Reciprocal modulation of adult beta cell maturity by activin A and follistatin
- First Online:
- Cite this article as:
- Szabat, M., Johnson, J.D. & Piret, J.M. Diabetologia (2010) 53: 1680. doi:10.1007/s00125-010-1758-0
- 637 Downloads
The functional maturity of pancreatic beta cells is impaired in diabetes mellitus. We sought to define factors that can influence adult beta cell maturation status and function.
MIN6 cells labelled with a Pdx1 monomeric red fluorescent protein–Ins1 enhanced green fluorescent protein dual reporter lentivirus were used to screen candidate growth and/or differentiation factors using image-based approaches with confirmation by real-time RT-PCR and assays of beta cell function using primary mouse islets.
Activin A strikingly decreased the number of mature beta cells and increased the number of immature beta cells. While activins are critical for pancreatic morphogenesis, their role in adult beta cells remains controversial. In primary islets and MIN6 cells, activin A significantly decreased the expression of insulin and several genes associated with beta cell maturity (e.g. Pdx1, Mafa, Glut2 [also known as Slc2a2]). Genes found in immature beta cells (e.g. Mafb) tended to be upregulated by activin A. Insulin secretion was also reduced by activin A. In addition, activin A-treated MIN6 cells proliferated faster than non-treated cells. The effects of endogenous activin A on beta cells were completely reversed by exogenous follistatin.
These results suggest that autocrine and/or paracrine activin A signalling exerts a suppressive effect on adult beta cell maturation and function. Thus, the maturation state of adult beta cells can be modulated by external factors in culture. Interventions inhibiting activin or its signalling pathways may improve beta cell function. Understanding of maturation and plasticity of adult pancreatic tissue has significant implications for islet regeneration and for in vitro generation of functional beta cells.
KeywordsBeta cellsGene expressionHuman and mouse isletsInsulinDifferentiationMIN6PDX1Reporter lentivirus
Bone morphogenetic protein
Enhanced green fluorescent protein
Monomeric red fluorescent protein
The loss of functional pancreatic beta cell mass is a hallmark of diabetes. A fundamental understanding of pancreatic beta cell fate decisions and the process of beta cell maturation is imperative to correct this defect. Using a dual reporter lentiviral system to perform single-cell analysis of beta cell differentiation, we recently characterised a dynamic immature beta cell state in adult islet cells from humans and mice, distinguished by Pdx1 promoter activity prior to Ins1 promoter activity (Pdx1+/Inslow) . MIN6 beta cell maturation is marked by the acquisition of Ins1 promoter activity and takes less than 12 h .
Functional beta cell mass can adapt to changes in metabolic demand resulting from obesity or pregnancy , suggesting that physiological and pathophysiological factors modulate the differentiation status of adult beta cells. In vitro and in vivo, pancreatic beta cells have also been stimulated to proliferate, dedifferentiate and transdifferentiate [3–5]. The plasticity of adult human pancreatic tissue has significant implications for islet regeneration and for in vitro generation of functional beta cells. However, the specific conditions and molecular cues that drive these mechanisms remain to be elucidated.
Here, our aim was to define factors that modulate the maturation state of adult beta cells. Factorial design of experiments  was used to compare multiple candidate growth and differentiation factors simultaneously. Based on factors reported to influence the development or differentiation of beta cells, we examined the effects of glucose , nicotinamide , exendin 4 , insulin [10, 11], IGF-1 , betacellulin , laminin-1 , epidermal growth factor , retinoic acid , gastrin17 , hepatocyte growth factor  and activin A . To date, the effects of these factors and their interactions have not been systematically compared in one study.
Activins, members of the TGFβ superfamily, elicit numerous context-dependent effects on growth and differentiation . Activins control embryonic patterning of foregut-derived organs , have important roles in pancreatic development  and have been implicated in the control of insulin secretion [21, 22]. Activins and their receptors are present in the developing pancreas and adult islet cells [23, 24]. Follistatin, a potent endogenous activin antagonist, is also produced in adult islets [23, 24]. These observations suggest that activins may play dynamic, tightly controlled autocrine and/or paracrine roles in adult islets. Here, using a novel image-based screening approach, activin A was found to dedifferentiate mature beta cells. Activin A decreased expression of the insulin gene and other mature beta cell genes, while increasing beta cell proliferation. These effects were fully reversed by follistatin, which augmented the mature beta cell phenotype. Our data point to a powerful local regulatory system within islets, which controls the maturity of adult beta cells.
