The development of a method for the preparation of rat ...Tissue culture flasks (T25 and T75) were purchased from Falcon, and 24 multi-well plates from Costar, Corning, Falcon and

Journal of Cell Science 101, 219-231 (1992)Printed in Great Britain © The Company of Biologists Limited 1992

219

The development of a method for the preparation of rat intestinal epithelial

cell primary cultures

G. S. EVANS1*, N. FLINT1, A. S. SOMERS1, B. EYDEN2 and C. S. POTTEN1

^Cancer Research Campaign Department of Epithelial Biology, Paterson Institute for Cancer Research, Christie Hospital and Holt RadiumInstitute, Christie Hospital, Withington, Manchester, M20 9BX, UK2Department of Pathology, Christie Hospital and Holt Radium Institute, Withington, Manchester M20 9BX, UK

•Author for correspondence

Summary

We describe a reproducible method for growing smallintestinal epithelium (derived from the suckling ratintestine) in short-term (primary) cultures. Optimalculture conditions were determined by quantitativeassays of proliferation (i.e. changes in cellularity andDNA synthesis). Isolation of the epithelia and, signifi-cantly, preservation of its three-dimensional integritywas achieved using a collagenase/dispase digestiontechnique. Purification of the epithelium was alsofacilitated by the use of a simple differential sedimen-tation method.

The results presented below support the idea thatproliferation of normal gut epithelium ex vivo is initiallydependent upon the maintenance of the structural

integrity of this tissue and upon factors produced byheterologous mescnchymal cells. Proliferation in vitrowas also critically dependent upon the quality of themedium and constituents used.

Cultures reached confluence within 10-14 days andconsisted of epithelial colonies together with varyingamounts of smooth-muscle-like cells. Cultures have beenmaintained for periods up to one month, but the longer-term potential for growth by sub-culturing has not beenexamined. Strategies for reducing the proliferation ofthese non-epithelial cells are also described.

Key words: intestine, epithelial, in vitro, smooth musclecells.

Introduction

The maintenance of proliferation of isolated intestinalepithelial cells ex vivo has proven to be a majorlimitation in the advancement of the understanding ofgrowth factor function in this tissue. There are manyepithelial cell lines derived from colonic adenocarci-noma (for recent review see Moyer et al., 1990) andthere have been occasional successes in establishingpermanent epithelial lines from the intestine of sucklinganimals (Quaroni and May, 1980; Blay and Brown,1984; Ne"grel et al., 1983). In these instances, theepithelial cells were selected (using cloning cylinders)from cultures containing a mixture of cell types thatformed over a period of weeks following the initialdigestion of the intestine in collagenase. Normal ortransformed cell lines cannot be considered adequatemodels of the intact epithelium, since these cells haveundergone changes and selection to facilitate long-termgrowth in vitro. This does not preclude their use forexperimental studies, but they clearly represent adifferent type of model to that of primary cultures.Consequently there is a need to develop methods for

propagating freshly isolated epithelium on a short-termbasis, allowing reproducible quantitative studies. Ourultimate aim is to identify growth factors and extracellu-lar matrix components that are required to maintain cellproliferation, differentiation and ultimately the func-tion of stem cells in the gut epithelium.

The suckling rat intestine was chosen for theseexperiments because this is the stage of developmentwhere there has been success in establishing cell lines(Quaroni and May, 1980). The in vitro requirements ofthis primary culture system that have been examinedbelow include the procedures for isolating the cells,quality of the culture medium, optimal serum con-ditions, CO2 concentration and pH, and the use ofbiological substrata. Where possible we have tried totake a quantitative, rather than anecdotal, approach tostudying the culture conditions for these cells. Theisolated cells were plated into 24 multi-well dishes andchanges in cell proliferation were determined by assaysof ceOularity (crystal violet staining, which preferen-tially binds to the nuclei of fixed cells) and DNAsynthesis (tritiated thymidine incorporation). Usingthese methods reproducible culture of intestinal epi-

220 G. S. Evans and others

thelium ex vivo has been achieved and procedures toestablish primary monoculture of this tissue have beenidentified.

Materials and methods

AnimalsMale and female six-day-old Wistar rats were used and thesewere housed under a twelve-hour light/dark cycle (light: 0700to 1900 hours, GMT), and given food and water ad libitum.The animals were killed by cervical dislocation and thecomplete length of the small intestine was removed (andcleaned free of mesentery).

Culture reagentsThe culture media tested were Minimum Essential Medium(single strength and 10x concentrate, with Earle's salts, andL-valine; Gibco BRL), high-glucose formulation Dulbecco'sModified Eagle's Medium (single and 10 x concentrates ofDMEM; Gibco BRL). Foetal calf serum (Biological Indus-tries) was chosen from batch tests for growth stimulation oftwo rat suckling intestinal epithelial cell lines (IEC 6 and IEC18; Quaroni and May, 1980).

Additives to some of the media included sodium pyruvate(Sigma), glutamine (Gibco), non-essential amino acids(Gibco), insulin (Sigma and Collaborative Research) andepidermal growth factor (EGF; Sigma); penicillin (Crystapen,Glaxo at 100 U/ml) and streptomycin (Evans Medical Ltd, at60 jig/ml). Porcine mucosal heparin (Sigma H8514) was addedat concentrations of 1-200 ^g/ml. All reagents were culturegrade.

Tissue culture flasks (T25 and T75) were purchased fromFalcon, and 24 multi-well plates from Costar, Corning, Falconand Nunc; gas-permeable dishes were obtained from Herar-eus. All of the cultures were incubated at 37°C in a humidifiedatmosphere in either a 5%, 7.5 or 10% CO2 incubator(Herareus).

Culture substratumVarious reagents were used to coat culture dishes to improvethe attachment of the intestinal cells in culture. They wereeither dissolved in DMEM medium or sterile water at thestated concentrations, left overnight in the incubators, themedium was removed and the plates allowed to air dry.Before use, the dishes were washed with a change of HBSS.The attachment factors examined were; 1-100 /ig/ml fibronec-tin (purified rat, Boehringer), laminin (Boehringer), 10-50/ig/ml collagen IV (Collaborative Research), 200 /ig/ml bovinedermal collagens (Vitrogen, Collagen Corporation), 200/ig/ml rat tail tendon collagens I + III (McAteer andCavanagh, 1983) In addition, dishes were prepared withMatrigel (Collaborative Research), bovine corneal endo-thelial cell matrix (Gospadarowicz et al. 1980) prepared by amodified procedure (Yvonne Barlow, personal communi-cation) using 0.5% Triton X-100 (Sigma) and 0.0125 Mammonium hydroxide (BDH), and also mouse Swiss 3T3fibroblast matrix (Hedman et al., 1979). Collagen gels wereprepared by the method described by McAteer and Cavanagh(1983).

