Spatiotemporal patterns of multipotentiality in Ptf1a-expressing cells during pancreas organogenesis and injury-induced facultative restoration

The inherent plasticity potential of adult pancreas is especially interesting considering the possibility of controllably reprogramming various cell types into β-cells as an alternative to embryonic stem cell differentiation and transplantation therapy for diabetes. Pancreatic multipotent progenitor cells (MPCs) are the basis for generating a functional pancreas with appropriate cell allocations and structures. Understanding the ontogeny and intermediate states of MPCs could provide insight into whether the differentiated adult pancreatic cells could be challenged to return to an embryonic multipotency under stimuli such as physiological injury or chronic diabetes.

The early pancreatic epithelium contains MPCs producing several transcription factors (TFs), including Ptf1a, associated with commitment toward pancreatic fate, and controlling early organ growth (Pan and Wright, 2011). At the beginning of morphogenetic maturation [around embryonic day (E) 12], the epithelium becomes compartmentalized into ‘tip’ and ‘trunk’ domains (Zhou et al., 2007; Villasenor et al., 2010). Genetic lineage-tracing via CpaI^CreER suggested that MPCs with the rudimentary signature Pdx1⁺Ptf1a⁺cMyc^HighCpaI⁺ are located at the tip domain, but rapidly disappear with the start of the secondary (2°) transition (Zhou et al., 2007), and the trunk epithelium between E12.5 and E15.5 was proposed to harbor bipotential progenitor pools for duct and endocrine cells (Zhou et al., 2007; Schaffer et al., 2010). Lineage-tracing analysis using Hnf1β^CreER (Solar et al., 2009) and Sox9^CreER (Kopp et al., 2011), however, suggested that trunk epithelium contains a progressively reducing number of MPCs, even after overt tip-trunk segregation. While the Hnf1β⁺ population is multipotent during the early 2° transition, but largely bipotential at later stages, relatively low numbers of Sox9⁺ cells were proposed to feed all pancreatic lineages throughout embryogenesis, suggesting persistence of Sox9⁺ MPCs after the 2° transition (Kopp et al., 2011). Because Sox9⁺ and Hnf1β⁺ cells were localized to the trunk epithelium, with CpaI⁺ cells at the tips, the precise location and number, and molecular signature, of MPCs during and after the 2° transition remains unclear.

The existence of adult facultative progenitors capable of regenerating endocrine cells under special stimuli has been contentious (Granger and Kushner, 2009). Recent lineage tracing in mice, however, yielded the surprising suggestion that many terminally differentiated cell types could be reprogrammed to other cell-types, either using exogenous factors or under injury conditions. Xu et al. (Xu et al., 2008) concluded that facultative Ngn3⁺ (Neurog3⁺ - Mouse Genome Informatics) endocrine progenitors, producing β-cells ex vivo, appeared in/at the duct following pancreatic duct ligation (PDL). By contrast, lineage-tracing results with duct/centroacinar-specific CreER lines (Hnf1β^CreER, Hes1^CreER, Sox9^CreER) provided evidence against duct and centroacinar cells deriving endocrine cells after PDL (Solar et al., 2009; Kopinke et al., 2011; Furuyama et al., 2011; Kopp et al., 2011). Detailed studies are required to determine rigorously the origin of the Ngn3⁺ duct cells and if they progress to islet/β-cells in vivo.

Because of their vast number, acinar cells could be an attractive source for replenishing β-cells to offset losses from disease or injury. Acinar-to-endocrine conversion was suggested with primary acinar cells cultured in vitro when treated with growth factors (Baeyens et al., 2005; Minami et al., 2005). In vivo lineage-tracing using Elastase^CreER, however, showed no evidence of such conversion upon regeneration/injury with partial pancreatectomy (Ppx), PDL, or cerulein-induced pancreatitis (Desai et al., 2007). Strikingly, overexpression of three β-cell TFs - Pdx1, Ngn3, MafA - apparently turned acini into β-like cells in vivo, suggesting convertibility under strong stimuli (Zhou et al., 2008). It remains unclear whether all acinar cells are equally susceptible to such conversion, and if it could be activated without genetic intervention or under injury conditions that might exist in diabetes or pancreatitis.

Ptf1a is expressed in early bud MPCs, with an instructive role in distinguishing pancreatic fate from the adjacent organs (Kawaguchi et al., 2002). In the 2° transition pancreatic epithelium, Ptf1a production is dynamically regulated. Over time, its activity changes from driving an MPC program to directing tip cells into a proacinar state. Moving from MPC to proacinar behavior is proposed to be linked to the switching of Ptf1a co-regulatory proteins in the trimeric PTF1 complex, from PTF1^RBP-J to PTF1^RBP-JL (Masui et al., 2007). An outstanding issue is whether small numbers of Ptf1a⁺ MPCs persist during/after the 2° transition, or if adult Ptf1a⁺ acini could somehow re-engage (aspects of) an embryonic Ptf1a-driven MPC program, to adopt facultative progenitor activity.

Here, we report a knock-in tamoxifen-inducible Ptf1a^CreERTM line for the dissection of gene regulatory networks critical for maintaining the MPC state and controlling tip-trunk compartmentalization. We reasoned that lineage-tracing analysis performed using the MPC-instructive factor Ptf1a would provide more certainty in addressing the intrinsic potentiality (mono, bi, or multi) of pancreatic progenitor populations during organogenesis. We describe a dynamic heterogeneity within the tip epithelium, based on distinct TF signatures, which changes during tip-trunk segregation concordant with the shift from multipotency to unipotency in Ptf1a-expressing cells. Pulse-labeling-type lineage-tracing using Ptf1a^CreERTM showed that the number of Ptf1a⁺ MPCs decreased progressively and rapidly during organogenesis, with rare MPCs present late into the 2° transition, and perinatal/adult stage Ptf1a⁺ cells produce only acinar cells. However, Ptf1a⁺ acinar cells were activated by PDL to become facultative progenitors, express several embryonic multipotency factors and become competent to produce duct and endocrine cells, including mature β-cells. Acinar-to-endocrine conversion was detected in substantial numbers only 2 months post-PDL, but their formation was enhanced by eliminating pre-existing β-cells with a selective toxin, streptozotocin (STZ). This is the first study to demonstrate, with in vivo lineage-tracing, that acinar cells give rise to endocrine cells under injury-induced reprogramming paradigms and without additional transcription factors or signaling molecules.

