Apoptosis of hepatic stellate cells: involvement inresolution of biliary fibrosis and regulation bysoluble growth factors

R Issa, E Williams, N Trim, T Kendall, M J P Arthur, J Reichen, R C Benyon, J P Iredale

AbstractBackground—Activated hepatic stellatecells (HSC) are central to the pathogen-esis of liver fibrosis, both as a source offibrillar collagens that characterise fibro-sis and matrix degrading metalloprotein-ases and their tissue inhibitors, theTIMPs.Aims—To test the hypothesis that HSCapoptosis is critical to recovery frombiliary fibrosis and that soluble growthfactors may regulate HSC survival andapoptosis.Methods—Rats (n=15) were subjected tobile duct ligation for 21 days, after whichbiliodigestive anastomosis was under-taken (n=13). Livers were harvested atfixed time points of recovery for periods ofup to 42 days. Numbers of activated HSCswere quantified after á smooth muscleactin staining and HSC apoptosis wasdetected by terminal UDP-nick end label-ling (TUNEL) staining and quantified ateach time point. HSC apoptosis wasquantified in vitro in the presence orabsence of insulin-like growth factor(IGF)-1, IGF-2, platelet derived growthfactor (PDGF), and transforming growthfactor â1 (TGF-â1).Results—Following biliodigestive anasto-mosis after 21 days of bile duct ligation,rat liver demonstrated a progressive reso-lution of biliary fibrosis over 42 days,associated with a fivefold decrease in acti-vated HSC determined by á smooth mus-cle actin staining. TUNEL stainingindicated that loss of activated HSCresulted from an increase in the rate ofapoptosis during the first two days postbiliodigestive anastomosis. Serum depri-vation and culture in the presence of50 µM cycloheximide was associated withan increase in HSC apoptosis which wassignificantly inhibited by addition of 10ng/ml and 100 ng/ml IGF-1, respectively(0.05>p, n=5). In contrast, 1 and 10 ng/mlof TGF-â1 caused a significant increase inHSC apoptosis compared with serum freecontrols (p<0.05, n=4). PDGF and IGF-2were neutral with respect to their eVect onHSC apoptosis.Conclusion—HSC apoptosis plays a criti-cal role in the spontaneous recovery frombiliary fibrosis. Both survival and apopto-sis of HSC are regulated by growth factorsexpressed during fibrotic liver injury.(Gut 2001;48:548–557)

Keywords: hepatic stellate cells; apoptosis; hepaticfibrosis; insulin-like growth factor; transforminggrowth factor â1

Liver fibrosis represents the common end pointof the majority of chronic liver injuries.1 2 Ulti-mately, it results in distortion of the liver archi-tecture (cirrhosis) which is associated with dis-turbance of liver function and significantmorbidity and mortality. At the cellular levelthere is now a wealth of evidence indicatingthat the hepatic stellate cell (HSC) representsthe pivot of the fibrotic process.1–3 During liverinjury, HSCs undergo transformation from aretinoid rich pericyte-like cell to amyofibroblast-like cell, a process termed acti-vation.1 2 Highly activated HSCs are morpho-logically indistinguishable from myofibrob-lasts. The activated HSCs express collagen Iand other extracellular matrix genes and arequantitatively the major source of the matrixwhich accumulates during fibrosis.3 4 Duringactivation, HSC enter the cell cycle with theresult that the hepatic matrix accumulation isthe result of an overall increase in the numberof HSCs in addition to changes in HSC geneexpression.1–4

We have previously demonstrated that recov-ery from relatively advanced CCl4 induced liverfibrosis can occur in a reproducible manner.5

Recovery is associated with remodelling of theexcess liver matrix resulting in restitution ofnear normal liver architecture. An essentialelement of this recovery process is apoptosis ofactivated HSCs.5 6 An understanding of thecontrol of HSC apoptosis is important pre-cisely because regulating this process may pro-vide a novel therapeutic approach to the treat-ment of advanced hepatic fibrosis.5–7 As part ofour previous study we showed that HSCscultured from rat liver respond to absoluteserum deprivation by undergoing apoptosis,indicating that activated HSCs undergo apop-tosis as a default pathway, once the cell isdeprived of specific survival signals. In the cur-rent study we have described the role of HSCapoptosis in a mechanistically distinct model ofreversal of fibrosis (bile duct ligation followed

Abbreviations used in this paper: HSC, hepaticstellate cell; IGF-1 (-2), insulin-like growth factor 1(2); TGF-â1, transforming growth factor â1; PDGF,platelet derived growth factor; BDL, bile duct ligation;TUNEL, terminal UDP-nick end labelling; FCS, fetalcalf serum; á-SMA, á smooth muscle actin; DMEM,Dulbecco’s modified Eagle’s medium; TE, tris EDTA;[3H]-TdR, [methyl-3H]-thymidine; FasL, Fas ligand.

Gut 2001;48:548–557548

Liver Research Group,Division of Cell andMolecular Medicine,Level D, South Laband Path Block,Southampton GeneralHospital,Southampton, UKR IssaE WilliamsN TrimT KendallM J P ArthurR C BenyonJ P Iredale

Department of ClinicalPharmacology,University of Berne,Murtenstrasse 35, 3010Berne, SwitzerlandJ Reichen

Correspondence to:J P Iredale, Mail Point 811,Level D, South Lab and PathBlock, Southampton GeneralHospital, Tremona Road,Southampton SO16 6YD,UK. jpi@soton.ac.uk

Accepted for publication17 October 2000

www.gutjnl.com

by reanastomosis8 9) and have also character-ised the response of activated HSCs to growthfactor deprivation in detail to determine thepotential roles of insulin-like growth factor(IGF)-1, IGF-2, transforming growth factor â1

(TGF-â1), and platelet derived growth factor(PDGF) in acting as survival factors for HSCs.Our data support the hypotheses that HSCapoptosis is a general response in the recoveryfrom fibrosis and that specific growth factorsand cytokines may regulate survival of HSCs.

