Mesenchyme-derived IGF2 is a major paracrine regulator of pancreatic growth and function

Summary

The genetic mechanisms that determine the size of the adult pancreas are poorly understood. Imprinted genes, which are expressed in a parent-of-origin-specific manner, are known to have important roles in development, growth and metabolism. However, our knowledge regarding their roles in the control of pancreatic growth and function remains limited. Here we show that many imprinted genes are highly expressed in pancreatic mesenchyme-derived cells and explore the role of the paternally-expressed insulin-like growth factor 2 (Igf2) gene in mesenchymal and epithelial pancreatic lineages using a newly developed conditional Igf2 mouse model. Mesenchyme-specific Igf2 deletion results in acinar and beta-cell hypoplasia, postnatal whole-body growth restriction and maternal glucose intolerance during pregnancy, suggesting that the mesenchyme is a developmental reservoir of IGF2 used for paracrine signalling. The unique actions of mesenchymal IGF2 are demonstrated by the absence of any discernible growth or functional phenotypes upon Igf2 deletion in the developing pancreatic epithelium. Additionally, increased IGF2 levels specifically in the mesenchyme, through conditional Igf2 loss-of-imprinting or Igf2r deletion, leads to pancreatic acinar overgrowth. Furthermore, ex-vivo exposure of primary acinar cells to exogenous IGF2 activates AKT, a key signalling node, and increases their number and amylase production. Based on these findings, we propose that mesenchymal Igf2, and perhaps other imprinted genes, are key developmental regulators of adult pancreas size and function.

Keywords:

Gene expression – Alleles – Body weight – Polymerase chain reaction – Mouse models – Pancreas – Genetically modified animals

Introduction

The mammalian pancreas plays a central role in energy homeostasis, which is achieved by functionally and morphologically distinct exocrine and endocrine components. Optimal pancreatic function requires a match between pancreas size and the physiological demands of the host organism. However, the genetic determinants of organ size, and the mechanisms that achieve adequate relative organ size, are poorly understood. The size of the pancreas is thought to be fixed early in development, limited by the size of the progenitor cell pool that is established in the developing pancreatic bud [1]. In addition to autonomous cues in the epithelium, pancreatic development is also dictated by non-autonomous signals from the mesenchyme. Mesenchymal cells overlie the developing pancreatic bud and provide critical signals for the expansion of both precursors and differentiated endocrine and exocrine cells [2]. These cells are present throughout pancreas organogenesis, but the relative proportion of mesenchyme to epithelium changes, with a dramatic reduction over time, as the mesenchyme differentiates into more specialized cell types such as pericytes [3] and epithelial cells expand. Recent elegant genetic manipulation approaches in the mouse have shown that mesenchymal cells regulate pancreatic growth and branching at both early and late in utero developmental stages [2,4]. However, many of the mesenchymal signals that control these processes remain un-identified. Moreover, the factors required for cell fate decisions that specify individual pancreas cell type subsets are fairly well established but less is known about signalling pathways involved in proliferation and survival of cell types.

Insulin-like growth factors (IGF1 and IGF2) are small mitogenic polypeptides (~7KDa) with structural homology to pro-insulin. IGF2 is also a pro-survival factor, protecting against apoptosis [5–7]. Although IGF expression is ubiquitous in many cell types, they are most abundant in the cells and tissues of mesodermal origin, and form important components of stem cell niches [8–13]. Igf2 transcription is regulated by genomic imprinting [14], an epigenetic process that causes a subset of genes in the genome to be transcribed according to parental origin. Imprinted genes regulate key aspects of mammalian physiology, from growth to energy homeostasis [15]. In most mouse and human tissues, Igf2 is only transcribed from the paternally inherited allele, with the maternal allele being repressed by a methylation sensitive CTCF-dependent boundary that restricts the access of downstream enhancers to the Igf2 gene promoters [16]. In humans, reduced Igf2 expression contributes to the intra-uterine growth restriction in patients with Silver-Russell syndrome [17]. Conversely, bialellic Igf2 expression caused by loss of Igf2 imprinting is observed in Beckwith-Wiedemann patients [18], a syndrome characterized by somatic overgrowth, neonatal hypoglycaemia with variable penetrance and increased predisposition to tumours. In mice, increased supply of embryonic Igf2 above biallelic expression can result in disproportionate overgrowth associated with heart enlargement, oedema and fetal death [19]. Igf2 null mice are viable, depending on the genetic background, but have almost half the adult body weight [19].

Key questions about how IGF2-mediated growth effects are developmentally programmed and how IGF2 contributes to organ growth control in-vivo remain unanswered. The normal timing, location and duration of IGF2 supply at the organ level are likely to be crucially important. The pancreas serves as an interesting model for the study of these processes, as it requires tight spatial and temporal regulation of proliferation, differentiation and morphogenesis. Very little is known about the main sites of Igf2 expression and imprinting during pancreas development. Moreover, most mouse transgenic studies conducted so far have been directed to the beta-cells, in which Igf2 is overexpressed [20,21] or knocked-out [22], or focused on mice that overexpress Igf2 [23,24] or lack Igf1 and Igf2 constitutively [25]. Here we describe the generation of a conditional knock-out allele for Igf2 and perform analyses of cell-type specific deletions that target the developing pancreatic mesenchyme (Nkx3.2-Cre), or epithelium (Ptf1a-Cre), with the overall aim of defining the paracrine and autocrine roles of IGF2 in pancreatic growth.

Results

The mesenchyme-derived cells are the main source of Igf2 in the developing perinatal and postnatal pancreas

We first sought to establish which major cell types within the developing mouse pancreas express Igf2 transcripts and to determine their relative levels. To achieve this aim, we used a recombinase inducible YFP reporter under the control of the Rosa26 locus (Rosa26YFP-stop^fl/fl) [26] and crossed this mouse model with transgenic strains that express Cre recombinase in the main cell types of the developing pancreas, i.e. Nkx3.2-Cre: mesenchyme [27]; Ptf1a-Cre: pancreatic epithelium [28]; RIP-Cre: beta-cells [29]. We confirmed the activity of each Cre line by performing immunofluorescence staining for YFP in combination with cell-type specific markers in histological sections (Fig 1A). We then performed an expression timeline analysis based on fluorescence activated cell sorting (FACS) of YFP-positive cells isolated from pancreas at various developmental stages, ranging from embryonic day 16 (E16) to adulthood, followed by qRT-PCR measurements of Igf2 mRNA levels (Fig 1B, S1A Fig). This analysis revealed that mesenchyme-derived cells express the highest levels of Igf2 mRNA when compared to beta-cells or non-mesenchyme cells (which is comprised of acinar, endocrine, ductal and endothelial cells) (Fig 1B). At E16, mesenchyme-derived cells express 380 fold more Igf2 than beta-cells and 1.5 fold more than non-mesenchyme cells. At P14, the difference between mesenchyme and non-mesenchyme is the greatest, with mesenchyme-derived cells expressing 700 fold more Igf2 than non-mesenchyme cells. The levels of Igf2 in mesenchyme-derived cells remain high in the neonatal period until the weaning period (at P21 levels are 3.2 fold reduced compared to E16, followed by a steep decline in adulthood) (Fig 1B). Interestingly, both non-mesenchyme cells and the beta-cells decrease their Igf2 expression levels after E16. To verify that the mesenchyme-derived cells express high levels of Igf2 mRNA, we next performed in situ hybridization analysis at E19, P2 and P5. Our results confirm that Igf2 transcripts are localized in mesenchyme-derived cells surrounding the vessels, pancreatic acini and ducts, with low to undetectable expression in endocrine and acinar cells (Fig 1C).

