Notch and Kras reprogram pancreatic acinar cellsto ductal intraepithelial neoplasiaJean-Paul De La Oa, Lyska L. Emersonb, Jessica L. Goodmana, Scott C. Froebea, Benjamin E. Illuma, Andrew B. Curtisa,and L. Charles Murtaugha,1

aDepartment of Human Genetics, University of Utah, 15 North 2030 East, Room 2100, Salt Lake City, UT 84112; and bDepartment of Pathology, Division ofAnatomic Pathology, University of Utah, 1950 Circle of Hope Drive, Room 3860, Salt Lake City, UT 84112

Communicated by Mario R. Capecchi, University of Utah, Salt Lake City, UT, October 9, 2008 (received for review June 23, 2008)

Efforts to model pancreatic cancer in mice have focused on mim-icking genetic changes found in the human disease, particularly theactivating KRAS mutations that occur in pancreatic tumors andtheir putative precursors, pancreatic intraepithelial neoplasia(PanIN). Although activated mouse Kras mutations induce PanINlesions similar to those of human, only a small minority of cells thatexpress mutant Kras go on to form PanINs. The basis for thisselective response is unknown, and it is similarly unknown whatcell types in the mature pancreas actually contribute to PanINs. Oneclue comes from the fact that PanINs, unlike most cells in the adultpancreas, exhibit active Notch signaling. We hypothesize thatNotch, which inhibits differentiation in the embryonic pancreas,contributes to PanIN formation by abrogating the normal differ-entiation program of tumor-initiating cells. Through conditionalexpression in the mouse pancreas, we find dramatic synergybetween activated Notch and Kras in inducing PanIN formation.Furthermore, we find that Kras activation in mature acinar cellsinduces PanIN lesions identical to those seen upon ubiquitous Krasactivation, and that Notch promotes both initiation and dysplasticprogression of these acinar-derived PanINs, albeit short of invasiveadenocarcinoma. At the cellular level, Notch/Kras coactivationpromotes rapid reprogramming of acinar cells to a duct-like phe-notype, providing an explanation for how a characteristicallyductal tumor can arise from nonductal acinar cells.

cancer � pancreas � Ras � PanlN � metaplasia

Pancreatic ductal adenocarcinoma (PDA) is the fourth lead-ing cause of cancer death in the United States, proving fatal

to nearly all who are diagnosed (1). PDA is thought to arise fromductal precursor lesions termed pancreatic intraepithelial neo-plasia (PanIN), which accumulate mutations and become pro-gressively dysplastic, finally forming metastatic tumors. Activa-tion of the KRAS proto-oncogene, which is seen in almost allPDA cases, occurs in many early PanINs and may represent aninitiating event (1).

PDA has been modeled in mouse by use of a Cre-dependentactivated Kras allele, KrasloxP-STOP-loxP-G12D (henceforth,KrasG12D), which induces mouse PanIN lesions (mPanINs) sim-ilar to those seen in humans (2–4). Interestingly, althoughprevious studies used Pdx1Cre or Ptf1aCre drivers to activateKrasG12D throughout the pancreas, mPanINs formed only fo-cally, and most of the organ appeared normal (2, 3). This suggeststhat only a subset of cells in the pancreas can be transformed byactivated Kras, although the basis of this specificity remainsunknown. Additional signaling pathways that may promoteKras-mediated transformation are active in human and mousePanINs, including the Notch pathway (1, 2). Notch is of partic-ular interest given its importance to embryonic pancreas devel-opment (5), and its involvement in phenotypic plasticity of adultexocrine cells (6, 7). In turn, exocrine cell plasticity is likelyrelevant to the long-standing question of where pancreaticcancer originates.

Under certain circumstances, including exposure to the EGFreceptor ligand TGF-�, acinar cells can assume duct-like char-

acteristics (6, 8, 9). Importantly, TGF-�-induced acinar-to-ductconversion requires Notch activity (6). Ras proteins are down-stream effectors of the EGF receptor, and overexpressing acti-vated Kras in acinar cells also causes ductal metaplasia andtumors of mixed acinar/ductal character (10, 11). Overexpressionof activated Kras in duct cells, by contrast, does not producePanINs or other dysplastic lesions (12). Conversion of acinarcells to a duct-like phenotype may therefore be an important firststep in pancreatic tumorigenesis.

This hypothesis receives support from several recent studies,using Cre-dependent endogenous KrasG12D or KrasG12V alleles.First, mPanIN formation in young KrasG12D;Ptf1aCre mice isassociated with acinar-ductal metaplasia, and similar metaplasticlesions can be found associated with human PanINs (13).Second, activation of KrasG12D specifically in acinar precursors,by using NestinCre, produces mPanINs identical to those seenwith Pdx1Cre and Ptf1aCre (14). Finally, and most critically,activation of KrasG12V in adult acinar and centroacinar cells(CACs) leads to acinar-ductal metaplasia and mPanIN forma-tion after experimentally induced pancreatitis (15). Experimen-tal pancreatitis activates the Notch pathway (7, 16, 17), and thispathway also appears to be activated during metaplasia inducedby KrasG12V (15). As this study involved activation in both aciniand CACs, however, it remains unclear which cell type is theactual mPanIN cell of origin. Furthermore, although Notchpromotes acinar-ductal metaplasia in vitro, its contribution topancreatic cancer initiation remains unknown.

