Nicotinamide is a potent inducer of endocrine differentiation in cultured human fetal pancreatic cells

Nicotinamide is a potent inducer of endocrine

differentiation in cultured human fetal

pancreatic cells.

T Otonkoski, … , C Ricordi, A Hayek

J Clin Invest.

1993;

92(3)

:1459-1466.

https://doi.org/10.1172/JCI116723

.

The effects of nicotinamide (NIC) on human fetal and adult endocrine pancreatic cells were

studied in tissue culture. Treatment of the fetal cells with 10 mM NIC resulted in a twofold

increase in DNA content and a threefold increase in insulin content. This was associated

with the development of beta cell outgrowths from undifferentiated epithelial cell clusters

and an increase in the expression of the insulin, glucagon, and somatostatin genes. DNA

synthesis was stimulated only in the undifferentiated cells. Half-maximal doses for the

insulinotropic and mitogenic effects of NIC were 5-10 and 1-2 mM, respectively. Islet-like

cell clusters cultured with NIC responded to glucose stimulation with a biphasic increase in

insulin release (fourfold peak), whereas control cells were unresponsive to glucose. Both

control and NIC-treated cells developed into functional islet tissue after transplantation into

athymic nude mice. As compared with adult islets, the insulinotropic action of NIC could

only be demonstrated in the fetal cells. Our results indicate that NIC induces differentiation

and maturation of human fetal pancreatic islet cells. This model should be useful for the

study of molecular mechanisms involved in beta cell development.

Research Article

Find the latest version:

Nicotinamide

Is

a

Potent

Inducer of Endocrine Differentiation

in

Cultured

Human Fetal

Pancreatic

Cells

Timo Otonkoski,*Gillian M.Beattie,*Martin1.Mally,*CamilloRicordi,t and Alberto Hayek*

*TheLucyThorneWhittier Children'sCenter, The Whittier InstituteforDiabetes andEndocrinology, LaJolla,California92037; andtUniversity ofPittsburgh Transplant Institute, Pittsburgh, Pennsylvania15260

Abstract

The effects of nicotinamide (NIC)on human fetal and adult endocrine

pancreatic

cellswerestudied in tissue culture. Treat-mentof the fetal cells with 10mMNIC resulted inatwofold increase in DNA content and athreefold increase in insulin content. This was associated with thedevelopment of

ft

cell outgrowths from undifferentiated

epithelial

cell clusters andan increase in the

expression

of the

insulin,

glucagon, and somato-statin genes.DNA

synthesis

wasstimulatedonly in the undif-ferentiated cells. Half-maximal doses for the insulinotropic and mitogenic

effects

of NICwere5-10 and 1-2

mM, respectively.

Islet-like cell clusters cultured with NIC

responded

to

glucose

stimulation witha

biphasic

increasein insulin release

(fourfold

peak),

whereas control cells were

unresponsive

to

glucose.

Both control and NIC-treated cells

developed

into functional islet tissue after

transplantation

into

athymic

nude mice. As

compared

with adult

islets,

the insulinotropic action of NIC couldonly be demonstrated in the fetal cells. Our results indi-cate that NIC induces

differentiation

andmaturation of human fetalpancreatic islet cells. This model should be useful for the

study

of molecular

mechanisms

involved in

,6

cell

development.

(J. Clin. Invest. 1993. 92:1459-1466.) Key words: islets of Langerhanse nicotinamide * tissue culture * cell

differentiation.

transplantation

Introduction

Nicotinamide (NIC) treatment preventsthedevelopmentof diabetes in

experimental

animals

following administration of

the

#

cell toxinsstreptozotocinand alloxan( 1),aswell asthe spontaneous diabetes of NOD mice(2). The mechanism of this effect has been attributedto inhibition of the nuclear en-zyme poly(ADP-ribose) synthetase which,ifactivatedduring DNA

repair synthesis,

could lead toacritical decreaseinthe

Parts of this work were presentedatthe 52nd Annual Meeting of the AmericanDiabetesAssociation,SanAntonio,TX, 20-23 June1992, and have appeared in abstract form (1992. Diabetes. 41[Suppl.

1]: 159A).

Address correspondence and reprint requests to Timo Otonkoski, M.D., The Whittier InstituteforDiabetes and Endocrinology, 9894 Genesee Avenue, La Jolla, CA 92037.

Receivedfor publication 20January 1993 and in revised form 9 April1993.

1.Abbreviations usedinthispaper: BrdU,bromodeoxyuridine; ICC, islet-like cell cluster; NIC,nicotinamide;nt, nucleotides.

NADlevel

of

the

,3

cells

(3).

Other

mechanisms,

however,

may be

operative

such as

free

oxygen

radical

scavenging (4)

and

inhibition of

MHC classII

antigen expression

ontheislet cells (5). NIC also enhances the recovery

from

diabetes

after

90% pancreatectomy both in rats

(6)

and

dogs

(7),

suggesting

a

stimulatory effect

on

islet

regeneration.

Further studies have

confirmed

thatNIC promotes

replication

ofmouseislet cells both in culture (8) and

after

transplantation

(9).

The

only

other

study addressing

the

effects of NIC

on human islets

showed

a

stimulatory effect

ontheformation of islet-like cell clusters

(ICC)

in human

fetal

pancreatic

cultures

(10).

Although the results of clinical trials using NIC in newly

diagnosed

type 1

diabetic

patients

have not been encouraging

(1 1-13),

recent experience in prediabetic individuals suggests that, when started early enough in the course of the disease, NICadministration may delay or prevent the development of typeI

diabetes

( 14). These

observations

havepromoted several

prospective

clinical studies that are currently in progress (

15

).

Considering

the many potent actions of NIC on pancreatic islet cells in

animal

models, as well as its use in clinical trials to preserve (3 cell function, we have characterized the effects of NIC on the

function,

growth, and differentiation of human islets.