The results presented above suggest that Igf2 is a mesenchymal-derived cell marker in the neonatal and post-natal period, with mostly restricted expression to these cell types compared to non-mesenchyme. To confirm this observation, we next performed RNA-seq analyses in mesenchyme-derived cells and non-mesenchymal cells isolated by FACS from pancreases of Rosa26YFP-stop^+/fl; Nkx3.2^Cre/+ mice at P2. We found over four thousand differentially expressed genes (DEGs, fold change >1.5; FDR adjusted p value <0.05) between the mesenchyme-derived cells and non-mesenchyme cells in wild-type Igf2^+/+ pancreases (S2A Fig, S1 Data). A number of these DEGs were validated by qRT-PCR in independent biological replicates (S2B Fig). Using DAVID functional annotation (see Materials and Methods) we identified over 20 significantly enriched GO terms (FDR adjusted p value <0.05), which included biological processes known to be involved in pancreatic mesenchyme function (e.g. Wnt signalling), early pancreas development (e.g. retinoic acid and SMAD signalling) and epithelial-mesenchymal interactions (integrin-mediated signalling, cell adhesion, collagen catabolism) (Fig 2A). This analysis also highlighted unexpected pathways, such as semaphorin-plexin and insulin-like growth factor receptor signalling (Fig 2A). Notably, several gene members of the IGF signalling family are highly enriched in mesenchyme-derived cells versus non-mesenchyme (Fig 2B). Interestingly, imprinted genes, with known roles in growth control, make up almost one third of the top 30 genes most highly expressed genes in the pancreatic mesenchyme-derived cells, with Igf2 being the highest (Fig 2C) and one of the most enriched compared to non-mesenchyme cells (53 fold) (S2A Fig, S1 Data). DAVID functional annotation identified 17 GO terms enriched in non-mesenchyme cells including expected pathways, such as digestion and endocrine pancreas development, as well as novel pathways, such as peptidyl-tyrosine phosphorylation (S2C Fig, S1 Data). Contrary to mesenchyme-derived cells, no imprinted genes are found amongst the top 30 most highly expressed genes in non-mesenchyme cells (S2D Fig).

The observation of high levels of expression in the mesenchyme-derived cells isolated at P2 for several imprinted genes, including Igf2, prompted us next to extend our analysis to earlier developmental time points. We analysed the recently published transcriptomes of E13.5, E15.5 and E17.5 pancreatic mesenchyme-derived cells [3] isolated by FACS in Rosa26YFP-stop^+/fl; Nkx3.2^Cre/+ genetic crosses, as also used in our study. Similar to our observations at P2, the eight imprinted genes exhibit very high levels of expression at all three embryonic stages (Fig 2D) and Igf2 ranks among the top 22 most highly expressed genes. Additionally, all gene members of the IGF signalling family identified in our analysis at P2 were expressed in E13.5, E15.5 and E17.5 pancreatic mesenchyme-derived cells, with ten of these genes reaching levels above 50 FPKM at all three developmental time points (Fig 2E). Our analysis further shows that at all three embryonic stages, as well as at P2, the pancreatic mesenchyme-derived cells express not only pericyte markers, as previously published [3], but also markers of other cell types, including smooth muscle cells, fibroblasts and myofibroblasts (S3 Fig), and in agreement with the highly heterogeneous histological features that we observed at P2 (Fig 1A).

To accurately measure the contribution of the various cell types to the overall pancreatic Igf2 levels in the neonatal period, we analysed cell-type specific deletions of Igf2 using floxed mice (Igf2^+/fl) that we generated (S4 Fig, S5 Fig). Deletion of Igf2 from mesenchymal-derived cells reduced its total pancreatic levels by more than 80% (Fig 2F). Notably, mesenchyme-derived cells represent 10.3 ± 1.2% (n = 7) of the pancreatic tissue weight at this time point, as measured by stereology. Ptf1a-Cre mediated deletion of Igf2 in exocrine, endocrine and ductal cells (S1B Fig) showed no discernible changes in the whole pancreas Igf2 mRNA, which is consistent with in situ hybridisation data shown (Fig 1C). Deletion of Igf2 from endothelial cells (mediated by Tek-Cre [30]) reduced Igf2 mRNA with approximately 20–25% (Fig 2F). Altogether, these results demonstrate that the mesenchyme-derived cells are the main source of Igf2 in the pancreas during late gestation and early postnatal life.

Mesenchyme-specific Igf2 signalling controls the growth of exocrine and endocrine pancreas

To assess whether mesenchymal Igf2 plays a role in controlling the early growth of the pancreas, Igf2^+/fl; Nkx3.2^Cre/+ knockout mice were first analysed at the neonatal stage P2 (Fig 3A–3D). As expected, paternal transmission of the mutation resulted in 99% reduction of Igf2 levels at the mRNA level in FACS-isolated mesenchymal cells (Fig 3A), and absence of protein by immunofluorescence (Fig 3B). Mice with deletion of the paternal Igf2 allele have normal body weight, but significantly lighter pancreases (69% of normal) compared to littermate controls (Fig 3C, S6A Fig). Deletion of the maternal Igf2 allele has no phenotypic effects, which is consistent with findings that the maternal allele is transcriptionally inactive due to imprinting (S6B and S6C Fig). Importantly, Nkx3.2-Cre or Igf2 floxed carrier mice are indistinguishable from wild-type littermates (S6A and S6B Fig). We next measured the number of cell nuclei in DAPI-stained paraffin sections to establish if the loss of pancreas weight was due to hypoplasia. We found that the number of cell nuclei was significantly reduced in mutants (74% of normal) compared to controls (Fig 3D). This finding suggests that mutant pancreases are smaller due a reduction in cell numbers, which is in agreement with the well-established role for IGF2 in the control of cell proliferation and survival. Stereological analysis revealed that loss of Igf2 in pancreatic mesenchyme leads to decreases in acinar and beta-cell mass, in line with the overall pancreas weight deficit, while the mesenchymal mass remained unchanged (Fig 3D). Also consistent with the decrease in acinar mass, mutant pancreases contained lower amounts of lipase (11,367 ± 5674 in n = 6 mutants versus 18,343 ± 4238 in n = 7 controls, p = 0.035). Thus, our data reveals that the mesenchyme-derived IGF2 is likely to control acinar and beta-cell growth through paracrine signalling from the mesenchyme.

The microRNA miR-483 is located within the intron 4 of Igf2, which is deleted alongside the Igf2 coding exons 4–6 upon Cre-mediated recombination. To rule out a contributory role of this microRNA to the phenotype, we analysed pancreases of mice carrying a miR-483 deletion [31], and found that it does not alter pancreas weight at P2 (S7 Fig). Furthermore, mir-483 knockout mice are normal sized, viable and show no evidence of impaired glucose homeostasis defects or other discernible phenotypes up to at least three months of age (Sekita et al. manuscript in preparation). We therefore conclude that the growth phenotype in the Igf2^+/fl; Nkx3.2^Cre/+ knockout mice can be attributed solely to IGF2 signalling actions.

