Lineage tracing reveals the dynamic contribution of Hes1⁺ cells to the developing and adult pancreas

The vertebrate pancreas comprises three major cell types: endocrine islets, which include insulin-producing β-cells; and a network of exocrine acinar and duct cells, which are responsible for producing and transporting digestive enzymes, respectively. The Notch signaling pathway has been implicated in several aspects of pancreatic cell fate determination, beginning with the finding that mouse embryos lacking various Notch components, including the downstream target gene Hes1, exhibit overproduction of endocrine cells (Apelqvist et al., 1999; Jensen et al., 2000). Hes1 can repress the promoter of Neurog3, a crucial pro-endocrine transcription factor, and de-repression of Neurog3 in the absence of Hes1 may drive excessive endocrine differentiation (Apelqvist et al., 1999; Jensen et al., 2000; Lee et al., 2001). In gain-of-function experiments, Notch also inhibits exocrine acinar cell development, promoting instead progenitor maintenance (Esni et al., 2004; Hald et al., 2003; Murtaugh et al., 2003). These findings are corroborated by studies in zebrafish (Esni et al., 2004; Yee et al., 2005; Zecchin et al., 2006), and conform to a generic conception of Notch as regulating cell fate throughout animal development (Lai, 2004).

The Notch pathway knockout phenotypes implied that the early pancreas comprised multipotent cells, the differentiation of which was held in check by Notch signaling (Apelqvist et al., 1999; Jensen et al., 2000). Lineage-tracing studies suggest that multipotent progenitors reside in the `tips' of the embryonic pancreatic epithelium, the expansion of which leaves behind `trunks' that give rise to ducts and islets (Kopinke and Murtaugh, 2010; Solar et al., 2009; Zhou et al., 2007). How Notch regulates this process is unknown, although it may signal through Hes1 to repress Neurog3 (Lee et al., 2001) and control the balance of duct and islet differentiation. Contradicting this model, however, deletion of Notch1 and Notch2, the major receptors expressed in the pancreas, has little effect on late embryonic islet development (Nakhai et al., 2008).

Whether progenitor cells persist in the adult pancreas, particularly for insulin-producing β-cells, remains controversial. Two lineage tracing approaches have been taken to address this issue: `pulse-chase' labeling of mature islet cells, to detect changes in labeling frequency caused by differentiation of new β-cells (neogenesis); or marking acini and/or ducts, to determine whether they can contribute to β-cells. The former studies argue against β-cell neogenesis (Dor et al., 2004), and the latter indicate that neogenesis is either non-existent (Desai et al., 2007; Kopinke and Murtaugh, 2010; Solar et al., 2009) or rare (Inada et al., 2008) in the uninjured pancreas. Pancreatic injury, in particular caused by ligation of the main duct, has been proposed to induce facultative neogenesis from acinar or duct cells (Wang et al., 1995; Xu et al., 2008), for which exist both contradictory and supporting lineage tracing data (Inada et al., 2008; Solar et al., 2009). As previous approaches used mature duct or acinar marker genes to drive Cre expression, however, they may have excluded specialized adult progenitor cells. These might include centroacinar cells, terminal elements of the exocrine ductal network in which Notch-Hes1 signaling appears particularly high (Miyamoto et al., 2003; Parsons et al., 2009; Stanger et al., 2005). These cells have been suggested to generate new β-cells following injury (Hayashi et al., 2003; Nagasao et al., 2003), and they can give rise to both acinar and islet cells following isolation and culture (Rovira et al., 2010).

To understand how and when Notch-Hes1 signaling regulates pancreatic progenitor cells, we generated `knock-in' mice in which the tamoxifen-dependent CreERT2 recombinase is targeted to the Hes1 locus. With these mice, we have analyzed the stage-specific differentiation potential of Notch-responsive cells in the embryonic pancreas, revealing a novel shift from multipotent to exocrine-restricted progenitor cells. This parallels a shift in the cellular response to Notch, from arresting differentiation to promoting duct cell specification. In the adult, we find that Hes1⁺ duct and centroacinar cells appear to be fixed in their fate, and do not detectably contribute to β-cells, even after duct ligation injury. Ours is the first study to address the fate of Notch-responsive cells in any adult tissue, and supports an emerging model that lineage boundaries in the pancreas are normally fixed at birth.