Methods

Tissue culture. Human fetal pancreases at 18-24 gestational weeks were obtained through nonprofit organ procurement centers (Ad-vanced Bioscience Resources, Oakland, CA, and International

Insti-tutefortheAdvancementofMedicine,Exton, PA). Patient consent for tissuedonationwasobtained by the procurement centers. In addition,

our ownInstitutional ReviewBoard had reviewed and approved the

useof fetal tissue forthesestudies.The pancreases, obtained by dilata-tionandextraction,wereshippedon iceinRPMI 1640medium (Ir-vine Scientific, Irvine, CA) containing 10% pooled normal human

serumandantibiotics(100U/ml penicillin, 0.1mg/ml streptomycin, and 1

_ttg/ml

amphotericinB) and received in the laboratory within 18-24 h.Digestionand culture of the tissue was carried out essentially

asdescribedpreviously( 16).Briefly,the pancreases were dissected free ofsurrounding tissue, weighed,and cut into

1-mm3

pieces. After wash-ingin HBSScontaining 10 mM Hepes (Sigma Immunochemicals, St. Louis,MO), the fragments were digested for 15 min in a shaking water bath

(370C,

200oscillations/min) in HBSS containing 5.5 mg/ml collagenase P (Boehringer Mannheim Biochemicals, Indianapolis, IN). Digested tissue was washed twice in cold HBSS and plated on 60-mmpetridishes(diSPo; Baxter, McGaw Park, IL) in RPMI 1640 supplementedwith 10% human serum and antibiotics. Various con-centrations of NIC(SigmaImmunochemicals) were added to the

cul-turemediumfrom the beginning. The NIC content of the standard mediumwithout additionswas8

MiM.

Medium was changed on the 4th dof culture and every 2-3 d thereafter. ICCs were harvested from the dishes after4-1 1d. Todetectpossiblecytotoxic effects, lactate dehydro-genaseactivitywasmeasured in the culture medium withacommercial kit(SigmaImmunochemicals).

J.Clin. Invest.

© The AmericanSocietyforClinicalInvestigation,Inc.

For monolayer culture, ICCs were transferred to tissue culture dishescoated withbovinecornealendothelial cell matrix as described previously ( 17). Nonattached cells were removed after 3 d.

Adult humanisletswereisolated with an automated method as describedpreviously (18) and shippedonice. Dithizonestaining was usedtodeterminethepurityof the islet preparations ( 19). The same method was used to detectinsulin-containingcellsin the fetalICCs.

Toquantitatetheeffect ofNIC on thedevelopment ofICCs and the (Bcell massafter7d in culture,pieces of10 consecutive pancreases were blotted with filter paper to remove excess moisture, weighed, and treated asabove.After7d, allICCs were harvestedandcounted, and theresults were correlated to theoriginaltissue weight.

Immunohistochemistry. ICCs were incubated for 16 h with 0.1mM bromodeoxyuridine (BrdU), fixedin 4%paraformaldehyde,and em-bedded inparaffin.

_10-,gm

sectionswerestainedusingthe immunoal-kalinephosphatasetechnique (20) for insulin,glucagon and somato-statin andtheimmunoperoxidase technique (21 ) forBrdU.Polyclonal guinea pig antiporcine insulin (dilution 1:100) (Chemicon, El Se-gundo, CA), rabbit antiglucagon(1:250)andrabbitantisomatostatin (1:150) (Biogenex Labs, San Ramon, CA) antibodieswereused. Cell nuclei whichhadincorporated BrdU duringDNAsynthesiswere iden-tified bybinding ofmousemonoclonal anti-BrdU(1:20) (Dako, Car-pintaria, CA).Thesurfaceareaofinsulinpositive cells,aspercentof the total ICC area, wasquantitatedwithacomputerized image analyzer (American Innovision, San Diego, CA).The same method was used forthedetermination ofBrdUlabelingindex.

DNA synthesis and insulin secretion. After various incubation timesinculture, ICCswere harvested ingroupsof50into35-mmpetri dishesin 1 mlofmediumcontaining 1juCiof

[methyl-3H]thymidine

(specific activity5.0Ci/mmol, Amersham Corp., Arlington, IL).After anovernight(16-h) incubationat370Cthe mediumwascollected for insulin measurements by a solid-phase RIA (Diagnostic Products Corp.,LosAngeles, CA).The ICCs were washedtwicein PBS, pH 7.4, resuspendedin300 Ml of distilledwater,and homogenized by sonica-tion. DNA wasmeasured from thesonicate fluorometrically as de-scribed (22). Insulin content was measured by RIA in dilutions (> 1:20) ofacid ethanol extracts.Incorporationof [3H]thymidine was determined by trapping 10% trichloroacetic acid precipitates of the sonicatesonglass fiber filters (Whatman GF/A, Maidstone, United Kingdom), drying,andliquid scintillation counting in 4 ml of Beta-Max(ICN, Irvine, CA).

Insulinrelease in response to glucose and glucose plustheophylline wasstudied by perifusionasdescribedpreviously(23). After 7 din culture, batches of 300-400ICCswereloadedinperifusionchambers filledwithBio-GelP2(Bio-Rad Laboratories, Richmond,CA) micro-beads. Krebs-Ringer bicarbonate buffer supplemented with 20mM Hepes, pH7.40,and0.2%bovineserumalbuminwas pumped through thechambersat0.25ml/min. Uptosixchambers wereperifused simul-taneously(Superfusion 600; Brandel, Gaithersburg,MD).

Islet hormone geneexpression. Islet hormone mRNA levelswere

measured witharibonucleaseprotectionassay(24).Total RNAwas

isolated(25)from fetal ICCscultured in the absenceorpresence of10 mMNIC for7 d. RNA(400 ng)was

hybridized

for18 hat

56°C

with 32P-radiolabeledantisenseprobesspecificfor human insulin(262

nu-cleotides[nt],cDNAprobeprovided byDr. P.Miettinen,

University

ofHelsinki, Finland), glucagon(389 nt, cDNAprovided byDr. D. Drucker, Toronto GeneralHospital,Toronto,Ontario, Canada),or

somatostatin(236nt,asubclone from the

genomic

sequenceobtained from American Type Culture Collection, Rockville, MD). Excess probe and nonhomologous RNA sequenceswereremovedby ribonu-clease digestion. Theprotected,double-stranded hybridswere sepa-ratedby denaturing PAGE,followedbyautoradiographytovisualize the target RNAs. Cyclophilin (135 nt, cDNA provided byDr. D. Bergsma, SmithKline BeechamPharmaceuticals, KingofPrussia, PA) wasused as aninternalquantitative probetoensureequivalentRNA

sampling.Probe-specificmRNAsignalswerequantitated by scanning densitometry(Ultroscan;LKBInstrumentsInc., Bromma,Sweden), and normalizedtothecyclophilin signal.YeasttransferRNA(10 g)

was used as a negative control. To allow for direct quantitation between transcript levels, riboprobes were adjusted to 1,000 (insulin, glucagon, and somatostatin) or 3,000 (cyclophilin) cpm per probe-specific uri-dine residue per microliter.

