Establishment of a tissue culture system for epithelial cells derived from

human pancreas: a model for the study of cystic fibrosis

ANN HARRIS and LINDSAY COLEMAN

Pacdialric Research Unit, United Medical and Dental Schools of Guy's and St Tttowas' Hospitals, 8th Floor Guy's Totcer,London Rndge, London SE1 9RT, UK

Summary

A tissue culture system for epithelial cells de-rived from human foetal pancreas has beenestablished. The cultured cells show many ultra-structural features of interlobular duct cells. Im-munocytochemical and histochemical evidenceis presented in support of the view that these cellsare ductal in origin. They are likely to be one of

the few cell types that express the basic defect ofcystic fibrosis in vitro. The cells may be passagedand sufficient material obtained to permit bio-chemical and molecular biological analysis.

Key words: pancreas, epithelial cells, in vitro, cysticfibrosis.

Introduction

The autosomal recessive disease cystic fibrosis (CF)affects three main organ systems: the respiratory tract,the digestive system and the sweat glands. The involve-ment of the digestive system is due primarily to fibrosisof the pancreas following blockage of small pancreaticducts with inspissated secretions. This fibrosis pre-vents normal production and release of pancreaticdigestive enzymes. To date very little research has beencarried out on pancreatic tissue in CF due to thedifficulty of obtaining material in a viable condition. Itis well known that pancreatic autolysis occurs veryrapidly post-mortem and the situation in CF is ham-pered further by the destruction of pancreatic tissueassociated with the disease. In order to circumventthese problems we have used pancreatic tissue frommid-trimester pregnancy terminations, from patientsnot known to be at risk for CF.

At 9 weeks gestation the pancreas is largely com-posed of a simple system of branched ducts surroundedby loose interstitial tissue containing largely undiffer-entiated cell types. Between 12 and 14 weeks, cells fromthe ducts invade the interstitial tissue and start forminglobular structures. Though primitive acini may be seenafter 12 weeks (Laitio et al. 1974) the first matureacinar cells do not appear until somewhere near the20th week of gestation, and the acini do not develop alumen until the 24th week (Adda et al. 1984). The mid-trimester pancreas is thus less susceptible to autolysisJournal of Cell Science 87, 695-703 (1987)Printed in Great Britain © The Company of Biologists Limited 1987

than post-natal pancreas, since functional digestiveenzymes are not yet being synthesized. From changesin pancreatic histology it appears that the CF gene isalready being expressed in 20-week-old foetuses (Bou6etal. 1986).

Until recently it was not certain which tissuesactually expressed the basic defect in CF. As a resultresearch was carried out on the most readily availableclinical materials, namely skin fibroblasts, erythro-cytes, blood serum and various exocrine secretions.From the pathology of CF it seemed reasonable tosuspect the basic defect of being a functional abnor-mality in the secretory epithelia of pancreas, sweatgland and the respiratory system. It has now beenshown that, in CF, the epithelia of sweat gland ductsand parts of the respiratory system show a reducedpermeability to chloride ions (Quinton, 1983; Knowleset al. 1983; Sato & Sato, 1984; Widdicombe et al.1985; Welsh & Liedtke, 1986; Frizzell et al. 1986). Thedefect seems to be caused by abnormal regulation ofchloride ion channels. By analogy with these datapancreatic duct epithelial cells would be expected toexpress similar ion-transport defects in CF (bearing inmind that the pancreatic duct epithelium secretesbicarbonate ions rather than chloride ions).

The pancreatic duct system is complex and it is notclear which of the epithelial cell types contained withinit might be functionally abnormal in CF. The acinicontain intercalated duct cells (centroacinar cells) thatlead into larger intralobular ducts. These in turn join

695

onto interlobular ducts that increase in size as they passtowards the main pancreatic duct. The latter forms thechannel between the pancreas and the duodenum. Theultrastructure of the epithelial cells lining the intra-lobular ducts, interlobular ducts (small and large) andthe main pancreatic duct has been studied extensivelyin adult pancreas by Kodama (1983). Apart fromindirect evidence of the main functions of each of thesecell types, on the basis of the nature of intracellularorgandies, little is known about the specific propertiesof the different ductal epithelial cells.