Ptf1a^CreERTM was generated using recombinase-mediated cassette exchange (Long et al., 2004) (supplementary material Fig. S1A). R26R^lacZ and R26R^EYFP mice were described previously (Soriano, 1999; Srinivas et al., 2001). Animals and embryos were PCR-genotyped. Experiments were under protocols approved by Vanderbilt University IACUC.

Tamoxifen (Tam, Sigma T-5648) was 30 mg/ml in corn oil (Sigma C8267). Tamoxifen administration and dosage used were described in detail in results section. PDL was performed as described (Xu et al., 2008). Streptozotocin (STZ; 10 mg/ml in sodium citrate pH 4.5) was given intraperitoneally (100 mg/kg) for 2 consecutive days at post-PDL D3.

Embryonic or adult pancreas was fixed (2-4 hours, 4% paraformaldehyde, 4°C), and for cryosections, washed twice in cold PBS, sucrose-equilibrated (30%, 4°C, overnight), and embedded in OCT compound (Tissue-Tek, Sakura). Paraffin sections, whole-mount β-galactosidase detection, H&E staining, processing was as described (Kawaguchi et al., 2002). Immunofluorescence used 10 μm cryosections (Kawaguchi et al., 2002). Antibodies used are listed in supplementary material Table S1.

Images from Zeiss confocal (LSM 510 META upright) or Apotome were analyzed with LSM Image Browser and Zeiss Axiovision 4.8 software, respectively. Systematic quantitation analysis was performed by Photoshop CS3 and NIH ImageJ software.

Short-term lineage tracing, E13.5 and E14.5: labeling of the acini, ducts or endocrine cells was manually scored as percentage of each EYFP⁺ cell-type of total cell number in each compartment, using all field of view, every other section (20 μm apart; ∼50% total pancreas). E18: every sixth section (60 μm apart, ∼16% total pancreas) was counted. Juvenile/adult mice: 150-200 random fields from 40 evenly spaced sections (100 μm apart; ∼10% of total pancreas) were analyzed. For E18.5, and juvenile/adult pancreata, percentage labeling was calculated based on EYFP⁺ area over total area for each cell type. Head/tail pieces of PDL pancreata were separately embedded and analyzed. All fields of view from every 100 μm section (∼40-45; ∼10% total pancreas) ensured coverage of all regions. At least three to four pancreata per stage/group were analyzed. Two-tailed Student's t-test used a significance of P<0.05.

Given that Ptf1a is expressed in early pancreatic MPCs (Kawaguchi et al., 2002), and later at E12 is seemingly restricted to MPC/proacinar progenitors in the tip epithelium of the remodeling epithelial plexus (Zhou et al., 2007), it was critical to determine quantitatively the dynamics of this shift from multipotential to unipotential behavior. To lineage-trace Ptf1a-expressing cells at different stages of pancreas organogenesis, we derived Ptf1a^CreERTM knock-in mice expressing CreER™ under endogenous Ptf1a promoter/enhancer elements (supplementary material Fig. S1A). CreER™ production recapitulates endogenous Ptf1a expression, with nuclear translocation induced in Ptf1a⁺ and CpaI⁺ acinar cells within 24 hours of tamoxifen (Tam) administration at E15.5 (supplementary material Fig. S2A,B).

Testing Ptf1a^CreERTM mice with reporter alleles (Rosa26R^lacZ [R26R^lacZ] or Rosa26R^EYFP [R26R^EYFP]) and Tam administered at designated developmental stages allowed indelible marking of the progeny of Ptf1a⁺ cells throughout subsequent development. Recombination of R26R^lacZ was strictly tamoxifen-dependent and no pancreatic β-gal⁺ cells were detected in non-CreER™ mice (supplementary material Fig. S3A,B; Fig. 1C,C′). For embryonic labeling, while single 3 mg Tam doses at E16.5 and E18.5 resulted in ∼50-60% labeling of acinar cells (supplementary material Fig. S3D; Fig. 3B-D), lower labeling (∼10-40%) occurred with single 3 mg Tam doses at earlier stages (E9.5-13.5; Fig. 1D′,E′,F′,G′). For adults, three 3 mg Tam doses over 6 days resulted in ∼60-80% recombination of R26R^lacZ or R26R^EYFP (supplementary material Fig. S3B; Fig. 3E-J). Ptf1a^CreERTM-induced R26R^lacZ recombination in adult pancreas was Tam-dose-dependent; labeling efficiency was enhanced moving from 1 to 3 mg (supplementary material Fig. S3C). Ptf1a^CreERTM is therefore valuable for pulsed-labeling studies, and conditional inactivation of genes involved in pancreas development/cancer, with excellent spatiotemporal control.

Ptf1a^CreERTM;R26R^EYFP/EYFP embryos were generated at various stages (Fig. 1A). Single 3 mg Tam dosing of Ptf1a^CreERTM;R26R^EYFP pregnant mice at E10.5 would label Ptf1a-expressing cells of the primary (1°) transition (E9.5-12); pulses between E12.5-14.5 would label 2° transition Ptf1a-expressing cells; and at E15.5 or later would test whether Ptf1a-expressing cells had completely switched from multipotential behavior (Fig. 1B). Pancreata were analyzed near the end of gestation (E18.5) for co-localization of EYFP (derivation from Ptf1a⁺ cells) with endocrine, ductal and acinar markers. The percentage of labeled (EYFP⁺) cells expressing each marker was quantified by a strategy taking into account the potential variation in proliferation and cell sizes in the acinar, duct or endocrine compartment. Because nuclear CreER localization starts ∼6-12 hours post-Tam-injection with a 36-hour end point (supplementary material Fig. S2) (Ahn and Joyner, 2004; Danielian et al., 1998), data are presented as each injection point having a labeling period offset by 12-36 hours.