Transplantation studies.500 ICCs, cultured either in standard or 10 mM NIC medium for 7 d, were transplanted under the left kidney capsule of 6-wk-old athymic nude (nu/nu) BALB/C mice. Five mice were transplanted in each experimental group. After 3 mo, fasted mice were given glucose intraperitoneally (3 g/kg) to release human C pep-tide, which was measured with an RIA that does not cross-react with mouse C peptide (DPC). After obtaining blood samples at 0 and 30 min,theanimals were sacrificed and thegraftsremoved for histological analysis.

Statistics. Statistical significances of observed differences were tested with software for theMacIntosh(Statview II; Abacus Concepts, Berkeley, CA). Student's unpaired t test was used when only two groups were involved, and the variation of data was close to normal. Mann-Whitney U test was used when the data were skewed. Multiple comparisonswere donewith one-way ANOVA and Fischer's protected least significance difference (PLSD) test using 95% level as the limit of significance.

Results

Culture characteristics.

Thedigested tissuestayed freefloating in the form of numerous small cell clusters of30-200 Mmin diameter (ICCs). The number of ICCs increased during the first 4 d in culture. In the control medium the ICCs remained rounded and regularly shaped. In contrast, when cultured in

medium containing 10 mM NIC, several small outgrowths

were seenafter4 d.When stained with dithizone, the majority of the cells within these budswere positive, indicating the pres-ence of insulin-containing cells (Fig. 1). In the control cul-tures, dithizone staining was seen weakly in small areas within the ICCs, and budding cells were rare. When the ICCs were plated on extracellular matrix-coated dishes, the majority of them attachedand spread out, forming monolayers as previ-ously described(17). Duringthefirst7 d, the monolayers con-sisted mainly of epithelioidcells, but fibroblast overgrowth oc-curred when the cultureswerecontinued. Thiswas more obvi-ousin thepresenceof 10 mMNIC.

10mM NIC induceda48%increase inthe number of ICCs developing during culture. The DNAcontent of theindividual ICCswas notsignificantly affected.As aresult, the total DNA contentof the 7-d culturedtissue was increased 2. 1-fold. In comparison, the total insulincontent wasincreased 3.3-fold, suggesting preferential development,orsurvival, of

_#3

cells (Ta-ble I).

Whenstained forthepancreatichormones after 7 d in cul-turewith 10 mM NIC, the majority of the outgrowing cells stained positively for insulin. Many BrdU-labeled cells were also presentwithinthe ICCs. As a rule, thehormone-positive cellswereBrdU-negative (Fig. 1). Only occasional cells posi-tive for both insulin and BrdUweredetected in both controls and NIC-treated cultures. In four separate experiments, the meanlabeling index forBrdU-positive nuclei was three times higher inNIC-treated cultures as compared with the controls (3.6vs1.2%,P <0.01).Inthesametissuesections, there was also a threefold increase in the cell surface areacovered by insulin-positivecells(9.3 vs 3.1%, P <

0.01,

Fig. 2).

simulta-a

d

Figure 1. Dithizonestaining ofhuman fetal ICCs culturedfor7d incontrol medium(A) ormedium containing 10 mM NIC (B). Positively stained (red)cellsarebuddingoutfrom the NIC-treatedICCs,whereasonly weakly positivecells can be seenwithinthe control ICCs. (C and D)Sectionsof 7-d cultured ICCs froma22-wk pancreas stainedimmunohistochemicallyfor insulin(pink)and BrdU (dark brown). (C) Cul-tured in standardmedium; (D)cultured with 10 mM NIC. Increased BrdU labeling and outgrowths of insulin-positive cells are evident in the NIC-treated ICCs. The cells that haveincorporatedBrdUduringthe 16-hincubationdo notcontaininsulin (A and B, x306; C and D, X733).

neously measured levels of lactate dehydrogenase were

consis-tently

lower in theNIC-supplemented medium (Fig. 3).

DNA

synthesis

and

(3-cellfunction.

Culture of ICCs for 7 d with various concentrations ofNICled to a dose-dependent increase in insulin content and basal insulin release.

Conse-quently,

the ratioof insulin release to content remained

un-changed.

Theresponse was

threefold

at 20 mM NIC. At least 5

TableI. Effect ofNicotinamideonthe Development ofICCs during7dinCulture

Control Nicotinamide P

10mM

Numberof ICCs/mg starting

tissue 14.7±2.1 21.8±1.4 0.01

TotalDNA(gg)/mg starting tissue 0.40±0.04 0.83±0.10 0.001

DNA/ICC(ng) 28.7±2.5 39.2±5.6 0.11

Totalinsulin (pmol)/mg starting

tissue 0.96±0.36 3.19±0.70 0.009

Pieces of human fetal pancreaswereweighed, digested,andcultured

asdescribed in Methods. After7d,theICCswerecounted and har-vested for measurement of insulin andDNAcontent. Data arethe

means(±SEM)of 10 pancreases.

mM NIC was required for a

significant effect

on the

insulin

levels

(Fig.

4). DNA

synthesis

was stimulated by NIC at a lower

concentration

range; 1 mM

concentration

was

sufficient

for

a

significant increase,

andthe

effect

was

maximal (2.4-fold)

at 5 mM. Higher

concentrations did

not further increase the DNA

synthesis.

In

fact, thymidine

incorporation was

signifi-cantly

lower at 20 mM, as compared with 5 or 10 mM NIC

(Fig. 4).

61

5-

4-2 3.

IC

i2-A lsn

2

10-8 c 5. a₃

.5

Control Nicotinamide

B

Control Nicotinamide

Figure 2. Morphometric analysisofimmunohistochemicallystained ICCsections.Labelingindex(percentofBrdU-labelednuclei,A)and surfaceareaofinsulin-positive cytoplasm (expressedaspercent of

to-talcell area,B)areshown for four separate

experiments.

Atleast

1,200 nuclei and 10 ICCswereevaluated for thedeterminationof each value.

*P

.40,

".- 0

A

5000 *

4000_

i3000-l|**..