We used bacterial recombineering (Liu et al., 2003) to generate a Hes1^CreERT2-neoR targeting vector, in which most of the Hes1 open reading frame, including the bHLH domain, is replaced by that of CreERT2 (Feil et al., 1997), linked to an FRT-flanked neo^R cassette (see Fig. S1A in the supplementary material). This was electroporated into R1 ES cells (Nagy et al., 1993), generously provided by Mario Capecchi (University of Utah, USA), and G418-resistant ES cell clones were screened by Southern blotting and PCR (see Fig. S1B in the supplementary material and data not shown). Germline chimeras were derived by the University of Utah Transgenic Core Facility. The neo^R cassette was excised in vivo by breeding to Rosa26^FLPe (Farley et al., 2000), obtained from the Jackson Laboratory. Cre reporter mice R26R^EYFP (Srinivas et al., 2001) and R26R^LacZ (Soriano, 1999) were obtained from the Jackson Laboratory. Rosa26^NIC (Murtaugh et al., 2003) and Pdx1Cre mice (Gu et al., 2002) were provided by Doug Melton (Harvard University, MA, USA). Ctnnb1^lox(ex3) mice (Harada et al., 1999) were provided by Makoto Mark Taketo (Kyoto University, Japan). Tamoxifen (Sigma T-5648) was dissolved in corn oil and administered by oral gavage.

Explant cultures were established from E11.5 dorsal pancreatic buds, with embryos genotyped by PCR immediately after dissection. Buds were cultured at the air-medium interface, on Millicell-CM filters (0.4 μm pore size; Millipore), in DMEM supplemented with 10% FBS and antibiotics. As indicated, explants were treated with the γ-secretase inhibitor DAPT (10 μM final; Calbiochem) to inhibit Notch signaling (Dovey et al., 2001), or with vehicle alone as a control (0.1% DMSO final), 24 hours prior to treatment with 250 nM 4-hydroxytamoxifen to activate CreERT2.

Pancreatic duct ligations were performed as described previously (Scoggins et al., 2000; Solar et al., 2009; Xu et al., 2008), following a protocol approved by the University of Utah IACUC. Briefly, following midline laparotomy, the stomach and ascending colon were deflected to access the dorsal pancreas. The dorsal pancreas was lifted and stretched slightly with a surgical probe, and a monofilament ligature placed near the base, sparing the splenic vessels. Surgery was performed under isoflurane anesthesia, and mice received an analgesic dose of buprenorphine (50 μg/kg body weight) immediately after surgery.

Tissues were fixed and immunostained essentially as described previously (Kopinke and Murtaugh, 2010). Immunofluorescence was performed on frozen sections (7-8 μm) of tissue fixed in 4% paraformaldehyde/PBS (4°C, 1-2 hours). Other analyses used paraffin sections (6 μm) of tissue fixed in zinc-buffered formalin (room temperature, overnight). Primary antibodies used in this study are listed in Table 1; where indicated, sections were also stained with Dolichos biflorus agglutinin (DBA) lectin (Vector Laboratories), which marks duct cells (Kobayashi et al., 2002). Staining was analyzed by compound fluorescent or light microscopy, using MicroSuite software (Olympus). Photomicrographic images were processed using Adobe Photoshop, with parallel images processed identically.

Our lineage analysis approach, following that of others (Gu et al., 2002; Solar et al., 2009; Zhou et al., 2007), is schematized in Fig. S2 in the supplementary material. One or two randomly chosen fields, comprising 800-1000 DAPI⁺ cells each, were photographed from each of four to ten widely separated sections. For studies of E17.5 embryonic pancreata, we collected six to eight sections per slide, separated by 80-100 μm and spanning the entire pancreas as well as adjacent stomach and duodenum. For uninjured adult pancreata, we embedded and sectioned one half of each dorsal pancreas, collecting four or five sections per slide spaced at least 100 μm apart. For duct ligations, we sectioned the entire dorsal pancreas (ligated and unligated regions), collecting eight to ten sections per slide separated by at least 120 μm. Total numbers of mice, fields and cells scored in each experiment are listed in Tables S2 and S3 in the supplementary material.

Labeling efficiencies were derived from co-immunofluorescence, using the Analyze Particles function of ImageJ (NIH) (Kopinke and Murtaugh, 2010). Counting accuracy was confirmed by eye in Adobe Photoshop for random samples. To exclude the possibility that cells were scored incorrectly due to poor resolving power of the compound microscope, two samples were scored according to the same criteria but using optical sections (less than 1 μm thick) generated by confocal microscopy (Fig. 3). Calculations and graphs were generated with Microsoft Excel and R (www.r-project.org).

