Develop. Growth Differ.

(2003)

45

, 39–50

Development and retroviral transduction of porcine neonatal pancreatic islet cells in monolayer culture

Krystyna Tatarkiewicz, Maria D. López-Avalos,* Kun-Ho Yoon, Nitin Trivedi, Robert R. Quickel, Susan Bonner-Weir and Gordon C. Weir

Research Division, Joslin Diabetes Center, Harvard Medical School, Boston, MA, USA.

To learn more about the potential of neonatal porcine pancreatic duct and islet cells for xenotransplantation, thedevelopment of these cells when cultured as monolayers was evaluated. Immunostaining for islet hormonesand cytokeratin-7 revealed that day eight monolayers consisted of approximately 70% duct cells and less than10% beta cells. Using Ki-67 immunostaining as a proliferation marker, the fraction of beta cells in the cell cyclewas shown to decrease from 20% at day three to 10% at day eight, and for duct cells from 36 to 19%. Insulinsecretion increased 2.4-fold upon glucose stimulation, and 38-fold when 10 m

M

theophylline was added,showing the responsiveness of the neonatal beta cells. Reaggregated monolayers consisted mostly of ductcells, but 4 weeks after transplantation, grafts contained predominantly endocrine cells, with duct cells beingalmost absent, suggesting

in vivo

differentiation of duct cells to endocrine cells. Monolayer susceptibility toretroviral transduction was also investigated using a Moloney Murine Leukemia Virus-based vector. Approxi-mately 60% of duct cells but less than 5% of beta cells expressed the transgene, indicating that precursor ductcells are better targets for transgene expression. These results show that porcine neonatal pancreatic cells canbe cultured as monolayers in preparation for transplantation. Furthermore, in such a culture setting, precursorduct cells have a high rate of proliferation and are more efficiently transduced with a retrovirus-based reportergene than are beta cells.

Key words:

beta cells, duct cells, monolayer culture, pancreatic development, transplantation.

Introduction

Porcine pancreatic cells have been proposed as asource of insulin-producing cells that might be usedas a xenograft for the treatment of human diabetes.Fetal and neonatal tissue has the advantage of a highgrowth capacity compared to adult pancreas. It hasbeen shown that islet-like cell clusters derived fromdispersed immature human or porcine pancreas cannormalize blood glucose levels after transplantationinto diabetic nude mice (Sandler

et al

. 1985; Oton-koski

et al

. 1993; Korbutt

et al

. 1996; Beattie

et al

.1997b; Yoon

et al

. 1999). In our earlier study, porcineneonatal pancreatic cell clusters (NPCC) cultured insuspension consisted of approximately 40% ductcells, and more than 90% of these developed intoinsulin-producing cells after transplantation into nude

mice (Yoon

et al

. 1999). In spite of the promisingadvantages of neonatal tissue, protection of the graftfrom immune rejection in a xenogeneic environmentremains a major obstacle.

In contrast to cell clusters in suspension, monolayercultures may provide a better opportunity to manipu-late cells before transplantation, through approachessuch as growth factor addition or genetic transduc-tion. The use of immunoprotective genes to preventrejection is of major interest. In general, cells can bemore efficiently expanded when grown as monolayersrather than as floating clusters. Monolayer cultures ofporcine pancreatic cells can be used for genetic,biochemical and physiological studies that investigatepancreas development.

The main goal of the present study was to gain abetter knowledge of the potential use in xeno-transplantation of porcine neonatal pancreatic cellscultured as monolayers. The different cell types in themonolayer were followed for 11 days and werecharacterized by immunocytochemistry. Functionalityof the beta cells was assessed by glucose stimulation.To investigate the possible applications of cellsexpanded in monolayers, we carried out retroviraltransduction of the neonatal pancreatic cells with a

*Author to whom all correspondence should be addressed.

Present address:

Departamento Biología Celular, Genética yFisiología, Facultad de Ciencias, Universidad de Málaga,Campus de Teatinos, Málaga 29071, Spain.

Email: lopezavalos@uma.esReceived 10 July 2002; revised 30 July 2002; accepted

25 September 2002.

40 K. Tatarkiewicz

et al.

�

-galactosidase (

�

-gal) reporter gene. Furthermore,after reaggregation and transplantation of the mono-layer cells, we followed development of the graft

in vivo

.

Materials and Methods

Media

Medium for NPCC (Ham’s F10+).

Ham’s F10 (GibcoLife Technologies, Grand Island, NY, USA) supple-mented with 10 m

M

glucose, 50

µ

M

3-isobutyl-1-methylxanthine (IBMX), 2 m

M

L

-glutamine, 10 m

M

nicotinamide, 25 mg/L CaCl

2

·2H

2

O and 0.5% bovineserum albumin (BSA; RIA grade, fraction V) fromSigma Chemical, St Louis, MO, USA; and 100 U/mLpenicillin and 100

µ

g/mL streptomycin (Pen/Strep)from Mediatech, Herndon, VA, USA.

Modified Hanks Balanced Salt Solution (HBSS+).

HBSS supplemented with 11 m

M

HEPES, 10 mg/mLphenol red and 0.25% BSA fraction V (all from Sigma).

Dispersion medium (DM).

Ca

+

- and Mg

+

-free HBSSsupplemented with 1 m

M

ethylenediaminetetraaceticacid (EDTA), 10 m

M

HEPES and 0.5% BSA fraction V(all from Sigma).

Culture medium (CM).

RPMI-1640 with 11 m

M

glu-cose, supplemented with 10% fetal calf serum (FCS),100 U/mL penicillin and 100

µ

g/mL streptomycin fromMediatech; and 10 m

M

nicotinamide and 10 m

M

HEPES from Sigma.

Medium for static insulin secretion studies (ISM).

RPMI-1640 without glucose but with 0.5% BSA, 10 m

M

HEPES, 100 U/mL penicillin, 100

µ

g/mL streptomycinand glucose added as required.

Animals

One- to 3-day-old Yorkshire pigs (Parsons Farm,Hadley, MA, USA) were used as donors of pancreatictissue. Six- to 8-week-old male Swiss Webster nudemice (Taconic, Germantown, NY, USA) were used astransplant recipients of reaggregated porcine pan-creatic cells. All animal procedures were approved bythe Animal Care Committee of the Joslin DiabetesCenter.

