Be-thesda Research Laboratories, Gaithersburg, MD. Restriction

en-zymesand RNaseT2 werepurchased from Bethesda Research

Labora-tories. SP6 and T7 RNA polymerases were from Promega Biotec, Madison, WI.Polyclonal rabbit anti-chicken desminwasfromDako

Corp., Santa Barbara, CA.

Procedures

Animal modelsoffibrogenesis. Mildhepatic fibrosis was induced in maleSprague-Dawleyratsby either bile ductligation(BDL)'orcarbon

tetrachloride administration (CCL). Bile ductligationand scission

were performed under ether anesthesia. Animalswere maintained postoperatively on food andwater ad lib. and killed for liver cell

isolationonpostoperative day5.Carbontetrachloride (50% vol/vol in cornoil,0.1 cm3/100gbodywti.p.)wasadministeredto unanesthe-tized rats ondays 1, 4, and7. Injectedanimalswereallowed7dto

recoverfromacutehepaticinjury,andkilledfor liver cell isolationon

day 14. Normal Sprague-Dawley rats servedas controls for both modelsoffibrosis.

Histology.Ratlivers subjectedtoexperimental injuryasdescribed

abovewereperfusion-fixedwith4%paraformaldehydein 0.1 M phos-phate(pH7.4) and embedded inparaffin.Cut sectionswerestained with Gordon and Sweet'sreticulin.

Liver cell isolation andpurification. Hepatocytes wereisolated from controlandexperimental animals by collagenase perfusionand

centrifugalelutriation,aspreviouslydescribed(16,21).Sinusoidal

en-dothelial cells andlipocytes wereisolated byin situ perfusionwith collagenase andpronase; thetwocellpopulationswereseparatedby density gradient centrifugationthrough LarexO,asdescribed(22).

Li-pocytes,which remainedatthetoptwointerfaces,werecollected and

washedtwice inculturemediumfollowed bycentrifugationat 500g to removehepatocytedebris. Endothelial cells, collected from the bottom two interfaces of the gradient, were purified further by centrifugal

elutriation at 2,500 rpm and a flow rate of 10-18 ml/min (J2-21

centrifuge, with JE6-B rotor andSanderson chamber; Beckman In-struments, Inc.,PaloAlto,CA). Purityofallisolateswas assessed by

directcellcounting,usingDil-conjugatedacetoacetylated low-density

lipoprotein (21), fluorescein-conjugated S.aureus (22), and anti-des-min(23)asmarkersfor endothelialcells,Kupffercells,and lipocytes,

respectively. Lipocytes couldalso bedistinguished bytheir intrinsic

fluorescenceunderultravioletillumination.

1. Abbreviations usedin this paper: BDL, bileduct ligation; CCL, carbon tetrachloride.

Probe preparation. cDNAs encodingrat al(I)collagen (paIRI) (24), ratal(III) collagen (PRGR5) (Makela,J.K.,andE. Vuorio,

un-publishedobservations), mouse al(IV) collagen (pPE123) (25), and

mouselaminin B2 chain (pPE9) (26) were donatedtothelaboratory. Restriction fragments of each were subcloned into pGEM@-4 (Pro-mega Biotec) to enable synthesisof labeledRNA for hybridization. The greatest sensitivity and specificity in RNase protection studies wereobtained with the following subclones: (a) for

paIRl,

a - 340-bp

Pst I-Bam HIfragment from the triple helical domain; (b) for pRGR5,

a- 280-bp Hinc II-EcoRIfragment from the 3' end of the translated sequence; (c)for pPE 123,a - 380-bpAvaI-EcoRIfragment at the 3'

endof the translated sequence; (d) for pPE9,a486-bpBamHI-SalI

fragment encompassing mostof the original cDNA. A - 900-bp cDNA for human

_#l-actin

(27), whichcross-hybridizes with ratfl-actin, was provided inSP64 by Dr. J. M. Friedman, The Rockefeller Univer-sity, New York.Radiolabeled probes were generated by transcription withSP6 or T7 polymerase using[a-32P]CTP (28). Specific activity of thetranscripts was estimatedat0.5 X109cpm/gg.

RNase protection(29). TotalRNA wasextractedfrom individual liver cell isolates bylysis in guanidine thiocyanate, followed by centrif-ugation throughcesium chloride (30). All extractions were performed immediatelyafter cell isolation and purification. The integrity of each samplewasverified by spectrophotometry

(A260/A280

> 1.9) and aga-rose gel electrophoresis withethidium bromide staining before use in molecular hybridization studies. 5 ug of totalRNA washybridized in solution with 0.25-0.75 X 106 cpm of32P-labeled cRNA for 16 h. Hybridization temperature was optimized for each probe, varying be-tween 560 and70°C. The mixturewasthen treated with T2 RNaseto

digest unhybridized cellular RNA and probe, followed by phenol/chlo-roform extraction and ethanol precipitation. The remaining intact hy-brids were denatured in electrophoresis buffer containing 80% form-amide, and separated by electrophoresis through a 5% polyacryl-amide-urea gel (21 X 40cm). Dried gelswere applied tox-ray film (Kodak X-OMat AR-5) for 1-24 h. Autoradiographic signals were

quantitated by scanning densitometry (Hoefer Scientific Instruments, SanFrancisco, CA).

