Aldh1-Expressing Endocrine Progenitor Cells Regulate Secondary Islet Formation in Larval Zebrafish Pancreas

Aldh1 expression is known to mark candidate progenitor populations in adult and embryonic mouse pancreas, and Aldh1 enzymatic activity has been identified as a potent regulator of pancreatic endocrine differentiation in zebrafish. However, the location and identity of Aldh1-expressing cells in zebrafish pancreas remain unknown. In this study we demonstrate that Aldh1-expressing cells are located immediately adjacent to 2F11-positive pancreatic ductal epithelial cells, and that their abundance dramatically increases during zebrafish secondary islet formation. These cells also express neurod, a marker of endocrine progenitor cells, but do not express markers of more mature endocrine cells such as pax6b or insulin. Using formal cre/lox-based lineage tracing, we further show that Aldh1-expressing pancreatic epithelial cells are the direct progeny of pancreatic notch-responsive progenitor cells, identifying them as a critical intermediate between multi-lineage progenitors and mature endocrine cells. Pharmacologic manipulation of Aldh1 enzymatic activity accelerates cell entry into the Aldh1-expressing endocrine progenitor pool, and also leads to the premature maturation of these cells, as evidenced by accelerated pax6b expression. Together, these findings suggest that Aldh1-expressing cells act as both participants and regulators of endocrine differentiation during zebrafish secondary islet formation.


Introduction
In mammals, pancreatic endocrine differentiation occurs through a defined sequence of progenitor cell types, which undergo progressive lineage restriction [1][2][3]. Detailed elaboration of this step-wise differentiation program has allowed progress towards the guided programming of human stem cells towards a β-cell fate [4], potentially providing a source of insulin-producing cells suitable for cell replacement therapy in diabetes.
In contrast to developmental programs responsible for formation of the principle islet, secondary islet formation in zebrafish appears to occur in a manner more similar to that observed during mammalian islet formation. As in mammals, notch-responsive cells associated with developing zebrafish ductal epithelium generate endocrine progenitor cells that delaminate and proliferate prior to differentiating and coalescing to form organized islets [10,20]. These zebrafish pancreatic notch-responsive cells (PNCs) have been implicated as precursors for α, β, and δ endocrine lineages, and are subject to precocious endocrine differentiation following pharmacologic or genetic inhibition of notch signaling [10,12,20,21]. However, the intermediate cell states associated with generation of mature endocrine cells from PNCs remain incompletely characterized.
In addition to notch, recent studies have also implicated retinoic acid signaling in the regulation of endocrine differentiation during zebrafish secondary islet formation. While Aldh1 enzymatic activity and associated retinoic acid production are known to exert an early positive influence on the pancreatic progenitor field [22][23][24], they have recently been shown to exert a negative influence on β-cell differentiation during secondary islet formation [25]. In addition to this functional data in zebrafish, Aldh1 isoform expression and enzymatic activity have also been utilized to mark candidate progenitor populations in murine pancreas [26,27]. Using a variety of transgenic lines to mark pancreatic cell types and lineages in embryonic and larval zebrafish, we have confirmed the presence and identity of Aldh1-expressing progenitor cells during zebrafish secondary islet formation, and have further placed these cells within a multi-step endocrine differentiation paradigm. Using both Cre/lox lineage tracing and pharmacologic manipulation, we further demonstrate that Aldh1 enzymatic activity regulates both entry and exit from the Aldh1expressing progenitor compartment. These results identify Aldh1-expressing cells as both a participant and a regulator of endocrine differentiation during zebrafish secondary islet formation.
Fluorescent images were acquired with Nikon A1 scanning confocal microscope, and areas and volumes of labeling for specific markers were calculated using NIS-Elements software.

FACS sorting
Pancreatic tissue was dissected from adult AB and Tg(TP1bglob:eGFP) fish, incubated in 0.5% bleaching solution for 2 min and digested in 1.4 mg/ml collagenase-P at 37 °C for 20 min. Following multiple washes with HBSS supplemented with 5% FBS, collagenase-digested pancreatic tissue was filtered through 500 µm polypropylene mesh (Spectrum Laboratories), then spun through a 30% FBS cushion. For FACS, pelleted-pancreatic tissue were resuspended in diluted trypsin (0.05%) (Invitrogen) and incubated at 37 °C for 15 min. Dispersed cells from AB and Tg(TP1bglob:eGFP), were then directly resuspended in Aldefluor buffer and 0.8 x PBS, respectively. To label Aldh1-positive cell for FACS-sorting, the Aldefluor Kit (Stem Cell Technologies) was utilized according to the manufacturer's protocol. Flow cytometry was performed using a FACSAria (Becton Dickinson) flow cytometer.