Preparation of porcine neonatal pancreatic cell clusters

The procedure for preparation of NPCC has beendescribed in detail previously (Korbutt

et al

. 1996;Yoon

et al

. 1999). Briefly, each pancreas was minced

into 1–2 mm pieces and digested for 15 min with4 mg/mL collagenase P (Boehringer-Mannheim, Indi-anapolis, IN, USA) and 2.5 or 5 mg/mL type V colla-genase (Sigma) at 37

�

C in a shaking water bath. After8 min, and at the end of the digestion, the suspensionwas hand-shaken for an additional 1 min, filteredthrough a 500

µ

m sieve (Newark Wire Cloth Co.,Newark, NJ, USA) and washed three times withHBSS+ (190

g

; 2 min). The final suspension from eachpancreas was cultured in four 150

�

15 mm bacterio-logical plates (Becton-Dickinson, Franklin Lakes, NJ,USA) in 35 mL of F10+ at 37

�

C in an atmosphere of5% CO

2

.

Monolayer culture

On the day after isolation, NPCC were dispersed intosingle cells using a modification of a previouslydescribed method (Pipeleers

et al

. 1985). NPCC fromone pancreas were collected and washed two timeswith DM (190

g

; 2 min), suspended in 40 mL in a75 cm

2

tissue culture flask and gently shaken(35 r.p.m.) in a water bath for 7 min at 37

�

C. Trypsin(Sigma) and DNase (Boehringer Mannheim) werethen added to obtain final concentrations of 25 and4

µ

g/mL, respectively, followed by the next incubationin a shaking water bath for 10 min. Cell clusters werebroken apart three times by gentle aspiration with a10 mL plastic pipette (Falcon; Becton Dickinson)resulting in a suspension with approximately half ofthe cells being single. To separate single cells fromremaining clumps, the suspension was centrifuged,and as soon as the centrifuge reached 190

g

, it wasstopped. The supernatant with single cells was collec-ted and spun down at 290

g

for 5 min, while the pelletwas again subjected to dissociation for 10 min asdescribed above. Cells collected from the two stepsof dissociation were filtered through a 100

µ

m sieve,counted, and incubated for 3 h at a density of approxi-mately 1

�

10

6

cells/mL in 15 cm cell culture dishes(Nalge Nunc International, Rochester, NY, USA) in CMat 37

�

C in a humidified atmosphere of 5% CO

2

. Fibro-blasts have a tendency to attach more rapidly to thebottom of culture dishes (Rabinovitch

et al

. 1980),therefore such preincubation allowed for selectiveelimination of many fibroblasts. The primary cellsuspension was transferred into new dishes at adensity of approximately 0.4

�

10

6

cells/mL and wasmaintained for the next 11 days with media changeson days three, six and eight after dispersion.

Insulin secretion

Eight days after dispersion, monolayers of neonatalporcine pancreatic cells were subjected to static

Porcine pancreatic cell monolayers 41

incubation to assess glucose-induced insulin secre-tion. Monolayers in 25 cm

2

tissue culture flasks werewashed twice with ISM and incubated for 2 h with5.6 m

M

glucose, 16.7 m

M

glucose, or 16.7 m

M

glucoseand 10 m

M

theophylline. Supernatants from eachincubation were collected, spun down (420

g

; 3 min)to eliminate any floating cells and frozen beforemeasurement of insulin with a radioimmunoassay(RIA) kit for porcine insulin (Linco, St Charles, MO,USA).

Transduction of

�

-gal reporter gene by retroviral vector

The production of Moloney Murine Leukemia Virus(MLV)-based vector expressing the

�

-gal reportergene was described by Ory

et al

. (1996). Briefly,concentrated vector stocks were prepared by ultra-centrifugation of conditioned medium harvested3–8 days after withdrawal of tetracycline from thestable producer cell line, GPR2nLacZ2. Viral titerswere determined by infection of NIH3T3 cells. Fortransduction, 3 days after dispersion, monolayers ofporcine neonatal pancreatic cells in 3 cm disheswere incubated for 6 h with 1 mL of serial dilutions ofvector stock supplemented with 8

µ

g/mL polybrene(Sigma). Mock infection was performed using theculture medium supplemented with the same con-centration of polybrene. After infection and mediachange, the cells were cultured for 48 h and

�

-galexpression was detected by immunostaining.

Reaggregation of monolayers and transplantation

Eight days after dispersion, monolayers were washedtwo times with phosphate-buffered saline (PBS)without Ca

2+

and Mg

2+

ions, and were lifted offculture dishes by a 20 min incubation at 37

�

C withnon-enzymatic dissociation solution (C-5789; Sigma).

Cells were collected, washed with CM and incubatedin 2 mL CM for 1 h in two 50 mL conical tubes in atissue culture incubator to enhance aggregation.Aggregates from one pancreas were then transferredto four 10 cm bacteriological Petri dishes and left inCM for overnight culture.

To follow their development

in vivo

, cell clusterswere transplanted under the kidney capsule of non-diabetic Swiss Webster nude mice as describedpreviously by Yoon

et al

. (1999). Briefly, aggregateswere sedimented in 200

µ

L plastic pipette tips withattached PE-50 polyethylene tubing (Becton Dickin-son, Sparks, MD, USA) and centrifuged to form atransplantable pellet inside the tubing. The prepar-ations consisted of aggregates and some single cells,and therefore could not be counted. To estimate thevolume of tissue for transplantation, we measured thelength of the pellet and estimated the volume in termsof 150

µ

m islet equivalents (IE). From our previousexperiments with transplantation of NPCC, we esti-mated that a 10 mm pellet corresponds to 1600 IE(K. Tatarkiewicz, unpubl. data, 1999). Pellets equiva-lent to approximately 3200 IE were implanted underthe kidney capsule of each mouse. Four weeks aftertransplantation, grafts with adjacent kidney tissuewere removed for histology.

Insulin and DNA content

Cells from monolayers of porcine neonatal pancreaticcells cultured for 3, 8 and 11 days in 35 mm tissueculture dishes were removed as described above.They were sonicated in acid ethanol or high saltbuffer for measurement of insulin and DNA content,respectively, as described in Davalli

et al

. (1995).Insulin was measured by RIA. DNA content wasmeasured fluorometrically using the method of DyNaQuant 200 (Hoefer Pharmacia Biotech, San Francisco,CA, USA).

Table 1.