Pancreatic duct epithelial cell culture systems havealready been set up for hamster, mouse, guinea pig andbovine pancreas (Stonere^ al. 1978; Jones et al. 1981;Hirataefa/. 1982; Resau etal. 1983; Sato<>/«/. 1983).The cultured cells have characteristic surface microvilliand secrete mucins as they do in vivo. There is also onereport of a short-term human pancreatic duct cellculture system (Jones et al. 1981). Since there is noadequate animal model for CF, it is essential to work onhuman tissue.

We have established a tissue culture system forepithelial cells derived from normal mid-trimesterfoetal pancreas. The system as described below resultsin the establishment of primary cultures of epithelialcells from about 80 % of pancreases (a total of morethan 80). The cells can be passaged up to five timesbefore reaching terminal crisis and may remain inculture for up to 20 weeks. They have been character-ized by morphological, histochemical and immuno-cytochemical criteria.

Materials and methods

MaterialsPancreases were obtained within 48 h from mid-trimesterprostaglandin-induced terminations or spontaneous abor-tions (either normal or with known abnormalities other thancystic fibrosis). The pancreas, still attached to the duo-denum, was removed from the foetus and was washed once intissue culture medium containing antibiotics. Mesenteryaround the pancreas was trimmed away and then 1-2 ml ofcollagenase (Sigma type IA at 0-5 mgml"1) was injected intothe organ. When visible, the main duct was micro-dissectedfrom the inflated pancreas. In early mid-trimester pan-creases, in which the main duct could not be clearly dis-tinguished, the central core of the head of the pancreas wasisolated instead.

Micro-dissected tissue was washed in tissue culture me-dium, cut into pieces of about 1-2 mm diameter and platedout in CMRL1066 medium containing 20 % foetal calf serum(FCS) (Gibco UK); penicillin (100 units ml" ' ) ; strepto-mycin (lOOj(gmr'); L-glutamine (4mM); insulin (0-2unitsml"1); cholera toxin ( 1 0 ~ ' ° M ) and hydrocortisone {\ fxgml"') , all from Sigma. Cultures were routinely maintained at37°C in a wet 5% CO2 incubator. Various cell types wereseen migrating from the primary explants after 3-10 days.

Culture media and substratesThe growth characteristics of the pancreatic epithelial cellcolonies were monitored in a variety of tissue culture media(see Table 1) and on several different culture substrates (seeTable 2). Tissue culture plastics were purchased as follows:multiwells from Nunclon, Gibco; flasks from Sterilin andFalcon, Becton Dickinson; and Primaria from Falcon,Becton Dickinson.

Removal oj fibroblasts from epithelial cell culturesThree alternative methods have been used to eliminatefibroblasts: first, an EDTA (0-01 %) wash, which removesfibroblasts from the substrate before epithelial cells, so longas the former have not infiltrated the epithelial colonies;second, ricin-conjugated anti-thy-1 (Paraskeva et al. 1985)because thy-1 antigen is expressed on fibroblasts but not onepithelial cells (Cotmore et al. 1981; Daar & Fabre, 1983);third, physical removal of fibroblasts with a rubber police-

ImmunocytochemistryImmunocytochemistry was done by standard procedures.LE61, a monoclonal antibody that is specific for the cytokera-tin intermediate filaments of simple epithelial cells, wasdonated by Dr Birgitte Lane. Monoclonal antibodies tocytokeratins 8 and 19 were purchased from Amersham.Polyclonal antisera against carbonic anhydrase I and II weredonated by Dr Norman Gregson. Monoclonal antibodies Ca2

Table 1. Tissue culture media

MediumEpithelial

cell growthFibroblast

growth

CMRL1066 (Gibco)+ 20% FCS with + + + +supplements*

CMRL1066 + 20% FCS + +Ham's F10 (Gibco)+ 20% FCS +Dulbecco's MEM (Gibco)+ 20% FCS +RPM1-1640 (Flow) + 20% FCS +MCDBf (no FCS + supplements) —

•Insulin (0-2unitsmr ' ) , cholera toxin ( 1 0 ~ ' ° M ) ,hydrocortisone (1 j igmP1).

t Hammond et al. (1984).