Cell isolationVarious techniques were employed for isolating the epithelialcells, these including a modified Weiser (Weiser, 1973)procedure at 4°C and 20°C (Flint et al. 1991), 1-5 mM EDTA

Step 1

Remove the small intestine, slit open and cut into 2-3 mmlengths. Transfer these to a T25 ml flask and wash at least8 times in 50 ml changes of HBSS with vigorous shaking*

Step 2

Transfer to a Petri dish and use a sharp scalpel blade todice the tissue into <1 mm3 pieces. Return to a T25 mlflask with 20 ml of enzyme solution (ES). Shakevigorously* for 25 minutes at 25°C

Step 3

Pipette solution vigorously (with >2 mm bore pipette)approximately 150 times and transfer contents to a 25 mlsterile universal tube (see Fig. 2A)

Step 4

Leave the contents to sediment under gravity for 60seconds, and carefully removal all but the bottom few ml;repeat this procedure twice

• » •

Step 5

Add 10 ml of DMEM-S to supernatant, mix and spin at200-300 revs/minute for 2 minutes; remove the supernatantcarefully and resuspend the pellet in 20 ml of DMEM-S

-•-

Step 6

Repeat this procedure at least 5-6 times until thesupernatant is completely clear and the pellet is welldefined (Fig. 2D); finally, resuspend in the appropriategrowth medium

Fig. 1. Flow diagram of the collagenase/dispase methodused to isolate the neonate rat epithelium; ES, 0.1 mg/mldispase and 300 U/ml collagenase in HBSS, pH 7.4.;DMEM-S, DMEM + 2.5% FCS and 2% sorbitol.•Shaking procedures carried out on an orbitol shakingplatform (80 cycles/min).

(disodium salt, BDH), 0.01-0.1% trypsin (WorthingtonBiochemicals) at 4°C and 20°C, crude collagenase (Sigma;Clostridium histolyticum type XI), dispase (neutral proteasetype I, Boehringer) and a combination of the crudecollagenase and dispase (Fig. 1); for details of these methodssee Table 1. Apart from the Weiser solution, all of these otheragents were dissolved in Hanks' buffered salt solution(HBSS", low calcium formulation; Northumbria BiologicalsLtd).

Cell proliferation and growth assaysOwing to the difficulty of estimating cell yield when theepithelium was isolated in the form of organoids (villi andcrypts), it was necessary to standardise the input of cells intoeach well. On average, the intestine from a single animal wasused for each 24-well plate. To assess the growth-promotingeffects of different media, additives etc, the number of cellswas estimated by the crystal violet staining method (Bra-saemle and Attie, 1988). The important advantage of thisassay is that it is non-destructive, and after solubilization of

Primary cultures of intestinal epithelium 221

Table 1. Methods of isolating intestinal epithelium for culture

Isolation method Temp. (°C) Time Results

1-10 niM EDTA in HBSS1-10 mM EDTA in HBSS

1-10 mM EGTA in HBSS

1-10 mM EGTA in HBSS

Weiser solution (modified)*

Weiser solution (modified)*

0.125% Trypsint in HBSS

4

20

4

20

4

20

4

45 minutes15 minutes

45 minutes

15 minutes

45 minutes

15 minutes

18 hours

0.125% Trypsin in HBSS 20 1 hour

300-1500 U collagenasel in 37 2 hours

HBSS

1 mg/ml dispase§ in HBSS 20 2 hours

0.1 mg/ml dispase and 300 U/ml 20 30 minutescollagenase in HBSS

•Hint et al. (1991).tBovine pancreatic (Worthington Biochemicals).tCollagenase type 1 (Sigma).§Dispase (neutral protease) type 1 (Boehringer).

Pure epithelial cells, mostly single cells and clumps, no cells attached

Pure epithelial cells, mostly single cells, no cells attached

Pure epithelial cells, mostly single cells and clumps, no cells attached

Pure epithelial cells, mostly single cells, no cells attached

Pure epithelial cells, mostly clumps, some cells attached but no growth

Pure epithelial cells, mostly single cells and clumps, no cells attached

Gumps of epithelial cells with some mesenchymal cells (low yield),some cells attached but no growth

Single epithelial cells, with some mesenchymal cells (low yield), a fewcells attached but no growth

Clumps of epithelial and mesenchymal cells (high yield), very goodattachment and growth

Single epithelial cells (low yield), these attached but did not grow

Intact organoids of epithelium with attached mesenchyme and singlemesenchymal cells (high yield), very good attachment and growth

the dye, the cell cultures can be restained with differenthistological/histochemical dyes so that the different cellpopulations of cells can be observed.

For this assay, the cells were seeded into 24 multi-welldishes, with triplicate wells for each variable, after appropri-ate periods of time, the dishes were fixed in methanol thenthey were stained in 0.1% crystal violet. The wells werewashed thoroughly in distilled water and the bound dye (inthe cell nuclei) solubilized and transferred to a 96 micro-wellplate and read at 540 nm absorbency on an automatedTitreTek Multiscan (MCQ340) plate reader. Using thisprocedure there is a linear relationship between cell numberand dye concentration between 5 x 103 to 105 cells/well.Beyond this number there is an increase in dye concentrationup to 2 x 105 but this is progressively non-linear. Theminimum number of cells that can be distinguished by thisassay is 5 x 103 cells/well. Following the extraction of the

crystal violet the cells were re-stained for morphologicalexamination.

The proliferation of cells was also assessed by continuous orpulse labelling with tritiated thymidine incorporation (Fresh-ney, 1978). A dose of 0.025 jCi (0.925 kBq)/ml of [methyl-3H]thymidine (6.7 mCi/mmol specific activity; 24.9 GBq/mmol, NEM) was added to each well at the start of the cultureand replenished every three days. The samples were countedon a Beckman LS 1801 scintillation counter.