Cell labeling with EdU
The Click-iT EdU Alexa Fluor647 Imaging Kit (Invitrogen) was utilized according to the manufacturer's protocol. Larvae were incubated in 1 mM EdU/1 %DMSO in embryo medium for 2 hours, fixed in 4% PFA in PBS at 4°C overnight, immersed in 30% sucrose/PBS, embedded in OCT compound, frozen in liquid nitrogen, and sectioned at 10 mm using a cryostat. After EdU staining according manufacturer's protocol, additional immunostaining was performed. The percentage of EdU+ cells in relevant tissue compartments was then determined at 15, 20 and 25dpf, using 10 fish at each developmental stage and a minimum of 3 pancreatic sections per fish.

Aldh1-positive cells progressively accumulate during zebrafish secondary islet formation
In order to evaluate the presence and possible function of Aldh1-expressing cells during secondary islet formation in larval zebrafish pancreas, we first surveyed Aldh1 expression in conjunction with other lineage markers. As previously reported, secondary islet formation in zebrafish larvae occurs between 5 and 20 dpf, with approximately 50% of larvae having pax6b-expressing secondary islet tissue by 10dpf, and essentially 100% of larvae having discernible secondary islet tissue by 20 dpf [10]. Between 5-15 dpf, pancreatic Aldh1 immunoreactivity was confined to the principle islet, where Aldh1-positive cells were often observed in proximity to Insulinpositive β-cells ( Figure 1A-C). By 20 dpf, small numbers of Aldh1-positive cells also became detectible outside of the principle islet. These cells were frequently observed in association with forming ductal epithelium, and were uniformly negative for ptf1a:eGFP, a marker of exocrine cells ( Figure 1D-F). They were also negative for Insulin as assessed by double immunofluorescent labeling, but were frequently observed to be immediately adjacent to Insulin-positive cells ( Figure 1E). When quantified as the total number of cells per section (n=10 fish with minimum 3 pancreatic sections per fish), non-principle islet associated Aldh1-positive cells were found to steadily increase between 15 and 30dpf ( Figure 1G). Beginning at 20 dpf, a similar steady increase in the number of non-principle islet associated Aldh1-positive cells in direct contact with Insulin-positive cells was also observed, raising the question of whether these duct-associated Aldh1-positive cells might represent a population of β-cell precursors.

Aldh1-positive cells express markers of endocrine progenitor cells
To further characterize Aldh1-positive cells in larval zebrafish pancreas, we examined Aldh1 expression in conjunction with additional endocrine markers. These included Tg(pax6:GFP) ulg515 (hereafter referred to as pax6b:GFP), a panendocrine marker [16], Tg(-2.6mnx1:GFP) ml5 (hereafter referred to as hb9:GFP), a marker of early β-cells [32]; and TgBAC(neurod:eGFP) nl1 , (hereafter referred to as neurod:eGFP), a marker of endocrine progenitor cells as well as differentiated endocrine cells [33]. These analyses revealed that, between 15 and 30dpf, Aldh1 immunoreactivity marked cells also labeled with neurod:eGFP, but not hb9:eGFP or pax6b:GFP (Figure 2A-L). While many Aldh1-negative, neurodpositive cells were identifiable within islet tissue, virtually all Aldh1-positive cells also expressed neurod:eGFP. Because neurod:eGFP is felt to label cells undergoing early commitment to the endocrine lineage, these findings suggest that Aldh1 is transiently expressed in at least a fraction of early endocrine progenitor cells.