Here, we test the hypothesis that Notch activation conferssensitivity to KrasG12D, and that this cooperative interactiondrives conversion of acinar cells to preinvasive PanIN lesions. Wefind that coactivation of Notch and Kras, by using a relativelynonspecific Pdx1CreERT driver, dramatically increases mPanINformation compared with Kras activation alone. Furthermore,we find that fully differentiated adult acinar cells are able to formmPanIN lesions after KrasG12D activation, and that coactivationof Notch accelerates this process. Notch and Kras coactivationcauses rapid reprogramming of acinar cells to a duct-like phe-notype, suggesting an explanation for how precursors of a ductalcancer can originate from nonductal differentiated cells.

ResultsNotch Promotes Kras-Induced mPanIN Formation. To determinewhether Kras and Notch cooperate to induce mPanIN forma-tion, we crossed a Cre-dependent Notch1 gain-of-function trans-gene, Rosa26Notch1IC-IRES-GFP (18) (henceforth, Rosa26NIC), into

Author contributions: J.-P.D.L.O. and L.C.M. designed research; J.-P.D.L.O., J.L.G., S.C.F.,B.E.I., and A.B.C. performed research; J.-P.D.L.O., L.L.E., and L.C.M. analyzed data; andJ.-P.D.L.O. and L.C.M. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

1To whom correspondence should be addressed. Email: murtaugh@genetics.utah.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0810111105/DCSupplemental.

© 2008 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0810111105 PNAS � December 2, 2008 � vol. 105 � no. 48 � 18907–18912

MED

ICA

LSC

IEN

CES

the KrasG12D background. To avoid lethality caused by theabsence of islet differentiation in Rosa26NIC;Pdx1Cre mice (18),we used a tamoxifen-inducible Pdx1CreERT transgene to gen-erate mosaic pancreata (19). On tamoxifen (TM) administrationin utero, a dose-dependent subset of Pdx1� progenitor cellsundergoes recombination, whereas nonrecombined cells con-tribute to normal pancreatic function. Pregnant females fromcrosses involving all three alleles (KrasG12D, Rosa26NIC, andPdx1CreERT) were administered 1.5 mg TM at embryonic day10.5, a dose producing recombination in 25–50% of duct, acinarand islet cells (data not shown).

The resulting offspring were killed at 1 month of ageand subjected to histological analysis (Fig. 1 A–F).Rosa26NIC;Pdx1CreERT pancreata contained undifferentiatedepithelia similar to those observed in Rosa26NIC;Pdx1Cre (18),albeit affecting much less of the organ (Fig. 1 A and D). Asexpected, pancreata of KrasG12D;Pdx1CreERT mice exhibitedfocal mPanIN formation (Fig. 1 B and E), similar to that seen inKrasG12D;Pdx1Cre (2, 3). When both Notch and Kras wereactivated (KrasG12D;Rosa26NIC;Pdx1CreERT), however, nearlythe entire organ was overtaken by PanIN-like epithelium (Fig. 1C and F). These results suggest that by blocking differentiation,activated Notch sensitizes pancreatic progenitor cells to Kras.The severity of the phenotype, however, precluded quantitativeanalysis.

When we analyzed offspring of mothers that did not receiveTM, we found that a low level of TM-independent recombina-tion induced numerous focal mPanIN lesions in KrasG12D;Rosa26NIC;Pdx1CreERT mice (Fig. 1I). By contrast, few or noabnormalities were found in untreated Kras-only or Notch-onlypancreata (Fig. 1 G and H). We counted mPanIN lesions acrossrandom sections from each animal, to derive a PanIN initiationindex (see Materials and Methods). At all ages (1–4 months),

there were significantly more mPanIN lesions inKrasG12D;Rosa26NIC coexpressing mice than mice expressingKrasG12D alone, indicating that Notch activation sensitizes cellsto transformation by Kras (Fig. 1J).

Like mPanINs described previously (2, 3), those found in bothKrasG12D;Pdx1CreERT and KrasG12D;Rosa26NIC;Pdx1CreERTmice express the duct marker cytokeratin-19 (CK19) and pro-duce periodic acid-Schiff (PAS)-reactive mucins (see supportinginformation (SI) Fig. S1 A–F). Rosa26NIC drives coexpression ofnuclear EGFP with Notch1IC (18), and we find that CK19�

mPanIN lesions in KrasG12D;Rosa26NIC;Pdx1CreERT mice ex-press EGFP, indicating activation of Rosa26NIC (Fig. S1 G–I).Although KrasG12D;Rosa26NIC coexpressing mPanINs are oftenlarger than those induced by KrasG12D alone, we find no consis-tent differences in proliferation or apoptosis (the latter of whichis extremely rare) between the different classes of lesions (datanot shown).