Tam pulses at E10.5 (E11 labeling; n=4) labeled EYFP⁺ cells equivalently in all three epithelium-derived compartments - acini/duct/endocrine - indicating that the Ptf1a⁺ population is multipotent at this stage (Fig. 1D,D′), and consistent with previous studies showing Ptf1a marks early pancreatic bud MPCs (Kawaguchi et al., 2002; Burlison et al., 2008). The percentage of EYFP⁺ cells per compartment was reproducibly comparable (Fig. 1D′; supplementary material Fig. S1B), suggesting that the early MPC population contributes equally to all three lineages, including all endocrine cell types (supplementary material Fig. S1F).

To determine the degree to which the Ptf1a⁺ population maintains multipotency over the 2° transition (E12.5-15.5), single Tam doses were given at E12.5, E13.5 and E14.5, with analysis at E18.5. Labeling at E13 and E14 yielded lineage-traced endocrine, duct and acinar cells (Fig. 1E,E′,F,F′), indicating that significant numbers of Ptf1a⁺ MPCs exist during the 2° transition. Quantitation showed a tendency to produce acinar over duct and endocrine cells (E13, n=3; E14, n=5; Fig. 1E,F; supplementary material Fig. S1C,D,E). In particular, the E13 result suggests that the tip epithelium at this stage already contains different populations of Ptf1a-expressing cells, in accordance with immuno-colocalization data below. These findings largely agree with CpaI^CreERT2 studies demonstrating diminution of the MPC pool during the 2° transition (Zhou et al., 2007). With labeling at E15, lineage-trace was predominantly restricted to acini and to a much lesser extent duct cells (n=3; Fig. 1G-G′). Taken together, these data indicate that Ptf1a-expressing cells have almost completely switched to unipotential acinar progenitor behavior by ∼E15, with a minimal late-stage-remaining contribution to the duct compartment.

Shorter-term lineage-tracing was performed to assess lineage allocation of Ptf1a-expressing MPCs during the 2° transition, specifically to determine the extent to which tip MPC yield progeny moving into the trunk epithelium (the presumptive location of the endocrine/duct bipotent pool). Twenty-four hours post-labeling (injecting at E12.5, E13.5; analyzing at E13.5 or E14.5), numerous CpaI⁺ proacinar/acinar cells were EYFP⁺ (of total CpaI⁺ tip cells: ∼8.3±3.2% at E13.5, ∼11.3±2.70% at E14.5; n=4; Fig. 2B,E,F,I). EYFP⁺Hnf1β⁺ cells in trunk epithelium were also regularly observed but at a relatively low frequency (of total Hnf1β⁺ trunk cells: ∼2.1±0.9% at E13.5, ∼1.5±0.9% at E14.5; n=4; Fig. 2D,E,H,I). Therefore, Ptf1a-lineage cells could contribute to both tip, and to a lesser extent, trunk compartments at ∼E13/14, the stage of progression into the secondary transition proper. EYFP⁺Ngn3⁺ endocrine progenitors and EYFP⁺Synaptophysin⁺ endocrine cells (synaptophysin labels endocrine precursors and differentiated hormone⁺ cells), although rare in number, were found 24 hours post-labeling [of total (Ngn3⁺+Synaptophysin⁺) cells: 0.2±0.1% at E13.5, ∼0.06±0.04% at E14.5; n=4; Fig. 2C,E,G,I]. These analyses provide direct evidence, consistent with the longer-term tracing above, that the Ptf1a-expressing tip population is already undergoing, near the initiation of the 2° transition, significant restriction of lineage multipotency. We next questioned to what extent the Ptf1a-expressing tip domain comprised heterogeneous populations of qualitatively distinct cell-types, based on their production of TFs associated with multipotency.

We explored the signature of putative Ptf1a⁺ MPCs between E11.5 and E15.5 to try to move beyond the general signature of Pdx1⁺Ptf1a⁺CpaI⁺cMyc^HI proposed by Zhou et al. (Zhou et al., 2007). We monitored co-localization and relative levels of Ptf1a with Sox9 and Hnf1β, two TFs having loss-of-function evidence of important roles in MPC formation (Seymour et al., 2007; Haumaitre et al., 2005). At E11 and E11.75, with the initiation of tip-trunk compartmentalization, the pancreatic epithelium displayed significant heterogeneity, with three populations observable: (1) Ptf1a^HISox9^LO cells located in peripheral regions of the branching bud epithelium; (2) Ptf1a^-Sox9⁺ cells concentrated in the prospective trunk domain; and (3) Ptf1a⁺Sox9⁺ cells scattered throughout the epithelium (Fig. 2J,K).

By E12.5, tip and trunk domains were relatively distinct based on Ptf1a and Sox9 production, respectively (Fig. 2L). Between E12.5 and E15.5, Ptf1a⁺Sox9⁺ tip cells were found distributed at both tip-trunk interface and intra-tip regions, intermingled with the Ptf1a^HISox9^LO (proacinar) population (Fig. 2L-O; supplementary material Fig. S4A-F). The numbers of these Ptf1a⁺Sox9⁺ decreased progressively over this period. Conversely, the Ptf1a^HISox9^LO population increased to comprise the entire tip domain by E15.5. Ptf1a⁺Sox9⁺ tip cells were also Hnf1β⁺ and CpaI⁺ (supplementary material Fig. S4A-F), and thus may represent the MPCs persisting during the early 2° transition, as marked by our lineage-tracing (Fig. 2P). Notably, the relative signal intensity of Ptf1a, Sox9 and Hnf1β in some of the tip MPC population seems lower compared to the robust signals in the trunk (for Sox9/Hnf1β) or proacini (for Ptf1a). But, determining if lower levels of these co-expressed TFs are necessarily associated with MPC character awaits future tools that allow isolation of this population. Our co-expression and lineage-tracing data fit the notion that multipotent Ptf1a⁺ tip populations diminish rapidly during the 2° transition as they convert to Ptf1a^HI proacinar cells.