Primary antibodies used for immunocytochemistry

Immunogen Marker Host Dilution Source

Insulin Beta cells Guinea pig 1:200 Linco Research, St Charles, MO, USAGlucagon Alpha cells Rabbit 1:3000 Donated by Dr M. Appel, University of

Massachusetts Medical School, Worcester, MA, USA

Somatostatin Delta cells Rabbit 1:300 Made in our laboratoryPancreatic polypeptide PP cells Rabbit 1:3000 Donated by Dr R. E. Chance, Eli Lilly,

Indianapolis, IN, USACytokeratin (CK)-7 Duct cells Mouse 1:200 DAKO, Glostrup, DenmarkPancytokeratin Epithelial cells Rabbit 1:200 DAKO, Glostrup, DenmarkKi-67 Proliferation Mouse 1:200 Calbiochem, La Jolla, CA, USAVimentin Fibroblasts Mouse 1:200 DAKO, Glostrup, Denmark

�

-galactosidase Reporter gene Rabbit 1:200 Clontech Laboratories, Palo Alto, CA, USA

42 K. Tatarkiewicz

et al.

Immunocytochemistry

Immunostaining of cells in monolayers and grafts wasdone to identify beta cells, non-beta endocrine isletcells and duct cells. Additionally, monolayers wereimmunostained for vimentin as a marker of fibroblasts.Ki-67 was used as a proliferation marker. The

�

-galreporter gene used for retrovirus transduction wasdetected by immunostaining.

Cells in dishes were fixed with 4% paraformalde-hyde in Sorensen’s buffer for 15–30 min and werepermeabilized for 10 min using a 0.25% solution ofTriton X-100 in PBS. Reaggregated monolayers andgrafts were fixed in 4% paraformaldehyde for 2 h. Cellaggregates were suspended in prewarmed 2% agar(Bacto-Agar, Difco Laboratories, Detroit, MI, USA) andwere spun down at 4

�

C to form compact pellets.These pellets and grafts were embedded in paraffinfor sectioning. PBS (pH 7.4) was used to wash allpreparations and to dilute the antibodies. Non-specific binding was blocked by preincubation in a1.5% solution of non-immune serum from the speciesin which the secondary antibody was raised. After thisblocking step, primary antibodies were applied andincubated overnight at 4

�

C. Table 1 presents the

primary antibodies used in this study, their dilutionsand sources.

Secondary antibodies conjugated with Texas Red,fluorescein isothiocyanate (FITC) or 7-amino-4-methylcoumarin-3-acetic acid (AMCA; JacksonImmunoresearch Laboratory, West Grove, PA, USA)were diluted 1:100 and applied for 1 h at roomtemperature. After a final wash, dishes and sectionswere mounted with a solution of 90% glycerolcontaining 0.1

M

antifade agent DABCO (triethyl-enediamine; Sigma). The DNA fluorescent dye 4

�

,6-diamidino-2-phenylindole (DAPI; 1

µ

g/mL) wasincluded in this mounting solution for countingcell nuclei. For double or triple staining, primary aswell as secondary antibodies were mixed for co-incubation.

Cytokeratin (CK)-7 and Ki-67 immunostaining wereamplified with the biotin–streptavidin method (VectorLaboratories, Burlingame, CA, USA). After overnightincubation with the primary antibody, biotinylated anti-mouse IgG followed by streptavidin conjugated withFITC or Texas Red was applied, each for 1 h at roomtemperature. If double immunostaining with either ofthese two antibodies was required, a two-step pro-tocol was used, staining first for any of the other

Fig. 1. Several cell phenotypeswere recognized in the monolayerculture of porcine pancreatic cellsat day eight. (A) Duct epithelialcells were the most common,identified by their characteristicintermediate filaments immuno-stained for cytokeratin (CK-7,green). Small round beta cells(insulin, blue) or alpha cells(glucagon, red) were often foundanchored to nest-like structuresmade by underlying duct cells.(B) Subepithelial fibroblasts couldbe distinguished by their intensevimentin staining (vimentin, red)and their characteristic morph-ology. Fibroblasts and duct cells(CK-7, green) usually grew asseparate but adjacent homog-eneous patches: the dual stainingfound at ruffled borders of ductcells are characteristic of rapidlygrowing cultured epithelial cells.(C) Beta cells (insulin, blue) andduct cells (CK-7, red) wereproliferating during the culture, asrevealed by nuclei staining posi-tively for the proliferation markerKi-67 (green). Bars, 25 µm.

Porcine pancreatic cell monolayers 43

markers and later for CK-7 or Ki-67 using the amplifi-cation protocol.

For CK-7 staining on paraffin sections, antigenretrieval pretreatment was done by microwaving theslides three times for 4 min in 0.01

M

sodium citratepH 6.0, using fresh cold solution each time (Leong &Milios 1993).

The immunofluorescently stained images wereacquired on a Zeiss LSM 410 microscope (Carl Zeiss,Thornwood, NY, USA) with excitations at 488 nm forFITC, 568 nm for Texas Red and 364 nm for DAPI andAMCA. For quantification purposes, systematicallytaken images (approximately 20 per dish) wereanalyzed and 1000–2000 cells per dish were counted.

Some cellular aggregates were fixed in glutaralde-hyde, embedded in plastic resin, sectioned at a thick-ness of 1 µm, and stained with toluidine blue.

Statistical analysis

Results are expressed as mean ± SEM. Statisticalsignificance was assessed with the unpaired Studentst-test or multiple group analysis of variance ANOVA inconjunction with the Fisher test.

Results

Cell phenotypes in porcine pancreatic cell monolayers

Immunocytochemical analysis revealed the presenceof several cell phenotypes in the monolayer cultures

of porcine pancreatic cells. Duct epithelial cells wererecognized by their characteristic intermediate cyto-keratin filaments (Yoon et al. 1999) that were stainedwith an antibody against CK-7, which has been shownto be a specific duct cell marker for pig pancreas(Bouwens et al. 1994). After attachment to the bottomof the dish, duct cells were also easy to distinguish bytheir large, very flat, rounded shape with rufflededges. They usually stayed in clusters rather than assingle cells (Fig. 1A). Beta and alpha cells weresmaller and had an oval shape. Under the phase-contrast microscope, they displayed a darker cyto-plasm than other cells, possibly due to insulin orglucagon granules. The endocrine cells were fre-quently found in limited groups, either attached tothe dish surface or very often anchored in nest-likestructures made by duct cells (Fig. 1A).