Mature Acinar Cells Represent a Cell of Origin for Kras-InducedmPanINs. While analyzing KrasG12D;Pdx1CreERT andKrasG12D;Pdx1Cre pancreata, we noted frequent areas of appar-ent acinar-ductal metaplasia, including cells coexpressing acinarand duct markers, sometimes adjacent to mPanINs (Fig. S2). Aswith mPanINs themselves, these lesions occur more frequentlywhen Rosa26NIC is coexpressed with KrasG12D (Fig. S3A). Similarfindings in KrasG12D;Ptf1aCre mice, and in human specimens,suggest that metaplasia may be an early step in the conversionof acini to PanINs (13).

To determine whether PanINs can arise from differentiatedacinar cells, we used transgenic mice in which the acinar-specificElastase1 promoter drives expression of CreERT. This ElaCreERTline has been extensively documented to induce acinar-specificrecombination, excluding duct or centroacinar cells (9, 20, 21).

E

du

BA

du

D

C

F

IG

du

is

Hdu

+ta

moxi

fen

(1.5

m

g

@

E10.5

)

Kras

(n=6)

Notch

(n=2)

Notch +

Kras

(n=8)

Kras

(n=10)

Notch

(n=7)

Notch

+

Kras

(n=8)

Kras

(n=10)

Notch

(n=9)

Notch

+

Kras

(n=8)

0

100

200

300

400

1 month 2 month 4 month

**

**

(mP

anIN

le

sion

s/m

m2)

x1

00

Age

*Rosa26NIC;

Pdx1CreERT

Rosa26NIC;

KrasG12D; Pdx1CreERT

KrasG12D; Pdx1CreERT

n

o

tam

oxi

fen

J

is

Fig. 1. Notch and Kras synergize to induce mPanIN initiation. (A–I) H&E-stained sections from mice of the indicated genotypes, administered TM in utero (A–F)or left untreated (G–I). Broad mosaic activation of Rosa26NIC, via in utero TM treatment, produces foci of convoluted, undifferentiated epithelium (D, bracket),whereas broad activation of KrasG12D produces isolated mPanIN lesions (E, arrow). Broad coactivation of both alleles results in almost complete replacement ofnormal exocrine tissue by mPanIN-like epithelium (C and F). Focal mosaic activation of NotchIC or Kras alone, in the absence of TM, produces no or fewabnormalities (G and H), whereas focal coactivation of both alleles induces numerous mPanIN lesions (I, arrow). Also indicated are normal ducts (du) and islets(is). (Magnification: A–C 100�, D–F 400�, G–I 200�.) (J) Stripchart of mPanIN initiation indices for individual TM-untreated Pdx1CreERT mice, expressing activatedNotch1 (Notch) and/or KrasG12D (Kras), at indicated ages. P-values (by t test): *, P � 0.05, **, P � 0.01.

18908 � www.pnas.org�cgi�doi�10.1073�pnas.0810111105 De La O et al.

Adult (6-weeks-old) KrasG12D;ElaCreERT mice were given 5 mgTM, inducing recombination in �50% of acinar cells, and killed1, 2, or 10 weeks thereafter (i.e., 7 weeks, 2 months, or 4 monthsof age). Like Pdx1CreERT, ElaCreERT confers a low level ofTM-independent recombination (20); we therefore examinedadditional 2- and 4-month-old mice that did not receive TM.Although most mice did not exhibit gross pancreatic abnormal-ities, regardless of TM treatment, histological analysis revealedthat they often contained mPanIN lesions identical to thoseobserved with general pancreatic Cre drivers (Fig. 2, and seebelow). In total, at least one mPanIN was found in 12/33KrasG12D;ElaCreERT pancreata examined, and the likelihood offinding mPanINs increased with the age of the mouse (P � 0.005by �2 test for trend). Whereas most mPanIN lesions found werelow-grade (mPanIN-1), more advanced mPanIN-2 andmPanIN-3 lesions were found in 6/33 (18%) and 1/33 (3%) ofKrasG12D;ElaCreERT pancreata, respectively. These results in-dicate that mature acinar cells are a potential source of PanINs,and possibly for ‘‘ductal’’ adenocarcinoma as well.

Notch Activation Sensitizes Acinar Cells to Kras-Driven mPanIN For-mation and Progression. Similar to KrasG12D;Pdx1Cre, the vastmajority of acinar cells are unaffected in KrasG12D;ElaCreERTmice, with or without TM treatment. As Notch and Krassynergize after activation by PdxCreERT, we asked whether theyalso synergize in adult acinar cells. In parallel with theKrasG12D;ElaCreERT experiments described above, littermatesalso inheriting Rosa26NIC were administered 5 mg of TM at 6weeks of age, and killed 1, 2, or 10 weeks later. Althoughactivation of Rosa26NIC alone in acini had no visible effect (Fig.3B and see below), it did induce robust expression of the Notchtarget genes Hes1 and HeyL (Fig. S4). By contrast, coactivationof Rosa26NIC and KrasG12D caused widespread mPanIN forma-tion, exceeding that seen with KrasG12D alone (Fig. 3 C–E).Similar synergy was observed with respect to acinar-ductalmetaplasia, for which Notch alone was also ineffective (Fig.S3B). Notch/Kras coactivation did not result in greater inductionof Hes1 or HeyL than Notch alone (Fig. S4), suggesting that thesynergy between these pathways is unlikely to occur at the levelof immediate Notch target genes.