To extend previous studies of Hes1 expression (Apelqvist et al., 1999; Esni et al., 2004; Jensen et al., 2000; Lammert et al., 2000), we performed Hes1 immunostaining on pancreata of different embryonic stages, spanning major developmental transitions. At E11.5, when most cells are undifferentiated progenitors, we found widespread but non-uniform expression of Hes1 in the epithelium (Fig. 1A), consistent with previous immunostaining and in situ hybridization studies (Jensen et al., 2000; Lammert et al., 2000; Murtaugh et al., 2005; Nakhai et al., 2008). We observed Hes1 downregulation from E13.5, the onset of the `secondary transition' wave of acinar and β-cell differentiation (Gittes, 2009; Pictet and Rutter, 1972). Hes1 was initially mosaic throughout the epithelium (Fig. 1B), partly overlapping with the `tip cell' marker carboxypeptidase A1 (Cpa1) (see Fig. S3A in the supplementary material) (Zhou et al., 2007), but became increasingly confined to Cpa1-negative ductal and centroacinar cells at later stages (Fig. 1C-D; Fig. S3B,C in the supplementary material). A similar restriction of Hes1 from tip cells has been seen previously (Esni et al., 2004), and the late ductal localization prefigures its expression in the adult (Miyamoto et al., 2003; Stanger et al., 2005). Antibody specificity was indicated by the lack of staining in Hes1-deficient embryos (generated as described below) (see Fig. S3D,E in the supplementary material). Together, these data unify previous studies, and support the hypothesis that Hes1 expression marks early pancreatic progenitors (Esni et al., 2004; Jensen et al., 2000).

In order to follow the fate of Hes1⁺ cells, we engineered a Hes1^CreERT2 allele (henceforth, Hes1^C2) by replacing most of the coding region with CreERT2, a tamoxifen-inducible recombinase (Feil et al., 1997) (Fig. 1E; Fig. S1 in the supplementary material). Hes1^C2/+ animals were viable and fertile, while most Hes1^C2/C2 embryos died between E12.5 and E13.5 (data not shown), as described for Hes1 knockouts (Ishibashi et al., 1995).

To confirm that Hes1^C2 was active in Hes1-expressing cells, we immunostained for Hes1, Cre and lacZ in E12.5 embryos double-transgenic for Hes1^C2 and the lineage reporter R26R^LacZ (Soriano, 1999), which had received tamoxifen (TM) by maternal gavage at E9.5. We observed close overlap between Cre and Hes1 (see Fig. S4A-F in the supplementary material), indicating that Hes1^C2 recapitulates endogenous Hes1 expression. Furthermore, the lacZ lineage marker was widely expressed in the pancreatic epithelium, including by numerous cells that continue to express Hes1 (see Fig. S4G-I, arrowheads, in the supplementary material). In organ cultures, 4-hydroxytamoxifen treatment of dorsal pancreatic buds from E11.5 embryos double-transgenic for Hes1^C2 and the lineage reporter R26R^EYFP (Srinivas et al., 2001) induced widespread epithelial EYFP expression. Labeling was almost abolished by pre-treatment with the γ-secretase inhibitor DAPT (Dovey et al., 2001) (see Fig. S5 in the supplementary material), indicating that Hes1^C2 expression requires Notch activity. Together, these results indicate that Hes1^C2 is expressed and regulated in the same way as endogenous Hes1, and that Hes1^C2 can be used to follow the fate of Notch-responsive cells.

We performed additional crosses between Hes1^C2 and R26R^LacZ, administered a single 2 mg tamoxifen dose to pregnant females between E9.5 and E15.5, and analyzed lacZ expression in the pancreas and gut at E17.5 (n=3-5 embryos per treatment group). No recombination was detected in the absence of tamoxifen (Fig. 2A,F). The labeling frequency in the pancreas was highest (w25%) when tamoxifen was given at E9.5 (Fig. 2B), and decreased with later treatment: ∼12% lacZ⁺ with TM at E11.5, ∼8% at E13.5 and ∼5% at E15.5 (Fig. 2C-E). [In this and other experiments, we analyzed multiple sections spaced throughout the specimen, to avoid errors due to stochastic variations in labeling efficiency (see Fig. S2 in the supplementary material).] The Hes1^C2 labeling pattern agrees with the downregulation of endogenous Hes1 expression (Fig. 1A-D), and was reproduced using the R26R^EYFP reporter (data not shown). In contrast to the pancreas, liver labeling by Hes1^C2 increased with later tamoxifen administration, particularly in cells adjacent to the portal veins (Fig. 2F-J). Almost all labeled cells in the liver expressed the ductal plate markers CK19 and E-cadherin (see Fig. S6 in the supplementary material and data not shown), consistent with studies showing that Notch and Hes1 promote intrahepatic bile duct development (Antoniou et al., 2009; Geisler et al., 2008; Kodama et al., 2004; Lozier et al., 2008; Zong et al., 2009).

To determine the fate of Hes1^C2-labeled cells in the pancreas, we stained these specimens for endocrine and exocrine differentiation markers, and calculated the fraction of labeled (lacZ⁺) cells expressing each marker. Scoring over 1000 lacZ⁺ cells in each experimental group (see Table S1 in the supplementary material), we found that Hes1⁺ cells labeled at E9.5 generate both endocrine and exocrine progeny, with roughly one-third of all lacZ⁺ cells co-expressing insulin (β-cells) or glucagon (α-cells), and the remainder comprising amylase⁺ acinar and DBA⁺ duct cells (Fig. 3A-D,I-J). [Note that although these counts were derived from images taken on a compound microscope, we obtained essentially identical numbers with images generated by confocal microscopy (Fig. 3I-J).] This finding implies that early Hes1⁺ cells are multipotent, a conclusion supported by a low-dose clonal labeling approach, previously used to demonstrate tip cell multipotency (Zhou et al., 2007) (see Fig. S7 in the supplementary material). Together, these results are consistent with Notch signaling through Hes1 to maintain early multipotent progenitors.

