Retinoic Acid Promotes the Generation of Pancreatic Endocrine Progenitor Cells and Their Further Differentiation into β-Cells

The identification of secreted factors that can selectively stimulate the generation of insulin producing β-cells from stem and/or progenitor cells represent a significant step in the development of stem cell-based β-cell replacement therapy. By elucidating the molecular mechanisms that regulate the generation of β-cells during normal pancreatic development such putative factors may be identified. In the mouse, β-cells increase markedly in numbers from embryonic day (e) 14.5 and onwards, but the extra-cellular signal(s) that promotes the selective generation of β-cells at these stages remains to be identified. Here we show that the retinoic acid (RA) synthesizing enzyme Raldh1 is expressed in developing mouse and human pancreas at stages when β-cells are generated. We also provide evidence that RA induces the generation of Ngn3+ endocrine progenitor cells and stimulates their further differentiation into β-cells by activating a program of cell differentiation that recapitulates the normal temporal program of β-cell differentiation.


Introduction
The identification of secreted signal(s) that promotes the selective generation of b-cells from stem and/or progenitor cells represent a significant step in the development of stem cell-based b-cell replacement therapy for type 1 diabetes. Such signals may be identified by studying how b-cells are normally generated during pancreas development. Transcription factors with important roles during pancreatic development and endocrine cell differentiation have been identified, and Notch signaling has been shown control pancreatic cell differentiation by regulating the expression of the proendocrine gene Neurogenin 3 (Ngn3) [1][2][3][4][5][6]. In contrast, no secreted signal has been identified that selectively promotes the generation of b-cells during pancreas development.
Ngn3 + endocrine progenitor cells give rise to different endocrine cell types at different stages of mouse embryonic development [1,2,5]. Glucagon positive (Glu + ) cells appear by embryonic day (e) ,9.5 whereas insulin positive (Ins + ) cells appear predominantly from ,e14.5 and onwards. Between e14.5 and neonatal stages Ins + cell numbers increase 4-10 fold more than other pancreatic endocrine cell types [7,8]. Why Ngn3 + endocrine progenitor cells preferentially start to differentiate into Ins + cells at ,e14.5 remains unknown. One possibility is that a cell intrinsic program in Ngn3 + progenitor cells regulates the temporal order of generation of differentiated endocrine cell types. Alternatively, extra-cellular signals may induce distinct classes of Ngn3 + endocrine progenitor cells at different developmental stages.
The vertebrate ventral spinal cord is a well studied system with respect to the mechanisms by which different classes of progenitor cells and differentiated progeny, inter-neurons and motor neurons, are generated [9][10][11]. Pancreatic endocrine cells and neurons share many common features and in particular b-cells and spinal cord motor neurons show striking similarities in patterns of gene expression of several basic-helix-loop-helix (Bhlh) and homeodomain transcription factors [11][12][13]. The secreted signal retinoic acid (RA) has been shown to control multiple, sequential steps of motor neuron differentiation in the ventral spinal cord, including initial patterning of progenitors cells and specification of motor neuron subtype identity [11]. The similarities between b-cells and motor neurons suggest that RA may also control specification and/or differentiation of insulin-producing b-cells.
Here we show that RA receptor signalling is required in early pancreatic progenitor cells for pancreatic development. We also show that the RA-synthesizing enzyme Raldh1 is expressed at stages when b-cells are normally generated in embryonic mouse and human pancreas. Using an explant assay of pancreatic endocrine cell differentiation, we provide evidence that RA promotes both the generation of Ngn3 + endocrine progenitor cells and their further differentiation into b-cells. The identification of a secreted signal that induces the generation insulin producing b-cells will aid in the establishment of a stem cell-based therapy for type 1 diabetes.

Raldh1 expression coincides with b-cell differentiation in both mouse and human fetal pancreas
To address a potential role for RA in generation of endocrine progenitor and/or b-cells we first investigated expression of Retinaldehyde dehydrogenase (Raldh) genes in the developing mouse pancreas by in situ hybridization. Neither Raldh 3 nor 4 was expressed in the developing pancreatic epithelium or mesenchyme at any stage examined (data not shown). Although Raldh2 was expressed in e8.0 somitic mesoderm and e10.5 duodenal mesenchyme (Fig. S1a), it was not expressed within the developing pancreas. Raldh1 was, however, expressed in the pancreatic epithelium from ,e13.5 and onwards (Fig. 1a), coinciding with the period of massive b-cell differentiation. Expression of Raldh1 in embryonic mouse pancreas was restricted to the branching ductal and acinar epithelium flanking the differentiating endocrine cells (Fig. S1b). qRT-PCR revealed that RALDH1 expression overlapped temporally with that of INSULIN also in the developing human pancreas, whereas expression of RALDH 2, 3, and 12 (the homolog of mouse Raldh4) was distinctly lower and/or barely detectable ( Fig. 1b and data not shown). In situ hybridization analyses confirmed expression of RALDH1 in fetal human pancreas (Fig. 1c). Thus, the expression of the RA-synthesizing enzyme Raldh1, coincides with the period of b-cell differentiation in both mouse and human pancreatic development.