The same relationship was observed in TM-untreated mice(Fig. S5 A and B). For reasons as yet unknown, we found thatuntreated KrasG12D;ElaCreERT mice exhibited more mPanINsthan TM-treated mice at the same age, although the synergy withRosa26NIC remains dramatic and robust. By contrast, the appar-ently greater number of mPanINs seen in untreatedKrasG12D;Rosa26NIC-expressing mice, compared to those receiv-ing TM (Fig. 3E and Fig. S5A), likely reflects the fact thatuntreated pancreata retain more normal exocrine tissue (Fig. S5C and D). Individual lesions are therefore more easily scored

A B

C D

PAS CK19

Fig. 2. Acinar-specific activation of KrasG12D induces mPanIN lesions similarto those induced with general pancreatic Cre drivers. (A and B) H&E-stainingof mPanIN lesions in KrasG12D;ElaCreERT mice. Both low-grade mPanIN-1lesions are found, with columnar cytoplasm and basally located nuclei (A), aswell as rarer high-grade lesions, including an mPanIN-3 lesion with nuclearatypia and luminal budding (B). (C) PAS staining (magenta) of acinar-derivedmPanIN. (D) Cytokeratin-19 staining (brown) of acinar-derived mPanIN. (Mag-nification: 400�.)

Notch

(n=9

)

Kra

s (n

=8

)

Notch

+

Kras

(n=4

)

Notch

(n=7

)

Kra

s (n

=7

)

Notch

+

Kras

(n=4

)

Notch

(n-5)

Kra

s (n

=7

)

Notch

+

Kras

(n=7

)

(mP

anIN

le

sion

s/m

m2

)

x10

0

0

100

200

300 *

***1 week 2 weeks 10 weeks

Time post-tamoxifenE

wildtypeRosa26NIC; ElaCreERT

Rosa26NIC;

KrasG12D; ElaCreERT

KrasG12D; ElaCreERT

B

DC

A

Fig. 3. Notch and Kras synergize to induce mPanINs from acinar cells. (A–D)Representative H&E-stained sections from pancreata of the indicated geno-types, 10 weeks postTM administration. Acinar-specific NotchIC expression hasno detectable effect (B), whereas acinar-specific KrasG12D activation inducesfocal mPanINs (C). Coactivation of the alleles induces many mPanIN lesions,replacing most of the normal exocrine tissue (D). (Magnification: 200�.) (E)Stripchart indicating mPanIN initiation indices of individual ElaCreERT mice,expressing NotchIC and/or KrasG12D, at indicated timepoints following tamox-ifen treatment. P-values: *, P � 0.05; ***, P � 0.0005.

De La O et al. PNAS � December 2, 2008 � vol. 105 � no. 48 � 18909

MED

ICA

LSC

IEN

CES

than in TM-treated mice, where the lack of normal tissue makesseparate lesions hard to distinguish.

Previous studies of KrasG12D;Pdx1Cre mice indicate that higher-grade mPanIN-2 and mPanIN-3 lesions are much rarer thanmPanIN-1, although they become more common as the overalllesion burden increases with age (2). To determine whether Notchaffects this process, we developed an mPanIN progression metricbased on observation by a blinded pathologist (L.E.), and weightedaccording to increasing lesion stage (see Materials and Methods). Inevery ElaCreERT experimental group (i.e., separated by age andTM treatment), the progression index was significantly higher withNotch and Kras together than with activated Kras alone (P � 0.05by Wilcoxon test). Although this suggested that Notch promotesmPanIN progression, it was also possible that, by enhancing mPa-nIN initiation, Notch activation would make it likelier to encounterrare advanced lesions even if the progression rate were unchanged.This would predict that the progression index should correlate withoverall mPanIN number, irregardless of Rosa26NIC status. As theindividual experimental groups were too small to model separatelythe effects of lesion number and genotype, we pooled all ElaCre-ERT samples that contained at least one mPanIN and compared theeffects of overall mPanIN number and genotype on progressionindex (Fig. S6). We developed a general linear model, in whichprogression index was influenced separately by mPanIN numberand Rosa26NIC genotype (see Materials and Methods), and foundthat Rosa26NIC activation significantly correlated with increasedprogression (P � 5 � 10�6, by F-test). Similar results were obtainedby analysis of our Pdx1CreERT samples, either separately or pooledwith ElaCreERT, suggesting that Notch promotes both mPanINinitiation and progression.