MethodsBILE DUCT LIGATION AND ANASTOMOSIS

Bile duct ligation (BDL) was undertaken in 15male Sprague-Dawley rats exactly as previouslydescribed.9 After 21 days of BDL, 13 animalsunderwent a further laparotomy at which aRoux-en-Y choledochal-jejunostomy was per-formed to eVect biliary drainage, as previouslydescribed.9 At time points before and followingreanastomosis, groups of animals were killedand the liver harvested, fixed in buVeredformalin, and blocked in paraYn. Livers wereharvested at 21 days post BDL (peak fibrosis,n=2) and at one day (n=3), two days (n=4),seven days (n=4), and 42 days (n=2) of recov-ery post biliary-jejunal anastomosis. In addi-tion, control livers subjected to a shamoperation (n=4) were harvested at 21 days andincluded in the analysis described below.

HISTOLOGICAL ANALYSIS AND QUANTITATION OF

ACTIVATED HSC DURING RECOVERY FROM BDL

Blocks containing four lobes from each fixedliver were sectioned and stained with haema-toxylin and eosin, Sirius red, and reticulin.Further sections were deparaYnised and sub-jected to microwave antigen retrieval beforebeing stained for á smooth muscle actin(á-SMA), as described previously.5 Afterwards,the number of á-SMA positive cells werecounted in 50 random high power fields of eachof the four lobes—that is, the mean number ofactivated HSCs per high power field was deter-mined from 200 individual observations.

DETERMINATION OF NUMBERS OF

NON-PARENCHYMAL CELLS UNDERGOING

APOPTOSIS FOLLOWING BILIARY-JEJUNAL

ANASTOMOSIS

Sections of each model liver were stained forDNA fragmentation characteristic of apoptosisby the terminal UDP-nick end labelling(TUNEL) reaction, as previously described,5

with the modifications recently described toreduce false positivity.10 Each slide was thenanalysed by a blinded observer who countedthe number of non-parenchymal TUNELpositive apoptotic figures per 50 high powerfield (×40) per liver lobe section—that is, themean number of apoptotic non-parenchymalcells per high power field was determined from200 observations. Apoptotic bodies in a distri-bution consistent with hepatocytes were notcounted. Dual staining for á-SMA andTUNEL was undertaken in representative liversections as previously described to localiseapoptosis to HSCs.5

HSC EXTRACTION AND CULTURE

HSCs were extracted from normal rat liver bysequential in situ perfusion with pronase andcollagenase, as previously described.11 Cellswere seeded onto uncoated tissue culture plas-tic in Dulbecco’s modified Eagle’s medium(DMEM) with 16% fetal calf serum (FCS) andgrown for 10–14 days until a monolayer ofmyofibroblast-like activated HSCs was ob-tained. Cells were then trypsinised and re-plated onto 24 well uncoated plastic and main-tained in DMEM containing 16% FCS.Experimental manipulations were performedafter passaging when cells had reached 80%confluence.

QUANTIFICATION OF APOPTOSIS IN HSC CULTURE

Passaged HSCs were washed twice in serumfree media before being preincubated in serumfree media for one hour. Thereafter the mediafrom the preincubation was discarded and cellswere cultured for defined periods of time inserum free DMEM containing 0.1% bovineserum albumin. In each experiment parallelcells were returned to culture in media supple-mented with 16% FCS. The eVect of the spe-cific growth factors on HSC survival wasdetermined by adding predetermined concen-trations of each growth factor to pairs of wellsfor periods of up to 24 hours. The growth fac-tors studied were PDGF-AB (Sigma, Poole,UK), IGF-1 and IGF-2 (Peprotech, London,UK), and TGF-â1 (Sigma, Poole, UK). Cellswere also cultured in the presence of cyclohex-imide (6.25–50 µM final concentration) in thepresence and absence of 16% FCS.

At the appropriate time apoptosis was quan-tified by inverted fluorescence microscopy afteraddition of acridine orange to previouslyundisturbed wells. The total number of normaland apoptotic cells was determined in each ofthree random high power fields (×100) fromthe centre of each well and duplicate wells wereexamined for each culture condition at eachtime point. Thus the mean rate of apoptosis foreach individual experiment is based on sixobservations and the results presented are themean of 3–5 independent experiments. Apop-totic cells were identified by their shrunken,absent, or budding cytoplasm, condensed bud-ding and fragmenting nuclei, and fluorescenceenhancement by condensed chromatin, asdescribed previously.5 12 Cells were countedfirst on the adherent monolayer and then in thesupernatant above each field to include anydetached apoptotic cells. To confirm that thecells rounding up on the surface of the mono-layer were apoptotic, the loosely adherent cellsfrom representative wells were harvested bygentle washing and a cytospin preparation wasmade and stained as previously described.5 Inaddition, further preparations of the looselyadherent condensed cells were washed andharvested by centrifugation. The cellular DNAwas extracted from the resulting pellet andsubjected to analysis by electrophoresis, asdescribed by Baker and colleagues,12 to provideconfirmatory evidence of apoptosis. To furthervalidate the acridine orange staining method of

Apoptosis of hepatic stellate cells 549

www.gutjnl.com

quantifying apoptosis and to confirm the anti-apoptotic aVect of IGF-1, further parallelstudies were undertaken. In these experiments,apoptosis was quantified using FACS analysisafter staining with an anti-single strandedDNA antibody (Chemicon, Temecula, Califor-nia, USA), propidium iodide staining, andannexin V binding (Calbiochem, Nottingham,UK) according to the manufacturer’s instruc-tions. In addition, caspase 3 activity was deter-mined in cytosolic extracts from HSCs fromparallel experiments using the CaspACE assaysystem according to the manufacturers’ in-structions (Promega, Chilworth, UK).