Next, we used a well-established mouse model of Igf2 loss-of-imprinting–the H19DMD conditional knockout [32]–to overexpress Igf2 specifically in the pancreatic mesenchyme (H19DMD^fl/+; Nkx3.2^+/Cre) and investigated the effects on acinar growth (Fig 4A–4D). As expected, deletion of the imprinting control region (DMD) from the maternal allele resulted in increased Igf2 and decreased H19 RNA levels (160% and 18% of controls, respectively–Fig 4A) and increased IGF2 protein (Fig 4B), and was associated with a 29% increase in pancreas weight, with similar body weights between genotypes at P2 (Fig 4C). The acinar cell mass was increased by 34%, in line with the overall pancreas weight increase (Fig 4D). Beta-cell mass remained unchanged, while the mesenchymal mass increased by 45% (Fig 4D). We then used an Igf2 gain-of-function model that does not lead to transcriptional changes in Igf2 mRNA but instead results in increased IGF2 protein (Fig 4E–4G). Accordingly, we deleted the IGF-type II receptor that is required for recycling IGF2 via lysosome degradation [33], specifically in the pancreatic mesenchyme (Igf2r^fl/+; Nkx3.2^+/Cre) (Fig 4E). Reduced Igf2r mRNA levels in the mesenchyme cells (25% of controls), was associated with a 20% increase in pancreas weight, with similar body weights between genotypes at P2 (Fig 4F). The acinar cell mass was also increased by 33% in this model (Fig 4G). Additionally, beta-cell mass was also increased, but the observed difference between groups (31%) did not reach statistical significance (p = 0.099), while mesenchymal mass remained unchanged (Fig 4G). Therefore, the two gain-of-function genetic models further demonstrate that IGF2 produced by the pancreatic mesenchyme-derived cells controls the growth of the acinar cells in a paracrine manner.

Mesenchyme-derived IGF2 exerts paracrine effects on pancreatic epithelium in vivo and acinar cells respond to exogenous IGF2 in vitro

To investigate the molecular signatures associated with the paracrine effects exerted by the mesenchyme-derived IGF2, we performed genome wide-transcriptional profiling by RNA-seq in both fractions of Igf2^+/fl; Nkx3.2^Cre/+ mutants (i.e. mesenchyme-derived cells with Igf2 deleted and the corresponding non-mesenchyme fraction), compared to mesenchyme and non-mesenchyme cells isolated from controls, at the neonatal P2 stage. We found that transcriptional changes are far more widespread in the non-mesenchyme fraction (498 DEGs with >1.5-fold change and FDR-adjusted p value <0.05) (Fig 5A, S2 Data) than in Igf2-deficient mesenchyme-derived cells (151 DEGs; Fig 5A, S3 Data). DAVID functional annotation identified 15 enriched GO terms in the non-mesenchyme fraction (Fig 5B, S2 Data) but only 2 in the Igf2-deficient mesenchyme fraction (namely, actin binding and inflammatory response, S3 Data). The top biological processes in the non-mesenchyme fraction enriched in genes up-regulated in mutants were related to digestion and inflammation/immune responses, and biological processes enriched in genes down-regulated in mutants were related to erythrocyte development and tRNA methylation (Fig 5B and 5C, S2 Data). These findings are in agreement with the stereological analyses showing that loss of Igf2 from the mesenchyme-derived cells impacts on the growth/function of the non-mesenchymal fraction through a paracrine effect.

To provide additional evidence that mesenchyme-derived IGF2 acts on acinar cells in a paracrine manner, we next isolated primary mesenchyme-derived and acinar cells from P2 pancreases (see Materials and Methods) (Fig 5D–5F). Mesenchymal-derived cells secreted higher levels of IGF2 into the conditioned media compared to acinar cells (Fig 5G), a finding that is consistent with higher levels of Igf2 mRNA expression in the mesenchyme-derived cells in-vitro (Fig 5F) and in-vivo (Fig 1, Fig 2). We then tested the effect of exogenous IGF2 treatment on acinar cells cultured ex-vivo. Isolated acinar cells treated with recombinant mouse IGF2 showed increasing levels of AKT phosphorylation (S473) in a concentration-dependent manner (Fig 5H). Additionally, treatment of acinar cells with 50 ng/ml IGF2 increased their cell numbers and led to a modest, but significant increase in amylase production (Fig 5I). Altogether, the ex-vivo data shows that mesenchyme-derived cells are capable of secreting IGF2 and that exogenous IGF2 induces intracellular signalling in acinar cells associated with increased enzymatic output.

Post-weaning growth and glucose homeostasis analyses in Igf2 pancreas-cell type specific knockouts

To investigate the impact of Igf2 loss-of-function from the developing mesenchyme and/or epithelium on post-weaning growth and glucose homeostasis, single and double Nkx3.2-Cre and Ptf1a-Cre Igf2 knockouts were analysed (Fig 6A). The pancreas weight deficit observed in single Igf2^+/fl; Nkx3.2^Cre/+ knockouts at P2 is maintained at weaning (S8A Fig) and is also observed at 9 weeks of age in single Igf2^+/fl; Nkx3.2^Cre/+ knockout males and females (Fig 6B). However, in contrast to the P2 time point, body weights are now reduced, by approximately 12% at P21 (S8B Fig), and by 5% and 10% for females and males, respectively, at 9 weeks of age (Fig 6C). After normalization for body weight, the mutant pancreases remain disproportionately smaller at both P21 and 9 weeks time points (S8A Fig, Fig 6B). Double Igf2^+/fl; Nkx3.2^Cre/+; Ptf1a^Cre/+ knockouts show similar pancreatic weights and body weight reductions to the single Igf2^+/fl; Nkx3.2^Cre/+ knockout (Fig 6B and 6C), thus suggesting that IGF2 produced by the endocrine and exocrine pancreas does not play major autocrine or paracrine growth roles that alter pancreas size. Consistent with this hypothesis, single Igf2^+/fl; Ptf1a^Cre/+ knockouts have normal pancreas sizes and body weights (Fig 6B and 6C).

Glucose homeostasis, assessed by oral glucose tolerance tests (OGTT), was unaltered in single Igf2^+/fl; Nkx3.2^Cre/+ and Igf2^+/fl; Ptf1a^Cre/+ knockouts, as well as in double Igf2^+/fl; Nkx3.2^Cre/+; Ptf1a^Cre/+ knockouts compared to Igf2^+/fl controls at 8 weeks of age in both sexes (Fig 6D and 6E). We next assessed glucose homeostasis during pregnancy, a naturally occurring metabolic stress state. We found that 8 week old Igf2^+/fl; Nkx3.2^Cre/+ knockout females at the E15 stage of pregnancy had similar levels of glycaemia and insulinemia compared to age-matched pregnant Igf2^+/fl littermate controls after fasting and prior to OGTT (time 0) (Fig 6F). However, during OGTT, pregnant Igf2^+/fl; Nkx3.2^Cre/+ knockout females became glucose intolerant compared to pregnant Igf2^+/fl littermate controls (Fig 6F). Beta-cell mass, measured at E15, was significantly smaller in pregnant Igf2^+/fl; Nkx3.2^Cre/+ knockout females compared to pregnant Igf2^+/fl littermate controls (Fig 6G). Thus, our results show that mesenchyme-specific Igf2 deletion results in postnatal whole-body growth restriction and maternal glucose intolerance during pregnancy.

Discussion

This study uncovers novel roles for IGF signalling in the developmental regulation of adult pancreas size, with consequences for energy homeostasis and post-natal whole body-growth. We first observed that gene members of the IGF signalling pathway, which include IGF2 and IGF1 ligands, are highly enriched in the neonatal mesenchyme-derived cells compared to the non-mesenchyme fraction. Moreover, IGF2 is the top expressed gene in the mesenchyme-derived cells, as revealed by RNA-seq profiling at postnatal day 2 (P2) and one of the top ~25 most highly expressed genes at E13.5, E15.5 and E17.5. Mesenchyme-derived cells express the highest levels of Igf2 mRNA compared to all other pancreatic cell types from E16 to 8 weeks postnatally. At P2 we estimate, based on conditional deletions of Igf2 from the diverse cell types, that ~75–85% of total Igf2 in the pancreas is expressed by the mesenchyme-derived cells, ~20–25% by the endothelium and less than 1% is derived from the epithelium. Therefore, we hypothesized that mesenchymal IGF2/IGF signalling may play previously unanticipated roles in pancreatic development and generated conditional specific mouse models mainly targeting IGF2 levels in the mesenchyme.