From birth until weaning, islet cell mass increases parallel to body mass in mouse and rat (Bouwens and Rooman, 2005). We found no evidence that endocrine cells arise from Ptf1a-expressing cells during this endocrine expansion period. Single 3 mg Tam doses were given to Ptf1a^CreERTM;R26R^EYFP pregnant mice near the end of gestation (E18.5) and pancreata were analyzed at postnatal day 21 (P21; n=4; Fig. 3A). EYFP was found exclusively in CpaI⁺ acinar cells (Fig. 3B-D), suggesting that postnatal Ptf1a-expressing cells contribute to neither endocrine nor duct compartments.

To determine if duct or endocrine cells arise from Ptf1a⁺ cells during normal homeostasis in young adults, we treated P30 mice with three doses of Tam (Fig. 3A). Two days after the last injection (P36), a high proportion of acinar cells were labeled (60.9±3.83%; n=4; Fig. 3E-G), increasing to 72.6±5.39% after 6 months post-Tam. No EYFP⁺ cells were found in the duct or endocrine compartments at either age (P36, 7-month; n=4; Fig. 3H-J). We also failed to detect Hnf1β⁺Ck19⁺ (Ck19 is also known as Krt19 - Mouse Genome Informatics) centroacinar cells (CACs) labeled with EYFP after 6-month chase (Fig. 3K,K′,L,M), consistent with a lack of Ptf1a immunoreactivity in Sox9⁺Ck19⁺ and Hnf1β⁺Ck19⁺ CACs (Fig. 3M-M‴; supplementary material Fig. S6A). These latter results suggest that Ptf1a expression does not mark CACs in vivo, contrasting a previous study reporting Ptf1a expression in flow-sorted CACs (Rovira et al., 2010). Together, these data suggest that Ptf1a⁺ cells self-replicate to maintain the acinar pool in the adult organ, with no contribution towards the duct or endocrine populations.

We endeavored to determine whether Ptf1a-expressing acinar cells could exhibit facultative progenitor activity under PDL, exhibiting multipotentiality similar to embryonic Ptf1a-expressing MPCs. PDL was reported to induce fairly efficient de novo generation of facultative Ngn3⁺ endocrine progenitors (Xu et al., 2008), and this model was chosen to evaluate whether Ptf1a⁺ acinar cells could convert toward the Ngn3⁺ population. Five-week-old Ptf1a^CreERTM;R26R^EYFP/EYFP mice given three doses of Tam had PDL performed 2 weeks after the last Tam injection, as described (Xu et al., 2008) (Fig. 4A). An average ∼60-80% labeling efficiency was observed and strictly restricted to acinar cells (Fig. 3J); high labeling efficiency was deliberately pushed so as to detect minor populations of facultative progenitors. Pancreata were analyzed 5 days (PDL D5), 1 week (PDL D7), 4 weeks (PDL D30) or 8 weeks (PDL D60) post-PDL.

The ligated distal part of the pancreas (hereafter ‘PDL tail’), which consists of ∼80% of the splenic lobe, underwent massive acinar regression by PDL D7 (supplementary material Fig. S5B,D). The unligated proximal pancreas (hereafter ‘PDL head’, comprising ∼20% of the remaining unligated splenic lobe, intact gastric lobe and ventral pancreas) remained unchanged morphologically after ligation and served as a control for labeling efficiency (supplementary material Fig. S5C). After PDL, the total pancreatic epithelial area of PDL tail was dramatically reduced, while overall pancreatic weight remained similar because acini were replaced by fibrotic and inflammatory cells; islets and duct cells were spared and the animals remained euglycemic (supplementary material Fig. S5C-H) (Xu et al., 2008). Post-PDL D5, tubular complexes composed of Ck19⁺ ducts started to form in the PDL tail. Ptf1a protein was detected in these tubular complexes at this stage but not in the ducts of the PDL head, or the control sham-operated tail (hereafter ‘sham tail’; Fig. 4B-D; supplementary material Fig. S5A). The presence of Ptf1a protein in ductal cells proximate to the regressing acini might reflect detection of the early stages of acinar-to-ductal conversion. An alternate explanation is that remodeling ducts activated Ptf1a expression; a hypothesis ruled out by two observations: (1) the lack of Ptf1a protein by immunolabeling in the Ck19⁺/Hnf1β⁺ tubular complexes at PDL D7, D30 and D60 (supplementary material Fig. S6B-M); and (2) Tam treatment of Ptf1a^CreERTM;R26R^EYFP/EYFP mice post-PDL D7, after acini involution was almost completed, such that CreER would only label duct cells activating Ptf1a expression. By this strategy, extremely rare Ptf1a-lineage-labeled duct cells (∼1-2 EYFP⁺Ck19⁺ every three sections, possibly derived from late incorporation into the tubular complexes of the 1% acinar cells that escape involution) were found at post-PDL D30 and D60 (supplementary material Fig. S7E-K). Taken together, these results indicate that duct cells did not activate Ptf1a expression.

One week post-PDL, ∼99% of acinar cells in the PDL tail had involuted, and the remodeled ducts formed highly proliferative tubular complexes. The PDL tail was fibrotic and infiltrated with inflammatory cells (supplementary material Fig. S5A-D). As previously reported (Xu et al., 2008), we detected Ngn3 protein in Ck19⁺ duct cells in the PDL tail at post-PDL D7 (Fig. 4G; supplementary material Fig. S8A-A′) and D30 (Fig. 4M-O; Fig. S8B-B′). A low Ngn3 signal was detected in islet cells, as reported (Wang et al., 2009), but not in the ducts of the sham tail or PDL head tissues (Fig. 4E,F). An average of 74 (PDL D7) and 56 (PDL D30) Ngn3⁺Ck19⁺ duct cells were found per section (∼10-15 sections counted per PDL tail pancreas) (n=3; Fig. 4H). Scattered rare Ngn3⁺Ck19⁺ duct cells were also found at post-PDL D60 (not shown), consistent with the previous post-PDL analysis showing Ngn3 mRNA expression peaking at post-PDL D7 and reducing by later time points (Xu et al., 2008). These data also imply that the majority of the early Ngn3⁺Ck19⁺ protoendocrine cells have moved on to another differentiation state at later time points post-PDL. Notably, the Ngn3 signal was much higher in duct cells than in the islet endocrine cells (Fig. 4F,G, inset). These data imply that while Ngn3 is required for proper function/maintenance of adult endocrine cells (Wang et al., 2009), Ngn3⁺ duct cells arose from de novo activation of endocrine progenitor behavior upon injury, resembling the embryonic endocrine neogenesis process.