Statisticalanalysis. Allnumerical data represent the mean±SEM of individual experiments performedon three separate animals. Re-sults with P<0.05 byMann-WhitneyU-test wereconsidered signifi-cant.

Results

To assess the degree of hepatic fibrosis induced by bile duct ligation and carbon tetrachloride administration, livers from both normal rats and those with experimentally induced fi-brosis were examined under light microscopy. As shown in Fig. 1,bile duct-ligated animals exhibited a moderate increase in periportal connective tissue relative to normal animals. An-imals exposed to three doses of carbon tetrachloride displayed a muchless pronounced change in hepatic connectivetissue, although central-central bridgingwasapparent. At these same time points, parenchymal and nonparenchymalcellswere iso-latedfor analysis of various extracellular matrix mRNAs. The purity of each cell isolate is indicatedinTable I. Hepatocytes exhibited the greatest homogeneity,with < 2% contamination by lipocytes. Endothelial cells were 95% pure,containing ap-proximately equal numbers ofKupffer cells and lipocytes. In animalswithexperimental fibrosis, the proportionoflipocytes inendothelial cell isolates sometimes reached 10%; this maybe due to increased penetration ofthedensity gradient by lipo-cytes, which releasetheir low-density vitaminAduring fibro-genesis(17, 31). The lipocyte fractionwas98% homogeneous, withKupffer cells being the primary contaminant.

Steady-state levels of mRNA for collagen types I, III, and

Figure1.Liverhistology. Photomicrographs illustrate the connective tissue distribution in livers from normal (a), bile duct-ligated (b), or CCL-treated (c)rats.Normal liver exhibits only mild connective

tis-suestainingaroundportal triads (P) and central veins (C).BDLafter 5 d results inanincrease in connective tissue in expanded portal

re-gions.CCLafter14dproducesanincrease in pericentral connective tissue,with evidenceof central-centralbridging.(Gordonand

Sweet'sreticulin, IOX.)

IV,laminin,and

f,-actin

were quantitated in liver cell isolates

by

RNase

protection

(seeMethods). Arepresentative

autora-diogram depicting

the results fortype I collagen is shownin Fig. 2. In normal liver, all three cell types exhibited weak

signals

fortype IcollagenmRNA, incomparison tofibroblast

Table I.PurityofIndividual Cell Isolates from Normal andFibrotic Liver*

Hepatocytes 98.8±0.2% Endothelial cells 95.0±0.5% Lipocytes 97.5±0.4%

Valuesrepresent mean±SEM for a minimum of three separate ani-mals.

*_Puritywas

assessed

_{by direct cellcounting,asdescribed in Methods.}

controls. After bile duct ligation,type I collagenmRNAwas

increasedinboth endothelial cells andlipocytes, butwas

un-changed in hepatocytes. Asimilarpattern wasobserved after carbontetrachloride administration. Additional experiments measuringmRNA fortypes III and IV collagen andlaminin

areillustrated in Fig.3. Innormal liver,type III and IV colla-gen mRNA were atthelimitof detection in normal

hepato-cytes; alamininsignalwasapparent butfaint. Endothelial cells and

lipocytes,

bycontrast,exhibitedstronghybridization

sig-nalsforall three matrixgenes. Induction of hepatic fibrogen-esis by eitherBDL or CCLwasnot accompanied bya gross

alteration in matrixgene expression by hepatocytes. mRNA levelsfortype III collagen, type IV collagen, andlamininwere all modulated,however, in endothelialcellsandlipocytes.

Molecular hybridization experiments were performed in

threesetsof animalsand

quantitated

by scanning

densitome-try(Table II; Fig. 4). Signal intensitywas normalized to the

lowest valuein eachgroup(uniformlythatin normal hepato-cytes), whichwasgivenanarbitraryvalueof1.0.Thisallowed

direct,

quantitative

comparisons of matrix gene expression

among different cell types and under different

experimental

conditions. Comparisonsare validonly forthe experimental

datafor eachindividual extracellular matrixgene(e.g.,type I

collagenin hepatocytes vs. type Icollagen in lipocytes);due to

differencesinprobe lengthand

specificity,

aswell as autoradio-graphic exposure time, the level ofexpression of individual extracellular matrix mRNAsfora

given

cell type cannot be compared (e.g., type I collagen vs. type IV collagen in lipo-cytes).

Fig.4 A illustratesthat innormal

liver,

steady-state levels

ofmRNAfortype Icollagenweresimilar inboth

parenchymal

andnonparenchymal cells. After bileductligation,type I

col-lagen geneexpression was increasedin both endothelial cells

and

lipocytes, reaching

7.8and49.8

times, respectively,

that in normal hepatocytes. Carbon tetrachloride administration re-sulted in a similar enhancement in type I collagen gene

ex-pression, witha14.8-fold and43.1-fold increaseinendothelial

cells andlipocytes, respectively.

Type III collagen mRNA was abundantin

nonparenchy-malcellsfrom normalliver(Fig. 4B). Endothelialcells

con-tained 28.5 timesand

lipocytes

29.0 timesthe amountoftype IIIcollagen mRNA asnormalhepatocytes. Type IIIcollagen

gene

expression

remained

high

in endothelial cells and

in-creased

dramatically

in lipocytes,

following

both BDL and

CCL

injury.

Type IVcollagengeneexpressioninnormal liverwas

dom-inated by endothelial cells, which containedalmost 100times

asmuch type

IV-specific

mRNA asnormalhepatocytes

(Fig.