Fibroblasts, identified by immunostaining vimentinfilaments, were also present in the monolayers. Therewere two phenotypes: large, flat and spindle-like, ora smaller, comma-shaped subepithelial phenotype.Vimentin staining was also found in the rufflededges of duct cells, a phenomenon that occursin some cell types when cultured in monolayers.Usually, groups of fibroblasts and duct cells weregrowing as separate, but adjacent, homogeneousgroups competing for the available surface area(Fig. 1B).

The mean yield after the dispersion of one neonatalporcine pancreas was 1.4 ± 0.2 � 108 single cells(n = 9). Abundant cellular debris was observed duringthe first 3 days of culture, which can be explainedmainly by the death of exocrine cells, a phenomenonreported in our previous studies on NPCC cultured insuspension (Yoon et al. 1999). On day one after dis-persion, most of the cells were still floating, and weobserved only some duct cells adhering to dishes;therefore, cultures were maintained undisturbed for3 days, at the end of which most cells were attachedand almost no live cells were found floating in themedia. At this point, the monolayers of adherent cellsconsisted of patches of flattened duct cells, single orgrouped endocrine cells and limited numbers offibroblasts. By day eight, in most dishes monolayersreached confluence, and at day 11, cells started tobuild up into double layers. Over the time of theculture, there was a decline in the fraction of allepithelial cells (pancytokeratin positive, a marker ofepithelial cells), including duct and endocrine cells,with a concomitant increase in the percentage offibroblasts (vimentin positive), which by day 11reached approximately half of the total population ofcells in the monolayer (Fig. 2). Therefore, day eightmonolayers were chosen for further studies, because

Fig. 2. Changes in the number of epithelial cells (pancytoker-atin positive; �) and fibroblasts (vimentin positive; �) in themonolayer culture of porcine pancreatic cells. These two cellpopulations are expressed as the percentage of total cellnumber, which was assessed by nuclei staining with 4�,6-diamidino-2-phenylindole (DAPI) at 3, 8 and 11 days of culture(n = 4 independent experiments). *P < 0.01 vs 3 days.

44 K. Tatarkiewicz et al.

at this time point a large proportion of epithelial cellswas still present in the culture.

Insulin and DNA content

The growth capacity of the neonatal pancreatic cellsin monolayer was assessed by measuring the DNAcontent (Fig. 3a). An initial notable decrease of DNAcontent can be mainly attributed to the death ofexocrine cells, although some loss of beta and otherendocrine cells can not be excluded. After thisdecline, a rebound of DNA content was observed,denoting cell proliferation, and by day 11, DNAcontent was essentially the same as at the com-mencement of culture. Insulin content expressed as afunction of the DNA content (Ins : DNA) served as anindicator of the enrichment of the culture in beta cells.The Ins : DNA ratio remained unchanged during thefirst week and significantly fell at day 11 (P < 0.05 vsday eight; Fig. 3b), at which point the fibroblastnumber had increased to approximately 50%. Inter-estingly, the number of beta and alpha cells roseapproximately threefold from days three to eight andremained unchanged until day 11 (Fig. 4). This finding

is consistent with the observation that the relativecontribution of beta cells to the total number of cellswas not changed (P = 0.12), although there was atendency to decrease from 10.9 ± 2.6% at day threeto 6.4 ± 0.7% at day eight. Additionally, during thisperiod of time the percentage of alpha cellsdecreased from 9.1 ± 1.3% to 5.1 ± 0.9% (P < 0.05).Protodifferentiated cells were detected by their co-staining for insulin and glucagon, but their contributionto the total cell number was less than 3% (Fig. 4).

Proliferation

Ki-67 is a proliferation-associated antigen, expressedin all active stages of the cell cycle. It can thus beused as a proliferation marker (Schluter et al. 1993;Starborg et al. 1996). Using Ki-67 immunostaining(Fig. 1C), we found that twice as many duct cells werein the replicative cell cycle as beta cells (Fig. 5). Thenumber of proliferating cells decreased for both celltypes as they reached confluence, but was stillsubstantial.

Insulin secretion

Neonatal pancreatic cell monolayers cultured for8 days were used to study the responsiveness of theneonatal beta cells to glucose stimulation. The amountof insulin secreted by a monolayer in a 25 cm2 flaskduring 2 h incubation in ISM with basal glucose(5.6 mM) was 4.9 ± 0.7 ng. In response to high

Fig. 3. DNA and insulin content of dispersed neonatal pancre-atic cell clusters cultured for 3, 8 and 11 days. DNA content (a)was expressed as a function of the surface area of the culturedish. Insulin content (b) was presented as the function of DNAcontent from the same dish (n = 4 independent experiments).*P < 0.01 vs 1 day and vs 11 days; **P < 0.05 vs 8 days.

Fig. 4. Number of beta (�) and alpha (�) cells in the mono-layer culture at day three (n = 8), day eight (n = 8) and day 11(n = 4). Co-stained cells (�) represent a small population ofprotodifferentiated cells immunostained with both insulin andglucagon antibodies. Data are expressed as a function of thesurface area of the culture dish. *P < 0.05 vs 3 days.

Porcine pancreatic cell monolayers 45

glucose (16.7 mM) the insulin release increased2.4-fold, and further increased up to 38-fold aftersupplementation with 10 mM theophylline (Fig. 6).

Transduction of �-gal reporter gene by retroviral vector

The results of two separate transduction experimentson the cultured porcine pancreatic cells are summar-ized in Table 2. On day three, cells were infected witha MLV-based vector at a multiplicity of infection (MOI)of 1, 10 and 100. Successful transduction events werescored by expression of �-gal carried within thevector. Interestingly, while approximately 20% of thebeta cells were replicating (based on staining with theKi-67 marker, Fig. 5), the percentage of beta cells

expressing �-gal did not exceed 5%, even at an MOIof 100. In contrast, approximately 60% of the ductcells showed positive staining for �-gal despite thefact that only approximately 36% of the duct cells werereplicating. Figure 7 shows immunostaining of amonolayer transduced with MLV–�-gal at an MOI of10. Few �-gal-positive beta cells can be found(Fig. 7A), while �-gal-positive duct cells are muchmore abundant, as seen by staining of most of theduct cells in a patch (Fig. 7B).

