MafA and MafB Regulate Genes Critical to β-Cells in a Unique Temporal Manner

Pancreas-wide MafA deletion mutant mice were generated using the Cre-loxP mediated recombination system. A conditional MafA allele was generated using a targeting vector consisting of two loxP sites inserted into the 5′ (PstI site [−699 bp] and 3′ (KpnI site [1,658 bp]) end of the MafA exon (supplementary Fig. 1A in the online appendix available at http://diabetes.diabetesjournals.org/cgi/content/full/db10-0190/DC1). The herpes simplex virus-thymidine kinase (TK) gene was placed outside of the MafA gene homology region for neomycin^R selection. After electroporation of 129S6-derived mouse embryonic stem cells, 327 clones survived chemical selection. Fifteen clones were correctly targeted, as determined by Southern blotting hybridization. Two clones were independently injected in mouse blastocysts, and then chimeric mice were bred with C57BL/6J mice for germline transmission screening. The FRT-flanked neomycin^R gene was then removed in transmitted MafA^fl/+ mice by crossing with FLPe mice (JAX human β-actin FLPe deleter strain; more information online at http://www.mmrrc.org/strains/29994/029994.html). To delete MafA in the developing pancreas, MafA^fl/fl animals were bred with Pdx1^5.5-Cre mice (a gift from Dr. Guoqiang Gu, Vanderbilt University), which produce Cre recombinase by E10.5 in a pattern analogous to endogenous Pdx-1 early in development (23, supplementary reference 1). MafA^fl/fl;Pdx1^5.5-Cre mice were referred to as MafA^ΔPanc, with MafA effectively removed from the pancreas by E18.5 (Table 1, supplementary Fig. 1B). MafB^−/− mice were generated by homologous recombination (24). For embryonic analysis, noon of the day of the vaginal plug discovery was designated E0.5. The Vanderbilt University Institutional Animal Care and Use Committee approved all of our studies in mice.

E18.5 pancreas anlagen from five independently derived wild-type MafA^ΔPanc, MafB^−/−, and MafA^ΔPanc;MafB^−/− mice were dissected in PBS and stored in RNAlater (Ambion). Tissue RNA was isolated using the ToTALLY RNA isolation kit (Ambion). RNA was further purified over RNeasy columns (Qiagen), and RNA quality was analyzed using an Agilent 2100 Bioanalyzer. Each RNA sample (100 ng) was amplified using the Ovation Aminoallyl RNA Amplification and Labeling System (NuGEN Technologies), and PancChip 6.1 (http://www.betacell.org/ma, supplementary reference 2) microarray analysis was performed at the Functional Genomics Core at the University of Pennsylvania.

Quantitative RT-PCR was conducted on total RNA from E18.5 pancreata or 12-week-old islets (n ≥ 3 for each genotype). The RNA (1 μg) was transcribed using iScript reverse transcriptase and iScript reaction mix (Bio-Rad). The PCRs were performed using the SYBR Green (with dissociation curve) program on an Applied Biosystems 7900 machine. All reactions were performed in duplicate with reference dye normalization, and median cycling threshold values were used for analysis. Primer sequences are available on request.

Tissue fixation, embedding, and immunofluorescence labeling were performed as previously described (25). The primary antibodies used were guinea pig α-insulin (1:2,000; Linco Research), mouse α-insulin (1:2,000; Invitrogen), guinea pig α-glucagon (1:2,000; Linco Research), sheep α-somatostatin (1:2,000; American Research), guinea pig α-pancreatic polypeptide (1:2,000; Linco Research), rabbit α-MafB (1:10,000; Bethyl Laboratories), rabbit α-MafA (1:1,000, Bethyl Laboratories), rabbit α-Slc30a8 (1:1,000, Mellitech), mouse α-Rbp4 (1:1,000, Abnova), rabbit α-G6PC2 (1:500, a gift from Dr. John Hutton at the University of Colorado), rabbit α-Pdx1 (1:5,000, a gift from Dr. Chris Wright at Vanderbilt University), mouse α-Isl1 (1:300, Developmental Studies Hybridoma Bank), rabbit α-Nkx6.1 (1:1,000, Beta Cell Biology Consortium), and rabbit α-Pax6 (1:200, Covance Research Products). The secondary Cy2-, Cy3-, or Cy5-conjugated donkey α-rabbit, α-guinea pig, and α-sheep IgGs were obtained from Jackson ImmunoResearch Laboratories. Nuclear counterstaining was performed using YoPro1 or DAPI (Invitrogen). Immunofluorescence images were acquired using confocal microscopy (LSM510, Carl Zeiss).