Human islets were kindly provided by G. Warnock and the Ike Barber Human Islet Transplant Laboratory (Vancouver General Hospital, Vancouver, BC, Canada) and cultured as described . Mouse islets were isolated from 10- to 12-week-old C57BL/6J mice as described  and cultured overnight in RPMI 1640 with 10% (vol./vol.) FBS. MIN6 cells were cultured as described . Activin A and follistatin were purchased from R&D Systems (Minneapolis, MN, USA). All doses of activin A used in this study were at a saturating level (Electronic supplementary material [ESM] Fig. 1). Culture reagents were from Invitrogen (Burlington, ON, Canada), unless otherwise stated. Animal and human cell protocols were approved by the University of British Columbia, Canada in accordance with national guidelines.
Lentiviral vector production and infection
The dual reporter pTiger Pdx1 monomeric red fluorescent protein (mRFP)–Ins1 enhanced green fluorescent protein (eGFP) and control pTigerCMVeGFP and pTigerIns1eGFP lentiviral vectors were used to label MIN6 cells; details on vector construction, virus generation, infection protocols and expression validation have been described elsewhere . Briefly, MIN6 cells were seeded in six-well plates the day before infection. Lentiviral vectors were added at a multiplicity of infection of ∼1 in serum-free DMEM (with insulin–transferrin–selenium supplement and 0.2% (wt/vol.) BSA) and 8 μg/ml protamine sulphate. Plates were centrifuged for 2 h at 1,500 g and 30°C, then cultured overnight at 32°C. Medium was changed to complete DMEM and expression was monitored at least 72 h post-infection. After infection with lentivirus, cells have stable integration of the transgene(s), allowing long-term monitoring of reporter gene expression . Infection efficiency ranged from 40% to 80%. However, populations of infected MIN6 cells with a particular infection efficiency were used for an individual biological replicate (i.e. treated with activin A or non-treated control) and results were always normalised to the control within the same labelled population of cells, thereby controlling for differences in infection efficiency between preparations.
Screening and factorial design of experiments
JMP 7.0.2 software (SAS Institute, Cary, NC, USA) was used to design two-level (i.e. zero dose and factor added) fractional factorial experiments to screen the effects of factors on Pdx1 and Ins1 promoter activities. Initially, 12 factors were chosen at concentrations based on previous reports or preliminary single factor experiments (data not shown; ESM Table 1). We then chose eight factors for the second screen. The factorial design is presented in ESM Tables 2 and 3. The day before treatment, labelled MIN6 cells were seeded at 10,000 cells/well (ViewPlate-96; Perkin Elmer, Waltham, MA, USA) as a heterogeneous unsorted population of cells containing Pdx1+/Inslow immature cells, Pdx1+/Ins+ mature cells and cells that were not labelled (i.e. negative for both reporters). Cells were washed with basal medium (DMEM containing 5.5 mmol/l glucose, 0.2% (wt/vol.) BSA, 4 mmol/l glutamine, 100 U/ml penicillin and 172 μmol/l streptomycin) and factors were added to basal medium at concentrations and combinations described in ESM Tables 1, 2 and 3. After 48 h of culture, the nuclear stain Hoechst 33342 (0.32 µmol/l; Invitrogen) was added 30 min prior to automated imaging using a high-content screening instrument (ArrayScan VTI; Cellomics, Pittsburgh, PA, USA). Hoechst-positive, GFP-positive and RFP-positive cells were identified using fluorescence intensity cut-offs and then automatically counted (Target Activation Bioapplication; Cellomics). Cell count and intensity results for each factorial run were analysed by JMP 7.0.2 statistical software to identify the significant effects within each experiment. This analysis included analysis of multiple internal replicates for each factor in various combinations. Graphs are presented as per cent effect of each factor on the given read-out relative to no factors added (i.e. basal medium).
