The MAFB transcription factor impacts islet α-cell function in rodents and represents a unique signature of primate islet β-cells

Abstract

Analysis of MafB^−/− mice has suggested that the MAFB transcription factor was essential to islet α- and β-cell formation during development, although the postnatal physiological impact could not be studied here because these mutants died due to problems in neural development. Pancreas-wide mutant mice were generated to compare the postnatal significance of MafB (MafB^Δpanc) and MafA/B (MafAB^Δpanc) with deficiencies associated with the related β-cell-enriched MafA mutant (MafA^Δpanc). Insulin⁺ cell production and β-cell activity were merely delayed in MafB^Δpanc islets until MafA was comprehensively expressed in this cell population. We propose that MafA compensates for the absence of MafB in MafB^Δpanc mice, which is supported by the death of MafAB^Δpanc mice soon after birth from hyperglycemia. However, glucose-induced glucagon secretion was compromised in adult MafB^Δpanc islet α-cells. Based upon these results, we conclude that MafB is only essential to islet α-cell activity and not β-cell. Interestingly, a notable difference between mice and humans is that MAFB is coexpressed with MAFA in adult human islet β-cells. Here, we show that nonhuman primate (NHP) islet α- and β-cells also produce MAFB, implying that MAFB represents a unique signature and likely important regulator of the primate islet β-cell.

pancreatic α- and β-cells are major hormone-producing cell types of the islets of Langerhans. The hormones secreted by these cells act in a counterregulatory manner to control the glucose concentration in the bloodstream, with α-cell-produced glucagon raising and β-cell-produced insulin lowering levels. Loss or dysfunction of islet β-cells leads to insulin insufficiency and hyperglycemia in type 1 (T1DM) and type 2 diabetes mellitus (T2DM) (44). The healthcare costs associated with diabetes care are enormous and rising, and consequently, efforts are focused on developing therapies to minimize these complications, including developing β-like cells from human embryonic stem cells or induced-pluripotent stem cells (26, 33, 35).

Identifying the key factors that mediate islet α- and β-cell formation is essential to understanding T1DM and T2DM disease processes and for improving cell-based interventions. Notably, many transcription factors serve as critical regulators of islet cell development and function (32), with alterations in their activity contributing to α- and β-cell dysfunction in T2DM (19). For example, pancreatic duodenal homeobox-1 (Pdx1) is expressed in all pancreatic progenitors before being restricted to islet β-cells. Mice (2, 25, 30) and humans (42) completely lacking this transcription factor fail to develop a pancreas, whereas cell identity is severely compromised upon conditional deletion from adult mouse islet β-cells (2). Ngn3, which is produced later during embryogenesis, is essential for the differentiation of endocrine progenitor cells into each of the hormone⁺ cell subtypes [i.e., β- and α-cells, δ-cells (which secrete somatostatin), pancreatic polypeptide cells (which secrete pancreatic polypeptide), and ε-cells (which secrete ghrelin)] (16, 17). MafA and MafB are unusual in not being expressed until the onset of hormone⁺ cell formation (21), with MafA exclusively in insulin⁺ cells (27, 29) and MafB also in glucagon⁺ cells (6).

MAFA, MAFB, and PDX1 levels were specifically decreased in human T2DM islet α- and β-cells, and these factors have been proposed to be major contributors to islet cell dysfunction (19). Interestingly, the combination of Pdx1, Ngn3, and MafA was sufficient to convert nonendocrine cells to β-like cells in mice (12, 47), whereas adenovirus-driven expression of MafA was able to induce glucose-stimulated insulin secretion (GSIS) in neonatal rat islets (1). Collectively, these data not only illustrate the stage-specific activities of key islet transcription factors but also provide support for MafA and MafB contributing to events involved in β-cell formation.

MafB^−/− pancreata exhibited an ∼50% loss of insulin⁺ and glucagon⁺ cells by embryonic day (E)15.5, whereas expression of key β-cell-enriched transcription factors was reduced by E18.5 (e.g., Pdx1, Nkx6.1). Islet β-cell-specific MafA production correlated with insulin⁺ cell formation in the MafB^−/− animals, and there was no reduction in total endocrine cell number (3). Although these results strongly suggested that MafB was a key islet cell regulator in vivo, this could not be assessed postnatally since MafB^−/− mice die at birth due to renal failure and central apnea (7, 37). In contrast to MafB^−/− mice, there was no impact on endocrine cell formation in null [MafA^−/− (46)], pancreas-specific [MafA^Δpanc (22)], or β-cell-specific [MafA^Δβ (41)] MafA-knockout mice. However, these MafA mutants postnatally develop glucose intolerance, changes in islet cell architecture, reduced β-cell mass, and compromised β-cell gene expression (5, 22, 41, 46). Because MafB is not produced in rodent β-cells after birth (6), these results indicated that MafB was necessary for β-cell development, and MafA was essential to adult islet β-cell function.