We found that mesenchymal IGF2 is principally required for normal growth of the exocrine pancreas. Deletion of Igf2 from the mesenchyme cells from E9.5 leads to hypoplasia of the postnatal pancreas (~69%N at P2, ~67%N at P21 and ~79–83%N at 9 weeks of age). The acinar mass is affected the most and is reduced, in line with the overall loss in pancreatic weight. This finding is in agreement with the reports of reduced exocrine mass in mice that lack both IGF ligands (i.e. mice lacking both Igf1 and Igf2 in all cells of the body), or their receptors (i.e. mice constitutively lacking both Insr and Igf1r) [25]. However, the contribution of the individual ligands or receptors to the loss of exocrine mass has remained unclear, mainly because Insr or Igf1r single total knockouts result in either normal or increased exocrine mass, respectively, and single Igf1 or Igf2 total knockouts were not studied for pancreatic phenotypes. Our work now demonstrates that IGF2 is a key factor promoting acinar growth during development. Crucially, we have ruled out a role for IGF2 expressed in the developing epithelium, which includes IGF2 expressed in acinar cells, as a determinant of exocrine pancreas growth. Accordingly, deletion of Igf2 using Ptf1a-Cre, results in normal pancreatic growth. We therefore conclude that the effect of IGF2 on pancreas growth is restricted primarily to signalling from the mesenchyme.

We next addressed the question of how mesenchymal IGF2 regulates pancreas size. We propose that this is mainly achieved through secretion of IGF2 from the mesenchyme and signalling to the neighbouring cells types in a paracrine manner (Fig 6H). This hypothesis is based on the following findings: in primary cell cultures we show that mesenchyme-derived cells secrete IGF2 into the culture media, and that exogenous IGF2 stimulates activation of a key signalling node, AKT, downstream of IGF1R and INSR in acinar cells, increasing their number and enzymatic production. In vivo, increased IGF2 levels specifically in the mesenchyme led to significant overgrowth of the exocrine pancreas, as shown in two distinct genetic models (H19DMD^fl/+; Nkx3.2^+/Cre, and Igf2r^fl/+; Nkx3.2^+/Cre). Importantly, deletion of Igf2 from the mesenchyme leads to a reduction in beta-cell mass (~82% N at P2 and ~84% N in pregnant females at E15), showing that the paracrine effects are not exclusive to the exocrine pancreas. We found that mesenchyme mass at postnatal day 2 was not affected by the deletion of mesenchymal Igf2 or overexpression of IGF2 in the mesenchymal Igf2r mutant model. These findings strongly support the hypothesis that the effects of the mesenchyme are mainly paracrine, thus suggesting that a) the pancreas growth restriction occurs independently from the loss of mesenchyme-derived cells; b) the function of the IGF2 producing mesenchyme-derived cells is mainly to act as a developmental reservoir for IGF2 paracrine signalling. The lack of mesenchymal IGF2 has important physiological consequences. It is well established that the pancreas size must match the physiological demands of the host organism to promote nutrient digestion and absorption in the gut and to maintain glucose homeostasis. In this study, we report that mesenchyme Igf2-deficient mice have a disproportionately smaller pancreas, with a 30% size reduction at P2, which is associated with a reduced output of secreted acinar enzymes, such as lipase. Our data suggests that nutrient malabsorption due to loss of exocrine mass and/or exocrine dysfunction might be causative of the postnatal growth restriction, which is first observed in these mice at weaning and maintained throughout adulthood. Mouse models with perturbations of lipase genes such as colipase (Clps) and pancreatic-lipase related protein 2 (Pnliprp2) exhibit a similar body weight growth defect phenotype due to an inability to process fat from mother’s milk [34,35]. It has been suggested that the lipase deficiency can programme a “set-point” of body weight, with an inability to catch-up in body weight later, due to the effects of poor weight gain during the suckling period [34,35]. Interestingly, RNA-seq shows that mesenchyme Igf2-deficient pancreases upregulate transcription of exocrine genes involved in the production of digestive enzymes. We suggest that these changes represent adaptive responses of the small exocrine pancreas to meet the demand for digestion and absorption of nutrients in early postnatal life. However, this compensatory mechanism might not be sufficient to match the demand, and whole-body growth restriction ensues. The RNA-seq studies also showed increased expression of chemokines, which are normally up-regulated by pro-inflammatory stimuli and attract immune cells by inducing a response from the innate and adaptive immune system [36,37]. Whether a pre-pancreatitis state exists, linked to the premature activation of digestive enzymes in the interstitial space of the pancreas, and how such a state might contribute to putative nutrient absorption abnormalities require further investigation. We also observed that the loss of IGF2 paracrine effects from the mesenchyme results in a modest reduction of the beta-cell mass that does not impact on glucose homeostasis control in young mice under a normal diet and physiological state. However, induced metabolic stress, as it occurs naturally during pregnancy, leads to glucose intolerance. We suggest that the loss of beta-cell mass in the mutant pregnant females compared to controls is a contributing factor for the observed glucose intolerance. Further work is required to understand the precise mechanisms, which includes the generation of a double knockout of Igf2 from the mesenchyme and the beta-cell. It is possible that a more severe glucose homeostasis in pregnancy will occur when the known autocrine IGF2 actions in the beta-cell are removed [22].

Our findings offer novel insights into mesenchymal factors that regulate pancreas organogenesis. Wnt, BMP, FGF and Hedgehog signalling pathways have been implicated as early mesenchymal factors regulating pancreas development and growth [2,38–42], but for the majority of these studies the specific roles in the mesenchyme could not be assessed (it has only recently become possible to start interrogating the roles that genes play in the mesenchyme with the generation of specific Cre recombinases). To our knowledge, this study, using a combination of loss and gain-of-function conditional mouse models, is the first to identify a mesenchymal-specific paracrine growth signal of the developing pancreas. It also reveals that IGF signalling plays previously unappreciated roles in the mesenchyme function and in the epithelial-mesenchymal interactions during pancreas organogenesis. Igf1, like Igf2, is highly enriched in the mesenchyme-derived cells compared to non-mesenchyme cells and paracrine release of IGF1 from fibroblasts, which derive from the primitive mesenchyme, stimulate acinar cell proliferation during regeneration from acute pancreatitis [43]. Interestingly, deletion of Igf1 from the developing epithelium also results in normal exocrine pancreas growth [44], similarly to the results reported here for Igf2. Our study has also highlighted a number of growth-related imprinted genes (including Plagl1/Zac1 a master regulator of an imprinting growth network [45]) that are highly enriched in the mesenchyme, and raises the interesting hypothesis that genomic imprinting contributes significantly to mesenchymal-specific functions during pancreas development.

Our postnatal studies provide proof of principle that lack of mesenchyme-derived IGF2 has physiological consequences related to whole-body growth (first observable around weaning) and glucose intolerance during pregnancy. It remains to be established if pregnancy-related beta-cell mass expansion is impaired and the extent to which other metabolic stress states, such as ageing, high fat diet, acute insulin resistance states, may uncover mesenchymal IGF2 effects on beta-cell function.