Table 2. Culture substrates

Primaryexplants

Passagedcolonies

Plastic multiwells or plastic flasksPnmariaCollagen-coated* plastic multiwellsCollagen-coated* glass covcrslipsGlass coverslipsFeeders-layersf (pancrcas-dcnvcd

fibroblasts or NIH-3T3 cells)

•Collagen (Sigma type IV, human placenta) at 1 mgml ' in1:1000, glacial acetic acid/water, placed on culture substrate anddried at 37°C for 16h.

•[Treated with mitomycin C (4^gml" ' ) for 2h.

696 A. Hams and L. Coleman

and HMFG2, both of which react with the same mucousglycoprotein, known to be expressed on certain ductalepithelia (Harris, 1987), were donated by Professor HenryHarris and Dr Joy Burchell, respectively. Peroxidase-conju-gated rabbit anti-mouse immunoglobulins and swine anti-rabbit immunoglobulins were purchased from DAKO-Patts.Fluorescein-conjugated goat anti-mouse immunoglobulinsand goat anti-rabbit immunoglobulins were purchased fromCappel laboratories.

MucinsMucins were stained by standard procedures using AlcianBlue-periodic acid-Schiff (PAS) (Cook, 1982). Under theseconditions acid mucins stain blue, neutral mucins stain pinkand mixtures stain purple.

Electron microscopyCells were fixed in 2 5 % phosphate-buffered glutaraldehyde,pH 7-3, for 15 min, while still attached to the plastic or glasssubstrates, and washed overnight in phosphate buffer,pH 7 3 , containing sucrose (0-25 M). Samples were immersedin 1% osmium tetroxide (Millonigs) for 20 min and thendehydrated through an ethanol series. Cells growing onplastic substrates were then directly embedded in TAABresin for 18 h at 60cC. Cells growing on glass were treatedtwice for 5 min in propylene oxide then once in 50%propylene oxide/50 % resin for 1 h before being embedded inresin. Thin sections (10—90^m) were cut with an LKBultratome and then stained in uranyl acetate saturatedsolution in 50% ethanol) and lead citrate (0-4%) by standardprocedures. Samples were examined in a Hitachi HU 12Atransmission electron microscope.

Results

Cell types

Two main epithelial cell types grow out of primarypancreatic explants as illustrated in Fig. 1: smalltightly packed cells (Fig. 1A) and larger, more looselypacked cells (Fig. IB). These two cell types are usuallyfound together in one culture and when this is observedthe larger cells often occur at the periphery of coloniesof tightly packed cells. Occasionally only one cell typemay be present in a culture. Generally the tightlypacked cells are the more abundant type. The cul-tures usually form characteristic structures in vitw(Fig. 2A,B) with areas of the smaller cells separated bylarge streaming cells that, though they often lookfibroblastic, in fact express epithelial cell cytokeratinmarkers such as those detected by the monoclonalantibody LE61. (These markers will be discussedfurther below.) As is shown in Table 1, CMRL1066with 20% FCS, insulin (0-2unitsml~'), cholera toxin( 1 0 ~ ' ° M ) and hydrocortisone ( l ^ g m P 1 ) was the bestculture medium. Primaria tissue culture flasks orcollagen-coated glass or plastic were the most efficientsubstrates (see Table 2).

Fibroblast contamination of epithelial cell culturesfrequently occurs in both primary cultures and insubsequent passages. The most successful method ofeliminating fibroblasts was by physical removal with arubber policeman.

Fig. 1. Light microscopy of pancreatic epithelial cell types in culture. A. Tightly packed colonies of small epithelial cells.X250. B. Loosely packed larger epithelial cells. X250.