RA-signalling is required for pancreatic development
To address the role for RA-signalling in the developing pancreas we next generated transgenic (tg) mice expressing a dominantnegative (dn) form of the retinoic acid receptor (RAR) a, denoted RAR403 [14,15], under the control of the Ipf1/Pdx1 promoter, which is active both in pancreatic progenitor cells and in b-cells [16]. The resulting Ipf1/RAR403 tg mice showed complete pancreatic agenesis, lacking both dorsal and ventral pancreas, and died at the neonatal stage (Fig. 1d). These data are consistent with Raldh2 expression in lateral plate and somitic mesoderm of e8.0 mouse embryos (Fig. S1a) and provides support for an intrinsic role for RA-signalling in specified Ipf1/Pdx1 + pancreatic progenitor cells. The perturbation of both dorsal and ventral pancreatic development in Ipf1/RAR403 tg mice contrasts with the Raldh2 2/2 mutant mice in which dorsal but not ventral pancreatic development was affected [17,18]. Taken together, our data provide evidence for a requirement for RA-signalling in early pancreatic progenitor cells of both the dorsal and ventral pancreatic anlage and suggest the existence of a Raldh2 independent source of RA that signals to the ventral pancreatic anlage.

Retinoic acid promotes the generation b-cells
The impaired pancreatic development in Ipf1/RAR403 tg mice precluded any analysis of a role for RA-signalling in b-cell differentiation, and attempts to block RA signalling in b-cells by expressing the dnRARa in b-cells of tg mice using the rat insulin 1 promoter 19 failed to generate tg offspring, which may be due to non-pancreatic expression of the transgene at embryonic stages [20]. The elucidation of the roles of RA in motor neuron generation benefited largely from the use of an ex vivo explant system [9,21]. Thus, to assess a potential role for RA signalling in b-cell generation we established an ex vivo explant system where e10.5 dorsal pancreatic buds were cultured in serum-free, defined media for six days in presence or absence of RA. Using these culture conditions the explants survive but do not grow.
In explants cultured for six days in absence of RA, only a few glucagon positive (Glu + ) and insulin positive (Ins + ) cells were observed and the majority of cells were of non-endocrine identity (Fig. 2a,b and data not shown). In explants cultured in presence of 25nM RA a .3-fold increase in the relative number and ,2,5 fold increase in absolute numbers of Ins + cells (67633 Ins + cells in control explants and 155623 Ins + cells in RA treated explants, presented as mean6S.E.M) was observed, whereas there was no significant increase in the number of Glu + cells (Fig. 2a,b and data not shown). Higher concentrations of RA (50 and 100 nM), failed to further stimulate the generation of Ins + cells, and even showed an inhibitory effect on endocrine cell differentiation (data not shown).
The expression of Raldh1 in the developing pancreatic epithelium (Fig. 1a-c) suggests that the epithelial cells themselves synthesize retinoids that participate in the induction of Ins + cells. Previous studies have, however, shown that explants grown in medium are deprived of both RA and the metabolic substrate required for synthesis of RA by RA-synthesizing enzymes [21]. Thus, to test for RA-synthesizing activity in the pancreatic explants the RA precursor retinol was added to the medium. In explants cultured in presence of 5 mM retinol or 100 nM retinol a ,10-fold increase in Ins + cells was observed as compared to that of control explants ( Fig. 2d and data not shown). Taken together, these results support the existence of an endogenous RA synthesizing system in the developing pancreas that promotes the generation of ins + cells from pancreatic progenitor cells.