CL

1000- 1 Ii I

4 6 8 11 13

Days

B

100 80

..J 60

0I

4 6 8 11 13

Days

Figure 3. Contentof in-sulin (A)and lactate dehydrogenase(B)in theculturemediumof fetal ICCs duringa 13-d cultureperiod in stan-dardmedium(open col-umns)orwith10 mM NIC(filledcolumns). Data are the mean±SEMfor five groupsof50 ICCs cul-tured intissue culture wells in1mlofmedium since day2. Media were changedat2-3 d inter-vals andfrozen for in-sulinand lactate dehy-drogenase assays. The ICCsattachedto the bottom of the dishand startedtoforma mono-layer byday 5. Fibro-blastovergrowthwas evident after 11 d. InB, thedetection limit of the LDHassay is de-notedby the horizontal line.*P<0.05;**P

<0.01; ***P<0.001, ascompared with con-trolusing Mann-Whit-ney U test.

A

I1

0 5 10 15 20

Nlcotnamld (mM)

B

z

a

I

5 10 15 20

In control cultures, monolayer formation was

accompa-niedbyamarked increaseintherateof DNAsynthesisanda

decreaseintheinsulin release after 7 d(Table II).Insulin

con-tentwasalsodecreased,but toalesser extent. Thepresenceof contaminating fibroblasts hampers the interpretation of

re-sults.Nevertheless,it is evident thatmonolayer-cultured acells

were in a different functional state, characterized by a 70%

lower fractional insulin release rate,ascomparedwith the ICCs (3.5vs12.3%, Table II). The fractional insulin release of

mono-layers cultured with NICwascloser to that of the ICCs(9.2vs

12.1%), possibly reflecting preservation ofahigherlevel of cell differentiation (Table II). Contrarytothe resultsin

suspen-sion culture, NIC mainly induced the growth of noninsulin containingcells in themonolayer,asjudged byadoublingof theDNA content and a decrease in the insulin content per

DNA(TableII).

Theacuteinsulinresponseof ICCs cultured for7d eitherin control mediumorwith 10 mM NICwastested inaperifusion

system.Insulin release fromcontrol ICCswasunresponsiveto

glucose alone, whereas the combination of 16.7 mMglucose and 10mMtheophyllineinduced apromptresponse(Fig. 5

A). Incontrast,ICCscultured with NIC hadasignificant

four-fold first-phase glucose response. Although attenuated,there

wasalsoevidence ofasecond-phaseresponse(Fig. 5 B).

To determine whether the effects of NIC could be

repro-duced with other poly(ADP-ribose) synthetase inhibitors,the ICCswerecultured with10mMbenzamideor10mM

3-amin-obenzamide, in parallel with 10 mM NIC.Benzamide hadan evenstrongerstimulatoryeffectonthe insulinlevels thanNIC,

Nicotinamkie(mM)

Figure 4. Response of insulin release (o) andcontent(., A) and DNAsynthesis (B)tovaryinglevels of NIC in the culture medium

duringa7-d cultureperiod.Dataarethe mean±SEMof 10-15

repli-catesfrom threeseparateexperiments.Insulin release andthymidine incorporationweremeasuredduringthe last 16 hasdescribedinthe Methods. *P<0.05,ascompared with the control medium

(con-taining 8

,M

NIC)usingonefactorANOVA and Fisher's PLSDtest.

whereas 3-aminobenzamide only stimulated the DNA synthe-sis but didnotaffect insulincontent(Table III).

Islet hormone gene expression. Insulin and glucagon

mRNAs were abundantly present in the ICCs, whereasthe level of somatostatin mRNAwasclearly lower. The somato-statin transcripts were seen astwo protected bands. NIC in-creased thetranscriptional levels of all three hormones, but therewas noobvious regulation of cyclophilin mRNA byNIC (Fig. 6). Afternormalization against cyclophilin expression, thedensitometrically determinedmeanfold increaseswere1.8

(range = 1.4-2.8), 1.7 (range = 1.3-2.9), and 1.7 (range

= 1.1-2.2) for insulin, glucagon and somatostatin, respectively

(P<0.05 for alleffects, Student's one-samplettest,n=5).

Graft development.HumanCpeptide could be detected in theserumof nude mice 3-5 moaftertransplantation of 500

controlorNIC-treated ICCs under the kidney capsule. In

ani-malsreceiving controlICCs, therewasa5.1-fold increase in the

Cpeptide level 30min afteranintraperitoneal glucose

chal-lenge. The difference inresponsesbetween animalsreceiving

control andNIC-treated ICCswasnotstatistically significant. Basal Cpeptide levelswerealsosimilar in bothgroups(Fig. 7).

zO

TableH.Effect ofTissueCulture Method(ICCvs.Monolayer)and NicotinamideonDNASynthesisand d Cell Function

Day 12, Day12

Day 5 Day 12 monolayer, Day 12 monolayer,

ICCM, NIC- ICCs, NIC- NIC- ICCs, NIC+ NIC+

DNAcontent

(Cg)

0.65±0.05 0.31±0.03 0.53±0.05 0.34±0.04 1.20±0.15*$

3H-Thymidine incorporation

(103cpm/dish) 3.2±0.2 1.1±0.3 9.9±1.5* 3.2±0.2* 20.3±3.7*$

3H-Thymidine incorporation

(103

cpm/jgg

DNA) 5.1±0.5 3.8±0.9 18.9±3.4$ 10.7±1.8* 16.4±1.5$

Insulinrelease

(pmol/dish) 0.25±0.03 0.26±0.03 0.04±0.01$ 0.49±0.02* 0.16±0.02*$

Insulinrelease

(pmol/;ig

DNA) 0.38±0.04 0.83±0.11 0.09±0.02* 1.62±0.21* 0.13±0.01*

Insulin content

(pmol/dish) 2.33±0.28 2.07±0.19 1.24±0.07* 4.27±0.40* 1.69±0.09$

Insulin content

(pmol/fg

DNA) 3.62±0.45 7.03±0.93 2.57±0.41* 13.5±1.53* 1.53±0.15*

Insulin release percentof

insulin content 11.3±1.4 12.3±1.2 3.5±0.6* 12.1±1.0 9.2±1.0*t

After 4 d inculture,groupsof50ICCsoriginating fromthree pancreaseswerecollectedeitherin culture wellinserts,wherethey remainedin suspension,or onECM-coateddishestoformamonolayer. Insulinrelease andthymidine incorporationwererecordedduringa16-h incubation eitherimmediately (day 5)or 7d later(day12).Dataarethe mean±SEMforeight replicates. Statistical significances usingone-way ANOVA andFischer's PLSDtest: *Significancelevel <95% whencompared with thesameculture methodwithout NIC. *Significancelevel.95% when comparedwith ICCs culturedin the same medium.