Reaggregation of monolayers and transplantation

Monolayers of porcine pancreatic cells were veryfirmly attached to the culture dishes and required arelatively long incubation time with the dissociationsolution. The more common approach of using trypsinwas not used in these experiments because it wouldimpair subsequent cell aggregation, probably due tothe digestion of adhesion molecules (G. M. Beattie,pers. comm.,1999). Cells from day eight monolayerswere lifted off the culture dish with a non-enzymaticsolution. After an overnight culture in non-adhesivedishes, the cells formed 30–200 µm aggregates;however, some cell loss was observed. Theseexperiments were conducted as a proof of principle,therefore no quantification of the yield, size or cellcomposition of the aggregates was made. Immuno-histochemical analysis showed that the aggregatesconsisted predominantly of duct cells, with only a fewinsulin-positive cells (Fig. 8A), and sometimes they

Fig. 5. Replication of beta (�) and duct (�) cells in the mono-layer culture at days three and eight. Cells were double stainedfor Ki-67 (replication marker) and insulin or cytokeratin (CK-7).Cells were considered as replicating when nuclei were positivefor Ki-67. The data are presented as percentage of total betacells or duct cells counted in the dishes (n = 5 independentexperiments). *P < 0.05 vs 3 days.

Fig. 6. Glucose-induced insulin release from the monolayer ofporcine pancreatic cells cultured for 8 days. Separate cultureflasks were incubated for 2 h at 5.6 mM glucose, 16.7 mM glu-cose or 16.7 mM glucose supplemented with 10 mM theophyl-line. Data are shown as amount of insulin released as a functionof DNA content (n = 4 independent experiments). *P < 0.01 vs16.7 mM glucose.

Table 2. Percentage of beta cells and duct cells expressing�-galactosidase after transduction with Moloney Murine Leu-kemia Virus (MLV) retroviral vector

MOI

Beta cells Duct cells Individual

values MeanIndividual

values Mean

0 0, 0 0.0 0, 0 01 1.2, 1.2 1.2 9.3, 13 1110 3.5, 2.5 3.0 47, 30 38100 3.3, 5.4 4.3 67, 57 62

Cells were transduced on day three and 48 h later weredouble stained for �-galactosidase and insulin or cytokeratin(CK)-7. Up to 1000 beta cells or duct cells were counted foreach infection condition (n = 2 independent experiments). MOI,multiplicity of infection.

46 K. Tatarkiewicz et al.

formed characteristic duct-like lumens (Fig. 8B). Fourweeks after transplantation of the aggregates intonude mice, grafts were localized under the kidneycapsule as discrete white patches. Within the grafts,cells formed well-vascularized clusters, and fibrosiswas sometimes seen between them (Fig. 8C).Immunostaining for CK-7 revealed few duct cells.Further staining showed that the grafts consisted of

approximately equal numbers of beta and non-betaendocrine cells (Fig. 8D).

Discussion

This study showed that monolayers of duct and endo-crine cells from the porcine neonatal pancreas main-tained in culture for 11 days can undergo substantial

Fig. 7. Neonatal pancreatic cellcluster monolayers at day threewere transduced with a �-galac-tosidase reporter gene using theMoloney Murine Leukemia Virus(MLV) retroviral vector at a multi-plicity of infection (MOI) of 10. (A)Only approximately 3% of betacells (insulin, red) were express-ing the �-gal transgene (�-gal,green). (B) Approximately half ofthe duct cells (CK-7, red) werepositive also for the reporter gene(�-gal, green). Bars, 25 µm.

Fig. 8. Neonatal pancreatic cellclusters reaggregated after 8 daysin culture were transplanted underthe kidney capsule of normal nudemice. (A) Cell aggregates beforetransplantation consisted mostly ofduct cells (CK-7, green) with a fewinsulin-positive cells (insulin, red).The sizes of aggregates rangedfrom 30 to 200 µm. (B) Duct cellssometimes formed structures withlumens, as seen on 1 µm sectionsstained with toluidine blue. (C)Hematoxylin staining of a repre-sentative graft removed 4 weeksafter transplantation: kidney cap-sule (left, top) and kidney cortex(right bottom) adjacent to the graftare visible. (D) Adjacent sectionshowing that the graft consistedmostly of endocrine cells, with onlya few duct cells. Insulin-positive(red) and non-beta cell hormoneslurry-positive (green) cells ten-ded to form clumps. All imageswere taken at the same magnifi-cation: bar, 50 µm.

Porcine pancreatic cell monolayers 47

expansion from days three to eight. During the first3 days of culture, there was a marked decrease inDNA content, probably due mostly to death of acinarcells, a phenomenon previously reported (Korbuttet al. 1996; Yoon et al. 1999). A decrease in the insulincontent during this time period indicates that there ispossibly also an important loss of beta cells, althoughdegranulation of existing beta cells might contribute tothe decrease. After stabilization of the culture at daythree, the replication rate of duct cells was relativelyhigh, accompanied by a retroviral-based transductionof the �-gal reporter gene into approximately 50% ofthese cells. During the last stage of the culture period,fibroblasts became as abundant as duct cells: ifallowed to continue to grow, they would presumablyhave overgrown the duct cells. Nonetheless, whencells were lifted off and reaggregated after 8 days inculture, aggregates were found to consist mostly ofduct cells. When transplanted into nude mice, theseaggregates developed into endocrine cells. Thus, theneonatal duct cells in monolayer culture appear as anendocrine precursor, with the advantage of having amuch greater proliferative capacity and better retro-viral transduction efficiency than beta cells. The defin-ition of these events in culture allows us to betterunderstand the development of duct cells in vitro, anddemonstrates the feasibility of gene transfer, whichcould help these cells withstand rejection afterxenotransplantation. Furthermore, this culture systemcan be used as an in vitro model for duct celldifferentiation.

We are aware of only a single published reportabout the monolayer culture of neonatal porcinepancreatic cells (Archer 1983); however, similarculture of neonatal rat cells was described more thantwo decades ago (Lambert et al. 1972; Chick et al.1973). More recently, several studies of fetal human orporcine pancreatic tissue in monolayer cultures havebeen published (Mandel et al. 1982; Sandler et al.1985; Simpson et al. 1990; Simpson et al. 1991;Korsgren et al. 1993; Kovarik et al. 1995; Beattie et al.1996; Beattie et al. 1997b; Hayek & Beattie 1997;Yoon et al. 1999).