Pancreatic epithelial cells 697

Passaging of cells

Passaging of epithelial cell colonies was achieved bytreating the latter with dispase, a neutral protease(Boehringer, Mannheim) at 2 units ml"1 in CMRL

1066 medium for 15-40min at 37°C. The colonieswere replated in 100% FCS for 24 h then changed to50% CMRL 1066 plus 50% FCS for 24h, and finallychanged to CMRL 1066 with supplements and 20%

Fig. 2. Light microscopy of pancreatic epithelial cell types in culture. A,B. Characteristic structures formed in primaryand subsequent passages of pancreatic epithelial cells. A, X125; B, X250.

698 A. Harris and L. Coleman

serum (the usual culture medium). Cell colonies sur-vive passaging if they remain intact (or physicallydivided into smaller colonies), but not if they aredisrupted into clumps of small numbers of cells. Thetightly packed smaller cell type appears to be selectedfor by the latter method of passaging. Cells survive upto five successive passages before reaching terminalcrisis, and they appear to retain constant morphologicalfeatures until this time. Cultures can generally bemaintained for up to 20 weeks in vitro.

Ultrastructural characteristics of cellsAn electron-microscopic analysis of both epithelial celltypes in the cultures showed them to have character-istic features of simple epithelial cells. They haveextensive cytokeratin intermediate filaments (tonofila-ments) and clear desmosomal plaques (Fig. 3A). Thecells have surface microvilli (Fig. 3B). Furthermore,the small, tightly packed epithelial cells have abundantlarge mitochondria (Fig. 4A,B) and are packed withsecretory vacuoles, probably containing mucins

(Fig. 4C) (see Histochemistry section, below).Together these features suggest that the tightly packedcells being cultured are likely to be large interlobularduct cells or a developmental precursor of them. It ispossible that the larger epithelial cells, seen in primarycultures but which passage much less efficiently thanthe small cells, may be cells derived from the mainpancreatic duct or another part of the duct system.

Immunocytochemistry

Immunocytochemical data are summarized in Table 3.Cytokeratins. Both small and large epithelial cell

types express the mixture of cytokeratin intermediatefilament proteins characteristic of simple epithelia;these are detected by the monoclonal antibody LE61(Lane, 1982) (Fig. 5A). Both cell types also expresscytokeratin 8 (Fig. 5B) and cytokeratin 19 (Fig. 5C).Expression of LE61, cytokeratin 8 and 19 is stable from2-5 weeks after the cultures are established to at least 7weeks (these being the earliest and latest dates at whichcytokeratin expression has been analysed). In adult

Fig. 3. Electron microscopy of pancreatic epithelial cells. A. Cytokeratin intermediate filaments and desmosomal plaques.X64000. B. Surface microvilli on epithelial cells. X22 000.

Pancreatic epithelial cells 699

pancreas, exocrine acini express predominantly cyto-keratins 8 and 18, while microdissected pancreaticducts also contain cytokeratins 7 and 19 (Moll et al.1983). Control cultures, stained only with the second

antibody used in the immunocytochemical procedures,showed no significant fluorescence (Fig. 5D). Thenegative controls for cytokeratin immunocyto-chemistry were human skin fibroblasts and the positive

Fig. 4. Electron microscopy of pancreatic epithelial cells. A. Interdigitated cells showing tight junctions and abundantmitochondria. X6400. B. Elongated mitochondrion: these are frequent. X 10200. C. Secretory vacuoles within the cells.X6400.

700 A. Harris and L. Colentan

s

B

Table 3. Summary of immunocytochemical data

LE61Cytokeratin

8Cytokeratin

19Carbonicanhydrase

Ca2 andHMFG2

Loosely packed epithelial cellsTightly packed epithelial cellsMCF7, control epithelial cellsControl fibroblasts

controls were MCF7 cells. MCF7 is a mammarycarcinoma cell line expressing cytokeratins 8, 18 and19.

Carbonic anhydrase. The only parts of the adultpancreas that have been shown in tissue sections toexpress carbonic anhydrase are the pancreatic ducts(Kumpulainen & Jalovaara, 1981; Spicer et al. 1982).Using polyclonal antisera against carbonic anhydrase,the epithelial cells described here were seen to expresscarbonic anhydrase in first-passage cultures (until atleast 9 weeks in culture, Fig. 6B). Cultured skinfibroblasts in which no reaction was detectable with thesame antisera were negative controls (data not shown).