To determine whether acinar cells contribute to regeneration of the duct and endocrine pancreas after PDL, we used Ptf1a^CreERTM to lineage-trace the Ptf1a⁺ acinar cells and their progeny. In PDL tail post-PDL D7, we observed small numbers of EYFP⁺ cells in Ck19⁺ ducts, suggesting that some of the Ptf1a-lineage-labeled acinar cells escaped involution but became integrated into the tubular complex (Fig. 4Q-S; supplementary material Fig. S7A-D; Fig. S8F-H). The number of EYFP⁺Ck19⁺ duct cells increased substantially from post-PDL D7 to D60 (from 5.8±1.1% to 25.4±5.14% of total Ck19⁺ duct, respectively; n=3; supplementary material Fig. S7A-D,A′-C′), consistent with a continued high proliferative capacity as detected by BrdU incorporation (supplementary material Fig. S5E-G,I). EYFP⁺Ck19⁺ duct cells were also found to produce Hnf1β (Fig. 4Q-S; supplementary material Fig. S8C-E) and Sox9 (supplementary material Fig. S8F-K) or both (supplementary material Fig. S8L-N), at post-PDL D7 and D30. Detection of Hnf1β and Sox9 protein in these EYFP⁺ ducts at post-PDL D7 suggests that, after PDL, the Ptf1a⁺ acinar cells de-differentiated to a more primitive epithelium resembling the embryonic trunk epithelium of the 2° transition. Furthermore, we did not observe any amylase⁺Hnf1β⁺ transitional cell state at PDL D7 (supplementary material Fig. S9A,B), suggesting that acinar enzymes were downregulated before the activation of multipotency factors.

The presence of EYFP⁺Ngn3⁺Ck19⁺ duct cells in the tail region post-PDL D7 (Fig. 4I-K) and D30 (Fig. 4M-O) suggested their derivation from Ptf1a⁺-acinar-derived duct cells. Of total Ngn3⁺Ck19⁺ duct population, 21.9±2.4% (PDL D7) and 19.4±3.3% (PDL D30) were EYFP⁺Ngn3⁺Ck19⁺ (Fig. 4P). Notably, not all EYFP⁺ duct cells were Ngn3⁺, indicating either: (1) incomplete R26R^Eyfp lineage labeling; or (2) not all Ngn3⁺ duct cells originated from Ptf1a-lineage cells (possibly originating from the duct proper); and/or (3) while some Ptf1a⁺-acinar-derived new duct cells went on to activate Ngn3 expression as they commit toward the endocrine lineage, others remain stably integrated as mature duct epithelial cells. The EYFP⁺Ngn3⁺Ck19⁺ cells comprised 38% of the total EYFP⁺Ck19⁺ cell number at post-PDL D7 and became reduced to 8% at post-PDL D30 (n=3; Fig. 4L). It is possible that there was a higher proliferative capacity of Ptf1a-lineage-derived (EYFP⁺Ck19⁺) duct cells, compared with protoendocrine EYFP⁺Ngn3⁺Ck19⁺ cells. Taken together, these data suggest that after PDL, significant numbers of Ptf1a⁺ acinar cells trans-differentiated into ductal cells with a portion of these duct cells being competent to activate expression of the endocrine commitment factor Ngn3.

To assess whether Ptf1a⁺ cells can give rise to endocrine cells after injury, particularly focusing on β-cells, pancreata from Ptf1a^CreERTM;R26R^EYFP/EYFP mice lineage-labeled by Tam treatment prior to PDL were analyzed, first at post-PDL D30 (Fig. 5A). In order to increase the sensitivity of detecting a potentially small population of endocrine cells derived from Ptf1a⁺ facultative progenitors, we performed an extensive, systematic analysis across a large portion of each pancreas. Thus, we analyzed ∼40-45 sections from each mouse (∼10% of total sections from each tail or head region) spaced evenly across the entire PDL tail, sham tail and PDL head region, and visually scanned the entire section for EYFP⁺endocrine-hormone⁺ cells (with an insulin, glucagon, somatostatin and pancreatic-polypeptide antibody cocktail) or EYFP⁺insulin⁺ cells (see supplementary material Fig. S10 for quantitation strategy). No EYFP⁺ endocrine cells were observed in the sham tail or PDL head after PDL (Fig. 5B,C,N,O), solidifying the absence of leakiness of Ptf1a^CreERTM allele and consistent with the adult lineage-tracing above (Fig. 3J). We did not detect any EYFP⁺endocrine-hormone⁺ cells or EYFP⁺insulin⁺ cells either in islets or ducts proper in post-PDL D30 tail (n=4; Fig. 5D,E).