EFFECT OF SOLUBLE GROWTH FACTORS ON3H-THYMIDINE INCORPORATION BY HSC

To examine the eVects of absolute serum dep-rivation and exogenous growth factors onDNA synthesis under conditions identical tothose used to quantify apoptosis, [methyl-3H]-thymidine ([3H]-TdR) incorporation wasdetermined. Passaged cells grown in 24 wellplates were washed three times in serum free

media as described. Cultures were then incu-bated in serum free DMEM with 0.1% bovineserum albumin alone, or supplemented withPDGF-AB (10 ng/ml), IGF-1 (10 ng/ml), orTGF-â1 (10 ng/ml) for 19 hours. At this time0.5 µCi [3H]-TdR (925 GBq/mmol) was addedto each well as 0.5 µl [3H]-TdR in 10 µl ofserum free DMEM. Cultures were thenincubated for a further five hours. Afterincubation, [3H]-TdR incorporation was deter-mined by scintillation counting of the cell layeras previously described.13

EFFECT OF GROWTH FACTORS ON CELL NUMBERS

To determine the net eVect of growth factormanipulation on overall HSC numbers, whichis a function of both cell proliferation and celldeath, genomic DNA content of cell monolay-ers was quantified by Picogreen staining(Molecular Probes, Ireland). Passaged HSCswere cultured in 12 well plates and exposed toserum deprivation or growth factors in dupli-cate in an identical manner to that describedabove. After 24 hours of incubation, the media

Figure 1 Histological analysis at peak fibrosis, 21 days after bile duct ligation (BDL). Livers harvested from animals 21days after BDL were stained with reticulin (A, ×10) and Sirius red (B, ×20). A highly distinct pattern of matrix andcollagen is apparent in the expanded portal tracts. This neomatrix extends in a septate manner into the parenchyma, inplaces linking portal tracts, and beginning to distort the liver architecture. Histological analysis of animals 42 days afterbiliary-jejunal anastomosis: following biliary-jejunal anastomosis there was a progressive diminution in the stainablecollagen and matrix both within the parenchyma and adjacent to the bile ducts. At 42 days post anastomosis, reticulinstaining (C, ×10) and Sirius red staining (D, ×10) demonstrated that considerable matrix remodelling had taken placewith a return to near normal patterns of collagen and matrix.

550 Issa, Williams, Trim, et al

www.gutjnl.com

was removed and adherent cells were harvestedby scraping into 150 µl of Tris EDTA (TE)(pH 7.5). The resulting suspension wassonicated for 15 minutes. Sonicated product(100 µl) was then diluted 1:1 with TE andmixed with 100 µl of Picogreen, prediluted1:200 in TE. The reaction mixture was placedin wells of a 96 well plate and incubated indarkness for five minutes at room temperature.Controls of Herring sperm DNA diluted from0 to 3 µg/ml were simultaneously prepared andanalysed in parallel. The fluorescence of thesamples was measured with a Cytofluor IIMicrowell fluorescence reader (PerSpectivesBiosytems, Framingham, Massachusetts,USA) and DNA concentration of each sampledetermined by reference to a standard curvegenerated from analysis of the control DNA.

DETECTION OF Fas/FasL EXPRESSION BY HSC

HSCs have been demonstrated recently toexpress both Fas and Fas ligand (FasL) and theautocrine production of FasL has been sug-gested as an apoptotic mechanism for activatedHSCs. We undertook ribonuclease protectionanalysis for Fas and FasL on HSC RNA. TotalRNA was extracted from HSCs using theRNeasy method (Qiagen, Crawley, UK) afterserum deprivation, with or without IGF-1 orPDGF treatment. The ribonuclease protectionassay for Fas/FasL was undertaken as previ-ously described.14 Riboprobes were transcribed

using the rAPO-1 multiprobe template (Am-bion, Oxon, UK) according to the manufactur-er’s instructions. The probe template con-tained a GAPDH control to confirm equalsample loading.

DATA ANALYSIS

Results are expressed as mean (SEM). Statisti-cal analysis was performed using the Students’t test (two tailed). Significance was assumedwhen p<0.05.

ResultsHISTOLOGY OF SPONTANEOUS RECOVERY FROM

BILE DUCT LIGATION

The histology of the results of BDL and recov-ery from BDL induced fibrosis has beendescribed in detail in a previous report.9 Theresult of Sirius red and reticulin stainingconfirmed that the collagen fibrils, whichextended out from the expanded portal triadsafter 21 days of BDL, became progressivelyremodelled during the 42 days of recovery fol-lowing biliary-jejunal anastomosis. At peakfibrosis (21 days following BDL) there was ahighly distinct pattern of collagen apparent inthe portal triads. This surrounded the prolifer-ating bile ductules and extended into theparenchyma in places forming distinct septaelinking portal triads (fig 1A, B). During the 42days of recovery following biliary-jejunal anas-tomosis there was progressive diminution of

15

10

5

021 1 2 7 42 Sham

C

α S

MA

po

siti

ve c

ells

per

hig

h p

ow

er f

ield

Days afterbile ductligation

Days afterbile duct

anastomosis

Figure 2 Alpha smooth muscle actin (á-SMA) stainingof each liver sample was used to demonstrate activatedhepatic stellate cells (HSCs). A progressive diminution ofthe number of positive cells was observed when fibrotic liver,21 days after bile duct ligation (BDL), was compared withliver samples harvested after biliary-jejunal anastomosis.Representative examples are shown of á-SMA staining atpeak fibrosis (21 days after BDL) (A) and after 42 daysof biliary-jejunal anastomosis (B) (each at ×10). ActivatedHSCs are observed within and around the periportalneomatrix and around the developing fibrotic septae (A).After 42 days of recovery, residual activated HSC arepresent in the periportal region but are considerablydiminished in number. (C) Overall results of quantificationof numbers of activated HSC defined by á-SMAexpression. There is progressive diminution in activatedHSCs following biliary-jejunal anastomosis, but even after42 days the number is still raised above that observed insham operated controls.

Apoptosis of hepatic stellate cells 551

www.gutjnl.com

stainable collagen. Even after seven days ofrecovery the septae became less defined andthe matrix surrounding the bile ductulesdiminished. This process continued until at 42days there remained only remnants of collagenand matrix in the portal triads (fig 1C, D).There was no notable histological abnormalityor evidence of stainable collagen in any of thesham operated control livers.