In contrast to mouse β-cells, MAFB is coexpressed with MAFA in adult human islet β-cells (14). However, the expression and distribution of other islet-enriched transcription factors (i.e., PDX1, NKX6.1) are similar between rodents and humans (14). Because islet-enriched transcription factors profoundly influence mouse islet cell function and identity, the distinctive human MAFB expression pattern could be contributing to dissimilarities in islet cell characteristics between humans and mice (14). This hypothesis derives from the biochemical and functional differences reported for the MafA and MafB dimeric activators (20). For example, misexpression of MafA in a mouse islet α-cell line activated insulin gene expression (4), whereas MafB induced glucagon in a mouse β-cell line (6). Similarly, only MafA stimulated insulin production in chick in ovo electroporation assays, although exchanging the MafB COOH-terminal DNA-binding dimerization (basic leucine zipper) spanning region with that of MafA enabled insulin induction (4).

Here, we produced pancreas-wide deletion mutants of MafB (MafB^Δpanc) and MafA/B (MafAB^Δpanc) to compare the postnatal contributions of these transcription factors to islet cell formation and function. Elevated blood glucose levels were observed just after birth in MafB^Δpanc mice, which soon resolved upon comprehensive production of MafA within the insulin⁺ cell population. In addition, there was a profound reduction in glucagon secretion levels from adult islet α-cells. In contrast, MafAB^Δpanc mice died from hyperglycemia shortly after birth due to loss of insulin⁺ cells. These results demonstrated that MafB primarily affects mouse islet α-cell function and not β-cell activity. Notably, MAFB was coexpressed with MAFA in nonhuman primate (NHP) islet β-cells, suggesting that this factor imparts distinct control properties in primates. The significance of MAFB to primate β-cells is supported by the recent observation showing that knockdown of this transcription factor suppressed GSIS in the human EndoC-βH1 β-cell line (41).

MATERIALS AND METHODS

Generation of the pancreas-specific and endocrine-specific knockout mice.

Pancreas-wide deletion mutants of MafA and MafB were generated by crossing MafA^fl/fl (5) and MafB^fl/fl (45) mice with Pdx1^5.5-Cre mice (17), which produces Cre in pancreatic progenitor cells prior to MafA and MafB expression. MafA^fl/fl;MafB^fl/fl;Pdx1^5.5-Cre and MafB^fl/fl;Pdx1^5.5-Cre were referred to as MafAB^Δpanc and MafB^Δpanc mice, respectively. MafAB^fl/fl and MafB^fl/fl mice were used as controls. Pancreatic endocrine cell-specific MafA and MafB deletion mutant mice were generated with neurogenin 3 (Ngn3)-Cre mice (39), referred to as MafAB^Δendo. Chi² analysis of expected vs. actual genotypes at time of genotyping was performed on MafAB^Δpanc and MafAB^Δendo litters. For embryonic samples, day 0.5 was counted as the day the vaginal plug was observed. All studies with mice were in compliance with protocols approved by the Vanderbilt Institutional Animal Care and Use Committee.

Islet isolation conditions.

Two-week-old mouse pancreata were partially digested with 1 mg/ml collagenase, followed by handpicking of islets; adult islets were collected as described previously (22). NHP (rhesus macaque) pancreata were obtained from five females and five males (average age 8.8 ± 1.3 yr, range 0.32–13 yr) as excess material under unrelated protocols approved by the Oregon National Primate Research Center Institutional Animal Care and Use Committee. Islet isolation was initiated within 10–15 min of exsanguination by cannulation and perfusion through the pancreatic duct with collagenase/neutral protease solution in a Ricordi apparatus. When islet release was evident by dithizone staining, islets in digestion solution were concentrated and washed by centrifugation and purified using a COBE 2991 cell processor. Human islets were provided by the Integrated Islet Distribution Network [http://iidp.coh.org/; 30 total preparations, 11 female and 19 male donors, age 38.4 ± 2.4 yr (range: 17–60), BMI 25.99 ± 0.55 kg/m² (range: 18.8–29.7)]. Cause of death was head trauma (n = 11), neurological events (stroke, subarachnoid hemorrhage, etc.; n = 10), anoxia (n = 2), or unknown (n = 7). The cold ischemia time before pancreas isolation was 9.9 ± 1.1 h (range: 1.5–24.9 h). Human and NHP islets were handpicked on the day of arrival as described (14). All studies with human and NHP islets were in compliance with the Vanderbilt Institutional Animal Care and Review Board Committee.

RNA analysis.