Ca2 and HMFG2. The monoclonal antibodies Ca2and HMFG2, which react with the same mucousglycoprotein, known to be expressed on certain ductalepithelia (Harris, 1987), react with the epithelial cellsdescribed here (data not shown). The negative controlswere human skin fibroblasts and the positive controlsHeLa cells, an epithelial cell line derived from a humanuterine carcinoma.

Histochemistry

Alcian Blue-periodic acid-Schiff (PAS) staining of theepithelial cell cultures revealed that the cells were richin neutral mucins (Fig. 6A). Under the conditionsused, neutral mucins stain pink. The tall columnarcells of interlobular ducts contain neutral mucins(Roberts & Burns, 1972).

Discussion

A tissue culture system for epithelial cells derived fromnormal mid-trimester foetal pancreases has been estab-lished. These cells could either be ductal epithelial cellsor acinar cells, which share common developmental

Fig. 5. Immunocytochemistry of epithelial cells.A. Fluorescein-conjugated LE61. X175. B. Fluorescein-conjugated anti-cytokeratin 8. X17S. C. Fluorescein-conjugated anti-cytokeratin 19. X17S. D. Backgroundfluorescence with second antibody only.Fig. 6. A. Alcian blue-periodic acid-Schiff staining ofpancreatic epithelial cells to show intracellular mucins.X875. Nuclei are counterstained with Giemsa.B. Pancreatic epithelial cells reacted with peroxidase-conjugated antisera against carbonic anhydrase II. X175.

pathways and have many structural and functionalcharacteristics in common. On the basis of morphologi-cal, histochemical and immunocytochemical character-istics it is likely that these cells are ductal in origin. It ispossible that they do not correspond directly to one ofthe differentiated duct cell types seen in adult humanpancreas, but rather are developmental precursors ofthese cells. However, the epithelial cells reported heredo show many characteristics of differentiated inter-lobular duct cells (Kodama, 1983). They containabundant mitochondria, many secretory vacuoles inthe apical region, some filled with mucins, and havemicrovilli on their luminal surface (Figs 4A-C, 3B).These cells also contain a dense network of intermedi-ate filaments terminating in desmosomal plaques(Fig. 3A). The immunocytochemical data with cyto-keratins 8 and 19 is further evidence that the cellscultured here are ductal in origin (Fig. 5). In prelimi-nary experiments with the same markers on frozensections of 18-week foetal pancreases, the only cellsfound to have the same characteristics as the cells wehave isolated were those lining the main pancreaticduct and cells in islands that are probably developingsmaller ducts. The fact that the cells being culturedhere express carbonic anhydrase and the antigendetected by the monoclonal antibodies Ca2 andHMFG2 provides strong support for their ductalorigin.

The cells we have isolated do not show the character-istic zymogen granules found in acinar cells, though ofcourse a developmental precursor of acinar cells mightnot yet contain these differentiated structures. Fur-thermore, the nuclei of the cells cultured here are notlocated in the basal region of the cytoplasm as iscommonly seen in acinar cells.

Finally, preliminary electrophysiological analyses ofour cell cultures by direct insertion of microelectrodesinto the cells, suggests that at least some of them show asmall depolarization in response to stimulation bysecretin and dibutyryl cyclic AMP (Dr Barry Argent,personal communication). In the rat and mouse pan-creas, ductal cells show electrophysiological responsesto secretin and dibutyryl cyclic AMP, while acinar cellsdo not.