SPONTANEOUS RECOVERY FROM FIBROSIS IS

ASSOCIATED WITH A DECREASE IN NUMBERS OF

ACTIVATED HSC

Because á-SMA is expressed by activatedHSCs, immunostaining for this protein wasused to detect and quantify numbers ofactivated HSCs.15 Fibrotic liver at 21 days wasdemonstrated to contain á-SMA positive cellsin the region that had been previously demon-strated to be strongly stained with Sirius red. Inaddition, á-SMA positive cells were observedextending into the parenchyma as well aswithin and surrounding fibrotic bands (fig 2A).In fibrotic liver 21 days after BDL there were14.4 (0.23) á-SMA positive cells per highpower field. The number of á-SMA positivecells then progressively decreased over the 42days of recovery (fig 2B, 2C). At all times dur-ing recovery, the distribution of á-SMA

positive cells was within and extending into theparenchyma from the fibrotic bands and theareas of fibrosis surrounding the expandedportal triads. The numbers of á-SMA positivecells in the sham operated control livers werenegligible (<0.5 per high power field).

HSC APOPTOSIS OCCURS DURING SPONTANEOUS

RECOVERY FROM BILIARY FIBROSIS

Liver sections were dual stained for TUNELand á-SMA. TUNEL positive condensednuclei in the distribution of the collagen fibrils,demonstrated by Sirius red and reticulin stain-ing, were demonstrated to be within theá-SMA positive cells (fig 3). Moreover, therewas no evidence of necrosis in á-SMA positivecells. These data strongly support the hypoth-esis that HSC apoptosis is occurring duringspontaneous recovery from fibrosis. The rate ofapoptosis in non-parenchymal cells duringrecovery from fibrosis was quantified by directcell counting of TUNEL positive nuclei. Thesewere counted only around collagen fibrils—that is, in the regions demonstrated to have thehighest density of á-SMA positive cells, asdescribed above. At 21 days after BDL, therewere 0.45 (0.02) TUNEL positive nuclei perhigh power field which increased after two daysof recovery before falling progressively to 0.08(0.01) TUNEL positive nuclei per high powerfield on day 42 (fig 4). There were negligiblenumbers of TUNEL positive cells in the shamoperated control livers.

ACTIVATED HSC UNDERGO APOPTOSIS FOLLOWING

SERUM DEPRIVATION

Apoptosis in cultured HSCs was quantified invitro by staining with acridine orange andcounting under fluorescence (as demonstratedin fig 5A). In addition, to confirm that thesecondensed cells had undergone apoptosis, theywere harvested by gentle washing, prepared asa cytospin, and Giemsa stained (fig 5B). DNAwas extracted from further preparations. The

Figure 3 Two examples of apoptotic hepatic stellate cells(HSCs) demonstrated by dual terminal UDP-nick endlabelling (TUNEL) and á smooth muscle actin (á-SMA)staining. Liver sections at two days after biliary-jejunalanastomosis were dual stained with á-SMA and TUNEL.Non-parenchymal TUNEL positive cell nuclei (red)demonstrating the characteristic condensation and halo areobserved within the á-SMA positive cells (blue), indicatingthat the apoptotic nuclei lie within activated HSCs (×40).

Figure 4 Quantification of non-parenchymal cellapoptosis during recovery from bile duct ligation (BDL).Terminal UDP-nick end labelling (TUNEL) positivenon-parenchymal cell apoptotic bodies in a distributionconsistent with activated hepatic stellate cells (HSCs) werequantified as described in the methods both before and afterbiliary-jejunal anastomosis. The results indicated that thefirst two days following biliary reanastomosis are associatedwith an increase in apoptosis, which coincides with theobserved diminution in activated HSCs.

0.75

0.50

0.25

0.0021 1 2 7 42 Sham

TU

NE

L p

osi

tive

cel

ls p

erh

igh

po

wer

fie

ld

Days afterbile ductligation

Days afterbile duct

anastomosis

552 Issa, Williams, Trim, et al

www.gutjnl.com

DNA thus obtained demonstrated the ladder-ing that results from oligonucleosomal frag-mentation and is considered to be characteris-tic of apoptosis (fig 5C). By determining therate of apoptosis at predetermined times afterserum withdrawal, as described above, wedemonstrated that deprivation of growth fac-tors was associated with a significant increasein HSC apoptosis at 6, 24, and 72 hours (fig 6).Cycloheximide has been reported to enhanceserum deprivation induced apoptosis in a vari-ety of cell types,12 16 therefore a further experi-ment was conducted in which 6.25–50 µMcycloheximide was added to the serum freemedia. Cycloheximide induced a concentra-tion dependent increase in apoptosis comparedwith serum withdrawal alone, which was maxi-mally eVective at 50 µM (data not shown).

IGF-1 BUT NOT PDGF OR IGF-2 RESCUES HSCs

FROM APOPTOSIS INDUCED BY SERUM

DEPRIVATION

The eVect of serum withdrawal suggested thatgrowth factors were important regulators ofHSC apoptosis in culture. To explore this con-cept further we undertook a series of experi-ments to examine the eVect of the growth fac-tors IGF-1, IGF-2, TGF-â1, and PDGF onHSC apoptosis, as these are expressed duringfibrotic injury. Incubation of HSC with IGF-1in serum free media in concentrations of0.1–10 ng/ml for 24 hours was associated witha modest but consistent dose dependentreduction in apoptosis compared with parallelcultures maintained serum deprived; similarresults were obtained after six hours ofexposure. This eVect was significant at 1 and10 ng/ml (fig 7A). In parallel studies, the eVectof IGF-1 (10 mg/ml) in reducing apoptosisrelative to serum free and serum free withcycloheximide (50 µM) control cultures wasdetermined by caspase 3 activity. Relative tocontrols, IGF-1 reduced caspase 3 activity by22% in serum free conditions, and by 80% inserum free conditions in the presence of 50 µMcycloheximide. Further studies using quantita-tive FACS analysis of IGF-1 treated HSCsrelative to serum free controls after stainingwith an anti-single stranded DNA antibody,annexin V binding assays, and propidiumiodide staining demonstrated a similar reduc-tion in apoptosis relative to controls in thepresence of IGF-1 (data not shown). Incuba-tion of HSCs with IGF-2 (0.125–12.5 ng/ml)or PDGF (0.1–100 ng/ml) resulted in nosignificant change in apoptosis (22.9 (3.6)% inserum free controls v 20.7 (3.1)% in cellsincubated with 12.5 ng/ml IGF-2, n=3; and