Flow cytometry and cell sorting
For FACS analysis and sorting, stably infected MIN6 cells were lifted off plates using trypsin–EDTA, resuspended in PBS containing 5% vol./vol. FBS and kept on ice. The influx sorter (BD Biosciences, San Jose, CA, USA) used was equipped with a tunable laser at 488 nm with filters 488LP and 531/40 for GFP, and a solid-state laser at 561 nm with filters 568LP and 624/40 for RFP. Pdx1+/Inslow and Pdx1+/Ins+ cells were simultaneously sorted into chilled medium before seeding and treatment with activin A. Analysis and sorting gates are shown in a sample FACS dot plot in ESM Fig. 2.
Quantitative real-time RT-PCR
Total RNA was purified using a kit (RNeasy; Qiagen, Mississauga, ON, USA) and used to prepare cDNA using SuperScript III First-Strand Synthesis SuperMix for quantitative RT-PCR (Invitrogen). Primers were designed to flank an intron and are listed in ESM Table 4. Quantitative RT-PCR was performed using SYBR GreenER qPCR SuperMix (Invitrogen) and 7500 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). For relative quantification of transcripts, cycle threshold values for each sample were normalised to Gapdh.
MIN6 cells were cultured for 48 h on 96-well microplates (ViewPlate-96; Perkin Elmer) with and without activin A in basal DMEM medium, after which 10 μmol/l BrdU (Kit I; Roche Applied Science, Laval, QC, Canada) was added for 30 min prior to fixation and staining. Cells were imaged using an inverted microscope (Axiovert 200 M) equipped with a FLUAR 20× objective (Carl Zeiss, Thornwood, NY, USA). Images were analysed and quantified using a software package (SlideBook; Intelligent Imaging Innovations, Boulder, CO, USA). A minimum of 500 cells per sample were imaged and quantified.
For static incubation, cells in 24-well plates were washed once and incubated for 1 h at 37°C in 3 mmol/l glucose Kreb’s buffer, which was then replaced for 2 h with 3 mmol/l glucose, 20 mmol/l glucose or 30 mmol/l KCl Kreb’s buffer at 37°C. Buffer samples were analysed for insulin. Secreted insulin was measured using RIA (Linco/Millipore, Billerica, MA, USA). Secreted activin A was quantified using an immunoassay (Human/Mouse/Rat Activin A; R&D Systems).
Data are presented as means ± SEM. Differences between means were evaluated by Student’s paired or unpaired t tests, as appropriate. A p value of 0.05 or less was considered significant. A minimum of three independent experiments was performed, as noted.
Screening for factors that modulate adult beta cell maturity
Interestingly, activin A had a strong negative effect on mature beta cells, yet a highly significant positive effect on the relative percentage of immature beta cells (Fig. 1). Upon removal of five factors in the second screen, this opposing effect of activin A was augmented twofold. In addition, negative and positive interactions between activin A and glucose or nicotinamide, respectively, became significant in the second screen. Similar factorial design results were observed using the INS-1 cell line labelled with the same Pdx1mRFP–Ins1eGFP dual reporter lentivirus (data not shown). Significant factor effects, as well as factor interactions, were uncovered using statistical design of experiments. For follow-up studies, we selected activin A as the most interesting candidate involved in modulating the maturation state of adult beta cells.
Activin A reduces beta cell maturity
Activin A decreases expression of insulin and mature beta cell genes
Activin A increases beta cell proliferation
Activin A decreases insulin secretion
Follistatin reverses the effects of activin A
Follistatin is a specific and potent antagonist of activin A, preventing binding to its receptor . Indeed, follistatin completely reversed the effects of activin A on Ins1 promoter activity, MIN6 maturation state, insulin gene expression and the expression of maturity-associated genes (Fig. 6). In some cases, follistatin not only reversed the effects of activin, but appeared to have significant effects of its own, suggesting the reversal of endogenous local activin signalling (Fig. 6b–e).