Quantitative real-time PCR was performed on RNA isolated from NHP, human, and mouse (e.g., C57BL/6J) islets (14) as well as MafB^Δpanc and MafB^fl/fl islets (22), using previously described conditions. Preloaded arrays (Applied Biosystems) of 16 genes were used to determine expression levels in NHP, human, and mouse islets in Fig. 7, with the 18S/18s, ACTB/ActB, TFRC/Tfrc, and TBP/Tbp genes used for normalization. Primer sequences are available upon request. Primer efficiency was similar between species as guaranteed by the manufacturer and verified empirically. For assessing the data, the Minimum Information for Publication of Quantitative Real-Time PCR Experiments was followed (10).

Fig. 7.

MAFB is expressed with MAFA in adult nonhuman primate (NHP) islet β-cells. A: relative gene expression of various transcription factors in NHP (n = 5), human (n = 10), and mouse (C57BL/6J) (n = 6) islets. *P < 0.05 and ***P < 0.001, NHP and mouse vs. human; ++P < 0.01 and +++P < 0.001, mouse vs. NHP. The islet-enriched gene mRNA signal was normalized to the 18S, ACTB, TFRC, and TBP signal. B: representative images of a NHP islets; MAFB (red), MAFA (red), Ins (green), and Gluc (blue). White arrows denote MAFB⁺INS⁺, blue arrows MAFB⁺GCG⁺, and yellow arrows MAFA⁺INS⁺ cells. Scale bar, 10 μm. C: illustration of functional domains in human MAFA and MAFB. Protein alignment and %identity of human MAFA and MAFB was determined using the EMBOSS Matcher tool. Amino acids of the transactivation domain are aligned below with residues that are common between the 2 proteins highlighted in gray. Amino acids 76–108 of human MAFA are not present in human MAFB. D: the endocrine hormone mRNA expression pattern is analogous between human and NHP islets, whereas mouse is different. NHP (n = 5), human (n = 10), and mouse (C57BL/6J) (n = 6) islets. *P < 0.05, **P < 0.01, and ***P < 0.001, mouse vs. human; +P < 0.05 and ++P < 0.01, mouse vs. NHP. No significant difference seen between NHP and human (P > 0.05). E: representative image of a NHP and mouse (C57BL/6J) islet; Ins (green), Gluc (red), and somatostatin (SOM, blue). Scale bars, 100 μm.

Tissue collection and immunohistochemistry.

The paraformaldehyde-embedding conditions for NHP, human, and mouse islets (14) as well as mouse pancreata (22) have been described. Primary antibodies utilized for mouse sections were as follows: insulin (guinea pig; Dako), glucagon (mouse; Sigma), glucagon (rabbit; Linco), MafB (rabbit; Bethyl), MafA (rabbit; Novus), Pax6 (rabbit; Covance), Pdx1 (goat; provided by Chris Wright, Vanderbilt), Nkx6.1 (rabbit; β-Cell Biology Consortium), Slc30A8 (rabbit; Pierce), Glut2 (goat; Santa Cruz Biotechnology), ghrelin (goat; Santa Cruz Biotechnology), Ki-67, and (mouse; BD Pharmingen). Primary antibodies used for NHP and human sections were as follows: insulin (guinea pig; Dako or Linco), glucagon (mouse; Sigma), glucagon (rabbit; Cell Signaling Technology), MafB (rabbit; Novus or Bethyl), MafA (rabbit; Novus), and somatostatin (goat; Santa Cruz Biotechnology). Briefly, the primary antibody-antigen complex was visualized on 6-μm sections by immunofluorescence using secondary antibodies conjugated with Cy2, Cy3, or Cy5 fluorophores (1:500; Jackson ImmunoResearch, West Grove, PA). Nuclear costaining was conducted with DAPI Fluoromount G (Southern Biotech). Immunofluorescent images were acquired with a Zeiss LSM510 confocal microscope or a Zeiss Axioimager M2 microscope.

Islet cell population analysis.

Mouse pancreatic sections at 133-, 288-, and 354-μm spacing were prepared from E15.5, postnatal day 1 (P1), and 2-wk-old samples, respectively. Insulin⁺, glucagon⁺, Pax6⁺, Pdx1⁺, and Nkx6.1⁺ cell images were counted manually and divided by the total number of pancreatic DAPI⁺ nuclei. At least 10,000, 20,000, and 100,000 pancreatic nuclei were counted for E15.5, P1, and 2-wk-old samples, respectively. The percentage of MafA⁺insulin⁺ and Ki-67⁺insulin⁺ cells was determined by dividing the number of copositive cells by the total count of insulin⁺ cells for MafA at P1 and 2 wk and for Ki-67 at P1. The percentage of Pdx1⁺insulin⁻ cells was determined by dividing the number of Pdx1⁺insulin⁻ cells by total Pdx1⁺ cells at P1 and 2 wk. Cell counting was performed blinded to prevent bias.

Intraperitoneal glucose tolerance test.