Our study has a number of limitations. Although we provide evidence that mesenchymal IGF2 is a long-sought promotor of acinar growth, we have not established the precise timing of the actions of IGF2. IGF2 is well known to promote cell proliferation and survival, with the main wave of these actions thought to occur in the mouse between E9-E10 [46]. Therefore, it is likely that many of the mesenchymal IGF2 actions in promoting pancreatic growth start during the earliest stages of pancreas development. However, based on RNA-seq data we also observed very high levels of Igf2 expression at E13.5, E15.5, E17.5 and P2. Further to the well-established mitogenic roles, IGF2 has anti-apoptotic actions in the endocrine pancreas during the neonatal life [7,24]. Future systematic analyses based on pancreas stereology and measurements of cell proliferation and apoptosis are therefore required to fully understand if IGF2 actions occur in the early mesenchyme and/or mesenchyme-derived cells in later fetal and perinatal development, and the extent to which it influences the differentiation programmes of progenitor cells into exocrine, endocrine and mesenchyme lineages. The genetic models we use in this study have some limitations as well. The effects of both H19DMD^fl/+; Nkx3.2^+/Cre, and Igf2r^fl/+; Nkx3.2^+/Cre models are not exclusively mediated through IGF2 signalling–H19DMD^fl/+; Nkx3.2^+/Cre results in downregulation of H19 ncRNA and its associated exonic miR-675 [47,48], and Igf2r serves as a mannose-6-phosphate receptor in addition to acting as an IGF2 clearance receptor/signalling through G proteins [33,49]. Importantly, the complete rescue of the overgrowth phenotypes due to deletions of maternal H19 or Igf2r alleles in an Igf2-null background [50,51] shows that the majority of the effects are mediated through IGF2 signalling. However, we cannot exclude that subtle effects caused by IGF2-independent actions, may explain some of the inconsistent observations regarding mesenchyme and beta-cell mass between these models (Fig 3, Fig 4).

In summary, we report in this study that IGF2 growth actions in the pancreas are limited to the highly expressing mesenchyme-derived cells, and that size-determining mechanisms are programmed in early life by the activity of this important growth factor in a paracrine fashion, with consequences for post-natal physiology. In general, this work provides novel insights into how growth factors control pancreatic architecture and is highly relevant to pancreas pathologies.

Materials and methods

Ethics statement

This research has been regulated under the Animals (Scientific Procedures) Act 1986 Amendment Regulations 2012 following ethical review by the University of Cambridge Animal Welfare and Ethical Review Body (AWERB). All mouse experiments were performed under PPL No. 80/2484 (study plan 2484/25/13) and PPL No. 70/7594 (study plan 7594/4/15).

Generation of the Igf2^fl/+ mouse

The Igf2 gene targeting vector carried a loxP site inserted 5' of exon 4 and a loxP-flanked neomycin resistance cassette (neoR) inserted 3' of exon 6 (S4A Fig). Details of the cloning procedures are available upon request. In brief, we used a 3.1-kb SalI-PacI genomic fragment (from exon 2 to intron 3 of Igf2) as the 5’ region of homology (5’-ROH), a 5.8-kb PacI-AfeI genomic fragment that includes the entire Igf2 coding sequence (exons 4 to 6) as internal ROH (Int-ROH), and a 3.1-kb AfeI-BlpI genomic fragment (3’ of exon 6) as 3’-ROH. The Int-ROH was flanked by a 5’ loxP site and a 3’ loxP-neoR-loxP sequence (S4A Fig). The targeting vector was linearized at a unique ScaI site outside the area of homology and 50 μg linearized vector were electroporated into passage 9, E14 129ola male ES cells, at 250V and 950 μF. Transfected cells were plated onto 10 gelatinized 100-mm dishes pre-seeded with fibroblast feeder cells. After 24 h in nonselective medium, cells were incubated for 8 days with G418 medium (200 μg/μl) to select for neomycin resistance. Resistant clones were picked at day 9 and expanded into 96-well plates pre-seeded with fibroblast feeder cells. We initially screened 384 G418-resistant clones by Southern blotting analysis of genomic DNA (gDNA) digested with EcoRI and hybridized the blots with a unique 511 bp 5' probe (located external to 5’-ROH and obtained by PCR amplification using primers 5’Pr-F: 5’-AACAACGCGGTGGTAGGGAA-3’ and 5’Pr-R: 5’-TCAGCAGAAAAAGAAGCAGGGC-3’). Two correctly targeted clones at the 5’ end were then verified by Southern blot (EcoRI digested DNA) using a 590 bp 3’ probe (located external to 3’-ROH and obtained by PCR amplification using primers 3’Pr-F: 5’-ACAAAGCCCAAGACAACTCC-3’ and 3’Pr-R: 5’-CTTCCACAGTTCAAGCAACC-3’) and an additional check for multiple integrations elsewhere in the genome using a 583 bp Internal probe (located in Igf2 exon 6 and obtained by PCR amplification with primers Int-F: 5’-AGAACCCAAGAAGAAAGGAAG-3’ and Int-R: 5’-AGAAAGACAGAACTAGCAGCC-3’). One clone with a single integration site and correctly targeted 5’ and 3’ loxP sites was thus identified (S4A Fig), with the loxP sequences further verified by Sanger sequencing. The neoR cassette was excised by transiently transfecting this ES cell clone with Cre recombinase (pMC-Cre vector), followed by two rounds of subcloning. Correctly excised clones that carry a single 3’ loxP site were verified by PCR screening (480 ES subclones) and gDNA digestion with SphI restriction enzyme, followed by Southern blot analysis using the same 583 bp Internal probe described above (S4C–S4F Fig). Primers used for PCR screening of in vitro neoR deletion were: F0 5’-TGACCTCAGCAATTCAAGTCC-3’; F1 5’-GGTAGTGGTCTTTGGCATCC-3’ and R1 5’-CAATAACTGGGGAAAAGGAGC-3’. Two independent ES clones were then microinjected into C57BL/6J blastocysts and transferred into (C57BL/6J X CBA/Ca)F1 pseudo-pregnant females to generate chimeric mice. 27 chimeras were born, all males. Germline transmitting mice were backcrossed into the C57BL/6J genetic background for more than 10 generations before being used as experimental animals. Paternal transmission of the Igf2 floxed allele did not have any impact on fetal and placental growth kinetics (S4G and S4H Fig). Homozygous Igf2^fl/fl mice were obtained by breeding heterozygous Igf2^fl/+ parents.

Mouse strains

Targeted ES cells for the miR-483 knock-out [31] were generated and provided by Dr. Haydn Prosser (Sanger Institute and International Knockout Mouse Consortium Project). Details on the generation of the miR-483 knock-out mice (depicted in S7 Fig; Sekita et al., manuscript in preparation) are available upon request. Rosa26YFP-stop^fl/fl mice [26] were kindly provided by Dr. Martin Turner (The Babraham Institute, Cambridge). H19DMD^fl/fl mice [32] and Igf2r^fl/fl mice [52] were generously provided by Prof. Bass Hassan (University of Oxford). CMV-Cre mice [53] were obtained from the Babraham Institute, Cambridge. This Cre recombinase is expressed soon after fertilization and allows ubiquitous deletion of floxed alleles in all tissues, including the germline [53]. Nkx3.2-knock-in-Cre mice [27] were kindly provided by Dr. Warren Zimmer (Texas A&M University). Nkx3.2 expression occurs in the mesenchyme of the developing pancreas, stomach and gut, as well as in the forming somites, but not in the endoderm-derived cells of these organs. Nkx3.2 is expressed in the pancreatic mesenchyme as early as E9.5 and, by E12.5, its expression becomes restricted to the mesenchymal area, which will give rise to the splenic bud [27]. The Nkx3.2-knock-in-Cre, in which a Cre recombinase cDNA cassette and a PGK-NeoR cassette were inserted in frame within exon 1 of Nkx3.2 (Bapx1) gene, faithfully replicates endogenous Nkx3.2 expression and directs Cre activity to the foregut mesenchyme and skeletal somites starting at E9.5 [2,27]. We verified the pancreatic cell-type specificity of Nkx3.2-Cre expression using the Rosa26YFP-stop^fl/fl reporter mice and demonstrated absence of recombinase activity in acinar cells, pancreatic endocrine cells, endothelial cells or duct cells (Fig 1A) at postnatal day P2. Tek-Cre transgenic mice, in which Cre expression is controlled by the endothelial-specific receptor tyrosine kinase Tek (Tie2) promoter/enhancer and therefore found uniformly in endothelial cells during embryogenesis (E7.5 onwards) and adulthood [30], were imported from the Jackson Laboratory (Maine, USA). Ptf1a-knock-in-Cre mice, in which the protein-coding region of Ptf1a (exon 1 and 2) was replaced with a Cre-cassette, express Cre in pancreatic ducts, exocrine and endocrine cells as early as E10.5 [28]. In newborn pups Ptf1a-Cre expression marks all acinar cells and roughly 95% of ductal and insulin-producing cells, as well as 75% of glucagon-producing cells [28]. RIP-Cre mice, carry a Cre transgene under the control of the rat Ins2 (insulin 2) promoter (RIP) that directs expression to insulin-positive beta-cells from approximately E8.5–9 onwards [29]. Ptf1a-Cre and RIP-Cre strains were obtained from Central Biomedical Services (CBS Transgenic Services, University of Cambridge). All lines were bred into an inbred C57BL/6J line for >10 generations.