4

C).

Lipocytes

also exhibited abundant type IV collagen

CONTROLS

II ₊

NORMAL BDL CCL4 tRNA 3T3

H E _Li | H E Li |

I

F L | lUg 2Ig5LIg1

****. olmm

)N

_

644

_ %

q404bp

JW. f

'1

(1)

'r~,%Now

Figure2.Type Icollagengeneexpression in individ-ualliver cellsasmeasuredby RNaseprotection(see Methods). Bandsrepresentsteady-state levels oftype IcollagenmRNA inhepatocytes(H), endothelial cells (E), andlipocytes (Li) from normal, bile duct-li-gated (BDL), and carbon tetrachloride(CCL)-treated

animals, respectively. Controls include tRNA from

yeastandmRNAfromNIH3T3fibroblasts,the

lat-terinincreasing concentrationstodocument linear-ityof the autoradiographicsignal.The band above 404 bprepresentsnonspecific hybridization.Type I collagenmRNAis visibleas adoubletat - 340 bp,

in parallel with the 3T3 controls.Signalsaregreatest

in thenonparenchymal cells frominjured livers.

hepatic injury with BDL or CCL, type IV collagen gene ex-pression remained high in both cell types; endothelial cells, however, were replaced by lipocytes as the population with the

highestlevelsoftype IV collagen mRNA.

Laminin gene expression was also highest in endothelial

cells fromnormalliver, measuredat13.1 timesthatinnormal hepatocytes (Fig. 4 D). Laminin mRNA in lipocytes was 5.8 times higher than in normal hepatocytes. As in the case of types III andIVcollagen,laminin mRNAremainedabundant in nonparenchymal cellsduring fibrogenesis although to dif-ferentdegrees in endothelial cells and lipocytes.

In order toidentifythe cell-specific changes that occur in extracellular matrix geneexpression during hepatic fibrosisin

vivo, mRNA levelsforcollagen types I, III, and IV, and

la-minininpurifiedcellpopulations from experimentally injured liverswerecompared tothoseforthecorrespondingcell type

from normal liver(Fig. 5). Viewedin this manner, the data

indicate that enhancement ofextracellular matrix gene

ex-pression during hepatic fibrogenesiswas largely afunctionof hepatic lipocytes. Asubstantial (>30-fold) increase intype I

collagengeneexpressionwasdemonstratedinbothBDLand

CCL-induced fibrogenesis; a smaller, albeit significant,

in-crease intype IIIcollagengeneexpressionwasobserved inthis

cell type. Type IVcollagengeneexpression wasincreasedin

lipocytes following BDLbut not CCLin the current

experi-NORMAL

BDL

CCL4

H E Li I H E Li H E Li

~~..

1w;..

I- l - c

(IV)

lo d_ ..f _*wo

4

O_"

W -LAM

B2

Figure 3. Steady-state levelsofmRNAfortype IIIcollagen (a(III)),

type IVcollagen

(a1(IV)),

andlaminin(L4MB2)incellsisolated from normalandfibrotic liver. Abbreviationsasin Fig.2. The stron-gesthybridization signalsareagainexhibited by nonparenchymalcells.

ments;laminingeneexpressionwas notsignificantly increased

in either model of fibrosis.

Enhancementof extracellular matrixgeneexpressionwas less pronouncedin endothelial cellsduring hepatic

fibrogen-esis(Fig. 5). TypeIcollagenmRNA wassignificantlyincreased inendothelial cellsfollowing both BDLandCCL-induced in-jury; mRNA for type III collagen, however, was unchanged fromnormalcells, and that fortype IVcollagenandlaminin actually decreasedtoless than 50%ofnormallevels.

Hepato-cytes were the leastresponsiveof all three cell types tohepatic injury. NeitherBDL norCCLinducedachangein steady-state mRNAlevelsforanyofthematrixgenestested.Noreductions

inextracellular matrix geneexpressionwereobserved in

hepa-tocytesfollowingliverinjury;itisunlikely, however,that any decreasebelownormalcouldbedetectedinthese

experiments,

giventhe lowhybridization signalsinallhepatocyte samples (Figs.2 and 3).

Table II. Steady-stateLevels ofExtracellular Matrix mRNA in LiverCellsIsolated

from

Normal and

Injured

Rat Liver

Normal BDL CCL

CollagenI

Hepatocytes 1.0±0* 1.4±0.2 0.9±0.2 Endothelial cells 1.9±0.3 7.8±2.5* 14.8±7.3$

Lipocytes 1.7±0.3 49.8±11.8* 43.1±13.6* CollagenIII

Hepatocytes 1.0±0* 1.2±0.2 2.1±0.9 Endothelial cells 28.5±8.0t 15.8±2.9t 35.6±10.0t Lipocytes 29.0±4.9* 150.0±49.8t 120.0±39.0t CollagenIV

Hepatocytes 1.0±0* 1.7±0.3 0.9±0.0 Endothelial cells 99.4±20.6t 38.5±5.4* 40.0±2.6t Lipocytes 35.7±4.2* 99.3±19.4* 66.8±10.4* Laminin

Hepatocytes 1.0±0* 1.7±0.4 1.0±0.1 Endothelial cells 13.1±8.4t 9.6±3.1* 16.2±3.7t Lipocytes 5.8±2.2* 13.6±4.3* 10.1±4.2*

*_Alldataareexpressedasrelativevalues,incomparisontonormal hepatocytes(arbitrarilyset at1.0).Numbersrepresent mean±SEM for threeseparateanimals.*P<0.05vs.normalhepatocytes,

by

Mann-WhitneyU-test.