Figure 5 Acridine orange staining illustrating normal and apoptotic activated hepatic stellate cells (HSCs). (A)Representative fields from passaged HSC incubated for four days in media with 16% serum (left) and in serum free conditions(right). Normal HSC have large pale nuclei and orange tinged cytoplasm (the result of RNA staining). Apoptotic HSC aredistinct with brightly condensed chromatin with shrunken or absent cytoplasm; two of several examples are arrowed. Apoptoticcells lift oV from the monolayer, therefore to focus on these the monolayer falls below the plane of focus. Mitotic figures are alsoidentifiable with this method (left field). Giemsa stain of cytospin preparation of condensed HSC: after exposure to conditions ofabsolute serum deprivation as described in the methods, the condensed rounded up cells on the monolayer surface, correspondingto the cells counted after acridine orange staining, were harvested by gentle washing and prepared as a cytospin, before beingGiemsa stained. The cells demonstrated clear evidence of chromatin condensation and fragmentation, and blebbing; arepresentative example is given in (B). DNA analysis of the condensed and rounded up cells was undertaken as described inthe methods and demonstrates the characteristic laddering associated with apoptosis (C).

Figure 6 EVect of prolonged serum deprivation on hepaticstellate cell (HSC) apoptosis. Rat HSCs were washed threetimes and incubated for 6, 24, and 96 hours in media withno additives (0% fetal calf serum (FCS)) or returned tomedia with 16% FCS. After incubation, the number ofapoptotic cells was counted and expressed as a percentage ofthe total cell number per field (mean (SEM), n=7;*p<0.05, **p<0.01 for 0% FCS v 16% FCS at the sametime point by paired t test).

30

25

20

15

10

5

06 h 24 h 72 h

Ap

op

tosi

s (%

)

** **

0% FCS

16% FCS

Apoptosis of hepatic stellate cells 553

www.gutjnl.com

23.8 (3.5)% in serum free controls v 22.1(3.1)% in cells incubated with 100 ng/mlPDGF; n=4). Similar results were obtainedafter six hours of exposure. The addition ofIGF-1 to HSCs incubated in serum free condi-tions in the presence of cycloheximide also sig-nificantly protected HSCs from apoptosis after24 hours of incubation, although the concen-tration of IGF-1 required to be significantlyeVective in this context was 100 ng/ml (fig 7B).

The addition of TGF-â1 (10 ng/ml) to serumdeprived HSC cultures resulted in a 40%increase in apoptosis relative to serum deprivedcontrols (n=4) which was significant at concen-trations of both 1 and 10 ng/ml; this eVect wasmaximal at six hours of exposure (fig 7A).

IGF-1 INDUCES AN INCREASE IN HSC NUMBERS BY

PROMOTING SURVIVAL

To determine the eVect on HSC proliferationand number when cultures were incubatedwith specific growth factors, further experi-ments were undertaken in which HSC wereexposed to IGF-1 or PDGF for 24 hours andthe rate of proliferation and DNA concentra-tion (as an index of overall cell numbers) weredetermined. As previously anticipated, bothserum and PDGF were eVective mitogens (fig8A). In contrast, IGF-1 was not mitogenic.Indeed, there was no significant diVerence inthe rate of proliferation observed with PDGFcompared with serum, whereas proliferationwas significantly reduced after incubation withIGF-1 compared with serum (fig 8A). Whentotal DNA concentration was analysed in par-allel wells to act as an indicator of changes incell numbers, in comparison with culturesmaintained in serum free conditions alone,FCS induced the greatest increase in cell num-bers (fig 8B). IGF-1 induced a similar increasein cell number. However, the increase in cellnumber observed with PDGF was minor incomparison with serum or IGF-1. These datasuggest that inhibition of apoptosis is a majorcontributor to the increase in HSCs in culturemodels. Incubation of HSCs with TGF-â1

resulted in a reduction in proliferation (fig 8C)and a 30% decrease in total DNA contentreflecting a net decrease in HSC numbers rela-tive to serum deprived controls (fig 8D).

RIBONUCLEASE PROTECTION ASSAY DETECTION

OF Fas/FasL

There was no diVerence in FasL mRNAexpression relative to GAPDH in culturesmaintained for 24 hours serum free, or treatedwith IGF-1 (10 ng/ml) or PDGF (10 ng/ml). Inthe same assays, Fas mRNA was not detected(n=2; data not shown).

DiscussionThe data presented in this paper demonstrateevidence of the critical role of HSC apoptosisin the spontaneous recovery from liver fibrosis.By showing that HSC numbers decreaseduring recovery and by providing clear evi-dence of HSC apoptosis in a model which ismechanistically distinct from that described inour previous report,5 we conclude that HSCapoptosis is a vital mechanism that contributesto recovery from hepatic fibrosis regardless ofaetiology. Moreover, as we have previously dis-cussed, there are clear paradigms of theseobservations in the form of case reports ofhuman fibrotic disease in vivo.17 18

á Smooth muscle actin, an intermediate fila-ment protein that is expressed by activatedHSC and is widely accepted to be a marker ofactivation, was used to identify and quantifyactivated HSCs. By staining the BDL liver

Figure 7 (A) EVect of specific growth factors on hepaticstellate cell (HSC) apoptosis. EVect of insulin-like growthfactor 1 (IGF-1) and transforming growth factor â1

(TGF-â1) on serum deprivation induced apoptosis of ratHSCs. For each concentration of growth factor (x axis,ng/ml) the percentage of cells demonstrating an apoptoticmorphology after staining with acridine orange is expressedas mean (SEM). For the IGF-1 experiments, counting wasundertaken after 24 hours (n=5, *p<0.05 v no additivesby paired t test). For the TGF-â1 experiments, counting wasundertaken after six hours (***p<0.001 by paired t test,n=4). †S, serum containing media. (B) EVects of IGF-1on cycloheximide induced apoptosis of rat HSCs. Washedrat HSC cultures were incubated for six hours in mediawith no additives, or media supplemented with 50 µMcycloheximide (CHX) with or without IGF-1. Thepercentage of cells undergoing apoptosis detected by acridineorange staining is given as mean (SEM) (n=4). *p<0.05for IGF-1 treated cultures v cycloheximide alone by paired ttest. SF, serum free.