To investigate rigorously whether new endocrine/β-cells could be generated from Ptf1a⁺ cells in a more chronic timeframe post-PDL, we analyzed Tam-treated Ptf1a^CreERTM;R26R^EYFP/EYFP pancreata 8 weeks post-PDL (D60), a period substantially longer than most previous PDL studies. More Ck19⁺ duct clusters than at post-PDL D60 were EYFP⁺. Many foci of tubular-complexes, individually almost completely labeled (Fig. 5L), were distributed throughout the tissue and accounted for 25.4±5.14% of the total Ck19⁺ duct area. This representation was 11.6±1.81% at post-PDL D30, suggesting that proliferation and expansion of the Ptf1a-lineage-labeled duct/tubular-complex cells continued from D30 to D60 (n=3; supplementary material Fig. S7B-D). We also detected in post-PDL D60 tissue a small but significant number of EYFP⁺endocrine-hormone⁺ cells in Ck19⁺ ducts (Fig. 5H), as well as in islets (Fig. 5I). Around 75% of the EYFP⁺hormone⁺ cells were located in/at the ducts, whereas ∼25% were found in the islets. Some of these endocrine cells, either in the ducts or in islets, were also insulin-producing (Fig. 5L,M), suggesting that β-cells can be derived from Ptf1a⁺ acinar cells after PDL. Most importantly, these Ptf1a-lineage-derived insulin⁺ cells produced several mature β-cell transcription factors, such as Pdx1^HI (Fig. 7A,B), Nkx6.1 (Fig. 7E,F), and MafA (Fig. 7I,J), indicating a mature β-cell status. Most of the EYFP⁺insulin⁺ cells were presents as singlet β-cells located at/in the ducts and did not produce GLUT2 (data not shown). The lack of β-β cell contact may prevent membrane localization of GLUT2 in these EYFP⁺ singlet β-cells as previously suggested (Orci et al., 1989; Gannon et al., 2000). Extensive, systematic quantification revealed 965±393 EYFP⁺hormone⁺ endocrine cells, per PDL tail analyzed (n=4; Fig. 5N). We only observed one to eight islets per 10% of PDL tail containing EYFP⁺ endocrine cells; in such islets only ∼5-10% of the cells were EYFP⁺ (n=4; Fig. 5O; data not shown). Because ∼99% of acinar cells were lost at post-PDL D7, and the persistent blockage of the pancreatic duct limits acinar regeneration, we believe that these EYFP⁺ endocrine cells were originated from Ptf1a⁺ acinar-derived duct cells, rather than arising directly from Ptf1a⁺ acini as no amylase⁺insulin⁺ transitional intermediate was found at PDL D7, D30 and D60 (supplementary material Fig. S9C-H). We did not detect any EYFP⁺ cells in the endocrine compartment for either sham tail or PDL head at post-PDL D60 (Fig. 5F,G,J,K,N,O). These in vivo lineage-tracing data support the proposal that Ptf1a⁺ acinar cells could be stimulated to become facultative progenitors and undergo a long-term (albeit inefficient) reprogramming to endocrine cells, via a ductal intermediate, in the PDL-injury/regeneration context (Fig. 5P).

Because islet cells were spared and mice remained euglycemic after PDL, we propose that the lack of hypothetical physiological stimuli to produce additional β-cells could explain the low level conversion to endocrine cells from Ptf1a⁺ acini-derived duct cells. We reasoned that eliminating some pre-existing β-cells could provide one such physiological cue (perhaps including the ensuing hyperglycemia), and addressed the deep injury context of PDL plus β-cell destruction. Tam-treated Ptf1a^CreERTM;R26R^EYFP/EYFP mice were given two consecutive doses of STZ three days post-PDL to kill a portion of β-cells, and pancreata were analyzed 30 and 60 days post-PDL (Fig. 6A). We observed ∼50-70% reduction in β-cell numbers with this protocol, and mice became hyperglycemic (ad lib feeding blood glucose level >450 mg/dl) 2-3 days after STZ treatment (supplementary material Fig. S10B).

To determine if [PDL+STZ] treatment could speed up acinar-to-endocrine conversion, we analyzed pancreata post-PDL D30. [PDL+STZ]-treated mice remained hyperglycemic at this time (supplementary material Fig. S10B). We did not treat mice with any β-cell-proliferation-stimulating agents (e.g. insulin, GLP or EGF/gastrin), under the logic that avoiding such agents might promote transdifferentiation and disfavor replication of β-cells. We found EYFP⁺endocrine-hormone⁺ cells mostly located either in ducts, or duct-apposed, in [PDL+STZ] tail tissue already by post-PDL D30, notably different from PDL alone at D30 (Fig. 6D; for more examples see supplementary material Fig. S11A). No EYFP⁺endocrine-hormone⁺ cells were found in the PDL head or STZ-treated sham tail (Fig. 6B,C), suggesting strongly that STZ alone did not induce acinar-to-endocrine transdifferentiation. The numbers of EYFP⁺endocrine-hormone⁺ cells found in [PDL+STZ] D30 tail tissue were similar to those found post-PDL D60 tail without STZ treatment (average 990±149 EYFP⁺endocrine-hormone⁺ cells per PDL tail in [PDL+STZ] D30, compared to 965±393 EYFP⁺ endocrine-hormone⁺ cells per PDL tail at PDL D60; n=3 separate animals; compare Fig. 6N with Fig. 5N). At post-[PDL+STZ] D60, the numbers of EYFP⁺endocrine-hormone⁺ cells had increased ∼3-4.5 fold over D30 (average 3653±1205 EYFP⁺endocrine-hormone⁺ cells per PDL tail tissue; n=3; Fig. 6N). EYFP⁺insulin⁺ cells were also found at post-[PDL+STZ] D30 and D60 (Fig. 6L,M; see more examples in supplementary material Fig. S11A,B). These EYFP⁺insulin⁺ cells were also Pdx1HI (Fig. 7C,D), Nkx6.1⁺ (Fig. 7G,H), and Mafa⁺ (Fig. 7K,L), suggesting that they are mature β-cells. Notably, fewer EYFP⁺endocrine-hormone⁺ cells were located within islets under the [PDL+STZ] condition compared to PDL alone at day 60 (average ∼1 islet per 10% PDL tail labeled in post-[PDL+STZ] D60, versus average ∼5 islets per 10% PDL tail in post-PDL D60; compare Fig. 5O; Fig. 6O). We hypothesize that repellents or toxics released from apoptotic β-cells after STZ treatment might render the islets unattractive to home Ptf1a-lineage-derived endocrine cells. Taken together, these findings demonstrated that the deep lesion enhanced acinar-to-endocrine transdifferentiation rate and efficiency compared to PDL alone.