Later Hes1⁺ cells continued to generate exocrine progeny, while appearing to lose endocrine differentiation capacity: almost no α-cells were labeled by TM treatment at E13.5, and β-cell labeling approached zero at E15.5 (Fig. 3E-J). Reduced islet contribution was also seen when comparing E13.5-labeled pancreata, 2 weeks after birth, with those labeled at E9.5 (Fig. 3K-N). Delivery of live pups in these experiments required lower tamoxifen doses, which resulted in decreased labeling overall but did not affect the distribution of labeled cells among differentiated lineages. The fact that labeling efficiency was uncoupled from label distribution implies that Hes1^C2 drives recombination within a cell population of relatively homogeneous potential for endocrine, duct and acinar differentiation, with the proportion of labeled cells depending on TM dose. We cannot exclude the existence of cells expressing Hes1 at levels too low for labeling by Hes1^C2, the developmental potential of which might differ from those observed here. Nonetheless, our results suggest that Hes1 expression shifts from multipotent to exocrine-restricted progenitors (Fig. 3O), consistent with Hes1 having to turn off before the pro-endocrine gene Neurog3 can turn on (Jensen et al., 2000; Lee et al., 2001).

The wave of acinar and islet cell differentiation that occurs at the secondary transition coincides with Hes1 downregulation (Fig. 1A-D), and we have previously shown that artificially preventing Notch downregulation blocks this differentiation process (Murtaugh et al., 2003). This was achieved by crossing the pan-pancreatic driver Pdx1Cre (Gu et al., 2002) to Rosa26^{Notch1IC-IRES-GFP} (henceforth, Rosa26^NIC), which drives co-expression of activated mouse Notch1 and GFP (Murtaugh et al., 2003) (Fig. 4A,G). To determine whether early and late Hes1⁺ cells are similarly susceptible to Notch, we crossed Hes1^C2 to Rosa26^NIC.

Pregnant females received a single 2 mg TM dose between E9.5 and E15.5, and double-transgenic offspring were analyzed at E17.5 (n=2 or 3 per timepoint, see Table S1 in the supplementary material). After TM at E9.5 or E11.5, GFP⁺ cells (expressing activated Notch) formed cystic structures lacking endocrine or acinar marker expression (Fig. 4B-C,H-I). This phenotype resembled that obtained with Pdx1Cre (Fig. 4A,G), and agrees with early Hes1⁺ cells representing multipotent, Notch-sensitive progenitors. The GFP⁺ epithelia stained with DBA lectin, which marks mature ducts as well as embryonic progenitors (Kobayashi et al., 2002), but their cystic morphology distinguished them from the narrow and highly-branched ducts normally present at these stages (Fig. 4G-K).

Although Hes1^C2 labeled very few islet cells at E13.5 or E15.5 (Fig. 3I), Rosa26^NIC prevented even this low level of islet differentiation (Fig. 4D,E). By contrast, exocrine differentiation of E13.5-E15.5 Hes1⁺ cells appeared to be partially Notch resistant, as GFP⁺ cells were found integrated into normal acini and ducts (Fig. 4J-K). To determine if the blunted effects of late Rosa26^NIC activation were secondary to the overall decrease in Hes1^C2 labeling efficiency (Fig. 2), we repeated E9.5 treatment with a lower tamoxifen dose (0.5 mg), to activate fewer cells. As previously, the rare GFP⁺ cells observed in this experiment formed abnormal cystic tubules (Fig. 4F,L), suggesting that early Notch activation can disrupt exocrine differentiation without a `critical mass' of affected cells.

When Rosa26^NIC is activated by Pdx1Cre, all cells exhibit a Pdx1^high `trapped progenitor' phenotype (Murtaugh et al., 2003) (see Fig. S8A,D in the supplementary material). When Notch was activated by Hes1^C2 at E15.5, GFP⁺ cells were negative for Pdx1, which instead was expressed only by β-cells (see Fig. S8C,F in the supplementary material). Cells in which Rosa26^NIC was induced at E9.5 exhibited modest Pdx1 upregulation (see Fig. S8B,E in the supplementary material), suggesting that they retained a partial progenitor-like identity (Fig. 4N).