Three-, six-, and eight-wk-old mice (n ≥ 7) were fasted for 6 h, and blood glucose level from tail blood was determined using a FreeStyle glucometer (Abbott Diabetes Care). The mice were then weighed, and 2 mg dextrose/g body wt (Fisher Biotech) in sterile PBS was injected intraperitoneally. Blood glucose levels were measured at 0, 15, 30, 60, 90, and 120 min postinjection. Fed blood glucose levels were determined prior to fasting.

Stimulated hormone secretion.

Handpicked islets from 2- and 8-wk-old MafB^Δpanc and control MafB^fl/fl mice were cultured overnight in RPMI 1640, 10% FBS, and penicillin-streptomycin supplemented with 5.6 (for insulin secretion) or 11 mM d-glucose (for glucagon secretion). The next day, islets were subjected to hormone secretion conditions in KRBH buffer (1.25 mM CaCl₂, 0.6 mM MgS0₄, 0.6 mM KH₂PO₄, 2.4 mM KCl, 64.0 mM NaCl, 20 mM HEPES, pH 7.9, and 5 mM NaHCO₃) with either 2.8 or 16.7 mM glucose for insulin secretion and 1 mM glucose or 1 mM glucose plus 10 mM arginine for glucagon secretion. Following 45 min of stimulation at 37°C, secretion media were collected and islets lysed in 1.5% HCl and 70% ethanol. Secreted (media) and cell content hormone levels were determined by ELISA (insulin, ALPCO; glucagon, RayBiotech). Hormone secretion was expressed as the percentage of total hormone content. The NHP, human, and mouse perifusion studies were conducted in the Vanderbilt Islet Isolation and Analysis Core under standard conditions (14), using as baseline 5.6 mM glucose and 16.7 mM glucose alone or with 500 μM 3-isobutyl-1-methylxanthine (IBMX) for stimulating insulin secretion.

Statistics.

Statistical significance was determined by one-way ANOVA between NHP, human, and mouse samples. The data with MafAB^Δpanc and MafB^Δpanc mice were presented as means ± SE, and statistical significance assessed by comparing data points between mutants and controls using the Student two-tailed unpaired t-test.

RESULTS

Loss of MafB alone results in delayed β-cell maturation.

The role of MafB in postnatal α- and β-cells remains unclear, since MafB^−/− mice die at birth (37). As a result, floxed MafB mice were crossed with early pancreas-driven Pdx1-Cre (MafB^Δpanc) transgenic mice to determine the specific influence on postnatal islet cells. MafB was effectively removed from MafB^Δpanc α- and β-cells by E15.5 (i.e., 3.65 ± 2.15% remains MafB⁺insulin⁺ and 6.62 ± 1.63% remains MafB⁺glucagon⁺), and there was no effect on mortality. However, MafB^Δpanc mice were mildly hyperglycemic at P1, a situation that resolved by 2 wk (Fig. 1A). Hormone⁺ and pan-endocrine Pax6⁺ cell numbers were quantified at E15.5 and 2 wk to determine whether insulin⁺ and/or glucagon⁺ cell production was impacted in MafB^Δpanc mice. Although there was no change in the Pax6 population that marks all islet cells, there was a significant decrease in the insulin⁺ (∼75%) and glucagon⁺ (∼50%) cell numbers at E15.5 (Fig. 1B). These results were expected from the analysis of the MafB^−/− mutant (3).

Fig. 1.

Postnatal β-cell maturation is delayed in MafB^Δpanc mice. A: postnatal day 1 (P1) and 2-wk-old random blood glucose measurements in MafB^Δpanc and control MafB^fl/fl littermates (n = 12–14). B: reduction in embryonic day (E)15.5 MafB^Δpanc insulin⁺ and glucagon⁺ cell numbers is ameliorated within 2 wk after birth. Approximately 10,000 and 100,000 pancreatic nuclei were counted at E15.5 and 2 wk, respectively (n = 3). C: MafB^Δpanc animals exhibit normal glucose tolerance by 3 wk. Glucose tolerance tests were performed at 3, 6, and 8 wk (n = 11–21). Statistical significance between MafB^fl/fl and MafB^Δpanc was analyzed by Student's t-test. *P < 0.05 and **P < 0.01 compared with MafB^fl/fl.