Retinoic acid induces the expression of transcription factors that control endocrine cell differentiation
In the ventral spinal cord RA induces a transcriptional program that culminates in the generation of a distinct set of motor neurons [9][10][11]. To begin to understand how RA promotes the generation of Ins + cells in our explant system, we analyzed expression of transcription factors that regulate endocrine cell differentiation. Pancreatic endocrine cell differentiation, including that of b-cells, is preceded by and dependent on transient expression of the proendocrine gene Ngn3 in progenitor cells [1][2][3][4][5][6]. Moreover, during motor neuron differentiation RA has been shown to regulate the expression of the pro-neural gene Ngn2 [9][10][11]. Thus, we examined Ngn3 expression in explants harvested after 2 or 4 days of in vitro culture in presence or absence of RA. After 2 days of culture, explants exposed to RA showed a ,4-fold increase in both relative (Fig. 3a,b) and absolute number of Ngn3 + cells (2468 Ngn3 + cells in control explants and 94624 Ngn3 + cells in RA treated explants, presented as mean6S.E.M) compared to controls. Consistent with the transient role for Ngn3 in pancreatic endocrine progenitor cell differentiation [1][2][3][4][5][6], the number of Ngn3 + cell was not significantly greater than that in control explants after an additional 2 days of RA exposure (Fig. 3b).
Ngn2 expression and thus motor neuron differentiation critically depends on the transient expression of Olig2, a Bhlh transcriptional repressor, induced by RA [9][10][11]. To elucidate whether expression of Olig2 or related Bhlh factors [22,23] correlated with Raldh1 expression in the pancreas we investigated expression of Olig1-3, and Bhlhb2-5 in e13.5 and e15.5 pancreas by qRT-PCR. Only Bhlhb2 was robustly expressed in the e13.5 and e15.5 pancreas (Fig. S2 and data not shown) and its expression was also increased by ,50% in RA-treated explants at day 2 of culture ( Fig. 3c). NeuroD is a Ngn3 target gene that is expressed in differentiated endocrine but not pro-endocrine cells. In agreement with the normal temporal order of expression of Ngn3 and NeuroD, no increase in NeuroD expression was observed at day 2 of RA exposure whereas an almost 3-fold increase in NeuroD expression was evident at day 4 of RA exposure (Fig. 3c). Thus, RA triggers the expected temporal sequence of events of transient Ngn3 expression followed by increased NeuroD expression.
During pancreatic and motor neuron development, activation of Notch results in increased expression of Hes genes that repress expression of the proendocrine gene Ngn3 and the pro-neural gene Ngn2, respectively, thus maintaining cells as undifferentiated progenitor cells [12,13]. Bhlhb2 has been shown to inhibit Notch mediated activation of the Hes1 promoter by interacting with the intracellular form of Notch [24]. In agreement with the increased expression of Bhlhb2 in RA treated explants at day2, Hes1 expression was reduced by RA to only ,50% of that of untreated controls at day 4 (Fig. 3c). Thus, sequential stimulation of Bhlhb2, Ngn3, and NeuroD expression, paralleled by reduced Hes1 expression, precedes RA-stimulated differentiation of b-cells.

Sequential roles for RA in the generation of b-cells
The induction of Ngn3 + endocrine progenitor cells and the preferential generation of b-cells by RA, suggests that RA may promote the generation of a set of Ngn3 + cells that consequently are destined to generate b-cells. Alternatively, similar to its role during motor neuron differentiation [9][10][11], RA may have distinct sequential functions during the generation of b-cells, i.e. inducing both Ngn3 expression and the further differentiation of Ngn3 + cells into b-cells. To distinguish between these two possibilities, e10.5 pancreatic buds were cultured in presence of RA for only the first 2 or 4 days of the total 6 day culture period. No significant increase in the number of Ins + cells was observed in explants exposed to RA for only the first 2 of the total 6 day culture period (Fig. 4a). In explants exposed to RA for the first 4 of the total 6 days, there was a significant increase in the number of Ins + cells (Fig. 4a), though smaller than that observed in explants exposed to RA the entire 6 day culture period (Fig. 2b and data not shown). Thus, RA needs to be added throughout the culture period to ensure Ins + cell differentiation.
Like the sequential expression of Bhlh genes (Fig. 3c), homeobox genes linked to b-cell differentiation were expressed in a sequential manner during RA-stimulated Ins + cell differentiation. The expression of Pax4, a critical factor in specification of b-cell identity [25], was not significantly increased after 2 days of RA exposure but a ,2.5 fold increase in Pax4 expression was observed after 4 days of RA exposure (Fig. 4a). The Ins + cells induced by RA expressed the additional b-cell markers Ipf1/Pdx1, Nkx6.1 and Hb9, as well as C-peptide and Glut2, after 6 days of RA exposure (Fig. 4b and data not shown). RA therefore mediates sequential induction of the Bhlh transcription factors Ngn3, Bhlhb2, and NeuroD that promote the generation of endocrine cells, followed by induction of the homeobox gene, Pax4, that determines b-cell identity, and finally the generation of insulin producing b-cells that express the homeodomain proteins Ipf1/ Pdx1, Nkx6.1 and Hb9 (Fig. 4c). Taken together these results provide evidence that, in analogy with the generation of motor neurons in the spinal cord, RA acts at multiple levels in the generation of b-cells in the pancreas.