The present study sought to identify factors that modulate adult beta cell maturity. Using dual labelled MIN6 cells , candidate factors were systematically screened for their effects on Pdx1 and Ins1 promoter activities. The major finding of our study was that activin A and follistatin had powerful reciprocal effects on beta cell maturity. These results contribute to the understanding of maturation and plasticity in adult beta cells.
Factorial design offers insight into the complexity of interaction of various in vitro culture conditions, factors and nutrients, while substantially decreasing the number of experiments required . For example, our fractional factorial design for factorial 1 required 212 − 6 = 64, instead of 212 = 4,096 individual treatments, if all combinations of 12 factors were screened. Thus, the statistical design of experiments, which is used much more frequently in engineering science, can be a valuable tool in screening experiments designed to elucidate important interaction effects [6, 32]. The factorial design method revealed a number of interesting effects on adult beta cell maturity. For example, high glucose had highly significant positive effects on the relative numbers of immature and mature beta cells. Nicotinamide is generally used in cell culture media as a vitamin supplement and has been shown to promote differentiation of pancreatic fetal cells into insulin-positive cells . In our screens, nicotinamide also had positive effects. While synergistic or negative interaction effects can be masked in conventional screening experiments, our statistical design uncovered hidden interactions of factors including synergy between nicotinamide and high glucose, FBS or exendin-4. However, upon removal of many negative or insignificant first screen factors (such as FBS, IGF-1, retinoic acid and gastrin17) from the second screen, these synergies were lost, but new synergies between nicotinamide and activin A or insulin were uncovered. While previous studies have generally shown beneficial effects of exendin-4 on beta cell function and survival [33, 34], we observed a significant negative effect of exendin-4 on beta cell maturity.
Emerging evidence of pancreatic tissue plasticity suggests that normally quiescent, terminally differentiated pancreatic cells retain the potential to dedifferentiate, transdifferentiate or increase their proliferation after specific molecular cues in vitro and in vivo [3–5, 35]. Our findings support the notion that beta cell plasticity can be modified by external molecular cues. Activin A decreased Ins1 promoter activity, downregulated expression of several mature beta cell genes, increased proliferation and dampened insulin secretion in primary islets and MIN6 cells. Conversely, follistatin completely reversed the dedifferentiating effects of activin A. Both activin A and follistatin are produced in islets [23, 24], suggesting paracrine control of beta cell maturity. Interestingly, a preliminary experiment showed that levels of the activin βA subunit were increased in human islets treated with activin A (M. Szabat, unpublished data), supporting a positive feedback for autocrine or paracrine regulation of activin A expression.
The mature beta cell phenotype is commonly defined by the expression of genes such as the insulin gene, Pdx1, Glut2, Mafa, Neurod1, glucokinase and Nkx6-1 . In our activin A-treated cells, genes associated with beta cell maturity, specifically Ins1, Ins2, Pdx1, Glut2, Nkx6-1 and Mafa were downregulated. Follistatin increased the expression of these genes. Consistent with these negative effects of activin A, the TGFβ/SMAD pathway is known to restrict pancreatic progenitor specification, in part by restraining Pdx1 expression during early embryonic development . The notion that activin A is a tonic negative regulator of beta cell differentiation and function in the adult islet is supported by a recent study showing that TGFβ/SMAD3 signalling repressed insulin gene and other mature beta cell genes and also that downregulation of Smad3 improved beta cell function . Bone morphogenetic protein (BMP) signalling, also mediated via SMADs, prevented beta cell differentiation in zebrafish . Thus, both the TGFβ and BMP superfamilies appear to have suppressive roles in adult beta cell maturation.
Activin A increased expression of Mafb and decreased that of Mafa in MIN6 cells, whereas follistatin had the opposite effect. This pattern of Mafa and Mafb expression in activin A-treated cells is consistent with a dedifferentiated, immature beta cell phenotype, as it occurs during endocrine development. In murine pancreas development, Mafb is expressed before Mafa, followed by Mafb downregulation in adult beta cells . Mafb is required for beta cell maturation and directly regulates expression of Mafa, Pdx1, Nkx6-1, Glut2 and insulin . Activin A did not increase Mafb expression in primary mouse islet cells, as it did in MIN6, probably because islet alpha cells express high levels of Mafb  and a small increase in beta cell Mafb expression with activin A treatment might not have been detectable above control.