GraftsofNIC-treated and control ICCsweremorphologically similar, consisting of endocrine and ductal tissue. We have shownpreviously that 85% ofthe surface areaofhistological sections cut through the control grafts was immunoreactive when exposedto a

cocktail of

antibodies against the

four

islet hormones

(insulin,

glucagon,

somatostatin,

and pancreatic

polypeptide) (Beattie,

G. M., F.

Levine,

M.I. Mally, T. Oton-koski,J. S.O'Brien, andA.Hayek,manuscript submitted for

publication.)

Effects

on

adult

human

islets.

Theinsulincontent per cellu-larDNA

of

the adultisletswas comparable topreviously pub-lished data (26). When culturedfor7dwith 10 mM NIC, the DNA content

of

the

islets

wasincreased by 62% (TableIV). DNA

synthesis,

asmeasured by [

3H]

thymidine incorporation,

was 30%of the level observed in fetal

ICCs,

andit tended to, be

increased

by NIC. The insulincontentpercellularDNAof the adult

islets

was40

times

higher

than in thefetal ICCs.In contrast totheexperiments with fetal cells, the insulincontent

wassimilar in the two groups, and the insulin release was signifi-cantly lower in

islets

cultured with 10mM NIC (TableIV).

Discussion

Our

experiments

show that addition of NICtothe culture

me-dium of

human

fetal

pancreatic

cellsboth

increases

the total cell number and the

frequency of

f cells.

Thus,

the total num-berof ( cells

is increased

by NIC. This could be caused either by the

formation of

new

,3

cells

through differentiation

or

by

increased survival

of

the 3 cells thatwere

originally

present. Our

morphological

observations strongly supportthe former

possibility.

The

development of

insulin-positive outgrowths

from

the

undifferentiated epithelial

cellclusters

closely

resem-bles the

islet

neogenesis

that has beendescribed in the rodent pancreas both in vivo

(27)

and in

vitro (28).

The observed

increases

in the

transcription of

the islet hormonegenes by NIC, and the NIC-induced maturation of

glucose-induced

in-TableIII.

Effects

of Poly (ADP-ribose) Synthetase Inhibitorson :CellFunction andDNA Synthesis ofHumanFetalIslet-likeCell Clusters

after

7dinCulture

[3H]thymidine Insulin release

incorporation Insulin release Insulincontent percentofcontent

cpm/MgDNA pmol/lLgDNA pmol//lgDNA

Control 1,884±342 0.16±0.02 1.97±0.33 8.4±1.0

Nicotinamide 10mM 5,016±205§ 0.35±0.08* 4.26±0.4* 8.0±1.1

Benzamide 10 mM 3,346±476* 0.72±0.13* 9.98±1.21§ 7.1±0.6

3-Aminobenzamide 10mM 6,178±753* 0.51±0.1* 3.27±0.99 17.0±2.5*

Insulin release and thymidineincorporation were measured during the last 16 h. Data are the mean±SEM of four replicate groups of 50 ICCs.

0

minutes

G1.67 I G16.7 I

OJ --.

0 10 20 30 40 50 60

minutes

Figure 5.Insulinrelease inperifusion of adult islets (0)orfetal ICCs cultured for 7 d in stan-dard medium(o)or10 mMNICmedium (o). A,expressed in absolute units, shows the re-sponse tosequential stimulation with 16.7 mMglucose(G16.7) and 10mM theophyl-line(T1O). Insulin re-leasefromNIC-treated ICCs is

significantly

higherthanin the con-trolICCsthroughout theexperiment.Bshows only theglucose re-sponseofthefetal ex-periments expressedas percentchanges from theprestimulatorybasal level. Data arethe mean oftwo separate experi-mentswithadultislets and the mean±SEMof sixfetal experiments; very small errorbars are not shown.*P<0.05; **p<0.01; ***P

<0.001,ascompared withcontrolusing Stu-dent'stwo-samplettest.

sulinsecretion, furthersupportthe view thatNIC inducestrue endocrine differentiationinthe fetalpancreas.This is the first clear evidence for neogenesis of human insulin-producing cells inculture. Untilnow,it hasnotbeen possibletoreproduce in human tissue the proliferation, differentiation, and maturation of, cells thatoccursduring culture ofthe fetal rodentpancreas

(29). The only previous study exploring the effects ofNICon

cultured humanfetal islets suggestedanincreasein cellmass,

butnofurther conclusions could be drawn, possibly because of

poortissueviability (10).

The markedNIC-inducedimprovementintheglucose

sen-sitivity ofinsulin releaseisinteresting,becausesignificant devel-opment ofaglucose-inducedinsulinresponsefromhuman fe-tal cellshasbeen previously shownonlyafter engraftment into nude mice (30, 31). Unlike fetal rat islets (32), tissue culture ofthe human fetal pancreasin the presence ofhigh glucose doesnotinduce glucose sensitivity (33). Growth

hor-mone (34) and retinoic acid (35) mayinducemarginal

im-provement inthe glucose sensitivity, but the effects havenot

beennearlyasmarkedasthat ofNIC observed in this study.

Inaccordance with previous studies ( 30, 31 ), transplanta-tion ofthe ICCs under thekidney capsuleofnude mice resulted in maturationof glucose-induced insulin secretion, asjudged

bytheserumhuman C peptide concentrations measured 3mo

after transplantation. However, the beneficial effect ofNIC couldnolonger be demonstrated in vivoatthistime. Our

re-sults would thussuggest that NIC-pretreatment of fetal islet cells fortransplantation has only limited value in improving the functionalend resultofthegraft.However,ourdata donot

A

C N C N C N tRNA

Glucagon 4 _b _ _^p- _ _

Insulin* ** * l

Cyclophilin ...

B

C N C N C N tRNA

Somatostatin A A:

Cyclophilin

Figure 6. RNAseprotectionanalysis ofhormonetranscriptsinICCs obtained fromthree pancreases and culturedfor7dinthe absence (C)orpresence(N) of 10mMNIC. Total RNA(400 ng)was hy-bridizedsimultaneouslytoriboprobes forinsulin,glucagon, and cy-clophilin,or tosomatostatin andcyclophilin,for 18 hat56°C,as described inMethods. YeasttransferRNA(10 sg)wasusedas a negativecontrol. Protectedfragments for insulin (262 nt), glucagon (389 nt), somatostatin(236 nt),andcyclophilin (135nt)are indi-cated.Exposure timeswere3 hfor insulin andglucagonand 8 hfor somatostatinandcyclophilin.

exclude the

possibility

that NIC-treated

grafts

would have reachedmaturefunction earlier than the control grafts,ashas been

recently

shown for fetal

porcine islets (36).