Mouse crosses and genotyping

Mice were fed standard chow diet with 9% of kcal from fat (SDS, Essex, UK) and housed with a 12-h light/dark cycle in a temperature-controlled room (22°C). For timed mating, the day of detection of a vaginal plug was noted as embryonic day 1 (E1) and the day of birth was noted as post-natal day 0 (P0). Mice were weaned at 3 weeks of age and ear notches were used for visual identification and to collect tissue for PCR genotyping.

Mouse crosses used for each experiment are listed in S1 Table. Throughout the paper, Igf2^+/fl represents genotype of offspring that inherited the floxed allele from father; Igf2^fl/+ represents genotype of offspring that inherited the floxed allele from mother; Igf2^+/+ represents genotype of offspring with wild-type alleles at the Igf2 locus; Nkx3.2-Cre represents genotype of offspring that are heterozygous Cre-carriers (^Cre/+ or ^+/Cre); Igf2^+/fl; Nkx3.2^Cre/+ represents offspring with mesenchymal Igf2 deletion on the paternal allele, having inherited the floxed allele from father and the Cre allele from mother; Igf2^fl/+; Nkx3.2^+/Cre represents offspring with mesenchymal Igf2 deletion on the maternal allele, having inherited the floxed allele from mother and the Cre allele from father; H19DMD^fl/+; Nkx3.2^+/Cre represents offspring with mesenchymal H19DMD deletion on the maternal allele, having inherited the floxed allele from mother and the Cre allele from father. All strains were genotyped by standard PCR using DNA extracted from ear biopsies (adult mice) or tail DNA (embryos or post-terminal). PCR was performed using the Red Taq Ready PCR system (Sigma Aldrich) using primers described in S2 Table, followed by separation of PCR amplicons by agarose gel electrophoresis.

Southern blotting screening of ES clones

gDNA were digested with EcoRI or SphI restriction enzymes and electrophoresed on 0.8% agarose gels in 1×TBE buffer, alkaline blotted onto Hybond N+ membranes (Amersham), and UV cross-linked (Stratalinker, Stratagene). Probes were obtained by PCR and radiolabeled (α-32P-CTP). Hybridisation and washing of Southern blots was performed as described [54]. Membranes were exposed overnight to MS film (Kodak).

Northern blotting analysis of Igf2 mRNA

Total RNA (10 μg) extracted from E19 placenta and liver samples using RNeasy midi kits (Qiagen) according to manufacturer’s protocol, was separated in low-percentage formaldehyde gels, blotted onto Nytran-plus membrane (Schleicher and Schuell), and UV cross-linked (Stratalinker, Stratagene). The RNA blots were hybridized with radiolabelled (α-32P-UTP) Igf2 and Gapdh cDNA probes. We carried out hybridization and washing of northern blots as described [55]. Transcript levels were quantified by PhosphorImager analysis (Molecular analyst software, Biorad).

Western blotting analyses

For Western blot analysis of IGF2, 5 μl serum samples collected at E19 were loaded into Bis-Tris gels (NuPAGE Novex, Life technologies) and transferred to nitrocellulose membranes (Invitrogen). 4 ng human recombinant IGF2 (292-G2-050, R&D Systems) was used as positive control, and the Novex Sharp protein ladder (Life technologies) was used as marker. The membranes were initially stained with Ponceau red (Invitrogen) and imaged with the Biorad GelDoc system, the intensity of bands being used as internal control for protein loading. The membranes were then blocked with 5% skimmed milk in TBS-T for 1 h at 4°C, after which were incubated overnight at 4°C with goat anti-human IGF2 antibody (R&D systems AF-292) diluted 1: 1,000 in TBS-T containing 0.3% skimmed milk. After 4x10 min washes with milliQ water, the blots were incubated for 1h at room temperature with the secondary antibody (1:2,000 rabbit anti-goat IgG coupled to HRP, SantaCruz) in TBS-T containing 3% skimmed milk. Blots were washed 4x10 min with milliQ water, exposed to substrate (Clarity ECL Western Blotting Substrate, Biorad) for 5 minutes and imaged with the Biorad GelDoc system. Bands were quantified using the ImageLab software (Biorad), the output values being normalized to the corresponding Ponceau red loading control values.

For Western blot analysis of AKT phosphorylation, acinar cell pellets were lysed in 30μl RIPA containing protease and phosphatase inhibitors. Protein concentration was determined by DC assay (Biorad) and 10μg/lane total cell lysates were electrophoresed through 4–12% Bis-Tris NuPAGE gels with MOPS running buffer. Proteins were transferred to nitrocellulose by iBlotII (Invitrogen) prior to being blocked for an hour in 3%BSA/TBST. Primary antibody to pAKT (S473) (CST 9271), total AKT (CST 2920), or Beta-actin (CST 4967) was incubated overnight at 4°C. Membrane was washed 5 times before detection of bound antibody by chemiluminescent signal associated with HRP-conjugated anti-rabbit or anti-mouse secondary antibody.

Pancreas immunostainings, imaging and stereology analysis

Pancreases were dissected under microscope (for embryonic and early post-natal analyses), fixed in 4% paraformaldehyde in PBS overnight, dehydrated and then embedded in paraffin. Paraffin blocks were cut at 5 μm thickness, sections were then deparaffinised, rehydrated, stained and mounted with coverslips. For all stereological analyses, in order to obtain accurate morphometric estimations without measuring more than once the same feature (e.g. pancreatic islands), sections spaced at 100 μm (P2) or 200 μm (pregnant females) were stained (6–8 sections/block for P2 and 20–24 sections/block for pregnant females). Conditions used for antigen retrieval, blocking and combinations of primary and secondary antibodies used for each staining are described in S3 Table. Counterstaining was performed with haematoxylin for light microscopy stains or DAPI for fluorescent stains.