Notably, MafB^Δpanc insulin⁺ and glucagon⁺ cell numbers were indistinguishable from wild-type littermate controls by 2 wk (Fig. 1B). However, the islet β-cell-enriched Pdx1 transcription factor was aberrantly detected in hormone-negative cells in earlier P1 MafB^Δpanc samples, which was resolved by 2 wk (Fig. 2, A and B). Since MafB is normally produced in embryonic insulin⁺ cells prior to MafA (6, 27), we examined whether MafA or another related large Maf transcription factor, c-Maf (28, 29), could be acting in a compensatory manner in MafB^Δpanc mice. MafA was detected only in MafB^Δpanc insulin⁺ cells (Fig. 2, A and C), whereas c-Maf was undetectable in both mutant and control samples (data not shown). Moreover, the inability to detect a change in the Ki-67⁺ cell numbers between MafB^Δpanc and MafB^fl/fl insulin⁺ cells at P1 suggests that enhanced proliferation did not lead to recovery of the β-cell population (Fig. 2D). Considering the importance of MafA to mouse β-cell function after birth (22, 46), we conclude that MafA compensates for MafB in MafB^Δpanc mice, with the later expression of MafA influencing the delayed production of insulin⁺ cell numbers and β-cell function. In support of this conclusion, the expression levels of two other known MafA- and MafB-activated gene targets, Slc30a8 and Glut2 (5, 22), were also only transiently compromised in neonatal MafB^Δpanc islets (Fig. 2E).

Fig. 2.

Pancreatic duodenal homeobox-1 (Pdx1) is not produced in all newly produced insulin⁺ cells, and Glut2 and Slc30a8 expression is also delayed in these cells. A: white arrows denote the Pdx1⁺insulin⁻ nuclei and yellow arrows MafA⁻insulin⁺ nuclei in MafB^fl/fl and MafB^Δpanc islets. B–D: percentage of Pdx1⁺insulin⁻ (B), MafA⁺insulin⁻ (C), and Ki-67⁺insulin⁻ cells (D). There was no difference in the proliferation rate of P1 MafB^fl/fl and MafB^Δpanc β-cells. E: islet-enriched Slc30a8 and Glut2 levels are reduced on P1 and normal by 2 wk in MafB^Δpanc islets; Glut2 (white) and Slc30a8 (red). Dashed yellow lines mark the endocrine⁺ cells. Scale bars, 10 μm. The same confocal settings were used to produce these images.

Our results indicate that embryonic loss of MafB delays islet β-cell maturation. Consequently, we analyzed whether GSIS was compromised in 2-wk-old compared with 8-wk-old MafB^Δpanc islets (Fig. 3A), when essentially no MafB⁺ β- or α-cells were present (data not shown). Insulin secretion was slightly elevated under nonstimulating 2.8 mM glucose conditions in 2-wk-old MafB^Δpanc islets, a basal secretion pattern of immature β-cells (8, 24, 36). Conversely, the insulin secretion profile was normal in 8-wk-old MafB^Δpanc islets. MafB^Δpanc and control mice also had indistinguishable blood glucose tolerance profiles at 3, 6, and 8 wk (Fig. 1C). In addition, the expression levels of a variety of MafA- and MafB-regulated genes were similar to controls in 8-wk-old MafB^Δpanc islets, including insulin, Nkx6.1, Glut2, Pdx1, and Slc30a8 (Fig. 3C).

Fig. 3.

Glucagon secretion is compromised in adult MafB^Δpanc islets but not insulin secretion. Static culture analysis of insulin (A) and glucagon secretion (B) in MafB^Δpanc (black bars) and MafB^fl/fl (open bars) islets. In B, the ratio of 1 mM glucose to 1 mM glucose + 10 mM Arg is 3.5 in MafB^fl/fl and 1.6 in MafB^Δpanc. *P < 0.05, ***P < 0.001, and ****P < 0.0001, basal vs. stimulating conditions (n = 4–6). C: glucagon and ghrelin mRNA levels are selectively reduced in 8-wk-old MafB^Δpanc islet α-cells (n ≥ 4). In contrast, many other α-cell, β-cell, and islet control factors are unchanged. *P < 0.05 compared with MafB^fl/fl controls.

MafB is important to ε-cell ghrelin expression.

Interestingly, glucagon and ghrelin mRNA levels were decreased significantly in 8-wk-old MafB^Δpanc islets, whereas somatostatin, pancreatic polypeptide, and insulin levels were similar to controls (Fig. 3C). The impact on glucagon expression was not surprising, since MafB binds to and activates at the G1 element found between −71 and −55 base pairs relative to the transcription start site (6). In contrast, it is unclear how MafB regulates ghrelin expression. Unfortunately, we could not determine whether there was a change in ghrelin staining levels due to the very few ε-cells in adult MafB^fl/fl or MafB^Δpanc islets. Immunostaining confirmed that MafB is present in wild-type ghrelin⁺ cells at P1 (Fig. 4), when ε-cells make up a small but larger proportion of the islet cell population than in the adult (34). There was no obvious change in the P1 islet ghrelin cell population (Fig. 4). Interestingly, MafB expression was observed in the gut but not in ghrelin⁺ cells (Fig. 5), the principal location of this hormone-expressing population.

Fig. 4.