It

should

also be notedthat we

transplanted

thesameamount

of

control and NIC-treated tissue.

Considering

the NIC-induced increase in ICC formation, the totalinsulin-producing graftmass

develop-ing

froma

single

pancreas maybe

increased

by

NIC

pretreat-ment.

1500- Figure7. Basal and

glu-cose-stimulated serum

0

ns human Cpeptidelevels

Si1o0004

measuredinnudemice

~~~~~~~~~~3-5moafter

transplan-0111 1 tation of 500 human

E00

n.s. 10 1 fetalICCs under the

T

iz z. kidney capsule. Before

|lucose

J transplantation, the

0~

asaS (Jlucos imulat ICCs were culturedfor

7 d in thestandard

me-dium(open columns)ormedium containing10mM NIC(filled

col-umns). Dataarethe mean±SEM of fiveobservations.Human C

peptidewasundetectable in micethat hadnotreceived human

trans-plants (not shown).

A

I

8

I

I

I

1250-

1000-

750-

500-

25

0-B

500,

400.

300-

200-

100-

L-I

TableIV.Effects ofNicotinamideonAdult Human Islets

Control Nicotinamide P

DNA content(ng/islet) 21.3±2.8 34.6±2.9 0.003

Insulinrelease(pmol/islet) 0.21±0.01 0.18±0.02 NS

Insulinrelease(pmol/flgDNA) 11.3±1.3 5.7±0.8 0.001

Insulincontent(pmol/islet) 1.54±0.20 2.05±0.23 NS

Insulincontent(pmol/ugDNA) 82.7±11.3 66.6±10.7 NS [3H]Thymidineincorporation

(cpm/islet) 16.8±2.8 31.2±5.7 NS

[3H]Thymidineincorporation

(cpm/AgDNA) 911±181 982±180 NS

Theisletswerecultured for7 deither in RPMImedium(5.6 mM glucose and 10% humanserum) alone,orwith10mM nicotinamide. Insulin release and[3H]thymidineincorporationwererecorded dur-ingthe last 16 h. Dataarethemeans(±SEM)of15

replicates

of 50 isletsfromthree pancreases.

Incontrast with the results ofSimpsonand Tuch(37),we found thatmonolayerculture of the ICCswasassociated witha decrease, andnot an

increase,

in insulin

production.

This is in accordancewith thegeneralview that increased cell prolifera-tioninvitrois

inversely proportional

tothestageof differentia-tion, which is also true for insulinoma cell lines

(38).

It is noteworthy that NICtreatment induced a simultaneous in-crease intherateof DNA

synthesis

and the level of endocrine

differentiation. However,

it is evident that the DNA

synthesis

was

mainly

stimulated

in the undifferentiated

cells,

whereas the BrdUlabeling of

hormone-containing

islet cellswas unde-tectable. This wouldsuggestthat NIC

triggered

aprocess in the uncommitted

epithelial

cells,

which

eventually

resultedin the

expression

ofanislet cell

phenotype.

Thus,

the observed insu-linotropic effects of NIC would

mainly

be causedby

prolifera-tion and

differentiation

ofprecursor

cells,

rather thana

mito-genic

effect on

existing f

cells. The functional maturation of the cells alsosupports

this

view.

Based onthese

observations,

theeffects of NIC should be

different

in the adult

islets,

becausethey consist

mainly

of

fully

differentiated endocrine

cells. In our

experiments,

the DNA content of the adult

islets

was

increased by NIC,

suggesting

a

stimulatory effect

oncell

proliferation. However,

incontrast to thefetal

cells,

thiswas

accompanied by

impairment of

insulin release. These resultsarein goodagreementwith

previous

stud-iesonadultrat

islets,

whereincreased cellreplicationwas asso-ciated with adverse effectson ( cell

function,

specifically

insu-lin release (8). If the insuinsu-linotropic action of NICwasmainly caused by differentiation of undifferentiated islet precursor

cells,

whichare presumably located within the ductal epithe-lium(39), it would be understandable thatthis effectcould not bedetected in the

isolated

islets.Itisnot

clear,

whetherislet cell

neogenesis

from suchprecursorcells couldcontribute signifi-cantlytotheisletcell mass in the adult pancreas. However,this is one ofthepossible mechanisms to explain the beneficial effects ofNIC on the

preservation

ofinsulin secretion in

experi-mentaldiabetesand inprediabetic individuals.

The effects ofNIC could be reproduced with two other inhibitors of poly(ADP-ribose)synthetase, benzamide, and 3-aminobenzamide (40), suggesting that decreased activity of thisenzyme wasthedirectcause of the observed changes. The

activity

of poly(ADP-ribose)synthetaseislinkedwith the level

of cellulardifferentiation in several cell types. Evidence for this comes from experiments with preadipocytes (41), intestinal epithelialcells(42), chondrocytes (43),teratocarcinoma cells (44),andthymocytes (45).Inall of thesecells, the ADP-ribo-sylating activity is lower in the more differentiated phenotype, and morphological evidence of differentiation is preceded by diminished activity of poly(ADP-ribose) synthetase. Further-more, chondrocyte and teratocarcinoma cell differentiation canbetriggered by inhibitors of the enzyme, such as NIC and 3-aminobenzamide(44, 46). These effects arelikelyto reflect changes in the ADP ribosylation of nuclearproteins,resulting inmodificationsinchromatin structure thatenablethe tran-scriptionof previouslymaskedregionsofDNA. However,the precise genes responsiblefordifferentiationremain elusive.