ImmunohistochemistryTo characterise the different cell types growing in thesecultures a number of antibodies were used (see Table 2) forimmunohistochemical staining. The primary culture cells (orcontrol cell lines) were plated onto sterile (acid-cleaned)collagen-coated glass coverslips (1 cm diameter, BDH), and

Table 2. Characterisation of cultured cells with antibodies

Antibody Specifications and source Neonate intestinal cells stained in vivo

Anti-keratin1

Anti-cytokeratin 82

Anti-cytokeratin 182

Anti-cytokeratin 192

Anti-vimentin3

Anti-desmin4

Anti-smooth muscle a actin5

Anti-rat aminopeptidase6

Anti-rat Thy-rAnti-rat endothelium7

RabbitMouseMouseMouseMouseMouseMouseMouseMouseMouse

polyclonal; (Dako) L/100monoclonal LE-61 1/100monoclonal LE-41 1/50monoclonals LP 2K and BA-16 1/50monoclonal V9 (I.C.N.) 1/100monoclonal D-ER-11 (Dako) 1/50monoclonal 1A-4 (Sigma) 1/250monoclonal M233 1/50monoclonal M226 1/50monoclonal OX-43 (Serotec) 1/50

Epithelial and mesothelial cellsSimple epithelium and mesothelial cellsSimple epithelium and mesothelial cellsSimple epithelium and mesothelial cellsFibroblast, smooth muscle cells and blood vesselsMuscle cells and myo-fibroblasts2

Smooth muscle cells and myoepitheliumApical surface of villus and pre-crypt epitheliumFibroblasts, smooth muscle cells and blood vesselsRat endothelium

'Franke et al. (1978).2Donated by Dr B. Lane (I.C.R.F.).3Osborn et al. (1984).4Debus et al. (1983).'Skalli et al. (1989).6Donated by Dr A. Ager (Department of Biochemistry, University of Manchester). This antibody cross-reacts with a putative rat Thy-1

antigen.7Barclay (1981).

222 G. S. Evans and others

grown in culture for various periods of time. The coverslipswere then removed, washed twice in serum free medium, andfixed in 50:50 (v/v) methanol: acetone (at 4°C) for 10 min. Thecoverslips were then permanently mounted on to glass slidesusing Glass Bond (Loctite), and stored in a -20°C freezerunder dry conditions until required.

For staining, the cultured cells were washed twice in bufferthen incubated with a 10% goat non-immune serum inphosphate-buffered saline (PBS) for 20 min followed by theprimary antibody (for dilutions see Table 2) in PBS and 0.5%BSA (Sigma) for 2 hours (at 20°C). The sections were thenwashed three times in PBS and fixed in cold acetone (4°C) for5 minutes. After air drying, two further washes in PBS werefollowed by a 1 hour (at 20°C) incubation in goat anti-mouseperoxidase diluted at 1/75 in PBS plus 0.5% BSA. Followingfive brief washes in PBS, the sections were incubated in thechromogenic substrate (0.2 mg/ml diamino-benzidine HC1,Sigma; 0.8 mg/ml nickel chloride, Sigma; and 0.003% H2O2BDH) in PBS for 5 minutes at 20°C. The slides werethoroughly washed in buffer and several changes of distilledwater, dehydrated, cleared in xylene, and mounted underlarge glass coverslips with XAM (BDH).

Controls included the omission of the primary antibody,and both the primary and secondary antibodies, sections ofintact 6-day gut and the inclusion of several indicator celllines; IEC 6,17 and 18; human breast carcinoma MCF-7 line;human smooth muscle cell line (from Dr A. Schor; Depart-ment of Oncology, Paterson Institute, Manchester), bovinecorneal endothelial cells, and mouse Swiss 3T3 fibroblasts. Insome cases verification of the antibody binding was also testedby Western blotting.

Histology and electron microscopySome material isolated during the enzymic digests was fixed inCarnoy's fluid, embedded in agar, and then dehydrated andembedded in Paraplast wax. Sections were cut at 4 fan andstained in haematoxylin and eosin. Specimens for electronmicroscopy were fixed in 2.5% glutaraldehyde in 0.15 MSorensons buffer, pH 7.4, followed by 0.1% osmiumtetroxide, and then dehydrated and embedded in Epon-Araldite. The sections were stained with lead citrate anduranyl acetate.

Results

Isolation of the intestinal epitheliumThe best isolation procedure (Table 1) was based upona mixture of crude collagenase (type XI) and dispase(type 1). At low concentrations (300 U ml collagenaseand 0.1 mg/ml dispase) the tissues could be rapidlydisaggregated at 20°C. To ensure a maximal recovery ofmaterial, the tissues were placed on a shaking platformfor 30 minutes during the enzymic digestion stage,followed by vigorous pipetting for 3 minutes using awide-mouthed pipette. As a result of this procedure theepithelial cells detached as intact units (villi and smallpre-crypt units; Fig. 2A,B) leaving behind large flaps ofcircular muscle (Fig. 2C). The epithelium in theseorganoid units remained as polarised intact layers (Fig.2D), and attached readily to both plastic and collagencoated dishes within 24-48 hours (Fig. 2E,F), forminglarge coalescing colonies of epithelium. Higher concen-

Fig. 2. (A) Haematoxylin- and eosin-stained section (4 fstn)of rat neonate small intestine (six-day-old) showing the villi(V) and small pre-crypt (C) regions (xlOO). Remainingphotographs (xl25 magnification; CM, circular muscle; M,muscularis mucosae; E, epithelium; LP, lamina propria)show cell pellets from the initial period after enzymicdigestion; (B) showing the mixture of intact epithelial unitsand cell debris (Fig. 1, step 3); (C) showing themesenchyme pellet that is removed (Fig. 1, step 4)following sedimentation under gravity; (D) from the laststage of the isolation (Fig. 1, step 6) showing the purifiedpellet of intact epithelial units (organoids of villi and cryptepithelium); (E) crystal violet-stained colony of theepithelium attaching after 24 hours in culture when placedinto a 24-well plate and grown in DMEM + 10% FCS(single-strength and high-glucose formulation; phase-contrast; X200); (F) crystal violet-stained colony ofepithelial cells that attached 48 hours after plating out intoa 24-well plate and grown in DMEM + 10% FCS (single-strength and high-glucose formulation; phase contrast,X200).

trations of collagenase and dispase allowed a morerapid isolation, but resulted in the epithelial cellsdissociating into single or clumps of cells. In this formthe cells did not attach efficiently or grow well.

Crude collagenase digestion also released manyclumps of epithelial cells and dissociated clumps ofmesenchymal cells but this procedure took much longer(several hours) and gave rise to cultures that from theoutset consisted of many cell types. All of the otherisolation methods tested, i.e. trypsin or divalent ionchelating solutions (Table 1) generally released purepopulations of dissociated epithelial cells, but theseusually failed to attach or survive in culture longer than24 hours. The deficiencies in these methods were notovercome by modifications in either the temperature orconcentration of the reagents in these protocols.

Culture mediumThe greatest difficulties with maintaining the prolifer-ation of isolated gut epithelium were associated withthe quality of the basal culture medium and theconstituents that were added to this medium. The bestresults were always obtained when the cells were grownin medium prepared with fresh single-strength medium.In contrast, less growth was seen when cells were grownin medium prepared from concentrated stocks and thiswas seen both in reduced cellularity in these cultures(crystal violet assay), as well as reduced incorporationof triated thymidine.