In situ hybridization assays were performed using Neuronatin (base pair 119 to 1,163 relative to coding ATG) and Slc30a8 (base pair 166 to 1,269) gene probes according to Henrique et al. (26). The Rbp4 cDNA plasmid (clone ID IMAGp998P0711141Q1) was obtained from the resource center of the human genome project (www.rzpd.de) and used to generate the probe (base pair 295–927).

Immunolabeling for insulin and MafA (or MafB) was performed on serial 6-μm pancreatic sections that were collected at 60-μm (i.e., at E14.5), 84-μm (E18.5), or 150-μm (postnatal) intervals. All insulin⁺ cells were imaged by confocal microscopy and presented as the percentage of MafA⁺insulin⁺ or MafB⁺insulin⁺ to total insulin⁺ cells. At least 10 random fields were counted from wild-type and MafA^ΔPanc pancreata (n = 3) to calculate the percentage of MafB⁺insulin⁺ cells in adult islets. Immunolabeling for insulin, glucagon, somatostatin, and pancreatic polypeptide-producing cells was performed on 12-week islet sections at 300-μm intervals. At least 10 random fields were counted from wild-type and MafA^ΔPanc pancreata (n = 4) to calculate the proportion of somatostatin⁺ δ-cells within islets. Mean differences were tested for statistical significance using a two-tailed Student t test.

There are two distinct waves of insulin⁺ and glucagon⁺ cell production within the pancreatic epithelium during embryogenesis, with the massive expansion of secondary transition cells starting at E13.5 populating the adult islet (27). Previous studies established that MafB is expressed in both waves (18,21), whereas MafA is only in the insulin⁺ cells of the second wave (22). Quantitative immunohistochemical analysis was performed at E15.5, E18.5, P14, and P28 to precisely define the MafA and MafB expression pattern during embryonic and postnatal β-cell differentiation.

MafB was found in almost all insulin⁺ cells produced at E14.5 and E18.5, whereas the fraction of MafA⁺ insulin⁺ cells increased during this period (Fig. 1,A). Our results are in accordance with previous data showing that MafA was expressed in only 54% of the insulin⁺ cells at E15.5 and MafB in nearly 90% (21). However, MafB was present only in very few insulin⁺ cells soon after birth (P14, 2.70 ± 0.11%) and was essentially absent in insulin⁺ cells after 4 weeks (P28, 0.69 ± 0.06%). In contrast, MafA was detected in most islet β-cells by 4 weeks (Fig. 1 A). The postnatal switch in β-cells from principally MafB to exclusively MafA occurs during a period of cell maturation in regards to islet architecture, islet cell mass, and glucose-sensitive insulin secretion.