Normal BDL CCL

Normal BDL CCL

Normal BDL CCL

Normal BDL CCL

Figure 4. Steady-state levels of mRNA for four extracellular matrix genes in cells from normal and fibrotic liver. Histograms illustrate thenumerical data in Table II for typeIcollagen (A),type III colla-gen(B), typeIVcollagen (C), and laminin (D).Notethat the

ordi-natein each graph isdifferent, making the results for hepatocytes in panelsBand C appearindistinct.Barsrepresent mean±SEM for he-patocytes

.,

endothelial cells *, and lipocytesofrom three separate animals. *P<0.05vs.normalhepatocytes.

TableIII. Steady-state Levels of(3-ActinmRNA inCells from Normal and InjuredRat Liver

Normal BDL* CCL*

Hepatocytes 1.0±0* 1.4±0.3 1.0±0.1 Endothelial cells 8.4±2.1 11.3±2.0 5.6±1.1 Lipocytes 3.1±0.5 3.2±0.4 2.0±0.6

*_{P >0.05 for each cell population fromBDLandCCL liver,in}

comparisontothesamecellpopulation fromnormalliver.tAll data areexpressedasrelative values, in comparisontonormalhepatocytes (arbitrarilyset at 1.0). Numbers represent mean±SEM for three sepa-rate animals.

Beta-actinmRNAwasalsoquantitatedin each of the cell

populations (TableIII;Fig. 6).The amountoftotal RNA that wasspecific for

_fl-actin

was notconstant among the three cell types: relative proportions of

fl-actin

mRNA were 1:8:3 in hepatocytes,sinusoidalendothelia, and lipocytes, respectively.

Steady-state levelsof

_fl-actin

mRNAdid not change in any of the cell types after experimental liverinjury, indicatingthat the alterationsnoted in extracellular matrix gene expression

were selective. Even when normalized for

_fl-actin,

mRNA levelsfortypes I and III collagen were significantly increased in

lipocytes from injured livers(data not shown).

Discussion

The preceding experiments demonstrate that collagen gene

expression, particularlythatfor interstitial collagen, increases dramaticallyinliverfollowingafibrogenic stimulus. Further-more,enhancement of interstitial collagen gene expression is moststriking ina single population of livercells, namely

he-patic lipocytes. After experimental liver injury, mRNAlevels

fortype Icollagen riseto over30timesthose in normal

lipo-cytes;type IIIcollagengeneexpression is increasedto approxi-mately 5 times over normal. Smaller, albeit significant,

changes are observed in sinusoidal endothelial cells with re-specttotype Icollagen.

The cellular source of extracellular matrix proteins and

their mRNAs in liver has been the subject ofsome debate.

MessengerRNAsfor collagentypesI,III, and IV, andlaminin

have beenlocalizedby insitu hybridizationtosinusoidal cells

in fibrotic liver (11-13); their precise cellular distribution, though, cannot be distinguished at the level oflight

micros-copy. By measuring extracellular matrix gene expression in

Collagen CollagenIII Collagen IV Lamninin

E

Normal BDL CCL Normal ODL CCL Normal BDL CCL Normal BOL CCL

Figure5.Cell-specificincreaseinextracellularmatrix

geneexpression during fibrogenesis. Graphsrepresent

specificmatrixmRNAabundanceincellsfrom BDL

andCCL-treatedliver,relativetothesamecells in

normal liver. Values for normal liver cellsare

arbi-trarilyset at1.0.Bars representmean±SEMfor

hepa-tocytes.,endothelial cells*,andlipocytesofrom threeseparate animals.*P<0.05vs.normal liver cells.

70

G Collagen A

50 40 _M E

20*

10 T

a

S 0

O,

C

-.

o

c

E

.> a 0t 120

40

S u

c 120

a

o

c-=_0i930

E

e

_

~Collagen l\

ri

*

0

25

a

c3

0e

"Oa

C-a

10

:.5

..

50

0 40

< 10

c

24

2

CONTROLS

-- + I

NORMAL. BDL tRNA 3T3

H F L _|I E Li | | g ig 5ulq

Figure6.Actin mRNA in liver cells. Autoradiogramillu

steady-statelevelsof

_fl-actin

m'RNAin cellsisolated fror andfibroticliver. AbbreviationsasinFig.2.Actin

signa

gestin endothelialcells,followedbylipocytesandhepat, relationshipdoesnotchangeafter eitherbile ductligatio

carbon tetrachloride' administration(not shown).

purified

liver cell

isolates,

wehavebeen ablenota ize

collagen

mRNA to

specific population's

of liv alsoto

quantitate

the relative amount of

collagei

each

'Cell.

type.