MafB is expressed in mouse P1 islet ghrelin⁺ ε-cells, although there is no apparent change in the P1 MafB^Δpanc ghrelin⁺ cell population. A: representative images of P1 MafB^fl/fl islets stained for ghrelin (white), MafB (red), and insulin (green). White arrows denote MafB⁺ghrelin⁺ cells. B: P1 MafB^fl/fl and MafB^Δpanc islets stained for insulin (green), glucagon (blue), and ghrelin (white). Scale bars, 20 μm.

Fig. 5.

MafB is produced in the developing gut, albeit not in gut ghrelin⁺ cells. A representative image of P1 wild-type mouse duodenum; MafB (red), ghrelin (white), and DAPI (blue). White arrow denotes a ghrelin⁺MafB⁻ cell. Scale bar, 20 μm.

MafB is important to islet α-cell function.

Glucose-stimulated glucagon secretion was measured in 8-wk-old MafB^Δpanc islets to determine whether secretion capacity was altered. Both glucagon content (Fig. 3B) and α-cell numbers (Fig. 1B) were unchanged from controls, although glucagon secretion was reduced in MafB^Δpanc islets upon exposure to a stimulating low glucose concentration and the potent secretagogue, arginine (Fig. 3B). There were no changes in islet α-cell-enriched transcription factors [i.e., Arx (13) and Brn4 (22a)] or several other α-cell-enriched gene products in MafB^Δpanc islets [i.e., HigD1a, Ndst4, Pdk4, and Ptprd (15); Fig. 3C]. These analyses demonstrate that MafB controls postnatal islet α-cell function, which is quite interesting considering that MAFB levels are reduced in T2DM islet α-cells and glucagon secretion is dysregulated in this state (19, 43).

Loss of MafA and MafB in islet cells results in death soon after birth due to hyperglycemia.

Floxed MafA and MafB mice were crossed with early pancreas-driven Pdx1-Cre (MafAB^Δpanc) transgenic mice to compare their phenotype to the individual mutants. In contrast to MafA^Δpanc (5) or MafB^Δpanc mutants, no viable MafAB^Δpanc mice were found within 2.5 wk of birth (Table 1), although these mice were born at the predicted frequencies (data not shown). This same property was found in MafAB^Δendo animals, where expression of Ngn3-driven Cre removes each transcription factor gene prior to expression in embryonic islet cell progenitors (Table 1). The blood glucose levels in MafAB^Δpanc P1 mice were extremely high (Fig. 6A), strongly indicating that deficiencies in β-cell mass and/or function led to their early death. In addition, blood glucose levels were elevated in MafA^ΔpancMafB^Δpanc/+ and MafA^Δpanc/+MafB^Δpanc mice.

Table 1.

MafAB^Δpanc and MafAB^Δendo mice die soon after birth

Potential Genotypes	Mendelian Ratio	No. of Pups Expected (Out of 54)	Actual No. of Pups	Actual Ratio
MafAfl/flMafBfl/fl × MafAΔpanc/+MafBΔpanc/+
MafABΔpanc	0.125	6.75	0*	0
MafAΔpanc/+ BΔpanc or MafAΔpanc BΔpanc/+	0.25	13.50	16	0.30
MafAΔpanc/+ MafBΔpanc/+	0.1	6.75	12	0.22
Pdx1-Cre−	0.50	27.00	26	0.48
Total			54
MafAfl/flMafBfl/fl × MafAΔendo/+MafBΔendo
MafABΔendo	0.25	14.75	0***	0
MafAΔendo/+ MafBΔendo	0.25	14.75	21	0.36
MafAfl/+ Bfl/fl	0.25	14.75	18	0.31
MafAfl/fl Bfl/fl	0.25	14.75	20	0.34
Total			59

The expected and actual no. of animals obtained from the MafA^fl/flMafB^fl/fl × MafA^Δpanc/+MafB^Δpanc/+ mating or MafA^fl/flMafB^fl/fl × MafA^Δendo/+MafB^Δendo mating at 2.5 wk.

^*P = 0.0438 for MafAB^Δpanc;

^***P = 0.000171 MafAB^Δendo.

Fig. 6.

MafAB^Δpanc mice have profoundly reduced insulin⁺ and glucagon⁺ cell numbers. A: neonatal MafAB^Δpanc animals are severely hyperglycemic at P1. Statistical significance between MafA^fl/flMafB^fl/fl and MafA^ΔpancMafB^Δpanc/+, MafA^Δpanc/+MafB^Δpanc, or MafAB^Δpanc is shown. *P < 0.05, **P < 0.005, and ***P < 0.0001. B: insulin⁺ and glucagon⁺ cell numbers are decreased in MafAB^Δpanc animals (n = 3). *P < 0.05 and **P < 0.01 compared with MafAB^fl/fl controls. C: most MafAB^Δpanc Pax6⁺ cells do not produce insulin at P1; insulin (Ins; green), glucagon (Gluc; white), and Pax6 (red). Scale bars, 10 μm. D: MafAB^Δpanc animals have fewer β-cell-enriched Pdx1⁺ and Nkx6.1⁺ cells (n = 3). Approximately 10,000 and 20,000 pancreatic nuclei were counted at E15.5 and P1, respectively. **P < 0.01 compared with MafAB^fl/fl control.