MafB mRNA expression at E18.5 was elevated 1.7 ± 0.1-fold in MafA^ΔPanc pancreata, yet no overt alterations in islet hormone⁺ cell development or Isl1, Nkx6.1, Pax6, or Pdx1 expression were found (supplementary Fig. 2). Also similar to MafA^−/− mice (19), changes in MafA^ΔPanc β-cell function and islet morphology were first observed postnatally (data not shown). Pan-endocrine Isl1 expression was unaffected in the MafA^ΔPanc, MafA^ΔPanc;MafB^−/−, and MafB^−/− mutants at E18.5 (Table 1), supporting evidence linking MafB to α- and β-cell maturation, but not islet cell formation or viability (18,21). As expected, insulin mRNA levels were compromised in MafB^−/− (18) and MafA^ΔPanc;MafB^−/− samples from E18.5, but not the MafA^ΔPanc (Table 1) or MafA^−/− mutants (19). The larger effect on Insulin mRNA expression in MafA^ΔPanc;MafB^−/− pancreata implies that MafA contribution to embryonic Insulin transcription becomes significant only in the absence of MafB. This was also indicated by the delay in residual insulin⁺ cell production until the beginning of the secondary transition at E13.5 in MafB^−/− mice (18).

Strikingly, immunohistochemical analyses revealed that MafB protein expression was retained in a fraction of insulin⁺ cells in adult MafA^ΔPanc mice (compare Fig. 1 B with C; 33.4 ± 5.6% of insulin⁺ cells express MafB). These results showed that the MafB protein produced in adult MafA^ΔPanc β-cells does not compensate for the loss of MafA.

Gene expression profiling studies of E18.5 wild-type, MafA^ΔPanc, MafB^−/−, and MafA^ΔPanc;MafB^−/− pancreata were performed to more broadly define the regulatory roles of MafB and MafA in β-cell differentiation. Thirty-five genes with mRNA levels altered by ≥1.5-fold (P ≤ 0.01, false discovery rate ≤25%) were identified using the PancChip 6 microarray platform in the MafB^−/− mutant, and 47 were identified in MafA^ΔPanc;MafB^−/− (supplementary Table 1). The profiles were similar between MafB^−/− and MafA^ΔPanc;MafB^−/− (29 genes were differentially expressed in both genotypes), with most changes more pronounced in MafA^ΔPanc;MafB^−/− (supplementary Table 1). In contrast, no differences were found between MafA^ΔPanc and the wild-type gene expression profile.

Gene ontology analysis revealed that the differentially expressed genes were mainly involved in aspects of mature β-cell function, such as ion binding and transport, signal transduction, and hormone secretion (supplementary Table 2). The mRNA changes of many candidate genes (like Slc30a8, G6pc2, Rbp4, Nnat, Insulin, and Slc2a2) were independently verified in E18.5 MafB^−/− and MafA^ΔPanc;MafB^−/− pancreata by quantitative RT-PCR (Table 1; data not shown). The studies described below examined the significance of MafA in adult expression of genes initially regulated by MafB. Expression within specific islet cell types was examined by in situ hybridization during development and by immunohistochemistry in adults; unfortunately, limitations in the technique (e.g., in situ) or protein abundance precluded examination of these temporal periods using both methods. Importantly, our results provide mechanistic insight into the mutually beneficial relationship between MafB and MafA and provide an explanation for why β-cell activity decreases in MafA^−/− and MafA^ΔPanc mice.

Slc30a8 or ZnT8 encodes an islet-specific zinc transporter that facilitates the accumulation of zinc from the cytoplasm into intracellular vesicles for insulin maturation and storage (28,29). Overexpression of Slc30a8 enhances glucose-stimulated insulin secretion in INS1 β-cells (29), whereas variants in this locus are linked to type 2 diabetes in humans (30,31). The transporter is also a type 1 diabetes autoantigen in humans (32).

The normal Slc30a8 expression pattern during pancreatic development was first determined by in situ hybridization. Slc30a8 mRNA transcripts were found in both developing β- and α-cells (Fig. 2,A–D, supplementary Fig. 3). Notably, Slc30a8 expression was lost in α- cells at E18.5 in the MafB^−/− mutant, but retained in the remaining insulin⁺ cells (Fig. 2,E and F, supplementary Fig. 3). Significantly, Slc30a8 expression was undetectable in the MafA^ΔPanc;MafB^−/− mutant (Fig. 2 G and H, supplementary Fig. 3), demonstrating that expression was mediated by MafA in the residual insulin⁺ cells in the MafB mutants. These results suggest that MafB is the primary activator of Slc30a8 expression in α- and β-cells during development.