It is noteworthy that some authors have notdescribed the presence of duct epithelial cells in theirmonolayer cultures of human or porcine fetal pancre-atic cells, while in other studies the presence of ductepithelial cells has been reported (Mandel et al. 1982;Simpson et al. 1990; Kovarik et al. 1995; Beattie et al.1996). This omission may be explained by the unavail-ability of duct cell markers at the time of the formerstudies or by the methods used for exclusion ofcontaminating fibroblasts. Fibroblasts are known toattach rapidly to the surface of culture dishes. There-

fore, to eliminate them, a preincubation step of morethan 12 h was used in some studies (Rabinovitch et al.1980; Archer 1983; Montesano et al. 1983). Wewonder if such a long preincubation allowed attach-ment of duct cells that were then discarded with thefibroblasts. In our study, preincubation was for only3 h, because after overnight culture, we observedmultiple patches of flattened duct cells alreadyadhered to the dish. Some authors have used specificreagents, such as iodoacetic acid (Rabinovitch et al.1980) or thimerosal (Bratten et al. 1974), to eliminatefibroblasts; however, these reagents could have had atoxic effect on duct cells, as we observed in ourcultures exposed to thimerosal (data not shown). Anegative toxic effect of iodoacetic acid and thimerosalon fetal human pancreatic cells survival in vitro hasalso been reported by Simpson et al. (1991).

In our study, the DNA content of the monolayersincreased fourfold from day three to day eight ofculture, but the insulin : DNA ratio remained stable for8 days, as beta cell number also increased almostfourfold. New beta cells appeared by proliferation, asdemonstrated by Ki-67 staining of insulin-positivecells, but probably also by differentiation of duct cells:a morphological sign of such differentiation could bethe presence of endocrine cells closely associated toduct cells. After 8 days, the number of beta cellsreached a plateau, and further cell growth in themonolayer was mostly attributed to fibroblasts. Beattieet al. (1997a) observed a comparable increase ofDNA in fetal human pancreatic cells cultured asmonolayers on HTB-9 extracellular matrix. However,they reported a dramatic decrease in insulin contentbetween 4 and 7 days, which might be ascribed to anincreased number of precursor cells or fibroblasts, orto dedifferentiation of beta cells (Beattie et al. 1997a).Rabinovitch et al. (1983) showed that approximately13% of neonatal rat beta cells were incorporating 3Hthymidine. An eightfold increase of fetal porcineepithelial cell number during a 2-week monolayerculture with a 19% contribution of beta cells has alsobeen reported, but it is difficult to appreciate theidentity of the epithelial cells because staining forvimentin and cytokeratin was not done (Simpson et al.1990). A comparison of the rate of proliferation ofendocrine and non-endocrine cells among thedifferent studies is difficult to do because differentmethodologies were used, but in general, duct cellshave a higher proliferation rate than endocrine cells(Lefebvre et al. 1998). Similar results were found in astudy of late gestational fetal human pancreas inwhich approximately 5% of duct cells but only 0.1% ofbeta cells were Ki-67 positive (Bouwens et al. 1997).Our present results based on Ki-67 immunostaining

48 K. Tatarkiewicz et al.

show that in neonatal porcine cell monolayers, thenumber of duct cells in the cell cycle is almost twiceas high as that of beta cells.

Neonatal beta cells cultured for 8 days in monolayerwere responsive to both glucose and theophylline.Previously, neonatal porcine islet monolayers werereported to be responsive to glucose at day six ofculture, but to have lost their responsiveness by day10 (Archer 1983), whereas neonatal porcine beta cellsresponded only to theophylline, with a 2.4-foldincrease in insulin release compared to basal levels(Simpson et al. 1990). This discrepancy might beexplained by the presence of nicotinamide in ourstudies. Nicotinamide was found by Otonkoski et al.(1993) to enhance the insulin responses of humanfetal beta cells.

In the present study, we supplemented culturemedium with nicotinamide, an inhibitor of poly(ADP-ribose) synthetase, because it has been shownto be one of the most beneficial factors for maturationand replication of pancreatic endocrine cells. In thepresence of 10 mM nicotinamide, DNA and insulincontent increased significantly in the culture of humanor porcine fetal islet-like clusters (Sandler et al. 1989;Korsgren et al. 1993; Otonkoski et al. 1993). In addi-tion, nicotinamide enhanced adhesion of adult por-cine beta cells and inhibited fibroblast proliferation(Ohgawara et al. 1998). After transplantation into nudemice, aggregated monolayers of human fetal pancre-atic cells treated with nicotinamide demonstratedbetter maturation to endocrine cells than thosewithout treatment (Beattie et al. 1996).

When cells from neonatal porcine pancreatic mono-layers were harvested, aggregated and transplantedunder the kidney capsule, we found that 1-month-oldgrafts were mostly composed of endocrine cells. Wepresume that duct cells (the major component ofaggregates) derived from monolayers differentiatedin vivo into endocrine cells. Similar differentiation tobeta cells was observed when fetal human pancreaticcells were reaggregated from monolayers and trans-planted: reaggregation was crucial for graft survival,because monolayers retrieved and transplanted assingle cells did not survive transplantation (Beattieet al. 1996). Several studies have found that neonatalpancreatic cell clusters can proliferate, mature andnormalize glycemia in diabetic recipients (Sandleret al. 1985; Otonkoski et al. 1993; Korbutt et al. 1996;Beattie et al. 1997b; Yoon et al. 1999).

One of the most important impediments for suc-cessful transplantation in diabetes is protection fromimmune rejection. Genetic modification before trans-plantation may be a useful strategy to protect isletgrafts from destruction by immune rejection and

autoimmunity. Potential approaches include alterationof islet recognition, secretion of immunomodulatoryproteins, or overexpression of anti-apoptotic proteins(Weir & Bonner-Weir 1997; Feng et al. 1999; Giannou-kakis et al. 1999). Among the commonly used tools forgene delivery, adenoviral vectors provide high levelsof gene transfer and gene expression, but thisexpression only lasts for a short period of time. Alter-natively, retroviral vectors allow sustained long-termgene expression by permanently inserting the genesinto the host genome. However, a wider application ofthe conventional MLV-based retroviral vectors hasbeen limited by the requirement of host cells needingto be replicating. Future work in transducing mature,non-dividing islets may benefit from the newly devel-oped lentiviral vectors, which infect non-dividing cellswith high efficiency (Naldini et al. 1996; Blomer et al.1997).