Light microscopy stained slides were imaged using the NanoZoomer whole slide scanner (Hamamatsu). Fluorescent immunostainings were imaged with a LSM510 Meta confocal laser scanning microscope and the ZEN 2009 software (Carl Zeiss, Germany) or scanned using an Axio whole slide scanner. Whole slide scans of stained sections were analysed using the Visiopharm automated image quantification software (Visiopharm, Denmark) or HALO AI (Indica Labs) to measure the area of positive cell-specific staining or to count individual nuclei. The mass of each cell-type was calculated using the formula:

mRNA in situ hybridization (ISH) for Igf2

ISH was performed as described [56], with minor modifications. Briefly, a region of 415 bp spanning coding Igf2 exons 4–6 was amplified by PCR using primers: F: 5’-CACGCTTCAGTTTGTCTGTTCG-3’ and R: 5’-GCTGGACATCTCCGAAGAGG-3’ and cDNA from whole P2 pancreases as template. The PCR product was cloned into a pCR2.1-TOPO plasmid (Invitrogen) containing M13 primers and a T7 RNA polymerase transcription initiation site. Sense (S) and antisense (AS) RNA probes were generated and labelled with Digoxigenin (DIG) by in vitro reverse transcription, according to manufacturer’s instructions (Roche). Pancreases were dissected in cold phosphate buffered saline (PBS) and fixed overnight in 4% paraformaldehyde in 0.1% diethylpyrocarbonate (DEPC)-PBS at 4°C. After rinsing in DEPC-PBS, tissues were dehydrated and then embedded in paraffin in RNase-free conditions. Pancreas sections (7 μm thick) mounted on polysine slides (VWR) were de-waxed, rehydrated in PBS, post-fixed in 4% paraformaldehyde for 10 minutes, treated with proteinase K (30 μg/ml) for 10 min at room temperature, acetylated for 10 minutes (acetic anhydride, 0.25%) and hybridized overnight at 65°C in a humidified chamber with DIG-labeled probes diluted in hybridization buffer (containing 200 mM sodium choride, 13 mM tris, 5 mM sodium phosphate monobasic, 5 mM sodium phosphate dibasic, 5 mM EDTA, 50% formamide, 10% dextran sulfate, 1 mg/ml yeast tRNA and 1× Denhardt's [1% w/v bovine serum albumin, 1% w/v Ficoll, 1% w/v polyvinylpyrrolidone]). Two 65°C post-hybridization washes (1× SSC, 50% formamide, 0.1% tween-20) followed by two room temperature washes in 1× MABT (150 mM sodium chloride, 100 mM maleic acid, 0.1% tween-20, pH7.5) were followed by 30 minutes RNAse treatment (400 mM sodium chloride, 10 mM tris pH7.5, 5 mM EDTA, 20 μg/ml RNAse A). Sections were blocked for 1 hour in 1×MABT, 2% blocking reagent (Roche), 20% heat inactivated goat serum and then incubated overnight with anti-DIG antibody (Roche; 1:2,500 dilution) at 4°C. After 4x20 min washes in 1× MABT, slides were rinsed in 1× NTMT (100 mM NaCl, 50 mM MgCl, 100 mM tris pH 9.5, 0.1% tween-20) and incubated with NBT/BCIP mix in NTMT buffer, according to manufacturer’s instructions (Promega). Slides were counterstained with nuclear fast red (Sigma), dehydrated and cleared in xylene and mounted in DPX mounting medium (Sigma). Pictures were taken with a camera attached to a light microscope.

Fluorescence-activated cell sorting (FACS)

Mice were sacrificed by decapitation (embryos and P2 neonates) or cervical dislocation (P5 neonates and older). Then, pancreases were dissected and dissociated into single cells using trypsin-EDTA (Sigma Aldrich) at 37°C for 20 min. After one wash with ice-cold PBS, the cells were passed through 70 μm strainers and single-cell suspensions were sorted into YFP-positive and YFP-negative fractions using an Aria-Fusion cell sorter (BD Bioscience). Dead cells were excluded based on forward and side scatter profiles and the uptake of 7AAD (7-Aminoactinomycin D dead cell stain, Life Technologies). Sorted YFP-positive and YFP-negative cells were pelleted by centrifugation and flash frozen in liquid nitrogen, then stored at -80°C until use.

Total RNA and microRNA extraction (pancreas and sorted cells)

Total RNA was extracted from whole pancreases, cells isolated by FACS or cultured cells using the RNEasy Mini kit (Qiagen) or the RNEasy Micro kit (Qiagen), respectively, according to manufacturer’s instructions, with additional removal of contaminating DNA using QIASpin DNA eliminator columns. microRNA was extracted using the miRNeasy kit (Qiagen), according to manufacturer’s instructions with an additional step of DNaseI treatment (Qiagen).

Quantitative real-time PCR (qRT-PCR)

Concentration and integrity of total RNA was verified by NanoDrop (Thermo Scientific) and agarose gel electrophoresis, respectively. For RNA extracted from whole pancreases, reverse transcription was performed using the RevertAid RT Reverse Transcription Kit (Life technologies), according to manufacturer’s instructions. In the case of total RNA extracted from sorted cells, cDNA was produced using the QuantiTect Whole Transcriptome Kit (Qiagen) following manufacturer’s instructions. Primers used for qRT-PCR are listed in S4 Table. Annealing temperatures were tested by gradient PCR using pancreas-cDNA as a template. qRT-PCR was performed using the SYBR Green JumpStart Taq Ready Mix (Sigma Aldrich) on an ABI Prism 7900 system (Applied Biosystems). Gene expression levels were normalized to the housekeeping gene Ppia (peptidylpropyl isomerase A or cyclophilin-A), which has previously been established as a good housekeeping gene for pancreas gene expression studies [57] and is stably expressed between various developmental time points. For microRNAs, reverse transcription was performed using the TaqMan Micro RNA Reverse Transcription kit (Applied Byosystems), according to manufacturer’s protocol. qRT-PCR was performed using TaqMan assays (TM: 002560 for mmu-miR-483; TM: 001232 for snoRNA202 and TM: 001234 for snoRNA234) and TaqMan 2x Universal PCR Master Mix (Applied Byosystems). Levels of expression were calculated using the 2^-ΔΔCt method [58].

Complementary DNA library preparation and RNA-seq analysis

Input RNA for genome-wide transcriptome analysis was verified for concentration and quality using Agilent RNA Pico chips, according to manufacturer’s instructions. All RNA samples had RNA integrity numbers (RIN) >7.5. Total RNA (2 ng) was whole-transcriptome amplified using the Ovation RNA–seq System V2 (NuGEN). To prepare the RNA–seq libraries, the amplified cDNA (2 μg per sample) was fragmented to 200 bp using a Bioruptor Sonicator (Diagenode), and barcode ligation and end repair were performed using the Ovation Rapid DR Library System (NuGEN). The barcoded libraries were combined and loaded onto an Illumina HiSeq 2500 system for single-end 50-bp sequencing at the Genomics Core Facility, Cambridge Institute, CRUK. The reads were aligned onto the mouse GRCm38 genome using TopHat 2.0.11 [59]. Gene abundance and differential expression were determined with Cufflinks 2.2.1 [60] and expressed in fragments per kilobase per million mapped reads (FPKM). The cut off for expression was set at ≥ 1 FPKM. Genes with a linear fold expression change greater than 1.5 and a Benjamini–Hochberg false discovery rate <5% were considered differentially expressed.

Functional annotation and enrichment analysis

DAVID (Database for Annotation, Visualization and Integrated Discovery; v6.8 http://david.abcc.ncifcrf.gov/, accessed March 2017) was performed to assess whether there was enrichment for genes implicated in particular biological processes within the differential expressed gene lists identified by RNA-seq. Enriched gene ontology (GO) terms with an FDR < 0.05 were considered significant. These terms were then clustered semantically using REViGO (Reduce and Visualize GO) [61], which removes redundancy. The results obtained by REViGO were ordered according to log10 p values.

Whole pancreas lipase content and amylase measurements

For whole pancreas lipase measurements, frozen samples were removed from -80°C and thawed on ice. 100 μL TK lysis buffer with proteinase inhibitors (Calbiochem) was added to each pancreas and tissues were disrupted on ice with a homogenizer (Biospec Tissue Tearor). Lysates were spun at 3,000 rpm for 15 min at 4°C and supernatants were used for analysis. Lipase levels were measured using a lipase activity assay (Dimension RXL, Siemens).