Next, immunohistochemical staining was performed to assess Slc30a8 expression in adult wild-type and MafA^ΔPanc islets. As reported while this work was in progress (33), Slc30a8 is present in both α- and β-cells (Fig. 2,J and K). (The specificity of the Slc30a8 antibody was confirmed upon comparing pancreatic samples from wild-type and Slc30a8^−/− mice [supplementary Fig. 4]). Expression was essentially absent in 3-month-old islet insulin⁺ cells from the MafA^ΔPanc mutant (Fig. 2,L). The remaining Slc30a8 in MafA^ΔPanc islets is primarily (if not exclusively) in α-cells (Fig. 2 M). These data imply that MafA and MafB function at distinct temporal stages to promote Slc30a8 expression in β-cells.

Glucose-6-phosphatase catalytic subunit-2 protein (G6pc2, also known as IGRP) is a major autoantigen in type 1 diabetes (34) and may regulate fasting plasma glucose levels in humans (35). Expression of G6pc2 mRNA was significantly compromised in E18.5 pancreata from MafB^−/−, MafA^ΔPanc;MafB^−/−, and MafA^ΔPanc embryos. However, the impact of MafA appears greater in adult MafA^ΔPanc islets than during development (see Fig. 3,A and Table 1), as a much lower level of G6pc2 was present in 3-month-old MafA^ΔPanc islet β-cells than wild-type (Fig. 3, compare D with G). These data indicate that both MafA and MafB play a role in G6pc2 expression and complement recent cell-line–based studies showing that MafA stimulates G6pc2 transcription by binding to the −177/−164 bp element in its promoter region (36).

Neuronatin (Nnat) is involved in modulating ion channel activity in β-cells, with overexpression increasing insulin secretion by elevating intracellular Ca²⁺ levels (37,38). In situ hybridization analysis showed that Nnat mRNA was expressed at E15.5 and E18.5 in insulin⁺ cells and in a population that was neither insulin⁺ nor glucagon⁺ (Fig. 4, see WT panel, supplementary Figs. 5 and 6). Nnat synthesis was predominately in insulin⁺ cells by E18.5. A profound reduction in steady-state Nnat levels was observed in E18.5 MafB^−/− and MafA^ΔPanc;MafB^−/− pancreata (Table 1), with the enduring Nnat detected only in the remaining mutant insulin⁺ cells (Fig. 4, supplementary Figs. 5 and 6).

These results suggest that Nnat expression in developing β-cells is only partially dependent on MafB. This was also true of Pdx1 transcription wherein the ∼50% reduction in E18.5 mRNA levels reflected the loss in the MafB^−/− insulin⁻ cell population (18). Notably, steady-state Nnat (Fig. 4,M) and Pdx1 (19) levels were unchanged in MafA^ΔPanc adult islets, a distinction from Slc30a8 (Fig. 2,I) or G6pc2 (Fig. 3 A). These data demonstrate that only MafB regulates Nnat expression.

Rbp4 contributes to the pathogenesis of type 2 diabetes, as its release from adipocytes causes systemic insulin resistance in both mice and humans (39). This circulating protein is also produced in the liver, although little was known about its effect on islets until recently. Rbp4 has now been found to repress glucose-stimulated insulin secretion in both isolated rat islets and in vivo (P. Fueger, C. Newgard, unpublished data).