Inadditiontoinsulin,thetranscriptionallevelsoftwo other islethormones, glucagon andsomatostatin,were alsoinduced byNIC.This isinaccordancewithanaction on the putative pluripotent isletstem cell. Allfour islethormones are thought to arise from a common precursor, asevidenced by celllineage analysis (47). Cells expressing insulin together with glucagon, somatostatin, or pancreatic polypeptide can be detected after day 10 of mouse fetaldevelopment (47), and such mixedcell types have recently also been identified inthemidgestational human fetalpancreas (48).

Insummary, we have shown that morphological and func-tionaldifferentiation of human pancreatic (3 cells can be in-ducedby NIC in tissue cultureof human fetalpancreas rich in undifferentiated cells. Thus, this tissueculture model may be usefulinfurther studiestodissect the molecular mechanisms of islet cell differentiation. Although it is difficultto assess the possible clinical implications of our in vitro findings, NIC treatmentof fetalislet cellsintended for transplantation may increase the functional graftmass. Also, increasedneogenesis of

fl

cellsshould beconsideredas apossible mechanism forthe reportedbeneficial effectof NIC on the preservation of insulin secretion ( 14). Further studies in vivoarerequiredtoclarify this issue.

Acknowledaments

We thank Ana Lopez for expert technical assistance and Fred Levine for helpfulcriticismofthemanuscript. This study was supported by the NationalInstituteof Digestive grant RO1-DK 39087, The Charles H. andAnnaS.SternFoundation, The Herbert0.Perry Fund, The Deutz Family, The Sigrid Juselius Foundation (T. Otonkoski), and The PauloFoundation (T. Otonkoski).

References

1.Uchigata,Y., H. Yamamoto, H.Nagai,and H.Okamoto. 1983.Effectof

poly(ADP-ribose)synthetase inhibitor administrationtorats before andafter

injection of alloxan and streptozotocinonislet proinsulin synthesis.Diabetes. 32:316-3 18.

2. Yamada, K., K. Nonaka, T. Hanafusa,A.Miyazaki,H.Toyoshima, and S. Tarui. 1982. Preventiveandtherapeuticeffects oflarge-dosenicotinamide injec-tionsondiabetes associated with insulitis. Diabetes. 31:749-753.

3.Okamoto,H. 1984.Molecular basis ofexperimental diabetes:

degenera-tion, oncogenesisandregeneration of pancreaticB-cellsof islets ofLangerhans.

Bioessays. 2:15-21.

4.LeDoux, S. P., C.R.Hall,P. M.Forbes,N. J.Patton,and G. L. Wilson.

1988.Mechanisms ofnicotinamide and thymidine protection fromalloxan and

streptozotocintoxicity. Diabetes. 37:1015-1019.

5.Yamada, K.,E.Miyajima, andK.Nonaka.1990.Inhibition

ofcytokine-in-ducedMHC classIIbut notclassImolecule expressionon mouseislet cellsby

niacinamide and 3-aminobenzamide. Diabetes. 39:1125-1130.

H. Okamoto. 1984.Ameliorationofdiabetes mellitus in partially depancreatized rats bypoly(adp-ribose) synthetaseinhibitors. Diabetes. 33:401-404.

7.Sugiyama,K., Y.Yonemura, and H.Okamoto.1991.Effects of

poly(ADP-ribose)synthetaseinhibitoronB-cellsofacanine pancreas after massive pancre-atectomy. Int. J.Pancreatol. 8:85-95.

8. Sandler,S., andA. Andersson. 1986. Long-term effects ofexposureof pancreaticislets tonicotinamide in vitroonDNAsynthesis, metabolismand B-cellfunction.Diabetologia. 29:199-202.

9. Sandler, S., andA. Andersson. 1988.Stimulation ofcellreplicationin transplantedpancreaticislets by nicotinamidetreatment.Transplantation

(Balti-more). 46:30-31.

10. Sandler,S., A. Andersson,0.Korsgren, J.Tollemar,B.Petersson,C.-G. Groth,and C.Hellerstrom.1989.Tissue culture of human fetal pancreas.Effects ofnicotinamideoninsulinproductionandformationof islet-likecell clusters.

Diabetes. 38(Suppl. 1):168-171.

1 1. Vague, P., R.Picq,M.Bernal,V.Lassman-Vague,and B.Vialettes. 1989. Effect ofnicotinamidetreatment ontheresidual insulin secretion intype 1

(insu-lin-dependent)diabetic patients.Diabetologia. 32:316-321.

12. Mendola, G.,R. Casamitjana,and R.Gomis. 1989. Effect of nicotin-amide therapyupon B-cellfunction in newlydiagnosedtype 1

(insulin-depen-dent) diabeticpatients.Diabetologia.32:160-162.

13.Lewis,C. M., D. M.Canafax,J. M.Sprafka,and J. J.Barbosa. 1992. Double-blind randomized trial of nicotinamideonearly-onset diabetes. Diabetes

Care. 15:121-123.

14.Elliott,R.B.,and H. P.Chase. 1991.Preventionordelayoftype 1

(insu-lin-dependent) diabetes mellitus in childrenusingnicotinamide.Diabetologia.

34:362-365.

15.Chase, P.,J.Dupre,J.Mahon,R.Ehrlich,E.Gale,H.Kolb,E.Lampeter,

andJ.Nerup. 1992.Nicotinamideandpreventionof diabetes. Lancet.

339:1051-1052.

16.Otonkoski, T.,M.Knip,P.Panula,S.Andersson,I.Wong,H.Goldman,

and0. Simell. 1988. Morphology,yieldand functionalintegrityof islet-like cell clusters in tissuecultureof humanfetal pancreata obtained after differentmeans

ofabortion.ActaEndocrinol. 118:68-76.

17.Beattie,G.M.,D.A.Lappi,A.Baird,andA.Hayek. 1991. Functional impact ofattachment andpurificationin the shorttermcultureofhuman

pancre-atic islets.J. Clin. Endocrinol. & Metab. 73:93-98.

18.Ricordi,C.,P. E.Lacy,E.H.Finke,B. J.Olack,and D. W.Scharp.1988. Automatedmethodfor isolation ofhumanpancreaticislets.Diabetes.

37:413-420.

19.Latif,Z. A., J.Noel, and R.Alejandro.1988.Asimplemethodofstaining

freshandcultured islets.Transplantation(Baltimore).45:827-830.