Discussion
In the developing mouse pancreas, b-cells increase in large numbers from ,e14.5 onwards, suggesting that intrinsic and/or extrinsic signals favour b-cell differentiation at these stages. Transcription factors that control specification of pancreatic endocrine progenitor cells and their further differentiation into b-cells have been identified [13]. In contrast, secreted signals that may promote the large increase in generation of b-cells remain to be identified. In this study, we show that the RA synthesizing enzyme Raldh1 is expressed in the developing mouse and human pancreas at stages when b-cells are predominantly generated. Moreover, using an explant assay of pancreatic endocrine cell differentiation, we provide evidence that RA activates a program of cell differentiation that recapitulates the normal temporal program of b-cell differentiation.
In Raldh2 mutant mice, Ipf1/Pdx1 expression is lost in the dorsal but not ventral pancreatic bud and mutant mice present with dorsal but not ventral pancreatic agenesis [17,18]. However, the dorsal but not ventral pancreatic mesenchyme fails to form in Raldh2 mutants [17], suggesting that this phenotype is indirect. Consistent with this possibility early pancreatic markers such as Hb9 and Isl1 are expressed in e9.5 Raldh2 mutant dorsal pancreatic endoderm [17]. Moreover, maternal RA administration restored the formation of dorsal pancreatic mesenchyme, Ipf1/Pdx1 expression and thus pancreatic development in Raldh2 mutant embryos [17]. Loss of dorsal pancreatic mesenchyme and Ipf1/Pdx1 expression are also observed in Isl1 [26] and N-cad [27] mutant embryos, and dorsal pancreatic development could be rescued in both mutants by co-culturing the pancreatic endoderm with wild-type mesenchyme [26,27]. In Xenopus and zebrafish, induction of pancreatic development is inhibited following treatment of early, pre-pancreatic embryos, with the RAantagonist BMS493, and embryos treated with exogenous RA prior to the initiation of the pancreatic development show ectopic expression of pancreatic markers such as insulin [28][29][30][31]. Collectively, these data provide evidence that RA is required during early (dorsal) pancreas development, at least in part by promoting the development of dorsal pancreatic mesenchyme. In addition, the apancreatic phenotype of Ipf1/RAR403 tg mice suggests a direct role for RA signaling in both the early dorsal and ventral pancreatic epithelium after the onset of Ipf1/Pdx1 expression. The lack of pancreas in these mice unfortunately precluded the establishment of a transgenic line that would allow detailed analysis of the pancreatic developmental defect.
Differentiation of all pancreatic endocrine cells requires the transient activity of the proendocrine gene Ngn3 in progenitor cells [1][2][3][4][5][6]. Different classes of endocrine cells are generated at different stages of embryonic development of the pancreas and forced expression of Ngn3 in early pancreatic progenitor cells generates predominantly glucagon-producing cells whereas later induction generates b-cells and other endocrine cell types [1,2,5]. Our data show that the temporal expression of the RA-synthesizing enzyme Raldh1 in the developing pancreas coincides with the large increase in b-cells from ,e14.5 and onwards. Using an explant assay, we also show that RA and RA precursors selectively stimulate b-cell differentiation, resulting in both a relative and absolute increase in b-cell numbers. Raldh1 mutant mice have been generated and are viable, suggesting that they are not devoid of b-cells [32]. Whether this reflects increased expression of any of the other Raldhs in the pancreas of Raldh1 mutant mice at the time of b-cell differentiation will have to await future analyses of Raldh2-4 expression in Raldh1 2/2 mice. An alternative, RALDH-independent mode of RA synthesis mediated by Cyp1b1 was recently reported [33] and qRT-PCR analyses detected low level expression of Cyp1b1 in mouse e13 and e15 pancreas and in 8-19 weeks human fetal pancreas (data not shown).
By using medium containing high levels of serum that favors growth of explants, previous studies suggest that RA antagonizes growth and exocrine cell differentiation leading to a relative but not absolute increase of b-cells [34,35]. In the developing pancreas, Fibroblast growth factor (FGF) signals provided by the mesenchyme stimulate growth and differentiation of the exocrine pancreas [30,36,37]. RA and FGFs are known to exert opponent activities [38] and under culture conditions promoting growth, RA is likely to primarily antagonize the stimulatory effect of FGFs on growth and differentiation of the exocrine pancreas. In support of this, FGF10 was shown to antagonize the negative effects of RA on pancreatic growth, morphogenesis, and exocrine differentiation [34]. b-cell differentiation in the mouse embryo becomes prominent when the mesenchyme/epithelium ratio, and thus growth decreases [7,8]. This phase, sometimes referred to as the secondary transition [7], coincides temporally with Raldh1 expression in the pancreas. Our culture conditions were designed to mimic this phase of b-cell differentiation by using defined, minimal media without serum, which allows survival but not growth of the pancreatic explants. Under these conditions, RA or RA precursors selectively stimulate differentiation of pancreatic progenitors into b-cells.
Together, our results provide strong evidence that RA promotes both the generation of Ngn3 + endocrine progenitor cells and their further differentiation into b-cells. The increase in Ngn3 + cells is evident already after 2 days of RA exposure although Hes1 expression is significantly reduced first after 4 days of RA exposure, leaving open the possibility that the increase in Ngn3 + cells is independent of Hes1 expression. Whether the early increase in Bhlhb2 expression already at day 2 of RA exposure, directly or indirectly, promotes a parallel increase in Ngn3 expression, similar to the positive effects of Olig2 on Ngn2 expression [9.11], will require further analyses. Nevertheless, the early increase in the expression of Bhlhb2, which has Notch antagonizing activity [24], is in agreement with the subsequent reduced expression of Hes1 observed at day 4 of RA exposure. The identification of RA as a secreted signal that promotes the selective generation of b-cells, and increased availability of human ES cell lines that respond to experimental protocols that allow efficient and reproducible generation of pancreatic endoderm, will help to develop stem cell-based b-cell replacement therapy.