In distinction to the genes described previously, Rbp4 mRNA expression was increased in MafB^−/− and MafA^ΔPanc;MafB^−/− pancreata at E18.5 (Table 1). In situ RNA analysis was performed to determine the distribution of Rbp4 in hormone⁺ cells during development. Rbp4 transcripts were found in the majority of wild-type insulin⁺ and glucagon⁺ cells at E15.5, but only in a small portion of insulin⁺, glucagon⁺, and somatostatin⁺ (δ) cells by E18.5 (Fig. 5, WT panel, supplementary Figs. 7 and 8). In contrast, most of the remaining hormone⁺ cell types, as well as hormone⁻ islet cells, synthesize Rbp4 in the MafB^−/− and MafA^ΔPanc;MafB^−/− mutants (Fig. 5 C–F, J–O, supplementary Figs. 7 and 8).

Quantitative RNA analysis demonstrated that adult MafA^ΔPanc islets contained elevated Rbp4 mRNA levels (Fig. 5,P). Notably, Rbp4 protein expression was detected only in somatostatin-producing δ-cells in both wild-type and MafA^ΔPanc islets (Fig. 5 Q–X). However, there was no change from the wild-type in the somatostatin⁺ δ-cell proportion produced in mutant islets (wild-type: 8.94 ± 0.73% vs. MafA^ΔPanc: 8.59 ± 0.89%). Collectively, the data indicate that MafA and MafB negatively regulate production of this effector of insulin secretion and insulin resistance in developing and mature islet cell types.

The expression pattern and functional characteristics of MafA and MafB is unusual when compared with all other islet transcriptional regulators. Hence, the MafA^−/− mutant affects only adult islet architecture and β-cell activity, whereas MafB and essentially all other pancreatic transcriptional regulators (at least) affect islet cell development. Here we have examined the relationship between these closely related factors in islet β-cell formation and function, an especially important issue considering that MafB is not produced within this islet cell population in adults.

Histologic and gene expression analyses of regulated genes identified from a microarray screen performed on E18.5 MafB^−/−, MafA^ΔPanc;MafB^−/−, and MafA^ΔPanc pancreata clearly illustrated that MafB is a more potent regulator of β-cell development than MafA, with the majority of differentially expressed genes identical in MafB^−/− and MafA^ΔPanc;MafB^−/− samples. Our data also indicate that MafB may serve in a compensatory capacity in the embryonic MafA^ΔPanc mutant, since MafB levels were amplified at E18.5. Notably, persistent MafB expression in a fraction of adult MafA^ΔPanc β-cells did not prevent the change in islet architecture and β-cell function first described in MafA^−/− mice (19). MafB may not compensate in this setting because of limiting protein levels, or more interestingly, because of differences in activation properties with MafA. In this regard, it is noteworthy that MafA was found to activate Insulin expression, whereas MafB specifically stimulated glucagon in experiments performed in endocrine cells (20,22,40).

Expression profiling analysis showed that MafB activates genes involved in mature endocrine function, including those significant to glucose sensing (e.g., Slc2a2), vesicle maturation (Slc30a8), Ca²⁺ signaling (Camk2b), and insulin secretion (Nnat) (Table 1, supplementary Table 2; data not shown). In contrast, the levels of transcription factors associated with β-cell differentiation (e.g., Pax6, NeuroD1, Pax4, and Isl1) were unaffected in the MafB^−/− (or MafA^ΔPanc;MafB^−/−) mutant, supporting a role for MafB in late events essential to cell maturation and not early islet cell specification (e.g., Ngn3) or cell lineage commitment steps (e.g., Arx, Isl1, Nkx2.2, Pax4, and Pax6) (18). This proposal is also supported by the fact that MafB^−/− mutant retained the expression of some α- and β-cell–enriched developmental markers while undergoing a loss in insulin and glucagon production (18). Interestingly, profiling E18.5 MafB^−/− or MafA^ΔPanc;MafB^−/− pancreata did not pick up the Gck (i.e., glucokinase), Pcsk1, or Glp1R products that were previously associated with MafA control in a candidate gene study performed with INS1 β-cells (41). These genes may be preferentially under later stage MafA control, since, for example, Gck is produced in high levels postnatally only in β-cells (42).