Furthermore,

because all studie

formedon cells

immediately following isolation,

relevanttotheliver invivo. The

pronounced

incr4

Iand III

collagen

mRNA inthe

lipocyte popullatil

central roleto'thesecellsin thefibrotic processin Extracellular matrix gene

expression

increase

lipocytes

aftera

fibrogenic

stimulus. Carbontetra, ministrationpromotesa marked increaseinintei

gengene

expression by day 14,

before the onseta

cirrhosis. Bile duct ligation results in a similar

only

5d,butsubstantialconnective tissue

depositi

apparent

histolo

preventing

a firm conclui

ing

the time course ofextracellular matrix gen with respectto matrix

protein

production in this

fibrosisis

_{preceded by}

anincrease in

collagen

ml

gested

by

other

studies,

which demonstrate that

tyi

geneexpression in whole liver rises within 24 h af

ligation (32).

Both bile duct

ligation

and carbon tetrachlori

tration,

two

mechanistically

distinct typesof liver duce similar

changes

inbymatrix gene expression mal liver cells. Bile duct ligation results in conni

deposition

around

portal

triads with~a

_paucityc

inflammation ornecrosis. Carbon

tetrachloride,

hand,

produces

a

pericentral pattern

of

fibrosis,

w

ent parenchymal

injury.

Small

quantitative

diffe

observed in

steady-state

levels of

collagen

mRNA

tiveas

compared

totoxin-induced fibrosis

(Fig. 4)

ably

reflect themore

rapid

time course offibrosism

(33,

34).

Apart

from the

quantitative results,

th,

type

I

collagen

gene expression in both models s

mesenchymal

cell activation is a central

pathoge

bothtypes of fibrotic liver

injury.

Sinusoidal endothelial cellsaswellas

lipocyte-nificant

increases

intypeI

collagen

mRNAaftere

liver injury.This i not supisionguin view ofe

endothelial cells from other organs are active in matrix synthe-1 sis (35, 36). It is possible that some ofthe type I collagen mRNA measured in endothelial cells derives from

contami-S

Trl

nating lipocytes; if so, however, similar increases would be

expected forcollagentypes IIIand IV andlaminin.The results indicate that mRNAs for the latter three matrix proteins do not modulate inparallelinendothelial cells and lipocytes; in

fact,type IVcollagen mRNA levels varyinverselyin the two 1

IActin

populations during fibrogenesis.Thus,endothelial cells them-j selves must not be ignored as potential contributors to hepatic

fibrosis in vivo.

s -609 bp In neither model of

hepatic

fibrogenesis

isextracellular

matrix gene expression modulated in hepatocytes. Indeed, mRNA for all collagen types is barely detectable in hepato-cytes fromeither normalorinjured liver,suggestingthat

par-istrating enchymal cells are minimally active in collagen synthesis. This

mnormal conclusion contradicts that of Brenner and colleagues (37), ls are stron- who have recently reported that hepatocytes from normal rats tocytes.This contain up to three times as much type I collagen mRNA as n(BDL) or nonparenchymal cells. The latter investigators prepared non-parenchymal cells by collagenase perfusion and centrifugal elutration, under conditions thatselectivelyremove both lipo-cytesandsinusoidal endothelialcells. Theremaining nonpar-inly to local- enchymalfractionwaspresumably rich inKupffercells, which tercells, but wouldnotbeexpectedtocontain type IcollagenmRNA.

He-n mRNAin patocytes in their study did exhibit strongsignals fortype I s were per- collagenmRNA; their methods, though, donotpermitarigid

the dataare comparison of collagen gene expression between hepatocytes easein types and endothelial cells or lipocytes. Earlierstudies from this ionassignsa group suggested thatparenchymal cells produce the bulk of

vivo. hepatic collageninvivo, bothin normal and carbon tetrachlo-srapidly in ride-treatedrats(10, 38). Theseconclusionswerebasedonin

chloridead- vivolabelingofnewlysynthesized collagen,followedby deter-rstitialcolla- mination of the cellular source of the protein upon extraction

ofhistologic fromtheintactorgan.Solubilizationand recoveryof collagen change after fromwholeliver, however, isnotquantitative, raisingthe pos-onis already

sibility

that the collagen analyzed in these studies was not

ision regard-

representative

oftotalhepatic collagen.Ourexperiments

con-e expression firmthe resultsof in situ

hybridization

studies,

in which

non-model.That parenchymal cells contribute most significantlytohepatic fi-RNAissug- brosisinvivo.

peIcollagen Theonly extracellular matrix mRNA

readily

detected in

fter bileduct hepatocytesinourexperimentswasthat for laminin. Identifi-cation of laminin mRNA in these cells is unlikelytoreflect

ide adminis- contamination of thepopulation by nonparenchymal cells,as

injury,pro- hepatocyte isolates prepared by

centrifugal

elutriation are ymesenchy- >98%pure(Table 1).NochangeinlamininmRNAwasnoted

ective tissue inhepatocytes after

experimental

liver

injury;

inview ofthe fassociated small changes in laminin gene expression in all liver cells,

ontheother however, the relevance of this

finding

to the intact liver re-vith anteced- mainstobeclarified (see below).

trences

were Type IVcollagen and laminin gene expression were not in obstruc- significantly enhanced in the current

experiments,

in any of i;these

prob-

the cell typesexamined.Theseresultsare

unexplained,

in view n theformer of previousdataindicatinganincreasein both

type

IV colla-ie pattern of genandlaminin

during

hepatic

fibrosisinvivo

(4).