Formation of PanINs from Acinar Cells via Reprogramming to a DuctalPhenotype. PanIN and PDA, in humans and mice, typically looklike ducts and express duct-specific markers. Nevertheless, ourresults indicate that mPanINs can arise from mature acinar cells,implying that their formation involves a switch from acinar toductal differentiation programs. To follow this switch, we ex-amined ElaCreERT mice two weeks following tamoxifen treat-ment, at which time Kras activation alone has had little or nodetectable effect, whereas Notch/Kras coactivation has inducedwidespread metaplasia (Fig. 3E and Fig. S3B). As GFP expres-

sion marks cells that have undergone Rosa26NIC activation (18),we performed coimmunofluorescence to monitor expression ofacinar and ductal markers in cells expressing activated Notch. Asexpected, Cre-negative pancreata exhibit no detectable GFP(Fig. 4 A–D and M–P). Also as expected, TM-treatedRosa26NIC;ElaCreERT pancreata exhibit widespread Rosa26NIC

activation, indicated by GFP expression (Fig. 4 E and Q).Nonetheless, these cells retain a normal acinar differentiationprogram: they express the digestive enzyme amylase and thetranscription factor Ptf1a (Figs. 4 F and R), a regulator of acinargene expression (5), and they are negative for the duct markerCK19 (Fig. 4 G and S). Therefore, Notch activation alone doesnot detectably perturb the differentiation state of mature acinarcells. Following coactivation of Notch and Kras, however, nearlyall GFP-expressing cells undergo reprogramming to a duct-likephenotype, including loss of amylase and Ptf1a expression (Fig.4 J and V) and up-regulation of CK19 (Fig. 4 K and W). A similarreprogramming event must occur in KrasG12D;ElaCreERT mice,as the PanIN lesions that form in these pancreata are positive forCK19 (Fig. 2D) and negative for amylase and Ptf1a (data notshown), but the efficiency of conversion is much higher whenNotch is activated together with Kras. These data implicate thereprogramming of acinar cells, including down-regulation of theacinar regulator Ptf1a, in the formation of ductal lesions andtumors from a nonductal cell type.

DiscussionNotch is a critical negative regulator of progenitor cell differ-entiation in the embryonic pancreas, and its up-regulation inPanINs and pancreatic cancer suggests that it represses thedifferentiation of tumor-initiating cells (2, 6, 13). This possibilityis particularly compelling given evidence that the ductal pheno-type of PDA might not reflect a strictly ductal origin. Forexample, whereas overexpression of activated Kras in acinar cellsleads to tumor formation, tumors are not observed on Krasoverexpression in ducts (10–12). In addition, activation of en-dogenous KrasG12D in acinar-restricted precursor cells, viaNestinCre, results in mPanIN formation similar to that observedwith more widespread pancreatic Cre drivers such as Pdx1Creand Ptf1aCre (14). These previous experiments, however, do notaddress the question of whether mature acinar cells can give rise

L

D

H

GFP

A B C

E F G

I J K X

T

P

Ro

sa

26

NIC

; K

ras

G1

2D

; E

la-C

reE

RT

R

osa

26

NIC

; E

laC

reE

RT

R

osa

26

NIC

; K

ras

G1

2D

M N O

Q R S

U V W

Ro

sa

26

NIC

; K

ras

G1

2D

; E

la-C

reE

RT

R

osa

26

NIC

; E

laC

reE

RT

R

osa

26

NIC

; K

ras

G1

2D

amylase CK19 merge GFP Ptf1a CK19 merge

Fig. 4. Notch and Kras synergize to induce rapid acinar-to-ductal reprogramming in vivo. Adult (6-week-old) mice of the indicated genotypes were treatedwith tamoxifen and analyzed two weeks later. (A–L) Immunofluorescence staining for GFP (green, indicating activation of Rosa26NIC), the acinar marker amylase(red) and the duct marker CK19 (white). Activation of Notch in acinar cells results in no detectable changes (E–H, note green nuclei in amylase� acinar cells,negative for CK19), whereas Notch/Kras coactivation results in ductal reprogramming specifically of GFP-positive cells, as indicated by loss of amylase andup-regulation of CK19 (I–L). (M–X) Immunofluorescence staining for GFP (green), the acinar regulatory transcription factor Ptf1a (red) and CK19 (white) providesfurther evidence for Notch/Kras-driven reprogramming, as NotchIC expression alone does not result in loss of Ptf1a expression (Q–T), whereas NotchIC/KrasG12D

coactivation down-regulates Ptf1a although inducing CK19 (U–X). (Magnification: 400�.)

18910 � www.pnas.org�cgi�doi�10.1073�pnas.0810111105 De La O et al.

to PanINs, as all of them involved KrasG12D activation in uteroor shortly after birth. Here, we have addressed both the contri-bution of Notch signaling to the initiation of pancreatic tumor-igenesis, and the possibility that the acinar cell represents asource for precancerous lesions in the adult.