We are still far from understanding how dynamic cell heterogeneity and intercommunication connect to epithelial morphogenesis and cell-fate allocation, and if certain mature cells retain homeostatic or facultative multipotency/plasticity. Our immuno-colocalization of TFs considered as progenitor-defining revealed heterogeneity as epithelial morphogenesis begins at E11.5-12.5, well before the 2° transition stage. The putative Ptf1a⁺Sox9⁺Hnf1β⁺ MPCs (the same signature as early MPCs), possibly under the influence of localized signals from the mesenchyme, become localized to the tip epithelium and progressively diminish between E12.5 and E15.5. The reduction in these putative MPCs parallels the quantitative decrease in progeny of Ptf1a-expressing cells that enter non-acinar lineages. After the 2° transition (E15.5 onward), Ptf1a⁺ MPC activity becomes very low, and Ptf1a-expressing cells are acinar-unipotent by late gestation and in adults.

More importantly, we showed that PDL or [PDL+STZ] induced restoration of an early-organogenesis-like MPC competence in adult acinar cells, with probably stepwise redifferentiation into ducts then endocrine cells. Ptf1a⁺ acinar cells reactivated embryonic multipotency markers, differentiating rapidly into long-lived duct cells. Some Ptf1a-derived duct cells activated Ngn3, and some Ptf1a-lineage-derived cells progressed to endocrine cells, including Pdx1^HI/Nkx6.1⁺/MafA⁺/insulin⁺ β-cells. This conversion occurred without exogenous factors such as β-cell-instructive TFs or intercellular signaling factors.

We have more carefully enumerated the multi-to-unipotential switch during organogenesis, with findings largely consistent but with novel aspects compared to the previous CpaI^CreER study (Zhou et al., 2007), but it is also important that we used an MPC-instructive gene Ptf1a, and not a Ptf1a surrogate encoding a digestive enzyme, which should have no regulatory role in MPC formation/maintenance. Furthermore, our E10.5 immuno-colocalization data (supplementary material Fig. S12) showed CpaI in ∼65% of Ptf1a⁺ cells (all CpaI⁺ cells were Ptf1a⁺), suggesting CpaI only partially marks MPC. Our Ptf1a^CreERTM-based analysis validates the CpaI^CreER study but elaborates a stronger understanding of the dynamic nature of MPC location during tip-trunk segregation in the 2° transition.

The CpaI^CreER study proposed that CpaI⁺ tip MPC contribute significantly to the expansion of the trunk epithelium (endocrine/duct bipotent progenitor compartment) after E12, but we found that Ptf1a-expressing cells seeded the trunk epithelium until E13.5, with only a small contribution at/after ∼E12.5. We can conclude that tip seeds trunk is a minor influence during the true 2° transition, consistent with the proposal that intra-trunk proliferation and plexus remodeling (Villasenor et al., 2010) cause most of the growth and extension of the trunk epithelium.

We note that the Ptf1a⁺Pdx1⁺CpaI⁺cMyc^HI MPC proposed in Zhou et al. (Zhou et al., 2007) did not refer to tip heterogeneity, and in fact differentiated acini retain that signature. Solar et al. (Solar et al., 2009) suggested that Hnf1β⁺ trunk cells are multipotent between E11.5-13.5, but did not rule out that Hnf1β^CreER-traced MPC could be an Hnf1β⁺CpaI⁺ tip subset. Furthermore, considering the tip-trunk interface Nkx6.1⁺Ptf1a⁺Sox9⁺Pdx1⁺ MPC proposed by Kopp et al. (Kopp et al., 2011), which lacked Hnf1β, the Ptf1a⁺Sox9⁺ putative tip MPC we detected were Hnf1β⁺, and distributed heterogeneously, both intra-tip and at the tip-trunk interface. The Ptf1a⁺Sox9⁺Hnf1β⁺ tip MPC could be the same as those detected by Solar et al. (Solar et al., 2009) and Kopp et al. (Kopp et al., 2011), leading to the generalization that multipotentiality is preserved in the tip, not trunk, domain. What controls tip-trunk compartmentalization, and which signals are present in the local tip microenvironment that determine this MPC niche, represent interesting aspects for future studies.

Postnatal Ptf1a⁺ cells only produced acinar cells, agreeing with the results of Elastase^CreERT2 studies showing homeostatic maintenance of this compartment comes from acinar replication (Desai et al., 2007). A Sox9^CreER knock-in study (Furuyama et al., 2011) seems contradictory, in suggesting that ∼80% of acinar cells came from Sox9⁺ duct or centroacinar cells following a one-year post-Tam chase. We found no significant reduction in the proportion of Ptf1a^CreERTM-labeled acini after a 6-month chase, arguing against dilution from other cell populations. This discrepancy could be reconciled, bearing in mind the lower level Sox9 expression in adult acini (Kopp et al., 2011; Dubois et al., 2011) (Fig. 3M,M′), such that the high tamoxifen dosage used (5′4 mg) in Furuyama et al. (Furuyama et al., 2011), would broadly and directly activating the reporter allele in acini.

The mass of acinar cells and their relatively close developmental history with endocrine cells lead to ideas that partial reprogramming in vivo might offset a β-cell deficit without compromising organ function (Pan and Wright, 2011). Zhou et al. (Zhou et al., 2008) reported reprogramming of acini to β-like fate using a viral cocktail of β-cell TFs: Pdx1, Ngn3 and MafA. Our evidence suggests that Ptf1a⁺ acinar cells possess facultative progenitor potential under PDL, allowing them quickly to produce long-lived ducts and later-arising endocrine/β-cells. Several in vitro studies suggested acinar-to-β-cell conversion with EGF, nicotinamide, TGFα or combination(s) treatment (Baeyens et al., 2005; Minami et al., 2005). Non-autonomous effects of tissue dissociation might substantially affect in vitro conversion. Genetic lineage-tracing was not included in the previous reports of in vivo acinar-to-endocrine transdifferentiation (purportedly involving amylase⁺insulin⁺ intermediates) in 24-hour post-PDL rats (Bertelli and Bendayan, 1997). Our study therefore provides rigorous in vivo evidence for PDL-induced acinar-to-endocrine conversion. We contradict the conclusion from Elastase^CreERT2 lineage-tracing in vivo that acini did not seed duct or endocrine fates after Ppx, PDL and cerulean-induced pancreatitis (Desai et al., 2007), but the 30% acinar labeling efficiency (compared to our ∼80% labeling, assessing large amounts of the organ and over a long period) might have failed to detect low-level contributions.