30

20

10

0SF 0.1 1 10

Ap

op

tosi

s (%

) * *

75

50

25

0+S SF 1 10

Ap

op

tosi

s (%

)

***

40

30

20

10

0SF 0 10 100

Ap

op

tosi

s (%

)

*

*

[IGF-1] ng/ml +50 µM CHX

IGF-1

TGF-β1

B

A

554 Issa, Williams, Trim, et al

www.gutjnl.com

before and after reanastomosis we demon-strated that the high numbers of activatedHSCs present at peak fibrosis rapidly dimin-ished while remodelling of the fibrosis occurredand a more normal architecture was restored.Indeed, during the first week following reanas-tomosis there was a fivefold diminution in acti-vated HSC numbers. The diminution of HSCswas not associated with an increase in inflam-matory activity and specifically there was noevidence of necrosis, and therefore we tried toestablish if there was evidence to indicate thatthe HSC loss was mediated by apoptosis.

In our initial studies, dual staining ofrepresentative samples of BDL liver which hadundergone reanastomosis with TUNEL andá-SMA consistently demonstrated morpho-logically apoptotic, TUNEL positive nucleiwithin á-SMA positive cells, providing strongevidence for the presence of HSC apoptosis.We therefore established a series of experi-ments to quantify HSC apoptosis followingBDL induced fibrosis and biliary reanastomo-sis. As before,5 we undertook TUNEL stainingand á-SMA staining independently on sequen-tial liver sections as this method allows optimalconditions to be used for each stain. We thenapplied strict morphological criteria to thequantification of TUNEL positive apoptoticbodies. Only those apoptotic bodies whichwere TUNEL positive and lay within the areasof fibrosis in the expanded portal triad werecounted. Reference to sections stained in

parallel for á-SMA indicated that the cells inthese areas were largely activated HSCs. Theresults indicate that there is a background rateof apoptosis in fibrotic livers 21 days afterBDL; these represent between 1% and 3% ofthe á-SMA positive HSC population. Follow-ing reanastomosis there is a clear increase inthe rate of apoptosis over 48 hours. Althoughthe absolute number of apoptotic figuresremains small and never represents at maxi-mum more than 8% of the total HSC popula-tion, an increase in the rate of apoptosis,particularly in association with either a con-stant rate of HSC proliferation or a reductionin HSC proliferation, was associated with adramatic overall reduction in HSC numbers inthe liver. As the total time for apoptosis toeliminate a single cell may be as little as 20minutes,12 these results also serve to emphasisethat the HSC pool is dynamic and overallnumbers reflect changes in the rate of apopto-sis in addition to representing changes in therate of proliferation, a feature emphasised inour cell culture studies. Interestingly, the chro-nology of HSC apoptosis diVers from thatobserved in biliary epithelial cells during BDLrecovery, suggesting cell specificity in responseto individual survival signals.9

We had previously proposed that apoptosismight be a default pathway for activated HSCsunless specifically signalled to survive by physi-cal (matrix) or soluble (cytokine and growthfactor) agents. To investigate this hypothesis

Figure 8 (A) EVect of growth factors on 3H-thymidine incorporation. After washing, rat hepatic stellate cells (HSCs)were cultured in DMEM with no additives (serum free (SF)), 16% fetal calf serum (FCS) (+S, positive control), orspecified growth factors for 24 hours, and the rate of 3H-thymidine incorporated was determined as described. Results areexpressed relative to control cultures (no additives, SF) which were given the arbitrary value of 100% (n=4, *p<0.05relative to 16% FCS containing cultures by paired t test). (B) EVect of growth factors on total DNA of cultures of ratHSCs. Washed cultures of rat HSCs were incubated for 24 hours in DMEM with no additives (SF) or supplemented with16% FCS (+S, positive control) or growth factors as shown and the total DNA concentration of each cell suspension wasdetermined as described in the methods. Results are expressed relative to control cultures (no additives, SF) which weregiven the arbitrary value of 100% (n=4, *p<0.05 relative to 16% FCS containing cultures by paired t test). (C, D) EVectof transforming growth factor â1 (TGF-â1) on HSC proliferation and total DNA content, respectively, were analysed asdescribed and are represented relative to control cultures (no additives, SF) which have been given the arbitrary value of100%.

700

600

500

400

300

200

100

0SF +S PDGF IGF-1

cpm

(%

)

**

A 175

150

125

100

75

50

25

0SF +S PDGF IGF-1

dsD

NA

(%

)

* *

B

500

400

300

200

100

0SF +S TGFβ

10 ng/ml

cpm

(%

)

C 150

125

100

75

50

25

0SF +S TGFβ

10 ng/ml

D

dsD

NA

(%

)

Apoptosis of hepatic stellate cells 555

www.gutjnl.com

and to complement our studies of the in vivorecovery models we have undertaken a detailedcharacterisation of the response of HSCs toserum deprivation in the presence and absenceof specific growth factors.

Apoptosis was quantified in vitro by directcell counting under fluorescence illuminationafter acridine orange staining. We and othershave consistently demonstrated that in fibro-blast, HSC, and myofibroblast cultures, therounded up condensed cells on the monolayersurface demonstrate features of apoptosis mor-phologically and demonstrate the specificnuclear changes of apoptosis.5 12 19 To furthervalidate this technique we have harvested thecondensed cells from the monolayer surfaceand showed that the DNA demonstrates theladdered pattern that results from oligonucleo-somal fragmentation which is characteristic ofapoptosis.19 20 In further experiments we usedcaspase activity assays and quantitative FACSanalysis after annexin V binding and propid-ium iodide staining to confirm the increase inapoptosis seen in serum deprivation.