By using Pdx1CreERT to activate recombination in progenitorcells, we find that Notch dramatically sensitizes cells to KrasG12D-driven mPanIN initiation. In addition to mPanINs, we findmetaplastic lesions induced by KrasG12D that may representinitial stages of acinar reprogramming to a ‘‘pre-PanIN’’ state.Furthermore, activation of Kras alone in adult acinar cellsinduces both metaplastic lesions and mPanINs, including high-grade mPanIN-2 and mPanIN-3, suggesting that acini can serveas a cell of origin for PDA. This process is dramatically enhancedwhen Notch is coactivated with Kras, producing rapid andwidespread reprogramming of acinar cells to a duct-like pheno-type, and apparently increased progression to high-grade lesions.Together, these results suggest that the observed up-regulationof Notch components in pancreatic cancer is not an epiphenom-enon, but reflects a supporting role for Notch in Kras-inducedreprogramming and transformation. Both positive and negativeinteractions between Notch and Ras have been described inother systems, and Notch can promote or inhibit tumorigenesisdepending on cellular context (22). In cultured human cells, Rasappears to activate Notch, and this activation is required fortransformation (23). In our model, by contrast, we find that Krasactivation alone does not induce Notch target genes in thepancreas (Fig. S4), suggesting that the pathways act convergentlyrather than linearly. Understanding the mechanism of thisconvergence may identify treatment targets for PDA.

Although several lines of evidence suggest that PanINs rep-resent the early stages of human PDA (1), it remains a formalpossibility that they instead represent ‘‘dead ends’’ in the geneticpathway that culminates in PDA. Similarly, although our dataand those of others (13) suggest that metaplastic acini are theprecursor to mPanIN lesions, we cannot yet rule out the possi-bility that overtly metaplastic structures, as described here, formindependently of mPanINs. Further analysis of the kinetics oflesion formation, preferably by using a Cre driver that exhibitslittle or no TM-independent recombination, should indicatewhether metaplasia clearly precedes mPanIN formation. Directproofs of metaplasia-to-mPanIN and mPanIN-to-PDA progres-sion will require more sophisticated lineage-tracing tools thancurrently exist.

Although our results suggest a model to unify diverse obser-vations in the literature, we note some potential discrepancies.Although Rosa26NIC activation in utero blocks acinar differen-tiation (18), we find that activation in mature acini is notsufficient for ductal reprogramming. In culture, however, Notchactivation is both necessary and sufficient for acinar-ductalreprogramming (6). These in vitro studies used adenoviralvectors to express activated Notch1, which may result in higherexpression than our Rosa26 knock-in strategy; alternatively,placing acinar cells in culture may enhance endogenous Krassignaling. In addition, whereas we show that KrasG12D activationin acinar cells is sufficient for at least a small number of mPanINsto form, Guerra et al. (15) find that mPanINs do not form at allon acinar activation of an endogenous KrasG12V allele, unless thepancreas is subjected to caerulein-induced pancreatitis. Thisgroup used a tetracycline-regulated system to direct acinar-specific recombination, although ours is tamoxifen-regulated,but given that 6/11 TM-untreated KrasG12D;ElaCreERT micedeveloped at least one mPanIN lesion in our study (versus 6/14of TM-treated mice at the same ages), the use of tamoxifencannot explain the discrepancy. A more plausible explanation isthat differences in the targeted Kras alleles affect their expres-sion or function, as KrasG12D has been shown to induce trans-

formation in lung under circumstances in which KrasG12V ap-pears less active (24, 25).

Nonetheless, chronic pancreatitis is an important risk factorfor human pancreatic cancer (1), and the fact that caeruleinpancreatitis sensitizes the mouse pancreas to activated Kras islikely relevant to our model. Caerulein pancreatitis promotesacinar-ductal metaplasia, and is associated with Notch pathwayup-regulation (7, 16, 17). As our study leaves open the questionof how endogenous Notch becomes activated in mPanINs in-duced by Kras alone, further investigation of the links betweenpancreatitis, Notch activation and PanIN initiation are clearlywarranted. Our work also leaves open the question of whetherNotch signaling is necessary, as well as sufficient, for KrasG12D-induced mPanIN initiation and progression. Finally, as our studydesign focused on relatively short-term effects of Notch andKras, we cannot determine whether the synergy between thesepathways extends to the formation of fully invasive tumors. AgedKrasG12D;Pdx1Cre mice do exhibit rare PDA (2), and we hypoth-esize that studies of older mice in our cohorts will reveal acontribution of Rosa26NIC to invasive tumor formation.

A final issue left unresolved by our work is whether acinar cellsrepresent the only cell of origin for mPanINs. In the uninjuredpancreas, the Notch target gene Hes1 is expressed primarily bycentroacinar cells (CACs), terminal cells of the duct network, andthese have also been proposed as cells of origin for PanINs andPDA (1, 6). Evidence for this derives from pancreas-specificdeletion of the Pten tumor suppressor, which causes expansion andtransformation of Hes1-expressing CACs, without acinar-ductalreprogramming (21). Importantly, no ductal lesions formed whenPten was deleted exclusively in acinar cells (using the same ElaCre-ERT driver used here). Although PTEN loss-of-function mutationshave not been reported in human pancreatic cancer, amplificationof AKT2, encoding a kinase epistatic to PTEN, is found in a minorityof PDAs (1). It is tempting to speculate that these tumors may arisefrom a different lineage, possibly through distinct mechanisms, thanthe acinar-to-ductal pathway identified here. Testing whether adultCAC or ductal cells are competent to form PanINs, on activationof Kras and/or Akt signaling, will require the development oftemporally regulated Cre drivers specific to these compartments, animportant priority for the field. In the meantime, our work allowsus to conclude that acinar cells, via Notch-mediated ductal repro-gramming, represent at least one potential major source for Kras-induced lesions. Future studies should elucidate the mechanism of,and requirement for, interaction between these critical signalingpathways.