Further analysis of late-induced Hes1^C2/+; Rosa26^NIC/+ pancreata revealed that Notch activation at E13.5 caused most GFP⁺ cells to adopt a ductal rather than acinar fate, whereas after E15.5 activation the proportions were reversed (Fig. 4M). The latter resembles the wild-type distribution observed with R26R^LacZ (Fig. 3J), suggesting that Notch activation at E15.5 does not perturb exocrine differentiation, while E13.5 activation drives bipotent progenitors toward a duct fate (Fig. 4N). Activated Notch can also respecify endocrine precursors to ducts (Greenwood et al., 2007), and this pro-ductal activity may underlie the pathological effects of Notch in pancreatic cancer (De La O et al., 2008).

The question of whether pancreatic progenitor cells persist after birth is a matter of long-standing controversy, particularly with respect to adult differentiation of β-cells. Although several lineage-tracing studies indicate that adult duct and acinar cells do not generate new β-cells in the resting or regenerating pancreas (Desai et al., 2007; Kopinke and Murtaugh, 2010; Solar et al., 2009), these have not excluded the existence of specialized progenitor cells. Centroacinar cells (CACs), in particular, have been proposed to behave as β-cell progenitors (Hayashi et al., 2003; Nagasao et al., 2003; Rovira et al., 2010). CACs express higher levels of Hes1 than do other exocrine cells (Miyamoto et al., 2003; Stanger et al., 2005), and we used Hes1^C2 to determine whether these or other Hes1⁺ cells behave as adult stem or progenitor-like cells.

To identify Hes1-expressing cells in the adult pancreas, we administered 10 mg tamoxifen to 2-month-old (P60) Hes1^C2/+; R26R^EYFP/+ mice (n=4), and analyzed EYFP expression ∼48 hours later. Consistent with prior studies of Hes1 expression, we found EYFP labeling of approx. one-quarter of CK19⁺ centroacinar cells (Fig. 5A), as well as a lesser proportion of labeled cells within larger ducts (Fig. 5B). We also observed a small fraction of EYFP⁺ acinar cells, suggesting Hes1 expression by rare, differentiated acinar cells (Fig. 5C). To follow the longer-term fate of adult (P60) Hes1⁺ cells, we compared quantitatively the labeling obtained at 7 days post-TM, reflecting the initial differentiation state of Hes1⁺ cells, to that observed after a 2-month `chase', in which time cells might have adopted new fates. As in the 48-hour chase experiment, we detected acinar, duct and CAC labeling after 7 days, which persists at 2 months (Fig. 5D-G). Scoring the labeling index of each differentiated cell type (n=3-10 mice analyzed per condition) (see Table S2 in the supplementary material), we found that ∼1% of acinar cells were EYFP⁺ at each timepoint [1.4±0.3% (s.e.m.) at 7 days, 1.3±0.3% at 2 months; Fig. 5H]. By contrast, the labeling index of duct cells (defined here as CK19⁺ epithelial cells not embedded within acini) increased by roughly twofold, from 8.0±0.6% EYFP⁺ at 7 days to 15.5±0.6% EYFP⁺ after 2 months (Fig. 5H). This might indicate that Hes1 marks a subpopulation of proliferating duct cells, consistent with a mitogenic role for Notch in this lineage (Golson et al., 2009). Regarding centroacinar cells specifically, we found a similar labeling index at both timepoints (27.0±2.9% EYFP⁺ at 7 days, 26.5±0.9% EYFP⁺ at 2 months). The relative labeling indices of ducts and CACs raises an alternative explanation of why duct labeling increases over time, namely that expansion of the ductal tree is driven by descent from CACs. Although anatomically plausible, this hypothesis requires work beyond the scope of this study.

Unexpectedly, we also detected Hes1^C2-labeled cells within islets. In islets and throughout the pancreas and other organs, Hes1^C2 labeled numerous endothelial cells (∼20% in all experiments) (see Fig. S9 in the supplementary material and data not shown), which we have not analyzed further. With respect to endocrine cells, we did not observe a single Hes1^C2-labeled β-cell in these experiments, out of over 2000 insulin⁺ cells scored at each timepoint (see Table S2 and Fig. S9A-B in the supplementary material). We did find rare glucagon⁺ α-cells marked by Hes1^C2 at both timepoints (3.2±0.4% EYFP⁺ at 7 days post-TM, 5.7±1.1% at 2 months) (see Fig. S9C-D in the supplementary material), as well as after very short chase periods (12-24 hours post-TM, data not shown). Although the increased α-cell labeling with time is statistically significant (P<0.05), its biological relevance is unclear: it could indicate rare neogenesis from the more highly labeled ducts or CACs, although this has not been seen in previous experiments (Kopinke and Murtaugh, 2010; Solar et al., 2009). Alternatively, and consistent with α-cells being labeled at short chase periods, Hes1 could be expressed by rare α-cells and mark a more proliferative subset of this population. Adult α-cells dynamics have received less attention than those of β-cells, although tools now exist to determine whether significant numbers of adult α-cells are born outside the islet (Thorel et al., 2010).