To investigate how insulin⁺ and glucagon⁺ cell formation was affected in MafAB^Δpanc mice, the number of hormone⁺ and Pax6⁺ cells was quantified at E15.5 and P1. The MafAB^Δpanc insulin⁺ and glucagon⁺ cell population was significantly reduced (Fig. 6B), whereas total Pax6⁺ endocrine cell numbers were unchanged. Note that very few Pax6⁺ cells produce either insulin or glucagon in MafAB^Δpanc islets (i.e., ∼75% change in hormone⁺ cell numbers; Fig. 6, B and C). However, β-cell-enriched Nkx6.1⁺ and Pdx1⁺ cell numbers were decreased by only ∼25% (Fig. 6D). These results suggest that MafA and MafB are the principal drivers of insulin gene expression and not Nkx6.1 (38) and Pdx1 (31). The overall developmental impact on hormone and transcription factor levels was comparable between MafAB^Δpanc and MafB^−/− (3) mice. Collectively, these data demonstrate not only that the combined actions of MafA and MafB are required for postnatal β-cell formation but that MafB alone (unlike MafA) has little (if any) impact on mouse β-cells.

MAFB is expressed in NHP islet β-cells, with islet architecture and cell function analogous to human.

Human islets are dissimilar to mice in several ways, including the presence of MAFB in human β-cells, GSIS properties, and islet cell composition and distribution (9, 11, 14). We next determined whether these islet properties were shared with NHPs. Like humans, MAFB was present in β-cells as well as α-cells (Fig. 7, A and B). MAFB mRNA expression was relatively high in NHP and human islets compared with mice, whereas NHP MAFA was present at a level similar to mice (Fig. 7A). PDX1 and pan-endocrine NEUROD1 transcription factor levels were comparable across species.

The relative mRNA levels of NHP INSULIN, GLUCAGON, and SOMATOSTATIN were similar to human islet levels (Fig. 7D). The differences in INSULIN, GLUCAGON, and α-cell-enriched ARX (Fig. 7A) transcription factor levels with mice closely correlated with the increased proportion of α-cells in human and NHP islets, with fewer β-cells and more α-cells in primates (Fig. 7E) (9). Moreover, the distribution of human (9) and NHP α-, β-, and δ-cells was scattered throughout the islet, which was different from mouse islets where the α- and δ-cells surround the β-cell-rich core (Fig. 7E). Insulin secretion from NHP and human islets was indistinguishable at 5.6, 16.7, and 16.7 mmol/l glucose plus IBMX and significantly different from mouse islets (Fig. 8). The higher basal and lower insulin secretion response to high glucose is a defining feature of human and NHP islets. These results suggest that MAFB expression in islet β-cells represents a unique feature of the primate islet and is consequential to cell function.

Fig. 8.

Insulin secretion properties of NHP and human islets are indistinguishable. A–C: glucose-stimulated insulin secretion characteristics of NHP (n = 9; A), human (n = 30; B), and C57 mouse (n = 12; C) islets in perifusion assays. D: basal insulin secretion levels. E and F: area under the curve (AUC) analysis of the 16.7 mM glucose (E) and 16.7 mM glucose + IBMX (F) data. ***P < 0.001, mouse vs. human; +P < 0.05 and +++ P < 0.001, mouse vs. NHP. No significant differences were found between NHP and human. G, glucose. IEQ, islet equivalent.

DISCUSSION

Mouse MafA and/or MafB were removed specifically in early pancreatic progenitor cells to obtain insight into the contribution of MafB individually and MafA and MafB in combination with postnatal islet β- and α-cells. Only transient deficiencies in MafB^Δpanc β-cell activity were found, although adult islet α-cell glucagon secretion levels as well as ghrelin mRNA were compromised. Compensation by MafA likely explains the limited effect on MafB^Δpanc β-cells, which is supported by the tight expression linkage with mutant insulin⁺ cells (Fig. 2). This proposal is also consistent with the substantially reduced islet insulin⁺ cell population in MafAB^Δpanc islets, which caused overt hyperglycemia and death. We conclude that MafB is essential only to postnatal mouse α-cell activity. This seems paradoxical since MAFB in human islet β-cells represents a unique transcription factor signature. Here we show that NHP islet β-cells also produce MAFB, and islet cell composition and β-cell function are essentially equivalent to that of humans.