Aldh1-positive endocrine progenitor cells are associated with ductal epithelium
Previous studies have suggested that, as in mammals, secondary islet formation in zebrafish originates from progenitor cells located within the forming ductal epithelium [10,20,21,34]. We therefore examined the relationship between Aldh1-positive, neurod-positive cells and labeling of 2F11, a marker of the forming pancreatic ductal epithelium [11,35,36]. These analyses revealed that all Aldh1-positive cells were also labeled with ( Figure 3A-D). 2F11 also labeled Aldh1-negative, ptf1a:eGFP-negative tubular structures ( Figure 3C), which were found to be distinct from vascular endothelium marked by kdrl:GRCFP ( Figure 3E). In further evaluating the overlap between 2F11 labeling and the expression of Aldh1, neurod and insulin in larval zebrafish pancreas, we further observed that 2F11 also marked Aldh1-positive cells expressing neurod ( Figure 3F and F'), as well as Aldh1-negative cells also expressing pax6b:GFP ( Figure 3G and G'), insulin ( Figure 3G and G'), and hb9 ( Figure 3H). Using Edu incorporation to mark proliferating cells, we further identified a subset of 2F11 cells undergoing active proliferation ( Figure 4A-F). However, EdU incorporation was limited to 2F11 cells not co-expressing Aldh1, neurod or pax6b, suggesting that commitment of 2F11 cells to the endocrine lineage is marked by relative or complete cell cycle exit. Further evaluation of the detailed spatial relationships between cells expressing the 2F11 epitope, Aldh1, neurod and pax6b suggested that proliferating 2F11 pos /EdU pos /Aldh1 neg /neurod neg cells were incorporated into tubular ductal epithelial structures, while non-proliferating 2F11 pos /EdU neg /Aldh1 pos /neurod pos endocrine progenitor cells appeared to delaminate directly from this epithelium ( Figure 4F, F'), in a manner similar to that described in mammalian pancreas development.

Aldh1-positive endocrine progenitor cells arise from Notch-responsive ductal progenitor cells
Within the emerging ductal epithelium, a subset of cells characterized by active Notch signaling are known to serve as multi-lineage pancreatic progenitor cells [10,21], referred to as pancreatic notch-responsive cells, or PNCs. We therefore sought to determine the relationship between PNCs, 2F11 labeling and Aldh1 expression in larval zebrafish pancreas. Using either Tg(Tp1bglob:eGFP) um14 or Tg(Tp1bglob:mCherry) jh11 to mark PNCs at 20 and 25dpf, we found that PNCs were 2F11-positive, and frequently incorporated EdU ( Figure S1A-D). In contrast, we observed no overlap between Notch reporter expression and Aldh1 labeling ( Figure S1E), suggesting that PNCs and Aldh1 pos cells represented distinct cell compartments. To further investigate potential overlap between Aldh1 pos cells and PNCs, we FACS sorted PNCs from Tg(Tp1bglob:eGFP) um14 adult zebrafish pancreas, and assessed Aldh1 expression following cytospin using immunofluorescent labeling. This revealed that Aldh1 expression was restricted to the eGFP neg , non-PNC population ( Figure S2C, D). In contrast, Aldh1 pos cells sorted using the Aldefluor reagent [27] displayed strong immunoreactivity for Aldh1, further validating use of this Aldh1 antibody in zebrafish ( Figure S2A, B). Additional analysis of gene expression using RT-PCR on FACS-sorted PNCs and FACS-sorted Aldh1 pos cells isolated from adult fish confirmed unique patterns of gene expression between these two subpopulations, with Aldh1 pos cells expressing high levels of aldh1a2 and low levels of the general stem cell markers cd133 and sca1 ( Figure S2 E), while PNCs expressed low levels of aldh1a2, high levels of cd133 and moderate levels of scal1 ( Figure S2B). Together, these results suggest that the Aldh1 pos and PNC subpopulations represent distinct sets of endocrine progenitor cells.
In order to determine whether Notch activation and Aldh1 expression might represent sequential steps in pancreatic progenitor cell differentiation, we next performed Cre/lox based lineage tracing of the PNC lineage, as previously described [21]. For these studies, we utilized the Tg(Tp1glob:creERT2) jh12 and Tg(βactin:loxP-stop-loxP-hmgb1-mCherry) jh15 alleles, and transiently activated Cre activity using 4-hydroxy-tamoxifen (4OHT) at either 3-5 dpf or 18-20 dpf, followed by harvest at 25dpf ( Figure 5A). These studies confirmed that PNCs labeled at either 3-5 dpf or 18-20 dpf gave rise not only to later appearing endocrine cells and ductal cells as previously reported, but also gave rise to Aldh1-positive cells ( Figure 5B-E), allowing us to order PNCs and Aldh1-positive, neurodpositive endocrine progenitor cells as sequential populations arising during endocrine differentiation. Together with the spatial labeling patterns observed for Aldh1, neurod and pax6b expression as described above, these data suggest a stepwise model of endocrine differentiation in the formation of Aldh1-Expressing Pancreatic Progenitor Cells PLOS ONE | www.plosone.org zebrafish secondary islet formation, as depicted schematically in Figure 5F.