An example of how important the quality of themedium is to the growth of these cultures is provided inFig. 3. This shows the difference for two mediaformulations (DMEM and MEM) prepared either from10 x concentrated stocks or used as single-strengthsolutions. The media used in these experiments were allfreshly manufactured. Very little growth was everobserved when medium was prepared from concen-trated stocks that had been stored for more than onemonth (unpublished data).

Primary cultures of intestinal epithelium 223

. -

and COz concentration cultures. This indicated that at three different serum In these studies we employed bicarbonate-buffered concentrations (10, 5 and 2.5% FCS) a C 0 2 concen- medium and varied the concentration of C 0 2 to tration between 5 and 7.5% was reproducibly the most establish the effect of p H on the growth of the intestinal suitable for growth (data not shown).

224 G. S. Evans and others

1.4-

1.2-

I 10-

3 0.8-g1 0.6-1to

o0.2-

0.0

• = 10XDMEM + LW*• " 10XDMEM+GW• - 1XDMEM

10 12 14

1.4-

1.2-

1 1-0-

3 0.8Hg1 0.6-SCO

O

B

0.4-

0.2-

0.0

• = 10XMEM + LWA - 10XMEM + GW• -1xMEM

4 6 8 10Time (days)

12 14

Fig. 3. (A-B) The effect of different medium on the growthof primary cultures of neonate rat intestinal epithelium.Epithelial organoids were plated in 24 multi-well dishes inthe following medium; (A) 10x strength DMEM medium(high-glucose formulation reconstituted with laboratorysupplied tissue culture water (LW, de-ionised, doubledistilled and autoclaved); 10 x strength DMEM medium(high-glucose formulation) reconstituted with themanufacturer's supply of tissue culture water (GW, de-ionised double-distilled then heated to 65°C prior to flashcooling and bottling); single strength DMEM (high-glucoseformulation). (B) 10x strength MEM mediumreconstituted with laboratory supplied water (LW), 10xstrength MEM medium reconstituted with themanufacturer's supply of water (GW); single strengthMEM. All six media were fresh batches from themanufacturer and contained 5% FCS, 10 ng/ml EGF and2.5 /ig/ml insulin, and other supplements and antibiotics asindicated in Materials and methods. The MEM mediumwas also supplemented with non-essential amino acids.Dishes were fixed at intervals up to day 14 and thecellularity measured by the crystal violet procedure wherea stain density of 0.2 corresponds approximately to 2X104

to 3X104 cells/well and a value of 1.2 to llxlO4 to 13x10"cells/well. Each data point represents the average reading(plus standard error of the mean) for triplicate wells.

Characterisation of the cultured cellsTo establish which types of cells were present in thecultures, a battery of antibody probes and functionaltests were applied. The results of this study indicatedthat after the initial isolation, less than 10% of the cellswere non-epithelial, and these fell into three categories,smooth-muscle-like cells, smaller populations of myo-fibroblast-like cells and endothelial cells. In the pres-ence of higher serum concentrations (i.e. >2.5% FCS)the smooth-muscle-like cells usually formed largecolonies that surrounded the areas of epithelium.

The epithelial cells were identified using a combi-nation of antisera against cytokeratins 8,18 and 19 (seeTable 2). In the sections of intact neonate epitheliumthe strongest reaction was to the cytokeratin 8 antisera,which was localised throughout the epithelium (Fig.4A), whereas the staining patterns were weaker and notuniform when antibodies to cytokeratin 18 and 19 wereused. In the early stages of culture, the epithelialcolonies were distinguished easily using these antisera(Fig. 4B), but with time as the epithelial cells spread outonto the substratum the reaction became weaker andmany of these cells began to express vimentin (V9monoclonal). Positive control epithelial lines alsoreacted with these cytokeratin antibodies, and thesmooth muscle, endothelial and 3T3 cell lines wereunreactive.

The smooth-muscle-like cells were identified by theirreactivity with the smooth muscle a actin (1A-4)antibody in sections of intact neonate gut; the cells thatreacted with this antiserum were present throughoutthe lamina propria (associated with the blood vessels)and the muscularis mucosa. Elements of the circularsmooth muscle layer were also reactive (Fig. 4C). Theantibody stained colonies of cells in the establishedprimary cultures (days 7-14; Fig. 4D) and a positivecontrol human pericyte smooth muscle line. A furtherfeature of some of these smooth-muscle-like cells wastheir ability to invade thick collagen gels, an activitythat is shared with foetal fibroblasts and some smooth-muscle-like cells (Schor, 1980). We have also been ableto show that the growth of these smooth-muscle-likecells was inhibited by the addition of exogenous heparin(Table 3), like many other smooth muscle-like cells invitro.

In addition, three other cell types were observed,usually as small colonies or single cells. The first typewas reactive with the OX-43 antibody, which binds to anantigen expressed by rat endothelial cells, and showeduptake of flourescently labelled acetylate low-densitylipoprotein (LDLa; which is regarded as a functionalmarker of endothelial cells; Table 3). Another set ofcells reacted with the anti-desmin antisera (Fig. 5A),which in tissue sections reacted with the peri-cryptalfibroblasts and cells associated with the capillaries in thelamina propria. Finally, we have observed colonies ofcells with morphologies analogous to developing ner-vous tissue, e.g. they exhibited extensive dendriticprocesses and only reacted with the anti-vimentinantibody (Fig. 5B).

Primary cultures of intestinal epithelium 225

B

M«

VE

4

CE

VE

LP

CM>-

Fig. 4. (A-D) Photographs of: (A) a frozen section (5 \ai\ thick) of rat duodenum (six-day-old pup) stained with a mixtureof antibodies against cytokeratins 8 (LE-41), 18 (LE-61) and 19 (LP-2K) all at 1/50 dilution. The primary antibody layerwas revealed with peroxidase-conjugated goat anti-mouse (at 1/75 dilution) and diamino-benzidine/nickel chloridesubstratum; CE, crypt epithelium; VE, villus epithelium; M, muscularis mucosae (xl50). (B) Primary culture colonies ofintestinal epithelial cells (after two days in culture) plated onto glass coverslips. The cells were fixed and stained accordingto the procedure described in above (phase-contrast, X300). (C) A frozen section (5 /an thick) of rat duodenum (six-day-old pup) stained with monoclonal antibody 1A-4 against smooth muscle a actin at 1/250 dilution. The primary antibodylayer was revealed with peroxidase-conjugated goat anti-mouse IgG (at 1/75 dilution) and diamino-benzidine/nickel chloridesubstratum; VE, villus epithelium; LP, lamina propria; M, muscularis mucosae; CM, circular muscle (xl50). (D) Primaryculture colonies of intestinal cells (after seven days in culture) plated onto glass coverslips. The cells were stained withmonoclonal antibody 1A-4 against smooth muscle a actin at 1/250 dilution according to the method described above(phase-contrast, X300).