MafB was broadly expressed in the P1 duodenum (Fig. 5), wherein a variety of metabolically interesting hormones are produced [e.g., cholecystokinin (CCK), gastric inhibitory peptide (GIP), glucagon-like peptide-1 (GLP-1), and ghrelin] (18). Notably, MafB was not expressed in ghrelin⁺ cells in the duodenum, presumably because these cells are not derived from Ngn3⁺ progenitors (23), as in the pancreas. With this consideration, it will be of interest to determine whether MafB is necessary for regulating Ngn3-derived CCK, GIP, and/or GLP-1 expression, with positive results supporting a wider role for MafB in defining cell identity decisions.

Notably, whereas MafA and MafB influence transcription of genes associated with glucose sensing and hormone secretion in mouse islet α- and β-cells (5, 22), both have distinct effects on cell activity (1, 5) and hormone gene expression (4). For example, induction of glucose-responsive insulin secretion is obtained upon misexpression of MafA in neonatal rat islets, with high MafB expression associated with immature rodent GSIS properties (1). Amino acid differences within the NH₂-terminal transactivation domain and COOH-terminal spanning basic leucine zipper DNA-binding/dimerization region likely contribute to the distinct functions of each of these proteins (see Fig. 7C). For example, exchanging the COOH-terminal region of MafA with MafB enabled this chimera to activate insulin levels in chick in ovo electroporation assays (4). Consequently, the increased level of MAFB in primate β-cells could contribute to differences in insulin secretion levels in rodents islets (Fig. 8) (14). Importantly, recent evidence suggests that both MAFA and MAFB mediate GSIS in human β-cells (41). Thus, insulin secretion was compromised upon reducing the levels of either MAFA or MAFB in human EndoC-βH1 cells. Furthermore, candidate gene analysis demonstrated that several genes important to human β-cell activity were regulated by MAFA and MAFB, but only MAFB activated SLC2A1 (GLUT1) levels, the primary glucose transporter of human β-cells. Future efforts will be focused on obtaining a better understanding of how MAFA and MAFB control human islet β- and α-cell functions.

Our data suggest that reduced MAFB levels in islet α-cells would contribute to the dysregulated glucagon secretion associated with T2DM patients (44). However, we predict that T2DM α-cell dysfunction does not derive solely from MAFB loss but rather from a compilation of changes that occur in the islet populations, including counterregulatory insulin signaling to the α-cell. It will be of value to explore the mechanism(s) by which MAFB controls glucagon secretion and whether these are dysregulated in diabetic islet α-cells.

GRANTS

This work was supported by grants from the National Institutes of Health (RO1-DK-090570 to R. Stein; DK-66636, DK-68854, DK-72473, DK-89572, and DK-089538 to A. C. Powers; T32-DK-007563 to J. Spaeth; R24-DK-093437 to D. M. Harlan, K. L. Grove, C. T. Roberts, Jr., A. C. Powers, and R. Stein; and P51-OD-0110921 to K. L. Grove and C. T. Roberts, Jr.) and the Vanderbilt Diabetes Research and Training Center (DK-20593). This work was also supported by a Merit Review Award from the Veterans Affairs Research Service (BX000666 to A. C. Powers) and the March of Dimes (1-FY08-381 to L. V. Goodrich) as well as grants from the Juvenile Diabetes Research Foundation (JDRF; 26-2008-863 to A. C. Powers). Imaging was performed with National Institutes of Health support from the Vanderbilt University Medical Center Cell Imaging Shared Resource (CA-68485, DK-20593, DK-58404, HD-15052, DK-59637, and EY-08126) and the Vanderbilt University Medical Center Islet Procurement and Analysis Core (DK-20593). Human islets were obtained from the Integrated Islet Distribution Program, which is supported by the National Institute of Diabetes and Digestive and Kidney Diseases and the JDRF.

DISCLOSURES

No potential conflicts of interest relevant to this article, financial or otherwise, are reported.

AUTHOR CONTRIBUTIONS

E.C., C.D., J.S., H.A.C., D.S., J.C., W.-M.Y., and R.S. conception and design of research; E.C., C.D., J.S., M.G., H.A.C., D.S., J.C., and W.-M.Y. performed experiments; E.C., C.D., J.S., H.A.C., D.S., W.-M.Y., L.V.G., and D.M.H. analyzed data; E.C., J.S., H.A.C., D.S., W.-M.Y., L.V.G., K.L.G., C.T.R.J., A.C.P., G.G., and R.S. interpreted results of experiments; E.C. and J.S. prepared figures; E.C., J.S., and R.S. drafted manuscript; E.C., J.S., H.A.C., D.M.H., K.L.G., C.T.R.J., A.C.P., G.G., and R.S. edited and revised manuscript; E.C., C.D., J.S., M.G., H.A.C., D.S., J.C., W.-M.Y., L.V.G., D.M.H., K.L.G., C.T.R.J., A.C.P., G.G., and R.S. approved final version of manuscript.