The lack of β-cell labeling suggests that Hes1⁺ cells do not behave as adult precursors for this cell type. Our TM treatment regimen appeared to capture most Hes1⁺ cells in the adult pancreas: with respect to centroacinar cells, a higher dose of tamoxifen (3×10 mg, over 3 days) conferred no more labeling than our standard 1×10 mg dose (24.9±0.1% EYFP⁺ at 1 month post-treatment, n=2). We propose that Hes1 expression marks only a subset of CACs in the adult pancreas, which does not contribute to β-cells in vivo.

Independent evidence for the efficiency of adult Hes1^C2 labeling comes from the adult intestine, in which we found that Hes1^C2 does mark stem cells. Low dose (2 mg) tamoxifen treatment of adult (P60) Hes1^C2/+; R26R^EYFP/+ mice (n=2) revealed labeling, within 12 hours, of single cells in the crypt base region as well as just above the crypts (Fig. 6A,B). The location of the former cells, representing approx. two-thirds of all EYFP⁺ cells in the intestinal epithelium, agreed with previous Hes1 expression studies (Jensen et al., 2000; Schroder and Gossler, 2002), whereas the latter may represent transit-amplifying cells in which Notch inhibits secretory lineage specification (Crosnier et al., 2006). After a long-term chase (30 days, n=2), we found entire crypt-villus units expressing EYFP, suggesting labeling of intestinal stem cells (Fig. 6C,D). We also found that Hes1^C2-expressing cells, like bona fide intestinal stem cells (Barker et al., 2009; Sangiorgi and Capecchi, 2008; Zhu et al., 2008), were susceptible to transformation by activated β-catenin [using a Ctnnb1^lox(ex3) gain-of-function allele (Harada et al., 1999)] (Fig. 6E,F). Importantly, a single 5 mg TM dose was sufficient to label most crypt-villus units along the intestinal tract of Hes1^C2/+; R26R^LacZ/+ mice (n=2) after a 6-month chase. Labeling was highest in the duodenum (∼90% of crypts), and decreased posteriorly to ∼50% labeling in the colon (Fig. 6G-J), frequencies that compare favorably with those of other intestinal stem cell Cre drivers (Barker et al., 2007; Sangiorgi and Capecchi, 2008; Zhu et al., 2008). This high efficiency implies that a 5 mg TM dose was sufficient to recombine most Hes1⁺ cells in the intestine; assuming a similar dose-response relationship in the pancreas, our 10 mg TM treatment regimen should also have labeled most Hes1⁺ cells in this organ. Taken together, our results highlight the contrasting cellular dynamics of the adult pancreas and intestine, and suggest that Hes1^C2 might be useful to mark and manipulate adult stem cells when they exist in other organs.

Although β-cell neogenesis may not be required in the healthy adult pancreas, it might be induced by injury, as in the case of pancreatic duct ligation (PDL) (Inada et al., 2008; Xu et al., 2008). PDL causes inflammation and acinar cell apoptosis distal to the ligation site (Scoggins et al., 2000; Watanabe et al., 1995), which is accompanied by a local increase in β-cell numbers and apparent reappearance of Neurog3⁺ β-cell precursors (Wang et al., 1995; Xu et al., 2008). We therefore sought to determine whether Hes1 marks cells capable of β-cell neogenesis in this model.

Ligation of the dorsal (splenic) pancreas lobe was performed as described by others (Scoggins et al., 2000; Solar et al., 2009; Watanabe et al., 1995; Xu et al., 2008); a full description of our observations in this model will be submitted elsewhere. At 7 days after surgery, acinar cells in the ligated region were completely replaced by fibro-inflammatory cells and epithelial tubules (Fig. 7A,B). These epithelia presented suggestive evidence of neogenesis, including association with small β-cell clusters, upregulation of Pdx1, and re-expression of Neurog3 (Fig. 7E-G and data not shown).