Strikingly, our results demonstrate that islet MafA controls many genes first regulated by MafB in developing β- cells. This conclusion is based on immunohistological and mRNA analysis of selected MafB^−/− microarray candidate genes within the embryonic MafB^−/− pancreas and adult MafA^ΔPanc islet. For example, Slc30a8 expression was compromised in MafB mutant glucagon⁺ and insulin⁺ cells during development and specifically in adult MafA mutant insulin⁺ cells (Fig. 2). Instances were also found when MafB was the primary (if not exclusive) activator, such as Nnat (Fig. 3) and Pdx1 (data not shown).

Given the significance of MafA and MafB mutual target genes (e.g., Slc30a8, Insulin, and Slc2a2) to β-cell function, insulin secretion defects associated with the MafA^−/− (19) and MafA^ΔPanc mutants would be expected. The recent studies showing that Rbp4 produced from rat islet cells inhibits glucose-stimulated insulin secretion (P. Fueger, C. Newgard, unpublished data) suggests that activation in δ-cells of MafA^ΔPanc mouse would also act to diminish β-cell function. (Rbp4 is also principally expressed in human islet δ-cells [supplementary Fig. 9]). Rbp4 attenuates insulin-induced phosphorylation of insulin receptor substrate-1 in human adipocytes (supplementary reference 3). Presumably, it also inhibits insulin-responsive signalling that is essential to β-cell function (supplementary reference 4). We believe that the increased Rbp4 expression in δ-cells is a byproduct of MafA^ΔPanc β-cell inactivity. Further investigation may reveal that islet Rbp4 expression is also induced in mouse models of type 2 diabetes wherein islet MafA expression was selectively lost (43,44).

Although other closely related transcription factors are synthesized within the pancreas (e.g., Pax4/Pax6, Nkx6.1/Nkx6.2, FoxA1/FoxA2), none act as specifically and in such a coordinated manner to affect islet cell function as MafA and MafB. For instance, Nkx6.1 and Nkx6.2 are expressed more broadly and earlier, with only Nkx6.1 significantly influencing β-cell development and function (10,12,45). Likewise, FoxA1 and FoxA2 have a similar early expression pattern to Nkx6.1/6.2 in the forming pancreas, with the dominant and general role in islet cell formation and function served by FoxA2 (7,–9). Perhaps parallels can be made between islet MafA and MafB and the basic helix-loop-helix Myf-5, MyoD, and Myogenin factors through their actions in skeletal muscle formation. Hence, gene knockout experiments in mice have shown that myogenin plays an essential role in the late differentiation of myoblasts to myotubes (46) (analogous to MafA in the maturing β-cell), whereas both Myf-5 and MyoD act upstream and are critical to myoblast formation (47) (acting similarly to MafB). However, expression of any of these myogenic factors can convert a variety of tissue culture cells into muscle cells (48,49), whereas only MafA appears capable of activating Insulin transcription in vitro (22,40). Interestingly, differences in myogenic potential was revealed when Myogenin was expressed from the Myf5 locus in vivo (48), suggesting that a knock-in analysis might also reveal elementary regulatory differences between the MafA and MafB proteins.

This work was supported by grants from the National Institutes of Health (P01 DK-42502 to R.S.; DK-72433 to M.A.M.), Juvenile Diabetes Research Foundation (research grant 1-2008-595 to R.S.; 2-2007-716 to I.A.), and the Swedish Research Council (to I.A.). Support for this work was provided to the Vanderbilt Transgenic Mouse/ES Cell Shared Resource by the Vanderbilt University Diabetes Research and Training Center (P60 DK-20593).

No potential conflicts of interest relevant to this article were reported.

I.A. and Y.H. researched data, contributed to discussion, wrote the manuscript, and reviewed/edited the manuscript. M.M., M.G., and J.L. researched data. T.Y. researched data and contributed to discussion. M.A.M. contributed to discussion and reviewed/edited the manuscript. R.S. contributed to discussion, wrote the manuscript, and reviewed/edited the manuscript.