Materials and MethodsMice. Pdx1CreERT (19), ElaCreERT (9, 20, 21), KrasG12D (2), and Rosa26NIC (18)mice have been described previously, and were genotyped by PCR on tailbiopsy DNA. As neither KrasG12D nor Rosa26NIC is expressed in the absence ofCre, Cre-negative littermates from these crosses served as wild-type controls.Tamoxifen (TM) was dissolved in corn oil and administered by oral gavage. Allexperiments were carried out according to institutional guidelines.

Tissue Processing and Histology. After euthanasia, pancreata were dissected,cut into 5–6 fragments, and processed to frozen or paraffin sections asdescribed (20). PAS staining was performed according to the manufacturer’sinstructions (Sigma). Primary antibodies included rat anti-cytokeratin 19 1:50(Developmental Studies Hybridoma Bank), sheep anti-amylase 1:2,500 (Bio-Genesis), rabbit anti-GFP 1:5,000 (Abcam), and guinea pig anti-Ptf1a 1:5,000(gift of Dr. Jane Johnson, U.T. Southwestern), and secondary antibodies werepurchased from Jackson Immunoresearch. Paraffin sections were subjected tohigh temperature antigen retrieval (Vector Unmasking Solution) before add-ing primary antibody. Vectastain reagents and Vector diaminobenzidine sub-strate were used for immunohistochemistry.

Lesion Scoring. Paraffin sections were cut from two independent dorsal and oneventral fragment of each pancreas. After H&E staining of a single section fromeach fragment, and photomicrography/photomerging (Adobe Photoshop), thearea of the section was determined with ImageJ software (NIH). mPanIN lesions

De La O et al. PNAS � December 2, 2008 � vol. 105 � no. 48 � 18911

MED

ICA

LSC

IEN

CES

were then tallied under the microscope, and the PanIN initiation index wasdefinedas thenumberof lesionspermm2,multipliedby100,andaveragedacrossthe three sections scored for each mouse. Metaplastic lesions were scored simi-larly. In severely affected areas, lesions were scored as separate only if theyappeared in anatomically distinct lobules. The same slides were subsequentlyevaluated, blindly, by a surgical pathologist (L.E.), who scored mPanIN lesions as1A, 1B, 2, or 3 (4), and recorded the five highest-grade lesions per animal. ThePanINprogressionindexwasderivedfromthisdataas:1� (#mPanIN-1A/1B)�2�(# mPanIN-2) � 3� (# mPanIN-3), resulting in scores from 0 (no mPanIN lesions atall) to 15 (five mPanIN-3 lesions found). Any score above 5 must reflect thepresence of at least one high-grade mPanIN-2 or mPanIN-3 lesion.

RNA Isolation and Quantitative RT-PCR. RT-PCR was performed on ElaCreERTmice 7 days after TM treatment, at which point minimal gross changes wereevident (Fig. 3E and data not shown). RNA was isolated from approximatelyhalf of the dorsal pancreas by guanidinium isothiocyanate (26). RealtimeRT-PCR was performed on an ABI 7900HT instrument (Applied Biosystems), byusing SYBR Green. Quantification was done by ��Ct method, normalizing tothe housekeeping gene Hprt1. Primers are listed in Table S1, and were selectedfrom the PrimerBank database (27).

Statistics. All statistical analysis was performed with the R software package(28). Comparisons of PanIN initiation indices were performed by Welch’s t test(i.e., not assuming equal variances), with Holm’s correction for multiplecomparisons. Realtime PCR expression levels were compared to wildtype byANOVA. PanIN progression indices were modeled as a function of PanINinitiation index (continuous) and Rosa26NIC genotype (categorical), by using ageneral linear model with quasipoisson error distribution, and the contribu-tions of the variables were evaluated by F-test.

ACKNOWLEDGMENTS. We thank Doug Melton (Harvard University, Cam-bridge, MA), Tyler Jacks (Massachusetts Institute of Technology, Cambridge,MA), and Jane Johnson (University of Texas Southwestern Medical Center,Dallas, TX) for providing reagents, Steve Leach for helpful discussions, andKristen Kwan and Daniel Kopinke for comments on the manuscript. This workwas supported by grants to L.C.M. from the Lustgarten Foundation for Pan-creatic Cancer Research (RFP06-059), Searle Scholars Program (06-B-116), andNational Cancer Institute (R21-CA123066), and by a National Cancer InstituteCancer Center Support grant (P30-CA042014) to the Huntsman Cancer Insti-tute. J.-P.D.L.O. is supported by National Institutes of Health DevelopmentalBiology Training Grant 5T32-HD07491.