For amylase measurements, cultured acinar cells were lysed in 100 μL RIPA buffer, spun at 3,000 rpm for 15 min at 4°C and the supernatants used for analysis. Amylase levels were measured in an autoanalyser (Siemens Dimension RXL) through a colorimetric reaction based on the ability of amylase to hydrolyse the chromogenic substrate 2-chloro-4-nitrophenol linked with maltotriose (CNPG3) into the coloured product 2-chloro-4-nitrophenol (CNP) plus CNPG2, maltotriose G3 and glucose. The absorbance increase at 405 nm is proportional to the amylase activity in the sample. Amylase levels (U/L) were normalized to the total protein content determined by a BCA assay (Pierce BCA Protein Assay Kit, Thermo Fisher Scientific, 23225).

Primary pancreatic acinar and mesenchymal cell isolation and culture

Primary pancreatic cells were isolated as previously described [62] and adapted here to P2 samples. Briefly, pancreases from an entire P2 litter were micro-dissected under microscope and sterile conditions and pooled in one tube containing HBSS, 0.25 mg/ml of trypsin inhibitor and 1% Penicillin-Streptomycin mix. After rinsing, pancreases were digested for 20–30 min at 37°C with a collagenase IA solution (HBSS containing 10mM HEPES, 200 U/ml of collagenase IA, and 0.25 mg/ml trypsin inhibitor). The digestion was stopped by placing the samples on ice and addition of FBS (fetal bovine serum) for a 2.5% final concentration. After additional washing steps, the cell suspension was passed through a 100 μm cell strainer, allowing the passage of acini and mesenchymal cells, while retaining the non-digested fragments and larger pancreatic islets. The cells were placed in the culture media (basal Waymouth's media containing 2.5% FBS, 1% Penicillin-Streptomycin mix, 0.25 mg/ml trypsin inhibitor, and 25 ng/ml of recombinant human EGF) and plated into six-well plates pre-coated with poly-L-lysine. After one hour incubation at 37°C under 5% (v/v) CO2 atmosphere, the floating acini were transferred to a new well, while the mesenchyme cells remained attached. The cells were cultured under the above conditions for up to four days, with daily media changes.

Measurement of IGF2 secretion by primary mesenchymal and acinar cells in vitro

The primary mesenchymal and acinar cells were isolated and cultured as described above until reaching confluency. Media (1mL) was collected 24h post-confluency and freeze-dried using a Christ Gamma 2–16 LSC Freeze dryer. Dry pellets were re-dissolved in 50 μl RIPA buffer and IGF2 measured by ELISA (Mouse IGF-II DuoSet ELISA kit, R&D Systems–DY792) using an assay adapted for the MesoScale Discovery electrochemiluminescence immunoassay platform (MSD), as recently described [63].

In vitro IGF2 treatments

To assess the impact of exogenous IGF2 on intracellular signalling via AKT, freshly isolated acinar cells were starved for two hours in basal Waymouth's media, without FBS and EGF. IGF2 treatments (with vehicle or 1 nM, 10 nM or 50 nM mouse recombinant IGF2) were performed for 10 min at 37°C under 5% (v/v) CO2 atmosphere, then the cells were immediately placed on ice, washed two times in ice-cold PBS, pelleted and flash-frozen on dry-ice and stored at -80°C until analysed by Western blotting.

To assess the impact of exogenous IGF2 on amylase production by acinar cells, subconfluent acinar cultures (48 hours after isolation) were placed in basal Waymouth's media containing 1% Penicillin-Streptomycin mix and 0.25 mg/ml of trypsin inhibitor. FBS was replaced with Serum Replacement (Sigma Aldrich S0638) that does not contain any IGF2. The treated cells received 50 ng/ml mouse recombinant IGF2 for 48 hours, with new media added every day. After the treatment, cells were harvested and dissociated into single-cell suspension using trypsin, counted with a Cedex XS Analyser (Roche), washed, pelleted and stored at -80°C until used for amylase measurement.

Oral glucose tolerance test

Conscious mice were used for an oral glucose tolerance test after 6 hours of fasting (from 8am to 2pm). Blood samples (≤5 μL) were taken from the tail vein immediately before administration of glucose by oral gavage (20% weight for volume, 2g/kg body weight based on average body weights for each experimental group) and thereafter at the time points indicated, and used for glucose measurements with a glucose meter (AlphaTRAK).

Statistical analysis

Numerical data used to generate all graphs presented in this study are shown in S5 Table. Statistical analysis was performed using GraphPad Prism 7 software. For two groups with up to 6 samples, statistical analysis was performed using Mann-Whitney or the Wilcoxon matched-pairs signed rank test. For two groups with ≥6 samples, to determine whether the data was of a Gaussian distribution, the Shapiro-Wilk test was first applied, followed by un-paired or paired Student’s t-tests, as appropriate. Where more than two groups exposed to the same treatment were analysed, 1-way ANOVA with Dunnett’s post-hoc or Friedman’s test with Dunn’s correction tests were used, comparing every mean to a control mean. OGTT data was analysed using two-way ANOVA with Sidak’s multiple comparison tests, using the time from administration and the genotype as factors (two genotypes) or by repeated measures 1-way ANOVA with Dunnett’s post hoc test for multiple comparisons against the controls (four genotypes). The area under the curve (AUC) was calculated by the trapezoidal rule and used for analyses by unpaired Student’s t tests (two genotypes) or by 1-way ANOVA with Dunnett’s post-hoc tests (four genotypes). Unless stated otherwise, data is shown as average values and error bars represent SEM or SD. N.S. p >0.05; * p ≤0.05; ** p ≤0.01, *** p ≤0.001.

Supporting information

S1 Data [xlsx]
List of genes significantly enriched (fold change >1.5; FDR adjusted p value <0.05) in mesenchyme and non-mesenchyme pancreatic cells at P2 in wild-type mice (Igf2+/+; Rosa26YFP-stop+/fl; Nkx3.2Cre/+) and associated functional pathway analyses (DAVID).

S2 Data [xlsx]
List of DEGs (fold change >1.5; FDR adjusted p value <0.05) in non-mesenchymal cells between Igf2+/fl; Rosa26YFP-stop+/fl; Nkx3.2Cre/+ knockouts and Igf2+/+; Rosa26YFP-stop+/fl; Nkx3.2Cre/+ controls at P2, and associated functional pathway analyses (DAVID).

S3 Data [xlsx]
List of DEGs (fold change >1.5; FDR adjusted p value <0.05) in mesenchyme-derived cells between Igf2+/fl; Rosa26YFP-stop+/fl; Nkx3.2Cre/+ knockouts and Igf2+/+; Rosa26YFP-stop+/fl; Nkx3.2Cre/+ controls at P2, and associated functional pathway analyses (DAVID).

S1 Fig [a]
Assessment of efficiency of pancreas-expressing Cre lines using the -stopfl/fl reporter mouse, at postnatal day 2 (P2).

S2 Fig [a]
Gene expression patterns of pancreatic mesenchyme-derived cells and non-mesenchyme cells identified by RNA-seq at postnatal day 2 (P2).

S3 Fig [a]
Expression of marker genes in mesenchyme-derived cells by RNA-seq.

S4 Fig [a]
Generation of conditional knockout mice.

S5 Fig [a]
-Cre-mediated deletion of the floxed allele in vivo.

S6 Fig [a]
Inheritance of -Cre or floxed alleles, or deletion of maternal allele do not affect total body or pancreas weights at postnatal day 2 (P2).

S7 Fig [a]
Paternally inherited deletion does not lead to reduction in pancreas size.

S8 Fig [a]
Deletion of paternal in pancreatic mesenchyme leads to reduced body and pancreas weights at weaning (P21).

S1 Table [docx]
Mouse strains and crosses.

S2 Table [docx]
Primers used for genotyping by PCR.

S3 Table [docx]
Conditions used for pancreas immunostaining.

S4 Table [docx]
Primers/Assays used for qRT-PCR.

S5 Table [xlsx]
Numerical data underlying graphs shown in this study.