Itis

possi-suggests that ble that enhancedtranslation of

existing

type IV and laminin

nic event in mRNAaccountforsuchanincrease;

alternatively,

alterations in type IVcollagenandlamininmRNA may bedelayed rela-sexhibit

sig-

tivetocollagentypes I and

III,

andthusnotobserved

during

xperimental

early

fibrogenesis

as

exemplified

in the current

experiments.

vidence that Pierce and colleagues (39) havefoundthat in wholeliver,type

IV collagen geneexpression does not increasesignificantly for 5 wkfollowing carbon tetrachlorideadministration. Afterbile duct ligation, however, type IVcollagen andlaminin mRNA areincreasedfourfoldwithin 5 d (32). This suggests that biliary ductular cells, which proliferate rapidly following bile duct ligation, may contribute to the increase in this model of fi-brosis. Milaniand colleagues (13) haverecently shownbyin

situhybridization that type IVcollagen mRNA is located in the vicinity ofbiliary epithelial cells in experimental biliary

fibrosis.

Basedonthe workofClaytonandDarnell (20),who dem-onstrated thatspecificcellular mRNAsarepreserved in hepa-tocytesfollowing collagenase perfusion, we areconfidentthat our assessmentofextracellularmatrixgene expression in these cellsisrepresentative ofthat invivo. While selective

degrada-tion of mRNA may occur in nonparenchymal cells during isolation,such alosswould resultinunderestimation of extra-cellularmatrixgeneexpression by endothelialcells and

lipo-cytes. Given that the existing results already demonstrate a

significant difference betweenhepatocytes and

nonparenchy-mal cells,thisconcern doesnotaffectouroverallconclusions. The quantitative measurements ofextracellular matrixgene

expressioninparenchymalandnonparenchymalcells

(Fig. 4)

suggestthat extracellular matrixmRNAsaremoreabundant inlipocytesand endothelialcells than in hepatocytes. This is

truefortypes III and IVcollagenandlamininin normal

liver,

and for all the matrix genes examined during fibrogenesis.

Thesecomparisonsassumethathepatocytes, sinusoidal

endo-thelial cells, and lipocytes contain similar amounts oftotal cellular RNA. In theeventthat total RNAcontentis different

in each cell type, ourresults cannot be usedto interpret the

relative contribution ofeach cell typetohepatic fibrosison a

per-cell basis.

Studies

are currently underway to clarify this

issue. Based on RNA/DNA ratios measured in hepatocytes (40) and mesenchymal cells (41), the total RNA content of lipocytes is estimated at 42% ofthat in hepatocytes.-Taking

into accountthisdifference,theincreaseininterstitialcollagen gene expression in lipocytes still far exceeds that in

hepato-cytes. Thus,

lipocytes,

andto alesserextentendothelial cells,

appeartobe themaincontributorstohepatic fibrosisin vivo.

Extracellular

matrixprotein productionwasnotmeasured inparallel withcollagengeneexpressionin theseexperiments.

Studies

of

lipocytes

inculture, however, demonstratethat in-creasesincollagenmRNAareassociated withincreased colla-gen synthesis (Maher, J. J., unpublished observations).

Al-though itcannot beexcluded that posttranscriptional events controlsynthesis ofextracellular matrixbylipocytesinvivo, it

isunlikelythattheir contribution tofibrosisis overshadowed

bythatofhepatocytes, which contain littleor no matrix-spe-cificmRNA.

Itisof interestthat innormalliver,both parenchymal and nonparenchymal cells contain very low levels of type I colla-gen mRNA. Evenusingasensitiveassay(RNaseprotection),

typeIcollagen mRNA isdifficulttoidentifyin anyofthe cell types examined. This supports the concept that in normal adult liver, synthesis oftype I collagen is minimally active.

Both bileduct ligation and carbon tetrachloride administra-tion alterthisstate.Although themechanismwherebymatrix

synthesis

is activated is unknown, our studies suggest that these different modes ofinjury elicit a similarresponse and thus may actthrough common signals. These could include

humoral

factors,

such as

cytokines

elaborated

by

recruited

macrophagesorresident livercells

(42-44),

orstimuli in the

form of altered cell-matrix

interactions, provoked

bylocal re-leaseof

matrix-degrading

proteinases

during

liver

injury

(45). Exploration ofthesepossibilities is currently underway,

using

theapproach outlinedhere.

Acknowledgments

Expert technical assistance wasprovided by Melanie Mark, Glenn Yamasaki,and John Niemiec.

ThisworkwassupportedbyU.S. Public Health Servicegrantsfrom theNational Institute of Alcohol Abuse andAlcoholism(AA-078 10)

andthe National Institutes of Health(DK-31198, 1 P-30DK-26743, andT32DK-07007).

References

1.Rojkind, M.,M. A.Giambrone,and L.Biempica. 1979. Colla-gen typesin normalandcirrhoticliver.Gastroenterology.76:710-719.

2. Murata, K., M. Kudo, R. Onuma, and T. Motoyama. 1984.

Changesofcollagentypesatvariousstagesof human liver cirrhosis.

Hepato-Gastroenterology.31:158-161.

3.Seyer,J.M.,E.T.Hutcheson,and A. H.Kang. 1977.Collagen polymorphismin normal and cirrhotichuman liver. J. Clin. Invest.

59:241-248.

4.Hahn,E.,G.Wick,D.Pencev,and R.Timpl.1980. Distribution

ofbasement membraneproteinsin normal and fibrotic human liver:

collagentypeIV,laminin,and fibronectin. Gut. 21:63-71.