Serum quality and concentrationSeveral batches of foetal calf serum were screened usingthe IEC 18 and 6 cell lines as an assay system, and theeffect of different concentrations of the optimal batchof foetal calf serum on the growth of the primarycultures was assessed. Using single-strength DMEMand in the presence of insulin (2.5 ^g/ml) and EGF (10ng/ml) there was an increase in the growth of theintestinal cultures from 2.5% to 10% FCS (Fig. 6).Serum levels lower than 1% were not able to sustain thegrowth of these primary cultures beyond day 10.

Although growth was most rapid at high serumconcentrations (Fig. 6), these conditions were not themost appropriate for the epithelial cells, since coloniesof the smooth-muscle-like cells were also more abun-dant (Fig. 5C). At concentrations of FCS <2.5% thegrowth of these smooth-muscle-like cells was reducedso that they formed smaller colonies surrounding theepithelial cells (Fig. 5D), and hence the lower concen-tration of serum has been used for the furtherdevelopment of a method to grow mono-cultures of thisepithelium.

226 G. S. Evans and others

Table 3. Characterisation of gut cells by functionalassays

Assay for different cell types

Attachment (within 20 minutes) of smooth-muscle-like cells toplastic and clean glass but not of epithelial cells

Specific internalisation of acetylated low density lipoprotein (LDL)by endothelial cells*

Ability to invade collagen gels a property of normal fetal likefibroblastst

Smooth muscle cells inhibited by doses of heparin as low as 5/

'Personal communication and gift from Dr D. West(Department of Paediatric Oncology, Christie Hospital,Manchester).

tSchor (1980).tWright et al. (1988).

Plastic and biological substrataTo investigate the role of extracellular matrix interac-tions upon the epithelial cultures we have examined therequirements for different types of substrata, includingplastic-coated with purified constituents of the extra-cellular matrix, or whole extracellular matrices. Theresults of these experiments suggest that the type ofsubstratum is important, but that the effect is primarilyupon the morphological appearance of the epithelialcells and not specifically upon proliferation.

The best results were obtained with completebiological matrices isolated from bovine corneal endo-thelial cells and Swiss Balb 3T3 cells, since thesepromoted attachment of most of the isolated cells.However, these matrices were less satisfactory forroutine applications. Good attachment was also facili-tated by the use of dried collagens (particularlyVitrogen and rat tail collagens), but did not affect the

y.'\'s TV

?M. > > . * *

Fig. 5. (A-D) Photographs of non-epithelial cells stained by: (A) anti-desmin antisera (1/50 dilution) 10 days after initiatingprimary cultures of rat intestinal epithelial cells; and (B) glia-like cells stained with anti-vimentin antisera (1/100 dilution)10 days after initiating primary cultures of rat intestinal epithelial cells. The primary antibody layer was revealed withperoxidase-conjugated goat anti-mouse IgG (at 1/75 dilution) and diamino-benzidine/nickel chloride substrate.(C,D) Photographs of cells fixed in methanol and stained with crystal violet and photographed under phase-contrast;(C) smooth-muscle-like cells in primary cultures of neonate rat intestinal cells (day 10; xl50); (D) a colony of epithelium(E) surrounded by smooth-muscle-like cells (SM) (day 14; x250).

Primary cultures of intestinal epithelium 227

Day 10 Day 14

Fig. 6. The effect of growing primary cultures of ratneonate intestinal epithelial cells in DMEM medium(bicarbonate-buffered, single-strength, high-glucoseformulation) at five different foetal calf serum (FCS)concentrations (0, 0.5, 1.0, 2.5, 5.0 and 10% FCS). Theepithelial organoids were plated out in 24 multi-wells andplates fixed at days 2,7 10 and 14. The changes incellularity were measured by the crystal violet procedurewhere a stain density of 0.2 corresponds approximately to2X104 to 3X104 cells/well and a value of 1.2 to 11x10* to13X104 cells/well. Each data point represents the averagereading (plus standard error of the mean) for triplicatewells.

growth rate of the cultures compared to plasticsubstratum (Fig. 7). Results from the tritiated thymi-dine assay showed similar levels of incorporation forboth plastic and collagen substrata, with the highestincorporation on day 21. Purified fibronectin andcollagens I and IV showed no improvement in facilitat-ing attachment compared to Vitrogen, and very poorattachment (and impairment of growth) was observedwith hydrated gels (1-2 mm thick) of Vitrogen andMatrigel. A comparison of 24 multi-well plates from avariety of manufacturers was made, but no substantive,reproducible differences were observed.

Discussion

There have been many previous attempts to develop amethod to gTOw in culture normal intestinal epithelialcells, and usually these reports have emphasised thedifficulties of maintaining the proliferation of these cellswhen isolated from the gut (Ke'dinger et al., 1987a;Neutra and Louvard, 1989). There have been a fewnotable exceptions where the propagation of normalcolonic cells for several months in culture has beenreported (Chopra and Yeh, 1981). However, this hasstill not led to the widespread use of these methods.Without stimulation of proliferation there has been no

1.4"

1.2-

1 1°"J 0.8-

•s•a 0.6-

IO 0.4-

0.2-

0

- Collagen-coated

- Plastic

I . I10 15 20

Time (days)25 30

Fig. 7. The effect of growing primary cultures of ratneonate intestinal epithelial cells in DMEM medium(bicarbonate-buffered, single-strength, high-glucoseformulation) with 2.5% foetal calf serum, 10 ng/ml EGF,2.5 /ig/ml insulin and supplements as described in Materialsand methods. The cells were grown in 24 multi-wells eitherwith or without a dried coating of bovine dermal collagensI + HI (Vitrogen; as described in Materials and methods)and plates fixed at days 1, 7, 14, 21 and 28. The changes incellularity were measured by the crystal violet procedurewhere a stain density of 0.2 corresponds approximately to2x10" to 3X104 cells/well and a value of 1.2 to 11x10* to13xlO4 cells/well. Each data point represents the averagereading (plus standard error of the mean) for triplicatewells.

objective way to identify the optimal in vitro conditions,and this paucity of hard data has inevitably led to amore anecdotal approach. Our aim has been, wherepossible, to use a quantitative approach to develop aprimary culture method for normal gut epithelium.