In conclusion, we believe that we are culturinghuman pancreatic duct cells, one of the cell types inwhich one would expect the basic defect in cystic

Pancreatic epithelial cells 701

fibrosis to be expressed. Clearly, until the CF geneproduct is isolated we cannot be certain that the cellswe are culturing do express the CF gene. However,using tissue from an organ known to be affected in CFto set up cell cultures, as opposed to using skin biopsiesor other tissues that are apparently unaffected in thisdisease, greatly increases the chance that we are indeedstudying cells expressing the basic defect in CF.Analysis by Northern blotting of mRNA synthesized inthese cells will clarify this. Further, it seems likely thatcells within the mid-trimester CF pancreas are alreadyexpressing this defect. In a recent extensive study offoetuses diagnosed as having CF on the basis ofmicrovillar enzyme and alkaline phosphatase assays(Boue' et al. 1986), it was found that 20-week-oldfoetuses already showed small amounts of PAS-positivematerial within some of the pancreatic acini. Theearliest lesions seen at post-mortem examinations of CFpancreases post-natally are deposits of eosinophilicmaterial in the acini and intralobular ducts (Oppen-heimer & Esterly, 1975). It is thought that the inspiss-ated secretions in the ducts are the primary cause ofacinar distension and fibrosis, that is to say the ductcells are defective rather than the acini.

We can now culture sufficient numbers of theseputative ductal epithelial cells from one organ to startscreening for biochemical and molecular biologicaldifferences between CF and normal cells.

This work was supported by the Cystic Fibrosis ResearchTrust. The project was approved by Guy's Hospital EthicalCommittee. We thank Dr Mary Seller, Dr Jean Keeling(John Radcliffe Hospital, Oxford) and Professor MatteoAdinolfi for their help; Derrick Lovell and Jocelyn Davies forelectron microscopy; Dr Bruce Ward for help with immuno-cytochemistry; Drs Birgitte Lane, Norman Gregson, JoyBurchell and Professor H. Harris for generous gifts ofantisera; Dr Philip Thorpe for ricin-conjugated anti-thy-1;and Adrienne Knight for help with preparation of themanuscript.

References

ADDA, G., HANNOUN, L. & LOYGUE, J. (1984).

Development of the human pancreas: variations andpathology. A tentative classification. Anat. Clin. 5,275-283.

BOUE, A., MOLLER, F., NEZELOF, C , OURY, J. F.,

DUCHATEL, F., DUMEZ, Y., AUBRY, M. C. & Boufi, J.(1986). Prenatal diagnosis in 200 pregnancies with a 1-in-4 risk of cystic fibrosis. Hum. Genet. 74, 288-297.

COOK, H. C. (1982). Theory and Practice ofHistologicalTechniques (ed. J. D. Bancroft & A. Stevens), 2nd edn.New York: Churchill Livingstone.

COTMORE, S. F., CROWHURST, S. A. & WATERFIELD, M. D.

(1981). Purification of thy-1 related glycoproteins fromhuman brain and fibroblasts: comparison between these

molecules and murine glycoproteins thy-1.1 and thy-1.2antigens. Eur.J. Immun. 11, 597-603.

DAAR, A. S. & FABRE, J. W. (1983). The membraneantigens of human colorectal cancer cells; demonstrationwith monoclonal antibodies of heterogeneity within andbetween tumours and of anomalous expression of HLA-DR. Eur.J. Cancer din. Oncol. 19, 209-220.

FRIZZELL, R. A., RECHKEMMER, G. & SHOEMAKER, R. L.

(1986). Altered regulation of airway epithelial cellchloride channels in cystic fibrosis. Science 233,558-560.

HAMMOND, S. L., HAM, R. G. & STAMPFER, M. R. (1984).

Serum-free growth of human mammary epithelial cells:rapid growth in defined medium and extended serialpassage with pituitary extract. Pmc. natn. Acad. Sci.U.SA. 81, 5435-5439.

HARRIS, H. (1987). The Ca antigen: structure, functionand clinical application. In Tumour Markers in ClinicalPractice (ed. A. S. Daar), pp. 115-128. Oxford:Blackwell Scientific Publications.

HIRATA, K., OKU, T. & FREEMAN, A. E. (1982). Duct,exocrine, and endocrine components of cultured fetalmouse pancreas. In Vitro 18, 789-799.

JONES, R. T., HUDSON, E. A. & RESAU, J. H. (1981). A

review of in vitro and in vivo culture techniques for thestudy of pancreatic carcinogenesis. Cancer 47,1490-1496.