If there is any cryptic destabilization of the acinar program from Ptf1a heterozygosity (Ptf1a^CreERTM is a null allele), there are consequences only under injury, and Ptf1a^CreER/+ mice have normal embryonic and adult pancreas. Distinct Ptf1a levels might control the interaction with the RBP-J or RBP-JL partner in the PTF1 complex. Ptf1a associates with RBPJ (PTF1^RBPJ) in MPC, but with RBPJL (PTF1^RBPJL) in proacinar/acinar cells (Masui et al., 2007), with the PTF1^RBPJ-to-PTF1^RBPJL switch likely essential for facilitating epithelial tip-trunk compartmentalization (Masui et al., 2007; Schaffer et al., 2010). Under injury, reduced Ptf1a levels could allow reformation of PTF1^RBPJ, which together with Sox9/Hnf1β production leads to reacquisition of MPC character. Future studies will be directed toward understanding how differential Ptf1a levels affect multipotency restoration in Ptf1a⁺ acini during injury.

Acinar-to-endocrine conversion could be aided by disturbances in the extracellular environment, including the chronic inflammation-associated signals in PDL. Cytokines may also influence pancreas regeneration (Homo-Delarche and Drexhage, 2004), and invading macrophages may suppress islet loss during exocrine degeneration (Tessem et al., 2008). Wang et al. (Wang et al., 2007) suggested a role for immune-reaction effects against adenovirus in liver-to-β-like cell conversions driven by pro-endocrine TFs (Pdx1, Ngn3). Our Ptf1a-lineage-labeled endocrine cells were often found next to duct-apposed small endocrine clusters. Such clusters are distributed throughout PDL pancreas, although some might arise from duct cells proper (Xu et al., 2008). Speculatively, acinar-derived endocrine cells could home to these small endocrine clusters rather than to islets proper, or endocrine cell migration from their birthplace in the tubular complexes might be hindered by the altered fibrotic stroma.

While we showed that Ptf1a⁺ acinar cells produced Ck19⁺Hnf1β⁺Sox9⁺ primitive ductal intermediates, Ngn3⁺ endocrine progenitors and hormone⁺ endocrine cells, showing directly a sequential movement through these transitional states would need either real-time tracing, or dual/multiplex genetic lineage-tracing methods (e.g. combining Cre/LoxP, Dre/Rox or FLPe/FRT). We did not observe any pan-endocrine/insulin co-label with acinar markers arguing against a direct rapid conversion from acini to endocrine cells.

Reproliferative CACs have been suggested as facultative progenitors under injury. Rovira et al. (Rovira et al., 2010) detected Ptf1a by RT-PCR in Aldh1⁺ CACs, and MPC competence in CAC-derived pancreatospheres. We failed to detect Ptf1a protein in CACs, and adult lineage-tracing with Ptf1a^CreERTM did not label CACs. It was still possible that Ptf1a expression could be activated in CACs post-PDL, which incorporated into tubular complexes and yielded duct/endocrine cells. However, Tam-induction at post-PDL D7 (when acinar involution is essentially complete) only labeled extremely scarce, if any, duct cells, and no endocrine cells, at post-PDL D30 and D60, strongly arguing against any relevant post-PDL acquisition of Ptf1a expression by CAC or duct. Lineage-tracing using lines that label CACs (Hnf1β^CreERT2, Sox9^CreERT2, Hes1^CreERT2) showed no evidence for endocrine cell production post-PDL (Solar et al., 2009; Kopp et al., 2011; Kopinke et al., 2011). The simplest conclusion is that our labeled endocrine cells originated from sequential acinar-to-duct-to-endocrine transdifferentiation.

β-Cell destruction by STZ enhanced PDL-induced transdifferentiation. We reasoned that the relative scarcity under PDL alone reflected the absence of a ‘call for more β-cells’. One candidate signal under [PDL+STZ] is glucose, a stimulant of β-cell replication after ablation (Porat et al., 2011). Under [PDL+STZ] condition, we observed no significant restoration of β-cell mass after 60 days nor improved glycemia (few β-cells being derived from Ptf1a⁺ cells). Future study could determine the signals (hyperglycemia and inflammatory signals from the attacked islets) that contribute to the increased stimulus for acinar to endocrine transformation under [PDL+STZ]. And, if reinduction of MPC character in the adult under injury recapitulates any of the steps of earlier embryonic endocrine ontogeny.

[PDL+STZ]-induced endocrine cell interconversion has been reported. Chung et al. (Chung et al., 2010) proposed an α/β intermediate as a precursor for β-cells post-injury, although there was no lineage-tracing analysis. Another non-PDL study (Thorel et al., 2010) used lineage tracing to show α-to-β conversion after near-total β-cell elimination. In our study, only a subset of cells derived from Ptf1a-expressing facultative progenitors were mature β-cells, and we are unable to comment on interconversion between hormone cell types.

Our evidence that adult pancreatic acini are facultatively plastic provides impetus for determining how injury epigenetically destabilizes the acinar program, influence of niche perturbation and signals from inflammatory cells, and if only certain acinar cells are conversion-competent. The low rate of endocrine conversion could be analogous to generating induced pluripotent stem cells (iPSCs), where a tiny fraction of the initial differentiated population is captured via their induced growth potential; indeed, endocrine conversion might increase under strategies similar to those applied to iPSCs. Our findings might lead to pharmacological intervention using small molecules that modulate the epigenetic landscape or mimic inflammatory signaling. Controlling plasticity might allow regulated reprogramming as an alternative to cell transplantation.

We thank Jill Lindner and Susan Hipkens for performing RMCE; Maike Sander, Ben Stanger, Jane Johnson for Ngn3, Ck19 and Ptf1a antibodies, and Guoqiang Gu and Wright Lab members for discussions. Immunofluorescent images were gathered in part using the VUMC Cell Imaging Shared Resource (supported by NIH grants CA68485, DK20593, DK58404 and DK59637).