Basic aspects of culture methodology have notreceived great attention in previous publications, butsome of these conditions were found to be critical in thisstudy and could well account for the observed difficul-ties in getting the isolated gut cells to proliferate invitro. In this study the basic culture conditions that gaverise to the proliferation of epithelial and smooth-muscle-like cells are described.

The optimal method of isolating the gut epitheliumwas found to be a mixture of crude collagenase anddispase, a combination of enzymes that has beenemployed in some previous studies (Gibson et al.,1989). The success of this method may be associatedwith the low level of tryptic activity and the ability tomaintain the integrity of the epithelium. In the six-daysuckling rat gut, the epithelium forms small blunt villibut there are no well-developed crypts. The pre-cryptpopulation (and hence proliferative cells) is present onthe inter-villus plateau regions, or as small buds at thebase of the villi (Fig. 2A). After incubation in thisenzymic solution, gentle pipetting releases the villi andthe pre-crypt cells, some of which remain attached tothe base of the villi. These units have been termedepithelial organoids, since they retain their morphologi-

228 G. S. Evans and others

cal integrity on isolation and the closely associatedmesenchymal cells remain attached at the core of thevilli (Fig. 2D). Isolation of intact crypt units from theadult rat gut has not yet been achieved with this methodto our satisfaction, but we have applied it successfully tothe isolation of pure populations of colonic crypts fromthe adult human colon (Evans, de Silva and Reddel,unpublished data).

The isolation of other epithelial cells in the form ofintact 'organoids' is also seen to be a key factor in themaintenance of growth and function in vitro (Barcellos-Hoff and Bissell, 1989). The apparent disadvantage ofisolating epithelial cells as organoids is the contami-nation by attached mesenchymal cells. Collagenase/dispase digestion releases large numbers of mesenchy-mal cells from the lamina propria, which are usuallypresent as single cells and smaller clumps. Theadvantage of the gentle sedimentation steps employedin this method is that it has proved to be a very effectiveway of removing unwanted types of cells and micro-organisms, without recourse to the use of gradedmeshes or cocktails of anti-biotic and anti-mycoticagents (Moyer et al., 1990). This differential sedimen-tation procedure can easily be adapted to enrich forsingle cells or very small cell clumps without a large lossof material.

The investigation of many other isolation methods(see Table 1) suggested that the maintenance of closecellular interactions is important if these cells are to begrown in culture. Methods that dissociated the epi-thelium by tryptic or chelating activity never yieldedcells that would grow in culture (Table 1). Apart frompossible toxic effects, this could support the idea thatthe maintenance and quality of the cellular interactionsis most important for proliferation of gut epithelium.

Immunocytochemical characterisation of these early-stage cultures showed that >90% of the attached cellswere positive for cytokeratin expression, whilst theremainder consisted of single cells and small clumps ofsmooth-muscle-like cells, myo-fibroblast-like cells andendothelial cells. Within a short period after the initialattachment and spread of these epithelial cells, theychanged to a flat/cuboidal morphology and lost thesupra-nuclear heterophagosomes and large Golgibodies that typify the normal enterocyte in the neonategut (Fig. 8A-C). These colonies of epithelium alsobecame unreactive to an antibody (Mab 233) raisedagainst surface aminopeptidase, suggesting further lossof functional activity. However, in the initial stages inthe development of this culture technique no significantproliferation of any of these attached epithelial cellswas observed, they often remained in culture for up totwo weeks and gradually detached and died (incommon with previous published observations; Gibsonet al., 1989; Kondo et al., 1984).

The major contributory factor to the declining stateof these cultures was found to depend upon the qualityof the basal culture medium (Fig. 3), and the need forhigh-quality reagents in this primary culture systemcannot be emphasised enough. The exquisite sensitivityof these primary cultures contrasted with that of the

Fig. 8. (A) Photograph of an organoid of epitheliumattaching in culture 24 hours after isolation, showing thecentral mass of the cells still unattached (T) and cellsspreading on to the dish at the periphery (P). The cellswere grown in DMEM (single-strength, high-glucoseformulation) + 10% FCS (day 1). The cells were fixed inmethanol and photographed under phase contrast (250 xmagnification). (B) Ultrastructure of cells taken from thetip (T) of an attaching organoid showing large supra-nuclear heterophagosomes that result from the apicalpinocytosis of milk proteins. The epithelial cells weregrown under the same conditions as described above, butfixed in 2.5% glutaraldehyde and 0.1% osmium tetroxideand embedded in Epon-Araldite (X2500; bar, 4 jun).(C) Ultrastructure of cells taken from the edge (P) of anattaching organoid of epithelium grown under the sameconditions (X7500; bar, 1.33 fan).

IEC 6 and 18 lines that were used to batch test many ofthe reagents that were initially chosen for these cultureexperiments (unpublished data). Storage of concen-trated medium stocks may lead to sequestration of ions(e.g. on buffer salts) that are essential for gut cells inprimary cultures. As short a period as 24 hours in amedium prepared from liquid concentrates is sufficientto suppress the proliferation of these primary cultures(unpublished data). These problems were overcome bythe use of fresh batches of single-strength DMEM (highglucose formulation) medium and plastic disposables.However, even the plastic disposables may need to beselected carefully, since there is evidence that volatileconstituents of some plastics can affect primary culturesof certain tissues (Knight, 1990).

With optimal conditions these intestinal primarycultures were able to proliferate very rapidly and wouldfill the multi-wells with cells within 10-14 days. In 10%FCS, this represents a starting population of about25,000 cells/well (on day one) leading to 250,000cells/well (on day 14). This enabled us to applyquantitative assays of growth and proliferation to assessthe affects of other changes in the culture environment.

Close examination of cultures indicated the presenceof heterogeneous populations of cells not all of whichwere epithelial in appearance (Fig. 5A-C). Most ofthese were smooth-muscle-like cells. Such smooth-muscle-like cells have been reported in a recentpublication by K6dinger et al. (1990) to be a hetero-geneous population within the intestinal lamina pro-pria. The presence of these heterologous cell types maybe important for epithelial cell proliferation, since onlyin these mixed cultures have growing colonies ofepithelium been observed previously (Kondo et al.,1984; Ke"dinger et al., 1987a). This contrasts withisolations of purified gut epithelium that can attach butdo not proliferate in vitro (Quaroni and May, 1980;Vidrich et al., 1988). It is significant that success inestablishing epithelial cell lines from the normal gut hasfollowed the use of crude collagenase digestion and spillculture techniques, with the colonies of epitheliumbeing isolated (with cloning cylinders) from a mixed cellculture (Quaroni and May, 1980).