We previously found that neurogenin 3 (NEUROG3) message and protein are present in adult human islets, mouse islets and MIN6 cells, where they are regulated by notch signalling . Subsequently, the presence of NEUROG3 protein has been confirmed by others and its role in the maintenance of adult beta cell maturity was suggested . It is possible that activin and follistatin have reciprocal effects on NEUROG3 levels in beta cells, but it remains to be determined whether NEUROG3 mediates the effects of activin or follistatin on key beta cell genes such as Mafa and Glut2.
GLUT2 is required for glucose sensing, normal glucose homeostasis and insulin secretion . Glut2 was one of the most highly regulated candidate genes in activin A- or follistatin-treated islets and MIN6 cells. Similarly, Glut2 was significantly decreased in Pdx1+/Inslow immature beta cells . Consistent with our findings, Glut2 was elevated in Smad3 knockout islets . A strong reduction in Glut2 expression is an early indicator of beta cell stress and possibly dedifferentiation in many mouse models of glucose intolerance or diabetes , including mice with reduced Pdx1 . It will be interesting in the future to examine the characteristics and plasticity of the population of adult pancreatic cells with low Glut2 expression.
Our proliferation results with MIN6 cells are also consistent with another study that showed an approximately threefold increase in proliferation of primary rat beta cells treated with activin A . Collectively, activin A and follistatin appear to modulate the maturation state of adult beta cells, with activin A driving mature beta cells to a more progenitor-like phenotype with increased proliferation and decreased differentiated function.
Reports on activin A effects on insulin secretion are conflicting. Acute activin A treatment of cultured rat and human islets was reported to increase insulin secretion [21, 22]; however acutely treated MIN6 cells did not show increased secretion . In our study, activin A had no acute stimulatory effect on glucose-stimulated insulin secretion in perifused isolated mouse islets, while prolonged activin A treatment significantly decreased insulin secretion. It is possible that any acute activin A effects on insulin secretion may be mediated through a SMAD-independent, non-transcriptional signalling mechanism .
In addition to its effects on beta cell function, activin A has been reported to be a differentiation factor. In vitro, it appeared to direct pancreatic fetal cells into insulin-positive cells . Although activin A maintains pluripotency and self-renewal of embryonic stem cells , it is also required to induce stem cell differentiation into insulin-positive cells . Our results support a dedifferentiating role for activin A in adult beta cells. In this regard, activin A receptors, activin type I receptor and activin type II B receptor, were found to be expressed at a higher level in adult vs neonatal beta cells (S. Bonner-Weir, Harvard, Boston, MA, USA; personal communication), supporting a differential role for activin signalling at different stages of beta cell development.
In summary, our results demonstrate that local factors, such as activin and follistatin, control the maturation status of adult beta cells. Identification of modulators of beta cell replication and maturation will help in the development of therapies designed to increase functional beta cell mass in vivo, as well as helping to find alternative sources of transplantable beta cells in vitro.
We thank T. Kieffer and R. Kay for advice and reagents, C. Hoesli and D. Luciani for advice, L. Marmolejo and X. Hu for technical assistance, and G. Warnock for human islets. Research was supported by operating grants to J .D. Johnson and J. M. Piret from the Stem Cell Network, the Juvenile Diabetes Research Foundation (JDRF) and the Canadian Institutes of Health Research (CIHR). Infrastructure support of the Michael Smith Foundation for Health Research (MSFHR)-funded Centre for Human Islet Transplantation and Beta-Cell Regeneration is acknowledged. J. D. Johnson was supported by salary awards from MSFHR, CIHR, JDRF and the CDA. M. Szabat was supported by studentships from NSERC, MSFHR, CIHR and the University of British Columbia.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.