5.Martinez-Hernandez,A. 1984.Thehepaticextracellular matrix. I. Electron immunohistochemical studies in normal rat liver. Lab. Invest.51:57-74.

6. Clement,'B., H. Emonard, M. Rissel, M. Druguet,J.-A.

Gri-maud,D.Herbage,M.Bourel,andA.Guillouzo. 1984. Cellularorigin

ofcollagenand fibronectininthe liver. Cell.Miol. Biol. 30:489-496. 7.Clement, B.,M.Rissel,S.Peyrol,'Y. Mazurier,J.-A.Grimaud,

and A. Guillouzo. 1985. A procedure forlightand electron

micro-scopicintracellular immunolocalizationofcollagenand fibronectin in

ratliver.J.Histochem.Cytochem. 33:407-414.

8.Sakakibara, K.,S.Igarashi,and T. Hatahara. 1985.Localization oftype IIIprocollagenaminopeptide antigenicityinhepatocytesfrom cirrhotic human liver. VirchowsArch. A Pathol. Anat. Histol. 408:219-228.

9.Martinez-Hernandez,A. 1985. Thehepaticextracellularmatrix.

II. Electron immunohistochemical studiesinratswithCCl4-induced

cirrhosis.Lab.Invest.53:166-186.

10.Chojkier,M.,K. D.Lyche,and M.Filip. 1988. Increased

pro-duction ofcollageninvivoby hepatocytesandnonparenchymalcells inratswithcarbontetrachloride-inducedhepaticfibrosis.Hepatology (Baltimore).8:803-814.

11.Milani, S.,H.Herbst,D.Schuppan,E. G.Hahn,andH.Stein. 1989.Insituhybridization forprocollagentypesI,IIIand IV mRNA in normal andfibroticratliver:evidence forpredominant expression

innonparenchymallivercells.Hepatology (Baltimore). 10:84-92. 12. Milani, S., H. Herbst, D. Schuppan, E. 0. Riecken, and H.

Stein. 1989. Cellularlocalizationoflaminin genetranscriptsinnormal andfibrotichumanliver.Am.J.Pathol. 134:1175-1182.

13.Milani, S.,H.Herbst,D.Schuppan,K. Y.Kim,E.0. Riecken,

and H. Stein. 1990. Procollagen expressionby nonparenchymal rat

liver cellsinexperimental biliaryfibrosis. Gastroenterology.

98:175-184.

14.Friedman,S.L.,F.J.Roll,J.Boyles,and D. M.Bissell. 1985. Hepatic lipocytes:theprincipalcollagen-producingcellsofnormalrat

liver. Proc.Natl. Acad.Sci. USA.82:8681-8685.

15. Senoo, H., R.-I. Hata, Y.Nagai,and K. Wake. 1984. Stellate cells(vitamin A-storing cells)aretheprimarysite ofcollagensynthesis

innonparenchymalcellsin theliver.Biomed.Res.5:451-458.

Collagen measured in primaryculturesofnormal rathepatocytes

de-rives fromlipocytes within the monolayer. J. Clin. Invest. 82:450-459. 17.Friedman,S. L., F. J. Roll, J.Boyles, D. M. Arenson, and D. M.

Bissell. 1989. Maintenance ofdifferentiated phenotype of cultured rat

hepatic lipocytes by basementmembrane matrix. J. Biol. Chem. 264:10756-10762.

18.Davis, B.H., B. M. Pratt, and J. A. Madri. 1987. Retinol and

extracellularcollagen matrices modulate hepaticItocellcollagen phe-notype and cellular retinol bindingprotein levels. J. Biol. Chem.

262:10280-10286.

19. Weiner, F. R., M. A. Giambrone, M.J. Czaja, A. Shah, G.

Annoni, S. Takahashi, M. Eghbali, andM. A. Zern.1990.Ito-cell gene

expression and collagenregulation. Hepatology (Baltimore). 11:

111-117.

20. Clayton,D. F.,andJ. E.Darnell. 1983. Changesin liver-spe-cific comparedtocommon genetranscription during primary culture ofmousehepatocytes. Mol. Cell. Biol. 3:1552-1561.

21.Irving,M. G., F. J.Roll, S. Huang, and D. M. Bissell. 1984.

Characterization and culture of sinusoidal endothelium fromnormal

ratliver:lipoprotein uptake and collagen phenotype. Gastroenterology. 87:1233-1247.

22.Friedman,S. L.,andF. J. Roll. 1987.Isolationandcultureof hepatic lipocytes, Kupffer cells, and sinusoidal endothelial cells by density gradient centrifugation with Stractan. Anal. Biochem.

161:207-218.

23. Maher,J.J.,S.L.Friedman,F.J.Roll, andD. M.Bissell. 1988. Immunolocalization of laminin in normalratliver and biosynthesis of laminin byhepatic lipocytes in primary culture. Gastroenterology. 94:1053-1062.

24.Genovese, C.,D.Rowe, and B. Kream. 1984. Construction of

DNA sequencescomplementaryto ratalanda2collagenmRNAand their usein studying the regulation oftype Icollagen synthesis by

1,25-dihydroxyvitaminD.Biochemistry.23:6210-6216.