Primary cultures of intestinal epithelium 229

' : w ..-./

'1

Reducing the growth of the smooth-muscle-like cellswas an important objective if this method was to beused as a primary culture assay for epithelial cells.

Lowering the serum level to concentrations <2.5%FCS was one way of keeping these cells in check. As aresult, the reduced level of platelet-derived growth

230 G. S. Evans and others

factor (present in the serum) may be one possibleexplanation for this effect upon the smooth-muscle-likecells. At low serum concentrations, the growth of theepithelial cells was still not expansive, and so theaddition of supplements, e.g. transferrin, lipids, hor-mones etc, to enhance the growth of these cells is nowbeing pursued. Initially EGF and insulin were added tothe serum because of their general application inepithelial cell culture. Subsequently we have examinedin greater detail the contribution of these two factors,and in their absence serum concentrations lower than2.5% are less able to maintain the gut epithelium inculture. EGF also promotes the migration of theattaching cells in the early stages of the culture, and thepromotion of growth by insulin and EGF is synergistic(unpublished data).

Heparin has been recognised as a potent inhibitor ofthe proliferation of many smooth muscle cells in vitro(Wright et al., 1988). At concentrations up to 100 jig/mlheparin was effective at preventing the proliferation ofcolonies of the smooth-muscle-like cells (unpublishedobservations).

The role of the substratum in these cultures wasanother aspect which warranted attention. With col-lagen substrata, the morphological appearance of theepithelial colonies was much more organised, and thecells had a compact cobblestone appearance. Incontrast the poor growth of the epithelial cells on thehydrated collagen gels suggests that this provided aninsufficiently stable substratum, and we have observedsimilar results with Matrigel. There is good evidencethat some of the more complex biological substratum,e.g. Matrigel, can induce differentiation of someepithelial cells in vitro (Montgomery, 1986). This is oneaspect that we have not investigated, but there isalready evidence to support this phenomenon incultured intestinal epithelial cells (Schor,1980; K6d-inger et al., 1987b).

The development of a primary culture system for thegut epithelium is a worthwhile objective, particularly ifit could be used as a reproducible assay to define furtherthe in vivo requirements of the stem cell 'niche'.Although progress in this direction has been slow,methods to maintain other tissue stem cells have bynecessity been complex and essentially uncharacter-ised, e.g. the use of long-term bone marrow cultures forhaematopoietic cells and of 3T3 fibroblasts for embryostem cells (Dexter, 1979; Evans and Kaufman, 1981).The complexity of these systems, however, has notprevented the eventual identification of key regulatorymolecules that maintain the haematopoietic and em-bryo stem cells (Zsebo et al., 1990; Smith et al., 1988).Apart from methodological considerations, the mainoutcome of this present work is the recognition that theepithelial cells need to be isolated in a way thatmaintains their integrity and organisation. Continuedassociation with mesenchymal cells in vitro is alsobeneficial, but the major problem is the growth of thesmooth-muscle-like cells at higher serum concen-trations. This can be partly overcome by the proceduresdescribed above. The need for these heterologous

interactions further supports the substantive evidencefor epithelial-mesenchymal interactions in the gut,produced by Kfdinger et al (1987a,b), and may indicatethe involvement of paracrine growth factors such asthose described for the epidermis (Finch et al., 1989).

Since there has been very limited success in inducingcharacteristic in vivo pathways of differentiation in theestablished epithelial lines derived from normal gut, itis important to determine if a primary culture modelwill be of more use for in vitro functional studies of thisepithelium. The ability of monolayer cultures torecapitulate features such as cell polarity, maintenanceof tight junctions and transmembrane electro-potentialdifferences, transport functions, crypt-to-villus entero-cyte maturation, and the production of different celllineages (e.g. Paneth and Goblet cells) are difficult andunresolved questions. We are currently using thisprimary culture model to investigate what effectsextracellular matrix constituents, hormones, growthfactors and known agents for inducing differentiation(e.g. retinoic acid and dimethyl sulphoxide) have uponthese aspects of epithelial cell function in vitro.Preliminary evidence suggests that the epithelial cellsremain polarized with junctional complex and apicalmicro-villi and retain expression of, e.g. alkalinephosphatase, which is seen strongly expressed in largegrowing epithelial colonies which develop between days5 and 10. With respect to questions of cell lineage, theearly stage of development (six days post-natal) of thegut must be borne in mind, since cell types such asentero-endocrine cells begin to appear only late in fetaldevelopment, and Paneth cells 5-6 days after birth. Itwill, therefore, be of interest to determine what effectsdifferentiation-promoting agents have upon a popu-lation of cells undergoing commitment events.

A long-term aim is to identify factors that morespecifically promote the proliferation of the epithelialcells, leading essentially to a monoculture. We haverecently identified feeder cells (bovine corneal endo-thelial cells) that are able to promote selectively theproliferation of rat intestinal epithelium. This activity isdue to a soluble factor found in the conditioned mediumproduced by this cell line (N. Flint and G. S. Evans,unpublished observations). In addition to the inhibitoryeffects upon the smooth-muscle-like cells, heparin hasbeen found to stimulate significantly the proliferation ofthe epithelial cells independently of a biological feederlayer (N. Flint and G. S. Evans, unpublished obser-vations). The mechanisms by which these signalsstimulate the proliferation of the intestinal epithelialcells is being investigated.

This work was supported by the Cancer Research Cam-paign. We also acknowledge the contribution of Mrs P. Taylorand the technical advice of Yvonne Barlow (for the BCEmatrix preparation method); Dr A. Quaroni for the IEC18,17 and 6 lines; Dr A. Schor for the smooth muscle line; DrA. Ager for the Mab 233 and 266 antibodies and Dr B. Lanefor the LE61, LE41, LP2K and BA19 antibodies. We areindebted to Dr C. Booth for reading the manuscript andhelpful discussions.

Primary cultures of intestinal epithelium 231

References

Barcellos-Horr, M. and BlsseU, M. J. (1989). Mammary EpithelialCells as a Model for Studies of the Regulation of Gene Expression:In Functional Epithelial Cells In Culture (ed. Matlin, K.S. andValentich, J.D.), pp. 363-398. Alan R. Liss, Inc., New York.

Barclay, A. N. (1981). The localisation of populations of lymphocytesdefined by monoclonal antibodies in rat lymphoid tissues.Immunology 42 (4), 593-600.

Blay, J. and Brown, K. D. (1984). Characterisation of an epithelioidcell line derived from rat small intestine: Demonstration ofcytokeratin filaments. Cell Biol. Int. Rep. 8, 551-560.

Brasaemle, D. L. and Attie, A. D. (1988). MicroELISA readerquantification of fixed, stained, solubilised cells in microtitredishes. Biotechniques 6, 418-419.