Inhibition of Aldh1 enzymatic activity increases the number of Aldh1 pos endocrine progenitor cells and induces their premature differentiation
Prior studies have demonstrated that, in addition to exerting a positive influence on the early pancreatic progenitor field [22][23][24], Aldh1 enzymatic activity and associated retinoic acid production exert a negative influence on β-cell differentiation during secondary islet formation [25]. We therefore examined how inhibition of Aldh1 enzymatic activity influenced Aldh1 cell behavior in conjunction with Notch signaling and neurod expression in larval Tg(Tp1bglob:hmgb1-mCherry) jh11 ; Tg(neurod:eGFP) zebrafish pancreas. Using the Aldh1 inhibitor DEAB, we observed that inhibition of Aldh1 enzymatic activity between 72 hpf and 120 hpf was associated not only with premature activation of neurod expression in PNCs, but also with a marked increase in the number of Aldh1-positive cells ( Figure 6A-D). When Aldh1 enzymatic activity was inhibited between 18 and 20 dpf, a similar increase in neurod expression was observed in PNCs ( Figure 6E, F and I). Pax6b-expressing cells also increased in number, although pax6b expression was confined to cells not expressing Tp1:mCherry ( Figure 6I). Activation of not only neurod, but also pax6b expression was observed in Aldh1-positive cells ( Figure 6E, F and J), suggesting an additional negative influence of Aldh1 enzymatic activity on the further differentiation of Aldh1-positive, neurodpositive endocrine progenitor cells.