Acridine orange staining demonstrated thatabsolute serum deprivation increased the rateof apoptosis above that observed in HSC inserum containing media, supporting the hy-pothesis that specific growth factors maypromote HSC survival. However, even in con-ditions of absolute serum deprivation for up tothree days, the absolute level of apoptosis neverrose above 25% for HSCs 72 hours postpassage or 50% for cells 24 hours post passage.This suggests that other influences, stage of thecell cycle, cell-matrix and cell-cell interactions,autocrine factors, or combinations of thesemay also influence HSC survival in this model.Moreover, mitotic figures were observed in theserum deprived cultures indicating that thistreatment did not halt cell proliferation inhighly activated HSCs. Therefore, a possibleexplanation for the relatively constant rate ofapoptosis after serum deprivation is that cellshave to be entering the growth cycle to be sus-ceptible to apoptosis. Incubation with cy-cloheximide has been reported to increaseapoptosis in a variety of cell types12 16 and wasassociated with a concentration dependentincrease in HSC apoptosis which reached100% after 24 hours in 50 µM cycloheximide.

Both IGF-1 and PDGF have been reportedto promote HSC growth; they and theirrespective receptors are expressed by HSC.During liver injury they may act in a paracrine(from hepatocytes and KupVer/infiltrating in-flammatory cells, respectively) and an auto-crine manner.21–23 We therefore incubatedHSCs in the presence of increasing concentra-tions of these growth factors in the absence ofserum. Our results demonstrate that IGF-1 butnot PDGF promotes HSC survival. Incubationof HSC with IGF-1 was associated with ahighly reproducible although modest reductionin apoptosis in concentrations of 10 and1 ng/ml. In contrast, PDGF had no consistenteVect on apoptosis. IGF-1 also proved to eVec-tively inhibit apoptosis in the presence ofcycloheximide although a higher concentration(100 ng/ml) was required to achieve this eVect

than with cultures in serum free media alone.Taken together these data suggest that IGF-1 isa candidate mediator promoting HSC survivalduring injury, an eVect which has beendocumented in a series of cell systems.24–27 Italso provides a mechanism in wounding inwhich the epithelium (hepatocyte) may directlyregulate the survival of the wound healingmyofibroblast (HSC) in a paracrine man-ner.21 28 While maximum inhibition of apopto-sis of 20% induced by IGF-1 appears small, ithas been suggested that IGF-1 mediatedinhibition of apoptosis is somewhat underesti-mated by in vitro work because more profoundeVects have been observed in experimentalmanipulations in vivo.26 27 In contrast withIGF-1, TGF-â1 actually increased apoptosis.As TGF-â1 is a powerful profibrogenic stimu-lus to HSCs, this observation suggests that thesame cytokine may simultaneously limit thesurvival of matrix sythesising HSCs.

Expression of Fas and FasL has recentlybeen demonstrated in HSCs.6 29 FasL mRNAwas not regulated by either IGF-1 or PDGFexposure when assayed by ribonuclease protec-tion, and Fas mRNA was not detectable usingthis method. This may suggest that the anti-apoptotic eVect of IGF-1 was not due to adecrease in FasL (a proapoptotic ligand) but bysome other, as yet, unidentified mechanism. Inour model, as in the previous FasL mediatedmodel reporter by Gong and colleagues,29

cycloheximide appears to promote HSC apop-tosis. One possible explanation is that inhibi-tion of a critical autocrine or intracellular sur-vival factor is necessary to prime highlyactivated HSCs for apoptosis through growthfactor withdrawal or TNF receptor superfamily(Fas and related receptors) stimulation.

As the overall cell number in a tissue reflectsthe net eVect of proliferation in addition toapoptosis/death, we determined the impact ofour manipulations of culture conditions onproliferation and total HSC number. In thesestudies experiments were performed in anidentical manner to those in which apoptosiswas quantified and without cell cycle coordina-tion. The 4.8-fold increase in 3H-thymidineseen with PDGF is consistent with the welldescribed mitogenic eVects of this cytokine28

and contrast markedly with the lack of eVect onapoptosis. Exactly the converse was observedwith IGF-1; a marginal eVect on proliferationwas observed (1.2-fold increase) in contrastwith the significant eVect on apoptosis. Viewedwith the data on total DNA, where incubationwith IGF-1 was associated with a 38%increase, these results emphasise theimportance of survival as a mechanism to pro-mote an overall increase in HSC numbers.

Compared with cultures returned to serumcontaining media, all serum free experimentalconditions showed a reduction in cell number,as determined by DNA concentration. Cul-tures supplemented with PDGF contained asimilar number of cells suggesting that cell lossmust have occurred simultaneously and rein-forcing the observation that the number withinthe HSC pool is a dynamic function of

556 Issa, Williams, Trim, et al

www.gutjnl.com

proliferation and apoptosis. In fact there is evi-dence that PDGF is proapoptotic for fibro-blasts in conditions of low serum.30 In physi-ological terms it is logical to propose that sucha potent inducer of proliferation should notconcurrently inhibit apoptosis—this combina-tion of events is a classical feature of tumori-genesis. Rather, that HSC proliferation andapoptosis are independently controlled andthat the net cell number reflects the overall bal-ance of these cell fates.

We have described HSC apoptosis in thespontaneous recovery from BDL inducedhepatic fibrosis. These data together with ourprevious report provide strong evidence for thecritical role of apoptosis of HSCs in spontane-ous recovery from liver fibrosis. We have alsodemonstrated that IGF-1 is a potent inhibitorof HSC apoptosis in vitro and may function indisease to regulate the increase in the pool ofactivated HSC. In contrast, PDGF promotesproliferation but is neutral with respect to HSCapoptosis while TGF-â1 was observed toincrease apoptosis. These data are consistentwith a model in which HSC apoptosis resultsfrom changes in the balance of pro and anti-apoptotic influences which impinge on the celland are regulated independently of prolifera-tive influences.

JPI is an MRC UK senior clinical fellow. The support of theWessex Medical Trust and Bayer AG are gratefully acknowl-edged. This study was also supported by a grant from the SwissNational Foundation for Scientific Research (45349.95) to JR.

Conflict of interest. This work was supported by Bayer AG.

1 Friedman SL. The cellular basis of hepatic fibrosis: Mecha-nisms and treatment strategies. N Engl J Med1993;328:1828–35.