25.Kurkinen, M.,M.R.Condon,B.Blumberg,D. P.Barlow, S. Quinones, J. Saus, and T. Pihlajaniemi. 1987. Extensive homology

between thecarboxyl-terminal peptidesofmouseal(IV) anda2(IV) collagen.J.Biol. Chem. 262:8496-8499.

26.Barlow, D. P.,N. M.Green,M.Kurkinen,andB. L. M.Hogan. 1984.Sequencingof laminin B chain cDNA's reveals C-terminal

re-gions of coiled-coil alpha-helix. EMBO(Eur. Mol. Biol. Organ.) J.

3:2355-2362.

27.Gunning,P.,P.Ponte,H.Okayama,J.Engel,H.Blau, andL.

Kedes. 1983. Isolation and characterization offull-length cDNA

clones for human a-,

P-,

and y-actinmRNAs:skeletal butnot

cyto-plasmic actinshave anamino-terminal cysteinethatis subsequently removed.Mol. Cell. Biol. 3:787-795.

28.Melton,D.A.,P. A.Krieg,M.R.Rebagliati, T.Maniatis,K.

Zinn,andM.R.Green. 1984. Efficient in vitrosynthesisof biologi-cally activeRNAand RNAhybridization probesfromplasmids

con-tainingabacteriophageSP6promoter. Nucleic AcidsRes. 12:7035-7056.

29.Gilman,M. 1987.Ribonucleaseprotectionassay. InCurrent Protocols in MolecularBiology.Vol. 1. F.M.Ausubel,R.Brent,R.E.

Kingston, D. D. Moore,J.G. Seidman,J. A. Smith,and K.Struhl,

editors. Greene Publishing Associatesand Wiley Interscience, John

Wileyand Sons, Inc., New York. 4.7.1-4.7.8.

30.Berger,S.L. 1987.Preparationandcharacterizationof RNA:

overview. Methods Enzymol. 152:215-227.

31.Mak, K. M., and C. S.Lieber. 1988.Lipocytesand transitional

cells in alcoholic liver disease: amorphometric study. Hepatology (Baltimore). 8:1027-1033.

32.Maher, J. J. 1989. Rapidinductionofextracellular matrix gene

expression and collagen synthesis following bileductligation inthe rat.

Hepatology (Baltimore). 10:631.(Abstr.)

33. Kountouras, J., B. H. Billing, and P.J. Scheuer. 1984.

Pro-longed bile duct obstruction:a newexperimentalmodelfor cirrhosis in therat. Br.J. Exp.Pathol. 65:305-311.

34. Perez-Tamayo, R. 1979. Cirrhosis ofthe liver: a reversible

disease? Pathol. Annu. 14:183-213.

35. Madri,J. A., andS.K.Williams. 1983. Capillaryendothelial

cell cultures: phenotypic modulation by matrixcomponents. J. Cell Biol. 97:153-165.

36. Tseng, S. C. G.,N.Savion,D.Gospodarowicz, and R. Stem. 1983. Modulation ofcollagensynthesis byagrowth factor andbythe extracellularmatrix:comparison of cellularresponse to twodifferent stimuli.J. Cell Biol. 97:803-809.

37. Brenner,D. A., J. M.Alcorn, S.P.Feitelberg,H. L.Leffert, and

M.Chojkier. 1990.Expression of collagengenes inliver. Mol. Biol. & Med. 7:105-115.

38.Chojkier,M. 1986.Hepatocytecollagenproduction in vivo in normal rats.J. Clin. Invest. 78:333-339.

39. Pierce, R. A.,M. R.Glaug,R.S.Greco, J.W.Mackenzie, C.D.

Boyd, and S. B. Deak. 1987. Increased procollagen mRNAlevelsin carbon tetrachloride-induced liver fibrosis in rats. J. Biol. Chem. 262:1652-1658.

40. Flaim, K. E., S. M. Hutson, C. E. Lloyd,J. M. Taylor, R.

Shiman, andL.S.Jefferson. 1985. Direct effect of insulinonalbumin

geneexpression in primary cultures ofrathepatocytes.Am.J.Physiol. 249:E447-E453.

41.Tolstoshev, P., R. Haber, B.C. Trapnell, andR. G.Crystal. 1981. Procollagen messenger RNA levels andactivity andcollagen synthesisduring the fetal development ofsheep lung,tendon, and skin. J. Biol. Chem. 256:9672-9679.

42. Friedman, S. L., and M. J. P. Arthur. 1989. Activation of culturedrathepaticlipocytesbyKupffercellconditioned medium.J.

Clin.Invest. 84:1780-1785.

43.Czaja, M. J., F.R.Weiner,K.C.Flanders,M. A.Giambrone,

R.Wind,L.Biempica,andM. A.Zern. 1989. The in vitro and in vivo association oftransforminggrowth

factor-#l

with hepaticfibrosis. J. CellBiol. 108:2477-2482.

44. Weiner, F.R.,A.Shah, L. Biempica,M. A. Zern,and M. J.

Czaja. 1989. Cellularsourcesof increasedcollagenandtransforming growth

factor-#

l geneexpressioninfibroticratliver.Hepatology (Bal-timore). 10:629. (Abstr.)

45.Arthur,M.J.P.,S. L.Friedman,F.J.Roll, andD. M.Bissell. 1989. Lipocytes from normal ratliver releaseaneutral metallopro-teinase that degradesbasementmembrane(typeIV) collagen.J.Clin.

Invest.84:1076-1085.