Chopra, D. P. and Yeh, K. Y. (1981). Long term culture of epithelialcells form the normal rat colon. In Vitro 17, 441-449.

Debus, E., Weber, K. and Osborn, M. (1983). Monoclonal antibodiesto desmin, the muscle-specific intermediate filament protein.EMBO J 2,(12) 2305-2312.

Dexter, T. M. (1979). Cell interactions in vitro. Clin. Haemal. 8, 453-460.

Evans, M. J. and Kaufman, M. (1981). Establishment in culture ofpluripotential cells from mouse embryos. Nature 292, 154-156.

Finch, P. W., Rubin, J. S., Mlki, T., Ron, D. and Aaronson, S. A.(1989). Human KGF is FGF-related with properties of a paracrineeffector of epithelial cell growth. Science 245, 752-755.

Flint, N., Cove, F. L. and Evans, G. S. (1991). A low temperaturemethod for the isolation of small intestinal epithelium along thecrypt-villus axis. Biochem. J. 280, 331-334.

Franke, W. W., Weber, K., Osborn, M., Schmid, E. andFreudenstein, C. (1978). Antibody to prekeratin. Decoration oftonofilament like arrays in various cells of epithelial character. ExplCell Res. 116 (2), 429-445.

Freshney, R. I. (1978). Cultures of animal cells. A Manual of BasicTechnique, chap. 18, p. 236, 2nd edn, Alan R. Liss Inc, New York.

Gibson, P. R., Vande Pol, E., Maxwell, L. E., Gabriel, A. and Doe,W. F. (1989). Isolation of colonic crypts that maintain structuraland metabolic viability in vitro. Gastroenterology 96, 283-291.

Gospadarowicz, D., Delgado, D. and Vlodavsky, I. (1980). Permissiveeffect of the extracellular matrix on cell proliferation in vitro. Proc.Nat. Acad. Sci. U.S.A. 77, 4094-4501.

Hedman, K., Kurkinen, M., Alitalo, K., Vaherl, A., Johansson, S.and H66k (1979). Isolation of the pericellular matrix of humanfibroblast cultures. /. Cell Biol. 81, 83-91.

K£dinger, M., Haffen, K. and Slmon-Assmann, P. (1987a). Intestinaltissue and cell cultures. Differentiation 36, 71-85.

Kldlnger, M., Simon Assmann, P., Alexandra, E. and Haffen, K.(1987b). Importance of fibroblastic support for in vitrodifferentiation of intestinal endodermal cells and for their responseto glucocorticoids. Cell Differ. 20, 171-182.

K&llnger, M., Simon Assmann, P., Boudges, F., Arnold, C ,Alexandra, E. and Haffen, K. (1990). Smooth muscle actinexpression during rat gut development and induction in fetal skinfibroblastic cells associated with intestinal embryonic epithelium.Differentiation 43, 87-97.

Knight, D. E. (1990). Disposable filters may damage your cells.Nature 343, 218.

Kondo, Y., Rose, I., Young, G. P. and Whitehead, R. H. (1984).Growth and differentiation of fetal rat small intestinal epithelium intissue culture. Exp. Cell Res. 153, 121-134.

McAteer, J. and Cavanagh, T. J. (1983). Medium hydrated collagengels as an explant support in organ culture. J. Tiss. Culture Meth. 7,117-122.

Montgomery, R. K. (1986). Differentiated function of cultured fetalintestinal epithelial cells determined by extra-cellular matrix.Gastroenterology 92, 1540-1548.

Moyer, P. M., Dixon, P. S., Culpepper, A. L. and Aust, J. B. (1990).Colon cancer cells. In In Vitro Propagation and Characterisation ofNormal, Preneoplastic, and Neoplastic Colonic Epithelial Cells,(ed. Moyer, M. P. and Poste,G. H.),chap. 5, Academic Press Inc.,New York.

NSgrel R., Rampal P., Nano, J-P., Cavenel, C , and Allhaud, G.(1983). Establishment and characterisation of an epithelialintestinal cell line form rat fetus. Exp. Cell Res. 143, 427-437.

Neutra, M. and Louvard, D. (1989). Differentiation of Intestinal CellsIn Vitro: In Functional Epithelial Cells In Culture (ed. Matlin, K.S.and Valentich, J.D.), pp. 363-398. Alan R. Liss, Inc., New York.

Osborn, M., Debus, E. and Weber, K. (1984). Monoclonal antibodiesspecific for vimentin, Eur. J. Cell Biol. 34 (1), 137-143.

Quaronl, A. and May, R. J. (1980). Establishment andcharacterisation of intestinal epithelial cell cultures. Meth. CellBiol. 21B, 403-427.

Schor, S. L. (1980). Cell proliferation and migration on collagensubstrata in vitro. J. Cell Sci. 41, 159-175.

Skalli, O., Schurch, W., Seemayer, T., Lagace, R., Montandon, D.,Pittet, B. and Gabblani, G. (1989). Myofibroblasts from diversepathologic settings are heterogenous in their content of actinisoforms and intermediate filament proteins. Lab. Invest. 60 (2),275-285.

Smith, A. G., Heath, J. K., Donaldson, D. D., Wang, G. G., Moreau,J., Stahl, M. and Rogers, D. (1988). Inhibition of pluripotentialembryonic stem cell differentiation by purified peptides. Nature336, 688-690.

Vidrich, A., Ravmdranath, R., Fassi, K. and Targan, S. (1988). Amethod for the establishment of normal adult mammalian colonicepithelial cell cultures: In Vitro Cell. Dev. Biol. 24, 188-194.

Weiser, M. M. (1973). Intestinal epithelial cell surface membraneglycoprotien synthesis. J. Biol. Chem. 248, 2536-2541.

Wright, T. C , Castellot, J. J., Diamond, J. R. and Karnovsky, M. J.(1988). Regulation of cellular proliferation by heparin and heparansulphate. In Heparin (ed. Land, D. and Lindahl, U.), chap 14, pp.295-316. Edward Arnold, London.

Zsebo, K. M., Wypych, J., McNiece, I. K., Lu, H. S., Smith, K. A.,Karkare, S. B., Sachdev, R. K., Yuschenkoff, V. N., Birkett, N. C ,Williams, L. R., Satyagal, V. N., Tung, W., Bosselman, R. A.,Mendlaz, E. A. and Langley, K. E. (1990). Identification,purification, and biological characterisation of hematopoietic stemcell factor from Buffalo Rat Liver-Conditioned Medium. Cell 63,195-201.

(Received 17 June 1991 - Accepted 25 September 1991)