1. Maitra A, Hruban RH (2008) Pancreatic cancer. Annu Rev Pathol 3:157–188.2. Hingorani SR, et al. (2003) Preinvasive and invasive ductal pancreatic cancer and its

early detection in the mouse. Cancer Cell 4:437–450.3. Aguirre AJ, et al. (2003) Activated Kras and Ink4a/Arf deficiency cooperate to produce

metastatic pancreatic ductal adenocarcinoma. Gene Dev 17:3112–3126.4. Hruban RH, et al. (2006) Pathology of genetically engineered mouse models of pancreatic

exocrine cancer: Consensus report and recommendations. Cancer Res 66:95–106.5. Jensen J (2004) Gene regulatory factors in pancreatic development. Dev Dyn 229:176–

200.6. Miyamoto Y, et al. (2003) Notch mediates TGF alpha-induced changes in epithelial

differentiation during pancreatic tumorigenesis. Cancer Cell 3:565–576.7. Siveke JT, et al. (2008) Notch signaling is required for exocrine regeneration after acute

pancreatitis. Gastroenterology 134:544–555.8. Means AL, et al. (2005) Pancreatic epithelial plasticity mediated by acinar cell trans-

differentiation and generation of nestin-positive intermediates. Development132:3767–3776.

9. Strobel O, et al. (2007) In vivo lineage tracing defines the role of acinar-to-ductal trans-differentiation in inflammatory ductal metaplasia. Gastroenterology 133:1999–2009.

10. Grippo PJ, Nowlin PS, Demeure MJ, Longnecker DS, Sandgren EP (2003) Preinvasivepancreatic neoplasia of ductal phenotype induced by acinar cell targeting of mutantKras in transgenic mice. Cancer Res 63:2016–2019.

11. Tuveson DA, et al. (2006) Mist1-KrasG12D knock-in mice develop mixed differentiationmetastatic exocrine pancreatic carcinoma and hepatocellular carcinoma. Cancer Res66:242–247.

12. Brembeck FH, et al. (2003) The mutant K-ras oncogene causes pancreatic periductallymphocytic infiltration and gastric mucous neck cell hyperplasia in transgenic mice.Cancer Res 63:2005–2009.

13. Zhu L, Shi G, Schmidt CM, Hruban RH, Konieczny SF (2007) Acinar cells contribute to themolecularheterogeneityofpancreatic intraepithelialneoplasia.AmJPathol171:263–273.

14. Carriere C, Seeley ES, Goetze T, Longnecker DS, Korc M (2007) The Nestin progenitorlineage is the compartment of origin for pancreatic intraepithelial neoplasia. Proc NatlAcad Sci USA 104:4437–4442.

15. Guerra C, et al. (2007) Chronic pancreatitis is essential for induction of pancreatic ductaladenocarcinoma by K-Ras oncogenes in adult mice. Cancer Cell 11:291–302.

16. Gomez G, Englander EW, Wang G, Greeley GH, Jr (2004) Increased expression ofhypoxia-inducible factor-1alpha, p48, and the Notch signaling cascade during acutepancreatitis in mice. Pancreas 28:58–64.

17. Jensen JN, Cameron E, Garay MV, Starkey TW, Gianani R, Jensen J (2005) Recapitulationof elements of embryonic development in adult mouse pancreatic regeneration.Gastroenterology 128:728–741.

18. Murtaugh LC, Stanger BZ, Kwan KM, Melton DA (2003) Notch signaling controlsmultiple steps of pancreatic differentiation. Proc Natl Acad Sci USA 100:14920 –14925.

19. Gu G, Dubauskaite J, Melton DA (2002) Direct evidence for the pancreatic lineage:NGN3� cells are islet progenitors and are distinct from duct progenitors. Development129:2447–2457.

20. Murtaugh LC, Law AC, Dor Y, Melton DA (2005) Beta-Catenin is essential for pancreaticacinar but not islet development. Development 132:4663–4674.

21. Stanger BZ, et al. (2005) Pten constrains centroacinar cell expansion and malignanttransformation in the pancreas. Cancer Cell 8:185–195.

22. Sundaram MV (2005) The love-hate relationship between Ras and Notch. Gene Dev19:1825–1839.

23. Weijzen S, et al. (2002) Activation of Notch-1 signaling maintains the neoplasticphenotype in human Ras-transformed cells. Nat Med 8:979–986.

24. Guerra C, et al. (2003) Tumor induction by an endogenous K-ras oncogene is highlydependent on cellular context. Cancer Cell 4:111–120.

25. Jackson EL, et al. (2001) Analysis of lung tumor initiation and progression usingconditional expression of oncogenic K-ras. Gene Dev 15:3243–3248.

26. MacDonald RJ, Swift GH, Przybyla AE, Chirgwin JM (1987) Isolation of RNA usingguanidinium salts. Methods Enzymol 152:219–227.

27. Wang X, Seed B (2003) A PCR primer bank for quantitative gene expression analysis.Nucleic Acids Res 31:e154.

28. Crawley, MJ (2007) The R Book (Wiley, Chichester, England; Hoboken, N.J.).

18912 � www.pnas.org�cgi�doi�10.1073�pnas.0810111105 De La O et al.