20. Erber, W.N., and D. Y. Mason. 1987. Immunoalkalinephosphatase

labeling of terminal transferasein hematologic samples. Am.J. Clin. Pathol. 88:43-50.

21.Sternberger,L. A., P. H.Hardy,J.J.Cuculis,and H. G.Meyer.1970. The

unlabeledantibodymethodofimmunohistochemistry.Preparationand

proper-ties ofsoluble antigen-antibody complex (horseradish

peroxidase-antiperoxi-dase) and its use in identification ofspirochetes. J. Histochem. Cytochem.

18:315-333.

22.Hinegardner,R. T.1971.AnimprovedfluorometricassayforDNA. Anal.

Biochem. 39:197-201.

23.Otonkoski,T.,S.Andersson,M.Knip,and0. Simell. 1988. Maturation of insulinresponsetoglucoseduringhumanfetaland neonataldevelopment. Diabetes. 37:286-291.

24.Singer, P. A., R.S. Balderas,and A. N.Theofilopoulos. 1990.Thymic

selection definesmultipleT cell receptorVj#'repertoire phenotypes'attheCD4/

CD8 subset level. EMBO(Eur. Mol.Biol.Organ.)J.9:3641-3648.

25.Chomczynski,P., and N.Sacchi.1987.Single-stepmethodofRNA

isola-tion byguanidiumthiocyanate-phenol-chloroformextraction. Anal. Biochem. 162:156-159.

26.Eizirik,D.L., G. S.Korbutt,and C.Hellerstrom.1992.Prolonged expo-sure of humanpancreaticisletstohighglucose concentrationsin vitroimpairsthe ,B cell function. J.Clin.Invest.90:1263-1268.

27.Pictet, R., and W. J. Rutter. 1972. Development of the embryonic

endo-crinepancreas. In HandbookofPhysiology.Section7:Endocrinology. D. Steiner and N.Freinkel, editors.Williams and Wilkins, Baltimore, MD. pp. 25-66.

28.Dudek, R. W., N. Freinkel, N. J. Lewis, C. Hellerstrom, and R. C. John-son.1980.Morphologicstudyofculturedpancreatic fetalislets during

matura-tion of theinsulin stimulus-secretion mechanism. Diabetes.29:15-21.

29.Hellerstrom,C.,N. J.Lewis, H. Borg, R. Johnson, and N. Freinkel. 1979. Methodforlarge-scaleisolation ofpancreatic islets by tissue-culture of fetal rat pancreas.Diabetes.28:766-769.

30.Tuch,B.E.,A.Jones, and J. R. Turtle. 1985. Maturation of the response

ofhumanfetal pancreaticexplants to glucose.Diabetologia.28:28-31. 31. Hullett, D. A., K. P. Bethke, A. S. Landry, D. K. Leonard, and H. W.

Sollinger. 1989. Successfullong-term cryopreservation and transplantation of humanfetalpancreas.Diabetes.38:448-453.

32.Dudek, R. W., T. Kawabe, J. E. Brinn, K.O'Brian,M. C. Poole, and C. R. Morgan.1984.Glucoseaffects invitro maturation of fetal rat islets. Endocrinol-ogy.114:582-587.

33.Otonkoski, T.,M.Knip, I.Wong, andO. Simell. 1991.Lack

ofglucose-in-ducedfunctional maturation duringlong-term cultureofhumanfetal isletcells.

Life Sci. 48:2157-2163.

34.Otonkoski,T., M.Knip,I.Wong,and0. Simell. 1988.Effects ofgrowth hormone andinsulin-like growth factor1 onendocrinefunction ofhumanfetal islet-like cellclusters during long-term tissue culture.Diabetes.37:1678-1683.

35. Tuch, B. E., and K. J. Osgerby. 1990.Maturationof insulinogenic re-sponse toglucosein humanfetalpancreaswith retinoic acid.Horm.Metab.Res.

25(Suppl.):233-238.

36. Korsgren, O., A.Andersson,andS.Sandler.1993. Pretreatmentof fetal porcinepancreasinculture with nicotinamide acceleratesreversal of diabetes after transplantationtonudemice. Surgery (St.Louis).113:205-214.

37.Simpson,A., B.Tuch, and P.Vincent.1991.Characterization of

endo-crine-rich monolayers ofhumanfetalpncreas thatdisplayreduced immunogenic-ity. Diabetes. 40:800-808.

38.Philippe,J.,D.J.Drucker,W. L.Chick,and J. F. Harbener.1987.

Tran-scriptional regulationofgenesencoding insulin,glucagon,andangiotensinogen

bysodium butyratein a ratisletcellline.Mol. Cell. Biol. 7:560-563. 39.Dudek,R.W.,I. E. Lawrence, Jr., R. S.Hill,and R.C. Johnson. 1991. Induction of isletcytodifferentiationby fetal mesenchymein adultpancreatic

ductalepithelium.Diabetes.40:1041-1048.

40. Ueda,K.,and0. Hayaishi. 1985.ADP-ribosylation.Ann.Rev.Biochem.

54:73-100.

41.Pekala, P. H., M. D. Lane, P.A.Watkins, and J.Moss. 1981. On the

mechanism ofpreadipocyte differentiation.J.Biol.Chem.256:4871-4876. 42. Porteous, J. W., H. M. Furneaux, C. K. Pearson, C. M.Lake,andA. Morrison.1979.Poly(adenosine diphosphate ribose) synthetaseactivityin nu-clei ofdividing andnon-dividingbutdifferentiatingintestinalepithelialcells. Biochem.J. 180:455-463.

43. Caplan, A.I., and M. J. Rosenberg. 1975. Interrelationshipbetween

poly(ADP-Rib)synthesis,intracellularNADlevels,andmuscleorcartilage dif-ferentiation from mesodermal cells ofembryonicchick limb.Proc.Natl. Acad. Sci.USA. 72:1852-1857.

44.Ohashi,Y.,K.Ueda,0. Hayaishi,K.Ikai, and0.Niwa. 1984. Induction ofmurine teratocarcinoma cell differentiation bysuppressionof

poly(ADP-ri-bose)synthesis.Proc.Natl.Acad.Sci. USA. 81:7132-7136.

45.Hoshino,J.,G. Beckmann,and H.Kroger. 1992.Sensitivityinvitroof

matureandimmaturemousethymocytestodexamethasonecytotoxicityandits correlationtopoly ADP-ribosylation. Biochem.Int. 27:105-116.

46.Nishio,A.,S.Nakanishi,J.Doull, and E. M.Uyeki. 1983. Enhanced

chondrocyticdifferentiation in chicklimbbud cell culturesbyinhibitors of poly(ADP-ribose) synthetase. Biochem.Biophys.Res. Commun. 111:750-759.

47.Alpert,S.,D.Hanahan,andG.Teitelman. 1988.Hybridinsulingenes revealadevelopmentallineageforpancreaticendocrinecells andimplya

rela-tionshipwithneurons.Cell. 53:295-308.