In this study, the high proliferation rate of duct cellson day three allowed for an impressively high genetransfer efficiency by the concentrated MLV vector. Itis possible that sequential infection might result in aneven higher level of transduction. In contrast, thetransduction efficiency of beta cells was surprisinglylow. While the fraction of cycling beta cells (deter-mined by Ki-67 staining) was approximately 50% ofthat of duct cells, the transduction efficiency of betacells was only approximately 5–10% of that found induct cells. It is not clear at the moment whether thelower percentage of beta cells expressing �-gal wasdue to less proviral integration, less �-gal geneexpression from integrated provirus, or both. Similarresults were reported in a study by Leibowitz et al.(1999), in which infectivity with an MLV-derived vectorwas much higher (approximately 20%) in monolayersof human fetal pancreatic cells than in beta cells only(<1%). The same authors also demonstrated thatinfectivity was enhanced when cells were cultured asmonolayers compared to clusters in suspension. Thediffusion of the viral particles to the center of the isletshas been found to be a problem to achieve goodinfectivity rates in islets (Feng et al. 1999), whichmakes monolayers a good alternative for viral trans-duction studies.

Success of transplantation of isolated adult humanislets in the treatment of diabetes has been limited,even with the use of multiple donors (Brendel et al.1999) and improved immunosuppression (Shapiroet al. 2000). Adult islet cells have a very low replicationrate (Lefebvre et al. 1998) and presumably lowerpotential for expansion than pancreatic fetal or neo-natal tissue. In a comparison of grafts 3 months aftertransplantation of equivalent volumes of human adultislets and fetal islet cell clusters into nude mice, it was

Porcine pancreatic cell monolayers 49

found that the fetal grafts contained more insulin thanadult grafts (Hayek & Beattie 1997). These findingssupport the idea that the presence of duct epithelialcells may improve graft survival. The fact that ductcells can be efficiently transduced with retroviralvectors to express immunoprotective genes makesthem more attractive for further investigation.

In summary, our study indicates that it is possible toexpand neonatal porcine pancreatic cells as mono-layers consisting mostly of duct epithelial cells. Thismonolayer culture approach provides a useful modelfor the study of duct cell growth and neogenesis ofislet cells. Moreover, these cells, particularly ductcells, grow rapidly in culture, and can therefore beused as targets for transfer of genes carried by retro-viral vectors. After transplantation, the grafts werefound to contain mostly endocrine cells, and thisfinding is consistent with the assumption that betacells originate from duct cells. Because transplantedpancreatic duct cells appear to differentiate into endo-crine cells in vivo, it seems possible that this strategy,with its efficient transduction, could lead to long-termgene expression in the beta cells contained in thegrafts.

Acknowledgements

The authors appreciate the kind donation of MLV-based virus by the Laboratory of Richard C. Mulliganfrom the Department of Genetics, Harvard MedicalSchool and Howard Hughes Medical Institute, TheChildren’s Hospital, Boston, MA, USA. The experttechnical assistance of Jennifer Lock, Magdalena S.Merrill and Christopher Cahill is also greatly appreci-ated.

This work was supported by grants from theNational Institutes of Health (DK-53087 to G. C. W. andDK-36836 to the Joslin Diabetes EndocrinologyResearch Center), from the Islet Core Laboratory ofthe Juvenile Diabetes Research Foundation to K. T.and the Center for Islet Transplantation at HarvardMedical School, and by an important group of privatedonors.

References

Archer, F. J. 1983. Monolayer culture of neonatal pig pancreaticislet cells. Diabetologia 24, 185–190.

Beattie, G. M., Cirulli, V., Lopez, A. D. & Hayek, A. 1997a. Ex vivoexpansion of human pancreatic endocrine cells. J. Clin.Endocrinol. Metab. 82, 1852–1856.

Beattie, G. M., Otonkoski, T., Lopez, A. D. & Hayek, A. 1997b.Functional �-cell mass after transplantation of human fetalpancreatic cells. Diabetes 46, 244–248.

Beattie, G. M., Rubin, J. S., Mally, M. I., Otonkoski, T. &Hayek, A. 1996. Regulation of proliferation and differenti-

ation of human fetal pancreatic islet cells by extracellularmatrix, hepatocyte growth factor, and cell-cell contact.Diabetes 45, 1223–1228.

Blomer, U. L., Naldini, L., Kafri, T., Trono, D., Verman, I. M. &Gage, F. H. 1997. Highly efficient and sustained genetransfer in adult neurons with a lentivirus vector. J. Virol. 71,6641–6649.

Bouwens, L., Lu, W. G. & Krijger, R. D. 1997. Proliferation anddifferentiation in the human fetal endocrine pancreas.Diabetologia 40, 398–404.

Bouwens, L., Wang, R.-N., De Blay, E., Pipeleers, D. G. &Kloppel, G. 1994. Cytokeratins as markers of ductal celldifferentiation and islet neogenesis in the neonatal ratpancreas. Diabetes 43, 1279–1283.

Bratten, J. T., Lee, M. J., Schenk, A. & Mintz, D. H. 1974.Removal of fibroblastoid cells from primary monolayercultures of rat neonatal endocrine pancreas by sodiumethylmercurithiosalicylate. Biochem. Biophys. Res. Com-mun. 61, 476–482.

Brendel, M. D., Hering, B. J., Schultz, A. O. & Bretzel, R. G., eds.1999. Longest graft function of islet allografts in diabetesmellitus. In International Islet Transplant Registry News-letter, pp. 4–18. Justus-Liebig-University, Giessen.

Chick, W. L., Lauris, V., Flewelling, J. H., Andrews, K. A. &Woodruff, J. A. 1973. Effects of glucose on beta cellsin pancreatic monolayer cultures. Endocrinology 92,212–218.

Davalli, A. M., Ogawa, Y., Scaglia, L. et al. 1995. Function, mass,and replication of porcine and rat islets transplanted intodiabetic nude mice. Diabetes 44, 104–111.

Feng, S., Quickel, R. R., Hollister-Lock, J. et al. 1999. Prolongedxenograft survival of islets infected with small doses ofadenovirus containing CTLA4Ig. Transplantation 67,1607–1613.