Human fetal pancreas isolation
The fetal pancreatic studies were approved by the Human Subject Research Office, Miami, Florida, in compliance with US legislation and the guidelines of our institution. Fetal pancreatic tissue was obtained following written maternal consent. Elective termination of pregnancy was performed by aspiration between 8 and 19 weeks of development. Warm ischemia lasted less than 30 minutes.

Animals and generation of transgenic mice
Animal studies were approved by the Institutional Animal Care and Use Committee of Umeå University and conducted in accordance with the Guidelines for the care and use of Laboratory Animals. Human RARa403 was subcloned behind the Ipf1/Pdx1 promoter and the RIP promoter respectively. Transgenic mice were generated as described 16 .

Immunohistochemistry and in situ hybridization
Immunohistochemical analyses were carried out as described elsewhere 16 . Primary antibodies used were: rabbit anti-Ipf1 39 (Ohlsson et al., 1993), rabbit anti-Ngn3 40 , guinea pig anti-Insulin (Linco), rabbit anti C-peptide (Linco), rabbit anti-Glucagon (Linco), rabbit anti-Nkx6.1 (raised against a GST-Nkx6.1 fusion protein by AgriSera AB, Vä nnä s, Sweden) and rabbit anti-Hb9 (kindly provided by T. M. Jessell). Secondary antibodies were: Cy3 anti-rabbit (Jackson), Alexa Fluor 488 anti-guinea pig and Alexa Fluor 594 anti-rabbit (Molecular Probes). In situ hybridizations were performed as previously described 16 . Images were acquired using a Zeiss Axioplan imaging microscope or a Nikon confocal microscope C1 fitted with an Ar, He/Ne and a blue diode laser. Digitalized confocal images were assigned red and green pseudocolors for Cy3 and Alexa Fluor 488 respectively.

In vitro cultures
Isolated e10.5 dorsal pancreatic buds were cultured on filters (Millipore) in 24-well plates (Costar) with 400 ml of medium consisting of DMEM-glutamax1 (Gibco), 16N 2 -supplement (Gibco) and pen/strep (Gibco), and different concentrations of all trans retinoic acid or retinol (Sigma). Explants were cultured for up to six days in 37uC and 5% CO 2 and medium was changed every other day. After culture, explants were washed twice in PBS, fixated in 4% PFA at 4uC for 20 minutes followed by two more washes in PBS. Explants were frozen in drops of Tissue Tek OCT applied on glass slides (Menzel) and stored at 280uC until sectioned (8 mm sections) and analyzed by immunohistochemistry.
Quantification of mRNA expression levels cDNA was prepared from explants, fetal mouse and human pancreas and qRT-PCR analysis of the cDNA was performed essentially as described 41 . Expression of TATA box binding protein (TBP) and human 18S ribosomal RNA (18S) was used to normalize expression levels. Primer sequences are found in Supplementary Table S1.

Statistical analysis
All data is presented as mean value6standard errors (S.E.M.). All statistical analyses were performed using the two-tailed, two sample Students t-test and P values ,0.05 were considered as significant.