KNOWLES, M. R., STUTTS, M. J., SPOCK, A., FISCHER, N.,

GATZY, J. T. & BOUCHER, R. C. (1983). Abnormal ionpermeation through cystic fibrosis respirator}'epithelium. Science 221, 1067-1070.

KNOWLES, M. R., STUTTS, M. J., YANKASKAS, J. R.,

GATZY, J. T. & BOUCHER, R. C. (1986). Abnormalrespirator)' epithelial ion transport in cystic fibrosis. Clin.Chest Med. 7, 285-297.

KODAMA, T. (1983). A light and electron microscope studyof the pancreatic ductal system. Ada path. jpn. 33,297-321.

KUMPULAINEN, T. & JALOVAARA, P. (1981).

Immunohistochemical localization of carbonic anhydraseisoenzymes in the human pancreas. Gastroentemlogy 80,796-799.

LAITIO, M., LEV, R. & ORLIC, D. (1974). The developinghuman fetal pancreas: an ultrastructural andhistochemical study with special reference to exocnnecells. J. Anat. 117, 619-634.

LANE, E. (1982). Monoclonal antibodies provide specificintramolecular markers for the study of epithelialtonofilament organization. .7. Cell Biol. 92, 665-673.

MOLL, R., KREPLER, R. & FRANKE, W. W. (1983).

Complex cytokeratin polypeptide patterns observed incertain human carcinomas. Differentiation 23, 256-269.

OPPENHEIMER, E. H. & ESTERLY, J. R. (1975). Pathologyof cystic fibrosis: review of the literature and comparisonwith one hundred and forty-six autopsied cases. Perspec.Paediatr. Path. 2, 241-278.

PARASKEVA, C , BUCKLE, B. G. & THORPE, P. E. (1985).

Selective killing of contaminating human fibroblasts inepithelial cultures derived from colorectal tumours usingan anti thy-1 antibody—ricin conjugate. Br.J. Cancer 51,131-134.

702 A. Harris and L. Coleman

QuiNTON, P. M. (1983). Chloride impermeability in cysticfibrosis. Nature, land. 301, 421-422.

RESAU, J. H., HUDSON, E. A. & JONES, R. T. (1983).

Organ explant culture of adult Syrian golden hamsterpancreas. In Vitm 19, 315-325.

ROBERTS, P. & BURNS, J. (1972). A histochemical study ofmucins in normal and neoplastic human pancreatictissue. J. Path. 107, 87-94.

SATO, T., MAMORU, S., HUDSON, E. A. & JONES, T.

(1983). Characterization of bovine pancreatic ductal cellsisolated by a perfusion-digestion technique. In Vitm 19,651-660.

SATO, K. & SATO, F. (1984). Defective beta-adrenergicresponse of cystic fibrosis sweat glands /;/ vivo and ///vitro. J. din. Invest. 73, 1763-1771.

SPICER, S. S., SENS, M. A. & TASHIAN, R. E. (1982).

Immunocytochemical demonstration of carbonic

anhydrase in human epithelial cells. J . Histochein.Cylochem. 30, 864-873.

STONER, G. D., HARRIS, C. C , BOSTWICK, D. G., JONES,

R. T., TRUMP, B. F., KINGSBURY, E. W., FINEMAN, E.

& NEWKIRK, C. (1978). Isolation and characterization ofepithelial cells from bovine pancreatic duct. /// Vitro 14,581-590.

WELSH, M. J. & LIEDTKE, C. M. (1986). Chloride andpotassium channels in cystic fibrosis airway epithelia.Nature, Land. 322, 467-470.

WIDDICOMBE, J. H., WELSH, M. J. & FINKBEINER, W. E.

(1985). Cystic fibrosis decreases the apical membranechloride permeability of nionolayers cultured from cellsof tracheal epithelium. Proc. natn. Acad. Sci. U.SA. 82,6167-6171.

(Received 5 February 1987-Accepted 23 March 1987)

Pancreatic epithelial cells 703