Giannoukakis, N., Rudert, W. A., Robbins, P. D. & Trucco, M.1999. Targeting autoimmune diabetes with gene therapy.Diabetes 48, 2107–2121.

Hayek, A. & Beattie, G. M. 1997. Experimental transplantation ofhuman fetal and adult pancreatic islets. J. Clin. Endocrinol.Metab. 82, 2471–2475.

Korbutt, G. S., Elliott, J. F., Ao, Z., Smith, D. K., Warnock, G. L.& Rajotte, R. V. 1996. Large scale isolation, growth, andfunction of porcine neonatal islet cells. J. Clin. Invest. 97,2119–2129.

Korsgren, O., Andersson, A. & Sandler, S. 1993. In vitroscreening of putative compounds inducing fetal porcinepancreatic �-cell differentiation: Implications for celltranpslantation in insulin-dependent diabetes mellitus. Ups.J. Med. Sci. 98, 39–52.

Kovarik, J., Koulmanda, M. & Mandel, T. E. 1995. Organ cultureof fetal pig pancreas for transplantation. Diabetes Nutr.Metab. 8, 361–368.

Lambert, A. E., Blondel, B., Kanazawa, Y., Orci, L. &Renold, A. E. 1972. Monolayer cell culture of neonatal ratpancreas: Light microscopy and evidence for immuno-reactive insulin synthesis and release. Endocrinology 90,239–248.

Lefebvre, V., Otonkoski, T., Ustinov, J., Huotari, M., Pipeleers, D.& Bouwens, L. 1998. Culture of adult human islet prepar-ations with hepatocyte growth factor and 804G matrix ismitogenic for duct cells but not for beta-cells. Diabetes 47,134–137.

Leibowitz, G., Beattle, G. M., Kafri, T. et al. 1999. Gene transferto human pancreatic endocrine cells using viral vectors.Diabetes 48, 745–753.

50 K. Tatarkiewicz et al.

Leong, A. S.-Y. & Milios, J. 1993. An assessment of the efficacyof the microwave antigen-retrieval procedure on a range oftissue antigens. Appl. Immunohistochem. 1, 267–274.

Mandel, T. E., Hoffman, L., Collier, S., Carter, W. M. &Koulmanda, M. 1982. Organ culture of fetal mouse andfetal human pancreatic islets for allografting. Diabetes 31,39–47.

Montesano, R., Mouron, P., Amherdt, M. & Orci, L. 1983.Collagen matrix promotes reorganization of pancreaticendocrine cell monolayers into islet-like organoids. J. CellBiochem. 97, 935–939.

Naldini, L., Blomer, U., Gallay, P. et al. 1996. In vivo genedelivery and stable transduction of nondividing cells by alentiviral vector. Science 272, 263–267.

Ohgawara, H., Shikano, T., Fukunaga, K., Yamagishi, M. &Miyazaki, S. 1998. Establishment of monolayer culture of pigpancreatic endocrine cells by use of nicotinamide. DiabetesRes. Clin. Prac. 42, 1–8.

Ory, D. S., Neugeboren, B. A. & Mulligan, R. C. 1996. A stablehuman-derived packaging cell line for production of hightiter retrovirus/vesicular stomatitis virus G psuedotypes.Proc. Natl Acad. Sci. USA 93, 11 400–11 406.

Otonkoski, T., Beattie, G. M., Mally, M. I., Ricordi, C. & Hayek, A.1993. Nicotinamide is a potent inducer of endocrinedifferentiation in cultured human fetal pancreatic cells.J. Clin. Invest. 92, 1459–1466.

Pipeleers, D. G., in’t Veld, P. A., Van De Winkel, M., Maes, E.,Schuit, F. C. & Gepts, W. 1985. A new in vitro model for thestudy of pancreatic A and B cells. Endocrinology 117,806–816.

Rabinovitch, A., Blondel, B., Murray, T. & Mintz, D. H. 1980.Cyclic adenosine-3�,5�-monophosphate stimulates islet Bcell replication in neonatal rat pancreatic monolayercultures. J. Clin. Invest. 66, 1065–1071.

Rabinovitch, A., Quigley, C. & Rechler, M. M. 1983. Growthhormone stimulates islet B-cell replication in neonatal

rat pancreatic monolayer cultures. Diabetes 32,307–312.

Sandler, S., Andersson, A., Korsgren, O. et al. 1989. Tissueculture of human fetal pancreas. Effects of nicotinamide oninsulin production and formation of isletlike cell clusters.Diabetes 38, 168–171.

Sandler, S., Andersson, A., Schnell, A. et al. 1985. Tissue cultureof human fetal pancreas. Diabetes 34, 1113–1119.

Schluter, C., Duchrow, M., Wohlenberg, C. et al. 1993. The cellproliferation-associated antigen of antibody Ki-67: a verylarge, ubiquitous nuclear protein with numerous repeatedelements, representing a new kind of cell cycle-maintainingproteins. J. Cell Biochem. 123, 513–522.

Shapiro, A. M., Lakey, J. R., Ryan, E. A. et al. 2000. Islettransplantation in seven patients with type 1 diabetesmellitus using a glucocorticoid-free immunosuppressiveregimen. N. Engl. J. Med. 343, 230–238.

Simpson, A. M., Tuch, B. E. & Vincent, P. C. 1990. Pig fetalpancreatic monolayers. Transplantation 49, 1133–1137.

Simpson, A. M., Tuch, B. E. & Vincent, P. C. 1991. Characteriz-ation of endocrine-rich monolayers of human fetalpancreas that display reduced immunogenicity. Diabetes40, 800–808.

Starborg, M., Gel, K., Brundell, E. & Hoog, C. 1996. The murineKi-67 cell proliferation antigen accumulates in the nucleolarand heterochromatic regions of interphase cells and atthe periphery of the mitotic chromosomes in a processessential for cell cycle progression. J. Cell Sci. 109,143–153.

Weir, G. C. & Bonner-Weir, S. 1997. Scientific and politicalimpediments to successful islet transplantation. Diabetes46, 1247–1256.

Yoon, K.-H., Quickel, R. R., Tatarkiewicz, K. et al. 1999.Differentiation and expansion of beta cell mass in porcineneonatal pancreatic cell clusters transplanted into nudemice. Cell Transplantation 8, 673–689.