2 Alcolado R, Arthur MJP, Iredale JP. Pathogenesis of liverfibrosis. Clin Sci 1997;92:103–12.

3 Friedman SL, Roll FJ, Boyles J, et al. Hepatic lipocytes: Theprincipal collagen-producing cells of normal rat liver. ProcNatl Acad Sci USA 1985;82:8681–5.

4 Maher JJ, McGuire RF. Extracellular matrix gene expres-sion increases preferentially in rat lipocytes and sinusoidalendothelial cells during hepatic fibrosis in vivo. J Clin Invest1990;86:1641–8.

5 Iredale JP, Benyon RC, Pickering J, et al. Mechanisms ofspontaneous resolution of rat liver fibrosis: hepatic stellatecell apoptosis and reduced hepatic expression of metallo-proteinase inhibitors. J Clin Invest 1998;102:538–49.

6 Saile B, Knittel T, Matthes N, et al. CD95/CD95L-mediated apoptosis of the hepatic stellate cell. Am J Pathol1997;151:1265–72.

7 Gressner AM. The cell biology of liver fibrogenesis—animbalance of proliferation, growth arrest and apoptosis ofmyofibroblasts. Cell Tissue Res 1998;292:447–52.

8 Abdel-Aziz G, Lebeau G, Rescan PY, et al. Reversibility ofhepatic fibrosis in experimentally induced cholestasis in rat.Am J Pathol 1990;137:1333–42.

9 Stahelin BJ, Marti U, Zimmermann H, et al. The interactionof Bcl-2 and Bax regulates apoptosis in biliary epithelialcells of rats with obstructive jaundice. Virchows Arch 1999;434:333–9.

10 Stahelin BJ, Marti U, Solioz M, et al. False positive stainingin the TUNEL assay to detect apoptosis in liver and intes-tine is caused by endogenous nucleases and inhibited bydiethyl pyrocarbonate. Mol Pathol 1998 51:204–8.

11 Arthur MJP, Friedman SL, Roll FJ, et al. Lipocytes fromnormal rat liver release a neutral metalloproteinase thatdegrades basement membrane (type IV) collagen. J ClinInvest 1989;84:1076–85.

12 Baker AJ, Mooney A, Hughes J, et al. Mesangial cellapoptosis: the major mechanism for resolution of glomeru-lar hypercellularity in experimental mesangial proliferativenephritis. J Clin Invest 1994;94:2105–16.

13 Vyas SK, Leyland H, Gentry J, et al. Transin (stromelysin) isexpressed in early rat lipocyte primary culture. Gastroenter-ology 1995;109:889–98.

14 Iredale JP, Benyon RC, Arthur MJ, et al. Tissue inhibitor ofmetalloproteinase-1 messenger RNA expression is en-hanced relative to interstitial collagenase messenger RNAin experimental liver injury and fibrosis. Hepatology1996;24:176–84.

15 Rockey DC, Boyles JK, Gabbiani G, et al. Rat hepaticlipocytes express smooth muscle actin upon activation invivo and in culture. J Submicrosc Cytol Pathol 1992;24:193–203.

16 Weil M, Jacobson MD, Coles HS, et al. Constitutive expres-sion of the machinery for programmed cell death. J CellBiol 1996;133:1053–9.

17 Rojkind M, Dunn MA. Hepatic fibrosis. Gastroenterology1979;76:849–63.

18 Hunt J. Long-term follow-up of patients with hepatitis Btreated with interferon. Interferons Cytokines 1992;20:6–9.

19 Fanidi A, Harrington EA, Evan GI. Cooperative interactionbetween c-myc and bcl-2 proto-oncogenes. Nature 1992;359:554–6.

20 Evan GI, Wyllie AH, Gilbert CS, et al. Induction of apopto-sis in fibroblasts by c-myc protein. Cell 1992;69:119–28.

21 Brenzel A, Gressner AM. Characterization of insulin-likegrowth factor (IGF)-I-receptor binding sites during in vitrotransformation of rat hepatic stellate cells to myofibrob-lasts. Eur J Clin Chem Clin Biochem 1996;34:401–9.

22 Pinzani M, Abboud HE, Aron DC. Secretion of insulin-likegrowth factor-I and binding proteins by rat liver fat-storingcells: regulatory role of platelet-derived growth factor.Endocrinology 1990;127:2343–9.

23 Valentinis B, Reiss K, Baserga R. Insulin-like growth factor-I-mediated survival from anoikis: role of cell aggregationand focal adhesion kinase. J Cell Physiol 1998;176:648–57.

24 Romano G, Prisco M, Zanocco-Marani T, et al. Dissocia-tion between resistance to apoptosis and the transformedphenotype in IGF-I receptor signaling J Cell Biochem 1999;72:294–310.

25 ResnicoV M, Burgaud JL, Rotman HL, et al. Correlationbetween apoptosis, tumorigenesis, and levels of insulin-likegrowth factor I receptors. Cancer Res 1995;55:3739–41.

26 ResnicoV M, Abraham D, Yutanawiboonchai W, et al. Theinsulin-like growth factor I receptor protects tumor cellsfrom apoptosis in vivo. Cancer Res 1995;55:2463–9.

27 Scharf JG, Schmidt-Sandte W, Pahernik SA, et al.Characterization of the insulin-like growth factor axis in ahuman hepatoma cell line (PLC). Carcinogenesis 1998;19:2121–8.

28 Pinzani M, Gesukado L, Sabbah GM, et al. EVects of plate-let derived growth factor and other polypeptide mitogenson DNA synthesis and growth of cultured rat liver fat stor-ing cells. J Clin Invest 1989;84:1786–93.

29 Gong WR, Pecci A, Roth S, et al. Transformation-dependent susceptibility of rat hepatic stellate cells toapoptosis induced by soluble Fas ligand. Hepatology 1998;28:492–502.

30 Kim H-RC, Upadhyay S, Li G, et al.Platelet-derived growthfactor induces apoptosis in growth-arrested murine fibro-blasts. Proc Natl Acad Sci USA 1995;92:9500–4.

Apoptosis of hepatic stellate cells 557

www.gutjnl.com