Discussion
In this study, we have identified a low-abundance population of Aldh1-expressing cells in larval zebrafish pancreas. These cells increase in abundance in association with secondary islet formation and are frequently located within the pancreatic ductal epithelial tree adjacent to insulin expressing cells, consistent with possible endocrine progenitor function. This Aldh1-Expressing Pancreatic Progenitor Cells PLOS ONE | www.plosone.org conclusion is further supported by the fact that Aldh1expressing cells co-express the endocrine progenitor marker neurod, but do not normally co-express either insulin, pax6 or other genes typically expressed in more mature endocrine cells. In addition, their abundance is significantly increased by inhibition of Aldh1 enzymatic activity, a manipulation known to promote the precocious differentiation of pancreatic progenitor cells and the acceleration of secondary islet formation [25]. Using formal Cre/lox lineage tracing, we also show that, like differentiated endocrine cells [21], Aldh1-expressing cells are the progeny of pancreatic notch-responsive progenitor cells (PNCs) located within the forming ductal system of the larval zebrafish pancreas. In addition to accelerating the formation of Aldh1-expressing cells from PNCs, inhibition of Aldh1 enzymatic activity using DEAB results in the premature activation of pax6 expression among cells expressing Aldh1.
These observations suggest that the increase in secondary islet formation observed following inhibition of Aldh1 enzymatic activity is associated with the premature activation of neurod and Aldh1 expression among PNCs, as well as the accelerated maturation of Aldh1-positive, neurod-positive cells, reflected by activation of pax6b expression. These findings place Aldh1-positive, neurod-positive cells as a critical intermediate between PNCs and more differentiated endocrine cells, and further suggest that Aldh1 enzymatic activity may play an important functional role in regulating both entry to and exit from this intermediate progenitor compartment.
Based on these observations, we now propose a model of zebrafish secondary islet formation, in which retinoic acid plays an integral role in controlling the progression of pancreatic progenitor cells to a differentiated endocrine phenotype ( Figure  7). During secondary islet formation, pancreatic epithelial expression of Aldh1 is initiated by former PNCs as they inactivate notch signaling, delaminate and start to express early markers of the endocrine lineage. As these cells continue to differentiate, they switch off Aldh1 expression and turn on markers of more mature endocrine cells, including pax6b and insulin. Blocking production of retinoic acid accelerates this process. We hypothesize that retinoic acid is produced following the activation of Aldh1 expression immediately following delamination from ductal epithelium and commitment to the endocrine lineage. Retinoic acid produced by these cells then acts to prevent additional cells from delaminating, and also impedes the further maturation of cells that have already These findings contribute to recent heightened awareness of Aldh1 expression and enzymatic activity as markers and possible regulators of pancreatic progenitor cells in adult and embryonic pancreas. In embryonic mouse pancreas, Aldh1 is expressed in "tip" progenitor cells, a zone known to harbor multi-lineage endocrine and exocrine progenitor cells [26,27]. In murine ES cell-derived pancreatic progenitor cells, activation of an inducible Ngn3 transgene upregulated expression of both NeuroD and the Aldh1 isoform Aldh1b1 [37], potentially echoing our findings regarding neurod and Aldh1 coexpression in zebrafish pancreatic progenitor cells. As in zebrafish [25], inhibition of Aldh1 enzymatic activity in developing mouse pancreas is associated with precocious endocrine differentiation [26]. In adult mouse pancreas, epithelial expression of Aldh1 is largely confined to centroacinar and terminal duct cells [27]. When isolated based upon high-level Aldh1 enzymatic activity, these adult cells display both endocrine and exocrine progenitor capacities in vitro, and are also uniquely able to contribute to the endocrine and exocrine lineages of the embryonic pancreas [27]. Thus the expression of Aldh1 isoforms appears to be a common feature of pancreatic progenitor cells in both mouse and zebrafish.
It is important to acknowledge that many aspects of our model remain speculative. With respect to our hypothesis that zebrafish Aldh1-expressing cells serve as direct precursors to the pax6 lineage, it has proven difficult to detect significant numbers of Aldh1-expressing cells prior to the onset of pax6:eGFP expression. This may reflect the fact that, under normal conditions, there is a very narrow temporal window of Aldh1 expression during endocrine differentiation; alternatively, it may reflect differential sensitivities in the assessment of gene and protein expression using antibodies vs. transgenic reporters. It also remains unclear whether all endocrine cell types emanate from progenitor cells expressing both Aldh1 and neuroD. This point is especially relevant given our finding that many neuroD-positive cells in larval zebrafish pancreas do not co-express Aldh1, as well as recent observations in the mouse suggesting that early pancreatic progenitor cells may already be compartmentalized into unipotent subpopulations [38]. In the mouse, adult pancreatic Aldh1-expressing cells are capable of generating both insulin-expressing cells and glucagonexpressing cells [27]; whether Aldh1-expressing cells in larval zebrafish pancreas can similarly serve as precursors for multiple endocrine cell types remains unknown.
In addition, while inhibition of Aldh1 enzymatic activity is associated with precocious endocrine differentiation in both mouse and zebrafish, the specific products of Aldh1 enzymatic activity responsible for mediating this effect remain unknown. While Aldh1 enzymatic activity may be linked to retinoic acid synthesis, the reagents currently utilized to isolate cells based on Aldh1 expression do not specifically test for retinal dehydrogenase activity. In addition, embryonic and adult mouse pancreatic cells marked by Aldh1 immunoreactivity have been shown to express some Aldh1 isoforms that carry  retinal dehydrogenase activity (e.g. Aldh1a1), and some isoforms (e.g. Aldh1a7, Aldh1B1) lacking this activity [26,27]. While we have shown that the zebrafish Aldh1 epitopes recognized by Aldh1 immunostaining are indeed associated with Aldh1 enzymatic activity ( Figure S2), the precise zebrafish Aldh1 isoforms recognized by currently available Aldh1 antibodies are unknown, further underscoring the difficulties in determining the relevance of retinoic acid production by these cells.
Together with prior reports [25], our findings suggest that Aldh1 enzymatic activity may serve as a critical gatekeeper of endocrine lineage commitment and maturation, and that pharmacologic inhibition of this activity may play a role in facilitating β-cell regeneration. :eGFP (C and D) fish. Wild-type cells were labeled with Aldefluor and then FACS sorted, while cells from Tp1:eGFP fish were sorted for eGFP. Sorted populations were then subjected to cytospin and immunofluorescent labeling for Aldh1. In the case of cells from Tp1:eGFP fish, labeling was also performed for eGFP. Note that Aldh1 protein is detected in Aldefluor pos but not Tp1:eGFP pos cells, while eGFP is detected in Tp1:eGFP pos but not Aldefluor pos cells. Arrows in (C) indicate low-abundance Aldh1-sorted cells present in Tp1:eGFP neg but not Tp1:eGFP pos cell fraction. E, RT-PCR analysis of gene expression in FACS sorting populations. Positive and negative populations of cells sorted for either Aldh1 activity or Tp1:eGFP expression displayed differential expression of aldh1a2, prom1, and sca1, with contrasting patterns of enrichment in Aldefluor pos vs. Aldefluor neg and Tp1:eGFP pos vs. Tp1:eGFP neg cell fractions, further documenting the non-overlapping nature of Aldh1-