To follow Hes1⁺ cells, we administered TM to Hes1^C2/+; R26R^EYFP/+ mice either 1 month pre-surgery (to mark Hes1⁺ cells in the healthy pancreas; n=8) or 3 days post-surgery (to capture progenitor-like cells upregulating Hes1 after injury; n=4), and analyzed EYFP expression 7 days after injury. The 7-day timepoint was chosen for consistency with previous studies indicating a doubling of β-cell mass at this stage, possibly owing to ductal neogenesis (Solar et al., 2009; Wang et al., 1995; Xu et al., 2008). From each mouse, we analyzed 8-10 sections spaced evenly throughout the ligated region (see Fig. S2 in the supplementary material), manually scanning the entire area of each section for EYFP⁺/insulin⁺ cells (encompassing ∼10,000 insulin⁺ cells scanned per ligated pancreas). We did not find a single Hes1^C2-labeled β-cell in any pancreas, regardless of labeling strategy (Fig. 7C,D and data not shown). We did observe widespread labeling of E-cadherin⁺ and CK19⁺ epithelial complexes in ligated pancreata [∼15-20%, a labeling index similar to that of duct and centroacinar cells in the absence of injury (Fig. 5H)], in both TM treatment groups (Fig. 7C,D) (see Table S2 in the supplementary material and data not shown). Although insulin⁺ cells were frequently associated with these complexes, they were always unlabeled (Fig. 7E). These results suggest that Hes1^C2 does not mark cryptic or injury-induced β-cell progenitors in the adult. Although these findings do not exclude the possibility of β-cell differentiation from cultured centroacinar cells in vitro (Rovira et al., 2010), or from Hes1-negative cells in vivo, they agree with an independent finding that ducts do not generate β-cells after PDL (Solar et al., 2009).

Hes1 is a major Notch target in diverse tissues (Kageyama et al., 2007), and we use its expression as a `tag' to determine where and when Notch is active in the pancreas. In the embryonic pancreas, our results suggest that early Hes1⁺ cells are multipotent progenitors, the differentiation of which is inhibited by Notch, whereas late Hes1⁺ cells are exocrine restricted, and respond to elevated Notch levels by becoming ducts (Fig. 3H; Fig. 4F). Our findings in the adult pancreas suggest that Hes1 expression is restricted to differentiated cells, most abundant within the centroacinar population, and that Hes1⁺ cells do not normally behave as cryptic or facultative stem cells for endocrine β-cells. Hes1^C2 provides a novel tool with which to analyze embryonic and adult cells in the pancreas, and our studies of the adult intestine suggest that it might be widely useful in marking and manipulating adult stem cells.

Our results confirm and significantly extend prior studies of Notch-responsive cells in the embryonic pancreas. For example, mapping Notch1 receptor activation in vivo, via N₁IP-Cre, reveals scattered labeling throughout the exocrine and endocrine pancreas (Vooijs et al., 2007). Although N₁IP-Cre identifies the range of cell types that had experienced Notch1 signaling at some prior stage, it cannot determine when that signaling occurred. As our CreERT2 approach allowed us to mark cells expressing Hes1 at specific developmental stages, we could show that most islet-fated cells had received Notch signals only before the `secondary transition', a wave of endocrine differentiation spanning ∼E13.5 to birth in the mouse (Herrera et al., 1991; Pictet and Rutter, 1972).

The fact that Hes1 can directly repress Neurog3 (Lee et al., 2001), together with the excessive α-cell differentiation observed in Hes1 mutants (Jensen et al., 2000), might suggest that Hes1 downregulation is rate-limiting for endocrine specification. However, Neurog3⁺ endocrine precursors continue to be generated throughout the secondary transition, with a peak at E15.5 (Gradwohl et al., 2000; Gu et al., 2002), and the majority of β-cells differentiate between E15.5 and birth (Herrera et al., 1991). Our results suggest that these cells must derive from progenitors that turn on Neurog3 several days after having turned off Hes1, arguing that endocrine specification is not immediately induced upon Notch-Hes1 downregulation. Indeed, the secondary transition appears to proceed normally in Notch1/Notch2 double mutants (Nakhai et al., 2008), suggesting that Notch-independent mechanisms control the timing of Neurog3 expression and islet differentiation in late embryogenesis.

Prior to the secondary transition, both Cpa1⁺ tip cells and Hnf1b⁺ ducts contain multipotent progenitors (Solar et al., 2009; Zhou et al., 2007). Hes1 is expressed in and labels both tip cells and ducts during early pancreas development (see Fig. S3A in the supplementary material and data not shown), suggesting that it marks multipotent progenitors regardless of anatomical location. After E13.5, Cpa1⁺ cells behave as acinar-restricted precursors, whereas Hnf1b⁺ cells become restricted to islet and duct fates (Solar et al., 2009; Zhou et al., 2007). To reconcile these observations with our hypothesis that Hes1 marks bipotent acinar/duct progenitors after E13.5, we suggest that acinar-restricted Cpa1⁺ cells derive from bipotent Hes1⁺ cells from ∼E13.5-E15.5. Indeed, analysis of GFP perdurance in Sox9-EGFP transgenic embryos suggests that ducts give rise to acini through at least E14 (Seymour et al., 2008). That late duct-to-acinar differentiation was not observed with Hnf1b-CreERT2 may reflect inefficient labeling by this transgene in utero (Solar et al., 2009), combined with the overall rarity of Hes1⁺ progenitors at these stages (Figs 1, 2). In zebrafish, Notch is required for duct specification of exocrine-restricted progenitors (Yee et al., 2005), and our results suggest that Notch also promotes duct development in late mouse embryogenesis.

Notch is implicated in self-renewal of adult stem cells (Chiba, 2006), and Hes1^C2 robustly labels stem cells in the intestinal crypts. Under conditions sufficient to label the majority of intestinal stem cells, however, we do not find evidence that Hes1^C2 labels stem-like cells in the adult pancreas. Instead, we find that Hes1^C2 is active in several mature cell types, of which centroacinar cells are the most highly labeled. Previous studies indicate that Hes1 expression and Notch activity are highest in CACs, and lower in more proximal ductal elements (Miyamoto et al., 2003; Parsons et al., 2009; Stanger et al., 2005), closely paralleling the Hes1^C2 labeling pattern. Although we have not obtained reliable Hes1 immunostaining in adult pancreata (data not shown), Hes1^C2 labeling suggests that it is also expressed by rare differentiated α-cells and acinar cells. Whether Notch has a functional role in these cells remains to be determined. Importantly, we never observe Hes1^C2 labeling of insulin⁺ β-cells, suggesting little or no contribution to these cells from the adult Hes1 lineage.

This result appears inconsistent with the finding that CACs, isolated based on high aldehyde dehydrogenase activity, can give rise to β-cells and other cell types in vitro (Rovira et al., 2010). Interestingly, however, the cells isolated in that study expressed only low levels of Hes1, suggesting that they represent a distinct subpopulation of CACs. Indeed, we find that only approx. one-quarter of CACs are labeled by Hes1^C2 using our standard tamoxifen dose, and that this proportion is not increased by a threefold higher dose. Our results therefore constitute in vivo evidence for heterogeneity within the duct and CAC compartments, and suggest that the Hes1⁺ subpopulation does not normally give rise to β-cells.

Alternatively, the failure of Hes1^C2 to label β-cells might reflect limitations imposed by the micro-environment of the mature pancreas, e.g. active Notch signaling reinforcing ductal fate, which could be removed in tissue culture or during regeneration. To address this, we adopted an injury model, pancreatic duct ligation, which has provided suggestive evidence of β-cell neogenesis from ductal progenitors (Wang et al., 1995; Xu et al., 2008). In rats and mice, PDL is reported to lead to a local doubling of β-cell mass within 1 week (Wang et al., 1995; Xu et al., 2008), together with inflammation, acinar cell apoptosis and ductal hyperplasia (Scoggins et al., 2000; Watanabe et al., 1995). At 7 days post-PDL, we find that Hes1^C2-labeled cells contribute to the abnormal ductal epithelium, but not to β-cells located either in islets or within or near ducts. We note that identical results were obtained using Hnf1b-CreERT2, which labels cells throughout the ductal network (Solar et al., 2009), suggesting that new β-cells arise after PDL either from pre-existing β-cells, or from a duct subpopulation that expresses neither Hnf1b nor Hes1 (Inada et al., 2008).

Our results do not exclude the possibility that CACs or other Hes1⁺ cells could give rise to β-cells more than 7 days post-PDL, although further increases in β-cell mass beyond this timepoint have not been reported (Wang et al., 1995; Xu et al., 2008), and adipocyte infiltration at later stages may produce secondary effects on islets (Watanabe et al., 1995). It is also possible that other injury models might evoke β-cell differentiation from Hes1⁺ cells, much as glucagon⁺ α-cells can transform into β-cells after extreme β-cell loss, despite an otherwise absolute barrier to interconversion (Thorel et al., 2010). We also note that, as Hes1^C2 labels a minority of CACs, our study does not definitively test CAC differentiation potential. Several other CreERT drivers, particularly those with stringent tamoxifen dependence, have been shown to recombine only a minority of their putative target cells (Desai et al., 2007; Dor et al., 2004; Solar et al., 2009), raising the possibility of unlabeled subpopulations. Our results, however, suggest that Hes1^C2 does label most Hes1⁺ cells in the adult, but that Hes1⁺ duct and centroacinar cells cannot generate β-cells under the conditions studied here. In sum, Hes1^C2 provides a new tool to test the role of Notch-responsive cells in physiological and pathological conditions, and our studies raise the issue of whether Notch activity functionally subdivides pancreatic duct cells into those with and without progenitor-like potential.

We thank Nadean Brown (University of Cincinnati) for the Hes1 antibody, Makoto Mark Taketo (Kyoto University) for Ctnnb1^lox(ex3) mice, and Kirk Thomas and Mario Capecchi (University of Utah) for reagents and advice on gene targeting. We thank Richard Dorsky, Edward Levine and Nadja Makki for helpful comments on this manuscript. This work was supported by grants from the Searle Scholars Foundation (06-B-116), NIH (R01-DK075072) and Beta Cell Biology Consortium (U01-DK072473, subaward VUMC35146) to L.C.M., and a Boehringer Ingelheim Fonds graduate fellowship to D.K. Deposited in PMC for release after 12 months.